Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

PERSPECTIVE

pubs.acs.org/JPCL

Protein Heat Capacity: An Anomaly that Maybe Never Was


Alan Cooper*
School of Chemistry, College of Science and Engineering, Joseph Black Building, University of Glasgow, Glasgow G12 8QQ, U.K.

ABSTRACT Protein unfolding in aqueous solution is usually accompanied by an


increase in heat capacity (ΔCp), and this has long been regarded as somewhat
anomalous. However, neither the absolute heat capacities (Cp) of folded globular
proteins nor the heat capacity increments upon unfolding (ΔCp) are unusual in
comparison to values observed for order-disorder (melting) transitions in other
organic substances. The consequences for protein stability, including cold dena-
turation, enthalpy-entropy compensation, and the temperature of maximum
stability, may seem counterintuitive but should not be unexpected. Nevertheless,
while perhaps not so anomalous as once thought, quantitative interpretation of
protein heat capacity and related effects remains a theoretical challenge.

A
round 1757, Joseph Black, a colleague from an earlier Heat Capacity ; The Basics. As the name implies, heat
era in Glasgow, was puzzled by his observations that capacity is a measure of the capacity of any object to take up
certain materials took longer to warm up than others. heat energy. At the molecular level, this heat energy will be
Ice, water, and various “spiritous liquors” were particularly distributed among the available degrees of freedom and
intriguing, and it is said that “He waited with impatience for partitioned into kinetic energies (related to vibrational, rota-
the winter” so that he could continue his investigations on tional, or translational motions) or potential energies (related
heating and cooling.1 (There was no ice house in Glasgow, and to changes in interatomic potential energies, bond stretching,
refrigeration had not yet been invented, though his professor bending, breaking, and so forth). It follows that the more ways
and mentor, William Cullen, was working on it.) This was no there are of distributing heat energy in a substance, the higher
idle pursuit; James Watt, Black's friend and “mathematical will be its heat capacity. This roughly explains why liquid water
instrument maker to the University”, at that time was con- has such a relatively high heat capacity; heat energy is diverted
cerned with improving the efficiency of steam engines. What into breaking residual intermolecular hydrogen bonds rather
Black showed was that substances undergoing heat-induced than raising the temperature.
phase transitions ; melting, boiling, and evaporation ; did Heat capacity is central to fundamental thermodynamics
so without an increase in temperature. This led to the concept because both the absolute enthalpy (H) and the entropy (S) of
of “latent” (i.e., hidden) heat, what we now call heat capacity, any object are integral functions of its heat capacity
Z
in which heat energy is somehow used to do something other
than simply raise temperature. HðT Þ ¼ Hð0Þ þ C p dT ð1Þ

Z
SðT Þ ¼ Sð0Þ þ ðC p =T Þ dT ð2Þ
The consequences for protein
where Cp is the heat capacity at constant pressure, T is the
stability may seem counterintuitive absolute temperature, and the integral is taken from absolute
but should not be unexpected. zero (0 K) to the required temperature. H(0) and S(0) are the
zero-point enthalpy and entropy, respectively. (Equivalent
expressions exist for heat capacity at constant volume, Cv.
Here, however, we shall normally assume constant pressure
Now, 250 years later, we are still intrigued by this, and the
because that is what mostly applies. Except for gases, the
properties of water continue to present fascinating challenges.
numerical difference between Cp and Cv is negligible for most
Much of this has been reflected in the discussions surrounding
practical purposes.)
the heat capacity of proteins and the heat capacity changes For enthalpy (H=U þ PV, with the internal energy, U, cor-
that accompany protein transitions and interactions. In what rected for pressure-volume effects), the physical interpretation
follows, I shall try to present a (mostly personal) view of the
current situation in relation to heat capacity effects in protein
unfolding and related processes in solution, mainly from a Received Date: August 27, 2010
physical chemistry perspective. More comprehensive and Accepted Date: November 1, 2010
authoritative reviews can be found elsewhere.2,3 Published on Web Date: November 04, 2010

r 2010 American Chemical Society 3298 DOI: 10.1021/jz1012142 |J. Phys. Chem. Lett. 2010, 1, 3298–3304
PERSPECTIVE
pubs.acs.org/JPCL

Figure 1. Typical DSC data for the unfolding of a small globular


protein (lysozyme) in solution at various pH values.8 The increase
in area under each endotherm with higher Tm and the higher heat
capacity baselines after the unfolding transitions are both indica-
tions of the significant positive ΔCp commonly associated with
such processes. (Adapted from ref 8, with permission; protein
structure drawn from pdb 1HEW.9)

of eq 1 is straightforward; starting from absolute zero,


each increment in temperature, dT, requires addition of
heat energy, dQ=Cp dT, with summation up to the required
temperature to give the total enthalpy. The entropy relation-
ship, eq 2, is perhaps less intuitive but derives from the Sec- Figure 2. Examples of commonly observed heat capacity effects
ond Law definition of the entropy increment, dS = dQ/T = associated with protein unfolding in solution. (A) DSC data for
Cp dT/T. In molecular thermodynamics/statistical mechanics thermal denaturation of yeast phosphoglycerate kinase, illustrat-
ing exothermic baseline distortion and heat capacity decrease
terms, this relates to the number of ways in which heat caused by irreversible precipitation of unfolded protein. (B) Re-
energy may be distributed within the object. peat DSC scans of thermal unfolding of lysozyme, showing gradual
Heat capacity also plays a fundamental role in determining accumulation of misfolded forms, thought to arise from cis-trans
isomerization of proline peptide bonds at high temperature.10
the size of thermodynamic fluctuations in any system. All Note how the apparent heat capacities for the less-well-folded
objects are subject to thermal fluctuations arising from forms (below Tm) are higher than those in the original native state.
Brownian-motion-like bombardment from surrounding atoms
and molecules. For any object of mass m and specific heat cv transition enthalpy and entropy changes at 0 K (or any other
(heat capacity per unit mass), the mean-square internal convenient reference temperature).
energy fluctuations at temperature T are given by Equation 4 is often used in its differential form as a con-
ÆδU 2 æ ¼ kB T 2 mcv ð3Þ venient way to estimate ΔCp from the temperature depen-
dence of observed heats of reaction
where kB is the Boltzmann constant. For most everyday ΔC p ¼ DΔH=DT ð6Þ
objects, these fluctuations are so small (in proportion to the
total energy) as to go unnoticed most of the time. However, for Protein Heat Capacity ; The Experimental Situation. The
much smaller objects, particularly microscopic or mesoscopic measurement of heat capacities and heat capacity changes
systems of the size of protein molecules, these fluctuations are for proteins in dilute (<1 mg mL-1) aqueous solutions is now
relatively large.4,5 This is why protein structures must be con- relatively straightforward using commercially available differ-
sidered as dynamic objects at the molecular level. ential scanning calorimeters (DSC) specifically designed for
For any thermodynamic transition (A f B) in which the the purpose, most usually in the 0-110 C temperature
different states (A, B) have different heat capacities, it follows range.6,7 “Typical” DSC data for the thermal unfolding of a
from eqs 1 and 2 that the enthalpy and entropy changes for simple globular protein in solution are shown in Figure 1. This
the transition will also be temperature-dependent illustrates various features that we shall consider in a moment,
Z
but first, a note of caution. The sort of behavior shown in
ΔHðA f BÞ ¼ H B -H A ¼ ΔHð0Þ þ ΔC p dT ð4Þ Figure 1 is usually only seen for relatively small proteins
Z undergoing reversible unfolding, typically at low pH and low
ΔSðA f BÞ ¼ S B - S A ¼ ΔSð0Þ þ ðΔC p =T Þ dT ð5Þ concentration. Unfortunately, most proteins do not behave
that way, especially near neutral pH. Unfolded protein is sticky
where ΔCp =Cp,B - Cp,A is the change in heat capacity of the stuff, and protein denaturation is frequently accompanied by
system in going from A to B and ΔH(0) and ΔS(0) are the aggregation or other irreversible effects. This gives rise to

r 2010 American Chemical Society 3299 DOI: 10.1021/jz1012142 |J. Phys. Chem. Lett. 2010, 1, 3298–3304
PERSPECTIVE
pubs.acs.org/JPCL

Figure 3. DSC thermogram of the thermal aggregation of insulin Figure 4. Typical thermodynamic data for reversible two-state
in solution under conditions where the unfolded protein under- unfolding of a globular protein, with Tm = 40 C, ΔHunf(Tm) =
goes a kinetically limited exothermic nucleation-growth transition 300 kJ mol-1, and ΔCp = 9 kJ mol-1 K-1.
to a condensed protein aggregate similar to amyloid fibrils.12,13 The
inset shows the more typical thermogram seen for protein un-
folding (without aggregation). Note the decreased heat capacity amino acid side chains is generally small with respect to
for the fibril state (arrow). (Adapted from ref 12, with permission; ΔHunf.
structures drawn from pdb files 4INS14 and 2KIB.15) This positive ΔCp has been taken as a defining feature of
experimental artifacts, as indicated in Figure 2, that can make protein unfolding in solution, but it is possibly better to ob-
accurate heat capacity analysis difficult. However, sufficient serve that the lower heat capacity of folded protein (below Tm)
data on reasonably representative samples are now available is rather more characteristic of a condensed state of the
so as to make thermodynamic analysis appropriate. polypeptide. This is indicated by DSC data on aggregated
protein (Figures 2 and 3), showing that condensed protein
phases, either nonspecific aggregates (Figure 2) or more
ordered amyloid-like precipitates (Figure 3) as well as the
Unfolded protein is sticky stuff. natively folded state (Figure 1), all show a relatively low heat
capacity in comparison with the more dynamic, unfolded
polypeptide.
As a consequence of the positive ΔCp, the thermody-
For well-behaved proteins undergoing reversible thermal namics of the protein unfolding transition (assuming a simple
transitions, the heat capacity curves obtained by DSC two-state equilibrium) show significant temperature effects,
(Figure 1) show several characteristic features. Starting from as illustrated in Figure 4. Historically, thermodynamic stability
low temperature, the apparent protein heat capacity gradually profiles such as these were first obtained by indirect spectro-
increases until the onset of thermal unfolding, accompanied scopic analyses, predating more direct calorimetric observa-
by a sharp peak in Cp corresponding to an endothermic un- tions, and are a tribute to the careful measurements done at
folding transition. The peak of the transition gives the mid- that time.16 What they show is that both enthalpic (ΔHunf) and
point temperature (Tm), and the area under the transition entropic (TΔSunf) contributions increase markedly with tem-
gives the enthalpy of unfolding (ΔHunf) at this temperature. perature, but in a compensatory fashion such that the unfold-
The heat capacity of the protein above the transition, once it ing free energy (ΔGunf = ΔHunf - TΔSunf) is relatively small
has returned to baseline, is typically higher than would be and much less sensitive to temperature variation. This “thermo-
anticipated from extrapolation of the low-temperature base- dynamic homeostasis”8 may be functionally useful and is
line. Consequently, the thermal unfolding shows a positive one example of more ubiquitous “enthalpy/entropy com-
heat capacity change, ΔCp. This is also indicated by the pensation” effects now seen in many systems.8,17,18 Here, it
marked increase in the area under the transition peak is simply a consequence of a positive ΔCp, which, by virtue of
(ΔHunf) with increasing Tm. In practice, because post-transi- the fundamental thermodynamic relationships, eqs 4 and 5,
tion baselines are often poorly defined, it is customary to requires that both ΔH and ΔS increase with temperature, with
perform a series of experiments, varying conditions such as approximately linear correlation over narrow temperature
pH, to give ΔHunf over a range of Tm values, from which ΔCp ranges.8,18 The midpoint temperature of the unfolding transi-
may be estimated using eq 6. The validity of this approach tion, Tm, equivalent to the “melting point” in conventional
has been demonstrated by Pfeil and Privalov11 and can be phase transitions, where ΔGunf = 0, is given by the point at
justified because the heats of proton ionization of acidic which the ΔHunf and TΔSunf lines intersect. Because the ΔH

r 2010 American Chemical Society 3300 DOI: 10.1021/jz1012142 |J. Phys. Chem. Lett. 2010, 1, 3298–3304
PERSPECTIVE
pubs.acs.org/JPCL

predictions,19 though the structure of the cold denatured


state is ambiguous.20
Anomaly? What Anomaly? The conclusions arising out of
the positive ΔCp for protein unfolding are in many ways
counterintuitive. For example, the positive slope of the ΔHunf
versus T plot (Figure 4) means that unfolding gets more
endothermic with increasing temperature; that is, the higher
the temperature, the more heat energy that is needed to
unfold the protein. Conversely, at sufficiently low tempera-
tures, ΔHunf becomes negative, and exothermic unfolding
(cold denaturation) becomes a possibility. Such apparently
anomalous effects were initially unexpected and led to much
discussion and (sometimes) confusion. However, they were
taken as a striking vindication of the hypothesis that protein
folding was driven by hydrophobic interactions or, more
cautiously, that a positive ΔCp reflects exposure of hydropho-
bic groups upon unfolding. For example, to quote Brandts and
Hunt in 1967, “...More specifically, there is an anomalously
large heat capacity associated with the unfolded protein
which apparently reflects a striking alteration in the process
of accommodation of the exposed hydrophobic side chains in
the solvent phase with changing temperature...”.21 The hydro-
phobic effect22 and its possible role in biomolecular interac-
tions was highlighted by Walter Kauzmann's influential
review,23 laying the foundations for much of what has been
Figure 5. Comparison of experimental specific heats (heat capa- discovered since. However, it is interesting to note that, even
city, Cp) at 298 K for pure organic solids and liquids24,25 (top panel) then, Kauzmann was careful to point out that unusual thermo-
with folded and unfolded proteins in aqueous solution26 (bottom
panel). Experimental data are for compounds up to and including dynamic effects are not just restricted solely to hydrophobic
C5 with elemental composition CHO, CHN, or CHNO, for which groups, and that similar thermodynamic signatures could be
data are available (118 crystalline solids and 159 pure liquids). The inferred for interactions involving hydration of salts and other
mean values ((s.d.) are 1.32 ((0.22) and 2.18 ((0.38) J K-1 g-1
for solids and liquids, respectively. Mean Cp values for native (n = polar groups; “It is evident that both salt linkages and hydro-
35) and denatured (n = 13) proteins are 1.49 ((0.13) and 1.95 phobic bonds are stabilized predominantly by entropy effects
((0.15) J K-1 g-1 for native and denatured states, respectively. The rather than by energy effects.”23 Therefore, let us re-examine
curves show Gaussian fits to the individual distributions. (Adapted
from ref 24, with permission.) the situation as regards protein heat capacity.

and TΔS lines are almost parallel on this scale (Figure 4), the
intersection point (Tm) will be very sensitive to relatively small
perturbations, analogous to the familiar moir e effect, where Unusual thermodynamic effects
slight changes in displacement between almost parallel lines
gives rise to much larger changes in fringe patterns. This are not just restricted solely
indicates why it is so difficult to predict the thermal stability of
proteins, and I am not aware of any molecular dynamics
to hydrophobic groups.
or other theoretical simulations that have yet addressed this
seriously.
The curvature of the free-energy line in the thermal stabil- Comparison with heat capacities of other substances is
ity profiles (Figure 4) is a direct consequence of the finite informative. Figure 5 shows the absolute specific heats (heat
positive ΔCp, and as a result, proteins have a temperature of capacity per gram) of various folded and unfolded proteins in
maximum thermodynamic stability (maximum ΔGunf) for solution together with a range of organic compounds in the
folding. ΔGunf falls off either side of this maximum, becoming solid or liquid state at room temperature. As would be
negative (i.e., thermodynamically favorable for unfolding) not expected from the increase in degrees of freedom in the fluid
only at high temperatures, as would be expected, but also, by state, organic liquids tend to have a higher specific heat
extrapolation, at low temperatures. This implies that native capacity than solids. Data for proteins in solution fall in the
proteins should unfold at sufficiently low temperatures, same range, with Cp for folded proteins somewhat higher, on
though this temperature is usually below the freezing point average, than organic solids and the more flexible, unfolded
of water and experimentally inaccessible under normal con- proteins slightly lower than organic liquids. One might ratio-
ditions. Alternatively, starting with unfolded protein at low nalize this, at least qualitatively, by simply observing that the
temperature, the protein should fold spontaneously as the folded protein state is probably a bit more flexible and
temperature is raised. The demonstration of “cold denaturation” dynamic than a crystalline organic solid, whereas the un-
in certain cases was a satisfying verification of these folded protein, though undoubtedly even more flexible and

r 2010 American Chemical Society 3301 DOI: 10.1021/jz1012142 |J. Phys. Chem. Lett. 2010, 1, 3298–3304
PERSPECTIVE
pubs.acs.org/JPCL

relatively little effect on heat capacity estimates at this crude


level of approximation.
It is therefore interesting to note that, if viewed as a simple
order-disorder transition, the heat capacity increase upon
protein unfolding is entirely unexceptional, with experimen-
tal values lying comfortably within the range that would be
expected from the properties of organic materials in general
(Figure 5), both theoretically and experimentally. This is
further indicated by comparison (Figure 6) with the heat
capacity increments upon melting of a range of simple
compounds, including water. Note that the ΔCp upon melting
of polar substances is significantly greater than that for simple
hydrocarbons, consistent with our understanding of residual
hydrogen-bonded structures in polar liquids, water being the
prime (but not only) example. Again, plotted on the same
scale, the ΔCp for protein unfolding is unexceptional. There-
fore, where is the anomaly? Furthermore, does this mean that
we might reasonably treat protein unfolding as a simple
Figure 6. Comparison of the specific heat capacity increments “melting” process, albeit with quite small lumps of material
upon melting of simple compounds (right) with protein unfolding immersed in a solvent environment?
(left).24 Absolute heat capacities for pure solid and liquid com-
pounds are plotted with respect to the normal melting point (ΔT =
T - Tm). For simplicity, the heat capacity plots for simple com-
pounds omit the large spike in Cp associated with heat uptake at
the melting point, ΔT = 0. For proteins, being smaller cooperative
units, this heat uptake takes place over a wider temperature range,
as indicated by the dashed line. (Adapted from ref 24, with The heat capacity increase upon
permission.)
protein unfolding is entirely
dynamic, is still a polymer and will have fewer available unexceptional.
degrees of freedom than a small molecule in the true liquid
phase. (Bear in mind also that thermally unfolded protein still
retains considerable structure which, though dynamic and Of course, protein unfolding is not really a phase transition
heterogeneous, is a long way from true random coil polymer.) in the classic sense. The size of the cooperative unit is finite,
To get a rough idea of the numbers involved, consider the determined by the size of the macromolecule in the simplest
following. Classical equipartition theory would give an aver- case, and this is why the thermal unfolding transition of the
age thermal energy of order ½kBT for each thermally protein is relatively broad compared to the (infinitely) sharp
accessible degree of freedom, corresponding to a Cp of around cooperative melting transitions of crystalline solids. Moreover,
½kB, or ∼4 J mol-1 K-1 per degree of freedom. For a typical the protein molecules are in solution, with all of the potential
amino acid (mean residue weight ≈ 110 Da), this converts to a solvation/hydration issues that this entails. Indeed, that is the
specific heat change of around 0.04 J g-1 K-1 per degree of picture that has dominated much of the thinking about
freedom. Consequently, the average ΔCp increment of order protein folding thermodynamics, driven by burial of hydro-
0.5 J g-1 K-1 (Figure 5) upon protein unfolding would phobic side chains in the relatively nonpolar interior of the
translate to an increase of about 12 degrees of freedom per globular protein structure. However, when I look inside of a
amino acid residue in this crude estimate. This seems not protein structure, it does not look particularly “hydrophobic”,
unreasonable. Consider also the classic empirical Dulong and certainly not to the extent that one sees inside of a detergent
Petit law, also rationalized by thermodynamic equipartition micelle, for example. Yes, there are more nonpolar groups
theory, which gives an upper limit to the heat capacity (for inside, but maybe not so many as one might expect, and with
solids) of 3R (=3NAkB =25 kJ mol-1 K-1) per mole of atoms. every amino acid side chain, polar or otherwise, there comes
For a typical atomic weight of 14 (e.g., C, N, and O in proteins, the polar amide backbone. In structural terms, this is what
ignoring H), this corresponds to a specific heat of ∼1.8 J g-1 K-1, usually dominates our description of the folded protein, an
which is remarkably (and probably fortuitously) close to intricate, closely packed network of hydrogen bonds, mostly
the observed values for organic liquids and unfolded protein involving backbone -NH and -CdO groups, in which the
(Figure 5). Heat capacities for real solids (and folded overriding rule seems to be “let no H-bond go unsatisfied”.
proteins?) would naturally be lower at normal temperatures Potential voids in this structure are filled by side chains,
because of quantum limitations on the accessibility of vibra- usually nonpolar (to avoid breaking the rule), but almost as
tional degrees of freedom. Of course, order-of-magnitude an afterthought ; a topologically complex, three-dimen-
estimates such as these have ignored complexities arising sional jigsaw puzzle in which the main aim is to fill space as
from intermolecular interactions in the liquid state and/or efficiently as possible to give a densely packed structure.
solvation effects, but such effects will generally only serve to Maybe this is the clue. Such structures inevitably rely on
replace one kind of degree of freedom with another, with cooperativity; not until the last piece is in place does the

r 2010 American Chemical Society 3302 DOI: 10.1021/jz1012142 |J. Phys. Chem. Lett. 2010, 1, 3298–3304
PERSPECTIVE
pubs.acs.org/JPCL

structure stabilize. Partially folded structures just will not hang Therefore, although the heat capacities of proteins are
together in the face of constant thermal buffeting; it is either perhaps less anomalous than was once thought, they are still
folded or it is not. (Okay, the two-state model for protein fundamental to our understanding of the thermodynamics of
folding is an approximation, and the real picture is one of a protein folding and binding. In the absence of suitable empiri-
complex, multidimensional conformational space popu- cal models, one might hope that accurate molecular dynamics
lated by different distributions, but the simpler view will simulations or other computational methods could provide
serve for now.) Perhaps, in terms of physical chemistry at some guidance. However, this remains a major computational
least, we should look upon the folded protein molecule as challenge. Although some progress is being made in estimation
just another (albeit small) chunk of organic solid. This is not a of free-energy changes, no realistic simulation has yet modeled
new idea (very little is in this Perspective). It has long been the experimentally observed heat capacity effects and asso-
recognized that the molecular packing densities within ciated variations in enthalpy and entropy. This is understand-
globular proteins are close to what is found in organic able; adequate computation at just one temperature is hard
solids.27,28 The adiabatic compressibilities and thermal ex- enough, especially with inclusion of solvent; therefore, repeti-
pansion coefficients of protein molecules are likewise com- tion at different temperatures just increases the burden. Inter-
parable to organic solids.29 So why not heat capacity? estingly, however, the direct relationship between heat capacity
Please be aware that I am being deliberately provocative and energy fluctuations (eq 3) might provide a convenient and
here. The conventional approach has been to explain ΔCp (and computationally economical method for estimation of Cp in
other thermodynamic properties) in terms of changes in the properly configured molecular dynamics simulations. The
environment of polar and nonpolar groups, often taking a two- “noise” in such simulations at a single temperature, once
dimensional analysis based on changes in solvent-accessible equilibrated with an appropriate thermal bath, should relate
surface areas (ΔASA). This seems reasonable and has the directly to the heat capacity of the system. This would probably
advantage that it can be parametrized, at least in principle, by be most easily implemented in a ligand binding simulation
comparison with the thermodynamics of small-molecule trans- where the energy fluctuations in the free protein at a particular
fer or by statistical analysis of ΔCp data from proteins of known temperature should be reduced in the (usually) more rigid
structure.2,3,30,31 However, this approach attributes essentially ligand-bound state. Calculation of the resulting decrease in heat
all of the effects to solvation changes, potentially at the expense capacity (negative ΔCp), using eq 3, in comparison with that in
of the neglect of more global, three-dimensional contributions experiment, would be an interesting exercise.
from conformational dynamics and other protein intramolec-
ular interactions.8,12,24 It is possibly for this reason that area- AUTHOR INFORMATION
based models for ΔCp have come to widely differing conclu-
sions about the relative contributions of polar versus nonpolar Corresponding Author:
*E-mail: alanc@chem.gla.ac.uk. Phone: (þ44) 141 330 5278. Fax:
groups, for example.2,3 All ΔASA models agree that exposure of
(þ44) 141 330 4888.
nonpolar groups during protein unfolding should result in an
increase in heat capacity, as classically expected for hydropho-
bic groups, though absolute magnitudes vary. However, totally Biographies
Alan Cooper is Professor of Biophysical Chemistry at the Uni-
disparate results are obtained for polar groups, with no agree- versity of Glasgow. Following B.Sc. (Theoretical Physics) and Ph.
ment as to the sign, let alone the magnitude of the ΔCp, to be D.(Biophysics) studies at Manchester, he did postdoctoral work at
attributed to their exposure. This probably reflects the paucity of Oxford and Yale, laying foundations for a lifetime's fascination (and
data for real proteins and the lack of appropriate model systems frustration) with the thermodynamics of biomolecular structures
in a statistically underdetermined system, together with un- and their interactions. His association with Joseph Black's old
certainties about ΔASA estimates in the absence of information department is entirely serendipitous. By a further quirk of history,
about polypeptide conformations in the thermally denatured his maternal ancestors were neighbors of Joseph Black, working
state. In the most comprehensive analysis to date, Robertson nearby in Glasgow as impoverished handloom weavers. Personal
and Murphy3 showed that there is little statistically significant website: http://www.chem.gla.ac.uk/staff/alanc/
difference between polar and nonpolar groups; both make ACKNOWLEDGMENT Much of this work evolved under the
similar positive contributions to ΔCp, as might be expected auspices of the Biological Microcalorimetry Facility in Glasgow,
from fundamental principles.8 The best correlation is with the funded by the U.K. Biotechnology and Biological Sciences
overall size or number of amino acid residues in the protein, Research Council, and I am grateful to numerous colleagues and
collaborators over many years, none of whom are to be blamed for
regardless of their polarity.3
the crazy ideas presented here.
For simplicity, and in keeping with the theme of this issue,
I have limited discussion to protein (un)folding. However,
similar heat capacity effects are seen in most protein-protein REFERENCES
and protein-ligand interactions, regardless of whether they
(1) Guerlac, H. Joseph Black's Work on Heat. In Joseph Black
involve hydrophobic groups or not.12 The consequences can
1728-1799: A Commemorative Symposium; Simpson,
be quite extreme, with heats of binding varying significantly A. D. C., Ed.; Royal Scottish Museum: Edinburgh, U.K., 1982.
and sometimes even changing sign over relatively small (2) Prabhu, N. V.; Sharp, K. A. Heat Capacity in Proteins. Annu.
temperature ranges. This implies that enthalpy (ΔH) mea- Rev. Phys. Chem. 2005, 56, 521–548.
surements at a single temperature might be less informative (3) Robertson, A. D.; Murphy, K. P. Protein Structure and the
than is sometimes assumed. Energetics of Protein Stability. Chem. Rev. 1997, 97, 1251–1267.

r 2010 American Chemical Society 3303 DOI: 10.1021/jz1012142 |J. Phys. Chem. Lett. 2010, 1, 3298–3304
PERSPECTIVE
pubs.acs.org/JPCL

(4) Cooper, A. Thermodynamic Fluctuations in Protein Mole- (23) Kauzmann, W. Some Factors in the Interpretation of Protein
cules. Proc. Natl. Acad. Sci. U.S.A. 1976, 73, 2740–2741. Denaturation. Adv. Protein Chem. 1959, 14, 1–63.
(5) Cooper, A. Protein Fluctuations and the Thermodynamic Un- (24) Cooper, A. Heat Capacity of Hydrogen-Bonded Networks: An
certainty Principle. Prog. Biophys. Mol. Biol. 1984, 44, 181–214. Alternative View of Protein Folding Thermodynamics. Bio-
(6) Plotnikov, V. V.; Brandts, J. M.; Lin, L.-N.; Brandts, J. F. A New phys. Chem. 2000, 85, 25–39.
Ultrasensitive Scanning Calorimeter. Anal. Biochem. 1997, (25) Domalski, E. S.; Hearing, E. D. Heat Capacities and Entropies
250, 237–244. of Organic Compounds in the Condensed Phase. Volume Iii.
(7) Privalov, G.; Kavina, V.; Freire, E.; Privalov, P. L. Precise J. Phys. Chem. Ref. Data 1996, 25, 1–525.
Scanning Calorimeter for Studying Thermal-Properties of (26) Makhatadze, G. I. Heat Capacities of Amino Acids, Peptides
Biological Macromolecules in Dilute-Solution. Anal. Biochem. and Proteins. Biophys. Chem. 1998, 71, 133–156.
1995, 232, 79–85. (27) Richards, F. M. The Interpretation of Protein Structures: Total
(8) Cooper, A.; Johnson, C. M.; Lakey, J. H.; Nollmann, M. Heat Volume, Group Volume Distributions and Packing Density.
Does Not Come in Different Colours: Entropy-Enthalpy J. Mol. Biol. 1974, 82, 1–14.
Compensation, Free Energy Windows, Quantum Confine- (28) Liang, J.; Dill, K. A. Are Proteins Well Packed? Biophys. J.
ment, Pressure Perturbation Calorimetry, Solvation and the 2001, 81, 751–766.
Multiple Causes of Heat Capacity Effects in Biomolecular (29) Chalikian, T. V. Volumetric Properties of Proteins. Annu. Rev.
Interactions. Biophys. Chem. 2001, 93, 215–230. Biophys. Biomolec. Struct. 2003, 32, 207–235.
(9) Cheetham, J. C.; Artymiuk, P. J.; Phillips, D. C. Refinement of (30) Makhatadze, G. I.; Privalov, P. L. Energetics of Protein Struc-
an Enzyme Complex with Inhibitor Bound at Partial Occu- ture. Adv. Protein Chem. 1995, 47, 307–425.
pancy. Hen Egg-White Lysozyme and Tri-N-Acetylchitotriose (31) Spolar, R. S.; Livingstone, J. R.; Record, M. T. Use of Liquid-
at 1.75 Å Resolution. J. Mol. Biol. 1992, 224, 613–628. Hydrocarbon and Amide Transfer Data to Estimate Contribu-
(10) Cooper, A. Thermodynamics of Protein Folding and Stability. tions to Thermodynamic Functions of Protein Folding from
In Protein: A Comprehensive Treatise; Allen, G., Ed.; JAI Press the Removal of Nonpolar and Polar Surface from Water.
Inc.: Greenwich, CT, 1999; Vol. 2; pp 217-270. Biochemistry 1992, 31, 3947–3955.
(11) Pfeil, W.; Privalov, P. L. Thermodynamic Investigations of
Proteins. I. Standard Functions for Proteins with Lysozyme as
an Example. Biophys. Chem. 1976, 4, 23–32.
(12) Cooper, A. Heat Capacity Effects in Protein Folding and
Ligand Binding: A Re-Evaluation of the Role of Water in
Biomolecular Thermodynamics. Biophys. Chem. 2005, 115,
89–97.
(13) Dzwolak, W.; Ravindra, R.; Lendermann, J.; Winter, R. Ag-
gregation of Bovine Insulin Probed by DSC/PPC Calorimetry
and FTIR Spectroscopy. Biochemistry 2003, 42, 11347–
11355.
(14) Baker, E. N.; Blundell, T. L.; Cutfield, J. F.; Cutfield, S. M.;
Dodson, E. J.; Dodson, G. G.; Hodgkin, D. M.; Hubbard, R. E.;
Isaacs, N. W.; Reynolds, C. D. The Structure of 2Zn Pig Insulin
Crystals at 1.5 Å Resolution. Philos. Trans. R. Soc. London, Ser.
B 1988, 319, 369–456.
(15) Nielsen, J. T.; Bjerring, M.; Jeppesen, M. D.; Pedersen, R. O.;
Pedersen, J. M.; Hein, K. L.; Vosegaard, T.; Skrydstrup, T.;
Otzen, D. E.; Nielsen, N. C. Unique Identification of Supra-
molecular Structures in Amyloid Fibrils by Solid-State NMR
Spectroscopy. Angew. Chem., Int. Ed. 2009, 48, 2118–2121.
(16) Brandts, J. F. The Thermodynamics of Protein Denaturation.
I. The Denaturation of Chymotrypsinogen. J. Am. Chem. Soc.
1964, 86, 4291–4301.
(17) Dunitz, J. D. Win Some, Lose Some ; Enthalpy-Entropy
Compensation in Weak Intermolecular Interactions. Chem.
Biol. 1995, 2, 709–712.
(18) Sharp, K. Entropy-Enthalpy Compensation: Fact or Artifact?
Protein Sci. 2001, 10, 661–667.
(19) Privalov, P. L. Cold Denaturation of Proteins. Crit. Rev. Bio-
chem. Mol. Biol. 1990, 25, 281–305.
(20) Davidovic, M.; Mattea, C.; Qvist, J.; Halle, B. Protein Cold
Denaturation as Seen from the Solvent. J. Am. Chem. Soc.
2009, 131, 1025–1036.
(21) Brandts, J. F.; Hunt, L. Thermodynamics of Protein Denatura-
tion. III. Denaturation of Ribonuclease in Water and in
Aqueous Urea and Aqueous Ethanol Mixtures. J. Am. Chem.
Soc. 1967, 89, 4826–4838.
(22) Tanford, C. The Hydrophobic Effect: Formation of Micelles and
Biological Membranes; John Wiley & Sons: New York, 1973.

r 2010 American Chemical Society 3304 DOI: 10.1021/jz1012142 |J. Phys. Chem. Lett. 2010, 1, 3298–3304

You might also like