ABSTRACT
The high-resolution structure of the 43 kDa N-terminal fragment of the DNA gyrase B protein shows a large cavity within the protein dimer. The approximate size of this cavity is 20 Å , suggesting it could accommodate a DNA helix. Computer-modelling studies of this cavity suggest that it contains a constriction, reducing the width to ~13 Å, principally caused by the side chain of Arg286. We have used site-directed mutagenesis to alter this residue to Gln. Gyrase bearing this mutation shows virtually no supercoiling activity and near-normal relaxation and DNA cleavage activities. The mutated protein has ATPase activity which cannot be stimulated by DNA. These data support the proposed role of the 43 kDa domain as an ATP-operated clamp which binds DNA during the supercoiling cycle. The lack of DNA-dependent ATPase of the mutant may indicate that binding of DNA within the clamp is a prerequisite for stimulation of the ATPase activity.
DNA gyrase is the enzyme from bacteria which can introduce supercoils into closed-circular DNA using the free energy of ATP hydrolysis (1 ,2 ). The enzyme from Escherichia coli consists of two proteins, GyrA and GyrB, of molecular masses 97 and 90 kDa respectively; the active enzyme is an A2B2 complex. Gyrase is a member of a group of enzymes called DNA topoisomerases which are involved in the control of DNA topology (3 ). These enzymes can be divided into two types: type I enzymes catalyse reactions involving transient single strand breaks in DNA while type II enzymes break both strands of the duplex. DNA gyrase is a type II enzyme and shares significant sequence similarity with other enzymes of the group (4 ). All topoisomerases can catalyse the relaxation of DNA (the removal of supercoils) but DNA gyrase is the only enzyme which can also catalyse the introduction of supercoils into DNA.
The mechanism of DNA supercoiling by gyrase involves the following steps: (i) the binding of gyrase to DNA and the wrapping of a segment of DNA (~130 bp) around the A2B2 complex; (ii) cleavage of this wrapped DNA in both strands, involving the formation of covalent bonds between the 5'-phosphates at the break sites and Tyr122 of the GyrA subunits; (iii) passage of another segment of DNA through this break; (iv) resealing of the break. Catalytic supercoiling requires the hydrolysis of ATP but in the presence of the non-hydrolysable ATP analogue ADPNP (5'-adenylyl [beta],[gamma]-imidodiphosphate) limited supercoiling can be achieved, indicating that nucleotide binding will promote one round of supercoiling and that hydrolysis is required for the enzyme to turnover.
The gyrase proteins have been shown to be organised as functional domains (1 ,2 ). The GyrA protein consists of an N-terminal domain (59-64 kDa) which contains the DNA breakage-reunion site and has interactions with the quinolone drugs, and a C-terminal domain (33 kDa) which is thought to be involved in the wrapping of DNA around the A2B2 complex. The GyrB protein consists of an N-terminal domain (43 kDa) which hydrolyses ATP and binds the coumarin drugs, and a C-terminal domain (47 kDa) which binds to the A protein and DNA. The structure of the 43 kDa N-terminal domain complexed with ADPNP has been solved to 2.5 Å resolution by X-ray crystallography (5 ).
Gyrase is the target of a number of antibacterial agents many of which belong to the quinolone and coumarin classes. Quinolone drugs (e.g. oxolinic acid and ciprofloxacin) inhibit DNA supercoiling by interrupting the breakage and reunion of DNA by gyrase (6 -8 ). Incubation of gyrase and DNA in the presence of a quinolone drug and termination of the reaction with SDS leads to the double strand cleavage of DNA with the GyrA proteins covalently attached to the 5'-phosphates. Coumarin drugs inhibit supercoiling by preventing the hydrolysis of ATP by gyrase (6 ,7 ,9 ). The structure of an N-terminal sub-domain of GyrB (24 kDa) complexed with novobiocin has recently been determined (10 ). These data and recent binding experiments (11 ) support the idea that these drugs are competitive inhibitors of ATP binding.
Although the mechanism of DNA supercoiling by gyrase is understood in outline, many details are unclear. For example, it is not clear how a double-stranded segment of DNA is passed through the gyrase tetramer (A2B2) during the supercoiling reaction. The crystal structure of the 43 kDa ATPase domain complexed with ADPNP revealed the presence of a cavity between the two monomers of the protein dimer. As the diameter of this cavity is ~20 Å, it was suggested that it is likely to be a DNA-binding pocket that binds the DNA segment to be transported during the strand-passage reaction (5 ). Experiments with yeast topoisomerase II (topo II), the eukaryotic homologue of DNA gyrase, suggested that the addition of ADPNP to the enzyme prevents the binding of closed-circular DNA (12 ,13 ). This has been interpreted as the binding of nucleoside triphosphate closing an ATP-operated clamp (the equivalent of the 43 kDa N-terminal domain of GyrB) and preventing access of DNA into the interior of the enzyme. The crystal structure of a large internal fragment of yeast topo II (residues 410-1202; 92 kDa) has recently been solved (14 ). This does not include the ATP-operated clamp (residues 1-409), but the authors suggest that this domain will have a structure analogous to that of the 43 kDa domain of GyrB, and function to capture DNA during the strand-passage reaction. At the present time there is no direct evidence to support the idea that either the 43 kDa domain of GyrB or the corresponding domain of topo II can bind DNA. Gel-shift experiments with the 43 kDa GyrB domain in the presence and absence of ADPNP failed to detect any DNA binding (15 ,16 ). In this paper we have used computer modelling and site-directed mutagenesis to investigate the DNA-binding capabilities of the 43 kDa N-terminal domain of the DNA gyrase B protein.
The mutation of Arg286 to Gln in GyrB was carried out using overlap-extension PCR (17 ) as described by Jackson and Maxwell (18 ). The oligonucleotides used in the first-stage PCR reaction were: (i) 5'-GCGCCGACATCATAACGGTTCTGGC and (ii) 5'-CATC- GCCGCCTGGAAGCCTGC; and (iii) 5'-ACGATAGAAGAAGGTCAACAGCAGCG and (iv) 5'-GCAGGCTTCCAGGCGGCGATG, which produced fragments of 963 and 701 bp respectively, amplified from the gyrB gene present in plasmid pAG111 (19 ). In the second-stage reaction these fragments were incubated with oligonucleotides (i) and (iii) to generate a 1008 bp product which was cleaved with NcoI and AatII and ligated into pAG111 cut with the same enzymes. The DNA sequence of the region surrounding the site of the mutation (Arg286 to Gln) in the resultant plasmid (pAG286Q) was confirmed (450 bp in total).
The gyrase A and B proteins were purified as described by Hallett et al. (19 ). The mutant protein, GyrBR286Q, was purified as follows. Escherichia coli strain JM109[pAG286Q] was grown at 37oC in Luria Broth to an A595 of 0.4 with slow agitation, and protein expression was induced by the addition of isopropyl- [beta]-D-thiogalactopyranoside to a final concentration of 50 [mu]M. Incubation was continued for a further 12 h. The cells were harvested and resuspended in 50 mM Tris-HCl pH 7.5, 10% (w/v) sucrose and extensively sonicated, and the pellet collected by centrifugation. The pellet from the sonication step (containing GyrBR286Q) was washed five times in 1% (v/v) Triton X-100 and re-pelleted by centrifugation each time. The pellet was subjected to successive 15 min washes of 2, 4, 6 and 8 M urea in Enzyme Buffer (50 mM Tris-HCl pH 7.5, 100 mM KCl, 2 mM dithiothreitol, 1 mM EDTA, 10% w/v glycerol) and the pellet collected by centrifugation after each wash. The soluble fraction from each wash and the final pellet were examined by SDS-PAGE. The mutated protein was eluted exclusively in the 6 and 8 M urea washes. The protein solution was then dialysed extensively against Enzyme Buffer to re-fold the protein. The protein was found to be ~90% pure. Negatively-supercoiled and relaxed forms of plasmid pBR322 were prepared as described previously (20 ). Linear pBR322 was prepared by digestion of the supercoiled form with EcoRI. Supercoiled dimeric catenanes were prepared using plasmid pSS4 and Tn21 resolvase (gifts of Dr M. Oram, University of Bristol) as previously described (21 ).
DNA supercoiling reactions were carried out as described previously (22 ), under the following conditions: 35 mM Tris-HCl pH 7.5, 24 mM KCl, 4 mM MgCl2, 5 mM dithiothreitol, 6.5% (w/v) glycerol, 1.8 mM spermidine, 1.4 mM ATP, 0.36 mg/ml BSA, 9 [mu]g/ml tRNA, 10 [mu]g/ml relaxed pBR322 DNA; incubations were for 1 h at 25oC. Relaxation reactions were carried out under the same conditions except that supercoiled DNA replaced relaxed DNA, ATP was omitted and, where indicated, spermidine was omitted also. Oxolinic acid-induced cleavage of DNA by gyrase was carried out as described by Tamura et al. (23 ). Catenation and decatenation reactions, using monomeric pSS4 and pSS4-derived dimeric supercoiled catenane respectively, were carried out with minor modifications of previously published methods (24 ,25 ). ATPase reactions were performed using a linked enzyme assay as previously described (16 ). The binding of GyrB to a novobiocin-affinity column was carried out as described by Gilbert and Maxwell (26 ).
The 43 kDa domain of GyrB has been crystallised in the presence of ADPNP and the structure solved to a resolution of 2.5 Å (5 ). This domain contains the site of ATP hydrolysis and a number of novel features and protein folds. The most striking feature is the cavity formed by the dimer (Fig. 1 A) which is lined with several arginine residues (5 ). Since the diameter of the hole is roughly comparable with that of B-form DNA (~20 Å) it was proposed that the arginine residues form some kind of binding surface for DNA and hence that the cavity interacts with the passage helix during strand passage (5 ), although no direct evidence for this exists to date.
Aside from the above modelling information, sequence alignment data also show that Arg286 is a highly-conserved residue in GyrBs from different bacterial species, the only deviations being to lysine (4 ). Modelling studies suggested that a change to glutamine would cause minimal disruption to the structure. Site-directed mutagenesis was carried out using overlap-extension PCR and involved changing a CGT codon to CAG (the most abundant codon for Gln in GyrB). The mutation was moved into plasmid pAG111 which overexpresses GyrB (19 ) and the new plasmid was named pAG286Q. We found that cells carrying this plasmid grew slowly and tended to die upon addition of inducer. [This problem had been encountered previously with mutations of Glu42 of GyrB, the catalytic residue of the ATPase reaction (18 ); in this case it was mutations which abolished catalytic activity which exhibited this behaviour.] To make the GyrBR286Q protein we had to adjust the growth conditions from those reported previously (see Materials and Methods).
Mutant GyrB expressed from E.coli JM109[pAG286Q] was found to be insoluble, i.e., the protein is likely to be in inclusion bodies. The insoluble protein was solubilised in 6-8 M urea and remained in solution after dialysis into Enzyme Buffer. When this protein was complexed with wild-type GyrA the resultant complex was found to have ~0.5% of the DNA supercoiling activity of the wild-type complex (Fig. 2 ). In addition, the residual supercoiling activity of the mutant enzyme appeared to be largely distributive, as compared with the processive reaction of the wild-type enzyme.
The DNA gyrase B protein has an intrinsic ATPase activity which is stimulated by the addition of the A protein and DNA (36 -38 ). This ATPase activity is competitively inhibited by novobiocin and other coumarin drugs (10 ,11 ). As shown in Figure 4 , GyrBR286Q shows a novobiocin-sensitive ATPase activity which is ~60% that of wild-type GyrB. Addition of GyrA and DNA stimulates the ATPase of the wild-type protein but has no effect on the activity of the mutant protein (Fig. 4 ). Addition of GyrA or DNA alone do not affect the rate (data not shown). It could be argued that urea denaturation and subsequent refolding of the mutant protein leads to disruption of the ATPase activity of gyrase containing GyrBR286Q. However, when wild-type protein was treated in the same way the DNA-dependent ATPase activity was still apparent (data not shown).
X-ray crystallography and computer-modelling studies suggest that the dimer of the 43 kDa N-terminal domain of GyrB contains a channel which could accommodate a DNA double helix. However, this channel is not uniform in diameter and contains a constriction where the diameter is reduced to ~13 Å, caused principally by the side chain of Arg286. Coupled with the fact that this is a highly-conserved residue in GyrBs, we proposed that Arg286 might have a role in DNA binding during strand passage by DNA gyrase. To probe the role of this residue we mutated it to Gln. We found that GyrB bearing this mutation tended to be expressed in an insoluble form, an observation made previously with GyrBs having very low supercoiling activity (18 ). However, the protein could be resolubilised and remained in solution. The mutant protein (when complexed with GyrA) had very low supercoiling activity (~0.5% compared with wild-type; Table 0 ). It is possible that this residual activity was a consequence of contaminating wild-type GyrB expressed from the chromosomal gene. However, wild-type GyrB does not normally form inclusion bodies and can be found in the soluble fraction after cell lysis. Secondly, the supercoiling activity of the wild-type enzyme is generally highly processive in contrast with the distributive supercoiling reaction supported by GyrBR286Q (Fig. 2 ). Therefore we favour the low-level supercoiling being intrinsic to the mutant protein.
In contrast with the supercoiling results we found that GyrBR286Q could support near-normal relaxation activity, and DNA cleavage activity which was 20-25% that of wild-type. Catenation and decatenation activity, which like supercoiling are ATP-dependent, could not be detected. These results suggest that the mutant enzyme can support DNA cleavage and ATP-independent strand passage but cannot efficiently support ATP-dependent strand passage. We also found that GyrBR286Q had novobiocin-sensitive ATPase activity which was not stimulated by the presence of DNA.
Eukaryotic topoisomerase II shares significant sequence similarity with DNA gyrase. In particular the N-terminal region of the eukaryotic enzyme is closely related in sequence to the N-terminus of GyrB (4 ). It is known that this domain is responsible for the ATPase activity. Experiments by Roca and Wang (12 ,13 ) support a model in which the N-terminal domain of topo II (and by analogy the 43 kDa N-terminal domain of GyrB) is an ATP-operated protein clamp which is able to capture a segment of DNA during strand passage. In this model the enzyme binds two segments of DNA, the `gate' segment which is bound at the DNA breakage-reunion site of the enzyme, and the `transport' segment which is captured by the ATP-operated clamp prior to its passage through the gate. The experiments described in this paper can be interpreted in terms of this model. The low-level supercoiling activity can be attributed to the inefficient capture of the transport segment caused by the loss of the Arg286 side chain which would normally be involved with interacting with DNA bound in the clamp. DNA cleavage would be expected to be largely unaffected as it is a manifestation of the breakage of the gate segment. However, it has been proposed that binding of the transport helix stimulates the cleavage reaction of eukaryotic topoisomerase II (40 ). It is possible, therefore, that a gyrase mutant which captures the passage helix inefficiently could also show a reduction in DNA cleavage activity. It is not clear whether the modest reduction in cleavage activity supported by GyrBR286Q indicates coupling between cleavage and binding of the transport helix in the case of gyrase.
That DNA relaxation is also unaffected by this mutation implies that the relaxation reaction of gyrase does not require the clamp to operate in the same manner as during the supercoiling reaction. DNA supercoiling operates against the torsional stress of supercoiled DNA and is driven by the free energy of ATP hydrolysis. Therefore it is an active process whereby DNA has to be captured by the enzyme and prevented from slipping out of the clamp. It is possible that Arg286 is involved in holding the bound DNA. The DNA gyrase relaxation reaction is an ATP-independent reaction and, by implication, does not require DNA to be captured and held by the clamp. It can be regarded as a passive process in which DNA is allowed to pass through the enzyme, in the opposite direction to strand passage during supercoiling, driven by the free energy of the supercoiled DNA substrate. Therefore relaxation might be expected to be unaffected by a mutation which destabilises the binding of DNA in the clamp.
The fact that GyrBR286Q retains ATPase activity but that this activity is DNA-independent might suggest that the binding of DNA in the clamp is a prerequisite for DNA-stimulated ATPase activity. It has been previously reported that the stimulation of the gyrase ATPase activity by DNA requires the binding of DNA to at least two sites on the enzyme (38 ). In this and subsequent work (41 ) it was found that stimulation of the ATPase activity could be achieved by either a single DNA molecule of >100 bp or two molecules of <70 bp. It was not clear whether these sites are two identical sites in the two (AB) half molecules or whether one is the breakage-reunion site and the other a site for binding of the DNA segment to be translocated. The present experiments support the latter proposal by tentatively identifying the DNA-binding site within the ATP-operated clamp (the 43 kDa domain of GyrB) as one of the DNA-binding sites. We assume that the other site would be the DNA breakage-reunion site of the enzyme and that DNA molecules of >100 bp would be required to bridge these two sites. This mechanism for DNA-dependent stimulation of the ATPase activity would ensure that maximal ATP hydrolysis would only occur when DNA was present at both the gate and clamp sites.
A degree of caution must be exercised in the interpretation of the data with the GyrBR286Q protein. Any mutant protein produced by protein engineering has the possibility of being misfolded such that it loses catalytic activity due to this non-specific effect. In the case of GyrBR286Q the appearance of the protein in inclusion bodies might be indicative of misfolding. However, the protein could be readily re-solubilised and retained some catalytic functions. The relaxation, DNA cleavage and DNA-independent ATPase activities were near normal and it retained its ability to bind novobiocin. These observations would tend to argue in favour of the protein being correctly folded and the loss of DNA supercoiling and DNA-dependent ATPase activities being a consequence of the mutation at Arg286. Therefore we suggest that the properties of GyrBR286Q support the proposed role of the 43 kDa domain of GyrB as an ATP-operated clamp and are suggestive of the binding of DNA within this clamp being required for the DNA-dependent ATPase activity of DNA gyrase.
We thank Andy Bates and Niall Gormley for helpful comments. A.P.T. acknowledges BBSRC and Glaxo-Wellcome for a CASE studentship, and A.M. is a Lister Institute Jenner Fellow.
*To whom correspondence should be addressed. Tel: +44 116 252 3464; Fax: +44 116 252 3369; Email: ony@leicester
+Present address: School of Animal and Microbial Sciences, University of Reading, Whiteknights, PO Box 228, Reading RG6 6AJ, UK
REFERENCES
Return