Thermal unfolding of proteins: principles, methods and applications – What is thermal unfolding?

von Agatha Rosenthal, Thursday, 3rd September 2026

Cell,And,Protein,Realistic,Structure,Science,Innovation,Biology,,Dna,3d

Thermal unfolding is the temperature-driven loss of a protein’s native three-dimensional structure. As thermal energy increases, non-covalent interactions that stabilise the folded state — hydrogen bonds, hydrophobic packing, van der Waals contacts — are progressively disrupted, and the protein transitions from an ordered, compact conformation to a disordered or partially structured state.

This transition is typically characterised by the melting temperature (T_m): the temperature at which the folded and unfolded populations are present in equal proportion. T_m is not a fixed physical constant of a protein in isolation — it depends on solution conditions, buffer composition, and the presence of ligands or excipients — which is precisely what makes it useful. Comparing T_m across conditions or variants provides a quantitative, reproducible measure of relative thermal stability. This context-dependence extends beyond the test tube: Leuenberger and colleagues’ cell-wide thermal profiling study, tracking the unfolding of thousands of proteins simultaneously inside intact cells, showed that thermostability correlates with structural and biological features such as protein size, quaternary structure, and aggregation propensity, rather than being determined by amino acid sequence alone.¹ Thermal stability, in other words, reflects a protein’s broader structural context, not just its fold in isolation — reinforcing why an in vitro Tm shift is such an informative, generalisable readout of underlying structural sensitivity.

A higher T_m generally means a more thermally stable molecule, and T_m shifts are one of the fastest ways to see how a mutation, buffer, excipient or formulation condition affects a protein’s stability.

In biopharmaceutical development, this matters because thermal stability correlates closely with a protein’s behaviour under real-world stress: elevated temperature during manufacturing, transport, or storage; freeze-thaw cycling; or simple time-at-temperature during shelf life. A lower or shifted Tm is an early indicator that a molecule may be more susceptible to these stresses.

Unfolding is also mechanistically linked to downstream risk. As a protein loses native structure, previously buried hydrophobic residues become solvent-exposed. These exposed surfaces are the primary driver of protein–protein association, making partially or fully unfolded species prone to aggregation. Aggregation, in turn, is associated with loss of biological activity, altered pharmacokinetics, and — in biologics — immunogenicity risk. Thermal unfolding is therefore rarely of interest on its own; it is a proxy for the structural instability that precedes these downstream failure modes, which is why Llowarch et al. describe thermal shift as one of the most sensitive and widely used readouts of protein stability in early drug discovery.²

How is thermal unfolding measured?

Differential scanning calorimetry (DSC) and differential scanning fluorimetry (DSF) are the two most widely used techniques for measuring thermal unfolding, and they probe fundamentally different physical properties.

DSC measures the heat directly. Unfolding is an endothermic process, so it appears as a heat capacity peak, from which T_m, the enthalpy of unfolding (ΔH), and heat capacity change (ΔC_p) can all be extracted without relying on a reporter molecule or assumed spectroscopic model. This makes DSC a model-independent, thermodynamically direct measurement — the reference technique against which other methods are typically validated.

DSF is optical and indirect: it tracks a fluorescence signal that changes as structure is lost. Conventional DSF uses an extrinsic dye that binds exposed hydrophobic surfaces on unfolding. Intrinsic DSF as a entirely label-free technique tracks the intrinsic fluorescence of tryptophan and tyrosine residues as the protein is heated. As those residues become more solvent-exposed during unfolding, their emission spectrum shifts, and plotting that shift against temperature reveals the onset temperature and the T_m.

The two are not interchangeable. NIST’s direct comparison found DSC and DSF T_m values often agree, but not always. Results depend on scan rate and method choice can meaningfully shift the transition temperatures obtained — a reminder that “how you measure” is as important as “what you measure”³, since unfolding is frequently under kinetic rather than purely thermodynamic control.³

This is why the techniques are complementary rather than substitutable: DSC provides thermodynamic depth, DSF provides throughput. Using DSF for broad early screening and DSC to confirm and characterise the resulting leads offsets each technique’s main limitation — reflecting current practice in drug discovery, where shift assays serve as the primary screen and calorimetric methods provide orthogonal confirmation.² DSC and DSF report on conformation, not on what happens once a protein starts to unfold. Dynamic light scattering (DLS) closes that gap: DLS measures the size of particles in solution by analysing fluctuations in scattered light, making it exquisitely sensitive to the appearance of higher-order species — dimers, oligomers, and larger aggregates — as a sample is heated. Run as a thermal ramp alongside DSF or DSC, DLS reveals the onset of aggregation (T_agg) alongside any changes in hydrodynamic radius, showing whether an unfolding event is followed by self-association.

DLS and SLS: completing the picture

Static light scattering (SLS) adds a further, thermodynamic layer. From the concentration dependence of scattering intensity, SLS gives direct access to the second virial coefficient, B₂₂ — a measure of net protein–protein interaction. This is a more rigorous readout than the diffusion interaction parameter, k_D, traditionally derived from DLS alone: k_D is empirical and convolves true thermodynamic interaction with hydrodynamic and frictional effects, sometimes distorting or even inverting the apparent interaction sign. B₂₂ is free of these hydrodynamic artefacts, making it a more direct and reliable predictor of colloidal stability, solubility, and aggregation propensity.

Together, a conformational analysis (DSF or DSC), a size analysis (DLS), and a thermodynamic interaction analysis (SLS) turn a single thermal scan into a complete picture — not just when a protein unfolds, but whether, how strongly, and under which conditions it goes on to self-associate.

Why combine DSF and high-throughput DLS/ SLS?

Unfolding and aggregation are related, but they are not the same event, and they don’t always track together. A molecule can be conformationally very stable and still aggregate at moderate temperatures because of colloidal instability; conversely, a molecule can unfold at a relatively low T_m without immediately forming visible aggregates. Measuring only one of the two risks missing the other failure mode entirely.

This is exactly the argument made in the drug discovery literature on thermal shift approaches: because these assays are built on the simple physical link between temperature and the folded/unfolded equilibrium of a protein, they are most informative when paired with orthogonal detection methods that confirm what that shifting equilibrium is actually doing to the sample.⁴ Running DSF and DLS/ SLS on the same thermal ramp, on the same plate, means every T_m value can be checked against a T_agg value from the identical sample and heating profile — no second experiment, no separate aliquot, no risk that differences between assays are really just differences between sample preparations.

The efficiency case is just as compelling as the scientific one. Instead of running a DSC campaign, a separate DSF screen, and then a DLS/SLS stability study, a combined DSF/ DLS/ SLS workflow collects conformational and colloidal data in one pass. For formulation screening and stability assessment — where dozens of buffer, pH, and excipient combinations need to be compared — that integration saves both sample and time, without sacrificing the depth and quality of the dataset.

Benefits of high-throughput analysis

High throughput isn’t just about speed for its own sake; it changes what kind of questions you can ask. A plate-based, high-throughput DSF/DLS/SLS workflow lets you compare dozens of formulations, buffers, or candidate molecules side by side, under identical conditions, in a single run.

That breadth matters most early in development, when the goal isn’t to fully characterise one molecule but to compare many and rank them. Generating robust, comparable data sets at this stage — rather than deep but narrow data on a handful of conditions — gives formulation and candidate-selection teams a much stronger basis for decisions. Faster screening means those decisions get made sooner, with fewer resources spent chasing candidates or conditions that were never going to work. The net effect is what every biologics program is chasing: shorter development timelines, and material and instrument time spent where it counts.

Applications in biologics

Formulation development. Screening excipients, buffers, pH, and ionic strength against T_m and T_agg in parallel lets formulators quickly identify conditions that protect a molecule against both unfolding and aggregation — not just one or the other.

Candidate screening. When multiple candidate variants or mutants are on the table, thermal stability profiling is one of the fastest ways to rank them, flagging which candidates are least likely to cause downstream manufacturability or stability headaches.

Stability studies. Stability studies combine unfolding and aggregation behaviour to build a risk profile for a molecule under thermal stress. Tracking T_onset, T_m, T_agg, and B₂₂ across candidate formulations or storage conditions shows not only where a protein starts to lose structure, but whether that loss of structure translates into self-association — the combination needed to flag aggregation-prone conditions before they surface later in development.

Combined insight, less effort. Across all three use cases, the appeal of pairing high-throughput intrinsic DSF with high-throughput DLS/ SLS is the same: richer, better-correlated data, generated with less sample and in less time than running the equivalent techniques separately.

Conclusion

Thermal unfolding is one of the most information-rich ways to interrogate protein stability — but only if you’re looking at the whole picture. Combining intrinsic DSF and high-throughput DLS captures conformational change and aggregation behaviour side by side, from the same sample, on the same thermal ramp. The practical upside is straightforward: lower sample consumption, faster screening across formulations or candidates, and more confidence in the stability data behind a formulation or molecule-selection decision.

References

Leuenberger P, Ganscha S, Kahraman A, et al. Cell-wide analysis of protein thermal unfolding reveals determinants of thermostability. Science. 2017;355(6327):eaai7825. https://www.science.org/doi/10.1126/science.aai7825
Llowarch P, Usselmann L, Ivanov D, Holdgate GA. Thermal unfolding methods in drug discovery. Biophysics Reviews. 2023;4(2):021305. https://pmc.ncbi.nlm.nih.gov/articles/PMC10903397/
Lang BE, Cole KD. Measurement of the Thermal Unfolding of Proteins by Differential Scanning Calorimetry and Fluorescent Reporter Dyes. National Institute of Standards and Technology. https://www.nist.gov/publications/measurement-thermal-unfolding-proteins-differential-scanning-calorimetry-and
Thompson A. Employing thermal protein unfolding to search for new drugs. Scilight. 2023;2023(20):201108. https://pubs.aip.org/aip/sci/article/2023/20/201108/2890803/Employing-thermal-protein-unfolding-to-search-for

Further reading