Vous êtes sur la page 1sur 8

Elastic model of DNA supercoiling

Nathaniel G. Hunt

in the infinite-length


Department of Ph)?sics,University of California, Division of Chemical Biodynamics, Lawrence Berkelq Laboratory, Berkeley, California 94720

John E. Hearst
Department of Chemistry, University of California, Division of Chemical Biodynamics, Lawrence Berkelqv Laboratory, Berkeley, California 94720

(Received 5 August 199 1; accepted 6 September 199 1) The energies of plectonemic and toroidal supercoiled DNA are calculated by treating DNA as an elastic rod with a finite radius. End effects are ignored and all extensive quantities (e.g., writhe, bend energy) are treated as linear densities (writhe per unit length, bend energy per unit length). Minimum energy configurations are found. For plectonemic DNA, the superhelical pitch angle czis in the range 45< ag90. For low values of specific linking difference, most superhelicity is in writhe. As specific linking difference increases, a greater proportion of superhelicity is in twist. Under physiological conditions, roughly 8X% of superhelicity is in writhe. Ionic strength effects are discussed, and it is found that variation of excluded volume with ionic strength has a large effect, resulting in significantly greater torsional stress in supercoiled DNA at low ionic strength. For biologically relevant values of specific linking difference, the plectonemic conformation is energetically favored over toroidal conformations. Results are compared with electron microscopy data. The application of the model to DNA conformational transitions is discussed.

1. INTRODUCTION Differences in DNA linking number ALk from the equilibrium linking number Lk, can be distributed between changes in twist, ATw, and writhe, Wr, of the DNA axis according to the equation1 2 (1) DNA with a nonzero ALk has net writhe and twist, resulting in torsional stress and bending of the molecular axis around itself in three dimensions. This latter effect, due to writhe, gives rise to the terms supercoiling and superhelicity to describe an excess or (more commonly) a deficit of linking number. DNA supercoiling plays a role in many important phenomena, including DNA conformational transitions,3-9 recombination, transcription, replication, and chro matin structure.13 There have been several previous studies of elastic models of DNA supercoiling. Numerical1620 and analytica12 studies based on the theory of elasticity have all been -26 for finite lengths of DNA and have included end effects. Including end effects gives greater accuracy than the longlength approximation used. in this work, but the results are not as general as those derived here. Monte Carlo studies of the writhe of closed circular DNA molecules7-36 may complement the results of this paper, since they include entropic effects, whereas this study does not. The results presented here are valid for relatively tight structures with large positive or negative specific linking differences. These structures have only small entropic contributions to their free energies. There is one previous study, by Camerini-Otero and ALk = ATw + Wr.

end effects in order to derive results of general validity.37 Camerini-Otero and Felsenfeld came to different conclusions from this paper, but we believe this is because of errors in their work mentioned in Sec. V of this paper. Many of the results in this paper can be summarized by two main points. The first is that the writhe per unit length, pWr, for a plectonemic helix has a maximum value,pWr,,, . For low values of specific linking difference 1~1 (a= ALk/Lk, ), most superhelicity can be taken up in writhe at very little energetic cost. At greater values of 101, where the writhe density is a substantial fraction ofpWr,, , superhelicity goes increasingly into twist, at greater energetic cost. The energy of supercoiling which results is not simply a quadratic function of specific linking difference, as has often been assumed based on extrapolations of measurements made over limited ranges of specific linking difference.384 The second main point is the importance of DNA excluded volume. Osmotic pressure measurements of the DNA excluded volume radius R, show that it is a strong function of ionic strength, varying from about 15 A at high ionic strength to over 100 A at low ionic strength.4244 pWrmax is 1/(4rR,, ), so that six times more superhelicity can be taken up by writhe at high ionic strength than at low ionic strength. Thus, at low ionic strength, superhelicity starts going into twist at much lower values of 1~1,resulting in significantly greater torsional stress. Since torsional stress can induce DNA conformational transitions,3-9 this effect may account for some of the ionic strength dependence of transitions induced under superhelical stress at low ionic

Felsenfeld,whichis very similar to this one in that it ignores


J. Chem. Phys. 95 (12), 15 December 1991 0021-9606/91/249329-08$03.00 @ 1991 American Institute of Physics 9329 Downloaded 21 Oct 2007 to Redistribution subject to AIP license or copyright, see http://jcp.aip.org/jcp/copyright.jsp


N. G. Hunt and J. E. Hearst: Elastic model of DNA supercoiling

II. FORMULATION A. Bend energy, twist energy, writhe, and twist Giving a DNA molecule twist or writhe requires energy. In this paper we calculate the bending and twisting energies of two DNA configurations: the interwound form, or plectonemic helix (Fig. 1); and the toroidal or solenoidal form (Fig. 2). In both cases we consider the limit of long molecules and ignore end effects. All extensive quantities (e.g., writhe, bend energy) are treated as linear densities (writhe per length, bend energy per length, etc.). DNA is treated as an elastic rod characterized by a torsional rigidity Ptwist, a bending rigidity fibend, and an excluded volume radius R, . Changing the twist from its equilibrium value has an energy cost per unit length

PEtwist @twist/21 = (2~)Z@AT~)2,


wherepEtwist is the twist energy per unit length andpATw is the deviation from the equilibrium twist per unit length. The factor of (2~)~ converts twist from turns to radians, as required by the usual definition Of Ptwist. The axis of a DNA molecule with net writhe curves in three dimensions. The bending energy per &it length, ~-%end of a DNA molecule is 9 (3) ~&end u3bend/2W, = where&,, is the bending constant and K is the curvature of themolecular axis. Changing the writhe of a DNA molecule changes the bend energy of the molecule, but there is no general relation between writhe and bend energy such as the relation Eq. (2) between twist and twist energy. The energy cost of adding or subtracting writhe depends on the molecule specific geometry and is different for the plectonemic s and toroidal configurations.

FIG. 2. Long-length limit of toroidal DNA. Cross-section showing excluded volume radius of the DNA. R, superhelix radius; R,,, excluded volume radius; LY,superhelix pitch angle; 27rRtan a, superhelix pitch.

B. Writhe per unit length The writhe of a space curve r(u) , where u is some parametrization of the curve, is given by 329 dr(u) x dr(u ) ~ du du Wr=J.du du 4n- s C ,C [r(S) -r(u )13 (4) Writhe is a global property of a curve. If u is the arc length rather than an arbitrary parametrization, we can define a local quantity, the writhe per unit length, by [r(zY) -r(d)] [r(uo) pWr =-& du I C - r(d)] [duo) (FJ, -r(u )]3 (5) For the infinite helices considered in this paper,pWr is independent of position on the curve, since the helices are spatially homogeneous modulo a translation and rotation, neither of which affect the writhe integral. The curve r (u) is taken to be the axis of the DNA double helix. For a discussion of other possible choices, see Sec. V. Fuller gave the writhe per turn for plectonemic and single helices, based on geometrical arguments (and ignoring end elects) . For the plectonemic helix, (6) where a is the pitch angle of the helix. For the single helix, (7) The sign of the writhe depends on the handedness of the helix. The writhe of a plectonemic helix is positive if the helix is left-handed, while the writhe of a single helix is positive if it is right-handed. In this paper we take a, cr, andpWr as positive. The results for a<0 are identical to those for 1~1,the only difference being the handedness of the helix. One turn of a helix of radius R has length 29~R/cos a. The writhe per unit length for a plectonemic helix is, therefore, W&urn= f (1 -sina). W&urn = f sin a, x!!!I$?*

2%!x c5 l-

2 n R&an

FIG. 1. Finite plectonemic helix with end loops. Each strand represents the DNA double helix. a, superhelix pitch angle: an-R, tan a, superhelix pitch; R,,, excluded volume radius.

J. Chem. Phys., Vol. 95, No. 12,155 December 1991

Downloaded 21 Oct 2007 to Redistribution subject to AIP license or copyright, see http://jcp.aip.org/jcp/copyright.jsp

N. G. Hunt and J. E. Hearst: Elastic model of DNA supercoillng


pWr =

sin a cos a 273-R

. 120 4 a 9) 6 E A

while for a single helix

(1 -sina) cosa -. (9) 277-R Equations (8) and (9) are exact for infinite helices, but only approximate for finite helices. The end effects can be calculated by comparing Eqs. (8) and (9) with the exact results given by Eq. (5). For helices of more than two turns, end effects are negligible. Therefore the results of this paper can be applied to finite helices without altering the expressions for pWr. pWr = III. ENERGETICS OF THE PLECTONEMIC HELIX Fuller pointed out in 1971 that a plectonemic helix has a high value of writhe per bend energy. In fact, the energy cost can be made arbitrarily small if the two strands are allowed arbitrarily close, i.e., if excluded volume effects (and end effects) are ignored. The plectonemic helix was therefore a promising candidate for the tertiary structure of superhelical DNA. One can also arrive at the interwound or plectonemic helix as a plausible structure by twisting a string or piece of tubing.26 For the plectonemic helix, R is taken to be the DNA excluded volume radius R,, . The curvature K of a helix with radius R,, is
K = cm2








l?IG. 3.p&,,,, (exx s),pE,,ist (squares), and@&,, (triangles) as functions of superhelix pitch angle. p&,, is zero when the writhe density pWr equals the linking difference per length, g/ho. R,, = 30 A; CT 0.05; = 9ergcm. -P B bend w&f =2X10- -


(10) (11)

so the bend energy per unit length pEbend is pEbend = Pbendcos4 a/2R zx. Using Eq. ( 1) in the form (12) where h, is the equilibrium helical repeat (of the DNA double helix, not of the superhelix) in units of length and the total energy per unit length, pEti,, for the plectonemic helix is PE,O, = p~%end+ retwist

cheap. As g increases, aQP, decreases towards 45 where writhe is more costly, butpWr is greater. Any configuration with a < 45 (Ref. 49) is not a minimum energy configuration, since there is then a configuration with pitch angle a = 90 - a which has the same pWr (and therefore the same twist energy) but with lower bend energy. The model can be parametrized in terms ofpWr in place of the pitch angle a. Using Eq. (8) we have cosz a = 1 - ( 1 - 16&R ZXpWr2) 2 (16)

pA Tw = a/h, -pWr,




Avist --y-


d, 60 d

2 .


pE,,, is plotted in Figs. 3-5 as a function of a for different values of c and R,, . The energy minimum for the plectonemic helix always lies in the pitch angle range (14) 45 -c aopt 690, where aopt denotes the pitch angle corresponding to minimum total energy. This is because the writhe per unit length, pWr [ Eq. (8) 1, has its maximum pWrmax= 1/(473-R,, ) at a = 4Y, whereas the writhe per bend energy per= Rex sin a -=Wr (15) E bend P&end rffibbend a cos3 becomes infinite as a-+ 90 For low values of superhelix den. sity a, aopt will be near 90, where writhe is energetically

a "0 E 40 h

0 65 75





FIG. 4. $I,,, as a function of pitch angle for different values of (T. u = 0 (Squares); (T= 0.02 (plus s); C= 0.04 (triangles); (T = 0.06 (open circles); c = 0.08 (solid circles). The optimum pitch angle decreases towards the limiting value of 45 as (T increases. For all curves; R, = 30 A;

Bbcnd pttwirt 2 x 10- I9ergcm. = =

J. Chem. Phys., Vol. 95, No. 12,15 December 1991 Downloaded 21 Oct 2007 to Redistribution subject to AIP license or copyright, see http://jcp.aip.org/jcp/copyright.jsp

9332 80

N. G. Hunt and J. E. Hearst: Elastic model of DNA supercoiling

Minimum energy configurations can be found by minimizing the total energy pE,,, with respect to pitch angle a. Setting dpEt,,,/da to zero gives g _ 2ccos sina a (18) ex . m - + sin a cos a, sin a - cosL a ho where c = fibend /fitwist. Equation ( 18 ) is an expression for the pitch angle aO,, which minimizes pE,,, for a given R,, and C. Using Eq. ( 18 ) , the values of pWr, pTw, pWrh, /c, aWEtot which correspond to aopt, are plotted as functions of g in Figs. 7-9. = If &end = j?tW*St /? (c = 1 ), the aforementioned expressions have a simpler form and optimum values can be given as explicit functions of g. For example, the values of pE,,, , pWr, and a for minimum energy configurations are 2rrR (h; + 16773R~xo-~)~- ho ~Etot ==B 4R&(h; f 167?R~xa2)2 h,(h; 47ia 2 + 167?R&~~=)~ > (19) (20) (21) 2n-b = + -h; pwr=


a 0 E 2 40
n 8 5









FIG. 5. pE,,, as a function of superheIix pitch angle for different values of the excluded volume radius R, . R, = 20 8, (pluses); R,, =40 %, (circles); R, = 60 A (exx's); R, = 80 .& (triangles). For allcurves: a = 0.05; &,, = fltwin = 2 X 10 - l9 erg cm.

(h; + 16;R:,~ )=

so that, substituting into Eq. (111, 1 - (1 - 16tiR &pWr2) ~ ~%end = PM 4R gxm

a opt=G-+Tan-


-2?r=pwr= > (17)

Equation ( 17) can be thought of as the energy cost of putting superhelicity into writhe. Eq. ( 17) is plotted with the corresponding expression for twist, Eq. (2)) in Fig. 6.

Equations ( 19)~(21) do not differ much qualitatively from optimized values obtained using Eq. ( 18 > for c in the range of its experimental values (roughly between 0.67 and 1.5). They can therefore be used to get an idea of the approximate behavior of pE,,, and pWr. IV. COMPARISON OF PLECTONEMIC AND TOROIDAL STRUCTURES An alternate structure to the plcctonemic helix is a toroidal structure. In the limit of infinite length, the toroidal



d 600 60

0 0






0 0

FIG. 6. Energy cost ofwrithe (exx and twist (triangles). The twist cost is s) given by Eq. (2); the writhe cost by Eq. (17). The curve for writhe does not go beyond the value of maximum writhe density, or h, pWr,,, = 0.094 in this case. At the point pWr =pWrmax, Eq. (17) has a finite maximum, fibend 8R 2, ), but the slope of the curve is inlinite. This figure should not be /( confused with Fig. 7, which gives the distribution of twist and bend energies for energy-minimum configurations, or with Fig. 8, which gives the dist$ bution of twist and writhe for the same configurations. R,, = 30 A; 9ergcm. -P P bend- fvhf =2X10-






FIG. 7. Bend (squares), twist (exx and total (triangles) energies as s), functions of ofor minimum-energy configurations. As aincreases, a greater proportion of superhelicity goes into twist, and twist energy dominates. = R, =30&P bend ptwirt = 2 X 10 - erg cm.

J. Chem. Phys., Vol. 95, No. 12,15 December 1991

Downloaded 21 Oct 2007 to Redistribution subject to AIP license or copyright, see http://jcp.aip.org/jcp/copyright.jsp

N. G. Hunt and J. E. Hearst: Elastic model of DNA supercoiling



volume radius R,, . We will momentarily disregard excluded volume effects while finding the minimum energy configuration for the single helix. With pWr given in Eq. (9)) pEbend from Eq. ( 11) (with R in place of R, ), and @Ztwistagain given by Eq. (2), the total energy per unit length pE,,, for the single helix is B bendcos4 Q + Ptwist -y(2?r) 2R 2

P&,, =

0 0

~g ( ho

(1 -sina) cosa 2 2?rR > to zero gives


Specific linking


For fixed R, setting dpE,,,/da o 27rR;=



FIG. 8. Superhelicity in twist (h,pATw) (circles) and writhe (h,pWr) (exx as functions of c for minimum-energy configurations. The sum of s) the two quantities is LXpWr approaches pWrmax as (r appioaches infinity.

2c cos3 a sin a + cos a - sin a cos a. cos2 a + sin a - sin a (23)

Fixing a and setting dpE,,, /dR to zero gives lc,,,a +~cos a - sin a cos a. (24) 0 Equations (23) and (24) give equal values only at the nonphysical point a = 90and R = 0. Therefore no energy minima for both R and a exist inside the interval 0 < a <go, and any such minima must occur at either a = 0 or 90. At the limit a-+0, Eq. (24) becomes 2nRf= RI(I+ c)h,/2m. (25) Substituting this into Eq. (22) at a = 00 gives P-L =

R,, = 30 k&c,


structure becomes equivalent to a single helix (Fig. 2). While most experimental evidence favors the plectonemic form for DNA in solution,50*5* there have been some reports of evidence for the toroidal structure.52,s3 Fuller noted that, at sufficiently high specific linking difference, the single helix must become energetically favored over the plectonemic helix, sincepWr has a greater maximum value for the single helix than for the plectonemic helix. By calculating the energy per unit length of a. single helix and, comparing the results with those obtained earlier for the plectonemic helix, we will show that the plectonemic helix is energetically favored over the single helix for realistic values of g and excluded volume radius R,, . For the single helix, R is not equal to the excluded

2 Wbend%+J
cl+ c12hZ,

++(2V)2(;(;Te+c)) . (26)

At a = 90, all of the linking difference is in twist, so that (27) Equation (27) is greater than Eq. (26) for any-c > 0, so the global minimum in energy is given by Eq. (26) with a = 0 and R given by Eq. (25). For c = 1, the total energy of the single helix without excluded volume, Eq. ( 26 ) , can be easily compared with the total energy of the plectonemic helix, Eq. ( 19). The plectonemic helix has lower total energy until a/h, exceeds 23 2pWrmax, generally quite a large value of 1~1. When excluded volume effects are included, as outlined later, the transition to the toroidal form takes place at an even greater value of 101. The finite excluded volume radius of the DNA double helix puts a lower limit on the pitch. The pitch of a helix is defined as the displacement per turn and is 2?rR tan a. In Fig. 2 it is shown that the minimum pitch with an excluded radius R,, is 2R,,/cos a. Setting the two pitch expressions equal gives


. E % c


0.4 0.2








as a function of 0;

FIG. 9. Proportion of superhelicity in writhe, h,pWr/g,

for minimum-energy ccnfigur%Cons. = 30 A; bbend= /Ttwist. R,,

R sin a = 2 , cosa-(l--$$)l z. (28) 5-R Equations (28) can be substituted into Eq. (22) to get the energy at the smallest allowable pitch angle when excluded

J. Chem. Phys., Vol. 95, No. 12,15 December 1991 Downloaded 21 Oct 2007 to Redistribution subject to AIP license or copyright, see http://jcp.aip.org/jcp/copyright.jsp


N. G. Hunt and J. E. Hearst: Elastic model of DNA supercoiling

volume is taken into account. The resulting expression can be minimized with respect to R to get the global energy minimum. We did this numerically and plotted the results in Fig. 10, together with the corresponding results for the plectonemic case. As Fuller predicted, the toroidal form becomes energetically favorable at sufficiently high Ial, but the values of [al at which this occurs are much greater than what is physiologically accessible. V. DISCUSSION A. Alternate topological


Equation (1) relates a property of two curves, linking number (or linking number difference ), to another property of two curves, the twist of one about the other, and a property of one of the curves, writhe. White and Bauer pointed out that, in applying Eq. ( 1) to DNA, there are several possible choices for the curves which do not give the same values for twist and writhe.54 The usual choice is to take one curve to be the axis of the DNA double helix and the other to be either of the sugar-phosphate backbones. In this case the writhe is the writhe of the helix axis and the twist is the twist of the backbone about the axis (which is not the same as the twist of the axis about the backbone) .54 In Ref. 54 it was suggested that the correct choice for analyzing ring closure experiments (and, by extension, any study of DNA energetics) would be to take the two curves to be the two sugar-phosphate backbones of the DNA. While this choice is perfectly self-consistent topologically, it does not lead to correct values of twist to apply to Eq. (2). It is not a convenient choice for bend energy and writhe, either. For example, the writhe of a straight piece of DNA would be nonzero by this definition, since the backbones are helical. The usual choice of the

DNA axis and one of the backbones gives the right twist and curvature for Eqs. (2) and (3), respectively. Another alternative to Eq. ( 1) was derived by White, Cozzarelli, and Bauer. In what was later called surface linking theory, linking number was expressed as the sum of two quantities; winding number and surface linking number. Both of these quantities are defined with respect to a surface which the DNA axis lies on. The winding number was shown to be the experimentally measured quantity in nuclease studies of helical repeat of DNA bound in nucleosomes.56 However, surface linking theory is not as useful for describing the configuration of plectonemic DNA. The surface linking number for plectonemic DNA is zero, so that the winding number equals the linking number.s6 Since this is true for any superhelical pitch angle, surface linking theory does not distinguish between configurations with differing values of twist and writhe. Twist and writhe are the relevant quantities for treating elastic energies of DNA, since twist energy is a function of twist and bend energy can be related to writhe. l3. Configuration of supercoiled DNA

The model of the plectonemic helix presented here makes several quantitative predictions about the configuration of superhelical DNA. At low specific linking difference 101,the superhelicity is almost entirely in writhe. As 1~1increases, a greater proportion of superhelicity is in twist. There is an absolute limit on the writhe per unit length, pwr,,, , which is 1/(4?rR,,); therefore the limit on the amount of specific linking difference 1~1which can be taken up by writhe is (h,,pWr,,, ) . This limit is approached as Ial approaches infinity (for biologically relevant values of 1~1, pWr doesn exceed 0.7pWr,, t ). A comparison of the plectonemic and toroidal structures shows that the plectonemic structure is energetically favored. C. Range of superhelical earlier model pitch angle differs from an


n 2 100 x w 0 0 0.04

difference, CT



FIG. 10. Total energies ofplectonemic (triangles) and toroidal (squares) configurations as functions of C, for minimum-energy configurations. R,, = 100 4;&,, = &,+,, = 2 x 10 - erg cm. For smaller values of R, , the difference between the two curves is-greater. The two curves cross at 0~0.15; PI!?,,,=: 1000 J/moI b.p.

Several previous models of DNA supercoiling have included some of the results in this paper, at least implicitly. l&z6 In particular, the paper by Camerini-Otero and Felsenfeld3 is very similar to this work in that it ignores end effects to derive properties of supercoiling in the limit of long pieces of DNA. Camerini-Otero and Felsenfeld came to different conclusions from this study, most notably in the statement that the superhelical pitch angle lies in the range 45 <a<W, whereas we conclude that the upper limit on a is 90. The differing conclusions stem from two related errors in Ref. 37. One is that a constraint on two parameters of their model, equation (7) of Ref. 37, is mistaken for an equation for the torsional constant (at that time unmeasured) and is not applied consistently as a constraint. The second error is that the free-energy expression derived in their model is equated with an empirical quadratic dependence of free energy on ALk. While such a relation does hold over a restricted range of ALk,38-41 it is only consistent with the model of Ref. 37 if the torsional constant and/or the superhelical radius are made to vary in a nonphysical manner.

J. Chem. Phys., Vol. 95, No. 12,15 December 1991

Downloaded 21 Oct 2007 to Redistribution subject to AIP license or copyright, see http://jcp.aip.org/jcp/copyright.jsp

N. G. Hunt and J. E. Hearst: Elastic model of DNA supercoiling


A simple phYsica argument against the upper limit on a given in Ref. 37 is that, in the limit where cr = 0, a configuration with a = 90 has zero bend and torsional energy, whereas a configuration at any other pitch angle (e.g., 59) has finite bend and twist energies. This energy is plotted in the u = 0 curve of Fig. 4. D. Comparison with experimental results

2. Energy of supercoiling The energy of supercoiling inferred from ethidium binding studies and the distribution of linking numbers resulting from ligation of linear pieces of DNA is (29) where 3.4 A is the length of DNA per basepair. This quantity gives larger energies than our model except at very high values of 1~1,as shown in Fig. 11. The lack of agreement is not too surprising, since Eq. (29) is based on measurements made at low 1~1, where entropic effects and end effects are large.jsdl 3. Salt concentration and conformational transitions

pE,,, = 10 RT u =/3.4 &

1. Electron microscopy The most basic prediction of this paper, that supercoiled DNA in solution is plectonemic, agrees with the branched plectonemic structures observed by electron micross~58 COPY* The most extensive electron microscopy study of DNA supercoiling is that of Boles et aLso Comparison of our model with their data on a more quantitative level is complicated by the fact that the observed superhelical radius is a function of a, whereas we hold the superhelix radius constant. Two possible extensions to the model presented here might explain the variation of superhelical radius with LT. The first would be to replace the hard-body model of the DNA excluded volume with a screened electrostatic repulsion.59 Then the two strands of the plectonemic helix would approach each other until the outward force per length balanced the inward force per length caused by superhelicity. This would result in a superhelix radius which decreased with increasing 1 The resulting changes in supera-1. helix radius would be on the order of the Debye length.60 A second modification to the model presented here would be to include entropic effects. Since there are more possible conformations at greater superhelical radius, the entropic contribution to free energy would tend to give larger values for superhelical radius than what results from the enthalpic contribution alone. As 1~1increases, the dependence of the enthalpy on superhelical radius becomes sharper, resulting again in a superhtelical radius which decreases with increasing Ial. This entropic effect on superhelical radius would be greater for small 1al. The justification for ignoring entropic effects at large values of In.1is presented in the previous paper.6 It is possible that the observed dependence of siuperhelix radius on G may be partly an experimental artifact. If the interaction of the DNA with the polylysine matrix used in Ref. 50 results in forces which disturb the solution conformation of the DNA, the resulting distortions would be greater for smaller values of 1al, since the bend and twist energies vary more slowly with superhelical radius at small 14. This work, together with data on Rex,4244 suggests a strong dependence of superhelical parameters on salt concentration. Thus, it is surprising that Boles et al. do not report a difference in results for two buffers of substantially different cationic concentrations. Adrian et al., using cryoelectron microscopy, report a strong dependence of superhelical diameter on salt concentration. Their results for superhelical diameter are somewhat smaller than values of the excluded volume obtained by osmotic pressure measurements.42-44

The model parameters fibend,fitwist, and R, are all functions of ionic strength. While&,, andfitwist may vary by as much as a factor of 2 in the range 1 mM< [ Na + ] < 1 M,b2 the excluded volume radius R,, varies by a factor of 5 over the same range.42-44The ionic strength dependence of R,, has effects which are generally greater than the effects of the ionic strength dependencies of fibendand fitwist. At low ionic strength, R,, is about 100 &42-44 which gives (h,pWr,,, ) = 0.028. Therefore, superhelicity starts going into twist at low values of Ial, resulting in large values of twist, twist energy, total energy, and superheiical stress. Conversely, at high salt concentrations, R,, is smaller, pWrmax is larger and more superhelicity goes into writhe, resulting in lower twist, twist and total energies, and lower superhelical stress. Experimental data on the salt-concentration dependence of conformational transitions agrees qualitatively with our work. B-Z transitions45,46 and cruciform formation47V48 are both induced by lower levels of superhelicity at low salt concentration than at higher salt concentration.

si d
a3 "0 E G 80

0 0










FIG. 11. Total energy of plectonemic configurations compared with the parabolic energy of supercoiling, Eq. (29 ) . Equatidn (29 ) lexx plectos); nemic helix with R,, = 100 A, fibend= 2 X 10 - I9 erg cm, fitWi,, = 3 X 10 - I9 erg cm (squares); plectonemic helix with R,, = 30 A; B bend= &,i,l = 2 X 10 - I9 erg cm (circles).

J. Chem. Phys., Vol. 95, No. 12,15 December 1997 Downloaded 21 Oct 2007 to Redistribution subject to AIP license or copyright, see http://jcp.aip.org/jcp/copyright.jsp


N. G. Hunt and J. E. Hearst: Elastic model of DNA supercoiiing M. Gellert, Ann. Rev. Biochem. 50,879 (1981). A. Baker, K. Sekimuzu, B. E. Funnell, and A. Komberg, Cell 45,53 zT (i986). i3Y. Zivanovic, I. Duband-Goulet, P. Schultz, E. Stofer, P. Oudet, and A. Prunell, J. Mol. Biol. 214,479 (1990). *M. Wadati and H. Tsuru, Physica D 21,213 (1986). Is H. Tsuru and M. Wadati, Biopolymers 25,2083 (1986). F. Tanaka and H. Takahashi, J. Chem. Phys. 83,6017 (1985). M. LeBret, J. Mol. Biol. 200,267 (1988). M.-H. Hao and W. K. Olson, Biopolymers 28,873 (1989). i9M.-H. Hao and W. K. Olson, J. Biomol. Str. Dynamics 7,661 (1989). W. K. Olson and J. Cicariello, Ann. N.Y. Acad. Sci. 482,69 (1986). C. J. Benham, Proc. Natl. Acad. Sci. USA 74, 2397 (1977). zC. J. Benham, Biopolymers 22,2477 (1983). 23C. J. Benham, Biopolymers 26,9 (1987). C. J. Benham, Biopolymers 18,609 ( 1979). M. LeBret, Biopolymers 18, 1709 (1979). 26M. LeBret, Biopolymers 23, 1835 ( 1984). * V. Klenin. A. V. Voloaodskii, V. V. Anshelevich, V. YIL Klishko, A. K. M. Dykhne, and M. D. Flank-Kamenetskii, J. Biomol. Str. Dynamics 6, 707 (1989). 28K. V. Klenin, A. V. Vologodskii, V. V. Anshelevich, A. M. Dykhne, and M. D. Frank-Kamenetskii, J. Biomol. Str. Dynamics 5, 1173 (1988). A V. Vologodskii, V. V. Anshelevich, A. V. Lakashin, and M. D. FrankKamenetskii, Nature 280,294 ( 1979). OK V. Klenin, A. V. Vologodskii, V. V. Anshelevich, A. M. Dykhne, and M. D. Frank-Kamenetskii, J. Mol. Biol. 217,413 (1991). 31M. LeBret, Biopolymers 19,619 (1980). J. Shimada and H. Yamakawa. Biouolvmers 27,657 (1988). 33S. D. Levene and D. M. Crothkrs, f Mol. Biol. i89,61 (1986). 34S. D. Levene and D. M. Crothers, J. Mol. Biol. 189, 73 -( 1986). $K. Iwata, J. Chem. Phys. 78,2778 (1983). 36Y-D. Chen, J. Chem. Phys. 75,2447 (1981). 37R. D. Came&i-Otero and G. Felsenfeld, Proc. Natl. Acad. Sci. USA 75, 1708 ( 1978 ) . T.-S. Hsieh and J. C. Wang, Biochem. 14,527 (1975). 9R. E. Depew and J. C Wang, Proc. Natl. Acad. Sci. USA 72, 4275 (1975). 40D. E. Pulleyblank, M. Shure, D. Tang, J. Viiograd, and H.-P. Vosberg, Proc. Natl. Acad. Sci. USA 72,428O ( 1975). D. S. Horowitz and J. C. Wang, J. Mol. Biol. 173,75 (1984). F! G. Yarmola, M. I. Zarudnaya, and Yu. S. Lazurkin, J. Biomol. Str. Dynamics 2,981 (1985). 43A A. Brian, H. L. Frisch, and L. S. Lerman, Biopolymers 20, 1305 (;981). D. Stigter, Cell Biophys. 11, 139 (1987). 45C. K. Singleton, J. Klysik, S. M. Stirdivant, and R. D. Wells, Nature 299, 312 (1982). 46A. Nordheim and A. Rich, Proc. Natl. Acad. Sci. USA 80, 1821 (1983). 4 R. Singleton, J. Biol. Chem. 258, 7661 (1983). C. 48S. A. Kozyavkin, A. I. Slesarev, S. R. Malkhosyan, and I. G. Panyutin, Eur. J. Biochem. 191, 105 (1990). 49J. Torbet and E. DiCapua, EMBO J. 8,435l (1989). s0T C. Boles, J. H. White, and N. R. Cozzarelli, J. Mol. Biol. 213, 931 (i990,. S. J. Spengler, A. Stasiak, and N. R. Cozzarelli, Cell 42,325 ( 1985). 52G. W. Brady, M. Satkowski, D. Foos, and C. J. Benham, J. Mol. Biol. 195, 185 (1987). 53G. W. Brady, D. B. Fein, H. Lambertson, V. Grassian, D. Foos, and C. J. Benham, Proc. Natl. Acad. Sci. USA 80,741 (1983). r4J. H. White and W. R Bauer, J. Mol. Biol. 189, 329 ( 1986). L. D. Landau and E. M. Lifshitz, Theory ofElnsticity (Pergamon, London, 1959). 56J. H. White, N. R. Cozzarelli, and W. R. Bauer, Science 241,323 ( 1988). * M H. Richardson, T. C. Boles, and N. R. Cozzarelli, Nucleic Acids S ies. ;6,6607 (1988). *M. Adrian, B. ten Heggeler-Bordier, W. Wahli, A. Z. Stasiak, A. Stasiak, and J. Dubochet, EMBO J. 9,455l (1990). 59N. G. Hunt, M. P. Klein, and J. E. Hearst (unpublished). P. W. Atkins, Physical Chemistry (Freeman, New York, 1990). 6J. E. Hearst and N. G. Hunt, J. Chem. Phys. 95,9322 (1991). P. J. Hagerman, Ann. Rev. Biophys. Biophys. Chem. 17,265 ( 1988).

E. Range of validity of the model The models presented in this paper of plectonemic and toroidal supercoiling are valid in the limit of long DNA molecules. The axis of a finite toroidal superhelix is circular, rather than straight, which results in greater bend energy. A finite plectonemic helix. has loops at the end which contribute to the bend energy while contributing little to the writhe. Since the infinite length models ignore these effects, they are not valid for pieces of DNA shorter than a few persistence lengths. This work does not entirely rule out toroidal structures for all lengths of DNA. It is possible that the finite length effects mentioned earlier result in a region of DNA length and D which would favor the toroidal structure. As noted earlier, these models do not take entropic effects into~ account. 61 In general, entropic effects are of greatest importance when the structure is loose, i.e., when pE,,, varies relatively slowly with R and a. This is the case for small values of 1~1.A more exact statement on the range of D for which entropic effects are significant would require calculating a partition function for all configurations near the enthalpy minimum ones obtained earlier, which is beyond the scope of this paper. Empirically, one can say that the model is .valid for 1~71 values high enough that interwound structures are observed experimentally. F. Future work Several aspects of the present model and its differences with experimental data suggest that a more detailed treatment of electrostatic interaction between strands is needed.59 The energy of supercoiling would be raised by including this energy, perhaps bringing the total energy into better agreement with Eq. (28). An electrostatic term in the total energy would also result in the superhelical radius being a function of 101,as observed in electron microscopy data. Finally, the sensitivity of the results of the model to the value of R, suggests that electrostatic repulsion between and within DNA strands could have significant effects. ACKNOWLEDGMENTS The authors thank David Cook for many useful discussions. This work was supported by National Institutes of Health Grant No. GM. 41911 and by the Office of Energy Research, Office of Health and Environmental Research of the Department of Energy, under Contract No. DE ACO376SFOO98.
H. White, Am. J. Math. 91, 693 (1969). J. z F. B. Fuller, Proc. Natl. Acad. Sci. USA 68, 815 ( 197 1). C J. Benham, Cell Biophys. 10, 193 ( 1987). 4 L. J. Peck and J. C. Wang, Proc. Natl. Acad. Sci. USA SO,6206 ( 1983). M. Gellert, K. Mizuuchi; M. H. O Dea, H. Ohmori, and J. Tomizawa, Cold Spring Harbor Symp. Quant. BioL 43,35 (1979). 6D. M. J. Lilley, Proc. Natl. Acad. Sci. USA 77,6468 ( 1980). N. Panayatatds and R.-D. Wells, Nature 289,466 ( 198 1). *F. S. Lee and W. R. Bauer, Nucleic Acids Res. 13, 1665 (1985). +Yu. L. Lyubchenko and L. S. Shlyakhtenko, Nucleic Acids Res. 16,3269 (1988). H. A. Nash, TIBS 15,222 ( 1990).

J. Chem. Phys., Vol. 95, No. 12,15 December 1991

Downloaded 21 Oct 2007 to Redistribution subject to AIP license or copyright, see http://jcp.aip.org/jcp/copyright.jsp