1: \documentstyle[12pt]{report}
2: \def\rbuildrel#1\over#2{\mathrel{\mathop{#2}\limits_{#1}}}
3:
4: \catcode`@=11
5: \def\underline#1{\relax\ifmmode\@@underline#1\else
6: $\@@underline{\hbox{#1}}$\relax\fi}
7:
8: \def\changefootnote{\def\thefootnote{\fnsymbol{footnote}} }
9: \def\titlepage{\pagestyle{empty}\c@page=0
10: \def\thefootnote{\fnsymbol{footnote}} }
11: \def\endtitlepage{\pagestyle{plain}\c@page=1
12: \def\thefootnote{\arabic{footnote}} \c@footnote\z@ }
13: \catcode`@=12
14: \def\sfootnote{\def\thefootnote{\fnsymbol{footnote}}}
15:
16: \def\Deltait{{\mit \Delta}}
17:
18: \newskip\humongous \humongous=0pt plus 1000pt minus 1000pt
19: \def\caja{\mathsurround=0pt}
20: \def\eqalign#1{\,\vcenter{\openup1\jot \caja
21: \ialign{\strut \hfil$\displaystyle{##}$&$
22: \displaystyle{{}##}$\hfil\crcr#1\crcr}}\,}
23: \newif\ifdtup
24: \def\panorama{\global\dtuptrue \openup1\jot \caja
25: \everycr{\noalign{\ifdtup \global\dtupfalse
26: \vskip-\lineskiplimit \vskip\normallineskiplimit
27: \else \penalty\interdisplaylinepenalty \fi}}}
28: \def\eqalignno#1{\panorama \tabskip=\humongous
29: \halign to\displaywidth{\hfil$\displaystyle{##}$
30: \tabskip=0pt&$\displaystyle{{}##}$\hfil
31: \tabskip=\humongous&\llap{$##$}\tabskip=0pt
32: \crcr#1\crcr}}
33:
34: \def\Dscr{{\cal D}}
35: \def\DsC{{\cal C}}
36: \def\DsS{{\cal S}}
37: \def\DsN{{\cal N}}
38: \def\DsE{{\cal E}}
39: \def\*{\hskip .06 cm}
40:
41: \def\thebibliography#1{\section*{\ \markboth
42: {REFERENCES}{REFERENCES}}\list
43: {{\arabic{enumi}}.}
44: {\settowidth\labelwidth{{#1}.}\leftmargin\labelwidth
45: \advance\leftmargin\labelsep
46: \usecounter{enumi}}
47: \def\newblock{\hskip .11em plus .33em minus -.07em}
48: \sloppy
49: \sfcode`\.=1000\relax}
50: \let\endthebibliography=\endlist
51: %
52: %
53: \def\thebibliographyp#1{\section*{\ \markboth
54: {Chan \& Dill, $\quad$ Polymer Principles in Protein Structure
55: and Stability}{Chan \& Dill, $\quad$ Polymer Principles in Protein Structure
56: and Stability}}\list
57: {\arabic{enumi}.}{\settowidth\labelwidth{#1.}\leftmargin\labelwidth
58: \advance\leftmargin\labelsep
59: \usecounter{enumi}}
60: \def\newblock{\hskip .11em plus .33em minus -.07em}
61: \sloppy
62: \sfcode`\.=1000\relax}
63: \let\endthebibliographyp=\endlist
64:
65: \def\sqr#1#2{{\vcenter{\vbox{\hrule height.#2pt
66: \hbox{\vrule width.#2pt height#1pt \kern#1pt
67: \vrule width.#2pt}
68: \hrule height.#2pt}}}}
69: \def\square{\mathchoice\sqr34\sqr34\sqr{2.1}3\sqr{1.5}3}
70:
71: %\def\@cite#1#2{({#1\if@tempswa , #2\fi})}
72:
73: \renewcommand{\textfraction}{0.0}
74: \renewcommand{\topfraction}{1}
75: \renewcommand{\bottomfraction}{1}
76: \setcounter{topnumber}{50}
77: \setcounter{bottomnumber}{50}
78: \setcounter{totalnumber}{50}
79: \setlength{\floatsep}{\baselineskip}
80: \setlength{\textfloatsep}{\baselineskip}
81: \renewcommand{\thefigure}{\arabic{figure}}
82: %
83: \def\baselinestretch{1.3}
84: \topmargin = +.1 in
85: \textheight 8.75 in
86: \oddsidemargin = 0.2in
87: \textwidth 450 pt
88: %
89: \begin{document}
90: %
91: %\titlepage
92: \noindent
93: $\null$
94: \hfill April 8, 2003
95: %
96:
97: \vskip 0.6in
98: %
99:
100: \begin{center}
101: %
102: {\Large\bf Contact Order Dependent Protein Folding Rates:}\\
103:
104: \vskip 0.3cm
105:
106: {\Large\bf Kinetic Consequences of a Cooperative Interplay}\\
107:
108: \vskip 0.3cm
109:
110: {\Large\bf Between Favorable Nonlocal Interactions and}\\
111:
112: \vskip 0.3cm
113:
114: {\Large\bf Local Conformational Preferences}\\
115:
116:
117:
118:
119: \vskip .5in
120: %
121: {\bf H\"useyin K{\footnotesize{\bf{AYA}}}}
122: and
123: {\bf Hue Sun C{\footnotesize{\bf{HAN}}}}$^\dagger$\\
124: %
125: $\null$
126:
127: Protein Engineering Network of Centres of Excellence (PENCE),\\
128: Department of Biochemistry, and
129: Department of Medical Genetics \& Microbiology,
130: Faculty of Medicine, University of Toronto,
131: Toronto, Ontario M5S 1A8, Canada\\
132:
133:
134: %
135:
136:
137: %{\tt Submitted to ""}
138: %{\tt To appear in ""}
139: %
140:
141: \end{center}
142: %
143:
144: \vskip 1cm
145:
146: \noindent
147: {\bf Running title:} Physics of Contact-Order Dependent Protein Folding \\
148:
149: \vskip 1cm
150:
151: \noindent {\bf Key words:}
152: calorimetry / chevron plot / G\=o models /
153: simple two-state kinetics /\\ single-domain proteins / nonadditivity
154:
155: $\null$\\
156: %
157:
158: %\vskip 0.8in
159:
160: \noindent
161: $^\dagger$ Corresponding author.\\
162: E-mail address of Hue Sun C{\footnotesize{HAN}}:
163: chan@arrhenius.med.toronto.edu\\
164: Tel: (416)978-2697; Fax: (416)978-8548\\
165: Mailing address: Department of Biochemistry, University of Toronto,
166: Medical Sciences Building -- 5th Fl., 1 King's College Circle,
167: Toronto, Ontario M5S 1A8, Canada.
168: %
169:
170: \vfill\eject
171: %\endtitlepage
172: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
173: %
174:
175: \def\thefootnote{\fnsymbol{footnote}}
176:
177: \noindent
178: {\large\bf Abstract}\\
179:
180: \vskip .2 in
181:
182: \noindent
183: Physical mechanisms underlying the empirical correlation between
184: relative contact order (CO) and folding rate among naturally-occurring
185: small single-domain proteins are investigated by evaluating postulated
186: interaction schemes for a set of three-dimensional 27mer lattice
187: protein models with 97 different CO values. Many-body interactions are
188: constructed such that contact energies become more favorable when short
189: chain segments sequentially adjacent to the contacting residues adopt
190: native-like conformations. At a given interaction strength, this
191: scheme leads to folding rates that are logarithmically well correlated
192: with CO (correlation coefficient $r=0.914$) and span more than 2.5 orders of
193: magnitude, whereas folding rates of the corresponding G\=o models with
194: additive contact energies have much less logarithmic correlation with CO
195: and span only approximately one order of magnitude.
196: The present protein chain models also exhibit calorimetric cooperativity
197: and linear chevron plots similar to that observed experimentally for
198: proteins with apparent
199: simple two-state folding/unfolding kinetics. Thus, our findings
200: suggest that CO-dependent folding rates of real proteins may arise
201: partly from a significant positive coupling between nonlocal contact
202: favorabilities and local conformational preferences.
203:
204: %**************************************************************************
205: \vfill\eject
206:
207: $\null$
208: \vskip -1cm
209:
210: %------------------------------------------------------------------------------
211:
212: \centerline{\bf INTRODUCTION}
213:
214: $\null$
215:
216: \noindent
217: {\bf Generic protein properties as energetic constraints}
218:
219: The folding of many small single-domain proteins is well approximated
220: by simple two-state thermodynamics and kinetics.$^{1,2}$ In
221: the past several years, we have shown that fundamental insights
222: into protein energetics can be gained by using these general, apparently
223: mundane properties as experimental constraints on protein chain
224: models.$^{3-10}$ This approach is based on the recognition
225: that model interaction schemes capable of producing these commonly observed
226: experimental properties are, somewhat surprisingly, not entirely
227: straightforward to come up with. To date, much advance has been made by
228: coarse-grained modeling of protein folding.$^{7,11-15}$ Nonetheless,
229: the interactions postulated by many existing models are insufficient for
230: calorimetric two-state cooperativity.$^{3,4}$ Furthermore, even common
231: G\=o models are not cooperative enough for simple two-state kinetics,
232: their explicit native biases notwithstanding. Specifically, we recently
233: found that several lattice$^{6,9,10}$ and continuum (off-lattice)$^{8}$
234: G\=o-like formulations with essentially additive interaction schemes all
235: led to chevron rollovers --- a hallmark of folding kinetics that are often
236: operationally referred to as non-two-state.$^{9}$ Apparently, many-body
237: interactions are needed to produce chevron plots with linear folding and
238: unfolding arms consistent with a two-state description of equilibrium
239: thermodynamics.$^{10}$
240: \\
241:
242: Small single-domain proteins are characterized as well by a significant
243: correlation between relative contact order (CO) and folding rate.$^{16}$
244: Therefore, it is only logical to require a model protein interaction
245: scheme to produce a similar correlation.$^{17,18}$ Ising-like$^{19,20}$ and
246: other$^{21,22}$ constructs without explicit chain representations have had
247: successes in this regard. However, as for thermodynamic and kinetic
248: cooperativities, achieving the CO dependence requirement in models with
249: explicit chain representations appears to be a nontrivial task.
250: Notably, an early lattice model study using a 20-letter alphabet suggested
251: that proteins with higher CO should fold faster,$^{23}$ thus predicting
252: a trend opposite$^{17}$ to that for real single-domain proteins.$^{16,18}$
253: A more recent 20-letter lattice model investigation, on the other hand,
254: found modest correlations between increasing CO and longer logarithmic
255: folding time (correlation coefficient $r\approx 0.70$--$0.79$ for chain
256: lengths $\ge 54$).$^{24}$ An earlier continuum G\=o model studies of 18
257: proteins also found a modest correlation between increasing CO and slower
258: logarithmic folding rates ($r=0.69$).$^{25}$ But the corresponding dispersion
259: in simulated folding rates covers only $\approx 1.5$ orders of magnitude,
260: which is much narrower than the $\approx 5$ orders of magnitude covered by
261: the real folding rates of the proteins in the given dataset. When a different
262: potential function was used in a more recent continuum G\=o model analysis,
263: however, no correlation between CO and simulated folding rates was
264: discerned.$^{26}$
265: \\
266:
267: Recently, based on lattice 27mer simulations, Jewett et al.$^{27}$ have
268: proposed that enhanced thermodynamic cooperativity and
269: many-body interactions --- which are basic properties of individual
270: two-state proteins to begin with$^{1-10}$ --- may also be a key to
271: understand the correlation between CO and folding rate across different
272: proteins. This is an attractive and insightful idea. However, the particular
273: way in which thermodynamic cooperativity was enhanced by these authors
274: led only to modest increases in folding rate dispersion relative to that
275: for the corresponding lattice G\=o models with pairwise additive contact
276: energies. Both the dispersion in folding rates and the correlation of
277: logarithmic folding rate with CO ($r=0.75$) for the most cooperative
278: interaction scheme they reported were
279: similar to that obtained from an earlier continuum G\=o model study,$^{25}$
280: as well as that from a recent simulation of 20-letter lattice models$^{24}$
281: with only pairwise additive contact energies (see above). In our view,
282: these results suggest that while CO-dependent folding may well derive from
283: certain intraprotein interactions that are also responsible for high
284: thermodynamic cooperativity, CO-dependent folding does not arise from
285: thermodynamic cooperativity {\it per se}. In other words, how cooperativity
286: is achieved can be critically important. Many {\it a priori} many-body
287: mechanisms are consistent with high thermodynamic cooperativity. An example
288: is the two rather different interaction schemes we considered
289: in ref.~10 --- one involves local-nonlocal coupling while the other
290: assigns an extra favorable energy to the ground-state structure
291: as a whole. But perhaps not all such mechanisms can mimic experimentally
292: observed CO dependencies to the same degree. Therefore, to shed light on
293: the physical mechanisms of CO-dependent folding, we endeavor to construct
294: an interaction scheme that would provide larger dispersions in folding
295: rates and better correlations with CO.
296: \\
297:
298: $\null$
299:
300: %------------------------------------------------------------------------------
301:
302: \centerline{\bf MODELS AND METHODS}
303:
304: $\null$
305:
306: The present study focuses on the idea of a cooperative interplay
307: between local conformational preferences and the contact-like interactions
308: that drive the packing of the protein core.$^{3,5,6,10}$ We have shown
309: that chain models embodying this idea can lead to calorimetric
310: cooperativity and simple two-state kinetics,$^{10}$ although our
311: exploration thus far has been limited to model proteins that are
312: mostly helical.$^{3,5,6,10}$ Here we consider a general formulation
313: of this idea, the basic ingredients of which are described by Fig.~1A.
314: This hypothesis may be viewed as a synthesis of the local-dominant and
315: the nonlocal-dominant perspectives.$^{28}$ We were motivated by the recognition
316: that both local$^{29,30}$ and nonlocal$^{31,32}$ intraprotein interactions
317: are important determinants of protein structure and stability. Yet local
318: conformational preferences alone are often insufficient for stable secondary
319: structures under physiological conditions. Secondary structure formation
320: is known to be context dependent;$^{33}$ they are stable when packed in the
321: core of a protein but are usually not stable in isolation (ref.~31 and
322: references therein). Furthermore, conformational space grows exponentially
323: with chain length, even when preferences arising from local excluded
324: volume effects are taken into account.$^{34}$ It follows that a large
325: part of the stability and uniqueness of protein native structures cannot
326: be explained by local interactions alone.$^{35}$ On the other hand,
327: our recent G\=o-model studies have shown that nonlocal contact-like
328: interactions by themselves are not cooperative enough for simple
329: two-state kinetics$^{6,8-10}$ if they are not coupled to local
330: conformational propensities.
331: \\
332:
333: \noindent
334: {\bf A simple model of local-nonlocal coupling}
335:
336: Here we explore the hypothesis in Fig.~1A by incorporating its form of
337: local-nonlocal coupling into a new interaction scheme in Fig.~1B
338: for explicit-chain models configured on three-dimensional simple
339: cubic lattices. This allows the idea to be tested quantitatively. Fig.~1B
340: may be viewed as a generalization of similar constructs we have
341: used previously in the context of helical proteins.$^{3,5,6,10}$
342: As a first step in our inquiry, we make the simplifying assumption
343: that the interactions are native-centric,$^{25-27,31,36-38}$ in that
344: only native interactions are favored, while nonnative interactions
345: are neutral (have zero energy). The local-nonlocal coupling in Fig.~1B
346: involves nonadditive many-body interactions. A chain segment which is
347: locally nativelike (with native bond and torsion angles) but make no
348: native contact is not stabilized (contributing zero energy). On the other
349: hand, nonlocal contact interactions between monomers far apart along
350: the chain sequence are more favorable when the chain segments around the
351: contacting residues are in their native conformations than when they are not.
352: As such, the present model differs from models that additively combine contact
353: energies and local favorabilities.$^{39}$ The importance of nonadditive
354: many-body effects in protein folding has been recognized,$^{3,5,6,10,40-44}$
355: but they have not been used extensively to model calorimetric two-state
356: cooperativity and linear chevron plots.$^{3-10}$ Our aim here is to
357: utilize extremely coarse-grained representations as a computationally
358: efficient means to explore the general principles linking CO-dependent
359: folding and proteinlike cooperativities. Many structural and
360: energetic details of real proteins are beyond the scope of this work.
361: In particular, the present work does not deal with the microscopic
362: physical origins of local-nonlocal coupling. Instead we just presume
363: that its presence in naturally occurring proteins could arise from
364: evolutionary design. Because of these, the simple interaction scheme
365: in Fig.~1B should be viewed only as a tentative model in this regard.
366: \\
367:
368: In order to examine the folding rates of a set of model proteins
369: whose native structures cover a diverse range of CO values, we now consider
370: chains of length $n=27$ configured on simple cubic lattices. For these
371: 27mers, there are 103,346 distinct maximally compact conformations
372: (not related by rotations or inversions)$^{45,46}$ confined to a
373: $3\times 3\times 3$ cube. The distribution of CO among these maximally
374: compact conformations covers 97 different values$^{27}$ from
375: CO $=$ $208/756=0.275$ to $402/756=0.532$ (inset of Fig.~2A,
376: where CO is computed using equation~1 of ref.~16). For each CO value,
377: we randomly choose a maximally compact 27mer conformation as
378: the native structure of a model protein (Table~I).\footnote{Since the
379: present choices of structures are independent of that by
380: Jewett et al.,$^{27}$ the structures listed in Table I do not necessarily
381: coincide with those used in their study.}
382: \\
383:
384: Folding and unfolding kinetics are modeled by standard Monte Carlo
385: simulations using the Metropolis criterion and the elementary
386: chain moves of end flips, corner flips, crankshafts, and rigid
387: rotations. The relative frequencies of attempting these moves are
388: 4.7\%, 58.3\%, 27\%, and 10\% respectively (c.f. ref.~6)\footnote{The
389: following typographical error in ref.~6 should be corrected. The relative
390: attempt frequencies of corner flips and crankshafts used in this prior
391: study of ours were, respectively, 60.6\% and 27\%, not the 27\% and 60.6\%
392: stated on p.~901 of ref.~6.}
393: Time is measured by the number of attempted Monte Carlo moves for a given
394: process. The set of elementary chain moves is chosen to mimic physically
395: plausible processes. Lattice model kinetics are dependent on
396: the choice of move set.$^{12}$ Nonetheless, we expect the general trend
397: predicted by the model is less sensitive to move set when kinetics are
398: not dominated by trapping events,$^{12}$ as is the case here and has
399: been verified by Jewett et al.$^{27}$
400: Progress towards the native state is tracked by the fractional
401: number of native contacts $Q$ (ref.~3--6). To ascertain the implications
402: of the local-nonlocal coupling we proposed, results from a highly cooperative
403: interaction scheme with $a=0.1$ are compared with that from the additive
404: scheme ($a=1$) of common G\=o models (c.f. Fig.~1B). Folding trajectories
405: are initiated at a randomly generated conformation; folding first passage
406: time is defined by the formation of the $Q=1$ ground-state conformation.
407: Unfolding trajectories are initiated at the ground-state conformation;
408: unfolding first passage time is the time it takes for the chain to
409: be left with three or fewer native contacts ($Q\le 3/28$); $Q=3/28$
410: is chosen to define unfolding because it coorresponds approximately
411: to the free energy minimum for the denatured state.
412: \\
413:
414: $\null$
415:
416: %------------------------------------------------------------------------------
417:
418: \centerline{\bf RESULTS}
419:
420: $\null$
421:
422: \noindent
423: {\bf Sensitivity of folding rate on CO enhanced by local-nonlocal coupling}
424:
425: Fig.~2 provides the correlation between CO and folding rate among
426: our 27mer models. It shows clearly that the local-nonlocal coupling
427: mechanism postulated in Fig.~1 can lead to a significant enhancement of
428: correlation as well as much increased sensitivity of folding rate to CO.
429: Whereas the dispersion in folding rates among the common additive
430: G\=o models in Fig.~2A covers only approximately one order of magnitude
431: (a factor of ten) and the logarithmic folding rates exhibit only a relatively
432: weak correlation with CO (correlation coefficient $r= 0.63$), the corresponding
433: dispersion among the $a=0.1$ cooperative models in Fig.~2B covers
434: approximately 2.5 to 3 orders of magnitude, with a strong correlation
435: between CO and logarithmic folding rate ($r=0.914$) comparable to that
436: observed among
437: a selection of real, small single-domain proteins.$^{18}$ Similar to
438: the corresponding experimental situations,$^{16,18}$ the comparisons in
439: Fig.~2 were performed under conditions for which folding relaxation is
440: essentially single-exponential, as is evident from the good agreements
441: in Fig.~2 between median first passage time divided by $\ln 2$
442: and the corresponding mean first passage time.$^{6,47}$ To better
443: delineate the effects of having weakened contact interactions when
444: the chain segments locally adjacent to the contacting residues are
445: nonnative, several $a$ values other than the $a=0.1$ used for the main
446: plot are compared in the inset of Fig.~2B. It shows CO-dependent
447: folding at different levels of local-nonlocal coupling (different
448: $a$ values) for several 27mers with representative CO's.
449: The $a=0$ case here corresponds to complete interdependence between
450: nonlocal contact and local structure. This inset indicates
451: that sensitivity of folding rate to CO increases
452: (the fitted line has a more negative slope) with decreasing $a$, and that
453: the behavior of the $a=0.1$ models is very similar to that of the
454: $a=0$ models. These results
455: further affirm that local-nonlocal coupling is a key ingredient for
456: the good correlation between CO and fold rate in these models. Nevertheless,
457: as for real proteins,$^{16,18}$ despite the good correlation, CO by itself
458: cannot predict folding rates of the present models with high accuracy.
459: Folding rates here can vary significantly for different structures with
460: the same CO as well. For example, for the particular 27mer with CO
461: $=346/756=0.458$ in Fig.~2B, the datapoint
462: $\log_{10}({\rm folding\ rate})=-5.75$ may be viewed as an ``outlier''
463: vis-\`a-vis the fitted line. However, for two other 27mers with the
464: same CO but do not belong to the randomly chosen set in Table~I (and
465: therefore not plotted and not used in the correlation analysis of
466: Fig.~2B), we found $\log_{10} ({\rm folding\ rate})=-7.26$ and $-7.60$,
467: which happen to be much closer to the fitted line in Fig.~2B.
468: The reasons behind variations in folding rates among structures
469: with same CO remain to be elucidated.
470: \\
471:
472: \noindent
473: {\bf A consistent model of thermodynamic and kinetic cooperativity}
474:
475: Fig.~3 provides further analyses of the folding/unfolding kinetics of
476: one example 27mer structure we choose to study in more detail.
477: Consistent with our previous results,$^{6,8-10}$ it shows that the model
478: chevron plot$^{48}$ predicted by the common additive G\=o potential (upper
479: plot) deviates significantly from simple two-state kinetics in that it
480: exhibits a severe rollover under only moderately native conditions. More
481: specifically, for this case rollover becomes significant at
482: ${\cal E}/k_{\rm B}T$ values that are only slightly more negative (more
483: favorable to folding) than that of the transition midpoint
484: (${\cal E}/k_{\rm B}T\approx -1.43$). In contrast, the chevron plot
485: predicted by the model with a substantial local-nonlocal coupling (lower
486: plot) is qualitatively similar to that of real, small single-domain
487: proteins that fold and unfold with simple two-state kinetics.$^{10}$
488: In particular, it has essentially linear folding and unfolding arms over an
489: extended range of ${\cal E}/k_{\rm B}T$ values. We have also obtained
490: for this model the equilibrium free energy of unfolding
491: $\Delta G_{\rm u}$ as a function of ${\cal E}/k_{\rm B}T$, where
492: $\Delta G_{\rm u}$ here is taken to be that between the unique $Q=1$
493: conformation and those with $Q\le 3/28$. (The same definition is used for
494: unfolding kinetics as stated above.) Because $\Delta G_{\rm u}$ is
495: essentially linear in ${\cal E}/k_{\rm B}T$, the linearity of the
496: chevron arms over an extended ${\cal E}/k_{\rm B}T$
497: range implies an essentially linear relationship between folding/unfolding
498: rates and $\Delta G_{\rm u}$ within the corresponding regime (i.e., the
499: model parameter ${\cal E}$ may be eliminated in favor of the lower horizontal
500: scale in Fig.~3). Furthermore, comparing the mean first passage times in Fig.~3
501: versus the corresponding median first passage times divided by $\ln 2$ shows
502: that folding or unfolding relaxation for this model is essentially
503: single exponential$^{6,47}$ for $\Delta G_{\rm u}<$ $10k_{\rm B}T$.
504: Essentially single-exponential folding under moderately folding conditions
505: is further demonstrated by an approximately linear logarithmic
506: distribution of first passage time$^{8,9,49}$ shown in the inset.
507: Similar to the cooperative models we recently investigated,$^{10}$
508: for the model with local-nonlocal coupling in Fig.~3, the thermodynamic
509: $\Delta G_{\rm u}$ values matches well with the kinetically obtained
510: quantity $-k_{\rm B}T\ln [({\rm folding\ rate})/({\rm unfolding\ rate})]$
511: for $\Delta G_{\rm u}$ ranging from $10k_{\rm B}T$ to $-6k_{\rm B}T$
512: (lower V-shape). In other words, the folding/unfolding kinetics of this
513: model is simple two-state$^{6,8-10}$ within a $\Delta G_{\rm u}$
514: range quite similar to that experimentally accessible to small
515: single-domain proteins.$^{10}$ Finally, the cooperative model in Fig.~3
516: is also calorimetrically two-state. Assuming that the interactions are
517: temperature independent, the model's van't Hoff to calorimetric enthalpy
518: ratio $\Delta H_{\rm vH}/\Delta H_{\rm cal}$ ($\kappa_2$ without baseline
519: subtraction$^{4}$) is determined to be $0.992$ (detailed calculation not
520: shown), satisfying the requirement of
521: $\Delta H_{\rm vH}/\Delta H_{\rm cal}\approx 1$ for two-state
522: thermodynamics.$^{3-5}$ Taken together, the above considerations imply that
523: the local-nonlocal coupling mechanism for enhanced CO-dependent folding
524: in Fig.~2B also provides --- as it should --- a consistent account of
525: thermodynamic and kinetic cooperativities$^{6,8-10}$ in simple two-state
526: proteins (Fig.~3).
527: \\
528:
529: As it stands, the transition midpoints of all 27mers considered here
530: with the local-nonlocal coupling parametrized by $a=0.1$ are very close
531: to one another. This is because the interaction scheme in Fig.~1B assigns
532: the same energy ($=28{\cal E}$) to every ground-state conformation.
533: This is a simplifying assumption in the present modeling setup.
534: Since the thermodynamic stabilities of real, small single-domain proteins
535: are quite diverse,$^{16,18}$ it is important to note that, in a broader
536: perspective, our hypothesis that significant CO-dependent folding can
537: emerge from local-nonlocal coupling is not contingent upon the different
538: proteins in question having very similar thermodynamic stabilities. In
539: more sophisticated models, for example, an extra favorable energy that
540: differs from one 27mer to another may be assigned to the ground-state
541: conformation (i.e., a different $E_{\rm gs}$ term as defined in ref.~10
542: for each 27mer). In that case, the
543: thermodynamic stabilities of different 27mers can be very different,
544: but their folding rates would not be affected by this extra feature of
545: the model. In other words, the correlation between CO and folding rate
546: in Fig.~2B would remain unchanged. As we have recently argued,$^{10}$
547: such extra stabilizing energies for the ground state as a whole are
548: physical plausible because experimental evidence$^{50}$ indicates that
549: in real proteins there is a partial separation between the driving
550: forces for folding kinetics and the interactions responsible for
551: thermodynamic stability.
552: \\
553:
554: $\null$
555:
556: %------------------------------------------------------------------------------
557:
558: \centerline{\bf DISCUSSION}
559:
560: $\null$
561:
562: Energy landscapes of the present models are further characterized in
563: Fig.~4 for three representative structures with low, intermediate,
564: and high CO values. In this figure, the low- and high-CO structures are,
565: respectively, the fastest and slowest folding among the 97 structures
566: in Table~I, whereas the intermediate-CO structure is the one analyzed
567: in Fig.~3. For the common additive G\=o potential, energy $E$ is
568: directly proportional to $Q$ ($E={\cal E}Q$). However, for the
569: cooperative models with local-nonlocal coupling, there are multiple
570: energy levels for each $Q$, with $E={\cal E}Q$ as the lower bound
571: (left panels of Fig.~4). This means that, on average, the energetic
572: separations between non-ground-state and ground-state conformations in the
573: cooperative models with local-nonlocal coupling are larger than that in the
574: additive G\=o models. This feature is demonstrated directly in the right
575: panels of Fig.~4, which show that the number of non-ground-state conformations
576: within a given energy range is smaller for the cooperative models than for
577: the additive G\=o models except for the highest energies ($E\approx 0$).
578: It follows that the overall thermodynamic cooperativities of the models
579: with local-nonlocal coupling are substantially higher than that of the
580: corresponding additive G\=o models. This behavior is expected as well from
581: our recent finding that simple two-state folding/unfolding kinetics (Fig.~3
582: above) requires ``near-Levinthal'' thermodynamic cooperativity.$^{10}$
583: Indeed, for the three models in Fig.~4 with local-nonlocal coupling,
584: the van't Hoff to calorimetric enthalpy ratios
585: $\Delta H_{\rm vH}/\Delta H_{\rm cal}$ are, from top to bottom,
586: $\kappa_2=$ $0.972$, $0.992$, and $0.998$. These values are extremely
587: high for model enthalpy ratios without baseline subtractions.$^{4}$
588: In contrast, the corresponding additive G\=o models are less cooperative,
589: with $\kappa_2=$ $0.751$, $0.861$, and $0.878$. Here it is noteworthy
590: that the additive G\=o models' $\Delta H_{\rm vH}/\Delta H_{\rm cal}$
591: ratios even after empirical baseline subtractions,$^{4}$
592: $\kappa_2^{({\rm s})}=$ $0.885$, $0.961$, and $0.962$, are lower than
593: the $\Delta H_{\rm vH}/\Delta H_{\rm cal}$ ratios of the
594: cooperative models in the absence of baseline subtractions.
595: \\
596:
597: \noindent
598: {\bf Contact-order dependence indicative of special mechanisms
599: of cooperativity}\\
600:
601: Obviously, thermodynamic cooperativity is a necessary ingredient
602: for any protein chain model that purports to rationalize the
603: generic properties of small single-domain proteins.$^{3-10}$
604: For the particular interaction scheme we consider, the above analysis shows
605: that features that give rise to significant CO-dependent folding also
606: lead to high thermodynamic cooperativity. However, the converse
607: is not necessarily true. More in-depth considerations and a comparison
608: of the present results with that of Jewett et al.$^{27}$ indicate that
609: higher thermodynamic cooperativity {\it per se} does not necessarily give
610: rise to more enhanced dependence of folding rate on CO. Our reasoning
611: is as follows. First, for the present set of 27mer structures we have
612: chosen randomly, the correlation between logarithmic folding rate and CO
613: is quantified by $r=0.63$ ($r^2=0.39$) for the additive G\=o interaction
614: scheme. Despite that this correlation happens to be weaker than that
615: of Jewett et al.'s collection of additive G\=o models (their $r^2=0.51$),
616: after cooperativity is enhanced by local-nonlocal coupling, the correlation
617: between logarithmic folding rate and CO for our $a=0.1$ models is much
618: higher ($r^2=0.84$, see Fig.~2 above, an improvement in $r^2$ value
619: of $0.33$)\footnote{
620: Because all the model chains in the present study have the same length
621: and the same number of native contacts, their correlation coefficient
622: between folding rate and CO is the same as that between folding rate
623: and the total contact distance (TCD) defined in ref.~51.
624: }
625: than the best case reported by Jewett et al.$^{27}$ ($r^2=0.57$
626: for their $s=3$, an improvement in $r^2$
627: value of $0.06$ over that for their additive G\=o models).\footnote{
628: If the $s=3$ interaction scheme of Jewett et al. is applied to the
629: present set of structures and kinetic models, we found $r^2=0.65$
630: for the correlation between CO and folding rate. In
631: this case, the folding rates span $\approx 1.8$ orders of magnitude;
632: see ref.~52 for details.
633: }
634: Second, the folding rates of our cooperative models are much more
635: sensitive to CO, covering 2.5 to 3 orders of magnitude, whereas those
636: of Jewett et al. cover only approximately 1.3 orders of magnitude.
637: This means that the present local-nonlocal coupling mechanism is
638: significantly more effective in enhancing CO dependence than the
639: nonlinear $E$--$Q$ relationship postulated by Jewett et al. (equation~1
640: of ref.~27). Their interaction scheme does not make direct reference to
641: chain conformations as such. Thermodynamic cooperativity is
642: enhanced in their models by stipulating that the total contact energy $E$ (for
643: a given conformation as a whole) does not decrease (does not become more
644: favorable) linearly with increasing $Q$ as in common G\=o models;
645: but rather decreases at progressively faster and faster rates when $Q$
646: is closer to unity.\footnote{
647: Jewett et al. suggested that the ``extraordinary
648: cooperativity in protein folding'' may originate from ``three-body
649: interactions.'' But how three-body interactions might lead to their
650: $E$--$Q$ relationship remains to be elucidated.}
651: Third, in fact, if thermodynamic cooperativity is further increased
652: in the interaction scheme of Jewett et al. by increasing their $s$ parameter,
653: the energy landscape will eventually become a Levinthal golf course in the
654: $s\rightarrow\infty$ limit. In that case, folding would be rate-limited
655: by random conformational search and CO-dependence would be all but
656: eliminated. Fourth, in this connection, we have recently considered three
657: 27mer models with CO $=0.28$, $0.40$ and $0.51$ in a separate study. The
658: thermodynamic cooperativity of these models are enhanced by
659: assigning an extra stabilizing energy to the ground state but without
660: local-nonlocal coupling.$^{10}$ For the energetic parameters we considered,
661: the folding rates of these models cover less than an order of
662: magnitude.$^{10}$ The same set of results also indicated that dispersion
663: in folding rates under moderately folding conditions would decrease if
664: thermodynamic cooperativity is increased by assigning an even stronger
665: stabilizing energy to the ground state, in a manner similar to greatly
666: increasing $s$ in Jewett et al.'s formulation. Taken together, these
667: observations lead us to the conclusion that while thermodynamic
668: cooperativity is certainly necessary, by itself it is not sufficient
669: to guarantee CO-dependent folding rates similar to that observed
670: experimentally$^{16,18}$ if the underlying mechanism for thermodynamic
671: cooperativity is not specified.
672: \\
673:
674: CO-dependent folding highlights the important role of local interactions
675: in determining folding rates.$^{16-18}$ It suggests that the mechanism
676: of folding may involve relatively fast formation of local structure.
677: In this regard, we note that under the general lattice scheme in Fig.~1B,
678: formation of strong (unattenuated) native contacts with contact order
679: $\vert j - i \vert=3$ is relatively easier than formation of strong
680: native contacts with higher contact orders. This is because in the
681: $\vert j - i \vert=3$ case there is an overlap between parts of the two
682: local segments that have to be nativelike in order
683: for the contact to be strong. Physically, how a general
684: mechanism similar to that in Fig.~1 may arise in real proteins from
685: solvent-mediated atomic interactions such as sidechain packing
686: and hydrogen bonding remains to be elucidated.
687: Many basic issues will have to be tackled to address this question.
688: For example, correlations between backbone and sidechain rotamer
689: conformations$^{53}$ may contribute to such a mechanism.
690: Another possibility is that aspects of {\it anti-cooperativity}
691: of certain types of hydrophobic interactions$^{54}$ may help disfavor
692: premature nonspecific hydrophobic collapse (which would lead to kinetic
693: trapping$^{14}$) when the sidechains are locally less well packed than
694: that in the native state. If this is the case, it could give rise to
695: local-nonlocal coupling mechanisms similar to that postulated in Fig.~1.
696: \\
697:
698: In summary, while the models used in the present study are rudimentary,
699: they provide strong evidence that a cooperative interplay between local
700: conformational preferences and nonlocal favorable contact-like
701: interactions is an important mechanism in accounting for experimentally
702: observed CO-dependent folding of small single-domain proteins.
703: We are optimistic that more rigorous applications of
704: the CO-dependence constraint as well as the thermodynamic and kinetic
705: cooperativity requirements would help further narrow down theoretical
706: possibilities and thus contribute to a more realistic understanding of protein
707: energetics.
708: \\
709:
710:
711:
712: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
713:
714: %---------------------------------------------------------------------------
715:
716:
717: $\null$
718:
719: %===========================================================================
720:
721: \noindent
722: {\Large Acknowledgments.}
723: We thank Robert L. Baldwin, Alan Davidson, Teresa Head-Gordon, Michael
724: Levitt, Vijay Pande, Kevin Plaxco, Steve Plotkin, Wes Stites and Yaoqi Zhou
725: for helpful discussions, and Vijay Pande and Kevin Plaxco for kindly sharing
726: their work (ref.~27) before publication. The research reported
727: here was partially supported by the Canadian Institutes of Health
728: Research (CIHR grant no. MOP-15323), a Premier's Research Excellence Award
729: from the Province of Ontario, and the Ontario Centre for Genomic Computing
730: at the Hospital for Sick Children in Toronto. H. S. C. is a Canada
731: Research Chair in Biochemistry.
732:
733: %===========================================================================
734: \vfill\eject
735:
736:
737: \par\vfill\eject
738:
739: \noindent
740: {\large\bf References}
741:
742: \kern -1.5cm
743:
744: \begin{thebibliography}{99}
745:
746: %XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
747:
748: \bibitem{1}
749: Jackson SE, Fersht AR.
750: Folding of chymotrypsin inhibitor 2. 1. Evidence for a two-state
751: transition. Biochemistry 1991;30:10428--10435.
752:
753: \bibitem{2}
754: Baker D.
755: A surprising simplicity to protein folding. Nature 2000;405:39--42.
756:
757: \bibitem{3}
758: Chan HS.
759: Modeling protein density of states: Additive hydrophobic
760: effects are insufficient for calorimetric two-state cooperativity.
761: Proteins 2000;40:543--571.
762:
763: \bibitem{4}
764: Kaya H, Chan HS.
765: Polymer principles of protein calorimetric
766: two-state cooperativity.
767: Proteins 2000;40:637--661 [Erratum: Proteins 2001;43:523].
768:
769: \bibitem{5}
770: Kaya H, Chan HS.
771: Energetic components of cooperative protein folding.
772: Phys Rev Lett 2000;85:4823--4826.
773:
774: \bibitem{6}
775: Kaya H, Chan HS.
776: Towards a consistent modeling of protein thermodynamic and kinetic
777: cooperativity: How applicable is the transition state picture to
778: folding and unfolding? J Mol Biol 2002;315:899--909.
779:
780: \bibitem{7}
781: Chan HS, Kaya H, Shimizu S. Computational methods
782: for protein folding: scaling a hierarchy of complexities.
783: In: Jiang T, Xu Y, Zhang MQ, editors. Current Topics in Computational
784: Molecular Biology. Cambridge, MA: The MIT Press; 2002. p 403--447.
785:
786: \bibitem{8}
787: Kaya H, Chan HS.
788: Solvation effects and driving forces for protein thermodynamic and
789: kinetic cooperativity: How adequate is native-centric topological
790: modeling? J Mol Biol 2003;326:911--931.
791:
792: \bibitem{9}
793: Kaya H, Chan HS.
794: Origins of chevron rollovers in non-two-state protein folding kinetics.
795: Submitted (2003); [cond-mat/0302305,\\
796: {\tt http://xxx.lanl.gov/abs/cond-mat/0302305}].
797:
798: \bibitem{10}
799: Kaya H, Chan HS.
800: Simple two-state protein folding kinetics requires near-Levinthal
801: thermodynamic cooperativity. Submitted (2003);
802: [cond-mat/0302306,
803: {\tt http://xxx.lanl.gov/abs/cond-mat/0302306}].
804:
805: \bibitem{11}
806: Bryngelson JD, Onuchic JN, Socci ND, Wolynes PG.
807: Funnels, pathways, and the energy landscape of protein folding: A
808: synthesis. Proteins 1995;21:167--195.
809:
810: \bibitem{12}
811: Dill KA, Bromberg S, Yue K, Fiebig KM, Yee DP, Thomas PD,
812: Chan HS.
813: Principles of protein folding --- A perspective from simple
814: exact models. Protein Sci. 1995;4:561--602.
815:
816: \bibitem{13}
817: Thirumalai D, Woodson SA. Kinetics of folding of proteins
818: and RNA. Acc Chem Res 1996;29:433--439.
819:
820: \bibitem{14}
821: Chan HS, Dill KA. Protein folding in the landscape
822: perspective: Chevron plots and non-Arrhenius kinetics.
823: Proteins 1998;30:2--33.
824:
825: \bibitem{15}
826: Mirny L, Shakhnovich E.
827: Protein folding theory: From lattice to all-atom models.
828: Annu Rev Biophys Biomol Struct 2001;30:361--396.
829:
830: \bibitem{16}
831: Plaxco KW, Simons KT, Baker D.
832: Contact order, transition state placement and the refolding rates
833: of single domain proteins. J Mol Biol 1998;227:985--994.
834:
835: \bibitem{17}
836: Chan HS.
837: Matching speed and locality. Nature 1998;392:761--763.
838:
839: \bibitem{18}
840: Plaxco KW, Simons KT, Ruczinski I, Baker D. (2000). Topology,
841: stability, sequence, and length: Defining the determinants of two-state
842: protein folding kinetics. Biochemistry 2000;39:11177--11183.
843:
844: \bibitem{19}
845: Alm E, Baker D. Prediction of protein-folding mechanisms from
846: free-energy landscapes derived from native structures.
847: Proc Natl Acad Sci USA 1999;96:11305--11310.
848:
849: \bibitem{20}
850: Mu{\~n}oz V, Eaton WA.
851: A simple model for calculating the kinetics of
852: protein folding from three-dimensional structures.
853: Proc Natl Acad Sci USA 1999;96:11311--11316.
854:
855: \bibitem{21}
856: Debe DA, Goddard WA.
857: First principles prediction of protein folding
858: rates. J Mol Biol 1999;294:619--625.
859:
860: \bibitem{22}
861: Makarov DE, Keller CA, Plaxco KW, Metiu H.
862: How the folding rate constant of simple, single-domain proteins depends
863: on the number of native contacts.
864: Proc Natl Acad Sci USA 2002;99:3535--3539.
865:
866: \bibitem{23}
867: Abkevich VI, Gutin AM, Shakhnovich EI.
868: Impact of local and nonlocal interactions on thermodynamics and kinetics
869: of protein folding. J Mol Biol 1995:252:460--471.
870:
871: \bibitem{24}
872: Faisca PFN, Ball RC. Topological complexity, contact order, and
873: protein folding rates. J Chem Phys 2002;117:8587--8591.
874:
875: \bibitem{25}
876: Koga N, Takada S.
877: Roles of native topology and chain-length scaling in protein folding:
878: A simulation study with a G\=o-like model.
879: J Mol Biol 2001;313:171--180.
880:
881: \bibitem{26}
882: Cieplak M, Hoang TX.
883: Universality classes in folding times of proteins.
884: Biophys J 2003;84:475--488.
885:
886: \bibitem{27}
887: Jewett AI, Pande VS, Plaxco KW.
888: Cooperativity, smooth energy landscapes and the origins of
889: topology-dependent protein folding rates. J Mol Biol 2003;326:247--253.
890:
891: \bibitem{28}
892: Uversky VN, Fink AL.
893: The chicken-egg scenario of protein folding revisited.
894: FEBS Lett 2002;515:79--83.
895:
896: \bibitem{29}
897: Baldwin RL, Rose GD.
898: Is protein folding hierarchic? I. Local structure and peptide
899: folding. Trends Biochem Sci 1999;24:26--33.
900:
901: \bibitem{30}
902: Shortle D. Composites of local structure propensities:
903: Evidence for local encoding of
904: long-range structure. Protein Sci 2002;11:18--26.
905:
906: \bibitem{31}
907: G\=o N, Taketomi H.
908: Respective roles of short- and long-range interactions in protein folding.
909: Proc Natl Acad Sci USA 1978;75:559--563.
910:
911: \bibitem{32}
912: Dill KA.
913: Dominant forces in protein folding.
914: Biochemistry 1990;29:7133--7155.
915:
916: \bibitem{33}
917: Minor DL, Kim PS.
918: Context-dependent secondary structure formation of a designed protein
919: sequence. Nature 1996;380:730--734.
920:
921: \bibitem{34}
922: Feldman HJ, Hogue CWV.
923: Probabilistic sampling of protein conformations: New hope for brute force?
924: Proteins 2002;46:8--23.
925:
926: \bibitem{35}
927: Shimizu S, Chan HS.
928: Origins of protein denatured state compactness and hydrophobic
929: clustering in aqueous urea: Inferences from nonpolar potentials of
930: mean force. Proteins 2002;49:560--566.
931:
932: \bibitem{36}
933: Micheletti C, Banavar JR, Maritan A, Seno F.
934: Protein structures and optimal folding from a geometrical variational
935: principle. Phys Rev Lett 1999;82:3372--3375.
936:
937: \bibitem{37}
938: Clementi C, Nymeyer H, Onuchic JN. Topological and energetic
939: factors: What determines the structural details of the transition state
940: ensemble and ``en-route'' intermediates for protein folding? An investigation
941: for small globular proteins. J Mol Biol 2000;298:937--953.
942:
943: \bibitem{38}
944: Linhananta A, Zhou Y.
945: The role of sidechain packing and native contact interactions in folding:
946: Discontinuous molecular dynamics folding simulations of an all-atom
947: G\=o model of fragment B of {\it Staphylococcal} protein A.
948: J Chem Phys 2002;117:8983--8995.
949:
950: \bibitem{39}
951: Thomas PD, Dill KA.
952: Local and nonlocal interactions in globular proteins and
953: mechanisms of alcohol denaturation.
954: Protein Sci 1993;2:2050--2065.
955:
956: \bibitem{40}
957: Kolinski A, Galazka W, Skolnick J.
958: On the origin of the cooperativity of protein folding: Implications
959: from model simulations. Proteins 1996;26:271--287.
960:
961: \bibitem{41}
962: Plotkin SS, Wang J, Wolynes PG.
963: Statistical mechanics of a correlated energy landscape model for protein
964: folding funnels. J Chem Phys 1997;106:2932--2948.
965:
966: \bibitem{42}
967: Liwo A, Kazmierkiewicz R, Czaplewski C, Groth M, Oldziej S, Wawak RJ,
968: Rackovsky S, Pincus MR, Scheraga HA. United-residue force field for
969: off-lattice protein structure simulations: III. Origin of backbone
970: hydrogen-bonding cooperativity in united-residue potentials.
971: J Comput Chem 1998;19:259--276.
972:
973: \bibitem{43}
974: Takada S, Luthey-Schulten Z, Wolynes PG.
975: Folding dynamics with nonadditive forces: A simulation study
976: of a designed helical protein and a random heteropolymer.
977: J Chem Phys 1999;110:11616--11629.
978:
979: \bibitem{44}
980: Eastwood MP, Wolynes PG.
981: Role of explicitly cooperative interactions in protein folding funnels:
982: A simulation study. J Chem Phys 2001;114:4702--4716.
983:
984: \bibitem{45}
985: Chan HS, Dill KA.
986: The effect of internal constraints on the configurations of chain
987: molecules. J Chem Phys 1990;92:3118--3135 [Erratum: J Chem Phys
988: 1997;107:10353].
989:
990: \bibitem{46}
991: Chan HS, Bornberg-Bauer E.
992: Perspectives on protein evolution from simple exact models.
993: Applied Bioinformatics 2002;1:121-144.
994:
995: \bibitem{47}
996: Gutin A, Sali A, Abkevich V, Karplus M, Shakhnovich EI.
997: Temperature dependence of the folding rate in a simple protein model:
998: Search for a ``glass'' transition.
999: J Chem Phys 1998;108:6466--6483.
1000:
1001: \bibitem{48}
1002: Matthews CR.
1003: Effect of point mutations on the folding of globular proteins.
1004: Methods Enzymol 1987;154:498--511.
1005:
1006: \bibitem{49}
1007: Abkevich VI, Gutin AM, Shakhnovich EI. Free energy
1008: landscape for protein folding kinetics: Intermediates, traps, and multiple
1009: pathways in theory and lattice model simulations.
1010: J Chem Phys 1994;101:6052--6062.
1011:
1012: \bibitem{50}
1013: Northey JGB, Di Nardo AA, Davidson AR.
1014: Hydrophobic core packing in the SH3 domain folding transition state.
1015: Nature Struct Biol 2002;9:126--130.
1016:
1017: \bibitem{51}
1018: Zhou HY, Zhou YQ. Folding rate prediction using total contact distance.
1019: Biophys J 2002;82:458--463.
1020:
1021: \bibitem{52}
1022: Chan HS, Shimizu S, Kaya H. Cooperativity principles in protein folding.
1023: Methods Enzymol, in press.
1024:
1025: \bibitem{53}
1026: Dunbrack RL. Rotamer libraries in the 21st century.
1027: Curr Opin Struct Biol 2002;12:431--440.
1028:
1029: \bibitem{54}
1030: Shimizu S, Chan HS. Anti-cooperativity and cooperativity
1031: in hydrophobic interactions: Three-body free energy landscapes and
1032: comparison with implicit-solvent potential functions for proteins.
1033: Proteins 2002;48:15--30 [Erratum: Proteins 2002;49:294].
1034:
1035:
1036: \end{thebibliography}
1037:
1038: %------------------------------------------------------------------------
1039: \vfill\eject
1040:
1041: \centerline{\large \bf Table I}
1042: \vskip .2 in
1043:
1044:
1045: {\footnotesize
1046:
1047: \begin{center}
1048: \begin{tabular}{|cc|cc|}
1049: \hline
1050: $\sum\Delta S_{ij}$ & conformation & $\sum\Delta S_{ij}$ & conformation \\
1051: \hline
1052: 208 & uufddfuurddbuubddruufddfuu & 306 & uufrrbldrfdflurullddburdbr \\
1053: 210 & uufddfuurbbdffdbbrffuubdbu & 308 & uufdfrbrbulddrfllfrruublfl \\
1054: 212 & ufdfuubbrddffubufrddbuubdd & 310 & ufrulblfddrrbbllfuburdrfub \\
1055: 214 & uuffdbdfrbufubbddrffuubdbu & 312 & uufrbbllffdrrdllbubdrurfdb \\
1056: 216 & ufdfuubbrddfuufddruubbddfu & 314 & uufdrubbdfdfllbbuffubbrddr \\
1057: 218 & ufdfuubbrdfufddbbruuffddbu & 316 & uffrddblbruufdllbdffrulubb \\
1058: 220 & uuffddburfdbbuuffrddbuubdd & 318 & uufrbddbuullffdrrdllbubdru \\
1059: 222 & uuffddburdfuubbddruuffdbdf & 320 & uufddfrruubbdfdbluuffldrbd \\
1060: 224 & uufddrbufubrfdbdfflurulldd & 322 & uffdbrbrfufullbbrdrufldfdr \\
1061: 226 & uufddfuurddbubdrffubbulfrf & 324 & ufrbddlfrflurullbbddffubrr \\
1062: 228 & uffdbrbuffdrbbuffubbllfrfl & 326 & ufrfddluulddbbuufdrdbrfubu \\
1063: 230 & ufdrbufublffddruurddbuubdd & 328 & ufrubdbuldldrrffllbufurblb \\
1064: 232 & uufrrblddrufdluldfurdruull & 330 & uufdrubblddlfubuffddrrbbuf \\
1065: 234 & uufddfuurddrbluurfdbbulddr & 332 & uuffrddruubbdfllfdbrrbluuf \\
1066: 236 & ufddbbuurrflfrdlbdfrbubldr & 334 & ufrrdbdfluldbbrruuflbldrfd \\
1067: 238 & uuffdbdfrrblbrulffrulbbrfd & 336 & uffrbrbuflblffrrddllbrrblu \\
1068: 240 & uffdbrfurdbblufrbuffllbbrf & 338 & uffurrdldrbblurullfrrdldlf \\
1069: 242 & uufrbddfflburflurrbbdffdbb & 340 & ufrrbbdffdlbrbllffurbubldr \\
1070: 244 & ufdrrbluulfrdrbuffllddrrul & 342 & ufrullbrrblldrrdllfufdrrbu \\
1071: 246 & ufdfurbdfruullbrrddblurull & 344 & uffdrdllbrbluuffdbrrdbuuff \\
1072: 248 & ufdfrbuflurblbrrdldrffuubd & 346 & uuffddrrbbuufdfuldbubddflu \\
1073: 250 & ufddbrfruublfdbrdblluurdru & 348 & uufrbbdlulddrrffllbuufdrrb \\
1074: 252 & uffdbrrflurbbdlufufrbbllff & 350 & ufubrrdfdfuldblfuurrbldbdr \\
1075: 254 & ufddbrblurrdfflubrfulbrbll & 352 & uffddrbllurrfubbddlluuffdd \\
1076: 256 & ufdfurdruullbbrddrfluurdbu & 354 & ufrfdrbufubbllfrflddbrburd \\
1077: 258 & ufrbdflfrrbbuullfrrdfulldr & 356 & uffrrbdbuullffrrbldbdflfrr \\
1078: 260 & uufddfurbbrdlffrbufubblffl & 358 & uufdrrubddffuulldrdlbrbuuf \\
1079: 262 & uuffdbrbufrfldrdllbrbrfubu & 360 & ufrddllfrruulldrblubddrruu \\
1080: 264 & ufdrurddbuuldblurrddllffrb & 362 & ufubrrdffldrbblflfuurrbldb \\
1081: 266 & ufdrurddllbrbluurrfldbrdfu & 364 & ufrfddlbblffubbuffrdbrdbuu \\
1082: 268 & uuffdbrubrfddbuldflfrrulur & 366 & ufrfdbdfllbbuufdfurdbdbruu \\
1083: 270 & uuffrrdllbdrbufrulbrddffll & 368 & uffurrbbddffllbrbuulfrdrfl \\
1084: 272 & uufrdfuldbdfrruubblddfrubd & 370 & uffurrddbbuufllbrddflfrubr \\
1085: 274 & ufdrubrfddllbbuurrdldrfuld & 372 & uufdrfdruubbddluufflddbrru \\
1086: 276 & ufdrbdlfrrubdblluurffrbbdl & 374 & uffrdrbbuullffrrdbuldbdflf \\
1087: \end{tabular}
1088: \end{center}
1089: \vskip .15 in
1090:
1091: }
1092:
1093: \vfill
1094:
1095: $\null$ \hfill $\dots$ {\it Table I to be cont'd}
1096:
1097: \vfill\eject
1098:
1099: \noindent{\large \bf Table I} $\dots$ ({\it cont'd from last page})
1100: \vskip .2 in
1101:
1102: {\footnotesize
1103:
1104: \begin{center}
1105: \begin{tabular}{|cc|cc|}
1106: 278 & ufddrrbllbrrullurrfflbdfrb & 376 & uffrddllbuubddrfrbuufdlflu \\
1107: 280 & uffrddlubdruubddllfubuffdd & 378 & ufdrfdlluubbdfdbrfrbuuffld \\
1108: 282 & ufrbdffuldlubbddrrfflbuldf & 380 & ufrfddlbrbllffubbuffrdbrbu \\
1109: 284 & uufrfldrrubbldrfdblfuldfrr & 382 & uufdrfdrbbuufdfullddbrbuuf \\
1110: 286 & uffubbrddrffuldlbrurbufflb & 384 & ufrbbullddfuurrfllddrrbblu \\
1111: 288 & ufdfrrubufldlubbrfdbdfrbuu & 386 & uffddrbllfuubbddrruuffdbll \\
1112: 290 & uufrbbldrfdbllfubuffddrurd & 388 & uffrburbddffllbrbuulffrrdb \\
1113: 292 & uffrddbbuufdldblffrulubbdf & 390 & ufrufddrbbuffubbllffddbrbu \\
1114: 294 & ufdrdfulurbbddlluufddrfluu & 392 & ufrufddrbbuffubbllffddbrbu \\
1115: 296 & ufdfrbdflbbruuffllddbbuufd & 394 & ufrrddlbburuflblddffurbrdb \\
1116: 298 & ufrrdblblurfrbddffluldbrbl & 396 & uffrddblflbufubbddrruufdlf \\
1117: 300 & ufdfrullddbuubddrffrbuubdd & 400 & ufrfddllubdrrblluuffrdbrbu \\
1118: 302 & ufdfurddlluubbdfdbrfrbuufd & 402 & ufrufrbbddffllbrbuulffdrrb \\
1119: 304 & ufrdlluurrbbdfdbllfuubdruf & & \\
1120: \hline
1121: \end{tabular}
1122: \end{center}
1123: \vskip .15 in
1124:
1125: }
1126: {\noindent {{\bf Table~I.}}} $\quad$
1127: The ground-state 27mer conformations ($n=27$) used in this investigation are
1128: given by sequences of 26 bond directions, where r = right ($+x$),
1129: l = left ($-x$), f = forward ($+y$), b = backward ($-y$), u = up ($+z$),
1130: d = down ($-z$). A structure is randomly selected for each of the 97 possible
1131: CO values amongst the compact 27mer structures with
1132: $t_{\rm max}=28$ contacts. Each integer $\sum\Delta S_{ij}$ is
1133: the sum of $\vert j-i \vert$ over the $(i,j)$ nearest-neighbor
1134: contacts in the given conformation ($j-i\ge 3$).
1135: Here CO $=\sum\Delta S_{ij}/(n t_{\rm max})$ $=\sum\Delta S_{ij}/756$.
1136:
1137: %------------------------------------------------------------------------
1138: \vfill\eject
1139:
1140: \noindent
1141: {\large\bf Figure Captions}\\
1142:
1143: \noindent
1144: {\bf Figure 1.} $\quad$
1145: (A) Schematics of local-nonlocal cooperative energetics in protein
1146: folding. The conformation in the solid box represents the native (N)
1147: structure; the two filled circles depict a pair of nonlocal residues
1148: interacting favorably in the native state. The interaction strength
1149: between a residue pair is strong and essentially the same as that in
1150: the native structure if the chain segments sequentially local
1151: to both residues are nativelike, as in (i). [Dotted boxes in (A) are
1152: used to mark nativelike chain segments.] However, the interaction strength is
1153: weakened if one or two chain segments sequentially local to the
1154: interacting residues are not nativelike, as in examples (ii)--(iv).
1155: (B) A lattice implementation of this protein folding scenario. Here
1156: the favorable energy for every contact (between residues $i$ and $j$,
1157: $\vert j-i \vert \ge 3$) in the ground-state native (N) structure is
1158: ${\cal E}$ ($<0$) when the relative positions of the five residues centered
1159: at $i$ (residues $i-2$, $i-1$, $i$, $i+1$, and $i+2$) as well as the relative
1160: positions of five residues centered at $j$ (residues $j-2$, $j-1$,
1161: $j$, $j+1$, and $j+2$) are the same as that in N [solid lines in (i)],
1162: irrespective of the relative orientations of the two five-residue chain
1163: segments. However, if the local conformation of one or both sets of
1164: five contiguous residues is nonnative, the contact energy is weakened by an
1165: attentuation factor $a$ ($0\le a< 1$). Examples of the latter situation
1166: is given by (ii)--(iv), where nonnative local chain segments are
1167: drawn as broken lines.
1168: \\
1169:
1170: \noindent
1171: {\bf Figure 2.} $\quad$
1172: Correlation between the common (base 10) logarithm of folding rate
1173: and CO for the 97 structures in Table~I under moderately folding
1174: conditions at ${\cal E}/k_{\rm B}T=-1.47$, using (A) the
1175: common additive G\=o potential and (B) the local-nonlocal
1176: cooperative interaction scheme with $a=0.1$. Solid lines are least-square
1177: fits. Here folding rate is the reciprocal of mean folding first passage
1178: time (folding rate $=$ 1/MFPT). Each MFPT is averaged from 500
1179: trajectories. Associated with each value of $\log_{10}(1/{\rm MFPT})$
1180: (filled circle) is an open circle marking the common logarithm of the
1181: median folding first passage time (FPT) divided by $\ln 2$. If the
1182: kinetics is single-exponential, MFPT $=$ (median FPT)/$\ln 2$.
1183: The inset in (A) is the distribution of CO among the
1184: 103,346 maximally compact 27mer conformations, wherein the number of
1185: conformations (vertical scale) is shown as a function of CO (horizontal
1186: scale). The inset in (B) uses six
1187: representative structures with different CO values ($\sum\Delta S_{ij}=$
1188: 208, 224, 268, 310, 348, and 386 entries in Table~I) to illustrate that
1189: $\log_{10}({\rm folding\ rate})$ (vertical scale) is more sensitive to CO
1190: (horizontal scale) when the local-nonlocal coupling is stronger. In this
1191: inset, different symbols denote different $a$ values; the lines fitted
1192: through the symbol are, from top to bottom, for $a=1$, $0.75$, $0.5$,
1193: $0.25$, $0.1$, and $0.0$.
1194: \\
1195:
1196: \noindent
1197: {\bf Figure 3.} $\quad$
1198: Model chevron plots for a CO $=0.410$ structure ($\sum\Delta S_{ij}=310$
1199: entry in Table~I) are given by negative natural logarithm of MFPT as a
1200: function of ${\cal E}/k_{\rm B}T$ (filled symbols). Values
1201: of (median FPT)/$\ln 2$ are shown by the open symbols. Squares (folding)
1202: and triangles (unfolding) are for the additive G\=o potential
1203: ($a=1$, upper plot), whereas circles (folding) and diamonds (unfolding)
1204: are for the $a=0.1$ local-nonlocal cooperative interaction scheme (lower
1205: plot). Each MFPT is averaged from 500 trajectories, except for the model
1206: with local-nonlocal coupling at ${\cal E}/k_{\rm B}T=-1.47$ (arrow).
1207: For this particular case, 7,500 folding trajectories were simulated to
1208: provide enriched statistics for the FPT distribution in the inset,
1209: wherein $P(t)\Delta t$ is the fraction of trajectories with
1210: $t-\Delta t/2<$ FPT $\le t+\Delta t/2$, and the bin size $\Delta t$ for FPT
1211: is equal to $5\times 10^6$. The free energy of unfolding
1212: $\Delta G_{\rm u}$ for the $a=0.1$ cooperative model is computed using
1213: Monte Carlo histogram techniques based on sampling at the transition
1214: midpoint ${\cal E}/k_{\rm B}T=-1.33$. $\Delta G_{\rm u}$ is essentially
1215: linear in ${\cal E}$ (lower horizontal scale). The dotted V-shape,
1216: which fits well to the kinetic datapoints of the $a=0.1$ cooperative
1217: model over an extended regime, is an hypothetical simple two-state
1218: chevron plot consistent with the dependence of $\Delta G_{\rm u}$ on
1219: ${\cal E}$.
1220: \\
1221:
1222: \noindent
1223: {\bf Figure 4.} $\quad$
1224: Energy landscapes of three representative models with local-nonlocal
1225: coupling ($a=0.1$, $\sum\Delta S_{ij}=224$, $310$, and $386$ entries
1226: in Table~I; ${\cal E}=-1$). The left panels show the correlation between
1227: $E$ and $Q$; each dot indicates that at least one conformation with the
1228: given $(E,Q)$ was encountered in our sampling. The right panels show these
1229: structures' logarithmic densities of states, where $g(E)$ is the number of
1230: conformations with energy $E$ for the cooperative models ($a=0.1$, dots).
1231: Included for comparison are the $\ln g(E)$ values of the corresponding
1232: additive G\=o models ($a=1$, open circles; ${\cal E}=-1$). The densities of
1233: states here are estimated by Monte Carlo
1234: sampling at the models' transition midpoints ${\cal E}/k_{\rm B}T=-1.33$
1235: ($a=0.1$) and ${\cal E}/k_{\rm B}T=-1.43$ ($a=1$). Note that the cooperative
1236: models have more energy levels than the additive models. Therefore, to
1237: compare their densities of states on an equal footing, the open squares
1238: provide the natural logarithm of the number of conformations in the $a=0.1$
1239: cooperative models with energies in the range $m-0.5\le E< m+0.5$,
1240: where $m=1,0,-1,-2,\dots$ is an integer. Now the densities of states
1241: represented by the open squares ($a=0.1$) are directly comparable to
1242: that represented by the open circles ($a=1$) because their values
1243: are based upon the same unity bin size for $E$.
1244:
1245: %===========================================================================
1246:
1247: \end{document}
1248: