You are on page 1of 13

Biophysical Chemistry 101 – 102 (2002) 281–293

Recent advances in helix–coil theory

Andrew J. Doig
Department of Biomolecular Sciences, UMIST, P.O. Box 88, Manchester M60 1QD, UK

Received 18 December 2001; received in revised form 1 March 2002; accepted 1 March 2002


Peptide helices in solution form a complex mixture of all helix, all coil or, most frequently, central helices with
frayed coil ends. In order to interpret experiments on helical peptides and make theoretical predictions on helices, it
is therefore essential to use a helix–coil theory that takes account of this equilibrium. The original Zimm–Bragg and
Lifson–Roig helix–coil theories have been greatly extended in the last 10 years to include additional interactions.
These include preferences for the N-cap, N1, N2, N3 and C-cap positions, capping motifs, helix dipoles, side chain
interactions and 310-helix formation. These have been applied to determine energies for these preferences from
experimental data and to predict the helix contents of peptides. This review discusses these newly recognised structural
features of helices and how they have been included in helix–coil models.
䊚 2002 Elsevier Science B.V. All rights reserved.

Keywords: a-Helix; Random coil; Protein stability; Protein structure; Zimm–Bragg; Lifson–Roig

1. Introduction or, most frequently, central helices with frayed coil

ends. In order to interpret experiments on helical
It is a pleasure to be a part of this tribute to peptides and make theoretical predictions on heli-
John Schellman. He has made numerous funda- ces, it is therefore essential to use a helix–coil
mental contributions to biophysical chemistry. In theory that considers every possible location of a
particular, he was the first person to analyse the helix within a sequence. The purpose of this review
thermodynamics of the helix–coil transition w1x, is to cover how helix–coil theories have been
thus founding the field that is the subject of this developed in the last 10 years, principally by
review. Peptides that form helices in solution do including additional structural features of the helix,
not show a simple two-state equilibrium between such as the distinct preferences that amino acids
a fully folded and fully unfolded structure. Instead have for particular locations in the helix, or helix
they form a complex mixture of all helix, all coil dipole interactions. I do not consider in detail the
results obtained from applying the models to
E-mail address: (A.J. Doig). empirical data, such as the values of preferences

0301-4622/02/$ - see front matter 䊚 2002 Elsevier Science B.V. All rights reserved.
PII: S 0 3 0 1 - 4 6 2 2 Ž 0 2 . 0 0 1 7 0 - 9
282 A.J. Doig / Biophysical Chemistry 101 – 102 (2002) 281–293

for helix interiors, or the origins of helix energet- 2. Structure of the a-helix
ics, such as the influence of solvation on helix
preferences. These areas are well covered in Proteins are built of regular local folds of the
reviews w2–8x, as well as numerous papers cited polypeptide chain called secondary structure. The
later in the text. a-helix was first described by Pauling, Corey and
The first wave of work on helix–coil theory Branson in 1950 w16x, and their model was quickly
was in the late 1950s and early 1960s, and this supported by X-ray analysis of haemoglobin w17x.
has been reviewed in detail by Poland and Scher- Irrefutable proof of the existence of the a-helix
aga w9x. Scheraga and co-workers, in particular, came with the first protein crystal structure of
continued to use and develop these models to myoglobin, in which most secondary structure is
determine helix preferences with hostyguest poly- helical w18x. a-Helices were subsequently found
peptides. In their work, long polypeptides of in nearly all globular proteins. It is the most
hydroxybutyl-L-glutamine or hydroxypropyl-L-glu- abundant secondary structure, with ;30% of res-
tamine containing 100s of amino acids were ran- idues found in a-helices w19x.
domly substituted with low amounts of a guest A helix combines a linear translation with an
amino acid. Application of helix–coil theory gave orthogonal circular rotation. In the a-helix the
helix preferences of the guest w10x. ˚ per turn of the
linear translation is a rise of 5.4A
In 1992, Qian and Schellman w11x reviewed helix and a circular rotation is 3.6 residues per
current understanding of helix–coil theories. Since turn. Side chains spaced i,iq3, i,iq4 and i,iq7
this time there has been a great deal more interest are therefore close in space, and interactions
in the field, primarily driven by two kinds of between them can affect helix stability. Spacings
experimental result. Firstly, following the pioneer- of i,iq2, i,iq5 and i,iq6 place the side chain
ing work of Marqusee and Baldwin w12x, peptide pairs on opposite faces of the helix, avoiding any
models have become available that form stable a- interaction. The helix is primarily stabilised by
helices, despite typically having only 15–20 amino i,iq4 hydrogen bonds between backbone amide
acids. Their sequences are mostly alanine, thus groups.
giving a relatively context-free environment to The conformation of a polypeptide can be
insert interactions of interest. Secondly, examina- described by the backbone dihedral angles f and
tion of protein crystal structures has shown that c. Most f,c combinations are sterically excluded,
amino acids have distinct preferences for different leaving only the broad b region and narrower a
environments within the helix. For example, Argos region. One reason why the a-helix is so stable is
and Palau w13x, Richardson and Richardson w14x that a succession of the sterically allowed a f,c
and Presta and Rose w15x showed in the 1980s angles naturally position the backbone NH and
that amino acid preferences for the first positions CO groups towards each other for hydrogen bond
in the helix are entirely different from interior formation. It is possible that a succession of the
preferences. Helix–coil theory was therefore most stable conformation of an isolated residue in
developed to (i) provide a quantitative interpreta- a polymer with alternative functional groups could
tion of the new experimental results from short point two hydrogen-bond donors or acceptors
peptide models and (ii) include the new interac- towards each other, making secondary structure
tions missing from the older models. It is the formation unfavourable. One reason why polypep-
purpose of this review to examine this recent work. tides may have been selected as the polymer of
First, I briefly cover the problem of the helix–coil choice for building functional molecules is that
transition and the older Zimm–Bragg (ZB) and the sterically most stable conformations also give
Lifson–Roig (LR) models. A more detailed expla- strong hydrogen bonds.
nation can be found in the Poland and Scheraga The residues at the N-terminus of the a-helix
w9x or Qian and Schellman w11x reviews. Then I are called N9–N-cap–N1–N2–N3–N4, etc., where
cover new structural features of the helix and how the N-cap is the residue with non-helical f,c
they are included within helix–coil models. angles immediately preceding the N-terminus of
A.J. Doig / Biophysical Chemistry 101 – 102 (2002) 281–293 283

Fig. 1. Zimm–Bragg and Lifson–Roig codes and weights for the a-helix.

an a-helix and N1 is the first residue with helical bond, and 0 otherwise. The first hydrogen-bonded
f,c angles w14x. The C-terminal residues are unit proceeding from the N-terminus has a statis-
similarly called C4–C3–C2–C1–C-cap–C9, etc. tical weight of ss, successive hydrogen bonded
The N1, N2, N3, C1, C2 and C3 residues are units have weights of s, and non-hydrogen-bonded
unique because their amide groups participate in units have weights of 1. The s-value is a propa-
i,iq4 backbone–backbone hydrogen bonds using gation parameter and s is an initiation parameter.
either only their CO (at the N-terminus) or NH The most fundamental feature of the thermody-
(at the C-terminus) groups. The need for these namics of the helix–coil transition is that the
groups to form hydrogen bonds has powerful initiation of a new helix is much more difficult
effects on helix structure and stability w15x. than the propagation of an existing helix. This is
because three residues need to be fixed in a helical
3. Helix–coil theory geometry to form the first hydrogen bond, while
adding an additional hydrogen bond to an existing
The simplest way to analyse the helix–coil helix requires that only one residue is fixed. These
equilibrium, still occasionally observed, is the two- properties are thus captured in the ZB model by
state model, in which the equilibrium is assumed having s smaller than s. The statistical weight of
to be between 100% helix conformation and 100%
a homopolymeric helix of a N hydrogen bonds is
coil. This is incorrect and its use gives serious
ss Ny1. The cost of initiation, s, is thus paid only
errors. This is because helical peptides are gener-
once for each helix, while extending the helix
ally most often found in partly helical conforma-
simply multiplies its weight by one additional s-
tions, often with a central helix and frayed,
value for each extra hydrogen bond. The complete
disordered ends, rather than in the fully folded or
helix–coil equilibrium is handled by determining
fully unfolded states.
the statistical weight for every possible conforma-
3.1. Zimm–Bragg model tion that contains a helix, plus a reference weight
of 1 for the coil conformation. The population of
The two major types of helix–coil model are each conformation is given by the statistical weight
(i) those which count hydrogen bonds, principally of that conformation divided by the sum of the
ZB w20x, and (ii) those that consider residue statistical weights for every conformation (i.e. the
conformations, principally Lifson–Roig w21x. In partition function). Thus, the greater the statistical
the ZB theory the units being considered are weight, the more stable the conformation. Partition
peptide groups and they are classified on the basis functions are extremely powerful concepts in sta-
of whether their NH groups participate in hydrogen tistical thermodynamics, since they include all
bonds within the helix. The ZB coding is shown properties of an equilibrium. Any property of the
in Fig. 1. A unit is given a code of 1 (e.g. peptide equilibrium can be extracted from the partition
unit 5 in Fig. 1) if its NH group forms a hydrogen function by applying the appropriate mathematical
284 A.J. Doig / Biophysical Chemistry 101 – 102 (2002) 281–293

function. This is analogous to quantum mechanics, at that position. In the ZB model, the initiation
where any property can be determined from a parameter s is associated with several residues
system by applying an operator function to a and s with a peptide group, rather than a residue.
wavefunction. In this case the properties could be It is therefore easier to use the LR model when
the mean number of hydrogen bonds, the mean making substitutions. Indeed, most recent work
helix length, the probability that each residue is has been based on this model. A further difference
within a helix, etc. Statistical weights can be is that the ZB model assigns weights of zero to all
regarded as equilibrium constants for the equilib- conformations that contain a chc or chhc sequence.
rium between coil and the structure (as the refer- This excludes a very large number of conforma-
ence coil weight is defined as 1). They can tions that contain a residue with helical f,c angles
therefore be converted to free energies as but with no hydrogen bond. In LR theory, these
yRTln(weight). are all considered. The ZB and LR weights are
related by the following formula w11x: sswy(1q
3.2. Lifson–Roig model v); ssv 2 y(1qv)4.
The treatment of peptide conformations is based
In the LR model, each residue is assigned a on Flory’s isolated-pair hypothesis w23x. This states
conformation of helix (h) or coil (c), depending that while f and c for a residue are strongly
on whether it has helical f,c angles. Every con- interdependent, giving preferred areas in a Rama-
formation of a peptide of N residues can therefore chandran plot, each f,c pair is independent of the
be written as a string of N ‘c’s or ‘h’s, giving 2N f,c angles of its neighbours. Pappu et al. exam-
conformations in total. Residues are assigned sta- ined the isolated-pair hypothesis in detail by
tistical weights depending on their conformations exhaustively enumerating the conformations of
and the conformations of surrounding residues. A poly(Ala) chains w24x. Each residue was consid-
residue in an h conformation with an h on either ered to populate 14 mesostates, defined by ranges
side has a weight of w. This can be thought of as of f,c values. By considering all 14N mesostate
an equilibrium constant between the helix interior strings, all conformations were considered for up
and the coil. Coil residues are used as a reference to seven alanines. The number of allowed confor-
and have a weight of 1. In order to form an i,iq mations was found to be considerably fewer than
4 hydrogen bond in a helix, three successive the maximum, thus showing that the isolated-pair
residues need to be fixed in a helical conformation. hypothesis is invalid. The chains mostly populated
M consecutive helical residues will therefore have extended or helical conformations, as many partly
My2 hydrogen bonds. The two residues at the helical conformations are sterically disallowed.
helix termini (i.e. those in the centre of chh or Such effects are not included in helix–coil theo-
hhc conformations) are therefore assigned weights ries, thus presenting a considerable challenge for
of v. The ratio of wyv gives the approximate effect the future. Helix–coil theories assign the same
of hydrogen bonding w1.7:0.036 for Ala w22x or y weight (1) to every coil residue; steric exclusion
RTln(1.7y0.036)sy2.1 kcal moly1x. A helical means that this should vary and be lower than 1
homopolymer segment of M residues has a weight in many cases.
of v 2w My2 and a population in the equilibrium of
v 2w My2 divided by the sum of the weights of 3.3. Single sequence approximation
every conformation (i.e. the partition function). In
this way the population of every conformation is Since helix nucleation is difficult, conformations
calculated and all properties of the helix–coil with multiple helical segments are expected to be
equilibrium are evaluated. The LR model is easier rare in short peptides. In the one-, or single-helical-
to handle conceptually for heteropolymers, since sequence approximation, peptide conformations
the w and v parameters are assigned to individual containing more than one helical segment are
residues. The substitution of one amino acid at a assumed not to be populated and are excluded
certain position thus changes the w- and v-values from the partition function (i.e. assigned statistical
A.J. Doig / Biophysical Chemistry 101 – 102 (2002) 281–293 285

weights of zero). As peptide length increases, the N-capping (as yRTln n) of all the amino acids
approximation is no longer valid, since multiple w27x.
helical segments can be long enough to overcome Similarly, the C-cap is the first residue in a non-
the initiation penalty. The single-sequence approx- helical conformation (c) at the C-terminus of a
imation also breaks down when a sequence with a helix. C-cap weights (c-values) are assigned to
high preference for a helix terminus is within the central residues in hcc triplets. Application of the
middle of the chain. The error from using the model to experimental data for which the C-
single-sequence approximation therefore shows a terminal amino acid of a helical peptide was varied
wide variation with sequence and could be poten- allowed the determination of the c-values, and
tially serious if a sequence has a high preference hence free energy of C-capping (as yRTln c) of
to populate more than one helix simultaneously. all the amino acids w27x.
Conformations with two or more helices may also A problem with the original definitions of the
often include helix–helix tertiary interactions that capping weights above is that they apply to isolat-
are ignored in all helix–coil models. ed h or hh conformations that are best regarded as
part of the random coil. A helical hydrogen bond
4. Extension of the helix–coil models can only form when a minimum of three consec-
utive h residues are present. The most stabilising
The majority of recent work has been based on class of N-caps (Asp, Asn, Ser, Cys and Thr)
the LR model w21x. In the original model, weights accept hydrogen bonds to the NH groups of the
are assigned to residues in the centre of hhh triplets N3 residue. This is clearly impossible if an N3
(a weight of w for propagation) or in the centre residue does not exist, which is the case if there
of chh or hhc triplets (a weight of v for initiation). are not three or more consecutive h residues. The
Residues in all other triplets have weights of 1. Doig et al. model w25x has the flaw that sequences
LR-based models have been extended by assigning such as Asp–Ala–Gly have a high population for
weights to additional conformations. In general, the chc conformation, as Asp has a high N-cap
preference. This is incorrect, however, as the Asp
this work has been motivated by the discovery of
N-cap preference results from hydrogen bonding
additional features that affect helix stability in
to the N3 position, so should not be apparent in
protein crystal structures. Their inclusion within
chc and chhc conformations. Andersen and Tong
helix–coil theory has allowed the measurement of
w28x and Rohl and Baldwin w6x therefore changed
the effect of these features on helix stability.
the definition of the N-cap to apply only to the c
residue in a chhh quartet. The N-cap residues in
4.1. N- and C-caps chc or chhc conformations have weights of zero
in the Andersen–Tong model and 1 in the Rohl–
The first residue in a helical (h) conformation Baldwin model. These modifications give different
at the N-terminus is called N1 and the preceding N-cap and helix interior energies after fitting.
residue, the last in a non-helical (c) conformation
before the beginning of the helix, is called the N- 4.2. Capping boxes
cap. Argos and Palau w13x first showed that N-cap
preferences differ from other helical positions, The N-terminal capping box w29x includes a side
although the terminology was originated by Rich- chain–backbone hydrogen bond from N3 to the
ardson and Richardson w14x. N-capping can there- N-cap (i,iy3). For example, a Ser–X–X–Glu
fore be included in LR theory by assigning a sequence has a high preference for a helix N-
weight of n to the central residue in a cch triplet terminus, as the Ser side chain accepts a hydrogen
w25x. Application of the model to experimental bond from the Glu backbone NH, while the Ser
data for which the N-terminal amino acid of a backbone NH donates a hydrogen bond to the Glu
helical peptide was varied w26x allowed the deter- side chain. This is included in the LR model by
mination of the n-values, and hence free energy of assigning a weight of w=r to the chhh confor-
286 A.J. Doig / Biophysical Chemistry 101 – 102 (2002) 281–293

mation, where r is the weight for the Ser backbone in i,iq4 backbone–backbone hydrogen bonds. The
to the Glu side-chain bond w6x. helix N-terminus shows significantly different res-
idue frequencies for the N-cap, N1, N2, N3 and
4.3. Side chain interactions helix interior positions w13,14,33x. Penel et al. w34x
made a detailed survey of the structures adopted
As helices have 3.6 residues per turn, side chains by the amino acids at N1, N2 and N3 and identified
spaced i,iq3 or i,iq4 are close in space. Side many new structural features. They found that the
chain interactions are thus possible when four or most significant structural preferences can be
five consecutive residues are in a helix. They are explained by short-range hydrogen bonding to the
included in the LR-based model by giving a weight free N-terminal NH groups, as foreseen by Presta
of w=q to hhhh quartets and w=p to hhhhh and Rose w15x, with the strongest trends being the
quintets. The side chain interaction is between the N2 amino acids Gln, Glu, Asp, Asn, Ser, Thr and
first and last side chains in these groups; the w His preferentially forming i,i or i,iq1 hydrogen
weight is maintained to maintain the equivalence bonds to the backbone. A complete theory for the
between the number of residues with w weighting helix should therefore include distinct preferences
and the number of backbone helix hydrogen bonds. for the N1, N2 and N3 positions.
Scholtz et al. w30x used a model based on the In the original LR model, the N1 and C1
one-helical-sequence approximation of the LR residues are both assigned the same weight, v.
model to quantitatively analyse salt-bridge inter- Shalongo and Stellwagen w31x separated these as
actions in alanine-based peptides. Only a single vN and vC. Andersen and Tong w28x did the same
interaction between residues of any spacing was and derived complete scales for these parameters
considered, although this was appropriate for the from fitting experimental data, although some val-
sequences they studied. ues were tentative. The helix initiation penalty is
Shalongo and Stellwagen w31x also proposed vN=vC, and so vN- and vC-values are all small
incorporating side-chain interaction energies into (;0.04).
the LR model, using a clever recursive algorithm. We added weights for the N1, N2 and N3 (n1,
In our model we have considered only i,iq3 and n2 and n3) positions as follows. The n1-value is
i,iq4 side chain interactions w32x; Shalongo and assigned to a helical residue immediately following
Stellwagen more generally consider side-chain a coil residue. The penalty for helix initiation is
interactions of any spacing. In their implementa- now n1Øv, instead of v 2, as v remains the C1
tion of the Lifson–Roig formalism, however, they weight. An N2 helical residue is assigned a weight
change the definition of the propagating and initi- of n2Øw, instead of w. The weight w is maintained
ating (capping) weights, such that the physical in order to keep the useful definition of the number
meaning of these parameters is lost and the number of residues, with a w weighting being equal to the
of propagating residues (w) no longer correlates number of residues with an i,iq4 main-chain–
with the number of hydrogen bonds formed. Their main-chain hydrogen bond. The n2-value is an
capping parameters are associated with residues in adjustment to the weight of an N2 residue that
helical conformations, instead of coil. In proteins, takes into account the structures that can be adopt-
however, N- and C-cap residues have non-helical ed by side chains uniquely at this position. Simi-
dihedral angles w14x. This loss of physical meaning larly, an N3 residue is now assigned the weight
means that their energies for substituting residues n3Øw, instead of w. Residues in the centre of the
at capping and interior positions in peptide helices rare hch conformation have a weight of 1. Appli-
are not directly transferable to helices in proteins. cation of the model to peptides designed to probe
the N1 position (CH3CO–XAAAAQAAAAQAA-
4.4. N1, N2 and N3 preferences GY–NH2 w35x) and the N2 position (CH3CO–
AXAAAAKAAAAKAAGY–NH2 w36x) has given
The N1, N2 and N3 helix positions are unique N1 and N2 preferences for most amino acids for
because their amide NH groups do not participate these positions, and these agree well with prefer-
A.J. Doig / Biophysical Chemistry 101 – 102 (2002) 281–293 287

ences observed in protein structures. Petukhov et well defined. The helix–coil parameters are deter-
al. similarly obtained N1, N2 and N3 preferences mined by fitting a model to experimental data,
for non-polar and uncharged polar residues by rather than by rigorously defining an area of the
applying AGADIR (see below) to experimental Ramachandran map as helical and assessing the
helical peptide data, and found almost identical population within that area.
results w37,38x. In our N1N2N3 model, an isolated helical resi-
due, in the centre of a chc conformation, is
4.5. Helix dipole assigned a weight n1, rather than v. This is because
the most important structures adopted by residues
The three non-hydrogen-bonded NH groups at at N1 are hydrogen bonds between the N1 side
the N-terminus of the helix give a net positive chain and the N1 backbone NH group w34x. These
charge; similarly, the three CO groups of C1, C2 interactions can form if only a single residue is in
and C3 give a negative charge at the C-terminus an h conformation. All N1 residues commencing
w39–41x. There is therefore a general trend for any helical segment have the weight n1. The N2
negative side chains to be favoured at the N- residue in a chhc sequence is assigned the weight
terminus of the helix and positive groups at the C- n2Øv for the same reason: N2 structures are typi-
terminus. Helix dipole effects were added to the cally hydrogen bonds between the N2 side chain
LR model by Scholtz et al. w30x, although they and the N2 backbone NH group w34x. These can
used the one-sequence approximation, so that only therefore form even if only two consecutive helical
one or no dipoles in total are present. In LR residues are present.
models, helix dipole effects are subsumed within Andersen and Tong w28x gave weights of zero
other energies. For example, N-cap, N1, N2 and to chc and chhc conformations. While it seems
N3 energies include a contribution from the helix strange to state that such conformations are never
dipole interaction, so the energy of interaction of populated, they did find a better fit to experimental
charged groups at this position with the dipole data than when using the Doig et al. model w25x.
should not be counted in addition. This may be because they also assigned N-cap
weights to a minimum of three helical segments,
4.6. chc and chhc conformations in the random as discussed above. Weights for these coil confor-
coil mations thus vary greatly. As all the models fit
well to experimental data, it is unclear at present
The assignment of weights to very short helical which choice is best.
segments chc and chhc requires careful consider-
ation. While each such conformation is disfavou- 4.7. 310- and p-helices
red, as they are residues that pay the entropic cost
of restriction to helical f,c space without forming The ideal a-helix has a periodicity of 3.6 resi-
stabilising bonds, the large number of places in dues per turn, encloses 13 atoms in a ring by
which they can form in the peptide means that formation of an i,iq4 C_ O∆H–N hydrogen
they make a significant contribution to the random bond, and is thus a 3.613-helix. The 310-helix is a
coil population. The original LR model assigns more tightly wound helix, stabilised by i,iq3
these conformations weights of v and v 2, implicitly C_ O∆H–N hydrogen bonds, while the p-(4.416)-
assuming that the initiation penalty for helix for- helix is wound less tightly, with i,iq5
mation also gives the correct weights for these C_ O∆H–N hydrogen bonds. The Lifson–Roig
coil conformations. If, however, residues have formalism can easily be adapted to describe helices
more conformational freedom within a chc confor- of other co-operative lengths w42x. To treat the
mation than within the N1 or C1 positions of the 310-helix–coil transition, the helical conformation,
helix, which seems reasonable, then the weights ha, is replaced with ht, reflecting the different f,c
of chc and chhc should be higher. The root of the angles of 310-helical residues. The fundamental
problem is that the c and h conformations are not difference between a 310-helix and an a-helix is
288 A.J. Doig / Biophysical Chemistry 101 – 102 (2002) 281–293

that the 310-helix has an i,iq3 hydrogen bonding between a- and 310-helical segments, and that it
pattern, rather than the i,iq4 pattern characteristic allows a 310-helix to extend only from the C-
of the a-helix. For a given number of units in terminus of an a-helix. N-terminal 310-helical
helical conformations, a 310-helix consequently has extensions to a-helices are often observed in crys-
one more hydrogen bond than an a-helix. An tal structures, however w19,45x.
equivalent description of this difference is that In a p-helix, formation of an i,iq5 hydrogen
initiation is easier for a 310-helix than for an a- bond requires that four units be constrained to the
helix: one fewer unit needs to be fixed in a helical p-helical conformation, hp. The p subscript des-
conformation before the first hydrogen bond is ignates the conformation and weights describing
formed. To include this difference in the 310-helix the p-helix, the dihedral angles of which are
theory, one of the a-helical initiating residues (i.e. distinct from a- and 310-helices. Assigning statis-
the central unit of either the hahac or the chaha tical weights to individual units requires consider-
triplet) must become a 310-helix-propagating resi- ation of the conformations of the unit itself and its
due. We arbitrarily chose to assign the propagating three nearest neighbours w42x. The initiating statis-
statistical weight, wt, to the central unit of the tical weight, vp, is assigned to a helical unit when
hahac triplet such that the helix-propagating unit i one or more of its two N-terminal and nearest C-
is associated with the hydrogen bond formed terminal neighbours are in the coil conformation.
between the CO of peptide iy2 and the NH of The definition of helix-initiating units as the two
peptide iq1. The remainder of the statistical N-terminal and one C-terminal units of each helical
weights applicable to the a-helix–coil theory are stretch is again arbitrary. Units in a p-helical
maintained. conformation with three helical neighbours are
The models described above for the a-helix– assigned the propagating statistical weight, wp. A
coil and 310-helix–coil transitions can be combined p-helix-propagating residue, i, is thus associated
to describe an equilibrium including pure a-heli- with the hydrogen bond between the NH of residue
ces, pure 310-helices and mixed a-y310-helices iq2 and the CO of residue iy3. Some recent
w42x. In this model, three conformational states are work suggests that the p-helix may be of more
possible, 310-helical (ht), a-helical (ha) and coil than theoretical interest w46–49x.
(c). Stretches of residues in ha conformation are
treated as in the pure a-helix model, and stretches 4.8. AGADIR
of residues in ht conformation are treated as in the
pure 310-helix model. Mixed helices consist AGADIR is an LR–based helix–coil model devel-
regions of a- and regions of 310-helical structure, oped by Serrano, Munoz ˜ and co-workers. The
and transitions between the two types of helices. original model w50x included parameters for helix
We defined two additional parameters, tN and tC, propensities excluding backbone hydrogen bonds
to describe the junction from 310- to a-helix and (attributed to conformational entropy), backbone
from a- to 310-helix, respectively. The pure 310- hydrogen bond enthalpy, side chain interactions
helix and the mixed a-y310-helix models were and a term for coil weights at the end of helical
subsequently extended to include side chain inter- sequences (i.e. caps). The single-sequence approx-
actions w43x. 310-Helices have only i,iq3 side imation was used. The original partition function
chain interactions, while both i,iq3 and i,iq4 are assumed that many helical conformations did not
possible in mixed helices. exist, as all conformations in which the residue of
Sheinerman and Brooks w44x independently pro- interest is not part of a helix were excluded
duced a model for the ay310 ycoil equilibrium, w32,50x. These were corrected in a later version,
based on the ZB formalism rather than the LR AGADIRms, which considers all possible confor-
model. They similarly extended the classification mations w51x. If AGADIR and LR models are both
of conformations from aycoil to ay310 ycoil. Their applied to the same data, to determine a side chain
model differs from ours primarily in that it does interaction energy, for example, the results are
not include additional parameters for junctions similar, showing that the models are now not
A.J. Doig / Biophysical Chemistry 101 – 102 (2002) 281–293 289

significantly different w51,52x. The treatment of of all the terms that contribute to helix stability,
the helix–coil equilibrium differs in a number of notably the 400 possible i,iq4 side-chain interac-
respects from the ZB and LR models, and these tions. Since only a few of these interactions have
have been discussed in detail in by Munoz ˜ and been accurately measured, the terms used cannot
Serrano w51x. The minimal helix length in AGADIR be precise. Further determination of energetic con-
is four residues in an h conformation, rather than tributions to helix stability is therefore still needed.
three. The effect of this assumption is to exclude
all helices that contain a single hydrogen bond;
4.9. Lomize–Mosberg model
only helices with two or more hydrogen bonds are
allowed. In practice, this probably makes little
difference, as chhhc conformations are usually Lomize and Mosberg also developed a thermo-
unfavourable and hence have low populations. dynamic model for calculating the stability of
Early versions of AGADIR considered that residues helices in solution w55x. Interestingly, they extend-
following an acetyl at the N-terminus or preceding ed it to consider helices in micelles or a uniform
an amide at the C-terminus were always in a c non-polar droplet to model a protein core environ-
conformation; this was changed to allow these to ment. Helix stability in water is calculated as the
be helical w53x. sum of main-chain interactions, which is the free
The latest version of AGADIR, AGADIR1s-2 w53x, energy change for transferring Ala from coil to
includes terms for electrostatics w53x, the helix helix, the difference in energy when replacing an
dipole w53,54x, pH dependence w54x, temperature Ala with another residue, hydrogen bonding and
w54x, ionic strength w53x, N1, N2 and N3 prefer- electrostatic interactions between polar side chains
ences w37x and capping motifs, such as the capping and hydrophobic side-chain interactions. An
box, hydrophobic staple, Schellman motif and entropic nucleation penalty of two residues per
Pro-capping motif w53x. The free energy of a helix is included. Different energies are included
helical segment, DGhelical-segment, is given by for N-cap, N1–N3, C1–C3, C-cap, hydrophobic
DGhelical-segmentsDGIntqDGHbondqDGSDqDGdipole staples, Schellman motifs and polar side-chain
qDGnonHqDGelectrost, which are terms for the interactions, based on known empirical data at the
energy required to fix a residue in helical angles time (1996). Hydrophobic interactions were cal-
(with separate terms for N1, N2, N3 and N4), culated from decreases in non-polar surface area
backbone hydrogen bonding, side-chain interac- when they are brought in contact. Helix stability
tions, excluding those between charged groups, in micelles or non-polar droplets is found by
capping and helix dipole interactions, respectively. calculating the stability in water, then adding a
Electrostatic interactions are calculated with Cou- transfer energy to the non-polar environment.
lomb’s equation. Helix dipole interactions were all
electrostatic interactions between the helix dipole
4.10. Extension of the Zimm–Bragg model
or free N- and C-termini and groups in the helix.
Interactions of the helix dipole with charged
groups located outside the helical segment were Following the discovery of short peptides that
also included. pH dependence calculations consid- ´
form isolated helices in aqueous solution, Vasquez
ered a different parameter set for charged and and Scheraga extended the ZB model to include
uncharged side chains and their pKa values. The helix dipole and side chain interactions w56x. The
single-sequence approximation (see above) is used model is very general, as it can include interactions
again, unlike in AGADIRms. of any spacing within a single helix. It was applied
AGADIR is at present the only model that can to determine i,iq4 and i,iq8 interactions. Long-
give a prediction of helix content for any peptide range interactions, beyond the scope of LR models,
sequence, thus making it very useful. It can also can thus be included. Roberts w57x and Gans et al.
predict NMR chemical shifts and coupling con- w58x also refined the ZB model to include side
stants. In order to do this, it must include estimates chain interactions.
290 A.J. Doig / Biophysical Chemistry 101 – 102 (2002) 281–293

5. Applications residues w77x. Knowledge of all the energies pres-

ent in a sequence allows the prediction of the helix
The use of a helix–coil model is essential to content of a peptide. In our experience, LR models
quantitatively interpret results on helical peptides. are accurate to within a few percent. AGADIR is
They have therefore been widely used to determine the only current method that can give a prediction
the forces that affect helix stability, including for any sequence.
interiors w5,22,59–62x, N-caps w25–27x, C-caps
w27,63,64x, N1 w35x, N2 w36x, N3 w37,38x, capping 6. The future
boxes w65–67x and side chain interactions w30–
32,52,68–75x. The results compare well to those
made in proteins by site-directed mutagenesis. There are several ways in which helix–coil
In general, helix–coil parameters are extracted models can continue to be developed. Firstly,
from experimental data by measuring the helix– additional interactions can be included, such as
coil content of a peptide that contains the interac- i,iq7 side chain interactions between side chains
tion of interest, and in which the parameters for separated by two turns of the helix, or more
all other terms in the sequence are known. For structurally complex C-capping motifs. Secondly,
example, the sequence Ac–AKAVAAAKAVAAA- conformations in addition to helix and coil can be
KAKAGY–NH2 was designed to measure the included. This approach has begun by considering
side-chain interaction energy between Val and Lys a-helix, 310-helix and coil, although it would
w74x. Two Val–Lys i,iq4 side chain interactions clearly be preferable to consider the b conforma-
are present. This peptide had a helix content of tion, as this is the next most frequent. Sheet–coil
34%. The control sequence, Ac–AKVAAAAK- models are inherently more difficult to develop, as
VAAAAKAKAGY–NH2 , is identical, except the essentially any residue can form an interaction
two Val residues are moved one place towards the with any other in a sheet, unlike helices that
N-terminus, thus removing the Val–Lys interac- contain only short-range interactions. Even the
tions. It had a helix content of 25%, thus showing simplest sheet–coil models w78–81x are far more
that the Val–Lys interactions are stabilising, as complex than helix–coil models. Thirdly, the sim-
they increase the helix content by 9%. All the ple division into helix or coil conformations could
parameters required to predict the helix content of be made more sophisticated by considering rota-
the control peptide using an LR-based model w22x meric states for each residue. Fourthly, tertiary
were already measured, namely the w-, n-, c- and structure could be included. Qian made a start to
v-values of Ala, Lys, Val, Gly and Tyr. The this problem by developing a model for coiled-
predicted helix content of 23% was in excellent coils that included a parameter for the interaction
agreement with experiment, providing confidence between two helices w82x. Finally, the treatment of
that the model and parameters are correct. The the random coil needs to be improved, as the
helix content of sequence with the Val–Lys inter- isolated-pair hypothesis appears to be invalidated
actions could not be predicted, since the weight by local steric effects w24x. This may be possible
for this interaction (p-value) was unknown. It was through the use of simulations of helical peptides
found by determining the Val–Lys p-value that or denatured proteins (for example w83–93x).
gave a prediction in agreement with the experi- Ultimately, we would like to be able to calculate
mental helix content. In this case it was 1.6, giving the stability of every possible conformation of a
a Val–Lys energy of yRTln 1.6sy0.25 kcal peptide or protein, thus solving the protein folding
moly1. problem. The problem that all of these develop-
The models can predict helix contents for indi- ments run into is that the calculation of the
vidual residues, as well as the mean helix content partition function rapidly becomes very complex
(typically measured by circular dichroism), so and unwieldy. Time will tell whether advances in
have been used to interpret hydrogen exchange theory and computer power will allow us to
data w76x and C13 chemical shifts at individual approach the partition function of a small protein
A.J. Doig / Biophysical Chemistry 101 – 102 (2002) 281–293 291

with tertiary structure, or whether the method is w14x J.S. Richardson, D.C. Richardson, Amino acid prefer-
already near to its practical limit. ences for specific locations at the ends of a-helices,
Science 240 (1988) 1648–1652.
w15x L.G. Presta, G.D. Rose, Helix signals in proteins,
Acknowledgments Science 240 (1988) 1632–1641.
w16x L. Pauling, R.B. Corey, H.R. Branson, The structure of
I thank all the people who I have worked with proteins: two hydrogen-bonded helical configurations of
in this field, namely Avi Chakrabartty, Carol Rohl, the polypeptide chain, Proc. Natl. Acad. Sci. USA 37
Buzz Baldwin, Tod Klingler, Ben Stapley, Jim (1951) 205–211.
w17x M.F. Perutz, New X-ray evidence on the configuration
Andrew, Eleri Hughes, Simon Penel, Duncan
of polypeptide chains, Nature 167 (1951) 1053–1054.
Cochran and Jia Ke Sun. Ben Stapley is thanked w18x J.C. Kendrew, R.E. Dickerson, B.E. Strandberg, et al.,
for reading the manuscript. Structure of myoglobin, Nature 185 (1960) 422–427.
w19x D.J. Barlow, J.M. Thornton, Helix geometry in proteins,
References J. Mol. Biol. 201 (1988) 601–619.
w20x B.H. Zimm, J.K. Bragg, Theory of the phase transition
w1x J.A. Schellman, The stability of hydrogen-bonded pep- between helix and random coil in polypeptide chains,
tide structures in aqueous solution, C. R. Trav. Lab. J. Chem. Phys. 31 (1959) 526–535.
Carlsberg Ser. Chim. 29 (1955) 230–259. w21x S. Lifson, A. Roig, On the theory of the helix–coil
w2x J.M. Scholtz, R.L. Baldwin, The mechanism of a-helix transition in polypeptides, J. Chem. Phys. 34 (1961)
formation by peptides, Annu. Rev. Biophys. Biomol. 1963–1974.
Struct. 21 (1992) 95–118. w22x C.A. Rohl, A. Chakrabartty, R.L. Baldwin, Helix prop-
w3x R.L. Baldwin, Helix formation by peptides of defined agation and N-cap propensities of the amino acids
sequence, Biophys. Chem. 55 (1995) 127–135. measured in alanine-based peptides in 40 volume per-
w4x V. Munoz, L. Serrano, Helix design, prediction and cent trifluoroethanol, Protein Sci. 5 (1996) 2623–2637.
stability, Curr. Opin. Biotechnol. 6 (1995) 382–386. w23x P.J. Flory, Statistical Mechanics of Chain Molecules,
w5x N.R. Kallenbach, P. Lyu, H. Zhou, CD spectroscopy Wiley, New York, 1969.
and the helix–coil transition in peptides and polypep- w24x R.V. Pappu, R. Srinivasan, G.D. Rose, The Flory isolat-
tides, in: G.D. Fasman (Ed.), Circular Dichroism and ed-pair hypothesis is not valid for polypeptide chains:
the Conformational Analysis of Biomolecules, Plenum implications for protein folding, Proc. Natl. Acad. Sci.
Press, New York, 1996, pp. 201–259. USA 97 (2000) 12565–12570.
w6x C.A. Rohl, R.L. Baldwin, Deciphering rules of helix w25x A.J. Doig, A. Chakrabartty, T.M. Klingler, R.L. Baldwin,
stability in peptides, Methods Enzymol. 295 (1998) Determination of free energies of N-capping in a-helices
1–26. by modification of the Lifson–Roig helix–coil theory
w7x C.N. Pace, J.M. Scholtz, A helix propensity scale based to include N- and C-capping, Biochemistry 33 (1994)
on experimental studies of peptides and proteins, Bio- 3396–3403.
phys. J. 75 (1998) 422–427. w26x A. Chakrabartty, A.J. Doig, R.L. Baldwin, Helix capping
w8x A.J. Doig, C.D. Andrew, D.A.E. Cochran, et al., in: A. propensities in peptides parallel those in proteins, Proc.
Berry, S.E. Radford (Eds.), From Protein Folding to Natl. Acad. Sci. USA 90 (1993) 11332–11336.
New Enzymes, vol. 68, Biochemical Society, London, w27x A.J. Doig, R.L. Baldwin, N- and C-capping preferences
2001, pp. 95–110. for all 20 amino acids in a-helical peptides, Prot. Sci.
w9x D. Poland, H.A. Scheraga, Theory of Helix–Coil Tran- 4 (1995) 1325–1336.
sitions in Biopolymers, Academic Press, New York and w28x N.H. Andersen, H. Tong, Empirical parameterization of
London, 1970. a model for predicting peptide helixycoil equilibrium
w10x J. Wojcik, K.H. Altman, H.A. Scheraga, Helix–coil populations, Protein Sci. 6 (1997) 1920–1936.
stability constants for the naturally occurring amino w29x E.T. Harper, G.D. Rose, Helix stop signals in proteins
acids in water. XXIV. Half-cysteine parameters from and peptides: the capping box, Biochemistry 32 (1993)
random poly(hydroxybutylglutamine-co-S-methylthio- 7605–7609.
L-cysteine, Biopolymers 30 (1990) 121–134. w30x J.M. Scholtz, H. Qian, V.H. Robbins, R.L. Baldwin,
w11x H. Qian, J.A. Schellman, Helixycoil theories: a compar- The energetics of ion-pair and hydrogen-bonding inter-
ative study for finite length polypeptides, J. Phys. Chem. actions in a helical peptide, Biochemistry 32 (1993)
96 (1992) 3987–3994. 9668–9676.
w12x S. Marqusee, V.H. Robbins, R.L. Baldwin, Proc. Natl. w31x W. Shalongo, E. Stellwagen, Incorporation of pairwise
Acad. Sci. USA 86 (1989) 5286–5290. interactions into the Lifson–Roig model for helix pre-
w13x P. Argos, J. Palau, Amino acid distribution in protein diction, Protein Sci. 4 (1995) 1161–1166.
secondary structures, Int. J. Peptide Protein Res. 19 w32x B.J. Stapley, C.A. Rohl, A.J. Doig, Addition of side
(1982) 380–393. chain interactions to modified Lifson–Roig helix–coil
292 A.J. Doig / Biophysical Chemistry 101 – 102 (2002) 281–293

theory: application to energetics of phenylalanine– ˜

w50x V. Munoz, L. Serrano, Elucidating the folding problem
methionine interactions, Protein Sci. 4 (1995) of helical peptides using empirical parameters, Nat.
2383–2391. Struct. Biol. 1 (1994) 399–409.
w33x S. Kumar, M. Bansal, Dissecting a-helices: position- w51x V. Munoz,˜ L. Serrano, Development of the multiple
specific analysis of a-helices in globular proteins, Pro- sequence approximation within the AGADIR model of a-
teins 31 (1998) 460–476. helix formation: comparison with Zimm–Bragg and
w34x S. Penel, E. Hughes, A.J. Doig, Side-chain structures in Lifson–Roig formalisms, Biopolymers 41 (1997)
the first turn of the a-helix, J. Mol. Biol. 287 (1999) 495–509.
127–143. ´
w52x J. Fernandez-Recio, ´
A. Vasquez, C. Civera, P. Sevilla,
w35x D.A.E. Cochran, S. Penel, A.J. Doig, Contribution of J. Sancho, The tryptophanyhistidine interaction in a-
the N1 amino acid residue to the stability of the a- helices, J. Mol. Biol. 267 (1997) 184–197.
helix, Protein Sci. 10 (2001) 463–470. w53x E. Lacroix, A.R. Viguera, L. Serrano, Elucidating the
w36x D.A.E. Cochran, A.J. Doig, Effects of the N2 residue folding problem of a-helices: local motifs, long-range
on the stability of the a-helix for all 20 amino acids, electrostatics, ionic-strength dependence and prediction
Prot. Sci. 10 (2001) 1305–1311. of NMR parameters, J. Mol. Biol. 284 (1998) 173–191.
w37x M. Petukhov, V. Munoz,˜ N. Yumoto, S. Yoshikawa, L. ˜
w54x V. Munoz, L. Serrano, Elucidating the folding problem
Serrano, Position dependence of non-polar amino acid of helical peptides using empirical parameters. II. Helix
intrinsic helical propensities, J. Mol. Biol. 278 (1998) macrodipole effects and rational modification of the
279–289. helical content of natural peptides, J. Mol. Biol. 245
w38x M. Petukhov, K. Uegaki, N. Yumoto, S. Yoshikawa, L. (1995) 275–296.
Serrano, Position dependence of amino acid intrinsic w55x A.L. Lomize, H.I. Mosberg, Thermodynamic model of
helical propensities II: non-charged polar residues: Ser, secondary structure for a-helical peptides and proteins,
Thr, Asn, and Gln, Protein Sci. 8 (1999) 2144–2150. Biopolymers 42 (1997) 239–269.
w39x A. Wada, The a-helix as an electric macro-dipole, Adv. ´
w56x M. Vasquez, H.A. Scheraga, Effect of sequence-specific
Biophys. 9 (1976) 1–63. interactions on the stability of helical conformations in
w40x polypeptides, Biopolymers 27 (1988) 41–58.
W.G.J. Hol, P.T. van Duijnen, H.J.C. Berendsen, The a-
w57x C.H. Roberts, A hierarchical nesting approach to
helix dipole and the properties of proteins, Nature 273
describe the stability of a-helices with side chain
(1978) 443–446.
w41x interactions, Biopolymers 30 (1990) 335–347.
J. Aqvist, H. Luecke, F.A. Quiocho, A. Warshel, Dipoles
w58x P.J. Gans, P.C. Lyu, P.C. Manning, R.W. Woody, N.R.
located at helix termini of proteins stabilize charges,
Kallenbach, The helix–coil transition in heterogeneous
Proc. Natl. Acad. Sci. USA 88 (1991) 2026–2030.
peptides with specific side-chain interactions: theory
w42x C.A. Rohl, A.J. Doig, Models for the 310-helixycoil, p-
and comparison with circular dichroism, Biopolymers
helixycoil, and a-helixy310-helixycoil transitions in iso- 31 (1991) 1605–1614.
lated peptides, Protein Sci. 5 (1996) 1687–1696. w59x S.H. Park, W. Shalongo, E. Stellwagen, Residue helix
w43x J.K. Sun, A.J. Doig, Addition of side-chain interactions parameters obtained from dichroic analysis of peptides
to 310-helixycoil and alpha-helixy310-helixycoil theory, of defined sequence, Biochemistry 32 (1993)
Protein Sci. 7 (1998) 2374–2383. 7048–7053.
w44x F.B. Sheinerman, C.L. Brooks, 310-Helices in peptides w60x A. Chakrabartty, T. Kortemme, R.L. Baldwin, Helix
and proteins as studied by modified Zimm–Bragg the- propensities of the amino acids measured in alanine-
ory, J. Am. Chem. Soc. 117 (1995) 10098–10103. based peptides without helix-stabilizing side-chain inter-
w45x E.N. Baker, R.E. Hubbard, Hydrogen bonding in glob- actions, Protrin Sci. 3 (1994) 843–852.
ular proteins, Prog. Biophys. Mol. Biol. 44 (1984) w61x J.K. Myers, C.N. Pace, J.M. Scholtz, A direct compar-
97–179. ison of helix propensity in proteins and peptides, Proc.
w46x W.A. Shirley, C.L. Brooks, Curious structure in ‘canon- Natl. Acad. Sci. USA 94 (1997) 2833–2837.
ical’ alanine-based peptides, Proteins Struct. Funct. w62x J. Yang, E.J. Spek, Y. Gong, H. Zhou, N.R. Kallenbach,
Genet. 28 (1997) 59–71. The role of context on a-helix stabilization: host–guest
w47x K.H. Lee, D.R. Benson, K. Kuczera, Transitions from analysis in a mixed background peptide model, Protein
a to p helix observed in molecular dynamics simula- Sci. 6 (1997) 1264–1272.
tions of synthetic peptides, Biochemistry 39 (2000) w63x J. Prieto, L. Serrano, C-capping and helix stability: the
13737–13747. Pro C-capping motif, J. Mol. Biol. 274 (1997) 276–288.
w48x T.M. Weaver, The p-helix translates structure into func- w64x A.R. Viguera, L. Serrano, Experimental analysis of the
tion, Protein Sci. 9 (2000) 201–206. Schellman motif, J. Mol. Biol. 251 (1995) 150–160.
w49x D.M. Morgan, D.G. Lynn, H. Miller-Auer, S.C. Mere- ˜
w65x V. Munoz, L. Serrano, Analysis of i,iq5 and i,iq8
dith, A designed Zn2q-binding amphiphilic polypeptide: hydrophobic interactions in a helical model peptide
energetic consequences of p-helicity, Biochemistry 40 bearing the hydrophobic staple motif, Biochemistry 34
(2001) 14020–14029. (1995) 15301–15306.
A.J. Doig / Biophysical Chemistry 101 – 102 (2002) 281–293 293

w66x V. Munoz, L. Serrano, The hydrophobic-staple motif w81x J.K. Sun, A.J. Doig, A statistical mechanical model for
and a role for loop residues in a-helix stability and b-sheet formation, J. Phys. Chem. B 104 (2000)
protein folding, Nat. Struct. Biol. 2 (1995) 380–385. 1826–1836.
w67x M. Petukhov, N. Yumoto, S. Murase, R. Onmura, S. w82x H. Qian, A thermodynamic model for the helix–coil
Yoshikawa, Factors that affect the stabilization of a- transition coupled to dimerization of short coiled-coil
helices in short peptides by a capping box, Biochemistry peptides, Biophys. J. 67 (1994) 349–355.
35 (1996) 387–397. w83x W. Schneller, D.L. Weaver, Simulation of a-helix coil
w68x E. Stellwagen, S.-H. Park, W. Shalongo, A. Jain, The transitions in simplified polyvaline—equilibrium prop-
contribution of residue ion pairs to the helical stability erties and Brownian dynamics, Biopolymers 33 (1993)
of a model peptide, Biopolymers 32 (1992) 1193–1200. 1519–1535.
w69x P.C. Lyu, P.J. Gans, N.R. Kallenbach, Energetic contri- w84x S.S. Sung, Helix folding simulations with various initial
butions of solvent-exposed ion pairs to a-helix structure, conformations, Biophys. J. 66 (1994) 1796–1802.
J. Mol. Biol. 223 (1992) 343–350. w85x L. Wang, T. O’Connell, A. Tropsha, J. Hermans, Ther-
w70x B.M. Huyghues-Despointes, T.M. Klingler, R.L. Bald- modynamic parameters for the helix–coil transition of
win, Measuring the strength of side-chain hydrogen oligopeptides—molecular dynamics simulation with the
bonds in peptide helices: the GlnØAsp (i,iq4) interac- peptide growth method, Proc. Natl. Acad. Sci. USA 92
tion, Biochemistry 34 (1995) 13267–13271. (1995) 10924–10928.
w71x A.R. Viguera, L. Serrano, Side-chain interactions w86x R.V. Pappu, W.J. Schneller, D.L. Weaver, Electrostatic
between sulfur-containing amino acids and phenylala- multipole representation of a polypeptide chain: an
nine in a-helices, Biochemistry 34 (1995) 8771–8779. algorithm for simulation of polypeptide properties, J.
w72x B.J. Stapley, A.J. Doig, Hydrogen bonding interactions Comp. Chem. 17 (1996) 1033–1055.
between glutamine and asparagine in a-helical peptides, w87x D. Poland, Discrete step model of helix–coil kinetics:
J. Mol. Biol. 272 (1997) 465–473. distribution of fluctuation times, J. Chem. Phys. 105
w73x J.S. Smith, J.M. Scholtz, Energetics of polar side-chain
(1996) 1242–1269.
interactions in helical peptides: salt effects on ion pairs w88x M. Pellegrini, N. Gronbech-Jensen, S. Doniach, Simu-
and hydrogen bonds, Biochemistry 37 (1998) 33–40.
w74x C.D. Andrew, S. Penel, G.R. Jones, A.J. Doig, Stabilis- lations of the thermodynamic properties of a short
polyalanine peptide using potentials of mean force,
ing non-polarypolar side chain interactions in the a-
Physica A 239 (1997) 244–254.
helix, Proteins Struct. Funct. Genet. 45 (2001) 449–455.
w89x S.L. Kazmirski, V. Daggett, Simulations of the structural
w75x C.A. Olson, Z. Shi, N.R. Kallenbach, Polar interactions
with aromatic side chains in a-helical peptides: CHO and dynamical properties of denatured proteins: the
H-bonding and cation–p interactions, J. Am. Chem. ‘molten coil’ state of bovine pancreatic trypsin inhibitor,
Soc. 123 (2001) 6451–6452. J. Mol. Biol. 277 (1998) 487–506.
w76x C.A. Rohl, R.L. Baldwin, Comparison of NH exchange w90x M. Schaefer, C. Bartels, M. Karplus, Solution confor-
and circular dichroism as techniques for measuring the mations and thermodynamics of structured peptides:
parameters of the helix–coil transition in peptides, molecular dynamics simulation with an implicit solva-
Biochemistry 36 (1997) 8435–8442. tion model, J. Mol. Biol. 284 (1998) 835–848.
w77x W. Shalongo, L. Dugad, E. Stellwagen, Analysis of the w91x T. Takano, T. Yamato, J. Higo, A. Suyama, K. Nagay-
thermal transitions of a model helical peptides using C- ama, Molecular dynamics of a 15-residue poly(L-ala-
13 NMR, J. Am. Chem. Soc. 116 (1994) 2500–2507. nine) in water: helix formation and energetics, J. Am.
w78x W.L. Mattice, H.A. Scheraga, Matrix formulation of the Chem. Soc. 121 (1999) 605–612.
transition from a statistical coil to an intramolecular w92x A. Mitsuake, Y. Okamoto, Helix–coil transitions of
antiparallel b-sheet, Biopolymers 23 (1984) 1701–1724. amino-acid homo-oligomers in aqueous solution studied
w79x S.-J. Chen, K.A. Dill, Statistical thermodynamics of by multicanonical simulations, J. Chem. Phys. 112
double-stranded polymer molecules, J. Chem. Phys. 103 (2000) 10638–10647.
(1995) 5802–5813. w93x A.V. Smith, C.K. Hall, a-Helix formation: discontinuous
w80x S.-J. Chen, K.A. Dill, Theory for the conformational molecular dynamics on an intermediate-resolution pro-
changes of double-stranded chain molecules, J. Chem. tein model, Proteins Struct. Funct. Genet. 44 (2001)
Phys. 109 (1998) 4602–4616. 344–360.