- Email: [email protected]
The new view of hydrophobic free energy
FEBS Letters 587 (2013) 1062–1066
journal homepage: www.FEBSLetters.org
Review
The new view of hydrophobic free energy Robert L. Baldwin Biochemistry Department, Beckman Center, Stanford University Medical Center, Stanford, CA 94305, USA
a r t i c l e
i n f o
Article history: Received 19 November 2012 Revised 7 January 2013 Accepted 8 January 2013 Available online 18 January 2013 Edited by Wilhelm Just Keywords: Transfer free energy Reference solvent Liquid alkane Folding analogy
a b s t r a c t In the new view, hydrophobic free energy is measured by the work of solute transfer of hydrocarbon gases from vapor to aqueous solution. Reasons are given for believing that older values, measured by solute transfer from a reference solvent to water, are not quantitatively correct. The hydrophobic free energy from gas–liquid transfer is the sum of two opposing quantities, the cavity work (unfavorable) and the solute–solvent interaction energy (favorable). Values of the interaction energy have been found by simulation for linear alkanes and are used here to find the cavity work, which scales linearly with molar volume, not accessible surface area. The hydrophobic free energy is the dominant factor driving folding as judged by the heat capacity change for transfer, which agrees with values for solvating hydrocarbon gases. There is an apparent conflict with earlier values of hydrophobic free energy from studies of large-to-small mutations and an explanation is given. Ó 2013 Federation of European Biochemical Societies. Published by Elsevier B.V. All rights reserved.
Ever since the pioneering work by Kauzmann [1] in 1959 and by Nozaki and Tanford [2] in 1971, hydrophobic free energy (DGh) has been determined from values of the transfer free energy when nonpolar solutes are transferred from water to non-polar (or semi-polar) reference solvents. In recent years the standard choice of a reference solvent has been a liquid alkane [3]: the liquid alkane provides both the alkane solute and the non-polar reference solvent. In 1984 a basic criticism of this method was made by BenNaim and Marcus [4]: the solvation free energy of a non-polar solute is determined from gas–liquid, not liquid–liquid, transfer. When an alkane solute is transferred from water through vapor to the liquid alkane, the liquid–liquid transfer may be written as two successive gas–liquid transfers [5]: first from water into vapor and then from vapor into liquid alkane. Kauzmann [1] and Tanford [2] expected that removing the hydrocarbon solute from water would account for most of the free energy change in liquid–liquid transfer but this expectation was not correct: data are available for the DG values of both gas–liquid transfers, which show that the ‘‘hydrophobic’’ transfer from water to vapor accounts for less than half of the total DG (5–7). This disconcerting fact was realized as early as 1976 when Wolfenden and Lewis [6] studied why water is a poor solvent for liquid hydrocarbons. They found that a strong favorable interaction among alkane molecules in liquid alkanes gives a strongly favorable transfer free energy for passage of an alkane solute from vapor into liquid alkane. Kauzmann [1] made an analogy between the protein folding process and the transfer of a non-polar solute from water into a
E-mail address: [email protected]
reference solvent, since folding transfers the non-polar side chains of an unfolded protein out of water into a non-aqueous environment, the protein interior. It was well-known that water is a poor solvent for hydrocarbons, and Kauzmann showed that the DG for transferring a hydrocarbon out of water into a liquid hydrocarbon is substantial compared to the net DG for folding a small protein. For example, the DG for transfer from water to liquid alkane is 4.8 kcal/mol for butane, a hydrocarbon which has four carbon atoms like the side chains of leucine and isoleucine. Thus, he argued, a similar DG should help to fold a protein when a non-polar side chain is buried by folding. The analogy breaks down, however, when the liquid–liquid transfer is divided into two gas–liquid transfers [5], because the ‘‘hydrophobic’’ transfer from water into vapor has less than half of the total DG. and the second transfer, from vapor to liquid alkane, is not obviously related to the protein folding process. For this and the following reasons, I argued recently [7] that hydrophobic free energy should be defined and measured by the ‘‘hydrophobic’’ transfer of alkane solutes from water into vapor. The new definition is consistent with the criticism of Ben-Naim and Marcus [4] that the solvation free energy of a non-polar solute must be measured by gas–liquid transfer. The new definition is also consistent with work by Makhatadze and Privalov [8] and Wu and Prausnitz [9], who used gas–liquid transfer data to solve folding problems involving the hydrophobic factor. The new definition of DGh accepts part of Kauzmann’s analogy between the folding process and solute transfer but stops after the first transfer from water into vapor. The second half of the analogy, which involves solute transfer from vapor into liquid alkane, is omitted because the transfer of a non-polar side chain into
0014-5793/$36.00 Ó 2013 Federation of European Biochemical Societies. Published by Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.febslet.2013.01.006
R.L. Baldwin / FEBS Letters 587 (2013) 1062–1066
the protein interior has different properties [7] from those of the transfer of an alkane solute from vapor into liquid alkane. These other properties include the topological connection of the non-polar side chain to the polypeptide chain, the presence of semipolar, H-bonded-NHCO– groups in the protein interior, and close packing of the groups inside a folded protein. 1. Hydrophobic free energy is the sum of two opposing quantities The gas–liquid transfer free energy, DGGL, which corresponds to the new DGh [7], is the sum of DGc (the cavity work) and Ea (the solute–solvent interaction energy), according to independent derivations of Eq. (1) (below) by Lee [10] and Pollack [11]. These authors analyzed the solute insertion model for equilibrium of a non-polar solute between vapor and liquid phases. In this model a cavity is first made for the solute in the liquid and then the solute is inserted; DGc is the work needed to make the cavity and Ea is the interaction energy between solute and solvent after insertion. Both derivations [10,11] start from the equality of the solute’s chemical potential between the gas and liquid phases. Lee [10] gave a detailed derivation, based on statistical mechanics, of Eq. (1) that includes the solvent reorganization after solute insertion, and he shows that this term is minor and can be dropped. He gives evidence [10], however, for important solvent reorganization as the cavity is made, but not after solute insertion
DGGL ¼ DGc þ Ea
ð1Þ
The transfer free energy is measured from the Ostwald coefficient L, which is the equilibrium ratio of the solute concentration (molar or number density concentration) in the liquid over the gas phase.
DGGL ¼ RT lnL
ð2aÞ
L ¼ ðM L Þ=ðM G Þ
ð2bÞ
Values of the gas–liquid transfer free energies of linear alkanes are given in [4], based on McAuliffe’s 1966 study [12] of the solubilities of gaseous hydrocarbons in water at 25 °C. McAuliffe used a gas chromatograph to determine hydrocarbon solubility and he found that separating impurities away from the test hydrocarbon was the key to getting accurate results. He maintained a pressure of 1 atmosphere gas while equilibrating the hydrocarbon gas with water. The ideal gas law was used [4,10] to give the molar concentration of the gas from its partial pressure, after subtracting the partial pressure of water, and the chromatograph results gave the molar concentration of the hydrocarbon in aqueous solution. Table 1 gives the values in water at 25 °C of DGc, Ea, DGGL, and DGLL for 4 linear alkanes, ethane through pentane. The values show
Table 1 Transfer energetics for linear alkanes.a Solute
Eab
DGcc
DGGLd
DGLLe
Ethane Propane Butane Pentane
4.80 6.60 8.29 9.99
6.57 8.57 10.4 12.3
1.77 1.98 2.15 2.34
3.14 3.92 4.79 5.74
a All values are in kcal/mol at 25 °C. Ea values are from [13], DGGL values are from Table V of [4], DGc values are found from Eq. (1) using DGGL and Ea values given here, DGLL values are found from Tables 1 and V of [4] (see [5]). DGGL is the sum of Ea and DGc. b The solute–solvent interaction energy from simulation [13]. c The cavity work. d The transfer free energy from vapor to water, which equals the hydrophobic free energy. e The transfer free energy from liquid alkane to water.
1063
that the liquid–liquid transfer energy DGLL is more than twofold larger than the gas–liquid transfer energy DGGL. The values of Ea in water are given by Jorgensen and coworkers [13]. They obtained them by exhaustive simulations of alkane conformations in water, using the TIP4P model of water and the Lennard–Jones potential for the interaction between water and alkane. Lee [10] recognized that, when Ea values are available from simulation, DGc values can be found from equation (1) by using experimental values of DGGL. An important result in Table 1 is that DGc and Ea are opposing factors in DGh and their difference determines the size of DGh. It has not always been recognized that opposing factors determine hydrophobic free energy (see below), although this is clear in the early work of Pratt and Chandler [14] and in their subsequent papers. A failure to recognize the roles of two different factors has sometimes led to misinterpretation of results. Kauzmann [1] originally spoke of ‘‘hydrophobic bonds’’. In 1968, Hildebrand [15] sharply criticized this term because he and coworkers had shown that hydrocarbons interact favorably (and strongly) with water. An oil drop added to a container of water does not remain a drop, rather it spreads out to form a film that interacts with as much water as possible. Kauzmann, Nemethy and Scheraga [16] replied to Hildebrand by agreeing that there are favorable van der Waals interactions between hydrocarbon and water, but they argued that the entropy-driven character of the hydrophobic interaction (at 25 °C) is best explained by ordering of water molecules around the non-polar groups [1,17,18]. Tanford did not reply to Hildebrand immediately but he was careful afterwards to refer to the ‘‘hydrophobic effect’’ [19] and later to the ‘‘hydrophobic factor’’ [20], not to ‘‘hydrophobic bonds’’. Eleven years later, he replied [21] to Hildebrand by showing that the cohesive energy of water is large compared either to that of hexane or to the adhesive energy of water and hexane. He argued that the hydrophobic effect arises from the large cohesive energy of water, which tends to squeeze out all solutes. Polar solutes resist by making strong interactions with water while hydrocarbons respond by having very low solubilities in water. There is a second factor besides cohesive energy that contributes to the low solubilities of hydrocarbons in water, which is the unusually small size of the water molecule [22]; the relative merits of the two factors are still being debated [23]. When values of DGc became available in 1991 for 3 alkane solutes [10], it became clear that the low solubilities of gaseous alkane molecules in water are caused by large values of the cavity work. The low solubilities of liquid alkanes in water are caused additionally [6] by large values of Ea for alkane solutes in liquid alkanes.
2. The cavity work scales with molar volume The correlation of hydrophobic free energy with accessible surface area (ASA) or molar volume can be analyzed reliably for those hydrocarbons whose values are known of DGc and Ea. There has been a controversy [24] about how DGLL from liquid–liquid transfer scales with molecular size. It has been widely believed that hydrophobic free energy correlates with surface area [25–27] of the solute. Chandler [28,29] argued persuasively, however, that the cavity work in water is not proportional to surface area in the small-size regime (see below) but instead scales linearly with molar volume. Until now, the controversy has been based on data for liquid–liquid transfer, which is more complex than gas–liquid transfer because DGLL depends on the values of DGc and Ea in both liquid alkane and water [5]. Gas–liquid transfer results from Table 1 are used in Fig. 1 to plot DGc (water) against molar volume (V) for 4 linear alkanes. McAuliffe [12] showed that a plot of this kind should be limited to hydrocarbons of a single type, such as linear or cyclic, because
1064
R.L. Baldwin / FEBS Letters 587 (2013) 1062–1066
water near room temperature and become more soluble at both higher and lower temperatures. Note, however, that the minimum solubility of liquid alkanes in water occurs at Th, the temperature at which the transfer enthalpy passes through 0, and Th changes from 25 °C for liquid–liquid transfer [31] to 95 °C for gas–liquid transfer [37]. Both Kauzmann [1,18] and Dill [30] emphasized the striking and similar properties of DCp (the change in heat capacity) when proteins unfold and when non-polar liquids are dissolved in water. The basic evidence that DCp has a large value in protein unfolding came as early as 1979 from Privalov [32] who measured calorimetrically the DH of protein unfolding at different temperatures, and obtained DCp from the plot of DH versus T. He varied pH in the range pH 2–5, where protein stability depends critically on pH and DH can be measured at varying temperatures; DCp is almost independent of pH in this range. The DH and DS values for protein unfolding at moderate temperatures (T) are fitted well by Eqs. (3) and (4) with DCp constant, independent of T; Tr is any convenient reference temperature. Fig. 1. A plot of the cavity work (DGc) versus molar volume (V) for the 4 linear alkanes in Table 1, and the data points for isobutane and neopentane from Lee’s study [10] have been added. All values of V are from McAuliffe [12]. The point (0,0) has also been added.
different types of hydrocarbons give different lines, even though the various types all scale linearly with molar volume. Since each type gives a different line, hydrocarbon solubility in water depends on more variables besides molar volume. Fig. 1 gives a good straight line, indicating that DGc is proportional to molar volume for linear alkanes. as expected both from McAuliffe’s 1966 results [12] and Chandler’s more recent work [28,29]. The solute–solvent interaction energy is expected to be proportional to surface area [13,27] but the linear alkanes discussed here are flexible molecules and average ASA values found by exhaustive simulation of all possible conformations should be used to probe this question; the simulated values are not presently available. It is interesting to consider today why hydrophobic free energy was expected in early proposals [25–27] to be proportional to surface area. It seems likely that a main reason was the perceived importance of the ordering of water structure around non-polar groups. Another reason [27] was the short-range distance dependence of the attractive Lennard–Jones term. Note, however, that the water-ordering effect (which is discussed below) has been used to explain the water-hating behavior of non-polar groups [1,17,18,20] whereas the solute–solvent interaction (Ea) between water and hydrocarbon groups is favorable [15]. In 1974 Tanford and coworkers [27] said ‘‘. . .the hydrophobic free energy arises from contacts between hydrocarbon and water at the solute–solvent interface and would thus be expected to be a function of the surface area. . .’’, although Tanford later contradicted this view [21]. 3. Hydrophobic free energy considered as the dominant factor driving protein folding Dill concluded in a widely cited 1990 review [30] that hydrophobic free energy is the dominant factor driving protein folding. His argument emphasizes similarities between the transfer energetics when non-polar liquids are dissolved in water and when proteins unfold, since the unfolding process transfers non-polar side chains from the protein interior to water. Dill’s argument is based especially on the unusual temperature dependence of protein stability and similarities with dissolving non-polar liquids in water. His review points out that some proteins can be denatured both by heating and cooling while non-polar liquids have minimum solubility in
DH ¼ DHr þ DCpðT T r Þ
ð3Þ
DS ¼ DSr þ DCp lnðT=T r Þ
ð4Þ
4. Quantitative comparisons between hydrophobic free energy and protein folding data Two quantitative comparisons between protein folding energetics and properties of hydrophobic free energy are discussed in [7] and reviewed briefly below. The first is a comparison between the DCp values found for protein unfolding and for transfer of hydrocarbons from water into the gas phase. The second is a test of the ability of gas–liquid transfer data to predict the temperature dependence of the enthalpy of protein unfolding. Bolen’s work [33,34] shows that osmolytes drive protein folding by acting chiefly on the peptide backbone, which suggests that osmolytes affect the formation of peptide H-bonds. Changes in hydrophobic free energy occur mainly in processes involving the non-polar groups of protein side chains, not the polar groups in the peptide backbone, and Bolen’s work raises the question of whether DGh is in fact the dominant factor in protein folding. To answer this question, the following approach was taken [7]. The free energy change that drives folding was taken from heat-induced unfolding since free energy is a state function. Then the temperature dependence of the free energy change, which is characterized by DCp, was used to find its source. An accurate value for heat-induced unfolding, (DCp/ n) = 13.9 ± 0.5 cal K1 mol1, was given by Robertson and Murphy [35], who surveyed the thermodynamic properties of 49 protein unfolding reactions. Their survey shows that DCp for protein unfolding is proportional to n, the number of residues in a protein. The protein DCp value has a positive sign, whereas Richardson and Makhatadze [36] found that DCp for unfolding a peptide helix and breaking peptide H-bonds is 7.6 cal K1 mol res1 with a negative sign, which rules out peptide H-bonds as the dominant source of the free energy that drives folding. To compare the protein DCp value with ones for the transfer of hydrocarbons from water into the gas phase, values of (DCp/ASA) can be compared [7]. Values of (DCp/ASA) for gas–liquid transfer of hydrocarbon gases were given by Makhatadze and Privalov [37] as follows: ethane 0.402, propane 0.395, and butane 0.410 cal K1 mol1 Å2. To find the protein unfolding value of (DCp/ASA), the (DCp/n) value above can be divided by 42 Å2, which is the average amount of non-polar surface area buried per protein residue [7]. The result is (DCp/ASA) = 0.33 cal K1 mol1 Å2 [7],
R.L. Baldwin / FEBS Letters 587 (2013) 1062–1066
which is close to the value for transferring hydrocarbons from water into the gas phase. This comparison strongly suggests that DGh is the dominant source of the free energy that drives folding. The likely reason [7] why osmolyte studies appear to give a different answer is that osmolyte-driven folding/unfolding is a different process from thermal folding/unfolding, since the osmolyte participates in the osmolyte-driven process. 5. Temperature dependence of protein stability To find out if gas–liquid transfer results correctly predict the temperature dependence of the enthalpy of protein unfolding, the following test was made [7]. Experimental values are available [8] both of the solvation enthalpies of protein non-polar and polar groups (for nine proteins, at temperatures from 5 to 100 °C) and of the corresponding enthalpy values for protein unfolding. The results for ribonuclease T1 were used in [7] to test whether gas–liquid transfer results, when used to estimate the solvation enthalpies of the polar and non-polar groups, correctly predict the large temperature dependence of the protein unfolding enthalpy. In making this test, the basic assumptions are: (1) the solvation enthalpies of the polar and non-polar groups have been estimated correctly, and (2) no other energetic variable (such as the van der Waals interactions) changes enough with temperature to affect seriously the DH of unfolding. If there is complete agreement between predicted and observed results in this test, the data points should fall on a straight line that passes through (0,0) and has slope 1.00. A good straight line was found [7] that passes through 0.0 and has slope 1.25. When the contribution from the polar peptide groups was omitted, the slope increased to 1.87, indicating that the effect of the polar peptide groups must be taken into account. Solvation of the non-polar groups upon unfolding is, however, the major cause of the strong temperature dependence of the enthalpy of protein unfolding [7]. The failure to obtain complete agreement (slope 1.25 observed versus 1.00 expected) remainns to be explained. A likely guess [7] is that the problem results from using the group additivity method to estimate the solvation enthalpies of the polar groups [8]. The group additivity method is known today to be unsatisfactory for solving this problem: see references in [7]. 6. Conflict between values of hydrophobic free energy from gas–liquid transfer and from studies of large-to-small mutations Beginning in the late 1980s, several workers independently undertook experimental tests of the hydrophobic free energy values found by liquid–liquid transfer. The test was to make large-to-small mutations (for example, V to A, I to A or L to A mutations) of buried non-polar residues in proteins and to measure the changes in unfolding free energy. The results of these studies showed good agreement between different laboratories and generally confirmed the hydrophobic free energy values found by liquid–liquid transfer: see the recent summary by Pace and coworkers [38]. The basic assumptions of this approach are: (1) the unfolding free energy change resulting from a large-to-small mutation contains a hydrophobic free energy term equivalent to the change in non-polar surface exposed by unfolding and (2) no other energetic variable (such as the van der Waals interactions) changes significantly upon mutation because the protein has a rigid structure. There is also a hidden assumption that hydrophobic free energy has the properties measured by liquid–liquid transfer and therefore is entirely entropic at 25 °C, as measured by liquid–liquid transfer [31]. A direct test of assumption 2 plus the hidden assumption was made in 2002 by Makhatadze and coworkers [39], who used
1065
isothermal titration calorimetry to measure the value of DH, as well as of DG, for unfolding. No change in the DH of unfolding is expected for a large-to-small mutation if assumption 2 and the hidden assumption are correct. Large changes in the DH of unfolding were, however, found [39]. In fact, the changes in DH were 2-fold larger than the changes in DG (see Fig. 5A of Ref. 39). Instead of DG being entirely entropic as expected, the entropy change had the opposite sign to that expected if DG corresponds to a change in hydrophobic free energy. Makhatadze and coworkers concluded [39], reasonably enough, that the dominant energetic change caused by a large-tosmall mutation is an unfavorable enthalpy change caused by a loss of stabilizing van der Waals interactions, and that the unexpected sign of the entropy change is caused by increased conformational freedom of residues neighboring on the mutational site. Assuming this analysis is correct, the conflict between gas–liquid transfer results and values obtained from studies of large-to-small mutations is caused by a false assumption about the mutational studies. 7. Related topics The hydrophobic factor in biology has an extraordinary range; only its role in protein folding is considered here, but its role in membrane studies is equally fundamental. A 1993 review of ‘‘hydrophobic effects’’ by Blokzijl and Engberts [41] has 371 references (!) and the authors comment that the reference list is restricted to recent studies of personal interest. A few of the other actively studied topics concerning the hydophobic factor in protein folding are mentioned next. There is a pairwise hydrophobic interaction between two nonpolar molecules in water that depends on the distance between the molecules. The pairwise interaction should be distinguished from hydrophobic hydration, which is measured by solubilities of individual non-polar molecules. According to experimental data, the pairwise interaction is quite weak compared to the values for hydrophobic hydration in Table 1. The weakness of the pairwise interaction can be illustrated by the dissociation constant for benzene dimers: KD = 0.035 mol fraction (2 molal) at 25 °C [42]. A connection between hydrophobic hydration and the pairwise hydrophobic interaction was found in 2001 by Raschke, Tsai and Levitt [40], who simulated the formation of contact clusters of hydrocarbon molecules in water and compared the results with the decrease in net ASA. When several hydrocarbon molecules are placed in a box of water molecules for a simulation study, the hydrocarbon contact clusters that are formed during simulation satisfy standard tests for being equilibrium clusters [40]. Thus, the association constant KA could be calculated for adding one new hydrocarbon molecule to a contact cluster and DASA, the change in net ASA, could also be calculated. Then DGA = RT ln KA and a plot was made of DGA versus DASA. The authors studied 4 hydrocarbon types, varying in size from methane to benzene. There was considerable scatter in the final plot of DGA versus DASA but, when all the data were included, the DGA values correlated reasonably well with the DASA values and the slope of the correlation line agreed with the 1974 value found by Chothia [26], 24 cal mol1Å2. Thus, hydrophobic free energy could be simulated from first principles by using a standard force field. A basic question posed by finding contact clusters of hydrocarbon molecules in water is: why do the clusters form? A paper by Wu and Prausnitz [9] gave a suggestive answer. They proposed a model for the pairwise hydrophobic interaction between two hydrocarbon molecules in water in which the distance dependence of the interaction depends on the overlap volume of the two solvation shells. Each shell was taken to be one water molecule deep and the hydrophobic free energy that drives the two hydrocarbon molecules together,
1066
R.L. Baldwin / FEBS Letters 587 (2013) 1062–1066
which they calculated from gas–liquid transfer data, is caused by unfavorable water ordering in the solvation shells. The water-ordering effect is expected to depend on molar volume (see above) for sufficiently small-sized spherical molecules, but to change to a dependence on surface area for large-sized spheres [24,28,29], as in the classical relation between liquid surface tension and surface area of an air bubble. This variation with length scale of the nature of hydrophobic free energy has become a fertile field for finding new phenomena. The reader is referred to a review [3] of recent theoretical advances concerning length scales. Remarkably, Ben-Amotz and coworkers [43] have been able to study a length scale phenomenon by a direct spectroscopic method, polarized Raman scattering. Quantum mechanics was used by Chopra and Levitt [44] to analyze the depth of water ordering around a ‘buckyball’: they found that water ordering extends to long range. A long-standing enigma in the study of the hydrophobic factor has been the ‘stacked-plates’ experiment [45] in which a weak, distance-dependent, attractive force is found between two flat nonpolar plates in water. To add to the puzzle, the attractive force is chiefly electrostatic in character, as judged by the effect of adding salt to the solution, which sharply reduces the attractive potential [45]. Despa and Berry [46] have given a plausible explanation in which the electrostatic potential arises from a disturbance of the water dipoles caused by the presence of a non-polar object. The discussion above points out the need for new methods of measuring hydrophobic free energy. Recently Li and Walker [47] have shown that mechanical pulling provides such a method. They used atomic force microscopy to pull out collapsed hydrophobic polymers in aqueous solution and to measure the force required. Their results [47] show interesting correlations with hydrophobic free energy values measured by solubility studies. Acknowledgments I thank B.-K. Lee for many past discussions of issues raised here and George Makhatadze for pointing out relevant data. References [1] Kauzmann, W. (1959) Factors in interpretation of protein denaturation. Adv. Protein Chem. 14, 1–63. [2] Nozaki, Y. and Tanford, C. (1971) The solubility of amino acids and two glycine peptides in aqueous ethanol and dioxane solutions: establishment of a hydrophobicity scale. J. Biol. Chem. 246, 2211–2217. [3] Ashbaugh, H.S. and Pratt, L.R. (2006) Colloquium: scaled particle theory and the length scales of hydrophobicity. Rev. Mod. Phys. 78, 159–178. [4] Ben-Naim, A. and Marcus, Y. (1984) Solvation thermodynamics of nonionic solutes. J. Chem. Phys. 81, 2016–2027. [5] Baldwin, R.L. (2012) Gas–liquid transfer data used to analyze hydrophobic hydration and find the nature of the Kauzmann Tanford hydrophobic factor. Proc. Natl. Acad. Sci. USA 109, 7310–7313. [6] Wolfenden, R. and Lewis Jr., C.A. (1976) A vapor phase analysis of the hydrophobic effect. J. Theor. Biol. 59, 235–321. [7] Baldwin, R.L (2013) Properties of hydrophobic free energy found by gas–liquid transfer. Proc. Natl. Acad. Sci. USA, in press. [8] Makhatadze, G.I. and Privalov, P.L. (1994) Hydration effects in protein unfolding. Biophys. Chem. 51, 291–309. [9] Wu, J. and Prausnitz, J.M. (2008) Pairwise-additive hydrophobic effect for alkanes in water. Proc. Natl. Acad. Sci. USA 105, 9512–9515. [10] Lee, B.-K. (1991) Solvent reorganization contribution to the transfer thermodynamics of small nonpolar molecules. Biopolymers 31, 993–1008. [11] Pollack, G.L. (1991) Why gases dissolve in liquids. Science 251, 1323–1330. [12] McAuliffe, C. (1966) Solubility in water of paraffin, cycloparaffin, olefin, acetylene, cycloolefin, and aromatic hydrocarbons. J. Phys. Chem. 70, 1267– 1275. [13] Jorgensen, W.L., Gao, J. and Ravimohan, C. (1985) Monte Carlo simulations of alkanes in water: hydration numbers and the hydrophobic effect. J. Phys. Chem. 89, 3470–3473.
[14] Pratt, L.R. and Chandler, D. (1977) Theory of the hydrophobic effect. J. Chem. Phys. 67, 3683–3704. [15] Hildebrand, J.H. (1968) A criticism of the term ‘‘hydrophobic bond’’. J. Phys. Chem. 72, 1841–1842. [16] Nemethy, G., Scheraga, H.A. and Kauzmann, W. (1968) Comments on the communication A criticism of the term ‘hydrophobic bond’. J. Phys. Chem. 72, 1842. [17] Frank, H.S. and Evans, M. (1945) Free volume and entropy in condensed systems. III: Entropy in binary liquid mixtures. J. Chem. Phys. 13, 507–532. [18] Kauzmann, W. (1987) Thermodynamics of unfolding. Nature 325, 763–764. [19] Tanford, C. (1980) The Hydrophobic Effect: Formation of Micelles and Biological Membranes, Wiley Publishing Ltd., New York. [20] Tanford, C. (1997) How protein chemists learned about the hydrophobic factor. Protein Sci. 6, 1358–1366. [21] Tanford, C. (1979) Interfacial free energy and the hydrophobic effect. Proc. Natl. Acad. Sci. USA 76, 4175–4176. [22] Lee, B. (1985) The physical origin of the low solubility of nonpolar solutes in water. Biopolymers 24, 813–823. [23] Lazaridis, T. (2001) Solvent size versus cohesive energy as the origin of hydrophobicity. Acc. Chem. Res. 34, 931–937. [24] Ball, P. (2011) More than a bystander. Nature 478, 467–468. [25] Hermann, R.B. (1972) Theory of hydrophobic bonding. II: The correlation of hydrocarbon solubility in water with solvent cavity surface area. J. Phys. Chem. 76, 2754–2759. [26] Chothia, C. (1974) Hydrophobic bonding and accessible surface area in proteins. Nature 248, 338–339. [27] Reynolds, J.A., Gilbert, D.B. and Tanford, C. (1974) Empirical correlation between hydrophobic free energy and aqueous cavity surface area. Proc. Natl. Acad. Sci. USA 71, 2925–2927. [28] Lum, K., Chandler, D. and Weeks, J.D. (1999) Hydrophobicity at small and large length scales. J. Phys. Chem. B 103, 4570–4577. [29] Chandler, D. (2005) Interfaces and the driving force of hydrophobic assembly. Nature 43, 640–647. [30] Dill, K.A. (1990) Dominant forces in protein folding. Biochemistry 29, 7133– 7155. [31] Baldwin, R.L. (1986) Temperature dependence of the hydrophobic interaction in protein folding. Proc. Natl. Acad. Sci. USA 83, 8069–8072. [32] Privalov, P.L. (1979) Stability of proteins. Small globular proteins. Adv. Protein Chem. 33, 167–241. [33] Auton, M. and Bolen, D.W. (2005) Predicting the energetics of osmolyteinduced protein folding/unfolding. Proc. Natl. Acad. Sci. USA 102, 15065– 15068. [34] Holthauzen, L.M.F., Rösgen, J. and Bolen, D.W. (2010) Hydrogen bonding progressively strengthens upon transfer of the protein urea denatured state to water and protecting osmolytes. Biochemistry 49, 1310–1318. [35] Robertson, A.D. and Murphy, K.P. (1997) Protein structure and the energetics of protein stability. Chem. Rev. 97, 1251–1267. [36] Richardson, J.M. and Makhatadze, G.I. (2003) Temperature dependence of the thermodynamics of helix–coil transition. J. Mol. Biol. 335, 1029–1037. [37] Makhatadze, G.I. and Privalo, P.L. (1994) Energetics of interactions of aromatic hydrocarbons with water. Biophys. Chem. 50, 285–291. [38] Pace, C.N., Fu, H., Fryar, K.L., Landua, J., Trevino, S.R., Shirley, B.A., Hendricks, M.M., Iimura, S., Gajwala, K., Scholtz, J.M. and Grimsley, G.R. (2011) Contribution of hydrophobic interactions to protein stability. J. Mol. Biol. 408, 514–528. [39] Loladze, V.V., Ermolenko, D.N. and Makhatadze, G.I. (2002) Thermodynamic consequences of burial of polar and nonpolar amino acid residues in the protein interior. J. Mol. Biol. 320, 343–357. [40] Raschke, T.M., Tsai, J. and Levitt, M. (2001) Quantification of the hydrophobic interaction by simulations of the aggregation of small hydrophobic solutes in water. Proc. Natl. Acad. Sci. USA 98, 5965–5969. [41] Blokzijl, W. and Engberts, J.B.F.N. (1993) Hydrophobic effects. Opinions and facts. Angew. Chem., Int. Ed. Engl. 32, 1545–1579. [42] Tucker, E.E., Lane, E.H. and Christian, S.D. (1981) Vapor pressure studies of hydrophobic interactions. Formation of benzene–benzene and cyclohexane– cyclohexanol dimers in dilute aqueous solution. J. Solution Chem. 10, 1– 20. [43] Davis, J.G., Gierszal, K.P., Wang, P. and Ben-Amotz, D. (2012) Water structural transformation at molecular hydrophobic interfaces. Nature 491, 582– 585. [44] Chopra, G. and Levitt, M. (2011) Remarkable patterns of surface water ordering around polarized buckminsterfullerene. Proc. Natl. Acad. Sci. USA 108, 14455–14460. [45] Meyer, E.E., Rosenberg, K.J. and Israelachvili, J. (2006) Recent progress in understanding hydrophobic interactions. Proc. Natl. Acad. Sci. USA 103, 15739–15746. [46] Despa, F. and Berry, R.S. (2007) The origin of long-range attraction between hydrophobes in water. Biophys. J. 92, 373–378. [47] Li, I.T.S. and Walker, G.C. (2011) Signature of hydrophobic hydration in a single polymer. Proc. Natl. Acad. Sci. USA 108, 16527–16532.
journal homepage: www.FEBSLetters.org
Review
The new view of hydrophobic free energy Robert L. Baldwin Biochemistry Department, Beckman Center, Stanford University Medical Center, Stanford, CA 94305, USA
a r t i c l e
i n f o
Article history: Received 19 November 2012 Revised 7 January 2013 Accepted 8 January 2013 Available online 18 January 2013 Edited by Wilhelm Just Keywords: Transfer free energy Reference solvent Liquid alkane Folding analogy
a b s t r a c t In the new view, hydrophobic free energy is measured by the work of solute transfer of hydrocarbon gases from vapor to aqueous solution. Reasons are given for believing that older values, measured by solute transfer from a reference solvent to water, are not quantitatively correct. The hydrophobic free energy from gas–liquid transfer is the sum of two opposing quantities, the cavity work (unfavorable) and the solute–solvent interaction energy (favorable). Values of the interaction energy have been found by simulation for linear alkanes and are used here to find the cavity work, which scales linearly with molar volume, not accessible surface area. The hydrophobic free energy is the dominant factor driving folding as judged by the heat capacity change for transfer, which agrees with values for solvating hydrocarbon gases. There is an apparent conflict with earlier values of hydrophobic free energy from studies of large-to-small mutations and an explanation is given. Ó 2013 Federation of European Biochemical Societies. Published by Elsevier B.V. All rights reserved.
Ever since the pioneering work by Kauzmann [1] in 1959 and by Nozaki and Tanford [2] in 1971, hydrophobic free energy (DGh) has been determined from values of the transfer free energy when nonpolar solutes are transferred from water to non-polar (or semi-polar) reference solvents. In recent years the standard choice of a reference solvent has been a liquid alkane [3]: the liquid alkane provides both the alkane solute and the non-polar reference solvent. In 1984 a basic criticism of this method was made by BenNaim and Marcus [4]: the solvation free energy of a non-polar solute is determined from gas–liquid, not liquid–liquid, transfer. When an alkane solute is transferred from water through vapor to the liquid alkane, the liquid–liquid transfer may be written as two successive gas–liquid transfers [5]: first from water into vapor and then from vapor into liquid alkane. Kauzmann [1] and Tanford [2] expected that removing the hydrocarbon solute from water would account for most of the free energy change in liquid–liquid transfer but this expectation was not correct: data are available for the DG values of both gas–liquid transfers, which show that the ‘‘hydrophobic’’ transfer from water to vapor accounts for less than half of the total DG (5–7). This disconcerting fact was realized as early as 1976 when Wolfenden and Lewis [6] studied why water is a poor solvent for liquid hydrocarbons. They found that a strong favorable interaction among alkane molecules in liquid alkanes gives a strongly favorable transfer free energy for passage of an alkane solute from vapor into liquid alkane. Kauzmann [1] made an analogy between the protein folding process and the transfer of a non-polar solute from water into a
E-mail address: [email protected]
reference solvent, since folding transfers the non-polar side chains of an unfolded protein out of water into a non-aqueous environment, the protein interior. It was well-known that water is a poor solvent for hydrocarbons, and Kauzmann showed that the DG for transferring a hydrocarbon out of water into a liquid hydrocarbon is substantial compared to the net DG for folding a small protein. For example, the DG for transfer from water to liquid alkane is 4.8 kcal/mol for butane, a hydrocarbon which has four carbon atoms like the side chains of leucine and isoleucine. Thus, he argued, a similar DG should help to fold a protein when a non-polar side chain is buried by folding. The analogy breaks down, however, when the liquid–liquid transfer is divided into two gas–liquid transfers [5], because the ‘‘hydrophobic’’ transfer from water into vapor has less than half of the total DG. and the second transfer, from vapor to liquid alkane, is not obviously related to the protein folding process. For this and the following reasons, I argued recently [7] that hydrophobic free energy should be defined and measured by the ‘‘hydrophobic’’ transfer of alkane solutes from water into vapor. The new definition is consistent with the criticism of Ben-Naim and Marcus [4] that the solvation free energy of a non-polar solute must be measured by gas–liquid transfer. The new definition is also consistent with work by Makhatadze and Privalov [8] and Wu and Prausnitz [9], who used gas–liquid transfer data to solve folding problems involving the hydrophobic factor. The new definition of DGh accepts part of Kauzmann’s analogy between the folding process and solute transfer but stops after the first transfer from water into vapor. The second half of the analogy, which involves solute transfer from vapor into liquid alkane, is omitted because the transfer of a non-polar side chain into
0014-5793/$36.00 Ó 2013 Federation of European Biochemical Societies. Published by Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.febslet.2013.01.006
R.L. Baldwin / FEBS Letters 587 (2013) 1062–1066
the protein interior has different properties [7] from those of the transfer of an alkane solute from vapor into liquid alkane. These other properties include the topological connection of the non-polar side chain to the polypeptide chain, the presence of semipolar, H-bonded-NHCO– groups in the protein interior, and close packing of the groups inside a folded protein. 1. Hydrophobic free energy is the sum of two opposing quantities The gas–liquid transfer free energy, DGGL, which corresponds to the new DGh [7], is the sum of DGc (the cavity work) and Ea (the solute–solvent interaction energy), according to independent derivations of Eq. (1) (below) by Lee [10] and Pollack [11]. These authors analyzed the solute insertion model for equilibrium of a non-polar solute between vapor and liquid phases. In this model a cavity is first made for the solute in the liquid and then the solute is inserted; DGc is the work needed to make the cavity and Ea is the interaction energy between solute and solvent after insertion. Both derivations [10,11] start from the equality of the solute’s chemical potential between the gas and liquid phases. Lee [10] gave a detailed derivation, based on statistical mechanics, of Eq. (1) that includes the solvent reorganization after solute insertion, and he shows that this term is minor and can be dropped. He gives evidence [10], however, for important solvent reorganization as the cavity is made, but not after solute insertion
DGGL ¼ DGc þ Ea
ð1Þ
The transfer free energy is measured from the Ostwald coefficient L, which is the equilibrium ratio of the solute concentration (molar or number density concentration) in the liquid over the gas phase.
DGGL ¼ RT lnL
ð2aÞ
L ¼ ðM L Þ=ðM G Þ
ð2bÞ
Values of the gas–liquid transfer free energies of linear alkanes are given in [4], based on McAuliffe’s 1966 study [12] of the solubilities of gaseous hydrocarbons in water at 25 °C. McAuliffe used a gas chromatograph to determine hydrocarbon solubility and he found that separating impurities away from the test hydrocarbon was the key to getting accurate results. He maintained a pressure of 1 atmosphere gas while equilibrating the hydrocarbon gas with water. The ideal gas law was used [4,10] to give the molar concentration of the gas from its partial pressure, after subtracting the partial pressure of water, and the chromatograph results gave the molar concentration of the hydrocarbon in aqueous solution. Table 1 gives the values in water at 25 °C of DGc, Ea, DGGL, and DGLL for 4 linear alkanes, ethane through pentane. The values show
Table 1 Transfer energetics for linear alkanes.a Solute
Eab
DGcc
DGGLd
DGLLe
Ethane Propane Butane Pentane
4.80 6.60 8.29 9.99
6.57 8.57 10.4 12.3
1.77 1.98 2.15 2.34
3.14 3.92 4.79 5.74
a All values are in kcal/mol at 25 °C. Ea values are from [13], DGGL values are from Table V of [4], DGc values are found from Eq. (1) using DGGL and Ea values given here, DGLL values are found from Tables 1 and V of [4] (see [5]). DGGL is the sum of Ea and DGc. b The solute–solvent interaction energy from simulation [13]. c The cavity work. d The transfer free energy from vapor to water, which equals the hydrophobic free energy. e The transfer free energy from liquid alkane to water.
1063
that the liquid–liquid transfer energy DGLL is more than twofold larger than the gas–liquid transfer energy DGGL. The values of Ea in water are given by Jorgensen and coworkers [13]. They obtained them by exhaustive simulations of alkane conformations in water, using the TIP4P model of water and the Lennard–Jones potential for the interaction between water and alkane. Lee [10] recognized that, when Ea values are available from simulation, DGc values can be found from equation (1) by using experimental values of DGGL. An important result in Table 1 is that DGc and Ea are opposing factors in DGh and their difference determines the size of DGh. It has not always been recognized that opposing factors determine hydrophobic free energy (see below), although this is clear in the early work of Pratt and Chandler [14] and in their subsequent papers. A failure to recognize the roles of two different factors has sometimes led to misinterpretation of results. Kauzmann [1] originally spoke of ‘‘hydrophobic bonds’’. In 1968, Hildebrand [15] sharply criticized this term because he and coworkers had shown that hydrocarbons interact favorably (and strongly) with water. An oil drop added to a container of water does not remain a drop, rather it spreads out to form a film that interacts with as much water as possible. Kauzmann, Nemethy and Scheraga [16] replied to Hildebrand by agreeing that there are favorable van der Waals interactions between hydrocarbon and water, but they argued that the entropy-driven character of the hydrophobic interaction (at 25 °C) is best explained by ordering of water molecules around the non-polar groups [1,17,18]. Tanford did not reply to Hildebrand immediately but he was careful afterwards to refer to the ‘‘hydrophobic effect’’ [19] and later to the ‘‘hydrophobic factor’’ [20], not to ‘‘hydrophobic bonds’’. Eleven years later, he replied [21] to Hildebrand by showing that the cohesive energy of water is large compared either to that of hexane or to the adhesive energy of water and hexane. He argued that the hydrophobic effect arises from the large cohesive energy of water, which tends to squeeze out all solutes. Polar solutes resist by making strong interactions with water while hydrocarbons respond by having very low solubilities in water. There is a second factor besides cohesive energy that contributes to the low solubilities of hydrocarbons in water, which is the unusually small size of the water molecule [22]; the relative merits of the two factors are still being debated [23]. When values of DGc became available in 1991 for 3 alkane solutes [10], it became clear that the low solubilities of gaseous alkane molecules in water are caused by large values of the cavity work. The low solubilities of liquid alkanes in water are caused additionally [6] by large values of Ea for alkane solutes in liquid alkanes.
2. The cavity work scales with molar volume The correlation of hydrophobic free energy with accessible surface area (ASA) or molar volume can be analyzed reliably for those hydrocarbons whose values are known of DGc and Ea. There has been a controversy [24] about how DGLL from liquid–liquid transfer scales with molecular size. It has been widely believed that hydrophobic free energy correlates with surface area [25–27] of the solute. Chandler [28,29] argued persuasively, however, that the cavity work in water is not proportional to surface area in the small-size regime (see below) but instead scales linearly with molar volume. Until now, the controversy has been based on data for liquid–liquid transfer, which is more complex than gas–liquid transfer because DGLL depends on the values of DGc and Ea in both liquid alkane and water [5]. Gas–liquid transfer results from Table 1 are used in Fig. 1 to plot DGc (water) against molar volume (V) for 4 linear alkanes. McAuliffe [12] showed that a plot of this kind should be limited to hydrocarbons of a single type, such as linear or cyclic, because
1064
R.L. Baldwin / FEBS Letters 587 (2013) 1062–1066
water near room temperature and become more soluble at both higher and lower temperatures. Note, however, that the minimum solubility of liquid alkanes in water occurs at Th, the temperature at which the transfer enthalpy passes through 0, and Th changes from 25 °C for liquid–liquid transfer [31] to 95 °C for gas–liquid transfer [37]. Both Kauzmann [1,18] and Dill [30] emphasized the striking and similar properties of DCp (the change in heat capacity) when proteins unfold and when non-polar liquids are dissolved in water. The basic evidence that DCp has a large value in protein unfolding came as early as 1979 from Privalov [32] who measured calorimetrically the DH of protein unfolding at different temperatures, and obtained DCp from the plot of DH versus T. He varied pH in the range pH 2–5, where protein stability depends critically on pH and DH can be measured at varying temperatures; DCp is almost independent of pH in this range. The DH and DS values for protein unfolding at moderate temperatures (T) are fitted well by Eqs. (3) and (4) with DCp constant, independent of T; Tr is any convenient reference temperature. Fig. 1. A plot of the cavity work (DGc) versus molar volume (V) for the 4 linear alkanes in Table 1, and the data points for isobutane and neopentane from Lee’s study [10] have been added. All values of V are from McAuliffe [12]. The point (0,0) has also been added.
different types of hydrocarbons give different lines, even though the various types all scale linearly with molar volume. Since each type gives a different line, hydrocarbon solubility in water depends on more variables besides molar volume. Fig. 1 gives a good straight line, indicating that DGc is proportional to molar volume for linear alkanes. as expected both from McAuliffe’s 1966 results [12] and Chandler’s more recent work [28,29]. The solute–solvent interaction energy is expected to be proportional to surface area [13,27] but the linear alkanes discussed here are flexible molecules and average ASA values found by exhaustive simulation of all possible conformations should be used to probe this question; the simulated values are not presently available. It is interesting to consider today why hydrophobic free energy was expected in early proposals [25–27] to be proportional to surface area. It seems likely that a main reason was the perceived importance of the ordering of water structure around non-polar groups. Another reason [27] was the short-range distance dependence of the attractive Lennard–Jones term. Note, however, that the water-ordering effect (which is discussed below) has been used to explain the water-hating behavior of non-polar groups [1,17,18,20] whereas the solute–solvent interaction (Ea) between water and hydrocarbon groups is favorable [15]. In 1974 Tanford and coworkers [27] said ‘‘. . .the hydrophobic free energy arises from contacts between hydrocarbon and water at the solute–solvent interface and would thus be expected to be a function of the surface area. . .’’, although Tanford later contradicted this view [21]. 3. Hydrophobic free energy considered as the dominant factor driving protein folding Dill concluded in a widely cited 1990 review [30] that hydrophobic free energy is the dominant factor driving protein folding. His argument emphasizes similarities between the transfer energetics when non-polar liquids are dissolved in water and when proteins unfold, since the unfolding process transfers non-polar side chains from the protein interior to water. Dill’s argument is based especially on the unusual temperature dependence of protein stability and similarities with dissolving non-polar liquids in water. His review points out that some proteins can be denatured both by heating and cooling while non-polar liquids have minimum solubility in
DH ¼ DHr þ DCpðT T r Þ
ð3Þ
DS ¼ DSr þ DCp lnðT=T r Þ
ð4Þ
4. Quantitative comparisons between hydrophobic free energy and protein folding data Two quantitative comparisons between protein folding energetics and properties of hydrophobic free energy are discussed in [7] and reviewed briefly below. The first is a comparison between the DCp values found for protein unfolding and for transfer of hydrocarbons from water into the gas phase. The second is a test of the ability of gas–liquid transfer data to predict the temperature dependence of the enthalpy of protein unfolding. Bolen’s work [33,34] shows that osmolytes drive protein folding by acting chiefly on the peptide backbone, which suggests that osmolytes affect the formation of peptide H-bonds. Changes in hydrophobic free energy occur mainly in processes involving the non-polar groups of protein side chains, not the polar groups in the peptide backbone, and Bolen’s work raises the question of whether DGh is in fact the dominant factor in protein folding. To answer this question, the following approach was taken [7]. The free energy change that drives folding was taken from heat-induced unfolding since free energy is a state function. Then the temperature dependence of the free energy change, which is characterized by DCp, was used to find its source. An accurate value for heat-induced unfolding, (DCp/ n) = 13.9 ± 0.5 cal K1 mol1, was given by Robertson and Murphy [35], who surveyed the thermodynamic properties of 49 protein unfolding reactions. Their survey shows that DCp for protein unfolding is proportional to n, the number of residues in a protein. The protein DCp value has a positive sign, whereas Richardson and Makhatadze [36] found that DCp for unfolding a peptide helix and breaking peptide H-bonds is 7.6 cal K1 mol res1 with a negative sign, which rules out peptide H-bonds as the dominant source of the free energy that drives folding. To compare the protein DCp value with ones for the transfer of hydrocarbons from water into the gas phase, values of (DCp/ASA) can be compared [7]. Values of (DCp/ASA) for gas–liquid transfer of hydrocarbon gases were given by Makhatadze and Privalov [37] as follows: ethane 0.402, propane 0.395, and butane 0.410 cal K1 mol1 Å2. To find the protein unfolding value of (DCp/ASA), the (DCp/n) value above can be divided by 42 Å2, which is the average amount of non-polar surface area buried per protein residue [7]. The result is (DCp/ASA) = 0.33 cal K1 mol1 Å2 [7],
R.L. Baldwin / FEBS Letters 587 (2013) 1062–1066
which is close to the value for transferring hydrocarbons from water into the gas phase. This comparison strongly suggests that DGh is the dominant source of the free energy that drives folding. The likely reason [7] why osmolyte studies appear to give a different answer is that osmolyte-driven folding/unfolding is a different process from thermal folding/unfolding, since the osmolyte participates in the osmolyte-driven process. 5. Temperature dependence of protein stability To find out if gas–liquid transfer results correctly predict the temperature dependence of the enthalpy of protein unfolding, the following test was made [7]. Experimental values are available [8] both of the solvation enthalpies of protein non-polar and polar groups (for nine proteins, at temperatures from 5 to 100 °C) and of the corresponding enthalpy values for protein unfolding. The results for ribonuclease T1 were used in [7] to test whether gas–liquid transfer results, when used to estimate the solvation enthalpies of the polar and non-polar groups, correctly predict the large temperature dependence of the protein unfolding enthalpy. In making this test, the basic assumptions are: (1) the solvation enthalpies of the polar and non-polar groups have been estimated correctly, and (2) no other energetic variable (such as the van der Waals interactions) changes enough with temperature to affect seriously the DH of unfolding. If there is complete agreement between predicted and observed results in this test, the data points should fall on a straight line that passes through (0,0) and has slope 1.00. A good straight line was found [7] that passes through 0.0 and has slope 1.25. When the contribution from the polar peptide groups was omitted, the slope increased to 1.87, indicating that the effect of the polar peptide groups must be taken into account. Solvation of the non-polar groups upon unfolding is, however, the major cause of the strong temperature dependence of the enthalpy of protein unfolding [7]. The failure to obtain complete agreement (slope 1.25 observed versus 1.00 expected) remainns to be explained. A likely guess [7] is that the problem results from using the group additivity method to estimate the solvation enthalpies of the polar groups [8]. The group additivity method is known today to be unsatisfactory for solving this problem: see references in [7]. 6. Conflict between values of hydrophobic free energy from gas–liquid transfer and from studies of large-to-small mutations Beginning in the late 1980s, several workers independently undertook experimental tests of the hydrophobic free energy values found by liquid–liquid transfer. The test was to make large-to-small mutations (for example, V to A, I to A or L to A mutations) of buried non-polar residues in proteins and to measure the changes in unfolding free energy. The results of these studies showed good agreement between different laboratories and generally confirmed the hydrophobic free energy values found by liquid–liquid transfer: see the recent summary by Pace and coworkers [38]. The basic assumptions of this approach are: (1) the unfolding free energy change resulting from a large-to-small mutation contains a hydrophobic free energy term equivalent to the change in non-polar surface exposed by unfolding and (2) no other energetic variable (such as the van der Waals interactions) changes significantly upon mutation because the protein has a rigid structure. There is also a hidden assumption that hydrophobic free energy has the properties measured by liquid–liquid transfer and therefore is entirely entropic at 25 °C, as measured by liquid–liquid transfer [31]. A direct test of assumption 2 plus the hidden assumption was made in 2002 by Makhatadze and coworkers [39], who used
1065
isothermal titration calorimetry to measure the value of DH, as well as of DG, for unfolding. No change in the DH of unfolding is expected for a large-to-small mutation if assumption 2 and the hidden assumption are correct. Large changes in the DH of unfolding were, however, found [39]. In fact, the changes in DH were 2-fold larger than the changes in DG (see Fig. 5A of Ref. 39). Instead of DG being entirely entropic as expected, the entropy change had the opposite sign to that expected if DG corresponds to a change in hydrophobic free energy. Makhatadze and coworkers concluded [39], reasonably enough, that the dominant energetic change caused by a large-tosmall mutation is an unfavorable enthalpy change caused by a loss of stabilizing van der Waals interactions, and that the unexpected sign of the entropy change is caused by increased conformational freedom of residues neighboring on the mutational site. Assuming this analysis is correct, the conflict between gas–liquid transfer results and values obtained from studies of large-to-small mutations is caused by a false assumption about the mutational studies. 7. Related topics The hydrophobic factor in biology has an extraordinary range; only its role in protein folding is considered here, but its role in membrane studies is equally fundamental. A 1993 review of ‘‘hydrophobic effects’’ by Blokzijl and Engberts [41] has 371 references (!) and the authors comment that the reference list is restricted to recent studies of personal interest. A few of the other actively studied topics concerning the hydophobic factor in protein folding are mentioned next. There is a pairwise hydrophobic interaction between two nonpolar molecules in water that depends on the distance between the molecules. The pairwise interaction should be distinguished from hydrophobic hydration, which is measured by solubilities of individual non-polar molecules. According to experimental data, the pairwise interaction is quite weak compared to the values for hydrophobic hydration in Table 1. The weakness of the pairwise interaction can be illustrated by the dissociation constant for benzene dimers: KD = 0.035 mol fraction (2 molal) at 25 °C [42]. A connection between hydrophobic hydration and the pairwise hydrophobic interaction was found in 2001 by Raschke, Tsai and Levitt [40], who simulated the formation of contact clusters of hydrocarbon molecules in water and compared the results with the decrease in net ASA. When several hydrocarbon molecules are placed in a box of water molecules for a simulation study, the hydrocarbon contact clusters that are formed during simulation satisfy standard tests for being equilibrium clusters [40]. Thus, the association constant KA could be calculated for adding one new hydrocarbon molecule to a contact cluster and DASA, the change in net ASA, could also be calculated. Then DGA = RT ln KA and a plot was made of DGA versus DASA. The authors studied 4 hydrocarbon types, varying in size from methane to benzene. There was considerable scatter in the final plot of DGA versus DASA but, when all the data were included, the DGA values correlated reasonably well with the DASA values and the slope of the correlation line agreed with the 1974 value found by Chothia [26], 24 cal mol1Å2. Thus, hydrophobic free energy could be simulated from first principles by using a standard force field. A basic question posed by finding contact clusters of hydrocarbon molecules in water is: why do the clusters form? A paper by Wu and Prausnitz [9] gave a suggestive answer. They proposed a model for the pairwise hydrophobic interaction between two hydrocarbon molecules in water in which the distance dependence of the interaction depends on the overlap volume of the two solvation shells. Each shell was taken to be one water molecule deep and the hydrophobic free energy that drives the two hydrocarbon molecules together,
1066
R.L. Baldwin / FEBS Letters 587 (2013) 1062–1066
which they calculated from gas–liquid transfer data, is caused by unfavorable water ordering in the solvation shells. The water-ordering effect is expected to depend on molar volume (see above) for sufficiently small-sized spherical molecules, but to change to a dependence on surface area for large-sized spheres [24,28,29], as in the classical relation between liquid surface tension and surface area of an air bubble. This variation with length scale of the nature of hydrophobic free energy has become a fertile field for finding new phenomena. The reader is referred to a review [3] of recent theoretical advances concerning length scales. Remarkably, Ben-Amotz and coworkers [43] have been able to study a length scale phenomenon by a direct spectroscopic method, polarized Raman scattering. Quantum mechanics was used by Chopra and Levitt [44] to analyze the depth of water ordering around a ‘buckyball’: they found that water ordering extends to long range. A long-standing enigma in the study of the hydrophobic factor has been the ‘stacked-plates’ experiment [45] in which a weak, distance-dependent, attractive force is found between two flat nonpolar plates in water. To add to the puzzle, the attractive force is chiefly electrostatic in character, as judged by the effect of adding salt to the solution, which sharply reduces the attractive potential [45]. Despa and Berry [46] have given a plausible explanation in which the electrostatic potential arises from a disturbance of the water dipoles caused by the presence of a non-polar object. The discussion above points out the need for new methods of measuring hydrophobic free energy. Recently Li and Walker [47] have shown that mechanical pulling provides such a method. They used atomic force microscopy to pull out collapsed hydrophobic polymers in aqueous solution and to measure the force required. Their results [47] show interesting correlations with hydrophobic free energy values measured by solubility studies. Acknowledgments I thank B.-K. Lee for many past discussions of issues raised here and George Makhatadze for pointing out relevant data. References [1] Kauzmann, W. (1959) Factors in interpretation of protein denaturation. Adv. Protein Chem. 14, 1–63. [2] Nozaki, Y. and Tanford, C. (1971) The solubility of amino acids and two glycine peptides in aqueous ethanol and dioxane solutions: establishment of a hydrophobicity scale. J. Biol. Chem. 246, 2211–2217. [3] Ashbaugh, H.S. and Pratt, L.R. (2006) Colloquium: scaled particle theory and the length scales of hydrophobicity. Rev. Mod. Phys. 78, 159–178. [4] Ben-Naim, A. and Marcus, Y. (1984) Solvation thermodynamics of nonionic solutes. J. Chem. Phys. 81, 2016–2027. [5] Baldwin, R.L. (2012) Gas–liquid transfer data used to analyze hydrophobic hydration and find the nature of the Kauzmann Tanford hydrophobic factor. Proc. Natl. Acad. Sci. USA 109, 7310–7313. [6] Wolfenden, R. and Lewis Jr., C.A. (1976) A vapor phase analysis of the hydrophobic effect. J. Theor. Biol. 59, 235–321. [7] Baldwin, R.L (2013) Properties of hydrophobic free energy found by gas–liquid transfer. Proc. Natl. Acad. Sci. USA, in press. [8] Makhatadze, G.I. and Privalov, P.L. (1994) Hydration effects in protein unfolding. Biophys. Chem. 51, 291–309. [9] Wu, J. and Prausnitz, J.M. (2008) Pairwise-additive hydrophobic effect for alkanes in water. Proc. Natl. Acad. Sci. USA 105, 9512–9515. [10] Lee, B.-K. (1991) Solvent reorganization contribution to the transfer thermodynamics of small nonpolar molecules. Biopolymers 31, 993–1008. [11] Pollack, G.L. (1991) Why gases dissolve in liquids. Science 251, 1323–1330. [12] McAuliffe, C. (1966) Solubility in water of paraffin, cycloparaffin, olefin, acetylene, cycloolefin, and aromatic hydrocarbons. J. Phys. Chem. 70, 1267– 1275. [13] Jorgensen, W.L., Gao, J. and Ravimohan, C. (1985) Monte Carlo simulations of alkanes in water: hydration numbers and the hydrophobic effect. J. Phys. Chem. 89, 3470–3473.
[14] Pratt, L.R. and Chandler, D. (1977) Theory of the hydrophobic effect. J. Chem. Phys. 67, 3683–3704. [15] Hildebrand, J.H. (1968) A criticism of the term ‘‘hydrophobic bond’’. J. Phys. Chem. 72, 1841–1842. [16] Nemethy, G., Scheraga, H.A. and Kauzmann, W. (1968) Comments on the communication A criticism of the term ‘hydrophobic bond’. J. Phys. Chem. 72, 1842. [17] Frank, H.S. and Evans, M. (1945) Free volume and entropy in condensed systems. III: Entropy in binary liquid mixtures. J. Chem. Phys. 13, 507–532. [18] Kauzmann, W. (1987) Thermodynamics of unfolding. Nature 325, 763–764. [19] Tanford, C. (1980) The Hydrophobic Effect: Formation of Micelles and Biological Membranes, Wiley Publishing Ltd., New York. [20] Tanford, C. (1997) How protein chemists learned about the hydrophobic factor. Protein Sci. 6, 1358–1366. [21] Tanford, C. (1979) Interfacial free energy and the hydrophobic effect. Proc. Natl. Acad. Sci. USA 76, 4175–4176. [22] Lee, B. (1985) The physical origin of the low solubility of nonpolar solutes in water. Biopolymers 24, 813–823. [23] Lazaridis, T. (2001) Solvent size versus cohesive energy as the origin of hydrophobicity. Acc. Chem. Res. 34, 931–937. [24] Ball, P. (2011) More than a bystander. Nature 478, 467–468. [25] Hermann, R.B. (1972) Theory of hydrophobic bonding. II: The correlation of hydrocarbon solubility in water with solvent cavity surface area. J. Phys. Chem. 76, 2754–2759. [26] Chothia, C. (1974) Hydrophobic bonding and accessible surface area in proteins. Nature 248, 338–339. [27] Reynolds, J.A., Gilbert, D.B. and Tanford, C. (1974) Empirical correlation between hydrophobic free energy and aqueous cavity surface area. Proc. Natl. Acad. Sci. USA 71, 2925–2927. [28] Lum, K., Chandler, D. and Weeks, J.D. (1999) Hydrophobicity at small and large length scales. J. Phys. Chem. B 103, 4570–4577. [29] Chandler, D. (2005) Interfaces and the driving force of hydrophobic assembly. Nature 43, 640–647. [30] Dill, K.A. (1990) Dominant forces in protein folding. Biochemistry 29, 7133– 7155. [31] Baldwin, R.L. (1986) Temperature dependence of the hydrophobic interaction in protein folding. Proc. Natl. Acad. Sci. USA 83, 8069–8072. [32] Privalov, P.L. (1979) Stability of proteins. Small globular proteins. Adv. Protein Chem. 33, 167–241. [33] Auton, M. and Bolen, D.W. (2005) Predicting the energetics of osmolyteinduced protein folding/unfolding. Proc. Natl. Acad. Sci. USA 102, 15065– 15068. [34] Holthauzen, L.M.F., Rösgen, J. and Bolen, D.W. (2010) Hydrogen bonding progressively strengthens upon transfer of the protein urea denatured state to water and protecting osmolytes. Biochemistry 49, 1310–1318. [35] Robertson, A.D. and Murphy, K.P. (1997) Protein structure and the energetics of protein stability. Chem. Rev. 97, 1251–1267. [36] Richardson, J.M. and Makhatadze, G.I. (2003) Temperature dependence of the thermodynamics of helix–coil transition. J. Mol. Biol. 335, 1029–1037. [37] Makhatadze, G.I. and Privalo, P.L. (1994) Energetics of interactions of aromatic hydrocarbons with water. Biophys. Chem. 50, 285–291. [38] Pace, C.N., Fu, H., Fryar, K.L., Landua, J., Trevino, S.R., Shirley, B.A., Hendricks, M.M., Iimura, S., Gajwala, K., Scholtz, J.M. and Grimsley, G.R. (2011) Contribution of hydrophobic interactions to protein stability. J. Mol. Biol. 408, 514–528. [39] Loladze, V.V., Ermolenko, D.N. and Makhatadze, G.I. (2002) Thermodynamic consequences of burial of polar and nonpolar amino acid residues in the protein interior. J. Mol. Biol. 320, 343–357. [40] Raschke, T.M., Tsai, J. and Levitt, M. (2001) Quantification of the hydrophobic interaction by simulations of the aggregation of small hydrophobic solutes in water. Proc. Natl. Acad. Sci. USA 98, 5965–5969. [41] Blokzijl, W. and Engberts, J.B.F.N. (1993) Hydrophobic effects. Opinions and facts. Angew. Chem., Int. Ed. Engl. 32, 1545–1579. [42] Tucker, E.E., Lane, E.H. and Christian, S.D. (1981) Vapor pressure studies of hydrophobic interactions. Formation of benzene–benzene and cyclohexane– cyclohexanol dimers in dilute aqueous solution. J. Solution Chem. 10, 1– 20. [43] Davis, J.G., Gierszal, K.P., Wang, P. and Ben-Amotz, D. (2012) Water structural transformation at molecular hydrophobic interfaces. Nature 491, 582– 585. [44] Chopra, G. and Levitt, M. (2011) Remarkable patterns of surface water ordering around polarized buckminsterfullerene. Proc. Natl. Acad. Sci. USA 108, 14455–14460. [45] Meyer, E.E., Rosenberg, K.J. and Israelachvili, J. (2006) Recent progress in understanding hydrophobic interactions. Proc. Natl. Acad. Sci. USA 103, 15739–15746. [46] Despa, F. and Berry, R.S. (2007) The origin of long-range attraction between hydrophobes in water. Biophys. J. 92, 373–378. [47] Li, I.T.S. and Walker, G.C. (2011) Signature of hydrophobic hydration in a single polymer. Proc. Natl. Acad. Sci. USA 108, 16527–16532.