The path of the DNA along the dimer interface of topoisomerase II.

The eukaryotic DNA topoisomerase II is a dyadic enzyme that, upon ATP binding, transports one duplex DNA (T-segment) through a transient double-stranded break in another (G-segment). The path of the T-segment involves the sequential crossing of three gates along the dimer interface: the entrance or N-gate, the DNA gate, and the exit or C-gate. Coordination among these gates is critical for dimer stability and the prevention of chromosome damage. This study examines DNA transactions by yeast topoisomerase II derivatives defective in gate function. The results indicate that, although the N-gate is not required for G-segment cleavage, the DNA gate per se is not able to widen unless ATP binds to the N-gate. Next, a captured T-segment cannot be held in the interdomainal region between the N-gate and the DNA gate. Finally, the G-segment can be religated while a T-segment is held in the central cavity of the enzyme between the DNA gate and the C-gate. These quaternary couplings for gate opening and closing suggest that topoisomerase II ensures a transient DNA gating state, during which dimer interface contacts are maximized and backtracking of the transported DNA is minimized.

DNA transactions, mostly replication and recombination, naturally knot and entangle the chromosomes. If such tangles are not removed when daughter chromosomes are pulled into newly forming cells, they lead to DNA breaks and cell death. Type II topoisomerases contribute to this removal; they cleave a DNA duplex, pass a second DNA duplex through the break, and religate the cleaved DNA (for recent reviews see Refs. 1 and 2). Biochemical and structural studies have led to a general model of how they transport one DNA duplex through another (reviewed in Ref. 3).
The eukaryotic type II topoisomerase (topo II) 1 is a homodimer with three distinct domain regions per subunit: an ATPase or N-terminal domain plus a DNA cleavage-religation core constituted by the BЈ and AЈ domains (reviewed in Ref. 4). This three-domain organization is identical to that of the heterotetrameric bacterial type II topoisomerases (DNA gyrase and topo IV) (5,6). Before interacting with DNA, topo II has the shape of an open clamp (see Fig. 1A), in which the N-terminal domains are located at the tips and the AЈ domains dimerize at the hinge (7,8). The double helix to be gated (named G-segment) first enters the open clamp and binds to the cleavage-religation core (9). The binding of ATP causes the N-terminal domains to interact, and the clamp closes (10). If this closure leads to the capture of a second double helix inside the clamp, a cascade of conformational changes ensues. The captured DNA (named T-segment) is transported through a transient double-stranded break in the G-segment (9) and is then expelled from the topoisomerase through the side opposite the one it entered by the AЈ-AЈ dimer interface at the hinge of the closed clamp (10,11). ATP hydrolysis and the release of the hydrolytic products allow the N-terminal domains to separate again, and the enzyme is ready for a new cycle.
The path of the T-segment along the dimer interface of topo II can be envisaged as the sequential crossing of three gates: the entrance or N-gate, the DNA gate, and the exit or C-gate. The N-gate, corresponding to the N-terminal domains, closes upon the binding of ATP or its nonhydrolyzable analogues, such as ADPNP (12,13). As the N-gate does not reopen when bound to ADPNP, the G-segment is clamped inside the topoisomerase. If such a G-segment is in a circular DNA molecule, a high salt-resistant protein-DNA complex is formed (9,14). The DNA gate corresponds to the DNA-protein platform constituted by the G-segment plus the active site for DNA cleavage and religation catalyzed by the BЈ and AЈ domains (15). Opening of the DNA gate requires, first, a chemical step and then, second, a mechanical step. The cleavage of both strands of the Gsegment is a trans-esterification in which a pair of tyrosyl residues, one in each AЈ domain, holds the severed 5Ј ends of the G-segment (16,17). Subsequently, the domain regions associated with the cleaved DNA ends have to move away from each other over 20 Å to permit the passage of the T-segment. The C-gate, corresponding to the AЈ-AЈ dimer interface near the C termini of the enzyme (15), opens transiently to allow the exit of the T-segment following its passage through the DNA gate (10,11).
Because each of the three gates of topo II constitutes a nonpermanent dimer interface, gate opening and closure must be tightly coordinated during DNA cleavage and transport. This study examines DNA transactions by three derivatives of the yeast Saccharomyces cerevisiae topoisomerase II defective in gate function. As depicted in Fig. 1B, an N-terminal-truncated topo II served as an N-gate-deleted enzyme. A topo II mutant, in which the pair of active site tyrosyl residues had been replaced by phenylalanines, provided an enzyme that could not cleave DNA and open the DNA gate. A topo II mutant, in which a pair of disulfide bonds form at the AЈ-AЈ dimer interface, was used to prevent the opening of the C-gate. The results revealed several quaternary couplings among the three gates by which topoisomerase II may ensure stability of the dimer when the G-segment is cleaved and unidirectional transport of the T-segment.

EXPERIMENTAL PROCEDURES
Topoisomerases-All the derivatives of the S. cerevisiae DNA topoisomerase II are described elsewhere. Their plasmid constructs originated from YEpTOP2-PGAL1, a multicopy plasmid used in the overex-pression of full-length yeast topoisomerase II (1429 amino acid residues) from an inducible yeast Gal1 gene promoter (18). The N-gatedeleted topo II was a 92-kDa fragment of the yeast topoisomerase spanning residues 410 -1202 (15). This fragment lacks the N-terminal or ATPase domain (residues 1-409) and a C-terminal region (residues 1203-1429), which is dispensable for DNA transport activity (19). The DNA gate mutant (YF) contained a Tyr 782 3 Phe substitution created by oligonucleotide-directed mutagenesis (20). The C-gate mutant topo II contained a double substitution, Ans 1043 3 Cys and Lys 1127 3 Cys, engineered by oligonucleotide-directed mutagenesis (11). Full-length S. cerevisiae DNA topoisomerase II and its derivatives were purified from the S. cerevisiae strain BCY123 harboring the corresponding expression vectors. Growth, protein over-expression by the addition of galactose, and cell disruption were executed as described by Worland and Wang (18). The enzymes were purified by phosphocellulose and Q-Sepharosecolumn chromatography (15). For purification of the C-gate mutant topo II, sulfhydryl reagents were omitted in all buffers. The purified enzymes were stored at a concentration of 5 mg/ml at Ϫ80°C.
DNA Substrates-For DNA transposition assays, a 3.1-kbp DNA fragment was obtained by linearization of a regular plasmid. A labeled 1.4-kb DNA fragment was generated by PCR amplification of the central region of the 3.1-kb fragment followed by 32 P end labeling using T4 polynucleotide kinase. For supercoil removal assays, a single negatively supercoiled topoisomer was prepared from a 2-kb pBR322 derivative. Supercoiled topoisomers of the 2-kb plasmid were fractionated by agarose gel electrophoresis (21), and an individual topoisomer with a specific linking number difference of about Ϫ0.05 was recovered from the gel slice and purified by phenol extraction and ethanol precipitation. For decatenation assays, the dimeric DNA catenane KS6 (N:S), consisting of a 32 P-labeled nicked circle singly linked to a supercoiled circle and containing distinct restriction sites in each, was constructed as follows. An EcoRI site of plasmid pKS5 (22) was converted into an XbaI site by linker insertion. The resulting plasmid, 7.2-kb pKS6, is a tandem dimer of two 3.6-kb sequences, differing in the presence of a single BamHI site in one half and a single XbaI site in the other half. Plasmid pKS6 was converted into a pair of 3.6-kb singly linked circles by Tn3 resolvase (22). Selective nicking of the circle containing the unique BamHI site was carried out by controlled digestion with BamHI in the presence of ethidium bromide (23). The singly nicked dimeric catenane was purified by density gradient centrifugation (24). Radiolabeling of the nicked circle with 32 P was done by nick translation.
Reactions, Filter Binding Assays, and Gel-blot Analysis-Reaction Buffer A contained 50 mM Tris⅐HCl (pH 8), 150 mM KCl, 8 mM MgCl 2 , 7 mM 2-mercaptoethanol, and bovine serum albumin (100 g/ml). Protein-mediated retention of DNA was carried out on Whatman GF/C glass fiber filters prepared and conducted as described previously (10,25). Following DNA electrophoresis, the gels were blotted to nylon membranes. The blots were exposed (either directly or after hybridization with 32 P-labeled DNA obtained by random priming) and quantified by phosphorimaging.

Dimer Stability during DNA Cleavage by Topoisomerase II
Lacking the N-gate Domain-Dimer stability of the N-gatedeleted topo II bound to DNA was examined in vitro by a DNA transposition assay (Fig. 2). A radiolabeled 1.4-kb DNA fragment and an unlabeled 3.1-kb DNA fragment were mixed at Exchange of topoisomerase II subunits linked to cleaved DNA is unlikely even when the N-gate domain is deleted. ϳ20 g of a 1.4-kb 32 P-labeled DNA fragment plus 20 g of a 3.1-kb unlabeled fragment were mixed with either 25 g of full-length or 25 g of N-gate-deleted topo II in a 40-l volume of reaction Buffer A. The mixtures were supplemented with 0 or 1 mM ATP. After 8 h of incubation at 30°C, 50 l of 2% SDS was mixed with each reaction. Proteinase K was added to a final concentration of 100 g/ml, and incubations continued for 1 h at 60°C. DNA was recovered by phenol extraction and ethanol precipitation and analyzed by agarose gel electrophoresis. The gel was stained with ethidium (left) and then transferred to a nylon membrane and exposed to visualize radiolabeled DNA (right). The presence of full-length (wt) or N-gate-deleted (⌬N) topo II and ATP for each reaction is denoted above each lane. Gel positions of the 1.4-and 3.1-kb fragments are indicated.
high concentration with stoichiometric amounts of either fulllength or N-gate-deleted topo II. Following 8 h of incubation at 30°C in the presence of 0 or 1 mM ATP, the reactions were quenched, and DNA fragments were examined by electrophoresis ( Fig. 2, left). Cleaved products of both DNA fragments were seen in the reaction containing full-length topo II (Fig. 2, left, lane 2) and increased in the reaction with ATP (lane 3). Cleaved products of both DNA fragments were also seen in the mixture containing the N-gate-deleted topo II (Fig. 2, left, lane 4), although cleavage was not affected by ATP (lane 5). When the gel was transferred to a membrane and the radiogram was obtained ( Fig. 2, right), only the labeled 1.4-kb fragment and its cleaved products were visible. Larger DNA molecules were not discernible in any reaction. Therefore, although the equilibrium level of DNA cleavage was comparable for the full-length and the N-gate-deleted topo II, DNA transposition (between labeled and unlabeled fragments), caused by the exchange of dimer halves linked to cleaved DNA, did not occur.
The N-gate Domain Is Required for Gating the Cleaved Gsegment-If, in addition to cleaving DNA, the N-gate-deleted topo II opened the DNA gate, passive transport of another DNA segment could eventually occur. This possibility was examined with DNA substrates in which DNA transport is energetically favorable and easy to detect (26). The N-gate-deleted topo II was assayed for the single linking number change of a supercoiled circle (Fig. 3A) and for one-step decatenation of a dimeric catenane (Fig. 3B).
A single linking number topoisomer of a 2-kb plasmid was mixed with full-length or N-gate-deleted topo II at molar ratios of enzyme per DNA of ϳ1. The reactions were supplemented with either ADPNP or ATP and incubated at 30°C for 8 h. Two-dimensional gel electrophoresis was used to discern linking number changes in the input DNA (Fig. 3A, lane 1). In the reactions containing full-length topo II (Fig. 3A, lanes 2-4), the linking number was altered only when ADPNP or ATP was added. With ADPNP (Fig. 3A, lane 3), single DNA transport events catalyzed in the direction of supercoil removal changed the linking number of most of the topoisomer by ϩ2 or ϩ4 units. With ATP (Fig. 3A, lane 4), enzyme turnover generated a distribution of topoisomers with linking number values differing in units of 2 near the relaxed state. However, the linking number did not change in any of the reactions containing the N-gate-deleted topo II (Fig. 3A, lanes [5][6][7] or in the reaction containing full-length topo II without nucleotide (lane 2).
The dimeric DNA catenane KS6 (N:S), consisting of a radiolabeled nicked circle singly linked to a supercoiled circle, was mixed with the full-length or the N-gate-deleted topo II at molar ratios of enzyme per catenane of ϳ1. The mixtures were incubated for 8 h at 30°C in the absence or presence of ADPNP or ATP. The reaction products were analyzed by gel electrophoresis. The gel position of the catenane was marked by the labeled nicked circle (Fig. 3B, lane 1). The nicked circle was released from the catenane by cutting the supercoiled circle with endonuclease XbaI (Fig. 3B, lane 2). In the reaction containing full-length topo II plus ADPNP (lane 4), single DNA transport events resulted in the unlinking of ϳ60% of the catenane. In reactions containing full-length topo II plus ATP (Fig. 3B, lane 5), enzyme turnover resulted in the complete unlinking of the catenane. In contrast, in the reaction lacking nucleotide (lane 3) and the reactions containing the N-gatedeleted topo II (lanes 6 -8), the circles were not at all unlinked. This and the above experiment demonstrate that although the N-gate-deleted topo II can cleave a G-segment, passive transport of another segment is unlikely to occur.
Capture  6 -8) in a 50-l volume of reaction Buffer A. Following the addition of either ADPNP (1 mM) or ATP (1 mM), the reactions were incubated for 8 h at 30°C and stopped by the addition of SDS to 0.5%. Proteinase K was added to 100 g/ml, and the mixtures were incubated for 30 min at 60°C before the samples were analyzed by gel electrophoresis. The gel was blotted and exposed. Hence, only the labeled nicked circle (depicted as linked or unlinked from the supercoiled ring) is visible in the autoradiogram. no E, no enzyme. incubated for 10 min at 30°C. Each reaction volume was then split into two halves. Endonuclease XbaI, which linearizes the supercoiled circle of the catenane, was added to one half. After 20 min of incubation at 30°C, NaCl at 1 M was added to all the reactions. The mixtures were then passed through glass fiber filters to retain high salt-resistant protein-DNA complexes. The protein-free DNA recovered from the filtrates and the protein-bound DNA recovered from the filters were examined by gel electrophoresis (Fig. 4).
In reactions with no topoisomerase added (Fig. 4, lanes 1-4), the dimeric catenane was found in the filtrate (lane 1). Freed nicked circles also appeared in the filtrate when the supercoiled circle of the catenane was linearized by XbaI (Fig. 4, lane  3). In reactions with full-length topo II (Fig. 4, wt, lanes 5-8), the nicked circle was unlinked from over 60% of the catenanes that reacted with topo II, and it was recovered in the filtrate (Fig. 4, lane 5). When the supercoiled ring was linearized, the nicked circle appeared mostly in the filtrate (Fig. 4, lane 7). These outcomes agreed with previous studies (10,11). The topoisomerase was bound to a G-segment in the supercoiled circle and, upon binding ADPNP, captured a T-segment located in the nicked circle with high probability. The T-segment was then transported and expelled through the C-gate. As a result, the unlinked nicked circle was found free in solution, whereas the supercoiled circle remained clamped by the topoisomerase.
In the reactions with the DNA gate mutant (Fig. 4, YF, lanes 9 -12), no unlinked circles were found in the filtrate (lane 9) or in the filter (lane 10). However, the YF mutant had closed the N-gate around one or both rings of the catenane because nearly 50% of it was retained in the filter (Fig. 4, lane 10). When the supercoiled circle was linearized, all of the nicked circle appeared in the filtrate (lane 11). No freed nicked circle was retained in the filter held by the topoisomerase (Fig. 4, lane 12). Hence, the YF mutant had closed the N-gate, but it did not capture the expected T-segment located in the nicked circle. This outcome indicated that a conformation in which a captured T-segment is entrapped between the closed N-gate and DNA gate domains is unstable or unfeasible.
The G-segment Can Be Religated before the T-segment Exits the Topoisomerase through the C-gate-Because a captured T-segment cannot be held between the N-gate and the DNA gate, it must readily move through the DNA gate toward the central cavity of the enzyme. In the experiments depicted in Fig. 5, it was examined whether the T-segment could be held in the central cavity of the topoisomerase between the DNA gate and the C-gate. The C-gate mutant topo II, in which the C-gate can be locked by a pair of disulfide links, was used to prevent the exit of the T-segment. The religation state of the G-segment was then examined to determine whether the DNA gate remained open or had closed.
The DNA catenane KS6 (N:S) was mixed with the C-gate mutant at a molar ratio enzyme per catenane of ϳ0.5. The  N:S)), were prepared. No enzyme (no E) was added to the first, 50 fmol of full-length topo II (wt) was added to the second, and 50 fmol of DNA gate mutant topo II (YF) was added to the third. The mixtures were supplemented with ADPNP at 1 mM. After 10 min of incubation at 30°C, each reaction was split into two halves, and endonuclease XbaI was added to one half. Following a 20-min incubation at 30°C, 50 l of NaCl (2 M) was added, and the reactions were passed through a glass fiber filter. DNA samples recovered from the filtrates (lanes F) and from the filters (lanes B), by washing them in 0.5% SDS, were analyzed by gel electrophoresis. As the gel was blotted and directly exposed, only the positions of the radiolabeled nicked circle (depicted as linked and unlinked from the supercoiled ring) were visible in the autoradiogram.

FIG. 5. The G-segment can be religated while the T-segment is in the central hole of the topoisomerase.
A, ϳ100 fmol of the DNA catenane (KS6 (N:S)) was mixed with 50 fmol of the C-gate mutant in 50-l volumes of reaction Buffer A containing either 0 or 7 mM 2-mercaptoethanol. After 5 min of pre-incubation at 30°C, the reactions were supplemented with ADPNP to 1 mM. Following 10 min of incubation at 30°C, 50 l of NaCl (2 M) was added, and each mixture was passed through a glass fiber filter. DNA recovered from the filtrates (lanes F) and from the filters (lanes B), by washing them in 0.5% SDS, was analyzed by gel electrophoresis. A mixture containing only the catenane is shown in lanes 1 and 2. Reactions containing the C-gate mutant (S:S), with (ϩ) and without (Ϫ) 2-mercaptoethanol (BME) added, were analyzed in lanes 3-6 as indicated above the lanes. The gel was blotted and directly exposed; therefore, only the radiolabeled nicked circle (depicted as linked and unlinked from the supercoiled ring) is visible in the autoradiogram. B, ϳ100 fmol of the DNA catenane KS6 (N:S) was mixed with 100 fmol of the C-gate mutant in a 150-l volume of reaction Buffer A without 2-mercaptoethanol. ADPNP to 1 mM was added. Following 10 min of incubation at 30°C, the reaction was split into three 50-l volumes. Endonuclease XbaI was added to one, endonuclease BamHI was added to another, and no endonuclease was added to the third volume. Following 30 min of incubation at 30°C, 50 l of NaCl (2 M) was added to all reactions, and the mixtures were passed through a glass fiber filter. DNA recovered from the filtrates (lanes F) and from the filters (lanes B), by washing them in 0.5% SDS, were analyzed by gel electrophoresis. The gel was blotted and probed with 32 P-labeled DNA. The positions of the catenane and of the nicked and supercoiled circles, either unlinked or linearized, are indicated. Reactions treated with endonucleases are indicated above the lanes. mixtures were pre-incubated either in the presence of 2-mercaptoethanol to reduce disulfide bonds at the C-gate or in the absence of the sulfhydryl reagent to keep the C-gate crosslinked. ADPNP was then added to promote single-step DNA decatenation. After 10 min of incubation at 30°C, NaCl at 1 M was added, and the reactions were passed through glass fiber filters to retain DNA circles trapped inside the topoisomerase. Protein-free DNA recovered from the filtrates and proteinbound DNA recovered from the filters were examined by gel electrophoresis (Fig. 5A). In the absence of the enzyme (Fig. 5A,  lanes 1 and 2), the catenane was recovered in the filtrate (lane 1). In the presence of the enzyme (Fig. 5A, lanes 3-6), decatenation took place, and the same amount of nicked circle was unlinked in the reactions with and without 2-mercaptoethanol. Yet, in the reaction containing the reducing agent (Fig. 5A,  lanes 3 and 4), the unlinked circles were found in the filtrate (lane 3). Conversely, in the reaction without reducing agent (Fig. 5A, lanes 5 and 6), the unlinked circles were retained by the enzyme in the filter (Fig. 5A, lane 6) because the T-segment could not exit through the C-gate.
To determine whether the G-segment was religated while the T-segment was inside the topoisomerase, a decatenation reaction in the absence of 2-mercaptoethanol was conducted as above. The molar ratio enzyme per catenane was ϳ1, and after the addition of ADPNP and 10 min of incubation, the reaction was split into three equal parts. Endonuclease XbaI, which linearizes the supercoiled circle of the catenane, was added to one part. Endonuclease BamHI, which linearizes the nicked circle of the catenane, was added to another part. No endonuclease was added to the third part. Following 30 min of incubation at 30°C, NaCl was added to a final concentration of 1 M, and the three mixtures were passed through a glass fiber filter. Protein-free DNA and protein-bound DNA were analyzed by gel electrophoresis (Fig. 5B). In the reaction not treated with endonucleases (Fig. 5B, lanes 1 and 2), the topoisomerase interacted with nearly 70% of the catenanes and, upon ADPNP binding, carried out decatenation with 60% probability. As before, the unlinked nicked circles were retained in the filter (Fig. 5B, lane 2), because they were held inside the topoisomerase with the C-gate locked. Remarkably, the supercoiled circles, which provided a G-segment for decatenation, were found fully religated in the filter fraction (Fig. 5B, lane 2). When the nicked circles were linearized with BamHI (lanes 3 and 4), all of them appeared in the filtrate (Fig. 5B, lane 3). Therefore, the entrapped T-segments had slid away from the central hole of the enzyme. When the supercoiled circles were linearized with XbaI ( lanes 5 and 6), all of them also appeared in the filtrate (Fig. 5B, lane 6). Hence, the G-segments gated during decatenation had also slid away from the closed topoisomerase. Consequently, the G-segment must have been religated while the T-segment was in the central hole of the enzyme. DISCUSSION The passage of the T-segment across the three gates of topoisomerase II implies that none of the dimer interfaces of the enzyme is permanent. Yet, the overall dimer must be firmly secured when the G-segment is cleaved and gated. Otherwise, dissociation or exchange of enzyme halves linked to cleaved DNA would cause chromosomal breaks or rearrangements. In this study, three different yeast topoisomerase II mutants were utilized to dissect quaternary couplings for gate opening and closing that control the DNA transport activity of the enzyme.
It has long been known that DNA transport by topoisomerase II is ATP-dependent. ATP usage is mandatory even when topological interconversions of DNA are energetically favorable. For DNA transport to occur, the G-segment has to be cleaved (a chemical step) and then gated (a mechanical step).
The BЈ and AЈ domains are sufficient to catalyze the cleavage step (15,27) but likely are insufficient to carry out the gating step. Mechanical widening of the DNA gate platform, involving the BЈ and AЈ domains plus the cleaved G-segment, is probably impaired until ATP binds to the N-terminal domains. Prior to ATP binding, either the closed conformation of the DNA gate per se is highly stable or, alternatively, the N-terminal domains play some role in suppressing the opening of the DNA gate. The experiments with an N-gate-deleted enzyme support the first alternative.
The DNA transposition assay with an N-gate-deleted enzyme demonstrates the endurance of topo II against subunit exchange when DNA is cleaved. The equilibrium level of DNA cleavage achieved by the N-gate-deleted dimer is comparable with that of the full-length topo II (in the absence of ATP). However, the absence of transposition indicates that the Ngate-truncated dimer rarely falls apart while DNA is cleaved. Such stability can be attributed to the main AЈ-AЈ dimer interface at the C-gate but also to a stable closure of the DNA gate. If the DNA gate were to open, passive transport of another DNA segment could then occur. The results show that passive transport of DNA is unlikely for the N-gate-deleted topo II. Although the G-segment can be cleaved, the DNA gate is either locked or does not open wide enough to permit the passage of another DNA segment. Therefore, binding of ATP to the Nterminal domains may not only serve up the capture of a T-segment (9) but may also unlock the DNA gate to open fully. This functional coupling could explain why DNA cleavagerejoining equilibrium is found altered as long as the N-gate is kept closed by ADPNP after DNA transport is completed (28,29).
Another functional dependence of the N-gate on the DNA gate is inferred from the YF mutant topo II, which cannot open the DNA gate. Upon nucleotide binding, this mutant closes the N-gate but does not capture a T-segment. The first implication of this result is that the closure of the N-gate does not require a transition state in which the G-segment has to be cleaved or gated. The second implication is that there is no space in which a captured duplex can be retained before the opening of the DNA gate. Structural studies have already indicated that a DNA double helix cannot be held in the dimer interface of the N-terminal domains. The crystal structure of the ADPNPbound N-terminal domain of S. cerevisiae topo II shows that the channel in this dimer interface is ϳ6 Å wide, insufficient to confine a DNA duplex (13). Also, the archway formed by the dimerized N-terminal domain of Escherichia coli DNA gyrase is nearly 20 Å wide, but it has constrictions that would prevent the accommodation of duplex DNA (30). However, there are no structural data available on how the N-terminal domains are connected to the BЈ domains in the DNA gate. Conceivably, the interdomainal region linking the N-gate and the DNA gate could be flexible enough to hold a captured duplex between the two gates. The present results preclude this last possibility. Consequently, the capture of a T-segment and the opening of the DNA gate must be mechanically coupled. Although ATP may be required for unlocking the DNA gate, the steric repulsion accompanying the entrapment of a T-segment by the Ngate may enforce the widening of the DNA gate.
After the T-segment passes through the DNA gate, it is promptly expelled through the C-gate (10,11). Before exiting the enzyme, the T-segment must have passed through the central hole of the dimer. This cavity has positive surface charges, and, when the DNA gate opens, it is enlarged (15). Hence, full entry of the T-segment into this space should be favored. Alternatively, the T-segment may remain in the vicinity of the open DNA gate until the opening of the C-gate allows it to exit. The experiments with the topo II mutant, in which the C-gate can be locked with disulfide bonds, examined this issue. The results show, first, that DNA transport probability is not affected and, second, that the DNA gate can close while a T-segment is trapped in the central hole of the dimer. These findings agree with that can be inferred from structural data. If the DNA gate and C-gate are closed, one DNA duplex could be accommodated, although tightly, in the central cavity of the enzyme (4,31). The fact that the G-segment is found entirely religated corroborates that the open conformation of the DNA gate must be thermodynamically unstable. Therefore, the DNA gate likely opens only when a T-segment is trapped by the N-gate, and even then it may close quickly once the T-segment has passed.
As to the question of what opens the C-gate, it is clear that it allows only the escape of the T-segment. The G-segment never escapes through the C-gate even if it is dissociated by high salt from a topoisomerase with the N-gate closed (9,10). Because the closure of the DNA gate reduces the central cavity, steric hindrance could drive the exit of the T-segment through the C-gate. It was suggested that conformational changes at a pair of hinge structures at the A:A dimer interface would lead to a transient opening of the C-gate (32). The results with the C-gate-locked enzyme suggested that the transition state, which leads to the opening of the C-gate, could be reached following the entrapment of the T-segment in the central cavity and the religation of the G-segment. Yet, because these experiments did not provide temporal information for the opening of the C-gate, it could not be excluded that the T-segment exits the enzyme before the G-segment has been religated. Fig. 6 summarizes a plausible pathway of conformational states for the three gates of topoisomerase II deduced from this study. One main inference is that quaternary couplings among the gates ensure the stability of the dimer during DNA cleavage. The various conformations of the enzyme-DNA complex seem to follow a "double lock" rule. Namely, a given gate cannot be open unless the other two are closed. Because the DNA gate and the C-gate may open only when they are triggered by the entrapment of a T-segment, transient opening of these gates would further maximize dimer interface contacts. The other main inference of this study is that quaternary couplings would enforce unidirectional transport of the T-segment along the dimer interface. Once the T-segment has reached the central cavity of the enzyme, backtracking of the T-segment is unlikely. If the DNA gate closes, reopening it by reverse transport of the T-segment could be energetically costly. Crystal studies indicate that widening of the DNA gate is coupled to a large rotation of the BЈ domains (32). However, even if the DNA gate were reopened, there would be no space to accommodate the returned T-segment between the N-gate and DNA gate. Unidirectional movement of the T-segment may ensure its entrapment in the central cavity of the enzyme, which may be needed to trigger the opening of the C-gate.
The transport of one DNA double helix through the complete dimer interface of topoisomerase II stands as a fascinating molecular mechanism, which demands energetically well balanced conformational states during the critical DNA cleavagereligation steps. Backtracking of the T-segment would cause unnecessary breakages of the G-segment and no net DNA transport. Dissociation of the enzyme halves linked to cleaved DNA would cause chromosomal breaks and rearrangements. This study clarifies how topoisomerase II maximizes dimer interface contacts during DNA cleavage and minimizes backtracking of the transported DNA. The conformational states inferred here may also provide new insights for rational design of chemotherapeutic agents targeting the DNA cleavage-religation equilibrium or the path of DNA along the dimer interface of topoisomerase II.  Fig. 1A. a, when the N-gate is open or deleted, the DNA gate and C-gate remain closed, and passive transport of DNA cannot occur. b, capture of the T-segment must be coupled to the opening of the DNA gate, because a duplex cannot be held between the closed N-gate and the DNA gate domains. c, when the T-segment reaches the central cavity of the enzyme, the DNA gate is able to close, and the G-segment can be religated. d, closure of the DNA gate and entrapment of the T-segment may trigger the opening of the C-gate. e, once the T-segment has escaped, the C-gate closes.