Mutational Analysis of the Preferential Binding of Human Topoisomerase I to Supercoiled DNA

Abstract

Human topoisomerase I binds DNA in a topology-dependent fashion with a strong preference for supercoiled DNAs of either sign over relaxed circular DNA. One hypothesis to account for this preference is that a second DNA binding site exists on the enzyme which mediates an association with the nodes present in supercoiled DNA. The failure of the enzyme to dimerize, even in the presence of DNA, appears to rule out the hypothesis that two binding sites are generated by dimerization of the protein. A series of mutant protein constructs was generated to test the hypotheses that the homeodomain-like core subdomain II (residues 233–319) provides a second DNA binding site, or that the linker or basic residues in core subdomain III are involved in the preferential binding to supercoiled DNAs. When putative DNA contact points within core subdomain II were altered or the domain was removed altogether, there was no effect on the ability of the enzyme to recognize supercoiled DNA as measured by both a gel-shift assay and a competition binding assay. However, the preference for supercoils was noticeably reduced for a form of the enzyme lacking the coiled-coil linker region or when pairs of lysines were changed to glutamic acids in core subdomain III. These results implicate the linker and solvent exposed basic residues in core subdomain III in the preferential binding of human topoisomerase I to supercoiled DNA.

Introduction

Type I DNA topoisomerases relax supercoils by introducing a transient single-strand break in the DNA. These enzymes are classified into type IA and type IB subfamilies based on the polarity of attachment to the cleaved DNA [1–3]. The members of the two subfamilies share no sequence homology and are further distinguished by their substrate requirements and mechanisms of relaxation. Type IA subfamily members require a single-stranded region to bind DNA, become attached to the 5′ end upon cleavage, and only relax negatively supercoiled DNA in the presence of divalent cations such as Mg²⁺. Escherichia coli DNA topoisomerase I is the prototype of the type IA subfamily. Type IB subfamily members bind double stranded DNA, become attached to the 3′ end of the cleaved strand, and relax both positive and negative supercoils. ATP or divalent cations are not required for the type IB enzymes, although Mg²⁺ and Ca²⁺ enhance the rate of relaxation [4].

The cleavage-religation reaction catalyzed by human DNA topoisomerase I, the prototypical type IB enzyme, is essential for many biological processes including DNA replication, transcription and recombination [2,3]. Strand cleavage is initiated by nucleophilic attack of the O-4 atom of the active site tyrosine on the scissile phosphate in the DNA, resulting in the covalent attachment of the enzyme to the 3′ end of the broken strand [2]. Rotation of the duplex region downstream of the break site relieves any supercoiling strain in the DNA prior to religation and release of the topoisomerase [5,6].

Human DNA topoisomerase I is composed of 765 amino acids and has a molecular mass of 91 kDa. Based on sequence comparisons, limited proteolysis studies, and the crystal structure of the enzyme [7,8], the following four domains have been identified in the protein: an NH₂-terminal domain (Met1-Gly214), a core domain (Ile215-Ala635), a linker domain (Pro636-Lys712), and a COOH-terminal domain (Gln713-Phe765) (Fig. 1). The NH₂-terminal domain is unstructured, poorly conserved, highly charged, and dispensable for the DNA relaxation activity in vitro. It contains nuclear localization signals and has been shown to interact with nucleolin, the SV40 large T antigen, p53, and possibly certain transcription factors [9–13]. Topo70 is atruncated form of human topoisomerase I that lacks residues 1–174 of the NH₂-terminal domain yet retains full enzymatic activity [7] and a preference for binding supercoiled DNA. The core domain is highly conserved and more protease-resistant than the other domains. The poorly-conserved linker domain is highly charged and forms an anti-parallel coiled-coil structure that connects the core domain to the COOH-terminal domain. The linker protrudes from the body of the protein and, instead of tracking with the axis of a bound DNA helix, angles away from DNA. The COOH-terminal domain is highly conserved and contains the active site Tyr723. When separately expressed, the COOH-terminal and core domains can associate in vitro to reconstitute wild type levels of enzymatic activity, showing that the linker domain is not required for activity [7,4,14].

Fig. 1.

Crystal structure of human topoisomerase I. Core subdomains I, II, and III are colored yellow, blue and red, respectively, with the linker and C-terminal domains colored orange and green, respectively. The Cap and Linkers regions are labeled along with the amino acid residues that were changed in the present study. Amino acids in core subdomain II (His266, Lys299 and Ser306) that were changed to glutamic acid in the combinations indicated in the text are shown in ball and stick and colored magenta. The three amino acids in the linker (Lys650/Lys654/Gln657) that were simultaneously changed to alanine are similarly depicted and colored brown. The four amino acids in the linker (Lys679/Lys682/Lys687/Lys689) that were simultaneously changed to serine are colored gray. Surface-exposed lysine residues in core subdomain III (Lys466/Lys468 and Lys545/Lys549) that were pairwise mutated to glutamic acid are colored black.

The co-crystal structure of human topoisomerase I with bound DNA indicates that the core domain can be further divided into three subdomains [8] (Fig. 1). Core subdomain I (residues 215–232, 320–433) and core subdomain II (residues 233–319) form the cap structure of the enzyme and cover one side of the DNA. Core subdomain III (residues 434–635) contains all the residues implicated in catalysis except Tyr723 and cradles the DNA on the side opposite of the cap [8,5]. Although there is little sequence similarity, the fold of the subdomain II is very similar to that of a homeodomain found in a family of DNA binding proteins. For example, residues 244 to 314 of the core subdomain II superimpose on the POU homeodomain of the Oct-1 transcription factor with an rms deviation of only 3.0 Å [15,8]. This observation suggests that core subdomain II, which forms part of the exposed Cap, could represent a second DNA binding site distinct from the substrate binding channel observed in the co-crystal structure. However, the conserved residues that are involved in base-specific contacts in the POU homeodomain are absent in core subdomain II of human topoisomerase I, suggesting that if subdomain II interacts with DNA, it does so with low affinity and likely without sequence specificity [15,16].

It has been proposed that topoisomerases relax the negative and positive supercoils generated by the translocation of an RNA polymerase along the DNA during transcription [17]. In support of this model, eukaryotic type IB topoisomerases have been found to associate with transcriptionally active genes and have been reported to interact directly with the transcription machinery [18–26]. Eukaryotic type IB topoisomerases have also been shown to provide the swivels for the relaxation of positive supercoils during DNA replication [27,20,28–31]. The mechanism for recruiting DNA topoisomerase I to transcriptionally active and replicating DNA remains unclear, but several labs have shown that the enzyme prefers to bind supercoiled over relaxed DNA [32–37]. Since the enzyme binds supercoiled DNA irrespective of the sign of the supercoils, Zechiedrich and Osheroff [36] hypothesized that topoisomerase I specifically binds at a node where two duplex regions of the supercoiled DNA cross and they provided electron microscopic evidence in support of this hypothesis [36].

The structural basis for the preferential binding of human topoisomerase I to supercoiled DNA is unknown, but if node recognition is important, then it seems likely that the binding involves an interaction with two regions of DNA at the point of crossing. One hypothesis to explain how the enzyme provides two DNA binding sites to stabilize an interaction with a DNA node is to suppose that it binds as a dimer as diagrammed in Fig. 2A. An alternative hypothesis is that in addition to the substrate binding channel identified in the crystal structure of the protein (Fig. 1) [8], there is a second DNA binding site present on the protein that stabilizes an interaction at a DNA node as shown schematically in Fig. 2B. Here we present the results of experiments designed to distinguish between these possible explanations for the preference of topoisomerase I for supercoils.

Fig. 2.

Alternative modes for topoisomerase I binding to a DNA node. Panel A: Node binding occurs through dimerization of topoisomerase I. Panel B: Node binding is mediated by two DNA binding sites on a single molecule of topoisomerase I.

Results

Human topoisomerase I does not dimerize in the absence or presence of DNA

We previously used a gel filtration assay to demonstrate that while a mutant form of topo70 missing a portion of the linker (topo70 ΔL) formed dimers through a domain swapping mechanism, no dimerization of WT topo70 was detectable under the same conditions [4,38]. Since these earlier experiments were carried out in the absence of DNA, we wanted to test whether dimers could form in the presence of DNA. Here we used a GST pull-down assay to ask whether topo70 that was already covalently bound to a DNA oligonucleotide could dimerize. GST-topo70 was incubated with free topo70 in the absence or presence of an oligonucleotide suicide substrate and any protein bound to GST-topo70 was collected by adsorption to glutathione S-Sepharose beads and analyzed by SDS-PAGE. Control experiments showed that free topo70 did not bind to either GST alone or to the beads (Fig. 3, lanes 6 and 7). Under the same conditions, no topo70 was found associated with the bead-bound GST-topo70 either in the absence or presence of DNA (Fig. 3, lanes 2 and 3, respectively). The slower migrating species of the doublet seen in lane 3 results from suicide cleavage and shows that approximately half of the GST-topo70 contained covalently bound oligonucleotide DNA. Thus, these results confirm our earlier finding that topo70 does not dimerize when free in solution and extend the results to show that even when bound to DNA after suicide cleavage, dimerization does not occur.

Fig. 3.

GST pull-down experiment to test for dimerization. The indicated combinations of GST-topo70, topo70, and GST were incubated with and without a suicide DNA oligonucleotide and mixed with glutathione Sepharose 4B beads (GSH beads). The beads were collected by centrifugation, washed and the samples were analyzed by SDS-PAGE. Lane 1, protein markers with sizes indicated in kilodaltons indicated along the left side of the gel. Lanes 4 and 5 contain GST-topo70 and topo70 size markers, respectively. The GST protein in lane 6 was run off the gel in this analysis. Although all of the samples were analyzed on the same gel, lanes with unrelated data were removed digitally at the places indicated by the thin vertical lines.

DNA binding properties of mutant proteins as measured by a gel shift assay

A structural alignment of core subdomain II of human topoisomerase I with the POU homeodomain of Oct-1 indicated that the residues making base specific contacts with the DNA in the homeodomain are not conserved in the core subdomain II. However, basic residues K25, R46 and R53 of the POU homeodomain that make hydrogen bonds with phosphates in the bound DNA correspond to residues His266, Lys299 and Ser306 in core subdomain II of human topoisomerase I (Fig 1). All three of these residues are conserved among known eukaryotic topoisomerase I sequences. To test whether these amino acids mediate an interaction with DNA that accounts for node binding by the enzyme, site-directed mutagenesis was used to replace these residues with glutamic acid in topo70. These changes would be predicted to disrupt an interaction with the DNA phosphate backbone, but have a minimal effect on the overall enzyme structure since all three are in a solvent exposed region. Since the assays to detect the preferential binding to supercoiled DNA require a catalytically inactive form of the protein [37], a mutation in the active site tyrosine (Y723F) was also introduced into the proteins. Topo70 capKS-E/Y723F and topo70 capHKS-E/Y723F were expressed and purified from recombinant baculovirus-infected insect SF-9 cells. Topo56/6.3 Y723F, a reconstituted form of the protein lacking the linker, the catalytically inactive Δcap [39], and topo31 were similarly purified for use in these assays (Fig. 4).

Fig. 4.

Human topoisomerase I fragments used in the DNA binding studies. (A) The four domains of full-length human topoisomerase I (topo I) are shown above the various constructs used in the binding studies: topo70, a 70 kDa NH₂-terminally truncated protein that starts with an engineered Met upstream of Lys175; topo58, a COOH-terminal deletion of topo70, ending at Ala659; topo31, a COOH-terminal deletion of topo70 ending at Ser433; Δcap, an NH₂-terminal truncation starting at Ser433; topo56/6.3, a reconstituted protein comprising the core domain from Lys175 to Thr639 and the COOH-terminal domain from Lys713 to the COOH terminus (Phe765). (B) SDS PAGE analysis of 2 μg of the indicated purified proteins. Lane 1, protein markers with sizes in kilodaltons indicated along left side of the panel; lane 2, topo70 Y723F; lane 3, topo70 capKS-E/Y723F; lane 4, topo70 capHKS-E/Y723F; lane 5, topo56/6.3 Y723F (6.3 kDa fragment of topo6.3 Y723F was run off the bottom of the gel); lane 6, Δcap; lane 7, topo31; lane 8, protein markers; lane 9, topo70 K466-468E/Y723F; lane 10, topo70 K545-549E/Y723F.

The various forms of the topoisomerase protein described above were mixed with an equimolar mixture of supercoiled, nicked circular, and linear pBluescript KSII (+) DNAs, and a gel shift assay [40–44] was used to analyze the preference of the proteins for the different topological forms of DNA. For the positive control protein, topo70 Y723F, the mobility of the supercoiled DNA was reduced with essentially no effect on the mobility of either the nicked or linear DNAs at the two lowest protein concentrations (Fig. 5A, compare lanes 2 and 3 with lane 1). As the amount of topo70 Y723F protein was increased, the supercoiled DNA was shifted further and, to a lesser extent, both the linear and nicked DNA bands became shifted as well (lanes 4 and 5). These results confirmed the earlier finding that topo70 Y723F has a preference for supercoiled over linear and nicked DNA [37]. Topo31, which corresponds to the Cap region of human topoisomerase I, provides a convenient nonspecific negative control for this analysis. As can be seen in Fig. 5A, lanes 22–26, all three forms of the plasmid DNA responded equally to increasing concentrations of the topo31 fragment, consistent with a lack of preference for one form over another. A higher concentration of topo31 was required to effect a gel shift, reflecting the lower affinity of the protein for DNA as compared to topo70.

Fig. 5.

DNA binding measured by an agarose gel shift assay. (A) Two-fold serial dilutions of the indicated proteins were incubated with equal amounts of supercoiled, linear, and nicked pBluescript KSII(+) plasmid DNA and analyzed by electrophoresis in an agarose gel as described in Experimental procedures. The mobilities of unshifted Supercoiled, Linear, and Nicked DNAs are indicated along the right side. Lanes 1, 6, 11, 16, 21 and 27, contain DNA alone; lanes 2–5 contain 0.88, 1.75, 3.5 and 7 pmol of topo70 Y723F, respectively; lanes 7–10 contain 0.88, 1.75, 3.5 and 7 pmol of topo70 capKS-E/Y723F, respectively; lanes 12–15 contain 0.88, 1.75, 3.5 and 7 pmol of topo70 capHKS-E/Y723F, respectively; lanes 17–20 contain 0.88, 1.75, 3.5, and 7 pmol of Δcap, respectively; and lanes 22–26 contain 0.88, 1.75, 3.5, 7 and 14 pmol of topo31, respectively. The white spaces demarcate separate gel analyses. (B) Same experimental design as panel A for the indicated proteins. Lanes 1, 6 and 11 are DNA alone; lanes 2–5 contain 0.88, 1.75, 3.5 and 7 pmol of topo70 Y723F, respectively; and lanes 7–10 contain 0.88, 1.75, 3.5 and 7 pmol of topo56/6.3 Y723F, respectively.

Both topo70 capKS-E/Y723F and topo70 capHKS-E/Y723F retained the preference for binding supercoiled DNA (Fig. 5A, lanes 7–10 and 12–15), ruling out Cap residues His266, Lys299, and Ser306 as contributors to the preferential binding to supercoils. To further test the possible involvement of the core subdomain II in the preferential binding to supercoiled DNA, the Δcap mutant lacking core subdomains I and II was also tested in the gel shift analysis (Fig. 5A, lanes 17–20). Δcap contains core subdomain III, the linker domain, and the COOH-terminal domain (residues 433–765) (Fig. 4A), and is catalytically inactive despite containing all of the residues known to be directly involved in catalysis [39]. At the lower concentrations of the Δcap protein, the supercoiled DNA was selectively shifted upon binding, although the magnitude of the shift was less compared to that observed with the topo70 protein (Fig. 5A, compare lanes 17–20 with lanes 1–5). This reduction in the shift most likely resulted from the two-fold lower affinity of the Δcap for DNA [39] and the lower molecular weight of Δcap (41 kDa) as compared to topo70 (71 kDa). Thus, deletion of the Cap region that includes subdomain II did not eliminate the preference for supercoiled DNA, indicating that core subdomain II is dispensable for the preferential binding of topoisomerase I to supercoils.

Although the band corresponding to the supercoiled DNA was selectively shifted in the presence of topo70 Y723F and all of the mutant proteins except topo31, we wanted to formally rule out the possibility that the proteins bound to the supercoiled, linear, and nicked DNAs equally well, but only the supercoiled DNA shift was detected visually because of its greater initial mobility. Therefore, the gel shift assay was repeated using topo70 Y723F or topo70 capHKS-E/Y723F that had been previously labeled with ³²P using protein kinase C. The autoradiograph of the agarose gel showed that the majority of the labeled proteins were associated with the shifted supercoiled DNA and that the amount of bound label correlated with the extent of the shift (Fig. 6, lanes 2, 3, 5 and 6). Furthermore, label was only associated with the nicked and linear DNAs at the protein concentration where a mobility shift of these species was also detected (lanes 3 and 6). These results validated the gel shift assay and confirmed that the selective shift of the supercoiled DNA band results from preferential binding.

Fig. 6.

Gel shift assay with ³²P labeled proteins. (A) Agarose gel shift assay as described for Fig. 5 using ³²P labeled topo70 Y723F and topo70 capHKS-E/Y723F. Lanes 1 and 4, DNA alone; lane 2 and 3 contain 1.75 and 3.5 pmol of topo70 Y723F, respectively; lanes 5 and 6 contain 1.75 and 3.5 pmol of topo70 capHKS-E/Y723F, respectively. (B) Autoradiogram of the gel shown in panel A. The mobilities of unshifted Supercoiled, Linear, and Nicked DNAs are indicated along the right side.

To further define the region that is involved in the preferential binding to supercoiled DNA, we repeated the assays using a form of human topoisomerase I reconstituted from a mixture of topo56 and topo6.3 Y723F (see Fig. 4A). This reconstituted protein contains only the core and COOH-terminal domains and completely lacks the linker region (Fig. 1). When tested in the gel shift assay, topo56/6.3 Y723F retained a preference for supercoiled DNA, but the preference was reduced when compared to that of the topo70 Y723F (Fig. 5B). For example, although only the supercoiled DNA was shifted by both topo70 Y723F and topo56/6.3 Y723F at the lowest protein concentration tested (Fig. 5B, lanes 2 and 7), at the higher protein concentrations where only the supercoiled DNA was shifted by topo70 Y723F, the reconstituted enzyme shifted the linear and nicked DNAs as well (Fig. 5B, in particular, compare lane 4 with lane 9). These results suggest that an intact linker region is necessary for the full manifestation of the preference for supercoiled DNA, but in its absence the enzyme can still distinguish to a limited extent a supercoiled from a nonsupercoiled DNA.

Competition binding assays

To verify these results by an independent method and to provide a more quantitative measure for the binding of the various proteins to supercoiled DNA, we employed a filter binding assay similar to the one we used previously [37]. Unlabeled nicked and supercoiled SV40 DNAs were used separately as competitors for the binding of ³H-labeled nicked SV40 DNA to catalytically inactive (Y723F) mutant forms of topo70. The competition assays were carried out for topo70 capHKS-E/Y723F and Δcap and the results were compared to those for topo70/Y723F. For all three proteins, the competition profile for the like competitor (nicked DNA) exhibited a half-maximum at the expected 1:1 ratio of competitor to labeled DNA (Fig. 7A, closed symbols), whereas only ~1/10 as much supercoiled competitor was required to reduce the binding of the labeled nicked DNA to the 50% level (Fig. 7A, open symbols). The competition profile of topo56/6.3 Y723F for the supercoiled DNA showed that the amount of supercoiled DNA needed to compete to the 50% level was about one-third as much as for the nicked DNA (Fig. 7B). These results are consistent with the gel shift assays and confirm that topo70 Y723F, topo70 capHKS-E/Y723F and Δcap have a strong preference for supercoiled DNA over nicked DNA, while the reconstituted topo56/6.3 Y723F lacking the linker has a reduced ability to discriminate supercoiled from nicked DNA.

Fig. 7.

Filter binding assays comparing unlabeled supercoiled and nicked SV40 DNAs as competitors for ³H-labeled nicked SV40 DNA binding to topoisomerase I constructs. (A) The results of the competition assay for topo70 Y723F (nicked competitor, solid squares; supercoiled competitor, open squares), topo70 capHKS-E/Y723F (nicked competitor, solid triangles; supercoiled competitor, open triangles), and Δcap (nicked competitor, solid diamonds; supercoiled competitor, open diamonds). (B) Results for the competition assay for topo56/6.3 Y723F (nicked competitor, solid circles; supercoiled competitor, open circles).

Since the above results implicate the linker in the preference for binding supercoiled DNA, we wanted to investigate whether the clusters of positively-charged amino acids in the linker region are required for this effect. To test this possibility, we generated two mutant forms of topo70 Y723F, each of which eliminates the positive charges associated with clusters of basic amino acids within one of the α helices of the linker region (α18). The changes in one of the mutant proteins were K650A/K654A/Q657A and in the second were K679S/K682S/K687S/K689S (Fig 1). These proteins are referred to as topo70 linkerKKQ-A/Y723F and topo70 linker4K-S/Y723F, respectively. When these proteins were used in the competition binding assay, the ratio of unlabeled supercoiled competitor to labeled nicked DNA that was required for half-maximal binding was offset from the ratio for the nicked or like competitor by the same amount for the mutants as for the topo70 Y723F protein (Fig. 8A). The magnitude of this offset was slightly less for the competition profiles in Fig. 8A as compared to what was observed in Fig. 7A because the preparation of unlabeled supercoiled competitor used in this experiment contained a slightly higher percentage of nicked molecules (~20% compared with ~5% previously, data not shown). Based on these results, we conclude that the absence of either of these two clusters of basic amino acid within the linker does not affect the ability of the protein to preferentially bind supercoiled DNA.

Fig. 8.

(A) Filter binding assays comparing unlabeled supercoiled and nicked SV40 DNAs as competitors for ³H-labeled nicked SV40 DNA binding to topoisomerase variants containing multiple amino acid changes in the linker domain: topo70 Y723F (nicked competitor, solid squares; supercoiled competitor, open squares); topo70 linker4K-S/Y723F (nicked competitor, solid triangles; supercoiled competitor, open triangles); topo70 linkerKKQ-A/Y723F (nicked competitor, solid diamonds; supercoiled competitor, open diamonds, dashed line). (B) Filter binding assays for topoisomerase variants containing mutations at exposed lysine residues in the core domain of the enzyme: topo70 Y723F (nicked competitor, solid diamonds, supercoiled competitor, open diamonds); topo70 K466-468E Y723F (nicked competitor, solid squares, supercoiled competitor, open squares); topo70 K545-549E Y723F (nicked competitor, solid triangles, supercoiled competitor, open triangles). For topo70 Y723F, the values plotted are the mean of 7 independent determinations and for the two mutant proteins, the values are the mean of 6 independent determinations.

The solvent exposed region of the core subdomain III distal from the Cap represents yet another region of the protein that might provide a binding interface for a second DNA binding site. To examine this possibility, we generated mutant proteins in which pairs of positively-charged lysine residues within core subdomain III were changed to glutamates (Fig. 1) and tested these proteins in the competition binding assay. As shown in Fig. 8B, the competition profiles of the nicked competitor DNA for the topo70 K466-468E/Y723F and topo70 K545-549E/Y723F proteins are identical to the profiles for the control topo70 Y723F protein (closed symbols), but importantly, the supercoiled DNA did not compete as well for the binding to the two mutant proteins as it did for the binding to the control topo70 Y723F protein (compare the open squares and triangles with the open diamonds). To be certain that these differences were significant, multiple experiments were performed to determine the mean value for the ratio of unlabeled nicked to supercoiled competitor required to reduce binding to the 50% level. For the positive control topo70 Y723F, this ratio was found to be 8.6 ± 3.9 (S.D.) (7 repeats), which is consistent with the earlier determinations, whereas the corresponding ratios for topo70 K466-468E/Y723F and topo70 K545-549E/Y723F were 4.1 ± 1.1 and 4.6 ± 1.7, respectively (6 repeats). By the T test, these differences of the ratios for the two mutant proteins from the control are significant at the P < 0.05 level and thus the mutant proteins have a reduced ability to discriminate supercoiled from nonsupercoiled DNA.

Discussion

Although protein-protein interactions have been implicated in targeting topoisomerase I to supercoiled substrates in vivo [21,24–26], when given a choice of supercoiled and relaxed substrates in the absence of other proteins in vitro, the enzyme exhibits a preference for binding to the supercoiled DNA [32–37]. Since this intrinsic preference for supercoils is independent of the sign of the supercoiling [37,45], it seems likely the DNA feature being recognized by the enzyme is a DNA node [36], a structural element that is shared by DNAs with positive and negative supercoils. In the absence of DNA, the topoisomerase I protein is bi-lobed structure that exists in an open clamp conformation [5]. Upon binding DNA, the clamp closes around the duplex to form a clearly-defined channel that interacts with the DNA backbone over a length of ~6 bp (Fig. 1) [8]. The simplest model to explain node recognition by the enzyme supposes that in addition to this well-characterized DNA binding channel, the protein has a second DNA binding region that stabilizes the interaction with a DNA crossing. Here we consider four structure-related hypotheses that could explain node binding. First, the bent structure of a supercoiled duplex could be a feature that is recognized by a single topoisomerase I protein without the need for a second DNA binding site. Second, a topoisomerase I homodimer could provide two DNA binding sites on the same protein molecule (Fig. 2A). Third, core subdomain II, which structurally resembles a homeodomain and is an exposed featured of the Cap (Fig. 1), could constitute a second DNA binding site on the protein. Fourth, clusters of basic residues in core subdomain III and the linker on the side of the protein distal from the Cap, could mediate DNA binding at a node.

For some proteins, the preference for binding to supercoiled DNA is related to the tendency of the proteins to cause DNA bending. For example HMG proteins [46–50,44] and the p53 protein [40–42,51,43] preferentially bind supercoiled DNA and it has been shown in both cases that the proteins bend DNA. Moreover, in the case of the HMG proteins, the DNA bending capacity correlates with the supercoiled DNA binding [50]. In the crystal structure of the human topoisomerase I-DNA complex, the 22 base pair DNA substrate does not show any bending deformation and is nearly a perfect B-shaped helix [8]. This observation suggests that the preference of human topoisomerase I for supercoiled DNA is not due to an attraction of the enzyme for bent DNA.

In an earlier study, we showed that the topo70ΔL form of human topoisomerase I missing part of the coiled-coil linker domain could form dimers through a domain swapping mechanism involving the core and COOH-terminal domains of the two subunits [38]. We hypothesized that the shortened linker in the mutant enzyme destabilized the interaction between the COOH-terminal and core domains, enabling the COOH-terminal domain of one protein to occupy its binding site in the core domain of the other protein and vice versa. Consistent with this suggestion, we were unable to detect dimerization of free WT enzyme containing the normal length linker [4,38]. However these results did not rule out the possibility that dimerization of the enzyme only occurs after the first molecule of enzyme is already bound to DNA. In this regard, it has been shown that a molecule of topoisomerase I that is covalently trapped on DNA after suicide cleavage recruits another molecule of enzyme to cleave ~13 bp upstream of the trapped enzyme [52]. Although the basis for dimerization in this case is unknown, this interaction between two enzyme molecules is unlikely to mediate node binding since the second molecule of enzyme is bound to the DNA immediately adjacent to the one already trapped on the DNA. For our GST pull-down assay, we deliberately chose an oligonucleotide too short to permit this type of side-by-side contact (total duplex length 14 bp) to assay for DNA-mediated dimerization. Importantly, under these conditions we show that a topoisomerase I molecule that is covalently bound to DNA after suicide cleavage does not bind another molecule of the enzyme. These results rule against the hypothesis that dimerization of topoisomerase I accounts for the preference of the enzyme for supercoiled DNA.

In previous studies that demonstrated a preference of topoisomerase I for supercoils, the full length enzyme was used [36,37]. Here we demonstrate that topo 70, a form of the enzyme missing residues 1–174 that constitute most of the N-terminal domain, also preferentially binds supercoiled over relaxed DNA. This observation rules out this portion of the N-terminus as a region of the enzyme that provides a second DNA binding site involved in node recognition.

In the present study, we tested whether the homeodomain-like region within the Cap of the enzyme (core subdomain II) constitutes a second DNA binding site on the enzyme that mediates the preference for supercoils (Fig. 2B). Alignment of the sequences of human topoisomerase I and the Oct-1 homeodomain revealed three amino acids within core subdomain II of the Cap that might be expected to interact with the negatively charged DNA backbone and form the basis for a second DNA binding site on the enzyme (His266, Lys299, and Ser306) (Fig. 1). Replacing all three of these residues with a glutamic acid residue or complete deletion of the Cap region (Δcap) had no effect on the ability of the resulting protein to preferentially bind supercoiled DNA when assayed by either a gel shift assay or a competition binding assay. These results rule out the hypothesis that an interaction with a node is mediated by a second DNA binding site localized to core subdomain II of the enzyme.

The results presented here with topo56/6.3 Y723F, a reconstituted enzyme completely missing the linker region, reveal that this form of the enzyme has a reduced preference for supercoiled DNA when compared to the wild type enzyme. In a study carried out prior to the availability of the co-crystal structure of topoisomerase I [8], we examined the substrate binding preference of topo58, a form of the protein now known to contain the core domain and one third of the linker region (residues 175–659) (see Fig. 4). At the time, we concluded that the binding properties of topo58 were similar to those of topo70 Y723F, but a reexamination of these older data [37] reveals that, like the reconstituted topo56/6.3 Y723F studied here, topo58 alone exhibits a reduced preference for supercoiled DNA. Together, these observations suggest that an intact linker region of the enzyme is necessary for the full manifestation of the preference for supercoils. It is noteworthy that the elimination of either of the clusters of basic amino acids within the linker region (Fig. 1) does not affect the preference of the enzyme for supercoiled DNA. Our interpretation of this finding is that the contribution of the linker to node binding relates to how the linker influences local protein structure rather than through the formation of a second DNA binding site that makes direct amino acid side chain contacts with the DNA backbone. In this regard, it is noteworthy that the linker region is not only remarkably flexible [53], but mutations that affect its flexibility can also influence the structure of the protein at distant sites [54].

Unlike the linker where the evidence rules out a direct interaction between basic amino acids and the DNA in node binding, mutational studies within core subdomain III indicate that reversing the charge on pairs of basic, surface-exposed amino acids (K466/K468 and K545/K549) (Fig. 1) has a significant impact on the preferential binding of the topoisomerase to supercoiled DNA. Notably, these lysine residues are conserved in the topoisomerase I protein in most higher eukaryotes. (Fig. 9). These results suggest that basic amino acids with core subdomain III contribute to node binding through direct contacts with the DNA. The observation that the pairwise mutation of these lysines to glutamic acid only partially eliminates the preference for supercoiled DNA suggests that other residues within this domain also contribute to the formation of a second DNA binding region in the protein. Taken together, the results presented here strongly support the node binding hypothesis to explain the preference of human topoisomerase I for supercoiled DNA [36].

Fig. 9.

Sequence alignment within core subdomain III of representative eukaryotic members of the type IB subfamily of topoisomerases. Human, Drosophila, Saccharomyces cerevisiae and vaccinia virus topoisomerase I sequences were aligned using ClustalW2 software provided online by the European Bioinformatics Institute (http://www.ebi.ac.uk/Tools/clustalw2/). Homology of the bacterial type IB enzymes to these eukaryotic members of the family was too weak for them to be included in the alignment. The key conserved active site residues Arg488 and Lys532 (human numbering) are marked with closed circles. The open circles identify the residues in the human enzyme (Lys466, Lys468, Lys545 and Lys549) that are implicated in the preferential binding to supercoils.

The related type IB topoisomerase from vaccinia virus also preferentially binds to node structures in duplex DNA [36,55]. In a recent study, it was found that the vaccinia topoisomerase binds cooperatively to DNA to form long filaments in a reaction that is nucleated by the formation of an intramolecular node on DNA [56]. Although it is not known whether the initial node binding event involves a monomer or dimer of the enzyme, if a monomer is sufficient for node binding, then a second DNA binding region must exist within the viral enzyme as we have suggested above for the human enzyme. If this were to be the case, it is noteworthy that the structural similarity between the human and vaccinia enzymes is confined to the region referred to as subdomain III in the human enzyme [57,58] and that the two of the residues in the human enzyme that we have implicated in node binding (Lys466 and Lys549) are conserved in the viral enzyme (Fig. 9). Thus it is conceivable that the structural basis for node binding by the two enzymes is similar.

Experimental Procedures

Generation of baculovirus constructs expressing mutant proteins

pFASTBAC1-topo70 K299E/S306E, pFASTBAC1-topo70 K299E/S306E/Y723F, pFASTBAC1-topo70 H266E/K299E/S306E, and pFASTBAC1-topo70 H266E/K299E/S306E/Y723F were generated as follows. The plasmid pGEX-topo70 [14] was the template for making site-directed mutations using the QuickChange^™ mutagenesis kit from Stratagene. A pair of oligonucleotides containing the nucleotide changes for replacing Lys299 and Ser306 with glutamic acid was used to generate pGEX-topo70 K299E/S306E. The resulting plasmid and another set of oligonucleotides that changed His266 to glutamic acid were similarly used to generate pGEX-topo70 H266E/K299E/S306E. Both pGEX-topo70 K299E/S306E and pGEX-topo70 H266E/K299E/S306E were digested with NdeI and NheI and the fragments that contain the point mutations were purified and used to replace the corresponding fragments in NdeI and NheI digested pFASTBAC1-topo70 [59]. The resulting constructs, pFASTBAC1-topo70 K299E/S306E and pFASTBAC1-topo70 H266E/K299E/S306E were used to generate baculoviruses with the Bac-to-Bac system (Invitrogen) according to the protocol provided by the manufacturer. Recombinant baculovirus infection of Sf9 cells was used to produce proteins referred to as topo70 capKS-E and topo70 capHKS-E, respectively. These same two pFASTBAC1 constructs were also digested with NdeI and PpuMI. and the fragments containing the mutations were purified by gel electrophoresis. The isolated fragments were used to replace the corresponding fragment of pFASTBAC1-topo70 Y723F [59] that had been digested with the same two restriction enzymes to generate pFASTBAC1-topo70 K299E/S306E/Y723F and pFASTBAC1-topo70 H266E/K299E/S306E/Y723F. The catalytically inactive proteins expressed in baculoviruses from these two constructs are referred to as topo70 capKS-E/Y723F and topo70 capHKS-E/Y723F, respectively.

Starting from pFASTBAC1-topo70, two sets of oligonucleotide pairs were used to introduce clustered mutations in the linker-coding region to produce pFASTBAC1-topo70 K650A/K654A/Q657A and pFASTBAC1-topo70 K679S/K682S/K687S/K689S using the QuickChange method as described above. Starting with these two pFASTBAC1 constructs, an oligonucleotide containing the codon change that generates the Y723F mutation was subsequently used to produce pFASTBAC1-topo70 K650A/K654A/Q657A/Y723F and pFASTBAC1-topo70 K679S/K682S/K687S/K689S/Y723F. The catalytically inactive versions of the two proteins produced in baculoviruses by these constructs are referred to as topo70 linkerKKQ-A/Y723F and topo70 linker4K-S/Y723F, respectively. Similarly, starting with pFASTBAC1-topo70/Y723F, oligonucleotide based mutagenesis was used to introduce two pairs of clustered mutations into the coding region for the core domain of topoisomerase I to produce pFASTBAC1-topo70 K466E/K468E/Y723F and pFASTBAC1-topo70 K545E/K549E/Y723F. The mutant proteins produced in baculoviruses by these constructs are referred to as topo70 K466-468E/Y723F, and topo70 K545-549E/Y723F, respectively. A construct of the topo70 K466E/K468E mutant containing the wild type Tyr723 codon (pFASTBAC1-topo70 K466E/K468E) was produced by replacing a NheI-EcoRI fragment with the corresponding fragment from a wild type clone. The Tyr723-containing version of pFASTBAC1-topo70 K545E/K549E was toxic to E. coli and therefore baculoviruses expressing this particular mutant protein could not be obtained.

Site-directed mutagenesis as described above was used to introduce a stop codon after residue 639 in pGEX-topo70. The resulting plasmid was digested with Eco0109I and NheI. The fragment containing the stop codon was purified and used to replace the corresponding fragment of pFASTBAC1-topo70 that had been digested with the same two restriction enzymes to generate pFASTBAC1-topo56. All the mutants were confirmed by dideoxy sequencing. The generation of Δcap and topo31 has been described previously [39]. The structures of the constructs described above are diagrammed in Fig. 4A.

Expression and purification of proteins

Expression and purification of GST-topo70 has been described previously [14]. Topo70, topo70 capKS-E, topo70 capHKS-E, topo70 capKS-E/Y723F, topo70 capHKS-E/Y723F, topo70 linkerKKQ-A, topo70 linker4K-S, topo70 linkerKKQ-A/Y723F, topo70 linker4K-S/Y723F, topo70 K466-468E, topo70 K466-468E/Y723F, topo70 K545-549E/Y723F and Δcap were expressed and purified as described previously for topo70 [7]. Topo56/6.3 Y723F was purified by the procedure previously described for topo58/6.3 [14]. The purification of top31 has been described [39]. SDS-PAGE analysis of the purified proteins is shown in Fig. 4B. The DNA binding assays were carried out with the various mutant proteins containing the Y723F inactivating mutation, but to ensure that the mutations did not affect the overall fold of the protein, the mutant proteins containing the active site Tyr723 were also purified and assayed for plasmid relaxation activity. In all cases tested, the proteins containing Tyr723 were found to retain near full WT activity. As explained above, we were unable to obtain the topo70 K545E/K549E mutant protein to test whether it was enzymatically active.

GST pull-down assay for dimerization

15 pmol of purified GST-Topo70 was incubated in the absence or presence of a two-fold molar excess of an oligonucleotide suicide substrate CP14/CL25 (5′ GAAAAAAGACTTAG/5′TAAAAATTTTTCTAAGTCTTTTTTC) [60] in 10 mM Tris-HCl, pH 7.5, 100 mM KCl, 1 mM EDTA, 0.1 mg/ml BSA for 60 min at 23°C. An equimolar amount of purified topo70 was added to the samples and the mixtures were incubated at 23°C for 30 min. Each reaction was added to a 15 μl packed volume of glutathione Sepharose 4B beads (Amersham Biosciences) and mixed by rotation for 2 h at 23°C. As a negative control, topo70 alone was incubated with either purified GST or the Sepharose beads. Reactions were centrifuged for 2 min at 10K rpm, and the supernatant containing unbound protein was discarded. The pelleted beads were washed one time with 1 ml of the same buffer, pelleted and resuspended in SDS gel loading buffer, and boiled. The samples were fractionated with size standards by 8% SDS-PAGE as previously described [39]. Proteins were visualized by Coomassie blue staining and photographed using the AlphImager IS-2200 (Alpha Innotech) digital imager.

Generation of nicked and linear DNA

Unlabeled and ³H-labeled supercoiled SV40 DNA, prepared as previously described [37], and pBluescript KSII(+) DNA were relaxed by topo70 in Relaxation Buffer (150 mM KCl, 1 mM EDTA, 1 mM DTT, 10 mM Tris-HCl, pH7.5, 50 μg/ml bovine serum albumin), and phenol/chloroform extracted before ethanol precipitation. The relaxed DNAs (100 μg/ml) were treated with BamH1 (1000 units/ml) in a buffer containing 150 mM NaCl, 10 mM Tris-HCl, Ph 7.9, 10 mM MgCl₂, 1 mM DTT, 100 μg/ml bovine serum albumin, and 100 μg/ml ethidium bromide. After incubation at 37°C for 1 h, the majority of the DNA was nicked by the restriction enzyme under these conditions [61]. The reaction was phenol/chloroform extracted, ethanol precipitated, and stored in TE buffer (10 mM Tris-HCl, pH 8.0, 1 mM EDTA) at 4°C. Linear DNAs were generated by digestion with BamH1 under the same conditions as above without the addition of ethidium bromide.

Labeling proteins with protein kinase C

50 pmol of the indicated mutant topoisomerase I proteins were incubated with 20 ng of protein kinase C (Upstate Biotechnology, Inc.) and 20 μCi of [γ-³²P]ATP (3000 Ci/mmol) in 20 μl of Labeling Buffer (20 mM Hepes, pH 7.4, 10 mM MgCl₂, 0.5 mM CaCl₂, 50 ng phosphatidylserine, 2 μl diacylglycerol). The reactions were incubated at 30°C for 30 min in amber Eppendorf tubes because of the light sensitivity of both phosphatidylserine and diacylglycerol. Labeled proteins were stored at 4°C.

Agarose gel shift assay

Equal molar amounts of supercoiled, linear, or nicked pBluescript KSII (+) DNA (0.08 pmol of each) were incubated with the indicated amounts of the various protein constructs in 10 μl of Reaction Buffer (10 mM Tris-HCl, pH 7.5, 50 mM KCl, 1 mM EDTA, 1 mM DTT) at 23°C for 20 min. 2.5 μl of 50% glycerol was added to the reaction before loading on a 1% agarose gel. The gel was run with 0.5 X TBE Buffer (45 mM Tris base, 45 mM boric acid, 1 mM EDTA) at 4°C. Bands were visualized with a UV illuminator after staining with ethidium bromide. For the gel shift assay with labeled proteins, the gel was stained with ethidium bromide, visualized under UV illuminator, and then dried before exposure to film.

Filter binding assay

Filter binding assays were carried out by incubating 0.08 pmol of nicked ³H-labeled SV40 DNA with 0.32 pmol of topo70 Y723F, topo70 capKS-E/Y723F, topo70 capHKS- E/Y723F, topo70 linkerKKQ-A/Y723F, topo70 linker4K-S/Y723F, or 0.64 pmol of Δcap in 10 μl of Reaction Buffer at 23°C for 20 min. 10 μl of Reaction Buffer containing the indicated amounts of either unlabeled supercoiled or unlabeled nicked SV40 DNA was added to the reaction as competitor DNAs. The reactions were further incubated at 23°C for 30 min. The 20 μl reactions were then applied to 0.45 μm, 13 mm nitrocellulose filter papers (Whatman D25AGR) which had been soaked in H₂O, positioned on a suction apparatus and filtered at the rate of 3–4 ml/min. The filters were immediately washed with 0.8 ml of buffer containing 10 mM Tris-HCl, pH 7.5, 50 mM KCl, dried and immersed in 5 ml toluene/Omnifluor (4 g/l) scintillation fluid before counting in a Beckman LS 3801 scintillation counter.

The abbreviations used are

topo70

NH₂-terminal truncation of human topoisomerase I missing the first 174 amino acids

topo58

COOH-terminal truncation of topo70 missing the last 106 amino acids

topo56

COOH-terminal truncation of topo70 missing the last 126 amino acids

topo58

COOH-terminal truncation of topo70 missing the last 106 amino acids

topo70ΔL

a form of topo70 missing linker residues from 660–688

topo31

a fragment of human topoisomerase I extending from residue 175 to 433

Δcap

NH₂-terminal truncation of human topoisomerase beginning at residue 433

WT

wild type

DTT

dithiothreitol

BSA

bovine serum albumin

GST

glutathione S-transferase

PAGE

polyacrylamide gel electrophoresis