Heat capacity effects in protein folding and ligand binding: a re-evaluation of the role of water in biomolecular thermodynamics

https://doi.org/10.1016/j.bpc.2004.12.011Get rights and content

Abstract

Large “anomalous” heat capacityCp) effects are a common feature of the thermodynamics of biomolecular interactions in aqueous solution and, as a result of the improved facility for direct calorimetric measurements, there is a growing body of experimental data for such effects in protein folding, protein–protein and protein–ligand interactions. Conventionally such heat capacity effects have been ascribed to hydrophobic interactions, and there are some remarkably convincing demonstrations of the usefulness of this concept. Nonetheless, there is also increasing evidence that hydrophobic interactions are not the only possible source of such effects. Here we re-evaluate the possible contributions of other interactions to the heat capacity changes to be expected for cooperative biomolecular folding and binding processes, with particular reference to the role of hydrogen bonding and solvent water interactions. Simple models based on the hydrogen-bonding propensity of water as a function of temperature give quantitative estimates of ΔCp that compare well with experimental observations for both protein folding and ligand binding. The thermodynamic contribution of bound waters in protein complexes is also estimated. The prediction from simple lattice models is that trapping of water in a complex should give more exothermic binding (ΔΔH−6 to −12 kJ mol−1) with lower entropy (ΔΔS0≈−11 J mol−1 K−1) and more negative ΔCp (by about −75 J mol−1 K−1) per water molecule. More generally, it is clear that significant ΔCp effects are to be expected for any macromolecular process involving a multiplicity of cooperative weak interactions of whatever kind.

Introduction

The thermodynamics of biomolecular processes can now be measured routinely and accurately using direct calorimetric methods and, with the growing accumulation of reliable data on a wide range of systems, the time is ripe for a re-evaluation of what the data might have in common and what this tells us about the underlying forces responsible for controlling the stability and interaction of biological (macro)molecules. Here I wish to summarize briefly the current situation regarding heat capacity changes and related thermodynamic quantities in relation to the conventional interpretations based on hydrophobic interaction. I will show that, despite the enormous success of the hydrophobic concept, so-called “anomalous” heat capacity effects are in reality much more ubiquitous than has previously been assumed, and are not restricted solely to the interactions between non-polar groups in water. Indeed, both experimental data and theoretical considerations show that changes in heat capacity quantitatively similar to those seen in biomolecular processes are an inevitable consequence in any system involving cooperative transitions of a multiplicity of weak interactions—regardless of the nature of the interaction.

Differential scanning calorimetry (DSC) measurements on the thermal unfolding of simple globular proteins in solution typify the general observation that disruption of biomolecular assemblies results in an increase in heat capacity (positive ΔCp). This is illustrated in Fig. 1, which shows how the heat capacity of unfolded polypeptide is generally higher than the compact native state from which it is derived. As a natural thermodynamic consequence, the higher the temperature of the unfolding transition, the more heat energy is required to unfold the protein. This is also shown by the data in Fig. 1, where the Tm for the unfolding transition has been varied by adjusting the pH of the sample. Not only does the baseline shift during the transition, but the magnitude of the transition endotherm increases with Tm. Such observations, coupled with structural information about the disposition of non-polar amino acid sidechains in globular protein structures, were among the first to support the hypothesis that protein folding is driven by hydrophobic interactions involving the burial of hydrophobic groups [1], [2], [3].

Detailed analysis of such experimental data gives the temperature dependence of the thermodynamic parameters that control folding stability. Such data (illustrated in Fig. 2) show how the very strong temperature dependencies of both ΔHunf and T·ΔSunf compensate to give relatively much smaller changes in ΔGunf. This is just one example of the widely reported phenomenon of entropy–enthalpy compensation, a natural consequence of weakly interacting systems itself [4], [5].

Nor are such heat capacity changes just a property of native protein (un)folding, since they are also seen in much less specific condensation processes such as protein aggregation. For example, Fig. 3 shows DSC data for the thermal aggregation of insulin in solution under conditions where it forms amyloid-like fibrils upon thermal denaturation. In this case the DSC thermogram shows a negative ΔCp effect, similar in magnitude but opposite in sign to the positive ΔCp seen for unfolding without aggregation. This indicates that heat capacity changes reflect general changes in polypeptide environment, and that condensed/closely packed polypeptides – either as a result of specific folding or less-specific aggregation – have a lower heat capacity than the unraveled chain exposed to water.

Large ΔCp effects are also seen in many protein–protein interactions. In some cases, such as illustrated in Fig. 4 for the binding of a small protein domain at the dimer interface of the E3 subunit of pyruvate dehydrogenase from B. stearothermophilus [8], [9], [10], the temperature dependence of the observed heat of binding is so great that it can even change sign over a relatively small temperature range. Again, the entropy component (T·ΔS°) of this interaction shows a similar compensating dependence, such that the overall Gibbs free energy of the interaction (ΔG°) scarcely changes. (We have termed this phenomenon “thermodynamic homeostasis” [5] since, whatever its origin, it might have evolutionary significance for organisms during adaptation to hostile environments.)

ΔCp for the E3/di-domain interaction depends somewhat on ionic strength [8], but is around −1.9 kJ mol−1 K−1 in 50 mM phosphate buffer, pH 7.4 (Fig. 4). Prevailing wisdom is that ΔCp effects may be correlated with changes in polar and non-polar exposed surface areas (ΔASA). The crystal structure of this complex [11] gives a calculated total surface area buried at the protein–protein interface ΔASAtot=1125 Å2, with a polar/non-polar ratio, ΔASAp/ΔASAnp=0.67. Using published algorithms [12], [13], [14] that relate ΔCp to changes in accessible surface areas (ΔASA), we would predict ΔCp=−640 to −780 J mol−1 K−1 (different algorithms have different coefficients), compared to the measured ΔCp=−1900 J mol−1 K−1. Such discrepancy is disturbing, and suggests that ΔCp/ΔASA correlations should be treated with caution when applied out with the limited data sets from which they were derived. Significantly, such a large ΔCp as observed here would normally be taken to indicate a dominant role for hydrophobic interactions, yet the crystal structure, supported by mutagenesis studies [9], [10], shows that the key interactions are predominantly electrostatic in this instance. Similar discrepancies between measured ΔCp and predicted surface area burial have been reported recently for protein–DNA interactions [15] and for drug–protein and inhibitor binding [16], [17].

The binding to proteins of seemingly very hydrophilic molecules also shows significant ΔCp effects. This has been demonstrated in an ITC study of the binding of specifically engineered tri-saccharides to the lectin, concanavilin-A (ConA) summarized in Fig. 5 [18].

These two tri-saccharide ligands (Fig. 5) were specifically designed and constructed [18] to probe the effects of displacement of a single water molecule identified in structural studies of ConA–ligand complexes. The water molecule trapped in the ConA/ligand 1 complex is displaced from the active site by the additional hydroxyethyl group in ligand 2. Two points to note here: (a) sugars are water-soluble molecules, not normally considered to have any substantial hydrophobic content, yet their heats of binding to ConA show significant temperature dependence/ΔCp behaviour; (b) incorporation of the larger hydroxyethyl group and consequent water displacement with ligand 2 reduces both the heat and entropy loss on binding. It also reduces the magnitude of the (negative) ΔCp effect. This is surprising since, all other things being equal, one might anticipate that the slight increase in hydrophobicity of ligand 2 compared to ligand 1 (from the –CH2–CH2– group) should have the opposite effect on ΔCp.

These are just a few examples of the ubiquitous nature of ΔCp effects. The classic explanation, stemming from the influential paper by Kauzmann [1], is that such effects are a consequence of the unusual properties of solvent water and the way in which water molecules – having a greater affinity for each other in the liquid phase – tend to repel non-polar molecules and groups: the hydrophobic effect. Simple empirical observations on model compounds in water (e.g. [19]) suggest that the thermodynamics of such interactions should show complicated temperature dependence. For example, the solubility of non-polar compounds in water – contrary to expectation – initially decreases with increase in temperature (hydrophobic association endothermic at low T) before behaving more rationally at higher temperatures (hydrophobic association exothermic at higher T). Consequently, the association of hydrophobic groups in water should have a negative ΔCp. Such observations have been variously described in terms of the structural re-organization (“ice-bergs”, “clusters”, “clathrates”, etc.) of solvent water molecules around hydrophobic groups but, regardless of the detailed explanation, hydrophobic interactions are expected to show this characteristic ΔCp “signature”. But is it unique? Can other interactions show similar effects? We have seen in some of the examples described above how sometimes it is difficult to reconcile observed ΔCp effects with hydrophobicity alone.

It is worth noting that Kauzmann [1] even in his seminal 1959 review recognized that the thermodynamic signature of hydrophobic interactions was not unique, in particular that both hydrophobic and electrostatic interactions (“salt linkages”) in water are expected to be endothermic and entropy-driven, at least at low temperatures. This has not always been appreciated in subsequent literature.

Section snippets

Positive ΔCp is normal for cooperative order–disorder transitions, especially in H-bonded lattices

A clue to the ubiquitous nature of ΔCp effects comes from examination of the heat capacity changes that take place upon melting of pure crystalline solids, particularly those involving weaker non-covalent lattice forces [20]. This is summarized in Fig. 6, which shows heat capacity data taken from the older literature [21] for simple compounds above and below their melting points.

This increase in heat capacity during order–disorder (solid–liquid) transitions is to be expected. The heat capacity

Protein–carbohydrate binding: the thermodynamics of trapped waters

We can extend these arguments to explore protein–ligand interactions, and in this section we will exemplify this by an analysis of the protein–carbohydrate interaction described above.

Determining the role of solvation/hydration changes in the thermodynamics of protein–ligand interactions is complicated because of the need to take account of all the waters in the system, not just those that might be most apparent in well-resolved crystal structures. A generic approach based on the thermodynamics

Conclusions

  • “Anomalous” heat capacity effects are ubiquitous, and are a natural property of any transition or binding process involving a multiplicity of weak interactions—not just hydrophobics.

  • Simple estimates based on cooperative lattice models give thermodynamic parameters comparable to observation.

  • This may require re-evaluation of the role of hydrophobic interactions.

  • In general it can be shown that large ΔCp effects are to be expected in any system comprising multiple, cooperative, weak (ca. kT)

Acknowledgements

The Biological Microcalorimetry facility in Glasgow is funded by BBSRC and EPSRC. I thank Samantha Rutherford for DSC data and Margaret Nutley for technical support.

References (27)

  • W.M. Jackson et al.

    Thermodynamics of protein denaturation. Calorimetric study of the reversible denaturation of chymotrypsinogen and conclusions regarding the accuracy of the two-state approximation

    Biochemistry

    (1970)
  • S.J. Rutherford, The kinetics and thermodynamics of insulin fibril formation, PhD thesis, Glasgow University,...
  • W. Dzwolak et al.

    Aggregation of bovine insulin probed by DSC/PPC calorimetry and FTIR spectroscopy

    Biochemistry

    (2003)
  • Cited by (182)

    • Intrinsic affinity of protein – ligand binding by differential scanning calorimetry

      2022, Biochimica et Biophysica Acta - Proteins and Proteomics
    • Equivalence of the transition heat capacities of proteins and DNA

      2022, Biochemical and Biophysical Research Communications
    View all citing articles on Scopus
    View full text