13.07.2015 Views

Elastic property of single double-stranded DNA molecules ...

Elastic property of single double-stranded DNA molecules ...

Elastic property of single double-stranded DNA molecules ...

SHOW MORE
SHOW LESS
  • No tags were found...

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

PHYSICAL REVIEW E VOLUME 62, NUMBER 1JULY 2000<strong>Elastic</strong> <strong>property</strong> <strong>of</strong> <strong>single</strong> <strong>double</strong>-<strong>stranded</strong> <strong>DNA</strong> <strong>molecules</strong>:Theoretical study and comparison with experimentsHaijun Zhou, 1,2, * Yang Zhang, 1 and Zhong-can Ou-Yang 1,31 Institute <strong>of</strong> Theoretical Physics, Academia Sinica, P.O. Box 2735, Beijing 100080, China2 State Key Laboratory <strong>of</strong> Scientific and Engineering Computing, Beijing 100080, China3 Institute for Advanced Study, Tsinghua University, Beijing 100084, ChinaReceived 7 December 1999This paper aims at a comprehensive understanding <strong>of</strong> the novel elastic <strong>property</strong> <strong>of</strong> <strong>double</strong>-<strong>stranded</strong> <strong>DNA</strong>ds<strong>DNA</strong> discovered very recently through <strong>single</strong>-molecule manipulation techniques. A general elastic modelfor <strong>double</strong>-<strong>stranded</strong> biopolymers is proposed, and a structural parameter called the folding angle is introducedto characterize their deformations. The mechanical <strong>property</strong> <strong>of</strong> long ds<strong>DNA</strong> <strong>molecules</strong> is then studiedbased on this model, where the base-stacking interactions between <strong>DNA</strong> adjacent nucleotide base pairs, thesteric effects <strong>of</strong> base pairs, and the electrostatic interactions along <strong>DNA</strong> backbones are taken into account.Quantitative results are obtained by using a path integral method, and excellent agreement between theory andthe observations reported by five major experimental groups are attained. The strong intensity <strong>of</strong> the basestacking interactions ensures the structural stability <strong>of</strong> <strong>DNA</strong>, while the short-ranged nature <strong>of</strong> such interactionsmakes externally stimulated large structural fluctuations possible. The entropic elasticity, highly extensibility,and supercoiling <strong>property</strong> <strong>of</strong> <strong>DNA</strong> are all closely related to this account. The present work also suggests thepossibility that negative torque can induce structural transitions in highly extended <strong>DNA</strong> from the right-handedB form to left-handed configurations similar to the Z-form configuration. Some formulas concerned with theapplication <strong>of</strong> path integral methods to polymeric systems are listed in the Appendixes.PACS numbers: 87.15.By, 05.50.q, 64.60.Cn, 75.10.HkI. INTRODUCTIONThe <strong>DNA</strong> molecule is the primary genetic material <strong>of</strong>most organisms. It is a <strong>double</strong>-helical biopolymer in whichtwo chains <strong>of</strong> complementary nucleotides the subunitswhose sequence constitutes the genetic message wind usuallyright-handedly around a common axis to form a <strong>double</strong>helicalstructure 1. Because <strong>of</strong> this unique structure, theelastic <strong>property</strong> <strong>of</strong> <strong>DNA</strong> molecule influences its biologicalfunctions greatly. There are mainly three kinds <strong>of</strong> deformationsin a <strong>DNA</strong> <strong>double</strong> helix: a stretching and bending <strong>of</strong> themolecule and a twisting <strong>of</strong> one nucleotide chain relative toits counterpart. All these deformations have vital biologicalsignificance. During <strong>DNA</strong> replication, hydrogen bonds betweenthe complementary <strong>DNA</strong> bases should be broken, andthe two nucleotide chains should be separated. This strandseparationprocess requires a cooperative unwinding <strong>of</strong> the<strong>double</strong> helix 2. In a <strong>DNA</strong> recombination reaction, RecAproteins recombination proteins <strong>of</strong> Type A polymerizealong <strong>DNA</strong> template and the <strong>DNA</strong> molecule is stretched to1.5 times its relaxed contour length 3,4. It is suspected thatthermal fluctuations <strong>of</strong> the <strong>DNA</strong> central axis might be veryimportant for RecA polymerization 5. Another importantexample is the process <strong>of</strong> chromosome condensation duringprophase <strong>of</strong> the cell cycle, where a long circular <strong>DNA</strong>chain wraps tightly onto histone proteins and is severely bent2. Furthermore, in living cells the <strong>DNA</strong> chain is usuallyclosed, i.e., the two ends <strong>of</strong> the molecule are linked togetherby covalent bonds and the molecule becomes endless. Withthis chain-closing process, all the quantities characterizing*Electronic address: zhouhj@itp.ac.cnthe topological state <strong>of</strong> the chain are fixed and can only bechanged externally by topoisomerases, which are capable <strong>of</strong>transiently cutting one <strong>DNA</strong> strand or both and making onestrand pass through the other at the cutting point in the case<strong>of</strong> type I topoisomerases 6 or one segment <strong>of</strong> <strong>DNA</strong> passesthrough another in the case <strong>of</strong> type II topoisomerases 7,8.It is possible that this kind <strong>of</strong> enzyme-induced topologychangingprocess is also closely related to the particular mechanic<strong>property</strong> <strong>of</strong> the <strong>DNA</strong> molecule. For example, the frequency<strong>of</strong> collisions between two distant <strong>DNA</strong> segments in acircular <strong>DNA</strong> molecule is influenced by the different knottypes and different linking numbers for a definition <strong>of</strong> thisquantity, see below and in Sec. II C. A thorough investigation<strong>of</strong> the deformation and elasticity <strong>of</strong> <strong>DNA</strong> will enable usto gain a better understanding <strong>of</strong> many important biologicalprocesses concerned with life and growth.A detailed study <strong>of</strong> <strong>DNA</strong> elasticity has now become possibledue to recent experimental developments, including,e.g., optical tweezer methods, atomic force microscopy, andfluorescence microscopy. These techniques make it possibleto manipulate <strong>single</strong> polymeric <strong>molecules</strong> directly, and torecord their elastic responses with a high precision. Experimentsdone on <strong>double</strong>-<strong>stranded</strong> <strong>DNA</strong> ds<strong>DNA</strong> have revealedthat this molecule has very interesting elastic properties9–16. When a torsionally relaxed <strong>DNA</strong> is pulled with aforce <strong>of</strong> less than 10 picoNetwon pN, its elastic responsecan be quantitatively understood by regarding the chain as aninextensible thin string with a certain bending rigiditynamely, the wormlike chain model 9,17,18. However, ifthe external force is increased up to 65 pN, the <strong>DNA</strong> chainbecomes highly extensible. At this force, the molecule transitsto an overstretched configuration termed S-<strong>DNA</strong>, whichis 1.6 times longer than the same molecule in its standard1063-651X/2000/621/104514/$15.00 PRE 62 1045 ©2000 The American Physical Society


1046 HAIJUN ZHOU, YANG ZHANG, AND ZHONG-CAN OU-YANGPRE 62B-form structure 11,12. In addition to external forces, it isalso possible to apply torsional constraints to a <strong>DNA</strong> <strong>double</strong>helixby external torques. The linking number <strong>of</strong> <strong>DNA</strong>, i.e.,the total topological times one <strong>DNA</strong> strand winds around theother, can be fixed at a value larger less than the molecule’srelaxed value. In such cases we say that the <strong>DNA</strong> molecule ispositively negatively supercoiled. It was shown experimentally13 that when the external force is less than a thresholdvalue <strong>of</strong> about 0.3 pN, the extension <strong>of</strong> <strong>DNA</strong> molecule decreaseswith increasing twist stress and the elastic response<strong>of</strong> positively supercoiled <strong>DNA</strong> is similar to that <strong>of</strong> negativelysupercoiled <strong>DNA</strong>, indicating that the <strong>DNA</strong> chain might beregarded as achiral. However, if the external force is increasedto be larger than this threshold, negatively and positivelysupercoiled <strong>DNA</strong> <strong>molecules</strong> behave quite differently.Under the condition <strong>of</strong> a fixed external force between 0.3 and3 pN, while positive twist stress keeps shrinking the <strong>DNA</strong>polymer, the extension <strong>of</strong> negatively supercoiled <strong>DNA</strong> is insensitiveto the supercoiling degree 13,14. In the higherforce region, it was suggested by some authors that positivelysupercoiled <strong>DNA</strong> may transit to a configuration calledPauling-like <strong>DNA</strong> (P-<strong>DNA</strong>) with exposed nucleotide bases15, while negative torque may lead to strand separation inthe <strong>DNA</strong> molecule the denaturation <strong>of</strong> the <strong>DNA</strong> <strong>double</strong> helix14. A very recent systematic observation performed byLéger et al. 16, on the other hand, suggested another possibility:that negative supercoiling may result in a lefthandedZ-form configuration in <strong>DNA</strong>.The above-mentioned complicated elastic <strong>property</strong> revealedby the experiments may be directly related to theversatile roles played by the <strong>DNA</strong> molecule in living organisms.Theoretically, to understand the <strong>DNA</strong> elastic <strong>property</strong>is <strong>of</strong> current interest. Models concerned with one or anotheraspect <strong>of</strong> <strong>DNA</strong> elasticity were proposed, and valuable insightswere obtained see, for example, Refs. 19,17,18,20–28, and now it is widely accepted that the competition between<strong>DNA</strong> bending and torsional deformations deservesconsiderable attention in order to understand the elastic <strong>property</strong><strong>of</strong> <strong>DNA</strong> <strong>molecules</strong>. However, it is still a great challengeto understand systematically and quantitatively all aspects <strong>of</strong>the <strong>DNA</strong> mechanical <strong>property</strong> based on the same unifiedframework. What is the intrinsic reason for the <strong>DNA</strong> molecule’sentropic elasticity and high degree <strong>of</strong> extensibility, aswell as its supercoiling <strong>property</strong>? Is it possible for negativetorque to stabilize left-handed <strong>DNA</strong> configurations? Theseare just some examples <strong>of</strong> unsolved questions.In the present work, we have tried to obtain a comprehensiveand quantitative understanding <strong>of</strong> <strong>DNA</strong> mechanicalproperties. We have postulated that the <strong>double</strong>-<strong>stranded</strong> nature<strong>of</strong> the <strong>DNA</strong> structure should be extremely important toits elastic <strong>property</strong>, and therefore we have constructed a generalelastic model in which this characteristic is properlytaken into account via the introduction <strong>of</strong> a structural parameter:the folding angle . The elastic <strong>property</strong> <strong>of</strong> longds<strong>DNA</strong> <strong>molecules</strong> was then studied based on this model,where the base-stacking interactions between <strong>DNA</strong> adjacentnucleotide base pairs, their steric effects, and the electrostaticinteractions along <strong>DNA</strong> backbones were all considered.Quantitative results were obtained by using the path integralmethod, and an excellent agreement between theory and theexperimental observations <strong>of</strong> several groups was attained. Itwas revealed that, on the one hand, the strong intensity <strong>of</strong> thebase-stacking interactions ensures the structural stability <strong>of</strong>the <strong>DNA</strong> molecule, while, on the other hand, the shortrangednature <strong>of</strong> such interactions makes externally stimulatedlarge structural fluctuations possible. The entropic elasticity,highly extensibility, and supercoiling <strong>property</strong> <strong>of</strong> theds<strong>DNA</strong> molecule are all closely related to this fact. Thepresent work also revealed the possibility that negativetorque can induce structural transitions in highly extended<strong>DNA</strong> from right-handed B-form-like configurations to lefthandedZ-form-like configurations. Some discussions on thisrespect were performed, and we suggested that a possibledirect way to check the validity <strong>of</strong> this opinion is to measurethe values <strong>of</strong> the critical torques under which such transitionsare anticipated to take place according to the present calculations.This paper is organized as follows: In Sec. II we introducethe elastic model for ds<strong>DNA</strong> biopolymers. At the action <strong>of</strong>an external force, the elastic response <strong>of</strong> ds<strong>DNA</strong> <strong>molecules</strong> isinvestigated in Sec. III, and compared with experimental observations<strong>of</strong> Smith and co-workers 9,12 and Cluzel et al.11. Section IV focuses on supercoiled ds<strong>DNA</strong> <strong>molecules</strong>,where the relationship between the extension and supercoilingdegree is obtained numerically and compared with theexperiment <strong>of</strong> Strick et al. 13. From the calculated foldingangle distribution, we infer that negative torque can causestructural transitions in ds<strong>DNA</strong> <strong>molecules</strong> from right-handed<strong>double</strong> helixes to left-handed ones. Section V is reserved forconclusions. Two appendixes are also presented: in AppendixA we review some basic ideas on the application <strong>of</strong> pathintegral method in polymer physics, and in Appendix B welist the matrix elements <strong>of</strong> the operators in Eqs. 15 and23. Some parts <strong>of</strong> this work were briefly reported in a previouspaper 29.II. ELASTIC MODEL OF DOUBLE-STRANDED <strong>DNA</strong>MOLECULEAs already stressed, the <strong>DNA</strong> molecule is a <strong>double</strong><strong>stranded</strong>biopolymer. Its two complementary sugarphosphatechains twist around each other to form a righthanded<strong>double</strong> helix. Each chain is a linear polynucleotideconsisting <strong>of</strong> the following four bases: two purines (A andG) and two pyrimidines (C and T) 1,2. The two chains arejoined together by hydrogen bonds between pairs <strong>of</strong> nucleotidesA-T and G-C. Hereafter, we refer to the two sugarphosphatechains as the backbones, and the hydrogenbondedpairs <strong>of</strong> nucleotides as the base pairs. In this sectionwe discuss the energetics <strong>of</strong> such an elastic system see Fig.1. First, the bending energy <strong>of</strong> the backbones <strong>of</strong> such<strong>double</strong>-<strong>stranded</strong> polymers will be considered; then we willdiscuss the interactions between <strong>DNA</strong> base pairs and energyterms related to external fields.A. Bending and folding deformationsBackbones can be regarded as two inextensible wormlikechains characterized by a very small bending rigidity k B Tl p , where k B is Boltzmann’s constant, T is the environmentaltemperature, and l p 1.5 nm is the bending persistencelength <strong>of</strong> <strong>single</strong>-<strong>stranded</strong> <strong>DNA</strong> ss<strong>DNA</strong> chains 12.The bending energy <strong>of</strong> each backbone is thus expressed as


PRE 62 ELASTIC PROPERTY OF SINGLE DOUBLE-STRANDED . . .1047as half the rotational angle from t 2 to t 1 , with b being therotational axis Fig. 1. We call the folding angle, and itcan vary in the range (/2, /2), with 0 correspondingto right-handed rotations and hence right-handed<strong>double</strong>-helical configurations, and 0 to left-handed ones.With the help <strong>of</strong> Eq. 1, we know thatanddbds t 2t 1 sin 2R R n 2FIG. 1. Schematic representation <strong>of</strong> the <strong>double</strong>-<strong>stranded</strong> <strong>DNA</strong>model used in this paper. The right part demonstrates the definition<strong>of</strong> on the local t-n plane, where t, t 1 and t 2 are the tangentialvectors <strong>of</strong> the central axis, and the two backbones, respectively; is the folding angle; and the unit vector btn is perpendicular tothe t-n plane.(/2) 0 L (dt i /ds) 2 ds, where t i (s) (i1 and 2 is the unittangent vector at arclength s along the ith backbone 30,31,and L is the total contour length <strong>of</strong> each backbone. The positionvectors <strong>of</strong> the two backbones are expressed as r i (s) s t i (s)ds.Since there are many relatively rigid base pairs betweenthe two backbones in many cases the lateral distance betweenthe backbones can be regarded to be constant andequal to 2R. 1 In this subsection we focus on the bendingenergy <strong>of</strong> the backbones; therefore, for the moment, we regardeach base pair as a rigid rod <strong>of</strong> length 2R linking betweenthe two backbones and pointing along a direction denotedby a unit vector b from r 1 to r 2 Fig. 1. Then r 2 (s)r 1 (s)2Rb(s). In B-form <strong>DNA</strong> the base pair plane isperpendicular to the <strong>DNA</strong> axis; therefore, in our model therelative sliding <strong>of</strong> the two backbones is not considered andthe base pair rod is thought to be perpendicular to both backbones31, with b(s)•t 1 (s)b(s)•t 2 (s)0. The centralaxis <strong>of</strong> the <strong>double</strong>-<strong>stranded</strong> polymer can be defined as r(s)r 1 (s)Rb(s) r 2 (s)Rb(s)(r 1 (s)r 2 (s))/2, andits tangent vector is denoted by t. In consistence with actual<strong>DNA</strong> structures, the central axial tangent t is also perpendicularto b, i.e., b(s)•t(s)0. Notice, however, that t(s)dr/ds; in this paper, s always refers to the arclength <strong>of</strong> thebackbones.Since all the tangent vectors t 1 , t 2 , and t lie on the sameplane perpendicular to b, we can write thatt 1 stscos snssin s,t 2 stscos snssin s,where n is also a unit vector, and nbt, and is defined1 In our present work, we have not taken into account the possibledeformations <strong>of</strong> the nucleotide base pairs. In many cases this maybe a reasonable assumption. However, under some extreme conditionssuch kinds <strong>of</strong> deformations may turn out to be important. Forexample, when a <strong>DNA</strong> <strong>double</strong> helix is stretched and at the sametime has a large positive torque applied to it, the nucleotide basepairs may collapse see Ref. 15 for a description.1drds 1 2 t 1t 2 t cos .Equation 3 indicates that cos measures the extent towhich the backbones are ‘‘folded’’ with respect to the centralaxis. Based on Eqs. 1–3, the total bending energy <strong>of</strong> thetwo backbones can be expressed in the following form:dt 1dsdt 22dsdsE b 2 0LL ds0 dtL ds0 dsdt2ds 20L2cosds2 sin 2dnds2dsd 2 R 2 sin 4 .2dsd 2The bending energy is thus decomposed into the bendingenergy <strong>of</strong> the central axis the first term <strong>of</strong> Eq. 4 plus thefolding energy <strong>of</strong> the backbones the second and third terms<strong>of</strong> Eq. 4. The physical meanings <strong>of</strong> these two energy contributionsare very clear, and Eq. 4 is very helpful for ourfollowing calculations. In Eq. 4, the bending energy <strong>of</strong> thecentral axis is very similar with that <strong>of</strong> a wormlike chain18, both <strong>of</strong> which are related to the square <strong>of</strong> the changingrate <strong>of</strong> the axial tangent vectors. But there are two importantdifferences: a in the derivative dt/ds <strong>of</strong> Eq. 4, thearclength parameter s is measured along the backbone, notalong the central axis; and b in the wormlike chain modelthe central axis is inextensible, while here the central axis isextensible.In deriving Eq. 4, the base pairs are models just as thinrigid rods <strong>of</strong> fixed length, i.e., the <strong>DNA</strong> molecule is viewedas a ladderlike structure see Fig. 1. Actually, however, basepairs form disklike structures and have a finite volume. Thesteric effects caused by the finite volume <strong>of</strong> base pairs areanticipated to hinder considerably the bending deformation<strong>of</strong> the central axis, and hence will increase its bending rigiditygreatly 1. Furthermore, the ds<strong>DNA</strong> molecule is a strongpolyelectrolyte, with negatively charged groups distributedregularly along the chain’s surface. The electrostatic repulsionforce between these negatively charged groups will alsoconsiderably increase the bending rigidity <strong>of</strong> the ds<strong>DNA</strong>chain 1. To quantitatively take into account the above mentionedtwo kinds <strong>of</strong> effects is very difficult. Here we treat thisproblem phenomenologically by simply replacing the bendingrigidity in the first term <strong>of</strong> Eq. 4 with a quantity *.It is required that *, and the precise value <strong>of</strong> * will34


1048 HAIJUN ZHOU, YANG ZHANG, AND ZHONG-CAN OU-YANGPRE 62then be determined self-consistently by the best fitting withexperimental data, as shown in Sec. III.B. Base-stacking interactions between base pairsIn Sec. II A, we have discussed in detail the bending energy<strong>of</strong> ds<strong>DNA</strong> polymers, which is caused by a bending <strong>of</strong>the backbones, as well as steric effects and electrostatic interactions.In the ds<strong>DNA</strong> molecule there is another kind <strong>of</strong>important interaction, namely, the base-stacking interactionbetween adjacent nucleotide base pairs 1,2. Base-stackinginteractions originate from the weak van der Waals attractionbetween the polar groups in adjacent nucleotide base pairs.Such interactions are short ranged, and their total effect isusually described by a potential energy <strong>of</strong> the Lennard-Jonesform 6-12 potential 1. Base-stacking interactions play asignificant role in the stabilization <strong>of</strong> the <strong>DNA</strong> <strong>double</strong> helix.The main reason why <strong>DNA</strong> can but RNA cannot form a long<strong>double</strong> helix is as follows 32: Because <strong>of</strong> the steric interferencecaused by the hydroxyl group attached to the 2carbon <strong>of</strong> RNA riboses, the stacking interaction between adjacentRNA nucleotide base pairs is very weak and cannotstabilize the formed <strong>double</strong>-helical structure; while in the<strong>DNA</strong> ribose, a hydrogen atom is attached to its 2 carbon,and serious steric interference is avoided fortunately.In a continuum theory <strong>of</strong> elasticity, the summed totalbase-stacking potential energy is converted into the form <strong>of</strong>the integrationN1E LJ i1U i,i1 0Lds,where U i,i1 is the base-stacking potential between the ithand (i1)th base pairs, N is the total number <strong>of</strong> base pairs,and the base-stacking energy density is expressed asr 0 cos 120cos 2 cos 60cos for 06cos 12 r 0 2 cos 6 0 for 0.0In Eq. 6, the parameter r 0 is the backbone arclength betweenadjacent bases (r 0 L/N); 0 is a parameter related tothe equilibrium distance between a <strong>DNA</strong> dimer (r 0 cos 03.4 Å ); and is the base stacking intensity which is generallybase-sequence specific 1. In this paper we focus onmacroscopic properties <strong>of</strong> long <strong>DNA</strong> chains composed <strong>of</strong>relatively random sequences; therefore, we just consider inthe average sense and take it as a constant, with 14.0 k B T averaged over quantum mechanically calculatedresults on all the different <strong>DNA</strong> dimers 1.The asymmetric base-stacking potential Eq. 6 ensuresthat a relaxed <strong>DNA</strong> molecule takes on a right-handed<strong>double</strong>-helix configuration i.e., the B form, with its foldingangle 0 . To change the local configuration <strong>of</strong> <strong>DNA</strong> considerablyfrom its B form generally requires a free energy <strong>of</strong>the order <strong>of</strong> per base pair. Thus, the <strong>DNA</strong> molecule will bevery stable under normal physiological conditions, and thermalenergy can only make it fluctuate very slightly around itsequilibrium configuration, since kT. Nevertheless, althoughthe stacking intensity in ds<strong>DNA</strong> is very strong5compared with the thermal energy, the base-stacking interactionby its nature is short ranged, and hence sensitive to thedistance between the adjacent base pairs. If a ds<strong>DNA</strong> chainis stretched by large external forces, which cause the averageinter-base-pair distance to exceed some threshold value determinedintrinsically by the molecule, the restoring forceprovided by the base-stacking interactions will no longer beable to <strong>of</strong>fset the external forces. Consequently, it will bepossible that the B-form configuration <strong>of</strong> ds<strong>DNA</strong> will collapse,and the chain will become highly extensible. Thus, onthe one hand, the strong base-stacking interaction ensuresthat the standard B-form configuration is very stable uponthermal fluctuations and small external forces this is requiredfor the biological functions <strong>of</strong> <strong>DNA</strong> molecule to beproperly fulfilled 2; however, on the other hand, its shortrangedness gives it considerable latitude to change its configurationto adapt to possible severe environments otherwise,the chain may be pulled apart by external forces, forexample, during <strong>DNA</strong> segregation 2. This <strong>property</strong> <strong>of</strong><strong>DNA</strong> base-stacking interactions is very important to theds<strong>DNA</strong> molecule. As we will see in Secs. III and IV, themechanical properties <strong>of</strong> a <strong>DNA</strong> chain are indeed closelyrelated to the above-mentioned insight.C. External forces and torquesIn the previous two subsections, we have described theintrinsic energy <strong>of</strong> the <strong>DNA</strong> <strong>double</strong> helix. Experimentally, toprobe the elastic response <strong>of</strong> linear <strong>DNA</strong> molecule, the polymerchain is <strong>of</strong>ten pulled by external force fields and/or untwistedor overtwisted by external torques. To study the mechanicresponse <strong>of</strong> the ds<strong>DNA</strong> molecule, in this subsectionwe consider the energy terms related to external forces andtorques in our theoretical framework.For the external force fields, here we constrained ourselvesto the simplest situation where one terminal <strong>of</strong> a <strong>DNA</strong>molecule is fixed and the other terminal is pulled with a forceF f z 0 along the direction <strong>of</strong> the unit vector z 0 9. In fact,hydrodynamic fields or electric fields are also frequentlyused to stretch semiflexible polymers 33, but we will notdiscuss such cases in this paper. The end-to-end vector <strong>of</strong> a<strong>DNA</strong> chain is expressed as L 0 t(s)cos (s)ds according to Eq.3. Then the total ‘‘potential’’ energy <strong>of</strong> the chain in theexternal force field isE f 0Lt cos ds•F 0Lf t•z0 cos ds.In the experimental setup, external torques can be appliedon a linear ds<strong>DNA</strong> molecule by the following procedure:first, <strong>DNA</strong> ligases are used to ligate all the possible <strong>single</strong><strong>stranded</strong>nicks; then the two strands <strong>of</strong> a ds<strong>DNA</strong> molecule atone end are fixed onto a template, while the two strands atthe other end are attached tightly to a magnetic bead; afterwards,torques are introduced into a <strong>DNA</strong> <strong>double</strong> helix byrotating the magnet bead with an external magnetic field11,15. The torque energy is then related to the topologicalturns caused by the external torque on a <strong>DNA</strong> <strong>double</strong> helix.The total number <strong>of</strong> times one <strong>DNA</strong> strand winds topologicallyaround the other, which is usually termed the total linkingnumber Lk 34–36, is expressed as the sum <strong>of</strong> the twist-7


PRE 62 ELASTIC PROPERTY OF SINGLE DOUBLE-STRANDED . . .1049ing number Tw„r 1 (r 2 ),r… <strong>of</strong> backbone r 1 or r 2 ) around thecentral axis r and the writhing number Wr(r) <strong>of</strong> the centralaxis, i.e., LkTwWr. According to Refs. 35–37 and Eq.2, we obtain thatTwr 1 ,r2 1 L dbtb•0 ds 2 ds1 Lsin 0 Rds. 8The writhing number <strong>of</strong> the central axis is generally muchmore difficult to calculate. It is expressed as the followingGauss integral over the central axis 34:Wrr4 1 drdr•rr. 9rr 3In the case <strong>of</strong> linear chains, provided that some fixed directionfor example, the direction <strong>of</strong> the external force, z 0 ) canbe specified, and that the tangent vector t never points toz 0 i.e., t•z 0 1), it was proved by Fuller that the writhingnumber Eq. 9 can be calculated alternatively accordingto the following formula 38:Wrr2 1 L z 0 t•dt/dsds. 100 1z 0 •tThe above equation can be further simplified for highly extendedlinear <strong>DNA</strong> chains whose tangent t fluctuates onlyslightly around z 0 . In this case, Eq. 10 leads to the approximateexpression thatWrr4 10Ldt xt yds t xdt ydsds,11where t x and t y are the two components <strong>of</strong> t with respect totwo arbitrarily chosen orthonormal directions (x 0 and y 0 )onthe plane perpendicular to z 0 , respectively.The energy caused by the external torque <strong>of</strong> magnitude is then equal toE t 2 Lk2TwWr.12To conclude this section, the total energy <strong>of</strong> a ds<strong>DNA</strong>molecule under the action <strong>of</strong> an external force and an externaltorque is expressed as*ds2 dtdsd 2 R 2 sin 4 EE b E LJ E f E t 0L f t•z 0 cos R sin z 0t•dt/dsds1z 0 •t0L *dsdt2dsd 2 R 2 sin 4 f t•z 0 cos R sin 2 t dt xyds 2 t dt yxdsds.1314Note that Eq. 13 can be applied only in the case <strong>of</strong> highlyextended <strong>DNA</strong>. In the following two sections, we will studythe mechanical <strong>property</strong> <strong>of</strong> <strong>single</strong> ds<strong>DNA</strong> <strong>molecules</strong> basedon the model energy Eqs. 13 and 14. The theoreticalresults will be compared with experimental observations anddiscussed.III. EXTENSIBILITY AND ENTROPIC ELASTICITYOF <strong>DNA</strong>In this section we investigate the elastic responses <strong>of</strong><strong>single</strong> <strong>DNA</strong> <strong>molecules</strong> under the actions <strong>of</strong> external forcesbased on the model introduced in Sec. II. There is no externaltorque acted upon; thus 0 in Eq. 13. The particularform <strong>of</strong> the energy function Eq. 13 <strong>of</strong> the present modelmakes it convenient for us to study its statistical <strong>property</strong> bythe path integral method. In Appendix A a detailed description<strong>of</strong> the application <strong>of</strong> the path integral method to polymerphysics is given 39. Our calculations in this section and thenext section are based on this method.For a polymer whose energy is expressed in the form <strong>of</strong>Eq. 13 with 0, according to the technique outlined inAppendix A see Eqs. A1 and A7, the Green equationgoverning the evolution <strong>of</strong> the ‘‘wave function’’ (t,;s)<strong>of</strong> the system is obtained to be <strong>of</strong> the formt,;ss 24l*t 22 p4l p 2 f cos k B Tt•z 0 k B T l pR 2 sin 4 t,;s,15where l p**/k B T and l/k B T. The spectrum <strong>of</strong> theabove Green equation is discrete and, for a long ds<strong>DNA</strong>molecule according to Eqs. A10 and A13, its averageextension can be obtained either by a differentiation <strong>of</strong> theground-state eigenvalue g 0 <strong>of</strong> Eq. 15 with respect to f,Z0Lt•z0 cos dsLk B T g 0 f ,16or by a direct integration with the normalized ground-stateeigenfunction, 0 (t,), <strong>of</strong> Eq. 15:ZL 0 2 t•z 0 cos dtd.17Both g 0 and 0 (t,) can be obtained numerically throughstandard diagonalization methods and identical results areobtained by Eqs. 16 and 17. Here we just briefly outlinethe main procedures in converting Eq. 15 into the form <strong>of</strong> amatrix.First, for our convenience, we perform the transformation˜ 2 ;18hence the new argument ˜ can change in the range from 0 to. Then we choose a combination <strong>of</strong> Y lm (t) and f n (˜ ) as thebase function <strong>of</strong> the Green equation Eq. 15:t,˜ ;slmnC lmn sY lm tf n ˜ .19


1050 HAIJUN ZHOU, YANG ZHANG, AND ZHONG-CAN OU-YANGPRE 62FIG. 2. Force-extension relation <strong>of</strong> torsionally relaxed <strong>DNA</strong>molecule. Experimental data are from Fig. 2A <strong>of</strong> Ref. 11 symbols.The theoretical curve is obtained by the following considerations:a l p 1.5 nm and 14.0k B T; b l p*53.0/2cos f0 nm, r 0 0.34/cos f0 nm and R(0.3410.5/2)tan f 0 nm; and c the value <strong>of</strong> 0 is adjusted to fit the data.For each 0 , the value <strong>of</strong> cos f0 is obtained self-consistently.The present curve is drawn with 0 62.0° in close consistencywith the structural <strong>property</strong> <strong>of</strong> <strong>DNA</strong>, and cos f0 is determinedto be 0.573840. The <strong>DNA</strong> extension is scaled with its B-form contourlength Lcos f0 .In the above expression, Y lm (t)Y lm (,) (l0,1,2, ...;m0,1,...,l) are the spherical harmonics40, where and are the two directional angles <strong>of</strong> t, i.e.,t(sin cos ,sin sin ,cos ); andf n ˜ 2 a sin n a ˜ n1,2,...20are the eigenfunctions <strong>of</strong> one-dimensional infinitely deepsquare potential well <strong>of</strong> width a. 2With the wave function being expanded using theabove mentioned base functions, the operator acting on i.e., the expression in the square brackets <strong>of</strong> Eq. 15 canalso be written in matrix form under these base functions.This matrix, whose elements are listed in Appendix B, isthen diagolized numerically to obtain its ground-state eigenvalueand eigenfunction. To simplify the calculation, we furthernote that in the present case <strong>of</strong> Eq. 15, the ground stateis independent <strong>of</strong> , i.e., m can be set to m0 in Eq. 19.The resulting force vs extension relation obtained fromEqs. 16 or 17 is shown in Fig. 2 in the whole relevantforce range, and compared with the experimental observation<strong>of</strong> Cluzel et al. 11. The theoretical curve in this figure isobtained with just one adjustable parameter see the caption<strong>of</strong> Fig. 2; the agreement with experiment is excellent. Figure2 In the actual calculations, the right boundary <strong>of</strong> the square well ischosen to be slightly less than to avoid flush <strong>of</strong>f <strong>of</strong> computermemory caused by the divergence <strong>of</strong> the base-stacking potentialEq. 6 at ˜ . Weseta0.95 in this paper. However, wehave checked that the results are almost identical for other values <strong>of</strong>a, provided that a165°.FIG. 3. Folding angle distribution for torsionally relaxed <strong>DNA</strong><strong>molecules</strong> under external forces.2 demonstrates that the high extensibility <strong>of</strong> the <strong>DNA</strong> moleculeunder large external forces can be quantitatively explainedby the present model.To further understand the force-induced extensibility <strong>of</strong><strong>DNA</strong>, in Fig. 3 the folding angle distribution <strong>of</strong> a ds<strong>DNA</strong>molecule is shown, with the external force kept at differentvalues. Here, according to Eq. A13 the folding angle distributionP() is calculated by the formulaP 0 t, 2 dt.21Figures 2 and 3, taken together, demonstrate that the elasticbehaviors <strong>of</strong> the ds<strong>DNA</strong> molecule are radically differentunder the condition <strong>of</strong> low and large applied forces. In thefollowing, we will discuss these separately.The low-force region. When the external force is low(10 pN), the folding angle is distributed narrowly aroundan angle <strong>of</strong> 57°, and there is no probability for thefolding angle to take on values less than 0° Fig. 3, indicatingthat the <strong>DNA</strong> chain is completely in the right-handedB-form configuration with small axial fluctuations. Thisshould be attributed to the strong base-stacking intensity, aspointed out in Sec. II B. Consequently, the elasticity <strong>of</strong> the<strong>DNA</strong> is solely caused by thermal fluctuations in the axialtangent t Fig. 2, and the <strong>DNA</strong> molecule can be regarded asan inextensible chain. This is the physical reason why, in thisforce region, the elastic behavior <strong>of</strong> the <strong>DNA</strong> can be welldescribed by the wormlike chain model 17,18,41. Indeed,as shown in Fig. 4, at forces 10 pN, the wormlike chainmodel and the present model give identical results. Thus wecan conclude with confidence that, when external fields arenot strong, the wormlike chain model is a good approximation<strong>of</strong> the present model to describe the elastic <strong>property</strong> <strong>of</strong>ds<strong>DNA</strong> <strong>molecules</strong>; the bending persistence length <strong>of</strong> themolecule is 2l p*cos , as indicated by Eqs. 3 and 4.The large-force region. With a continuous increase <strong>of</strong> externalpulling forces, the axial fluctuations become more andmore significant. For example, at forces 50 pN, althoughthe folding angle distribution is still peaked at 57°, thereis also a considerable probability for the folding angle to bedistributed in the region 0° Fig. 3. Therefore, at thisforce region, the <strong>DNA</strong> polymer can no longer be regarded asinextensible. At f 65 pN, another peak in the folding angledistribution begins to emerge at 0°, marking the onset <strong>of</strong>


PRE 62 ELASTIC PROPERTY OF SINGLE DOUBLE-STRANDED . . .1051In Sec. III, we have discussed the elastic response <strong>of</strong> long<strong>DNA</strong> chains under the action <strong>of</strong> external forces. In thepresent section, we study the elasticity <strong>of</strong> a supercoiled <strong>DNA</strong><strong>double</strong> helix. For this purpose, in the experimental setup, allthe possible nicks in the <strong>DNA</strong> nucleotide strands are ligated13, and a torque as well as an external pulling force act onone terminal <strong>of</strong> the <strong>DNA</strong> <strong>double</strong> helix, which typically untwistsor overtwists the original B-form <strong>double</strong> helix to someextent and makes its total linking number refer to Sec. II Cfor the definition <strong>of</strong> the linking number less or greater thanthe equilibrium value. We say that such <strong>DNA</strong> <strong>molecules</strong>with deficit excess linking numbers are negatively positivelysupercoiled, and define the degree <strong>of</strong> supercoiling as LkLk 0Lk 0, 22FIG. 4. Low-force elastic behavior <strong>of</strong> <strong>DNA</strong>. Here the experimentaldata are from Fig. 5B <strong>of</strong> Ref. 9, the dotted curve is obtainedfor a wormlike chain with bending persistence length 53.0nm, and the parameters for the solid curve are the same as those inFig. 2.a cooperative transition from B-form <strong>DNA</strong> to overstretchedS-form <strong>DNA</strong> 11,12. This is closely related to the shortrangednature <strong>of</strong> the base-stacking interactions 1 see Sec.II B. At even higher forces ( f 80 pN), 3 the <strong>DNA</strong> moleculechanges completely into an overstretched form, with its foldingangle peaked at 0°.The force-induced axial fluctuations in the <strong>DNA</strong> <strong>double</strong>helix can be biologically significant. For example, it has beendemonstrated that axial fluctuations in ds<strong>DNA</strong> considerablyenhance the polymerization <strong>of</strong> RecA proteins along the <strong>DNA</strong>chain 5,42,43. A quantitative study <strong>of</strong> the coupling betweenRecA polymerization and <strong>DNA</strong> axial fluctuation is anticipatedto be helpful.It seems that in experiments 11,12 the transition toS-<strong>DNA</strong> occurs even more cooperatively and abruptly thanpredicted by the present theory see Fig. 2. This may berelated to the existence <strong>of</strong> <strong>single</strong>-<strong>stranded</strong> breaks nicks inthe ds<strong>DNA</strong> <strong>molecules</strong> used in the experiments. Nicks in<strong>DNA</strong> backbones can lead to strand separation or to a relativesliding <strong>of</strong> backbones 11,12, and this can make the transitionprocess more cooperative. However, the comprehensiveagreement achieved in Figs. 2 and 4 indicates that such effectsare only <strong>of</strong> limited significance. The elasticity <strong>of</strong> the<strong>DNA</strong> is mainly determined by the competition between thefolding angle fluctuation and the tangential fluctuation,which are governed, respectively, by the base-stacking interactions() and the axial bending rigidity (*) in Eq. 13.IV. ELASTIC PROPERTY OF SUPERCOILED <strong>DNA</strong>where Lk 0 represents the linking number <strong>of</strong> a relaxed <strong>DNA</strong>molecule <strong>of</strong> the same contour length. In living organisms,<strong>DNA</strong> <strong>molecules</strong> are <strong>of</strong>ten negatively supercoiled, with alinking number deficit <strong>of</strong> about 0.06. Thus, a detailedinvestigation on the mechanical <strong>property</strong> <strong>of</strong> supercoiled<strong>DNA</strong> <strong>molecules</strong> is not only <strong>of</strong> academic interest, but can alsohelp us to understand the possible biological advantages <strong>of</strong>negative supercoiling.A. Relationship between extension and supercoiling degreeWe focus on the <strong>property</strong> <strong>of</strong> highly extended <strong>DNA</strong> <strong>molecules</strong>whose tangent vectors fluctuate only slightly aroundthe force direction z 0 . According to what we mentioned inSec. II C, in this case an approximate energy expression Eq.14 can be used. The external stretching force is restrictedto be greater than 0.3 pN to make sure that the end-to-enddistance <strong>of</strong> the <strong>DNA</strong> chain approaches its contour length seeFig. 5. Based on Eqs. 14 and A7, the Green equation forhighly stretched and supercoiled ds<strong>DNA</strong> is then obtained tobet,;ss 24l*t 22 p4l p 2 f cos k B Tt•z 0 k B T l pR 2 sin 4 sin Rk B T 4k B Tl*p 216l*k p B T 2sin2 t,;s,23where (,) are the two directional angles <strong>of</strong> t, as mentionedin Sec. III. Similar to what we did in Sec. III, we cannow express the above Green equation in matrix form usingthe combinations <strong>of</strong> spherical harmonics Y lm (,) andf n () as the base functions. The ground-state eigenvalue andeigenfunction <strong>of</strong> Eq. 23 can then be obtained numericallyfor given applied force and torque and the average extensionbe calculated through Eq. 16 or through the formulaZL 0 t,t•z 0 cos 0 t,dtd, 24where 0 (t,) is the ground-state left-eigenfunction <strong>of</strong> Eq.23. 4 The writhing number Eq. 11 is calculated accordingto Eq. A16 to be3 This threshold f t <strong>of</strong> the overstretch force is also consistent with aplain evaluation from a base-stacking potential <strong>of</strong> f t r 0 , i.e., f t90 pN.4 As remarked in Appendix A, because the operator in the squarebrackets <strong>of</strong> Eq. 23 acting on (t,;s) is not Hermitian, the resultingmatrix form <strong>of</strong> the operator may not be diagonalized byunitary matrices. Consequently, in general, 0 (t,) 0*(t,).


1052 HAIJUN ZHOU, YANG ZHANG, AND ZHONG-CAN OU-YANGPRE 62FIG. 5. Extension vs supercoiling relations at fixed pullingforces for torsionally constrained <strong>DNA</strong>. The parameters for thecurves are the same as in Fig. 2, and experimental data are fromFig. 3 <strong>of</strong> Ref. 13 symbols.WrL16l*k p B T 0 t,sin 2 0 t,dtd,and the average linking number is then calculated to beLkTwWr2R L 0 sin 0 dtdL16l*k p B T 0 sin 2 0 dtd.2526Thus, after we have obtained the ground-state eigenvalueas well as its left and right eigenfunctions numerically, wecan calculate numerically all the quantities <strong>of</strong> our interest,for example, the average extension, the average supercoilingdegree, and the folding angle distribution see also AppendixA. The relation between the extension and supercoiling degreecan also be obtained by fixing the external force andchanging the value <strong>of</strong> the applied torque. To calculate theground-state eigenvalue and eigenfunctions <strong>of</strong> an asymmetricmatrix turns out to be complicated and time consuming.Fortunately, as we have calculated in Appendix B, eacheigenfunction <strong>of</strong> Eq. 23 shares the same quantity m; thematrix for m0 is still Hermitian, and can be diagonalizedby unitary matrices. The ground-state eigenvalue for m0 islower by several orders than those for m0 in the wholerelevant region <strong>of</strong> external torque from 5.0k B T to5.0k B T. Thus, actually, we only need to consider the case <strong>of</strong>m0, and in this case we still have 0 (t,) 0*(t,). Thewhole procedure we performed in Sec. III can safely be repeatedin this section, and the relationships between forceand extension and torque and linking number can consequentlybe calculated.To make the calculation further easier, and also to makesure that the above-mentioned calculation is indeed correct,here we introduce an approximate method which reduces thecomputational complexity considerably. It turns out that thecalculated results using this method are in considerableagreement with the above-mentioned precise method. In theexperiment <strong>of</strong> Ref. 13, the applied external forces changein the region <strong>of</strong> 0.3–10 pN. In this region, as demonstrated inFigs. 3 and 4, both the tangential (t) and folding angle ()fluctuations <strong>of</strong> ds<strong>DNA</strong> <strong>molecules</strong> are small. Taking this factinto account, in Eq. 14 the energy term *(dt/ds) 2 can beapproximately calculated to be *„(dt x /ds) 2 (dt y /ds) 2 …,and f t•z 0 cos f cos fcos (t x 2 t y 2 )/2. Thus Eq. 14 isdecomposed into two ‘‘independent’’ parts. The first part isrelated only to . At each value <strong>of</strong> f and , we can calculatethe average quantities cos and sin based on this energyusing the path integral method. The second part is quadraticin t x and t y , therefore the average values <strong>of</strong> t•z 0 andWr Eq. 11 can be obtained analytically. Using this decompositionand preaveraging technique, the average extensionand average supercoiling degree can both be calculatedat each value <strong>of</strong> the external force and torque, and the relationbetween the extension and linking number at fixedforces can be then obtained.The theoretical relationship between extension and supercoilingdegree is shown in Fig. 5 and compared with theexperiment <strong>of</strong> Strick et al. 13. In obtaining these curves,the values <strong>of</strong> the parameters are the same as those used inFig. 2, and no adjustment has been made to fit the experimentaldata. We find that in the case <strong>of</strong> negatively supercoiled<strong>DNA</strong>, the theoretical and experimental results are inquantitative agreement, indicating that the present model iscapable <strong>of</strong> explaining the elasticity <strong>of</strong> negatively supercoiled<strong>DNA</strong>; in the case <strong>of</strong> positively supercoiled <strong>DNA</strong>, the agreementbetween theory and experiment is not so good, especiallywhen the external force is relatively large. In thepresent work, we have not considered possible deformations<strong>of</strong> the nucleotide base pairs. While this assumption might bereasonable in the negatively supercoiled case, it may fail fora positively supercoiled <strong>DNA</strong> chain, especially at largestretching forces. The work done by Allemand et al. 15suggested that a positive supercoiled and highly extended<strong>DNA</strong> molecule can take on Pauling-like configurations withexposed bases. To better understand the elastic <strong>property</strong> <strong>of</strong>positively supercoiled <strong>DNA</strong>, it is certainly necessary for usto take into account the deformations <strong>of</strong> base pairs.For a negatively supercoiled <strong>DNA</strong> molecule, both theoryand experiment reveal the following elastic aspects: aWhen the external force is small, the <strong>DNA</strong> molecule canshake <strong>of</strong>f its torsional stress by writhing its central axis,which can lead to an increase in the negative writhing number,and hence restore the local folding manner <strong>of</strong> <strong>DNA</strong>strands to that <strong>of</strong> B-form <strong>DNA</strong>. b However, a writhing <strong>of</strong>the central axis causes a shortening <strong>of</strong> <strong>DNA</strong> end-to-end extension,which becomes more and more unfavorable as theexternal force is increased. Therefore, at large forces, thetorsional stress caused by negative torque supercoiling degreebegins to unwind the B-form <strong>double</strong> helix, and triggersthe transition <strong>of</strong> <strong>DNA</strong> internal structure, where a continuouslyincreasing portion <strong>of</strong> <strong>DNA</strong> takes on some certain configurationas supercoiling increases, while its total extensionremains almost invariant. Our Monte Carlo simulations havealso confirmed the above insight 44.What is the new configuration? According to Ref. 13,such a configuration corresponds to denatured <strong>DNA</strong> segments,i.e., negative torque leads to a breakage <strong>of</strong> hydrogenbonds between the complementary <strong>DNA</strong> bases, and consequentlyto strand separation. The authors <strong>of</strong> Ref. 13 alsoperformed an elegant experiment in which short <strong>single</strong>-


PRE 62 ELASTIC PROPERTY OF SINGLE DOUBLE-STRANDED . . .1053FIG. 6. Folding angle distributions for negatively supercoiled<strong>DNA</strong> molecule pulled with a force <strong>of</strong> 1.3 pN.<strong>stranded</strong> homologous <strong>DNA</strong> segments were inserted into theexperimental buffer 14. They found, in confirmation <strong>of</strong>their insight, that these homologous <strong>DNA</strong> probes indeed bindonto negatively supercoiled ds<strong>DNA</strong> <strong>molecules</strong>. Recently,Léger et al. 16 also performed experiments on <strong>single</strong>ds<strong>DNA</strong> <strong>molecules</strong>. To explain their experimental resultqualitatively, they found that a left-handed Z-form <strong>DNA</strong>molecule should be considered a possible configuration for anegatively supercoiled ds<strong>DNA</strong> chain, while the <strong>molecules</strong>need not be denatured. As seen in Fig. 5, although thepresent model has not taken into account the possibility <strong>of</strong>strand separation, it can quantitatively explain the behavior<strong>of</strong> negatively supercoiled <strong>DNA</strong>. Therefore, it may be helpfulfor us to investigate the possibility <strong>of</strong> formation <strong>of</strong> lefthandedconfigurations based on our present model. This effortis made in Sec. IV B, where the energetics <strong>of</strong> such configurationswill also be discussed.B. Possible left-handed <strong>DNA</strong> configurationsWe mentioned in Sec. II A that left-handed configurationscorrespond to 0 in our present model see also Fig. 1.Based on the present model, then information about the newconfiguration mentioned in Sec. IV A can be revealed by thefolding angle distributions P(), as discussed in Sec. III. Inthe present case, P() is calculated asP 0 t, 0 t,dt, 27where, as mentioned in Sec. IV A, 0 (t,) and 0 (t,) are,respectively, the ground-state right and left eigenfunctions <strong>of</strong>Eq. 23, and, actually, 0 (t,) 0*(t,).The calculated folding angle distribution 45 is shown inFig. 6, which is radically different from that <strong>of</strong> the torsionallyrelaxed ds<strong>DNA</strong> <strong>molecules</strong> shown in Fig. 3. This distributionhas the following aspects: When the torsional stress issmall with a supercoiling degree 0.025), the distributionhas only one steep peak at 57.0°, indicating thatthe <strong>DNA</strong> is completely in B form. With an increase <strong>of</strong> torsionalstress, however, another peak appears at 48.6°and the total probability for the folding angle to be negativelyvalued increases gradually with supercoiling. Sincenegative folding angles correspond to left-handed configurations,the present model suggests that, with an increase <strong>of</strong>FIG. 7. a The sum <strong>of</strong> the average base-stacking and torsionalenergy per base pair at a force <strong>of</strong> 1.3 pN. For highly extended <strong>DNA</strong>only these two interactions are sensitive to torque. b The relationbetween the <strong>DNA</strong> supercoiling degree and the external torque at aforce <strong>of</strong> 1.3 pN.supercoiling, the left-handed <strong>DNA</strong> conformation is nucleatedand then elongates along the <strong>DNA</strong> chain as B-<strong>DNA</strong> disappearsgradually. The whole chain becomes completely lefthanded at 1.85.It is worth noting that, a as the supercoiling degreechanges, the positions <strong>of</strong> the two peaks <strong>of</strong> the folding angledistribution remain almost fixed; and b between these twopeaks, there exists an extended region <strong>of</strong> folding angle from0to/6 which always has only an extremely small probability<strong>of</strong> occurrence. Thus a negatively supercoiled <strong>DNA</strong> canhave two possible stable configurations: a right-handed Bform and a left-handed configuration with an average foldingangle 48.6°. A transition between these two structuresfor a <strong>DNA</strong> segment will generally lead to an abrupt and finitevariation in the folding angle.To obtain the energetics <strong>of</strong> such transitions, we have calculatedhow the sum <strong>of</strong> the base-stacking energy and torsionalenergy, (/R 2 )sin 4 ()(/R)sin , changes withexternal torque 45. Figure 7a shows the numerical result,and Fig. 7b demonstrates the relation between the supercoilingdegree and external torque. In both figures the externalforce is fixed at 1.3 pN. From these figures we can


1054 HAIJUN ZHOU, YANG ZHANG, AND ZHONG-CAN OU-YANGPRE 62TABLE I. For the torque-induced left-handed <strong>DNA</strong> configuration,the average rise per base pair (d), the pitch per turn <strong>of</strong> thehelix, and the number <strong>of</strong> base pairs per turn <strong>of</strong> the helix Num. arecalculated and listed under different external forces and torques.The last row contains the corresponding values for Z-form <strong>DNA</strong>.Force pN Torque (k B T) d(Å ) Pitch (Å ) Num.1.3 5.0 3.59 41.20 11.481.0 5.0 3.57 40.93 11.441.3 4.0 3.83 46.76 12.191.0 4.0 3.82 46.38 12.15Z form 3.8 45.6 125 Another possibility may be that we have overestimated the value<strong>of</strong> the critical torque. It may be possible that the transition fromright- to left-handed configurations is initiated in the weaker ATrichregions, whose value <strong>of</strong> should be less than the average valuetaken by the present paper.infer that a for negative torque less than the critical value c 3.8k B T, <strong>DNA</strong> can only stay in B-form state; b nearthis critical torque, <strong>DNA</strong> can either be right or left handedand, as negative supercoiling increases see Fig. 7b moreand more <strong>DNA</strong> segments will stay in the left-handed form,which is much lower in energy (2.0k B T per base pairbut stable only when the torque reaches c ; and c fornegative torque greater than c , the <strong>DNA</strong> is completely lefthanded.Nevertheless, we should emphasize that the above calculationsare all based on our present model, which assumedthat nucleotide base pairs do not break. Figure 5 indicatesthat for negatively supercoiled <strong>DNA</strong> chains, the extension vssupercoiling degree relation can be quantitatively explainedby the present model. Figure 6 reveals the reason for thequantitative agreement is that the present model allows forthe possibility <strong>of</strong> the occurrence <strong>of</strong> left-handed <strong>DNA</strong> configurations.At the present time, to state that the negativelysupercoiled <strong>DNA</strong> will ‘‘prefer’’ left-handed configurationsrather than denaturation and strand separation is premature.To clarify this question, the present model should be improvedto consider deformations <strong>of</strong> <strong>DNA</strong> base pairs. On theexperimental side, it might also be helpful to measure preciselythe critical torque at which the elastic behavior <strong>of</strong>negatively supercoiled <strong>DNA</strong> changes abruptly, and to comparethe measured results with the value calculated in thepresent work. In the earlier experiment <strong>of</strong> Allemand et al.15, the critical torque was estimated to be 2k B T,byassuming the torsional rigidity <strong>of</strong> ds<strong>DNA</strong> to be 75 nm, andassuming that torsional stress builds up linearly along the<strong>DNA</strong> chain. This value, however, may not be preciseenough, since the torsional rigidity <strong>of</strong> ds<strong>DNA</strong> molecule isnot a precisely determined quantity, and the values given bydifferent groups are scattered widely. 5The structural parameters <strong>of</strong> the left-handed configurationsuggested by Figs. 6 and 7 are listed in Table I 45, andcompared with those <strong>of</strong> Z-form <strong>DNA</strong> 2. The strong similaritybetween these parameters suggests that the torqueinducedleft-handed configurations, if they really exist, belongto Z-form <strong>DNA</strong> 2,16,45.V. CONCLUSIONIn this paper, we have presented an elastic model for<strong>double</strong>-<strong>stranded</strong> biopolymers such as <strong>DNA</strong> <strong>molecules</strong>. Thekey progress is that the bending deformations <strong>of</strong> the backbones<strong>of</strong> <strong>DNA</strong> <strong>molecules</strong> and the base-stacking interactionsexisting between adjacent <strong>DNA</strong> base pairs are quantitativelyconsidered in this model, with the introduction <strong>of</strong> a structuralparameter: the folding angle . This model has also qualitativelytaken into account the effects <strong>of</strong> the steric effects <strong>of</strong><strong>DNA</strong> base pairs and electrostatic interactions along <strong>DNA</strong>chain. In the calculation technique, the model is investigatedusing the path integral method; in addition, Green equationssimilar in form to the Schrödinger equation in quantum mechanicsare derived, and their ground-state eigenvalues andeigenfunctions are obtained by precise numerical calculations.The force-extension relationship in torsionally relaxed<strong>DNA</strong> chains, and the extension-linking number relationshipin torsionally constrained <strong>DNA</strong> chains are studied and comparedwith experimental results. A <strong>DNA</strong> molecule’s entropicelasticity and highly extensibility, as well as the elastic <strong>property</strong><strong>of</strong> negatively supercoiled <strong>DNA</strong>, can all be quantitativelyexplained by the present theory. The comprehensive agreementbetween theory and experiments indicated that shortrangedbase-stacking interactions are very important in determiningthe elastic response <strong>of</strong> <strong>double</strong>-<strong>stranded</strong> <strong>DNA</strong><strong>molecules</strong>. The present work showed that highly extendedand negatively supercoiled <strong>DNA</strong> <strong>molecules</strong> can be lefthanded, probably in the Z-form configuration. A possibleway to check the validity <strong>of</strong> this opinion is to measure thecritical external torque at which the transition betweenB-form <strong>DNA</strong> and the new configuration takes place.The present work regarded <strong>DNA</strong> basepairs as rigid objects,and did not consider their possible deformations andthe possibility <strong>of</strong> strand separation. The comprehensiveagreement between theory and experiments indicates thatthis approach is well justified in many cases. However, asmentioned in Sec. II A, under some extreme conditions thisassumption may not be appropriate. For example, recent experimentalwork by Allemand et al. 15 showed that positivelysupercoiled <strong>DNA</strong> under high applied force can take onPauling-like configurations with exposed bases. In this casethe base pairing <strong>of</strong> <strong>DNA</strong> is severely distorted, and becausethe present model has not taken into account the possibledeformations <strong>of</strong> the base pairs, the theoretical results onpositively supercoiled <strong>DNA</strong> <strong>molecules</strong> are not in quantitativeagreement with experiment see Fig. 5. Furthermore, althoughthe present work showed that left-handed <strong>DNA</strong> configurationscan be stabilized by negative torques, much theoreticalwork is still needed to calculate the denaturation freeenergy and to compare with the free energy <strong>of</strong> left-handed<strong>DNA</strong> configurations.ACKNOWLEDGMENTSIt gives us great pleasure to acknowledge the help andvaluable suggestions <strong>of</strong> our colleagues, especially Lou Ji-Zhong, Liu Lian-Shou, Yan Jie, Liu Quan-Hui, Zhou Xin,Zhou Jian-Jun, and Zhang Yong. The numerical calculations


PRE 62 ELASTIC PROPERTY OF SINGLE DOUBLE-STRANDED . . .1055are performed partly at ITP-Net and partly at the State KeyLab <strong>of</strong> Scientific and Engineering Computing.APPENDIX A: PATH INTEGRAL METHODIN POLYMER PHYSICSIn this appendix we review some basic ideas on the application<strong>of</strong> the path integral method to the study <strong>of</strong> polymericsystems 39. Consider a polymeric string, and supposeits total ‘‘arclength’’ is L, and along each arclengthpoint s one can define an n-dimensional ‘‘vector’’ r(s) todescribe the polymer’s local state at this point. 6 We furtherassume that the energy density per unit arclength <strong>of</strong> thepolymer can be written in the general form e r,s m 2dsdr 2Ar• drds Vr,A1where V(r) is a scalar field and A(r) is a vectorial field. Thetotal partition function <strong>of</strong> the system is expressed by theintegrationLdr f f r f Gr f ,L;r i ,0 i r i dr i , A2where i (r) and f (r) are the probability distributions <strong>of</strong>the vector r at the initial (s0) and final (sL) arclengthpoint points, respectively. G(r,s;r,s) is called the Greenfunction, and is defined asGr,s;r,s rrDrsexp ssdse r,s ,A3where integration is carried over all possible configurations<strong>of</strong> r, and 1/k B T is the Boltzmann coefficient. It can beverified that the Green function defined above satisfies therelation 39Gr,s;r,s drGr,s;r,sGr,s;r,ssss.The total free energy <strong>of</strong> the system is then expressed asA4systems. Suppose the value <strong>of</strong> at arclength point s is relatedto its value at s through the following formula suchthat 7r,s drGr,s;r,sr,s ss; A6then we can derive, from Eqs. A1, A3, and A6, that 39r,ss “ r 22m VrAr•“ r “ r•Arm 2m A2 r2mr,sĤr,s. A7Equation A7 is called the Green equation; it is very similarto the Schrödinger equation <strong>of</strong> quantum mechanics 40.However, there is an important difference. In the case <strong>of</strong>A(r)0, the operator Ĥ in Eq. A7 is not Hermitian. Therefore,in this case the matrix form <strong>of</strong> the operator Ĥ may notbe diagonalized by unitary matrix.Denote the eigenvalues and the right eigenfunctions <strong>of</strong>Eq. A7 as g i and i i (r) (i0,1,...),respectively.Then it is easy to see, from the approach <strong>of</strong> quantum mechanics,thatsie g i (ss) isiss, A8where i i (r) (i0,1,...)denote the left eigenfunctions<strong>of</strong> Eq. A7, which satisfy the following relation:ii dr i r i r ii .In the case where Ĥ is Hermitian i.e., A(r)0, then wecan conclude that i r i*r.From Eqs. A2, A6, A7, and A8 we know thatL dr drGr,L;r,s f r i rFk B T ln .A5i f ii i e g i L e g 0 L f 00 i To calculate the total partition function , we define anauxiliary function (r,s) and call it the wave function, because<strong>of</strong> its similarity to the true wave function <strong>of</strong> quantumfor L1/g 1 g 0 .A9Consequently, for long polymer chains the total free energydensity is just expressed as6 For example, in the case <strong>of</strong> a flexible Gaussian chain, r is athree-dimensional position vector; in the case <strong>of</strong> a semiflexiblechain, such as a wormlike chain 18,41, r is the unit tangent vector<strong>of</strong> the polymer and is two dimensional.7 In fact, the choice <strong>of</strong> the wave function (r,s) is not limited.Any function determined by an integration <strong>of</strong> the form <strong>of</strong> Eq. A6can be viewed as a wave function.


1056 HAIJUN ZHOU, YANG ZHANG, AND ZHONG-CAN OU-YANGPRE 62F/Lk B Tg 0 ,A10and any quantity <strong>of</strong> interest can then be calculated by differentiation<strong>of</strong> F. For example, the average extension <strong>of</strong> a polymerunder external force field f can be calculated as ZF/ f Lk B Tg 0 / f .We continue to discuss another very important quantity,the distribution probability <strong>of</strong> r at arclength s, P(r,s). Thisprobability is calculated from the following expression:Pr,s dr f dr i f r f Gr f ,L;r,sGr,s;r i ,0 i r i . dr f dr i f r f Gr f ,L;r i ,0 i r i A11Based on Eqs. A6 and A8 we can rewrite Eq. A11 in thefollowing form:Pr,s dr f dr i dr f r f Gr f ,L;r,srrGr,s;r i ,0 i r i dr f dr i f r f Gr f ,L;r i ,0 i r i mn f mn i m r n rexpg m Lsg n sm f mm i expg m L. A12For the most important case <strong>of</strong> 0sL, Eq. A12 thengives that the probability distribution <strong>of</strong> r is independent <strong>of</strong>arclength s, i.e.,Pr,s 0 r 0 r for 0sL.A13With the help <strong>of</strong> Eq. A13, the average value <strong>of</strong> a quantitywhich is a function <strong>of</strong> r can be obtained. For example,Qs drQrPr,s dr 0 rQr 0 rand0Q00LQ„rs…ds 0LQsdsL0Q0for L1/g 1 g 0 .A14A15Finally, we list the formula for calculating B(r)•dr/ds; here B(r) is a given vectorial field. The formulareads Br• drds 1 2 0r•ĤBB•Ĥr0for L1/g 1 g 0 .APPENDIX B: MATRIX FORMALISMOF THE GREEN EQUATION „23…A16In Eq. 23 and also Eq. 15 the variables t and arecoupled together. Denote the operator in the square brackets<strong>of</strong> Eq. 23 as Ĥ. We express this operator in matrix form. Tothis end we choose the base functions <strong>of</strong> this system to becombinations <strong>of</strong> spherical harmonics Y lm (,) and f n (˜ )see Eq. 20.For m0, the base functions are expressed to beil•N nl;nY l0 ,f n ˜ ;B1and for m1,2,...,thebase functions are expressed to bei2lm•N 2n1kl;n;kY (k) lm ,f n ˜ .B2In the above two equations, n1,2,...,N , with N set tobe 60 in our present calculations; l0,1,...,N l 1 for thecase <strong>of</strong> m0 with N l set to 30); or lm,m1,...,mN l 1 for the case <strong>of</strong> m0 with N l set to 15). k1 and2, andY l0 , 2l14 P lcos ,Y (1) lm ,1 2l1 lm!m2 lm! P l m cos cosm,Y (2) lm ,1 2l1 lm!m2 lm! P l m cos sinm.With the above definitions, we then obtain that, for m0,


PRE 62 ELASTIC PROPERTY OF SINGLE DOUBLE-STRANDED . . .1057i p Ĥil p ;n p Ĥl;n ll p nn pll14l p* n2 24l p a 2fk B T l l1 apl,0 l1 lpa l1,0 n p sin ˜ nl lp n p ˜ k B T l pR 2 cos 4 ˜ nl lpRk B T n pcos ˜ n 2n np16l*k p B T 2 ll p1a l,02 2a l1,0 l2 lpa l,0 a l1,0 l2 lpa l1,0 a l2,0 ;B3and, for m0,li p Ĥil p ;n p ;k p Ĥl;n;k lp n knp kp ll14l*p n2 24l p a 2fk B T k k p l1 lpa l,m l1 lpa l1,m ln p sin ˜ n lp k kp n p ˜ k B T l pR 2 cos 4 l˜ n lpl lpn npk kp4l*k p B T kk pmRk B T n pcos ˜ n n 2knp kp16l*k p B T 2 ll p1a l,m2 2a l1,m l2 lpa l,m a l1,m l2 lpa l1,m a l2,m .In Eqs. B3 and B4, the following notations are used:a l,m l12 m 22l12l3 ,2 an p y˜ na0sin n p˜a y˜ sin n˜a ,B4where y(˜ ) is any function <strong>of</strong> ˜ .The ground-state eigenvalues <strong>of</strong> the matrices Eqs. B3and B4 have been calculated at force 1.3 pN in the wholerelevant region <strong>of</strong> from 5.0k B T to 5.0k B T. We havefound that the ground-state eigenvalues for the case <strong>of</strong> m0 are <strong>of</strong> the order <strong>of</strong> 10 nm 1 , while those for the case<strong>of</strong> m0 are <strong>of</strong> the order <strong>of</strong> 10 6 nm 1 data not shown.Thus we can conclude with confidence that the ‘‘low energy’’eigenstates <strong>of</strong> the system all have the same m0.This can greatly reduce the calculation tasks.As a final point, here we demonstrate how to calculate theaverage values <strong>of</strong> the quantities <strong>of</strong> our interest. Supposey(t,˜ ) is a quantity whose average value we are interestedin. Denote U as the unitary matrix which can diagonalize thematrix Eq. B3. Because the matrix Eq. B3 is real symmetric,U is actually a real orthogonal matrix. Then its columnvector u(i) j U( j,i) j (i1,2,...) corresponds tothe ith eigenvector <strong>of</strong> Eq. B3. Consequently, we can calculatebased, on Eq. A14, thatyt,˜ i,ipUi p ,1Ui,1i p yt,˜ i. B51 W. Saenger, Principles <strong>of</strong> Nucleic Acid Structure Springer-Verlag, New York, 1984.2 J. D. Watson, N. H. Hopkins, J. W. Roberts, J. A. Steitz, andA. M. Weiner, Molecular Biology <strong>of</strong> the Gene, 4th ed.Benjamin/Cummings, Menlo Park, CA, 1987.3 T. Nishinaka, Y. Ito, S. Yokoyama, and T. Shibata, Proc. Natl.Acad. Sci. USA 94, 6623 1997.4 T. Nishinaka, A. Shinohara, Y. Ito, S. Yokoyama, and T. Shibata,Proc. Natl. Acad. Sci. USA 95, 110711998.5 J. F. Léger, J. Robert, L. Bourdieu, D. Chatenay, and J. F.Marko, Proc. Natl. Acad. Sci. USA 95, 122951998.6 L. Stewart, M. R. Redinbo, X. Qin, W. G. J. Hol, and J. J.Champoux, Science 279, 1534 1998.7 V. V. Rybenkov, C. Ullsperger, A. V. Vologodskii, and N. R.Cozzarelli, Science 277, 690 1997.8 J. Yan, M. O. Magnasco, and J. F. Marko, Nature London401, 932 1999.9 S. B. Smith, L. Finzi, and C. Bustamante, Science 258, 11221992.10 D. Bensimon, A. J. Simon, V. Croquette, and A. Bensimon,Phys. Rev. Lett. 74, 4754 1995.11 P. Cluzel, A. Lebrun, C. Heller, R. Lavery, J.-L. Viovy, D.Chatenay, and F. Caron, Science 271, 792 1996.12 S. B. Smith, Y. Cui, and C. Bustamante, Science 271, 7951996.13 T. R. Strick , J.-F. Alleman, D. Bensimon, A. Bensimon, andV. Croquette, Science 271, 1835 1996.14 T. R. Strick, V. Croquette, and D. Bensimon, Proc. Natl. Acad.Sci. USA 95, 105791998.15 J. F. Allemand, D. Bensimon, R. Lavery, and V. Croquette,Proc. Natl. Acad. Sci. USA 95, 141521998.16 J. F. Léger, G. Romano, A. Sarkar, J. Robert, L. Bourdieu, D.Chatenay, and J. F. Marko, Phys. Rev. Lett. 83, 1066 1999.17 C. Bustamante, J. F. Marko, E. D. Siggia, and S. Smith, Science265, 1599 1994.18 J. F. Marko and E. D. Siggia, Macro<strong>molecules</strong> 28, 87591995.19 M.-H. Hao and W. K. Olson, Biopolymers 28, 873 1989;Macro<strong>molecules</strong> 22, 3292 1989.20 B. Fain, J. Rudnick, and S. Östlund, Phys. Rev. E 55, 73641996.21 J. F. Marko, Europhys. Lett. 38, 183 1997; Phys. Rev. E 57,2134 1998.22 J. D. Moroz and P. Nelson, Proc. Natl. Acad. Sci. USA 94, 14418 1997.


1058 HAIJUN ZHOU, YANG ZHANG, AND ZHONG-CAN OU-YANGPRE 6223 A. V. Vologodskii and J. F. Marko, Biophys. J. 73, 123 1997.24 B.-Y. Ha and D. Thirumalai, J. Chem. Phys. 106, 4243 1997.25 P. Cizeau and J.-L. Viovy, Biopolymers 42, 383 1997.26 C. Bouchiat and M. Mézard, Phys. Rev. Lett. 80, 1556 1998;e-print cond-mat/9904018.27 P. Nelson, Phys. Rev. Lett. 80, 5810 1998.28 Zhou Haijun and Ou-Yang Zhong-Can, Phys. Rev. E 58, 48161998; J. Chem. Phys. 110, 1247 1999.29 Zhou Haijun, Zhang Yang, and Ou-Yang Zhong-Can, Phys.Rev. Lett. 82, 4560 1999.30 R. Everaers, R. Bundschuh, and K. Kremer, Europhys. Lett.29, 263 1995.31 T. B. Liverpool, R. Golestanian, and K. Kremer, Phys. Rev.Lett. 80, 405 1998.32 L. Stryer, Biochemistry, 4th ed. Freeman, New York, 1995.33 T. T. Perkins, D. E. Smith, and S. Chu, Science 276, 20161997, and references cited therein.34 J. H. White, Am. J. Math. 91, 693 1969.35 F. B. Fuller, Proc. Natl. Acad. Sci. USA 68, 815 1971.36 F. H. C. Crick, Proc. Natl. Acad. Sci. USA 73, 2639 1976.37 J. H. White and W. R. Bauer, Proc. Natl. Acad. Sci. USA 85,772 1988.38 F. B. Fuller, Proc. Natl. Acad. Sci. USA 75, 3557 1978.39 Zhou Haijun unpublished; also see M. Doi and S. F. Edwards,The Theory <strong>of</strong> Polymer Dynamics Clarendon Press,Oxford, 1986; H. Kleinert, Path Integrals in Quantum Mechanics,Statistics, and Polymer Physics World Scientific,Singapore, 1990.40 J. Y. Zeng, Quantum Mechanics, 2nd ed. Academic Press,Beijing, 1999 in Chinese, Vol. I.41 K. Kroy and E. Frey, Phys. Rev. Lett. 77, 306 1996.42 G. V. Shivashankar, M. Feingold, O. Krichevsky, and A.Libchaber, Proc. Natl. Acad. Sci. USA 96, 7912 1999.43 M. Hegner, S. B. Smith, and C. Bustamante, Proc. Natl. Acad.Sci. USA 96, 101091999.44 Zhang Yang, Zhou Haijun, and Ou-Yang Zhong-Can, Biophys.J. 78, 1979 2000.45 Zhou Haijun and Ou-Yang Zhong-Can, Mod. Phys. Lett. B 13,999 1999.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!