Direct heat measurement of a ligand-target titration — extracting ΔH, ΔS, ΔG, Ka and stoichiometry from a single calorimetric experiment
An isothermal titration calorimeter is, at its core, an extraordinarily sensitive differential thermometer. Two identical coin-sized cells — one holding the macromolecule of interest in buffer, one holding matched buffer alone — are held inside an adiabatic jacket at a fixed, user-set temperature. Nothing is titrated yet; the instrument first has to prove it can hold both cells at exactly the same temperature before it can trust any deviation as a real binding signal.
Modern ITC instruments (MicroCal-style power-compensation design) use two matched coin-shaped cells, machined from a highly conductive alloy and suspended inside an adiabatic jacket that isolates them from room-temperature fluctuations. The sample cell contains the macromolecule (protein, nucleic acid, receptor fragment) dissolved in buffer; the reference cell contains identical buffer with no macromolecule.
Both cells sit under independent resistive heaters. A baseline "reference power" is continuously applied to the reference cell heater, and a feedback circuit adjusts the sample cell heater power moment-by-moment so that the temperature difference between the two cells stays at essentially zero. Before any ligand is added, this feedback power is flat and small — the two cells are thermally identical.
Without a reference cell, the instrument would have to compare the sample cell to the surrounding jacket, which drifts with lab temperature, air currents and the heat leaking from the syringe motor. Using a matched buffer-only reference cell cancels these common-mode effects almost perfectly, because both cells experience the same environmental noise. What remains is only the differential signal: whatever it costs, in heater power, to keep the sample cell exactly as warm as the reference cell.
This differential design is what allows detection of heat pulses on the order of a fraction of a microcalorie — roughly a million times smaller than the heat released by a lit match — riding on top of a milliwatt-scale baseline.
The dominant commercial platforms are the MicroCal PEAQ-ITC and iTC200 (Malvern Panalytical) and the TA Instruments Affinity ITC / Nano ITC. Cell volumes range from ~200 µL (low-volume instruments, ideal for precious protein) to ~1.4–2 mL (classical VP-ITC style cells, higher sensitivity per injection).
Sample preparation is unglamorous but decisive: macromolecule and ligand must be dissolved in exactly matched buffer (often from the same dialysis or gel-filtration step), thoroughly degassed to prevent bubble-induced noise spikes, and present at concentrations high enough to produce measurable heat but low enough to keep the titration in a fittable regime — a balance formalized later as the "c-value."
A precision syringe — itself acting as the stirrer paddle, spinning at a few hundred rpm to keep the cell contents mixed — delivers a small, exactly known volume of ligand solution into the sample cell. If the ligand binds the macromolecule, the binding reaction releases (exothermic) or absorbs (endothermic) heat, and the feedback circuit must inject or withdraw compensating power to keep ΔT at zero. That compensating power, plotted against time, is the raw ITC signal.
By instrument convention, an exothermic binding reaction (heat released, ΔH negative) means the sample cell would warm up relative to the reference cell — so the feedback circuit must remove compensating power, producing a downward deflection (spike) in the recorded power trace. An endothermic reaction (heat absorbed, ΔH positive) does the opposite: the circuit must add extra power, producing an upward deflection.
Each injection therefore looks like a sharp spike that rises (or falls) within a couple of seconds of ligand delivery, then decays back to baseline over 1–3 minutes as the cell re-equilibrates — this decay time is itself informative, since it reflects mixing and reaction kinetics, not just thermodynamics.
The entire purpose of the experiment is to measure the three quantities linked by the fundamental relationship of binding thermodynamics:
ΔG = ΔH − TΔS = −RT·lnKa
ΔG (free energy) sets the overall favorability and, through Ka, the affinity (KD = 1/Ka). ΔH (enthalpy) reflects the direct chemical bookkeeping of the interaction — hydrogen bonds, van der Waals contacts, electrostatics, and desolvation. ΔS (entropy) reflects the reorganization of the system — loss of conformational and translational freedom on binding, offset by the release of ordered water molecules from both surfaces (the hydrophobic effect).
Critically, ITC is the only common method that measures ΔH directly, from the actual calorimetric heat released — it does not need to assume how Ka varies with temperature, which is what indirect ("van't Hoff") enthalpy determinations require.
Not all the heat measured on the first few injections is binding heat. Diluting concentrated ligand stock into the cell buffer, mixing, and even the mechanical stirring itself all produce small heats of their own. A proper experiment always includes a control titration — ligand injected into buffer alone, with no macromolecule present — and this blank trace is subtracted from the experimental trace before any fitting is done, isolating the heat that comes specifically from binding.
No single injection tells the whole story. It is the pattern across the full series — spikes shrinking as free binding sites become scarce — that lets the underlying binding curve, the Wiseman isotherm, be reconstructed and fit. Whether that curve has a sharp, well-defined transition or a shallow, uninformative slope depends on one dimensionless number: the c-value.
Early in the titration, essentially every ligand molecule injected finds an empty binding site and binds — the heat per injection is close to the maximal per-mole binding enthalpy, scaled by injection volume. As injections accumulate, an increasing fraction of macromolecule is already occupied, so a shrinking fraction of each new pulse of ligand actually binds; the rest remains free in solution, contributing little heat. The result is the characteristic descending staircase of spike heights seen on the raw thermogram.
Integrating each spike gives one heat value per injection; plotting these values against the molar ratio of total ligand to total macromolecule traces out a sigmoidal binding curve — the Wiseman isotherm. Its steepness is governed by the dimensionless c-value:
c = n · Ka · [M]
where n is stoichiometry, Ka the association constant, and [M] the macromolecule concentration in the cell.
• c too low (<1): the curve is nearly flat and featureless — Ka and n become poorly constrained by the fit, though ΔH can sometimes still be estimated from the total heat. • c too high (>1000): the curve becomes a near-vertical step at the stoichiometric point — great for measuring n precisely, but Ka is barely constrained because binding is essentially stoichiometric (complete) at every point before saturation. • c in the sweet spot (roughly 10–100): the transition is gradual enough to define curvature precisely, giving confident, simultaneous estimates of n, Ka and ΔH from one fit.
Because c depends on concentration, experimentalists routinely dilute or concentrate their macromolecule stock specifically to land in this window before running the real titration.
Unlike surface plasmon resonance (SPR), which requires immobilizing one binding partner on a sensor chip, or fluorescence-based assays, which require a fluorophore label or reporter tag that can itself perturb binding, ITC measures both partners free in solution, in their native, unmodified state. There is no surface to introduce mass-transport artifacts or orientation bias, and no chemistry needed to attach a probe that might occlude the very binding site under study. This is a major reason ITC remains the gold-standard method for validating affinities measured by other techniques.
Toward the end of a well-designed titration, nearly every binding site is occupied. The remaining injections still deliver ligand, but there is almost nothing left to bind — the measured heat collapses to a small, nearly constant residual: the heat of dilution. Reaching this flat plateau is not wasted experimental time; it is essential, because it anchors the mathematical fit and confirms the reaction has genuinely gone to completion.
As occupancy approaches 100%, the binding-derived component of each pulse shrinks toward zero, and what remains is dominated by the heat of diluting concentrated ligand stock into the cell — a small, roughly constant value across the final several injections. A good titration design deliberately continues injecting well past the point of visual saturation, typically adding ligand to 2–3× the expected stoichiometric molar ratio, so that this flat baseline is well characterized rather than assumed.
The simplest and most common model is a single set of independent, identical binding sites, described by one Ka, one ΔH and one n. Many systems are more complex: sequential binding models allow different sites to have different affinities and enthalpies (common in multi-subunit receptors), and competitive or linked-equilibria models are needed when a third species (proton, metal ion, cofactor) is coupled to the binding event. Choosing the wrong model — for instance forcing a two-site system into a one-site fit — typically shows up as systematic residuals: the fitted curve consistently over- or under-shoots the data in a patterned way rather than scattering randomly around it.
Experimentalists tune three levers before running the real titration: macromolecule concentration in the cell (to hit a target c-value), ligand concentration in the syringe (typically 10–20× the cell concentration, so that saturation is reached within the syringe volume), and the injection schedule itself — often a small first injection (discarded from analysis, since it is distorted by diffusion at the needle tip during equilibration) followed by 15–25 regular pulses. For very high-affinity, high-c systems, a "low-c" strategy — displacement titration or working at much lower macromolecule concentration — is sometimes used deliberately to still resolve Ka.
The raw thermogram is only an intermediate product. The actual measurement comes from integrating the area under each heat pulse to get one number per injection — heat released or absorbed, in microcalories — then fitting the full set of these numbers, plotted against molar ratio, to a binding model. A single nonlinear regression on this curve returns stoichiometry, affinity and enthalpy simultaneously; entropy and free energy follow by simple algebra.
Each raw power-vs-time spike is numerically integrated (area under the curve, corrected against the pre-injection baseline) to yield a single heat value, Qi, in microcalories, for injection i. Plotting Qi (normalized per mole of injected ligand) against the cumulative molar ratio [ligand]total/[macromolecule]total produces the sigmoidal Wiseman isotherm introduced earlier. A nonlinear least-squares fit of this curve to the chosen binding model (most simply, the standard single-site binding isotherm derived from mass-action and mass-balance equations) returns best-fit values for n, Ka and ΔH simultaneously, along with their statistical uncertainties — all from one titration, without needing separate experiments at multiple temperatures.
With ΔH and Ka in hand, ΔG follows immediately from ΔG = −RT·lnKa, and ΔS is then obtained by rearranging the master equation: ΔS = (ΔH − ΔG)/T. This complete thermodynamic signature — how much of the binding free energy comes from enthalpy versus entropy — is often more informative for drug design than affinity alone, because two compounds with identical KD can have opposite thermodynamic "fingerprints."
In fragment-based drug discovery and lead optimization, chemists use this signature to classify binders: "enthalpy-driven" binders make specific, directional interactions — hydrogen bonds, precisely shaped van der Waals contacts — that tend to be more selective and more optimizable through medicinal chemistry. "Entropy-driven" binders often gain affinity mainly from burying large hydrophobic surface area (displacing ordered water), which is easier to achieve early but harder to make truly selective. A frequent optimization trajectory is watching a fragment series shift from entropy-driven to enthalpy-driven binding as specific polar contacts are engineered in — a phenomenon linked to the broader "enthalpy-entropy compensation" seen across many binding series, where improving one term often partially erodes the other.
ITC is slower and requires more material than surface- or fluorescence-based methods, so in practice it is used alongside, not instead of, other binding techniques. The table below places it among the most common binding-characterization methods used in structural biology and drug discovery.
| Product | Indication | Trial Design | Key Result |
|---|---|---|---|
| Isothermal Titration Calorimetry | Any binding pair, free in solution | Direct heat measurement of the binding reaction via power compensation | Only method giving n, Ka, ΔH, ΔS, ΔG in one experiment — fully label-free |
| Surface Plasmon Resonance (SPR) | Immobilized ligand + flowed analyte | Refractive-index shift at a sensor surface tracks bound mass in real time | Real-time on/off kinetics (kon, koff), very low sample consumption |
| NMR Chemical Shift Perturbation | Isotope-labeled protein + small-molecule ligand | Ligand binding perturbs the local chemical environment, shifting HSQC peaks | Maps the binding site residue-by-residue; ideal for weak, fragment-sized binders |
| Fluorescence Polarization (FP) | Fluorescently labeled tracer + competitor | Bound (large, slow-tumbling) vs. free tracer show different polarization | High-throughput, low reagent cost, plate-based screening format |