J. Mol. Biol. (1993) 232, 1130-1140 Monte Carlo Analysis of the Conformation of DNA Catenanes Alexander V. Vologodskii Institute of Molecular Genetics Russian Academy of Sciences Moscow 123182, Russia and Nicholas R. Cozzarelli Department of Molecular and Cell Biology 401 Barker Hall, University of California Berkeley, CA 94720, U.S.A. (Received 11 December 1992; accepted 16 April 1993) We used a Monte Curlo method to study the conformational properties of eatenanes between two nicked DNA rings. We calculated the writhe induced by eatenation as a fanction of the linking number between the two tings. The simulations modeled eatenated rings of equal size as well as rings differing in length by a factor of 3. For both classes of catenanes, the calculated values of writhe agreed very well with the experimental measurements of cater ion-induced supercoi 1g made by Wasserman ef al. Therefore, the equilibrium value of DNA twist is not changed significantly by catenation, We found that the induced writhe increased linearly with catenane linking number, but was independent of DNA length and of effective helical diameter. We conclude that induced writhe is a general feature of catenati n, and that it depends primarily on the ratio of lengths of the linked rings and the number of vatenane interlocks, In contrast, eatenane conformation varied qualitatively with catenation linking number, DNA length, and donble helix diameter. At the values of these parameters for catenanes’ isolated from cells, catenane conformations were strikingly irregular, Nonetheless, the local concentration of two sites on separate but linked rings inoreased greatly with catenane linking number. This increase is similar to that brought about by (—) superes Ke ing to DNA sites in cis. pords: DNA; catenanes; Monte Carlo; supercoiling 1. Introduction Linked circular duplex DNAs, or catenanes, are a common cellular feature (Fig. I). They were d covered in 1967 in Vinograd’s laboratory as natur- ally occurring linked dimers in the mitochondria of ‘malignant cells (Hudson & Vinograd, 1967; Clayton & Vinograd, 1967). Catenanes were subsequently shown to be intermediates in the terminal stage of replication of ciroular DNA (Sundin & Varshavaky 1981) that accumatate when a type 2 topoisomerase is inhibited (Yanagida & Sternglanz, 1990; Adams et al., 1992). Catenanes also arise from site-specific recombination between two directly’ repeated sites (Bliska & Cozzarelli, 1987). Pethaps the striking example of catenation is found in lasts, the mitochondria of eortain unicellular par ‘0922-2836 /95/16) 130-11 508.00/0 sites. Kinetoplast DNA is a network of thousands of linked rings (Ryan et al., 1988) With the exception of the mitochondri cellular catenanes are short-lived. Some feature of catenanes results in their instantaneous untin! by topoisomerases in vivo (Bliska & Cozzarelli, 1087). ‘This is in contrast to DNA supercoiling, which is maintained at a constant level by topoiso- merases (Wang, 1991; Drlica, 1992 Although ina strict sense only circular DNA can be eatenated, linear DNA in vita is similarly linked because it is divided into topologically constrained domains (Stonington & Pettijohn, 1971; Benyajati & Woreel, 1976; Paulson & Laemmli, 197). Catenated rings provide an excellent model for such ‘tanyled linear chromosomes: in yeast, both plas: mids and chromosomes remain entangled in the forms, © 1003 Academie Prost LimitedMonte Carlo Analysis of DNA Catenane Conformation 1131 Figure 1. A torus eatenane with a regular geometr ‘Two nicked duplex DNA rings are shown as black and stippled tubes winding around a grey torus. Torus eate- naiies are rigorously defined topologically (Rolisen, 1976). "They are so named because each ring can lie on the surface of a torus without overlapping of DNA segments In the example shown, each ring winds 4 times in a right haded helical fashion around the torus. The cutenane lin cing number. ZE is equal to half of the algebraic sum of cercssings of the 2 DNA rings in any projection. 4 is equal to 4 in the example shown, ‘The sign of 24 depends on the relative orientation of the 2 rings, and we arbi trorily ascribe a parallel orientation. completely regular keometry is shown, but the molecule remains a torus catenane with Af of 4 after any deformation short of backbone breakage and reunion absence of topoisomerase IT Stornglanz, 1990) In addition to their biological significance, cate hanes are interesting structures in their own right as well as useful biochemical tools. The constituent rit gs of a DNA eatenane have no direct physical or chomical interaction, but. the topological linkage nonetheless changes the conformation of each ring (Wasserman ef ai., 1988). Moreover, the requirement that two DNA sites be on the same DNA molecule (ir cis) can often be met by sites on separate (in ircns) but catenated rings (Dunaway & Drige, 1989: Wedel ef af., 1990). Particularly intriguing from the standpoint ‘of enzyme mechanism and catenane structures is the requirement in some cases for the two recombination sites to be on multiply linked catenanes with @ particular handedness and orien- ta.ion Craigie & Mizuuchi, 1986; Kanaar et al 1989; Benjamin & Cozzarelli, 1990; Adams et al., 1992) The topology of DNA catenanes generated from a variety of processes has been determined by elee- tron microscopy (reviewed in Wasserman & Cozzarelli, 1986; Drage & Cozzarelli, 1992). There are hcwever, little experimental data about eatenane conformation. High-resolution methods such as X ray erystallography and nuclear magnetic res aree cannot be used because of the large size and flexible structure of eatenated DNA. Hydrodynamic methods stich as sedimentation and light-scattering give only average information about overall shape. (Yanagida & Electron microscopy is potentially valuable, but thus far has not heen as informative for eatenane conformation as for the study of DNA supercoiling because the observed structures are irregular and the component rings are difficult to resolve (Levene. 5. D., Donahue, C. & Cozzarelli, N. R.. unpublished results). Due to the absence of suitable experimental methods, it is unknown whether the links that maintain a catenane are close together or dispersed, whether catenanes have a structural regularity, and the extent to which their structure is similar to that of supercoiled DNA ‘The conformations of large flexible molecules can, however, be analyzed effectively by computer simu- lation. We have successfully applied this technique to demonstrate that the computed and measured properties of (—) supercoiled DNA are the same within the error limits of the methods (Vologodskii et al., 1992). The properties examined in these simu- lations inelude writhe, the number of superhelical turns, and the length of the supethelix axis. We have now simulated DNA eatenanes using the Monte Carlo procedure of Metropolis et al. (1953) that was also used in our study of supercoiled DNA. The subject of our study is the family of eatenanes known as torus eatenanes. A defining topological feature of these forms is that they ean be positioned without overlap on the surface of a torus, a doughnut-shaped object. Tn a right-handed torus catenane, as shown in Figure 1, the two molecules wind in right-handed helices around each other. Catenanes found in biological systems often belong to the right-handed torus family (Wasserman & relli, 1986), ‘We chose to study the conformations of eatenanes between two nicked DNA rings, because the eontri- butions of catenation to DNA conformations are easier to detect when the complicating effects of supercoiling are removed. Moreover, eatenanes of interrupted DNA are intermediates in DNA replication ‘A Monte Carlo simulation, in principle, generates an equilibrium distribution of molecular eonforma- tions from which one can extract as detailed information about the system as desired. In pra tive, however, all simulations are based on a simpli- fied made! of the system and involve computations of finite length. As a result, important features of real molecules can be missed. Therefore, it is essen tial to determine the validity of simulations by a comparison with quantitative experimental data For the present study, the experimental results of Wasserman et al. (1988) are particularly appro- priate, These workers measured the supercoiling induced by eatenation as a. funetion of the number of catenane interlocks and the relative size of the two DNA rings The very good agreement we found between our Monte Carlo results, these experimental data, and theoretical expectations supports the conclusion that our simulations reflect the true conformational equilibrium of DNA eatenanes. We found that eate- nanes generally have an irregular conformation but1132 Monte Carlo Analysis of DNA Catenane Conformation that the mutual distortions of the paths of the rings give rise to characteristic values of average writhe. ‘The writhe increased linearly with the number of interlocks, and for catenanes between equal-sized rings its value was insensitive to DNA length and effective diameter. Multiple interlocking of cate- nanes increased the local concentration of a DNA site in one ting relative to a site in a catenated ring by almost two orders of magnitude. The simulations also allowed us to identify three classes of catenane conformation whose structure depends on ring size and on the number of eatenane interlocks. 2, Simulation Procedure ur approach was based on the Metropotis Monte Carlo procedure that has previously been used successfully in. the study of DNA supercoiling (Klenin et al., 1989, 1991; Vologodskit et al., 1992). These papers describe the methods in detail, here we just summarize the main features of the procedure. ‘The DNA double helix is modeled as a chain consisting of nk rigid cylindrical segments of equal length and of diameter d, where & is the number of segments in a Kuhn statistical length and x is the length of DNA expressed in Kuhn lengths. The elastic energy of the chain, B, is specified by the equation: E=RTay 0, w nis over all the joints between the rigid segments, Ris the gas constant, 7’ is the absolute temperature, 0s the angular displacement (in radians) of segment. relative to segment i, and a is the bending rigidity constant for DNA. The parameter a is defined 50, that there are & segments per Kuhn statistical length (Frank-Kamenetskii et al., 1985). There is no torsional stress between the segments of the nicked DNA that was simulated ‘The connection between an actual DNA molecule and a ‘model chain is specified by 2 parameters, One is the Kuhn statistical length for DNA, whieh is twice the persistence length, and is equal to 100 nm (300 bp) independent of ionic conditions beyond a minimal concentration of about, 10mM monovalent ions (reviewed by Hagerman, 1988). ‘The 35kb DNA rings that we consider in this paper contain 12 Kuhn lengths. The 2nd parameter is the DNA effective diameter, The value of this parameter is larger than the geometric diameter because of the electrostatic repulsion between the charged DNA backbones. Thus, its value depends strongly on ionic conditions (Stigter, 1977: Brian ef al, 1081; Yarmola ef al., 1985; Ryhenkov et al. 1993) and was varied in the simulations from 3nm to 1Wnm, corresponding to the DNA effective diameters in 1M and 0-02 M NaCl solutions, respectively. ‘The starting conformation used in the simulations was a regular torus eatenane with a certain value of the linking number, £4, between the rings (Fig. 1). This ‘choice was made only for convenience and the final distri- bution of conformations was independent of the initial conformation. The notation for eatenane linking number is in script to distinguish it from the linking number between the two strands of an intact duplex DNA ring, which we abbreviate as Lk. To set the sign of 24, we arbitrarily define the orientation of the two rings of the simulated eatenanes as parallel; ie. the rings are con- sidered to be oriented in the same direction. -24 specifies where the summat uuniguely the topology of a torus catenane of known orientation. For a right-handed parallel torus eatenane, 6 is a positive integer. We designate catenanes in whieh the 2 rings have the same length as symmetric catenanes, and those consisting of chains of different length as asym- ‘metric catenanes. Successive deformation of the chains was then carried ‘out in accordance with the Metropolis ef al. (1953) pro- cedure. In each step of the deformation, a portion of the chain containing an arbitrary number of segments was rotated by an angle, @, around a line connecting the ends of the chain portion. The value of @ was uniformly distributed over a range chosen so that about half of the moves were accepted. ‘The trial conformation was then tested for physical overlap of chain segments. In the Metropolis procedure, & system has infinite energy if 2 non-adjacent segments of @ ‘rial conformation are closer than the value of d, the segment diameter, Therefore, if the minimum distance between any pair of non-adjacent segments of a trial conformation was less than d, the trial was rejected Because the chains were allowed to pass through each ‘other during the displacement, it was nevessary to ensure that the topology of the trial conformation was the same fas the current one. To describe the topology of the trial and current conformations, we ealculated the value of the Alexander polynomial, A(s,), at s = —1, ¢ = —1 (Rolfsen. 1976; Vologodskii et al., 1975). The Alexander polynomial «8 topological invariant that describes the linking of 2 rings of @ catenane. It has the same value for all topo- logically equivalent catenanes. Any 2 catenanes with a rent value of the Alexander polynomial are different. ‘The value A(s,) for @ particular eatenane conformation can be ealeulated from a projection of the chains into a plane (Vologodskii et ai., 1975). For torus catenane AIA(—1, —D] =L24l. Therefore, if a trinl conformation had a value of |A(—1, —1)] different from that of the current conformation, the displacement must have changed catenane topology and thus the trial was rejected Tf the trial conformation passed the excluded volume and topology tests, its clastic energy was calculated in accordance” with equation (1). ‘The probability of accepting a trial conformation was calculated by applying the classical rules of the Metropolis eal. (1953) procedure. If the energy of the trial conformation, Bug, Was lower than that of the current conformation, By, then the trial conformation was accepted, Tf the energy of the trial conformation was greater than the energy of the current conformation, then the trial conformation was accepted with a probability equal to expl(E.gy—B)/RT] The analysis of chain writhe, Wr, has an important place in our work. JP is a funetion of the geometry of the double helix axis alone (White, 1989; Vologodskii, 1992), and is independent of DNA twist, Tw, for a nicked DNA. We used the algorithm of Le Bret (1980) in order to calculate the Ir of particular chain conformations. Our DNA model is an approximation of the continuous wormlike chain model; the approximation improves as k. the number of segments per Kuhn length, increases, Because the computer time for simulations increases roughly with the second power of &, we chose a value of & lange enough to ensure reliabte results but small enough to keep the computation time within reasonable bounds, Even for highly supereoiled molecules, a k value equal to 19 is sufficient for reliable determinations of Wr (Vologodskii ef al, 1992). Accordingly, & was set equal to 10 in this work and the total number of segments was, usually 240.