1: \documentstyle[multicol,aps,prl,epsf,psfig]{revtex}

2: \begin{document}

3: \def\Dscr{{\cal D}}

4: \def\DsC{{\cal C}}

5: \def\DsS{{\cal S}}

6: \def\DsN{{\cal N}}

7: \def\DsE{{\cal E}}

8: \renewcommand{\textfraction}{0.0}

9: \renewcommand{\topfraction}{1}

10: \renewcommand{\bottomfraction}{1}

11: \setcounter{topnumber}{50}

12: \setcounter{bottomnumber}{50}

13: \setcounter{totalnumber}{50}

14: \setlength{\floatsep}{\baselineskip}

15: \setlength{\textfloatsep}{\baselineskip}

16: \renewcommand{\thefigure}{\arabic{figure}}

17: %

18: %\def\baselinestretch{1.3} 

19: \topmargin = +.1 in 

20: %\textheight 8.75 in 

21: %\oddsidemargin = 0.2in 

22: %\textwidth 450 pt 

23: %

24: 

25: \title{Towards A Consistent Modeling Of Protein

26: Thermodynamic And Kinetic Cooperativity:

27: How Applicable Is The Transition State Picture To

28: Folding and Unfolding?}

29: 

30: \author{H\"useyin KAYA  and Hue Sun CHAN}

31: \address{Department of Biochemistry, and \\

32: Department of Medical Genetics \& Microbiology \\

33: Faculty of Medicine, University of Toronto \\

34: Toronto, Ontario M5S 1A8, Canada}

35: 

36: \maketitle

37: 

38: \begin{abstract}

39: 

40: To what extent do general features of folding/unfolding kinetics of small 

41: globular proteins follow from their thermodynamic properties? To address

42: this question, we investigate a new simplifed protein chain model

43: that embodies a cooperative interplay between local conformational 

44: preferences and hydrophobic burial. The present four-helix-bundle 55mer 

45: model exhibits proteinlike calorimetric two-state cooperativity. It 

46: rationalizes native-state hydrogen exchange observations. Our

47: analysis indicates that a coherent, self-consistent physical account of 

48: both the thermodynamic and kinetic properties of the model leads naturally

49: to the concept of a native state ensemble that encompasses considerable 

50: confomational fluctuations. Such a multiple-conformation native state 

51: is seen to involve conformational states similar to those revealed by 

52: native-state hydrogen exchange. Many of these conformational 

53: states are predicted to lie below native baselines commonly used in 

54: interpreting calorimetric data.  Folding and unfolding kinetics are 

55: studied under a range of intrachain interaction strengths as in 

56: experimental chevron plots. Kinetically determined transition midpoints 

57: match well with their thermodynamic counterparts.

58: Kinetic relaxations are found to be essentially single exponential 

59: over an extended range of model interaction strengths. This includes the 

60: entire unfolding regime and a significant part of a folding regime 

61: with a chevron rollover, as has been observed for real proteins that

62: fold with non-two-state kinetics. The transition state picture of

63: protein folding and unfolding is evaluated by comparing 

64: thermodynamic free energy profiles with actual kinetic rates.

65: These analyses suggest that some chevron rollovers may arise from an 

66: internal frictional effect that increasingly impedes chain motions with 

67: more native conditions, rather than being caused by discrete deadtime folding 

68: intermediates or shifts of the transition state peak as previously posited.

69: \\

70: 

71: \noindent

72: {\bf Running title:} Transition State Picture of Protein Folding

73: 

74: \noindent {\bf Key words:} 

75: calorimetric cooperativity / single-exponential kinetics / rugged landscape /

76: unfolding / chevron plot / four-helix bundle /

77: heat capacity / lattice protein models 

78: 

79: \end{abstract}

80: 

81: \begin{multicols}{2}

82: 

83: \section{INTRODUCTION}

84: 

85: Our physical knowledge of protein folding is measured by the 

86: extent to which current ideas of elemental polypeptide interactions 

87: are capable of reproducing experimental data. 

88: Tremendous experimental progress has been made in the

89: recent past.$^{1-6}$ During the same time, 

90: investigations of simplified self-contained polymer models 

91: have contributed much physical insight.$^{7-16}$

92: These models are successful in physically rationalizing many general 

93: features of proteins in terms of polymer

94: properties, building a foundation for future advances. To 

95: move forward in theoretical development, it is necessary to 

96: recognize what common protein properties have not been predicted by models 

97: to date and target our efforts towards rectifying such deficiencies.

98: A prime example is the current lack of chain models that can quantitatively 

99: reflect the extreme kinetic and thermodynamic cooperativity of small 

100: single-domain proteins.$^{5,17}$ This highlights 

101: our insufficient understanding of protein energetics even at a 

102: ``big-picture'' level, suggesting that as heteropolymers natural proteins 

103: may be quite special in some respects.

104: 

105: We have recently investigated the severe constraints imposed on protein polymer

106: models by the experimental observations of calorimetric and other hallmarks

107: of thermodynamic two-state cooperativity.$^{18-20}$ 

108: These cooperativity requirements appear to be more stringent than

109: most other generic protein properties studied so far. A variety of flexible 

110: heteropolymer models with additive residue-based contact energies 

111: are able to explain significant aspects of the folding process.$^{7-16}$

112: But such additive models --- at least for the

113: several examples evaluated to date --- are found to be insufficient to 

114: satisfy the thermodynamic cooperativity requirements, even though deviations 

115: from proteinlike thermodynamics can be lessened in some instances by 

116: enhancing interaction heterogeneity through using larger numbers of letters 

117: and repulsive energies in model alphabets.$^{19}$ As far

118: as thermodynamic cooperativity is concerned, in scenarios 

119: tested thus far, we find that proteinlike thermodynamics can arise from 

120: nonadditive cooperative contributions that originate from an interplay 

121: between local conformational preferences and (mostly nonlocal) interactions 

122: that favor formation of protein cores.$^{18,20}$

123: 

124: In the folding literature, an intimate correspondence between protein 

125: thermodynamics and kinetics has figured prominently in 

126: theoretical,$^{7,21-23}$ modeling,$^{24-27}$

127: and interpretative$^{2,4}$ discourses. Therefore, we ask:

128: Given that a heteropolymer model has already been constrained to satisfy

129: a set of proteinlike thermodynamic properties, to what extent proteinlike 

130: kinetic behavior would follow automatically? For instance, would

131: such a model be sufficient for two-state kinetics as observed for 

132: many small single-domain proteins$^{5}$? More generally, 

133: what improvement in proteinlike kinetics would such a model enjoy over other 

134: models that are now known to be thermodynamically less cooperative? 

135: 

136: In analyses of folding/unfolding kinetics experiments, free energy profiles 

137: are used extensively to provide useful insight$^{2,4,28-33}$

138: and as a picturesque device to summarize data.  However, 

139: other than the folding free energy and rate measurements themselves, 

140: independent experimental techniques to accurately define and determine such 

141: profiles are currently lacking. Moreover, many of these profiles were

142: proposed without considering explicit chain representations. Therefore, the 

143: applicability and generality of their implied physical pictures remain to be 

144: ascertained. Coarse-grained protein chain models are well suited to shed light 

145: on this fundamental issue because they allow for broad conformational sampling.

146: Free energy profiles in coarse-grained models can be 

147: obtained directly from chain population distributions, without regard to 

148: (and therefore independent of) kinetic rates. It follows that a rigorous 

149: evaluation of the applicability of transition state theory to protein 

150: folding can be conducted by comparing the transition-state-predicted 

151: rates and the actual kinetic rates in these models. We study one such model 

152: below.

153: 

154: \section{\bf A MODEL FOR THERMODYNAMIC COOPERATIVITY}

155: 

156: \noindent

157: The present analysis is based on a thermodynamically cooperative 

158: 55mer lattice protein model that folds to a ground state with

159: a four-helix core (Fig.~1). The intrachain interaction

160: scheme includes additive 5-letter contact energies,$^{20,34}$ 

161: repulsive interactions disfavoring left-handed helices and sharp turns at the 

162: end of a helix, as well as cooperative ``native hydrogen bond burial'' 

163: terms$^{20}$ (c.f. refs.~35, 36). 

164: The total energy $E$ is defined by Eq.~(1) in ref.~20.

165: Although ``native-centric'' interactions were introduced to enhance 

166: thermodynamic cooperativity in the present model, unlike the usual G\=o 

167: construction, they are not necessary for recognizing the ground state.

168: This is because the general, non-native-centric terms in the model 

169: (all terms in $E$ except the ``native hydrogen bond burial'' terms) are

170: sufficient to provide global favorability to the proteinlike 

171: four-helix ground state. We note that several other studies$^{36-39}$ 

172: have also emphasized cooperative interactions in protein folding; and

173: nonadditive aspects of hydrophobic effects are being explored.$^{40-45}$ 

174: As we have emphasized,$^{20}$

175: although the present model is useful for exploring the issues at hand,

176: it should be regarded as tentative, partly because it does not provide an

177: explicit account of other possible physical origins of protein 

178: thermodynamic cooperativity such as sidechain packing.$^{19,46,47}$

179: Furthermore, in view of the current lack of definitive understanding of 

180: hydrogen bonding energetics (see discussion and references in ref.~20), 

181: the cooperative ``native hydrogen bond burial'' energy in the present model 

182: should be broadly interpreted as representative of a general favorable 

183: coupling between local conformational preference and formation of proximate 

184: tertiary contacts, the physical mechanisms of which remain to be

185: further elucidated.

186: 

187: 

188: \begin{figure}

189: \begin{center}

190: \leavevmode

191: \psfig{figure=kayajmb1.eps,width=8cm,height=8cm,angle=0}

192: \end{center}

193: \narrowtext   

194: \caption{Thermodynamic stabilities and definitions of native

195: and denatured states. One of the eight iso-energetic ground-state

196: conformations is shown in the inset, where black beads denote

197: nominally hydrophobic residues.$^{20}$ Free energies of

198: unfolding $\Delta G=k_B T\ln(P_{\rm N}/P_{\rm D})$.

199: Solid curves (labeled by $E_t$) classify conformations with

200: $E\le E_t$ and $E>E_t$ as native and denatured, respectively.

201: Dashed curves show the free energy of denatured conformations,

202: defined as those with $E$ greater than the values shown, relative

203: to the ground-state-only native state with $E_t=-52.04$. All results

204: presented in this paper were obtained using model energetic parameters

205: $\gamma_{\rm lh}=6.0$, $\gamma_{\rm st}=5.0$, ${\cal E}_{\rm Helix}=0$,

206: $b{\cal E}_{\rm Hb}=-0.8$, and $b=1.5$ as specified ref.~20.}

207: \label{step}

208: \end{figure}

209: 

210: In addition to the formulation in ref.~20,

211: here we introduce a parameter $\epsilon$ to model protein behaviors at 

212: different intrachain interaction strengths, such that the {\it effective}

213: energy of a conformation with energy $E$ is equal to $-\epsilon E$, hence

214: its Boltzmann weight equals $\exp(\epsilon E/k_B T)$, where $k_BT$ is

215: Boltzmann's constant times absolute temperature. It follows that

216: the partition function $Q=\sum_{E}g(E)\exp(\epsilon E/k_B T)$, 

217: where $g(E)$ is the number of conformations with 

218: energy $E$, and $g(E)$ is estimated by a parameter-space Monte Carlo 

219: histogram technique.$^{20}$

220: The formulation in ref.~20 corresponds 

221: to $\epsilon=-1$. Because of the peculiar and significant

222: temperature dependence of the solvent-mediated interactions in real proteins,

223: varying $-\epsilon/k_B T$ serves better as a model for how

224: effective intrachain interactions are modulated at constant

225: temperature by denaturant concentration$^{48}$ or denaturant activity$^{37}$ 

226: than for how Boltzmann weights changes with temperature.$^{11,49-51}$

227: Here, as a first approximation, denaturant effects are simply taken to be 

228: uniform over different interaction types. Models with different 

229: denaturant effects on different interaction types remain to be explored.

230: 

231: \begin{figure}

232: \begin{center}

233: \leavevmode

234: \psfig{figure=kayajmb2.eps,width=8cm,height=8cm,angle=0}

235: \end{center}

236: \narrowtext

237: \caption{Free energies profiles are given by negative logarithmic distributions 

238: of energy (solid curves), plotted here for the $\epsilon/k_B T$'s shown.

239: $P(E)$ is the sum of Boltzmann weights of conformations

240: with energies $E^\prime$ in the range $E-0.5<E^\prime\le E+0.5$.

241: The vertical dashed line marks the $E=-34$ free energy peak for

242: $\epsilon/k_B T=-1.56$. The inset shows the peak region of this profile,

243: where lines joining a pair of open circles [$-\ln P(E)$ values] record

244: all single-chain-move interconversions between a conformation with $E<-34$

245: and one with $E>-33$ in our simulation. These kinetic

246: connections suggest identifying the shaded area ($-38\le E \le -30$) 

247: as a transition state region. }

248: \end{figure}

249: 

250: As we have discussed from a general polymer perspective,$^{19}$

251: matching theoretical considerations with the experimental practice of 

252: calorimetric baseline subtractions

253: necessitate a multiple-conformation native state that entails considerable 

254: fluctuations beyond small-amplitude 

255: vibrations. Here we further investigate the implications of native-state 

256: conformational diversity. To that end, we study different definition

257: of ``native'' and ``denatured'' states by assigning different values for a 

258: ``transition'' energy $E_t$ demarcating the native and denatured ensembles,

259: such that $P_{\rm N}$ $=\sum_{E\le E_t}g(E)\exp(\epsilon E/k_B T)/Q$

260: is the fractional native population and $P_{\rm D}=1-P_{\rm N}$ 

261: is the fractional denatured population (Fig.~1). 

262: For each of these definitions to be tested, a range of energies is spanned by 

263: the native state, except for the special case when $E_t$ is chosen to be 

264: equal to the ground-state energy.

265: Remarkably, despite the differences in the definitions of ``native'' 

266: and ``denatured'' states, 

267: the thermodynamic ($\Delta G=0$) transition midpoints of the different $E_t$'s 

268: in Fig.~1 are very similar, all at $\epsilon/k_B T\approx -1.56$. 

269: 

270: Stabilities of different denatured ensembles relative to that of the 

271: ground state are shown in Fig.~1 (dashed curves). These quantities

272: correspond roughly to native state hydrogen 

273: exchange (HX) free energies,$^{52-54}$ (see also ref.~55),  for it

274: is reasonable to expect that certain amides become exposed and 

275: exchangeable when the effective energy of a conformation is 

276: above a certain threshold. Our results share the same general trend as 

277: that observed in these experiments,$^{52-54}$

278: suggesting that some of the fluctuations 

279: observed by HX may be considered as part of a multiple-conformation 

280: native state.$^{18,19,56}$

281: Figure~1 indicates that 

282: linear extrapolation of $\Delta G$ from the transition region to the strongly 

283: native regime (more negative $\epsilon/k_B T$) is only valid for the 

284: free energy difference between the set of ``fully unfolded'' open conformations 

285: and the ground state (top dashed curve).

286: 

287: \section{\bf FREE ENERGY PROFILES AND CHEVRON PLOTS}

288: 

289: \noindent

290: Consistent with the model's thermodynamically two-state character,$^{20}$

291: its energy distribution is bimodal under denaturing conditions and moderately 

292: native conditions (Fig.~2), although the 

293: distribution becomes one-sided or ``downhill''$^{7,57,58}$ 

294: under strongly native conditions (e.g., 

295: when $\epsilon/k_B T=-2.0$). We emphasize that here the determination of 

296: $\ln P(E)$ is entirely independent of any kinetic consideration. Therefore, 

297: the present $\ln P(E)$ function reflects the actual thermodynamics of the 

298: model. As such, its physical origin is fundamentally different from 

299: free energy profiles that have been empirically constructed or postulated 

300: to fit rate data. We can therefore use these $\ln P(E)$'s to 

301: assess the transition state picture, with $E$ as the reaction 

302: coordinate. Different reaction coordinates have been used in other 

303: investigations.$^{59,60}$

304: 

305: We employ standard Metropolis Monte Carlo dynamics$^{19,22}$ to explore 

306: physically plausible kinetic scenarios, using the number of attempted moves 

307: as the model time. This approach has been proven useful$^{7-11,13,61}$

308: despite its obvious limitations.$^{11,16,50}$

309: The present set of elementary chain moves 

310: consists of end flips, corner flips, crankshafts,$^{59-61}$ 

311: rigid rotations,$^{11}$ 

312: and local moves that transform two turns of a right-handed helix among 

313: its three possible orientations while holding its two end monomers fixed. 

314: The relative frequencies of attempting 

315: these moves are 2.3\%, 27\%, 60.6\%, 10\%, and 0.1\%, respectively.

316: Some chain moves can lead to large changes in energy, hence movements along 

317: the model free energy profile need not be continuous (inset of Fig.~2).

318: Therefore, it is more justified to regard the transition state as a region 

319: rather than a single highest point along this particular

320: reaction coordinate.$^{22,26,60}$ The group of conformations represented

321: by the shaded area in Fig.~2 clearly serves the role of a transition state 

322: because all conformational interconvertions between the native and 

323: unfolded sides of the population distribution must pass through 

324: one or more conformations in the shaded area.

325: 

326: 

327: \begin{figure}

328: \begin{center}

329: \leavevmode

330: \psfig{figure=kayajmb3.eps,width=8cm,height=8cm,angle=0}   

331: \end{center}

332: \narrowtext

333: \caption{Model chevron plots. Average logarithmic rates are given by negative

334: logarithms of mean first passage time (MFPT). Each folding trajectory

335: starts from a randomly generated conformation; unfolding trajectories are

336: initiated from the ground state.  Each data point is averaged from

337: $\sim 50$--$1,000$ trajectories. Solid curves through data points

338: are mere guides for the eye. Larger squares on the right show unfolding

339: MFPT's for attaining energies $E>-34$ (open) and $E>-4$ (filled).

340: Unfolding MFPT's for $E>-10$ are essentially identical to that for $E>-4$.

341: Other data points (on the left) show folding MFPT's for reaching

342: ({\it from top to bottom}) $E\le -34$, $-40$, $-42$, $-44$, $-46$, and

343: the ground state. The vertical dashed line is the approximate

344: transition midpoint. The inclined dashed-dotted line shows folding rates

345: if kinetics were two-state for the present model, an hypothetical

346: situation in which the ground state thermodynamic stability relative to the

347: fully unfolded conformations (c.f. dashed line labeled by ``$-10$''

348: in Fig.~1) is given by the difference between logarithmic folding rates 

349: (dashed-dotted line) and unfolding rates extrapolated from the solid squares. }

350: \end{figure}

351: 

352: Figure~3 reports simulated mean first passage times$^{59,60}$ for

353: a range of intraprotein interaction strength on both sides of the 

354: transition midpoint, in a format identical to typical 

355: experimental chevron plots.$^{29,31,32,49,51,62}$

356: We explore a variety of kinetic folding and unfolding criteria by 

357: monitoring the time it takes for the chain to first cross several 

358: different ``finish lines.'' This results in an appreciable variation in 

359: apparent rates under strongly native conditions (Fig.~3, more negative 

360: $\epsilon/k_B T$). Similar effects may be operative when multiple 

361: experimental probes are used to monitor kinetics.$^{63-65}$ 

362: 

363: 

364: The trajectory in Fig.~4 (upper panel) depicts the model's 

365: heuristically ``two-state'' behavior at the transition midpoint. For the 

366: two chain properties shown, native and denatured parts of the trajectory 

367: can be easily discerned, with very little time spent in between; strongly 

368: suggesting that the kinetics is first order. Fluctuations in $E$ is 

369: considerable within the native (low $E$) part of the trajectory, underscoring

370: the utility and necessity of a multiple-conformation native state (see 

371: Fig.~1 and below). Another facet of the native-denatured interconversion 

372: is provided by the $R_g$ trace. Consistent with a recent kinetic $R_g$ 

373: measurement on a small protein,$^{66}$

374: it shows that the chain undergoes 

375: sharp kinetic transitions between a native state that has minimal fluctuations 

376: in $R_g$ and a denatured state that spans a wide range of $R_g$'s.

377: 

378: 

379: A more quantitative test introduced by Gutin et al.$^{67}$ is performed in 

380: the lower panel of Fig.~4. It indicates that folding kinetics is essentially 

381: first order (i.e., single-exponential) for an extended range of 

382: model intraprotein interaction strength, covering moderately native conditions 

383: ($\epsilon/k_B T\approx -1.80$) through conditions that are less favorable 

384: to folding (less negative $\epsilon/k_B T$), although deviations from 

385: single-exponential behavior occur under strongly native conditions 

386: in the model ($\epsilon/k_B T< -1.85$). 

387: Using the same technique, unfolding kinetics 

388: (Fig.~3) is found to be essentially single-exponential for the entire 

389: range of unfolding $\epsilon/k_B T$ investigated (detailed results not shown). 

390: We have confirmed these conclusions by analyzing first passage time (FPT) 

391: distributions as in ref.~68 at

392: several $\epsilon/k_B T$'s, paying special attention to folding kinetics under 

393: moderately native conditions that are not far from the onset of drastic

394: chevron rollover and non-single-exponential behavior (Fig.~3). For example, 

395: we have obtained the logarithmic FPT distribution at $\epsilon/k_B T=-1.72$ 

396: by binning 1,080 simulated trajectories into time slots of $10^6$, and

397: found that 98\% of these trajectories can be fitted by a single exponential 

398: with a correlation coefficient $r=0.95$. 

399: If one assumes that the unit model time needed for each elementary chain 

400: move corresponds roughly to a real time scale

401: of $10^{-11}$--$10^{-9}$ sec (ref.~69),

402: the fastest model folding rate in Fig.~3 is in the 

403: order of $10^2$--$10^4$ sec$^{-1}$.

404: 

405: The contrast between the present model and its corresponding G\=o model is

406: intriguing. We have shown that the G\=o model in Fig.~4 is 

407: thermodynamically significantly less cooperative,$^{20}$ yet 

408: it folds faster than our model. This scenario of 

409: a {\it negative} correlation between folding speed and thermodynamic 

410: cooperativity may bear on the issue of folding rate overestimation 

411: in folding theories that use G\=o-like potentials.$^{23}$

412: It also raises a more basic question as to whether and when

413: the G\=o prescription is sufficiently adequate for capturing minimal 

414: frustration$^{21}$ mechanisms in real proteins.

415: 

416: 

417: \begin{figure}

418: \begin{center}

419: \leavevmode

420: \psfig{figure=kayajmb4.eps,width=8cm,height=8cm,angle=0}

421: \end{center}

422: \narrowtext

423: \caption{{\bf Upper panel:} A typical trajectory at

424: $\epsilon/k_B T=-1.56$. $R_g$ is radius

425: of gyration in units of lattice bond length.

426: {\bf Lower panel:} Folding MFPT's (filled circles) through the ground

427: state for the model in Fig.~3 (lower curves) are compared to that for

428: a G\=o model (upper curves) that has the same transition midpoint

429: (vertical dashed line), uses the same move set, and assigns a $-1.5$

430: energy for every contact in the conformation in Fig.~1 and zero

431: energy otherwise. Lines through

432: data points are mere guides for the eye. Open squares are median first

433: passage times divided by $\ln 2$, which equals MFPT for

434: single-exponential kinetics. Hence a discrepancy between the circles and

435: squares signals a deviation from single-exponential kinetics.$^{67}$}

436: \end{figure}

437: 

438: 

439: 

440: 

441: The present model is proteinlike in that it predicts a mild chevron rollover 

442: concomitant with single-exponential folding kinetics, consistent 

443: with experiments.$^{29,31,62}$ But the drastic chevron 

444: rollover (with an appreciable decreasing folding rate with increasing 

445: native conditions, and non-single-exponential folding kinetics)

446: predicted for strongly native conditions in the present model 

447: ($\epsilon/k_B T< -1.85$ in Fig.~3) has not been documented for real 

448: proteins. This suggests that such conditions, which coincide with

449: downhill folding$^{7,57,58}$ in the present model (see above), may not

450: be realizable. If so, this is not surprising. Native 

451: stability can be arbitrary high in the model ($-\epsilon/k_B T$ can be 

452: arbitrarily large), but for real proteins native stability is limited 

453: by the actual chemistry at zero denaturant. It follows that the experimental 

454: zero-denaturant situation for most proteins most likely corresponds 

455: to $\epsilon/k_B T> -1.80$ in Fig.~3.

456: It would be interesting to explore whether special experimental situations 

457: corresponding to the very strongly folding conditions in the model 

458: can be found for some proteins such that similar drastic chevron rollovers 

459: can be observed.

460: 

461: 

462: \section{\bf ASSESSING THE TRANSITION STATE PICTURE}

463: 

464: 

465: \noindent

466: Consistent with the thermodynamics of our model, kinetic rate of folding 

467: to the ground state and of unfolding to an open state meet at the 

468: approximate transition midpoint determined thermodynamically

469: in Fig.~1 (c.f. the lower ``V'' in Fig.~3). 

470: Interestingly, a similar consistency is seen by matching

471: rates of crossing the peak of the free energy profile 

472: in the inset of Fig.~2 in the folding and unfolding

473: directions (upper ``V'' in Fig.~3). Near the transition midpoint, 

474: folding rates defined by crossing several different finish lines at low 

475: but non-ground-state energies are very close to one 

476: another, and are only slightly faster than the rate of folding to the ground 

477: state (Fig.~3). This implies that, under midpoint to moderately native 

478: conditions, kinetics is rapid once the folding chain has cleared the 

479: shaded transition state region in Fig.~2 and proceeds on to the native side. 

480: But the folding rates for different finish lines

481: are very different under strongly native conditions ($\epsilon/k_B T\approx

482: -2.0$), indicative of glassy dynamics (Figs.~3 and 4).

483: 

484: Despite the essentially single-exponential and heuristically ``two-state'' 

485: kinetics discussed above, the folding/unfolding kinetics of the present model 

486: differs from the strictly two-state variety observed for an increasing 

487: number of small single-domain proteins$^{4,5,24,25,70}$

488:  such as a 64-residue form of 

489: chymotrypsin inhibitor 2.

490: Figure~3 shows that folding rates under moderately to 

491: strongly native conditions are slower than that required for such strictly 

492: thermodynamically {\it and} kinetically two-state proteins (inclined 

493: dashed-dotted line). In fact, Fig.~3 is reminiscent of experimental 

494: chevron plots with rollovers.$^{31,32,71}$ Examples include wildtype 

495: barnase$^{29}$ and ribonuclease A.$^{62}$

496: Hence we believe 

497: the present lattice construct may serve as a tool for better understanding 

498: the folding kinetics of these proteins.

499: 

500: How much kinetic information can be inferred from free energy profiles 

501: such as those in Fig.~2? The conventional transition state picture of

502: protein folding$^{2,28,30}$ stipulates that

503: $$

504: {\rm rate} = F \exp\biggl(-{\frac {\Delta G^\ddagger} {k_B T}}\biggr) \; ,

505: \eqno(1)

506: $$

507: where $\Delta G^\ddagger$ is the activation free energy for the kinetic

508: process in question. We call $\Delta G^\ddagger/k_B T$ the 

509: transition-state exponent. $F$ is the pre-exponential front factor$^{11}$

510: or prefactor,$^{23}$ which depends on solvent viscosity (not considered

511: here) but is often taken to be insensitive to intraprotein interaction 

512: strength.$^{30}$ Figure~5 examines whether this picture 

513: applies to the present model. It does so by investigating the dependence of $F$ 

514: on $\epsilon/k_B T$. For the sake of generality, several physically 

515: motivated $P(E)$-based definitions of ``transition state,'' ``folded state'' 

516: and ``unfolded state'' are evaluated. In the tests conducted here,

517: ``transition state'' is defined by either the shaded area in Fig.~2 

518: ($-38\le E\le -30$) or $E=-34$ (peak of barrier); 

519: ``folded state'' is defined by the ground state only ($E=-52$) or by 

520: $E\le -34$ (left of the barrier); and ``unfolded state'' is defined by 

521: the approximate position of the denatured free energy minimum ($E=-4$), 

522: by the bulk of the open conformations ($E>-10$), or by $E\ge -34$ 

523: (right of the barrier).

524: Transition-state exponents for folding and unfolding are then computed,

525: respectively, by the [transition/unfolded] and [transition/folded]

526: population ratios at the given $\epsilon/k_B T$ (Fig.~5).

527: 

528: The scaling of $F$ with respect to $\epsilon/k_B T$ is found to be

529: not sensitive to these variations in definition (Fig.~5, upper panel). 

530: Our results show that the simple transition state picture does not apply 

531: to folding in this model. For the quasi-linear part of the chevron 

532: plot in Fig.~3, the relationship between $\Delta G^\ddagger/k_B T$ 

533: and $-\ln({\rm MFPT})$ is approximately linear (Fig.~5), with slope

534: $\approx -1.5$ (filled symbols) or $\approx -1.9$ (open symbols).

535: This implies that $F$ has approximately the same functional form as the 

536: activation factor in this regime, but with an exponent of opposite sign, viz.,

537: $F\sim \exp(-\mu\Delta G^\ddagger/k_B T)$, where $\mu$ $\approx -0.33$

538: for folding to the ground state (filled symbols). On the other hand, 

539: unfolding appears to be well described by the simple transition state 

540: picture. The corresponding slope for unfolding $\approx -1.0$, hence 

541: $F\approx$ constant. Folding and unfolding front factors are approximately 

542: equal near the thermodynamic midpoint (Fig.~5, lower panel), reflecting

543: the fact that in that region folding and unfolding rates are essentially 

544: equal (Fig.~3). The role of front factors in understanding 

545: folding rates has recently been emphasized by Portman et al.$^{23}$

546: The pattern in Fig.~5 is similar in some respects to the results of

547: a recent model study by Nymeyer et al.$^{22}$ Using a different reaction 

548: coordinate for 2-, 3-letter and G\=o 27mer models, these authors found 

549: approximate linear relations with non-unity slopes between kinetically 

550: simulated rates and rate quantities deduced from free energy profiles, although 

551: they did not consider a broad range of overall interaction strengths 

552: as that in chevron plots.

553: 

554: Figure~5 shows that $F$ for folding decreases with increasingly native 

555: conditions. $F$ may be identified as an effective diffusion coefficient.

556: It corresponds to an internal friction term arising from the 

557: impediment to motion imposed by the chain segments on one 

558: another.$^{2,7,23,60,71-74}$ 

559: Plaxco and Baker$^{72}$ have experimentally investigated 

560: internal friction in protein folding, and concluded insightfully that internal

561: friction effects are limited for strictly two-state proteins. But the 

562: functional form they considered is different from the novel approximate 

563: exponential form $F\sim \exp(-\mu\Delta G^\ddagger/k_B T)$ discovered here 

564: for the quasi-linear part of chevron plots and the part with a mild

565: rollover (corresponding to a limited range 

566: of $\epsilon/k_B T$) for proteins that are not kinetically two-state. 

567: In the present model, when a wider range of $\epsilon/k_B T$ is considered 

568: (Fig.~5, lower panel), the folding $\ln F$ reveals 

569: its nonlinear character, a feature anticipated by energy landscape 

570: theory$^{21,23}$ and consistent with a pioneering simulation 

571: study of Socci et al.$^{60}$

572: 

573: 

574: \begin{figure}

575: \begin{center}

576: \leavevmode

577: \psfig{figure=kayajmb5.eps,width=8cm,height=8cm,angle=0}

578: \end{center}

579: \narrowtext

580: \caption{Correlations between rates and transition-state

581: exponents. Filled symbols are for reaching the ground state in folding and

582: reaching $E>-4$ in unfolding, open symbols are for crossing the

583: $-\ln P(E)$ peak at $E=-34$ in either directions.

584: {\bf Upper panel:} Folding transition-state exponent

585: $\Delta G^\ddagger/k_B T=$ $-\ln[P(-38\le E\le -30)/P(E\ge -10)]$ (circles),

586: $-\ln[P(-38\le E\le -30)/P(E\ge -34)]$

587: (triangles), $-\ln[P(E=-34)/P(E=-4)]$ (squares), or

588: $-\ln[P(E=-34)/P(E\ge -10)]$

589: (diamonds). Data points shown are for $\epsilon/k_B T\ge-1.75$.

590: For unfolding, $\Delta G^\ddagger/k_B T=$

591: $-\ln[P(-38\le E\le -30)/P(E=-52)]$ (squares),

592: $-\ln[P(-38\le E\le -30)/P(E\le -34)]$

593: (circles), or $-\ln[P(E=-34)/P(E=-52)]$ (diamonds).

594: The straight lines are fitted.

595: {\bf Lower panel:} $\ln F$ $\equiv$ $-\ln({\rm MFPT})$ $+$

596: $\Delta G^\ddagger/k_B T$ (horizontal variable) vs.

597: $\epsilon/k_B T$. Filled and open

598: circles (for folding) and squares (for unfolding) identify

599: the folding and unfolding $\Delta G^\ddagger/k_B T$ used, as defined in the 

600: upper panel. The dashed line marks the approximate transition midpoint.}

601: \end{figure}

602: 

603: 

604: 

605: \section{\bf THE CALORIMETRIC CONNECTION: WHAT IS THE NATIVE STATE?}

606: 

607: \noindent

608: A thermodynamic consideration of the model's free energy 

609: profiles (Fig.~2) and the above kinetic analysis suggest that 

610: a natural way to define the ``native'' and ``denatured'' states

611: is to have their demarcation line at $E_t=-34$. Figure~6 investigates

612: the calorimetric implications of different $E_t$ choices. As a first 

613: test of principles, in this section we take the effective intraprotein 

614: energies as temperature independent. Our conclusions are not expected 

615: to be significantly altered by the incorporation of proteinlike

616: temperature-dependent effective interactions.$^{18}$

617: 

618: 

619: \begin{figure}

620: \begin{center}

621: \leavevmode

622: \psfig{figure=kayajmb6.eps,width=8cm,height=8cm,angle=0}

623: \end{center}

624: \narrowtext

625: \caption{Heat capacity vs. temperature (from ref.~20).

626: Here nonlinear denatured (high $T$) dashed baselines$^{19,75}$ are computed

627: for ({\it from top to bottom}) $E_t=$ $-52.04$, $-48$, $-42$, $-38$,

628: and $-34$ (c.f. Fig.~1). Corresponding native baselines are plotted but

629: are too close to one another to be distinguishable. The vertical

630: dotted line marks the approximate transition midpoint used throughout

631: this work. Also shown are example compact non-ground-state conformations

632: with $E=-36.0$ (top) and $E=-46.0$ (bottom). The beads mark monomers

633: that are not in their folded environment, i.e., do not have their full

634: set of contacts as that in the ground state.}

635: \end{figure}

636: 

637: We have argued that empirical calorimetric baseline subtractions correspond 

638: essentially to an operational definition of the native and denatured 

639: ensembles.$^{19}$

640: The demarcation energy (or enthalpy) between the native 

641: and denatured states may be ascertained by matching empirical baselines$^{19}$

642: to the nonlinear ``formal two-state'' baselines of Zhou et al.$^{75}$

643: Figure~6 shows that the baselines from empirically extrapolating the native 

644: and denatured tails of the heat capacity curve of the present model$^{20}$

645: essentially coincide with the formal two-state baselines for $E_t=-34$,

646: implying that by adopting such empirical baselines

647: $E_t=-34$ is effectively adopted. In that case, the two chain conformations

648: shown in Fig.~6 

649: would belong to the native state and therefore may be regarded as

650: sitting ``below the calorimetric baseline.''$^{19}$

651: These chains may model

652: the sparsely populated conformations revealed by native-state HX.$^{52-54}$

653: A multiple-conformation native state view 

654: is supported by a recent experimental observation of pretransitional 

655: conformational changes in ribonuclease A.$^{76}$

656: 

657: 

658: 

659: 

660: \section{\bf DISCUSSION}

661: 

662: \noindent

663: The logic of the present analysis is premised on a comparison between 

664: simulated folding/unfolding rates and transition-state 

665: predictions based on independently obtained free energy profiles.

666: The conventional transition state picture of folding posits a weak

667: or nonexistent dependence of the front factor on

668: a protein's intrachain interaction strength. In the present model, which

669: exhibits a chevron rollover, we find that the conventional picture holds 

670: approximately for unfolding but not for folding. In particular,

671: for the quasi-linear part of the folding arm of the chevron plot,

672: the folding front factor adopts a re-scaled form of the exponential

673: factor, harboring an exponent opposite in sign to that of the activation 

674: term. These findings are consistent with internal friction and diffusive 

675: folding dynamics ideas from energy landscape theory. They suggest that

676: simple transition state theory with a constant front factor may not

677: be generally applicable in the presence of a chevron rollover,

678: even if the kinetics is apparently first order. In this regard, future 

679: single-molecule measurements$^{77}$ of folding time distributions 

680: may provide important insight into the physics underlying approximate 

681: single-exponential folding kinetics, since these measurements may detect 

682: small deviations from first-order kinetics$^{11}$ (e.g., 

683: a possible small non-exponential tail in the distribution) that would 

684: otherwise be difficult to ascertain from traditional measurements of 

685: ensemble-averaged folding rates.

686: 

687: Chevron rollovers have been rationalized by dead-time discrete 

688: intermediates$^{29}$

689: and by movements of the transition-state peak 

690: on broad activation barriers.$^{31}$ We have not been able 

691: to detect these proposed features in our model free energy profiles. 

692: Instead, the present results offer an alternate rationalization in terms 

693: of diffusive dynamics and an interaction-dependent folding front factor.

694: It follows that, in general, analyses that focus exclusively on free energy 

695: profiles may be incomplete. Inasmuch as chevron rollovers are a manifestation 

696: of an interaction-dependent front factor, as suggested here, experimental 

697: observations of significant mutational effects on rollover behavior$^{29}$

698: imply that mutations can have a significant 

699: effect not only on the free energy profile itself, but also on front 

700: factors not afforded by such profiles.

701: 

702: The present account of salient features of chevron rollover and 

703: native-state HX in terms of an essentially continuous energy 

704: distribution (Fig.~3) is similar in spirit to the recent idea that these 

705: features may originate from a ``burst phase continuum.''$^{32,33}$ 

706: However, the burst phase continuum view is based on postulated

707: free energy profiles, not free energy profiles derived from models with 

708: explicit chain representations. Further effort will be required to elucidate 

709: the relationship between the burst phase continuum and

710: the present chain-based perspectives, as there are apparent differences 

711: between the two. For instance, the present study suggests that some of the 

712: states detected by native-state HX are on the native side of the conformational 

713: distribution (Figs.~3 and 6) rather than on the denatured side as envisioned 

714: by the burst phase continuum scenario. 

715: 

716: In summary,

717: we emphasize that while the current study proposes a new 

718: physical rationalization for chevron rollover, it does not by itself rule out 

719: other mechanisms.  Chevron rollovers in real proteins may arise from a 

720: combination of effects. Obviously, the generality of the present 

721: interaction-dependent front factor scenario should be further tested 

722: using model proteins with non-helical native topologies as well as using 

723: geometrically more realistic off-lattice continuum models.$^{26,78}$ 

724: 

725: As for the relationship between generic features of folding/unfolding 

726: kinetics and thermodynamics of small globular proteins, 

727: the qualitative agreement between Fig.~3 and typical chevron rollover plots

728: for real proteins supports the idea that proteinlike thermodynamics

729: necessarily lead to proteinlike folding/unfolding kinetics.

730: A case in point is the folding kinetics of a set of 20-letter model 

731: sequences reported by Gutin et al.$^{67}$ Our test calculations show that 

732: random 

733: sequences of this particular 20-letter alphabet with additive contact

734: energies are not calorimetrically 

735: two-state (data not shown). Although much useful insight has been 

736: gained from them$^{10,14,68}$ (see also ref.~16), recent calculations$^{19}$

737: indicate that even some designed sequences in this 20-letter model are 

738: thermodynamically less cooperative than the present model.$^{20}$

739: Apparently, the folding kinetics of these 20-letter model sequences are 

740: less proteinlike as a result, in that their folding 

741: rates {\it decrease} when native stability is increased from the 

742: transition midpoint (Figs.~2, 5, 7 and 8 in ref.~67). 

743: This is because the maximum folding rates and the onset of drastic rollover

744: in these 20-letter models occur around the thermodynamic transition midpoint,

745: rather than under strongly native conditions as in the cooperative model 

746: studied here (Fig.~3 of the present work). Thus, the chevron trend 

747: predicted by these 20-letter models under transition midpoint through 

748: moderately folding conditions is opposite to that observed 

749: experimentally,$^{49,51}$ because experiments

750: almost invariably show an increasing folding rate when native stability 

751: is increased from the transition midpoint.

752: 

753: The present model's kinetics is proteinlike but not two-state. In this 

754: respect, it is reassuring that the exercise here fares no worse than 

755: Nature's. This is because a protein's calorimetric two-state 

756: cooperativity,$^{79}$ such as that of hen 

757: lysozyme,$^{63,71}$ barnase,$^{29}$ 

758: and ribonuclease A,$^{62}$ is no 

759: guarantee for two-state kinetics.$^{4,5}$ 

760: However, the present exercise 

761: does suggest that additional or alternate interaction mechanisms have to 

762: be discovered to account for the strictly two-state behavior of many small 

763: single-domain proteins. In that regard, it would be interesting to 

764: investigate the connection between the strictly two-state proteins' 

765: apparently nonglassy kinetics$^{80}$

766: and the possibility that their front 

767: factors might be minimally sensitive to the variation in intrachain 

768: interaction strength. 

769: 

770: \noindent

771: {\bf Acknowledgments}

772: 

773: We thank Alan Davidson, Julie Forman-Kay, Yuji Goto, Bob Matthews, and 

774: Tetsuya Yomo for helpful discussions,

775: and David Baker, Martin Gruebele, Walid Houry, Kevin Plaxco, and 

776: Peter Wolynes for their critical

777: reading of an earlier draft of this paper and their very useful 

778: suggestions. This work 

779: was supported by Medical Research Council of Canada grant no. MT-15323 

780: and a Premier's Research Excellence Award (Ontario). H. S. C. is a 

781: Canada Research Chair in Biochemistry.

782: 

783: %\vfill\eject

784: 

785: %\par\vfill\eject

786: 

787: %\noindent

788: %{\large\bf References}

789: 

790: %\kern -1.5cm

791: 

792: \begin{thebibliography}{99}

793: 

794: \bibitem{1}

795: Baker, D. (2000).

796: A surprising simplicity to protein folding.

797: {\it Nature} {\bf 405}, 39--42.

798: 

799: \bibitem{2}

800: Bilsel, O. \& Matthews, C. R. (2000).

801: Barriers in protein folding reactions.

802: {\it Adv. Protein Chem.} {\bf 53}, 153--207.

803: 

804: \bibitem{3}

805: Eaton, W. A., Mu{\~n}oz, V., Hagen, S. J., Jas, G. S., Lapidus, L. J., 

806: Henry, E. R. \& Hofrichter, J. (2000).

807: Fast kinetics and mechanisms in protein folding. 

808: {\it Annu. Rev. Biophys. Biomolec. Struct.} {\bf 29}, 321--359.

809: 

810: \bibitem{4}

811: Fersht, A. R. (2000). 

812: Transition-state structure as a unifying basis in protein-folding mechanisms: 

813: Contact order, chain topology, stability, and the extended nucleus mechanism.

814: {\it Proc. Natl. Acad. Sci. USA} {\bf 97}, 1525--1529.

815: 

816: \bibitem{5}

817: Plaxco, K. W., Simons, K. T., Ruczinski, I. \& Baker, D. (2000). Topology,

818: stability, sequence, and length: Defining the determinants of two-state

819: protein folding kinetics. {\it Biochemistry} {\bf 39}, 11177--11183.

820: 

821: \bibitem{6}

822: Radford, S. E. (2000).

823: Protein folding: progress made and promises ahead.

824: {\it Trend Biochem. Sci.} {\bf 25}, 611--618.

825: 

826: \bibitem{7}

827: Bryngelson, J. D., Onuchic, J. N., Socci, N. D. \& Wolynes, P. G. (1995).

828: Funnels, pathways, and the energy landscape of protein folding: A

829: synthesis. {\it Proteins Struct. Funct. Genet.} {\bf 21}, 167--195.

830: 

831: \bibitem{8}

832: Dill, K. A., Bromberg, S., Yue, K., Fiebig, K. M., Yee, D. P., Thomas, P. D.

833: \& Chan, H. S. (1995).

834: Principles of protein folding --- A perspective from simple

835: exact models. {\it Protein Sci.} {\bf 4}, 561--602.

836: 

837: \bibitem{9}

838: Thirumalai, D. \& Woodson S. A. (1996). Kinetics of folding of proteins 

839: and RNA. {\it Acc. Chem. Res}. {\bf 29}, 433--439. 

840: 

841: \bibitem{10}

842: Shakhnovich, E. I. (1997). Theoretical studies of protein folding

843: thermodynamics and kinetics. {\it Curr. Opin. Struct. Biol.} {\bf 7},

844: 29--40.

845: 

846: \bibitem{11}

847: Chan, H. S. \& Dill, K. A. (1998). Protein folding in the landscape 

848: perspective: Chevron plots and non-Arrhenius kinetics. {\it Proteins Struct. 

849: Funct.  Genet.} {\bf 30}, 2--33.

850: 

851: \bibitem{12}

852: Micheletti, C., Banavar, J. R., Maritan, A. \& Seno, F. (1999).

853: Protein structures and optimal folding from a geometrical variational 

854: principle. {\it Phys. Rev. Lett.} {\bf 82}, 3372--3375.

855: 

856: \bibitem{13}

857: Onuchic, J. N., Nymeyer, H., Garcia, A. E., Chahine, J. \& Socci, N. D.

858: (2000). The energy landscape theory of protein folding: Insights into folding 

859: mechanisms and scenarios. {\it Adv. Protein Chem.} {\bf 53}, 87--152.

860: 

861: \bibitem{14}

862: Mirny, L. \& Shakhnovich, E. (2001).

863: Protein folding theory: From lattice to all-atom models.

864: {\it Annu. Rev. Biophys. Biomol. Struct.} {\bf 30}, 361--396.

865: 

866: \bibitem{15}

867: Thirumalai, D. \& Lorimer, G. H. (2001).

868: Chaperonin-mediated protein folding.

869: {\it Annu. Rev. Biophys. Biomol. Struct.} {\bf 30}, 245--269.

870: 

871: \bibitem{16}

872: Chan, H. S., Kaya, H. \& Shimizu, S. (2002). Computational methods for protein

873: folding: Scaling a hierarchy of complexities. In

874: {\it Current Topics in Computational Molecular Biology}, eds. Jiang, T., 

875: Xu, Y.  \& Zhang, M. Q. (The MIT Press, Cambridge, MA), pp.~403-447.

876: 

877: \bibitem{17}

878: Chan, H. S. (1998).

879: Matching speed and locality. {\it Nature} {\bf 392}, 761--763. 

880: 

881: \bibitem{18}

882: Chan, H. S. (2000). Modeling protein density of states: Additive hydrophobic

883: effects are insufficient for calorimetric two-state cooperativity.

884: {\it Proteins Struct. Funct. Genet.} {\bf 40}, 543--571.

885: 

886: \bibitem{19}

887: Kaya, H. \& Chan, H. S. (2000).

888: Polymer principles of protein calorimetric

889: two-state cooperativity. {\it Proteins Struct. Funct. Genet.} {\bf 40},

890: 637--661 [Erratum: {\bf 43}, 523 (2001)].

891: 

892: \bibitem{20}

893: Kaya, H. \& Chan, H. S. (2000). Energetic components of cooperative protein

894: folding. {\it Phys. Rev. Lett.} {\bf 85}, 4823--4826.

895: 

896: \bibitem{21}

897: Bryngelson, J. D. \& Wolynes, P. G. (1987). 

898: Spin glasses and the statistical mechanics of protein folding.

899: {\it Proc. Natl. Acad. Sci. USA} {\bf 84}, 7524--7528.

900: 

901: \bibitem{22}

902: Nymeyer, H., Socci, N. D. \& Onuchic, J. N. (2000). Landscape approaches for

903: determining the ensemble of folding transition states: Success and failure

904: hinge on the degree of frustration. {\it Proc. Natl. Acad. Sci. USA} {\bf

905: 97}, 634--639.

906: 

907: \bibitem{23}

908: Portman, J. J., Takada, S. \& Wolynes, P. G. (2001). Microscopic theory of 

909: protein folding rates. II. Local reaction coordinates and chain dynamics.

910: {\it J. Chem. Phys.} {\bf 114}, 5082--5096.

911: 

912: \bibitem{24}

913: Alm, E. \& Baker, D. (1999). Prediction of protein-folding mechanisms from 

914: free-energy landscapes derived from native structures. {\it Proc. Natl.

915: Acad. Sci. USA} {\bf 96}, 11305--11310.

916: 

917: \bibitem{25}

918: Mu{\~n}oz, V. \& Eaton, W. A. (1999).

919: A simple model for calculating the kinetics of

920: protein folding from three-dimensional structures. 

921: {\it Proc. Natl. Acad. Sci. USA} {\bf 96}, 11311--11316.

922: 

923: \bibitem{26}

924: Clementi, C., Nymeyer, H. \& Onuchic, J. N. (2000). Topological and energetic 

925: factors: What determines the structural details of the transition state

926: ensemble and ``en-route'' intermediates for protein folding? An investigation

927: for small globular proteins. {\it J. Mol. Biol.} {\bf 298}, 937--953.

928: 

929: \bibitem{27}

930: Myers J. K. \& Oas, T. G. (2001).

931: Preorganized secondary structure as an important determinant of

932: fast protein folding. {\it Nature Struct. Biol.} {\bf 8}, 552--558.

933: 

934: \bibitem{28}

935: Matthews, C. R. \& Hurle, M. R. (1987).

936: Mutant sequences as probes of protein folding mechanisms.

937: {\it BioEssays} {\bf 6}, 254--257.

938: 

939: \bibitem{29}

940: Matouschek, A., Kellis, J. T., Serrano, L., Bycroft, M. \& Fersht, A. R. 

941: (1990). Characterizing transient folding intermediates by protein engineering.

942: {\it Nature} {\bf 346}, 440-445.

943: 

944: \bibitem{30}

945: Fersht, A. R., Matouschek, A. \& Serrano, L. (1992).

946: The folding of an enzyme. I. Theory of protein engineering analysis of

947: stability and pathway of protein folding.

948: {\it J. Mol. Biol.} {\bf 224}, 771--782.

949: 

950: \bibitem{31}

951: Oliveberg, M. (1998). Alternative explanations for ``multistate'' kinetics 

952: in protein folding: Transient aggregation and changing transition-state 

953: ensembles. {\it Acc. Chem. Res.} {\bf 31}, 765--772.

954: 

955: \bibitem{32}

956: Parker, M. J. \& Marqusee, S. (1999). The cooperativity of burst phase 

957: reactions explored. {\it J. Mol. Biol.} {\bf 293}, 1195--1210. 

958: 

959: \bibitem{33}

960: Parker, M. J. \& Marqusee, S. (2000). A statistical appraisal of

961: native state hydrogen exchange data: Evidence for a burst phase

962: continuum? {\it J. Mol. Biol.} {\bf 300}, 1361--1375.

963: 

964: \bibitem{34}

965: Wang, J. \& Wang, W. (1999).

966: A computational approach to simplifying the protein folding alphabet.

967: {\it Nature Struct. Biol.} {\bf 6}, 1033--1038.

968: 

969: \bibitem{35}

970: Klimov, D. K., Betancourt, M. R. \& Thirumalai, D. (1998).

971: Virtual atom representation of hydrogen bonds in minimal       

972: off-lattice models of $\alpha$ helices: effect on stability,

973: cooperativity and kinetics.

974: {\it Fold. Des.} {\bf 3}, 481--496.

975: 

976: \bibitem{36}

977: Takada, S., Luthey-Schulten, Z. \& Wolynes, P. G. (1999).

978: Folding dynamics with nonadditive forces: A simulation study of a designed

979: helical protein and a random heteropolymer.

980: {\it J. Chem. Phys.} {\bf 110}, 11616--11629.

981: 

982: \bibitem{37}

983: Kolinski, A., Galazka, W. \& Skolnick, J. (1996). On the origin of the 

984: cooperativity of protein folding: Implications from model simulations.

985: {\it Proteins Struct. Funct. Genet.} {\bf 26}, 271--287.

986: 

987: \bibitem{38}

988: Hao, M.-H. \& Scheraga, H. A. (1997).

989: Characterization of foldable protein models:

990: Thermodynamics, folding kinetics and force field. {\it J. Chem. Phys.} {\bf

991: 107}, 8089--8102.

992: 

993: \bibitem{39}

994: Eastwood, M. P. \& Wolynes, P. G. (2001).

995: Role of explicitly cooperative interactions

996: in protein folding funnels: A simulation study. {\it J. Chem. Phys.} {\bf

997: 114}, 4702--4716.

998: 

999: \bibitem{40}

1000: Rank, J. A. \& Baker, D. (1997), A desolvation barrier to hydrophobic cluster 

1001: formation may contribute to the rate-limiting step in protein folding.

1002: {\it Protein Sci.} {\bf 6}, 347--354.

1003: 

1004: \bibitem{41}

1005: Czaplewski, C., Rodziewicz-Motowid{\l}o, S., Liwo, A., Ripoll, D. R., Wawak,

1006: R. J. \& Scheraga, H. A. (2000). Molecular simulation study of cooperativity in

1007: hydrophobic association. {\it Protein Sci.} {\bf 9}, 1235--1245.

1008: 

1009: \bibitem{42}

1010: Shimizu, S. \& Chan, H. S. (2000). 

1011: Temperature dependence of hydrophobic interactions: A mean force 

1012: perspective, effects of water density, and non-additivity of thermodynamic 

1013: signatures. {\it J. Chem. Phys.} {\bf 113}, 4683--4700. 

1014: 

1015: \bibitem{43}

1016: Shimizu, S. \& Chan, H. S. (2001). Configuration-dependent heat capacity of

1017: pairwise hydrophobic interactions. {\it J. Am. Chem. Soc.} {\bf 123},

1018: 2083--2084.

1019: 

1020: \bibitem{44}

1021: Shimizu, S. \& Chan, H. S. (2001).

1022: Anti-cooperativity in hydrophobic interactions:

1023: A simulation study of spatial dependence of three-body effects

1024: and beyond. {\it J. Chem. Phys.} {\bf 115}, 1414--1421.

1025: 

1026: \bibitem{45}

1027: Shimizu, S. \& Chan, H. S. (2001).

1028: Statistical mechanics of solvophobic aggregation: Additive and

1029: cooperative effects. {\it J. Chem. Phys.} {\bf 115}, 3424--3431.

1030: 

1031: \bibitem{46}

1032: Bromberg, S. \& Dill, K. A. (1994).

1033: Sidechain entropy and packing in proteins.

1034: {\it Protein Sci.} {\bf 3}, 997--1009.

1035: 

1036: \bibitem{47}

1037: Klimov, D. K. \& Thirumalai, D. (1998).

1038: Cooperativity in protein folding: From lattice models with sidechains

1039: to real proteins.

1040: {\it Fold. \& Des.} {\bf 3}, 127--139.

1041: 

1042: \bibitem{48}

1043: Alonso, D. O. V. \& Dill, K. A. (1991).

1044: Solvent denaturation and stabilization of globular proteins.

1045: {\it Biochemistry} {\bf 30}, 5974--5985.

1046: 

1047: \bibitem{49}

1048: Scalley, M. L. \& Baker, D. (1997).

1049: Protein folding kinetics exhibit an Arrhenius

1050: temperature dependence when corrected for the temperature dependence of

1051: protein stability. {\it Proc. Natl. Acad. Sci. USA} {\bf 94}, 10636--10640. 

1052: 

1053: \bibitem{50}

1054: Chan, H.S. (1998). 

1055: Modelling protein folding by Monte Carlo dynamics:

1056: Chevron plots, chevron rollover, and non-Arrhenius kinetics. 

1057: In {\it Monte Carlo Approach to Biopolymers and 

1058: Protein Folding} (Grassberger, P., Barkema, G.T. \& Nadler, W., eds)

1059: pp.~29--44, World Scientific, Singapore.

1060: 

1061: \bibitem{51}

1062: Kuhlman, B., Luisi, D. L., Evans, P. A. \& Raleigh, D. P. (1998).

1063: Global analysis of the effects of temperature and denaturant on the 

1064: folding and unfolding kinetics of the N-terminal domain of the protein L9. 

1065: {\it J. Mol. Biol.} {\bf 284}, 1661--1670.

1066: 

1067: \bibitem{52}

1068: Kim, K. S. \& Woodward, C. (1993). Protein internal flexibility and global

1069: stability: Effect of urea on hydrogen-exchange rates of bovine pancreatic

1070: trypsin inhibitor. {\it Biochemistry} {\bf 32}, 9609--9613.

1071: 

1072: \bibitem{53}

1073: Bai, Y., Sosnick, T. R., Mayne, L. \& Englander, S. W. (1995). Protein folding

1074: intermediates: Native-state hydrogen exchange. {\it Science} {\bf 269},

1075: 192--197.

1076: 

1077: \bibitem{54}

1078: Llin{\'a}s, M., Gillespie, B., Dahlquist, F. W. \& Marqusee, S. (1999).

1079: The energetics of T4 lysozyme reveal a hierarchy of conformations.

1080: {\it Nature Struct. Biol.} {\bf 6}, 1072--1078.

1081: 

1082: \bibitem{55}

1083: Miller, D. W. \& Dill, K. A. (1995). A statistical mechanical model for

1084: hydrgen exchange in globular proteins. {\it Protein Sci.} {\bf 4},

1085: 1860--1873.

1086: 

1087: \bibitem{56}

1088: Sadqi, M., Casares, S., Abril, M. A., L\'opez-Mayorga, O., Conejero-Lara, F.

1089: \& Freire, E. (1999).

1090: The native state conformational ensemble of the SH3 domain from

1091: $\alpha$-spectrin.

1092: {\it Biochemistry} {\bf 38}, 8899--8906.

1093: 

1094: \bibitem{57}

1095: Eaton, W. A. (1999). Searching for ``downhill scenarios'' in protein folding. 

1096: {\it Proc. Natl. Acad. Sci. USA} {\bf 96}, 5897--5899.

1097: 

1098: \bibitem{58}

1099: Sabelko, J., Ervin, J. \& Gruebele, M. (1999). Observation of strange 

1100: kinetics in protein folding. 

1101: {\it Proc. Natl. Acad. Sci. USA} {\bf 96}, 6031--6036.

1102: 

1103: \bibitem{59}

1104: {\v S}ali, A., Shakhnovich, E. \& Karplus, M. (1994). How does a protein fold.

1105: {\it Nature} {\bf 369}, 248--251.

1106: 

1107: \bibitem{60}

1108: Socci, N. D., Onuchic, J. N. \& Wolynes, P. G. (1996). Diffusive dynamics 

1109: of the reaction coordinate for protein folding funnels. 

1110: {\it J. Chem. Phys.} {\bf 104}, 5860--5868.

1111: 

1112: \bibitem{61}

1113: Pande, V. S. \& Rokhsar, D. S. (1999). Folding pathway of a lattice model for 

1114: proteins. {\it Proc. Natl. Acad. Sci. USA} {\bf 96}, 1273--1278.

1115: 

1116: \bibitem{62}

1117: Houry, W. A., Rothwarf, D. M. \& Scheraga, H. A. (1995).

1118: The nature of the initial step in the conformational folding of

1119: disulphide-intact ribonuclease A.

1120: {\it Nature Struct. Biol.} {\bf 2}, 495--503.

1121: 

1122: \bibitem{63}

1123: Dobson, C. M., Evans, P. A. \& Radford, S. E. (1994). Understanding how 

1124: proteins fold: The lysozyme story so far. {\it Trend Biochem. Sci.} {\bf 19},

1125: 31--37.

1126: 

1127: \bibitem{64}

1128: Arai, M. \& Kuwajima, K. (2000). Role of the molten globule state in protein

1129: folding. {\it Adv. Protein Chem.} {\bf 53}, 209--282.

1130: 

1131: \bibitem{65}

1132: Kuwata, K., Shastry, R., Cheng, H., Hoshino, M., Batt, C. A., Goto, Y. \&

1133: Roder, H. (2001). Structural and kinetic characterization of early folding 

1134: events in $\beta$-lactoglobulin. {\it Nature Struct. Biol.} {\bf 8}, 151--155.

1135: 

1136: \bibitem{66}

1137: Plaxco, K. W., Millett, I. S., Segel, D. J., Doniach, S. \& Baker, D.

1138: (1999). Chain collapse can occur concomitantly with the rate-limiting step 

1139: in protein folding. {\it Nature Struct. Biol.} {\bf 6}, 554--556.

1140: 

1141: \bibitem{67}

1142: Gutin, A., Sali, A., Abkevich, V., Karplus, M. \& Shakhnovich, E. I. (1998).

1143: Temperature dependence of the folding rate in a simple protein model:

1144: Search for a ``glass'' transition. {\it J. Chem. Phys.} {\bf 108}, 

1145: 6466--6483.

1146: 

1147: \bibitem{68}

1148: Abkevich, V. I., Gutin, A. M. \& Shakhnovich, E. I. (1994). Free energy 

1149: landscape for protein folding kinetics: Intermediates, traps, and multiple 

1150: pathways in theory and lattice model simulations. 

1151: {\it J. Chem. Phys.} {\bf 101}, 6052--6062.

1152: 

1153: \bibitem{69}

1154: Creighton, T. E. (1984).

1155: {\it Proteins --- Structures and Molecular Properties}

1156: pp.~180--181, W. H. Freeman and Co., New York.

1157: 

1158: \bibitem{70}

1159: Debe, D. A. \& Goddard, W. A. (1999).

1160: First principles prediction of protein folding

1161: rates. {\it J. Mol. Biol.} {\bf 294}, 619--625.

1162: 

1163: \bibitem{71}

1164: Kiefhaber, T. (1995).

1165: Kinetic traps in lysozyme folding. {\it Proc. Natl. Acad. Sci.

1166: USA} {\bf 92}, 9029--9033.

1167: 

1168: \bibitem{72}

1169: Plaxco, K. W. \& Baker, D. (1998).

1170: Limited internal friction in the rate-limiting

1171: step of a two-state protein folding reaction. {\it Proc. Natl. Acad. Sci.

1172: USA} {\bf 95}, 13591--13596.

1173: 

1174: \bibitem{73}

1175: T\"uzel, E. \& Erzan, A. (2000).

1176: Glassy dynamics of protein folding. 

1177: {\it Phys.  Rev. E} {\bf 61}, R1040--R1043.

1178: 

1179: \bibitem{74}

1180: Jacob, M., Geeves, M., Holtermann, G. \& Schmid, F. X. (1999).

1181: Diffusional barrier crossing in a two-state protein folding reaction. 

1182: {\it Nature Struct.  Biol.} {\bf 6}, 923--926.

1183: 

1184: \bibitem{75}

1185: Zhou, Y., Hall, C. K. \& Karplus, M. (1999). The calorimetric criterion for a 

1186: two-state process revisited. {\it Protein Sci.} {\bf 8}, 1064--1074.

1187: 

1188: \bibitem{76}

1189: Stelea, S. D., Pancoska, P., Benight, A. S. \& Keiderling, T. A. (2001). 

1190: Thermal unfolding of ribonuclease A in phosphate at neutral pH: Deviations 

1191: from the two-state model. {\it Protein Sci.} {\bf 10}, 970--978.

1192: 

1193: \bibitem{77}

1194: Lu, H. P., Xun, L. Y. \& Xie, X. S. (1998).

1195: Single-molecule enzymatic dynamics.

1196: {\it Science} {\bf 282} 1877--1882.

1197: 

1198: \bibitem{78}

1199: Thirumalai, D. \& Klimov, D. K. (1999).

1200: Deciphering the timescales and mechanisms of protein folding using minimal

1201: off-lattice models.

1202: {\it Curr. Opin. Struct. Biol.} {\bf 9}, 197---207.

1203: 

1204: \bibitem{79}

1205: Makhatadze, G. I. \& Privalov, P. L. (1995). Energetics of protein structure. 

1206: {\it Adv. Protein Chem.} {\bf 47}, 307--425.

1207: 

1208: \bibitem{80}

1209: Gillespie, B. \& Plaxco, K. W. (2000).

1210: Nonglassy kinetics in the folding of a simple single-domain protein. 

1211: {\it Proc. Natl. Acad. Sci. USA} {\bf 97}, 12014--12019.

1212: 

1213: \end{thebibliography}

1214: \end{multicols}

1215: \end{document}

1216: 

1217: