A scanning calorimeter holds a sample cell and a reference cell at matched temperature while ramping both, and records the differential power needed to keep them matched. What that power difference buys is heat. The technique measures the heat capacity of states and the excess heat associated with transitions that can be induced by temperature change, and the integral of the excess heat capacity is the enthalpy for that process [1]. Nothing stands between the conformational change and the signal.
Contrast the dye-reported melt. In differential scanning fluorimetry the temperature at which a protein unfolds is measured by an increase in the fluorescence of a dye with affinity for hydrophobic parts of the protein, which are exposed as the protein unfolds, and a simple fitting procedure returns the transition midpoint [2]. The quantity that comes out is described in that method's own protocol literature as an apparent melting temperature [3]. The adjective is doing real work. What follows is about what a calorimetric scan licenses you to conclude instead, and about the several ways a thermogram can look clean and still not mean what you want it to.
What is actually in the thermogram
Three quantities come off a scan. The transition midpoint, read as the temperature of maximum excess heat capacity. The calorimetric enthalpy ΔHcal, obtained by integrating the excess heat capacity across the transition against a baseline interpolated between the pre- and post-transition states [1]. And the heat capacity change on unfolding: in a calorimetric analysis of carbonic anhydrase, the heat capacity of unfolding was estimated from the dependence of the calorimetric enthalpy on Tm [4] — that is, from a series of scans under different conditions rather than from one.
ΔHcal is what makes DSC different in kind. It is an area, not a parameter returned by a fit, and it does not care whether your model of the transition is right. The shape of the peak carries more than the midpoint does: the experimental output of a DSC scan reflects the energetics of all conformations that become minimally populated during thermal unfolding, which is why, for ultrafast-folding proteins whose conformational ensembles are naturally broad and whose unfolding is minimally cooperative, the technique opens the possibility of probing the complete folding free-energy landscape — at the cost of requiring high-quality experiments and analytical methods built on statistical mechanics [5].
Calorimetric enthalpy against van't Hoff enthalpy
The second enthalpy in a DSC analysis is not measured. It is inferred from the steepness of the transition under an assumed model, and the model is almost always two-state. That model assumes only two types of molecule in solution, native and unfolded; in reality unfolding is a multistep process even when intermediate states are only sparsely populated, and for apolipoprotein A-1, lysozyme and nine further protein examples taken from the literature, the transition curves and enthalpies predicted by the two-state model differed considerably from the experimental results [6].
When the two enthalpies disagree, the productive question is how many cooperative units the molecule contains. Streptokinase makes the point cleanly. Under conditions where its thermal unfolding occurs unequivocally under equilibrium, the heat capacity curve of the intact protein was quantitatively described by three independent two-state transitions, each of which compared well with the two-state transition observed for the corresponding isolated domain, indicating that each structural domain behaves as a single cooperative unfolding unit [7]. The total calorimetric enthalpy of that molecule is distributed across three van't Hoff enthalpies, none of which equals it.
The equality is therefore an input, not a result. It was the stated premise of the decomposition of the rat brain hexokinase endotherm in high-ionic-strength buffer, which was resolved into two transitions with Tm values of about 48 and 51 °C and enthalpies of about 109 and 112 kcal/mol only after assuming two-state behaviour with calorimetric enthalpy equal to van't Hoff enthalpy [8]. Report the ratio you obtain. Do not report a two-state ΔH as though the fit that assumed two-state behaviour had tested it.
A peptide the method was worked out on
A 56-residue fragment of GCN4 containing the leucine zipper and part of the basic region was studied by high-sensitivity DSC and circular dichroism. Its thermal unfolding was a reversible process well represented by a reaction of the form N2 ↔ 2U, indicating that the helices are stable only in the coiled-coil conformation. The transition temperature was concentration dependent: at pH 7.06 it was close to 70 °C at 5 × 10−4 M and close to 50 °C at 5 × 10−6 M, with an unfolding enthalpy of 31.5 kcal mol−1 at 70 °C [9].
That concentration dependence is a diagnostic you get for free. A unimolecular transition has a midpoint that does not move with concentration; a transition coupled to dissociation does, and in this peptide the midpoint moved roughly twenty degrees across a hundred-fold concentration change [9]. A Tm quoted without the concentration it was measured at is, for any self-associating peptide, not a usable number.
The same coiled coil later showed how much a satisfying two-state description can conceal. Studied by circular dichroism, spectrofluorimetry and heat capacity scanning calorimetry, temperature-induced unfolding of the leucine zipper was shown not to represent a simple two-state transition but to consist of several stages: a first transition starting at the very beginning of heating from 0 °C, sensitive to modification of the N terminus and independent of concentration; a second, unimolecular transition sensitive to modification of both termini that involves the whole dimer without dissociating it; and only then the concentration-dependent cooperative unfolding and dissociation of the two strands. The structural changes taking place before cooperative unfolding comprised almost 40% of the total enthalpy and entropy of unfolding of the completely folded coiled coil, the state in which it appears to be below 0 °C [10].
The practical consequence is about where the pre-transition baseline goes. If a peptide is already partly frayed at your starting temperature, the enthalpy you integrate is the enthalpy of what remains, not of unfolding from a fully folded state.
Reversibility, and the rescan that tests it
Thermodynamic parameters obtained from DSC are meaningful only for reversible unfolding at equilibrium [11]. The test is a rescan: cool the cell from beyond the transition, run the ramp again, and compare the recovered peak area with the first.
A partial scan sharpens it. In rat brain hexokinase, samples in which the domain corresponding to the first transition had been selectively denatured by stopping the scan showed only the second transition on rescanning, confirming selective denaturation of the first domain and retention of the folded structure of the second [8]. The rescan is not merely a pass/fail gate; it can be used to isolate one part of a multi-stage thermogram.
Aggregation is what usually destroys reversibility, and it does not simply shrink the second scan — it corrupts the first. In streptokinase at pH 7.0 and high sample concentration, or at pH 6.0 at any concentration investigated, unfolding of one domain was accompanied by time-dependent formation of aggregates; this produced a severe deformation of the DSC curves, which became concentration dependent and kinetically controlled, and precluded their proper analysis by standard deconvolution methods [7]. A synthetic peptide can do the same: DSC of the soluble linear peptide S4, a fragment inspired by human tropoelastin, in solution in TBS revealed irreversible aggregation into amyloid-like fibres [12]. If your rescan is flat, the shape of the first scan is evidence about a kinetic process, not about an equilibrium.
Scan rate, and transitions under kinetic control
The reference case is thermolysin, whose calorimetric traces were irreversible and highly scan-rate dependent. Both the shape of the thermograms and their scan-rate dependence were explained by a kinetic scheme in which the native state converts with a first-order rate constant that changes with temperature according to the Arrhenius equation to a final state reached irreversibly; on that basis the rate constant as a function of temperature and the activation energy were calculated, and the scheme was shown to be a particular case of the Lumry–Eyring model. That work explicitly framed the general use of the observation: the scan-rate effect on calorimetric traces serves as an equilibrium criterion [13]. Run at two or three rates. If the midpoint moves, equilibrium analysis is off the table.
Simulation bounds how far slow scanning rescues the situation. For a three-state kinetic model in which the reversible unfolding step is followed by an irreversible one, the apparent transition temperature is always reduced by the irreversible step; it is independent of the reduced scan rate at sufficiently slow rates even when denaturation is highly irreversible, but at fast rates the excess specific heat profile is determined by the rate of unfolding. Accurate ΔH and ΔS for the reversible step can be obtained only when the unfolding rate constant exceeds the irreversible rate constant by more than 2000- to 50,000-fold [11].
Scan-rate dependence is not confined to irreversible transitions, which is the trap. For Bacillus circulans xylanase and a disulfide-bridge mutant, calorimetric transitions reappeared in the second heating cycle in the presence of low, non-denaturing urea concentrations, indicating reversible denaturation under those conditions — yet the wild-type DSC curves remained scan-rate dependent there, much as in the absence of urea. The kinetic factors responsible can produce significant distortion and asymmetry of endotherms, especially at higher scanning rates, which is why the check must be made even for reversible denaturation before any asymmetric DSC curve is analysed by standard thermodynamic procedures [14].
On a peptide, the two regimes have been separated explicitly. The collagen-like peptide (Pro-Pro-Gly)10 was scanned over a range of rates and concentrations. At 24 mg/ml and 0.025 °C min−1, helicity approached a quasistatic state in which it reached its equilibrium value at all temperatures, and the endotherm peak temperature increased with chain concentration but was independent of scanning rate. At 0.25–1.0 mg/ml and 0.1 °C min−1 and above, the peak temperature increased linearly with the logarithm of the scanning rate and was independent of concentration. At intermediate rates two peaks were apparent, a nascent quasistatic peak and a nascent rate peak [15].
Treat that as a map of your own experiment. Which regime you are in is set jointly by concentration and rate, and the two regimes give opposite dependences. A midpoint that tracks the logarithm of the scan rate and ignores concentration is reporting an unfolding rate.
Concentration, buffer and baseline
A calorimeter needs enough material in the cell to generate a heat effect it can resolve, and the concentrations in published peptide work are not modest: the GCN4 fragment was scanned at concentrations including 5 × 10−4 M [9] and the quasistatic collagen-peptide scans ran at 24 mg/ml [15]. More is not automatically better. For streptokinase, unambiguous equilibrium unfolding required pH 7.0 with sample concentration below approximately 1.5 mg/mL, or pH 8.0; outside that window aggregation took over [7]. The working range has to be found by scanning at several concentrations, not assumed.
Buffer enters the measurement twice over. Instrumentally, the reference cell should hold the solvent the sample was dialysed into, because the instrument reports a difference and anything the two cells do not share lands in the baseline. This is not hypothetical. In the hexokinase study, transition temperatures were still resolved in low-ionic-strength buffer, but enthalpies were not routinely determined there because of erratic baseline behaviour under those conditions [8].
Thermodynamically, solvent composition moves the transition itself. For the collagen model peptide (Pro-Pro-Gly)10 the transition temperatures were 25.4 and 28.7 °C in H2O and D2O respectively; the increase in Tm measured calorimetrically at 1.0 °C min−1 in a low pH solution on going from protonated to deuterated solvent was 5.1 °C, against 4.2 °C for (Gly-Pro-4(R)Hyp)9 and 2.2 °C for pepsin-extracted type I collagen [16]. A solvent substitution nobody would call a formulation change moved a peptide midpoint by several degrees. Two midpoints measured in different solvents are not comparable numbers.
Deconvolution, and when it is allowed
A thermogram with a shoulder, a second peak or a stubbornly asymmetric one invites decomposition into sub-transitions, and the fit will usually oblige. The streptokinase analysis shows what makes such a decomposition defensible: the effects of pH, sample concentration and heating rate on the thermograms were analysed first to find conditions where unfolding occurs unequivocally under equilibrium, and only then was the intact heat capacity curve described as three independent two-state transitions — each compared against the two-state transition measured on the corresponding isolated domain [7]. The model was checked against separately prepared fragments rather than against itself.
The hexokinase decomposition was validated by chemistry instead. After a partial scan through the first transition, markedly increased susceptibility to tryptic digestion identified the denatured region, and immunoblotting with monoclonal antibodies recognising epitopes at defined positions assigned the first transition to unfolding of the C-terminal half of the molecule and the second to the more stable N-terminal half [8].
Both are proteins with real structural domains. A short peptide rarely has domains, yet it can still give a multi-stage thermogram — as the leucine zipper did, where the stages were separated by their concentration dependence and by sensitivity to modification of one terminus or of both, not by curve-fitting alone [10]. Where the peak shape itself is the object, more is available than a sum of two-state terms, but it requires high-quality data and analysis grounded in statistical mechanics [5].
Against the dye-reported melt
The point is not that one technique beats the other; they report different observables. The fluorimetric signal is an extrinsic dye binding hydrophobic surface exposed on unfolding, read on a conventional real-time PCR instrument, with the midpoint extracted by a simple fitting procedure [2]; the stated output is an apparent melting temperature, and the protocol literature for the method devotes space to its common pitfalls and how to identify them [3]. The calorimetric signal is heat, and the enthalpy is an integral [1].
Three consequences. A dye melt cannot return ΔHcal at all — DSC is the method of choice for measuring the unfolding enthalpy, and deducing an enthalpy from a CD transition curve under the two-state model can give erroneous results [6]. A dye that binds hydrophobic surface needs hydrophobic surface to appear, so a short, highly polar or already-disordered peptide can unfold without generating much signal, whereas heat is absorbed either way. And the well contains more than the peptide: any equilibrium binding ligand present increases thermal stability by an amount proportional to its concentration and affinity [4], which is one reason a fluorimetric midpoint is protocol-dependent [3].
What the thermogram does not tell you
It does not tell you what the unfolded state is, only how much heat separated it from the folded one. It does not tell you about stability on the bench: Tm is a midpoint under a ramp, and where the transition is kinetically controlled that midpoint is set by an unfolding rate at the scan rate you chose [15, 13]. It does not tell you the peptide is monomeric — only a concentration series does that [9]. It does not tell you the two-state model holds; a fit that assumed the model cannot test it [8], and a sparsely populated intermediate can leave the curve looking two-state while the predicted enthalpy is wrong [6]. And a scan under kinetic control licenses a ΔG, ΔH or ΔS only in the limiting case noted above, where the reversible unfolding step outruns the irreversible one by orders of magnitude [11]; short of that it licenses at most an activation energy under an explicitly stated kinetic model [13].