What the papers say

Knot what we thought before: the twisted story of replication

Lisa Postow, Brian J. Peter, and Nicholas R. Cozzarelli*
Summary

DNA replication requires the unwinding of the parental duplex, which generates ( ) supercoiling ahead of the replication fork. It has been thought that removal of these ( ) supercoils was the only method of unlinking the parental strands. Recent evidence implies that supercoils can diffuse across the replication fork, resulting in interwound replicated strands called precatenanes. Topoisomerases can then act both in front of and behind the replication fork. A new study by Sogo et al. [J Mol Biol 1999;286:637643 (Ref. 1)], using a topological analysis, provides the best evidence that precatenanes exist in negatively supercoiled, partially replicated molecules in vivo. BioEssays 21:805808, 1999. 1999 John Wiley & Sons, Inc.
Introduction Replication of a topologically bound DNA molecule causes a host of topological problems. Most notably, as helicases unwind the parental strands at the replication fork, ( ) supercoils build up in front of the moving fork (Fig. 1A). These ( ) supercoils must be removed by topoisomerases for replication to continue. As replication advances, the length of unreplicated, ( ) supercoiled DNA will progressively get shorter, offering a smaller and smaller substrate for the action of topoisomerases. Eventually there will not be a long enough stretch of unreplicated DNA for the topoisomerases to act upon; the cell will have painted itself into a corner. Champoux and Been(2) proposed a way out of this dilemma. They pointed out that supercoils in front of a replication fork should be able to diffuse across the fork and take the form of interwound replicated DNA (Fig. 1B). These interlinks have been termed precatenanes,(3) because if they are not removed by the end of replication, they will result in catenated products (two circular molecules linked to each other). The ( ) precatenanes behind the fork should be in equilibrium with the ( ) supercoils ahead of the fork, as diagrammed in Figure 1B. These precatenanes could be removed by decatenating topoisomerases during replication, allowing topoisomerases to act over the entire replicating DNA region, rather than merely on the shrinking unreplicated stretch. While the idea of precatenanes is very attractive, early experiments on plasmid and viral replication implied that it was false. Electron microscopy (EM) of theta-type replication intermediates using the Kleinschmidt method of spreading DNA(4) showed that the DNA in front of the fork was supercoiled, but that the replicated region was not interwound at all.(57) Textbook descriptions of DNA replication have used these examples to claim that supercoils only build up in front of the replication fork.(8) A recent article by Peter et al.(9) showed, however, that puried partially replicated molecules produced both in vitro and in vivo had the expected precatenane links in the replicated region in addition to the supercoiled unreplicated region, as diagrammed in Figure 1B. This view is supported by studies which imply that decatenationlike reactions are important during replication in vitro.(10,11) The problem remained, however, that these structural and enzymatic results did not prove that precatenanes exist inside the living cell. It was possible that the replication fork is not free to rotate in vivo and that all the ( ) topological strain is, in fact, ahead of the fork until deproteinization allowed it to spread throughout the molecule and form precatenanes. Sogo et al.(1) addressed this issue of replication intermediate structure inside the cell, and deduced from the topology of knotted partially replicated molecules that precatenanes do exist in vivo. We will discuss next in more detail the recent articles supporting the importance of precatenanes in replication intermediates: rst, enzymatic studies that indicate that decatenating topoisomerases are important for replication in vitro; second, structural studies of naked DNA that offer direct

Department of Molecular and Cell Biology, University of California, Berkeley, California. Funding agency: NIH; Grant number: GM31655. *Correspondence to: Nicholas R. Cozzarelli, Department of Molecular and Cell Biology, University of California, Berkeley, CA 94720. E-mail: ncozzare@socrates.berkeley.edu

BioEssays 21:805808,

1999 John Wiley & Sons, Inc.

BioEssays 21.10

805

What the papers say

evidence that precatenanes and supercoils are able to diffuse across a replication fork; and nally, the results of Sogo et al. showing that precatenanes likely exist inside a living cell.

Enzymatic studies in vitro indicate that precatenanes are important in unlinking during replication The rst indication of a role for precatenanes in replication came from studies by Hiasa and Marians.(10,11) They studied in vitro replication of plasmid DNA using puried Escherichia coli proteins to determine the requirements for the two classes of topoisomerases in that system, type 1A and type 2 (reviewed by Wang(12)). They concluded that unlinking was taking place both ahead of and behind the replication fork. First, Topoisomerase (Topo) III will support complete replication and decatenation of a negatively supercoiled plasmid in the absence of any other topoisomerases.(11) Since Topo III is a type 1A topoisomerase, and thus cannot relax positive supercoils, Hiasa and Marians concluded that this enzyme supported replication by unlinking precatenanes at singlestranded regions behind the replication fork. Topo IIImediated replication began with a lag not present in DNA gyrase-mediated replication. The lag was interpreted to result from precatenanes forming slowly at the beginning of replication and more readily as the replicated region formed a larger proportion of the molecule.(10) While this interpretation may indeed be correct, Kowalczykowski and coworkers have provided an alternative possibility.(13) They showed that duplex DNA circles can be catenated by Topo III in the presence of a DNA helicase. Thus, it is possible that the helicase may provide a single-stranded region for removal of ( ) supercoils ahead of the replication fork. It was demonstrated previously that a type 1A topoisomerase can use single-stranded regions to remove ( ) supercoils.(14) This interpretation does not explain the lag in elongation that Hiasa and Marians observed, and the conclusion that Topo III unlinks precatenanes still seems to be the more attractive hypothesis. A second argument for the existence of precatenanes during in vitro replication involves the salt sensitivity of the type 2 topoisomerases, DNA gyrase and Topo IV.(10) DNA gyrase is a unique topoisomerase in that it removes ( ) supercoils and inserts (-) supercoils; thus, it is very efficient at unlinking during replication. Topo IV, alternatively, excels at decatenation. Decatenation by Topo IV is salt sensitive, while ( ) supercoil removal by both DNA gyrase and Topo IV is less sensitive to salt concentration. Hiasa and Marians found that while early replication was relatively insensitive to salt, late replication became increasingly salt sensitive, implying that decatenation-like reactions may be important during late replication. These in vitro replication studies suggested that ( ) precatenanes exist during replication elongation as an alternative conformation to ( ) supercoils. However, more direct structural evidence was needed.

Structural studies involving stalled replication forks Structural studies of replication intermediates were made more facile by the generation of a homogeneous population with replication forks stalled at a dened point. These partially replicated molecules model actual intermediates of replication. Partially replicated plasmids of the same molecular weight will migrate on agarose gels as ladders of topoisomers molecules that differ only in linking number or knot structure. Two methods have been used recently to prepare a population of such molecules: the Tus/ter system,(9) which uses the protein Tus to stop replication at a ter site, and opposing ColE1 replication origins,(15) which begin replication from one origin and stall upon reaching the second origin. The E. coli Tus protein will prevent replication fork progression in a polar manner when bound to a ter site.(16,17) When Peter et al.(9) analyzed deproteinized plasmids in which replication had been stalled in this fashion, they found (-) supercoils in the unreplicated region as well as (-) precatenanes in the replicated region. All such molecules they observed using the polylysine spreading method for EM(18) showed supercoils and precatenane crossings. In addition, agarose gel electrophoresis showed that extensively replicated molecules migrate like catenanes, whereas those with a shorter replicated region have a mobility more like supercoiled plasmids, a result that supports the view that precatenanes become increasingly important as replication continues. Peter et al. found that earlier studies(57) missed precatenanes because of an artifact of the EM spreading procedure used. While this structural analysis of naked DNA offered incontrovertible evidence that precatenanes can exist in partially replicated molecules, the question remained whether DNA inside the cell, complete with proteins, could also assume this structure.

Topological analysis of knotted, partially replicated molecules implicates precatenanes in vivo In a plasmid with two opposing ColE1 origins, replication will initiate from one origin and pause when it encounters the other origin.(15) This arrest may result from an inability of the DnaB helicase to separate the RNA/DNA duplexes that initiate ColE1 DNA replication.(19) A subset of these arrested plasmids migrates on two-dimensional agarose gels as a series of spots that were shown to be knotted. These knots are, almost certainly, a byproduct of type 2 topoisomerase action on accumulated replication bubbles inside the cell and do not have a function during replication. However, they do provide an opportunity to study the structure of replication intermediates by topological analysis. Topological analysis uses changes in DNA topology to deduce an enzymatic mechanism or DNA structure.(20) This is an extremely powerful method, since topology is invariant

806

BioEssays 21.10

What the papers say

Figure 1. Partition of ( ) superhelical strain in vivo. During replication from a unidirectional origin (ori), helicase action on the parental Watson and Crick strands (shown as blue and red parallel lines) creates a positive superhelical stress on the DNA. This results in ( ) supercoils in front of the fork, as shown in A. However, this tension may equilibrate across the replication fork to create also ( ) intertwinings of the daughter duplexes, or precatenanes, as shown in B. Note that each ( ) supercoil node in the unreplicated DNA is topologically equivalent to two ( ) precatenane nodes, as each daughter duplex contains only one parental strand. The daughter strands (shown in grey) are incomplete and therefore topologically irrelevant in puried DNA. Figure 2. Generation of ( ) twist knots from (-) precatenanes by a type 2 topoisomerase in vivo. Inside a cell, stalled partially replicated molecules may accumulate (-) supercoils in the unreplicated region and (-) precatenanes in the replicated region. Here, replicated duplexes are illustrated as single red or blue lines, whereas parental duplexes are shown in purple. Shown at the top is a random movement of the red strand (illustrated by black arrows) which causes it to fold over. The same could occur between the blue strands. Topoisomerase-catalyzed strand passages in this overlap region will capture the topology of precatenane nodes, forming two series of twist knots, depending on which of the two overlap nodes is inverted by the topoisomerase. On the left is shown the simplest example of one series of knots which has 2 (-) nodes and 2n ( ) plectonemic nodes. On the right is shown the simplest example of the series of twist knots which has 1 2n ( ) nodes. In both examples shown, n 1. The analysis of the knots is simplied by cleaving off the unreplicated region at restriction sites, shown here as white arrowheads. The twist knots are generated from 2n precatenane crossings captured between the overlapping segments of the red strand. Note that the (-) precatenane nodes are converted to ( ) knot nodes, purely because of the convention for assigning node sign.(24) In precatenanes and catenanes, node sign is assigned assuming the constituent DNA segments (red and blue here) are parallel to each other. For knots, the red and blue strands are antiparallel as they are considered to be part of the same circle.

unless the DNA is broken. The mechanism of a reaction or the DNA substrate structure can be deduced by observing which of the enormous number of topological forms possible for a molecule are actually found. This method has been used to study the mechanism of recombinases(21) and topoisomerases.(22) A knot formed by a topoisomerase reects the number and pattern of DNA crossings trapped between the two segments that participate in the strand passage event. The topology of a knot can be determined by coating the DNA with RecA protein followed by EM examination.(23) A knot can only exist in a closed circle. Theta-type replication intermediates can be cleaved in the unreplicated region, and the replication bubble will still be a circle (Fig. 2). Sogo et al.(1) used this to their advantage, cleaving off all but the replication bubble in order to simplify the analysis of the knots. This cleavage will remove all supercoils and precatenanes. If there were no precatenane crossings of the replicated DNA, as depicted in the textbooks, then knotting in the replication bubble would be rare. Instead, extra spots were found on two-dimensional gels of partially replicated DNA, and shown to be knots by EM.(1,15) Most of the knots

BioEssays 21.10

807

What the papers say

that Sogo et al. examined were part of a series of ( ) twist knots, which implies that they resulted from topoisomerase linking in vivo of (-) precatenated replication bubbles. Their explanation for this result is diagrammed in Figure 2. DNA gyrase acting on the stalled molecules inside the cell will result in (-) supercoils in front of the fork and (-) precatenanes behind it. If a type 2 topoisomerase links these precatenanes to form a knot, it will generate a series of twist knots, as shown in Figure 2. Twist knots consist of a plectonemic region that is composed of either all ( ) or all (-) nodes, and a linker region that can contain either two ( ) or two (-) nodes. In this case, the twist knots have a ( ) plectonemic region even though they are from (-) precatenanes, as explained in the legend to Figure 2. The knots examined by Sogo et al. indicate that stalled, partially replicated molecules inside living cells have negative precatenanes in their replicated regions. Precatenanes and replication While the above experiments provide good evidence for a role for precatenanes in replication, there are some limitations to the conclusions that can be drawn from them. One problem is that DNA gyrase continues to catalyze strand passages on the partially replicated molecules used in the structural studies, introducing (-) supercoils and (-) precatenanes. These molecules cannot be true replication intermediates, as removal of a (-) precatenane or a (-) supercoil by a topoisomerase will act in opposition to strand separation. Therefore, one major disadvantage of using stalled replication forks to study replication intermediate structure is that the behavior of these molecules may be signicantly different from that of actively replicating, ( ) supercoiled plasmids. Another problem is that, for a number of reasons, the structure of a moving replication fork may be different from that of a stalled one. Finally, while the interpretation of the origin of the knots by Sogo et al. is reasonable, alternatives cannot be ruled out. In our view, however, the studies discussed above complement each other in their support of the existence of precatenanes, making it very likely that they are important during chromosomal replication.

References
1. Sogo J, Stasiak, A, Martnez-Robles ML, Krimer DB, Hernandez P, Schvartzman JB. Formation of knots in partially replicated DNA molecules. J Mol Biol 1999;286:637643. 2. Champoux JJ, Been MD. Topoisomerases and the Swivel Problem. In: Alberts B, editor. Mechanistic studies of DNA replication and genetic recombination. New York: Academic Press; 1980. p 809815.

3. Ullsperger CJ, Vologodskii AV, Cozzarelli NR. Unlinking of DNA by topoisomerases during DNA replication. In: Lilley DMJ, Eckstein F, editors. Nucleic acids and molecular biology. Berlin: Springer-Verlag; 1995. p 114142. 4. Kleinschmidt AM, Martinson HG. Structure of nucleosome core particles containing uH2A (A24). Nucleic Acids Res 1981;9:24232431. 5. Sebring ED, Kelly TJ, Thoren MM, Salzman NP. Structure of replicating simian virus 40 deoxyribonucleic acid molecules. J Virol 1971;8:478490. 6. Fuke M, Inselburg J. Electron microscopic studies of replicating and catenated colicin factor E1 DNA isolated from minicells. Proc Natl Acad Sci USA 1972;69:8992. 7. Roman A, Champoux JJ, Dulbecco R. Characterization of the replicative intermediates of polyoma virus. Virology 1974;57:147160. 8. Kornberg A, Baker TA. DNA replication, 2nd ed. New York: W.H. Freeman; 1992. 9. Peter BJ, Ullsperger C, Hiasa H, Marians KJ, Cozzarelli NR. The structure of supercoiled intermediates in DNA replication. Cell 1998;94:819827. 10. Hiasa H, Marians KJ. Two distinct modes of strand unlinking during theta-type DNA replication. J Biol Chem 1996;271:2152921535. 11. Hiasa H, Marians KJ. Topoisomerase III, but not topoisomerase I, can support nascent chain elongation during theta-type DNA replication. J Biol Chem 1994;269:3265532659. 12. Wang JC. DNA topoisomerases. Annu Rev Biochem 1996;65:635692. 13. Harmon FG, DiGate RJ, Kowalczykowski SC. RecQ helicase and topoisomerase III comprise a novel DNA strand passage function: a conserved mechanism for control of DNA recombination. Mol Cell 1999;3:611620. 14. Kirkegaard K, Wang JC. Bacterial DNA topoisomerase I can relax positively supercoiled DNA containing a single-stranded loop. J Mol Biol 1985;185:625637. 15. Viguera E, Hernandez P, Krimer DB, Boistov AS, Lurz R, Alonso JC, Schvartzman JB. The ColE1 unidirectional origin acts as a polar replication fork pausing site. J Biol Chem 1996;271:2241422421. 16. Hill TM, Kopp BJ, Kuempel PL. Termination of DNA replication in Escherichia coli requires a trans-acting factor. J Bacteriol 1988;170:662 668. 17. Khatri GS, MacAllister T, Sista PR, Bastia D. The replication terminator protein of E. coli is a DNA sequence-specific contra-helicase. Cell 1989;59:667674. 18. Williams RC. Use of polylysine for adsorption of nucleic acids and enzymes to electron microscope specimen films. Proc Natl Acad Sci USA 1977;74:23112315. 19. Santamara D, de la Cueva G, Martnez-Robles ML, Krimer DB, Hernan dez P, Schvartzman JB. DnaB helicase is unable to dissociate RNA-DNA hybrids. J Biol Chem 1998;273:3338633396. 20. Wasserman SA, Cozzarelli NR. Biochemical topology: applications to DNA recombination and replication. Science 1986;232:951960. 21. Crisona NJ, Weinberg RW, Peter BJ, Sumners D, Cozzarelli NR. The topological mechanism of phage integrase. J Mol Biol 1999;289:747 775. 22. Wasserman SA, Cozzarelli NR. Supercoiled DNA-directed knotting by T4 topoisomerase. J Biol Chem 1991;266:2056720573. 23. Sogo J, Stasiak A, DeBernardin W, Losa R, Koller T. Binding of proteins to nucleic acids as studied by electron microscopy. In: Sommerville J, Scheer U, editors. Electron microscopy. Oxford: IRL Press; 1987. p 6179. 24. White JH. An introduction to the geometry and topology of DNA structure. In: Waterman MS, editor. Mathematical methods for DNA sequences. Boca Raton, FL: CRC Press; 1989. p 225253.

808

BioEssays 21.10