| Nucleic Acids Research | Pages |
Structural analysis of a mutational hot-spot in the EcoRV restriction endonuclease: a catalytic role for a main chain carbonyl group
Introduction
Materials And Methods
Materials
Methods
Results
Crystal structures
Main chain carbonyl of Gln69
Molecular dynamics
Discussion
Acknowledgements
References
Structural analysis of a mutational hot-spot in the EcoRV restriction endonuclease: a catalytic role for a main chain carbonyl group
Received May 26, 1999; Revised and Accepted June 25, 1999
ABSTRACT Following random mutagenesis of the EcoRV endonuclease, a high proportion of the null mutants carry substitutions at Gln69. Such mutants display reduced rates for the DNA cleavage step in the reaction pathway, yet the crystal structures of wild-type EcoRV fail to explain why Gln69 is crucial for activity. In this study, crystal structures were determined for two mutants of EcoRV, with Leu or Glu at residue 69, bound to specific DNA. The structures of the mutants are similar to the native protein and no function can be ascribed to the side chain of the amino acid at this locus. Instead, the structures of the mutant proteins suggest that the catalytic defect is due to the positioning of the main chain carbonyl group. In the enzyme-substrate complex for EcoRV, the main chain carbonyl of Gln69 makes no interactions with catalytic functions but, in the enzyme-product complex, it coordinates a metal ion bound to the newly liberated 5[prime]-phosphate. This re-positioning may be hindered in the mutant proteins. Molecular dynamics calculations indicate that the metal on the phosphoryl oxygen interacts with the carbonyl group upon forming the pentavalent intermediate during phosphodiester hydrolysis. A main chain carbonyl may thus play a role in catalysis by EcoRV.
INTRODUCTION
The EcoRV endonuclease is a type II restriction enzyme that cleaves DNA specifically at its recognition sequence, GAT[darr]ATC (where [darr] marks the point of cleavage), in a reaction that requires only Mg2+ as a cofactor (1,2). It exists as a homodimer (2,3) and binds to the recognition sequence in a symmetrical fashion, so that its two active sites are juxtaposed against the two scissile bonds in the DNA, one in each strand (4). During the reaction at the recognition site, the DNA cleavage step is faster than the subsequent (rate-limiting) dissociation of the cleaved product (5). Consequently, EcoRV usually cleaves both DNA strands within one DNA-binding event (6). Under optimal reaction conditions, it cleaves its recognition sequence over 106 times faster than any other sequence (7,8). Yet under the same conditions lacking Mg2+, EcoRV binds all sequences with equal affinity and shows no preference for the recognition site (9-12). Ca2+ ions fail to support DNA cleavage by EcoRV but can mimic Mg2+ in promoting sequence-specific binding (13,14). The discrimination against DNA cleavages at sites other than the recognition site is thus achieved in the ternary EcoRV-DNA-metal ion complex rather than the binary EcoRV-DNA complex (13-15).
In the crystal structures of EcoRV bound to its recognition site in the absence of Mg2+ (4,16,17), the DNA lies in a cleft between the two protein subunits and is contacted primarily by two peptide loops from each subunit, though several other segments of the protein contact the DNA phosphates (18,19). The R (for recognition) loop, residues 182-188, lies in the major grove of the DNA and makes several hydrogen bonds to bases in the recognition sequence. The Q loop, so-called because it contains two glutamines, comprises residues 67-72. It interacts extensively with the sugar-phosphate backbone in the minor grove. When bound to EcoRV, the GATATC sequence is distorted from B-form DNA, with a marked bend towards the major grove (13,20), which places the scissile phosphodiester bond in the active site of the enzyme. In the crystal structure of EcoRV bound to non-specific DNA, the DNA is again located in the cleft between the subunits but it retains a B-form structure and fails to enter the active site (4). The R loop makes no contacts to the non-specific DNA but the Q loop still makes intimate contacts with the sugar-phosphate backbone. In the structure of the free protein without DNA, both R and Q loops are largely disordered (4). The Q loop is thus a conformationally mobile segment of polypeptide that can adjust to a variety of DNA structures.
The Q loop has been probed by random mutagenesis followed by selection for null mutants on the basis of viability in cells lacking the EcoRV methyltransferase (21). While many of the residues in the Q loop could be mutated to yield the null phenotype, mutations at Gln69 occurred at much higher frequency than at other positions. Mutants in which Gln69 had been replaced by Glu or Leu (Q69E and Q69L, respectively) were analysed further (22). In the absence of divalent metals, Q69L binds DNA with the same affinity and lack of specificity as wild-type EcoRV while Q69E binds less strongly. With Ca2+ present, both mutants bind DNA specifically in the same manner as native EcoRV. For Q69E, the rate of the DNA cleavage step in the reaction pathway is ~104 times lower than wild-type EcoRV while the cleavage step with Q69L is 100-fold slower than wild-type. Nevertheless, the reduced rate of phosphodiester hydrolysis by Q69L, relative to wild-type, results in the steady-state turnover of the mutant being rate-limited by this step rather than by product dissociation. Thus, Q69L has a steady-state kcat just 5-fold lower than wild-type, but its initial product is DNA cut in one strand, in contrast to the double strand breaks by wild-type EcoRV. In addition, Q69L requires higher concentrations of Mg2+ than wild-type and shows a sigmoidal dependence upon the Mg2+ concentration, indicating two metal ions per strand scission. Transient kinetics on Q69L also showed that its reaction involves a slow conformational change preceding DNA cleavage that has no equivalent with wild-type EcoRV (22).
However, the crystal structures of EcoRV do not explain why mutations at Gln69 should affect the DNA cleavage step in the reaction pathway. In the complexes of EcoRV bound to either specific or non-specific DNA (4), the side chain of Gln69 lies parallel to the sugar-phosphate backbone of the DNA and its methylene groups make van der Waals contacts to both the DNA and to other regions of the protein, but the carbonyl (CO) and amide groups at the end of the side chain do not make hydrogen bonds to the DNA. In addition, the main chain CO and amino groups of Gln69 do not seem to have any specific function in the enzyme-substrate complex. In contrast, the structure of EcoRV bound to the product from its DNA cleavage reaction shows the peptidyl CO of Gln69 coordinated to one of the two Mg2+ ions liganded to the 5[prime]-phosphate of the cleaved DNA (16). The interaction between the peptidyl CO and the metal ion is observed only in the crystal structure of the enzyme-product complex and not in the structures obtained after adding divalent metal ions to crystals of EcoRV bound to its recognition site. Both the metal ions and the scissile phosphate are positioned differently in the EcoRV-substrate-metal complex compared to the EcoRV-product-metal complex (see below; Fig. 4).
The objective of this study was to identify a role for Gln69 in catalysis by EcoRV. The Q69L and Q69E mutants were crystallised as complexes with specific DNA and their structures determined by X-ray crystallography. Since the diffraction data for the mutants were collected at 100K, valid comparisons between the mutant and wild-type structures also required the acquisition of structures for the wild-type enzyme bound to specific DNA at 100K. Further insights into possible roles for Gln69 were obtained by molecular dynamics (MD) modelling of the pentavalent intermediate during phosphodiester hydrolysis by EcoRV.
MATERIALS AND METHODS
Materials
Wild-type EcoRV and the Q69L and Q69E mutants (21) were purified as described previously (3,22). The oligodeoxynucleotide AAAGATATCTT was synthesised on a Millipore Expedite system with reagents from Cruachem (Glasgow, UK), purified by HPLC and annealed to form the duplex (23).
Methods
All crystallisations were by the sitting drop method. The protein (2 µl, 15 mg/ml, ~260 µM dimer) in 10 mM phosphate, pH 7.5, 250 mM NaCl, 1 mM EDTA, 0.1 mM dithiothreitol and 0.1% (w/v) sodium azide was mixed with the duplex form of AAAGATATCTT (2 µl, 3 mg/ml, ~460 µM) in 10 mM cacodylate, pH 6.0, and, in some instances, with CaCl2 at a final concentration of 5 mM. Solutions of PEG 4000 (0.7 µl) were added to yield final concentrations of PEG of 0.8-1.4% (w/v). Crystals of wild-type (±Ca2+) and mutant proteins grew in 5-6 days at 18°C. All of the crystals grew isomorphously in space group P1 with one enzyme-substrate complex in the asymmetric unit. The cryoprotectant solution used during data collection was the same as the crystallisation solution except that the concentration of PEG 4000 was 8.8% (w/v) and that glycerol was present at 23.4% (v/v). Diffraction data were collected at 100K at station PX7.2 of the CLRC synchrotron radiation source (Daresbury, Cheshire, UK) on a MAR30 image plate or in Bristol using CuK[alpha] radiation from a rotating anode and a 20 cm Mac-Science image plate and processed through Denzo/Scalepack (24). Cell constants and data statistics are given in Table 1. Data for Q69E were collected from two crystals: data for the other structures were from single crystals.
Table 1. Data and refinement statistics
| EcoRV with DNA | EcoRV with DNA + Ca2+ | Q69L with DNA | Q69E with DNA | |
| Space group | P1 | P1 | P1 | P1 |
| Cell constants | ||||
| a, b, c (Å) | 48.5, 48.6, 63.9 | 48.6, 49.4, 63.9 | 48.7, 49.2, 64.0 | 48.4, 48.8, 63.6 |
| [alpha], [beta], [gamma] (°) | 97.0, 108.9, 107.0 | 96.9, 108.7, 107.5 | 96.7, 108.9, 107.5 | 96.6, 108.9, 107.4 |
| Observed reflections | 73 981 | 158 169 | 113 353 | 148 523 |
| Unique reflections | 29 850 | 37 718 | 38 240 | 21 703 |
| Rmerge | 2.3% | 5.0% | 5.2% | 9.9% |
| Completeness (overall) | 95.1% (30.0-2.05 Å) | 91.9% (20.0-1.90 Å) | 92.7% (15.0-1.90 Å) | 96.5% (12.5-2.30 Å) |
| Completeness (outer shell) | 86.6% (2.21-2.05 Å) | 82.7% (2.05-1.90 Å) | 87.5% (1.99-1.90 Å) | 94.8% (2.38-2.30 Å) |
| Rcryst | 20.6% | 20.6% | 21.9% | 18.9% |
| Rfree | 28.2% | 27.0% | 27.9% | 28.2% |
| r.m.s.d. from ideal bond lengths (Å) | 0.012 | 0.010 | 0.013 | 0.011 |
| r.m.s.d. from ideal bond angles (°) | 1.6 | 1.7 | 1.3 | 2.3 |
| Number of water molecules | 357 | 350 | 358 | 257 |
| Other atoms | - | 2 Ca2+ | - | - |
The structures were solved by molecular replacement using AMoRe (25) with the PDB (Protein Data Bank, Brookhaven; 26) entry 1RVA (16) as the search molecule. Structures were refined with the CCP4 suite of programs (27), chiefly with Refmac (28). Electron density interpretation and model building used the program O (29). Waters were added through ARP (30) in three cycles with [sigma] cut-offs set at 3.5, 3.25 and 3.0, with cycles of refinement through Refmac between each cycle through ARP. Further refinement and rebuilding led to the removal of several water molecules due to high B-factors, lack of electron density and/or poor hydrogen bonding contacts. In the case of crystals obtained in the presence of Ca2+, the areas of density picked by ARP as being waters were changed to Ca2+ ions if all of the following criteria were met: larger than expected for a water molecule; visible at greater than 5[sigma] in an Fo - Fc map; lower than expected B-factors for waters in those positions; hexa-coordination to other ligands. Distances between the Ca2+ ions and coordinating waters were restrained to 2.56 Å and to protein side chain oxygens and DNA phosphate oxygens to 2.74 Å (31). This was followed by further refinement through Refmac. Refinement statistics are given in Table 1.
Superpositions were calculated using lsqkab (32) in the CCP4 suite (27). MD simulations of a sphere of radius 25 Å, centred on the active site of the B subunit, were carried out in Discover 2.97 under the Consistent Valence Force Field, as described previously (15).
RESULTS
Crystal structures
The following structures of EcoRV proteins bound to the EcoRV recognition sequence were determined from X-ray diffraction data collected at 100K: wild-type EcoRV (noted as ED); wild-type EcoRV in the presence of Ca2+ ions (EDC); a mutant EcoRV with Leu in place of Gln69 (Q69L); a mutant EcoRV with Glu in place of Gln69 (Q69E). In all four cases, the duplex oligodeoxynucleotide with the recognition sequence was the undecamer used by Kostrewa and Winkler (16), AAAGATATCTT. Though Q69E binds DNA weakly in the absence of Ca2+ (22), it still crystallised as a DNA-protein complex, presumably due to the high concentrations of both DNA and protein in the crystallisation drop. However, no crystals were obtained for the mutant proteins bound to DNA in the presence of either Ca2+ or Mg2+ ions.
The four structures reported in this paper have been deposited at the PDB, with the following entry codes (and structure factors): for ED, 1B95 (R1B95SF); for EDC, 1B94 (R1B94SF); for Q69L, 1B97 (R1B97SF); for Q69E, 1B96 (R1B96SF). A brief account of the overall architecture of the new structures and their similarities to previous structures for EcoRV from room temperature data (4,15,16) is given here. The two protein chains of the dimeric enzyme are labelled A and B and are numbered 2-245 from N- to C-termini (the N-terminal methionine in the gene sequence is not present in the mature protein; 33). The two DNA chains are labelled C and D and are numbered 1-11 from 5[prime]- to 3[prime]-termini, with the scissile bond between nucleotides 6 and 7 on both strands. The catalytic residues E45, D74 and D90 of the A chain act on the C strand and those of the B chain on the D strand. The two subunits are related by a non-crystallographic dyad and have numerous differences in conformational details.
The four structures have good geometries (Table 1) and are very similar to each other, with the protein C[alpha] atoms superimposing closely (Fig. 1). The side chains and the DNA, except around the relatively disordered regions (residues 15-19, 97-101, 143-153 and nucleotide 1), also overlay well. The structures of ED and EDC are at resolutions equal to or greater than the equivalent structures reported before, with greater completeness in the data, lower values of Rmerge and comparable values of Rcryst and (where available) Rfree. There are thus more water molecules in these structures, which results in more complete hydrogen-bonding networks on the surface of the protein and the DNA. Nonetheless, the current structure of wild-type EcoRV complexed with specific DNA in the absence of Ca2+ (ED) is essentially identical to that from Kostrewa and Winkler (1RVA; 16). An overlay of all 488 C[alpha] atoms in ED and 1RVA yields a root mean square displacement (r.m.s.d.) of 0.29 Å. The structures differ only in the rotation of some of the amino acid side chains, the position of some atoms in the Q loop and in the position and number of water molecules. This comparison shows that the structure is not affected by the temperature of data collection.
Figure 1. An overlay of the C[alpha] atoms in both peptide chains from the following structures of EcoRV proteins bound to specific DNA: wild-type EcoRV (ED), in red; wild-type EcoRV + Ca2+ (EDC), in green; the Q69L mutant, in blue; the Q69E mutant, in yellow. For clarity, the DNA has been omitted in all four cases.
EDC has two Ca2+ ions: one (visible at 7[sigma] in an Fo - Fc omit map) in the B subunit, octahedrally coordinated to Asp74, Asp90, the scissile phosphate and three water molecules, at a locus called the 90/74 site; the other (refined at 50% occupancy and visible at 5.5[sigma] in an Fo - Fc omit map) at the 90/74 site of the A subunit. In the previous structure determinations of EcoRV with DNA and Ca2+, the Ca2+ ions were either soaked into pre-existing crystals (16) or, as in this work, present during the crystallisation (17). The previous studies used higher concentrations of CaCl2 (30-40 mM) than that used here (5 mM), the latter being similar to the concentration used in DNA-binding studies by gel-retardation (13). This accounts for why EDC, unlike the earlier structures, shows incomplete occupancy of the 90/74 site in one subunit. In one model for the EcoRV-DNA-Ca2+ complex (17), several amino acids were excluded, due to high B-factors and/or lack of electron density. Nonetheless, the 466 C[alpha] atoms in this model superimpose on the equivalent atoms in EDC with a r.m.s.d. of 0.33 Å. There is also good agreement in the positions of most side chains.
A superposition of the C[alpha] atoms of ED and EDC (Fig. 1) yields a r.m.s.d. of 0.25 Å. The binding of Ca2+ to the complex of EcoRV with its recognition site thus has virtually no effect on the conformation of the protein or the DNA (not shown). The lack of divalent metal ions in the co-crystals with the mutant proteins is thus unlikely to have a significant effect on their structures.
The overall structures of the two mutant EcoRV proteins bound to specific DNA are also very similar to wild-type (Fig. 1). The C[alpha] atoms in ED overlay those in Q69E and Q69L with r.m.s.d. values of 0.25 and 0.24 Å, respectively. The similarity includes the DNA structure and the positions of most of the backbone and side chain atoms of the Q loop. In Q69E, the configuration of the Glu side chain at residue 69 follows that for the Gln in wild-type EcoRV. In both Q69E and ED, the terminal atoms of the side chain of residue 69 are within hydrogen bonding distance of two residues in the opposite subunit of the dimer: Thr37, both backbone nitrogen and side chain oxygen; Arg140, one or more of the side chain nitrogens. In Q69L, Leu cannot make these hydrogen bonds, but this has no detectable effect on the orientation of the side chain at position 69 or on either side chain or backbone positions of Thr37 and Arg140. However, either Glu or Leu at position 69 result in slightly closer van der Waals contacts between the methylene side chain and the deoxyribose backbone of the DNA than is the case with the wild-type Gln (not shown).
Main chain carbonyl of Gln69
Since the only discernible function for Gln69 is the interaction of its peptidyl CO with one of the two Mg2+ ions bound to the 5[prime]-phosphate in the enzyme-product complex (16), the enzyme-substrate complexes for both wild-type EcoRV and the two mutants were examined to see if the positioning of the main chain CO at residue 69 might account for the effect of the mutations. In one of the two subunits of the enzyme-substrate complexes for wild-type EcoRV, with and without Ca2+ (ED and EDC), and for Q69L, the electron density maps (Fig. 2) are consistent with equal proportions of two alternative configurations at residue 69: an `OUT' orientation where the main chain CO points away from the active site (Fig. 3a); an `IN' orientation where the CO is flipped through ~150° so that it points into the active site (Fig. 3b). In each of these three structures, the electron density for the opposite subunit of the dimer is consistent only with the OUT orientation for the main chain CO at residue 69, as in Figure 3a (Table 2). However, the enzyme-substrate complex for Q69E has the OUT orientation for the main chain CO in both subunits (Table 2). In contrast, the enzyme-product complex for wild-type EcoRV displays the IN orientation in both subunits (Fig. 3c). Indeed, the main chain CO at residue 69 must have the IN orientation if it is to interact with a Mg2+ bound to the 5[prime]-phosphate.
Figure 2. Stereo diagrams of the electron density maps for Leu69 and Asn70 in the B subunit of the Q69L mutant of EcoRV, showing the two alternative orientations for the main chain CO at residue 69. In each case, the 2Fo - Fc map (shown in blue) is calculated with coefficients from the accompanying coordinates. However, the corresponding Fo - Fc map (2[sigma] contour shown in red) demonstrates that the alternative conformation for the main chain CO is also present in the crystal. (a) The model with the main chain CO in the OUT orientation and (b) the model with the main chain CO in the IN orientation. The active site region is located on the left-hand side of these pictures.
Figure 3. Alternative conformations at the active site of EcoRV. (a and b) The two possible orientations for the main chain carbonyl of L69 and the side chain of N70, in the B subunit of Q69L bound to the duplex form of AAAGATATCTT. In (a), the main chain CO of L69 points away from the active site and the side chain of N70 points towards the recognition sequence (cf. Fig. 2a). In (b), the main chain CO of L69 points into the active site and the side chain of N70 points away from the recognition sequence (cf. Fig. 2b). (c) The enzyme-product complex for wild-type EcoRV (1RVC; 16), with the main chain CO of Q69 pointing into the active site and the side chain of N70 pointing away from the recognition sequence. All three panels show, in addition to residues 69 and 70: the active site residues, D90, E45 and D74; the TpA step (T6 and A7) cleaved by EcoRV; the thymidine immediately following the recognition sequence (T10). Also shown in (c) are two Mg2+ ions (purple crosses).
Figure 4. Structures for catalysis by EcoRV. (a) The active site in the B subunit from the crystal structure of EcoRV bound to its recognition sequence and to two Co2+ ions (D.Kostrewa and F.Winkler, personal communication). (b) The MD model for the active conformation of EcoRV, derived from the crystal structure in (a) (15; for MD, the Co2+ ions were converted to Mg2+). (c) The MD model for the pentavalent intermediate during phosphodiester hydrolysis, derived from the scheme in (b) utilising the water noted in (b) between the right-hand Mg2+ and E45. (d) The active site in the B subunit from the crystal structure of EcoRV bound to the products of its DNA cleavage reaction and to two Mg2+ ions (1RVC; 16). All four stereo diagrams show: the TpA step (T6 and A7) cleaved by EcoRV; the active site residues, D90, D74 and E45; Q69; two metal ions (purple crosses).
Table 2. Orientation of the main chain carbonyl at residue 69 and the side chain of Asn70
| Main chain CO of Residue 69 | Side chain of Asn70 | |||
| Subunit A | Subunit B | Subunit A | Subunit B | |
| ED | IN/OUT | OUT | IN | OUT |
| EDC | OUT | IN/OUT | IN | IN |
| Q69L | OUT | IN/OUT | IN | OUT |
| Q69E | OUT | OUT | IN | IN |
| 1RVC | IN | IN | OUT | OUT |
The overall configuration of the peptide backbone in the Q loop is largely unaffected by the flip in the orientation of the main chain CO at residue 69 (Fig. 2). Nevertheless, the two alternative configurations for the peptidyl CO at 69 are generally accompanied by an alternation in the position of the side chain of the adjacent residue, Asn70 (Table 2). In most cases where the main chain CO at 69 points away from the active site, such as in both subunits of Q69E, the side chain of Asn70 enters the DNA minor groove (Fig. 3a) and makes a sequence-specific hydrogen bond with the O2 carbonyl group of the cytosine at the 3[prime]-end of the recognition sequence (4; not shown). Conversely, in most cases where the main chain CO at 69 points into the active site, such as in both subunits of the EcoRV-product complex, a rotation around the C[beta]-C[gamma] bond of Asn70 leaves its side chain inappropriately positioned for any base-specific contact to the recognition sequence (Fig. 3b and c). Instead, the side chain of Asn70 then makes van der Waals interactions with the deoxyribose of the first nucleotide after the recognition sequence.
Molecular dynamics
Phosphodiester hydrolysis by EcoRV proceeds with inversion of configuration at the target phosphorus and thus involves the direct attack of a water molecule in line with the 3[prime]-leaving group, presumably via a pentavalent intermediate (34). To investigate whether the position of the main chain CO at residue 69 affects catalysis by EcoRV, molecular mechanics were used to generate a model of the pentavalent intermediate (Fig. 4).
In the crystal structures of EcoRV bound to its DNA substrate in the presence of Ca2+, one Ca2+ ion is found in each subunit, at the 90/74 site between Asp90 and Asp74 and the proS oxygen of the scissile phosphate (16,17; this study, EDC). In contrast, both the BamHI and BglI endonucleases incorporate two Ca2+ ions per active site (35,36). However, when both Ca2+ and Mn2+, or just Co2+ ions, are added to EcoRV-DNA crystals (16), metal ions bind to both the 90/74 site and to an additional site between Asp74 and Glu45, the 74/45 site (Fig. 4a). Both wild-type EcoRV and Q69L need at least two metal ions per catalytic event (15,22,23,37), but the metal at the 74/54 site is ~5 Å away from the scissile phosphate. Yet no DNA cleavage occurs in the EcoRV crystals with two metal ions per active site (16), even though Co2+ gives almost the same rates as Mg2+ (15), while the combination of Ca2+ and Mn2+ gives higher activity than Mn2+ alone (37). The crystal structure of the EcoRV-substrate complex (Fig. 4a) is of an inactive conformation.
A model for the active conformation was derived previously by applying MD to the crystal structure of the EcoRV-DNA-Co2+ complex (15; the structure used as the start point for MD already had the IN orientation for the main chain CO at 69). During the MD trajectory, a spontaneous transition occurred to a new state: the pucker of the deoxyribose 3[prime] to the scissile bond flipped from C3[prime]-endo to C2[prime]-endo and the target phosphate moved 3.4 Å, to a position close to its location in the crystal structure of the enzyme-product complex (Fig. 4b). [The phosphate in the enzyme-product complex (Fig. 4d) is 4.5 Å away from its position in the enzyme-substrate complex (16).] After the transition, the metals at both the 90/74 and the 74/45 sites contact oxygens on the scissile phosphate: the proR oxygen replaces the proS moiety on the metal at the 90/74 site while the proS oxygen moves to the metal at the 74/45 site, but the latter metal is pulled partly out of its site so that Glu45 becomes an outer shell ligand. However, the inner shell water between Glu45 and the 74/45 metal is placed to attack the phosphorus (15).
To convert the scheme for the active conformation (Fig. 4b) into a model for the pentavalent intermediate (Fig. 4c), the water between the 74/45 metal and Glu45 was reconfigured in silico as an OH and a H and these were attached to the phosphorus at the scissile bond and the carboxylate of Glu45, respectively. The resultant structure was subjected to energy minimisation followed by MD for 200 ps. In the minimisation, the aberrant bond lengths created by the reconfiguration reverted to standard lengths. In the subsequent MD, the coordination geometry of the phosphorus changed from tetrahedral to trigonal bipyramidal in the first 15 ps of the trajectory and essentially no further changes in the structure occurred during the remainder of the trajectory. During these adjustments, the phosphorus at the scissile bond remained in the same position but the oxygens surrounding the phosphorus took up new positions as the co-ordination geometry of the phosphorus changed. This in turn re-positioned the metal ions attached to the phosphoryl oxygens. In particular, the metal attached to the proS oxygen, that had been near the 74/45 site, moved from an apical to an equatorial position that was ~6 Å away from Glu45, but which was now only 2 Å away from the main chain CO of Gln69, the correct distance for an inner shell ligand. The model for the pentavalent intermediate (Fig. 4c) thus has marked similarities to the crystal structure of the enzyme-product complex (Fig. 4d).
DISCUSSION
Random mutagenesis of EcoRV had identified Gln69 as a hot-spot for the generation of null mutants, which indicates that Gln69 plays a crucial role in DNA recognition and/or catalysis (21). The biochemical analysis of EcoRV proteins mutated at residue 69 had shown that the defect could be isolated primarily to the DNA cleavage step in the reaction pathway (22). The replacement of Gln69 by Glu has a graver effect on catalysis than the substitution with Leu (22). Hence, it might have been anticipated that the overall structure of the Q69E protein would display a greater perturbation from the wild-type structure than Q69L. However, the high resolution crystal structures of the DNA-protein complexes for Q69E and Q69L show that this is not the case. The overall structures of the complexes with the mutant proteins are very similar to each other and to wild-type EcoRV (Fig. 1), with respect to both the protein and the DNA. In addition, while the amide and CO groups at the end of the side chain of Gln69 are within hydrogen bonding distance of Thr37 and Arg140 in the opposite subunit of the dimer, Leu at this position cannot duplicate these interactions, yet the Gln to Leu substitution has no effect on the orientation of the side chain at residue 69 or on the positioning of Thr37 and Arg140. Moreover, the severe consequences of mutations at Thr37 (18) seem not to be due to interactions across the subunit interface (38). The effect of mutations at Gln69 cannot therefore be assigned to the loss of the functional groups at the end of the side chain or to any gross perturbation to either protein or DNA conformation.
Discrete perturbations were, however, observed among the structures analysed here in the orientation of the main chain CO at residue 69 (Fig. 3 and Table 2). As noted previously (16), the peptide backbone of the EcoRV protein has two alternate configurations at residue 69: either the OUT orientation, in which main chain CO points away from the active site (Fig. 3a), or the IN orientation, in which the main chain CO points into the active site (Fig. 3b). The two orientations for the main chain CO at residue 69 are usually accompanied by alternate conformations for the side chain of the adjacent residue, Asn70 (Table 2). In most cases where the main chain CO at 69 has the OUT orientation, the side chain of Asn70 enters the minor groove of the DNA and contacts the cytosine base at the 3[prime]-end of the GATATC recognition sequence for EcoRV. However, this sequence-specific contact is usually missing when the main chain CO at 69 has the IN orientation: instead, the side chain of Asn70 makes van der Waals contacts with the deoxyribose of the nucleotide immediately after the 6 bp recognition sequence. DNA recognition and catalysis by EcoRV might thus entail the following sequence of events: Asn70 is perhaps involved first in the recognition of the specific DNA sequence; at a later stage in the reaction pathway, after the enzyme has identified its recognition sequence, a segment of the Q loop undergoes a flip in its conformation, so as to bring the main chain CO of Gln69 into the active site where it plays a role in catalysis; by this stage, the base-specific contact from Asn70 is no longer required.
For wild-type EcoRV and the moderately active Q69L mutant, the different configurations at residues 69 and 70 are both populated in the crystal structures, which suggests that the alternate conformations of these two proteins have similar free energies. The reaction pathway for Q69L features a conformational change preceding DNA cleavage that is much slower than any equivalent process with the wild-type enzyme (22). Hence, the difference between Q69L and wild-type EcoRV may be the energy barrier between the conformational states, rather than the free energy per se, perhaps as a consequence of the more intimate van der Waals contacts between the Q loop and the DNA backbone with Leu at position 69 instead of Gln. In contrast, only the OUT configuration for the main chain CO at residue 69 is populated in the crystal structure of Q69E (Table 2). Q69E has virtually no DNA cleavage activity (22), so this result is consistent with the view that catalysis by EcoRV requires the IN orientation for the main chain CO at residue 69.
A model for the pentavalent intermediate during phosphodiester hydrolysis by EcoRV was generated by two cycles of MD simulations (Fig. 4b and c), starting from the crystal structure of the enzyme-substrate complex for EcoRV. The resultant model (Fig. 4c) duplicates several of the distinctive features seen in the crystal structure of the enzyme-product complex for EcoRV (Fig. 4d), but which are absent in the crystal structure of the enzyme-substrate complex (Fig. 4a). MD thus appears to have yielded a plausible structure for this short-lived intermediate. If so, then the main chain CO of Gln69 is not just the final resting place for the metal ion attached to the proS phosphoryl oxygen after the hydrolytic reaction, but is instead a coordination ligand for the key intermediate in the reaction. If the interaction between the main chain CO of Gln69 and the metal ion on the proS phosphoryl oxygen is established upon the formation of the pentavalent intermediate, as suggested here, then this interaction would stabilise the transition state of the reaction and thus enhance the catalytic rate. Such an interaction demands the IN orientation for the main chain CO at position 69.
Other proposals for the mechanism of phosphodiester hydrolysis by EcoRV lack roles for either the main chain or the side chain functionalities of Gln69 (39,40). However, such proposals do not lead directly to a structure for the EcoRV-product complex akin to the crystal structure of the enzyme-product complex. In contrast, the scheme proposed here provides a direct route from the crystal structure of an inactive form of the enzyme-substrate complex, in which no DNA cleavage occurs, to the crystal structure of the enzyme-product complex (Fig. 4). A rearrangement at the active site allows the scissile phosphate to interact with metal ions at both the 90/74 and the 74/45 sites (15), while the subsequent formation of the pentavalent intermediate leads to the transfer of the second metal onto the main chain CO of Gln69 (Fig. 4c).
Many enzymes employ main chain amino or CO groups for stereospecific interactions with their substrates or cofactors. The EcoRV restriction enzyme provides one example: the R loop makes numerous hydrogen bonds to the bases in the recognition sequence, but several of these bonds emanate from peptidyl amino or carbonyl groups (4). Other examples are provided by Ca2+-binding proteins, that often use main chain CO groups to coordinate Ca2+ ions (41). It is, however, substantially less common for peptidyl amino or CO groups to play direct roles in catalysis, as proposed here for the main chain CO of Gln69 in EcoRV. Nonetheless, there exist precedents for this behaviour. In both T4 endonuclease V and penicillin acylase, a key catalytic function is the N-terminal amino group of the polypeptide (42,43)
ACKNOWLEDGEMENTS
We thank John Barker, Mark Banfield, Louse Hancox and Neil Stanford for advice and discussions, Janet Baber for technical support, the staff of the Synchrotron Radiation Source at Daresbury for assistance and the BBSRC and the Wellcome Trust for financial support.
REFERENCES
*To whom correspondence should be addressed. Tel: +44 117 928 7429; Fax: +44 117 928 8274; Email: s.halford{at}bris.ac.uk
This page is run by Oxford University Press, Great Clarendon Street, Oxford OX2 6DP, as part of the OUP Journals
Comments and feedback: jnl.info{at}oup.co.uk
Last modification:
Copyright© Oxford University Press, 1999.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  A. Pingoud and A. Jeltsch Structure and function of type II restriction endonucleases Nucleic Acids Res., September 15, 2001; 29(18): 3705 - 3727. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 N. A. Gormley, A. J. Bath, and S. E. Halford Reactions of BglI and Other Type II Restriction Endonucleases with Discontinuous Recognition Sites J. Biol. Chem., March 15, 2000; 275(10): 6928 - 6936. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||