Structure and Energetics of the Hydrogen-Bonded Backbone in ...

bolivialodgeInternet και Εφαρμογές Web

14 Δεκ 2013 (πριν από 3 χρόνια και 7 μήνες)

156 εμφανίσεις

Structure and Energetics
of the Hydrogen-Bonded
Backbone in Protein Folding
D.Wayne Bolen
1
and George D.Rose
2
1
Department of Biochemistry and Molecular Biology and The Sealy Center
for Structural Biology,The University of Texas Medical Branch,Galveston,
Texas 77555-1052;email:dwbolen@utmb.edu
2
T.C.Jenkins Department of Biophysics,Johns Hopkins University,Baltimore,
Maryland 21218;email:grose@jhu.edu
Annu.Rev.Biochem.2008.77:339–62
The Annual Review of Biochemistry is online at
biochem.annualreviews.org
This article’s doi:
10.1146/annurev.biochem.77.061306.131357
Copyright c2008 by Annual Reviews.
All rights reserved
0066-4154/08/0707-0339$20.00
Key Words
mvalue,protein denaturation,organic osmolyte,solvent quality,
Tanford Transfer Model
Abstract
We seek to understand the link between protein thermodynamics
and protein structure in molecular detail.Aclassical approach to this
probleminvolves assessingchanges inproteinstabilityresultingfrom
added cosolvents.Under any given conditions,protein molecules
in aqueous buffer are in equilibrium between unfolded and folded
states,U(nfolded) 

N(ative).Addition of organic osmolytes,small
uncharged compounds found throughout nature,shift this equilib-
rium.Urea,a denaturing osmolyte,shifts the equilibriumtoward U;
trimethylamine N-oxide (TMAO),a protecting osmolyte,shifts the
equilibriumtoward N.Using the Tanford Transfer Model,the ther-
modynamic response tomany suchosmolytes has beendissectedinto
groupwise free energy contributions.It is found that the energetics
involving backbone hydrogen bonding controls these shifts in pro-
tein stability almost entirely,with osmolyte cosolvents simply dial-
ing between solvent-backbone versus backbone-backbone hydrogen
bonds,as a function of solvent quality.This reciprocal relationship
establishes the essential link between protein thermodynamics and
the protein’s hydrogen-bonded backbone structure.
339
AR Further
Click here for quick links to
Annual Reviews content online,
including:
• Other articles in this volume
• Top cited articles
• Top downloaded articles
• AR’s comprehensive search
Contents
INTRODUCTION.................340
Two Paradigms for Analyzing
Protein Folding...............342
TERMS ANDCONCEPTS........343
Solvent Quality and Protein
Conformation.................343
Organic Osmolytes and
Solvent Quality...............344
Transfer Free Energies:
Quantifying Solvent Quality
Effects on Proteins............345
The Transfer Model:Identifying
Driving Forces and Predicting
Protein Behavior..............345
WHICHCONTRIBUTES MORE
TOTHE mVALUE,
BACKBONE OR SIDE
CHAINS?.......................346
Dissection of Backbone and Side
Chain m-Value Contributions:
A Case Study..................346
DOES THE PEPTIDE
HYDROGENBOND
STABILIZE PROTEINS?........349
Selected Highlights from
Seven Decades of
Hydrogen-Bonding History....349
Lingering Doubts and
Alternative Explanations.......351
THE SOLVENTQUALITY
OF WATER.....................353
How Solvated Is the
Peptide Unit in Water?........354
A SIMPLE STRUCTURAL
ORIGINFOR OBSERVED
OSMOLYTE
THERMODYNAMICS..........355
INTRODUCTION
“Function follows structure” is an oft-
repeated truism in biology.A vast number of
biological components attain their structure
via spontaneous,hierarchic self-assembly,a
modular,bottom-up process rooted in the
spontaneous foldingof individual proteins (1).
Under suitable conditions,a globular protein
will experience a spontaneous drive toward its
folded native conformation fromany accessi-
ble initial conformation (2).Although not all
proteins work in exactly this way (3,4),most
do,or almost so.Without the spontaneous
disorder


order folding reaction,life as we
know it could not exist.
The driving force for this protein folding
reaction,U(nfolded)


N(ative),is exerted
by a gradient in Gibbs free energy (2),like a
ball rolling downhill under the influence of a
gravitational potential.This thermodynamic
description has characterized our thinking
about protein folding in buffered solution
for decades.Arguably,thermodynamics is our
most powerful tool for understanding bio-
logical processes,and it has been applied to
protein folding for more than 75 years (5,6).
Yet,the field still lacks physical-chemical un-
derstanding of protein folding at a molecular
level.Are we missing something essential?
Over the past half century,views about
the important forces in the thermodynam-
ics of protein folding have been continu-
ously changing targets.The driving force
in folding was initially thought to be in-
tramolecular hydrogen bonding (7),then the
hydrophobic effect (8).In recent times,it
has been argued that intramolecular hydro-
gen bonding is destabilizing (9,10),par-
tially stabilizing and destabilizing (11) and,
once again,an important driving force (12,
13).During the past two decades,the ad-
vent of protein engineering brought a hope
that the newfound ability to introduce site-
directed mutants at will would provide ready
answers to such unresolved questions (14).
However,despite thousands of mutational
experiments,disagreement about the ener-
getic role of hydrogen bonding remains,and
the likelihood of success in resolving these
questions using previous approaches is fad-
ing.This inability to resolve the energetic
role of hydrogen bonding in protein folding
has been a lingering conceptual obstacle that
340 Bolen
·
Rose
continues to impede our understanding of
protein stability.
The strategy that nature uses to regu-
late protein stabilization/destabilization in-
troduces a viewpoint that differs radically
from familiar ideas about protein folding in
buffer solution,so well explored during the
last half century.Nature’s strategy operates
predominantly on the backbone in the un-
folded state,with much less involvement of
the native state.Whenanalyzedfromthis per-
spective,the pivotal role of intramolecular hy-
drogenbondinginbothstability andstructure
formation is revealed with clarity.
Throughout the course of evolution,na-
ture has successfully modulated protein fold-
ing using organic osmolytes,small organic
molecules whose intracellular presence pro-
tects cells from environmental stress condi-
tions that would otherwise threaten survival
(15).Temperature,pressure,and denaturing
cosolvents can unfold proteins in a natural
setting,just as they do in the laboratory.For
example,thermal stresses subject intracellular
proteins to destabilization from hot or cold
denaturation,or desiccation.In such cases,a
repertoire of polyols with the ability to coun-
teract denaturation under such conditions has
been selected for intracellular accumulation.
Similarly,chemical stress is commonplace in
the urea-richcells of sharks and rays,and even
more so in the kidney cells of mammals.Urea
is a denaturing osmolyte,and in these cases,
methylamine osmolytes with the ability to
counteract solvent denaturation have been se-
lected for intracellular accumulation (16,17).
In general,naturally occurring osmolytes are
used extensively throughout all three king-
doms of life,and understanding their mode of
action can provide insight into the way nature
regulates folding and stability so successfully.
Accordingly,this review departs from the
typical retrospective of protein folding ener-
getics in buffer solution.Taking our cue in-
stead from the time-tested rigors of natural
selection,we reviewprotein folding primarily
in relation to the way that organic osmolytes
modulate the U


Nequilibrium.
Protein folding:
the spontaneous
acquisition of native
structure under
solution conditions
that favor the native
state
Organic osmolyte:
small uncharged
molecules,found
throughout nature,
which are involved in
cell volume
regulation and
modulate solvent
quality
Urea:a denaturing
osmolyte,which is
highly effective in
inducing protein
unfolding
Solvent quality:the
character of any
solvent relative to a
reference solvent,
often described on a
scale ranging from
good to poor
The overall goal of this review is to relate
the thermodynamics of folding to the organi-
zation of molecular structure in mechanistic
detail.Our inquiry into the molecular origin
of protein structure is guided by a large body
of experimental evidence,muchof it involving
the relationship between solvent quality and
protein stability.Ultimately,we suggest that
these seeminglydisparate data inthe literature
canbe rationalizedby a single mechanismthat
originates primarily in the hydrogen-bonded
backbone.
Putting the “bottom line” at the top of
the review,we argue that the half century
quest to solve the protein folding problem
has been impeded primarily by failure to
correctly assess the energy of the peptide
hydrogen bond in water.Water is a poor
solvent for unfoldedproteins,i.e.,intramolec-
ular protein interactions are favored over pro-
tein:water interactions.Were this not so,pro-
teins in water would not fold spontaneously.
Recent experimental evidence adds a specific
and surprising new twist to this self-evident
inference:Predominantly,water is a poor sol-
vent for the peptide backbone,as discussed at
length in this review.Of course,water is also
a poor solvent for apolar side chain groups.
Indeed,the ongoing focus on the hydropho-
bic effect as the driving force in protein fold-
ing has tended to obscure the crucial role of
the protein backbone,an oversight we seek to
remedy.
In particular,evidence fromsolution ther-
modynamics suggests that intramolecular hy-
drogen bonds are marginally favored over wa-
ter:backbone hydrogen bonds (18).In folded
proteins,approximately two-thirds of the
intramolecular hydrogen bonds are within
repetitive elements of secondary structure
(19).From these observations,it is plausible
that desolvation of the peptide unit in water is
more than compensated by formation of in-
tramolecular peptide hydrogen bonds in the
α-helix and β-sheet,the hydrogen-bonded
scaffold elements upon which proteins are
built.If true,this plausible mechanism
would constitute a fundamental link between
www.annualreviews.org

Hydrogen Bonding in Protein Folding 341
solvation thermodynamics and the protein’s
hydrogen-bonded backbone structure (20).
Contrary to this hypothesis,many pa-
pers in the literature conclude that,in fact,
a large desolvation penalty overwhelms the
transfer of a peptide unit from water into
a folded protein,and therefore without ad-
ditional,compensating energetic factors,the
protein would not fold.This conclusion has
prompted many complex ideas about fold-
ing;for insightful discussions,see several re-
cent reviews by Kallenbach and coworkers
(21),Baldwin (22) and Bowler (23).Here,we
review literature pertaining to a simple hy-
drogen bond-based mechanism for protein
folding.
Two Paradigms for Analyzing
Protein Folding
As defined by Anfinsen (2) and Mirsky &
Pauling before him (5),protein folding is in-
herently a thermodynamic problem.The core
of this problem is to enumerate and quantify
the factors that link protein energetics to pro-
tein structure and stability.
Since Kauzmann’s seminal review (8),the
“energy-ledger” paradigm has conditioned
our thinking about the relationship between
protein energetics and protein structure.In
the energy-ledger approach,the folding pro-
cess is partitioned into additive components
that can be quantified or estimated inde-
pendently.In a two-state folding reaction,
U(nfolded)


N(ative),the folding equilib-
riumconstant,K = [N]/[U],canbe measured
under some set of physical-chemical condi-
tions and used to derive a free energy differ-
ence between the unfolded and folded pop-
ulations,G
U→N
= −RT ln K (24).Many
contributions to G
U→N
can then be esti-
mated using analogous model compounds,as
Kauzmann showed.
However,there is a practical stumbling
block in the energy-ledger approach,as re-
vealed in Brandts’ early papers on the en-
ergetics of chymotrypsinogen stability (25,
26).G
U→N
is the small difference between
large,opposing terms.Under folding con-
ditions,entropy (TS
U→N
) favors the un-
folded state,enthalpy (H
U→N
) favors the
native state,and each contributes ∼100–
200 kcal/mol (see figure 4 in Reference 26),
resulting in G
U→N
= H
U→N
− TS
U→N
of order −10 kcal/mol.Uponevaluating these
quantities,even small errors could easily re-
sult in the wrong sign at the bottom of the
energy ledger,as Brandts pointed out.
The energyledger is a tried-and-true strat-
egy to assess a state function by partitioning it
into separable components,measuring them
individually,and then summing these values
to obtain net system behavior.Hess’s law for
the heat of combustion is a familiar example.
However,unlike Hess’s law,where the data
can be measured at high accuracy,the exper-
imental quantities in protein folding are es-
timates and averages that depend on the mi-
croenvironment,and significant inaccuracies
cannot be eliminated.Beset by such practi-
calities,a pure energy-ledger approach is ill
matched to the protein folding problem.
Of even greater concern though,the
energy-ledger approach may also be inher-
ently ill matched to the question of interest.
In this review,that question is whether the
transition between the unfolded and folded
populations can be ascribed to a predominant
molecular origin.In a two-state folding reac-
tion,U


N,a populationof foldedmolecules
still persists under destabilizing conditions
(G > 0),and therefore the information
needed to encode the fold must survive ener-
getic variationwithinthe range set by G
U→N
(27).This analysis suggests that some energy-
ledger quantities are decisive and that others
are only secondary.
This abstract conclusion is nicely illus-
trated by a concrete example:The Protein
Data Bank (28) holds ∼400 lysozyme mutants
spanning a range of energies,but all have es-
sentially identical overall conformation (29).
A straightforward interpretation of this phe-
nomenon is that the fold is encoded robustly,
i.e.,small changes infreeenergyhavelittleim-
pact on the gross native structure.In general,
342 Bolen
·
Rose
the fidelity of the fold is maintained by a ro-
bust hydrogen-bonded network that typically
survives conformational changes arising from
small,localized perturbations,as is shown
below.
Indeed,nature can compensate for a broad
diversity of destabilizing conditions simply by
shifting the folding equilibrium to favor this
hydrogen-bonded network.Nature achieves
this result by use of organic osmolytes,a nat-
ural strategy that has nowbeen deciphered,as
discussed next.
In addition to the energy ledger,another
time-tested way to assess a system is the
stimulus-response approach.Here,one intro-
duces a perturbation and then tries to intuit
whether one or a fewprincipal underlyingfac-
tors are responsible for the system’s response
to that perturbation.When most successful,
this approach facilitates construction of a pre-
dictive model of systembehavior.
The two approaches are of opposite type:
The energy ledger is from the bottom up
and reductive,whereas stimulus-response is
from the top down and inductive.Many—
perhaps most—foldingstudies are of the latter
type,involving perturbations of either physi-
cal (temperature,volume,pressure) or chem-
ical (pH,mutations,solvent quality) factors,
with results often quantified as Gs.
In particular,much of the thermodynamic
data covered in this review comes from ex-
periments involving the measurement and in-
terpretation of osmolyte-induced changes in
solvent quality,using the Tanford Transfer
Model (or,simply,the Transfer Model) (30,
31).These are stimulus-response-type experi-
ments inwhichcosolvent additions affect pro-
tein stability in an informative way.Taken to-
gether,data from these experiments prompt
an obvious folding mechanism.Before pro-
ceeding,it is necessary to define a few terms
and concepts.
TERMS ANDCONCEPTS
The “protein folding problem,” an ongoing
challenge for protein folders,is not a prob-
Tanford Transfer
Model:a
thermodynamic
cycle comparing the
difference between
the native 

unfolded state
equilibriumin
osmolyte and in
buffer
lem for nature.Natural conditions subject
cells to a wide range of physical and chem-
ical extremes,stress conditions that can de-
nature most proteins.To counteract such ef-
fects,nature utilizes a diverse assortment of
organic osmolytes,as mentioned previously.
Acentral purpose of this reviewis to distill in-
sights fromosmolyte-induced protein folding
and use themto cast light on the energetics of
classical protein folding studies.To this end,
we introduce concepts fromadaptive biology
and the theory of polymer solutions that are
needed for this inquiry
Solvent Quality and
Protein Conformation
In this review,water or dilute buffer at pH7
is defined as a reference solvent for the pro-
tein.The term solvent quality,from poly-
mer science,describes the character of some
other solvent with respect to this reference
condition,and the terms “poor” and “good”
are used to designate any change in sol-
vent character (32).In a poor solvent,protein
solubility is reduced,and the mean radius
of gyration,R
G
,of the unfolded ensemble
contracts.Conversely,in a good solvent,pro-
tein solubility is increased,and the R
G
 of
the unfolded ensemble expands.Protein in
poor solvent is said to be solvophobic,and in
a good solvent,solvophilic.The protein’s sol-
vent response indicates that in a good solvent,
protein-solvent interactions are enhanced at
the expense of protein intramolecular inter-
actions and/or solvent-solvent interactions,
whereas ina poor solvent,proteinintramolec-
ular interactions predominate.
The protein’s solvent response canbe mea-
sured without regard to the molecular origin
of the response,and we employ the solvent
quality paradigm for just this reason.How-
ever,ultimately we seektointerpret suchmea-
surements at the molecular level by decom-
posing the overall solvent response into the
respective contributions made by individual
chemical groups.
www.annualreviews.org

Hydrogen Bonding in Protein Folding 343
Trimethylamine
N-oxide (TMAO):
a protecting
osmolyte that is
highly effective in
inducing protein
folding
Organic Osmolytes
and Solvent Quality
A basic stress response in living systems is to
adjust protein stability by regulating solvent
quality.Understanding this natural process
can provide deep insights into the mechanism
of protein folding.
In greater detail,essentially all organisms
can experience various types of water stress,
i.e.,stesses such as high or low tempera-
ture,desiccation,and external osmotic pres-
sure (33).To avoid osmotic catastrophe under
such conditions,cell volume is maintained os-
motically via carefully controlled changes in
the intracellular concentrations of organic os-
molytes,small uncharged molecules that can
also modulate solvent quality (15,34–36).Os-
molytes may be classified as either denatur-
ing or protecting.In the equilibrium fold-
ing reaction,U


N,the denaturing osmolyte
urea stabilizes U relative to N;protecting os-
molytes do the opposite,stabilizing Nrelative
to U (38,39).Often,water stress is coupled
with protein-denaturing stress,e.g.,from ei-
ther elevated temperature or urea.In such
cases,natural selection has provided a reper-
toire of protecting osmolytes,which do dou-
ble duty—both regulating cell volume and
counteracting denaturation (15).A clear ex-
ample of the latter occurs in the urea-rich
cells of sharks and rays,where the protecting
osmolyte trimethylamine N-oxide (TMAO)
is used to counteract protein denaturation
(16).In the papilla of mammalian kidney,
urea and salt concentrations can reach several
molar in some species (38),and correspond-
ingly highintracellular concentrations of pro-
tecting osmolytes are deployed in such cases
(39,40).These examples underscore the vital
role that solvent plays in protein stability and
folding.
How does solvent quality modulate pro-
tein stability?It is important to realize that
the folding equation,U


N,is not an ordi-
nary chemical reaction;no covalent bonds are
made or broken.Instead,the folded fraction
is simply dialed up or down in response to
physicochemical conditions such as solvent
quality.In thermodynamic terms,protecting
osmolytes like TMAO are solvophobic;they
effect a substantial increase in the free energy
of the U state above that of the N state
relative to the protein’s free energy in buffer
(41,42).This increase shifts the U


N
folding equilibrium toward N by changing
the ground states of Nand Uon transfer from
water to osmolyte solution,as illustrated in
Figure 1 (37,43).Conversely,denaturing
osmolytes like urea are solvophilic;they
effect a substantial decrease in the free energy
of the U state below that of the N state
relative to the free energy in buffer,shifting
the folding equilibrium toward U.The
thermodynamic consequence of this natural
mechanism is to repopulate denatured or
intrinsically disordered states relative to the
native state and,importantly,without altering
any information encoded in the native fold
(34).
G
Protecting Denaturing
N
aq
U
aq
U
aq
N
aq
U
TMAO
N
TMAO
N
urea
U
urea
Figure 1
Solvent quality,free energy and protein folding.
Predominantly,organic osmolytes affect the
unfolded state (U),much more than the folded
state (N).(left) Relative to the protein in buffer,
addition of a protecting osmolyte,e.g.,
trimethylamine N-oxide (TMAO),raises the free
energy of Uabove that of N,shifting the U

N
equilibriumtoward N.(right) Conversely,addition
of the denaturing osmolyte,urea,lowers the free
energy of Ubelow that of N,shifting the U


N
equilibriumtoward U.Abbreviations:N
aq
represents the native state in aqueous buffer;U
aq
represents the unfolded state in aqueous buffer.
344 Bolen
·
Rose
Transfer Free Energies:Quantifying
Solvent Quality Effects on Proteins
A basic level of quantifying protein-solvent
interactions involves the use of transfer free
energies.The transfer free energy,G
tr
,of
any solute (protein in this case) from wa-
ter to a second solvent system will be ei-
ther favorable or unfavorable.By definition,
an unfavorable transfer free energy,G
tr
>
0,means the protein becomes solvopho-
bic on transfer to a poor solvent,whereas
a favorable transfer free energy,G
tr
<0,
means the protein becomes solvophilic on
transfer to a good solvent.The sign and
magnitude of the measured G
tr
quantifies
the protein’s response to changes in solvent
quality.
However,as inall thermodynamic descrip-
tions,G
tr
values do not address the molec-
ular details that account for these solvent
quality changes.Clues to the underlying
molecular origins can be discovered by dis-
secting the overall G
tr
values into com-
ponent contributions corresponding to indi-
vidual protein groups.Accordingly,apparent
transfer free energies from water to 1 Mos-
molyte solutions have been determined for
the amino acids,their side chains,and the
peptide backbone unit (44–48).Such data can
thenbe appliedtomolecular hypotheses using
energy-ledger approaches.
The Transfer Model:Identifying
Driving Forces and Predicting
Protein Behavior
The Transfer Model,shown in Scheme 1,
compares the extent to which the native (N)
and denatured (U) state equilibrium in 1 M
osmolyte (given by G
1M
N→U
) differs from
N(
aq
) 

U(
aq
) in buffer solution (given by
G
o
N→U
) (30,31).The free energy differ-
ence between the two reactions,G
1M
N→U

G
o
N→U
,represents the m value,an exper-
imentally obtained quantity that measures
osmolyte efficacy in folding/unfolding a pro-
tein (24).G
tr,U
and G
tr,N
represent the
G
N→U
0
U
(aq)
U
(os)
N
(aq)
N
(os)
Gtr, N
Gtr, U
G
N→U
1M
Scheme 1
The Tanford Transfer Model.A thermodynamic
cycle that compares the extent to which the native
and denatured state equilibriumin 1 Mosmolyte,
N(
os
)


U(
os
),(given by G
1M
N→U
) differs fromthe
corresponding states in aqueous buffer,
N(
aq
)

U(
aq
) (given by G
o
N→U
).The free
energy difference between the two reactions,
G
1M
N→U
−G
o
N→U
,represents the mvalue,an
experimentally obtained quantity that measures
osmolyte efficacy in folding/unfolding a protein.
The perpendicular reactions,G
tr,U
and G
tr,N
,
represent the transfer free energies of unfolded
and native protein frombuffer solution to 1 M
osmolyte.Fromthe fact that the Transfer Model is
a thermodynamic cycle,it follows that
G
1M
N→U
−G
o
N→U
= G
tr,U
−G
tr,N
,and
therefore,the free energy difference of the
perpendicular reactions,G
tr,U
−G
tr,N
,is also
equal to the mvalue.
mvalue:an
experimentally
obtained measure of
osmolyte efficacy in
folding/unfolding a
protein;also a
measure of the
folding cooperativity
GTFE:group
transfer free energy
transfer free energies of unfolded and na-
tive protein from buffer solution to 1 M os-
molyte.Scheme 1 is a thermodynamic cycle,
so G
1M
N→U
− G
o
N→U
= G
tr,U
− G
tr,N
,
and therefore the free energy difference of
the perpendicular reactions,G
tr,U
−G
tr,N
,
is also equal to the m value (45).Two goals
of the Transfer Model are to evaluate G
tr,U
andG
tr,N
individually and,fromtheir differ-
ence,to predict the experimentally obtained
mvalue of the protein.
Assuming additivity,G
tr,U
and G
tr,N
can be evaluated by summing their com-
ponents,the experimentally obtained indi-
vidual apparent group transfer free ener-
gies (GTFEs) of side chains and the peptide
www.annualreviews.org

Hydrogen Bonding in Protein Folding 345
backbone.For the N state,the accessible
surface area (ASA) of the native structure is
computed (49) fromatomic coordinates (28),
using a solvent probe of radius 1.4
˚
A.The cal-
culatedASAof eachconstituent backbone and
side chain group is normalized by its standard
state ASA,defined as the area of that group
in a Gly-X-Gly tripeptide (50).These nor-
malized values are summed by group,and the
groupwise sums are then multiplied by their
corresponding GTFEs,providing groupwise
free energy subtotals that total G
tr,N
.Fur-
ther details of the procedure are described in
Auton &Bolen (45).
Similarly for the U state,evaluation of
G
tr,U
requires additivity of component
groups and,in this case,a denatured state
model from which accessible surface areas
can be obtained.Creamer et al.(51) pro-
posed two extreme models that are intended
to bracket denatured state solvent accessi-
bilities between credible boundaries:(a) a
random coil in a good solvent,represent-
ing an expanded,solvent-exposed limit,and
(b) a compact coil in a poor solvent,repre-
senting a collapsed,solvophobic limit.Fol-
lowing Schellman (52),Creamer’s groupwise
limits were averaged,multiplied by their cor-
responding GTFEs,and summed to obtain
G
tr,U
.Reassuringly,the numerical average
of Creamer’s two extremes closely resembles
the ASA of Goldenberg’s self-avoiding ran-
domcoil model (53).
These m-value calculations constitute a
rigorous test of the Transfer Model.GTFEs
are measurements—not fitted parameters,
andthe model requires that theybe strictlyad-
ditive.These are stringent constraints,espe-
cially so considering that GTFE values differ
from one osmolyte to another and that both
osmolyte-induced folding and urea-induced
unfolding are encompassed within a single,
invariant framework.Accurate m-value pre-
diction would be impossible if either of the
assumptions,additivity or Schellman’s model,
is invalid or if the measurements are in er-
ror.At this writing,predicted m values,cov-
ering a range of nine kcal/mol M
−1
,have been
obtained for osmolyte-induced folding and
urea-induced unfolding of over 40 proteins
using seven different osmolytes,with correla-
tion coefficient of 0.97 between the predicted
and experimental m values.Accurate predic-
tion over this large range provides persua-
sive evidence for the validity and utility of the
Transfer Model.
WHICHCONTRIBUTES MORE
TOTHE mVALUE,BACKBONE
OR SIDE CHAINS?
As described above,the Transfer Model pro-
vides a practical procedure to predict the
osmolyte-induced m value of a protein,and
in doing so,it partitions the overall free en-
ergy of transfer,G
tr
,into groupwise compo-
nents.When G
tr
is partitioned in this way,
it becomes apparent that the backbone alone
controls the U

Nfolding transition,a con-
clusion that has far-reaching implications for
the mechanismof protein folding,as is shown
below.We begin by dissecting the mvalue for
the Notch ankyrin domain,a representative
protein (54).
Dissection of Backbone and Side
Chain m-Value Contributions:
A Case Study
Notch ankyrin domain m values in urea and
TMAO have been measured by Mello &
Barrick (55) and predicted using the Transfer
Model.Experimental and predicted mvalues
are shown in Table 1.
Applying the unfolded state model (de-
scribed above in The Transfer Model:
Identifying Driving Forces and Predicting
Table 1 mvalues for ankyrin in kcal/mol M
−1
Urea
Trimethylamine
N-oxide
Experimental
−2.76
6.52
Tanford Transfer
Model
−2.98
6.11
346 Bolen
·
Rose
Protein Behavior),the difference in surface
area between native and unfolded states,
ASA,is 15,084
˚
A
2
,summed over both back-
bone and side chains.Groupwise changes
in area and the resultant group transfer
free energies are depicted graphically in
Figure 2 for urea (panel a) and TMAO(panel
b).In particular,the abscissa denotes ASA
A
F
L
I
V
P
W
S
T
Y
Q
M
N
D
E
H
K
R
Backbone
1
0
1 2 3 4 5 6 7
g transfer / ASA
(cal mol
–1
M
–1
Å
–2
)
g transfer / ASA
(cal mol
–1
M
–1
Å
–2
)
0.5
0.5
0
Side chains
8 9 10
11
12
13
14 15
A
F
L
I
V
P
W
S
T
Y
Q
M
N
D
E
H
K
R
Backbone
1
1
0
1 2 3 4 5 6 7
ASA (Å
2
∙ 10
–3
)
ASA (Å
2
∙ 10
–3
)
0.5
0.5
0
Side chains
8 9 10 11 12 13 14 15
2
1.5
a
1 M urea
b
1 M TMAO
Apolar
Polar
Acidic
Basic
Apolar
Polar
Acidic
Basic
Figure 2
Dissection of the mvalue in the Notch ankyrin domain.Applying the unfolded state model (described in
The Transfer Model:Identifying Driving Forces and Predicting Protein Behavior),the difference in
surface area between native and unfolded states,ASA,is 15,084
˚
A
2
,summed for both the backbone and
side chains.Groupwise changes in area and the resultant group transfer free energies are depicted
graphically for urea (panel a) and trimethylamine N-oxide (TMAO) (panel b).The abscissa measures
ASA and the ordinate measures the component mvalue (g transfer) per unit surface area.Filled
rectangles,colored by class,show the total free energy contribution of each respective component.By
way of mental calibration,the free energy of transfer of a backbone unit fromwater to 1 Murea is −40
cal/mol fromAuton &Bolen (44),and the standard state area of the backbone,averaged over the 20
residue types,is ∼40
˚
A
2
fromLesser &Rose (50).The resulting g transfer is −1 cal/mol M
−1
/
˚
A
2
,
corresponding to the value of the backbone component indicated on the ordinate in panel a.By far,the
largest free energy contributions to the mvalue are fromthe backbone,as described in the text.
www.annualreviews.org

Hydrogen Bonding in Protein Folding 347
for the backbone and each of the 18 types
of side chains (Notch ankyrin domain lacks
Cys residues and Gly lacks a side chain);the
ordinate denotes the m value per unit sur-
face area (kcal/mol M
−1
/
˚
A
2
) contributed by
these components upon transfer from water
to osmolyte;and therefore,the filled rectan-
gles (colored by polarity class) denote the to-
tal free energy contribution of each respective
component.
Groupwise dissection reveals a remarkable
fact:The backbone free energy contribution
alone controls the direction of the U


N
folding transition in both urea-induced un-
folding and TMAO-induced folding.It is a
striking realization that net side chain con-
tributions oppose the overall direction of the
reaction,disfavoring unfolding in urea and
disfavoring folding in TMAO.Furthermore,
hydrophobic groups (showninFigure 2) con-
tribute negligibly to the m value in either
osmolyte.
Elaborating,the overall urea mvalue is the
sum of both side chain and backbone con-
tributions.The net free energy contributed
by all side chains is small and positive;col-
lectively,Notch ankyrin domain side chains
oppose denaturation,though only slightly.In
opposition,the free energy contributed by the
newly exposed backbone is large and negative,
overcomingside chaincontributions anddriv-
ing urea-induced unfolding (48,56–60).Cor-
respondingly,in the overall TMAO m value,
the net free energy contributed by side chains
is small and negative;in this case,side chains
oppose folding but,again,only slightly.Here
too,it is the free energy of the backbone,now
large and positive,that overcomes side chain
contributions and drives the equilibrium to-
ward the native state.
In the general case,the backbone/side
chain proportions of energies and areas in
globular proteins resemble those inthe Notch
ankyrin domain (45,46).As in the Notch
ankyrin domain,the backbone contributes
most of the free energy,but the side chains
contribute most of the area.Specifically,side
chains contribute about 75%of the newly ex-
posed surface area,with the backbone con-
tributing the remaining 25%(45,46).
Proportionately,free energy contributions
are the other way around.In the Notch
ankyrin domain,the peptide backbone con-
tributes −3.49 kcal/mol M
−1
to the pre-
dicted urea m value,amounting to 117% of
the net total,whereas side chains contribute
+0.51 kcal/mol M
−1
,a −17% offset.In the
corresponding case of TMAO,the peptide
backbone contributes 8.06 kcal/mol M
−1
to
the mvalue,132%of the total,and side chains
contribute −1.95 kcal/mol M
−1
,a −32%off-
set.In either case,backbone contributions
alone control the direction of the U

 N
transition.
We note in passing that data depicted in
Figure 2 challenges the widely held propo-
sition that the m value is highly correlated
with ASA,the surface area newly exposed
on urea-induced protein unfolding.Myers
et al.(61) calculated that this relation-
ship has a correlation coefficient of 0.84.
However,the ostensible correlation is de-
ceptive.In fact,the largest contributor to
ASA on unfolding (e.g.,side chains) cor-
relates negatively with the m value;it is
only the smaller ASA—but larger trans-
fer free energy—of the peptide backbone
that contributes positively to the urea m
value and accounts for urea’s ability to unfold
proteins.Correspondingly,side chain burial
correlates negatively with the m value on
TMAO-inducedfoldingof the Notchankyrin
domain.
With regard to the mechanism of protein
folding,the opposing contributions of back-
bone and side chains serve to resolve any am-
biguity about which chemical groups control
the foldingreaction.Inbothurea-inducedun-
folding and TMAO-forced folding,the non-
covalent interactions that control the U

N
equilibrium are localized to the polar back-
bone,where hydrogen bonds are the most
likely candidate involved in determining the
relative populations of [N] and [U].Accord-
ingly,we turn now to the peptide hydrogen
bond.
348 Bolen
·
Rose
DOES THE PEPTIDE
HYDROGENBOND
STABILIZE PROTEINS?
Fewtopics in biochemistry are as confused,as
confusing,or as important as the contribution
that hydrogen bonds make to protein stabil-
ity.In his introduction to a recent monograph
on this topic (62),Baldwin asserted,“...the
drive for continued rapid progress in protein
structure prediction,which requires a fuller
understanding of protein-folding energetics,
brings peptide H-bonds and peptide solvation
into central focus.”
Undoubtedly,there is a favorable distance-
and orientation-dependent interaction
between an electro-negative atom and a
hydrogen covalently bonded to another
electro-negative atom (e.g.,C=O· · ·H–N).
However,the comparison between a hy-
drogen bond to water in the unfolded
polypeptide chain and a corresponding
intramolecular hydrogen bond in the
folded protein (i.e.,–O–H· · ·O = C+H–
O· · ·H–N


C=O· · ·H–N+2H
2
O) has
been highly controversial:Is it stronger,
weaker,or equivalent?An energy-ledger
approach to answering this question requires
(a) quantitative consideration of the enthalpy
of hydrogen bond formation in water (a high-
dielectric medium) and in the protein (a
lower-dielectric medium),and accurate
assessment of (b) the desolvation penalty
paid by a peptide unit upon transfer from
the solvent-accessible aqueous environment
to the solvent-shielded protein environ-
ment,(c) the changes in configurational
entropy in both the polypeptide chain and
in water molecules,(d ) the conformational
dependence of intramolecular hydrogen
bond strength,and (e) hydrogen bond
cooperativity.All are complex and elusive
issues.
In this as in all venues,our mindset con-
ditions our thinking.Seeking perspective,
we first sketch a history of the hydrogen
bond in protein chemistry.Then,we attempt
to identify—and possibly to resolve—some
NMA:N-methyl
acetamide
sources of confusion.Often,a zeal for energy-
ledger clarity has led the field to disregard in-
ferential evidence fromstimulus-response ex-
periments,as discussed above.
Selected Highlights from
Seven Decades of
Hydrogen-Bonding History
Contemporary ideas about hydrogen bond-
ing date back to Bernal & Megaw (63) and
Huggins (64) in the 1930s,but more than
others,it was Pauling (65) who brought these
concepts to the forefront.For a scholarly ac-
count,see Jeffrey &Saenger (66).
Guided by his ideas about the forma-
tive role of hydrogen bonding,Pauling &
Corey proposedmodels for proteinsecondary
structures,the α-helix (7) and β-sheet (67),
arguing,“The energy of an N–H· · · O=C
hydrogen bond is of the order of 8 kcal
mol
−1
,and such great instability would result
from the failure to form these bonds that we
may be confident of their presence” (7).See
Eisenberg (68) for an insightful account
of Pauling’s methods and discoveries.Soon
after,Schellman (69) estimated the peptide
hydrogen bond enthalpy in water to be ap-
proximately −1.5 kcal/mol,a value obtained
by analyzing the thermodynamics of dilute
urea solutions under the plausible assump-
tion that deviations from ideality are caused
by hydrogen-bonded dimerization.This early
estimate is remarkably close to contemporary
values (see below).
Kauzmann’s influential review (8) in 1959
overturned the existing mindset.Part III of
his classic energy-ledger-motivatedanalysis of
intramolecular forces arguedpersuasively that
the hydrophobic effect drives protein folding.
Also,see Dill’s later review (9).Kauzmann’s
conclusions were soon corroborated by ex-
perimental results of Klotz & Franzen (70)
and Susi et al.(71).Using the dimerization
of N-methyl acetamide (NMA) monomers as
a model for hydrogen bonding between pep-
tide units,Klotz & Farnam (72) found that
the hydrogen-bonding reaction in water is,
www.annualreviews.org

Hydrogen Bonding in Protein Folding 349
in fact,disfavored (G
o
= +3.1 kcal/mol).
Kauzmann’s analysis and Klotz’s measure-
ments anchoredthe plausible idea that folding
is driven largely by the burial of hydropho-
bic groups in the protein interior.These ideas
were widely accepted,and they conditioned
thinking in the field for decades to come.
If the hydrogen bond in water is dis-
favored,then one would expect that short,
protein-length helices lack autonomous sta-
bility.Consistent with this expectation,helix
propagation parameters (73) near unity (i.e.,
neither helix favoring nor helix disfavoring)
were found in an experimental host-guest sys-
tem (see Reference 74 and the references
therein).A clear inference from these mea-
surements was that 100 or more residues
would be needed to stabilize a helix in water,
well beyondthe 12-residue average lengthof a
proteinhelix.Yet,contrarytothis expectation,
Brown & Klee (75),using circular dichro-
ism,demonstrated a measurable helix content
in C-peptide,the cyanogen-bromide cleavage
product of residues 1–13 of ribonuclease A.
Importantly,residues 2–12 of C-peptide are
helical inthe intact protein(76),implying that
forces that stabilize the helix in the protein re-
semble those at work in the isolated peptide.
Regrettably,the impact of this paper was slow
in coming:Of 254 total references to date,
only 7 appearedinthe first twoyears,andonly
21 within the first five years.
If the peptide hydrogen bond in water
is sufficiently favorable,then quasi-stable
helices of protein-length should be de-
tectable.Over much of the 1980s,Baldwin’s
lab analyzed the stability and determinants
of short helices in water (see Reference 77),
starting with C-peptide (78) and culminating
in alanine-based peptides (79).Among the
naturally occurring amino acids,alanine has
both the highest helix propensity (80–83) and
the least-complicated chiral side chain,mak-
ing polyalanine an attractive candidate for
experimentation.However,pure polyalanine
is insoluble in water,requiring alanine-based
peptides to be punctuated by solubilizing
polar residues and thereby complicating
the analysis of helix-stabilizing factors (84).
Eventually,Kallenbach and coworkers (85)
eliminated most of these complications by de-
signing an uninterrupted 13-residue polyala-
nine sequence withchargedresidues relegated
to flanking positions;this peptide is soluble in
water,and it is helical.Other evidence,from
binding studies involving tRNA synthetase,
also demonstrated that hydrogen bonding is
favorable by approximately −1 kcal/mol (86).
Additionally,α-helices often terminate in
stabilizing capping motifs (87,88),and these
too are hydrogen-bonded structures (89).
Summarizing,it is now known that short
polyalanyl peptides can populate helical con-
formations in water.In such a simple ho-
mopolymer,what energetic factors are avail-
able to buttress helical structure against the
disorder engendered by conformational en-
tropy (90)?The most likely explanation is
an energetically favorable peptide hydrogen
bond,although other explanations cannot be
ruled out,including vander Waals and/or side
chain interactions (91,92).These issues are
revisited in the following section.
In the 1990s,Scholtz et al.(93) and later
Lopez et al.(94) determined the enthalpy
change for the helix to coil transition of a 50-
residue alanine-based peptide in water.If this
measured enthalpy change is attributed to hy-
drogen bonding,then the peptide hydrogen
bond in water is favorable,with an enthalpy
approximating −1 kcal/mol,in good agree-
ment with values obtained by Fersht (86) and,
much earlier,by Schellman (69).Meanwhile,
in an extensive series of mutations,Pace and
coworkers (13,95) replaced dozens of polar
residues with apolar residues of similar size
and shape (e.g.,Tyr →Phe,Thr →Val) and
measured the conformational free energy dif-
ferences,G
U→N
,betweenthe proteinwith
the polar residue andits apolar counterpart af-
ter suitable correction for conformational en-
tropy,volume changes,polar groups stranded
without partners,and the like.From this se-
ries,it can be concluded that buried hydro-
gen bonds stabilize proteins substantially,by
as much as −3 kcal/mol in some cases.
350 Bolen
·
Rose
Measurements of hydrogen bond strength
over the decades prior to 2000 were ex-
tracted from systems in which the possibil-
ity of cryptic variables could not be easily
eliminated.Ideally,one would like the means
to selectively toggle a hydrogen bond in an
experimental system without perturbing any
other aspect of that system.That ideal could
only be realized in computational systems,
which come with their own set of inherent
uncertainties.However,experiments using
time-resolved fluorescence resonance energy
transfer (FRET) are now providing a less
ambiguous way to quantify intramolecular
interactions,e.g.,Reference 96.Recently,
Kiefhaber and coworkers (97) were able to
detect nonspecific,intramolecular hydrogen
bonds in an unfolded polypeptide chain us-
ing FRET.In their study,the unfolded state
was modeled by a long Gly-Ser polymer,too
flexible to adopt a folded conformation.In
this model system,intramolecular hydrogen
bonds formin water and break upon addition
of GdmCl,a goodsolvent.This workprovides
solid evidence that intramolecular hydrogen
bondformationis governedbysolvent quality,
and that,in comparison,water is a poor sol-
vent.Thesameconclusionwas recentlydrawn
for the collapsed conformation of monomeric
polyglutamine in water (98).
Lingering Doubts and
Alternative Explanations
Against this historical backdrop,it nowseems
likely that,in fact,the peptide hydrogen bond
is somewhat stabilizing in water.Yet,argu-
ments to the contrary persist,many of them
tracking back to Klotz and coworkers (70,
72,99),who used the dimerization of NMA
monomers as a model system for hydrogen
bonding.Among these arguments,a common
denominator is the use of Klotz’s thermo-
dynamic cycle (Scheme 2),which compares
the solvation of a donor (D) and acceptor (A)
in water (D
w
,A
w
) and in a nonpolar solvent
(D
n
,A
n
) with the hydrogen-bonded donor-
acceptor pair in the same medium (D
w
· · ·A
w
and D
n
· · ·A
n
).
1
G
1
= +2.4
D
w
∙∙∙A
w
D
n
∙∙∙A
n
D
w
, A
w
D
n
, A
n
2
Helix folded in protein
G2 = –6.12
4
G4 = +0.62
3III
II
IV
I
G
3
= +3.1
Scheme 2
The Klotz cycle.A thermodynamic cycle that
compares the solvation of a hydrogen bond donor
(D) and hydrogen bond acceptor (A) in water (D
w
,
A
w
) and in a nonpolar solvent (D
n
,A
n
) with the
hydrogen-bonded donor-acceptor pair in the same
medium(D
w
· · ·A
w
and D
n
· · ·A
n
,respectively).
Inaninfluential paper,Roseman(100) pro-
posed a correction to Klotz &Farnham(72),
used it to reanalyze their thermodynamic cy-
cle,and set the stage for later conclusions
by others (e.g.,9–11).Taking CCl
4
as the
nonpolar solvent and values from Klotz &
Farnham (72),the measured free energies
in Scheme 2 are as follows:G
1
= +2.4,
G
2
= −6.12 (Roseman’s corrected value),
G
3
= +3.1,and therefore G
4
= +0.62
(all units in kcal/mol).Consequently,trans-
fer of the hydrogen-bonded species between
water and nonpolar solvent—or by inference,
the interior of a protein—is essentially neu-
tral,or even slightly disfavored,as Roseman
points out.
The largest free energies in this cycle in-
volve desolvation penalties:−G
2
measures
the cost of drying up two polar groups upon
transfer to nonpolar solvent,and its magni-
tude rivals the overall free energy,G
U→N
,of
a typical protein.Although smaller in magni-
tude,G
3
measures a desolvationpenalty that
is nevertheless sufficient to strongly disfavor
the hydrogen-bonded species in water.Taken
at face value,these substantial desolvation
www.annualreviews.org

Hydrogen Bonding in Protein Folding 351
penalties indicate that peptide hydrogenbond
formation in water is substantially disfavored
(9–11).
How can the observed stability of iso-
lated polyalanyl helices in water be rec-
onciled with a desolvation penalty that
exceeds the hydrogen bond energy?Expla-
nations invoke stabilizing interactions from
hydrophobic side chains and/or stabilizing
dispersion forces within the close-packed,
hydrogen-bonded helical core.The former
explanation seems unlikely,especially for a
polyalanyl helix.Relative to the coil state,a
C
β
atomburies little,if any,surface upon he-
lix formation,and longer side chains,forced
to protrude from the bulky backbone cylin-
der,can actually gain surface (101).Further-
more,the distance between any two C
β
atoms
exceeds 5.4
˚
A at the distance of closest ap-
proach,a separation of almost two water
diameters,precluding the possibility of hy-
drophobic burial between or among them.
An N–H· · ·O = C hydrogen bond favors
an Nto Odistance that is closer than the sum
of their van der Waals radii.Consequently,el-
ements of hydrogen-bonded secondary struc-
ture are expected to be tightly packed,and de-
tailed analysis of protein structure confirms
this expectation (102).However,tight pack-
ing would serve to stabilize the folded state
only if it exceeds corresponding protein:water
packing interactions in the unfolded state.
Whether or not this is the case remains an
open question.
With the benefit of hindsight,it is clear
that conclusions about hydrogen bonding
modeled on NMA are misleading for many
reasons,some of which are now discussed.
1.Owingtowell-knownend-groupeffects
in polymers,small molecules,such as
NMA,are unsuitable models for the
solvation energy of an internal pep-
tide backbone unit in a longer polypep-
tide.Avbelj et al.(103) drive this point
home in figures 1 and 2 of their pa-
per,and they state,“...amides are
not close models for the interaction
of the peptide group with water.The
ESF(electrostatic solvationfree energy)
value for N-methylacetamide differs by
−0.4 kcal/mol from the value for an
interior alanine peptide group.” Avbelj
et al.’s conclusions are based on calcu-
lations.An equivalent point,on the ba-
sis of measurements,is made for short
peptides by Auton & Bolen in figure 3
of their paper (44),and they state “For
peptides with small chain lengths,ac-
etamide and N-acetylglycinamide,the
solubility is high,resulting in large
differences in G
o
int
between concen-
tration scales,while at longer chain
lengths,the lower solubility of the pep-
tide causes G
o
int
values to converge.”
2.Protein-water interaction energy is
conformation dependent in polypep-
tide chains (104–111) but not in rigid
molecules like NMA.In grand canoni-
cal ensemble Monte Carlo simulations
using explicit solvent,Mezei et al.(105)
computed the per residue energy differ-
encebetweenleft-handedpolyprolineII
(PII) and a β-strand to be 0.7 kcal/mol,
approximating physiological RT(where
R is the gas constant,and T is the
temperature in degrees Kelvin).Both
PII and the β-strand lack intramolec-
ular hydrogen bonds.Yet,changes in
backbone dihedral angles,from φ,ψ=
(−139

,135

) to (−78

,149

),were
foundtomake a 7 kcal/mol difference in
solvation free energy for the central 10
residues in a longer polyalanyl peptide
(105).
3.Unlike NMA dimers,hydrogen bond
energies in longer polymers can be
highly cooperative.In his review
of cooperative interactions (112),
Dannenberg points out that the length
andstrengthof hydrogenbonds inanα-
helix are a function of residue position:
As chain length increases,bond length
decreases,and bond strength increases.
Dannenberg reports that the enthalpy
of adding an alanine to the α-helix,
352 Bolen
·
Rose
H,can be as much as 4.2 kcal/mol.
His findings are fromcalculations using
density functional theory,with the helix
treated as a solid and the coil as a liquid
in the helix


coil equilibrium.In yet
another recent density functional the-
ory calculation,Baker and colleagues
(113) report a high degree of hydrogen
bond cooperativity in amyloid fiber for-
mation,with an astonishing hydrogen
bond energy of −9.1 kcal/mol.
4.NMA molecules in solution are freely
diffusible,and their dimerization would
come at a higher entropy penalty than
an intrachain N–H· · ·O=C hydrogen
bond between partners for which the
loop-closingentropyis facilitatedbyco-
valent constraints (114).
5.The positive sign of G4 in Scheme 2
is derived from the other legs in the
thermodynamic cycle and involves the
small difference between substantially
larger measured quantities.Unavoid-
able experimental error could easily re-
sult in a change of sign from positive
(i.e.,unfavorable) to negative (i.e.,fa-
vorable) (25,26).
6.The measured free energies for NMA
are based on concentration differences,
uncorrected for activity coefficients,a
nontrivial issue in many cases,e.g.,
Reference 46.
It often happens that newly introduced
models are carefully qualifiedby their origina-
tors;then,withthepassageof time,themodels
persist but the cautions fade.The use of NMA
dimerization as a model for peptide hydro-
gen bond formation in water is subject to this
caveat.Essentially,the issues and arguments
raised in this section refocus the discussion on
one central question:Is water a good solvent
or a poor solvent for backbone polar groups?
THE SOLVENT QUALITY
OF WATER
For polypeptide chains,water is neither as
good a solvent as urea nor as poor a solvent
Neutral solvent
Good solvent
Poor solvent
Water?
Figure 3
Solvent quality scale.The termsolvent quality
describes the character of a solvent with respect to
a defined reference condition (see Solvent Quality
and Protein Conformation in TERMS AND
CONCEPTS).Situated in the continuum
between extremes of good (urea) and poor
(TMAO) is a reference condition (theta solvent)
defined as the solvent quality at which the
polymer-polymer interaction energy is
compensated exactly by the influence of volume
exclusion (115).The position of water within the
extremes of this scale is the issue of interest.
as TMAO.This comparison serves to bracket
water’s relative position on a solvent quality
scale (Figure 3) but fails to resolve its abso-
lute position.
Indeed,most thermodynamic quantities
in experimental systems are inaccessible as
absolute energy-ledger values.Instead,these
quantities are typically calibrated as differ-
ences in a stimulus-response-type paradigm.
Often,one obtains a G
o
N→U
for the system
of interest and a corresponding G

N→U
for
a perturbed version of the system (where the
prime designates the perturbedsystem).Then
G = G

N→U
−G
o
N→U
measures the de-
gree to which the perturbation either stabi-
lizes or destabilizes the original system.
The mvalue,as determined by the free en-
ergy difference G
1M
N→U
−G
o
N→U
,is such a
G;it measures the degree to which added
cosolvents dial solvent quality either up or
down.Using Scheme 1,m values obtained
in this way have been dissected,showing
that osmolytes operate predominantly on the
www.annualreviews.org

Hydrogen Bonding in Protein Folding 353
protein backbone (45,46),as described above.
A broad panel of osmolyte cosolvents,rang-
inginsolvent qualityfromTMAOtourea,can
be described by a single backbone-dependent,
hydrogen-bonding mechanism (116).Water
falls within this solvent-quality continuum,
and it too should operate according to this
same mechanism.
A helix of average length in proteins is ap-
proximately 12 residues (117).At this length,
short polyalanyl-based peptides populate he-
lical conformations in water (85,118).This
observation indicates that,by definition,wa-
ter is a poor solvent relative to a neutral sol-
vent in which protein:solvent interactions are
neither favored nor disfavored.An implicit
assumption undergirds this conclusion:Es-
sentially,all backbone polar groups form hy-
drogen bonds,either to solvent or to protein
partners,just as Pauling et al.anticipated (7).
A completely unsatisfied backbone polar
group would come at a free energy cost of
a magnitude that rivals G
U→N
for a typical
protein (119).
Even low-complexity polypeptide chains,
too soluble and too flexible to promote helix
formation,nevertheless form nonspecific in-
tramolecular hydrogen bonds in water (97),
as described above.Such intrachain hydro-
gen bonding is unencumbered by any of the
energetic complexity that might be thought
to accompany helix formation.Using flu-
orescence correlation spectroscopy,Pappu
and colleagues (120) determined that poly-
glutamine,another low-complexity polymer,
forms collapsed structures in water,despite
the absence of any hydrophobic groups.
These results,and many others,provide
clear evidence that water is a comparatively
poor solvent for polypeptide chains.Why is
this so?In β-strand or P
II
conformation,the
chain has a large negative enthalpy of solva-
tion (103,105),an ostensible indication that
water is,in fact,a good solvent.Yet,un-
der folding conditions in water,the chain
also visits other conformations for which the
solvation energy is much less favorable,as
described above.Osmolytes operate on this
equilibriumdistribution between conformers
that favor self-interaction and those that favor
chain:solvent interaction,dialing it either up
or down.Tounderstandthesethermodynamic
issues at a molecular level,the interactions
among water,osmolyte,and the polypeptide
chain must be examined in detail (59,116).
We begin by reassessing the number of water
molecules that are stripped away upon chain
desolvation.
How Solvated Is the
Peptide Unit in Water?
The thermodynamics of protein solvation has
long been a topic of keen interest (52,121).
In numerous studies of peptide and protein
folding,U 

N,it is often assumed,usually
implicitly,that each peptide unit in the un-
folded state is hydrogen-bonded to three wa-
ter molecules,one to the amide hydrogen and
one to each carbonyl oxygen lone pair.This
assumption has persisted over many decades
(122) andhas anchorednumerous assessments
of the desolvation energy.Is it valid?
Conclusions about hydration of the pep-
tide unit are often based on simple models,
such as NMA (see above) or small-molecule
crystal structures (123),which can be domi-
nated by nonadditive end effects (124).Water
molecules at polypeptide termini have a larger
effective interaction volume than they do at
interior peptide groups,where covalent chain
connectivity imposes additional excluded vol-
ume restrictions.Enhanced solvation at the
chain termini is apparent in both measure-
ments (44) and calculations (103),suggesting
that a polypeptide chain will experience par-
tial dewetting of its interior peptide groups.If
so,an assumption of three hydrogen-bonded
water molecules per peptide unit would lead
to an overestimate of the desolvation penalty.
Furthermore,given that water is a some-
what poor solvent for polypeptide chains,in-
tramolecular hydrogen bonds are expected
to compete with solvent hydrogen bonds,
stripping some water from the backbone.
The calculated difference in solvation free
354 Bolen
·
Rose
energy between a residue in an isolated β-
strand versus an isolated polyproline helix
is approximately 0.7 kcal/mol (103,105) (as
described above in Lingering Doubts and
Alternative Explanations).Clearly,this en-
ergy difference is not large enough to trap a
conformer in the P
II
basin exclusively (125),
and with ambient temperature fluctuations,
the chainwould visit other alternatives readily
(106,126,127),adjusting its hydration state
accordingly.
A SIMPLE STRUCTURAL
ORIGINFOR OBSERVED
OSMOLYTE
THERMODYNAMICS
At long last,Tanford’s Transfer Model (30,
31) has been validated by successful predic-
tionof mvalues for bothprotecting osmolytes
(45) and urea (46)—a stringent test.Using
the Transfer Model,the G
tr
of protein so-
lutes betweenbuffer andosmolyte-containing
solvents can be reliably dissected into group-
wise components,g
tr
(45).Equally,the
same,self-consistent thermodynamic frame-
work holds for urea-induced denaturation,
as shown recently (46).Thus,the Transfer
Model accounts for protectinganddenaturing
effects on protein stability,both lying on the
same solvent-quality continuum,withpeptide
hydrogen bonding as the dominant variable.
This thermodynamic framework lends it-
self to a straightforward structural interpreta-
tion:Protecting osmolytes,such as TMAO,
promote intramolecular hydrogen bonding
in the unfolded protein.Several early fold-
ing models are based on explicit intramolec-
ular hydrogen bonding in unfolded states,
e.g.,References 128–130.In common among
these models,flickering elements of sec-
ondary structure initiate a native-like scaffold
within the denatured species (see References
131–133).In accord with these ideas,incipi-
ent hydrogen bonding can be observed exper-
imentally (97),and it is known that TMAO
promotes helix formation in alanine-based
peptides (134).In contrast,denaturing os-
molytes,like urea,work in the opposite way,
disrupting incipient intramolecular hydrogen
bonding and suppressing secondary structure
formation;although even in this case,evi-
dence of incipient secondary structure in the
presence of high denaturant concentrations
has been reported (135).
Extending previous models,we hypothe-
size that the molecular origin of the osmolyte
effect can be ascribed to backbone hydrogen
bonding (20,116).In brief,the addition of
a protecting osmolyte reduces solvent quality
for the backbone,decreasing the equilibrium
population of solvent:backbone hydrogen
bonds in U,but not in N,and raising the
free energy of U relative to N accordingly.
Given the high energetic cost of even one
completely unsatisfied hydrogen bond (119),
loss of hydrogen-bonding capacity in U
would shift the U


Nequilibriummarkedly
toward N,where almost all backbone polar
groups are satisfied within units of local,
hydrogen-bonded secondary structure:α-
helix,β-sheet,3
10
-helix,and β-turns (136).
To a first approximation,TMAOsimply dials
down the number of solvent:backbone hydro-
gen bonds,relative to buffer (see Figure 1).
Reciprocally,urea dials up the number of sol-
vent:backbone hydrogenbonds,againrelative
to buffer.This reciprocal relationship—the
osmolyte hypothesis for protein folding—
establishes the fundamental link between
solvent thermodynamics and the protein’s
hydrogen-bonded backbone structure.
Anfinsen’s thermodynamic hypothesis
states that “the three-dimensional structure
of a native protein in its normal physiological
milieu (solvent,pH,ionic strength,presence
of other components such as metal ions or
prosthetic groups,temperature,and other)
is the one in which the Gibbs free energy of
the whole systemis the lowest” (2).The ther-
modynamic hypothesis posits a link between
free energy and native structure,and using
the Transfer Model,the osmolyte hypothesis
describes that link in quantitative detail.
It is important to emphasize that shifting
hydrogen bonding between U and N is an
www.annualreviews.org

Hydrogen Bonding in Protein Folding 355
Table 2 Intrinsically disordered proteins
Protein name
Reference
α-Synuclein
144
T62P staphylococcal nuclease
37
Reduced carboxyamidated RNase
T1
37
Truncated notch ankyrin protein
145
Glucocorticoid receptor AF1
domain
146
RNase P protein
147
exchange reaction,i.e.,solvent hydrogen
bonds in U are exchanged for peptide hy-
drogen bonds in N.No covalent bonds are
made or broken in this reaction,individ-
ual hydrogen bonds need not be intrinsically
strengthened or weakened,and the structure
of water is not altered in any significant way
(59,116).Although cooperative strengthen-
ing of hydrogen bonds may further stabi-
lize secondary structures (112),the basic ex-
change reaction does not depend upon this.
If it did,one would not expect the percentage
of residues involved in hydrogen-bonded β-
turns to exceed that of residues in α-helices,
as it does in globular proteins (137).
The structure-promoting efficacy of back-
bone hydrogen bonding is especially conspic-
uous in the forced folding of intrinsically dis-
ordered proteins.To date,at least six proteins
have been studied (Table 2),some of particu-
lar physiological relevance (e.g.,α-synuclein,
AF1,and RNase P protein),in which the ad-
dition of a protecting osmolyte can shift the
population froma disordered ensemble to the
native one.Clearly withsuchproteins,the hy-
drophobic interaction,although in water and
under usual foldingconditions,is nevertheless
insufficient to stabilize the folded form(138).
But,upon addition of TMAO,which pro-
motes folding by increasing polar backbone
interactions,the folded population emerges
spontaneously despite modest enegetic oppo-
sition fromside chain burial.
Proteins are poised between order and dis-
order.Under foldingconditions,the stabiliza-
tionfree energy,G
U→N
,of a typical globular
protein ranges between −5 to −15 kcal/mol
(139),the energetic equivalent of one or two
water:water hydrogen bonds.Consequently,
small changes in energy can leverage substan-
tial changes in chain organization.Osmolytes
work in exactly this way.According to table 2
in Reference 44,the free energy of trans-
ferring a backbone unit from water to 1 M
urea is favored by 40 cal/mol.This seemingly
modest contribution amounts to −4 kcal/mol
when summed over 100 residues in 1 Murea,
approaching the total G
U→N
of a charac-
teristic protein.With the urea concentration
further increased to typical denaturing con-
ditions (e.g.,6 M),destabilization becomes
overwhelming.
On consideration,it is apparent that hy-
drogen bonding acts as the conformational
pivot in the U

Nequilibrium.On one side
of the balance,the decrease in conformational
entropy that accompanies the folded confor-
mationfavors the Ustate under all conditions.
On the opposing side of the balance,the se-
questering of apolar residues from water ac-
cess favors the N state under all conditions.
It is only hydrogen bonding that pivots be-
tween unfolding and folding conditions with
changes in solvent quality fromosmolyte,and
this shifting balance can be monitored experi-
mentally using hydrogen exchange and NMR
(140–142).
We conclude this review with the recog-
nition that the osmolyte effect is universal
throughout all three kingdoms of life (15).
A broad repertoire of biologically active os-
molytes has been assembled via natural se-
lection,enabling each organism to select
for variants that are best suited to its cel-
lular microenvironment and external condi-
tions (143).Since Darwin,we have come to
regardmacroscopic characteristics,suchas or-
ganelles or opposable thumbs,as the province
of evolutionary biology.But osmolyte adap-
tation shows that natural selection is also at
work on a strictly physicochemical level as
well (15).Nature has been practicing success-
ful bioengineering since the beginning of life
on Earth.
356 Bolen
·
Rose
SUMMARY POINTS
1.The relationship between the thermodynamic forces responsible for protein folding
and the spontaneous emergence of protein structure is still not well understood at the
molecular level.In this review,we focus on the peptide hydrogen bond and its role
in protein folding.Our analysis reveals that the transition between the unfolded and
folded populations,U(nfolded)


N(ative),is mediated predominantly via peptide
hydrogen bonding.
2.Organic osmolytes,foundthroughout nature,modulate proteinfolding.The denatur-
ing osmolyte urea shifts the U

Nequilibriumtoward U,and protecting osmolytes
shift the equilibriumtoward N.This universal osmolyte effect provides essential clues
to the nature of the folding reaction.
3.The Tanford Transfer Model is used to dissect stability differences between a protein
inanosmolyte solutionand inbuffer alone.Specifically,the Transfer Model quantifies
the degree to which an osmolyte either stabilizes or destabilizes the protein relative
to buffer,and it partitions the overall stability difference into energetic contributions
fromdifferent side chains and the backbone.
4.Fromthe Transfer Model,the free energy differences that control the folding reaction
are contributed overwhelmingly by the backbone,with only small contributions from
side chains and negligible contributions fromhydrophobic groups.The exclusive role
of the peptide backbone in the energetics of osmolyte-induced folding and unfolding
identifies peptide hydrogen bonding as the control point in these processes.
5.Conclusions in the literature about the energetics of peptide hydrogen bonding have
often been based on simple monomers,which turn out to be misleading models.
6.Aqueous buffer alone is a somewhat poor solvent for the protein backbone.As a result,
intramolecular peptide hydrogen bonds are marginally favored over backbone:solvent
hydrogen bonds in water.Addition of protecting osmolytes,such as TMAO,fur-
ther diminish water solvent quality,increasing the population of peptide backbone
hydrogen-bonded species.Conversely,addition of urea increases solvent quality,fa-
voring protein:solvent hydrogen bonds at the expense of intramolecular hydrogen
bonds.In either case,hydrogen-bonding is pivotal in shifting the N

Uequilibrium.
7.The repertoire of sterically accessible hydrogen-bonded scaffold elements in proteins
is limitedlargely toα-helices,β-sheets,andβ-turns,andall globular proteins are built
upon such scaffolds.Consequently,solvent quality that disfavors backbone:water hy-
drogen bonds necessarily favors the hydrogen-bonded scaffold,thereby promoting
protein structure.Shifting hydrogen bonding between Uand Nis an exchange reac-
tion in which solvent:protein hydrogen bonds in Uare exchanged for peptide hydro-
genbonds inN.This reciprocal relationship establishes the fundamental link between
protein thermodynamics and the protein’s hydrogen-bonded backbone structure.
FUTURE ISSUES
1.Where does aqueous buffer fall on the solvent quality scale?Is water a good solvent
or a poor solvent for the protein backbone?
www.annualreviews.org

Hydrogen Bonding in Protein Folding 357
2.How solvated is the peptide unit under unfolding conditions?Many studies make an
implicit assumptionthat eachpeptide unit inthe unfolded state is hydrogenbonded to
three water molecules,one to the amide hydrogen and one to each carbonyl oxygen
lone pair.Is this assumption correct?If not,how does solvation vary with peptide
chain length?
3.How organized is the polypeptide backbone under unfolding conditions of interest?
What fraction of the backbone participates in intramolecular hydrogen bonds under
unfolding conditions,and how does this fraction vary with solvent quality?
4.To what extent is hydrogen bonding a paramount driving force in protein folding?
5.What is the relationship between osmolyte-induced protein stabilization/
destabilization and the past 70 years of classical protein folding studies?
DISCLOSURE STATEMENT
The authors are not aware of any biases that might be perceived as affecting the objectivity of
this review.
ACKNOWLEDGMENTS
We thank Buzz Baldwin,Neville Kallenbach,and J
¨
org R
¨
osgen for many formative discussions;
Neville Kallenbach and J ¨org R¨osgen for reading the manuscript,and Buzz Baldwin for sending
his review prior to publication.Support from the NIHGM49760 (D.W.B.) and the Mathers
Foundation (G.D.R.) is gratefully acknowledged.
LITERATURE CITED
1.Pollard TD,Earnshaw WC.2002.Cell Biology.Philadelphia:Saunders
2.Anfinsen CB.1973.Science 181:223–30
3.Radivojac P,Iakoucheva LM,Oldfield CJ,Obradovic Z,Uversky VN,Dunker AK.2007.
Biophys.J.92:1439–56
4.Uversky VN.2002.Protein Sci.11:739–56
5.Mirsky AE,Pauling L.1936.Proc.Natl.Acad.Sci.USA 22:439–47
6.Wu H.1931.Chin.J.Physiol.V:321–44
7.Pauling L,Corey RB,Branson HR.1951.Proc.Natl.Acad.Sci.USA 37:205–10
8.Kauzmann W.1959.Adv.Protein Chem.14:1–63
9.Dill KA.1990.Biochemistry 29:7133–55
10.Honig B,Yang AS.1995.Adv.Protein Chem.46:27–58
11.Honig B.1999.J.Mol.Biol.293:283–93
12.Takano K,Scholtz JM,Sacchettini JC,Pace CN.2003.J.Biol.Chem.278:31790–95
13.Pace CN,Trevino S,Prabhakaran E,Scholtz JM.2004.Philos.Trans.R.Soc.Lond.Ser.B
359:1225–34;discuss.34–35
14.Oxender DL,Fox CF.1987.Protein Engineering.New York:Wiley
15.Hochachka PW,Somero GN.2002.Biochemical Adaptation.Oxford:Oxford Univ.Press
16.Yancey PH,Somero GN.1979.Biochem.J.183:317–23
17.Garcia-Perez A,Burg MB.1990.Hypertension 16:595–602
358 Bolen
·
Rose
18.Richardson JM,Lopez MM,Makhatadze GI.2005.Proc.Natl.Acad.Sci.USA 102:1413–
18
19.Stickle DF,Presta LG,Dill KA,Rose GD.1992.J.Mol.Biol.226:1143–59
20.Rose GD,FlemingPJ,Banavar JR,MaritanA.2006.Proc.Natl.Acad.Sci.USA103:16623–
33
21.Shi Z,Chen K,Liu Z,Kallenbach NR.2006.Chem.Rev.106:1877–97
22.Baldwin RL.2007.J.Mol.Biol.371:283–301
23.Bowler BE.2007.Mol.Biosyst.3:88–99
24.Pace CN.1986.Methods Enzymol.131:266–80
25.Brandts JF.1964.J.Am.Chem.Soc.86:4291–301
26.Brandts JF.1964.J.Am.Chem.Soc.86:4302–14
27.Lattman EE,Rose GD.1993.Proc.Natl.Acad.Sci.USA 90:439–41
28.Berman HM,Westbrook J,Feng Z,Gilliland G,Bhat TN,et al.2000.Nucleic Acids Res.
28:235–42
29.Matthews BW,Remington SJ.1974.Proc.Natl.Acad.Sci.USA 71:4178–82
30.Tanford C.1964.J.Am.Chem.Soc.86:2050–59
31.Tanford C.1970.Adv.Protein Chem.24:1–95
32.Flory PJ.1969.Statistical Mechanics of Chain Molecules.New York:Interscience.432 pp.
33.Yancey PH,Clark ME,Hand SC,Bowlus RD,Somero GN.1982.Science 217:1214–22
34.Bolen DW,Baskakov IV.2001.J.Mol.Biol.310:955–63
35.Chamberlin ME,Strange K.1989.Am.J.Physiol.257:C159–73
36.Clark ME.1985.In Transport Processes,Iono- and Osmoregulation,ed.R Gilles,MGilles-
Baillien,pp.412–23.Berlin/Heidelberg:Springer-Verlag
37.Baskakov I,Bolen DW.1998.J.Biol.Chem.273:4831–34
38.MacMillen RE,Lee AK.1967.Science 158:383–85
39.Bagnasco S,Balaban R,Fales HM,Yang YM,Burg M.1986.J.Biol.Chem.261:5872–77
40.Burg MB,Kwon ED,Peters EM.1996.Kidney Int.57(Suppl.):S100–4
41.Lin TY,Timasheff SN.1994.Biochemistry 33:12695–701
42.Wang X,Vitalis A,Wyczalkowski MA,Pappu RV.2006.Proteins 63:297–311
43.Arakawa T,Bhat R,Timasheff SN.1990.Biochemistry 29:1914–23
44.Auton M,Bolen DW.2004.Biochemistry 43:1329–42
45.Auton M,Bolen DW.2005.Proc.Natl.Acad.Sci.USA 102:15065–68
46.Auton M,Holthauzen LMF,Bolen DW.2007.Proc.Natl.Acad.Sci.USA 104:15317–22
47.Nozaki Y,Tanford C.1963.J.Biol.Chem.238:4074–81
48.Wang A,Bolen DW.1997.Biochemistry 36:9101–8
49.Lee B,Richards FM.1971.J.Mol.Biol.55:379–400
50.Lesser GJ,Rose GD.1990.Proteins 8:6–13
51.
Creamer TP,Srinivasan R,Rose GD.1997.Biochemistry 36:2832–35
52.Schellman JA.2003.Biophys.J.85:108–25
53.Goldenberg DP.2003.J.Mol.Biol.326:1615–33
54.Zweifel ME,Barrick D.2001.Biochemistry 40:14344–56
55.Mello CC,Barrick D.2003.Protein Sci.12:1522–29
56.Courtenay ES,Capp MW,Record MTJr.2001.Protein Sci.10:2485–97
57.Courtenay ES,Capp MW,Saecker RM,Record MTJr.2000.Proteins 41(Suppl.4):72–85
58.Hong J,Capp MW,Saecker RM,Record MTJr.2005.Biochemistry 44:16896–911
59.O’Brien EP,Dima RI,Brooks B,Thirumalai D.2007.J.Am.Chem.Soc.129:7346–53
60.Robinson DR,Jencks WP.1965.J.Am.Chem.Soc.87:2462–70
61.Myers JK,Pace CN,Scholtz JM.1995.Protein Sci.4:2138–48
www.annualreviews.org

Hydrogen Bonding in Protein Folding 359
62.Baldwin RL.2006.Adv.Protein Chem.72:ix–xi
63.Bernal JD,Megaw HD.1935.Proc.R.Soc.Lond.Ser.A 151:384–420
64.Huggins ML.1936.J.Phys.Chem.40:723–31
65.Pauling L.1939.The Nature of the Chemical Bond.Ithaca:Cornell Univ.Press
66.Jeffrey GA,Saenger W.1991.Hydrogen Bonding in Biological Structures.Berlin:Springer-
Verlag
67.Pauling L,Corey RB.1951.Proc.Natl.Acad.Sci.USA 37:251–56
68.Eisenberg D.2003.Proc.Natl.Acad.Sci.USA 100:11207–10
69.Schellman JA.1955.C.R.Trav.Lab.Carlsburg S´er.Chim.29:230–59
70.Klotz IM,Franzen JS.1962.J.Am.Chem.Soc.84:3461–66
71.Susi H,Timasheff SN,Ard JS.1964.J.Biol.Chem.239:3051–54
72.Klotz IM,FarnhamSB.1968.Biochemistry 7:3879–82
73.ZimmBH,Bragg JK.1959.J.Chem.Phys.31:526–35
74.Denton JB,Powers SP,Zweifel BO,Scheraga HA.1982.Biopolymers 21:51–77
75.Brown JE,Klee WA.1971.Biochemistry 10:470–76
76.Wyckoff HW,Tsernoglou D,Hanson AW,Knox JR,Lee B,Richards FM.1970.J.Biol.
Chem.245:305–28
77.Chakrabartty A,Baldwin RL.1995.Adv.Protein Chem.46:141–76
78.Bierzynski A,KimPS,Baldwin RL.1982.Proc.Natl.Acad.Sci.USA 79:2470–74
79.Marqusee S,Robbins VH,Baldwin RL.1989.Proc.Natl.Acad.Sci.USA 86:5286–90
80.Lyu PC,Liff MI,Marky LA,Kallenbach NR.1990.Science 250:669–73
81.PadmanabhanS,MarquseeS,RidgewayT,LaueTM,BaldwinRL.1990.Nature 344:268–
70
82.O’Neil KT,DeGrado WF.1990.Science 250:646–51
83.Merutka G,Lipton W,Shalongo W,Park SH,Stellwagen E.1990.Biochemistry 29:7511–
15
84.Scheraga HA,Vila JA,Ripoll DR.2002.Biophys.Chem.101–102:255–65
85.Spek EJ,Olson CA,Shi Z,Kallenbach NR.1999.J.Am.Chem.Soc.121:5571–72
86.Fersht AR.1987.Trends Biochem.Sci.12:301–4
87.Bang D,Gribenko AV,Tereshko V,Kossiakoff AA,Kent SB,Makhatadze GI.2006.Nat.
Chem.Biol.2:139–43
88.Zhou HX,Lyu P,Wemmer DE,Kallenbach NR.1994.Proteins 18:1–7
89.Aurora R,Rose GD.1998.Protein Sci.7:21–38
90.Creamer TP,Rose GD.1994.Proteins 19:85–97
91.Yang AS,Honig B.1995.J.Mol.Biol.252:351–65
92.Yang AS,Honig B.1995.J.Mol.Biol.252:366–76
93.Scholtz JM,Marqusee S,Baldwin RL,York EJ,Stewart JM,et al.1991.Proc.Natl.Acad.
Sci.USA 88:2854–58
94.
Lopez MM,Chin DH,Baldwin RL,Makhatadze GI.2002.Proc.Natl.Acad.Sci.USA
99:1298–302
95.Thurlkill RL,Grimsley GR,Scholtz JM,Pace CN.2006.J.Mol.Biol.362:594–604
96.Sinha KK,Udgaonkar JB.2007.J.Mol.Biol.370:385–405
97.Moglich A,Joder K,Kiefhaber T.2006.Proc.Natl.Acad.Sci.USA 103:12394–99
98.Vitalis A,Wang X,Pappu RV.2007.Biophys.J.93:1923–37
99.Krescheck GC,Klotz IM.1969.Biochemistry 8:8–12
100.Roseman MA.1988.J.Mol.Biol.201:621–23
101.Creamer TP,Srinivasan R,Rose GD.1995.Biochemistry 34:16245–50
102.Schell D,Tsai J,Scholtz JM,Pace CN.2006.Proteins 63:278–82
360 Bolen
·
Rose
103.Avbelj F,Luo P,Baldwin RL.2000.Proc.Natl.Acad.Sci.USA 97:10786–91
104.Shi Z,Woody RW,Kallenbach NR.2002.Adv.Protein Chem.62:163–240
105.Mezei M,Fleming PJ,Srinivasan R,Rose GD.2004.Proteins 55:502–7
106.Kentsis A,Mezei M,Gindin T,Osman R.2004.Proteins 55:493–501
107.Drozdov AN,Grossfield A,Pappu RV.2004.J.Am.Chem.Soc.126:2574–81
108.Hamburger JB,Ferreon JC,Whitten ST,Hilser VJ.2004.Biochemistry 43:9790–99
109.Jha AK,Colubri A,Zaman MH,Koide S,Sosnick TR,Freed KF.2005.Biochemistry
44:9691–702
110.Avbelj F,Baldwin RL.2004.Proc.Natl.Acad.Sci.USA 101:10967–72
111.Creamer TP,Campbell MN.2002.Adv.Protein Chem.62:263–82
112.Dannenberg JJ.2006.Adv.Protein Chem.72:227–73
113.Tsemekhman K,Goldschmidt L,Eisenberg D,Baker D.2007.Protein Sci.16:761–64
114.Go N,Scheraga HA.1970.Macromolecules 3:178–87
115.Flory PJ.1953.Principles of Polymer Chemistry.Ithaca:Cornell Univ.Press
116.Street TO,Bolen DW,Rose GD.2006.Proc.Natl.Acad.Sci.USA 103:13997–4002
117.Presta LG,Rose GD.1988.Science 240:1632–41
118.Makhatadze GI.2006.Adv.Protein Chem.72:199–226
119.Fleming PJ,Rose GD.2005.Protein Sci.14:1911–17
120.Crick SL,Jayaraman M,Frieden C,Wetzel R,Pappu RV.2006.Proc.Natl.Acad.Sci.USA
103:16764–69
121.Schellman JA.2002.Biophys.Chem.96:91–101
122.Baker EN,Hubbard RE.1984.Prog.Biophys.Mol.Biol.44:97–179
123.Allen FH,Bellard S,Brice MD,Cartwright BA,Doubleday A,et al.1979.Acta Crystallogr.
B35:2331–39
124.Avbelj F,Baldwin RL.2006.Proteins 63:283–89
125.Zaman MH,Shen MY,Berry RS,Freed KF,Sosnick TR.2003.J.Mol.Biol.331:693–711
126.Makowska J,Rodziewicz-Motowidlo S,Baginska K,Vila JA,Liwo A,et al.2006.Proc.
Natl.Acad.Sci.USA 103:1744–49
127.Zagrovic B,Lipfert J,Sorin EJ,Millett IS,van Gunsteren WF,et al.2005.Proc.Natl.
Acad.Sci.USA 102:11698–703
128.Anfinsen CB,Scheraga HA.1975.Adv.Protein Chem.29:205–300
129.Karplus M,Weaver DL.1976.Nature 260:404–6
130.KimPS,Baldwin RL.1982.Annu.Rev.Biochem.51:459–89
131.Srinivasan R,Rose GD.1999.Proc.Natl.Acad.Sci.USA 96:14258–63
132.Baldwin RL,Rose GD.1999.Trends Biochem.Sci.24:26–33
133.Myers JK,Oas TG.2002.Annu.Rev.Biochem.71:783–815
134.Celinski SA,Scholtz JM.2002.Protein Sci.11:2048–51
135.Matsuo K,Sakurada Y,Yonehara R,Kataoka M,Gekko K.2007.Biophys.J.92:4088–96
136.Panasik NJr,Fleming PJ,Rose GD.2005.Protein
Sci.14:2910–14
137.Rose GD,Gierasch LM,Smith JA.1985.Adv.Protein Chem.37:1–109
138.Athawale MV,Dordick JS,Garde S.2005.Biophys.J.89:858–66
139.Kumar MD,Bava KA,Gromiha MM,Prabakaran P,Kitajima K,et al.2006.Nucleic Acids
Res.34:D204–6
140.Udgaonkar JB,Baldwin RL.1988.Nature 335:694–99
141.Roder H,Elove GA,Englander SW.1988.Nature 335:700–4
142.Rumbley J,Hoang L,Mayne L,Englander SW.2001.Proc.Natl.Acad.Sci.USA98:105–
12
143.Yancey PH.2005.J.Exp.Biol.208:2819–30
www.annualreviews.org

Hydrogen Bonding in Protein Folding 361
144.Uversky VN,Li J,Fink AL.2001.FEBS Lett.509:31–35
145.Holthauzen LM,Bolen DW.2007.Protein Sci.16:293–98
146.Baskakov IV,Kumar R,Srinivasan G,Ji YS,Bolen DW,Thompson EB.1999.J.Biol.
Chem.274:10693–96
147.Henkels CH,Kurz JC,Fierke CA,Oas TG.2001.Biochemistry 40:2777–89
362 Bolen
·
Rose
Annual Review of
Biochemistry
Volume 77,2008
Contents
Prefatory Chapters
Discovery of GProtein Signaling
Zvi Selinger ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ 1
Moments of Discovery
Paul Berg ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ 14
Single-Molecule Theme
In singulo Biochemistry:When Less Is More
Carlos Bustamante ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ 45
Advances in Single-Molecule Fluorescence Methods
for Molecular Biology
Chirlmin Joo,Hamza Balci,Yuji Ishitsuka,Chittanon Buranachai,
and Taekjip Ha ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ 51
How RNA Unfolds and Refolds
Pan T.X.Li,Jeffrey Vieregg,and Ignacio Tinoco,Jr.♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ 77
Single-Molecule Studies of Protein Folding
Alessandro Borgia,Philip M.Williams,and Jane Clarke ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ 101
Structure and Mechanics of Membrane Proteins
Andreas Engel and Hermann E.Gaub ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ 127
Single-Molecule Studies of RNA Polymerase:Motoring Along
Kristina M.Herbert,WilliamJ.Greenleaf,and Steven M.Block ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ 149
Translation at the Single-Molecule Level
R.Andrew Marshall,Colin Echeverría Aitken,Magdalena Dorywalska,
and Joseph D.Puglisi ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ 177
Recent Advances in Optical Tweezers
Jeffrey R.Moffitt,Yann R.Chemla,Steven B.Smith,and Carlos Bustamante ♣ ♣ ♣ ♣ ♣ ♣ 205
Recent Advances in Biochemistry
Mechanismof Eukaryotic Homologous Recombination
Joseph San Filippo,Patrick Sung,and Hannah Klein ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ 229
v
Structural and Functional Relationships of the XPF/MUS81
Family of Proteins
Alberto Ciccia,Neil McDonald,and Stephen C.West ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ 259
Fat and Beyond:The Diverse Biology of PPARγ
Peter Tontonoz and Bruce M.Spiegelman ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ 289
Eukaryotic DNA Ligases:Structural and Functional Insights
TomEllenberger and Alan E.Tomkinson ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ 313
Structure and Energetics of the Hydrogen-Bonded Backbone
in Protein Folding
D.Wayne Bolen and George D.Rose ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ 339
Macromolecular Modeling with Rosetta
Rhiju Das and David Baker ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ 363
Activity-Based Protein Profiling:FromEnzyme Chemistry
to Proteomic Chemistry
Benjamin F.Cravatt,Aaron T.Wright,and John W.Kozarich ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ 383
Analyzing Protein Interaction Networks Using Structural Information
Christina Kiel,Pedro Beltrao,and Luis Serrano ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ 415
Integrating Diverse Data for Structure Determination
of Macromolecular Assemblies
Frank Alber,Friedrich Förster,Dmitry Korkin,Maya Topf,and Andrej Sali ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ 443
Fromthe Determination of Complex Reaction Mechanisms
to Systems Biology
John Ross ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ 479
Biochemistry and Physiology of Mammalian Secreted
Phospholipases A
2
G´erard Lambeau and Michael H.Gelb ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ 495
Glycosyltransferases:Structures,Functions,and Mechanisms
L.L.Lairson,B.Henrissat,G.J.Davies,and S.G.Withers ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ 521
Structural Biology of the Tumor Suppressor p53
Andreas C.Joerger and Alan R.Fersht ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ 557
Toward a Biomechanical Understanding of Whole Bacterial Cells
Dylan M.Morris and Grant J.Jensen ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ 583
How Does Synaptotagmin Trigger Neurotransmitter Release?
Edwin R.Chapman ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ 615
Protein Translocation Across the Bacterial Cytoplasmic Membrane
Arnold J.M.Driessen and Nico Nouwen ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ 643
vi Contents
Maturation of Iron-Sulfur Proteins in Eukaryotes:Mechanisms,
Connected Processes,and Diseases
Roland Lill and Ulrich Mühlenhoff ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ 669
CFTR Function and Prospects for Therapy
John R.Riordan ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ 701
Aging and Survival:The Genetics of Life Span Extension
by Dietary Restriction
WilliamMair and Andrew Dillin ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ 727
Cellular Defenses against Superoxide and Hydrogen Peroxide
James A.Imlay ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ 755
Toward a Control Theory Analysis of Aging
Michael P.Murphy and Linda Partridge ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ 777
Indexes
Cumulative Index of Contributing Authors,Volumes 73–77 ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ 799
Cumulative Index of Chapter Titles,Volumes 73–77 ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ ♣ 803
Errata
An online log of corrections to Annual Review of Biochemistry articles may be found
at http://biochem.annualreviews.org/errata.shtml
Contents vii