DSC and the meaning of Tm – DSC protein stability

par Stoyan Milev, Tuesday, 17th November 2015

We can extract multiple thermodynamic parameters from DSC data, such as ΔH, ΔH_vH, ΔCp, and ΔG … but the most widely used parameter is T_m. Incidentally this is also the easiest and most accurate value to obtain – it is simply the temperature at the peak maximum.

DSC protein stability

How does it relate to protein stability?

There is more than one definition of “protein stability”. Most commonly, for industrially important proteins, the term refers to their functional (or operational) stability at physiological temperature; i.e. how long they can perform their function at 37°C? This can be assessed by isothermal studies that can take days or weeks to complete, or alternatively, if Differential Scanning Calorimetry (DSC) is used instead, the protein can be denatured in a matter of minutes.

Which thermodynamic parameter obtained by DSC correlates best with functional stability? It turns out that this is T_m. Thermodynamic stability (ΔG), on the other hand, is a poor predictor of functional stability; technically ΔG applies only to reversible unfolding, moreover it is calculated from T_m, ΔH and ΔCp, the latter of which can be difficult to obtain. An example is the correlation of T_m and ΔG with aggregation half time (as a measure of functional stability) of human botulinum protein antigen serotype C, used as a model protein. The ΔG vs t_1/2 aggr. correlation coefficient (R) is only 0.4, while the T_m vs t_1/2 aggr. is 0.92 (data from J Pharm Sci. 2011 Mar;100(3):836-48).

Here is one way of thinking about T_m:

If we assume these two scans show a protein in two different formulations or two different protein constructs, the negative T_m shift of 5°C actually reflects an increase from 2% to 3% of the unfolded fraction (Fu) at 37°C. Fu (T) can be estimated graphically as the percentage of the shaded area below T vs the total area under the peak. Unlike routinely assigned DSC baselines, the one applied here accounts for non-zero Fn (native fraction) and Fu (unfolded fraction) outside the cooperative transition peak as well. The integral of the curves depicts more realistic Fu distribution (inset).

Since the formation of the aggregates is likely a concentration-dependent process, the higher concentration of unfolded protein (red scan) will lead to faster aggregation (larger pool of U to convert to I) – see schematics below.

A corollary of this interpretation is that the overall shapes of the curves should be similar. We can assume this to be the case for a protein in different formulations or similar constructs stemming from one parental molecule, but it is likely not the case for completely different proteins where the use of T_m as a comparative stability predictor should be used with caution.

Related content: