A binding curve that rises unusually steeply is often labelled cooperative. But the Hill coefficient fitted to that curve is not automatically the number of binding sites on a protein. It describes the steepness of a selected mathematical response model, not a direct molecular headcount.
This guide works through Hill equations with exponents one and two, reconstructs the logarithmic Hill-plot slope, and tests why similar-looking binding curves can arise from different mechanisms. The aim is to interpret cooperativity without inventing molecular detail that the experiment did not measure.
eduKateSG · The Gold Standard of Biochemistry
Choose your reading route
Understand the curve · Chapters 1–3: slope, occupancy and worked examples.
Test the interpretation · Chapters 4–6: site count, heterogeneity and experiment design.
Check and transfer · Chapters 7–8: FAQs, linked material and sources.
Chapter index
Curve and calculation · Chapters 1–3
Interpretation and experimental design · Chapters 4–6
Questions and sources · Chapters 7–8
A protein may bind a ligand over a narrow concentration range, producing a sigmoidal occupancy or activity curve. Students sometimes see such a curve, fit a Hill coefficient near two and conclude that the protein must contain exactly two binding sites. This is a tempting but unreliable interpretation.
The Hill equation is a compact empirical description of the concentration dependence of a fractional response. Its exponent captures steepness in a defined model. Depending on what was measured, a fitted slope may reflect cooperative binding, conformational change, heterogeneous populations, detection amplification or other kinetic and assay processes.
The IUPAC Gold Book defines a Hill plot in terms of the logarithm of occupancy odds plotted against ligand concentration on a logarithmic scale. In a noncooperative ideal binding system the slope is one; a slope greater than one can indicate cooperative behaviour under suitable conditions. The original thermodynamic and stoichiometric story still needs separate evidence.
Research reviews on the uses and misuses of Hill fits stress that the parameter should not be equated uncritically with the number of occupied sites or subunits. A receptor with more than two sites can show a Hill slope near two, and not every sigmoidal measured response directly represents binding-site occupancy.
The central question is therefore what was measured: fraction of physical ligand sites occupied, enzyme turnover rate, fluorescence output, transcriptional response or product accumulation? A beautiful mathematical fit to one observable cannot give a unique mechanistic account of a different quantity.
The Gold Standard keeps the Hill coefficient, binding stoichiometry, functional output and assay conditions separately identified. We will first calculate exactly what the empirical curve predicts, then examine which conclusions it cannot establish.
One common two-parameter Hill form is θ=Lⁿ/(Kⁿ+Lⁿ), where θ is a dimensionless fractional occupancy between zero and one, L is a ligand concentration, K is the concentration at half-saturation in this parametrisation, and n is the fitted Hill exponent. Here L and K must use the same units.
Choose K=4 micromolar and consider n=1 versus n=2. With n=1 and L=2 µM, θ=2/(4+2)=0.3333. With n=2, the same concentration produces θ=2²/(4²+2²)=4/20=0.2000.
At L=4 µM both models predict θ=0.5 by construction. At L=8 µM the n=1 model gives θ=8/12≈0.6667, whereas n=2 gives 64/(16+64)=0.8000. The higher exponent produces a steeper transition around K, not a different half-saturation concentration in this chosen comparison.
The arithmetic reveals what the exponent controls. Below K the n=2 curve lies further toward zero than n=1; above K it rises further toward one. Yet both reach the same half-saturation value at L=K.
Nothing in these numbers says how many binding sites exist on a real protein. The toy equation itself does not specify a physically detailed sequence of intermediate binding states, the concentrations of distinct protein conformations or the kinetics of ligand association and dissociation.
Use this model as a summary of an observed, properly normalised binding curve. To infer molecular site counts, independent binding stoichiometry or structural measurements are necessary.
| Ligand L | n=1 occupancy | n=2 occupancy |
|---|---|---|
| 1 µM | 0.2000 | ≈0.0588 |
| 2 µM | ≈0.3333 | 0.2000 |
| 4 µM | 0.5000 | 0.5000 |
| 8 µM | ≈0.6667 | 0.8000 |
| 16 µM | 0.8000 | ≈0.9412 |
Rearrange the Hill equation: θ/(1−θ)=(L/K)ⁿ. Taking logarithms gives log[θ/(1−θ)]=n log L−n log K, when L and K are compared in consistent units or expressed as a dimensionless ratio.
The transformation converts the synthetic exact Hill model into a straight line with slope n. For n=2, at L=2 µM the fractional occupancy is 0.2, so occupancy odds are θ/(1−θ)=0.2/0.8=0.25. At L=8 µM occupancy is 0.8, giving odds 0.8/0.2=4.0.
The odds change by a factor of 4/0.25=16 while ligand concentration changes by a factor of 8/2=4. Thus the logarithmic slope is log(16)/log(4)=2, exactly the model’s exponent.
This is evidence of steepness in the constructed model. It is not a measurement of two separate binding events. A binding experiment containing multiple ligand-bound species can generate a curved Hill plot whose local slope changes with concentration, and a single globally fitted n may hide that variation.
At very low or high occupancy the log odds can become unstable because values approach zero or one. Noise and uncertainty in the fitted top and bottom plateaus also affect the transformed plot. A straight line drawn through a limited concentration range can be misleading if the rest of the response was never observed.
Modern analysis can fit a suitable response model directly using documented errors. A Hill plot remains a valuable diagnostic for shape, but a mechanistic claim requires independent evidence of the underlying protein and ligand states.
| Toy point | L | θ | Occupancy odds |
|---|---|---|---|
| Below half occupancy | 2 µM | 0.20 | 0.25 |
| Above half occupancy | 8 µM | 0.80 | 4.00 |
| Ratio | 4-fold L increase | — | 16-fold odds increase |
A protein with several ligand sites may show positive cooperativity: binding of a molecule may change the favourability of subsequent binding events. Yet the shape of an observed bulk response does not generally reveal the complete sequence of microscopic equilibrium constants or the total site stoichiometry.
For example, haemoglobin has four oxygen-binding haem sites while its commonly discussed cooperative Hill slope is below four, often around the mid-twos under suitable conditions. The number of binding sites and the empirical slope are distinct concepts. Quoting n=2.4 for a four-site molecule is not a contradiction.
Weiss’s methodological review argues that the Hill equation generally fails to represent a detailed, physically realistic multi-site mechanism and that site-number interpretation is justified only under restrictive conditions such as marked positive cooperativity. A more mechanistic scheme can fit the same overall response with different microscopic assumptions.
A fitted n>1 can support a discussion of steep or cooperative behaviour when occupancy has been measured properly, but ‘two sites’ is not the logical consequence of a fitted n=2. In some constrained equilibrium settings a Hill slope may inform bounds on interacting sites; that does not make the rounded exponent a literal molecular census.
If the scientific question is how many ligand molecules a protein can bind, seek a saturation binding capacity with a trustworthy protein-concentration basis, direct stoichiometry, structural data or other independent evidence. An arbitrary fluorescence normalisation to 100% at the highest dose is not enough to establish absolute occupied-site counts.
Similarly, a catalytic enzyme can have substrate-linked allostery without the measured enzyme reaction rate being identical to fractional ligand occupancy. To connect a Hill kinetic response to mechanism, investigate substrate binding, conformational states and catalysis as separate experimental layers.
The practical rule is short: a Hill coefficient describes curve steepness under a defined fit. Counting actual sites requires a different type of evidence.
An apparent Hill exponent below one is sometimes described as negative cooperativity. One ligand-binding event can reduce the favourability of later binding events in an appropriate multi-site system. But the same shallow overall response can arise from distinct, noninteracting site classes with different affinities or from heterogeneity between molecules.
Imagine two groups of protein molecules, each individually showing a simple noncooperative saturation curve, but one group has a much lower K than the other. The population average begins responding at low ligand concentration and continues changing over a broad range as the weaker-affinity group catches up. The combined curve can look shallower than either component.
The fitted parameter can also be distorted by incomplete baseline subtraction, ligand depletion, insufficient equilibration, fluorescence quenching or incorrectly fixed upper and lower plateaus. An assay artefact that broadens the apparent response is not molecular negative cooperativity.
Research on ligand-protein Hill coefficients shows that distinguishing negative cooperativity from mixtures of nonidentical binding sites requires additional assumptions or experiments. It is therefore not enough to label a fitted n=0.7 as proof that one particular binding event weakens the next.
Experimental context matters. A response measured at steady binding equilibrium is not the same as a reaction that is still approaching equilibrium while the instrument collects data. Kinetic trapping or slow transitions can change the dose-response curve observed over a fixed time.
A strong research report names the possible mechanisms and asks what independent perturbation would differentiate them: changes in protein concentration, ligand competition, mutant constructs, kinetic time courses or directly resolved binding states.
The most useful outcome is not choosing the most sophisticated interpretation by default. It is acknowledging what the curve alone does and does not distinguish.
Begin with an explicitly defined signal. If θ is claimed to be site occupancy, show how the measurement was calibrated to the fraction of occupied sites. If the measured response is enzyme velocity or fluorescence, report the actual quantity and avoid silently renaming it occupancy.
Collect ligand concentrations below, around and above the anticipated half-response value, within the validated solubility and detector range. A concentration series confined to the near-flat upper plateau may not reliably estimate either the half-response concentration or the exponent.
Control the free ligand concentration. When protein concentration is comparable to added ligand concentration, substantial ligand depletion can make total added ligand differ from free ligand in solution. Fitting free-ligand formulas directly to total-ligand concentrations can distort inferred parameters.
Check protein preparation, purity, temperature, pH, ionic strength, incubation time and equilibration. A conformationally heterogeneous preparation can broaden an aggregate curve independently of genuine interactions among sites on a single protein.
Test detector response with matched standards and appropriate blanks. If high ligand concentrations quench fluorescence or cause turbidity, the signal may cease to be proportional to the intended molecular quantity. The reported Hill coefficient would then describe the instrument’s combined output rather than a clean binding relationship.
Fit an appropriate model with replicates and uncertainties. Examine residuals, verify that apparent plateaus are supported by data, and report whether the curve was binding occupancy, enzyme reaction rate, receptor activation or another readout. A four-parameter logistic fit has a different set of estimated quantities from the simple two-parameter occupancy equation used in this article.
Finally, compare at least one mechanistically explicit binding model when the scientific objective is molecular interpretation. A Hill fit may remain useful for concise reporting, but it should not be allowed to replace the hard work of model discrimination.
If n=2, does the protein have two ligand sites? No. A fitted exponent near two describes a steepness pattern and does not uniquely determine binding-site stoichiometry.
Does n>1 always prove true positive cooperativity? It can be consistent with cooperative response under suitable binding assumptions, but linked readout and population effects can also produce steep curves. Independent mechanism tests remain important.
Does n=1 prove a protein has only one binding site? No. Multiple independent equivalent binding sites can produce noncooperative occupancy behaviour with Hill slope near one.
Does n<1 prove negative cooperativity? No. Heterogeneous affinities, nonequilibrium measurements and assay distortions are alternative explanations.
What is K in θ=Lⁿ/(Kⁿ+Lⁿ)? Under this parametrisation K is the concentration at which the fitted response reaches θ=0.5. It is not automatically a microscopic dissociation constant.
Can the Hill exponent change with the concentration range fitted? Yes. A real multi-state binding curve may not be an exact Hill equation, and the apparent slope can depend on which portion of the data is measured and fitted.
Should occupancy be plotted against total added ligand? Only if it is a reasonable approximation to free ligand or if the model explicitly accounts for depletion and binding mass balance.
Can an enzyme’s measured activity curve be called a binding isotherm? Not automatically. A functional response may depend on catalytic kinetics and coupling between binding and activity.
How should a fitted coefficient be reported? Alongside the mathematical equation, readout definition, concentration units, fitting method, uncertainty, sampling range and any independent evidence of molecular stoichiometry.
The most important lesson is the separation of a numerical model parameter from a physical count of molecular objects. A Hill slope is a concise description of observed response steepness under an explicit fit; it is not a photograph of the protein’s occupied ligand pockets.
Recompute the n=1 and n=2 toy curves at L=2 and 8 µM. Under n=2 the odds ratio rises by sixteen as concentration rises fourfold. Explain why that correctly establishes the model’s logarithmic slope but does not reveal how many protein subunits were present.
Connected reading includes The Gold Standard of Biochemistry, Substrate Inhibition and Nonlinear Rates, Active Enzyme Fraction and Turnover Number, and The Gold Standard of Chemical Biology. Broader model reasoning can be continued through Learning Transfers into Life and How X Works.
Authoritative sources reviewed on 23 October 2026: IUPAC Gold Book: Hill Plot, Weiss: The Hill Equation Revisited—Uses and Misuses, Prinz: Hill Coefficients, Dose–Response Curves and Allosteric Mechanisms, and Hill Coefficient and Negative Cooperativity.
All n, K and ligand values in the tables are invented examples. They are not kinetic parameters, stoichiometric site counts or binding measurements from a particular experimental protein.