What question does this answer? How strong is the binding between the QTY variant and its native ligand — and how does it compare to the native receptor? After purification we have a folded, detergent-free protein. But folding is not function. MST is the step that turns “this protein looks right” into a number: a dissociation constant in nanomolar units, measured in free solution and compared against the literature value for the native receptor.
The principle: thermophoresis in a tiny temperature gradient
MST uses infrared laser light to create a microscopic temperature gradient inside a capillary. Molecules migrate along that gradient — this directed motion is thermophoresis — and its magnitude and direction depend on the molecule's size, charge and hydration shell. The instrument does not measure mass or heat; it measures the fluorescence distribution in the capillary before, during and after the infrared heating pulse. When the labelled molecule binds a partner, its thermophoretic behaviour changes, and that change is read out as a shift in the fluorescence ratio.
The decisive advantage for our system is that MST works in free solution with no immobilisation. Chemokine receptors are notoriously difficult to surface-tether, and QTY variants are membrane-protein derivatives whose behaviour is easily disturbed by a solid support. Requiring no surface removes an entire class of artefact and lets us titrate a genuine solution-phase interaction. Because only one partner must be fluorescent, we label the receptor once and titrate everything else unlabelled — a versatile design that lets one labelled batch answer many ligand questions.
Assay design: hold the label low and constant
The design principle is simple but non-negotiable: keep the fluorescent partner at a concentration far below the expected Kd, and vary the unlabelled partner widely. We hold 5 nM NT647-labelled receptor constant and titrate unlabelled ligand across nM-range gradients. Holding the label « Kd » means the fraction bound is governed almost entirely by the ligand concentration, and the fit returns a clean Kd rather than a mixture of Kd and stoichometric depletion. If the labelled partner were near or above Kd, both partners would be depleted simultaneously and the midpoint of the curve would no longer equal the true affinity.
Each ligand was titrated over a range chosen to span the expected Kd with points both well below and well above it — the hallmark of a well-planned titration:
| Ligand | Concentration range (nM) | Role |
|---|---|---|
| Rantes / CCL5 | 0.0916–3000 | Cognate ligand |
| SDF1α / CXCL12 | 0.0916–3000 | Cognate ligand |
| Insulin | 0.153–5000 | Non-cognate negative control |
| CCL28 / CXCL11 | 0.0651–2000 | Cognate-family ligand |
| CCL27 | 0.012–400 | Cognate ligand |
| gp41-120 | 0.0153–500 | Non-cognate negative control |
Instrument settings and why each one
- Monolith NT.115 pico. The pico detector gives the sensitivity needed for a 5 nM NT647-labelled sample; without it the low label concentration would be buried in noise.
- 80% MST power. Sets the infrared heating. High power maximises the thermophoretic signal and the size of the binding-induced change, at the cost of a larger temperature jump.
- 15% LED power. Sets excitation. Kept low to avoid photobleaching a precious labelled sample during repeated runs.
- Premium capillaries. Surface-treated to suppress non-specific adsorption of protein to the capillary wall, the single most common source of flat, uninterpretable curves.
- 25 °C. Physiologically relevant and constant — thermophoresis itself is temperature-dependent, so the ambient temperature must be fixed.
- 1× PBS + 5 mM DTT. Matches the buffer used throughout the pipeline and, crucially, keeps the cysteine-rich receptor reduced; the four variants carry 12/9/10/14 cysteines, so a reducing environment is mandatory to avoid disulfide-driven aggregation and avidity artefacts.
Negative controls: proving it is not just “everything sticks”
A binding curve alone is not evidence of specificity. A sticky, partially unfolded protein will produce a “bound” signal against almost anything. We therefore run non-cognate ligands — insulin and gp41-120 — as negative controls. These are unrelated proteins of comparable size and should not engage the receptor. If the cognate ligands give sigmoidal, saturable curves while insulin and gp41-120 give flat responses, we have shown that the interaction depends on molecular recognition rather than generic adhesion. This is the same logic that guards the whole pipeline: each measurement must be able to fail, and the negative control is what makes the positive result meaningful.
Fitting and confidence
The data are fitted to a standard 1:1 binding model, and the Kd is read from the midpoint of the dose–response curve — the ligand concentration at which half the receptor is bound. The bottom and top plateaus define the signal window, and a healthy curve spans a large fraction of it. Confidence comes from replicates: independent titrations should return Kd values that agree with each other and with literature values for the native receptor–ligand pair. It is that agreement, not the elegance of a single curve, that lets us claim the QTY variant retains native-like binding.
Troubleshooting and pitfalls
- Aggregation. Aggregated protein scatters far-red light and distorts the fluorescence readout, producing erratic or steeply dropping curves. Keep DTT in the buffer, keep the sample cold between runs, and check homogeneity by SEC.
- Capillary adsorption. Protein sticking to the capillary wall depletes the free pool and flattens the curve. Use premium (treated) capillaries and include a low level of carrier detergent or BSA only if validated not to perturb binding.
- Inner-filter effect. At the high end of a titration, unlabelled ligand can absorb or scatter enough light to distort the fluorescence signal independent of binding. Keep the upper range justified and watch the plateau — a plateau that drifts downward at high ligand often signals inner-filter, not weak binding.
- Binding to the tube. Hydrophobic ligands and proteins can adsorb to plastic, lowering the effective free concentration and shifting the apparent Kd. Use low-binding tubes and pre-rinse with buffer containing carrier.
- Labelled partner too concentrated. If the constant fluorescent concentration is not « Kd, the fitted midpoint overestimates affinity. Validate that the on-plate concentration is truly ~5 nM.
- Temperature drift. Ambient fluctuations move the thermophoresis baseline; let the instrument equilibrate at 25 °C and run controls on the same plate.
Done well, MST delivers the central functional claim of the QTY case study: a detergent-free variant that binds its native ligand with an affinity indistinguishable from the native receptor, proven in free solution, with negative controls closing the specificity gap.
References
- Wienken CJ, Baaske P, Rothbauer U, Braun D, Duhr S. Protein-binding assays in biological liquids using microscale thermophoresis. Nat Commun. 2010;1:100. doi:10.1038/ncomms1093
- Seidel SAI, Dijkman PM, Lea WA, et al. Microscale thermophoresis quantifies biomolecular interactions under previously challenging conditions. Methods. 2013;59(3):301–315. doi:10.1016/j.ymeth.2012.12.005
- Jerabek-Willemsen M, André T, Wanner R, et al. MicroScale Thermophoresis: interaction analysis and beyond. J Mol Struct. 2014;1077:101–113. doi:10.1016/j.molstruc.2014.03.009
- Zhang S, et al. QTY code enables design of detergent-free chemokine receptors that retain ligand-binding activities. Proc Natl Acad Sci USA. 2018;115(37):E8652–E8659. doi:10.1073/pnas.1811031115
- Brinkley M. A brief survey of methods for preparing protein conjugates with dyes, haptens, and cross-linking reagents. Bioconjug Chem. 1992;3(1):2–13. doi:10.1021/bc00013a001
- Folta-Stogniew E. Oligomeric states of proteins determined by size-exclusion chromatography coupled with light scattering, absorbance, and refractive index detectors. Methods Mol Biol. 2006;328:97–112. doi:10.1385/1-59745-026-x:97
- Pace CN, Vajdos F, Fee L, Grimsley G, Gray T. How to measure and predict the molar absorption coefficient of a protein. Protein Sci. 1995;4(11):2411–2423. doi:10.1002/pro.5560041120