Thermostability of Proteins

What Does It Depend On and How Can It Be Modified?


 It is important to distinguish between two types of stability, namely thermodynamic and kinetic. The former refers to the extent to which the free energy of the native conformation differs from that of the (reversibly) unfolded form. Its value can in principle be changed by engineering a protein so that it has a more stable native state (eg by introducing new salt bridges, hydrogen bonds etc) or by destabilising the unfolded state (eg by limiting its conformational entropy) or a mixture of the two. Moreover it should be possible, by examining the 3-D structures of two related proteins, to predict their relative conformational stabilities from the stabilising interactions in the native forms (although this might be difficult because solvent effects should be taken into account).

What about kinetic thermostability? By this I mean the rate at which a protein becomes irreversibly inactivated on heating. Again, it is usually assumed that this depends on the nature of the stabilising interactions in the native state and that by increasing these interactions a more thermostable protein can be produced. This may be true if

  1. the rate limiting step of inactivation proceeds from the native state of the protein
  2. the rate of the process depends essentially on the activation energy (as defined by the Arrhenius equation) and
  3. if the transition state is essentially unchanged by these extra interactions in the ground state.

To see if criterion i) applies it is clearly necessary to know the kinetic pathway to inactivation. It is usually assumed that thermal inactivation proceeds via a reversible unfolding process folded by an irreversible step and moreover that the unfolded intermediate is some sort of random coil, i.e. that it resembles the form produced in reversible unfolding in response to changing solvent composition. If this is indeed the case, and if the unfolding is rate limiting, then some correlation might be expected between rate and stabilising interactions in the native state. However for a protein that unfolds rapidly but undergoes irreversible inactivation only slowly the rate of irreversible inactivation cannot depend on interactions in the native state since those have already been lost. Such a case was reported by Zale and Klibanov (Biotechnol. Bioeng. 1983, 25, 2221-2230) but the implications of these important observations seem to have been largely ignored.

We have been looking at the relative thermostabilities of cytosolic and mitochondrial aspartate aminotransferases. The cytosolic form has long been known to be the more thermally stable of the two proteins but the extent of the greater stability had not been properly quantified and the pathway to inactivation had not been established.

In one study we measured the rates of thermal inactivation over a range of temperatures (Twomey and Doonan, Biochim. Biophys. Acta, 1342 (1998) 37-44). The process apparently followed first order kinetics suggesting that the proteins inactivate in a single step pathway. This is not, in fact, the case. In a second study the rates of loss of CD signal at 225 nm were measured (Twomey, Kelly, Price and Doonan, submitted for publication) and the results showed that at all temperatures examined the rate of loss of secondary structure was much greater than the rate of irreversible thermal inactivation. Quantitatively the results were consistent with a model in which there is a rapid and reversible loss of secondary structure followed by a rate-determining irreversible step. Interestingly the rapidly formed intermediate did not seem to be extensively unfolded but rather appeared to be compact and essentially devoid of secondary structure.

Also of interest is the fact that "melting temperatures" measured for loss of secondary structure were considerably lower than those obtained by differential scanning calorimetry. This is further evidence that the initial fast step in irreversible thermal inactivation is not the result of extensive unfolding. It also means that an increase in the calorimetric Tm for a modified protein should not be taken to show that the protein has increased kinetic stability. The two measure different things.

The essential fact is, then, that the rate determining step in the irreversible thermal inactivation of these isoenzymes proceeds not from the native structure but from an intermediate that does not resemble that structure; hence nothing can be said about the relative thermostabilities of the isoenzymes on the basis of interactions in the native structures.

Going further our results show that it is not in fact the activation energy that is the main determinant of the rate of slow step in inactivation. We measured the rate of irreversible thermal inactivation of the two isoenzymes as a function of temperature. The Arrhenius plots for the two proteins are shown below.






Arrhenius plots for the thermal inactivation of cytosolic (lower line) and mitochondrial (upper line) aspartate aminotransferases

 Both plots show a break at a temperature of about 60oC which is interpreted to mean that the pathways to inactivation are different above and below this temperature. Also of interest is the fact that, at about 45oC the lines cross; that is, the mitochondrial isoenzyme is the more stable at low temperatures.

The central point, however, is that in both temperature ranges the slope of the Arrhenius plot is greater for the mitochondrial than for the cytosolic isoenzyme. That is the activation energy for inactivation of the mitochondrial isoenzyme is greater than that for the cytosolic form. What then is the reason for the slower rate of inactivation or greater thermostability (at most temperatures) of the cytosolic form? Given the Arrhenius equation k1 = Ae then the origin of this must lie in a smaller value of the pre-exponential term A.

The significance of the pre-exponential factor is not entirely clear for complex reactions in solution but for simple reactions transition state theory provides a direct link between A and the entropy of activation. If it is assumed that the same relationship holds for these much more complex reactions then the enthalpy and entropies of the thermal inactivation of the two isoenzymes can be calculated as

Temperature range

D H (kJ/mol)

D S (J/mol/k)

















 The point is clear that in both temperature ranges the relative rates of inactivation are determined by the entropy of activation rather than by the enthalpy. That is, what counts is not the stabilities of the intermediates on the reaction pathway but rather the entropy change on going to the transition state between the intermediate and the irreversibly inactivated form. The entropy of activation for the mitochondrial protein is much larger than that for the cytosolic form implying a more disordered transition state. The reason for the much lower entropies of activation in the low temperature range may reflect either less disordered transition states or the fact that the hydrophobic effect is grater at lower temperature i.e. the increase in configurational entropy is at least in part offset by ordering of the solvent.

The basic contention then is that the relative thermostabilities of the aspartate aminotransferases have nothing to do with stabilising interactions in the native state because the slow step of inactivation does not proceed form the native state. What is important is the degree of unfolding in the transition state between a rapidly formed intermediate on the reaction path and the final irreversibly inactivated form. If this is so more generally then it means that 3-D structures of thermostable proteins will not tell us why they are thermostable or indeed how to engineer in extra thermostability.

Please send any views on this to


Doonan's home page