Human Topoisomerase IIIα Is a Single-stranded DNA Decatenase That Is Stimulated by BLM and RMI1*

Human topoisomerase IIIα is a type IA DNA topoisomerase that functions with BLM and RMI1 to resolve DNA replication and recombination intermediates. BLM, human topoisomerase IIIα, and RMI1 catalyze the dissolution of double Holliday junctions into noncrossover products via a strand-passage mechanism. We generated single-stranded catenanes that resemble the proposed dissolution intermediate recognized by human topoisomerase IIIα. We demonstrate that human topoisomerase IIIα is a single-stranded DNA decatenase that is specifically stimulated by the BLM-RMI1 pair. In addition, RMI1 interacts with human topoisomerase IIIα, and the interaction is required for the stimulatory effect of RMI1 on decatenase activity. Our data provide direct evidence that human topoisomerase IIIα functions as a decatenase with the assistance of BLM and RMI1 to facilitate the processing of homologous recombination intermediates without crossing over as a mechanism to preserve genome integrity.

Human topoisomerase III␣ is a type IA DNA topoisomerase that functions with BLM and RMI1 to resolve DNA replication and recombination intermediates. BLM, human topoisomerase III␣, and RMI1 catalyze the dissolution of double Holliday junctions into noncrossover products via a strand-passage mechanism. We generated single-stranded catenanes that resemble the proposed dissolution intermediate recognized by human topoisomerase III␣. We demonstrate that human topoisomerase III␣ is a single-stranded DNA decatenase that is specifically stimulated by the BLM-RMI1 pair. In addition, RMI1 interacts with human topoisomerase III␣, and the interaction is required for the stimulatory effect of RMI1 on decatenase activity. Our data provide direct evidence that human topoisomerase III␣ functions as a decatenase with the assistance of BLM and RMI1 to facilitate the processing of homologous recombination intermediates without crossing over as a mechanism to preserve genome integrity.
Topoisomerases are ubiquitous enzymes conserved from bacteria to humans. Their roles in modulating DNA topology in replication, transcription, and other cellular processes (1,2) make them indispensable for cell viability. There are four subfamilies of topoisomerases as follows: IA, IB, IIA, and IIB. Type IA topoisomerases change DNA topological states in discrete steps of one via an enzyme-bridging mechanism (1,2). The catalytic tyrosine residue initiates a transesterification reaction in a single-stranded region to generate a transient DNA break, allowing for the passage of the intact strand through the break. After religation of the broken strand by a reversal of the reaction, the enzyme is free to engage in another round of catalysis (1,2). Members of the type IA topoisomerase family include Escherichia coli topoisomerase I (EcTop1) and III (EcTop3), yeast topoisomerase III (Top3), and two isoforms of topoisomerase III, ␣ (Topo 2 III␣) and ␤ (Topo III␤), in higher eukaryotes. These enzymes exhibit high sequence similarity in the N-terminal catalytic core domain, whereas the C-terminal tails are variable (1). In addition to the ability to relax negatively supercoiled DNA, EcTop1 is capable of catalyzing knotting, unknotting, and interlinking of DNA substrates that contain exposed single-stranded regions (1). Because single-stranded DNA gaps are a common feature found at DNA replication forks, replication termination sites, and replication and repair sites, it is believed that the main function of type IA topoisomerases is to unlink DNA catenanes.
Type IA topoisomerases function in concert with RecQ helicases to control recombination events (3). RecQ helicases are a highly conserved family of DNA helicases that are required for the maintenance of genome integrity (4). Human topoisomerase III␣ (hTopo III␣) physically interacts with BLM, one of the five RecQ helicases in humans (5,6). Biallelic mutations of BLM give rise to a clinically defined cancer predisposition disorder, Bloom syndrome (BS) (7). BS cells display signs of genome instability, featuring an ϳ10-fold elevation in the frequency of sister chromatid exchanges (SCE) (8), events that arise from the processing of recombination intermediates (9). The hTopo III␣ interacting domain of BLM is required for suppression of SCE in BS cells (10), suggesting that hTopo III␣ plays an anti-recombination role with BLM. Indeed, in vitro biochemical data show that BLM and hTopo III␣ catalyze the dissolution of double Holliday junctions (DHJs), a DNA structure that can arise as an intermediate during homologous recombination (11). Dissolution occurs via a strand-passage mechanism that prevents genetic exchange between flanking sequences and is presumed to mimic the in vivo role of BLM-hTopo III␣ in suppressing SCE (11). In the simplest case, the dissolution reaction is believed to have two components as follows: the helicase activity of BLM catalyzes branch migration of the Holliday junctions toward each other, resulting in collapse of the Holliday junctions, and generation of two duplex DNAs interlinked via catenated single strands. This structure, termed a hemicatenane, is then decatenated by hTopo III␣ to complete the dissolution of the DHJ (11,12). Direct evidence that hTopo III␣ possesses the relevant decatenase activity, however, is currently lacking.
RecQ helicases and type IA topoisomerases also cooperate to resolve converging replication forks. EcTop3 together with RecQ and the single-stranded DNA-binding protein SSB catalyzes the unlinking of late replication intermediates in vitro (13). EcTop3 also catalyzes efficient decatenation of daughter DNA molecules during oriC and pBR322 replication in vitro (14). Although similar assays have not been performed within eukaryotes, several lines of evidence suggest that eukaryotic topoisomerase III functions in a similar role. First, Schizosaccharomyces pombe Top3 is required for normal chromosome segregation (15). Second, Topo III␣ depletion in chicken DT40 cells causes accumulation of metaphase cells with chromosome gaps and breaks (16). Finally, hTopo III␣ localizes to ultrafine anaphase DNA bridges in a BLM-dependent manner (17). In each of these cases, the failure of Topo III to decatenate and thereby resolve converging replication forks could lead to interlinked sister chromatids after replication and improper sister chromatid disjunction in mitosis.
In eukaryotes, Topo III functions in concert with RMI (RecQ-mediated genomic instability) proteins (18 -24). In Saccharomyces cerevisiae, deletion of the gene encoding Rmi1 results in phenotypes similar to top3⌬, including sensitivity to DNA-damaging agents, and hyper-recombination (19,20), consistent with Rmi1 and Top3 functioning in the same pathway. In humans, RMI1 binds to hTopo III␣ via its conserved N-terminal domain (21,22), and this interaction appears to be important for hTopo III␣ stability in vivo (23). In vitro, RMI1 stimulates DHJ dissolution by hTopo III␣ and BLM (21,22). Although other type IA topoisomerases can function in DHJ dissolution in vitro, stimulation of dissolution by RMI1 specifically requires hTopo III␣ (22). RMI1 associates with a second RMI protein, RMI2, via the OB-fold domain at the C terminus to form a heterodimeric RMI complex (24). The RMI complex is required for the phosphorylation of BLM in mitotic cells and for the recruitment of BLM to nuclear foci in response to DNA damage (18,23,24). Moreover, cells that are depleted of the RMI complex show an elevated level of SCE (23,24), the hallmark of BS cells, indicating that the RMI complex is critical for BLM function. Thus, the complex of BLM-hTopo III␣-RMI1-RMI2 likely represents the functional unit in higher eukaryotes and is termed the BLM core complex (25).
To define the biochemical role of hTopo III␣ in DHJ dissolution and in the resolution of converging replication forks, we have generated a catenane of single-stranded DNA circles to mimic the proposed target of hTopo III␣ in the final step of these reactions. Using an improved procedure for the expression and purification of hTopo III␣, we demonstrate that hTopo III␣ acts as a single-stranded DNA decatenase in DHJ dissolution and that this decatenase activity is specifically stimulated by RMI1. Surprisingly, we also find that BLM stimulates decatenation, and that BLM and RMI1 together synergistically stimulate decatenation by hTopo III␣. We propose that hTopo III␣ is the cellular decatenase that functions in dissolution of DHJs and resolution of converging replication forks and that the optimal decatenase activity depends on the formation of a complex with RMI1 as well as BLM helicase activity.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-hTopo III␣ cDNA was subcloned from pET29H2-hTopo III (26) into pRS424-Gal-GST (27) to generate pRS424Gal-GST-hTopoIII␣. This plasmid was transformed along with plasmid pMTL4 (28) into the protease-deficient S. cerevisiae strain BJ2168 (MATa ura3-52 trp1-289 leu2-3 prb1-1122 prc1-407 pep4-3). Methods for cell growth, protein induction, and extract preparation were essentially the same as described previously (24), except that the cells were lysed in HEPES 100 lysis buffer (50 mM HEPES-NaOH (pH 7.5), 100 mM NaCl, 10% glycerol, 0.1% Tween 20, 1 mM EDTA, 1 mM EGTA, 3 mM DTT, 2.5 mM sodium pyrophosphate, 1 mM ␤-glycerophosphate, 10 mM sodium bisulfite, 10 g/ml leupeptin, and 1 g/ml pepstatin A). To separate DNA from soluble proteins, the extract was treated with 0.45% polymin-P and 600 mM ammonium sulfate at 4°C for 10 min, followed by centrifugation at 20,000 rpm for 45 min. The soluble fraction was then precipitated with 2 M ammonium sulfate, and the precipitate was resuspended and dialyzed overnight against HEPES 150 lysis buffer (which has the same composition as HEPES 100 , except with 150 mM NaCl). The soluble fraction was loaded onto a 5-ml heparin-Sepharose column (HiTrap heparin, GE Healthcare) using fast protein liquid chromatography (GE Healthcare), and the column was washed with HEPES 150 for 10 column volumes prior to elution with a NaCl gradient from 150 mM to 1 M over 10 column volumes. Fractions containing hTopo III␣ were identified by immunoblotting and fractionated on a 1-ml glutathione-Sepharose column (GSTrap 4B; GE Healthcare). The column was washed with HEPES 545 (545 mM NaCl) for 10 column volumes before the bound material was eluted by the addition of HEPES 545 containing 20 mM glutathione.
Generation of RMI1 Mutants-RMI1 mutants, LLTD and K166A were generated from p426Gal1-hRMI1-TAP by QuikChange mutagenesis. The oligonucleotides used for this study were purchased from Integrated DNA Technologies. Their sequences are in supplemental Table 1. All mutants were confirmed by sequencing.
Generation of Single-stranded Catenanes-The single strand catenane substrate was constructed as described previously (30), with several modifications. Briefly, oligos C1 and C2 were annealed in the presence of oligo B1 to form the catenane. Control reactions that contained only oligo C2 also contained oligo B2 to allow circularization of C2 to occur efficiently. Singlestranded oligonucleotides were radiolabeled individually with [␥-32 P]ATP (BLU502A; PerkinElmer Life Sciences) using T4 polynucleotide kinase (M0201S, New England Biolabs) prior to use. After ligation of the catenated oligos, the single-stranded catenane was purified by electrophoresis on a 12% polyacrylamide gel containing 8 M urea in TBE buffer for 75 min at 175 V, followed by staining with SYBR Green I (S-7563; Invitrogen) for 30 min. Bands containing the single-stranded catenanes were identified and excised under UV light. Catenanes were recovered by electroelution into dialysis bags in 0.25ϫ TBE buffer for 60 min at 120 V. The eluates were concentrated using spin filters (BIOMAX-10K NMWL, Millipore). The final product appeared as a double band on 12% denaturing polyacrylamide gels because there was likely a mixture of full-length and truncated oligos in the starting material due to the long runs of dT residues. The sequences of the DNA oligonucleotides used for the generation of single-stranded catenanes are in supplemental Table 2.
Single-stranded DNA Decatenation Assay-Purified singlestranded catenane (400 fmol) was incubated with the indicated proteins in 15 l of reaction buffer containing 50 mM Tris-HCl (pH 7.5), 40 mM NaCl, 5 mM MgCl 2 , 1 mM DTT, and 0.1 mg/ml BSA at 37°C for 30 min, unless indicated otherwise. Reactions were stopped by the addition of 1% SDS and 20 mM EDTA. Samples were deproteinized with 1 mg/ml proteinase K at 50°C for 1 h before being subjected to 12% polyacrylamide ϩ 8 M urea gel electrophoresis in TBE buffer at 175 V for 75 min. Gels were fixed in 10% MeOH, 5% acetic acid for 30 min and equilibrated in 1% glycerol, 10% acetic acid for another 30 min prior to drying. Gels were then exposed to a storage phosphor screen and subsequently analyzed with a Typhoon scanner (GE Healthcare) and ImageJ software. For reactions containing BLM, 5 mM ATP was added to initiate the reactions, unless indicated otherwise.
DNA Relaxation Assay-pBR322 (21 fmol) (N3033S, New England Biolabs) was incubated with the indicated proteins in 15 l of reaction buffer containing 50 mM Tris-HCl (pH 7.5), 40 mM NaCl, 5 mM MgCl 2 , 1 mM DTT, and 0.1 mg/ml BSA at 37°C for 30 min. Reactions were stopped by the addition of 3 l of Stop Buffer containing 2.5% SDS, 25 mM EDTA, 0.05% bromphenol blue, 0.05% xylene cyanol, and 15% glycerol. Samples were subjected to 1% agarose gel electrophoresis in TAE buffer with buffer re-circularization at 70 V for 16 h. The gels were stained with SYBR Green I (S-7563; Invitrogen) for 30 min and visualized under UV light.
Peptide Binding Assay-Peptide arrays were prepared by the CRUK Peptide Synthesis Laboratory. Peptides were synthesized on a Multipep Synthesizer (Intavis Bioanalytical Instruments, Cologne, Germany) on derivatized cellulose (amino-Peg500 UC540, Intavis). After synthesis and deprotection, the membranes were washed with dichloromethane, N-methylpyrrolidone, and then ethanol. Binding of hTopo III␣ to the immobilized peptides was analyzed by the far-Western technique using hTopo III␣ as probe and detecting with the D6 anti-hTopo III␣ antibody, as described previously (5).

RESULTS
Expression and Purification of Recombinant hTopo III␣-Because hTopo III␣ does not display a strong biochemical activity when purified from Escherichia coli, we developed a procedure for the expression and purification of hTopo III␣ in yeast (supplemental Fig. 1A). hTopo III␣ tagged at the N terminus with GST was expressed in S. cerevisiae using a galactose-inducible expression system (32). Using affinity chromatography together with two rounds of conventional column chromatography, we purified hTopo III␣ to near homogeneity (supplemental Fig. 1B). The purified hTopo III␣ was catalytically active as evidenced by its ability to relax negatively supercoiled pBR322 plasmid DNA (Fig. 1A). Furthermore, the DNA relaxation activity of hTopo III␣ was dependent on Mg 2ϩ (Fig.  1B), as reported previously (26), indicating that the preparation was not contaminated by detectable amounts of Top1.
RMI1 Specifically Stimulates the DNA Relaxation Activity of hTopo III␣-It has been proposed that RMI1 promotes BLM-hTopo III␣-mediated DHJ dissolution by influencing the activity of hTopo III␣ (22). To test this directly, we expressed TAPtagged RMI1 in an S. cerevisiae strain lacking SGS1 and TOP3 to prevent potential Top3 contamination. RMI1 was purified to near-homogeneity (supplemental Fig. 1, C and D) (24), and we measured the effect of RMI1 on hTopo III␣-mediated DNA relaxation. At the low topoisomerase concentrations used here, hTopo III␣ exhibited a weak DNA relaxation activity (Fig. 1C). Relaxation activity was stimulated by the addition of increasing amounts of RMI1. hTopo III␣ relaxation activity was not stimulated when it was incubated with the same concentrations of BSA suggesting that the stimulation, despite requiring a large molar excess of RMI1, was specific (Fig. 1C).
Other type IA topoisomerases, such as EcTop1, are able to function with BLM in DHJ dissolution (11). However, RMI1 does not stimulate the BLM-EcTop1-mediated DHJ dissolution (22). Consistent with this observation, the relaxation activity of EcTop1 was not stimulated by RMI1 (Fig. 1, D and E). This further suggests that the stimulation of hTopo III␣ by RMI1 is specific, and indicates that the stimulation of hTopo III␣ in both relaxation and DHJ dissolution requires the cognate topoisomerase-RMI1 pair.
RMI2, a member of the BLM core complex, is critical for hTopo III␣ and RMI1 protein function in vivo (18,24). We therefore measured the effect of RMI2 on hTopo III␣mediated DNA relaxation. We found that RMI2 did not enhance hTopo III␣ relaxation activity on its own, nor did it stimulate hTopo III␣ relaxation activity in concert with RMI1 (supplemental Fig. 2A).
Generation of Single-stranded Catenanes-Although relaxation assays are typically used to measure topoisomerase activity in vitro, it is unlikely that DNA relaxation is the biologically relevant activity of hTopo III␣. Rather, it has been proposed that the role of hTopo III␣ in resolving recombination intermediates is to facilitate the resolution of hemicatenanes that result from the BLM helicase-mediated, convergent migration of two Holliday junctions (11). Hemicatenanes, which contain topologically linked DNA single strands, could be resolved by a single strand decatenase activity of hTopo III␣, although this activity has not been identified directly. We generated singlestranded DNA catenanes consisting of a 60-nucleotide oligomer (circle 1) and a 40-nucleotide oligomer (circle 2) to mimic the hemicatenanes that hTopo III␣ is proposed to resolve in DHJ dissolution ( Fig. 2A). Each oligonucleotide contained a 12-bp complementary region that was flanked by flexible poly(dT) linkers to allow the formation of a mixture of singly linked and doubly linked single-stranded catenanes (30). Consistent with the properties of a closed circular singlestranded DNA catenane, the final product exhibited slower mobility than the monomer circles on a denaturing polyacrylamide gel (Fig. 2B, lanes 4 and 5 versus lane 6) and was resistant to denaturant (Fig. 2B, lane 3 versus lane 6) and exonuclease digestion (Fig. 2B, lanes 6 and 12 versus lanes 3 and 9). The purified catenane is shown in Fig. 2C.
hTopo III␣ Is a Single-stranded DNA Decatenase-We used the single-stranded DNA catenanes to test whether hTopo III␣ displays decatenase activity in vitro. When we incubated hTopo III␣ with the single-stranded catenanes, we observed an appearance of bands corresponding to the sizes of individual circles 1 and 2 and a concomitant disappearance of the catenated species (Fig. 2D), indicating that hTopo III␣ indeed possesses singlestranded DNA decatenase activity.
Unlike type IB topoisomerases, members of the type IA DNA topoisomerase family from non-human sources are able to substitute for hTopo III␣ in BLM-catalyzed DHJ dissolution (11). This suggests that type IA but not type IB topoisomerases are able to catalyze the relevant reaction in DHJ dissolution. We reasoned that if the relevant activity in DHJ dissolution is singlestranded DNA decatenation, then type IA but not type IB topoisomerases should catalyze this decatenation reaction. We therefore repeated the decatenation experiment with EcTop1, a type IA topoisomerase, and hTopo I, a type IB topoisomerase. We found that although EcTop1 displayed clear decatenase activity (Fig. 2E), hTopo I was unable to resolve single-stranded catenanes even at a concentration ϳ5 times higher than that at which relaxation activity was detected (supplemental Fig. 3, A and B). Together, these data suggest that the hTopo III␣ activity that is relevant to DHJ dissolution by BLM and hTopo III␣ is single-stranded DNA decatenation and not supercoil relaxation.
Characterization of hTopo III␣-mediated Decatenation of Single-stranded Catenanes-Next, we characterized the biochemical properties of the decatenase activity of hTopo III␣. The decatenase activity of hTopo III␣ was maximal when the reactions were incubated at 30°C, although it was almost completely abolished at 50°C (supplemental Fig. 4A). The decatenase activity of hTopo III␣ decreased linearly with both increasing NaCl concentration and increasing pH. Activity was strongest at NaCl concentrations below 20 mM and at a pH of less than 6.5 (supplemental Fig. 4, B and C). hTopo III␣ decatenase activity was greatly diminished when Mg 2ϩ concentration was reduced to less than 1.25 mM (supplemental Fig. 4D), consistent with previous reports that type IA DNA topoisomerases require Mg 2ϩ as a cofactor for catalytic activity. Of the divalent cations tested, only Ca 2ϩ was able to substitute for Mg 2ϩ as a cofactor in mediating the decatenation of singlestranded catenanes by hTopo III␣ (supplemental Fig. 4, E and F).
RMI1 Stimulates the Decatenase Activity of hTopo III␣ but Not That of EcTop1-Because RMI1 promotes hTopo III␣-mediated relaxation of supercoiled DNA and the BLM-hTopo III␣-mediated dissolution of DHJs, we tested the effect of RMI1 on hTopo III␣ decatenase activity. We found that the decatenation by hTopo III␣ was stimulated by RMI1, in a concentrationdependent manner, by up to 2-fold (Fig. 3A). Furthermore, although EcTop1 decatenated single-stranded DNA, RMI1 did not promote EcTop1-mediated decatenation (Fig. 3B). Together, these data suggest that RMI1 stimulates singlestranded DNA decatenation by hTopo III␣ specifically and argues against models of stimulation in which RMI1 modifies the DNA substrate, for example, to make it more accessible to the topoisomerase.  1 and 2, respectively, as markers. The percent of catenated substrate converted to circular products is indicated. E, decatenation reactions containing increasing concentrations of EcTop1 (0.375-24 nM, lanes 2-8) were fractionated on a denaturing polyacrylamide gel and autoradiographed. The percent of catenated substrate converted to circular products is indicated.

Identification of hTopo III␣-interacting Domain in RMI1-
Because specific stimulation of hTopo III␣-mediated decatenation by RMI1 suggests that the stimulatory effect is via a direct physical interaction between the two proteins, we tested the requirement for binding of hTopo III␣ by RMI1 in stimulation of decatenation and of DHJ dissolution. We first performed an in vitro peptide binding assay (5) to identify putative hTopo III␣ interacting regions in RMI1 (supplemental Fig. 5). Candidate binding regions of RMI1 were subjected to mutagenesis, and alanine substitution mutations that eliminated binding in the peptide assay were introduced into the full-length RMI1. Co-immunoprecipitation analysis was performed with the mutant proteins to test the role of each region in the hTopo III␣-RMI1 interaction. We found that RMI1 harboring an LLTD (residues 57-60) to AAAA mutation in the conserved DUF1767 domain (RMI1-LLTD) (Fig. 4A and supplemental Fig. 5) exhibited a defect in binding to hTopo III␣ (Fig. 4B), indicating that these four amino acids contribute to the physical interaction between hTopo III␣ and RMI1. In addition to the LLTD mutation, RMI1 harboring a K166A mutation (RMI1-K166A), which was previously shown to be defective in hTopo III␣ interaction (33), displayed reduced binding to hTopo III␣ (Fig. 4C).

Stimulation of DHJ Dissolution and Decatenation Depends on Physical Interaction between hTopo III␣ and RMI1-RMI1-
K166A is defective in stimulating the DHJ dissolution activity of BLM-hTopo III␣ (33). We therefore asked whether RMI1-LLTD displays a similar defect in DHJ dissolution. Although wild type RMI1 readily stimulated DHJ dissolution at 25 nM, both RMI1-LLTD and -K166A failed to stimulate DHJ dissolution even at Ͼ300 nM (Fig. 5, A-D), confirming that the stimulatory effect of RMI1 on BLM-hTopo III␣-mediated DHJ dissolution requires a physical interaction between hTopo III␣ and RMI1.
We tested whether RMI1-LLTD and -K166A could promote single-stranded decatenation by hTopo III␣. The decatenase activity of hTopo III␣ was not stimulated by either RMI1-LLTD or RMI1-K166A (Fig. 5E), indicating that the physical interaction between hTopo III␣ and RMI1 is critical for the RMI1 stimulation of hTopo III␣-mediated decatenation. These data are consistent with RMI1 stimulating hTopo III␣ in DHJ dissolution, and with the stimulation being the result of RMI1 modulation of hTopo III␣ activity, rather than an indirect effect of RMI1 on DNA substrate accessibility or conformation.
BLM, but Not RMI2, Stimulates hTopo III␣ Decatenase Activity-Like hTopo III␣ and RMI1, BLM and RMI2 are members of the BLM core complex (25). BLM promotes hTopo III␣mediated relaxation of negatively supercoiled DNA in vitro (34) and is required for the proper localization of hTopo III␣ in vivo (10,17). We tested the effect of BLM on decatenation by hTopo III␣. Addition of increasing concentrations of BLM stimulated decatenation by at least 2.5-fold (Fig. 6A). The stimulation depended on the presence of ATP (Fig. 6B), suggesting that it required the catalytic activity of BLM. BLM also stimulated EcTop1-mediated decatenation, but more weakly (ϳ0.5-fold; Fig. 6C). Together, these data suggest that BLM stimulates decatenation in the following two ways: unwinding of the substrate results in a modest nonspecific stimulation, but further stimulation of the cognate topoisomerase is also apparent. Thus it appears that BLM alters the conformation of the substrate in addition to influencing the activity of hTopo III␣ directly.
Addition of increasing concentrations of RMI2 had no detectable effect on hTopo III␣ decatenase activity (supplemental Fig. 2B), and it did not have an effect on the decatenase activity of the hTopo III␣-RMI1 pair (supplemental Fig. 2C). Hence decatenation by hTopo III␣ is independent of RMI2. It is likely that RMI2 plays a role in a higher order function of the BLM core complex that is not revealed by these highly specific biochemical assays.
BLM and RMI1 Stimulate hTopo III␣-mediated Decatenation Synergistically-RMI1 enhances DHJ dissolution by ϳ8-fold (22), whereas the stimulatory effect of RMI1 on decatenation by hTopo III␣ was a relatively modest 2-fold. Therefore, it is possible that there is additional interplay between BLM, hTopo III␣, and RMI1. When decatenation by hTopo III␣ was assayed in the presence of both BLM and RMI1, the decatenase activity increased by a striking ϳ20-fold (Fig. 7A,  5th lane versus 2nd lane). Under the same conditions, BLM or RMI1 alone stimulated decatenation by only 2-fold (Fig. 7A, 3rd  and 4th lanes versus 2nd lane), indicating that BLM and RMI1 Black stripes in the RMI1 mutants indicate the amino acid positions where amino acid substitution mutations were made. B and C, extracts from cells expressing the indicated epitope-tagged proteins were treated with glutathione 4B-Sepharose to precipitate GST or GST-hTopo III␣. 1% of input (I) extracts and the precipitate (P) was fractionated by SDS-PAGE. Immunoblots were probed with anti-GST antibody to detect GST-hTopo III␣ or with peroxidase/anti-peroxidase to detect RMI1-TAP. Identities of the bands are indicated on the right. The amount of RMI1 in the precipitates relative to that in the input extracts was calculated as a percentage and is indicated below each immunoblot.
stimulate the decatenase activity of hTopo III␣ synergistically. Stimulation of hTopo III␣ by BLM-RMI1 requires physical interaction between hTopo III␣ and RMI1 because the bindingdefective RMI1-LLTD mutant together with BLM did not stimulate above the level seen with BLM alone (Fig. 7B, 7th lane  versus 3rd lane). Thus, the physical interaction between hTopo III␣ and RMI1 is required for the optimal activity of the BLM-hTopo III␣-RMI1 complex in both decatenation and DHJ dissolution. Furthermore, these data indicate a role for BLM function in the very last step of the reactions catalyzed by the BLM core complex.

DISCUSSION
Using DNA substrates that resemble replication and recombination intermediates, numerous in vitro studies have made significant advances in elucidating the mechanistic details of RecQ-Topo III-mediated resolution of replication and recombination intermediates. Most notably, the helicase-topoisomerase partnership resolves converging replication forks (13) and resolves double Holliday junctions without crossing over (11,12). In this study, we generated single-stranded catenanes that resemble the DNA structure that is likely recognized by hTopo III␣ in these reactions. We provide direct evidence that hTopo III␣ is a single-stranded DNA decatenase and that the decatenase activity is stimulated by members of the BLM core complex, BLM and RMI1. Modulation of decatenation and DHJ dissolution reactions requires physical interaction between RMI1 and hTopo III␣. The specificity of singlestranded DNA decatenation and DHJ dissolution for type IA topoisomerases, as well as the specificity of the RMI1 stimula- tion of both reactions, indicates that the relevant activity of hTopo III␣ in the dissolution of DHJs is single-stranded DNA decatenation and not relaxation of superhelical stress.
Improved Purification of hTopo III␣ and RMI1-We developed improved purification methods for both hTopo III␣ and RMI1 by the use of S. cerevisiae as a host for expression. Several lines of evidence suggest that S. cerevisiae is a better tool than E. coli for the expression and purification of hTopo III␣ and RMI1. First, although RMI1 is insoluble in E. coli (22), the protein was soluble when overexpressed in S. cerevisiae. In addition, yeast-purified RMI1 was ϳ5 times more active in stimulation of DHJ dissolution by BLM and hTopo III␣ (compare this study with Ref. 22). hTopo III␣ purified from S. cerevisiae was also more active in both the decatenation ( Fig. 2D and supplemental Fig. 6A) and relaxation assays (compare this study with Ref. 26). Furthermore, in contrast to previous work with the yeast homologs Top3 and Rmi1, where stimulation of relaxation was only apparent when the proteins were co-expressed and co-purified (35), we were able to reconstitute stimulation of hTopo III␣ from individually purified hTopo III␣ and RMI1 proteins in both DNA relaxation and decatenation assays. Our data indicate that expressing and purifying hTopo III␣ and RMI1 in S. cerevisiae yield functional proteins with optimal biochemical activities.
Type IA Topoisomerases Are Single-stranded DNA Decatenases-We have developed an elegant system to directly assay the single-stranded DNA decatenase activity of type IA topoisomerases. Previous work has indirectly assayed decatenation of type IA topoisomerases on complex substrates in the presence of RecQ helicases. Decatenase activity of type IA topoisomerases was demonstrated in an in vitro replication system in which EcTop3 converted plasmid replication intermediates into monomeric daughter molecules (14,36). However, the complex nature of this system made it difficult to directly identify the DNA substrate on which the enzyme was acting. Recently, Suski and Marians (13) used an improved system to trap replication intermediates right before replication forks converge, allowing for the study of enzyme activities at a more defined stage. Using this system, they demonstrated that RecQ and EcTop3 catalyze the resolution of converging replication forks via a strand-passage mechanism similar to that in DHJ dissolution, where RecQ-mediated DNA unwinding is followed by EcTop3-mediated decatenation to liberate plasmid DNAs with single-stranded gaps (13). Similarly, the second component of the DHJ dissolution activity reported by Wu and Hickson (11) is proposed to be an hTopo III␣-catalyzed decatenation of a hemicatenane. The simple single-stranded DNA catenane used here resembles the topoisomerase IA substrate at the latest stage in both the resolution and the dissolution reactions. Significantly, although both type IA and IB topoisomerases catalyze the relaxation of negatively supercoiled DNA by introducing a single-stranded nick (1), the decatenation of single-stranded DNA reported here was specific to type IA topoisomerases. The specificity observed in the decatenation assay is consistent with that in the DHJ dissolution assay where type IA, but not type IB, topoisomerases are able to replace hTopo III␣ in catalyzing DHJ dissolution with BLM (11). Both decatenation and DHJ dissolution require a type IA topoisomerase, and stimulation of both reactions by RMI1 requires the cognate topoisomerase hTopo III␣. Stimulation of decatenation occurs at an hTopo III␣:RMI1 stoichiometry near the expected 1:1 and is compromised in RMI1 mutants that bind hTopo III␣ poorly. Together, these data suggest that decatenation is the biologically relevant activity of hTopo III␣ in processing replication and recombination intermediates. Failure to decatenate these intermediates in vivo would generate interlinked sister chromatids and homologous chromosomes, leading to abnormal mitoses, chromosome breaks, and rearrangements. Consistent with this, hTopo III␣ localizes to ultra-fine DNA bridges during anaphase (17), and depletion of hTopo III␣ causes an increase in the number of DNA bridges (37).

BLM and RMI1 Cooperate with hTopo III␣ in Decatenation-
We find that both BLM and RMI1 stimulate the decatenase activity of hTopo III␣, although the underlying mechanisms of stimulation by the two proteins could be quite different. BLM stimulates the decatenase activity of both hTopo III␣ and EcTop1 in an ATP-dependent manner, suggesting that unwinding of the substrate by BLM activity contributes to the stimulation indirectly. However, the stimulation of EcTop1 decatenase activity is modest in comparison with the stimulation of hTopo III␣ (0.5versus 2.5-fold), so it appears that a specific interaction of BLM with hTopo III␣ contributes to the stimulation. RMI1, on the other hand, is not able to promote EcTop1-mediated decatenation, and hTopo III␣-binding mutants of RMI1 fail to stimulate hTopo III␣mediated decatenase activity. This strongly suggests that the stimulation occurs via specific interactions between hTopo III␣ and RMI1. We envisage two possible scenarios. In the first, RMI1 targets hTopo III␣ to the single-stranded catenane via a DNA binding activity of RMI1. Whether RMI1 has such a DNA binding activity, however, remains controversial. A DNA binding domain has been localized to the C terminus of human RMI1, but this domain is dispensable for the stimulation of DHJ dissolution (33), and it seems unlikely that an activity critical for RMI1 function would be located outside of the conserved N-terminal region. Other studies of RMI1 failed to detect significant DNA binding activity (24). Budding yeast Rmi1, which lacks the C-terminal region, has weak single-stranded DNA binding activity and can bind Holliday junction structures (20,35). It is possible that the N terminus of RMI1 contains a cryptic DNA binding activity that is activated upon binding to hTopo III␣ or that RMI1 recognizes catenated single-stranded DNA specifically. Alternatively, the interaction between hTopo III␣ and RMI1 could induce conformational changes in hTopo III␣ that enhance its decatenase activity. Consistent with this possibility, the Holliday junctions and DHJ binding activity of Topo III in yeast and human systems are stimulated by RMI1 (22,35).
Several lines of evidence have suggested that Topo III and RMI1 form a tight complex that might then have a more peripheral association with RecQ. For example, co-expressed Top3 and Rmi1 have increased solubility compared with the individual subunits (35); knockdown of RMI1 destabilizes hTopo III␣, whereas knockdown of BLM does not (18,24); the Top3-Rmi1 complex binds to Sgs1 (19,35); deletion of rmi1 in yeast closely resembles the phenotype of top3 deletion rather than that of sgs1 deletion (19,20); stimulation of DHJ dissolution by RMI1 requires the cognate type IA topoisomerase, hTopo III␣ (22); and Topo III␣ and RMI1 perform meiotic functions that are independent of RecQ4a, a homolog of BLM helicase in Arabidopsis thaliana (38,39).
In this context, and because other type IA topoisomerases can substitute for hTopo III␣ in DHJ dissolution, a stepwise model for the dissolution in which the helicase and the topoisomerase perform distinct roles has been proposed (40). Although this is likely the case at the enzymatic level, because BLM lacks decatenase activity and hTopo III␣ lacks helicase activity, our data indicate that there is an important interplay between BLM, hTopo III␣, and RMI1 even at the latest stage of the reaction where only decatenase activity is required. This interplay is apparent when the three proteins are incubated together in the decatenation assay. Although both BLM and RMI1 alone display modest stimulation on hTopo III␣ decatenase activity (by ϳ2-fold), the BLM-RMI1 pair synergistically stimulates the activity by at least 20-fold. Moreover, this synergy depends on the physical interaction between RMI1 and Topo III␣. Thus, the maximum level of decatenation requires all members of the BLM-Topo III␣-RMI1 complex. The specificity for the entire BLM-Topo III␣-RMI1 complex for optimal decatenation that we observe is also seen in Holliday junction unwinding by BLM, which requires hTopo III␣-RMI1 for optimal activity (41). Because unwinding and decatenation both contribute to DHJ dissolution, it follows that maximum activity in DHJ dissolution is likewise only obtained when all three compo-  , lanes 3, 5, and 6), and RMI1 (100 nM, lanes 4, 5, and 7) as indicated were fractionated on a denaturing polyacrylamide gel and autoradiographed. ATP (5 mM) was added to initiate the reactions. Quantification of the decatenation products is presented in the histogram, normalized to the reaction in lane 2 (hTopo III␣ alone). The percent of catenated substrate converted to circular products is indicated. B, decatenation reactions containing hTopo III␣ (15 nM, lanes 2-7), BLM (33 nM, lanes 3 and 6 -8), wild type RMI1 (100 nM, lanes 4, 6, and 9), and RMI1-LLTD mutant (100 nM, lanes 5, 7, and 10) as indicated were fractionated on a denaturing polyacrylamide gel and autoradiographed. ATP (5 mM) was added to initiate the reactions. Quantification of the decatenation products is presented in the histogram, normalized to the reaction in lane 2 (hTopo III␣ alone). The percent of catenated substrate converted to circular products is indicated.
nents are present (21,22). Furthermore, in a scenario where the two junctions are separated by multiple topologically constrained linkages, the activity of hTopo III␣ would likely be required at earlier stages to relieve superhelical stress generated by BLM-dependent branch migration. Therefore, although the steps in catalyzing DHJ dissolution appear to be conceptually distinct, the coordinated actions of all the members in the BLM core complex are likely required for optimal activity in suppressing SCEs in vivo. Finally, our results point to an added complexity in reactions catalyzed by RecQ-Topo III complexes in eukaryotes. Whereas work with late replication intermediates indicates functional cooperation between RecQ and Topo III in E. coli (13), our data indicate that in eukaryotes the functional specificity and cooperation extend also to RMI1.
The strand passage activity of BLM-hTopo III␣-RMI1 is likely important for a number of cellular processes in which Holliday junctions are a feature (Fig. 8). These include the repair of DNA double-stranded breaks, the processing of stalled replication forks, the lengthening of telomeres via the alternative lengthening of telomeres pathway, and the resolution of converging replication forks, which are topologically similar to Holliday junctions. All of these processes share a common terminal step, decatenation. We propose that hTopo III␣ catalyzes the decatenation of these structures in vivo and that the decatenation is performed in cooperation with BLM and RMI1. Such a role is consistent with the recent association of two single-nucleotide polymorphisms in the hTopo III␣ gene with increased risk of cancer development (42). The details of decatenation catalyzed by hTopo III␣ presented here shed light on the role of hTopo III␣ and its cooperating partners BLM and RMI1 in the maintenance of genome integrity.