1: \documentclass[aps,rmp,twocolumn]{revtex4}

2: 

3: \usepackage{amsmath}

4: \usepackage{amssymb}

5: \usepackage{graphicx}

6: \usepackage{psfrag}

7: \usepackage{mathrsfs}

8: \usepackage[usenames]{color}

9: 

10: \newcommand{\ecoli}{\emph{E.coli\/} }

11: \newcommand{\tr}{\mathsf{t}}

12: 

13: 

14: \newcommand{\AC}{{\it Acta Crystallogr.} }

15: \newcommand{\AM}{{\it Acta Metall.} }

16: \newcommand{\AP}{{\it Ann. Phys., Lpz.} }

17: \newcommand{\APNY}{{\it Ann. Phys., NY\/} }

18: \newcommand{\APP}{{\it Ann. Phys., Paris\/} }

19: \newcommand{\CJP}{{\it Can. J. Phys.} }

20: \newcommand{\JAP}{{\it J. Appl. Phys.} }

21: \newcommand{\JCP}{{\it J. Chem. Phys.} }

22: \newcommand{\JJAP}{{\it Japan. J. Appl. Phys.} }

23: \newcommand{\JP}{{\it J. Physique\/} }

24: \newcommand{\JPhCh}{{\it J. Phys. Chem.} }

25: \newcommand{\JMMM}{{\it J. Magn. Magn. Mater.} }

26: \newcommand{\JMP}{{\it J. Math. Phys.} }

27: \newcommand{\JOSA}{{\it J. Opt. Soc. Am.} }

28: \newcommand{\JPSJ}{{\it J. Phys. Soc. Japan\/} }

29: \newcommand{\JQSRT}{{\it J. Quant. Spectrosc. Radiat. Transfer\/} }

30: \newcommand{\NC}{{\it Nuovo Cimento\/} }

31: \newcommand{\NIM}{{\it Nucl. Instrum. Methods\/} }

32: \newcommand{\NP}{{\it Nucl. Phys.} }

33: \newcommand{\PL}{{\it Phys. Lett.} }

34: \newcommand{\PR}{{\it Phys. Rev.} }

35: \newcommand{\PRE}{{\it Phys. Rev. E.} }

36: \newcommand{\PRL}{{\it Phys. Rev. Lett.} }

37: \newcommand{\PRS}{{\it Proc. R. Soc.} }

38: \newcommand{\PS}{{\it Phys. Scr.} }

39: \newcommand{\PSS}{{\it Phys. Status Solidi\/} }

40: \newcommand{\PTRS}{{\it Phil. Trans. R. Soc.} }

41: \newcommand{\RMP}{{\it Rev. Mod. Phys.} }

42: \newcommand{\PNAS}{{\it Proc. Natl. Acad. Sci. U.S.A.\/} }

43: %

44: \newcommand{\etal}{{\it et al\/}\ }

45: \newcommand{\dash}{------}

46: \newcommand{\nonum}{\par\item[]}                %\par added 1/9/93

47: \newcommand{\mat}[1]{\underline{\underline{#1}}}

48: %

49: % abbreviations for IOPP journals

50: %

51: \newcommand{\CQG}{{\it Class. Quantum Grav.} }

52: \newcommand{\CTM}{{\it Combust. Theory Modelling\/} }

53: \newcommand{\DSE}{{\it Distrib. Syst. Engng\/} }

54: \newcommand{\EJP}{{\it Eur. J. Phys.} }

55: \newcommand{\HPP}{{\it High Perform. Polym.} }              % added 4/5/93

56: \newcommand{\IP}{{\it Inverse Problems\/} }

57: \newcommand{\JHM}{{\it J. Hard Mater.} }                    % added 4/5/93

58: \newcommand{\JO}{{\it J. Opt.} }

59: \newcommand{\JOA}{{\it J. Opt. A: Pure Appl. Opt.} }

60: \newcommand{\JOB}{{\it J. Opt. B: Quantum Semiclass. Opt.} }

61: \newcommand{\JPA}{{\it J. Phys. A: Math. Gen.} }

62: \newcommand{\JPB}{{\it J. Phys. B: At. Mol. Phys.} }      %1968-87

63: \newcommand{\jpb}{{\it J. Phys. B: At. Mol. Opt. Phys.} } %1988 and onwards

64: \newcommand{\JPC}{{\it J. Phys. C: Solid State Phys.} }   %1968--1988

65: \newcommand{\JPCM}{{\it J. Phys.: Condens. Matter\/} }    %1989 and onwards

66: \newcommand{\JPD}{{\it J. Phys. D: Appl. Phys.} }

67: \newcommand{\JPE}{{\it J. Phys. E: Sci. Instrum.} }

68: \newcommand{\JPF}{{\it J. Phys. F: Met. Phys.} }

69: \newcommand{\JPG}{{\it J. Phys. G: Nucl. Phys.} }         %1975--1988

70: \newcommand{\jpg}{{\it J. Phys. G: Nucl. Part. Phys.} }   %1989 and onwards

71: \newcommand{\MSMSE}{{\it Modelling Simulation Mater. Sci. Eng.} }

72: \newcommand{\MST}{{\it Meas. Sci. Technol.} }                 %1990 and onwards

73: \newcommand{\NET}{{\it Network: Comput. Neural Syst.} }

74: \newcommand{\NJP}{{\it New J. Phys.} }

75: \newcommand{\NL}{{\it Nonlinearity\/} }

76: \newcommand{\NT}{{\it Nanotechnology} }

77: \newcommand{\PAO}{{\it Pure Appl. Optics\/} }

78: \newcommand{\PM}{{\it Physiol. Meas.} }                        % added 4/5/93

79: \newcommand{\PMB}{{\it Phys. Med. Biol.} }

80: \newcommand{\PPCF}{{\it Plasma Phys. Control. Fusion\/} }      % added 4/5/93

81: \newcommand{\PSST}{{\it Plasma Sources Sci. Technol.} }

82: \newcommand{\PUS}{{\it Public Understand. Sci.} }

83: \newcommand{\QO}{{\it Quantum Opt.} }

84: \newcommand{\QSO}{{\em Quantum Semiclass. Opt.} }

85: \newcommand{\RPP}{{\it Rep. Prog. Phys.} }

86: \newcommand{\SLC}{{\it Sov. Lightwave Commun.} }               % added 4/5/93

87: \newcommand{\SST}{{\it Semicond. Sci. Technol.} }

88: \newcommand{\SUST}{{\it Supercond. Sci. Technol.} }

89: \newcommand{\WRM}{{\it Waves Random Media\/} }

90: \newcommand{\JMM}{{\it J. Micromech. Microeng.\/} }

91: \newcommand{\NAT}{{\it Nature\/} }

92: 

93: 

94: \begin{document}

95: 

96: \title{Single DNA conformations and biological function}

97: 

98: \author{Ralf Metzler}

99: \email{metz@nordita.dk}

100: \affiliation{NORDITA---Nordic Institute for Theoretical Physics,

101: Blegdamsvej 17, 2100 Copenhagen \O, Denmark}

102: \affiliation{Department of Physics, University of Ottawa, 150 Louis

103: Pasteur, Ottawa, Ontario, K1N 6N5, Canada (New address)}

104: \author{Tobias Ambj{\"o}rnsson}

105: \affiliation{NORDITA---Nordic Institute for Theoretical Physics,

106: Blegdamsvej 17, 2100 Copenhagen \O, Denmark}

107: \affiliation{Department of Chemistry, Massachusetts Institute of Technology,

108: 77 Massachusetts Avenue, Cambridge, Massachusetts 02139 (New address)}

109: \author{Andreas Hanke}

110: \affiliation{Department of Physics and Astronomy, University of Texas at

111: Brownsville, 80 Fort Brown, Brownsville, TX 78520}

112: \affiliation{Institute of Biomedical Sciences and Technology, University of

113: Texas at Dallas, Richardson, TX 75083}

114: \author{Yongli Zhang}

115: \affiliation{Department of Physiology and Biophysics, Albert Einstein

116: College

117: of Medicine, Bronx, NY 10461}

118: \author{Stephen Levene}

119: \affiliation{Institute of Biomedical Sciences and Technology and

120: Department of Molecular and Cell Biology, University of Texas at Dallas,

121: Richardson, TX 75083}

122: 

123: \begin{abstract}

124: Abstract. From a nanoscience perspective, cellular processes and their

125: reduced in vitro imitations provide extraordinary examples for highly

126: robust few or single molecule reaction pathways. A prime example are

127: biochemical reactions involving DNA molecules, and the coupling of these

128: reactions to the physical conformations of DNA. In this review,

129: we summarise recent results on the following phenomena: We investigate the

130: biophysical properties of DNA-looping and the

131: equilibrium configurations of DNA-knots, whose relevance to biological

132: processes are increasingly appreciated. We discuss how random DNA-looping

133: may be related to the efficiency of the target search process of proteins for

134: their specific binding site on the DNA molecule. And we dwell on the

135: spontaneous formation of intermittent DNA nanobubbles and their importance

136: for biological processes, such as transcription initiation. The physical

137: properties of DNA may indeed turn out to be particularly suitable for the

138: use of DNA in nanosensing applications.\\[0.2cm]

139: Key words: DNA, single molecules, DNA looping, DNA denaturation, knots,

140: gene regulation

141: \end{abstract}

142: 

143: \maketitle

144: 

145: \vspace*{8cm}

146: 

147: \newpage

148: \clearpage

149: 

150: \tableofcontents

151: 

152: \clearpage

153: 

154: \section{Introduction}

155: 

156: Deoxyribonucleic acid (DNA) is the molecule of life as we know

157: it.\footnote{Our DNA world during

158: biotic and prebiotic evolution was supposedly preceded by an RNA world and,

159: quite likely, by sugarless nucleic acids.} It contains all information of

160: an entire organism.\footnote{A small fraction of genetic information is stored

161: on DNA that is kept at other regions of the cell and not replicated on cell

162: division, such as mitochondrial or ribosomal DNA.}

163: This information is copied during cell division with an extremely

164: high fidelity by the replication mechanism. Despite the rather high

165: chemical and physical stability of DNA, due to constant action of enzymes

166: and other binding proteins (mismatches, rupture) as well as potential

167: environmentally induced damage (radiation, chemicals), this low error rate,

168: i.e., the suppression of the liability to mutations, is only possible with the

169: constant action of repair mechanisms \cite{alberts,snustad,kornberg,kornberg1}.

170: Although DNA's structural and mechanical properties are rather well established

171: for isolated DNA molecules (starting with Rosalind Franklin's X-ray diffraction

172: images \cite{franklin}), the characterisation

173: of DNA in its cellular environment, and even in vitro during interaction with

174: binding proteins, is subject of ongoing investigations.

175: 

176: %\begin{figure}

177: %\includegraphics[width=8cm]{franklin.eps}

178: %\caption{The diffraction photograph of the B form of DNA taken by Rosalind

179: %Franklin in May 1952 \cite{franklin}. Data derived from this photograph were

180: %instrumental in allowing James Watson and Francis Crick to construct their

181: %Nobel Prize­winning model for DNA \cite{watsoncrick}. $\copyright$ Nature.

182: %\label{franklin}}

183: %\end{figure}

184: 

185: Recent advances in experimental techniques such as fluorescence methods,

186: atomic force microscopy, or optical tweezers have leveraged the potential

187: to both probe and manipulate the equilibrium and out of equilibrium

188: behaviour of \emph{single\/} DNA molecules, making it possible to explore DNA's

189: physical and mechanical properties as well as its interaction with other

190: biopolymers, such as the DNA-protein interplay during gene regulation or repair

191: processes. An important ingredient is the coupling to thermal activation due

192: to the highly

193: Brownian environment. Although mostly performed in vitro, these experiments

194: provide access to increasingly refined information on the nature of DNA and

195: its environment-controlled behaviour.

196: 

197: In addition to chromosomal packaging inside the nucleus of eukaryotic cells

198: and the concentration of DNA in the membraneless nucleoid region of

199: prokaryotes, the global

200: structure of the DNA molecule can be affected by topological entanglements.

201: Thus, by error or design a DNA molecule can attain a knotted or concatenated

202: state, reducing or inhibiting biologically relevant functions, for instance,

203: replication or transcription. Such entangled states

204: can be actively reduced by enzymes of the topoisomerase family. Their precise

205: action, in particular, how they determine the presence of an entangled state,

206: is not fully known. Current studies therefore aim at shedding light on

207: possible mechanisms, in particular, in view of the importance of

208: topoisomerase action (or better, its inhibition) in tumour proliferation.

209: Other applications may be directed towards the treatment of viral deceases

210: by modifying the packaging of viral DNA to create knots in the virus capsid

211: and prevent ejection of the DNA into a host, and thereby infection.

212: DNA knots are also being recognised as a potential complication in the use

213: of nanochannels for DNA separation and sequencing. In such confined geometries

214: DNA knots are created with appreciable probability, affecting the reliability

215: of these techniques. Similarly to DNA knots, DNA looping is intimately

216: connected to the function of DNA. Current results on DNA looping and DNA

217: knot behaviour are summarised in the first parts of this review.

218: 

219: The Watson-Crick double-helix represents the thermodynamically stable state

220: of DNA at moderate salt concentrations and below the melting temperature.

221: This stability is effected by Watson-Crick hydrogen bonding and the

222: stronger base stacking of neighbouring base-pairs (bps). However, even at room

223: temperature DNA locally opens up intermittent flexible single-stranded domains,

224: so-called DNA-bubbles. Their size typically ranges from a few broken

225: bps, increasing to some 200 broken bps closer to the

226: melting temperature. The thermal melting of DNA has traditionally been used

227: to obtain the sequence-dependent stability parameters of DNA. More recently,

228: the role of intermittent bubble domains has been investigated with respect

229: to the liability of DNA-denaturation induced by proteins that selectively

230: bind to single-stranded DNA. It has been speculated that due to the liability

231: to denaturation of the TATA motif bubble formation may add in

232: transcription initiation. The dynamics of single bubbles can be monitored

233: by fluorescence methods, opening a window to both study the breathing of

234: DNA experimentally, but also to obtain high precision DNA stability data.

235: Finally, bubble dynamics has been suggested as a useful tool in optical

236: nanosensing. DNA breathing is the topic of the second part of this work.

237: 

238: Essentially all the biological functions of DNA rely on site-specific

239: DNA-binding proteins locating their targets (cognate sites) on the DNA

240: molecule, and therefore require searching through megabases of non-target

241: DNA in a highly efficient manner. For instance, gene regulation is performed

242: by specific regulatory proteins. On binding to a promoter area on the DNA,

243: they recruit or inhibit binding of RNA polymerase and subsequent

244: transcription of the associated gene. The search for the cognate site is

245: in fact facilitated by the DNA molecule: in addition to three-dimensional

246: search it enables the proteins to also move one-dimensionally along the

247: DNA while being non-specifically bound. Moreover, at points where the DNA

248: loops back on itself, this polymeric conformation provides shortcuts for

249: the proteins in the chemical coordinate along the DNA, approximately giving

250: rise to search-efficient L{\'e}vy flights. Target search is currently a very

251: active field of research, and single molecule methods have been shown to

252: provide essential new information. Moreover, the architecture of more complex

253: promoters relying on the simultaneous presence of several regulatory

254: proteins is being investigated to create in silico circuits for highly

255: sensitive chemical probes in small volumes. Such nanosensing applications

256: are expected to be of great importance in microarrays or other nano- and

257: microapplications. The third part of this review deals with diffusional

258: aspects of gene regulation.

259: 

260: At the same time DNA's role in classical polymer physics is increasingly

261: appreciated. With the possibility to reproduce DNA with extremely low

262: error rate by the PCR\footnote{Polymerase Chain Reaction: thermal

263: denaturation of a DNA molecule into two single strands and subsequent

264: cooling in a solution of single nucleotides and invariable primers,

265: produces two new complete double-stranded DNA molecules. Cycling of this

266: process produces large, monodisperse quantities of DNA.}, monodisperse

267: samples can be prepared. While shorter single-stranded DNA can be

268: used as a model for flexible polymers, the double strand exhibits a

269: semiflexible behaviour with a persistence length, that can be easily

270: probed experimentally. Moreover, DNA

271: is orders of magnitude longer than conventional polymers. Combined with

272: the potential of single molecule probing, DNA is advancing as a model

273: polymer.

274: 

275: After an introduction to the properties of DNA we address these functional

276: properties of DNA from the perspective of biological relevance, physical

277: behaviour and nanotechnological potential. Most emphasis will be put on

278: the single molecular aspects of DNA. We note that this is not intended to

279: be an exhaustive review on the physical properties of DNA. Rather, we

280: present some important features and their consequences from a personal

281: perspective.

282: 

283: 

284: \section{Physical properties and biological function of DNA}

285: 

286: Biomolecules, that occur naturally in biological systems, can be grouped

287: into unspecific oligo- and macromolecules and biopolymers in the stricter sense

288: \cite{alberts}. Unspecific biomolecules are

289: produced by biological organisms in a large range of molecular weight and

290: structure, such as polysaccharides (cellulose, chitin, starch, etc.),

291: higher fatty acids, actin filaments or microtubules. Also

292: the natural `india-rubber' from the {\em Hevea Brasiliensis\/} tree,

293: historically important for both industrial purposes and the development of

294: polymer physics \cite{treloar} belongs to this group.

295: 

296: \begin{figure}

297: \includegraphics[width=6.4cm]{mycrick.eps}

298: \caption{Central dogma of molecular biology after F. Crick:

299: Potentially, information flow is completely symmetric

300: between the three levels of cellular biopolymers (DNA, RNA, proteins).

301: However, the recognised pathways are only those represented here,

302: where solid lines represent probable transfers, and dotted lines for

303: (in principle) possible transfers \protect\cite{crick}.

304: \label{crick}}

305: \end{figure}

306: 

307: Biopolymers in the stricter sense we are going to assume here comprise

308: the polynucleotides DNA and RNA consisting of the four-letter nucleotide

309: alphabet with A-T and G-C (A-U and G-C for RNA) bps, and the

310: polypeptidic proteins consisting of 20 different amino acids, each coded for

311: by 3 bases (codons) in the RNA \cite{alberts,snustad,kornberg,kornberg1}.

312: We will come back to proteins later when reviewing binding protein-DNA

313: interactions. Biopolymers are copied and/or created according

314: to the information flow sketched in figure \ref{crick}, the so-called central

315: dogma of molecular biology, a term originally coined by Frances Crick

316: \cite{crick1}. Accordingly, starting from the genetic code stored in the DNA

317: (in some cases in RNA) DNA is copied by DNA polymerase (replication), and

318: the proteins as the actually task-performing biopolymers are created via

319: messenger RNA (created by DNA transcription through RNA polymerase) and

320: further by translation in ribosomes to proteins.\footnote{Alternatively, the

321: genetic code can be transcribed into transfer and ribosomal RNA that is not

322: translated into proteins.}

323: 

324: DNA is made up of the four bases

325: \cite{alberts,snustad,kornberg,kornberg1,bloomfield,rnaworld}:

326: A(denine), G(uanine), C(ytosine), and

327: T(hymine) that form the DNA ladder structure shown in figure \ref{dh}.

328: \begin{figure}

329: \includegraphics[height=7.68cm]{dna_bb.eps}

330: \caption{Ladder structure of the DNA formed by its four building-blocks

331: A, G, C, and T, giving rise to the typical double-helical structure

332: of DNA. A-T bps establish 2 H-bonds, G-C bps 3 H-bonds.

333: \label{dh}}

334: \end{figure}

335: These building-blocks A, G, C, T bp according to the key-lock principle

336: as A-T and G-C, where the AT bond is weaker than the GC bond in terms of

337: stability. Apart from the Watson-Crick base-pairing energy, the stability

338: of dsDNA is effected by the stacking interactions, the specific matching

339: of subsequent bps along the double-strand, i.e., bp-bp

340: interactions. In standard literature, the stacking interactions are listed

341: for pairs of bps (e.g., for AT-GC, AT-AT, AT-TA, etc.), see

342: below.\footnote{Longer ranging bp-bp interactions are most likely small in

343: comparison.}

344: 

345: \begin{figure*}

346: \includegraphics[width=18cm]{Ecoli_structureatlas.ps}

347: \caption{Structure atlas of the \ecoli genome. Figure courtesy David Ussery,

348: Technical University of Denmark. The structure atlas is available under

349: the URL {\tt www.cbs.dtu.dk/services/GenomeAtlas/}.

350: \label{structure}}

351: \end{figure*}

352: 

353: Based on this AGCT alphabet, the

354: primary structure of DNA can be specified. DNA's six local structural elements

355: twist, tilt, roll, shift, slide, and rise are effected by the stacking

356: interactions between vicinal bps. In figure \ref{structure}, we show a map

357: with the structure elements of the entire \ecoli genome, demonstrating the

358: degree of structural information currently available. These structural

359: elements define the local geometrical structure of DNA within a typical

360: correlation (persistence) length\footnote{The persistence length of a polymer

361: chain defines the characteristic length scale above which the polymer is

362: susceptible to bending induced by thermal fluctuations, i.e., it is the

363: length scale above which the tangent-tangent correlation decays along the

364: chain, see the Appendix.} of about 150 bps corresponding to 50 nm

365: (the bp-bp distance measures 3.4 {\AA}, reflecting the rather complex chemical

366: structure of a nucleotide in comparison to the monomer size of man-made

367: polymers such as polyethylene) \cite{frank,frank1,marko,marko1}.

368: On a larger scale, much longer than the persistence length,

369: DNA becomes flexible. On this level, tertiary structural elements come

370: into play. One example is DNA looping, that is the formation of polymeric

371: lasso loops induced by chemical bonds between binding proteins attached to

372: the DNA at specific bps which are remote along the DNA backbone

373: \cite{alberts,snustad,ptashne1,revet,bell,bell1,hame_looping}.

374: An extreme limit of tertiary structure is the

375: packaging of DNA onto histones and further wrapping into the chromosomes

376: of eukaryotic cells \cite{schiessel,kreth,alberts}. At the same time, dsDNA may

377: locally open into floppy ssDNA bubbles, with a persistence length of a few

378: bases.\footnote{In fact, it has been questioned whether there is a meaningful

379: value of the persistence length of ssDNA at all, due to its significant

380: apparent sequence dependence \cite{goddard}.} These fluctuation-induced bubbles

381: increase their statistical weight at higher temperatures, until the dsDNA fully

382: denatures (melts). We will come back to DNA denaturation

383: bubbles below. Depending on the

384: external conditions, DNA occurs in several configurations. Under physiological

385: conditions,

386: one is concerned with B-DNA, but there are other states such as A, B', Z, ps,

387: triplex DNA, quadruplex DNA, cruciform, and H, reviewed, for instance, in

388: \cite{frank,frank1}. DNA occurs naturally in a large

389: range of length scales. In viruses, DNA is of the order of a few $\mu$m

390: long. In bacteria, it already reaches lengths of several mm, and in mammalian

391: cells it can reach the order of a few m, roughly 2 m in a human cell and 35 m

392: in a cell of the South American lungfish, albeit split up into the

393: individual chromosomes \cite{kornberg}. DNA in bacteria in

394: vivo, or extracted from bacteria and higher cells for our purposes can

395: therefore be viewed a fully flexible polymer with a persistence length of

396: roughly 50 nm, being governed by generic effects independent of the detailed

397: sequence. On short scales DNA becomes semiflexible and governed by the

398: worm-like chain model (Kratky-Porod model) \cite{grosberg}; on even shorter

399: scales, local structural elements become important (in particular, for

400: recognition by binding proteins \cite{alberts}), and eventually molecular

401: resolution is reached.

402: 

403: Stacking interactions govern the local structure of dsDNA. Globally,

404: an additional constraint arises due to the circular nature of the DNA,

405: since it has to satisfy the conservation law \cite{calu,white1,fuller}

406: \begin{equation} \label{ltw}

407: {\rm Lk}={\rm Tw}+{\rm Wr},

408: \end{equation}

409: where Lk stands for the linking number, Tw for the twist, and Wr for the

410: writhe of the double helix. The linking number $Lk$ is an integer and

411: formally

412: given by one-half the number of signed crossings of one DNA strand with the

413: other in any regular projection of the molecule. $Lk$ is a topological

414: property, and no deformation of a closed DNA, without breaking and rejoining

415: the DNA strands, will alter it. $Tw$ is equal to the number

416: of times that the two strands of DNA wind about the central

417: axis of the molecule, and $Wr$ is a number whose absolute value equals

418: approximately the number of times that the DNA axis winds about

419: itself.\footnote{For details about the calculation of Tw and Wr for

420: representative models of DNA, see \cite{white}.}

421: Whereas Tw is a property of the double-helical structure of DNA,

422: Wr is a property of the DNA axis alone. Tw and Wr do not need to

423: be integers and are not conserved, but coupled through

424: Lk by equation (\ref{ltw}).

425: A nicked circular DNA, i.e., when the twist can fully relax, carries

426: ${\rm Lk}_0=N/h$ links, where $N$ is the number of bp and $h$ ($h\simeq 10.5$

427: in B-DNA) the number of bps per turn.

428: 

429: \begin{figure}

430: \includegraphics[width=7.6cm]{wind1.ps}

431: \caption{Right-handed (negative), normal, and left-handed (positive)

432: superhelix. The DNA of virtually all terrestrial organisms is

433: negatively supercoiled.}

434: \label{wind1}

435: \end{figure}

436: 

437: \begin{figure}

438: \includegraphics[height=6.0cm]{plasmid1.eps}

439: \includegraphics[height=6.0cm]{plasmid2.eps}

440: \caption{Electron micrographs of nicked (left) and supercoiled (right)

441: 6996-bp plasmid DNAs. The supercoiled example is from a population of DNA

442: molecules with an average superhelix density, $\bar {\sigma }$= -0.027,

443: close to the value expected \textit{in vivo}.}

444: \label{wind2}

445: \end{figure}

446: 

447: The degree of \emph{supercoiling\/} of DNA can be expressed in terms of

448: the linking number difference, $\Delta Lk=Lk-Lk_0$.

449: The DNA of virtually all terrestrial organisms is underwound or

450: {\em negatively supercoiled}, i.e., $\Delta Lk<0$

451: (figures \ref{wind1} and \ref{wind2}).\footnote{An exception are

452: thermophilic organisms living near

453: undersea geothermal vents that have positively supercoiled DNA in order to

454: stabilise the double helix at extreme temperatures.}

455: Often, the superhelical density $\sigma=\Delta Lk/Lk_0$ is used; most

456: supercoiled DNA molecules isolated from either prokaryotes or eukaryotes

457: have $\sigma$ values between $-0.05$ and $-0.07$ \cite{bauer}.

458: Negative supercoiling is regulated in prokaryotes by DNA gyrase; eukaryotes

459: lack gyrase but maintain negative supercoiling through winding of DNA around

460: nucleosomes and interactions with DNA-unwinding proteins.

461: There are two forms of intracellular supercoiling, the {\it plectonemic}

462: form, characteristic of plasmid DNA and accessible, nucleosome-free

463: regions of chromatin, and the {\it toroidal} or {\it solenoidal} form,

464: where supercoiling is

465: attained by DNA wrapped around histone octamers or prokaryotic non-histone

466: DNA-binding proteins (figure \ref{wind3}).

467: The former is the active form of supercoiled DNA and

468: is freely accessible to proteins involved in transcription, replication,

469: recombination and DNA repair. The latter is the stored form of supercoiled

470: DNA and is largely responsible for the extraordinary degree of compaction

471: required to condense typical genomes into the cell's nucleus.\footnote{The

472: nucleus of a human cell has a radius of circa 5 $\mu$m and stores the 2 m

473: of the human genome \cite{sun}.}

474: Negative supercoiling facilitates the local unwinding of DNA by providing

475: a ubiquitous source of free energy that augments the unwinding free energy

476: accompanying the interactions of many proteins with their cognate DNA

477: sequences. The local unwinding of DNA, in turn, is an integral part of many

478: biological processes such as gene regulation and DNA replication

479: (see section \ref{generegulation}). Therefore, understanding the interplay of

480: supercoiling and local helical structure is essential to the

481: understanding of biological mechanisms

482: \cite{marko1,benham,goetze,levene,levene1}.

483: 

484: \begin{figure}

485: \includegraphics[width=7.2cm]{wind3.ps}

486: \caption{Toroidal (left) and plectonemic (right) forms of supercoiled DNA.}

487: \label{wind3}

488: \end{figure}

489: 

490: Ribonucleic acid (RNA)

491: consists of the same building blocks as DNA, with the exception

492: that T(hymine) is replaced by U(racile) \cite{rnaworld}. RNA typically occurs

493: in single-stranded form. Therefore, its secondary structure is richer, being

494: characterised by sequences of hairpins: Smaller regions in which

495: chemically remote sequences

496: of bases match, pair and form hairpins which are stiff and energy-dominated,

497: similar to dsDNA. The remaining regions form entropy-dominated floppy loops,

498: analogous to the ssDNA bubbles. Additional

499: tertiary structure in RNA comes about by the formation of so-called

500: pseudoknots, chemical bonds established between bases sitting on chemically

501: distant segments of the secondary structure. In RNA-modelling

502: the incorporation of pseudoknots is a non-trivial problem, which currently

503: receives considerable interest; see, for instance,

504: references \cite{rnaworld,orland,orland1,baiesips}.

505: 

506: 

507: \section{DNA-looping}

508: 

509: The formation of DNA loops mediated by proteins bound at distant sites 

510: along a single molecule is an essential mechanistic aspect of many biological 

511: processes including gene regulation, DNA replication, and recombination 

512: (for reviews, see \cite{loop1,loop2}).

513: In {\em E. coli}, DNA looping represses gene expression 

514: at the {\em ara}, {\em gal}, {\em lac}, and {\em deo} 

515: operons \cite{loop3,loop4,loop5,loop6}

516: and activates transcription from 

517: the {\em gln}ALG operon \cite{loop7}.  

518: The size of DNA loops formed in these systems varies 

519: between approximately 100 and 600 bps.  In eukaryotes, a variety of

520: transcription factors bind to enhancers that are hundreds to several 

521: thousand bps away from their promoters and interact with RNA 

522: polymerases directly or through mediators in order to achieve combinatorial 

523: gene regulation \cite{loop8}.

524: DNA looping is required to juxtapose two recombination 

525: sites in intramolecular site-specific recombination 

526: \cite{loop9,loop10,loop11} and is also

527: employed by a number of restriction endonucleases such 

528: as {\em Sfi}I and {\em Ngo}MIV, 

529: which recognise and cut two copies of well-separated cognate sites 

530: simultaneously \cite{loop12,loop13,loop14}.

531: Here we describe a recent statistical-mechanical theory of 

532: loop formation that 

533: connects global mechanical and geometric properties of both DNA and 

534: protein and demonstrates the importance of protein flexibility in 

535: loop-mediated protein-DNA interactions \cite{loop26,loop51}.

536: 

537: 

538: \subsection{Biological significance of DNA looping}

539: \label{loopbio}

540: 

541: The biological importance of DNA loop formation 

542: is underscored by the abundance of architectural proteins in the cell such 

543: as HU, IHF, and HMG, which facilitate looping by bending the intervening

544: DNA between protein-recognition sites \cite{loop15}.

545: Moreover, DNA looping has been 

546: shown to be subject to regulation through the binding of effector 

547: molecules that alter protein conformation or protein-DNA 

548: interactions \cite{loop16}.

549: 

550: Two characteristics of DNA looping have been demonstrated by {\em in vitro}

551: and {\em in vivo} experiments. One is cooperative binding of a protein 

552: to its two cognate sites, which can be demonstrated by footprinting 

553: methods \cite{loop17}.  

554: DNA looping can increase the occupancies of both binding sites; in particular,

555: it can significantly enhance protein association to the lower-affinity site

556: because of the tethering effect of DNA looping.  This is a general mechanism 

557: by which many transcription factors recruit RNA polymerases in gene

558: regulation. Another hallmark is the helical dependence of loop formation

559: \cite{loop1,loop3}, which arises because of DNA's limited torsional 

560: flexibility and the requirement for correct 

561: torsional alignment of the two protein-binding sites.

562: Although many methods have been developed to directly observe DNA looping 

563: {\em in vitro}, such as scanning-probe \cite{loop7}

564: and electron microscopy \cite{bloomfield}, and single-molecule techniques

565: \cite{loop19}, assays based on helical dependence have 

566: been the only way to identify DNA looping {\em in vivo}. In these experiments, 

567: the DNA length between two protein binding sites is varied and the yield 

568: of DNA loop formation is monitored, for example by the repression or 

569: activation of a reporter gene \cite{loop20}.

570: Using this helical-twist assay, DNA 

571: looping in the {\em ara} operon was first discovered \cite{loop3}.

572: 

573: Our knowledge about the roles of DNA bending, twist, and their respective 

574: energetics in DNA looping has come largely from analyses of DNA 

575: cyclisation \cite{loop1,loop21,loop22}.

576: Circularisation efficiencies of DNA fragments, which are 

577: quantitatively described by $J$-{\em factors}, 

578: oscillate with DNA length and therefore torsional 

579: phase \cite{loop23,loop24}. The $J$-factor is defined as the

580: ratio of the partition function of a circularised polymer chain

581: to that of an open chain. Since there is a dimension reduction due to

582: circularisation constraints (two polymer ends have to meet), the ratio

583: has a unit of concentration, or $1/ L^3$ with $L$ representing length;

584: see \cite{loop26} for details.

585: In the present context, the $J$-factor is equal to the free DNA-end

586: concentration whose bimolecular ligation efficiency

587: equals that of the two ends of a cyclising DNA molecule \cite{loop25}.

588: For short DNA fragments $J$-factors are 

589: limited by the significant bending and twisting energies required to form 

590: closed circles, whereas for long DNA, the chain entropy loss during

591: circularisation exceeds the elastic-energy decrease and reduces the 

592: $J$-factor.  Because of this competition between bending and twisting 

593: energetics and entropy, there is an optimal DNA length for 

594: cyclisation \cite{loop26}.

595: Analogous behaviour has been expected for DNA looping, especially with 

596: respect to the helical dependence discussed above.

597: 

598: Quantitative analyses of DNA looping and cyclisation are challenging 

599: problems in statistical mechanics and have been largely limited to 

600: Monte Carlo or Brownian dynamics simulations 

601: \cite{loop27,loop28,loop29,loop30,loop31}. Analytical 

602: solutions are available only for some ideal and special cases.

603: An important contribution in this area is the theory of Shimada and 

604: Yamakawa \cite{loop32}, which is based on a 

605: homogeneous and continuous elastic 

606: rod model of DNA.  This theory has been applied extensively to DNA 

607: cyclisation \cite{loop23,loop33} and also DNA looping 

608: \cite{loop21,loop22,loop34}. The Shimada-Yamakawa

609: theory makes use of a perturbation approach, in which small configurational 

610: fluctuations of a DNA chain around the most probable configuration are 

611: accounted for in the evaluation of the partition function.

612: 

613: The elastic-equilibrium conformation is obvious for the homogeneous DNA 

614: circle studied by Shimada and Yamakawa \cite{loop32}.

615: However, the search for the 

616: elastic-energy minimum of homogeneous DNA molecules with complex geometry, 

617: such as in DNA looping, supercoiling, and the case of inhomogeneous DNA 

618: sequences containing curvature and nonuniform DNA flexibility, is not 

619: trivial \cite{loop4,loop35,loop36}. 

620: Recently, a statistical-mechanical theory for sequence-dependent 

621: DNA circles has been developed \cite{loop26} and applied to the

622: problem of DNA cyclisation \cite{loop26} and DNA looping \cite{loop51}. 

623: In this model, the DNA configuration is described by 

624: parameters defined at dinucleotide steps, i.e., tilt, roll, and twist, 

625: which allows straightforward incorporation of intrinsic or protein-induced 

626: DNA curvature at the bp level. Following Shimada and Yamakawa's

627: method, the theory first determines the mechanical equilibrium configuration

628: in small DNA circles (i.e., less than $\sim 1000$ bp) under certain

629: constraints; fluctuations around the equilibrium configuration are then

630: taken into account using an harmonic approximation.

631: The new method is much more computationally 

632: efficient than Monte Carlo simulation, has comparable accuracy, 

633: and has been applied successfully to analyse experimental results from 

634: DNA cyclisation \cite{loop26}. 

635: 

636: The basis of the extension of the model

637: to DNA looping \cite{loop51} is 

638: to treat the protein subunits as connected rigid bodies and to allow for 

639: a limited number of degrees of freedom between the subunits.  Motions of 

640: the subunits are assumed to be governed by harmonic potentials and an 

641: associated set of force constants, neglecting the anharmonic terms often 

642: required for proteins undergoing large conformational fluctuations among 

643: their modular domains.  Indeed, the use of a harmonic approximation is 

644: supported by the success of continuum elastic models that are based only 

645: on shape and mass-distribution information in descriptions of protein 

646: motion \cite{loop37}.

647: Similar to the description used for individual DNA bps 

648: in the model, protein geometry and dynamics are described by three 

649: rigid-body

650: rotation angles (tilt, roll, and twist).  Therefore, DNA looping can be 

651: viewed as a generalisation of DNA cyclisation in which the protein component 

652: is characterised by a particular set of local geometric constraints and 

653: elastic constants. This treatment not only unifies the theoretical 

654: descriptions of DNA cyclisation and looping, but also allows consideration 

655: of flexibilities at protein-DNA and protein-protein interfaces and 

656: application of the concepts of linking number and writhe. In previous 

657: work, proteins were considered rigid and their effects on DNA configuration 

658: were represented by a set of constraints applied to DNA ends 

659: \cite{loop1,loop38,loop39}.

660: With the present approach, programs developed for analysing DNA cyclisation 

661: can be used to analyse DNA looping with only minor modifications.

662: 

663: The new method \cite{loop26,loop51} is most applicable to the problem 

664: of short DNA loops, in which 

665: the free energy of a wormlike chain is dominated by bending and torsional 

666: elasticity \cite{loop26,loop51}.

667: Possible modes of 

668: DNA self contact and contacts between protein and DNA at positions other 

669: than the binding sites are not considered.

670: For large loops 

671: contributions to the free energy from chain entropy and DNA-DNA contacts 

672: can become highly significant. Several alternative treatments of DNA 

673: looping have appeared recently. One of these addresses the excluded-volume 

674: contribution to DNA looping within large open-circular molecules 

675: \cite{hame_looping},

676: whereas two others consider the effect on looping of traction at the ends 

677: of a DNA chain \cite{loop41,loop42}.

678: None of these treatments includes helical phasing

679: effects on DNA looping.  In contrast, a method based on the Kirchhoff 

680: elastic-rod model, which includes the helical-phase dependence, has been 

681: presented \cite{loop39,loop43}.

682: However, this approach does not include thermal 

683: fluctuations {\em per se} and therefore is not directly applicable to 

684: calculations of the $J$-factor.  The comprehensive treatment of small DNA

685: loops described in \cite{loop26,loop51}

686: is thus far unique to the extent that it accounts for 

687: sequence- and protein-dependent conformational and flexibility parameters, 

688: thermal fluctuations, and helical phasing effects.

689: 

690: 

691: \subsection{DNA loop model}

692: 

693: The protein subunits that mediate loop 

694: formation are modelled as two identical and connected rigid 

695: bodies, as shown in figure \ref{loopfig1} \cite{loop51}.

696: There are three additional sets of rigid-body 

697: rotation angles that are defined in addition to those for dinucleotide 

698: steps: two sets for the interfaces between protein and the last (DP) and 

699: first bps (PD) of the DNA and one set for the interface between the 

700: two protein domains (PP), where the symbols in parentheses are used to 

701: indicate the corresponding angles through subscripts. The local 

702: Cartesian-coordinate frames for protein subunits are defined such that 

703: their origins coincide with vertices of a circular chain and their 

704: $z$-axes point toward the next vertex in succession.  Thus protein 

705: dimensions can be modelled in terms of a non-canonical value for the helix 

706: rise corresponding to particular segments within a circular polymer chain.

707: 

708: \begin{figure}

709: \begin{center}

710: \includegraphics[width=7.6cm]{bk_loopfig1.ps}

711: \end{center}

712: \caption{Rigid-body models for studies of protein-mediated DNA looping.

713: (a) A prototype 137-bp DNA loop generated by interactions with a pair of 

714: rigid, DNA-binding protein subunits is shown.  DNA bps are 

715: represented by rectangular slabs (red) with axes (blue) that indicate 

716: the orientation of the local Cartesian coordinate frame whose origin 

717: lies at the centre of each bp.  Two sets of coordinate axes 

718: (green) represent the local coordinate frames embedded in the protein 

719: subunits (gold ellipsoids) that mediate DNA looping.  The coupling of 

720: protein and DNA geometry is characterised by tilt, roll, and twist 

721: values for the DNA-protein, protein-protein, and protein-DNA interfaces.  

722: Three of these variables are shown here: the DNA-protein roll angle, 

723: $\phi_{DP}$; the protein-protein twist angle, $\tau_{PP}$; 

724: and the protein-DNA roll angle, $\phi_{PD}$.  

725: (b) Prototype 179-bp loop with protein-protein twist angle, $\tau_{PP}$,

726: equal to $-60$ degrees.  The view is from the base of the loop toward 

727: the DNA apex.  (c) Loop conformation shown in (b) viewed from the side, 

728: perpendicular to the loop dyad axis.

729: \label{loopfig1}}

730: \end{figure}

731: 

732: Angles are expressed in degrees, and length in units of the DNA helical 

733: rise, $\ell_{bp} = 3.4$ {\AA}. All 

734: calculations used canonical mechanical parameters for duplex DNA: helical 

735: twist $\tau_{0} = 34.45^{\circ}$, a sequence-independent twist-angle 

736: standard deviation, or twisting flexibility, $\sigma_{\tau} = 4.388^{\circ}$,

737: and standard deviations, or bending flexibilities, for all tilt and roll 

738: angles,

739: $\sigma_{\theta}$ and $\sigma_{\phi}$, respectively, of $4.678^{\circ}$

740: (equivalent to a persistence length of 150 bp). Except for specific cases 

741: where intrinsic DNA bending is considered, the average values of tilt and 

742: roll are taken to be zero.

743: 

744: 

745: \subsection{Simplified protein geometries and flexibility parameters}

746: \label{loopsection}

747: 

748: For DNA loops with either zero or nonzero end-to-end distances, constraints 

749: are directly applied to the DNA ends, as in the case of DNA cyclisation.  

750: We modelled DNA loops formed during site synapsis using protein-dependent 

751: parameters

752: ${\rm{roll }} = \phi_{DP}  = \phi_{PD} = 90^{\circ}$ and

753: ${\rm{twist}} = \tau_{DP}  = \tau_{PD} = 34.45^{\circ}$.

754: The angle was considered an adjustable parameter that we denote the 

755: {\em axial angle} and, unless specified, all other protein-related angular 

756: parameters were set equal to $0^{\circ}$. In these cases the DNA ends 

757: (the centres of two protein-binding sites on DNA) are separated by twice 

758: the protein-arm length $\ell_p$ and displaced from one another along 

759: the $+x$ direction, or toward the major groove of DNA. Projected along 

760: the $x$-axis, the axial angle is the included angle between the tangents 

761: to the DNA at the two protein binding sites and is altered by varying the 

762: twist between protein subunits (figure \ref{loopfig1} b, c). 

763: An axial angle equal to $0^{\circ}$

764: corresponds to antiparallel axes at the ends as shown in 

765: figure \ref{loopfig1}a.  

766: The case of a rigid protein assembly is modelled by setting the standard 

767: deviations of the DP, PP, and PD sets of rigid-body rotation angles to 

768: $1 \cdot 10^{-8}$ deg.

769: 

770: 

771: \subsection{DNA loops having zero end-to-end distance and antiparallel 

772: helical axes}

773: 

774: DNA loops containing $N$ bps in which the two ends meet in an 

775: antiparallel orientation can be empirically described by the following 

776: formula:

777: %

778: \begin{eqnarray} \label{1}

779: & & {\rm{Tilt:}}  \quad \, \, \, \, 

780: \theta_i = - A_i \, \cos(180 + \delta) \\ \nonumber

781: & & {\rm{Roll:}}  \quad \, \, \, \phi_i = A_i \, 

782: \sin (180 + \delta ) \\ \nonumber

783: & & {\rm{Twist:}} \quad \tau_i  = \tau^0 \nonumber

784: \end{eqnarray}

785: %

786: where $\tau^{0}$ is the intrinsic DNA twist and $\delta$ an arbitrary angle 

787: related to the unconstrained torsional degree of freedom of DNA. The 

788: coefficients $A_i$ are given by

789: %

790: \begin{equation} \label{2}

791: A_i = \frac{1}{N} \, f\left( {\frac{i}{{N - 1}}} \right) \, ,

792: \quad i = 0, \ldots, N - 1

793: \end{equation}

794: %

795: with

796: %

797: \begin{equation} \label{3}

798: f(x) = \left\{ \begin{array}{l}

799: g(x) \, \, , \qquad \quad \, \, \, 0 \le x \le 0.5 \\

800: g(0.5 - x) \, \, , \quad 0.5 < x \le 1 \\

801: \end{array} \right.

802: \end{equation}

803: %

804: where

805: %

806: \begin{equation} \label{4}

807: g(x) = \sum\limits_{i = 1}^5 {a_{i\,} x^i } ,\quad 0 \le x \le 0.5 \, \, .

808: \end{equation}

809: %

810: \begin{figure}

811: \begin{center}

812: \includegraphics[width=7.6cm]{bk_loopfig2.ps}

813: \end{center}

814: \caption{Conformation of an antiparallel, 150-bp DNA loop with zero 

815: end-to-end distance.  (a) Computed space-filling model of the loop 

816: generated with 3DNA \cite{loop49}.  

817: The ends of the DNA juxtapose exactly with 

818: antiparallel helical axes and exact torsional phasing. (b) Equilibrium 

819: roll and magnitude of the loop shown in (a).  The bending magnitude of 

820: each dinucleotide step is defined as $\sqrt{\theta_i^2 + \phi_i^2}$

821: where $\theta_i$ and $\phi_i$ are the tilt and roll of $i$-th 

822: dinucleotide step, respectively.

823: \label{loopfig2}}

824: \end{figure}

825: %

826: The coefficients in equation (\ref{4}) were obtained by fitting the space 

827: curve corresponding to the DNA helical axis that gives the minimum elastic 

828: energy conformation of DNA loops of different sizes and are as follows:

829: $a_0 = -335.0142$, $a_1 = 2318.881$, $a_2 = -1299.164$, 

830: $a_3 = -4483.366$, $a_4 = 38169.74$, $a_5 = -54753.5$. 

831: The error for end-to-end distances computed using 

832: equation (\ref{1}) is less

833: than $2\%$ of DNA length from 50 bp to 100 bp, and less than $0.5\%$ 

834: from 100 bp to 500 bp. The torsional phase angle between two ends is

835: $\xi  =  - \left( {N - 2} \right)\tau  - 2\delta$. The entire loop lies 

836: in a plane, and the angle between the normal vector of the plane and

837: the $x$-axis of the external coordinate can be shown to be 

838: $\psi  = 180 + \tau  - \delta$. The expressions for $\xi$ and $\psi$ 

839: suggest that $\delta$ is related to DNA bending isotropy. Loop 

840: configurations with different $\delta$ values are related to each other 

841: by globally twisting DNA molecules. Since the orientation of the first 

842: bp is fixed, this global twist is equivalent to rotation of the 

843: loop plane, which corresponds to the rotational symmetry met in DNA 

844: cyclisation of homogeneous DNA with bending isotropy \cite{loop26}. 

845: Therefore, $J$-factors for configurations with different $\delta$ values 

846: are identical.

847: 

848: If DNA looping needs to be torsionally in-phase, only two degenerate 

849: loop configurations are available, breaking the rotational symmetry. 

850: These loop geometries can be expressed

851: by equation (\ref{1}) with two different $\delta$ values:

852: $\delta_1 = - (N - 2) \tau / 2$ and

853: $\delta_2  = 180 - (N - 2) \tau / 2$,

854: which satisfy the torsional phase requirement 

855: $\xi  = 360 \cdot n,\;n = 0, \pm 1, \pm 2, \ldots$

856: In contrast to DNA cyclisation, no twist change is involved in forming these 

857: ideal DNA loops for any DNA length and thus the helical dependence vanishes 

858: in this case.  From the expression given above for $\psi$ it is clear that 

859: the helical axes of the two loops are coincident and their directions are 

860: reversed. Figure \ref{loopfig2}

861: shows the bending profile of the loop configuration 

862: corresponding to $\delta_1$ for a 150 bp DNA. Surprisingly, the maximal 

863: $J$-factor occurs at approximately the same DNA length, or 460 bp (data 

864: not shown), as in DNA cyclisation \cite{loop26}.

865: This can be partly explained by 

866: the fact that the total bending magnitude of the loop is $290$ degrees, 

867: close to a full circle, instead of $180$ degrees.

868: 

869: 

870: 

871: \subsection{DNA looping with finite end-to-end distance, antiparallel 

872: helical axes, and in-phase torsional constraint}

873: 

874: \begin{figure}

875: \begin{center}

876: \includegraphics[width=6.8cm]{fig9.eps}

877: \end{center}

878: \caption{The DNA-length-dependent $J$-factor and loop configuration as a 

879: function of end-to-end separation (the $J$-factor is defined

880: in section \ref{loopbio}).

881: (a) The helical dependence of DNA looping 

882: is shown for values of the end-to-end separation equal to 10 bp and 30 bp. 

883: The two configurations for the 10-bp separation are obtained from 

884: corresponding configurations with zero end-to-end separation by using an 

885: iterative algorithm. Therefore the two configurations are designated by the 

886: initial configurations with phase angles 

887: $\delta = -(N-2) \tau / 2 + 0$ ($0^{\circ}$, dashed line) and

888: $\delta = -(N-2) \tau / 2 + 180$ ($180^{\circ}$, solid line)

889: as described in the text. (b) and (c) show

890: stereo models of the two equilibrium 

891: configurations for 210-bp (b) and 215-bp (c) antiparallel DNA loops with 

892: end-to-end separation equal to 10 bp. The 210- and 215-bp DNA correspond 

893: to an adjacent peak and valley of the curve in (a), respectively. 

894: Conformations shown in blue correspond to $\delta = 0$; 

895: those shown in red are for $\delta = 180^{\circ}$. Note that for $N$-bp DNA, 

896: the chain contour length is equal to $(N-1)\ell_{bp}$.

897: \label{loopfig3}}

898: \end{figure}

899: 

900: 

901: Separation of the DNA ends breaks the rotational symmetry, restoring the 

902: dependence on helical twist. Figure \ref{loopfig3}a 

903: shows the $J$-factor as a function 

904: of DNA length for end-to-end distances of $10$ bp and $30$ bp. The helical 

905: dependence increases with end-to-end separation.  Starting from the two 

906: loop configurations (corresponding to $\delta_1$ and $\delta_2$) with zero

907: end-to-end distance and in-phase torsional alignment as initial

908: configurations, two mechanical equilibrium configurations are obtained by 

909: using the iterative algorithm described in \cite{loop26}.

910: The $J$-factor 

911: in figure \ref{loopfig3}a is the sum of separate $J$-factors 

912: calculated for the two 

913: configurations. Note that in all cases involving configurations that differ 

914: in linking number, equilibration between the two forms requires breakage of

915: at least one of the protein-DNA interfaces. The contributions from each of 

916: these configurations are shown in detail for the case where the ends are 

917: separated by $10$ bp. Interestingly, the length dependence of $J$ computed 

918: from the individual configurations are out of phase and have a periodicity 

919: of 2 helical turns, which results from the half-twist dependence of the 

920: phase angles $\delta_1$ and $\delta_2$.  However, their sum displays a 

921: periodicity of one helical turn. Figures \ref{loopfig3} b and c

922: show two such configurations for DNA molecules that are torsionally 

923: in-phase ($N = 210$ bp) or out-of-phase ($N = 215$ bp).

924: 

925: In the case of cyclisation, the helical-phase dependence of the $J$-factor 

926: persists at DNA lengths well beyond that corresponding to the maximum 

927: value of $J$, which lies near $500$ bp. This is clearly not the case for 

928: DNA looping.  In figure \ref{loopfig3}a, 

929: the periodic dependence of $J$ on DNA length 

930: for $10$-bp end-to-end separation decays nearly to zero well before the 

931: maximum $J$ value is reached.  Although the periodicity of $J$ is not 

932: attenuated quite as strongly for $30$-bp end separation, there is less 

933: than four-fold variation in the value of $J$ near $300$ bp, as opposed 

934: to the more than ten-fold variation in cyclisation $J$-factors expected

935: in this length range.  The differences between looping and cyclisation 

936: are largely due to substantial differences in the relative contributions 

937: of DNA writhe in the two processes, as discussed below.

938: 

939: 

940: \subsection{DNA looping in synapsis}

941: \label{loopsyn}

942: 

943: Intramolecular reactions of most site-specific recombination systems 

944: \cite{loop9,loop10,loop11}

945: and a number of DNA restriction endonucleases such as {\em Sfi}I and 

946: {\em Ngo}MIV \cite{loop12},

947: proceed through protein-mediated intermediate structures 

948: in which a pair of DNA sites are brought together in space and the

949: intervening DNA is looped out.  The intermediate nucleoprotein complex 

950: involved in site pairing and strand cleavage (and also exchange, in the 

951: case of recombinases) is termed the {\em synaptic complex}. In these systems, 

952: two characteristic geometric parameters are of interest: the average 

953: through-space distance between the sites and the average crossing angle 

954: between the two ends of the loop, which we denote the axial angle

955: (see section \ref{loopsection}).

956: The latter quantity can be described in terms of the twist angle between 

957: the protein domains, $\tau_{PP}$ (figure \ref{loopfig1}b), 

958: and we use these terms interchangeably.

959: 

960: Figure \ref{loopfig4} shows the helical dependence of looping 

961: (figure \ref{loopfig4}a) and the 

962: elastic-minimum configuration of DNA loops 

963: (figure \ref{loopfig4}b) for different 

964: values of the axial angle. The most prominent feature of these results 

965: is that the phase of the helical dependence is shifted as a function of the

966: axial angle, characterised by a relative global shift of the curve along 

967: the $x$-axis. This implies that DNA looping does not always occur most 

968: efficiently when two sites are separated by an integral number of helical 

969: turns, as has been suggested for some simple DNA looping systems studied 

970: previously. The axial angle also globally modulates $J$-factors, which 

971: is apparent from the vertical shift in the $J$ versus length curve and 

972: effects on the amplitude of the helical dependence. 

973: The torsion-angle-independent value of $J$, averaged over a full helical 

974: turn, decreases with increasing axial angle, whereas the amplitude of 

975: the helical dependence increases.  The above observations can be 

976: qualitatively explained by analogous results from DNA cyclisation.  

977: As in cyclisation, DNA forms loops most efficiently when the number of 

978: helical turns in the loop is close to an integer value. It is therefore 

979: appropriate to consider this issue in terms of the linking number for the

980: looped conformation, $Lk$, which involves contributions from the geometries 

981: of both the protein and DNA.

982: 

983: \begin{figure}

984: \begin{center}

985: \includegraphics[width=7.6cm]{bk_loopfig4.ps}

986: \end{center}

987: \caption{Dependence of the $J$-factor on axial angle 

988: (the $J$-factor is defined in section \ref{loopbio},

989: and the axial angle is defined as 

990: the average crossing angle between the two 

991: ends of the loop, see section \ref{loopsection}).

992: (a) DNA-length dependence of $J$ for axial angles of 

993: $0^{\circ}$, $60^{\circ}$, and $120^{\circ}$ 

994: with the end-to-end separation set equal to 40 bp.

995: Note that the positions 

996: of the extrema shift to the left with increasing values of the axial angle.  

997: (b) Stereo models of minimum elastic-energy conformations of 179-bp loops 

998: colour coded in accord with the corresponding axial-angle values in (a).

999: \label{loopfig4}}

1000: \end{figure}

1001: 

1002: We define the {\em loop helical turn} $H_{t,loop}$ as the sum of the 

1003: DNA twist and the twist introduced by the protein subunits, divided by 

1004: $360$.  Therefore, changing the twist angle, the axial angle will 

1005: shift the phase of the helical dependence relative to that of

1006: the DNA alone.  For a loop with $N = 179$ bp and $\tau_{PP} = 0$, the 

1007: total twist is simply equal to that for the DNA loop.  Because this loop 

1008: has $17.0$ helical turns, only one loop topoisomer contributes to the 

1009: $J$-factor.  The value of $J$ is a local maximum at $\tau_{PP} = 0$ and, 

1010: as shown in figure \ref{loopfig5}a,

1011: decreases monotonically for both $\tau_{PP} > 0$

1012: and $\tau_{PP} < 0$.  Contributions to $J$ from other topoisomers of 

1013: the $179$-bp loop are less than $5$ percent over the range 

1014: $-135^{\circ} < \tau_{PP} < +120^{\circ}$. The twist for the planar 

1015: equilibrium conformation of a $173$-bp loop is $16.5$ helical turns;

1016: thus there are two alternative loops that can be efficiently formed 

1017: (figure \ref{loopfig5}a): 

1018: either a loop with $H_{t,loop} = 17.0$ and $\tau_{PP} > 0$, 

1019: or a loop with $H_{t,loop} = 16.0$ and $\tau_{PP} < 0$. The $J$ value 

1020: at $\tau_{PP} = 0$ is a local minimum and there is a bimodal dependence 

1021: on axial angle for loops in which the DNA twist is half-integral. 

1022: We investigated the phase shift of the $J$-factor and found that 

1023: this quantity is a non-linear function of the axial angle.  From 

1024: figure \ref{loopfig4}a, 

1025: the calculated phase shifts for $60^{\circ}$ and $120^{\circ}$ 

1026: axial angles relative to $0^{\circ}$ are approximately $52^{\circ}$ 

1027: and $103^{\circ}$, respectively.  Moreover, the local maxima for the total

1028: $J$ curve for $N = 173$ shown in figure \ref{loopfig5}a

1029: are located at $-58.5^{\circ}$ 

1030: and $63^{\circ}$, positions that are not in agreement with predicted 

1031: angle values based solely on $H_{t,loop}$ ($-166^{\circ}$ and 

1032: $194^{\circ}$, respectively).

1033: 

1034: \begin{figure}

1035: \begin{center}

1036: \includegraphics[width=7.6cm]{bk_loopfig5.ps}

1037: \end{center}

1038: \caption{$J$-factor, loop-geometry parameters, and elastic-free energies as 

1039: functions of axial angle; compare figure \ref{loopfig4}.

1040: (a) $J$-factor values for loop topoisomers 

1041: corresponding to 179-bp and 173-bp loops in figure \ref{loopfig4}.

1042: The principal contribution to $J$ for $N = 179$ bp comes from a single 

1043: loop topoisomer with $Lk = 17$.  For $N = 173$ bp, the overall $J$-factor 

1044: is the sum of contributions from two loop topoisomers with $Lk$ values of 

1045: 16 and 17, generating a bimodal dependence of $J$ on axial angle as 

1046: described in the text. (b) Excess helical twist, $\Delta H_t$, 

1047: and writhe of the loop formed by the $Lk = 16$ topoisomer for $N = 173$ bp

1048:  as a function of axial angle.  Excess twist is computed from the expression 

1049: $H_{t, loop} - 16$, where $H_{t, loop}$ is the loop helical turn value 

1050: described in the text, and depends linearly on the axial angle.  

1051: The writhing number of the loop was calculated using the method of 

1052: Vologodskii \cite{loop28,loop50}. (c) Elastic-free energies of the 

1053: $Lk = 16$ loop topoisomer for $N = 173$ bp calculated according to 

1054: equation 38 of Zhang and Crothers \cite{loop26}. The individual 

1055: contributions of bending and twisting energies are shown along 

1056: with their sum.

1057: \label{loopfig5}}

1058: \end{figure}

1059: 

1060: These deviations can be explained by the fact that writhe makes an 

1061: important contribution to the overall $Lk$ for the loop.  This aspect 

1062: of DNA looping is dramatically different from that in the cyclisation 

1063: of small DNA molecules.  The conformations of small DNA circles are

1064: close to planar and the writhe contribution is small relative to DNA 

1065: twist \cite{loop26,loop30,loop44,loop45}.

1066: In the case of protein-mediated looping, nonzero 

1067: values of the axial angle impose an intrinsically nonplanar conformation 

1068: on the DNA.  The relative contributions of loop writhe and twist for 

1069: the $Lk = 16$ topoisomer of a $173$-bp loop are shown as a function of 

1070: axial angle in figure \ref{loopfig5}b.

1071: 

1072: In figure \ref{loopfig5}c, 

1073: we plot the axial-angle-dependent values of the bending and 

1074: twisting free energies for the $Lk = 16$ topoisomer and their sum, which 

1075: is the total elastic-free energy of the loop. The minimum value of the 

1076: total elastic energy occurs at $\tau_{PP} = -58.5^{\circ}$, coincident 

1077: with the position of the $J$-factor maximum for this topoisomer 

1078: (figure \ref{loopfig5}a).

1079: This mechanical state can be achieved with very little twist deformation 

1080: of the loop, but at the expense of significant bending energy.  Further 

1081: reduction of the axial angle requires even less twisting energy; however, 

1082: the bending energy increases monotonically. In contrast, for 

1083: $\tau_{PP} > -58.5^{\circ}$, somewhat less bending energy is required,

1084: but the twisting energy begins to increase significantly with increasing 

1085: axial angle. Since the sense of the bending deformation for 

1086: $\tau_{PP} > 0$ opposes the needed reduction in loop linking number, 

1087: the elastic energy cannot be decreased by increasing the axial angle.

1088: The only way that the loop geometry can compensate for this is through 

1089: twist deformation. This asymmetry arises because we are considering the 

1090: contribution of only one loop topoisomer to the elastic free energy.

1091: 

1092: \subsection{Conclusion}

1093: 

1094: The statistical-mechanical theory for DNA looping discussed above

1095: \cite{loop26,loop51} 

1096: suggests that the helical dependence of DNA looping is 

1097: affected by many factors and leads to the conclusion that whereas a 

1098: positive helical-twist assay can often confirm DNA looping, a negative 

1099: result cannot exclude DNA looping.  Since it is difficult to explore 

1100: the architecture of DNA loops with current experimental techniques, 

1101: this theory will be useful for more reliably analysing DNA looping

1102: with limited experimental data.  The model has advantages over 

1103: previous approaches based exclusively on DNA mechanics, particularly 

1104: when protein flexibility is taken into account. In these cases, entropy 

1105: effects become important and are responsible for the observed decay of 

1106: looping efficiency with DNA length.  

1107: 

1108: 

1109: \section{DNA knots and their consequences: entropy and targeted knot

1110: removal}

1111: 

1112: 

1113: Bacterial DNA occurs largely in circular form. Notably, instead of a simply

1114: connected ring shape (the unknot), the DNA often exhibits permanently

1115: entangled states, such as catenated and knotted DNA. An example for a DNA

1116: trefoil knot is shown in figure \ref{dna_trefoil}. Such configurations have

1117: potentially devastating effects on the cell development. Conversely, however,

1118: knots might have designed purposes in gene regulation, separating different

1119: regions of the genome, or, alternatively, locking chemically

1120: remote parts of the genome proximate in geometrical space. In eukaryotic

1121: cells additional topological effects occur in the likely entanglement of

1122: individual chromosomes. Here, we concentrate on the prokaryotic case.

1123: 

1124: \begin{figure}

1125: \includegraphics[height=5.4cm]{dna_trefoil.ps}

1126: \caption{Electron microscope image of a DNA trefoil knot, from

1127: \cite{wassermann2}. $\copyright$ Science, with permission.

1128: \label{dna_trefoil}}

1129: \end{figure}

1130: 

1131: 

1132: \subsection{Physiological background of knots}

1133: 

1134: The discovery how one can use molecular biological tools to create knotted

1135: DNA resolved a long-standing argument against the

1136: Watson-Crick double helix picture of DNA \cite{frank}, namely that the

1137: replication of DNA could not work as the opening up of the double helix

1138: would produce a superstructure such that the two daughter strands could

1139: not be separated. In fact, the topology of both ssDNA and dsDNA is continuously

1140: changed in vivo, and this can readily be mimicked in vitro, although the

1141: activity of enzymes in vivo is much more restricted than in vitro

1142: \cite{deibler,dna_topo}: Different concentrations of enzymes versus knotted

1143: DNA molecules accessible in vitro,

1144: that is, makes it possible to probe topology-altering effects by enzymes

1145: which in vivo do not contribute to such effects.

1146: 

1147: Although it would be likely

1148: with a probability of roughly $\frac{1}{2}$ that the linear DNA injected by

1149: bacteriophage $\lambda$ into its host \ecoli would create a knot before

1150: cyclisation, it turned out to be difficult to detect \cite{frank}. First studies

1151: therefore concentrated on the fact that under physiological conditions knots

1152: are introduced by enzymes, DNA replication and recombination, DNA repair, and

1153: topoisomerisation, using these enzymes to prove both knotting and unknotting

1154: \cite{liu,mizuuchi,pollock,spengler,wassermann,wassermann1,wassermann2}.

1155: DNA-knotting is also prone to occur behind a stalled replication fork

1156: \cite{viguera,sogo}. Some of the typical topology-altering reactions

1157: undergoing in \ecoli are summarised in figure \ref{topogyro}.

1158: Knots can efficiently be created from nicked\footnote{One of the two strands

1159: is cut.} dsDNA under action of

1160: topoisomerase I at non-physiological concentrations \cite{topoknot}.

1161: Another possibility is by active packaging of a DNA mutant into phage

1162: capsids \cite{arsuaga}, and then denaturing the capsid proteins. Both

1163: methods produce a distribution of different knot types. They can be

1164: separated by electrophoresis \cite{electro}.

1165: 

1166: \begin{figure}

1167: \includegraphics[width=8.6cm]{topo_gyro.eps}

1168: \caption{Enzymes changing the topology of dsDNA by cutting and pasting of one

1169: or both strands (example for \ecoli): (A) Torsional stress resulting from

1170: the Lk deficit causes the DNA double helix to writhe about itself (negative

1171: supercoiling). In \ecoli, gyrase introduces negative supercoils into DNA

1172: and is countered by topoisomerase I (topo I) and topo IV, which relax negative

1173: supercoils. (B) Topo IV unlinks catenanes generated by replication or

1174: recombination in vivo. (C) Topo IV unknots DNA in vivo. After \cite{deibler}.

1175: \label{topogyro}}

1176: \end{figure}

1177: 

1178: The existence of DNA-knots has far-reaching effects on physiological processes,

1179: and knottedness of DNA has therefore to be eliminated in order to

1180: maintain proper functioning of the cell. Among other possible effects, it

1181: is immediately clear that the presence of a knot in a circular DNA impedes

1182: replication of the DNA, i.e., the full separation of the two daughter

1183: strands \cite{alberts,frank}. Moreover, even transcription is impaired

1184: \cite{portugal}. The presence of knots inhibits the assembly of chromatin

1185: \cite{rodriguez}, knotted chromosomes cannot be separated during mitosis

1186: \cite{alberts}, and knots in a chromosome may serve as topological barriers

1187: between different sections of chromosomes, such that the genomic structural

1188: organisation is altered, and certain sections of the chromosomal DNA may no

1189: longer interact \cite{staczek}. Conversely, it is conceivable that knots,

1190: analogously to protein induced DNA looping, lock remote segments of the genome

1191: close together in geometric space. Finally, knots may lead to double-strand

1192: breaks, as they weaken biopolymers considerably due to creation

1193: of localised sharp bends \cite{arai,pieranski,saitta,stasiak1}

1194: as well as macroscopic lines and ropes \cite{mcnally}.\footnote{The weakness

1195: of strings at the site of the knot can be experienced easily by pulling apart

1196: a linear nylon string in comparison to a knotted one \protect\cite{pieranski}.}

1197: 

1198: Above we said that knots can be introduced, inter alia, by the different

1199: enzymes of the topoisomerase family. To remove a knot

1200: from a dsDNA, it is necessary to cut both strands, and then pass one

1201: segment through the created gap, before resealing the two open ends. In

1202: vivo, this is usually achieved by topoisomerases II and IV. A reconstruction

1203: of topo II is shown in figure \ref{topoII}, indicating the upper clamp

1204: holding a segment of the DNA, while the bulge-clamp introduces the cut

1205: through which the upper segment is passed. In the figure, the segment

1206: visible in the pocket of the lower clamp has already been passed through

1207: the gap. After resealing, topo II detaches. This process requires energy,

1208: provided by ATP. Notably, topo II is extremely efficient, for circular dsDNA

1209: of length $\simeq 10$ kbp it was found that topo reduced the knotted state in

1210: between 50 and 100-fold, in comparison to a `dumb' enzyme, which would simply

1211: pass segments through at

1212: random \cite{rybenkov}. We note that the step-wise action of topoisomerase

1213: II was recorded in a single molecule setup using magnetic tweezers

1214: \cite{stasiak,strick}. Topoisomerases are surveyed in the review of

1215: \cite{wang}.

1216: 

1217: \begin{figure}

1218: \includegraphics[width=6.8cm]{topo2_bollen.ps}

1219: \caption{Topoisomerase II. This enzyme can

1220: actively change the topology of DNA by cutting the double-strand and

1221: passing another segment of double-stranded DNA through the gap before

1222: resealing it. The image depicts a short stretch of DNA (horizontally

1223: at the bulge of the enzyme, as well as another segment in the lower

1224: clamp (perpendicular to the image) after passage through the gap from

1225: the upper clamp. This mechanism makes sure that no additional strand

1226: passage through the open gap can take place \cite{berger,wang_to}.

1227: Figure courtesy James M Berger, UC Berkeley.

1228: \label{topoII}}

1229: \end{figure}

1230: 

1231: 

1232: \subsection{Classification of knots}

1233: 

1234: Knottedness can only be defined on a closed (circular) chain. This is

1235: intuitively clear as in an open linear chain a knot can always be tied, or

1236: an existing knot released. Mathematically, this means that knot invariants

1237: are only well-defined for a closed space-curve. However, a linear chain

1238: whose ends are permanently attached to one, or two walls, or whose ends are

1239: extended towards infinity, can be considered as (un)knotted in the proper

1240: mathematical sense, i.e., their knottedness cannot change. In a loser

1241: sense, we will also speak of knots on an open piece of DNA, appealing to

1242: intuition.

1243: 

1244: The classification of knots, or graphs in general, in terms of invariants

1245: can essentially be traced back to Euler, recalling his graph theoretical

1246: elaboration in connection with the Bridges of K{\"o}nigsberg problem

1247: \cite{euler}, determining a closed path by crossing each K{\"o}nigsberg bridge

1248: exactly once. However, the first investigations of topological problems in

1249: modern science is most probably due to Kepler, who studied surface tiling to

1250: great detail (therefore the notion of Kepler tiling in mathematical literature)

1251: \cite{kepler}. Further initial steps

1252: were due to Leibniz, Vandermonde and Gauss, in whose collection of papers

1253: drawings of various knots were found\footnote{Probably copies from an English

1254: original.} whose linking (`Umschlingungen'=windings) number is indeed a knot

1255: invariant \cite{adams,kauffman,reidemeister}. Gauss' student, Listing, in fact

1256: introduced the term `topology', and his work on knots may be viewed as the

1257: real starting point of knot theory \cite{listing}, although his complexions

1258: number was proved by Tait not to be an invariant.

1259: 

1260: Inspired by Helmholtz' theory of an ideal fluid and building on Listing's

1261: early contributions to knot theory, Scotsmen and chums Maxwell, Tait and

1262: Thomson (Lord Kelvin) started to discuss the possible implications of

1263: knottedness in physics and chemistry, ultimately distilled into Thomson's

1264: theory of vortex atoms \cite{thomson,thomson1}. Out of this endeavour emerged

1265: Tait's interest in knots, and he devoted most of his career on the

1266: classification of knots. Numerous charts and still unresolved conjectures

1267: on knots document his pioneering work \cite{tait,tait1,tait2,tait3}. The

1268: studies were carried on by Kirkman and Little

1269: \cite{kirkman,kirkman1,little,little1}.

1270: A more detailed historical account of knot theory may be found in the review

1271: article by van de Griend \cite{degriend}, and on the St. Andrews history of

1272: mathematics webpages\footnote{The MacTutor History of Mathematics

1273: archive, URL: {\tt http://turnbull.mcs.st-and.ac.uk/~history/}}.

1274: 

1275: Planar projections of knots were rendered unique by Listing's introduction

1276: of the handedness of a crossing, i.e., the orientational information assigned

1277: to a point where in the projection two lines intersect. With this information,

1278: projections are the standard representation for knot studies. On their basis,

1279: the minimum

1280: number of crossings (`essential crossings') can be immediately read off as

1281: one of the simplest knot invariants. To arrive at the minimum number, one

1282: makes use of the Reidemeister moves, three fundamental permitted moves of the

1283: \begin{figure}

1284: \includegraphics[width=8cm]{reidemeist.eps}

1285: \caption{The three Reidemeister moves. All topology-preserving moves of a

1286: knot projection can be decomposed into these three fundamental moves.

1287: \label{reidemeister}}

1288: \end{figure}

1289: lines in a knot projection, as shown in figure \ref{reidemeister}. More complex

1290: knot invariants include polynomials of the Alexander, Kauffman and HOMFLY types

1291: \cite{adams,kauffman,reidemeister}.\footnote{These polynomials all start to

1292: be degenerate for higher order knots, i.e., above a certain knot complexity

1293: several knots may correspond to one given polynomial \cite{adams,kauffman}.

1294: In the case of the simpler knots attained in most DNA configurations and in

1295: knot simulations, the Alexander polynomials are unique, in contrast to the

1296: Gauss or Edwards invariant, compare, e.g., reference \protect\cite{volo}.}

1297: Here, we will only employ the number of essential crossings as

1298: classification of knots, in particular, we do not concern ourselves with the

1299: question of degeneracy for a given knot invariant. However, the bookkeeping

1300: of knot types is vital in knot simulations.

1301: 

1302: 

1303: \subsection{Long chains are almost always entangled.}

1304: 

1305: During the polymerisation and final cyclisation of a polymer grown

1306: in a solvent under freely floating conditions, a knot is created

1307: with probability 1. This Frisch-Wassermann-Delbr{\"u}ck conjecture

1308: \cite{frisch,delbrueck} could be mathematically proved for a self-avoiding

1309: chain \cite{sumners,pippenger}, compare also \cite{vanderzande}. This is

1310: consistent with numerical findings

1311: that the probability of unknot formation decreases dramatically with

1312: chain length \cite{frank3,volo}. Indeed, recent simulations results indicate

1313: that the probability of finding the unknot in such a cyclisised polymer

1314: decays exponentially with chain length \cite{koniaris,michels,michels1,janse}:

1315: \begin{equation}

1316: P_\emptyset(N)\propto \exp\left(-\frac{N}{N_c}\right).

1317: \end{equation}

1318: However, there exist theoretical arguments and simulations results indicating

1319: that the characteristic number of monomers $N_c$ occurring in this relation

1320: may become

1321: surprisingly large \cite{frank2,grosberg4,klenin,shimamura1}. The probability

1322: to find a given knot type ${\cal K}$ on random circular polymer formation has

1323: been fitted with the functional form \cite{katritch,dobay,shimamura1}

1324: \begin{equation}

1325: P_{\cal K}(N)=a\big(N-N_0\big)^b\exp\left(-\frac{N^c}{d}\right),

1326: \end{equation}

1327: where $a$, $b$, and $d$ are free parameters depending on ${\cal K}$, and $c

1328: \approx 0.18$. $N_0$ is the minimal number of segments required to form a

1329: knot ${\cal K}$, without the closing segment \cite{dobay}. The tendency

1330: towards knotting during polymer cyclisation creates problems in industrial

1331: and laboratory processes.

1332: 

1333: 

1334: \subsection{Entropic localisation in the figure-eight slip-link structure.}

1335: 

1336: To obtain a feeling for how and when entropy leads to the localisation of

1337: a permanently entangled structure, we consider the simplest polymer object

1338: with non-trivial (non-unknot) geometry, the 

1339: figure-eight structure (F8) displayed in figure \ref{fig8}. In this compound,

1340: a pair contact is enforced by a slip-link, separating off two loops in the

1341: circular polymer, such that none of the loops can fully retract, and both

1342: loops can freely exchange length among each other. We denote the loop sizes

1343: by $n$ and $N-n$, where $N$ is the (conserved) total length of the polymer

1344: chain. For such an object, we can actually perform a closed statistical

1345: mechanical analysis based on results from scaling theory of polymers,

1346: and compare the result with Monte Carlo simulations of the F8.

1347: \begin{figure}

1348: \includegraphics[height=4cm]{figu-8.eps}

1349: \caption{Figure-eight structure, in which a slip-link separates two loops

1350: of size $n$ and $N-n$, such that they can freely exchange length among

1351: each other, but none of the loops can

1352: completely retract from the slip-link. On the right, a schematic drawing

1353: of a slip-link, which may be thought of as a small belt buckle.

1354: \label{fig8}}

1355: \end{figure}

1356: 

1357: The statistical quantities that are of particular interest are the gyration

1358: radius, $R_g$, and the number of degrees of freedom, $\omega$

1359: \cite{degennes}. $R_g$, as defined in equation (\ref{appgyr}),

1360: measures the root mean squared distance of the monomers along the chain to the

1361: gyration centre,

1362: and is therefore a good measure of its extension. It can, for instance, be

1363: measured by light scattering experiment. The degrees of freedom $\omega$ count all possible

1364: different configurations of the chain. For a circular polymer (i.e., a polymer

1365: with $\mathbf{r} (0)=\mathbf{r}(N)$), the gyration radius becomes

1366: \begin{equation}

1367: \label{gyrra}

1368: R_g\simeq AN^{\nu}

1369: \end{equation}

1370: with exponent $\nu=1/2$ for a Gaussian chain, and $\nu=0.588$ in the 3D

1371: excluded volume case ($\nu=3/5$ in the Flory model, and $\nu=3/4$

1372: in 2D). Whereas in 2D this scaling contains truly a ring polymer, in 3D the

1373: exponent $\nu$ emerges from averaging over all possible topologies, and

1374: necessarily includes knots of all types \cite{duplantier1,deutsch,degennes}.

1375: For a circular chain, the number of degrees of freedom contains the number of all possible

1376: ways to place an $N$-step walk on the lattice with connectivity $\mu$

1377: (e.g., $\mu=2d$ on a cubic lattice in $d$ dimensions),

1378: $\mu^N$, and the entropy loss for requiring a closed loop, $N^{-d\nu}$,

1379: involving the same Flory exponent $\nu$. For the Gaussian case, we recognise

1380: in this entropy loss factor the returning probability of the random walk. In

1381: the excluded volume case, $N^{-d\nu}$ is an analogous measure

1382: \cite{degennes,grosberg,hughes}. Thus, for a circular chain embedded in

1383: $d$-dimensional space, the number of degrees of freedom is

1384: \begin{equation}

1385: \label{dof}

1386: \omega\simeq\mu^NN^{-d\nu}.

1387: \end{equation}

1388: Let us evaluate these measures for the F8 from figure \ref{fig8}.

1389: 

1390: As a first approximation, consider the F8 as a Gaussian (phantom) random walk,

1391: demonstrating that, like in the charged knot case \cite{paul}, entropic effects

1392: give rise to long-range interactions. The two loops correspond to returning

1393: random walks, i.e., the number of degrees of freedom for the F8 in the phantom chain case

1394: becomes \cite{degennes,flory1,grosberg}\footnote{Here and in the following we

1395: consider two configurations of a polymer chain different if they cannot be

1396: matched by translation. In addition, the origin of a given structure is fixed

1397: by a vertex point (see below), i.e., a point where several legs of the polymer

1398: chain are joint. In the F8-structure, this vertex naturally coincides with the

1399: slip-link. For a simply connected ring polymer, such a vertex is a two-vertex

1400: anywhere along the chain. \label{counting}}

1401: \begin{equation}

1402: \label{fig8id}

1403: \omega_{\rm F8,PC}\simeq\mu^Nn^{-d/2}(N-n)^{-d/2},

1404: \end{equation}

1405: where $d$ is the embedding dimension. We note that normalisation of this

1406: expression produces the probability density function for finding

1407: the F8 with a given loop size $\ell=na$ ($L=Na$),

1408: \begin{equation}

1409: \label{f8_pdf}

1410: p_{\rm F8,PC}(\ell)\simeq\mathscr{N}\ell^{-d/2}(L-\ell)^{-d/2},

1411: \end{equation}

1412: where $\mathscr{N}$ denotes a normalisation factor.

1413: The conversion from expressing the chain size in

1414: terms of the number of monomers to its actual length is of advantage in what

1415: follows, as it allows to more easily keep track of dimensions. Here, we use

1416: the length unit $a$, which may be interpreted as the monomer size (lattice

1417: constant), or as the size of a Kuhn statistical segment.

1418: 

1419: \begin{figure}

1420: \includegraphics[width=6.8cm]{sl2D_equilibrium.eps}

1421: \caption{Bead-and-tether chain used in Monte Carlo simulation, showing a

1422: typical equilibrium configuration for a self-avoiding chain: the localisation

1423: of the smaller loop is distinct. Note that in this 2D simulation the slip-link

1424: is represented by the three tethered black beads.

1425: \label{sl}}

1426: \end{figure}

1427: 

1428: To classify different grades of localisation, we follow the convention from

1429: references \cite{slili2d,paraknot}.

1430: The average loop size $\langle \ell\rangle$ determined through

1431: $\langle \ell \rangle = \int_{a}^{L-a} \ell p(\ell) d\ell$

1432: is trivially $\langle\ell\rangle=L/2$ by symmetry of

1433: the structure. Here, we introduce a short-distance cutoff set by the lattice

1434: constant $a$. However, as the probability density function

1435: is strongly peaked at $\ell=0$ and $\ell=L$,

1436: the two poles caused by the returning probabilities, and therefore a {\em

1437: typical\/} shape consists of one small ({\em tight\/}) and one large ({\em

1438: loose\/}) loop, compare figure \ref{sl}. This can be quantified in terms of

1439: the average size of the smaller loop,

1440: \begin{equation}

1441: \label{sls_f8}

1442: \langle \ell \rangle_<\equiv 2\int_a^{L/2} \ell p(\ell) d\ell \, .

1443: \end{equation}

1444: In $d=2$, we obtain

1445: \begin{equation}

1446: \langle \ell \rangle_<\sim \frac{L}{|\log(a/L)|},

1447: \end{equation}

1448: such that with the logarithmic correction the smaller loop is only marginally

1449: smaller than the big one. In contrast, one observes {\em weak localisation}

1450: \begin{equation}

1451: \label{fig8d3}

1452: \langle \ell \rangle_<\sim a^{1/2}L^{1/2}

1453: \end{equation}

1454: in $d=3$, in the sense that the relative size $\langle \ell \rangle_</L$

1455: tends to zero for large chains. By comparison, for $d>4$ one encounters

1456: $\langle \ell \rangle_<\sim a$, corresponding to {\em strong localisation},

1457: as the size of the smaller loop does not depend on $L$ but is set by the

1458: short-distance cutoff $a$. Above four dimensions, excluded volume effects

1459: become negligible, and therefore both Gaussian and self-avoiding chains are

1460: strongly localised in $d\ge 4$.\footnote{Consideration of higher than the

1461: physical 3 space dimensions is often useful in polymer physics.}

1462: 

1463: To include self-avoiding interactions, we make use of results for general

1464: polymer networks obtained by Duplantier \cite{duplantier,duplantier1}, which

1465: are summarised in the appendix at the end of this review. In terms of such

1466: networks, our F8-structure corresponds to the following parameters: the

1467: number ${\cal N}=2$ of polymer segments with lengths $s_1=\ell=na$ and $s_2=

1468: L-\ell=(N-n)a$, forming ${\cal L}=2$ physical loops, connected by $n_4=1$

1469: vertex of order four.

1470: By virtue of equation (\ref{network}), the number of configurations

1471: of the F8 with fixed $\ell$ follows the scaling form

1472: \begin{equation}

1473: \label{t_scaling}

1474: \omega_{\rm F8}(\ell)\simeq \mu^L (L-\ell)^{\gamma_{\rm F8}-1}

1475: {\cal X}_{\rm F8}\left(\frac{\ell}{L-\ell}\right),

1476: \end{equation}

1477: with the configuration exponent $\gamma_{\rm F8}=1-2d\nu+\sigma_4$. In the

1478: limit $\ell\ll L$, the contribution of the large loop

1479: in equation (\ref{t_scaling})

1480: should not be affected by a small appendix, and therefore should exhibit the

1481: regular Flory scaling $\sim (L-\ell)^{-d\nu}$ \cite{kafri,kafri1,duplantier3}.

1482: This fixes the scaling behaviour of the scaling function ${\cal X}_{\rm F8}

1483: (x)\sim x^{\gamma_{\rm F8}-1+d\nu}$ in this limit ($x \to 0$ in dimensionless

1484: variable $x$), such that

1485: \begin{equation}

1486: \label{sl_limit}

1487: \omega_{\rm F8}(\ell)\simeq\mu^L(L-\ell)^{-d\nu}\ell^{-c},

1488: \quad \ell\ll L,

1489: \end{equation}

1490: where $c=-(\gamma_{\rm F8}-1+d\nu)=d\nu-\sigma_4$. Using $\sigma_4=-19/16$

1491: and $\nu = 3/4$ in $d=2$ \cite{duplantier,duplantier1}, we obtain

1492: \begin{equation}

1493: \label{expon}

1494: c=43/16=2.6875, \quad d=2.

1495: \end{equation}

1496: In $d = 3$, $\sigma_4\approx -0.48$ \cite{schaefer,kafri,kafri1} and

1497: $\nu\approx 0.588$, so that

1498: \begin{equation}

1499: c\approx 2.24.

1500: \end{equation}

1501: In both cases the result $c>2$ enforces that the loop of length

1502: $\ell$ is strongly localised in the sense defined above.

1503: This result is self-consistent with the a priori assumption $\ell\ll L$.

1504: Note that for self-avoiding chains, in $d=2$ the

1505: localisation is even {\em stronger\/} than in $d=3$, in contrast

1506: to the corresponding trend for ideal chains.

1507: 

1508: We performed Monte Carlo (MC) simulations of the 2D figure-eight structure,

1509: in which the slip-link was represented by three tethered

1510: beads enforcing a sliding pair contact such that the loops cannot fully

1511: retract (see figure \ref{fig8_sim}). We used a 2D hard core bead-and-tether

1512: chain with 512 monomers, starting off from a symmetric initial condition

1513: with $\ell=L/2$.

1514: Self-crossings were prevented by keeping a maximum bead-to-bead

1515: distance of 1.38 times the bead diameter, and a maximum step length of 0.15

1516: times the bead diameter.

1517: As shown in figure \ref{sl1}, the size distribution

1518: for the small loop can be fitted to a power law with exponent $c=2.68$

1519: in good agreement with equation (\ref{expon}).

1520: 

1521: \begin{figure}

1522: \includegraphics[width=6.8cm]{loops.eps}

1523: \caption{Monte Carlo simulation of an F8-structure in 2D: loop

1524: sizes $\ell$ and $L-\ell$ as a function of Monte Carlo steps for a chain

1525: with 512 monomers. The symmetry breaking after the symmetric initial

1526: condition is distinct.

1527: \label{fig8_sim}}

1528: \end{figure}

1529: 

1530: \begin{figure}

1531: \includegraphics[width=6.8cm]{pdf_slN512.eps}

1532: \caption{Power-law fit to the probability density function of the smaller

1533: loop. The fit

1534: produces a slope of -2.68, in excellent agreement with the calculated value.

1535: \label{sl1}}

1536: \end{figure}

1537: 

1538: An experimental study of entropic tightening of a macroscopic F8-structure

1539: was reported in reference \cite{bennaim}. There, a granular chain consisting

1540: of hollow steel spheres connected by steel rods was once twisted and then

1541: put on a vibrating table. From digital imaging, the distribution of loop

1542: sizes could be determined and compared to a power-law with index 43/16

1543: as calculated for the 2D excluded volume chain. The agreement was found to

1544: be consistent \cite{bennaim}.

1545: 

1546: 

1547: \subsection{Simulations of entropic knots in 2D and 3D.}

1548: 

1549: Much of our knowledge about the interaction of knots with thermal fluctuations

1550: is based on simulations of knotted chains. Before going further into the

1551: theoretical modelling of knotted chains, we report some of the results based

1552: on simulations studies of both Gaussian and self-avoiding

1553: walks.\footnote{Although per se a Gaussian chain cannot have a fixed topology

1554: due to its phantom character, such simulations introduce a fixed topology

1555: by rejecting moves that result in a different knot type.}

1556: Such simulations

1557: either start with a given knot configuration and then perform moves of

1558: specific segments, each time making sure that the topology is preserved;

1559: or, each new configuration emanates from a new

1560: random walk, whose correct topology may be checked by calculating the

1561: corresponding knot invariant, usually the Alexander polynomial, and

1562: created configurations that do not match the desired topology are discarded.

1563: We note that it is of lesser significance that knot invariants such as the

1564: Alexander polynomials in fact are no longer unique for more complex knots,

1565: because for typical chain lengths with the highest probability simpler knots

1566: are created, for which

1567: the invariants are unique. For more details we refer to the works

1568: quoted below.

1569: 

1570: In fact, the fixed topology turns out to have a highly non-trivial effect

1571: on chains without self-excluded volume. As conjectured in \cite{descloizeaux},

1572: a Gaussian circular chain, whose permitted set of configurations is restricted

1573: to a fixed topology, will exhibit self-avoiding behaviour. This was proved in a

1574: numerical analysis in \cite{deutsch}. The required number of monomers to

1575: reach this self-avoiding exponent was estimated to be of the order of 500.

1576: Keeping this non-trivial scaling of a Gaussian chain at fixed topology in mind,

1577: knot simulations on the basis of phantom Gaussian chains were performed in

1578: \cite{dobay1}, always making sure that the configurations taken into the

1579: statistics fulfil the desired knot topology.

1580: 

1581: The dependence of the gyration radius $R_g$ on the knot type was investigated

1582: for simpler knots in 3D in reference \cite{janse}. On the basis of the

1583: expansion

1584: \begin{equation}

1585: R_g^2\simeq A_{\cal K}\Big(1+B_{\cal K}N^{-\Delta}+C_{\cal K}N^{-1}+o(1/N)\Big)

1586: N^{2\nu_{\cal K}},

1587: \end{equation}

1588: including a confluent correction \cite{janse,leguillou,zinn} in comparison to

1589: the standard expression (\ref{gyrra}), it was found that the Flory exponent

1590: $\nu_{\cal K}$ is independent of the knot type ${\cal K}$ and has the

1591: 3D value $0.588$. This was interpreted via a {\em localisation\/} of knots

1592: such that the influence of tight knots on $R_g$ is vanishingly small. In fact,

1593: $\Delta$ is

1594: of the order of $0.5$ according to the investigations in references

1595: \cite{leguillou,zinn,janse1,li}.

1596: Based on longer chains in comparison to reference

1597: \cite{janse}, the study of \cite{orlandini} thus corroborates the

1598: independence of $\nu_{\cal K}\approx 0.588$ of the knot type ${\cal K}$.

1599: In recent AFM experiments analysing single DNA knots, the Flory scaling

1600: $R_g\simeq N^{\nu}$ was confirmed for both simple and complex knots

1601: \cite{valle}.

1602: 

1603: For the number of degrees of freedom $\omega_{\cal K}$, it was found for the

1604: form\footnote{Note that

1605: we changed the exponent by 1 in comparison to the original work, making the

1606: counting of non-translatable configurations consistent with the counting

1607: convention specified in footnote \protect\ref{counting}.}

1608: \begin{equation}

1609: \label{orland}

1610: \omega_{\cal K}\simeq A_{\cal K}N^{\alpha_{\cal K}-2}\mu_{\cal K}^N

1611: \left(1+\frac{B_{\cal K}}{N^{\Delta_{\cal K}}}+\ldots\right)

1612: \end{equation}

1613: with confluent corrections, that while for the unknot with $\alpha_{\emptyset}

1614: \approx 0.27$ expression (\ref{orland}) is consistent with the standard

1615: result (\ref{dof}) ($[0.27-2]/3\approx -0.58\approx \nu$), for prime knots

1616: $\alpha_{\cal K}=\alpha_{\emptyset}+1$, and for composite knots with $N_f$

1617: prime components,

1618: \begin{equation}

1619: \alpha_{\cal K}=\alpha_{\emptyset}+N_f.

1620: \end{equation}

1621: This finding is in agreement

1622: with the view that each prime component of a knot ${\cal K}$

1623: is tightly localised and statistically able to move around one central loop,

1624: each prime component counting an additional factor $N$ of degrees of freedom. The fact that

1625: for a chain of finite thickness the size of the big central loop is in fact

1626: diminished by the size of the tight knot is a confluent effect, such that the

1627: confluent exponent $\Delta$ should be related to the size distribution of

1628: the knot region.

1629: Not surprisingly, the connectivity factor $\mu_{\cal K}\approx 4.68$ was found

1630: to be independent of ${\cal K}$, assuming the standard value for a cubic

1631: lattice \cite{guttmann}. Also the amplitude $A_{\cal K}$ and the exponent

1632: $\Delta_{\cal K}$ of the confluent correction turned out to be ${\cal

1633: K}$-independent. We note that a similar analysis in (pseudo) 2D\footnote{The

1634: simulated polymer chain moves in 2D, however, crossings are permitted at

1635: which one chain passes underneath another. In that, the simulated polymers

1636: are in fact equivalent to knot projections with a certain orientations of

1637: individual crossings.} also strongly points towards tight localisation of

1638: the knot \cite{guitter}.

1639: 

1640: In contrast to the above results, 3D simulations undertaken in \cite{quake}

1641: (also compare \cite{quake1}) show the dependence

1642: \begin{equation}

1643: \label{quake}

1644: R_g\simeq N^{3/5}C^{-4/15}

1645: \end{equation}

1646: of the gyration radius on the knot type, characterised by the number $C$ of

1647: essential crossings. $R_g$, that is, decreases as a power-law with $C$,

1648: where the exponent $-4/15=1/3-\nu$ \cite{quake}. The functional form

1649: (\ref{quake})) was derived from a Flory-type argument for a polymer construct

1650: of $C$ interlocked loops of equal length $N/C$ by arguing that each loop

1651: occupies a volume $\simeq (N/C)^{3\nu}$, and the volume of the knot is given

1652: by $V\simeq C(N/C)^{3\nu}$ (i.e., assuming that due to self-avoiding repulsion

1653: the volume of individual loops adds up to the total volume). Equation

1654: (\ref{quake}) then follows immediately. This model of equal loop sizes is

1655: equivalent to a completely delocalised knot. It may therefore be speculated,

1656: albeit rather long chain sizes of up to 400 were used, whether the numerical algorithm

1657: employed for the simulations in \cite{quake} causes finite-size effects that,

1658: in turn, prevent a knot localisation. We note that the Flory-type

1659: scaling assumed to derive expression (\ref{quake}) is consistent with a

1660: modelling brought forward in reference \cite{grosberg2}, in which the knot is

1661: quantified by the aspect ratio in a configuration corresponding to a maximally

1662: inflated tube with the given topology (i.e., a state corresponding to

1663: complete delocalisation). In reference \cite{quake}, the temporal relaxation

1664: behaviour of a given knot was also studied. While regular

1665: Rouse behaviour was found for the case of the unknot, the knotted chains

1666: displayed somewhat surprising long time contributions to the relaxation

1667: time spectrum \cite{treloar,ward,ferry,erman}, a phenomenon

1668: already pointed out by de Gennes within an

1669: activation argument to create free volume in a tight knot in order to move

1670: along the chain \cite{degennesma}. Note that relatively lose knots in shorter

1671: chains do not appear to exhibit such extremely long relaxation time behaviour

1672: \cite{lai1}.

1673: 

1674: Simulation of a 3D knot with varying excluded volume showed, if only the

1675: excluded volume becomes large enough, the gyration radius of the knot is

1676: independent of the knot type \cite{shimamura}. The picture of tight knots

1677: is further corroborated in the study by Katritch {\it et al.} using a Gaussian

1678: chain model with fixed topology to demonstrate that the size distribution

1679: of the knot is distinctly peaked at rather small sizes \cite{katritch}.

1680: 

1681: Apart from determining the statistical quantities $R_g$ and $\omega_{\cal K}$

1682: from simulations, there also exist indirect methods for quantifying the size

1683: of the knot region in a knotted polymer. One such method is to confine an

1684: open chain containing a knot between two walls, and measuring the finite

1685: size corrections of the force-extension curve due to the knot size. This is

1686: based on the idea that the gyration radius for a system depending on more

1687: than one length scale (i.e., apart from the chain length $N$) shows above

1688: mentioned confluent corrections, such that \cite{farago}

1689: \begin{equation}

1690: R_g=AN^{\nu}\Phi\left(\frac{N_0}{N},\frac{N_1}{N},\ldots\right)\simeq

1691: AN^{\nu}\Big(1-BN^{-\Delta}\Big)

1692: \end{equation}

1693: when only the largest correction is considered, and in 3D $\Delta\approx 0.5$

1694: is supposed to be universal \cite{leguillou,zinn,janse1,li}.

1695: If this leading correction is due to the argument $N_0/N$ in the scaling

1696: function $\Phi$, the length scale $N_0$ depends on $N$ through the scaling

1697: $N_0\sim N^t$ with $\Delta=1-t$. From Monte Carlo simulations of a bead and

1698: tether chain model, it could then

1699: be inferred that the size of the knot scales like \cite{farago}

1700: \begin{equation}

1701: N_k\sim N^t, \quad t=0.4\pm0.1

1702: \end{equation}

1703: This, in turn, enters the force-extension curve $f'=G(R')$ with the

1704: dimensionless force $f'=fAN^{\nu}/(k_BT)$ and distance $R'=R/(AN^{\nu})$ of

1705: the walls, in the form with confluent correction

1706: \begin{equation}

1707: f'\simeq G(R')\Big(1+g(R')N^{-\Delta}\Big).

1708: \end{equation}

1709: From the simulation, $t=0.4$ corresponds to the best data collapsing, assuming

1710: the validity of

1711: the scaling arguments. An argument in favour of this approach is the

1712: consistency of the exponent $t=0.4$ with the inferred $\Delta=0.6$, which is

1713: close to the known value. Note that the force-extension of a chain with a

1714: slip-link was discussed in reference \cite{pull} and shown that a loop

1715: separated off by a slip-link is confined within a Pincus-de Gennes blob.

1716: We also note that results

1717: corresponding to delocalisation in force-size relations were reported in

1718: \cite{sheng,sheng1}. An entropic scale was conceived in \cite{roya}: Separating

1719: two chains with fixed topology but allowing them to exchange length (e.g.,

1720: through a small hole in a wall) would enable one to infer the localisation

1721: behaviour of a knot by comparing the equilibrium balance of this knot with

1722: a slip-link construct of known degrees of freedom until the

1723: average length on both sides coincides. The preliminary results in \cite{roya}

1724: are shadowed by finite-size effects of the accessible system size,

1725: as limited by computation power. The analysis in reference \cite{marcone} of a

1726: self-avoiding polygon model uses the

1727: method of closure of a short fragment of the knot and subsequent determination

1728: of its Alexander polynomial to obtain the scaling exponent $t=0.75$; in a

1729: second variant, the authors find a consistent result by a variant of the

1730: knot scale method. Another recent study uses a more realistic model for a

1731: polymer chain, namely, a simplified model of polyethylene; with up to 1000

1732: monomers in the simulation, the exponent $t\approx0.65$ is found (and

1733: delocalisation is obtained in the dense phase) \cite{virnau}.

1734: 

1735: Thus, there exist simulations results pointing in both directions, knot

1736: localisation and delocalisation. As the latter may be explained by finite

1737: size effects, it seems likely that (at least simple) knots in 2D and 3D

1738: localise in the sense that the knot region occupies a portion of the chain

1739: that is significantly smaller in comparison to the entire chain. In particular,

1740: this would imply that the average size of the knot region $\langle\ell\rangle$

1741: scales with the chain length $Na$ with an exponent less than one, such that

1742: \begin{equation}

1743: \lim_{N\to\infty}\frac{\langle\ell\rangle}{Na}=0.

1744: \end{equation}

1745: Below, we show from analytical grounds that such a localisation is a natural

1746: consequence of interactions of a chain of fixed topology with fluctuations.

1747: We note, however, that conclusive results for knot localisation may in fact

1748: come from experiments: Manipulation of single chains such as DNA can be

1749: performed for rather long chains, making it possible to reach beyond the

1750: finite-size corrections inherent in, e.g., the force-extension simulations

1751: mentioned above. The aforementioned AFM studies on single DNA knots indeed

1752: reveal knot localisation of flattened knots \cite{valle}; due to experimental

1753: limitations, presently only one DNA length was investigated, such that the

1754: scaling exponent $t$ currently cannot be obtained.

1755: 

1756: Before proceeding to these analytical approaches, we note that there have

1757: also been performed simulations of knotted chains under non-dilute conditions

1758: \cite{stella,orlanddense}.

1759: In (pseudo) 2D, these have found delocalisation of the knot, i.e., $\lim_{

1760: N\to\infty}\langle\ell\rangle/N=const$. We come back to these

1761: simulations below in connection with the modelling of dense and $\Theta$-knots.

1762: 

1763: 

1764: \subsection{Flattened knots in dilute and dense phases.}

1765: 

1766: Analytically, knots are a hard problem to tackle. Statistical mechanical

1767: treatments of permanently entangled polymers are so difficult to treat

1768: since topological restrictions cannot be formulated as a Hamiltonian problem

1769: but appear as hard constraints partitioning the phase space

1770: \cite{degennes,grosberg,vilgis,kholodenko}.\footnote{For comparison,

1771: self-avoidance in 3D is usually treated as a perturbation, i.e., as a ``soft

1772: constraint'', in analytical studies \protect\cite{degennes}.} A segment of a 3D

1773: knot, in other words, can move without feeling the constraints due to the

1774: non-trivial topology of a knotted state, until it actually collides with

1775: another segment. The accessible phase space of degrees of freedom is therefore characterised

1776: by inequalities.\footnote{Although a similar statement is true for polymer

1777: networks in 3D, the field theoretical results for their critical exponents are

1778: in fact obtained as {\em averages\/} over {\em all\/} topologies. For

1779: instance, the exponent $\nu$ entering the gyration radius of a a 3D polymer

1780: ring counts all knotted states \protect\cite{duplantier1}.}

1781: 

1782: Consequently, only a relatively small range of problems have

1783: been treated analytically, starting with the seminal papers by Edwards

1784: \cite{edwards1,edwards2}, in which he considers the classification of

1785: topological constraints in polymer physics. De Gennes addressed the problem

1786: of tight knot motion along a polymer chain using scaling arguments for the

1787: activation of free length inside the knot region, producing a double-exponential

1788: expression for the corresponding time scale \cite{degennesma}, which might

1789: explain the extreme long-time contributions in the relaxation time

1790: spectrum of permanently entangled polymers \cite{doi,treloar,ward,erman,ferry}.

1791: Some analytical results were obtained for a pair, or an `Olympic' gel of

1792: entangled polymer rings, see for instance,

1793: \cite{otto,otto1,vilgis1,ferrari,ferrari1}. In a mean field approach based on

1794: the Kauffman invariant the entropy of knots was investigated in references

1795: \cite{grone,grone1,grone2}. Similarly, some statistical properties of random

1796: knot diagrams were investigated in \cite{nechaev,nechaev1}. However, some

1797: insight

1798: can be gained on the basis of phenomenological models, which we will come back

1799: to below. Here, we continue with an analytical study of flat knots.

1800: 

1801: One possibility to treat knotted polymer chains analytically is to confine

1802: the degrees of freedom of the knot to motion in 2D, only. The knot, that is, is

1803: preserved, as at the crossings the chain is allowed to form an

1804: over-/underpassing, while the rest of the knot is confined to 2D. Such a

1805: confinement can in fact be experimentally realised in various ways. Thus,

1806: the chain can be confined between two close-by glass slabs, as demonstrated

1807: in \cite{craighead}; it can be pressed flat on a surface by gravitation or

1808: similar forces, for instance in macroscopic systems \cite{bennaim,bennaim1};

1809: the chain can be adhesively bound to a membrane and still reach

1810: configurational equilibrium, as experimentally shown for DNA in references

1811: \cite{valle,maier}. Or it can be adsorbed to a mica surface either by APTES

1812: coating or by providing bivalent Mg ions in solution, as shown in

1813: figure \ref{valle}. From such flat knots as discussed in

1814: the remainder of this section, we will

1815: be able to infer certain generic features also for 3D knots.

1816: 

1817: \begin{figure}

1818: \begin{center}

1819: \includegraphics{knot42.ps}

1820: \end{center}

1821: \caption{AFM tapping-mode images of flattened, complex DNA-knots with

1822: approximately 30-40 essential crossings, see \protect\cite{valle}.

1823: The substrate surface used is AP-mica (freshly cleaved mica

1824: reacted with an amino terminal silane to obtain a positively charged surface).

1825: The DNA knots used are extracted from bacteriophage P4; the DNA is a 11.4

1826: kbp molecule (with a 1.4 kbp deletion resulting in a final length of 10 kbp)

1827: which has two cohesive ends. They are not covalently closed, thus no

1828: supercoiling is present. The knot adsorbed out of the 3D bulk

1829: on to the surface is strongly trapped, i.e., the knot is `projected' onto the

1830: surface without any equilibration. The knot appears rather delocalised.

1831: Courtesy F. Valle and G. Dietler.

1832: \label{valle}}

1833: \end{figure}

1834: 

1835: A flat knot therefore corresponds to a polymer network in 2D, but the

1836: orientation of the crossings is preserved, such that the network graph

1837: actually coincides with a typical knot projection

1838: \cite{kauffman,reidemeister,adams}, as shown in figure \ref{tref_proj}

1839: on the left. This projection of the trefoil, and similar projections for

1840: all knots, displays the knot with the essential crossings. A flat knot can,

1841: in principle acquire an arbitrary number of crossings by Reidemeister moves;

1842: for instance, the bottom left segment of the flat trefoil can slide under

1843: the vicinal segment, creating a new pair of vertices, and so on. However,

1844: we suppose that such transient additional loops are sufficiently short-lived

1845: so that we can neglect them in our analysis. Then, we can apply results from

1846: scaling analysis of polymer networks of the most general type shown in

1847: figure \ref{network}, see the primer in the appendix. We note that from the

1848: Monte Carlo simulations we performed it may be concluded that such additional

1849: vertices can in fact be neglected.

1850: 

1851: 

1852: \subsubsection{Flat knots in dilute phase.}

1853: 

1854: We had previously found that for the F8-structure the probability density

1855: function for the size of

1856: each loop is peaked at $\ell\to 0$ and $\ell\to L$.

1857: From the scaling analysis for

1858: self-avoiding polymer networks, we concluded strong localisation of one

1859: subloop. For more complicated structures, the joint probability to find

1860: the individual segments with given lengths $s_i$ is expected to peak at

1861: the edges of the higher-dimensional configuration hyperspace. Some analysis

1862: is necessary to find the characteristic shapes. Let us consider here the

1863: simplest non-trivial knot, the (flat) trefoil knot $3_1$ shown in figure

1864: \ref{tref_proj}.

1865: \begin{figure}

1866: \begin{center}

1867: \includegraphics[width=7.2cm]{tref_proj.eps}

1868: \end{center}

1869: \caption{Flat trefoil knot with segment labels. On the right, a schematic

1870: representation of a localised flat trefoil with one large segment is

1871: shown.

1872: \label{tref_proj}}

1873: \end{figure}

1874: Each of the three crossings is replaced with a vertex with four

1875: outgoing legs, and the resulting network is assumed to separate into

1876: a large loop and a multiply connected region which includes the vertices.

1877: Let $\ell=\sum_{i=1}^5 s_i$ be the total length of all segments contained

1878: in the multiply connected knot region. Accordingly, the length of the large

1879: loop is $s_6=L-\ell$.

1880: 

1881: In the limit $\ell\ll L$, the number of configurations of this network can be

1882: derived in a similar way as in the scaling approach followed for the F8. This

1883: procedure determines the concrete behaviour of the scaling form

1884: \begin{equation}

1885: \omega_{\rm III}\simeq\mu^L{\cal W}_{\rm III}\left(L-\ell,\ell,

1886: \frac{s_1}{\ell},\frac{s_2}{\ell},\frac{s_3}{\ell},\frac{s_4}{\ell}\right)

1887: \end{equation}

1888: including the scaling function ${\cal W}$ that depends on altogether six

1889: arguments. The index III is chosen according to figure \ref{knot_zoo},

1890: where the flat trefoil configuration in the dense phase appears at

1891: position III of the scheme (explained below). After

1892: some manipulations, the number of degrees of freedom yields in the form \cite{slili2d}

1893: \begin{equation}

1894: \label{t_limit}

1895: \omega_{\text{III}}(\ell,L)\sim\mu^L (L-\ell)^{-d \nu}\ell^{-c},

1896: \end{equation}

1897: with the scaling exponent

1898: \begin{equation}

1899: c=-(\gamma_{\text{III}}-1+d\nu)-m, \qquad m=4.

1900: \end{equation}

1901: Here, $m=4$ corresponds to the number of independent integrations over the

1902: segments $s_i$ ($i=1,\ldots,4$) of the knot region, as we only retain the

1903: cumulative size $\ell=\sum_{i=1}^5s_i$ of the knot region. Putting numerical

1904: values, we find $c=65/16$, i.e., {\em strong localisation}.

1905: 

1906: \begin{figure*}

1907: \begin{center}

1908: \includegraphics{comschem.eps}

1909: \end{center}

1910: \caption{Hierarchy of the flat trefoil knot $3_1$. Upper row: dilute phase.

1911: Middle row: $\Theta$-phase. Bottom row: dense phase. To the left of each

1912: row, the trefoil projection is shown. It splits up into the hierarchies of

1913: configurations, with exponents $c$ below each contraction. The small protruding

1914: legs represent the big central loop, compare, for instance, figure

1915: \protect\ref{tref_proj} on the right with position III of the top row. See

1916: text for details.

1917: \label{knot_zoo}}

1918: \end{figure*}

1919: 

1920: However, some care is necessary in performing these integrations,

1921: since the scaling function ${\cal W}_{\rm III}$ may

1922: exhibit non-integrable singularities if one or more of the arguments

1923: $s_i/\ell$ tend to $0$. The geometries corresponding to these

1924: limits (edges of the configuration hyperspace)

1925: represent {\em contractions\/} of the original

1926: trefoil network ${\cal G}_{\text{III}}$ in the sense that the length of

1927: one or more of the segments $s_i$ is of the order of the

1928: short-distance cutoff $a$. If such a short segment connects

1929: different vertices, they cannot be resolved on larger length

1930: scales, but appear as a single, new vertex. Thus, each

1931: contraction corresponds to a different network ${\cal G}$,

1932: which may contain a vertex with up to eight outgoing legs. For

1933: the flat trefoil knot, there exist six different contractions,

1934: as grouped in figure \ref{knot_zoo} around the original flat

1935: trefoil at position III. As an example, in the top row of figure \ref{knot_zoo}

1936: contraction VI follows from the original trefoil III if the uppermost segment

1937: becomes very small, and similarly the network VII emanates from contraction VI

1938: if one of the four symmetric segments becomes very small.

1939: For each of these networks, one can calculate the corresponding

1940: exponent $c$ in a similar way as above, leading to the general

1941: expression

1942: \begin{equation}

1943: \label{euler}

1944: c=2+\sum_{N\ge 4}n_N\left\{\frac{N}{2}\Big(d\nu-1\Big)+

1945: \Big(|\sigma_N|-d\nu\Big)\right\}.

1946: \end{equation}

1947: The $\sigma_N$ are given in equation (\ref{top_coeff}). In figure

1948: \ref{knot_zoo}, the various contractions are arranged in increasing

1949: exponent $c$.

1950: 

1951: Our scaling analysis relies on an expansion in

1952: $a/\ell\ll 1$, and the values of $c$ determine a sequence of

1953: contractions according to higher orders in $a/\ell$: The {\em smallest\/}

1954: value of $c$ corresponds to the most likely contraction, while the

1955: others represent corrections to this leading scaling behaviour, and

1956: are thus less and less probable (see figure \ref{knot_zoo}).

1957: To lowest order, the trefoil behaves like a large ring polymer at whose fringe

1958: the point-like knot region is located. At the next level of resolution,

1959: it appears contracted to the figure-eight shape ${\cal G}_\mathrm{I}$.

1960: For more accurate data, the higher order

1961: shapes II to VII may be found with decreasing

1962: probability. Interestingly, the original uncontracted trefoil

1963: configuration ranks third in the hierarchy of shapes.

1964: 

1965: These predictions were checked by MC simulations with the same conditions as

1966: described above, to prevent intersection. The flat trefoil knot was prepared

1967: from a symmetric, harmonic 3D representation with 512 monomers, which was

1968: collapsed and then kept on a hard wall by the

1969: ``gravitational'' field $V=-k_BTh/h^*$ perpendicular to the wall, where $h$ is

1970: the height of a monomer, and $h^*$ was set to 0.3 times the bead diameter.

1971: Configurations corresponding to contraction I

1972: are then selected by requiring that besides a large loop, they contain

1973: only one segment larger than a preset cutoff length (taken to be 5 monomers),

1974: and similarly for contraction II.

1975: The size distributions for such contractions, as well as for all

1976: possible knot shapes are shown in figure \ref{figsl1}.

1977: \begin{figure}

1978: \begin{center}

1979: \includegraphics[width=8cm]{trefoil_pdf.eps}

1980: \end{center}

1981: \caption{Power law tails in probability density functions

1982: for the size $\ell$ of tight segments:

1983: As defined in the figure, we show results for the smaller loop in a

1984: figure-eight structure, the overall size of the trefoil knot,

1985: as well as the two leading contractions of the latter.

1986: \label{figsl1}}

1987: \end{figure}

1988: The tails of the distributions are indeed consistent with the predicted

1989: power laws, although the data (especially for contraction II) is too

1990: noisy for a definitive statement.

1991: 

1992: Our scaling results pertain to all flat prime knots. In particular, the

1993: dominating contribution for {\em any} prime knot corresponds to the

1994: figure-eight contraction ${\cal G}_{\text{I}}$, as equation (\ref{euler}) predicts

1995: a larger value of the scaling exponent $c$ for any network ${\cal G}$ other

1996: than ${\cal G}_{\text{I}}$. Accordingly, figure \ref{fig_mcknots} demonstrates

1997: \begin{figure}

1998: \begin{center}

1999: \includegraphics[height=4.4cm]{k3.eps}

2000: \includegraphics[height=4.4cm]{k8.eps}

2001: \includegraphics[height=4.4cm]{k33.eps}

2002: \end{center}

2003: \caption{Typical configurations of 256-mer chains for the trefoil $3_1$,

2004: the prime knot $8_{19}$, and the composite knot $3_1$\#$3_1$

2005: consisting of two trefoils, in $d=2$. The initial conditions were symmetric

2006: in all cases.

2007: \label{fig_mcknots}}

2008: \end{figure}

2009: the tightness of the prime knot $8_{19}$. Composite knots, however, can

2010: maximise the number of configurations by

2011: splitting into their prime factors as indicated

2012: in figure \ref{fig_mcknots} for $3_1$\#$3_1$. Each prime factor is

2013: tight and located at the fringe of one large loop, and accounts for

2014: an additional factor of $L$ for the number of configurations

2015: as compared to a ring of length $L$ without a knot.

2016: Indeed, this gain in entropy leads to the tightness of knots.

2017: Flat knots can experimentally be produced by `projecting' a dilute 3D knot

2018: from the bulk onto a mica surface, on which the knot is adsorbed. Variation

2019: of the ionic strength in the solution determines whether the knot is going to

2020: be strongly trapped on the surface such that, once captured on the surface, it

2021: is completely immobilised (small ionic strength); or whether the adsorption

2022: is weaker such that the knot can (partially) equilibrate while being confined

2023: to 2D, i.e., equilibrate as a flat knot.

2024: Figure \ref{valle} shows a strongly trapped complex knot, whereas figure

2025: \ref{erika} depicts a weakly adsorbed simple knot, compare \cite{valle}.

2026: 

2027: \begin{figure}

2028: \includegraphics[width=6.8cm,angle=90]{erika.eps}

2029: \caption{Flat knot imaged by AFM, similar to the one shown in figure

2030: \ref{valle}. However, this knot is rather simple (likely a trefoil)

2031: and was allowed to relax while attaching to freshly cleaved mica in

2032: the presence of bivalent Mg counterions. Courtesy E. Ercolini, J. Adamcik,

2033: and G. Dietler. Note the close resemblance to

2034: the trefoil configuration shown in figure \ref{fig_mcknots}.

2035: \label{erika}}

2036: \end{figure}

2037: 

2038: 

2039: \subsubsection{Flat knots under $\Theta$ and dense conditions.}

2040: 

2041: In many situations, polymer chains are not dilute.

2042: Polymer melts, gels, or rubbers exhibit fairly high densities of

2043: chains, and the behaviour of an individual chain in such systems

2044: is significantly different compared to the dilute phase

2045: \cite{degennes,doi,erman}. Similar considerations apply to biomolecules:

2046: in bacteria, the gyration radius of the almost freely floating

2047: ring DNA may sometimes be larger than the cell radius itself.

2048: Moreover, under certain conditions, there is a non-negligible

2049: osmotic pressure due to vicinal layers of protein molecules,

2050: which tends to confine the DNA \cite{walter,vasilevskaya,lerman}.

2051: In protein folding studies, globular proteins in their native

2052: state are often modelled as compact polymers on a lattice

2053: (see \cite{garel} for a recent review).

2054: 

2055: A polymer is considered dense if, on a lattice, the fraction $f$ of occupied

2056: sites has a finite value $f > 0$. This can be obtained by considering a single

2057: polymer of total length $L$ inside a box of volume $V$ and taking the limit

2058: $L \to \infty$, $V \to \infty$ in such a way that $f = L/V$ remains

2059: finite \cite{duplantier2,duplantier3,jacobsen}.

2060: Alternatively, dense polymers can be obtained in an infinite volume

2061: through the action of an attractive force between monomers.

2062: Then, for temperatures $T$ below the collapse

2063: (Theta) temperature $\Theta$, the polymers collapse to a dense phase,

2064: with a density  $f > 0$, which is a function of $T$

2065: \cite{duplantier3,duplantier4,owczarek,owczarek1}.

2066: For a dense polymer in $d$ dimensions, the exponent

2067: $\nu$, defined by the radius of gyration $R_g \sim L^{\nu}$, becomes

2068: $\nu = 1 / d$.

2069: The limit $f = 1$ is realised in Hamiltonian paths,

2070: where a random walk visits every site of a given

2071: lattice exactly once \cite{duplantier7,kondev}. Dense polymers may be related

2072: to 2D vesicles and lattice animals (branched polymers)

2073: \cite{fisher,seno,dekeyser,cardy}.

2074: 

2075: As studied in reference \cite{dense},

2076: the value of the exponent $c$ for the 2D dense F8 is (compare to the appendix)

2077: \begin{equation} \label{c8}

2078: c=-\gamma_{\rm F8}=11/8=1.375,

2079: \end{equation}

2080: implying that the smaller loop is {\em weakly localised\/}. This means

2081: that the probability for the size of each loop is peaked

2082: at $\ell=0$ and, by symmetry, at $\ell=L$.

2083: An analogous reasoning for the 2D F8 at the $\Theta$ point gives

2084: \begin{equation} \label{c8_theta}

2085: c=11/7=1.571.

2086: \end{equation}

2087: In both cases the smaller loop is weakly localised in the sense that $\langle

2088: l\rangle_</L\to 0$. Figure \ref{fig-8} shows

2089: \begin{figure}

2090: \begin{center}

2091: \includegraphics[width=4.2cm]{dense_f8_1.eps}

2092: \includegraphics[width=4.2cm]{dense_f8_3.eps}

2093: \end{center}

2094: \caption{Symmetric ($\ell=L/2=128$) initial configuration of a 2D dense

2095: F8 (left) and equilibrium configuration (right) with

2096: periodic boundary conditions. The two different grey values correspond to the

2097: two subloops created by the slip-link. The slip-link itself is

2098: represented by the three (tethered) black dots.\label{fig-8}}

2099: \end{figure}

2100: the symmetric initial and a typical equilibrium configuration for periodic

2101: boundary conditions obtained from Monte Carlo (MC) simulations, see reference

2102: \cite{dense} for details.

2103: In figure \ref{fig-8}, the lines represent the bonds (tethers) between the

2104: monomers (beads, not shown here). The three black dots mark the locations

2105: of the tethered beads forming the slip-link in 2D.

2106: The initial symmetric configuration soon gives way to a

2107: configuration with $\ell\ll L$ on approaching equilibrium.

2108: Figure \ref{fig-8_1} shows the development of this symmetry breaking as a

2109: function of the number of MC steps.

2110: We note, however,

2111: that the fluctuations of the loop sizes in the ``stationary'' regime

2112: appear to be larger in comparison to the dilute case studied in reference

2113: \cite{paraknot}, compare figure \ref{fig8_sim}.

2114: We checked that for densities (area coverage) above

2115: 40\% the scaling behaviour becomes independent of the density.

2116: (The above simulation results correspond to a density of 55\%.)

2117: The size distribution data is well fitted to a power law (for over 1.5 decades

2118: with 1024 monomers),

2119: and the corresponding exponent  with 512 and 1024

2120: monomers in figure \ref{fig-8_1}

2121: is in good agreement with the predicted value (\ref{c8}).

2122: 

2123: For our MC analysis, we again used a hard core

2124: bead-and-tether chain, in which self-crossings were prevented by keeping a

2125: maximum bead-to-bead distance of 1.38 times the bead diameter, and a maximum

2126: step length of 0.15 times the bead diameter. To create the dense F8 initial

2127: condition, a free F8 is squeezed into a quadratic box with hard

2128: walls. This is achieved by starting off from the free F8, surrounding it by a

2129: box, and turning on a force directed towards one of the edges. Then,

2130: the opposite edge is moved towards the centre of the box, and so on.

2131: During these steps, the slip-link is locked,

2132: i.e., the chain cannot slide through it, and the two loops are of equal length

2133: during the entire preparation. Finally, when the envisaged density is

2134: reached, the hard walls are replaced by periodic boundary conditions,

2135: and the slip-link is unlocked.

2136: After each step, the system is allowed to relax for times larger than the

2137: localisation times occurring at the main stage of the run.

2138: 

2139: \begin{figure}

2140: \begin{center}

2141: \includegraphics[height=6cm]{pdf_f8_dense.eps}

2142: \end{center}

2143: \caption{The loop size probability distribution  $p(\ell)$ at $\rho=55\%$ area

2144: coverage, for the F8 with 512 (top) and 1024 (bottom) monomers.

2145: The power law with with the predicted exponent $c=1.375$ in equation (\ref{c8})

2146: is indicated by the dotted line.

2147: \label{fig-8_1}}

2148: \end{figure}

2149: 

2150: A similar analysis as for the dense/$\Theta$-F8 structure and the dilute flat

2151: trefoil above, reveals the number of degrees of freedom for the flat dense trefoil

2152: in the form \cite{dense}

2153: \begin{equation}

2154: \label{3_limit}

2155: \omega_{3}(\ell,L)\sim\omega_{0}(L)\ell^{-c}

2156: \end{equation}

2157: with $c=-\gamma_3-m$, where $\gamma_3=-33/8$ from equation

2158: (\ref{nexp}) in the appendix (${\cal L} = 4$, $n_4 = 3$)

2159: and $m = 4$ is the number of independent integrations over chain

2160: segments. Thus, $c = 1/8 < 1$ which implies that the dense 2D

2161: trefoil is {\em delocalised}.

2162: As above, we have to consider the various possible contractions of the flat

2163: knot. For dense polymers, the present scaling results show

2164: that both the original trefoil shape ($c = 1/8 < 1$, see above)

2165: and position II ($c = 3/4 < 1$) are in fact {\em delocalised\/} and

2166: represent equally the leading scaling order (cf.~top part of figure

2167: \ref{knot_zoo}). The F8 is only found at the third position and is weakly

2168: localised ($c = 11/8 > 1$).

2169: In an MC simulation of the dense 2D trefoil, we predict that

2170: one mainly observes delocalised shapes corresponding to the original

2171: trefoil and position II in figure \ref{knot_zoo}, and further,

2172: with decreasing probability, the weakly localised F8 and the other

2173: shapes of the hierarchy (top part) in figure \ref{knot_zoo}.

2174: 

2175: These predictions are consistent with the numerical simulations of

2176: reference \cite{stella}, who observe that the mean

2177: value of the second largest segment of the simulated 2D dense trefoil

2178: configurations grows linearly with $L$, and conjecture the same

2179: behaviour also for the other segments, corresponding to the

2180: delocalisation of the trefoil obtained above.

2181: 

2182: An analogous reasoning can be applied to the 2D trefoil in the

2183: $\Theta$ phase. We find that

2184: in this case that the leading shape is again the

2185: original (uncontracted) trefoil, with $c = 5/7 < 1$. This

2186: implies that the 2D trefoil is {\em delocalised\/} also at the

2187: $\Theta$ point. All other shapes are at least weakly localised,

2188: and subdominant to the leading scaling order represented by the

2189: original trefoil. The resulting hierarchy of shapes is shown

2190: in figure \ref{knot_zoo} (bottom part).

2191: 

2192: 

2193: \subsection{3D knots defy complete analytical treatment.}

2194: 

2195: As already mentioned, 3D knots correspond to a problem involving hard

2196: constraints that defy a closed analytical treatment. It may be possible,

2197: however, that by a suitable mapping to, for instance, a field theory,

2198: an analytical description may be found. This may in fact be connected

2199: to the study of knots in diagrammatic solutions in

2200: high energy physics \cite{gambini}. There exists a fundamental relation

2201: between knots and gauge theory as knot projections and Feynman graphs

2202: share the same basic ingredients corresponding to a Hopf algebra

2203: \cite{kauffman}. However, up to now no such mapping has been found, and

2204: a theoretical description of 3D knots based on first principles is

2205: presently beyond hope. To obtain some insight into the statistical

2206: mechanical behaviour of knotted chains, one therefore has to resort

2207: to simulations studies or experiments. In addition, a few phenomenological

2208: models for both the equilibrium and dynamical behaviour of knots have been

2209: suggested such as in references

2210: \cite{quake,slili3d,lai1,sheng1,sheng,grosberg1,grosberg2,grosberg4}.

2211: 

2212: When discussing numerical knot studies, we already mentioned the Flory-type

2213: model brought leading to equation \ref{quake} \cite{quake,quake1}.

2214: One may argue that the

2215: differences in the knot size for the different knot types corresponding to

2216: the same $C$ may be included in the prefactor, that is independent of the

2217: chain length $N$. Obviously, this model of equal loop sizes is equivalent to

2218: a completely delocalised knot.

2219: This statement is in fact equivalent to another Flory-type approach to

2220: knotted polymers reported in \cite{grosberg2}. In this model, the knot is

2221: thought of as an inflatable tube: for a very thin tube diameter, the

2222: tube is equivalent to the original knot conformation; inflating the tube more

2223: and more will increasingly smoothen out the shape until a maximally inflated

2224: state is reached. The knot is then characterised by the aspect ratio

2225: \begin{equation}

2226: p=\frac{L}{D}\,\,, \, \mbox{therefore} \,\,

2227: 1\le p\le N,

2228: \end{equation}

2229: between length $L$ and maximum tube diameter $D$. It appears that $p$ is a

2230: (weak) knot invariant, and can be used to characterise the gyration radius of

2231: the knot. It is clear that, by construction, the aspect ratio described a

2232: totally delocalised knot, and indeed it turns out that in good solvent, the

2233: gyration radius shows the dependence $R_g\simeq AN^{3/5}\tau^{1/5}p^{-4/15}$,

2234: where $\tau$ is the (dimensionless) deviation from the $\Theta$ temperature

2235: \cite{grosberg2}. Obviously, the aspect ratio appears to be proportional to

2236: the number of essential crossings in comparison to expression (\ref{quake}).

2237: We note that similar considerations are employed in reference \cite{grosberg1},

2238: including a comparison to the entropy of a tight knot, finding comparable

2239: entropic likelihood. The modelling based on the aspect ratio $p$ is further

2240: refined in \cite{grosberg4}.

2241: 

2242: Knot localisation is a subtle interplay between the degrees of freedom of one big loop,

2243: and the internal degrees of freedom of the various segments in the knot region. Under

2244: localisation, the number of degrees of freedom

2245: \begin{equation}

2246: \omega\simeq\mu^NN^{1-d\nu}

2247: \end{equation}

2248: includes an additional factor $N$ from the knot region encircling the

2249: big loop. For flat knots, the competition between the single big loop and

2250: the knot region is indeed won by the big loop.

2251: In the case of 3D knots, this balance is presently not resolved for

2252: knots of all complexity. Probably only detailed simulations studies of higher

2253: order knots will make it possible to decide for the various models of 3D knots.

2254: Major contributions are also expected from single molecule experiments, for

2255: instance, from force-extension measurements along the lines of the simulations

2256: study in \cite{farago}, the advantage of experiments being the fact that it

2257: should be possible to go towards rather high chain lengths that are inaccessible

2258: in simulations.

2259: To overcome similar difficulties in the context of the entropic

2260: elasticity for rubber networks, Ball, Doi, Edwards and coworkers

2261: replaced permanent entanglements by slip-links

2262: \cite{edwards2a,edwards2b,edwards2c,edwards2d}.

2263: Gaussian networks containing slip-links have been successful in the prediction

2264: of important physical quantities of rubber networks \cite{erman}, and they

2265: have been used to study a small number of entangled chains \cite{sommer}. In a

2266: similar fashion, one may investigate the statistical behaviour of single polymer

2267: chains in which a fixed topology is created by a number of slip-links. Such

2268: `paraknots' can be studied analytically using the Duplantier scaling results

2269: \cite{paraknot}. As mentioned previously, knowledge of the statistical

2270: behaviour of paraknots can be used to create a knot scale for calibrating

2271: the degrees of freedom of real knots, and therefore also important to understand or design

2272: indirect experiments on knot entropy, such as by force-extension measurements

2273: \cite{pull}. Paraknots may also be useful in the design of entropy-based

2274: functional molecules \cite{hame_cpl,meamb_ctn}.

2275: 

2276: 

2277: \section{DNA breathing: local denaturation zones and biological implications}

2278: 

2279: "A most remarkable physical feature of the DNA helix, and one that is

2280: crucial to its functions in replication and transcription, is the ease

2281: with which its component chains can come apart and rejoin. Many techniques

2282: have been used to measure this melting and reannealing behaviour. Nevertheless,

2283: important questions remain about the kinetics and thermodynamics of denaturation

2284: and renaturation and how these processes are influenced by other molecules in

2285: the test tube and cell'' \cite{kornberg}. This remarkable quotation, despite

2286: 30 years old, still summarises the challenge of understanding local and global

2287: denaturation of DNA, in particular, its dynamics. In this section, we report

2288: recent findings on the spontaneous formation of intermittent denaturation

2289: zones within an intact DNA double helix. Such denaturation \emph{bubbles\/}

2290: fluctuate in size by (random) motion of the zipper forks relative to each

2291: other. The opening and subsequent closing of DNA bubbles is often called

2292: \emph{DNA breathing}.

2293: 

2294: \begin{figure}

2295: \includegraphics[width=8.6cm]{melt.eps}

2296: \caption{Fraction $\theta_h$ of double-helical domains within the DNA as a

2297: function of temperature. Schematic representation of $\theta_h(T)$,

2298: showing the increased formation of bubbles and unzipping from the ends,

2299: until full denaturation has been reached.

2300: \label{melting_curve}}

2301: \end{figure}

2302: 

2303: 

2304: \subsection{Physiological background of DNA denaturation}

2305: \label{breathintro}

2306: 

2307: The Watson-Crick double-helix is the

2308: thermodynamically stable configuration of a DNA molecule under physiological

2309: conditions (normal salt and room/body temperature). This stability is effected

2310: (a) by Watson-Crick H-bonding, that is essential for the specificity of

2311: base-pairing, i.e., for the key-lock principle according to which the

2312: nucleotide Adenine exclusively binds to Thymine, and Guanine only to

2313: Cytosine. Base-pairing therefore guarantees the high level of fidelity

2314: during replication and transcription. (b) The second contribution to

2315: DNA-helix stability comes from base-stacking between neighbouring bps:

2316: through hydrophobic interactions between the planar aromatic bases, that

2317: overlap geometrically and electronically, the bp stacking stabilises

2318: the helical structure against the repulsive electrostatic force between the

2319: negatively charged phosphate groups located at the outside of the DNA

2320: double-strand. While hydrogen bonds contribute only little

2321: to the helix stability, the major support comes from base-stacking

2322: \cite{kornberg,delcourt}.

2323: 

2324: The quoted ease with which its component chains can come apart and rejoin,

2325: without damaging the chemical structure of the two single-strands, is

2326: crucial to many physiological

2327: processes such as replication via the proteins DNA helicase and polymerase,

2328: and transcription through RNA polymerase. During these processes, the proteins

2329: unzip a certain region of the double-strand, to obtain access to the genetic

2330: information stored in the bases in the core of the double-helix

2331: \cite{kornberg,revzin,watsoncrick}. This unzipping corresponds to breaking

2332: the hydrogen bonds between the bps. Classically, the so-called melting

2333: and reannealing behaviour of DNA has been studied in solution in vitro by

2334: increasing the temperature, or by titration with acid or alkali. During thermal

2335: melting, the stability of the DNA duplex is related to the content of

2336: triple-hydrogen-bonded G-C bps: the larger the fraction of G-C pairs,

2337: the higher the required melting temperature or pH value. Thus, under thermal

2338: melting, dsDNA starts to unwind in regions rich in A-T bps, and then

2339: proceeds to regions of progressively higher G-C content

2340: \cite{kornberg,delcourt}. Conversely, molten, complementary chains of

2341: single-stranded DNA (ssDNA) begin to reassociate

2342: and eventually reform the original double-helix under incubation at roughly

2343: 25$^\circ$ below the melting temperature $T_m$ \cite{kornberg}. The relative

2344: amount of molten DNA in a solution can be measured by UV spectroscopy,

2345: revealing large changes in absorption in the presence of perturbed

2346: base-stacking \cite{wartell}. Careful melting studies allow one to obtain

2347: accurate values for the stacking energies of the various combinations

2348: of neighbouring bps, a basis for detailed thermodynamic modelling

2349: of DNA-melting and DNA-structure per se \cite{blake,blossey}. In fact,

2350: thermal melting data have been successfully

2351: used to identify coding sequences of the genome due to the different

2352: G-C content \cite{yeramian,yeramian1,carlon}.

2353: 

2354: Complementary to thermal or pH induced denaturation, dsDNA can be driven

2355: toward denaturation mechanically, by applying a tensional stress along

2356: the DNA in an optical tweezer trap \cite{williams}. As shown in figure

2357: \ref{mark_stretch}, the force per extension increases in worm-like chain

2358: fashion, until a plateau at approximately 65 pN is reached. This plateau

2359: is sometimes interpreted as new DNA configuration, the S form \cite{busta}.

2360: By a series of experiments, it appears more likely that the plateau corresponds

2361: to the mechanical denaturation transition \cite{williams1}. To first order,

2362: the effect of

2363: the longitudinal pulling translates into an external torque $\mathfrak{T}$,

2364: whose effect is a decrease in the free energy for melting a bp:

2365: \begin{equation}

2366: \Delta G_F=\Delta G_{F=0}-\mathfrak{T}\theta_0,

2367: \end{equation}

2368: where $\theta_0=2\pi/10.35$ is the twist angle per bp of the double helix

2369: \cite{hwa}.

2370: 

2371: \begin{figure}

2372: \includegraphics[width=8cm]{williams.eps}

2373: \caption{Overstretching of double-stranded DNA. The black curve shows the

2374: typical force-extension behaviour of DNA following the rapid worm-like chain

2375: increase until at around 65 pN a plateau is reached. Crossing of the plateau

2376: corresponds to progressive mechanical denaturation. See text for details.

2377: Figure courtesy Mark C. Williams.}

2378: \label{mark_stretch}

2379: \end{figure}

2380: 

2381: An important application of thermal DNA melting is the Polymerase Chain

2382: Reaction (PCR). In PCR, dsDNA is melted at elevated temperatures into two

2383: strands of ssDNA. By lowering the temperature in a

2384: solution of invariable primers and single nucleotides, each ssDNA is completed

2385: to dsDNA by the key-lock principle of base-pairing \cite{pcr,pcrrev}.

2386: By many such cycles,

2387: of the order of $10^9$ copies of the original DNA can be produced within the

2388: range of hours.\footnote{Most proteins denature

2389: at temperatures between 40 to 60$^\circ$C, including polymerases. In early

2390: PCR protocols, after each heating step new polymerase had to be washed into

2391: the reaction chamber. Modern protocols make use of heat-resistant polymerases

2392: that survive the temperatures necessary in melting. Such heat-resistant proteins

2393: occur, for instance, in bacteria dwelling near undersea thermal vents.}

2394: Again, the error rate due to the underlying biochemistry can be considered

2395: negligible for most purposes. In particular, from the viewpoint

2396: of polymer physics/chemistry, the obtained sample is monodisperse and free

2397: of parasitic reactions, creating (almost) ideal samples for physical studies,

2398: in particular, as any designed sequence of bases can be custom-made in modern

2399: molecular biology labs \cite{snustad}.

2400: 

2401: While the double-helix is the thermodynamically stable configuration

2402: of the DNA molecule below $T_m$ (at non-denaturing pH), even at

2403: physiological conditions there exist

2404: local denaturation zones, so-called DNA-bubbles, predominantly in A-T-rich

2405: regions of the genome \cite{wartell,poland}. Driven by ambient thermal

2406: fluctuations, a DNA-bubble is a dynamical entity whose size varies by

2407: thermally activated zipping and unzipping of successive bps at the two

2408: forks where the ssDNA-bubble is bordered by the dsDNA-helix. This incessant

2409: zipping and unzipping leads to a random walk in the bubble-size coordinate,

2410: and to a finite lifetime of DNA-bubbles under non-melting conditions, as

2411: eventually the bubble closes due to the energetic preference for the closed

2412: state \cite{wartell,poland}. DNA-breathing typically

2413: opens up a few bps \cite{gueron,krueger}. It has been demonstrated

2414: recently that by fluorescence correlation methods the fluctuations of

2415: DNA-bubbles can be explored on the single molecule level, revealing a

2416: multistate kinetics that corresponds to the picture of successive

2417: zipping and unzipping of single bps.\footnote{Essentially, the zipper

2418: model advocated by Kittel \protect\cite{kittel}.} At room temperature, the

2419: characteristic closing time of an unbounded bp was found to be in

2420: the range 10 to 100 $\mu$sec corresponding to an overall bubble lifetime in

2421: the range of a few msec \cite{altan}. The multistate nature of the

2422: DNA-breathing was confirmed by a UV-light absorption study \cite{zocchi}.

2423: The zipping dynamics of DNA is also investigated by NMR methods

2424: \cite{nmr,nmr1,nmr2}, revealing considerably shorter time scales than the

2425: fluorescence experiments. An interesting finding from NMR studies is the

2426: dramatically different denaturation dynamics in B' DNA, where more than

2427: three AT bps occur in a row \cite{russu}. It is conceivable that fluorescence

2428: correlation and NMR probe

2429: different levels of the denaturation dynamics. Our analysis of the single

2430: DNA fluorescence data reported below demonstrates that, albeit the much

2431: longer time scale, the dependence of the measured autocorrelation function

2432: on the stacking along the sequence is very sensitive, and agrees well with

2433: the quantitative behaviour predicted from the stability data.

2434: 

2435: The presence of fluctuating DNA-bubbles is essential to the

2436: understanding of the binding of single-stranded DNA binding proteins

2437: (SSBs) that selectively bind to ssDNA, and that play important roles

2438: in replication, recombination and repair of DNA \cite{kornberg1}. One

2439: of the key tasks of SSBs is to prevent the formation of secondary

2440: structure in ssDNA \cite{alberts,snustad}. From the thermodynamical

2441: point of view one would therefore expect SSBs to be of an effectively

2442: helix-destabilising nature, and thus to lower $T_m$

2443: \cite{vonhippel}. However, it was found that neither the gp32 protein

2444: from the T4 phage nor E.coli SSBs do

2445: \cite{vonhippel,vonhippel1,rich}. An explanation to this apparent

2446: paradox was suggested to consist in a kinetic block, i.e., a kinetic regulation

2447: such that the rate constant for the binding of SSBs is smaller than

2448: the one for bubble closing \cite{vonhippel1,rich1}. This hypothesis

2449: could recently be verified in extensive single molecule setups using

2450: mechanical overstretching of dsDNA by optical tweezers in the presence

2451: of T4 gene 32 protein \cite{pant,pant1,pant2}, as detailed below.

2452: 

2453: 

2454: \subsection{The Poland-Scheraga model of DNA melting}

2455: 

2456: The most widely used approach to DNA melting in bioinformatics is the

2457: statistical, Ising model-like Poland-Scheraga model (sometimes also referred

2458: to as Bragg-Zimm model) and its variations \cite{poland,cantor,wartell};

2459: see also \cite{kafri,kafri1,richard,hanke}. It

2460: defines

2461: the partition function $\mathscr{Z}$ of a DNA molecule in a grand canonical

2462: picture with arbitrary many bubbles. For simplicity, we will restrict the

2463: following discussion to a single bubble. Below the melting temperature

2464: $T_m$, the one bubble picture is a good approximation: due to the high

2465: energy cost of bubble initiation, the distance between bubbles on a DNA

2466: molecule is large, and bubbles behave statistically independently. In

2467: typical experimental setups for measuring the bubble dynamics (see below),

2468: the used DNA construct is actually designed to host an individual bubble.

2469: For a homopolymer stretch of double-stranded DNA with 400 bps, figure \ref{400}

2470: shows the probabilities to find zero, one or, or two bubbles as a function

2471: of the Boltzmann factor $u=\exp(\Delta G/RT)$ for denaturation of a

2472: single bp.\footnote{In biochemistry, energies are usually measured in calories

2473: per mol. Instead of the Boltzmann factor $\beta=1/k_BT$ commonly used in physics

2474: and engineering, it is therefore convenient to replace the Boltzmann constant

2475: $k_B$ by the gas constant $R=k_BN_A$, where $N_A$ is the Avogadro-Loschmidt

2476: number.} Even at the denaturation transition $\Delta G=0$, it is

2477: quite unlikely to find two bubbles simultaneously.

2478: 

2479: \begin{figure}

2480: \includegraphics[width=8cm]{bubble_M400.eps}

2481: \caption{Probability of having 0, 1, or 2 bubbles as a function of $u$

2482: for a DNA region of chain length 400 bps. The cooperativity parameter

2483: was $\sigma_0=10^{-3}$ and the loop correction exponent c=1.76 (see text).}

2484: \label{400}

2485: \end{figure}

2486: 

2487: The free energy $\Delta G$ to break an individual bp are constructed as

2488: follows. We mention two different approaches. Common for both is the

2489: Poland-Scheraga construction of the partition function. We start with

2490: the case that a linear DNA molecule denatures from one of its ends. The

2491: corresponding partition function is \cite{poland,wartell}

2492: \begin{equation}

2493: \label{endpart}

2494: \mathscr{Z}_{\mathrm{end}}(m)=\prod_{x=1}^me^{\Delta G_{x,x+1}/RT}

2495: \end{equation}

2496: where $m$ is the number of broken bps, and $\Delta G_{x,x+1}$ is the stacking

2497: free energy for disruption of the bp at position $x$ measured from the end

2498: of the DNA. The notation explicitly refers to the stacking between the bp at

2499: $x$ and $x+1$.

2500: The first closed bp is located at $x=m+1$. For a homonucleotide, $\Delta

2501: G_{x,x+1}=\Delta G$, while for a given sequence of bps, there come into play

2502: the different stacking energies for the possible combinations of pairs of bps

2503: in sequence\footnote{I.e., an AT bp followed by another AT as different from

2504: an AT followed by a TA, etc.} The stacking energies $\Delta G_{x,x+1}$ have

2505: the following contributions.

2506: 

2507: The more traditional way to determine the stacking interactions is by fit of

2508: bulk melting curves of DNA constructs containing exclusively pairs of the

2509: specific bp-bp combination such as $(\mathrm{AT/TA})_n$

2510: (see, e.g., \cite{santalucia,delcourt} and references therein). The free energy

2511: used in this automated fit procedure using the MELTSIM algorithm \cite{blake},

2512: \begin{equation}

2513: \label{eblake}

2514: \Delta G^{\mathrm{Mix}}_{x,x+1}=\Delta H^{\mathrm{ST}}_{x,x+1}-T\Delta

2515: S_{x,x+1},

2516: \end{equation}

2517: combines the stacking enthalpy difference $\Delta H^{\mathrm{ST}}_{x,x+1}$

2518: for both hydrogen bond and actual stacking energies, and the entropy difference

2519: $\Delta S_{x,x+1}$ chosen to explicitly depend on the nature of the broken bp.

2520: A recent alternative to determine the stability parameters of DNA was

2521: developed in the group of Frank-Kamenetskii, leading to the free energy

2522: \cite{krueger} 

2523: \begin{equation}

2524: \label{ekrueger}

2525: \Delta G^{\mathrm{Sep}}_{x,x+1}=\Delta G^{\mathrm{ST}}_{x,x+1}+\Delta G

2526: ^{\mathrm{HB}}_x

2527: \end{equation}

2528: where the Gibbs free energies $\Delta G^{\mathrm{ST}}_{x,x+1}$ and

2529: $\Delta G^{\mathrm{HB}}_x$ measure the stacking of bps $x$ and $x+1$ and

2530: the hydrogen bonding of bp $x$ \emph{including} the entropy release on

2531: disruption. Note that $\Delta G^{\mathrm{HB}}_x$ is chosen such that it only

2532: depends on the broken bp and has two values for AT and

2533: GC bps, irrespective of the orientation (3' or 5'). The stacking free energies

2534: $\Delta G^{\mathrm{ST}}$ were determined from denaturation at a DNA nick and

2535: show a pronounced asymmetry between AT/TA and TA/AT bonds \cite{krueger}. For an

2536: end-denaturing DNA both descriptions are equivalent (though somewhat different

2537: when one puts numbers), as the breaking of each bp involves the disruption of

2538: one hydrogen bonds of bp $x$ and one stacking with its neighbour.

2539: 

2540: The difference between the two approaches becomes apparent when we consider

2541: the initiation of a bubble, i.e., a denatured coil enclosed by intact

2542: double-helix. Now, the partition function for a bubble with left fork

2543: position at $x_L$ and consisting of $m$ broken bps,

2544: \begin{equation}

2545: \label{midpart}

2546: \mathscr{Z}_{\mathrm{mid}}(x_L,m)=\Lambda\Sigma(m)\prod_{x=x_L}^{x_L+m}

2547: e^{\Delta G_{x,x+1}/RT},

2548: \end{equation}

2549: differs from (\ref{endpart}) in three respects: (i) While the bubble consists

2550: of $m$ molten bps, $m+1$ stacking interactions need to be broken to create

2551: two boundaries between intact double-strand and the single-strand in the

2552: bubble; the extra stacking interaction is effectively incorporated into

2553: $\Lambda$. (ii) The polymeric nature of the flexible single-stranded

2554: bubbles involves the entropy loss factor $\Sigma(m)=1/(m+D)^c$ with

2555: critical exponent $c$ and the parameter $D$ to take care of finite

2556: size effects\footnote{Usually, $D=1$ is chosen.} \cite{richard,blake,fixman};

2557: (iii) the factor $\Lambda$: In the standard notation, $\Lambda

2558: \equiv\sigma_0=\exp(-F_s/RT)\simeq10^{-4}-10^{-5}$ \cite{lazurkin,blake,%

2559: wartell,poland}, while according to \cite{krueger}, $\Lambda=\xi\simeq10^{-3}$

2560: is called the ring factor.

2561: Interestingly, the cooperativity parameter $\sigma_0$ is of the order of

2562: what corresponds to the singular unbalanced stacking enthalpy for breaking

2563: the first bp to initiate the bubble.

2564: The new stability data lead to a more pronounced asymmetry in opening

2565: probabilities between different bp-bp combinations. The analysis in

2566: references \cite{tobias_prl,tobias_long} demonstrates that the parameters

2567: from \cite{krueger} appear to support better the biological relevance

2568: of the TATA motif\footnote{The four bp {\small $\begin{array}{l}\cdot\mathtt{

2569: TATA}\cdot\\[-0.16cm]\cdot\mathtt{ATAT}\cdot\end{array}$} sequence is one of

2570: the typical codes marking where RNA polymerase starts the transcription process

2571: \cite{alberts,snustad}.} in natural sequences, that is, show a more pronounced

2572: simultaneous opening probability for the TATA motif.

2573: 

2574: As demonstrated for the autocorrelation function measuring the breathing

2575: dynamics in figure \ref{autocorr}, the description in terms of the

2576: partition function $\mathscr{Z}$ based on the stability parameters from

2577: \cite{krueger} reproduces the experimental data well. The analysis in

2578: \cite{tobias_prl,tobias_long} also indicates that the accuracy of the

2579: model predictions for the bubble dynamics is rather sensitive to the

2580: parameters. It is therefore conceivable that improved fluorescence

2581: measurement of the bubble dynamics may be employed to obtain accurate DNA,

2582: stability parameters, complementing the more traditional melting, NMR, and

2583: gel electrophoresis

2584: bulk methods. It should be noted that the dynamics is strongly influenced

2585: by local deviations from the B configuration of the DNA double helix. Thus,

2586: in local stretches of more than three AT bps in sequence, the B' structure

2587: is assumed, leading to pronouncedly different zipping dynamics \cite{russu}.

2588: 

2589: Two major questions remain in the thermodynamic formulation of DNA denaturation

2590: and its dynamics. Namely, the exact origin of the bubble initiation factor

2591: $\sigma_0$ (or, alternatively, the ring factor $\xi$ from \cite{krueger}), and

2592: a method to properly calibrate the zipping rate $k$. The

2593: factor $\sigma_0$ is related to the entropic imbalance on opening the

2594: first bp of a bubble: While this requires the breaking of two stacking

2595: interactions, only one bp has access to an increased amount of degrees

2596: of freedom. Still, these degrees of freedom must be influenced by the

2597: fact that the single open bp is coupled to two zipper forks. Currently,

2598: $\sigma_0$ remains a fit parameter. The exact value of the zipping rate

2599: $k$ remains open. While NMR experiments indicate much faster rates in

2600: the nanosecond range ($\simeq 10^8\mbox{ sec}^{-1}$), the fluorescence

2601: correlation measurements produce values in the microsecond range ($\simeq

2602: 10^4-10^5\mbox{ sec}^{-1}$). This large discrepancy may be based on the

2603: fact that both methods have different sensitivity to the amplitude of

2604: intra-bp separation. Currently, $k$ is taken as a fit parameter. In the

2605: analysis in \cite{tobias_prl,tobias_long}, we use the stacking parameters

2606: from \cite{krueger} including the value of the ring factor $\xi$, so that

2607: $k$ (a shift along the logarithmic abscissa) is the only adjustable

2608: parameter.

2609: 

2610: It has been under debate what exact value should be taken for the critical

2611: exponent $c$ entering in the entropy loss factor for a denaturation bubble.

2612: This is connected to the fact that $c>2$ would imply a first order denaturation

2613: transition on melting, while $1<c<2$ would stand for a second order

2614: transition \cite{fisher,richard}. Speculations about a possible first order

2615: transition are connected to the rather distinct spikes in the differential

2616: melting curves \cite{wartell}.\footnote{Due to the rather small DNA samples

2617: used in melting experiments (5000 bp and less \cite{wartell}), claims about

2618: the order of the underlying thermodynamical phase transition should be

2619: considered with some care.} Theoretical polymer physics approaches to explain

2620: a $c>2$ are either based on the inclusion of polymeric self-avoidance

2621: interactions of the bubble with the remainder of the chain \cite{kafri}; or

2622: built on a directed polymer model \cite{garel1}. Despite the elegance of both

2623: approaches, it is an open question how truly they represent the detailed

2624: denaturation behaviour of real DNA \cite{hame_comment,somendra_comment}.

2625: Applying the MELTSIM algorithm to typical sequences, it was found that there

2626: is a connection between the fit result for the cooperativity parameter $\sigma

2627: _0$, whose value is reduced from $\approx 10^{-5}$ to $\approx 10^{-3}$

2628: by assuming $c=2.12$ instead of $1.76$ \cite{blossey}. Below the melting

2629: transition, the typical bubble size is only a few bps, and in that regime the

2630: polymeric treatment of the loop entropy loss is of approximative nature.

2631: Indeed, in the analysis of \cite{krueger} no entropy loss due to polymer

2632: ring formation was included. For the breathing dynamics, we include $c$,

2633: to cover higher temperatures with somewhat larger bubbles,

2634: but find no significant change in the behaviour between $c<2$ and $c>2$,

2635: as long as the exponent is sufficiently close to 2.

2636: We therefore use the value $c=1.76$, that is consistent with the traditional

2637: data fits employed in the determination of the stacking parameters.

2638: 

2639: 

2640: \subsection{Fluctuation dynamics of DNA bubbles: DNA breathing}

2641: 

2642: Below the melting temperature $T_m$, DNA bubbles are intermittent, i.e.,

2643: they form spontaneously due to thermal fluctuations, and after some time close

2644: again. DNA-breathing can be thought of as a biased random walk in the phase

2645: space spanned by the bubble size $m$ and its position denoted, e.g., by the

2646: left zipper fork position $x_L$ \cite{tobias_prl,tobias_long}. The bubble

2647: creation can be viewed

2648: as a nucleation process \cite{landau}, whereas the bubble lifetime corresponds

2649: to the survival time of the first passage problem of relaxing to the $m=0$

2650: state after a random walk in the $m>0$ halfspace

2651: \cite{hame_jpa,tobias_jpc,tobias_prl,tobias_long,tobias_pre}. Bubble

2652: breathing on the

2653: single DNA-bubble level was measured by fluorescence correlation spectroscopy

2654: in \cite{altan}. This technique employs a designed stretch of DNA, in which

2655: weaker AT bps form the bubble domain, that is clamped by stronger GC bonds.

2656: In the bubble domain, a fluorophore-quencher pair is attached. Once the bubble

2657: is created, fluorophore and quencher are separated, and fluorescence occurs.

2658: A schematic of this setup is shown in figure \ref{fig:bubbles}.

2659: 

2660: \begin{figure}

2661: \begin{center}

2662: \includegraphics[width=6.8cm]{DNA_cartoon_xT.eps}

2663: \end{center}

2664: \caption{Clamped DNA domain with internal bps $x=1$ to $M$,

2665: statistical weights $u_{\rm hb}(x)$, $u_{\rm st}(x)$, and tag position

2666: $x_T$. The DNA sequence enters through the statistical weights $u_{\rm

2667: st}(x)$ and $u_{\rm hb}(x)$ for disrupting stacking and hydrogen bonds

2668: respectively. The bubble breathing process consists of the initiation

2669: of a bubble and the subsequent motion of the forks at positions $x_L$

2670: and $x_R$. See \cite{tobias_long} for details.}

2671: \label{fig:bubbles}

2672: \end{figure}

2673: 

2674: The zipper forks move stepwise $x_{L/R}\rightarrow x_{L/R}\pm 1$ with rates

2675: $\mathsf{t}^{\pm}_{L/R}(x_{L/R},m)$. We define for bubble size decrease

2676: \begin{equation}

2677: \label{eq:t_L_plus}

2678: \mathsf{t}^+_L(x_L,m)=\mathsf{t}^-_R(x_L,m)=k/2 \qquad (m\ge 2)

2679: \end{equation}

2680: for the two forks.\footnote{Due to intrachain coupling (e.g., Rouse), larger

2681: bubbles may involve an additional `hook factor' $m^{-\mu}$ \cite{tobias_jpc}.}

2682: The rate $k$ characterises a single bp zipping.

2683: Its independence of $x$ corresponds to the view that bp closure requires the

2684: diffusional encounter of the two bases and bond formation; as sterically AT

2685: and GC bps are very similar, $k$ should not significantly vary with bp stacking.

2686: The rate

2687: $k$ is the only adjustable parameter of our model, and has to be determined

2688: from experiment or future MD simulations. The factor $1/2$ is introduced for

2689: consistency \cite{tobias_jpc,tobias_pre,tobias_prl,tobias_long,tobias_leif}.

2690: Bubble size increase is controlled by

2691: \begin{eqnarray}

2692: \nonumber

2693: \mathsf{t}_{L}^{-}(x_L,m)&=&ku_{\rm st}(x_L) u_{\rm hb}(x_{L})

2694: s(m)/2,

2695: \label{eq:t_L_minus}\\

2696: \mathsf{t}_{R}^{+}(x_L,m)&=&ku_{\rm st}(x_R+1) u_{\rm hb}(x_{R})

2697: s(m)/2,

2698: \label{eq:t_R_plus}

2699: \end{eqnarray}

2700: for $m\ge 1$, where $s(m)=\{(1+m)/(2+m)\}^c$. Finally, bubble initiation and

2701: annihilation from and to the zero-bubble ground state, $m=0 \leftrightarrow

2702: 1$ occur with rates

2703: \begin{eqnarray}

2704: \nonumber

2705: \mathsf{t}_G^+(x_L)&=&k\xi's(0)u_{\rm st}(x_L+1)u_{\rm hb}(x_L+1)u_{\rm

2706: st}(x_L+2)\\

2707: \mathsf{t}_G^-(x_L)&=&k.

2708: \label{eq:t_G_plus}

2709: \end{eqnarray}

2710: The rates $\mathsf{t}$ fulfil detailed balance conditions. The annihilation

2711: rate $\mathsf{t}_G^-(x_L)$ is twice the zipping rate of a single fork, since

2712: the last open bp can close either from the left or right. Due to the clamping,

2713: $x_L\ge 0$ and $x_R\le M+1$, ensured by reflecting conditions $\mathsf{t}

2714: _L^-(0,m)=\mathsf{t}_R^+(x_L,M-x_L)=0$. The rates $\mathsf{t}$ together

2715: with the boundary conditions fully determine the bubble dynamics.

2716: 

2717: In the FCS experiment fluorescence occurs if the bps in a

2718: $\Delta$-neighbourhood of the fluorophore position $x_T$ are open

2719: \cite{altan}. Measured fluorescence time series thus correspond to

2720: the stochastic variable $I(t)$, that takes the value 1 if at least all bps

2721: in $[x_T-\Delta,x_T+\Delta]$ are open, else it is 0. The time averaged

2722: ($\overline{\,\,\cdot\,\,}$) fluorescence autocorrelation

2723: \begin{equation}

2724: \label{autocf}

2725: A_t(x_T,t)=\overline{I(t)I(0)}-\overline{I(t)}^2

2726: \end{equation}

2727: for the sequence AT9 from \cite{altan} are rescaled in figure \ref{autocorr}.

2728: 

2729: \begin{figure}

2730: \begin{center}

2731: \includegraphics[width=9.1cm]{autocorr_at9.eps}

2732: \end{center}

2733: \caption{Time-dependence of the autocorrelation function $A_t(x_T,t)$ for

2734: the sequence AT9 measured in the FCS setup of reference \cite{altan} at 100mM

2735: NaCl. The full

2736: lines show the result from the master equation, based on the DNA stability

2737: parameters from Krueger et al.~\cite{krueger}. The inset shows the broadening

2738: of the relaxation time spectrum with increasing temperature.}

2739: \label{autocorr}

2740: \end{figure}

2741: 

2742: We note that an alternative method to obtain precise DNA stability data

2743: may be provided by a DNA construct with two AT-rich zones

2744: between which a shorter GC-rich bridge is located. The first passage

2745: problem corresponding to bubble merging at temperatures between the

2746: melting temperatures of the AT and GC zones was recently calculated

2747: \cite{hansen}, and provides the framework for modified fluorescence

2748: correlation setups similar to the one from reference \cite{altan}.

2749: 

2750: 

2751: \subsection{Probabilistic modelling---the master equation (ME)}

2752: 

2753: DNA breathing is described by the probability

2754: distribution $P(x_L,m,t)$ to find a bubble of size $m$ located at $x_L$

2755: whose time evolution follows the ME

2756: \cite{tobias_jpc,tobias_pre,tobias_prl,tobias_long,tobias_leif}

2757: \begin{equation}

2758: \label{master}

2759: \frac{\partial}{\partial t}P(x_L,m,t)=\mathbb{W} P(x_L,m,t).

2760: \end{equation}

2761: The transfer matrix $\mathbb{W}$ incorporates

2762: the rates $\mathsf{t}$. Detailed balance guarantees equilibration toward

2763: $\lim_{t\to\infty}P(x_L,m,t)=\mathscr{Z}(x_L,m)/\mathscr{Z}$, with

2764: $\mathscr{Z}=\sum_{x_L,m}\mathscr{Z}(x_L,m)$

2765: \cite{vankampen}. The

2766: ME and the explicit construction of $\mathbb{W}$ are discussed at

2767: length in references \cite{tobias_jpc,tobias_long}. Eigenmode analysis and matrix

2768: diagonalisation produces all quantities of interest such as the ensemble

2769: averaged autocorrelation function

2770: \begin{equation}

2771: \label{auto}

2772: A(x_T,t)=\langle I(t)I(0)\rangle-(\langle I\rangle)^2.

2773: \end{equation}

2774: $\langle I(t)I(0)\rangle$ is proportional to the survival density that the

2775: bp is open at $t$ and that it was open initially \cite{tobias_prl,tobias_long}.

2776: 

2777: In figure \ref{autocorr} the blue curve shows the predicted behaviour of $A(

2778: x_T,t)$, calculated for $T=49^\circ$C with the parameters from

2779: \cite{krueger}. As in the experiment we assumed that fluorophore and quencher

2780: attach to bps $x_T$ and $x_T+1$, that both are required open to produce

2781: a fluorescence signal. From the scaling plot, we calibrate the zipping

2782: rate as $k=7.1\times 10^4/$s, in good agreement with the findings from

2783: reference \cite{altan}. The calculated behaviour reproduces the data within

2784: the error bars, while the model prediction at $T=35^\circ$C shows more

2785: pronounced deviation. Potential causes are destabilising effects of the

2786: fluorophore and quencher, and additional modes that broaden the decay

2787: of the autocorrelation. The latter is underlined by the fact that for

2788: lower temperatures the relaxation time distribution $f(\tau)$, defined

2789: by $A(x_T,t)=\int\exp(-t/\tau)f(\tau)d\tau$, becomes narrower

2790: (figure \ref{autocorr} inset).

2791: Deviations may also be associated with the correction for diffusional motion

2792: of the DNA construct, measured without quencher and neglecting

2793: contributions from internal dynamics \cite{oleg1}. Indeed, the black

2794: curve shown in figure \ref{autocorr} was obtained by a 3\% reduction of the

2795: diffusion time;\footnote{For diffusion time $\tau_D=150\mu$s measured for an RNA

2796: construct of comparable length in \cite{oleg1}.} see details in

2797: \cite{tobias_long}.

2798: 

2799: A remark on a prominent alternative approach to DNA breathing appears in order. 

2800: This is the Peyrard Bishop Dauxois (PBD) model \cite{peyrard,dauxois} based

2801: on the set of Langevin equations \cite{peyrard_cm}

2802: \begin{eqnarray}

2803: \nonumber

2804: m\frac{d^2y_n}{dt^2}&=&-\frac{dV(y_n)}{dy_n}-\frac{dW(y_{n+1},y_n)}{dy_n}-

2805: \frac{dW(y_n,y_{n-1})}{dy_n}\\

2806: &&-m\gamma\frac{dy_n}{dt}=\xi_n(t).

2807: \label{peyrard}

2808: \end{eqnarray}

2809: Here, $V(y_n)=D_n\left[\exp(-a_ny_n)-1\right]^2$ is a Morse potential

2810: for the hydrogen bonding, $D_n$ and $a_n$ assuming two different values

2811: for AT and GC bps; $W(y,y')=\frac{k}{2}\left[1+\rho\exp\{-\beta(y+y')\}

2812: \right](y-y')^2$ is a nonlinear potential to include bp-bp stacking

2813: interactions between adjacent bps $y$ and $y'$. The parameters $k$, $\rho$,

2814: $\beta$, $\gamma$, and $m$ are invariant of the sequence. Usually, the 

2815: stochastic equations (\ref{peyrard}) is integrated numerically

2816: \cite{peyrard_cm}.

2817: Due to its formulation in terms of a set of Langevin equations, the DPB model

2818: is very appealing, and it is a useful model to study some generic features of

2819: DNA denaturation. The disadvantage of the current formulation of the DBP model

2820: is the fact that it does not include enough parameters to account for the

2821: known independent stability constants of double stranded DNA (in fact, only

2822: two parameters are allowed to vary with the sequence, in contrast to the 12

2823: independent parameters needed to fully describe the bp stacking and hydrogen

2824: bonding \cite{krueger}).

2825: Moreover, there appear to be certain ambiguities in the proper formulation

2826: of boundary conditions in the stochastic integration \cite{the_prl_footnote},

2827: and also with respect to the interpretation of the biological relevance and

2828: computational limitations of the PBD model \cite{peyrard_comment}.

2829: The master equation and Gillespie approach brought forth in references

2830: \cite{tobias_jpc,tobias_pre,tobias_prl,tobias_long,tobias_leif} bridges the

2831: gap between the thermodynamic data for

2832: the bp stacking and hydrogen bonding obtained by various experimental methods,

2833: and the dynamical nature of DNA breathing. It is hoped that both dynamic

2834: models will synergetically be developed further and eventually lead to a

2835: better understanding of DNA denaturation fluctuations.

2836: 

2837: 

2838: \subsection{Stochastic modelling---the Gillespie algorithm}

2839: 

2840: \begin{figure*}

2841: \includegraphics[width=14cm]{ts_bubble.eps}

2842: \caption{Time series of single bubble-breathing dynamics for

2843: $\sigma_0=10^{-3}$, $M=20$, and (a) $u=0.6$ and (b) $u=0.9$. The lower

2844: panel shows a zoom-in of how single bubbles of size $m(t)$ open up and close.}

2845: \label{suman}

2846: \end{figure*}

2847: 

2848: Despite its mathematically simple form, the master equation (\ref{master})

2849: needs to be solved numerically by inverting the transfer matrix

2850: \cite{tobias_jpc,tobias_long}. Moreover, it produces

2851: ensemble-averaged information. Given the access to single molecule data,

2852: it is of relevance to obtain a model for the fully stochastic time evolution

2853: of a single DNA-bubble, providing a description for pre-noise-averaged

2854: quantities such as the step-wise (un)zipping. With this scope, we

2855: introduced a stochastic simulation scheme for the (un)zipping dynamics,

2856: using the Gillespie algorithm to update the state of the system by determining

2857: (i) the random time between individual (un)zipping events, and (ii) which

2858: reaction direction (zipping, $\leftarrow$, or unzipping, $\rightarrow$) will

2859: occur \cite{suman}. This scheme is efficient computationally, easy to

2860: implement, and amenable to generalisation.

2861: 

2862: To define the model, we denote a bubble state of $m$ broken bps by the

2863: occupation numbers $b_m=1$ and $b_{m'}=0$ ($m'\neq m$). The stochastic

2864: simulation then corresponds to the nearest-neighbour jump process

2865: \begin{equation}

2866: \label{eq1}

2867: b_0\rightleftarrows b_1\rightleftarrows\ldots

2868: \rightleftarrows b_m\rightleftarrows\ldots

2869: \rightleftarrows b_{M-1}\rightleftarrows b_M,

2870: \end{equation}

2871: with reflecting boundary conditions at $b_0$ and $b_M$. Each jump away

2872: from state $b_m$ occurs after a random time $\tau$, and in random direction

2873: to either $b_{m-1}$ or $b_{m+1}$, governed by the reaction probability

2874: density function\footnote{The original expression for the reaction probability

2875: density function,

2876: $P(\tau,\mu)=b_m\tr^{\mu}(m)$ $\exp\left(-\tau\sum_{m,\mu}b_m\tr^{\mu}(m)

2877: \right)$, that is relevant for consideration of multi-bubble states,

2878: simplifies here due to the particular choice of the

2879: $b_m$.} \cite{gillespie,gillespie1}

2880: \begin{equation}

2881: \label{rpdf}

2882: P(\tau,\mu)=\tr^{\mu}(m)e^{-\left(\tr^+(m)+\tr^-(m)\right)\tau},

2883: \end{equation}

2884: where $\mu\in\{+,-\}$ denotes the unzipping ($+$) or zipping ($-$) of a

2885: bp, and the

2886: jump rates $\tr^{\pm}(m)$ are defined below. From the joint probability

2887: density function (\ref{rpdf}),

2888: the waiting time probability

2889: density function that a jump away from $b_m$ occurs is given by $\psi(\tau)

2890: =\sum_{\mu}P(\tau,\mu)$, i.e., it is Poissonian. The probability that the

2891: bubble size does not change in the time interval $[0,t]$ is given by

2892: $\phi(t)=1-\int_0^t\psi(\tau)d\tau$. The fork position $x_L$ (and thereby

2893: the sequence of bps) is straightforwardly incorporated

2894: \cite{tobias_prl,tobias_long}.

2895: 

2896: We start the simulations from the completely zipped state, $b_0=1$ at

2897: $t=0$, and measure the bubble size at time $t$ in terms of

2898: $m(t)=\sum_{m=0}^Mmb_m(t)$. The time series of $m(t)$ for a single stochastic

2899: realization is shown in figure \ref{suman}. It is distinct that the bubble

2900: events are very sharp (note the time windows of the zoom-ins), and

2901: most of the time the zero-bubble state $b_0$ prevails due to $\sigma_0\ll 1$.

2902: Moreover, raising the temperature increases the bubble size and lifetime, as

2903: it should. By construction of the simulation procedure, it is guaranteed that an

2904: occupation number $b_m=1$ ($m\neq 0$) corresponds to exactly one bubble.

2905: 

2906: In a careful analysis, it was shown that the stochastic simulation method

2907: provides accurate information of the statistical quantities of the bubble,

2908: such as opening probability and autocorrelation function \cite{suman}. It

2909: can therefore be used to obtain the same information as the master equation

2910: with the advantage of also giving access to the noise in the system.

2911: With the Gillespie technique, we also obtained the data points in the

2912: graphs in this section.

2913: 

2914: 

2915: \subsection{Bacteriophage T7 promoter sequence analysis.}

2916: 

2917: An example from the analysis of the promoter sequence

2918: \begin{equation}

2919: \begin{array}{l}

2920: \mbox{\small\texttt{\textcolor{white}{AAAA}1\textcolor{white}{%

2921: AAAAAAAAAAAAAAAAAA}20}}\\[-0.1cm]

2922: \mbox{\small\texttt{\textcolor{white}{AAAA}|\textcolor{white}{%

2923: AAAAAAAAAAAAAAAAAA}|\textcolor{white}{AAAAAAAAA}\textcolor{white}{AAA}}}\\[-0.1cm]

2924: \mbox{\small\texttt{5'-aTGACCAGTTGAAGGACTGGAAGTAATACGACTC}}\\

2925: \mbox{\small\texttt{\textcolor{white}{AAA}AG}\textcolor{red}{

2926: \texttt{TATA}}\texttt{GGGACAATGCTTAAGGTCGCTCTCTAGGAg-3'}}\\[-0.1cm]

2927: \mbox{\small\texttt{\textcolor{white}{AAAAA}|\textcolor{white}{AAA}|

2928: \textcolor{white}{AAAAAAAAAAAAAAAAAAAAAAAAAA}|\textcolor{white}{AAA}}}\\[-0.1cm]

2929: \mbox{\small\texttt{\textcolor{white}{AAAAA}\textcolor{red}{38}%

2930: \textcolor{white}{AA}\textcolor{blue}{41}\textcolor{white}{%

2931: AAAAAAAAAAAAAAAAAAAAAAAAAA}68\textcolor{white}{AAA}}}

2932: \end{array}

2933: \label{promoter}

2934: \end{equation}

2935: of bacteriophage T7 is shown in figure \ref{signal} \cite{tobias_prl}. It

2936: depicts the time series of $I(t)$ for the tag positions

2937: $x_T=38$ at the beginning of TATA, and $x_T=41$ at the first GC bp after

2938: TATA. It is distinct how frequent bubble events are in TATA in comparison

2939: to the vicinal GC-rich domain (note that AT/TA bps are particularly weak

2940: \cite{krueger}). This is quantified by the waiting time density $\psi(\tau)$,

2941: whose characteristic time scale is more than an order of magnitude longer

2942: for the $x_T=41$ position. In contrast, we observe almost identical

2943: behaviour for the bubble survival density $\phi(\tau)$. Due to the proximity

2944: of $x_T=41$ to TATA, the typical bubble sizes for both tag positions are

2945: similar, and therefore the relaxation time. However, as shown in

2946: figure \ref{signal} bottom, the variation of the mean lifetime obtained from

2947: the master equation is quite small (within a factor 2) for the entire sequence.

2948: The latter graph also shows the insignificant variation according to the

2949: earlier stability parameters by Blake et al.~\cite{blake}.

2950: 

2951: The results summarised in figure \ref{signal} and further studies in

2952: \cite{tobias_prl,tobias_long} may indicate that it is not solely the

2953: increased opening probability at the TATA motif, as studied in \cite{choi}.

2954: Given the rather short bubble opening times of order of a few $k^{-1}$,

2955: it might be sufficient to induce binding of transcription enzymes (or

2956: other single stranded DNA binding proteins) if only bubble events are

2957: repeated often enough. In the present example, the waiting time between

2958: individual bubble events is increased by a factor of 25 inside the TATA

2959: motif. Guided by such results, detailed future studies combining optical

2960: tweezers overstretching and monitoring transcription initiation may be a

2961: step toward better understanding of this important biochemical process.

2962: 

2963: We note that the influence of noise (e.g., due to repetition of single

2964: molecule experiments) on the bubble dynamics can also be studied in the

2965: weak noise limit by a WKB method \cite{hans}. This model provides 

2966: information, for instance, about the time it takes a DNA to denature under

2967: temperatures above $T_m$ (mathematically corresponding to a finite time

2968: singularity). Bubble breathing can be mapped on the Coulomb problem of

2969: the Schr{\"o}dinger equation, and the corresponding phase transition

2970: studied \cite{hans_short}.

2971: 

2972: \begin{figure}

2973: \begin{center}

2974: \includegraphics[width=8.8cm]{signal.eps}

2975: \end{center}

2976: \caption{Time series $I(t)$ for the T7 promoter, with

2977: $x_T=38$, 41. Middle: Waiting time ($\psi(\tau)$) and bubble

2978: survival time ($\varphi(\tau)$) densities. Bottom: Mean bubble

2979: survival time, $\Delta=2$.}

2980: \label{signal}

2981: \end{figure}

2982: 

2983: 

2984: \subsection{Interaction of DNA bubbles with selectively single-strand

2985: binding proteins}

2986: 

2987: Let us now come back to the destabilising effect of single-stranded DNA

2988: binding proteins (SSBs) mentioned in section \ref{breathintro}. In a

2989: homopolymer approach, this was studied in a master equation approach

2990: in references \cite{tobias_pre,tobias_jpc}. The quantity of interest is

2991: the joint probability $P(m,n,t)$ to have a bubble consisting of $m$ broken

2992: bps, and $n$ SSBs bound to the two arches of the bubble. In addition to the

2993: rates $\mathsf{t}^{\pm}$ for bubble increase and decrease, the rates

2994: $\mathsf{r}^{\pm}$ for SSB binding and unbinding are necessary to define

2995: the breathing dynamics in the presence of SSBs. On the statistical level,

2996: the effect of the SSBs becomes coupled to the motion of the zipper forks.

2997: Thus, the rate for bubble size decrease is proportional to the probability

2998: that no SSB is located right next to the corresponding zipper fork; and

2999: the rate for SSB binding is proportional to the probability that there is

3000: sufficient unoccupied space on the bubble. Binding is allowed to be

3001: asymmetric, and is related to a parking lot problem in the following sense.

3002: The number $\lambda$ of bps occupied by a bound SSB is usually (considerably)

3003: larger than one. In order to be able to bind in between two already bound SSBs,

3004: the distance between these two SSBs must be larger than $\lambda$. The larger

3005: $\lambda$ the less efficient the SSB-binding becomes, similar to parking

3006: large cars on a parking lot desgined for small vehicles. Apart from the

3007: binding size $\lambda$ of the SSBs, two additional physical parameters

3008: come into play: the unbinding rate $q$ of the SSBs, and their binding

3009: strength $\kappa=c_0K^{\mathrm{eq}}$ consisting of the volume concentration

3010: $c_0$ of SSBs and the equilibrium binding constant $K^{\mathrm{eq}}=v_0\exp

3011: \left(\beta|E_{\mathrm{SSB}}|\right)$, with the typical SSB volume and binding

3012: energy $E_{\mathrm{SSB}}$.

3013: 

3014: \begin{figure}

3015: \includegraphics[width=8cm]{F_m_n_kappa_0p5.eps}

3016: \includegraphics[width=8cm]{F_m_n_kappa_1p5.eps}

3017: \caption{Effective free energy in the limit $\gamma\gg 1$ (---),

3018: and `free energy' for various

3019: fixed $n$ ($u=0.6$, $M=40$, $c=1.76$, $\lambda=5$). Top: $\kappa=0.5$; bottom:

3020: stronger binding, $\kappa=1.5$.

3021: \label{ssb_breath}}

3022: \end{figure}

3023: 

3024: The coupled dynamics of SSB-binding and bubble breathing is discussed in

3025: references \cite{tobias_pre,tobias_jpc}; similar effects in end-denaturing

3026: DNA was studied in \cite{tobias_jpca} in detail. Here, we report

3027: the behaviour of the effective free energy landscape in the limit of fast

3028: SSB-binding in the sense that the dimensionless parameter $\gamma\equiv q/

3029: k$ of SSB-unbinding and bubble zipping rates is large, $\gamma\gg 1$. This

3030: limit allows one to average out the SSB-dynamics and to calculate an effective

3031: free energy, in which the bubble dynamics with the slow variable $m$ runs

3032: off. The result for two different binding strengths $\kappa$ is shown in

3033: figure \ref{ssb_breath}, along with the free energies corresponding to keeping

3034: $n$ fixed. It is distinct that while for lower $\kappa$ the presence of SSBs

3035: diminishes the slope of the effective free energy, for larger $\kappa$ the

3036: slope actually becomes negative. In the first case, that is, the bubble

3037: opening is more likely, but still globally unfavourable. In the latter

3038: case, the presence of SSBs indeed leads to full denaturation. One observes

3039: distinct finite size effects due to $\lambda>1$: only when the bubble reaches

3040: a minimal size $m\ge\lambda$, SSB-binding may occur, a second SSB is allowed

3041: to bind to the same arch only once $m\ge2\lambda$, etc. This effect also

3042: produces the nucleation barrier for full denaturation in the lower plot of

3043: figure \ref{ssb_breath}.

3044: Similar finite size effects were investigated for biopolymer translocation

3045: in references \cite{tobias_pb,amlomme}.

3046: We note that the transition to denaturation could also be achieved by

3047: reaching a smaller positive slope of the effective free energy in the

3048: presence of SSBs, and additional titration or change of the effective

3049: temperature through actual temperature change or mechanical stretching

3050: as performed in the experiments reported in references \cite{pant,pant1,pant2}.

3051: 

3052: 

3053: \section{Role of DNA conformations in gene regulation}

3054: \label{generegulation}

3055: 

3056: Our current understanding of gene regulation to large extent is based on the

3057: experiments by Andr{\'e} Lwoff at Institut Pasteur more than 50 years ago

3058: \cite{lwoff}. Lwoff and his collaborators discovered that while a strain of

3059: \emph{E.coli}, a common intestinal bacterium, divided regularly when

3060: undisturbed,

3061: an unexpected phenomenon occurred when the strain was exposed to UV light:

3062: the bacteria stop growing and after some 90 minutes they burst (lyse),

3063: releasing a load of viruses. These viruses then invade new

3064: \emph{E.coli}.\footnote{Often these viruses are called phages or

3065: bacteriophages---bacteria eaters.} Some

3066: of the newly infected bacteria immediately lyse again, while the rest

3067: divides normally---while carrying the virus in them. This dormant state

3068: (lysogeny) of the bacterium can then be driven toward lysis by renewed UV

3069: exposure.

3070: 

3071: \begin{figure}

3072: \includegraphics[width=8.2cm]{ptashne.eps}

3073: \caption{Gene regulation, here the example of the (divergent) bacteriophage

3074: $\lambda$ switch after infection of \ecoli. Figure from \cite{ptashne1},

3075: with permission from M. Ptashne. This figure was modified by the author

3076: from the corresponding figure in M. Ptashne, A Genetic Switch: Phage Lambda

3077: and Higher Organisms, 2nd edition  \copyright Blackwell Science, Malden, MA

3078: and Cell Press, Cambridge, MA, with permission.

3079: \label{ptashne}}

3080: \end{figure}

3081: 

3082: The UV exposure-induced transition from lysogeny to lysis occurs as sketched

3083: in figure \ref{ptashne}. On infection, the bacteriophage $\lambda$ injects its

3084: DNA into \ecoli. In the lysogenic pathway, the viral DNA is integrated into

3085: the host DNA. During lysogeny, repressor dimers bind to certain operator

3086: sites on the $\lambda$ part of the DNA, recruiting RNA polymerase to bind

3087: to the overlapping promoter region(s) and blocking of the vicinal promoter

3088: for the divergent \emph{cro\/} gene. RNA polymerase then transcribes the

3089: \emph{cI} gene to the left of the operator, leading to the expression of

3090: new repressor molecules. UV light, however, leads to cleavage of the

3091: repressor dimers.\footnote{By activation of RecA proteins.} Now the basal

3092: transcription of the gene \emph{cro}, opposite to the \emph{cI\/} gene with

3093: respect to the operator region, leads to the expression of the Cro protein.

3094: Cro bound to the operator then recruits RNA polymerase to the operator,

3095: stabilising the Cro production and blocking \emph{cI}. Simultaneously, a whole

3096: sequence of genes

3097: is being expressed, and the virus reproduces itself inside \ecoli until

3098: lysis occurs. UV light flips the switch from transcription of the

3099: gene \emph{cI\/} maintaining the dormant lysogenic pathway, inducing lysis

3100: that is fostered by transcription of the \emph{cro\/} gene

3101: \cite{ptashne,ptashne1}.

3102: 

3103: The activity of a gene can be monitored even on the single genome level,

3104: by combining the targeted gene \emph{gI} with the gene leading to synthesis

3105: of GFP, the green fluorescent protein, i.e., when \emph{gI} is transcribed,

3106: then so is the gene for GFP. Occurrence of fluorescence then reports

3107: transcription of \emph{gI}. Connected to questions such as the stability of

3108: a genetic pathway is the search process of a specific gene by regulatory

3109: proteins, that is, how dynamically the binding protein actually locates

3110: the operator on the genome. We address these points in what follows.

3111: 

3112: 

3113: \subsection{Physiological background of gene regulation and expression}

3114: 

3115: The $\lambda$ switch from figure \ref{ptashne} is an example of a relatively

3116: simple mechanism. Even simpler is the well-studied Lac repressor. There,

3117: the \emph{lacZ} gene is expressed by recruitment through the CAP protein

3118: when \ecoli is starved of glucose and exposed to lactose. This enables

3119: \ecoli to digest lactose. In absence of lactose, \emph{lacZ} is blocked

3120: by the rep protein. In general, the expression of a certain gene is just

3121: one element in a cascade of simultaneous and/or hierarchical control units,

3122: such as in the developmental regulatory network of the sea urchin embryo

3123: \cite{davidson}. The basic physiological background is common to all of them:

3124: 

3125: Genes are the blueprints of proteins. They control physiological processes

3126: but also developmental pathways: from a fertilised egg cell, eventually

3127: all cell types (skin, hair, liver, brain, etc cells) of a human body emerge,

3128: or a skin cell changes colour on sunlight exposure. A gene is but a stretch

3129: of a DNA molecule, typically comprising some 200-1000 bps. Roughly

3130: speaking, a gene is on when it is being transcribed by RNA polymerase,

3131: otherwise it is off. RNA polymerase binds at the promoter region consisting

3132: of some 60 bps close to the beginning of the gene. It then converts

3133: the A, T, G, C code of the gene into a complementary messenger RNA (from

3134: which, in turn, the protein is produced during translation). The stop of

3135: transcription is triggered by a certain sequence at the end of the gene.

3136: Depending on specific conditions of the recruitment by regulatory proteins,

3137: RNA polymerase binding to the promoter of a certain gene is either blocked

3138: (the gene is off), facilitated (high binding affinity of RNA polymerase due

3139: to the (simultaneous) presence of certain protein(s)), or basal (in absence

3140: of any bound regulatory protein, RNA polymerase can still have a minor

3141: affinity to the promoter and then autonomously start transcription).

3142: The transcription mechanism is part and parcel of the central dogma of

3143: molecular biology summarised in figure \ref{crick}.

3144: 

3145: Molecular switches such as the $\lambda$ switch are surprisingly stable

3146: against noise, despite the fact that there are only about 100 repressor

3147: dimers in the entire bacteria cell \cite{reichardt}. Thus, apart from

3148: external induction, lysis occurs by spontaneous induction due to absence of

3149: CI from the operators \cite{metzi,mewo}.

3150: Such noise-induced errors are estimated to occur once

3151: in $10^7$ cell generations \cite{rozanov,little2}. The stability of the

3152: $\lambda$ switch against noise was analysed in terms of a Wentzell-Freidlin

3153: approach \cite{aurell} and by a simulation analysis \cite{aurell1}. The

3154: latter confirmed that the currently known molecular mechanisms used in

3155: modelling the $\lambda$ switch appear sufficient. While the classical

3156: Shea-Ackers model based on a statistical mechanical approach \cite{shea}

3157: is well established and studied numerically \cite{arkin},

3158: it relies on the knowledge of 13 fundamental Gibbs

3159: free binding energies composed to 40 different binding states of regulatory

3160: proteins and RNA polymerase at the two promoters of $\lambda$. Simulation

3161: of the complete $\lambda$ regulatory system proved the understanding of the

3162: mechanisms of the switch \cite{arkin}. Two more recent studies show that the

3163: $\lambda$ switch remains stable even when each of the fundamental Gibbs free

3164: energies is varied within its (appreciable) experimental error. Moreover,

3165: effects of potential mutations resulting in more significant changes of

3166: the binding energies were studied, and it was shown that certain mutations

3167: can even be compensated by parallel mutations influencing other binding

3168: energies (suppressors) \cite{bakk2,bakk3}. A typical result is shown in

3169: figure \ref{bakk}.

3170: 

3171: \begin{figure}

3172: \includegraphics[width=8.6cm]{Bakk_BJ.ps}

3173: \caption{Activity of the two $\lambda$ promoters as function of repressor

3174: concentration for vanishing Cro concentration. The full line corresponds

3175: to wild type data, whereas the dashed lines correspond to "mutations". The

3176: thin vertical line corresponds to lysogenic CI concentration. See

3177: reference \cite{bakk2} for details.

3178: \label{bakk}}

3179: \end{figure}

3180: 

3181: 

3182: \subsection{Binding proteins: Specific and nonspecific binding modes}

3183: 

3184: Given their very specific function, DNA-binding proteins must recognise

3185: a specific (cognate) sequence of nucleotides along the genome. In fact,

3186: without opening the double helix, the outside of the DNA can be read by

3187: proteins, as the edge of each bp is exposed at the surface. These

3188: patterns are unique only in the major groove of the DNA, this being the

3189: reason why gene regulatory proteins generally bind to the major groove.

3190: Apart from single bp pattern recognition, the protein binding is

3191: sensitive to the special surface features of a certain DNA region. This

3192: local structure of DNA needs to be complementary to the protein structure.

3193: Typical structure patterns (motifs) include helix-turn-helix, zinc 

3194: fingers, leucine zipper, and helix-loop-helix motifs \cite{alberts}.

3195: In bacteria, typical DNA-binding proteins cover some 20 bps or

3196: less. For instance, the lac repressor has a cognate sequence of 21 base

3197: pairs, the CAP protein 16, and the $\lambda$ repressor cI 17 bps

3198: \cite{alberts}. Although the interaction with a single nucleotide within

3199: such a DNA-protein bond is relatively weak, the sum of all matching

3200: nucleotides reaches appreciable values for the overall binding enthalpy,

3201: see below. Moreover, regulatory proteins bound simultaneously can 

3202: significantly enhance the stability of their individual bonds.

3203: 

3204: A simple model for the binding interaction goes back to the work of Berg

3205: and von Hippel \cite{bvh,berg}. Accordingly, the binding free energy is comprised

3206: of two contributions: (i) the (average) non-specific binding free energy

3207: due to electrostatic interaction with the DNA; and (ii) additional binding

3208: free energy if the sequence of the binding site is sufficiently close to the

3209: best (perfectly matching) sequence. The transition to the non-specific

3210: binding is supposed to occur via a conformational change of the regulatory

3211: protein from one that allows more hydrogen bond-formation to another that

3212: permits closer contact between the positive charges of the binding protein

3213: to the negatively charged DNA backbone \cite{wbvh}. This is supported by

3214: more recent structural studies: While in the non-specific binding mode

3215: the Lac repressor is bound to DNA in a rather loose and fuzzy way

3216: \cite{kaladimos}, it appears much more ordered in the specific mode. In

3217: fact, in the latter the protein induces a bend in the DNA \cite{lewis}.

3218: In case (ii), the additional binding free energy can, to good approximation,

3219: be considered independent and additive. reference \cite{gerland} provides a

3220: review of these issues, and derives the following result.

3221: Accordingly, to satisfy both the thermodynamic and kinetic constraints of

3222: the DNA-binding protein interaction, each additional base mismatch in

3223: comparison to the best sequence amounts to the loss of roughly 2 $k_BT$,

3224: and the optimal value for the transition between best specific binding to

3225: the cognate site and non-specific binding is shown to be some 16 $k_BT$

3226: below the energy of the best binding. This value is quite close to the

3227: $\approx 14$ $k_BT$ found for the difference between specific and nonspecific

3228: binding in \cite{bakk,bakk1}.

3229: 

3230: The fact that regulatory proteins bind with varying affinity is an important

3231: ingredient in gene regulation: Not all promoters should have the same activity,

3232: because some proteins are required by the cell at much higher levels than

3233: others. Thus, one given regulatory protein, that controls the recruitment to

3234: the promoters of several genes, can act with different strength depending

3235: on the degree of matching with the local sequence.

3236: 

3237: Non-specific binding can become quite appreciable. It was discovered in

3238: reference \cite{kao} that in the case of the Lac repressor less than 10\% of

3239: the proteins were unbound. In a more recent study using in vivo data of

3240: the $\lambda$ switch, it was found that in a lysogen nearly 90\% of the

3241: repressor protein cI is non-specifically bound. This implies that only

3242: 10-20 free cI dimers exist in the \ecoli cell at any time, pointing at

3243: the important role of non-specific binding in the search process of the

3244: cognate site addressed in the following subsection. Under different 

3245: conditions, both cI and Cro are always non-specifically bound by more than

3246: 50\%. The corresponding non-specific binding energies were estimated as

3247: 7 $k_BT$ \cite{bakk,bakk1}.

3248: 

3249: We note that in contrast to regulatory proteins, restriction enzymes

3250: have an approximate all-or-nothing matching

3251: condition: If a defined sequence matches the restriction enzyme, it will cut,

3252: otherwise not. Even a single mismatch reduces the action of the restriction

3253: enzyme by orders of magnitude. This distinction from regulatory

3254: protein makes sense as restriction enzymes are survival mechanisms and 

3255: should not just cut the cell's own DNA \cite{stormo}. This does not mean

3256: that restriction enzymes do not bind non-specifically---in fact, this is

3257: an important ingredient of their search process in total analogy to 

3258: regulatory proteins. However, their sole active role occurs on complete

3259: matching.

3260: 

3261: 

3262: \subsection{The search process for the specific target sequence}

3263: 

3264: \begin{figure}

3265: \includegraphics[width=8cm]{ownmechanisms.eps}

3266: \caption{Schematic of the search mechanisms in equation (\ref{eq:themodel}).}

3267: \label{fig:ownmechanisms}

3268: \end{figure}

3269: 

3270: To find their specific (cognate) binding site along the genome, DNA-binding

3271: proteins such as restriction enzymes or transcription factors have to search

3272: megabases along the DNA molecule. The high accuracy of gene expression control

3273: by binding proteins such as in the $\lambda$-switch requires

3274: a fast search and recognition of the target sequence by the proteins.

3275: A simple 3-dimensional (3D) search of the target sequence by the proteins

3276: is not sufficient to explain experimentally measured

3277: target search rates. It has been suggested relatively early \cite{adam,eigen}

3278: that additional search mechanisms such as 1D sliding along the genome are

3279: needed to account for the actual efficiency of the search process. In their

3280: pioneering work, Berg, von Hippel and coworkers established a statistical

3281: model for target search comprising the four fundamental steps, as shown in

3282: figure \ref{fig:ownmechanisms}: (i) 3D macrohops during which the 

3283: protein fully detaches from the genome until after a volume excursion it

3284: rebinds to the DNA (as a good approximation, the landing site on the DNA

3285: after a macrohop can be assumed to be equidistributed and uncorrelated);

3286: (ii) microhops

3287: during which the protein detaches from the DNA but always stays very close

3288: to it (i.e., the microhop takes place within a cylinder whose radius

3289: corresponds to the escape distance of the protein from

3290: the DNA, see \cite{berg}); (iii) 1D sliding along the genome (while

3291: preserving a certain bonding to the DNA due to nonspecific binding);

3292: and (iv) intersegmental jumps.

3293: The latter are mediated by DNA-loops bringing two chemically remote segments

3294: of the DNA close in Euclidean space, see, for instance, \cite{hame_looping} and

3295: references therein. A protein like Lac repressor, which can establish bonds to

3296: two different stretches of dsDNA simultaneously, can then jump from one to

3297: the neighbouring segment.\footnote{Possibly, also other binding proteins are

3298: able to perform intersegmental jumps.} This process might lead to a

3299: paradoxical diffusion behaviour \cite{paradox,dirk}. However, if the

3300: conformational changes in the DNA

3301: are not too slow, both the bulk mediated macrohops and the intersegmental

3302: transfer lead to fast mixing of the enzymes' positions along the chain (as

3303: it was shown for the related problem in \cite{JLum}), and on the mean-field

3304: level can be described by a desorption followed by the absorption at a

3305: random place.

3306: 

3307: Recently, there has been renewed interest in the targeting problem, both

3308: theoretically (see, for instance, \cite{gerland,slutsky,marko2,coppey}) and

3309: experimentally (e.g., \cite{grillo,record}), including single molecule studies

3310: \cite{shima,pant1,pant2}. Despite the extensive knowledge of specific binding

3311: rates and both specific and non-specific binding free energies, the precise

3312: relative

3313: contributions of the different search mechanisms (and, to some extent, also

3314: the stringent criteria to define these four elementary interactions) are not

3315: fully resolved. Moreover, it has been suggested that under tight(er) binding

3316: conditions, the sliding of the protein becomes subdiffusive due to the local

3317: structure landscape of a heteropolymer DNA \cite{slutsky1}. This complication,

3318: however, is expected to be relaxed in a more loosely bound search mode of the

3319: searching protein \cite{slutsky}. We here adopt the latter view of normal

3320: diffusion, which is corroborated by the experimental study in the next

3321: subsection.

3322: 

3323: 

3324: \subsection{A unique situation: Pure one-dimensional search of SSB mutants}

3325: 

3326: In previous studies, the 1D sliding problem had always been considered

3327: as a problem of 3D diffusion which is enhanced by 1D diffusion. Thus,

3328: workers such as Berg, Winter, and von Hippel \cite{berg} assumed that proteins

3329: nonspecifically bound would on average unbind before finding their

3330: specific binding sites. This results in an enhancement of specific binding

3331: rates that is proportional to the 1D sliding rate, but the overall specific

3332: binding rate depends linearly on protein concentration. These studies neglect

3333: the possibility that the protein finds its specific site before unbinding.

3334: Given the experimental conditions under which transcription factor binding has

3335: been previously studied, this approximation is appropriate. However, as

3336: demonstrated in \cite{igor_prot}, this mechanism, in which the unbinding rate is

3337: much lower than the specific binding rate, occurs for the 1D search of DNA by

3338: the single-stranded DNA binding protein T4 gene 32 protein (gp32). This fast 1D

3339: search rate is essential for gp32 to be able to quickly find specific locations

3340: on DNA molecules that are undergoing replication, and which have large sections

3341: of single-stranded DNA exposed for gp32 binding. The resulting

3342: nonlinear concentration dependence of gp32 binding will likely

3343: have significant effects on gp32's ability to find its replication sites as

3344: well as its ability to recruit other proteins during replication.

3345: If these nonlinear effects also occur for TFs, this

3346: characteristic will strongly affect regulatory processes governed by protein

3347: binding.

3348: 

3349: \begin{figure}

3350: \begin{center}

3351: \includegraphics[width=6.2cm,angle=270]{figure5.eps}

3352: \end{center}

3353: \caption{Dimensional binding rate $k_a$ in 1/s as function of protein

3354: concentration $C$ in M, for parameters corresponding to 100 mM salt. The

3355: fitted 1D diffusion constant

3356: for sliding along the dsDNA is $D_{\mathrm{1d}}=3.3\cdot 10^{-9}\mathrm{cm}^2/

3357: \mathrm{sec}$, located nicely within the experimental value $10^{-8}\ldots

3358: 10^{-9}\mathrm{cm}^2/\mathrm{sec}$ \protect\cite{pant}.}

3359: \label{rates}

3360: \end{figure}

3361: 

3362: Results from the single DNA overstretching experiment are shown in

3363: figure \ref{rates} along with the results from the theoretical and simulations

3364: analysis from references \cite{igor_prot}. The scaling of search

3365: rate as function of concentration is described by the relation

3366: \begin{equation}

3367: k_a=D_{\mathrm{1d}}n_0^2

3368: \end{equation}

3369: obtained for the pure 1D search of random walkers of line density $n_0^2$

3370: searching along the DNA. For low concentrations, the McGhee and von Hippel

3371: isotherm \cite{mcghee} predicts a linear relation between $n_0$ and the

3372: volume concentration $C$; thus, $k_a\propto C^2$. The experimental evidence

3373: for the purely linear search process, as shown in figure \ref{rates} for 100

3374: mM salt, was found for a large range of salt concentrations, see

3375: references \cite{igor_prot,pant2} for details. The case of high line density

3376: of proteins was discussed in \cite{igor_finite}.

3377: 

3378: 

3379: \subsection{L{\'e}vy flights and target search}

3380: 

3381: We now address the general search process with interchange of 1D and 3D

3382: diffusion, and intersegmental jumps. To this end, we first quickly review

3383: the definition of L{\'e}vy flights \cite{klafternature,physicstoday,report,%

3384: report1,coffey}.

3385: 

3386: L{\'e}vy flights (LFs) are random walks whose

3387: jump lengths $x$ are distributed like $\lambda(x)\simeq|x|^{-1-\alpha}$ with

3388: exponent $0<\alpha<2$ \cite{hughes}. Their probability density

3389: to be at position $x$ at time $t$ has the characteristic function

3390: $P(q,t)\equiv\int_{-\infty}^{\infty}e^{iqx}P(x,t)dx=\exp\left(-D_{\mathrm{L}}

3391: |q|^{\alpha}t\right)$, a consequence of the generalised central limit theorem

3392: \cite{levy,gnedenko}; in that sense, LFs are a natural extension of normal

3393: Gaussian diffusion ($\alpha=2$). LFs occur in a wide range of systems

3394: \cite{report}; in particular, they represent an optimal search

3395: mechanism in contrast to locally oversampling Gaussian search

3396: \cite{stanley}. Dynamically, LFs can be described by a space-fractional

3397: diffusion equation $\partial P/\partial t=D_{\mathrm{L}}\partial^{\alpha}

3398: P(x,t)/\partial|x|^{ \alpha}$, a convenient basis to introduce additional

3399: terms, as shown below. $D_{\mathrm{L}}$ is a diffusion constant of

3400: dimension $\mathrm{cm}^{\alpha}/\mathrm{sec}$, and the fractional derivative is

3401: defined via its Fourier transform, $\mathscr{F}\{\partial^\alpha P(x,t)

3402: /\partial |x|^\alpha\}=-|q|^\alpha P(q,t)$ \cite{report,report1,coffey}. LFs

3403: exhibit

3404: superdiffusion in the sense that $\langle|x|^{\zeta}\rangle^{2/\zeta}\simeq

3405: D_{\mathrm{L}}t^{2/\alpha}$ ($0<\zeta<\alpha$) \cite{report}, spreading

3406: faster than the linearly growing mean squared displacement of standard

3407: diffusion ($\alpha=2$).

3408: A prime example of an LF is linear particle diffusion to next neighbour

3409: sites on a fast folding (`annealed') polymer that permits intersegmental

3410: jumps at chain contact points (see figure \ref{fig:ownmechanisms}) caused

3411: by polymer looping \cite{paradox,dirk}. In fact, the contour length $|x|$

3412: stored in a loop between such contact points is distributed in 3D like

3413: $\lambda(x)\simeq|x|^{-1-\alpha}$, where $\alpha=1/2$ for Gaussian chains

3414: ($\theta$ solvent), and $\alpha\approx 1.2$ for self-avoiding walk chains

3415: (good solvent) \cite{duplantier}.

3416: 

3417: In our description of the target

3418: search process, we use the density per length $n(x,t)$ of proteins on

3419: the DNA as the relevant dynamical quantity ($x$ is the distance along

3420: the DNA contour). Apart from intersegmental transfer, we include 1D

3421: sliding along

3422: the DNA with diffusion constant $D_{\mathrm{B}}$, protein dissociation

3423: with rate $k_{\mathrm{off}}$ and (re)adsorbtion with rate $k_{\mathrm{on}}$

3424: from a bath of proteins of concentration $n_{\mathrm{bulk}}$.

3425: The dynamics of $n(x,t)$ is thus governed by the equation

3426: \cite{michael}

3427: \begin{eqnarray}

3428: \nonumber

3429: \frac{\partial}{\partial t}n(x,t)=&&\left(D_{\rm B} \frac{\partial^2}{\partial

3430: x^2}+D_{\rm L} \frac{\partial^\alpha}{\partial |x|^\alpha}-k_{\rm

3431: off}\right)n(x,t)\nonumber\\

3432: &&+k_{\rm on}n_{\rm bulk}-j(t)\delta(x).

3433: \label{eq:themodel}

3434: \end{eqnarray}

3435: Here, $j(t)$ is the flux into the target located at $x=0$. We determine

3436: the flux $j(t)$ by assuming that the target is perfectly absorbing: $n(

3437: 0,t)=0$ \cite{chechkin_bvp}. Be initially the system at equilibrium, except

3438: that the target is

3439: unoccupied; then, the initial protein density is $n_0=n(x,0)=k_{\rm on}

3440: n_{\rm bulk}/k_{\rm off}$.\footnote{Note that the dimension of the on and

3441: off rates differ; while

3442: $[k_{\mathrm{off}}]=\mathrm{sec}^{-1}$, we chose $[k_{\mathrm{on}}]=

3443: \mathrm{cm}^2/\mathrm{sec}$.} The total number

3444: of particles that have arrived at the target up to time $t$ is

3445: $J(t)=\int_0^t d t'\;j(t')$. We derive explicit analytic

3446: expressions for $J(t)$ in different limiting regimes, and study

3447: the general case numerically. We use $J(t)$ to obtain

3448: the mean first arrival time $T$ to the target; in

3449: particular, to find the value of $k_{\rm off}$ that minimises $T$.

3450: 

3451: \begin{figure}

3452: \includegraphics[width=6.8cm]{k_opt.epsi}

3453: \caption{Optimal choice of off rate $k_{\rm off}$ as function of the

3454: LF diffusion constant, from numerical evaluation of the model in reference

3455: \cite{michael}. The circle on the abscissa marks where $k_{\rm off}^{\rm opt}$

3456: becomes 0 in the case $\alpha<1/2$.}

3457: \label{searcheff}

3458: \end{figure}

3459: 

3460: The various regimes of target search embodied in equation (\ref{eq:themodel})

3461: are discussed in detail in references \cite{michael,michael_long}. The main

3462: result for the efficiency of the related search process is summarised in

3463: figure \ref{searcheff}, i.e., which protein unbinding rate $k_{\mathrm{off}}$

3464: optimises the mean search time $T$. Three regimes can be distinguished:

3465: 

3466: (i) Without L{\'e}vy flights, we obtain $k_{\rm off}^{\rm opt}=k_{\rm on}'$:

3467: the proteins should spend equal amounts of time in bulk and on the DNA.

3468: This corresponds to the result obtained for single protein searching on

3469: a long DNA \cite{slutsky,coppey}. 

3470: 

3471: (ii) For $\alpha>1$, i.e., when DNA is in the self-avoiding regime,

3472: we find

3473: \begin{equation}

3474: k_{\rm off}^{\rm opt}\sim (\alpha-1)k_{\rm on}':

3475: \end{equation}

3476: The optimal off rate shrinks linearly with decreasing $\alpha$.

3477: 

3478: (iii) For $\alpha<1$, i.e., when DNA leaves the self-avoiding phase

3479: (e.g., by lowering the temperature or introducing attractive interactions)

3480: the value of $k_{\rm off}^{\rm opt}$ approaches zero as the frequency of

3481: intersegmental jumps ($\propto D_L$) increases: The L{\'e}vy flight

3482: mechanism becomes so efficient that bulk excursions become irrelevant.

3483: At $\alpha=1/2$, the case of the ideal Gaussian chain, we observe a

3484: qualitative change:

3485: When $\alpha<1/2$, the rate $k_{\rm off}^{\rm opt}$ reaches

3486: zero for \emph{finite} values of the rate for intersegmental jumps.

3487: Note that when $\alpha<1$, the spread of the L{\'e}vy flight ($\simeq

3488: t^{1/\alpha}$) grows

3489: faster than the number of sites visited ($\simeq t$), rendering the mixing

3490: effect of bulk excursions insignificant. A scaling argument to understand

3491: the crossover at $\alpha=1/2$ relates the probability density of first

3492: arrival with the width ($\simeq t^{1/\alpha}$) of the Green's function of a

3493: L{\'e}vy flight

3494: $p_{\mathrm{fa}}\simeq t^{-1/\alpha}$. We see that the associated mean

3495: arrival time becomes finite for $0<\alpha<1/2$, even for the infinite chain

3496: limit considered here.

3497: 

3498: We remark that this model is valid for an annealed DNA only. This means

3499: that the chain can equilibrate (at least, locally) on the typical time

3500: scale between intersegmental jumps. Even though real DNA in solution

3501: might not be fully annealed, features of this analysis will reflect on

3502: the target search. A more detailed study of different regimes of DNA

3503: is under way.

3504: 

3505: 

3506: \subsection{Viruses---extreme nanomechanics}

3507: 

3508: Viruses have played an important role in the discovery of the mechanisms

3509: underlying gene regulation, see, for instance, reference \cite{lwoff}. From

3510: a nanoscience perspective, viruses are of interest on their own part.

3511: During the assembly of many viruses, the viral DNA of several $\mu$m length

3512: is packaged into the capsid, the protein container making up most of the virus, 

3513: by a motor protein. This motor packages the DNA by exerting forces of up to

3514: 60 pN or more, causing pressures building up in the capsid of the order

3515: of 6 MPa \cite{smith,zandi}. The size of the capsid spans few tens

3516: of $\mu$m, and is therefore comparable to the persistence length of DNA

3517: \cite{feiss,morita,smith,hud1,hud}. Therefore, fluctuation-based undulations are

3518: suppressed, and the chain can be approximately thought of as being wound

3519: up helically like thread on a bobbin, or like a ball of

3520: yarn. Ultimately, a relatively highly ordered 3D

3521: configuration of the DNA inside the capsid is achieved, which under certain

3522: conditions may even lead to local crystallisation of the DNA

3523: \cite{feiss,morita,pack,hud,kindt,kondev1,ali}.

3524: It is generally argued that this

3525: ordered

3526: arrangement helps to avoid the creation of entanglements or even knots of the

3527: wound-up DNA, thus enabling easy ejection, i.e., release of the DNA once

3528: the phage docks to a new host cell; this ejection is not assisted by the

3529: packaging motor, but it can be facilitated by host cellular DNA polymerase,

3530: which starts to transcribe the DNA and thereby pulls it out of the capsid

3531: \cite{alberts,snustad,kindt,arsuaga}.

3532: Details on the packaging energetics can be found in reference \cite{pack}, and

3533: the works cited therein. Model calculations for the

3534: entropy loss, binding and twist energy, and electrostatic forces that need

3535: to be overcome on packaging reveal, that at higher packaging ratios the

3536: packaging force almost exclusively comes from the electrostatic repulsion.

3537: 

3538: 

3539: \section{Functional molecules and nanosensing}

3540: 

3541: Complex molecules can be endowed with the distinct feature

3542: that they contain subunits which are linked to each other

3543: mechanically rather than chemically \cite{schill}. The investigation of

3544: the structure and properties of such interlocked {\em topological

3545: molecules\/} is

3546: subject of the growing field of chemical topology \cite{frisch};

3547: while speculations about the possibility of catenanes

3548: \footnote{{\em catena\/} (lat.), the chain.} (Olympic rings) date

3549: back to the early 20th century lectures of Willst{\"a}tter, the

3550: actual synthesis of catenanes and rotaxanes

3551: \footnote{{\em rota\/} (lat.), the wheel; {\em axis\/} (lat.),

3552: the axle.} succeeded in 1958 \cite{schill}. Modern organic chemistry

3553: has seen the development of refined synthesis methods to generate topological

3554: molecules.

3555: 

3556: 

3557: \subsection{Functional molecules}

3558: 

3559: In parallel to the miniaturisation in electronics \cite{bishop} and the

3560: possibility of manipulating single (bio)molecules \cite{strick1},

3561: supramolecular chemistry which makes use of chemical topology properties is

3562: coming of age \cite{lehn,dietrich}. Thus, rotaxane-type molecules are believed

3563: to be the building blocks for certain nanoscale machines and motors

3564: \cite{blanco}, so-called hermaphrodite molecules have been shown to perform

3565: linear relative motion (``contraction and stretching'') \cite{jimenez},

3566: and pirouetting molecules have been synthesised \cite{raehm}.

3567: Moreover, topological molecules are thought to become

3568: components for molecular electronics switching devices in memory and

3569: computing applications \cite{pease,lehn1}.

3570: These molecular machines are usually of lower molecular weight, and their

3571: behaviour is essentially energy-dominated in the sense that their

3572: conformations and dynamical properties are governed by external and thermal

3573: activation in an energy landscape. The understanding of the physical

3574: properties and the theoretical modelling of such designer molecules and their

3575: natural biological counterparts has increasingly gained momentum, and the

3576: stage is already set for the next generation of applications

3577: \cite{bishop,strick1,lehn,dietrich,blanco,jimenez,raehm,pease,lehn1,%

3578: montemagno,soong,yurke,metz,motors,frey}.

3579: 

3580: In reference \cite{hame_cpl} we introduced some basic

3581: concepts for functional molecules whose driving force is entropic rather

3582: than energetic, see also the more recent publications in chemistry journals

3583: \cite{functional,functional1}.

3584: Entropy-functional molecules will be of higher molecular weight

3585: (hundred monomers or above)

3586: in order to provide sufficient degrees of freedom such that entropic

3587: effects can determine the behaviour of the molecule.

3588: The potential for such {\em entropy-driven functional molecules\/}

3589: can be anticipated from the classical Gibbs Free energy

3590: \begin{equation} \label{free}

3591: \mathscr{F} = U - T S \, ;

3592: \end{equation}

3593: in functional molecules, $\mathscr{F}$ is minimised mainly by variation of the

3594: internal energy $U$ representing the shape of the energy landscape of

3595: the functional unit. New types of molecules were proposed for which

3596: $\mathscr{F}$ is minimised by variations of the entropy $S$, while the energies

3597: and chemical bondings are left unchanged \cite{hame_cpl}. The entropy-functional

3598: units of such molecules can be specifically controlled by external parameters

3599: like temperature, light flashes, or other electromagnetic fields

3600: \cite{lehn,dietrich}. We note that DNA is already being studied as

3601: a macromolecular prototype

3602: building block for molecular machines \cite{yurke}.

3603: 

3604: \begin{figure}

3605: \includegraphics[width=7.8cm]{fig4.eps}

3606: \caption{Molecular muscle consisting of two interlocked rings

3607: 1 and 2 with attached rod-like molecules. Within this structure, sliding

3608: rings, 3, can be placed, which, if activated, tend to contract the muscle

3609: by entropic forces.}

3610: \label{muscle}

3611: \end{figure}

3612: 

3613: A typical example is the molecule shown in figure \ref{muscle}.

3614: According to the arrangement of the sliding rings 1 and 2, this compound

3615: exhibits the unique feature of a molecule that it can slide {\em laterally\/}.

3616: Suggested as precursors of molecular muscles \cite{jimenez}, this compound

3617: could be propelled with internal entropy-motors, which entropically

3618: adjust the elongation of the muscle. In the configuration shown in

3619: figure \ref{muscle}, the sliding ring 3 creates, if activated, an

3620: entropic force which tends to contract the ``muscle'';

3621: at $T = 300 \, {\mbox K}$ and on a typical scale $x = 10 \, \mbox{nm}$,

3622: the entropic force $k_B T / x$

3623: is of the order of pN, and thus comparable to the force created

3624: in biological muscle cells \cite{gittes}. Molecular muscles of such a make

3625: can be viewed as the nano-counterpart of macroscopic muscle models proposed

3626: by de Gennes \cite{degennesm}, in which the contraction is based on

3627: the entropy difference between the isotropic and nematic phases in liquid

3628: crystalline elastomer films \cite{thomsen}.

3629: 

3630: Similarly, one might speculate

3631: whether the DNA helix-coil transition \cite{poland}

3632: in multiplication setups could be

3633: facilitated in the presence of pre-ring molecules which in vitro attach

3634: to an opened loop of the double strand and close, creating an entropy

3635: pressure which tends to open up the vicinal parts of the DNA which are

3636: still in the helix state. Finally, considering molecular motors,

3637: it would be interesting to design an externally controllable,

3638: purely entropy-driven rotating nanomotor.

3639: 

3640: Numerous additional nanoapplications of biopolymers appear in current

3641: literature. An interesting example is the nanomotor created by a DNA

3642: ring in a periodically driven external field, for instance, a focused

3643: light beam inducing localised temperature variations \cite{kulic_schiessel,%

3644: kulic_schiessel_jpc}. The speeds possibly attained by such a device are

3645: of the order of those reached by biological organisms. Such a nanorotor

3646: could be used to stir smallest volumes in higher viscous environments.

3647: 

3648: 

3649: \subsection{Nanosensing}

3650: 

3651: The advances in minituarization of reactors and devices also brings along

3652: the need of probes, by which smallest volumes can be tested. For instance,

3653: microarrays used in genomics require sensors to detect the presence of

3654: certain proteins (often at small concentrations) in a microdish, without

3655: disturbing the environment in the small volume too much. Similarly, single

3656: molecule experiments require specific local detection possibilities.

3657: 

3658: \begin{figure}

3659: \includegraphics[width=6.8cm]{colourbeacon.eps}

3660: \caption{Molecular beacon based on local DNA denaturation. The green blobs

3661: may represent single-stranded DNA binding proteins, or more specifically

3662: binding proteins binding or other molecule to a custom designed DNA sequence

3663: along the denaturation fork. Bound proteins stabilize the denatured fork

3664: and change the spectrum of the beacon.}

3665: \label{beacon}

3666: \end{figure}

3667: 

3668: A fine example for a potential nanosensore is

3669: the blinking behaviour of a fluorophore-quencher pair mounted on

3670: the denaturation wedge as shown in figure \ref{beacon}. This setup, similar

3671: to the ones described in references \cite{altan,oleg1} works as follows.

3672: As long as the dsDNA is intact, fluorophore and quencher are in close

3673: proximity. Once they come apart from one another when the denaturation

3674: wedge opens up, the incident laser light causes fluorescence of the dye.

3675: The on/off blinking of this "molecular beacon" can be monitored in the

3676: focus of a confocal microscope, or, depending on the intensity of the

3677: emitted light, by a digital camera.

3678: The blinking renders immediate information about the state of the bp,

3679: that is tagged by the dye-quencher pair. Fluorescence, that is, indicates that

3680: the bp is currently broken. It is therefore advantageous to define

3681: the random variable $I(t)$ with the property

3682: \begin{equation}

3683: I(t)=\left\{\begin{array}{ll}

3684: 0 & \mbox{if base-pair at $x=x_T$ is closed}\\

3685: 1 & \mbox{if base-pair at $x=x_T$ is open}\end{array}\right.,

3686: \end{equation}

3687: and in experiments one typically measures the corresponding blinking

3688: autocorrelation function

3689: \begin{equation}

3690: A(t)=\langle I(t)I(0)\rangle-\langle I\rangle_{\mathrm{eq}^2},

3691: \label{A_t}

3692: \end{equation}

3693: where $\langle I\rangle_{\mathrm{eq}}$ is the (ensemble) equilibrium value,

3694: or its spectral decomposition

3695: \begin{equation}

3696: A(t)=\int_0^{\infty}f(\tau)\exp\left(-\frac{t}{\tau}\right)d\tau,

3697: \end{equation}

3698: where

3699: \begin{equation}

3700: f(\tau)=\sum_{p\neq 0}T_p^2\delta\left(\tau-\tau_p\right).

3701: \end{equation}

3702: is called the relaxation time spectrum.

3703: 

3704: Figure \ref{blinkbeac} shows an example for the achievable sensitivity of

3705: such nanobeacons, in an example where the denaturation wedge is in

3706: solution together with a certain concentration (proportional to $\kappa$,

3707: compare Sec. IVG) of selectively single-stranded

3708: DNA binding proteins, as discussed previously. It is distinct how both

3709: measurable signals, $A(t)$ and $f(\tau)$ change with varying SSB-concentration.

3710: \begin{figure}

3711: \includegraphics[width=7.8cm]{f_tau_gamma2.eps}

3712: \includegraphics[width=7.8cm]{A_t_gamma2.eps}

3713: \caption{Spectral response of the denaturation beacon in the presence of

3714: single-stranded DNA binding proteins. Top: Relaxation time spectrum, bottom:

3715: blinking autocorrelation function.}

3716: \label{blinkbeac}

3717: \end{figure}

3718: 

3719: 

3720: \section{Summary}

3721: 

3722: Biopolymers such as DNA, RNA, and proteins are indispensable for their

3723: specificity and robustness in all forms of life. Given their detailed

3724: physical properties such as DNA's persistence length of some 50nm or

3725: its local denaturation in nano-bubbles already at room temperature,

3726: and biochemically relevant interfaces such as 10-20 bps, they 

3727: deeply stretch into the nanoscience domain. This statement is twofold

3728: in the following sense. Firstly, nanotechniques such as atomic force

3729: microscopes become important tools to manipulate and probe biomolecules

3730: and their interaction even on the single molecule level. Secondly,

3731: biomolecules are entering the stage as nanotools such as nanosensors,

3732: functional molecules, or highly sensitive force transducers.

3733: 

3734: The possibility to perform controlled experiments on biomolecules, for

3735: instance, to measure the force-extension curves of single biopolymers,

3736: also opens up novel possibilities to test new physical theories. The

3737: foremost examples may be the exploration of persistence lengths and

3738: other polymer physics properties, and the statistical mechanical

3739: concepts relevant for small system sizes. The latter are known under

3740: the keyword of the Jarzynski relation connecting the non-equilibrium

3741: work performed on a physical system with the difference in the thermodynamic

3742: (i.e., equilibrium) potential between initial and final states

3743: \cite{jarzinsky,udo}.

3744: However, there exist by now several similar theories addressing different

3745: physical quantities, such as the concept of entropy production along a single

3746: particle trajectory \cite{udo1}.

3747: 

3748: This review summaries fundamental physical properties of DNA, and their

3749: relevance for both biological processes and technological applications.

3750: The extensive list of references will be useful for further studies on

3751: specific topics covered herein. We are confident that the role of 

3752: biomolecules in technology, not at least for biomedical applications,

3753: will experience a dramatic increase during the coming years and will

3754: enable us to extend current physical understanding of fundamental

3755: processes.

3756: 

3757: 

3758: \acknowledgments

3759: 

3760: RM and TA acknowledge many helpful and enjoyable discussions with

3761: Jozef Adamcik, Audun Bakk,

3762: Suman Banik, Erika Ercolini, Giovanni Dietler, Hans Fogedby, Yacov Kantor,

3763: Mehran Kardar, Joseph Klafter, Oleg Krichevsky, Michael Lomholt, Maxim

3764: Frank-Kamenetskii, Igor Sokolov, Andrzej Stasiak, Francesco Valle, and

3765: Mark Williams.

3766: RM acknowledges partial funding from the Natural

3767: Sciences and Engineering Research Council (NSERC) of Canada and the Canada

3768: Research Chairs programme. TA acknowledges partial funding from the Wallenberg

3769: foundation.

3770: AH acknowledges funding by the AFOSR (FA9550-05-1-0472) and by the NIH

3771: (SCORE program GM068855-03S1).

3772: SDL acknowledges funding by the Joint DMS/NIGMS

3773: Mathematical Biology Initiative (NIH GM 67242)

3774: and by the National Foundation for Cancer Research

3775: through the Yale-NFCR Center for Protein and Nucleic Acid Chemistry.

3776: 

3777: \clearpage

3778: 

3779: \section*{\large Biographical notes}

3780: 

3781: \begin{minipage}[t]{3.6cm}

3782: \unitlength=1cm

3783: \begin{picture}(4,0)

3784: \put(-5.48,-11.72){\special{psfile=metz.eps vscale=65 hscale=65 voffset=0

3785: hoffset=0}}

3786: \end{picture}

3787: \end{minipage}

3788: \begin{minipage}[t]{4.6cm} \noindent

3789: \textbf{Ralf Metzler} received his doctoral degree in physics from the

3790: University of Ulm, FRG. He then went to Tel Aviv University as Humboldt

3791: Feodor Lynen fellow and later as Minerva Amos de Shalit fellow, to work

3792: with Joseph Klafter. As DFG Emmy Noether fellow, after a period as visiting

3793: scientist at the University of Illinois at Urbana-Champaign with Peter Wolynes,

3794: Ralf moved to the Massachusetts Institute of\\\vspace*{-0.26cm}

3795: \end{minipage}

3796: \noindent Technology (MIT) in Cambridge,

3797: MA, where he worked with Mehran Kardar. In 2002 Ralf was appointed Assistant

3798: Professor at

3799: the Nordic Institute for Theoretical Physics (NORDITA) in Copenhagen,

3800: Denmark. In summer 2006, Ralf assumed his post as

3801: Associate Professor and Canada Research Chair in Biological Physics at

3802: the University of Ottawa, Canada.

3803: Ralf works extensively on biological physics problems and anomalous

3804: stochastic processes. These include DNA physics such as DNA denaturation,

3805: DNA topology and the role of DNA conformations in gene regulation, as well

3806: as anomalous diffusion in biological systems. The latter questions are connected

3807: with L{\'e}vy statistics leading to, respectively, subdiffusion or L{\'e}vy

3808: flights. In his spare time, Ralf enjoys the company of his daughter and wife,

3809: he listens to classical music and is a keen reader of murder mysteries. In

3810: Canada, Ralf is looking forward to nature walks, cross-country skiing and

3811: ice-skating.

3812: 

3813: \hspace{0.2cm}

3814: 

3815: \begin{minipage}[t]{3.6cm}

3816: \unitlength=1cm

3817: \begin{picture}(4,0)

3818: \put(-10.82,-19.78){\special{psfile=tobias.eps vscale=118 hscale=118 voffset=0

3819: hoffset=0}}

3820: \end{picture}

3821: \end{minipage}

3822: \begin{minipage}[t]{4.6cm} \noindent

3823: \textbf{Tobias Ambj{\"o}rnsson} received his PhD from Chalmers and

3824: Gothenburg University, where he worked on the electromagnetic response of

3825: matter. In 2003 he became a NORDITA postdoctoral fellow, working with

3826: Ralf Metzler on the modelling of single biomolecule problems such as

3827: DNA breathing and biopolymer translocation through nanopores. Tobias

3828: re-\\\vspace*{-0.26cm}

3829: \end{minipage}

3830: \noindent

3831: recently received a prestigious fellowship from the Wallenberg foundation,

3832: allowing him to join the group of Robert Silbey at MIT in autumn 2006,

3833: to pursue studies on nanosensors. Tobias enjoys listening to pop music,

3834: travelling, and spending time with friends.

3835: 

3836: \hspace{0.2cm}

3837: 

3838: \begin{minipage}[t]{3.6cm}

3839: \unitlength=1cm

3840: \begin{picture}(4,0)

3841: \put(-5.32,-11.12){\special{psfile=andreas1.eps vscale=64 hscale=64 voffset=0

3842: hoffset=0}}

3843: \end{picture}

3844: \end{minipage}

3845: \begin{minipage}[t]{4.52cm} \noindent

3846: \textbf{Andreas Hanke}

3847: received his doctoral degree in physics from the University of

3848: Wuppertal in Germany, FRG. He then went to the Massachusetts Institute of

3849: Technology (MIT) for his postdoctoral studies with Mehran Kardar.

3850: After a period as visiting scientist with Michael Schick at the University of

3851: Washington, Se\-attle, he moved to a postdoc-\\\vspace*{-0.26cm}

3852: \end{minipage}

3853: \noindent

3854: toral position in John Cardy's

3855: group at the University of Oxford, UK, followed by postdocs with Udo Seifert

3856: in Stuttgart, FRG, and again with John Cardy in Oxford. In 2004 Andreas

3857: became Assistant Professor of Physics at the University of Texas at

3858: Brownsville. He is also Adjunct Assistant Professor at the Department of

3859: Physics at The University of Texas at Dallas and a member of the Institute

3860: of Biomedical Sciences and Technology at UT Dallas. In addition, his

3861: research includes summer appointments at the UT Dallas NanoTech Institute.

3862: Andreas has worked in the fields of mesoscopic quantum systems, soft

3863: condensed matter physics, and biological physics. Currently he is building

3864: up a theory division in Molecular Biophysics and Nanoscience at the

3865: University of Texas at Brownsville. In his spare time, Andreas enjoys

3866: latin and salsa music and dancing, all

3867: kinds of outdoor activities, and excursions to Mexico.

3868: 

3869: \hspace{0.2cm}

3870: 

3871: \begin{minipage}[t]{4.26cm}

3872: \unitlength=1cm

3873: \begin{picture}(4,0)

3874: \put(-7.72,-15.28){\special{psfile=Yongli_passport_photo.ps vscale=92

3875: hscale=92 voffset=0

3876: hoffset=0}}

3877: \end{picture}

3878: \end{minipage}

3879: \begin{minipage}[t]{3.82cm} \noindent

3880: \textbf{Yongli Zhang} is a postdoctoral fellow of the Jane Coffin Childs

3881: Memorial Fund for Medical Research in Carlos Bustamante’s group at University

3882: of California at Berkeley. He received his Ph.D. in Molecular Biophysics

3883: and Biochem-\\[-0.26cm]

3884: \end{minipage}

3885: \noindent

3886: Biochemistry from Yale University in 2003, under supervision of Donald

3887: Crothers. In December 2006, he will move to the Department of Physiology

3888: and Biophysics at Albert Einstein College of Medicine as an assistant

3889: professor. Yongli has been doing both experimental and theoretical

3890: research in DNA mechanics, chromatin dynamics, and mechanism of chromatin

3891: remodeling. His lab will mainly use single-molecule techniques, such as

3892: optical tweezers and atomic force microscopy, to study mechanism of

3893: molecular motors and dynamics of macromolecular assemblies.

3894: 

3895: \hspace{0.2cm}

3896: 

3897: \begin{minipage}[t]{4.26cm}

3898: \unitlength=1cm

3899: \begin{picture}(4,0)

3900: \put(-6.72,-14.48){\special{psfile=stephen.eps vscale=83 hscale=83 voffset=0

3901: hoffset=0}}

3902: \end{picture}

3903: \end{minipage}

3904: \begin{minipage}[t]{3.74cm} \noindent

3905: \textbf{Stephen Levene}

3906: is Associate Professor of Molecular and Cell Biology at

3907: The University of Texas at Dallas. Dr. Levene received his Ph.D. in

3908: Chemistry from Yale University in 1985 and was an American Cancer Society

3909: postdoctoral fellow with Bruno Zimm at University of

3910: California,\\\vspace*{-0.26cm}

3911: \end{minipage}

3912: \noindent

3913: San Diego

3914: until 1989. He then spent one year as a staff scientist at the Human Genome

3915: Center at Lawrence Berkeley Laboratory and was a Program in Mathematics and

3916: Molecular Biology Fellow at University of California, Berkeley in Nicholas

3917: Cozzarelli’s laboratory. Dr. Levene's research interests are in the area of

3918: nucleic-acid structure and flexibility, mechanisms of DNA recombination, the

3919: structural organization of human telomeres, and applications of these areas

3920: to biotechnology. He leads the focus group in Molecular Diagnostics and

3921: Bioimaging in the UT-Dallas Institute of Biomedical Sciences and Technology,

3922: is recipient of an Obermann Interdisciplinary Research Fellowship, a member

3923: of the Biophysical Society, and has served on several NIH study sections.

3924: In addition to scientific pursuits, Dr. Levene is an avid snow skier and

3925: cyclist, having previously competed in both disciplines.

3926: 

3927: \clearpage

3928: 

3929: \begin{appendix}

3930: 

3931: \section*{A polymer primer.}

3932: \label{appA}

3933: 

3934: In this section, we introduce some basic concepts from polymer physics.

3935: Starting from the random walk model, we define the fundamental measures

3936: of a polymer chain, before introducing excluded volume. For more details,

3937: we refer to the monographs \cite{degennes,doi,grosberg,flory}.

3938: 

3939: The simplest polymer model is due to Orr \cite{orr}. It models the polymer

3940: chain a a random walk on a periodic lattice with lattice spacing $a$. Then,

3941: each monomer of index $i$ is characterised by a position vector $\mathbf{R}_i$

3942: with $i=0,1,\ldots,N$. The distance between monomers $i$ and $i+1$ is called

3943: $\mathbf{a}_{i+1}=\mathbf{R}_{i+1}-\mathbf{R}_i$. Consequently, the end-to-end

3944: vector of the polymer is

3945: \begin{equation}

3946: \mathbf{r}=\sum_i\mathbf{a}_i.

3947: \end{equation}

3948: Different $\mathbf{a}_i$ have completely independent orientations, such that

3949: we immediately obtain the average ($\langle\cdot\rangle$ over different

3950: configurations) squared end-to-end distance

3951: \begin{equation}

3952: \mathbf{R}_0^2=\langle\mathbf{r}^2\rangle=\sum_{i,j}\langle\mathbf{a}_i\cdot

3953: \mathbf{a}_j\rangle=\sum_i\langle\mathbf{a}_i^2\rangle=Na^2.

3954: \end{equation}

3955: $R_0\simeq N^{1/2}a$ is a measure for the size of the random walk.

3956: An alternative measure of the size of a polymer chain is provided by its radius

3957: of gyration $R_g$, which may be measured by light scattering experiments. It is

3958: defined by

3959: \begin{equation}

3960: \label{appgyr}

3961: R_g^2=\frac{1}{1+N}\sum_{i=0}^N\langle\left(\mathbf{R}_i-\mathbf{R}_G\right)^2

3962: \rangle,

3963: \end{equation}

3964: and measures the average squared distance to the centre of gravity,

3965: \begin{equation}

3966: \mathbf{R}_G=\frac{1}{1+N}\sum_{i=0}^N\mathbf{R}_i.

3967: \end{equation}

3968: Expression (\ref{appgyr}) can be rewritten as

3969: \begin{equation}

3970: R_g^2=(1+N)^{-2}\sum_{i=0}^{N-1}\sum_{j=i+1}

3971: ^N\langle\left(\mathbf{R}_i-\mathbf{R}_j\right)^2\rangle.

3972: \end{equation}

3973: With $\mathbf{R}_j-\mathbf{R}_i=\sum_{n=i+1}^j\mathbf{a}_n$, one can easily

3974: show that $R_g^2=a^2N(N+2)/[6(N+1)]$. For large $N$, that is, $R_g\simeq

3975: \frac{a^2}{6}N$, and therefore:

3976: \begin{equation}

3977: R_g\sim R_0\sim aN^{1/2}.

3978: \end{equation}

3979: 

3980: On a cubic lattice in $d$ dimensions, each step can go in $2d$ directions,

3981: and for a general lattice, each vector $\mathbf{a}_i$ will have $\mu$ possible

3982: directions. The number of distinct walks with $N$ steps is therefore $\mu^N$.

3983: Denote $\mathfrak{N}_N(\mathbf{r})$ the number of distinct walks with

3984: end-to-end vector $\mathbf{r}$, the probability density function for a

3985: given $\mathbf{r}$ is

3986: \begin{equation}

3987: p(\mathbf{r})=\frac{\mathfrak{N}_N(\mathbf{r})}{\sum_{\mathbf{r}}\mathfrak{N}_N(

3988: \mathbf{r})}.

3989: \end{equation}

3990: For large $N$, due to the independence of individual $\mathbf{a}_i$, this

3991: probability density function will acquire a Gaussian shape,

3992: \begin{equation}

3993: p(\mathbf{r})=\left(\frac{d}{2\pi Na^2}\right)^{d/2}\exp\left(-\frac{dr^2}{

3994: 2Na^2}\right),

3995: \end{equation}

3996: where the normalisation is such that $\langle\mathbf{r}^2\rangle=Na^2$.

3997: From this expression, we can deduce that the number of degrees of freedom

3998: of a closed random walk chain is proportional to $N^{-d/2}$, the entropy

3999: loss suffered by a chain subject to the constraint $\mathbf{r}=0$. On a

4000: general lattice,

4001: \begin{equation}

4002: \omega\simeq \mu^NN^{-d/2}

4003: \end{equation}

4004: with the connectivity constant $\mu$, a measure for in how many different

4005: directions the next bond vector can point ($\mu=2d$ in a cubic lattice).

4006: At fixed end-to-end distance, the entropy of the random walk becomes

4007: $S(\mathbf{r})=S_0-dr^2/(2Na^2)$ where $S_0$ absorbs all constants.

4008: For the free energy $\mathscr{F}(\mathbf{r})=E-k_BTS(\mathbf{r})$ we therefore

4009: obtain

4010: \begin{equation}

4011: \mathscr{F}(\mathbf{r})=\mathscr{F}_0+\frac{dk_BTr^2}{2R_0^2},

4012: \end{equation}

4013: i.e., the random walk likes to coil, the restoring force $-\nabla

4014: \mathscr{F}(\mathbf{r})$ being linear in $\mathbf{r}$. This is often called

4015: the entropic spring character of a Gaussian polymer. Note that the `spring

4016: constant' increases with temperature (`entropy elasticity').

4017: 

4018: In this random walk model of a polymer chain, it is straightforward to define

4019: the persistence length of the chain. By this we mean that successive vectors

4020: $\mathbf{a}_i$ are not independent, but tend to be parallel. Over long

4021: distance, this correlation is lost, and the chain behaves like a random

4022: walk. Due to the quantum chemistry of the monomers, an adjacent pair of

4023: vectors $\mathbf{a}_i,\mathbf{a}_{i+1}$ includes preferred angles, for

4024: carbon chains leading to the trans/gauche configurations. This feature is

4025: captured schematically in the freely rotating chain as depicted in figure

4026: \ref{fjc}.

4027: \begin{figure}

4028: \includegraphics[height=6cm]{f_j_chain.eps}

4029: \caption{Freely jointed chain, in which successive bond vectors include

4030: an angle $\theta$.

4031: \label{fjc}}

4032: \end{figure}

4033: Following \cite{doi}, we can obtain the correlation $\langle\mathbf{a}_n

4034: \cdot\mathbf{a}_m\rangle$ as follows. If we fix all vectors $\mathbf{a}_m,

4035: \ldots,\mathbf{a}_{n-1}$, then the average $\langle\mathbf{a}_n\rangle_{

4036: \mathbf{a}_m,\mathbf{a}_{m+1},\ldots,\mathbf{a}_{n-1}\mbox{ fixed}}=

4037: \mathbf{a}_{n-1}\cos\theta$. Multiplication by $\mathbf{a}_m$ produces

4038: \begin{equation}

4039: \langle\mathbf{a}_m\cdot\mathbf{a}_n\rangle_{\mathbf{a}_m,\ldots,\mathbf{a}

4040: _{n-1}\mbox{ fixed}}=\mathbf{a}_m\cdot\mathbf{a}_{n-1}\cos\theta.

4041: \end{equation}

4042: Averaging

4043: over the $\mathbf{a}_m,\ldots,\mathbf{a}_{n-1}$ leads to the recursion

4044: relation $\langle\mathbf{a}_m\cdot\mathbf{a}_n\rangle=\langle\mathbf{a}_m

4045: \cdot\mathbf{a}_{n-1}\rangle\cos\theta$. With the initial condition

4046: $\langle\mathbf{a}^2\rangle=a^2$, we find

4047: \begin{equation}

4048: \langle\mathbf{a}_m\cdot\mathbf{a}_n\rangle=a^2\cos^{|n-m|}\theta.

4049: \end{equation}

4050: Thus, if $\theta=0$, we obtain a rigid rod behaviour, while for $\theta\neq 0$,

4051: there occurs an exponential decay of the correlation between any two bond

4052: vectors $\mathbf{a}_n$ and $\mathbf{a}_m$. This defines a length scale

4053: \begin{equation}

4054: \ell_p\equiv\frac{a}{\log\cos\theta},

4055: \end{equation}

4056: the `persistence length' of the chain. It diverges for $\theta\to 0$, while

4057: for $\theta=90^\circ$, it vanishes, corresponding to the random walk model

4058: discusses above (`freely jointed chain'). As

4059: \begin{equation}

4060: \sum_{k=-\infty}^{\infty}\langle\mathbf{a}_{n+k}\cdot\mathbf{a}_n\rangle=

4061: a^2\left(1+2\sum_{k=1}^{\infty}\cos^k\theta\right)=a^2\frac{1+\cos\theta}{

4062: 1-\cos\theta},

4063: \end{equation}

4064: we find $R_0^2=a^2N(1+\cos\theta)/(1-\cos\theta)$, i.e., statistically, the

4065: freely jointed chain behaves the same as the random walk chain, but with a

4066: rescaled monomer length. The statistical unit in a polymer chain is often

4067: taken to be the Kuhn length $\ell_K=2\ell_p$.

4068: 

4069: Above chain models are often referred to as being phantom, i.e., the chain

4070: can freely cross itself. A physical polymer possesses an excluded volume

4071: and behaves like a so-called self-avoiding chain. Mathematically, this can

4072: be modelled by self-avoiding walks. To include the major effects, it is

4073: sufficient to follow a simple argument due to Flory. Consider a chain with

4074: unknown radius $R$ and internal monomer concentration $c_{\rm int}\simeq N/

4075: R^d$. Assuming that the self-avoiding character is due to monomer-monomer

4076: interactions, the repulsive energy is proportional to the squared concentration,

4077: i.e.,

4078: \begin{equation}

4079: \mathscr{F}_{\rm rep}=\frac{1}{2}Tv(T)c^2,

4080: \end{equation}

4081: with the excluded volume parameter $v(T)$ ($v(T)\equiv(1-2\chi)a^d$ in

4082: Flory's notation, where the $\theta$ condition $\chi=1/2$ corresponds to

4083: ideal chain behaviour). To obtain the total averaged repulsive energy

4084: $\mathscr{F}_{\rm rep|tot}$, we need to average over $c^2$. In a mean field

4085: approach, we take $\langle c^2\rangle\longrightarrow\langle c\rangle^2\sim

4086: c_{\rm int}^2$. We therefore obtain

4087: \begin{equation}

4088: \mathscr{F}_{\rm rep|tot}\simeq Tv(T)c_{\rm int}^2R^d=Tv(T)\frac{N^2}{R^d},

4089: \end{equation}

4090: favouring large values of $R$. This `swelling' competes with the entropic

4091: elasticity contribution $\mathscr{F}_{\rm el}\simeq TR^2/(Na^2)$. The total

4092: free energy becomes

4093: \begin{equation}

4094: \frac{\mathscr{F}}{T}\simeq v(T)\frac{N^2}{R^d}+\frac{R^2}{Na^2},

4095: \end{equation}

4096: with a minimum at $R_F^{d+2}=v(T)a^2N^3$, so that the Flory radius scales

4097: like

4098: \begin{equation}

4099: R_F\sim AN^{\nu} \,\,, \, \mbox{therefore} \,\, \nu=\frac{3}{2+d} \,.

4100: \end{equation}

4101: The values of the exponent $\nu(d=2)=3/4$ and $\nu(d=3)=3/5$ are extremely

4102: close to the best known values $0.75$ and $0.588$.\footnote{An interesting

4103: discussion about the flaws underlying this reasoning can be

4104: found in reference \protect\cite{degennes}.}

4105: 

4106: 

4107: \subsection*{Polymer networks.}

4108: \label{duplantier}

4109: 

4110: A linear excluded volume polymer chain has the size

4111: \begin{equation}

4112: R_g^2\simeq AN^{2\nu}

4113: \end{equation}

4114: with $\nu=0.75$ in $d=2$, and $\nu=0.588$ in $d=3$. Its number of degrees of freedom is

4115: given in terms of the configuration exponent $\gamma$ such that

4116: \begin{equation}

4117: \omega\simeq\mu^NN^{\gamma-1},

4118: \end{equation}

4119: where $\gamma=1.33$ in $d=2$ and $\gamma=1.16$ in $d=3$.

4120: 

4121: Remarkably, similar critical exponents can be obtained for a general polymer

4122: network of the type shown in figure \ref{network}, as originally

4123: by Duplantier \cite{duplantier,duplantier1},

4124: compare also and in references \cite{ohno,schaefer}: In a network ${\cal G}$

4125: consisting of ${\cal N}$ chain segments of lengths $s_1,\ldots,s_{\cal N}$ and

4126: total length $L=\sum_{i=1}^{\cal N}s_i$, the number of configurations

4127: $\omega_{\cal G}$ scales as

4128: \begin{equation}

4129: \label{network}

4130: \omega_{\cal G}(s_1,\ldots,s_{\cal N})=\mu^{L}s_{\cal N}^{\gamma_{\cal G}-1}

4131: {\cal Y}_{\cal G}\left( \frac{s_1}{s_{\cal N}},\ldots,\frac{s_{{\cal N}-1}}{s_

4132: {\cal N}}\right),

4133: \end{equation}

4134: where ${\cal Y}_{\cal G}$ is a scaling function, and $\mu$ is the effective

4135: connectivity constant for self-avoiding walks. The exponent

4136: $\gamma_{\cal G}$ is given by $\gamma_{\cal G}=1-d\nu{\cal L}+\sum_{N

4137: \ge 1}n_N\sigma_N$, where $\nu$ is the swelling exponent, ${\cal L}$

4138: is the number

4139: of independent loops, $n_N$ is the number of vertices with $N$ outgoing legs,

4140: and $\sigma_N$ is an exponent associated with such a vertex.

4141: In $d=2$, this exponent is given by \cite{duplantier,duplantier1}

4142: \begin{equation}

4143: \label{top_coeff}

4144: \sigma_N=\frac{(2-N)(9N+2)}{64}.

4145: \end{equation}

4146: 

4147: In the dense phase in 2D

4148: \cite{duplantier2,duplantier3,owczarek,owczarek1,duplantier4,duplantier5},

4149: and at the $\Theta$ transition \cite{duplantier6}, analogous results can be

4150: obtained.

4151: 

4152: First, consider the dense phase in 2D. If all segments have equal

4153: length $s$ and $L = {\cal N} s$, the configuration number

4154: $\omega_{\cal G}$ of such a network scales as

4155: \cite{duplantier2,duplantier3}

4156: \footnote{Note that due to the factor $\omega_0(L)$ the exponent of

4157: $s$ is $\gamma_{\cal G}$, and not $\gamma_{\cal G}-1$ like in the

4158: expressions used in the dilute phase \cite{duplantier} or at the

4159: $\Theta$ point, for which  $\omega_0(L) \sim L^{- d \nu}$. However,

4160: for 2D dense polymers one has $d \nu = 1$, so that both definitions of

4161: $\gamma_{\cal G}$ are equivalent, cf.~section 3 in reference

4162: \cite{duplantier3}.}

4163: \begin{equation} \label{noc_md}

4164: \omega_{\cal G}(s) \sim \omega_0(L) \, s^{\gamma_{\cal G}} \, \, ,

4165: \end{equation}

4166: %

4167: where $\omega_0(L)$ is the configuration number of a simple ring of

4168: length $L$. For dense polymers, and in contrast to the dilute phase

4169: or at the $\Theta$ point, $\omega_0(L)$ (and thus $\omega_{\cal G}$)

4170: depends on the boundary conditions and even on the shape of the system

4171: \cite{duplantier3,owczarek,owczarek1,duplantier4,duplantier5}. For example,

4172: for periodic boundary conditions (which

4173: we focus on in this study) corresponding to a 2D torus, one finds

4174: $\omega_0(L) \sim \mu^L \, L^{\Psi - 1}$ with a connectivity

4175: constant $\mu$ and $\Psi = 1$ \cite{duplantier3}. However, the network

4176: exponent

4177: %

4178: \begin{equation} \label{nexp}

4179: \gamma_{\cal G} = 1 - {\cal L} + \sum_{N\ge 1}n_N\sigma_N

4180: \end{equation}

4181: %

4182: is {\em universal\/} and depends only on the topology of the

4183: network by the number ${\cal L}$ of independent loops, and by

4184: the number $n_N$ of vertices of order $N$ with vertex exponents

4185: $\sigma_N = (4 - N^2)/32$ \cite{duplantier2,duplantier3}. For a linear chain,

4186: the corresponding exponent $\gamma_{\rm lin} = 19/16$ has been

4187: verified by numerical simulations \cite{duplantier3,grassberger}. For a network

4188: made up of different segment lengths $\left\{s_i\right\}$ of total

4189: length $L = \sum_{i=1}^{\cal N} s_i$, equation (\ref{noc_md})

4190: generalises to (cf.~section 4 in reference \cite{duplantier3})

4191: %

4192: \begin{equation} \label{noc}

4193: \omega_{\cal G}(s_1, \ldots, s_{\cal N})

4194: \, \sim \, \omega_0(L) \, s_{\cal N}^{\gamma_{\cal G}} \,

4195: {\cal Y}_{\cal G}\left(\frac{s_1}{

4196: s_{\cal N}},\ldots,\frac{s_{{\cal N}-1}}{s_{\cal N}}

4197: \right) \,  ,

4198: \end{equation}

4199: %

4200: which involves the scaling function ${\cal Y}_{\cal G}$.

4201: 

4202: \begin{figure}

4203: \includegraphics[width=6.8cm]{network.eps}

4204: \caption{Polymer network ${\cal G}$ with vertices ($\bullet$) of different

4205: order $N$, where $N$ self-avoiding walks are joined ($n_1=5$, $n_3=4$, $n_4=3$,

4206: $n_5=1$).

4207: \label{netw}}

4208: \end{figure}

4209: For polymers in an infinite volume and endowed with an attractive

4210: interaction between neighbouring monomers, a different scaling behaviour

4211: emerges if the system is not below but right at the $\Theta$ point

4212: \cite{duplantier3}. In this case the number of configurations of a general

4213: network ${\cal G}$ is given by

4214: %

4215: \begin{equation} \label{noc_theta}

4216: \overline{\omega}_{\cal G}(s_1, \ldots, s_{\cal N})

4217: \sim \mu^L \, s_{\cal N}^{\overline{\gamma}_{\cal G} - 1} \,

4218: \overline{{\cal Y}}_{\cal G}\left(\frac{s_1}{s_{\cal N}},

4219: \ldots,\frac{s_{{\cal N}-1}}{s_{\cal N}}

4220: \right),

4221: \end{equation}

4222: %

4223: with the network exponent

4224: \begin{equation} \label{nexp_theta}

4225: \overline{\gamma}_{\cal G} =

4226: 1 - d \nu {\cal L} + \sum_{N\ge 1}n_N\overline{\sigma}_N \, \, .

4227: \end{equation}

4228: %

4229: Overlined symbols refer to polymers at the $\Theta$ point.

4230: In $d = 2$, $\nu = 4/7$ and $\overline{\sigma}_N = (2-N)(2N+1)/42$

4231: \cite{duplantier3}.

4232: 

4233: 

4234: \end{appendix}

4235: 

4236: 

4237: \clearpage

4238: 

4239: \begin{thebibliography}{999}

4240: 

4241: \bibitem{alberts} B.~Alberts, A.~Johnson, J.~Lewis, M.~Raff,

4242: K.~Roberts, and P.~Walter, \emph{Molecular biology of the cell\/}

4243: (Garland, New York, 2002).

4244: 

4245: \bibitem{snustad} D.~P.~Snustad and M.~J.~Simmons,

4246: \emph{Principles of Genetics} (John Wiley \& Sons, New York, 2003).

4247: 

4248: \bibitem{kornberg} A.~Kornberg, \emph{DNA synthesis} (W.~H.~Freeman,

4249: San Francisco, CA, 1974).

4250: 

4251: \bibitem{kornberg1} A.~Kornberg and T.~A.~Baker, \emph{DNA Replication}

4252: (W.~H.~Freeman, New York, 1992).

4253: 

4254: \bibitem{franklin} R.~E.~Franklin and R.~G.~Gosling, \NAT \textbf{171},

4255: 740 (1953).

4256: 

4257: \bibitem{watsoncrick} J.~D.~Watson and F.~H.~C.~Crick, \NAT \textbf{171},

4258: 737 (1953).

4259: 

4260: \bibitem{treloar} L.~R.~G.~Treloar, \emph{The physics of rubber elasticity\/}

4261: (Clarendon Press, Oxford, 1975).

4262: 

4263: \bibitem{crick} F.~Crick, \NAT \textbf{227}, 561 (1970).

4264: 

4265: \bibitem{crick1} F.~C.~H.~Crick, in {\em Symp.~Soc.~Exp.~Biol.,

4266: The Biological Replication of Macromolecules}, XII, 138 (1958).

4267: 

4268: \bibitem{bloomfield} V.~A.~Bloomfield, D.~M.~Crothers, and I.~Tinoco,

4269: {\em Physical Chemistry of Nucleic Acids} (Harper \& Row, New York, 1974).

4270: 

4271: \bibitem{rnaworld} R.~F.~Gesteland and J.~F.~Atkins, editors, {\em The

4272: RNA world} (Cold Spring Harbor Laboratory Press, Cold Spring Harbor,

4273: New York, 1993).

4274: 

4275: \bibitem{frank} M.~D.~Frank-Kamenetskii, {\em

4276: Unraveling DNA: The Most Important Molecule of Life\/} (Perseus,

4277: Cambridge, MA, 1997).

4278: 

4279: \bibitem{frank1} M.~D.~Frank-Kamenetskii, Phys.~Rep.~{\bf 288}, 13 (1997).

4280: 

4281: \bibitem{marko} J.~F.~Marko and E.~D.~Siggia, Macromolecules {\bf 28}, 8759

4282: (1995).

4283: 

4284: \bibitem{marko1} J.~F.~Marko and E.~D.~Siggia, Phys.~Rev.~E {\bf 52}, 2912

4285: (1995).

4286: 

4287: \bibitem{ptashne1} M. Ptashne and A. Gunn, {\em Genes and signals}

4288: (Cold Spring Harbor Laboratory Press, Cold Spring Harbor,

4289: New York, 2002).

4290: 

4291: \bibitem{revet} B. R{\'e}vet, B. v. Wilcken-Bergmann, H. Besset, A. Barker,

4292: and B. M{\"u}ller-Hill, Curr. Biol. {\bf 9}, 151 (1999).

4293: 

4294: \bibitem{bell} C. E. Bell and M. Lewis, J. Mol. Biol. {\bf 314}, 1127 (2001).

4295: 

4296: \bibitem{bell1} C. E. Bell, P. Frescura, A. Hochschild, and M. Lewis, Cell

4297: {\bf 101}, 801 (2000).

4298: 

4299: \bibitem{hame_looping} A. Hanke and R. Metzler, Biophys. J. \textbf{85}, 167

4300: (2003).

4301: 

4302: \bibitem{schiessel} H. Schiessel, J. Phys. Cond Mat. \textbf{15}, R699 (2003).

4303: 

4304: \bibitem{kreth} G. Kreth, J. Finsterle, J. von Hase, M. Cremer, and C. Cremer,

4305: Biophys. J. \textbf{86}, 2803 (2004).

4306: 

4307: \bibitem{goddard} N. L. Goddard, G. Bonnet, O. Krichevsky, and A. Libchaber,

4308: Phys. Rev. Lett. \textbf{85}, 2400 (2000).

4309: 

4310: \bibitem{grosberg} A.~Yu. Grosberg and A.~R.  Khokhlov, {\em Statistical

4311: Mechanics of Macromolecules\/} (AIP Press, New York, 1994).

4312: 

4313: \bibitem{calu} G.~C\u{a}lug\u{a}reanu, Czech.

4314: Math.~J.~{\bf 11}, 588 (1961).

4315: 

4316: \bibitem{white1} J.~H.~White, Am.~J.~Math.~{\bf 91}, 693 (1969).

4317: 

4318: \bibitem{fuller} F.~B.~Fuller, Proc.~Natl.~Acad.~Sci.~USA {\bf

4319: 75}, 3557 (1971).

4320: 

4321: \bibitem{white} J. H. White and W. R. Bauer, J. Mol. Biol. \textbf{189},

4322: 329 (1986).

4323: 

4324: \bibitem{bauer} W. R. Bauer, Ann. Rev. Biophys. Bioeng. \textbf{7},

4325: 287 (1978).

4326: 

4327: \bibitem{sun} H. B. Sun, J. Shen, and H. Yokata, Biophys. J. \textbf{79}, 184

4328: (2000).

4329: 

4330: \bibitem{benham} C.~J.~Benham, Comput.~Appl.~Biosci.~{\bf 12}, 375 (1996).

4331: 

4332: \bibitem{goetze} S.~Goetze, A.~Gluch, C.~Benham, and J.~Bode,

4333: Biochemistry {\bf 42}, 154 (2003).

4334: 

4335: \bibitem{levene} A.~V.~Vologodskii, S.~D.~Levene, K.~V.~Klenin,

4336: M.~Frank-Kamenetskii, and N.~R.~Cozzarelli, J.~Mol.~Biol.~{\bf 227},

4337: 1224 (1992).

4338: 

4339: \bibitem{levene1} H.~Tsen, A. Hanke, and S.~D.~Levene, in preparation

4340: (2006).

4341: 

4342: \bibitem{orland} M.~Pillsbury, H.~Orland, and A.~Zee, \PRE

4343: \textbf{72}, 011911 (2005).

4344: 

4345: \bibitem{orland1} H.~Orland and A.~Zee, Nucl.~Phys.~B {\bf 620}, 456 (2002).

4346: 

4347: \bibitem{baiesips} M.~Baiesi, E.~Orlandini, and A.~L.~Stella, \PRL

4348: \textbf{91}, 198102 (2003).

4349: 

4350: \bibitem{loop1}

4351: R. Schleif, Annu. Rev. Biochem. {\bf 61}, 199 (1992).

4352: 

4353: \bibitem{loop2}

4354: K.~S. Matthews, Microbiol. Rev. {\bf 56} 123 (1992).

4355: 

4356: \bibitem{loop3}

4357: T.~M. Dunn, S. Hahn, S. Ogden, and R.~F. Schleif,

4358: Proc. Nat. Acad. Sci. USA {\bf 81}, 5017 (1984).

4359: 

4360: \bibitem{loop4}

4361: M. Geanacopoulos, G. Vasmatzis, V.~B. Zhurkin, and S. Adhya,

4362: Nat. Struct. Biol. {\bf 8}, 432 (2001).

4363: 

4364: \bibitem{loop5}

4365: M.~C. Mossing and M.~T. Record, Science {\bf 233} 889 (1986).

4366: 

4367: \bibitem{loop6}

4368: P. Valentin-Hansen, B. Albrechtsen, and J.~E. Love Larsen,

4369: Embo J. {\bf 5} 2015 (1986).

4370: 

4371: \bibitem{loop7}

4372: K. Rippe, M. Guthold, P.~H. von Hippel, and C. Bustamante,

4373: J. Mol. Biol. {\bf 270}, 125 (1997).

4374: 

4375: \bibitem{loop8}

4376: K. Ogata, K. Sato, and T. Tahirov,

4377: Curr. Opin. Struct. Biol. {\bf 13}, 40 (2003).

4378: 

4379: \bibitem{loop9}

4380: G.~D. van Duyne, Annu. Rev. Biophys. Biomolec. Struct. {\bf 30}, 87 (2001).

4381: 

4382: \bibitem{loop10}

4383: H. Benjamin and N. Cozzarelli, in

4384: {The Robert A. Welch Foundation Conferences on Chemical Research},

4385: vol. 26, p. 107 (1986).

4386: 

4387: \bibitem{loop11}

4388: H. Tsen and S.~D. Levene, Proc. Natl. Acad. Sci. USA. {\bf 94}, 2817 (1997).

4389: 

4390: \bibitem{loop12}

4391: M.~A. Watson, D.~M. Gowers, and S.~E. Halford, J. Mol. Biol. {\bf 298},

4392: 461 (2000).

4393: 

4394: \bibitem{loop13}

4395: M.~L. Embleton, S.~A. Williams, M.~A. Watson, and

4396: S.~E. Halford, J. Mol. Biol. {\bf 289}, 785 (1999).

4397: 

4398: \bibitem{loop14}

4399: M. Deibert, S. Grazulis, G. Sasnauskas, V. Siksnys, and R. Huber,

4400: Nat. Struct. Biol. {\bf 7}, 792 (2000).

4401: 

4402: \bibitem{loop26}

4403: Y.~L. Zhang and D.~M. Crothers, Biophys. J. {\bf 84}, 136 (2003).

4404: 

4405: \bibitem{loop51}

4406: Y. Zhang, A.~E. McEwen, D.~M. Crothers, and S.~D. Levene,

4407: Biophys. J. {\bf 90}, 1903 (2006).

4408: 

4409: \bibitem{loop15}

4410: E.~D. Ross, P.~R. Hardwidge, and L.~J. Maher,

4411: Mol. Cell. Biol. {\bf 21}, (2001).

4412: 

4413: \bibitem{loop16}

4414: R. Schleif, Trends Genet. {\bf 16}, 559 (2000).

4415: 

4416: \bibitem{loop17}

4417: A. Hochschild, Method Enzymol. {\bf 208}, 343 (1991).

4418: 

4419: \bibitem{loop19}

4420: L. Finzi and J. Gelles, Science {\bf 267}, 378 (1995).

4421: 

4422: \bibitem{loop20}

4423: J. M\"uller, S. Oehler, and B. M\"uller-Hill,

4424: J. Mol. Biol. {\bf 257}, 21 (1996).

4425: 

4426: \bibitem{loop21}

4427: K. Rippe, P.~H. von Hippel, and J. Langowski,

4428: Trends Biochem. Sci. {\bf 20}, 500 (1995).

4429: 

4430: \bibitem{loop22}

4431: K. Rippe, Trends Biochem. Sci. {\bf 26}, 733 (2001).

4432: 

4433: \bibitem{loop23}

4434: D. Shore, J. Langowski, and R.~L. Baldwin,

4435: Proc. Nat. Acad. Sci. USA. {\bf 78}, 4833 (1981).

4436: 

4437: \bibitem{loop24}

4438: D. Shore and R.~L. Baldwin, J. Mol. Biol. {\bf 170}, 957 (1983).

4439: 

4440: \bibitem{loop25}

4441: D.~M. Crothers, J. Drak, J.~D. Kahn, and S.~D. Levene,

4442: Method Enzymol. {\bf 212}, 3 (1992).

4443: 

4444: \bibitem{loop27}

4445: K. Klenin, H. Merlitz, and J. Langowski, Biophys. J. {\bf 74}, 780 (1998).

4446: 

4447: \bibitem{loop28}

4448: A.~V. Vologodskii, V.~V. Anshelevich, A.~V. Lukashin,

4449: and M.~D. Frank-Kamenetskii, Nature {\bf 280}, 294 (1979).

4450: 

4451: \bibitem{loop29}

4452: P.~J. Hagerman, Biopolymers {\bf 24}, 1881 (1985).

4453: 

4454: \bibitem{loop30}

4455: S.~D. Levene and D.~M. Crothers, J. Mol. Biol. {\bf 189}, 61 (1986).

4456: 

4457: \bibitem{loop31}

4458: L.~M. Edelman, R. Cheong, and J.~D. Kahn, Biophys. J. {\bf 84}, 1131 (2003).

4459: 

4460: \bibitem{loop32}

4461: J. Shimada and H. Yamakawa, Macromolecules {\bf 17} 689 (1984).

4462: 

4463: \bibitem{loop33}

4464: A. Bacolla, R. Gellibolian, M. Shimizu, S. Amirhaeri, S. Kang, K. Ohshima,

4465: J.~E. Larson, S.~C. Harvey, B.~D. Stollar, and R.~D. Wells,

4466: J. Biol. Chem. {\bf 272}, 16783 (1997).

4467: 

4468: \bibitem{loop34}

4469: S.~M. Law, G.~R. Bellomy, P.~J. Schlax, and M.~T. Record,

4470: J. Mol. Biol. {\bf 230}, 161 (1993).

4471: 

4472: \bibitem{loop35}

4473: M.~H. Hao and W.~K. Olson, Macromolecules {\bf 22}, 3292 (1989).

4474: 

4475: \bibitem{loop36}

4476: B.~D. Coleman, W.~K. Olson, and D. Swigon,

4477: J. Chem. Phys. {\bf 118}, (2003).

4478: 

4479: \bibitem{loop37}

4480: D. Ming, Y. Kong, M.~A. Lambert, Z. Huang, and J. Ma,

4481: Proc. Natl. Acad. Sci. USA {\bf 99}, 8620 (2002).

4482: 

4483: \bibitem{loop38}

4484: I. Tobias, B.~D. Coleman, and W.~K. Olson,

4485: J. Chem. Phys. {\bf 101}, 10990 (1994).

4486: 

4487: \bibitem{loop39}

4488: A. Balaeff, L. Mahadevan, and K. Schulten,

4489: Phys. Rev. Lett. {\bf 83}, 4900 (1999).

4490: 

4491: %\bibitem{loop40}

4492: %A. Hanke and R. Metzler, Biophys. J. {\bf 85}, 167 (2003).

4493: 

4494: \bibitem{loop41}

4495: J. Yan, R. Kawamura, and J.~F. Marko, Phys. Rev. E {\bf 71}, 061905 (2005).

4496: 

4497: \bibitem{loop42}

4498: S. Blumberg, A.~V. Tkachenko, and J.~C. Meiners,

4499: Biophys. J. {\bf 88}, 1692 (2005).

4500: 

4501: \bibitem{loop43}

4502: A. Balaeff, C.~R. Koudella, L. Mahadevan, and K. Schulten,

4503: Philos. Transact. A {\bf 362}, 1355 (2004).

4504: 

4505: \bibitem{loop44}

4506: D. Shore and R.~L. Baldwin, J. Mol. Biol. {\bf 170}, 983 (1983).

4507: 

4508: \bibitem{loop45}

4509: D.~S. Horowitz and J.~C. Wang, J. Mol. Biol. {\bf 173}, 75 (1984).

4510: 

4511: \bibitem{loop49}

4512: X.~J. Lu and W.~K. Olson, Nucleic Acids Res. {\bf 31}, 5108 (2003).

4513: 

4514: \bibitem{loop50}

4515: K. Klenin and J. Langowski, Biopolymers {\bf 54}, 307 (2000).

4516: 

4517: \bibitem{wassermann2} S.~A. Wassermann,

4518: J.~M. Dungan, and N.~R. Cozzarelli, Science {\bf 229}, 171 (1985).

4519: 

4520: \bibitem{deibler} R.~W. Deibler, S. Rahmati, and

4521: E.~L. Zechiedrich, Genes \& Development {\bf 15}, 748 (2001).

4522: 

4523: \bibitem{dna_topo} N.~R. Cozzarelli and J.~C. Wang, {\em DNA topology and

4524: its biological effects} (Cold Spring

4525: Harbor Laboratory Press, Cold Spring Harbor, NY, 1990).

4526: 

4527: \bibitem{liu} L.~F. Liu, C.-C. Liu, and B.~M. Alberts,

4528: Cell {\bf 19}, 697 (1980).

4529: 

4530: \bibitem{mizuuchi} K. Mizuuchi, L.~M. Fisher, M.~H.~O. Dea,

4531: and M. Gellert, Proc. Natl. Acad. Sci. {\bf 77}, 1847 (1980).

4532: 

4533: \bibitem{pollock} T.~J. Pollock and H.~A. Nash,

4534: J. Mol. Biol. {\bf 170}, 1 (1983).

4535: 

4536: \bibitem{spengler} S.~J. Spengler, A. Stasiak, and

4537: N.~R. Cozzarelli, Cell {\bf 42}, 325 (1985).

4538: 

4539: \bibitem{wassermann} S.~A. Wassermann and

4540: N.~R. Cozzarelli, Science {\bf 232}, 951 (1986).

4541: 

4542: \bibitem{wassermann1} S.~A. Wassermann and

4543: N.~R. Cozzarelli, Proc. Natl. Acad. Sci. USA {\bf 82}, 1079 (1984).

4544: 

4545: \bibitem{viguera} E. Viguera, P. Hernandez, D.~B.

4546: Krimer, A.~S. Boistov, R. Lurz, J.~C. Alonso, and J.~B. Schvartzman,

4547: J. Biol. Chem. {\bf 271}, 22414 (1996).

4548: 

4549: \bibitem{sogo} J. Sogo, A. Stasiak, M.~L. Mart\'{\i}nez-Robles, D.~B. Krimer,

4550: P. H{\'e}rnandez, and J.~B. Schvartzman, J. Mol. Biol. {\bf 286}, 637 (1999).

4551: 

4552: \bibitem{topoknot} F. B. Dean, A. Stasiak, T. Koller, and N. R. Cozzarelli,

4553: J. Biol. Chem. {\bf 260}, 4975 (1985).

4554: 

4555: \bibitem{arsuaga} J. Arsuaga, M. Vazquez, S. Trigueros, D. Sumners,

4556: and J. Roca, Proc. Natl. Acad. Sci. USA {\bf 99}, 5373 (2002).

4557: 

4558: \bibitem{electro} A. V. Vologodskii, N. J. Crisona, B. Laurie, P. Pieranski,

4559: V. Katritch, J. Dubochet, and A. Stasiak, J. Mol. Biol. {\bf 278}, 1 (1998).

4560: 

4561: \bibitem{portugal} J. Portugal and A.

4562: Rodriguez-Campos, Nucleic Acids Res. {\bf 24}, 4890 (1996).

4563: 

4564: \bibitem{rodriguez} A. Rodriguez-Campos, J. Biol.

4565: Chem. {\bf 271}, 14150 (1996).

4566: 

4567: \bibitem{staczek} P. Staczek and N.~P.

4568: Higgins, Mol. Microbiol. {\bf 29}, 1435 (1998).

4569: 

4570: \bibitem{arai} Y. Arai, R. Yasuda, K.-I. Akashi, Y. Harada,

4571: H. Miyata, K. Kinosita Jr., and H. Itoh, Nature {\bf 399}, 446 (1999).

4572: 

4573: \bibitem{pieranski} P. Pieranski, S. Kasas, G.

4574: Dietler, J. Dubochet, and A. Stasiak, New J. Phys. {\bf 3}, 101 (2001).

4575: 

4576: \bibitem{saitta} A.~M. Saitta, P.~D. Soper, E. Wasserman,

4577: and M.~L. Klein, Nature {\bf 399}, 46 (1999).

4578: 

4579: \bibitem{stasiak1} A. Stasiak, A. Dobay, J. Dubochet,

4580: and G. Dietler, Science {\bf 286}, 11a (1999).

4581: 

4582: \bibitem{mcnally} T. McNally, {\em The complete book of fly

4583: fishing\/} (Mountain Press, Camden, ME).

4584: 

4585: \bibitem{rybenkov} V.~V. Rybenkov, C. Ullsperger,

4586: A.~V. Vologodskii, and N.~R. Cozzarelli, Science {\bf 277}, 690 (1997).

4587: 

4588: \bibitem{stasiak} A. Stasiak, Curr. Biol. {\bf 10}, R526

4589: (2000).

4590: 

4591: \bibitem{strick} T.~R. Strick, V.~Croquette, and D.

4592: Bensimon, Nature {\bf 404}, 901 (2000).

4593: 

4594: \bibitem{wang} J.~C. Wang, Annu. Rev. Biochem. {\bf 65}, 635

4595: (1996).

4596: 

4597: \bibitem{berger} J.~M. Berger, S.~J. Gamblin, S.~C. Harrrison, and J.~C. Wang,

4598: Nature {\bf 379}, 225 (1996).

4599: 

4600: \bibitem{wang_to} J.~C. Wang, Quart. Rev. Biophys. {\bf 31}, 107 (1998).

4601: 

4602: \bibitem{euler} L. Euler, {\em Solutio problematis ad geometriam

4603: situs pertinentis} Comment. Academiae Sci. Imp. Petropolitanae {\bf 8}, 128

4604: (1736); English translation: Sci. Amer. {\bf 189}, 66 (1953).

4605: 

4606: \bibitem{kepler} J. Kepler, in W. v. Dyck and M. Caspar

4607: (Editors), {\em Johannes Kepler, Gesammelte Werke\/} (Beck, M{\"u}nchen, 1937).

4608: 

4609: \bibitem{adams} A.~C.~C. Adams, {\em The knot book: an elementary

4610: introduction to the mathematical theory of knots\/} (Freeman, New York, 1994).

4611: 

4612: \bibitem{kauffman} L.~H. Kauffman, {\em Knots and physics\/},

4613: Series on Knots and Everything, Vol. I (World Scientific, Singapore, 1993).

4614: 

4615: \bibitem{reidemeister} K. Reidemeister, {\em

4616: Knotentheorie\/} (Springer, Berlin, 1931) [{\em Knot theory} (BSC Assocs.,

4617: Moscow, Idaho, 1983)].

4618: 

4619: \bibitem{listing} J.~B. Listing, {\em Vorstudien zur Topologie},

4620: G{\"o}ttinger Studien (Vandenhoeck und Ruprecht, G{\"o}ttingen, 1848).

4621: 

4622: \bibitem{thomson} W. Thomson, Lord Kelvin, Phil. Mag. {\bf 34},

4623: 15 (1867).

4624: 

4625: \bibitem{thomson1} W. Thomson, Proc. Roy. Soc. Edinb.

4626: {\bf 27}, 59 (1875-76).

4627: 

4628: \bibitem{tait} P.~G. Tait, Trans. Roy. Soc. Edinb. {\bf 28},

4629: 145 (1876-77).

4630: 

4631: \bibitem{tait1} P.~G. Tait, Trans. Roy. Soc. Edinb.

4632: {\bf 32}, 327 (1883-4).

4633: 

4634: \bibitem{tait2} P.~G. Tait, Trans. Roy. Soc. Edinb. {\bf 33},

4635: 493 (1884-85).

4636: 

4637: \bibitem{tait3} P.~G. Tait, {\em Scientific papers\/}

4638: (Cambridge University Press, London, 1898).

4639: 

4640: \bibitem{kirkman} T.~P. Kirkman, Proc. Royal Soc. Edinb.

4641: {\bf 13, 120}, 363 (1884-5).

4642: 

4643: \bibitem{kirkman1} T.~P. Kirkman, Proc. Royal Soc. Edinb.

4644: {\bf 32}, 483 (1884-85).

4645: 

4646: \bibitem{little}  C.~N. Little, Trans. Connecticut Acad. Sci.

4647: 18, {\bf 7}, 27 (1885).

4648: 

4649: \bibitem{little1} C.~N. Little, Trans. Roy. Soc. Edinb.

4650: {\bf 35}. 771 (1890).

4651: 

4652: \bibitem{degriend} P. van de Griend, in {\em History

4653: and science of knots,} edited by J. C. Turner and P. van de Griend

4654: (World Scientific, Singapore, 1996).

4655: 

4656: \bibitem{volo} A.~V. Vologodskii, A.~V. Lukashin,

4657: M.~D. Frank-Kamenetskii, and V.~V. Anshelvich, Sov. Phys. JETP {\bf 39},

4658: 1059 (1974).

4659: 

4660: \bibitem{frisch} H.~L. Frisch and E. Wassermann,

4661: J. Am. Chem. Soc. {\bf 83}, 3789 (1961).

4662: 

4663: \bibitem{delbrueck} M. Delbr{\"u}ck, in {\em Mathematical

4664: problems in biological sciences} (Proc. Symp. Appl. Math. {\bf 14}, 55 (1962)),

4665: edited by R. E. Bellman.

4666: 

4667: \bibitem{sumners} D.~W. Sumners and S.~G.

4668: Whittington, J. Phys. A {\bf 21}, 1689 (1988).

4669: 

4670: \bibitem{pippenger} N. Pippenger, Discrete Appl. Math.

4671: {\bf 25}, 273 (1989).

4672: 

4673: \bibitem{vanderzande} C. Vanderzande, J. Phys. A {\bf 28},

4674: 3681 (1995).

4675: 

4676: \bibitem{frank3} M.~D. Frank-Kamenetskii,

4677: A.~V. Lukashin, and A.~V. Vologodskii, Nature {\bf 258}, 398 (1975).

4678: 

4679: \bibitem{koniaris} K. Koniaris and

4680: M. Muthukumar, Phys. Rev. Lett. {\bf 66}, 2211 (1991).

4681: 

4682: \bibitem{michels} J.~P.~J. Michels and F.~W.

4683: Wiegel, Proc. Roy. Soc. (London) A {\bf 403}, 269 (1986).

4684: 

4685: \bibitem{michels1} J.~P.~J. Michels and F.~W.

4686: Wiegel, Phys. Lett. {\bf 90A}, 381 (1982).

4687: 

4688: \bibitem{janse} E.~J. Janse van Rensburg

4689: and S.~G. Whittington, J. Phys. A {\bf 24}, 3935 (1991).

4690: 

4691: \bibitem{frank2} M.~D. Frank-Kamenetskii,

4692: A.~V. Lukashin, V.~V. Anshelevich, and A.~V. Vologodskii, J. Biomol. Struct.

4693: Dyn. {\bf 2}, 1005 (1985).

4694: 

4695: \bibitem{grosberg4} A. Yu. Grosberg, E-print cond-mat/0207427 (2002).

4696: 

4697: \bibitem{klenin} K.~V. Klenin, A.~V. Vologodskii,

4698: V.~V. Anshelevich, A.~M. Dykhne, and M.~D. Frank-Kamenetskii, J. Biomol.

4699: Struct. Dyn. {\bf 5}, 1173 (1988).

4700: 

4701: \bibitem{shimamura1}  M.~K. Shimamura and

4702: T. Deguchi, Phys. Lett. A {\bf 274}, 184 (2000).

4703: 

4704: \bibitem{katritch} V. Katritch, W.~K. Olson, A.

4705: Vologodskii, J. Dubochet, and A. Stasiak, Phys. Rev. E {\bf 61}, 5545 (2000).

4706: 

4707: \bibitem{dobay} A. Dobay, P.~E. Sottas, J. Dubochet,

4708: and A. Stasiak, Lett. Math. Phys. {\bf 55}, 239 (2001).

4709: 

4710: \bibitem{degennes} P.~G. de Gennes, {\em Scaling concepts in

4711: polymer physics\/} (Cornell University Press, Ithaca, New York, 1979).

4712: 

4713: \bibitem{duplantier1} B. Duplantier, J. Stat. Phys.

4714: {\bf 54}, 581 (1989).

4715: 

4716: \bibitem{deutsch} J.~M. Deutsch, Phys. Rev. E {\bf 59}, R2539

4717: (1999).

4718: 

4719: \bibitem{paul} P. G. Dommersnes, Y. Kantor, and M. Kardar, Phys. Rev. E

4720: \textbf{66}, 031802 (2002).

4721: 

4722: \bibitem{hughes} B.~D. Hughes, {\em Random Walks and Random

4723: Environments, Volume 1: Random Walks\/} (Oxford University Press, Oxford,

4724: 1995).

4725: 

4726: \bibitem{flory1} P.~J. Flory, {\em Statistical Mechanics of

4727: Chain Molecules} (Oxford University Press, Oxford, 1989).

4728: 

4729: \bibitem{slili2d} R. Metzler, A. Hanke, P.~G.  Dommersnes, Y. Kantor,

4730: and M. Kardar, Phys. Rev. Lett. {\bf 88}, 188101 (2002).

4731: 

4732: \bibitem{paraknot} R. Metzler, A. Hanke, P.~G.

4733: Dommersnes, Y. Kantor, and M. Kardar, Phys. Rev. E {\bf 65}, 061103 (2002).

4734: 

4735: \bibitem{duplantier} B. Duplantier, Phys. Rev. Lett.

4736: {\bf 57}, 941 (1986).

4737: 

4738: \bibitem{kafri} Y. Kafri, D. Mukamel, and L. Peliti, Phys.

4739: Rev. Lett. {\bf 85}, 4988 (2000).

4740: 

4741: \bibitem{kafri1} Y. Kafri, D. Mukamel, and L. Peliti,

4742: Eur. Phys. J. B {\bf 27}, 135 (2002).

4743: 

4744: \bibitem{schaefer} L. Sch{\"a}fer, C. v.

4745: Ferber, U. Lehr, and B. Duplantier, Nucl. Phys. B {\bf 374}, 473 (1992).

4746: 

4747: \bibitem{bennaim} M.~B. Hastings, Z.~A. Daya, E.

4748: Ben-Naim, and R.~E. Ecke, Phys. Rev. E {\bf 66}, 025102(R) (2002).

4749: 

4750: \bibitem{descloizeaux} J. des Cloizeaux, J. Phys.

4751: (France) Lett. {\bf 42}, L433 (1981).

4752: 

4753: \bibitem{dobay1} A. Dobay, J. Dubochet, K. Millett,

4754: P.-E. Sottas, and A. Stasiak, Proc. Natl. Acad. Sci. USA {\bf 100},

4755: 5611 (2003).

4756: 

4757: \bibitem{leguillou} J.~C. LeGuillou and J.

4758: Zinn-Justin, Phys. Rev. B {\bf 21}, 3976 (1989).

4759: 

4760: \bibitem{zinn} J.~C. LeGuillou and J.

4761: Zinn-Justin, J. Physique (France) {\bf 50}, 1365 (1989).

4762: 

4763: \bibitem{janse1} E.~J. Janse van Rensburg,

4764: S.~G. Whittington, and N. Madras, J. Phys. A {\bf 23}, 1589 (1990).

4765: 

4766: \bibitem{li} B. Li, S.~N. Madras, and A.~D. Sokal, J. Stat.

4767: Phys. {\bf 80}, 661 (1995).

4768: 

4769: \bibitem{orlandini} E. Orlandini, M.~C. Tesi, E.~J.

4770: Janse van Rensburg, and S.~G. Whittington, J. Phys. A {\bf 31}, 5953 (1998).

4771: 

4772: \bibitem{valle} E. Ercolini, F. Valle, J. Adamcik, G. Witz, R. Metzler, P.

4773: de los Rios, J. Roca, and G. Dietler, E-print cond-mat/0609084.

4774: 

4775: \bibitem{guttmann} A.~J. Guttmann, J. Phys. A {\bf 22},

4776: 2807 (1989).

4777: 

4778: \bibitem{guitter} E. Guitter and E. Orlandini,

4779: J. Phys. A {\bf 32}, 1359 (1999).

4780: 

4781: \bibitem{quake} S.~R. Quake, Phys. Rev. Lett. {\bf 73}, 3317

4782: (1994).

4783: 

4784: \bibitem{quake1} S.~R. Quake, Phys. Rev. E {\bf 52}, 1176 (1986).

4785: 

4786: \bibitem{grosberg2} A.~Yu. Grosberg, A. Feigel, and

4787: Y. Rabin, Phys. Rev. E {\bf 54}, 6618 (1996).

4788: 

4789: \bibitem{erman} B. Erman and J.~E. Mark, {\em

4790: Structures and properties of rubberlike networks\/} (Oxford, New York, 1997).

4791: 

4792: \bibitem{ferry} J.~D. Ferry, {\em Viscoelastic properties of

4793: polymers\/} (Wiley, New York, 1970).

4794: 

4795: \bibitem{ward} I.~M. Ward and D. Hadley, {\em

4796: An Introduction to the Mechanical Properties of Solid Polymers\/}

4797: (John Wiley and Sons Ltd., New York).

4798: 

4799: \bibitem{degennesma} P.-G. de Gennes, Macromolecules {\bf 17},

4800: 703 (1984).

4801: 

4802: \bibitem{lai1} P.~K. Lai, Y.~J. Sheng, and H.~K. Tsao,

4803: Phys. Rev. Lett. {\bf 87}, 175503 (2001).

4804: 

4805: \bibitem{shimamura} M.~K. Shimamura and

4806: T. Deguchi, Phys. Rev. E {\bf 64}, 020801 (2001).

4807: 

4808: \bibitem{farago} O. Farago, Y. Kantor, and M. Kardar,

4809: Europhys. Lett. {\bf 60}, 53 (2002).

4810: 

4811: \bibitem{pull} R. Metzler, Y. Kantor, and M. Kardar,

4812: Phys. Rev. E {\bf 66}, 022102 (2002).

4813: 

4814: \bibitem{sheng} Y.~J. Sheng, P.~K. Lai, and H.~K. Tsao,

4815: Phys. Rev. E {\bf 61}, 2895 (2000).

4816: 

4817: \bibitem{sheng1} P.~K. Lai, Y.~J. Sheng, and H.~K. Tsao,

4818: Macromol. Theor. Simul. {\bf 9}, 578 (2000).

4819: 

4820: \bibitem{roya} R. Zandi, Y. Kantor, and M. Kardar,

4821: E-print cond-mat/0306587 (2003).

4822: 

4823: \bibitem{marcone} B. Marcone, E. Orlandini, A. L. Stella, and F. Zonta,

4824: J. Phys. A \textbf{38}, L15 (2005).

4825: 

4826: \bibitem{virnau} P. Virnau, Y. Kantor, and M. Kardar, J. Amer. Chem. Soc.

4827: \textbf{127}, 15102 (2005).

4828: 

4829: \bibitem{stella} E. Orlandini, A.~L. Stella, and C.

4830: Vanderzande, Phys. Rev. E {\bf 68}, 031804 (2003).

4831: 

4832: \bibitem{orlanddense} E. Orlandini, A.~L. Stella, and C. Vanderzande,

4833: J. Stat. Phys. {\bf 115}, 681 (2004).

4834: 

4835: \bibitem{vilgis} T.~A. Vilgis, Phys. Rep. {\bf 336}, 167 (2000).

4836: 

4837: \bibitem{kholodenko} A.~L. Kholodenko and T.~A.

4838: Vilgis, Phys. Rep. {\bf 298}, 251 (1998).

4839: 

4840: \bibitem{edwards1} S.~F. Edwards, Proc. Phys. Soc. (London)

4841: {\bf 91}, 513 (1967).

4842: 

4843: \bibitem{edwards2} S.~F. Edwards, J. Phys. A {\bf 1}, 15

4844: (1968).

4845: 

4846: \bibitem{doi} M. Doi and S.~F. Edwards {\em The Theory

4847: of Polymer Dynamics\/} (Clarendon Press, Oxford, 1986).

4848: 

4849: \bibitem{otto} M. Otto and T.~A. Vilgis,

4850: Phys. Rev. Lett. {\bf 80}, 881 (1998).

4851: 

4852: \bibitem{otto1} M. Otto and T.~A. Vilgis, J. Phys. A

4853: {\bf 29}, 3893 (1996).

4854: 

4855: \bibitem{vilgis1} T.~A. Vilgis and M. Otto, Phys.

4856: Rev. E {\bf 56}, R1314 (1997).

4857: 

4858: \bibitem{ferrari} F. Ferrari, H. Kleinert, and I.

4859: Lazzizzera, Int. J. Mod. Phys. B {\bf 14}, 3881 (2000).

4860: 

4861: \bibitem{ferrari1} F. Ferrari, H. Kleinert, and I.

4862: Lazzizzera, Phys. Lett. A {\bf 276}, 31 (2000).

4863: 

4864: \bibitem{grone2} A.~Yu. Grosberg and S. Nechaev,

4865: Adv. Polym. Sci. {\bf 106}, 1 (1993).

4866: 

4867: \bibitem{grone1} A.~Yu. Grosberg and S. Nechaev,

4868: Europhys. Lett. {\bf 20}, 613 (1992).

4869: 

4870: \bibitem{grone} A.~Yu. Grosberg and S. Nechaev,

4871: J. Phys. A {\bf 25}, 4659 (1992a).

4872: 

4873: \bibitem{nechaev} S. Nechaev, in {\em Topological aspects of

4874: low dimensional systems\/}, Lecture notes of Les Houches 1998 summer school;

4875: extended version available as E-print cond-mat/9812205.

4876: 

4877: \bibitem{nechaev1} O. Vasilyev and S. Nechaev,

4878: JETP {\bf 93}, 1119 (2001).

4879: 

4880: \bibitem{craighead} S.~W.~P. Turner, M. Cabodi, and H.~G.

4881: Craighead, Phys. Rev. Lett. {\bf 88}, 128103 (2002).

4882: 

4883: \bibitem{bennaim1} E. Ben-Naim, Z.~A. Daya, P.

4884: Vorobieff, and R.~E. Ecke, Phys. Rev. Lett. {\bf 86}, 1414 (2001).

4885: 

4886: \bibitem{maier} B. Maier and J.~O. R{\"a}dler,

4887: Phys. Rev. Lett. {\bf 82}, 1911 (1999).

4888: 

4889: \bibitem{walter} H. Walter and D.~E. Brooks, FEBS

4890: Lett. {\bf 361}, 135 (1995).

4891: 

4892: \bibitem{vasilevskaya} V.~V. Vasilevskaya, A.~R.

4893: Khokhlov, Y. Matsuzawa, and K. Yoshikawa, J. Chem. Phys. {\bf 102},

4894: 6595 (1995).

4895: 

4896: \bibitem{lerman} L. Lerman, Proc. Natl. Acad. Sci. USA {\bf 68},

4897: 1886 (1971).

4898: 

4899: \bibitem{garel} T. Garel, {\em Remarks on homo- and hetero-polymeric

4900: aspects of protein folding}, E-print cond-mat/0305053 (2003).

4901: 

4902: \bibitem{duplantier2} B. Duplantier, J. Phys. A {\bf 19},

4903: L1009 (1986).

4904: 

4905: \bibitem{duplantier3} B. Duplantier and H. Saleur,

4906: Nucl. Phys. B {\bf 290}, 291 (1987).

4907: 

4908: \bibitem{jacobsen} J.~L. Jacobsen, N. Read, and H.

4909: Saleur, Phys. Rev. Lett. {\bf 90}, 090601 (2003).

4910: 

4911: \bibitem{duplantier4} B. Duplantier, Phys. Rev. Lett.

4912: {\bf 71}, 4274 (1993).

4913: 

4914: \bibitem{owczarek} A.~L. Owczarek, T. Prellberg, and

4915: R. Brak, Phys. Rev. Lett. {\bf 70}, 951 (1993).

4916: 

4917: \bibitem{owczarek1} A.~L. Owczarek, T. Prellberg,

4918: and R. Brak, Phys. Rev. Lett. {\bf 71}, 4275 (1993).

4919: 

4920: \bibitem{duplantier7} B. Duplantier and F.

4921: David, J. Stat. Phys. {\bf 51}, 327 (1988).

4922: 

4923: \bibitem{kondev} J. Kondev and J.~L. Jacobsen,

4924: Phys. Rev. Lett. {\bf 81}, 2922 (1998).

4925: 

4926: \bibitem{fisher} M.~E. Fisher, Physica D {\bf 38}, 112 (1989).

4927: 

4928: \bibitem{seno} F. Orlandini, F. Seno, A.~L. Stella,

4929: and M.~C. Tesi, Phys. Rev. Lett. {\bf 68}, 488 (1992).

4930: 

4931: \bibitem{dekeyser} R. Dekeyser, E. Orlandini, A.~L.

4932: Stella, and M.~C. Tesi, Phys. Rev. E {\bf 52}, 5214 (1995).

4933: 

4934: \bibitem{cardy} J. Cardy, J. Phys. A {\bf 34}, L665 (2001).

4935: 

4936: \bibitem{dense} A. Hanke, R. Metzler, P.~G. Dommersnes,

4937: Y. Kantor, and M. Kardar, Eur. Phys. J. E {\bf 12}, 347 (2003).

4938: 

4939: \bibitem{gambini} R. Gambini and J. Pullin,

4940: {\em Loops, knots, gauge theories and quantum gravity} (Cambridge University

4941: Press, Cambridge, UK, 1996).

4942: 

4943: \bibitem{slili3d} R. Metzler, New J. Phys. {\bf 4}, 91.1 (2002).

4944: 

4945: \bibitem{grosberg1} A.~Yu. Grosberg, Phys. Rev. Lett.

4946: {\bf 85}, 3858 (2000).

4947: 

4948: \bibitem{edwards2a} R.~C. Ball, M.~Doi, S.~F. Edwards,

4949: and M. Warner, Polymer {\bf 22}, 1010 (1981).

4950: 

4951: \bibitem{edwards2d} M. Doi and S.~F. Edwards, J. Chem.

4952: Soc. Farad. Trans. {\bf 274}, 1802 (1978).

4953: 

4954: \bibitem{edwards2c} S.~F. Edwards and T.~A. Vilgis,

4955: Polymer {\bf 27}, 483 (1986).

4956: 

4957: \bibitem{edwards2b} P.~G. Higgs and R.~C. Ball,

4958: Europhys. Lett. {\bf 8}, 357 (1989).

4959: 

4960: \bibitem{sommer} J.~U. Sommer, J. Chem. Phys. {\bf 97}, 5777

4961: (1992).

4962: 

4963: \bibitem{hame_cpl} A. Hanke and R. Metzler, Chem. Phys. Lett. \textbf{359},

4964: 22 (2002).

4965: 

4966: \bibitem{meamb_ctn} R. Metzler and T. Ambj{\"o}rnsson, J. Comp. Theoret.

4967: Nanosc. \textbf{2}, 389 (2005).

4968: 

4969: \bibitem{revzin} A. Revzin, Editor, {\em The biology of

4970: non-specific DNA-protein interactions} (CRC Press, Boca Raton, FL, 1990).

4971: 

4972: \bibitem{delcourt}

4973: S.~G.~Delcourt and  R.D.~Blake, J.~Biol.~Chem.~{\bf 266}, 15160 (1991).

4974: 

4975: \bibitem{wartell} R. M. Wartell and A. S. Benight, Phys. Rep. \textbf{126},

4976: 67 (1985).

4977: 

4978: \bibitem{blake} R.~D.~Blake, R.~D., Bizzaro, J.~W., Blake, J.~D., Day, G.~R.,

4979: Delcourt, S.~G., Knowles, J., Marx, K.~A., \& SantaLucia, J., Jr. (1999)

4980: Bioinf.~{\bf 15}, 370 (1999).

4981: 

4982: \bibitem{blossey} R. Blossey and E. Carlon, Phys. Rev. E \textbf{68}, 061911

4983: (2003).

4984: 

4985: \bibitem{yeramian} E. Yeramian, Gene {\bf 255}, 139 (2000).

4986: 

4987: \bibitem{yeramian1} E. Yeramian, Gene {\bf 255}, 151-168 (2000).

4988: 

4989: \bibitem{carlon} E. Carlon, M. L. Malki, and R. Blossey,

4990: Phys. Rev. Lett. {\bf 94}, 178101 (2005).

4991: 

4992: \bibitem{williams} M. C. Williams, Biophys. Textbook online

4993: {\em Optical tweezers: measuring piconewton forces},

4994: {\tt www.biophysics.org/btol/}.

4995: 

4996: \bibitem{busta} S. B. Smith, Y. J. Cui, and C. Bustamante, Science \textbf{271},

4997: 795 (1996).

4998: 

4999: \bibitem{williams1} M. C. Williams, I. Rouzina, and V. A. Bloomfield,

5000: Accts. Chem. Res. \textbf{35}, 159 (2002).

5001: 

5002: \bibitem{hwa}  T. Hwa, E. Marinari, K. Sneppen, and L.-H. Tang,

5003: Proc. Natl. Acad. Sci. USA {\bf 100}, 4411 (2003).

5004: 

5005: \bibitem{pcrrev} C. W. Dieffenbach and G. S. Dycksler, {\em PCR primer: a

5006: laboratory manual} (Cold Spring Harbor Laboratory Press, Cold Spring Harbor,

5007: NY, 1995).

5008: 

5009: \bibitem{pcr} M. J. McPherson and S. G. M{\o}ller, {\em PCR Basics}

5010: (Springer-Verlag Telos, New York, 2000.

5011: 

5012: \bibitem{gueron} M. Gu{\'e}ron, M. Kochoyan, and

5013: J.~L. Leroy, Nature {\bf 328} 89 (1987).

5014: 

5015: \bibitem{krueger} A.~Krueger, E.~Protozanova, and M.~D.~Frank-Kamenetskii,

5016: Biophys. J., at press. (2006).

5017: 

5018: \bibitem{kittel} C. Kittel, Am. J. Phys. {\bf 37}, 917 (1969).

5019: 

5020: \bibitem{altan} G. Altan-Bonnet, A. Libchaber,

5021: and O. Krichevsky, \PRL {\bf 90}, 138101 (2003).

5022: 

5023: \bibitem{zocchi} Y.~Zeng, A.~Montrichok and G.~Zocchi, J. Mol. Biol.

5024: {\bf 339}, 67 (2004).

5025: 

5026: \bibitem{nmr} M. Gu{\'e}ron, M. Kochoyan, and J. L. Leroy, Nature \textbf{328},

5027: 89 (1987).

5028: 

5029: \bibitem{nmr1} M. Gu{\'e}ron and J. L. Leroy, Methods Enzymol. \textbf{261},

5030: 383 (1995).

5031: 

5032: \bibitem{nmr2} I. M. Russu, Methods Enzymol. \textbf{261}, 383 (1995).

5033: 

5034: \bibitem{russu} C. Chen and I. M. Russu, Biophys. J. \textbf{87}, 2545 (2004).

5035: 

5036: \bibitem{vonhippel} D.~E.~Jensen and P.~H.~von Hippel, J. Biol. Chem.

5037: {\bf 251}, 7198 (1976).

5038: 

5039: \bibitem{rich} R.~L.~Karpel, in reference \protect\cite{revzin}.

5040: 

5041: \bibitem{vonhippel1} D.~E.~Jensen, R.~C.~Kelly and P.~H.~von Hippel, J. Biol.

5042: Chem. {\bf 251}, 7215 (1976).

5043: 

5044: \bibitem{rich1} R.~L.~Karpel, IUBMB Life {\bf 53}, 161 (2002).

5045: 

5046: \bibitem{pant} K.~Pant, R.~L.~Karpel and M.~C.~Williams, J. Mol. Biol. {\bf

5047: 327}, 571 (2003).

5048: 

5049: \bibitem{pant1} K.~Pant, R.~L.~Karpel, I.~Rouzina and M.~C.~Williams, J. Mol.

5050: Biol. {\bf 336}, 851 (2004).

5051: 

5052: \bibitem{pant2} K. Pant, R. L. Karpel, I. Rouzina, and M. C. Williams, J. Mol.

5053: Biol. \textbf{349}, 317 (2005).

5054: 

5055: \bibitem{poland} D. Poland and H.~A. Scheraga,

5056: {\em Theory of Helix-Coil Transitions in Biopolymers\/} (Academic Press,

5057: New York, 1970).

5058: 

5059: \bibitem{cantor} C. R. Cantor and P. R. Schimmel, Biophysical Chemistry

5060: (W. H. Freeman and Company, New York, NY, 1980).

5061: 

5062: \bibitem{richard} C. Richard and A.~J. Guttmann, J. Stat. Phys. {\bf 115},

5063: 925 (2004).

5064: 

5065: \bibitem{hanke} A. Hanke, J. Phys.: Condens. Matter {\bf 17}, S1731 (2005).

5066: 

5067: \bibitem{fixman} M. Fixman and J.~J. Freiere, Biopol. {\bf 16}, 2693 (1977).

5068: 

5069: \bibitem{santalucia} J. SantaLucia Jr., Proc.~Natl.~Acad.~Sci.~{\bf 95},

5070: 1460 (1998).

5071: 

5072: \bibitem{lazurkin}  Yu. S. Lazurkin, M. Frank-Kamenetskii, and

5073: E. N. Trifonov, Biopol. \textbf{9}, 1253 (1970).

5074: 

5075: \bibitem{tobias_prl} T.~Ambj{\"o}rnsson, S.~K.~Banik, O.~Krichevsky, and

5076: R.~Metzler, Phys.~Rev.~Lett. \textbf{97}, 128105 (2006).

5077: 

5078: \bibitem{tobias_long} T.~Ambj{\"o}rnsson, S.~K.~Banik, O.~Krichevsky, and

5079: R.~Metzler, in preparation.

5080: 

5081: \bibitem{garel1} T. Garel, C. Monthus, and H. Orland, Europhys. Lett. {\bf 55},

5082: 132 (2001).

5083: 

5084: \bibitem{hame_comment} A. Hanke and R. Metzler,

5085: Phys. Rev. Lett. {\bf 90}, 159801 (2003).

5086: 

5087: \bibitem{somendra_comment} S.~M. Bhattacharjee, Europhys. Lett. {\bf 57}, 772.

5088: 

5089: \bibitem{landau} L.~D. Landau and E.~M. Lifshitz, {\em Physical Kinetics}

5090: (Butterworth-Heinemann, Oxford, UK, 1981).

5091: 

5092: \bibitem{hame_jpa} A. Hanke and R. Metzler, J. Phys. A \textbf{36}, L473

5093: (2003).

5094: 

5095: \bibitem{tobias_jpc} T. Ambj{\"o}rnsson and R. Metzler, J. Phys. Cond. Mat.

5096: \textbf{17}, S1841 (2005).

5097: 

5098: \bibitem{tobias_pre} T. Ambj{\"o}rnsson and R. Metzler, Phys. Rev. E

5099: \textbf{72}, 030901(R) (2005).

5100: 

5101: \bibitem{tobias_leif} R. Metzler and T. Ambj{\"o}rnsson, J. Biol. Phys.

5102: \textbf{31}, 339 (2005).

5103: 

5104: \bibitem{hansen} T. Novotn{\'y}, J. N. Pedersen, M. S. Hansen, T.

5105: Ambj{\"o}rnsson, and R. Metzler, E-print cond-mat/0610752.

5106: 

5107: \bibitem{vankampen} N.~G.~van Kampen, Stochastic Processes

5108: in Physics and Chemistry (North-Holland, Amsterdam, 1992).

5109: 

5110: \bibitem{oleg1} O.~Krichevsky and G. Bonnet, Rep.~Prog.~Phys.~\textbf{65},

5111: 251 (2002).

5112: 

5113: \bibitem{peyrard} M. Peyrard and A. R. Bishop, Phys. Rev. Lett. \textbf{62},

5114: 2755 (1989).

5115: 

5116: \bibitem{dauxois} T. Dauxois, M. Peyrard, and A. R. Bishop, Phys. Rev. E

5117: \textbf{44}, R44 (1993).

5118: 

5119: \bibitem{peyrard_cm} B. S. Alexandrov, L. T. Wille, K. {\O}. Rasmussen, A. R.

5120: Bishop, and K. B. Blagoev, E-print cont-mat/0601555.

5121: 

5122: \bibitem{campa} A. Campa and A. Giansanti, Phys. Rev. E \textbf{58}, 3585

5123: (1998).

5124: 

5125: \bibitem{the_prl_footnote} S. Ares, N. K. Voulgarakis, {\O}. Rasmussen,

5126: and A. R. Bishop, Phys. Rev. Lett. \textbf{94}, 035504 (2005); cf. reference

5127: [19].

5128: 

5129: \bibitem{peyrard_comment} T. S. van Erp, S. Cuesta-Lopez, J.-G. Hagmann,

5130: and M. Peyrard, Phys. Rev. Lett. \textbf{95}, 218104 (2005).

5131: 

5132: \bibitem{suman} S. K. Banik, T. Ambj{\"o}rnsson, and R. Metzler, Europhys.

5133: Lett. \textbf{71}, 852 (2005).

5134: 

5135: \bibitem{gillespie} D. T. Gillespie, J. Comp. Phys. \textbf{22}, 403 (1976).

5136: 

5137: \bibitem{gillespie1} D. T. Gillespie, J. Phys. Chem. \textbf{81}, 2340 (1977).

5138: 

5139: \bibitem{choi} Ch. H. Choi, G. Kalosakas, K. {\O}. Ramsussen, M. Hiromura, A. R.

5140: Bishop, and A. Usheva, Nucl. Acids Res. \textbf{32}, 1584 (2004).

5141: 

5142: \bibitem{hans} H.~C.~Fogedby and R.~Metzler, in preparation.

5143: 

5144: \bibitem{hans_short} H.~C.~Fogedby and R.~Metzler, E-print cond-mat/0608458.

5145: 

5146: \bibitem{tobias_jpca} T.~Ambj{\"o}rnsson and R. Metzler, J. Phys. Cond. Mat.

5147: \textbf{17}, S4305 (2005).

5148: 

5149: \bibitem{amlomme} T.~Ambj{\"o}rnsson, M. A. Lomholt, and R. Metzler, J. Phys.

5150: Cond. Mat. \textbf{17}, S3945 (2005).

5151: 

5152: \bibitem{tobias_pb} T.~Ambj{\"o}rnsson and R. Metzler, Phys. Biol. \textbf{1},

5153: 77 (2004).

5154: 

5155: \bibitem{lwoff} A.~Lwoff, \emph{Bacteriol.~Rev.\/} \textbf{17}, 269 (1953).

5156: 

5157: \bibitem{ptashne} M.~Ptashne, \emph{A Genetic Switch\/}, 3rd edition

5158: (Cold Spring Harbor Laboratory Press, Cold Spring Harbor, New York, 2004)

5159: 

5160: \bibitem{davidson} E.~H.~Davidson et al., \emph{Science\/} \textbf{295},

5161: 1669 (2002).

5162: 

5163: \bibitem{reichardt} L.~Reichardt and A.~D.~Kaiser, \PNAS \textbf{68},

5164: 2185 (1971).

5165: 

5166: \bibitem{metzi} R.~Metzler, \PRL \textbf{87}, 068103 (2001).

5167: 

5168: \bibitem{mewo} R.~Metzler and P.~G.~Wolynes, \emph{Chem. Phys.} \textbf{284},

5169: 469 (2002).

5170: 

5171: \bibitem{little2} J.~W.~Little, D.~P.~Shipley, and D.~W.~Wert, \emph{EMBO J.}

5172: \textbf{18}, 4299 (1999).

5173: 

5174: \bibitem{rozanov} D.~V.~Rozanov, R.~D'Ari, and S.~P.~Sineoky, \emph{J.

5175: Bacteriol.} \textbf{180}, 6306 (1998).

5176: 

5177: \bibitem{aurell} E.~Aurell and K.~Sneppen, \PRL \textbf{88}, 048101 (2002).

5178: 

5179: \bibitem{aurell1} E.~Aurell, S.~Brown, J.~Johanson, and K.~Sneppen, \PRE

5180: \textbf{65}, 051914 (2002).

5181: 

5182: \bibitem{shea} M.~A.~Shea and G.~K.~Ackers, \emph{J. Mol. Biol.} \textbf{181},

5183: 211 (1985).

5184: 

5185: \bibitem{arkin} A.~Arkin, J.~Ross, and H.~H.~McAdams, \emph{Genetics}

5186: \textbf{149}, 1633 (1998).

5187: 

5188: \bibitem{bakk2} A.~Bakk, R.~Metzler, and K.~Sneppen, \emph{Biophys. J.}

5189: \textbf{86}, 58 (2004).

5190: 

5191: \bibitem{bakk3} A.~Bakk, R.~Metzler, and K.~Sneppen, \emph{Isr. J. Chem.}

5192: \textbf{44}, 309 (2004).

5193: 

5194: \bibitem{berg} O.~G.~Berg, R.~B.~Winter, and P.~H.~von Hippel, Biochem.

5195: \textbf{20}, 6929 (1981).

5196: 

5197: \bibitem{wbvh} R.~B.~Winter, O.~G.~Berg, and P.~H.~von Hippel, \emph{Biochem.}

5198: \textbf{20}, 6961 (1981).

5199: 

5200: \bibitem{bvh} P.~H.~von Hippel and O.~G.~Berg, \PNAS \textbf{83}, 1608 (1986).

5201: 

5202: \bibitem{kaladimos} C.~G.~Kalodimos, N.~Biris, A.~M.~Bonvin, M.~M.~Levandoski,

5203: M.~Guennuegues, R.~Boelens, and R.~Kaptein, \emph{Science\/} \textbf{305},

5204: 350 (2004).

5205: 

5206: \bibitem{lewis} M.~Lewis, G.~Chang, N.~C.~Horton, M.~A.~Kercher, H.~C.~Pace,

5207: M.~A.~Schumacher, R.~G.~Brennan, and P.~Lu, \emph{Science\/} \textbf{271},

5208: 1247 (1996).

5209: 

5210: \bibitem{gerland} U.~Gerland, J.~D.~Moroz, and T.~Hwa, \PNAS \textbf{99}, 12015

5211: (2002).

5212: 

5213: \bibitem{bakk} A.~Bakk and R.~Metzler, \emph{FEBS Lett.} \textbf{563}, 66

5214: (2004).

5215: 

5216: \bibitem{bakk1} A.~Bakk and R.~Metzler, \emph{J. Theor. Biol.} \textbf{231},

5217: 525 (2004).

5218: 

5219: \bibitem{kao} Y.~Kao-Huang, A.~Revzin, A.~P.~Butler, P.~O'Conner, D.~W.~Noble,

5220: and P.~H.~von Hippel, \PNAS \textbf{74}, 4228 (1977).

5221: 

5222: \bibitem{stormo} G.~D.~Stormo, \emph{Bioinformatics\/} \textbf{16}, 16

5223: (2000).

5224: 

5225: \bibitem{adam} G. Adam and M. Delbr{\"{u}}ck, in A. Rich, and N. Davidson,

5226: Eds., Structural Chemistry and Molecular Biology  (W. H. Freeman, San

5227: Francisco, CA, 1968).

5228: 

5229: \bibitem{eigen} P. H. Richter and M. Eigen, Biophys. Chem. \textbf{2}, 255

5230: (1974).

5231: 

5232: \bibitem{paradox} I.~M. Sokolov, J. Mai and A. Blumen,

5233: Phys. Rev. Lett. {\bf 79}, 857 (1997).

5234: 

5235: \bibitem{dirk} D.~Brockmann and T.~Geisel, Phys.~Rev.~Lett.~\textbf{91},

5236: 048303.

5237: 

5238: \bibitem{JLum} I.M. Sokolov, J. Mai, and A. Blumen, J. Luminesc. \textbf{76-77},

5239: 377 (1998).

5240: 

5241: \bibitem{slutsky} M. Slutsky and L. A. Mirny, Biophys. J. {\bf 87}, 4021

5242: (2004).

5243: 

5244: \bibitem{coppey} M. Coppey, O. B{\'e}nichou, R. Voituriez,

5245: and M. Moreau, Biophys. J. {\bf 87}, 1640 (2004).

5246: 

5247: \bibitem{marko2} J. Yan and J.~F. Marko, Phys.~Rev.~Lett.~{\bf 93}, 108108

5248: (2004).

5249: 

5250: \bibitem{grillo} A. O. Grillo, M. P. Brown, and C. A. Royer, J. Mol. Biol.

5251: \textbf{287}, 539 (1999).

5252: 

5253: \bibitem{record} R. S. Spolar and M. T. Record, Science \textbf{263}, 777

5254: (1994).

5255: 

5256: \bibitem{shima} N. Shimamoto, J. Biol. Chem. \textbf{274}, 15293 (1999).

5257: 

5258: \bibitem{slutsky1} M. Slutsky, M. Kardar, and L. A. Mirny, Phys. Rev. E

5259: \textbf{69}, 061903 (2004).

5260: 

5261: \bibitem{igor_prot} I. M. Sokolov, R. Metzler, K. Pant, and M. C. Williams,

5262: Biophys. J. \textbf{89}, 895 (2005).

5263: 

5264: \bibitem{mcghee} J. D. McGhee and P. H. von Hippel, J. Mol. Biol. \textbf{86},

5265: 469 (1974).

5266: 

5267: \bibitem{igor_finite} I. M. Sokolov, R. Metzler, K. Pant, and M. C. Williams,

5268: Phys. Rev. E \textbf{72}, 041102 (2005).

5269: 

5270: \bibitem{klafternature} M. F. Shlesinger, G. M. Zaslavsky, and J. Klafter,

5271: Nature {\bf 363}, 31 (1993).

5272: 

5273: \bibitem{physicstoday} J. Klafter, M. F. Shlesinger and G. Zumofen, Phys.

5274: Today \textbf{49} (2), 33 (1996).

5275: 

5276: \bibitem{report} R. Metzler and J. Klafter, Phys. Rep. \textbf{339}, 1

5277: (2000).

5278: 

5279: \bibitem{report1} R. Metzler and J. Klafter, J. Phys. A \textbf{37}, R161

5280: (2004).

5281: 

5282: \bibitem{coffey} A. V. Chechkin, V. Yu. Gonchar, R. Metzler, and J. Klafter,

5283: Adv. Chem. Phys., at press (2006).

5284: 

5285: \bibitem{stanley} G.~M.~Viswanathan, S.~V.~Buldyrev, S.~Havlin, M.~G.~.E~.~da

5286: Luz, E.~P.~.~Raposo, and H.~E.~Stanley, Nature {\bf 401}, 911 (1999).

5287: 

5288: \bibitem{levy}  P. L{\'{e}}vy, \emph{Th{\'{e}}orie de l'addition des

5289: variables al{\'{e}}atoires\/} (Gauthier-Villars, Paris, 1954).

5290: 

5291: \bibitem{gnedenko}  B. V. Gnedenko and A. N. Kolmogorov \emph{Limit

5292: Distributions for Sums of Random Variables\/} (Addison--Wesley, Reading,

5293: 1954).

5294: 

5295: \bibitem{michael} M.~A.~Lomholt, T.~Ambj{\"o}rnsson, and R.~Metzler,

5296: \PRL \textbf{95}, 260603 (2005).

5297: 

5298: \bibitem{michael_long} M.~A.~Lomholt, T.~Ambj{\"o}rnsson, and R.~Metzler,

5299: in preparation

5300: 

5301: \bibitem{chechkin_bvp} A. V. Chechkin, R. Metzler, J. Klafter, V.~Y. Gonchar,

5302: and L.~V. Tanatarov, J. Phys. A {\bf 36}, L537 (2003).

5303: 

5304: \bibitem{smith}  D. E. Smith, S. J. Tans, S. B. Smith, S. Grimes,

5305: D. L. Anderson, and C. Bustamante, Nature \textbf{413}, 748 (2001).

5306: 

5307: \bibitem{zandi} R. Zandi and D. Reguera, Phys. Rev. E \textbf{}, (2005).

5308: 

5309: \bibitem{feiss} C. E. Catalano, D. Cue, and M. Feiss, Mol. Microbiol.

5310: \textbf{16}, 1075 (1995).

5311: 

5312: \bibitem{morita} H. Fujisawa and M. Morita, Genes to Cells \textbf{2},

5313: 537 (1997).

5314: 

5315: \bibitem{hud1} N. V. Hud and K. H. Downing, Proc. Natl. Acad. Sci. USA

5316: \textbf{98}, 14925 (2001).

5317: 

5318: \bibitem{hud} N. V. Hud, Biophys. J. \textbf{69}, 1355 (1995).

5319: 

5320: \bibitem{pack} M. E. Cerritelli, N. Cheng, A. H. Rosenberg, C. E.

5321: McPherson, F. P. Booy, and A. C. Steven, Cell \textbf{17}, 271 (1997).

5322: 

5323: \bibitem{kindt} J. Kindt, S. Tzlil, A. Ben-Shaul, and W. M. Gelbert,

5324: Proc. Natl. Acad. Sci. USA \textbf{98}, 13671 (2001).

5325: 

5326: \bibitem{kondev1} P. K. Purohit, M. M. Inamdar, P. D. Grayson, T. M. Squires,

5327: J. Kondev, and R. Phillips, Biophys. J. \textbf{88}, 851 (2005).

5328: 

5329: \bibitem{ali} I. Ali, D. Marenduzzo, and J. M. Yeomans, J. Chem. Phys.

5330: \textbf{121}, 8635 (2004).

5331: 

5332: \bibitem{schill} G. Schill, Catenanes, rotaxanes and knots,

5333: Academic Press, New York, 1971.

5334: 

5335: \bibitem{bishop} D. Bishop, P. Gammel, C. R. Giles, Phys. Today 54(10) (2001).

5336: 

5337: \bibitem{strick1} T. Strick, J.-F. Allemand, V. Croquette, D. Bensimon,

5338: Phys. Today 54(10) (2001) 46.

5339: 

5340: \bibitem{lehn} J.-M. Lehn, Supramolecular Chemistry, VCH, Weinheim, 1995.

5341: 

5342: \bibitem{dietrich} J.-P. Sauvage and C. Dietrich-Buchecker (editors),

5343: Molecular catenanes, rotaxanes, and knots: a journey

5344: through the world of molecular topology, Wiley-VCH, Weinheim, 1999.

5345: 

5346: \bibitem{blanco} M.-J. Blanco, M. C. Jim{\'e}nez, J.-C. Chambron, V. Heitz,

5347: M. Linke, J.-P. Sauvage, Chem. Soc. Rev. 28 (1999) 293.

5348: 

5349: \bibitem{jimenez} M. C. Jim{\'e}nez, C. Dietrich-Buchecker, J.-P. Sauvage,

5350: Angew. Chem. Int. Ed. 39 (2000) 3284.

5351: 

5352: \bibitem{raehm} L. Raehm, J.-M. Kern, J.-P. Sauvage, Chem. Eur. J. 5 (1999)

5353: 3310.

5354: 

5355: \bibitem{pease} A. R. Pease, J. O. Jeppesen, J. F. Stoddart, Y. Luo,

5356: C. P. Collier, J. R. Heath, Acc. Chem. Res. 34 (2001) 433.

5357: 

5358: \bibitem{lehn1} J.-M. Lehn, Chem. Eur. J. 6 (2000) 2097.

5359: 

5360: \bibitem{montemagno} C. Montemagno, G. Bachand, Nanotechnology 10 (1999) 225.

5361: 

5362: \bibitem{soong}

5363: R. K. Soong, G. D. Bachand, H. P. Neves, A. G. Olkhovets,

5364: H. G. Craighead, C. D. Montemagno, Science 290 (2000) 1555.

5365: 

5366: \bibitem{yurke} B. Yurke, A. J. Turberfield, A. P. Mills jr., F. C. Simmel,

5367: J. L. Neumann, Nature 406 (2000) 605.

5368: 

5369: \bibitem{metz} R. Metzler, Phys. Rev. E 63 (2001) 12103.

5370: 

5371: \bibitem{motors} A. Ajdari, F. J{\"u}licher, J. Prost, Rev. Mod. Phys.

5372: 69 (1998) 1269.

5373: 

5374: \bibitem{frey} A. Vilfan, E. Frey, F. Schwabl, M. Thormahlen, Y. H. Song,

5375: E. Mandelkow, J. Mol. Biol. 312 (2001) 1011.

5376: 

5377: \bibitem{functional} G. Bottari, F. Dehez, D. A. Lehigh, P. J. Nash, E. M.

5378: Perez, J. K. Y. Wong, and F. Zerbetto, Angew. Chemie Intl. Ed. \textbf{42},

5379: 5886 (2003).

5380: 

5381: \bibitem{functional1} S. Garaudee, S. Silvi, M. Venturi, A. Credi, A. H.

5382: Flodd, and J. F. Stoddart, Chem. Phys. Chem. \textbf{6}, 2145 (2005).

5383: 

5384: \bibitem{gittes} F. Gittes, E. Meyhofer, S. Baek, J. Howard, Biophys. J.

5385: 70 (1996) 418.

5386: 

5387: \bibitem{degennesm} P.-G. de Gennes, Comptes Rendus Acad. Sci. Paris S{\'e}r.

5388: II 324 (1997) 343.

5389: 

5390: \bibitem{thomsen} D. L. Thomsen, P. Keller, J. Naciri, R. Pink, H. Jeon,

5391: D. Shenoy, B. R. Ratna, Macromol. 34 (2001) 5868.

5392: 

5393: \bibitem{kulic_schiessel} I. M. Kulic, R. Thaokar, and H. Schiessel, Europhys.

5394: Lett. \textbf{72}, 527 (2005).

5395: 

5396: \bibitem{kulic_schiessel_jpc} I. M. Kulic, R. Thaokar, and H. Schiessel,

5397: J. Phys. Cond. Mat. \textbf{17}, S3965 (2005).

5398: 

5399: \bibitem{jarzinsky} C. Jarzinsky, Phys. Rev. Lett. \textbf{78}, 2690 (1997).

5400: 

5401: \bibitem{udo} O. Braun, A. Hanke, and U. Seifert, Phys. Rev. Lett. \textbf{93},

5402: 158105 (2004).

5403: 

5404: \bibitem{udo1} U. Seifert, Phys. Rev. Lett. \textbf{95}, 040602 (2005).

5405: 

5406: \bibitem{flory}  P.~J. Flory, {\em Principles of Polymer

5407: Chemistry\/} (Cornell University Press, Ithaca, New York, 1953).

5408: 

5409: \bibitem{orr} W.~J. Orr, Trans. Faraday Soc. {\bf 43}, 12 (1947).

5410: 

5411: \bibitem{ohno} K. Ohno and K. Binder, J. Phys.

5412: (Paris) {\bf 49}, 1329 (1988).

5413: 

5414: \bibitem{duplantier5} B. Duplantier and F. David, J. Stat.

5415: Phys. {\bf 51}, 327 (1988).

5416: 

5417: \bibitem{duplantier6} B. Duplantier and H. Saleur, Phys.

5418: Rev. Lett. {\bf 59}, 539 (1987).

5419: 

5420: \bibitem{grassberger} P. Grassberger and R. Hegger, Ann. Phys. {\bf 4}, 230

5421: (1995).

5422: 

5423: \end{thebibliography}

5424: 

5425: 

5426: \end{document}

5427: