The many lives of type IA topoisomerases

The double-helical structure of genomic DNA is both elegant and functional in that it serves both to protect vulnerable DNA bases and to facilitate DNA replication and compaction. However, these design advantages come at the cost of having to evolve and maintain a cellular machinery that can manipulate a long polymeric molecule that readily becomes topologically entangled whenever it has to be opened for translation, replication, or repair. If such a machinery fails to eliminate detrimental topological entanglements, utilization of the information stored in the DNA double helix is compromised. As a consequence, the use of B-form DNA as the carrier of genetic information must have co-evolved with a means to manipulate its complex topology. This duty is performed by DNA topoisomerases, which therefore are, unsurprisingly, ubiquitous in all kingdoms of life. In this review, we focus on how DNA topoisomerases catalyze their impressive range of DNA-conjuring tricks, with a particular emphasis on DNA topoisomerase III (TOP3). Once thought to be the most unremarkable of topoisomerases, the many lives of these type IA topoisomerases are now being progressively revealed. This research interest is driven by a realization that their substrate versatility and their ability to engage in intimate collaborations with translocases and other DNA-processing enzymes are far more extensive and impressive than was thought hitherto. This, coupled with the recent associations of TOP3s with developmental and neurological pathologies in humans, is clearly making us reconsider their undeserved reputation as being unexceptional enzymes.

In B-form DNA, the two complementary strands are associated with one another via hydrogen bonds that form between the bases of each strand. These paired strands periodically coil around their longitudinal axis in a clockwise orientation, giving rise to a so-called right-handed double helix. The intertwining that results from this conformation prevents the separation of the strands and thereby ensures that the genetic information is tightly protected within the center of the DNA fiber. However, to gain access to the sequences of the bases, the double helix must be opened, which requires the topological disentanglement of the two strands. In that respect, although topological intertwines provide stability to the double helix, they nevertheless can be viewed as a fundamental obstacle to the functionality of DNA. Hence, organisms must strictly control the degree to which their genomic DNA is intertwined to guarantee both the protection and the functionality of their genetic material.
B-form DNA is widely recognized for its elegant double-helical structure. However, this double helix is not as static as it is generally represented. It is instead constantly manipulated to extract, duplicate, and repair the genetic information that it encodes. The reciprocal interplay between DNA topology and DNA metabolism has an overwhelming complexity because many, if not all, factors involved in these processes alter the topological entanglements of the genome. Among these factors, the machineries that translocate along DNA clearly play one of the most prominent roles. When a translocating protein machinery is not free to rotate around the axis of the DNA, it instead behaves as a topological barrier whose translocation drives the rotation of the DNA fiber itself. In a topologically constrained DNA molecule, this DNA rotation leads to the overwinding (positive supercoiling) of the DNA segment located ahead of the translocase, whereas the DNA in its wake becomes progressively underwound (negatively supercoiled) (Fig. 1A). This process describes the so-called "twin supercoiling domain" model in which the absolute degree of interstrand entanglement of the DNA molecule is not altered, but instead is distributed differently relative to the position of the translocase (1). In living cells, the twin supercoiling domain model primarily applies to the translocation of the transcription machinery, whose rotation is thought to be limited not only by its size, but also by steric constraints, such that RNA polymerases introduce substantial levels of both positive and negative supercoiling (2-6) (Fig. 1B). This superhelical constraint has been shown to impact many DNA processes, including DNA replication, chromatin and chromosome architecture, and transcription itself (4,5,7). Indeed, because they represent an excess of topological linkages, the positive supercoils generated ahead of any RNA polymerase prevent DNA opening required for transcription initiation and would ultimately block the progression of the polymerase if not dealt with (8 -11). Conversely, the negative supercoils generated behind the transcription machinery correspond to a deficit in topological linkages and can lead to the destabilization of the double helix, an effect that favors the formation of R-loops and secondary DNA structures, which threaten the continued stability of the genome (12)(13)(14)(15).
The twin supercoiling domain model also applies to the translocation of the DNA replication machinery (replisome), although the potential topological problems arising during this process are more complicated due to the additional presence of the two sister dsDNA molecules that form behind the replication fork (16,17). Most often, it would seem that the rotation of the replisome is prevented, such that DNA synthesis is associated with the generation of positive supercoils in the unreplicated template DNA (18) (Fig. 1C). If allowed to accumulate, these supercoils would prevent further progression of the replisome and promote fork stalling and potentially fork reversal (19). The deficit in topological entanglements behind the translocating replisome manifests as a limited degree of winding of each of the newly replicated sister chromatids. However, under special circumstances, such as during replication termination or at sites of replication-transcription conflicts, the build-up of positive supercoils has been proposed to trigger fork rotation (20 -22) (Fig. 1D). In such cases, the intramolecular topological linkages between the template strands are converted into intertwines between the newly replicated sister DNA segments (23). These intertwines (known as precatenanes) would be converted into catenanes upon completion of replication and are generally considered to be the main obstacle to the faithful segregation of the sister chromatids during cell division.
Given the topological challenges that are continually generated during DNA processes such as transcription and replication, it is inconceivable that B-form DNA could have evolved as a carrier of genetic information without the co-evolution of efficient ways to actively relieve topological stress (24). This is particularly critical for circular genomes and for long linear chromosomes whose organization is effectively constrained topologically because their structure does not allow the torsional energy to be dissipated by free rotations of their ends (5,10,25). By promoting DNA breakage, strand rotation, and ligation within a three-step catalytic cycle, DNA topoisomerases are the enzymes responsible for manipulating the absolute degree of topological entanglement of DNA (26). They achieve this prodigious feat by catalyzing a reversible transesterification reaction involving an active-site tyrosine residue and a phospho-ester bond of the DNA substrate. This reaction leads to the formation of the so-called topoisomerase cleavage complex, a topoisomerase-DNA intermediate in which the enzyme is covalently attached to an extremity of the broken DNA. This bond can be reversed by another transesterification reaction, enabling the nick to be resealed and the enzyme to be released from its covalent association with the substrate. Using this elegant biochemical trick, DNA topoisomerases are able to manipulate intra-and intermolecular DNA topology without exposing their substrates to the dangers of long-lived strand breaks.
DNA topoisomerases belong to a select group of enzymes that are conserved in all living organisms and that existed during the earliest stages of cellular life. All bacteria, archaea, and eukaryotic cells are equipped with a set of DNA topoisomerases that is sufficient to cope with the topological by-products of their DNA metabolism by releasing negative and positive supercoils, as well as intermolecular catenation. In eukaryotic cells, these activities are performed by TOP1 and TOP2, which are members of the Type IB and Type IIA families of topoisomerases, respectively (Fig. 2). TOP1 introduces a transient nick into one strand of dsDNA, which leads to the formation of a covalent linkage to the 3Ј end of the nick ( Fig. 2A). Efficient release of both negative and positive supercoils is then performed via a swiveling mechanism during which the 5Ј end of the scissile strand rotates freely around the intact DNA strand to dissipate torsional stress and reset the molecule to its most stable topological conformation (27) (Fig. 2B). Type II topoisomerases can also relax both positive and negative supercoils, but they do so via an intrinsically different mechanism, which involves active manipulation of the DNA entanglements by the enzyme. Eukaryotic TOP2s are homodimeric enzymes that catalyze transesterification reactions on both strands of a dsDNA substrate, leading to the formation of a cleavage complex in which each protomer is covalently associated with a 5Ј end of a double-stranded break (Fig. 2C). The 3Ј extremities of this break, which are not covalently bound to the catalytic tyrosine residues, are nevertheless tightly associated with the enzyme and thereby prevented from rotating. Instead, changes in topology occur because an intact dsDNA molecule is actively transported through the transient dsDNA break (26). This mode of action is intrinsically distributive and potentially riskier than a swiveling mechanism for DNA relaxation, but it enables Type II topoisomerases to modulate both intra-and intermolecular DNA topology, the latter being critical for the decatenation of sister chromatids formed during DNA replication (Fig. 2D).
The different biochemical activities of eukaryotic Type IB and Type II topoisomerases have evolved to conduct at least some distinct functions in vivo. Whereas Type IB topoisomer- Ϫve SC) behind the translocase. B, in the context of transcription, the overwinding accumulated ahead of the RNA polymerase prevents strand opening and can ultimately block transcription elongation. Underwinding generated behind the polymerase can promote strand opening and lead to the stabilization of R-loops and other secondary structures. C, during DNA replication, and when fork rotation is prevented, the degree of entanglement between the newly replicated DNA molecules is limited, but overwinding ahead of the fork can prevent replisome progression. D, during DNA replication, overwinding of the template can be limited by fork rotation, but this leads to formation of precatenanes behind the fork, which represent an obstacle during segregation. JBC REVIEWS: DNA topoisomerases ases are well-suited to roles in relieving torsional stress generated during transcription and replication, Type II topoisomerases are indispensable for the decatenation of DNA that is necessary during chromosome condensation and segregation. The well-characterized relaxation and decatenation activities of Type IB and Type II topoisomerases appear sufficient to manage all topological constraints associated with DNA metabolism in eukaryotes (29 -34). It is perhaps surprising, therefore, that every eukaryotic cell also expresses another class of topoisomerase, the Type IA enzymes (35)(36)(37). Indeed, Type IA topoisomerases are the only topoisomerases conserved in virtually all living organisms. They are further divided into three subfamilies, the TOP3 enzymes, bacterial TOPA enzymes, and reverse gyrases. Whereas the distribution of TOPAs and reverse gyrases is restricted to bacteria and hyperthermophilic organisms (38), most organisms encode for at least one TOP3. Although it is commonly accepted that TOP3s are absolutely critical for the maintenance of genomic stability, their precise cellular function has remained quite mysterious. This is probably because they essentially lack the canonical relaxation activity shared by all other DNA topoisomerases and hence have not been considered important for the release of torsional stress generated by transcription and replication. These Type IA topoisomerases can nevertheless manipulate a broad range of topological substrates in vitro. The physiological relevance of this catalytic versatility is now being progressively uncovered, which has defined the involvement of TOP3s in some unexpected cellular processes. In this review, we will focus on our current knowledge of this unique family of topoisomerases and the many roles that they play in the maintenance of the stability of our genomes.

Protein structure and catalytic cycle
All Type IA topoisomerases share a very similar architecture within their core topoisomerase domain, which comprises one topoisomerase-primase subdomain (TOPRIM; 3 subdomain I) and two catabolite activator protein subdomains (CAP; subdomains III and IV), which are connected by two topo-folds (subdomain II) (39 -45) (Fig. 3A). Overall, this core domain 3 The abbreviations used are: TOPRIM, topoisomerase-primase subdomain; CAP, catabolite activator protein subdomain; T-segment, transported segment; UFB, ultrafine anaphase bridge; cen-UFBs, ultrafine anaphase bridges that arise from centromeres; mtDNA, mitochondrial DNA; RMI, RecQ-mediated genome instability; OB-fold, oligonucleotide-oligosaccharide-binding fold; ssDNA and ssRNA, single-stranded DNA and RNA, respectively. iii, a second set of transesterification reactions religate the break and free the enzyme from its covalent interaction with the DNA. D, when the G-segments (purple) and T-segments (yellow) are located on the same molecule, Type II topoisomerase activity leads to the relaxation of negative and positive supercoils (top). When the G-and T-segments are located in trans, Type II topoisomerases can modify the degree of catenation between two dsDNA molecules (bottom).

Figure 2. Mechanism of action and catalytic activities of
JBC REVIEWS: DNA topoisomerases resembles a toroidal clamp in which subdomains I, III, and IV are associated at the base of an arc formed by subdomain II. The clamp can adopt either a closed or open conformation upon association or dissociation of a gate formed at the interface between subdomain III and subdomains I and IV. Besides the catalytic tyrosine, which is located within subdomain III, the transesterification reaction requires 10 highly conserved residues distributed throughout subdomains I, III, and IV (35). As a result, the catalytic pocket can only be constituted at the interface of the gate upon closure of the clamp (Fig. 3A). Although many aspects of the mechanism by which Type IA topoisomerases manipulate DNA topology remain to be understood, a model for their mode of action has been proposed (37, 46) (Fig. 3B). In this model, the enzyme associates with its substrate via a DNA-binding groove primarily created by subdomains I and IV, which guides the scissile DNA strand (or gated segment; G-segment) toward the catalytic pocket, such that it is in line with the active tyrosine of the CAP subdomain III (47,48). After G-segment nicking, Type IA topoisomerases and their substrates are engaged in a covalent interaction between the catalytic tyrosine of subdomain III and the terminal 5Ј-phosphate of the nick. A series of structural rearrangements involving the flexible subdomain II lead to a swinging motion of subdomain III relative to the subdomains I/IV, which opens the gate and drives the subdomains III and I/IV and their associated DNA termini apart (49 -52). In this conformation, the noncovalent association between the enzyme and the 3Ј terminus of the G-segment can still resist forces up to 40 picoNewtons, indicating that the enzyme and the 3Ј terminus must remain very tightly associated (50,53,54). This strong interaction, which is most likely mediated via the DNA-binding groove, also prevents the swiveling of the noncovalently attached DNA end. Instead, changes in DNA topology occur via the active transport of another nucleic acid molecule through the gate (46). Once the transported segment (T-segment) has been passed through the gate, closure of the clamp leads to the reconstitution of the catalytic cycle. This in turn enables a second transesterification reaction to restore the continuity of the scissile strand and to free the enzyme and its substrate from their covalent interaction. Full dissociation is achieved by a final opening of the gate, allowing the T-segment to exit the cavity of the enzyme.

Catalytic versatility
Type IA topoisomerases are able to act on a broad range of topological structures, which contrasts with the limited number of substrates identified for other topoisomerase families (Fig. 4). Indeed, Type IA topoisomerases require only that their substrate exhibits a single-stranded segment to be used as a G-segment (55). This substrate restriction is determined by the nature of the G-segment-binding groove, whose dimensions can only accommodate a single-stranded nucleic acid segment with the bases facing toward the groove (40,47,48). This requirement explains why Type IA topoisomerases can relax supercoils only when the density of negative supercoiling is sufficiently high to promote strand separation, whereas they are A, the structure of the core topoisomerase domain resembles a toroidal clamp in which the topo-fold subdomain II forms an arc. The catalytic site is reconstituted at the base of this arc, by the association of residues from the topoisomerase-primase (TOPRIM) subdomain I and the two catabolite activator (CAP-Y and CAP) subdomains III and IV. The deep ssDNA (and ssRNA) binding groove (G-segmentbinding groove) is formed between subdomains I and IV. B, a, Type IA topoisomerases bind to single-stranded segments of DNA via the G-segment-binding groove that directs the G-segment in line with the catalytic tyrosine (i, open red circle). b, a transesterification reaction leads to the formation of a single strand break with the 5Ј termini covalently associated with the catalytic tyrosine (ii, closed red circle). The 3Ј end of the nick forms a tight association with the G-segment-binding groove. c, protein conformational changes enable the opening of a gate via the separation of the CAP domains and their associated DNA termini (iii). d, another nucleic acid segment (here the single-strand DNA complementary to the G-segment) is passed though the gate toward the cavity of the enzyme (iv). e, after this transport, the closure of the gate reconstitutes the catalytic cycle (v). f, a second transesterification reaction reseals the nick and frees the enzyme from its covalent interaction with the substrate (vi). g, full dissociation is enabled by opening of the gate. JBC REVIEWS: DNA topoisomerases much less potent on moderately negative and positive supercoils (unless a stable unpaired segment has been introduced into the substrate) (46,(55)(56)(57)(58)(59). Similarly, the activity of these topoisomerases is promoted by the presence of DNA secondary structures, such as mismatches, G-quadruplexes, D-loops, and R-loops, which generate ssDNA regions in the substrate (60 -63). Hence, Type IA topoisomerases are often described as single strand-specific DNA topoisomerases (61). In addition to catalyzing topological conversions in cis, which lead to substrate relaxation, they can perform topological conversions in trans and thereby display decatenation activities (Fig. 4). Furthermore, the nature of the T-segment also appears to be unrestricted by the properties of the conserved core topoisomerase domain. In particular, although the ability of Type IA topoisomerases to transfer dsDNA molecules has not been demonstrated formally, both the width of the opened gate and the size of the cavity (around 25 Å) are sufficient to accommodate dsDNA (50). Moreover, Escherichia coli Top3 has been shown to be able to entrap a dsDNA T-segment in its closed cavity (51). Last, but not least, Type IA topoisomerases have been shown to accept RNA both as a G-and as a T-segment and are therefore able to manipulate the topology of RNAs in vitro (64 -69). Hence, depending on the exact nature of the T-segment, mem-bers of this family have been shown to mediate the resolution of the full range of DNA and RNA intermolecular entanglements, including dsDNA catenanes, ssDNA catenanes, precatenanes, and hemicatenanes as well as knotted and catenated RNAs (64 -76) (Fig. 4). Hence, the conserved core catalytic domain of Type IA topoisomerases is characterized by impressive versatility, being able to modulate the topology of a broad range of DNA and RNA substrates. However, it is worth mentioning that some individual members of the type IA DNA topoisomerases appear to have become specialized for a subset of these reactions. This is well-illustrated by the catalytic, structural, and functional divergences of the two eukaryotic TOP3s, TOP3A and TOP3B, and will be discussed further below.
The precise mechanism by which the transfer of the T-segment through the gate occurs is unknown. Type IA topoisomerases can relax both negative and positive supercoils in substrates with unpaired DNA regions. This property indicates that they can both increase and decrease intramolecular DNA entanglements and suggests that the directionality of any topological change is driven primarily by the existing topology of the substrate (77). Similar to their Type II counterparts, Type IA topoisomerases have been proposed to modify DNA topology via a so-called sign inversion mechanism. An alternative JBC REVIEWS: DNA topoisomerases hypothesis is that these topoisomerases differentially alter DNA topology by the direction in which they transfer the T-segment either into or out of the cavity. Consistent with this alternative model, ecTop3 can capture a dsDNA segment into its closed cavity in the absence of a G-segment, indicating that the enzyme may be able to initiate its catalytic cycle when a T-segment is already located inside the cavity (51) (Fig. 3B).

Synergy with other enzymes
Because of the intrinsic lack of substrate specificity and strand transfer directionality, the activity of Type IA topoisomerases is prone to be influenced by any factor that can alter the dynamics of dsDNA denaturation (including changes in temperature, the presence of ssDNA-binding proteins, or the activities of other DNA-processing enzymes) (73)(74)(75)(77)(78)(79)(80)(81)(82). Most notably, TOP3s are known for their ability to tightly cooperate with DNA translocases to catalyze complex topological conversions that could not be performed by a stand-alone topoisomerase. A good example is the so-called "dissolvasome" that results from the association of a TOP3 and a RecQ helicase (Schizosaccharomyces pombe Rqh1, Saccharomyces cerevisiae Sgs1, and human BLM) (83-85) (Fig. 5). In this multienzyme complex, the activities of the RecQ helicase and TOP3 coordinate with each other to catalyze the dissolution of double Holliday junctions, which proceeds by the migration of the junctions toward each other until they merge to form a hemicatenated structure that is processed by TOP3 (86). In addition to its role during the unlinking of this hemicatenane, TOP3 also assists the convergent branch migration reaction primarily driven by the ATP-dependent helicase activity of Rqh1/Sgs1/ BLM (87)(88)(89). The mechanism by which TOP3 manipulates DNA topology during the convergent branch migration is still not clear, although it seems likely that the activity of the RecQ helicase specifically enables TOP3 to efficiently release the tor-sional stress associated with convergent branch migration (90). Besides its ability to dissolve double Holliday junctions, the dissolvasome has also been shown to catalyze the resolution of other complex substrates, such as D-loops, late-replication intermediates, and catenated dsDNA (53,63,71,(91)(92)(93)(94) (Fig.  5B). Altogether, therefore, this helicase-topoisomerase combination must be viewed as a multistructure dissolvasome, specialized in the resolution of various types of complex intermolecular entanglements (95).
More recently, the human topoisomerase III␣, TOP3A, was shown to cooperate with an ATPase of the SNF2 family, the dsDNA translocase PICH, to catalyze the introduction of positive supercoils into DNA (96) (Fig. 6). This supercoiling reaction is driven by the ability of PICH to act as a translocating topological barrier and to redistribute torsional stress on a dsDNA substrate, in the same manner as discussed above in the context of the twin supercoiling domain model (Fig. 1A). Unexpectedly, PICH was shown to extrude DNA loops concomitant with this torsional stress redistribution, such that negative supercoils are able to accumulate in the extruded loop located in its wake, whereas the rest of the substrate becomes progressively more positively supercoiled (96). Such a loop extrusion mechanism enables PICH to promote the local melting of the DNA in the loop, which creates the necessary hypernegatively supercoiled substrate for TOP3A to catalyze DNA relaxation. The final result of this combined activity of PICH and TOP3A is the introduction of an impressively high density of positive supercoiling in a processive manner (Fig. 6).
The multienzyme complex formed by PICH and TOP3A is not the first example of synergistic cooperation between a Type IA topoisomerase and an ATPase to induce positive supercoiling. Indeed, such an activity is the hallmark of the so-called reverse gyrases, an atypical family of Type IA topoisomerases that are essential for the survival of hyperthermophilic organisms at high temperatures (97,98). Reverse gyrases are chimeric enzymes that combine a Type IA topoisomerase domain and an ATPase domain of the superfamily 2 helicases within the same polypeptide, which is reminiscent of the association between the human TOP3A and PICH proteins (44). In contrast to their human analog, the hyperthermophilic reverse gyrases are distributive, and the degree of positive supercoiling they introduce  JBC REVIEWS: DNA topoisomerases is quite modest, illustrating a fundamentally different mechanism of action. Indeed, the ATPase domain of reverse gyrase is believed to function as a nucleotide switch that facilitates local DNA melting via a helicase-like mechanism (98). However, it is worth mentioning that some hyperthermophilic reverse gyrases, as exemplified by Sulfolobus solfataricus TopR2, introduce a high density of positive supercoiling in a processive manner (99). These observations are consistent with the existence of two reverse gyrase subfamilies that possess fundamentally different mechanisms of action and may indicate that the TopR2 reverse gyrases introduce positive supercoiling via a loop extrusion mechanism analogous to that of PICH and TOP3A.

The cellular functions of TOP3 enzymes
For DNA topoisomerases, just as for any other DNA-manipulating enzyme, a clear definition of their biochemical activities is a prerequisite for understanding their roles in vivo. The large repertoire of activities catalyzed by TOP3s alone and within multienzymatic complexes supports the notion that these topoisomerases are pleiotropic, being involved in a broad range of cellular processes that require the manipulation of specific topological structures that cannot be resolved by Type IB and Type II topoisomerases. Although TOP3s have been well-characterized in vitro, our understanding of their physiological roles is relatively superficial. However, several studies, in particular of the eukaryotic isoforms of TOP3s, are progressively revealing multiple roles of these unique topoisomerases in many aspects of cellular metabolism associated with the maintenance of genomic stability. Most higher eukaryotes encode two isoforms of TOP3, topoisomerases III␣ and III␤, designated TOP3A and TOP3B, respectively (Fig. 7A). Phylogenetic analysis indicates that the presence of these two isoforms resulted from a duplication early in the eukaryotic lineage (38). In the next section, we will review the known physiological roles of the two eukaryotic TOP3 isoforms.

Roles of TOP3A during homologous recombination
The function of eukaryotic TOP3 during homologous recombination has long been associated with its role as a member of the dissolvasome alongside RecQ helicases. The fission yeast top3 ϩ gene is essential for cell viability, and the lethal phenotype of top3 mutants is suppressed by deletion of the gene encoding the Rqh1 RecQ helicase (100). In budding yeast, top3 mutants show a slow-growth phenotype that is suppressed by SGS1 gene mutations (85,101). Moreover, top3 (and sgs1) mutants are synthetically lethal when combined with strains lacking the Mus81-Mms4 structure-selective nuclease, and this lethality can be suppressed by eliminating homologous recombination, indicating that Top3 and Sgs1 play an important role during some aspect of this process (102). Although the dissolvasome could prevent early recombination events by dissolving D-loops, it clearly also acts during the later stages of some homologous recombination events by catalyzing double Holliday junction dissolution (86,103). Through its activity as a dissolution enzyme, not only for double Holliday junctions but also various other substrates (Fig. 5B), the dissolvasome influences many aspects of genome maintenance, including DNA replica-tion, DNA repair, telomere maintenance, and meiosis (104 -111). Of the two eukaryotic TOP3 isoforms, only TOP3A has strongly conserved homology with its yeast Top3 counterpart, whereas TOP3B appears to have diverged significantly (Fig.  7A). TOP3A has an essential function, and its inactivation leads to embryonic lethality in mice, Caenorhabditis elegans, Drosophila melanogaster, and Arabidopsis thaliana (56,78,112). Human subjects carrying hypomorphic TOP3A mutations are characterized by short stature and microcephaly (113). Cells A, all Type IA topoisomerases share a highly conserved catalytic domain (blue) and sometimes an additional C-terminal extension (CTD), which contains multiple zinc finger motifs (black boxes) involved in protein-DNA and protein-protein interactions. In addition to putative zinc finger motifs, the C-terminal domain of HsTOP3B also exhibits RGG box motifs (green) that can be methylated and mediate interactions with RNA and the Tudor domain of TDRD3. An alternative start codon leads to the addition of a mitochondrial targeting sequence (MTS) to the TOP3A polypeptide, such that TOP3A encodes both nuclear and mitochondrial isoforms. In eukaryotes, TOP3 forms heterodimers with members of the RMI family. RMI members are characterized by a conserved association between a DUF1767/OB-fold domain (DUF-OB). In humans, the nuclear isoforms of TOP3A and TOP3B interact with their own cognate RMI protein, RMI1 and TDRD3, respectively. The RMI1 CTD exhibits a second OB-fold domain and mediates interactions with other proteins, including RMI2. TDRD3 CTD is characterized by the presence of multiple protein-protein interaction motifs including a ubiquitin-associating domain (UBA) and a Tudor domain. The Tudor domain of TDRD3 mediates interactions with methylated proteins, including histones, RNA polymerase, and TOP3B, and with the fragile X mental retardation protein. Ec, E. coli; Sc, S. cerevisiae; Hs, Homo sapiens. B, RMI1 and TDRD3 interact with the arc of TOP3A and TOP3B, respectively. RMI1 inserts a loop into the cavity of TOP3A, which restricts its size. A similar insertion loop is present in TDRD3, but this does not appear to significantly reduce the size of the TOP3B cavity.
with defective TOP3A have increased levels of sister chromatid exchanges and mitotic abnormalities, highlighting an essential role of TOP3A in the maintenance of genome stability that is reminiscent of BLM deficiency (113)(114)(115)(116). Consistently, TOP3A has been shown to be the topoisomerase component of the human dissolvasome together with the BLM helicase. In agreement with yeast genetic studies, this eukaryotic dissolvasome appears to be involved in multiple aspects of DNA metabolismthatrequirehomologousrecombinationoranalogousprocesses, such as recovery from replication stress, alternative lengthening of telomeres, and regulation of meiotic crossovers (117)(118)(119)(120)(121)(122)(123)(124)(125)(126)(127).

Roles of TOP3A at ultrafine anaphase bridges
Cells expressing defective TOP3A exhibit an increased frequency of ultrafine anaphase bridges (UFBs) (96,113). UFBs are thin DNA structures that are stretched between the segregating sister genomes during the anaphase of mitosis (116,128,129). They arise due to the presence of topological entanglements between sister chromatids that fail to be resolved prior to anaphase onset. A variety of topological entanglements give rise to UFBs, such as those arising in replication intermediates, recombination intermediates, and fully catenated DNA (130 -133). Hence, the increased frequency of UFBs in TOP3A-defective cells might be at least partly due to the possible roles of this topoisomerase in replication and homologous recombination during interphase. Furthermore, all members of the dissolvasome localize to UFBs, and all of the DNA structures predicted to give rise to UFBs have been shown to be substrates for the dissolvasome complex in vitro, suggesting that the dissolvasome might additionally act in the resolution of these intertwined structures directly during mitosis (Fig. 5) (53,116).
In addition to a possible role in resolving UFBs together with BLM, TOP3A may also act on these structures in conjunction with PICH. Indeed, PICH and TOP3A co-localize on UFBs during the anaphase of mitosis, and depletion of TOP3A or PICH inactivation leads to an increased frequency of the ultrafine anaphase bridges that arise from centromeres (cen-UFBs) (96,132). These are the only UFBs frequently observed during an unperturbed cell cycle and are therefore believed to represent physiological structures (128). They most likely correspond to fully catenated sister chromatids that are resolved shortly after anaphase onset (134). Although efficient resolution of cen-UFBs is mediated primarily by TOP2A, it is facilitated by PICH and TOP3A (96,132). Because the decatenation activity of TOP2 has been proposed to be most efficient on positively supercoiled catenanes, there is the intriguing possibility that PICH and TOP3A catalyze positive supercoiling to facilitate the rapid decatenation of cen-UFBs through stimulation of TOP2A (135)(136)(137). In this context and by analogy with the function of prokaryotic reverse gyrases critical to protect DNA against thermal denaturation (38,97,138), the positive supercoiling activity of the PICH-TOP3A complex might also contribute to protecting DNA against denaturation when exposed to mitotic spindle forces (139). Moreover, although PICH function is most likely restricted to mitosis due to its localization in the cytoplasm during interphase, one cannot exclude the possibility that other topoisomerase-associated complexes catalyze posi-tive supercoiling during interphase. Several scenarios in which a positive supercoiling activity might be necessary can be envisaged. For instance, it might be exploited to down-regulate any process that requires DNA melting for its initiation, such as transcription, replication, and homologous recombination (8,(140)(141)(142)(143). Supercoiling could also influence genome architecture at different levels, such as nucleosome dynamics, the formation and dynamics of topologically associated domains, and chromosome condensation (4,144,145).

Possible roles during replication
Early studies, largely performed in E. coli, suggested that Top3 plays a redundant role in DNA replication with the bacterial Type II topoisomerase, Topo IV (72,73,146,147). It was proposed that TOP3s could participate in decatenation of newly replicated DNA by acting on gapped or nicked precatenanes directly behind the replication fork, a model that was confirmed recently in E. coli (91). However, whether this also occurs in eukaryotic cells is unknown. Besides a role in precatenane removal, the activities of TOP3s could be relevant to other topological aspects associated with DNA replication. For example, in common with any DNA annealing process, the synthesis of a new dsDNA molecule during DNA replication has to be coupled with the de novo formation of topological intertwining to give rise to a double helix. Whereas on the lagging strand this introduction of intertwining can be achieved by free rotation of the Okazaki fragment ends, the synthesis of the leading strand has to be associated with an active introduction of topological linkages. The moderate relaxation activity exhibited by TOP3 may be well-suited to the relaxation of the extreme density of negative supercoils generated on the newly synthesized leading strand, while maintaining a degree of supercoiling compatible with nucleosome assembly. In eukaryotes, replication termination has been proposed to be initiated by convergent fork rotation and stalling resulting from the accumulation of positive supercoils in the short unreplicated segment that can no longer be accessed by TOP1 or TOP2 (20,148). The ability of the dissolvasome to resolve late replication intermediates may be ideally suited to processing these structures formed during the latter stages of DNA replication (92,(149)(150)(151). Finally, it should be mentioned that the dissolvasome has been previously proposed to play a role in establishment of sister chromatid cohesion via a pathway that depends on the homologous recombination protein Rad51 and independently of other known cohesion pathways (152,153). Although highly speculative, it is at least conceivable that the dissolvasome could perform this function by introducing, rather that resolving, entanglements between sister chromatids.

Roles of mitochondrial TOP3A
In higher eukaryotes, the TOP3A gene encodes both a nuclear and a mitochondrial isoform, the latter being expressed from an alternative start codon that creates a mitochondrial targeting sequence at the N terminus of the TOP3A polypeptide (56,154) (Fig. 7A). To our knowledge, it is only in Drosophila that the mitochondrial isoform has been genetically ablated. This led to infertility and premature aging, but did not affect organismal viability. These findings indicate that only the JBC REVIEWS: DNA topoisomerases nuclear isoform of TOP3A is required during embryonic development, but that mtTOP3 is essential for mtDNA stability (56,155,156). Some of the patients described above who harbor TOP3A mutations display severe cardiomyopathy reminiscent of that seen in disorders associated with mitochondrial abnormalities (113). Furthermore, a compound heterozygous mutation of TOP3A has been identified in an individual with a mitochondrial disorder (157). Significantly, the only functional protein expressed in this case carried an amino acid substitution in the TOP3A core catalytic domain (M100V). Cells harboring this defect exhibit perturbed mtDNA segregation and extensive mtDNA rearrangements, which is consistent with a key function of TOP3A in the resolution of hemicatenanes formed specifically during mitochondrial genome replication (157). Importantly, none of the core nuclear TOP3A partners (BLM, RMI1, and RMI2; see below for a description of RMI1 and RMI2) has been convincingly shown to reside in mitochondria, suggesting that TOP3A operates independently of the dissolvasome in the mitochondria (157)(158)(159).

Roles of TOP3B in RNA metabolism
TOP3B is the least studied of the human DNA topoisomerases probably because it is the only eukaryotic nuclear topoisomerase that does not appear to be essential for life (160). TOP3B is nonetheless highly relevant for human health, as revealed by studies showing that TOP3B deficiency is linked to the development of neurodevelopmental defects. Indeed, the TOP3B gene resides on chromosome 22q11.2 in the human genome, a region frequently affected by deletions or duplications leading to congenital heart disease, facial malformation, and cognitive dysfunction (68,(161)(162)(163)(164)(165). Consistent with this, copy number variants and de novo mutations in TOP3B have been linked to an increased risk of neurodevelopmental and cognitive disorders, including autism and schizophrenia (68,(166)(167)(168)(169). The involvement of TOP3B in cognition is likely conserved across species because TOP3B has been shown to be essential for correct synaptic formation in both mice and flies (67,166). These phenotypes, also observed when TOP1 and TOP2B are defective, suggest that TOP3B has a role in releasing topological constraints associated with transcription (170). Such a role is supported by the efficient relaxation activity of TOP3B in vitro, which appears to be specifically targeted to sites of transcription in vivo where TOP3B acts to prevent the accumulation of R-loops (60,171).
Somewhat surprisingly, given the above discussion, TOP3B has been shown to localize predominantly to the cytosol, where it associates with mRNA within polyribosomes and stress granules (67,68,166). Such a localization suggests a role of TOP3B outside of the nucleus and independent of its ability to interact with DNA and hints at the possible physiological relevance of the RNA topoisomerase activity of TOP3s. Indeed, whereas TOP3A has apparently lost its RNA topoisomerase activity, TOP3B seems to have evolved to specifically process RNA-containing substrates (see below). Although the cellular role of this RNA topoisomerase activity remains to be defined precisely, TOP3B may resolve RNA torsional stress and/or (pseudo-)knots to facilitate the translation of long mRNAs in vivo (67).

Roles of TOP3B in genome stability maintenance
TOP3B null mice have a shorter lifespan, which has been attributed to the development of inflammatory lesions in multiple organs resulting from autoimmunity hyperactivation probably driven by an accumulation of apoptotic cells (160,172). TOP3B-defective mice also display a progressive reduction in fecundity over generations due to an increased incidence of aneuploidy in germ cells (173). These observations highlight a role of TOP3B in the long-term maintenance of genomic stability (172). Consistent with this, a recent study identified a homozygous deletion mutation in TOP3B in a patient with a late-onset, bilateral renal cancer (174). TOP3B-deficient cells exhibit an increased frequency of DNA damage associated with an accumulation of R-loops, but impaired activation of DNA damage response effectors (60,174,175). Although one cannot exclude the possibility that TOP3B plays a direct role in DNA replication or repair, this increase in DNA damage may be directly related to the function of TOP3B in preventing R-loop accumulation during transcription. Indeed, R-loops have been shown to represent a threat to genome stability, in particular in proliferating cells, where they can clash with the replication machinery (60,171,176).
All of these findings suggest that TOP3B plays a role during transcription, alongside TOP1 and TOP2. Similar to TOP3A, TOP3B may also be a multifunctional topoisomerase potentially involved in many more cellular processes. For instance, a mitotic/meiotic role of TOP3B has long been speculated, and it was recently reported to intervene in RNAi-mediated heterochromatin formation and transcriptional silencing in Drosophila (173,(177)(178)(179).

Regulation of TOP3 activity
The different functions supported by TOP3A and TOPB in vivo suggest that these isoforms serve as nonredundant topoisomerases. Although no systematic, side-by-side comparison of the activities of TOP3A and TOP3B has been performed, they do apparently display marked differences in their substrate preferences. In particular, TOP3B can fully relax a negatively supercoiled plasmid, which strikingly contrasts with the weak relaxation activities of all other TOP3s in general and of TOP3A in particular (171,(179)(180)(181). Although these observations are in contradiction with the initial biochemical characterization of Drosophila TOP3B, which seemed to indicate inefficient DNA relaxation, they raise the interesting possibility that this isoform exhibits unique core DNA topoisomerase properties (182). Furthermore, whereas TOP3A has apparently lost the ability to manipulate the topology of RNA substrates, TOP3B seems to have specifically retained the RNA topoisomerase activity conserved within all other TOP3s (64 -69). It is possible that differences in the biochemical properties of TOP3A and TOP3B are at least partially supported by some subtle divergence within their core topoisomerase domains. However, it is clear that their respective C-terminal domains and obligatory partners play a prominent role in the specialization of activities of these two TOP3 isoforms.

C-terminal domain
The activity of the catalytic core of Type IA topoisomerases is often modulated by the presence of a less well-conserved C-terminal domain (Fig. 7A) (183). In yeast TOP3, the C-terminal domain is short, but it has been extended considerably in all TOP3s found in higher eukaryotes. Importantly, TOP3A and TOP3B have distinct C-terminal domains, which are nonetheless relatively well-conserved for each subclass across species, suggesting that their specialization is at least partly determined by this region. In TOP3A, the C-terminal domain is characterized by the presence of at least three zinc-binding motifs. The first, located directly after the core topoisomerase domain, comprises a four-cysteine-zinc motif (Zn-C4) that is involved in ssDNA binding. It is conserved in several Type IA topoisomerases and has been proposed to correspond to a fifth subdomain of the core topoisomerase domain (36). Two other zinc-binding regions are defined by tandem glycine-argininephenylalanine (GRF) motifs that mediate interaction with nucleic acids and are potentially involved in targeting the different topoisomerases to their favored substrate(s) in vivo and in vitro (184 -186). Besides multiple zinc finger motifs, the C-terminal domain of TOP3B harbors arginine-glycine-glycine (RGG) box motifs that are present in certain RNA-binding proteins and are required for TOP3B activity and its function in vivo (Fig. 7A) (166). In addition to their nucleic acid-binding properties, these C-terminal domains might also be involved in mediating protein-protein interactions. In particular, the interaction of TOP3B with the Tudor domain-containing 3 protein (TDRD3) is partly mediated via its RGG box motifs. Similarly, the C-terminal domain of Drosophila TOP3A is involved in an interaction with the BLM helicase (88). Whereas it is clear that the C-terminal domains of TOP3 enzymes are dispensable for their core topoisomerase activity, their ability to mediate interactions with their substrates and protein partners likely plays a key role in specifying cellular functions. Moreover, based on current knowledge, it is not possible to rule out a more direct role for this domain in modulating the biochemical activity of the topoisomerase core domain, such as in the coordination of the successive steps of the catalytic cycle or even in providing directionality to the movement of the T-segment (187). In the future, it might be particularly important to gain more understanding of the structure of the C-terminal domain of TOP3s, in particular in the context of its association with the core topoisomerase domain.

The RMI partners of eukaryotic TOP3s
Eukaryotic TOP3 enzymes form stable heterodimers with conserved RMI (RecQ-mediated genome instability) accessory proteins, with TOP3A and TOP3B each having a specific partner: RMI1 and TDRD3 (Tudor domain-containing 3 protein), respectively (Fig. 7A) (67,68). Both of these RMI proteins comprise a poorly characterized DUF1767 domain and an oligonucleotide-oligosaccharide-binding fold (OB-fold). OB-folds are known to be present in several proteins involved in DNA metabolism, where they mediate interaction with singlestranded nucleic acids and/or with other OB-fold-containing protein domains. These RMI-associated factors are critical for the stability of their cognate topoisomerase in vivo and should, therefore, be considered as obligatory components of the active topoisomerase (57,60,188). This contention is further supported by the structural features of their interaction with the clamp of the core topoisomerase domain of TOP3A or -B, at the hinge located between subdomains II and IV (Fig. 7B) (39, 42). RMI1 and TDRD3 are clearly able to modulate the catalytic activity of their cognate topoisomerase, although via different mechanisms. RMI1 promotes the activity of TOP3A on multiple DNA substrates, apparently by stabilizing the cleavage complex in its open state (42, 71,90). The crystal structure of truncated versions of RMI1 and TOP3A revealed that RMI1 inserts a loop into the cavity of the topoisomerase domain of TOP3A (42,189). In addition to modulating the dynamic opening and closing of the gate, this association leads to a significant reduction in the size of TOP3A cavity (Fig. 7B). This might enable RMI1 to exert some steric control over the nature of the T-segment, such as to favor single-stranded nucleic acids (71). In contrast, TDRD3 binding to TOP3B does not appear to significantly reduce the size of the topoisomerase cavity ( Fig. 7B) (39). However, it does exhibit an intrinsic affinity for both ssDNA and ssRNA and has been shown to increase the processivity of the relaxation activity of TOP3B (190).
In addition to modulating catalytic activities, the RMI1 and TDRD3 proteins are important for mediating interactions of their type IA topoisomerase partners with additional proteins. For the TOP3A partner RMI1, this involves a second OB-fold that interacts with another OB-fold protein, RMI2. Hence, together, the RMI1-RMI2 heterodimer contains three OBfolds, which mediate the interaction of TOP3A with other DNA-processing enzymes, such as the BLM and FANCM helicases (Fig. 7A) (28,189,(191)(192)(193). The TOP3B partner, TDRD3, acts as a multiprotein platform characterized by the presence of multiple domains mediating protein-protein interactions with actors in RNA metabolism, such as the fragile X mental retardation protein and the RNA-induced silencing complex (Fig. 7A) (67,68,177). In particular, the Tudor domain of TDRD3 interacts with methylated arginine modifications of histones and RNA polymerase II, which enables the recruitment of TOP3B to transcription start sites and its integration into downstream mRNA metabolism taking place in the cytoplasm (60,67,68). Interestingly, the RGG box motifs of TOP3B are known to be methylated, which promotes an interaction with the Tudor domain of TDRD3 and leads to a stimulation of TOP3B activity (171). This indicates that TOP3B and TDRD3 can interact via their respective C-terminal domains in a manner that depends on post-translational modifications, which might have relevance to how DUF1767/OB-fold domain interactions shape the structural organization of TOP3 complexes.

Conclusion and perspectives
Anyone who has wrestled with an uncooperative hosepipe while trying to water their beloved rose garden will know how frustrating the topological entanglements of a long polymer can be. If this annoying aspect of home life were as simple to solve as the problems of DNA topology in cells, we would be able simply to cut the hose and rejoin it again just as a topoisomerase does to DNA. Alas, the skill of the topoisomerase far exceeds that of JBC REVIEWS: DNA topoisomerases mere humans, and our attempts to emulate the topoisomerase would inevitably end in failure. It is remarkable how few mistakes topoisomerases make and how they have evolved more than one different mechanism to achieve topological transitions, each one being based on a simple concept that nevertheless is extremely challenging to perform rapidly and with high fidelity. Topoisomerases are indeed nature's DNA magicians.
In humans, supercoiling homeostasis is thought to result from the interplay between supercoiling-inducing processes (primarily transcription) and the supercoiling relaxation activities of DNA topoisomerases. Despite the fact that they are highly conserved enzymes and are often essential for viability, TOP3s are generally not included in this paradigm because they lack the canonical relaxation activity shared by all other DNA topoisomerases. Instead, these unique enzymes are able to efficiently manipulate the topology of a wide range of DNA and RNA substrates and are necessary during most cellular processes associated with maintenance of genomic stability.
TOP3s have been shown to cooperate with other DNA-processing enzymes to perform complex topological transactions. These illustrate underappreciated aspects of the biochemistry of DNA-processing enzymes. First, some DNA manipulating proteins, such as SNF2 translocases, could exert their physiological role through an active alteration of DNA topology and thereby contribute, alongside DNA topoisomerases, to the regulation of supercoiling homeostasis in vivo. Such activities are de facto ignored when biochemical assays are performed on DNA substrates that are not topologically constrained (such as oligonucleotides). The reintroduction of DNA topology into biochemical assays could reveal that many of the well-known or noncharacterized DNA-processing enzymes directly alter DNA topology through their catalytic activity. Second, the biochemical characterization of TOP3-associated multienzyme complexes has revealed that the canonical activities of a DNA topoisomerase can be modulated in a highly sophisticated fashion by its partners, giving rise to new topological transactions. This opens the interesting possibility that the activities displayed by stand-alone DNA topoisomerases do not reflect the actual diversity of topoisomerase-mediated transactions. We anticipate that future work focusing on the identification and characterization of topoisomerase-associated multienzyme complexes will uncover the many hidden lives that DNA topoisomerases share with their different partners.
The field of DNA topology was developed nearly 60 years ago, together with the discovery of the double-helical structure of DNA. Nonetheless, our understanding of how DNA supercoiling is regulated and exploited by eukaryotic cells still requires refinement. In particular, the repertoire of topological transactions catalyzed by topoisomerases and non-topoisomerase DNA-processing enzymes is probably more elaborate than is currently envisaged. In this context, it will be essential to develop new methods for monitoring DNA topology in vivo while refining the existing ones, to build up a more detailed and dynamic view of the topological landscape of our genomes. Altogether, these efforts will likely contribute to a renaissance in the field of DNA topology in which DNA topoisomerases are no longer considered as simple swivels that act to release torsional constraints associated with DNA metabolism, but are celebrated as versatile enzymes that promote highly sophisticated cellular functions.