Differential Scanning Fluorimetry

Ten microlitres per capillary on the Prometheus, and the same on Tycho. Typical working concentration is 0.2-5 mg/ml, though the instrument reads intrinsic fluorescence over an enormous dynamic range and samples from a few µg/ml to well over 100 mg/ml have been measured.

For a 30 kDa protein at 1 mg/ml that is 10 µg per condition. A 48-condition buffer screen therefore costs about half a milligram, which is the reason this method has largely replaced DSC for screening work.

Send 20-30% more than the arithmetic suggests: capillaries are filled by capillary action from a small drop, and the dead volume in the tube is not negligible at these scales.

It helps a great deal. The measurement follows the ratio of intrinsic fluorescence at 350 and 330 nm, which shifts as buried aromatic residues become solvent-exposed. Tryptophan dominates that signal; tyrosine contributes but gives a smaller amplitude, so a protein with tyrosines and no tryptophan usually still works, with a noisier curve.

A protein with no aromatic residues at all cannot be measured by intrinsic fluorescence. Options then are the back-reflection channel, which detects aggregation by light scattering without needing any fluorophore and often gives a usable transition on its own, or classical DSF with SYPRO Orange in a qPCR instrument.

Where the tryptophans sit matters as much as how many there are. A single Trp on the surface may barely change environment on unfolding, while a buried one in a small domain gives a textbook transition.

The melting temperature Tm from the inflection of the 350/330 ratio, and where several domains unfold separately, one Tm per transition. The onset temperature Ti, where the ratio first departs from baseline, is reported alongside it and is often the more relevant number for formulation work, since it marks where the protein begins to lose structure rather than where half of it is gone.

The back-reflection channel gives the aggregation onset temperature Tagg independently, on the same capillary at the same time.

Reproducibility on Tm is 0.1-0.3 °C between replicate capillaries, which is what makes small ΔTm values from ligand binding or buffer changes worth taking seriously. Absolute Tm values depend on scan rate and should be compared within a run rather than against a literature number obtained elsewhere.

Almost anything. Label-free detection is what makes nanoDSF tolerant where dye-based DSF is not: detergents above their CMC, high salt, up to 6 M guanidinium or 8 M urea, DMSO at several percent, glycerol, reducing agents, nucleotides and metals are all workable.

That tolerance is the main practical reason to choose it over SYPRO Orange thermal shift. SYPRO binds hydrophobic surfaces, so detergent micelles compete with it and membrane proteins usually give no signal at all.

Two things do interfere. Compounds that fluoresce or absorb strongly around 330-350 nm distort the ratio, so highly coloured or aromatic ligands at high concentration need a buffer-only control in the same capillary set. And anything that makes the protein precipitate before the scan starts gives a curve with no transition, only scattering.

To find one, yes. A ligand that binds the folded state stabilises it and raises Tm, and a ΔTm of 1-2 °C above the noise of the replicates is a real signal. Fragment screening and ligand confirmation are among the most common uses.

To measure affinity, no, not directly. ΔTm depends on the affinity at the melting temperature, on the unfolding enthalpy and on the ligand concentration, and the relation is not monotonic in any convenient way. A nanomolar binder and a micromolar binder can produce similar shifts, and a genuine binder that binds folded and unfolded states equally produces no shift at all. Treat the shift as a yes/no with a rough ranking, and go to ITC or SPR for a number.

Isothermal chemical denaturation on the same instrument does give ΔG of unfolding, and titrating ligand into that experiment is one route to a thermodynamic affinity, at the price of a much longer measurement.

Usually one of three things. The protein has two domains that unfold at different temperatures, which is real structural information. The preparation contains two species — a truncated form, a partly misfolded population, a mixture of oligomeric states. Or a bound cofactor dissociates at a temperature below the unfolding of the apo protein, which produces a small early transition that is easy to over-read.

Comparing the fluorescence ratio curve with the scattering curve helps. A transition visible in the ratio but not in scattering is unfolding without immediate aggregation; a transition visible only in scattering is aggregation of already-unfolded material.

Running the same sample at two concentrations separates concentration-dependent processes such as oligomer dissociation from concentration-independent unfolding.

Thermal or conformational stability is how hot the protein has to get before it unfolds — the Tm. Colloidal stability is how readily the molecules stick to one another, whether folded or not, and shows up as the aggregation onset Tagg or as growth in scattering at constant temperature.

They come apart more often than people expect. A buffer change can raise Tm by 4 °C and simultaneously lower Tagg, because the same ionic conditions that stabilise the fold also reduce the electrostatic repulsion that keeps molecules apart. If you are choosing a storage buffer, the one that keeps the protein in solution for six months at 4 °C is not always the one with the highest Tm.

Measuring both channels on the same capillary is why the technique is used for formulation rather than just for melting points.

Forty-eight capillaries in parallel on the Prometheus, twenty-four on the smaller format. A ramp from 20 to 95 °C at 1 °C/min takes about 75 minutes for all of them at once; at 4 °C/min it drops to twenty minutes, at some cost in resolution between close transitions.

Tycho is faster still — a fixed 35-95 °C ramp in three minutes on one sample — and is the right instrument for a quick identity and folding check on a purification fraction rather than for a careful comparison.

A 96-condition screen is therefore two runs and half a day including preparation. What takes the time is pipetting the conditions, not the scanning.

For a single melting curve: 50 µl of protein at 0.5-2 mg/ml plus its buffer. For a screen: concentrated protein stock and the list of conditions you want, or the conditions themselves if they are not standard. Working at 2-4 times the final concentration lets us dilute into each condition rather than dialysing 48 times.

Tell us the concentration, the molecular weight, the number of tryptophans and tyrosines, whether any ligand or cofactor is present, and whether the protein has ever been frozen. If a specific comparison is the point — two constructs, two batches, apo against holo — say which comparison, because that determines how the capillaries are arranged and which replicates matter.

Samples should be centrifuged before loading; visible precipitate in a capillary gives a scattering trace and no melting curve.

Yes, and it is the technique users most often end up running on their own. Loading capillaries takes a little practice and everything after that is software.

Two sessions are usually enough: one to load and run, one to go through the analysis and the failure modes — a ratio curve that starts already rising, a transition cut off at the top of the ramp, replicates that disagree by more than half a degree. After that, booking instrument time directly is the normal arrangement for groups doing regular construct or buffer screening.

Occasional users are better served by sending samples, particularly for a first experiment where the choice of conditions matters more than the pipetting.