Solution structure of the catalytic domain of γδ resolvase. Implications for the mechanism of catalysis1

doi:10.1006/jmbi.2001.4821 available online at http://www.idealibrary.com on

J. Mol. Biol. (2001) 310, 1089±1107

Solution Structure of the Catalytic Domain of g d Resolvase. Implications for the Mechanism of Catalysis Borlan Pan, Mark W. Maciejewski, Assen Marintchev and Gregory P. Mullen* Department of Biochemistry University of Connecticut Health Center, 263 Farmington Avenue, Farmington CT 06032, USA

The site-speci®c DNA recombinase, gd resolvase, from Escherichia coli catalyzes recombination of res site-containing plasmid DNA to two catenated circular DNA products. The catalytic domain (residues 1-105), lacking a C-terminal dimerization interface, has been constructed and the NMR solution structure of the monomer determined. The RMSD of the NMR conformers for residues 2-92 excluding residues 37-45 and 64-73 is Ê for backbone atoms and 0.88 A Ê for all heavy atoms. The NMR sol0.41 A ution structure of the monomeric catalytic domain (residues 1-105) was found to be formed by a four-stranded parallel b-sheet surrounded by three helices. The catalytic domain (residues 1-105), de®cient in the C-terminal dimerization domain, was monomeric at high salt concentration, but displayed unexpected dimerization at lower ionic strength. The unique solution dimerization interface at low ionic strength was mapped by NMR. With respect to previous crystal structures of the dimeric catalytic domain (residues 1-140), differences in the average conformation of active-site residues were found at loop 1 containing the catalytic S10 nucleophile, the b1 strand containing R8, and at loop 3 containing D67, R68 and R71, which are required for catalysis. The active-site loops display high-frequency and conformational backbone dynamics and are less well de®ned than the secondary structures. In the solution structure, the D67 side-chain is proximal to the S10 side-chain making the D67 carboxylate group a candidate for activation of S10 through general base catalysis. Four conserved Arg residues can function in the activation of the phosphodiester for nucleophilic attack by the S10 hydroxyl group. A mechanism for covalent catalysis by this class of recombinases is proposed that may be related to dimer interface dissociation. # 2001 Academic Press

*Corresponding author

Keywords: site-speci®c DNA recombination; multidimensional NMR; three-dimensional structure; catalytic mechanism; monomer-dimer

Introduction Site-speci®c DNA recombination occurs in a wide variety of biological processes where the Present address: Borlan Pan, Protein Engineering Department, Genentech, Inc, 1 DNA Way, South San Francisco, CA 94080, USA. Abbreviations used: res, resolvase DNA-binding site; HSQC, heteronuclear single-quantum coherence; HPLC, high-performance liquid chromatography; NOE, nuclear Overhauser effect; NOESY, NOE spectroscopy; R1, longitudinal relaxation; R2, transverse relaxation; J(o), spectral density function. E-mail address of the corresponding author: [email protected] 0022-2836/01/051089±19 $35.00/0

controlled rearrangement of DNA segments provides for the genetic diversity or a selective advantage for the organism.1 gd resolvase is a 20 kDa site-speci®c DNA recombinase that catalyzes the ®nal stage of gd DNA transposition in Escherichia coli, where a circular DNA cointegrate, formed by transposition and DNA replication, is converted into two singly linked circular DNA products.2 During the recombination process, six dimers of resolvase interact with two directly repeated resolvase DNA-binding site (res) DNA sequences on a circular, negatively supercoiled DNA plasmid (Figure 1(a)). In the resolvase-res site interaction, each of three resolvase dimers binds imperfect inverted repeat # 2001 Academic Press

1090

Structure of the Catalytic Domain of  Resolvase

Figure 1. The subunit rotation model for the catalytic reaction of resolvase and conserved residues in three homologous TnpR recombinases. (a) Resolvase binds as dimers to sites I, II and III in the two res sequences to form a catalytic synaptosome. Although not shown as such, the resolvase dimers bound to sites I (center of lower left panel) associate with dimers bound at sites II and III. After cleavage, the DNA strands crossover and are religated (panel at lower right). Catenated DNA products are released upon dissociation of the synaptosome. (b) Primary sequence and alignment of the full length gd-resolvase (gd RES) to a representative member of the DNA invertases, Hin (HIN)6 and a DNA resolvase-invertase, b-recombinase (b REC),7 which display limited sequence identity. Residues that are the same as resolvase are colored red, and residues of similar type are colored blue. The NMR determined secondary structure of the resolvase catalytic domain (1-105) is shown above the sequence. Residues forming the dimerization helix aE and the helical arm region determined from the X-ray structure of the dimeric resolvase DNA complexes are highlighted in yellow. Residues forming the DNA-binding domain are highlighted in blue. Two segments of residues that form a signi®cant part of the active site are highlighted in red.

sequences designated sites I, II and III in res to form a resolvasome, and two resolvasomes interact to form a synaptosome, through which catalysis is mediated (Figure 1(a)). Covalent catalysis is performed by a resolvase dimer bound at each site I of res. The S10 active-site residue forms a 50 phosphoester covalent intermediate with DNA, which results in a corresponding 30 overhang of two nucleotides on the opposing strand.3 The conserved Y6 does not act as the

nucleophile in catalysis.4 Of the mechanisms suggested to explain DNA strand crossover, subunit rotation at two interacting dimers (Figure 1(a)) has been most widely postulated. An alternative mechanism would entail a concerted migration of the two 30 -OH strands of the cleaved and 50 covalently attached duplex at the res site, from one dimer to an adjacent dimer, with the adjacent dimer similarly passing its 30 -OH strands of the covalently attached duplex.5

1091

Structure of the Catalytic Domain of  Resolvase

The resolvase protein folds into two domains (Figure 1(b)) that were ®rst characterized by limand later characterized ited proteolysis8 structurally.5,9 ± 12 In the crystal structure of the resolvase catalytic domain (residues 1-140), three resolvase monomers (termed 1, 2, and 3) pack within an asymmetric unit and display a 1,2 dimer interface formed by the parallel pairing of C-terminal helices (residues 102-120) and a 2,3 dimer interface. Additionally, the 2,3 and symmetryrelated 20 ,30 dimers within asymmetric units pack to form a 2,30 (and 20 ,3) interface, and the pairing of the 2,3 and 20 ,30 dimers results in a crystallographic tetramer containing three orthogonal C2 axes of symmetry. The 1,2 dimer interface has been characterized in solution by mutational studies and disul®de crosslinking.13 The S10 residues in the 1,2 Ê apart, which was too far for catdimer were >30 A alysis at respective phosphodiesters separated by two base-pairs at the res DNA site I.9 This observation has suggested that reorganization at the dimer interface may be necessary for catalysis. The 2,3/20 ,30 tetramer revealed cooperativity interfaces important for interaction between resolvase dimers in formation of the synaptic recombination complex.14 Other crystal structures, including a detailed re®nement of the original catalytic domain structure and diffraction data for the catalytic domain dimer on crystallization of the full-length protein,5,11 have provided additional structural information on sites of potential contacts between monomers in the synaptosome and on potential ¯exibility in the resolvase catalytic domain structure. A cluster of conserved residues (Y6, R8, V9, S10, T11, Q14, Q19, G40, D67, R68 and R71) is located at or near the active-site S10 residue, and mutagenesis studies have shown these residues to be essential for several steps of recombinational catalysis.4,14,15 The structure of the intact dimeric resolvase in complex with a symmetric 34 base-pair res-like DNA cleavage site showed that DNA was bent away from the catalytic domains with helical arm segments traversing the minor groove and connecting via C-terminal extended segments to the DNAbinding domain.12 The overall dimeric structure is rather asymmetric with one active site showing a closer proximity to the phosphodiester cleavage site (although not suf®cient for catalysis) than the other active site of the dimer. The DNA-binding domain at the C terminus is formed by three helices, with helices 2 and 3 being characterized as a helix-turn-helix motif.10 In the dimeric resolvaseDNA complex, the dimerization helices extend toward the DNA-binding domain.12 The helixturn-helix motif was found to display speci®c contacts with the 34 bp res-like DNA sequence. Further recognition was mediated through an extended segment between the arm helices and the DNA-binding domain, which made speci®c contacts in the minor groove. As stated above, strand exchange during resolvase catalyzed recombination has been postulated

to involve either subunit rotation (requiring steps that include dimer dissociation, switching of monomer partners, and re-association) or alternatively, res site DNA association and crossover in the absence of subunit rotation and likely in conjunction with large distortions in the DNA. Topological consequences of recombination are consistent with subunit rotation,16 as is the ®nding that the resolvase dimer acts in cis at a res site I.17,18 In the subunit rotation model (Figure 1(a)), the exchange of protein subunits would require signi®cant changes in protein-protein interactions at the dimer interface. Mutational and biochemical data have localized mutations that enhance catalytic activity to the dimerization region of the protein,19 ± 22 and biophysical evidence23 suggests that disruption of the initial dimer interface is involved in the catalytic activity of the TnpR family of recombinases. Additional interactions would be necessary to prevent complete dissociation of the covalently bound DNA ends on disruption of the 1,2 dimer interface during subunit rotation, and a suggestion has been put forward that during subunit rotation, transition state contacts among the opposing monomers may occur that could prevent dissociation.24 In previous X-ray structures, all monomers contained a 1,2 dimer interface in which the dimerization helix from each monomer appears to be integral to the structural stability of the catalytic domain. Additionally, in the intact resolvase-DNA complex, residues from the dimerization helix of an adjacent monomer make contacts with residues that form the active site. Thus, an unstructured (or absent) dimerization helix could be expected to induce changes in the structure and/or the dynamics of the catalytic domain and the active site that would be of signi®cance in evaluating the subunit exchange model for recombination. Indeed, prior to the structural characterization reported here, a structured resolvase catalytic domain in the absence of dimerzation was unknown. We report on the solution structure of the monomeric catalytic domain of resolvase, which, in combination with previous biochemical, mutational, and crystallographic structures, implies that a dimer dissociation mechanism may activate the catalytic active site.

Results Characterization of the catalytic domain monomer Initially, our efforts toward a structure determination of a monomeric form of resolvase had been directed toward the I110R mutant of resolvase, which was defective in dimerization at the 1,2 dimer interface. However, conformational exchange effects were found to result in loss of 1 H-15N cross-peaks for many of the amides in the catalytic domain of the I110R resolvase monomer.25 Proteolytic cleavage of the I110R mutant to yield the catalytic domain was found to eliminate these

1092 conformational exchange effects in the catalytic domain. Thus, limited tryptic digestion of the I110R mutant resolvase for up to 90 minutes produced major products of 12, 8 and 5 kDa, with the 12 and 5 kDa fragments corresponding to the masses of the catalytic domain and the DNA-binding domain, respectively (Figure 2). Matrix-assisted laser desorption/ionization (MALDI) mass spectroscopic analysis indicated that the 12 kDa fragment consisted of residues 1-105. The precise molecular masses for the remaining proteolytic fragments were not determined. The coding sequence for residues 1-105 was engineered for overexpression in the DE3(BL21) E. coli strain. The resolvase catalytic domain was isolated and puri®ed from the soluble fraction on cell lysis in yields of 100 mg/l cell culture. The 1 H-15N HSQC spectrum of the domain showed somewhat broadened resonances in 0.1 M NaCl (Figure 3(a)), but exhibited narrow resonances characteristic of the monomeric form (see below) at high ionic strength (Figure 3(b)). The potential association of the gd resolvase catalytic domain (residues 1-105) as suggested by the line-widths at low salt was characterized by quantitative gel-®ltration HPLC. Unlike the dimeric catalytic domain and dimeric full-length resolvase, which precipitate at low ionic strength and are puri®ed by urea-extraction of the insoluble fraction from E. coli, the resolvase catalytic domain (residues 1-105) is fully soluble at high ionic strength (50 mM Tris, 1 M NaCl) and at lower ionic strength (50 mM phosphate, 100 mM NaCl). Thus, we performed an analysis of the native molecular mass as assessed from the gel-®ltration analysis at each ionic strength condition. The retention time for the resolvase catalytic domain was used to calculate the apparent molecular mass on the basis of a comparison to known standards. At high ionic strength and at protein concentrations of 40 mM, 200 mM, and 1200 mM, the resolvase catalytic domain displayed an apparent molecular mass of 15 kDa at each concentration, indicating a monomer. At lower ionic strength (50 mM phosphate,

Structure of the Catalytic Domain of  Resolvase

100 mM NaCl), the molecular mass of the resolvase catalytic domain displayed a concentrationdependence, migrating at 15 kDa at the lowest concentration and 28 kDa at the highest concentration. The concentration-dependence of the molecular mass indicated a monomer-oligomer equilibrium. Using the equations described in Materials and Methods, a monomer-dimer equilibrium with a dissociation constant of 700 mM was found to give a good ®t to the data (Figure 3(c)). Solution structure of the catalytic domain monomer The solution structure of the monomeric catalytic domain (residues 1-105) of gd resolvase was determined using multidimensional NMR methods. Assignments of backbone and side-chain 1H, 15N and 13C resonances, as previously reported and deposited in the BioMagResBank,26 were used in a complete analysis of 15N-edited and 13C-edited 3D nuclear Overhauser effect spectroscopy (NOESY) spectra. Representative nuclear Overhauser effects (NOEs) from the 15N-edited 3D NOESY are shown for residues 17-26 that form aA (Figure 4(a)). Longrange NOEs are shown for side-chains that, in part, de®ne the active-site region (Figure 4(b)). From the NOESY data, a total of 2109 NOE distance restraints were obtained including 767 longrange restraints for structure determination (Table 1). In addition, 69 f and 45 w1 torsion angles were assessed from 3JNHa and 3JNbH coupling constants and NOE patterns. The NOE distances and torsion angles were supplemented by 35 hydrogen bonds assessed from slowly exchanging amide protons and initial structures calculated in the absence of hydrogen bond restraints. The NMRdetermined structure is represented by the superimposition of 19 conformers and displays excellent precision within the folded core (Figure 5(a) and (b)). Three loops and the C-terminal tail on one side of the protein structure are less well determined. The RMSD for residues 2-92 excluding resiÊ for backbone atoms dues 37-45 and 64-73 is 0.41 A Ê for all heavy atoms, and the structure and 0.88 A

Figure 2. Time-course for proteolytic digestion of the I110R dimerization-defective mutant of gd resolvase as analyzed by SDS-PAGE. Molecular mass markers are shown in the left lane. Times for digestion are shown above each lane. The apparent molecular masses of the resulting fragments are indicated on the right side.

1093

Structure of the Catalytic Domain of  Resolvase Table 1. Structural statistics hSAi Ê ) with respect to mean RMSD (A Secondary structure regions Heavy backbone atoms All heavy atoms Structured regions (2-36,46-63,74-92) Heavy backbone atoms All heavy atoms Number of experimental restraints Inter-residue short-range (ji ÿ jj ˆ 1) Inter-residue medium-range (1 < ji ÿ jj 4 5) Inter-residue long-range (ji ÿ jj > 5) Intra-residue Dihedral angles (deg.) Hydrogen bonds Restraint violationsa Ê NOE distances with violations >0.2 A Dihedrals with violations >2  Ê ) for experimental restraintsb RMSD (A All distance restraints Torsion angles X-PLOR energies from SAc FNOE (kcal molÿ1)d Ftor (kcal molÿ1) Frepel (kcal molÿ1) FL ÿ J (kcal molÿ1) RMSD from idealized covalent geometry Ê) Bonds (A Angles (deg.) Impropers (deg.) Ramachandran Analysis Residues in favored regions (%) Residues in additionally allowed regions (%) Residues in generously allowed regions (%) Residues in disallowed regions (%)

(SA)r

0.35  0.07 0.85  0.09 0.41  0.08 0.88  0.07 403 609 767 330 114 35 0.45  0.60 1.50  0.95

1 1

0.011  0.001 0.244  0.063

0.011 0.220

11.4  2.6 1.5  0.5 60.4  3.5 ÿ68.0  15.8

11.4 0.4 54.0 ÿ63.5

0.003  0.000 0.371  0.011 0.248  0.012

0.003 0.368 0.241

68  2.2 29  2.9 3  1.1 00

64 27 2 0

hSAi represents the 19 structures calculated in DYANA and re®ned by simulated annealing in X-PLOR using energy terms for NOE distance restraints, dihedral angle restraints, bonds, angles, impropers, and hard sphere van der Waals contacts. (SA)r represents the structure calculated by restrained minimization of the mean of the hSAi structures. a Ê and no dihedral angle restraints were violated by >4  . No distance restraints were violated by >0.4 A b The number of each class of experimental restraints is given in parentheses. c The force constants were as described.10 d 1 cal ˆ 4.184 J.

displays minimal Ramachandran violations (Table 1). The structure is formed by a parallel four-stranded b sheet (b2134), three a-helices (aA, aB, and aD), and a short 310-helix. There is no short helix C as seen in the X-ray structure but, for consistency, we have retained the original aA, aB, and aD helix labeling. Loop 3 in the solution structure, which displays considerable dynamic motion, as will be described, is found in place of helix C. The solution structure displays a unique 310-helix formed by residues 29-32 (Figure 5(c)), that was not seen in the X-ray structure. The overall fold is formed by the packing of helix aA on one side of the (b2134) sheet and the two parallel helices, aB and aD, packing on the opposite side of the central b-sheet (Figure 5(c)). Helix aA and the loop connecting aA and b2 display hydrophobic contacts from L16, V20, L23 A26 and V28 to A7 of b1 and V9, to I33 and T35 of b2, and to L62 of b3. Helix aB and the loop connecting aB and b3 display hydrophobic contacts from L48, L50, L51 and V55 to F4 and Y6 of b1, to F34 of b2, and to L79 and F83 of aD. M76, L79, I80 and F83 of aD display

contacts to I61 and V63 of b3 and to V88, I90, and F92 of b4. Particularly striking is a ®nding that the L66 side-chain (Figure 5(b)) interacts with the catalytic core of the monomer, whereas in the catalytic domain dimer X-ray structures, the L66 side-chain was found to pack with the dimerization helix of the opposing monomer. This difference in packing affects the positioning of the 64-75 loop in the monomer as compared to the dimer, as discussed below. The N terminus and three connections at the Cterminal ends of the a-helices are found along the N-terminal side of the b-strands. Helix aA and the b2 strand are connected by residues 26-32, which form an extended coil and the short 310-helix. The connection between ab and b3 is formed by a turn consisting of residues 55-58. Helix aD and the b4 strand are connected by a turn consisting of residues 86-87. These connections are structurally well de®ned and are stabilized by hydrogen bonds and hydrophobic contacts. In striking contrast, three dynamic loops, (see the relaxation rate measurements and reduced spectral density map-

1094

Structure of the Catalytic Domain of  Resolvase

Figure 3. 1H-15N HSQC spectra of the resolvase catalytic domain (1-105) dimer and the catalytic domain (1-105) monomer, and gel-®ltration HPLC analysis of the monomer-dimer equilibrium. (a) 1H-15N HSQC spectrum of the resolvase catalytic domain dimer (1 mM) in 50 mM phosphate (pH 6.5), 100 mM NaCl. The small dots indicate the cross-peak positions for the monomeric state at high salt. (b) 1H-15N HSQC spectrum of the resolvase catalytic domain monomer (2 mM) in 50 mM phosphate (pH 6.5), 1 M NaCl. (c) Apparent molecular mass versus concentration of the resolvase catalytic domain (1-105) expressed as monomers as determined by the average position of the chromatographic peak in a gel-®ltration HPLC experiment in 50 mM phosphate, pH 7.0, 100 mM NaCl. The calculated curve was ®t using Kdimer ˆ 700 mM. D

ping section), are found at the connections between the C-terminal side of the b-strands and the ahelices (bottom side of the structure in Figure 5). These loops form part of the active site. The dynamic C-terminal residues are found at the C-terminal side of b4 with residues 93-96 forming a partially ordered turn immediately C-terminal to b4.

Amide 15N relaxation rates, 1H-15N NOEs, and reduced spectral density mapping Amide 15N longitudinal relaxation (R1) and transverse relaxation (R2) relaxation rates and 1 H-15N NOEs at 600 MHz and 500 MHz (Figure 6) are indicative of considerable dynamic motion within several backbone segments of the catalytic

Structure of the Catalytic Domain of  Resolvase

1095

Figure 4. Representative regions from NOESY spectra used to obtain structural distance restraints for resolvase. (a) Strips from the 3D 15N-edited NOESY showing sequential connectivities in helix aA. 15N chemical shifts for each plane from which the strips are taken are shown at the top. (b) Peaks from the 3D 13C-edited NOESY showing longrange restraints from L66 to S10 (left) and from Val 63 to Y6 and F92 (right).

domain of gd resolvase. Increased R1 values were observed for amide protons at the N terminus (residues 2-3), at loop 1 (residues 10-11 and 13-15) as well as the amino-terminal residue of aA (residue 16), at loop 2 (residues 38-41 and 43-45), and at loop 3 (residues 67, 69, 72-74). The increase in R1 in the loops is consistent with increased motion (an overall decrease in the local correlation time) for these amide protons. In comparison to secondary structures, R2 relaxation rates were decreased dramatically for backbone amide protons in loop 2 and in C-terminal residues 95-105, due to increased motion. Conversely, increased R2 values were observed at residue 13 and for several of the residues in loop 3, consistent with exchange contributions to the transverse relaxation. From the R2/ R1 ratio for amide protons in regular secondary structure, an average tm of 7.8(0.3) ns was calcu-

lated, which, under conditions of 25  C, is indicative of a protein of 105 residues as determined from a plot of tm versus protein size.28 The 1H-15N NOEs, which are on a scale ranging from 1.0 for no NOE to ÿ5.5 for full NOE, were quite large for backbone amide protons in loops 1, 2, and 3, with the values approaching 0.2 for backbone amide protons in loop 2. 1H-15N NOEs were large for backbone amide protons in the C-terminal segment with negative values observed for residues 100105. The relaxation and NOE data were used to quantify the spectral density function at hJ(oH)i, J(oN), and J(0), thereby providing direct information on the motional contributions at each of these frequencies at the two ®eld strengths (Figure 7). For residues in loops 1, 2, and 3, and for residues at the C terminus, hJ(oH)i showed

1096

Structure of the Catalytic Domain of  Resolvase

Figure 5. The NMR solution structure of the gd-resolvase catalytic domain (residues 1-105). (a) Stereoview of a backbone superimposition of 19 structural conformers. The superposition used all a-helices (cyan) and b-strands (dark blue). Loops 1, 2 and 3 are colored green, yellow and red, respectively. The C-terminal tail is colored magenta. (b) Stereoview of the 19 structural conformers showing hydrophobic packing of side-chains. The coloring is blue for the backbone, gold for side-chains that pack at the helix 1 side of the b-sheet and green for side-chains that pack at the helix 2/helix 3 side of the b-sheet. The Leu66, Leu69 and Ala74 residues located in loop 3 are colored red with Leu66 located at the left side of loop 3. (c) Ribbon representation of the minimized average solution structure with coloring as in Figure 3(a). Structures were displayed using the molecular graphics program MOLMOL.27

Structure of the Catalytic Domain of  Resolvase

1097

Figure 6. Relaxation and 1H-15N NOE data for the catalytic domain (1-105) of gd resolvase. Plots of (a) R1, (b) R2, and (c) 1H-15N NOE values versus residue number at 500 and 600 MHz. The 500 and 600 MHz data are shown as circles and triangles, respectively.

increased values indicative of high-frequency (picosecond motion) internal motion. Loops 1, 2, and 3 also showed increased values for J(oN) indicative of dynamic motion on a time-scale that was ten times slower than that of the high-frequency motion. The backbone amide protons of residues 2 and 3 also showed increased values for J(oN), but the hJ(oH)i values were not increased signi®cantly. On comparing amide protons in the loops to those in the structured segments, the J(0) spectral density was found to be reduced signi®cantly in loop 2, consistent with the signi®cantly increased hJ(oH)i values in loop 2. An increased J(0) value was observed for residue 13 of loop 1 due to apparent exchange contributions. Increased J(0) values were observed for several backbone amides in loop 3, including 63, 64, 66, 69, 72, and 73, due to apparent exchange contributions. Consistent with this, the backbone 1H-15N cross-peaks for 65, 67, 68, 70, and

71 (i.e. neighboring residues to those with high J(0) values) are broadened severely at 25  C, and relaxation data for these residues were not obtained. While anisotropic rotation can also give rise to increased values of J(0), this seems unlikely in this case, where the effect was observed only in unstructured loops. These results are consistent with motion on a millisecond time-scale in loop 3. In contrast, the backbone amide protons in the loops that make connections between the C termini of the a-helices and the N-terminal side of the bstrands (top side of the structure in Figure 5), showed no signi®cant differences in the values of hJ(oH)i, J(oN), and J(0) in comparison to the secondary structural elements. The C-terminal residues showed relatively high hJ(oH)i and low J(0) values, indicative of considerable dynamic motion. Loops 1 and 3 contain several conserved residues (Figure 1(b)) that have been shown by muta-

1098

Structure of the Catalytic Domain of  Resolvase

Figure 7. Reduced spectral density mapping for the catalytic domain (1-105) of gd resolvase. Plots of (a) J(0), (b) J(oN), and (c) Jh(oH)i and their uncertainties versus residue number at 500 and 600 MHz. The 500 and 600 MHz data are shown as circles and triangles, respectively.

tional analysis to be important for the catalytic activity.4,14,15 S10 in loop 1 functions as the catalytic nucleophile and forms a phosphoester bond with the scissile phosphate group. R8 at the C terminus of b1 and D67, R68 and R71 in loop 3 are required for catalysis. R45 in loop 2 is conserved and appears to be positioned for activation of the scissile phosphate bond as described further below. Thus, the motional properties of these three loops, all on one side of the protein (bottom side in Figure 5), are of interest, particularly in view of the restricted dynamic motion in connecting loops between helices and strands on the opposite side of the protein (top side in Figure 5).

Chemical shift mapping of a salt-sensitive dimerization interface As described above, the resolvase catalytic domain, de®cient in the C-terminal helical dimerization region, was found to display a monomerdimer equilibrium in 50 mM phosphate and 150 mM NaCl. The catalytic domain was monomeric at high ionic strength (50 mM Tris, 1 M NaCl). These results are indicative of a salt-sensitive ionic dimerization interface. The dimerization interface was mapped by measuring chemical shift changes in the 1H-15N HSQC spectrum of the resolvase catalytic domain at concentrations of 100, 200,

Structure of the Catalytic Domain of  Resolvase

400, 800, 1200, and 1600 mM at the lower ionic strength condition. The determined dissociation constant of 700 mM for the monomer-dimer equilibrium (Figure 3(c)) indicates that the resolvase catalytic domain is predominantly monomeric at 100 mM and predominantly dimeric at 1600 mM. The chemical shift changes were easily tracked for nearly all cross-peaks over the 100-1600 mM concentration range, and the change in the position of the 1H-15N HSQC cross-peaks as a function of concentration were indicative of fast exchange on the chemical shift time-scale between the monomeric and dimeric forms. The sum of the absolute value of the chemical shift changes in the 1H and 15N dimensions for each residue are shown in Figure 8(a). Changes greater than 1.5 times the standard deviation from the mean value (excluding

1099 the largest chemical shift change for residue 55) were found to map to the C-terminal residues in aB, the adjoining C-terminal 55-58 loop, adjacent residues R2 and F4 on b1, the adjacent F34 on b2, K29 on the loop connecting to the N-terminal side of b2, and V88 at the N-terminal side of b4 (Figure 8(b)). These chemical shift effects mapped to a contiguous set of residues at one side of the resolvase catalytic domain (Figure 8(b)), and predict a surface for interaction that includes several charged side-chains (R2, K29, R52, and E56), consistent with its dissociation under high-salt conditions. The charged R32 side-chain forms part of this patch, and the backbone amide proton for R32 appears to have a large chemical shift effect. Due to cross-peak proximity, an assignment for the chemical shift change in R32 is ambiguous, but is

Figure 8. Mapping of the salt-sensitive dimerization interface in the resolvase catalytic domain (1-105). The hydrophobic dimerization interface (residues 106-140) has been removed from the catalytic domain. (a) Sum of the 1H and 15 N chemical shift changes (Hz) on comparing the 1H-15N HSQC spectra for resolvase (1-105) at 200 mM and 1200 mM in 50 mM phosphate (pH 7.0), 100 mM NaCl. The mean change in the chemical shifts and the values for 1.5 s and 2 s greater than the mean are shown. (b) Divergent stereoview of the structure of the resolvase catalytic domain (1105) showing side-chains (light green) for residues showing amide chemical shift changes at the salt-sensitive dimerization interface. An unambiguous change was not determined for R32 (side-chain in dark green), but is assumed. The amide groups (R2, F4, K29, R32, F34, R52, M53, V55, E56, and V88) with chemical shift changes are shown as van der Waals spheres with the nitrogen atoms in blue and the amide protons in gray. Residues that form the saltsensitive dimer interface are on the opposite side from the active-site residues, which are also shown (R8, R45, R68, and R71 in blue; D67 in magenta; and S10 in green).

1100 likely, since two other possible assignments for the cross-peak would result in the only residue with a substantial effect (1.5 s) not mapping to the determined interaction surface. The charged K54 sidechain is also part of this surface patch, but the backbone amide proton for K54 shows only a small chemical shift effect. In addition to charged residues, chemical shift effects mapped to M53, which displays an exposed hydrophobic side-chain and to the amide protons of residues with only partly exposed or buried hydrophobic side-chains (F4, F34, M53, V55, V88), all of which form the contiguous interaction surface. Comparison to the X-ray crystal structures of the resolvase dimer A superimposition of the b-sheet of the solution structure of resolvase with the available crystal structures5,11,12 in the DNA-complexed and uncomplexed forms shows relatively similar overall structure but localized structural differences on comparing the 1,2 dimeric form and the monomeric form of resolvase (Figure 9(a)). Although the b-sheet region of the catalytic domain shows good agreement between the monomeric and the dimeric forms, differences occur primarily in the relative orientations of the a-helices, the catalytic loops and the C-terminal region. Unlike the dimeric forms of the catalytic domain, L66, T73, I77, I80 and F92 in the solution structure cannot interact intramolecularly with L111 or intermolecularly with A113 and V114. The absence of these contacts previously seen in the dimer results in a shift of aD and a change in the conformation of loop 3. Instead of packing against the dimerization helix (Figure 9(b)), L66 packs near aA (Figure 9(c)), bringing the N-terminal segment of loop 3 towards the active-site S10 residue. Conformational heterogeneity is found for L66, M76, and I97 that are at or proximal to the dimer interface in the three monomers of the re®ned crystal structure. In comparison, the L66 side-chain adopts different rotomeric conformations in the three monomers of the re®ned crystal structure. The 310-helix C (residues 65-70) in the X-ray structure is not present in solution but instead forms the ¯exible loop 3, which shows conformational contributions to relaxation (see above). The lack of the 310-helix C in the solution structure is likely a result of an absence of side-chain contacts by L66 at a dimerization interface that is not present in the monomeric catalytic domain. In the dimeric structure determined previously by crystallography, the L66 side-chain is proximal to the dimerization helix of the adjacent monomer. Residues 37-45 that form loop 2 display high-frequency dynamics and show no electron density for monomer 1 in the re®ned X-ray structure, but do show de®ned conformations in monomers 2 and 3.11 Residues 94-97 of the resolvase catalytic domain (1-105) show increased hJ(oH)i and J(oN) values characteristic of high-frequency (picosecond)

Structure of the Catalytic Domain of  Resolvase

dynamics, and residues 93-96 form a partially ordered turn C-terminal to b4. In the crystal structure of the resolvase catalytic domain (1-140), residues 93-99 form a turn and an antiparallel b-strand (b5) that pack against the dimerization helices (Figure 9(b)). Further, in the crystal structure, residues I97, S98 and T99 contact V108, M103, V107 and L111 of the intramolecular dimerization helix, and I97 contacts M106 of the intermolecular dimerization helix (Figure 9(b)). These contacts are obviously absent due to deletion of the dimerization helix in the solution monomer, indicating that such contacts are not required for stable catalytic domain structure. Interestingly, in the re®ned crystal structure of resolvase, the dimerization helix aE displays considerable differences in packing with respect to the b-sheet on comparing monomers 1 and 2 in a superimposition that utilized aA and aB. The ®nding that the dimerization helix aE is not required for a structured domain in solution is consistent with such changes in the packing of aE in these monomers of the crystal.

Discussion The solution structure of the catalytic domain of gd resolvase (residues 1-105) in the monomeric state displays local conformational differences at the active site in comparison to the catalytic domain dimer (residues 1-140) and in comparison to the intact resolvase-DNA complex (Figure 10(a)). In the solution structure reported here, the carboxyl group of D67 shows a close proximity to the hydroxyl group of S10, with distances ranging Ê to 4.4 A Ê . Distances between the carfrom 2.4 A boxyl group of D67 and the hydroxyl group of S10 Ê for the two monomers in the crystal are 8 to 9 A Ê for the structure of the DNA complex and 10-12 A three monomers in the re®ned crystal structure of the catalytic domain dimer (Figure 10(a)). In each case, the distance between the D67 carboxylate oxygen atom displaying the closest approach to the S10 oxygen atom was calculated. The range of distances between S10 and D67 in the solution structure ensemble (as well as in the different crystal forms) is consistent with the motional properties deduced from the dynamics analysis for loops Ê approach (2.7 A Ê in the average 1 and 3. The 2.3 A structure) and the dynamic behavior of D67 would allow the carboxylate group to act as a general base in the abstraction of the S10 hydroxyl proton (Figure 10(b) and (c)). In the solution structure, several catalytically important Arg residues are highly accessible. R68 and R71 do not display de®ned conformations in loop 3, and loop 3 amide protons display both conformational exchange and high-frequency motion that may be of importance in promoting scissile phosphate contacts. The catalytic domain of gd resolvase shares signi®cant structural similarities with 50 -30 exonucleases.29 Interestingly, two divalent metal cations are positioned analogously to the Arg resi-

Structure of the Catalytic Domain of  Resolvase

1101

Figure 9. Comparison of the structure of the catalytic domain of resolvase in the dimeric and monomeric form. (a) Superposition of the b-sheet backbone atoms of the average monomeric solution structure of resolvase (yellow) with the crystal structures of three uncomplexed resolvase molecules11 (light blue) and the two monomers in the site I complex12 (dark blue). (b) Crystal structure of dimeric resolvase at the DNA-binding site I12 using the same orientation as in (a). The side-chains for the catalytic residues (R8 in yellow, R68 in light blue, R71 in dark blue, D67 in red, and S10 in magneta) are shown together with the position of L66 in gold. The DNA cleavage site (DNA phosphate in red) is also indicated. Hydrophobic side-chains that make contacts between the catalytic domain (1-105) and the dimerization helices are shown in cyan. (c) Comparative view of the solution structure of resolvase catalytic domain (1-105) oriented as shown in (a) and (b) with the active-site residues colored as in (b). The ten lowest-energy conformations of the side-chains of the catalytic residues and the backbone of loop 3 are shown. The a-helices, b-strands, and the catalytic residues are labeled.

1102

Structure of the Catalytic Domain of  Resolvase

Figure 10. Active site comparisons for the NMR solution structure and X-ray structures and a proposed mechanism for covalent catalysis. (a) Comparison of sidechain conformations for S10, D67 and R68 in the solution structure of the resolvase catalytic domain (1105) (yellow backbone and thinner, dark-colored side-chains and unprimed labels) and in the dimeric crystal structures (gray backbone and thicker, light-colored side-chains and primed labels). The ten lowest-energy conformations from the NMR structural ensemble are shown. (b) Divergent stereoview of the conformation of the catalytic side-chains in the minimized average structure determined by NMR. (c) The proposed mechanism for covalent catalysis by gd resolvase. In this mechanism, R8, R68 and R71 form charge interactions with the scissile phosphate group and promote nucleophilic attack by the catalytic S10 at the phosphodiester. D67 acts as a general base abstracting the S10 hydroxyl proton, and the S10 hydroxyl group forms a covalent bond to the DNA.

dues at the active site (Figure 11). A mechanism for DNA cleavage and phosphoester bond hydrolysis by a two-metal-ion nuclease was originally proposed for the 30 -50 exonuclease of DNA polymerase I, which binds two metal ions at the 30 50 exonuclease active site.31 ± 33 By analogy to the 30 50 exonuclease mechanism, we propose a mechanism for resolvase catalysis (Figure 10(c)) that is consistent with the positions of the catalytic residues at the active site (Figure 10(b)). In this mechanism, R68 and R71 in resolvase could function similarly to metal ion B in the 30 -50 exonuclease, while R8 and R45 could function similarly to metal ion A. Deprotonation of S10 by D67 is analogous to deprotonation of water by the E357 carboxylate group in the 30 -50 exonuclease mechanism for the Klenow fragment of Pol I.32 As already mentioned, mutagenesis has shown that R8, D67, R68, and R71 are essential for catalysis. S10 has been shown to be the catalytic nucleophile. R45 is conserved throughout this recombinase family (Figure 1(b)) and may functionally electrostatically (Figure 10(c))

or through a direct contact (not shown). Precedent for a similar DNA cleavage mechanism is seen in the structure of the XerD34 and Cre35 ± 37 recombinases, where Lys, Arg or His residues are positioned for electrostatic activation of the phosphodiester group for nucleophilic attack at phosphorus by a Tyr hydroxyl group. Resolvases and invertases are activated by perturbation of the hydrophobic C-terminal dimerization interface. Thus, we have considered the possibility that the resolvase catalytic domain, with the dimerization interface removed, might provide insight into the activated state. Partial disruption of the dimer interface by non-denaturing detergents has been shown to promote activation of the ®rst chemical step of catalysis for the homologous Hin invertase.23 Furthermore, mutational perturbation of the dimer interface can either enhance the ®rst chemical step of catalysis or produce a variety of activated forms of the resolvases and invertases. For the invertases, mutations at the dimer interface have been found to induce Fis-independent activity

1103

Structure of the Catalytic Domain of  Resolvase

Figure 11. A structural comparison of gd resolvase (gold and yellow ribbon) to the T5 50 -30 exonuclease (cyan backbone), two enzymes showing no sequence homology. For resolvase, the side-chains for R8, R45, R68, and R71 are shown in blue; the side-chain for D67 is in magenta; and the side-chain for S10 is in green. For the T5 50 -30 exonuclease, the Asp side-chains that coordinate the active-site metal ions are shown in red. The superimposition suggests a similarity of function for the two metal ions coordinated by the Asp residues in the T5 exonuclease and the Arg residues in gd resolvase. R68 and R71 of resolvase are positioned analogously to one of the metal ions (gray spheres), while R8 and R45 are positioned analogously to the second metal ion. Structurally homologous regions were determined using the program DALI 2.0.30

or to induce enhancer-independent activity on linear substrates.19,20,38 For Tn3 resolvase, a D102Y/E124Q double mutation at the dimerization interface was found to result in an activated form of the resolvase, that did not require the accessory sites II and III for recombination activity.22 While perturbation of the hydrophobic dimer interface is activating, complete disruption of the hydrophobic dimer interface results in an inactive DNA recombinase,13 raising the possibility that the X-ray structure of the resolvase dimer and the solution structure of the resolvase monomer are at two extremes of the catalytic reaction pro®le. Dimerization in an initial complex at site I is required for cooperative binding and bending of the DNA, since mutations that completely disrupt dimerization13 or perturb DNA bending39 inactivate resolvase. Re-organization at the dimer interface through ¯exibility or possible unfolding could explain the residual nicking activity seen in a disul®de-linked dimer of the recombination defective M106C(SS) mutant of resolvase13 and would be consistent with a hypothesis that complete dissociation of the resolvase 1,2 dimer is not necessary for activation. Previously, we had shown that the wild-type resolvase dimer dissociates in the absence of DNA with greater than 50 % dissociation of the dimer at concentrations of less than 10 mM.40 Here, we show that the resolvase catalytic domain lacking the hydrophobic dimerization segment displays an independent salt-sensitive dimerization interface indicative of ionic contacts. Interestingly, the ionic dimerization interface determined in solution by chemical shift mapping is similar to the 2,30 (and

the 20 ,3) interface in a crystallographic tetramer formed by a 2,3 dimer and a dyad-related 20 ,30 dimer in the X-ray structure of the catalytic domain (residues 1-140). For resolvase, resolvasome formation is mediated by contributions from the 2,30 (and 20 ,3) interface between 1,2 dimers bound at sites II and III. Additionally, the 2,30 (and 20 ,3) contacts in resolvase have been characterized as contributing to an interaction between the right half-sites of site I (I-R) and site III (III-R)-bound dimers in the synaptosome.41

Materials and Methods Sample preparation Production of the uniformly 15N or 15N/13C-labeled 12 kDa catalytic core of gd resolvase was achieved by mutation of Met106 to a stop codon in a pET23 overexpression plasmid that had been constructed previously.25 The resulting plasmid construct (pDJ1.3) was transformed into E. coli strain BL21/DE3 containing the accessory plasmid pLysS (Novagen). 15N-labeled and 15 N/13C-labeled resolvase catalytic domain (1-105) was produced by IPTG induction in a minimal medium supplemented with vitamins and 15NH4Cl or U-13C6 glucose.42 The 15N/13C-labeled protein was expressed at levels greater than 100 mg/l cell culture and was puri®ed as described.26 NMR samples were prepared at a concentration of 2 mM in 50 mM phosphate, 1 M NaCl, pH 6.5, 0.01 % (w/v) sodium azide and 10 % (w/v) 2H2O. Size-exclusion HPLC analysis The native molecular mass was determined using HPLC and a Superose 12 (Pharmacia) column. A ¯ow-

1104

Structure of the Catalytic Domain of  Resolvase

rate of 1 ml/minute was used and the retention time after injection was determined from the chromatogram. Molecular mass standards from BioRad (bovine gamma globulin, chicken ovalbumin, myoglobin, and vitamin B12) were used to construct a plot of the log (molecular mass) versus retention time. Experiments at high ionic strength (50 mM Tris (pH 6.8), 1 M NaCl) were performed at 22  C using protein concentrations of 40, 200, and 1200 mM. Experiments at lower ionic strength (50 mM phosphate, pH 7.0, 100 mM NaCl) were performed at 22  C using protein concentrations of 10, 40, 100, 200, 400, 800, 1200, and 2200 mM. In the gel-®ltration HPLC experiment in 50 mM phosphate (pH 7.0), 100 mM NaCl, the average observed molecular mass resulting from the averaging between the monomeric and the dimeric forms were ®t according to: MAVG ˆ a…28† ‡ b…15† where a and b are the mole fractions of dimer and monomer, respectively, and 28 kDa and 15 kDa represent the observed and ®tted molecular masses for the monomeric and dimeric migration, respectively. The mole fractions, a and b, were calculated according to the expressions: a ˆ ‰…x ÿ y†=2Š=‰…x ÿ y†=2 ‡ yŠ ˆ …x ÿ y†=…x ‡ y† b ˆ y=f‰…x ÿ y†=2Š ‡ yg ˆ 2y=…x ‡ y† where x represents the total concentration of resolvase, expressed as the monomer, and y represents the concentration of free monomers. The monomer-dimer equilibrium was ®tted according to: KD ˆ y2 =‰…x ÿ y†=2Š which was rearranged as the quadratic equation and in which x and y are as described above. NMR data collection and data analysis for structure determination Experiments were performed at 25  C on 500 MHz or 600 MHz Varian INOVA spectrometers equipped with triple resonance probes and pulse ®eld gradients. All experiments utilized sensitivity enhancement, gradient selection of 1H-15N coherence, gradient water suppression, and States-TPPI phase cycling in the indirect dimensions. The high-salt conditions posed no problem for tuning of the Varian triple resonance gradient probe and resulted in modest increases (30 %) in pulse widths and similar reductions (30 %) in sensitivity. Short spin lock times with minimal necessary power for the required excitation bandwidth were used in 13C total COSY (TOCSY) transfer experiments in order to prevent sample heating. Nearly complete assignments of 1H, 15N, and 13C resonances were made using 3D HNCA, HN(CO)CA, HNCACB, CBCA(CO)NH, HCC(CO)NH, CC(CO)NH, and HCCH-TOCSY spectra as reported.26 The 15N and 13C-edited NOESY-heteronuclear single quantum coherence (HSQC) experiments were performed on an INOVA 600 using a mixing time of 100 ms in each case. The 1H sweep widths were set to 12 ppm with the carrier at the water resonance. The 15N and 13C sweep widths were 28 ppm and 66.3 ppm, respectively. The 15N and 13C carriers were set at 117 ppm and 43 ppm, respectively. The 15N-edited NOESY-HSQC was acquired with 512 (t3)  64 (t1)  32 (t2) complex points

and 48 scans per increment. The 13C-edited NOESYHSQC was acquired with 512 (t3)  128 (t1)  64 (t2) complex points and 16 scans per increment. Parameters used in the collection of the 1H-15N and 1H-13C HSQC experiments were similar except that either 96 or 128 complex points were collected in the indirect dimension depending on the experiment performed. The data were processed using Felix95 (MSI, San Diego, CA, USA) and analyzed using XEASY43 on Silicon Graphics computers with R4400 or R10000 processors. Two-fold or higher zero-®lling was employed in the indirect dimensions. Nearly complete backbone and side-chain assignments26 were used to assign NOE cross-peaks in 3D 15N-edited and 13C-edited NOESY-HSQC spectra. Slowly exchanging amide protons were identi®ed from a time-course of 1H-15N HSQC spectra after transfer of the protein into 2 H2O. The values for 69 3JNHa and 45 3JNbH coupling constants were determined from an analysis of a 3D HNHA spectrum and a 3D HNHB spectrum as described.44,45 Structure calculations Structures were calculated with the program DYANA46 using upper distance constraints from NOESY cross-peak intensities and dihedral angle constraints derived from the 3JHNHa and 3JNbH coupling constants. NOESY cross-peak intensities were calibrated to daiNi ‡ 1 distances in structures obtained from preliminary rounds of calculations, in which distances were initially estimated for the NOE intensities, through use of the program CALIBA47 within DYANA. Stereospeci®c assignments were obtained by direct analysis of NOEs and relative couplings within a residue as obtained from the HNHB experiment. Hydrogen bonding restraints were added on the basis of slowly exchanging amide protons after analysis of initially determined structures. NOE crosspeak assignments were obtained either from interactive analysis or by using the program NOAH.48,49 NOAH assignments that were incorrect were corrected as required. Additional iterative rounds of structure calculations incorporating additional NOE assignments were performed with DYANA to re®ne the structure. In the ®nal round of calculations, 200 structures were generated from the data within the program DYANA. The 40 structures with the lowest target function were selected for re®nement using simulated annealing and energy minimization within X-PLOR.50 Of these structures, 19 structural conformers were selected that disÊ , no dihedral angle played no NOE violations >0.4 A restraint violations >4  , and low overall X-PLOR energy (Table 1). 1

H-15N relaxation rates, 1H-15N NOEs, and reduced spectral density mapping 1

H-15N correlation spectra for determination of 15N R1, N R2, and 1H-15N NOEs were collected at 25  C on Varian INOVA 500 MHz and 600 MHz NMR spectrometers as described.51 The relaxation rate spectra and 1H-15N NOE and ``no'' NOE spectra were processed and the cross-peak volumes analyzed in Felix95 (MSI, San Diego, CA, USA). The NMR experimental parameters were essentially as described.28 R1 delays of 10, 20, 30, 40, 60, 80, 100, 200, 300, 400, 600, 800, 1000, 1200, 1600, 2000, 2400, and 2800 ms were used, and R2 delays of 10, 170, 90, 25, 30, 190, 110, 290, 50, 210, 130, 250, 70, 230, 150, and 410 ms were used. The data over the complete set of time-delays were collected interleaved. Similarly, NOE

15

Structure of the Catalytic Domain of  Resolvase and ``no'' NOE data were collected interleaved. The number of scans was 16 for the relaxation rate data and 128 for the NOE and ``no'' NOE data. The R1 and R2 relaxation rates were determined using a non-linear least-squares ®t to the data using the equations I(t) ˆ I1 ÿ [I1 ÿ I0]exp(ÿR1t) and I(t) ˆ I0 exp(ÿR2t) as described.28 Uncertainties in the measured cross-peak volumes were extracted from the root-meansquare baseline noise. Uncertainties in the R1 and R2 rates were assessed from 500 Monte Carlo simulations using the estimated errors in the peaks heights.52 For reduced spectral density mapping, published methods were utilized in the analysis of the relaxation data53 ± 56 and software generously provided by Professor Gerhard Wagner and co-workers was employed. Reduced spectral densities at J(0), J(oN) and hJ(oH)i were calculated from the data collected at 500 and 600 MHz. Uncertainties in the reduced spectral density were determined from a Monte Carlo simulation on the basis of the errors in the measured R1, R2 and 1H-15N NOE values using the method described.52 Chemical shift mapping of the dimerization interface The resolvase catalytic domain (1-105) at a concentration of 100 mM was prepared in 50 mM phosphate (pH 7.0), 100 mM NaCl containing 10 % 2H2O. Buffer exchanges from the original puri®cation conditions were performed by several steps of concentration followed by dilution using a Centricon-3 (Amicon). The domain was subsequently further concentrated to 200, 400, 800, 1200, and 1600 mM. A 1H-15N HSQC spectrum was collected at each concentration of the resolvase catalytic domain. The changes in the average 1H and 15N chemical shifts for each cross-peak, resulting from fast averaging in the monomer-dimer equilibrium, were readily tracked as a function of the protein concentration. Assignments were made by comparing the 1H-15N HSQC at 100 mM and 1 M NaCl. Protein Data Bank accession numbers The coordinates for the 19 NMR conformers and the minimized average structure of the resolvase catalytic domain have been deposited in the Protein Data Bank under accession codes 1HX7 and 1GHT, respectively.

Acknowledgments This research was supported by grants GM48607 and GM52738 from the National Institutes of Health (to G.P.M.) and by postdoctoral fellowship grants GM18942 and GM18956 from the National Institutes of Health (to B.P. and M.W.M., respectively).

References 1. Stark, W. M., Boocock, M. R. & Sherratt, D. J. (1992). Catalysis by site-speci®c recombinases. Trends Genet. 8, 432-439. 2. Grindley, N. D. F. (1994). Resolvase-mediated sitespeci®c recombination. In Nucleic Acids and Molecular Biology (Eckstein, F. & Lelley, D. M. J., eds), vol. 8, pp. 236-267, Springer-Verlag, Berlin.

1105 3. Hatfull, G. F. & Grindley, N. D. F. (1986). Analysis of gd resolvase mutants in vitro: evidence for an interaction between serine-10 of resolvase and site I of res. Proc. Natl Acad. Sci. USA, 83, 5429-5433. 4. Leschziner, A. E., Boocock, M. R. & Grindley, N. D. F. (1995). The tyrosine-6 hydroxyl of gd resolvase is not required for the DNA cleavage and rejoining reactions. Mol. Microbiol. 15, 865-870. 5. Rice, P. A. & Steitz, T. A. (1994). Model for a DNA-mediated synaptic complex suggested by crystal packing of gd resolvase subunits. EMBO J. 13, 1514-1524. 6. van de Putte, P. & Goosen, N. (1992). DNA inversions in phages and bacteria. Trends Genet. 8, 457-462. 7. Alonso, J. C., Ayora, S., Canosa, I., Weise, F. & Rojo, F. (1996). Site-speci®c recombination in grampositive y-replicating plasmids. FEMS Microbiol. Letters, 142, 1-10. 8. Abdel-Meguid, S. S., Grindley, N. D., Templeton, N. S. & Steitz, T. A. (1984). Cleavage of the sitespeci®c recombination protein gd resolvase: the smaller of two fragments binds DNA speci®cally. Proc. Natl Acad. Sci. USA, 81, 2001-2005. 9. Sanderson, M. R., Freemont, P. S., Rice, P. A., Goldman, A., Hatfull, G. F., Grindley, N. D. F. & Steitz, T. A. (1990). The crystal structure of the catalytic domain of the site-speci®c recombination Ê resolution. Cell, 63, enzyme gd resolvase at 2.7 A 1323-1329. 10. Liu, T., DeRose, E. F. & Mullen, G. P. (1994). Determination of the structure of the DNA binding domain of gd resolvase in solution. Protein Sci. 3, 1286-1295. 11. Rice, P. A. & Steitz, T. A. (1994b). Re®nement of gd resolvase reveals a strikingly ¯exible molecule. Structure, 2, 371-384. 12. Yang, W. & Steitz, T. A. (1995). Crystal structure of the site-speci®c recombinase gd resolvase complexed with a 34 bp cleavage site. Cell, 82, 193-207. 13. Hughes, R. E., Rice, P. A., Steitz, T. A. & Grindley, N. D. F. (1993). Protein-protein interactions directing resolvase site-speci®c recombination: a structurefunction analysis. EMBO J. 12, 1447-1458. 14. Hughes, R. E., Hatfull, G. F., Rice, P., Steitz, T. A. & Grindley, N. D. F. (1990). Cooperativity mutants of the gd resolvase identify an essential interdimer interaction. Cell, 63, 1331-1338. 15. Newman, B. J. & Grindley, N. D. F. (1984). Mutants of the gd resolvase: a genetic analysis of the recombination function. Cell, 38, 463-469. 16. Stark, W. M., Sherratt, D. J. & Boocock, M. R. (1989). Site-speci®c recombination by Tn3 resolvase: topological changes in the forward and reverse reactions. Cell, 58, 779-790. 17. Droge, P., Hatfull, G. F., Grindley, N. D. F. & Cozzarelli, N. R. (1990). The two functional domains of gd resolvase act on the same recombination site: implications for the mechanism of strand exchange. Proc. Natl Acad. Sci. USA, 87, 5336-5340. 18. Boocock, M. R., Zhu, X. & Grindley, N. D. F. (1995). Catalytic residues of gd resolvase act in cis. EMBO J. 14, 5129-5140. 19. Haffter, P. & Bickle, T. A. (1988). Enhancer-independent mutants of the Cin recombinase have a relaxed topological speci®city. EMBO J. 7, 3991-3996. 20. Klippel, A., Cloppenborg, K. & Kahmann, R. (1988). Isolation and characterization of unusual gin mutants. EMBO J. 7, 3983-3989.

1106 21. Klippel, A., Kanaar, R., Kahmann, R. & Cozzarelli, N. R. (1993). Analysis of strand exchange and DNA binding of enhancer-independent Gin recombinase mutants. EMBO J. 12, 1047-1057. 22. Arnold, P. H., Blake, D. G., Grindley, N. D. F., Boocock, M. R. & Stark, W. M. (1999). Mutants of Tn3 resolvase which do not require accessory binding sites for recombination activity. EMBO J. 18, 1407-1414. 23. Haykinson, M. J., Johnson, L. M., Soong, J. & Johnson, R. C. (1996). The Hin dimer interface is critical for Fis-mediated activation of the catalytic steps of site-speci®c DNA inversion. Curr. Biol. 6, 163-177. 24. Droge, P. (1992). Recombination of nicked DNA knots by gd resolvase suggests a variant model for the mechanism of strand exchange. Nucl. Acids Res. 20, 6159-6166. 25. Pan, B., Deng, Z., Liu, D., Ghosh, S. & Mullen, G. P. (1997). Structural characterization of gd resolvase in solution: Comparison of the wild-type enzyme, I110R mutant and C-terminal DNA binding domain. Protein Sci. 6, 1237-1247. 26. Pan, B. & Mullen, G. P. (1999). 1H, 15N, and 13C resonance assignments for the catalytic core of gd Resolvase. J. Biomol. NMR, 13, 307-308. 27. Koradi, R., Billeter, M. & WuÈthrich, K. (1996). MOLMOL: a program for display and analysis of macromolecular structures. J. Mol. Graph. 14, 51-55. 28. Maciejewski, M. W., Liu, D., Prasad, R., Wilson, S. H. & Mullen, G. P. (2000). Backbone dynamics and re®ned solution structure of the N-terminal domain of DNA polymerase b. Correlation with DNA binding and dRP lyase activity. J. Mol. Biol. 296, 229-253. 29. Artymiuk, P. J., Ceska, T. A., Suck, D. & Sayers, J. R. (1997). Prokaryotic 50 -30 exonucleases share a common core structure with gd resolvase. Nucl. Acids Res. 25, 4224-4229. 30. Holm, L. & Sander, C. (1993). Protein structure comparison by alignment of distance matrices. J. Mol. Biol. 233, 123-138. 31. Beese, L. S. & Steitz, T. A. (1991). Structural basis for the 30 -50 exonuclease activity of Escherichia coli DNA polymerase I: a two metal ion mechanism. EMBO J. 10, 25-33. 32. Derbyshire, V., Grindley, N. D. F. & Joyce, C. M. (1991). The 30 -50 exonuclease of DNA polymerase I of Escherichia coli: contribution of each amino acid at the active site to the reaction. EMBO J. 10, 17-24. 33. Mullen, G. P., Serpersu, E. H., Ferrin, L. J., Loeb, L. A. & Mildvan, A. S. (1990). Metal binding to DNA polymerase I, its large fragment, and two 30 ,50 exonuclease mutants of the large fragment. J. Biol. Chem. 265, 14327-14334. 34. Subramanya, H. S., Arciszewska, L. K., Baker, R. A., Bird, L. E., Sherratt, D. J. & Wigley, D. B. (1997). Crystal structure of the site-speci®c recombinase, XerD. EMBO J. 16, 5178-5187. 35. Gopaul, D. N., Guo, F. & Van Duyne, G. D. (1998). Structure of the Holliday junction intermediate in Cre-loxP site-speci®c recombination. EMBO J. 17, 4175-4187. 36. Guo, F., Gopaul, D. N. & van Duyne, G. D. (1997). Structure of Cre recombinase complexed with DNA in a site-speci®c recombination synapse. Nature, 389, 40-46. 37. Guo, F., Gopaul, D. N. & Van Duyne, G. D. (1999). Asymmetric DNA bending in the Cre-loxP site-

Structure of the Catalytic Domain of  Resolvase

38.

39.

40.

41.

42.

43.

44.

45.

46.

47.

48.

49.

50. 51.

52.

speci®c recombination synapse. Proc. Natl Acad. Sci. USA, 96, 7143-7148. Nanassy, O. Z. & Hughes, K. T. (1998). In vivo identi®cation of intermediate stages of the DNA inversion reaction catalyzed by the Salmonella Hin recombinase. Genetics, 149, 1649-1663. Hatfull, G. F., Noble, S. M. & Grindley, N. D. F. (1987). The gd resolvase induces an unusual DNA structure at the recombinational crossover point. Cell, 49, 103-110. Liu, T., Liu, D., DeRose, E. F. & Mullen, G. P. (1993). Studies of the dimerization and domain structure of gd resolvase. J. Biol. Chem. 268, 1630916315. Murley, L. L. & Grindley, N. D. F. (1998). Architecture of the gd resolvase synaptosome: oriented heterodimers identity interactions essential for synapsis and recombination. Cell, 95, 553-562. Weber, D. J., Gittis, A. G., Mullen, G. P., Abeygunawardana, C., Lattman, E. E. & Mildvan, A. S. (1992). NMR docking of a substrate into the X-ray structure of staphylococcal nuclease. Proteins: Struct. Funct. Genet. 13, 275-287. Bartels, C., Xia, T., Billeter, M., GuÈntert, P. & WuÈthrich, K. (1995). The program XEASY for computer-supported NMR spectral analysis of biological macromolecules. J. Biomol. NMR, 6, 1-10. Vuister, G. & Bax, A. (1993). Quantitative J correlation: a new approach for measuring homonuclear three-bond J(HNHa) coupling constants in 15Nenriched proteins. J. Am. Chem. Soc. 115, 7772-7777. DuÈx, P., Whitehead, B., Boelens, R., Kaptein, R. & Vuister, G. W. (1997). Measurement of 15N-1H coupling constants in uniformly 15N-labeled proteins: application to the photoactive yellow protein. J. Biomol. NMR, 10, 301-306. GuÈntert, P., Mumenthaler, C. & WuÈthrich, K. (1997). Torsion angle dynamics for NMR structure calculation with the new program DYANA. J. Mol. Biol. 273, 283-298. GuÈntert, P., Qian, Y. Q., Otting, G., MuÈller, M., Gehring, W. J. & WuÈthrich, K. (1991). Structure determination of the Antp (C39S) homeodomain from nuclear magnetic resonance data in solution using a novel strategy for the structure calculation with the programs DIANA, CALIBA, HABAS and GLOMSA. J. Mol. Biol. 217, 531-540. Mumenthaler, C. & Braun, W. (1995). Automated assignment of simulated and experimental NOESY spectra of proteins by feedback ®ltering and selfcorrecting distance geometry. J. Mol. Biol. 254, 465-480. Mumenthaler, C., GuÈntert, P., Braun, W. & WuÈthrich, K. (1997). Automated combined assignment of NOESY spectra and three-dimensional protein structure determination. J. Biomol. NMR, 10, 351-362. BruÈnger, A. T. (1992). X-PLOR Manual, Version 3.0, Yale University, New Haven. Farrow, N. A., Muhandiram, R., Singer, A. U., Pascal, S. M., Kay, C. M. & Gish, G. et al. (1994). Backbone dynamics of a free and phosphopeptidecomplexed Src homology 2 domain studied by 15N NMR relaxation. Biochemistry, 33, 5984-6003. Kamath, U. & Shriver, J. W. (1989). Characterization of thermotropic state changes in myosin subfragment-1 and heavy meromyosin by UV difference spectroscopy. J. Biol. Chem. 264, 5586-5592.

Structure of the Catalytic Domain of  Resolvase 53. Farrow, N. A., Zhang, O., Szabo, A., Torchia, D. A. & Kay, L. E. (1995). Spectral density function mapping using 15N relaxation data exclusively. J. Biomol. NMR, 6, 153-162. 54. Ishima, R. & Nagayama, K. (1995). Protein backbone dynamics revealed by quasi spectral density function analysis of amide N-15 nuclei. Biochemistry, 34, 3162-3171.

1107 55. Ishima, R. & Nagayama, K. (1995). Quasi-spectraldensity function analysis for nitrogen-15. J. Magn. Reson. Ser. B, 108, 73-76. 56. Lefevere, J.-F., Dayie, K. T., Peng, J. W. & Wagner, G. (1996). Internal mobility in the partially folded DNA binding and dimerization domains of GAL4: NMR analysis of the N-H spectral density functions. Biochemistry, 35, 2674-2686.

Edited by P. E. Wright (Received 13 February 2001; received in revised form 8 May 2001; accepted 15 May 2001)