A surface plasmon resonance (SPR) sensorgram is only useful once you can turn its curves into numbers. Running the assay wrong — an overloaded sensor surface, an analyte series that doesn’t bracket the affinity, incomplete regeneration between cycles — produces a sensorgram that looks plausible but fits a binding model badly, and any ka, kd or KD you pull from it will not reproduce. This guide walks through the assay procedure step by step, then works the rate-constant arithmetic through on real sensorgram numbers so you can check your own fit against a worked calculation rather than trusting whatever the analysis software reports.
What SPR actually measures, in one paragraph
SPR detects binding by refractive index, not by a label. One binding partner (the “ligand”) is immobilized on a thin gold film on a sensor chip; the other (the “analyte”) flows over it in solution. Polarized light reflecting off the back of the gold film excites surface plasmons at an angle that depends on the refractive index immediately at the surface. As analyte molecules bind the immobilized ligand, mass accumulates at the surface, the local refractive index shifts, and the resonance angle shifts with it. The instrument tracks that angle shift continuously and reports it in resonance units (RU) — roughly 1,000 RU corresponds to about 1 ng/mm² of bound mass for a typical protein, though the exact conversion depends on the analyte’s refractive index increment. Because the signal is proportional to bound mass and is recorded continuously, a single injection cycle generates a full time-course of binding and unbinding, which is what makes SPR a kinetics technique rather than just an endpoint-affinity technique.
Assay components and instrument setup
- Sensor chip — a gold-coated chip with a hydrogel matrix, most commonly a carboxymethylated dextran surface (the format most SPR platforms, including Cytiva’s Biacore series, sell as their standard research-grade chip). The dextran layer extends the usable binding surface off the bare gold and reduces nonspecific adsorption.
- Immobilization chemistry — amine coupling (EDC/NHS activation of the dextran carboxyl groups, followed by covalent capture of the ligand’s primary amines) is the default because it works for most proteins without further engineering. Where amine coupling would damage an active site or where reversible, oriented capture is preferred, capture-based immobilization is used instead — an anti-Fc antibody or Protein A/G surface for antibody ligands, or a streptavidin surface for a biotinylated ligand.
- Ligand vs. analyte assignment — the smaller, more stable, more available binding partner is usually immobilized as the ligand, and the other flowed as the analyte in a dilution series. This is a practical convention, not a physical requirement of the interaction itself.
- Reference (blank) flow cell — a second surface on the same chip, either left unmodified or immobilized with an irrelevant protein at a matched density. Every analyte response is subtracted against this reference channel before fitting; without it, bulk refractive index changes and nonspecific binding are indistinguishable from real signal.
- Running buffer — filtered and degassed, matched as closely as possible to the analyte dilution buffer to minimize bulk refractive index mismatch at the start and end of each injection. See the site’s buffer and solution preparation guide for general buffer-matching practice.
Running the assay: step-by-step procedure
- Dock and prime the sensor chip. Equilibrate the chip in running buffer until the baseline response is stable (flat, with no significant drift over several minutes) before starting immobilization.
- Activate the surface. Inject an EDC/NHS mixture (typically a short pulse, on the order of 5-10 minutes contact) to convert the dextran carboxyl groups to reactive esters.
- Immobilize the ligand. Dilute the ligand into a low-ionic-strength acetate buffer near or below its isoelectric point (this promotes electrostatic pre-concentration into the dextran matrix before covalent coupling) and inject until the target response level is reached. For kinetic experiments, aim for a low immobilization level — enough to give a workable signal-to-noise ratio, but well short of the surface’s theoretical maximum capacity, because a dense surface causes mass-transport limitation and rebinding artifacts (see the troubleshooting table below).
- Deactivate remaining reactive esters. Inject ethanolamine to cap unreacted activated groups on both the active and reference surfaces.
- Equilibrate baseline. Flow running buffer over both the active and reference channels until the response is stable; this becomes the zero point for the binding cycles that follow.
- Inject the analyte dilution series. Flow a series of analyte concentrations — typically five to seven points prepared by serial dilution (see the site’s serial dilution guide) spanning roughly 10-fold above and below the expected KD — over both channels, from lowest to highest concentration (or in randomized/single-cycle order, depending on the kinetic method used). Each injection has a defined association contact time, commonly in the range of 60-300 seconds.
- Monitor the dissociation phase. After each injection, running buffer alone flows over the surface and bound analyte dissociates. Dissociation is monitored for long enough to see a clear decay — often several minutes, and considerably longer for high-affinity interactions with slow off-rates, since an under-sampled dissociation phase is the single most common cause of an unreliable kd estimate.
- Regenerate the surface. Inject a short pulse of a regeneration solution — commonly a low-pH glycine-HCl buffer (roughly pH 1.5-2.5), or a high-salt or mild alkaline wash for interactions that don’t tolerate low pH — to strip remaining bound analyte without denaturing the immobilized ligand, returning the surface to baseline for the next cycle.
- Repeat across the full concentration series, plus buffer-only (“zero”) blank injections. The blank cycles are used for double referencing — subtracting both the reference-channel response and a buffer-only injection from every analyte response — which removes instrument drift and injection-to-injection noise before fitting.
Reading the sensorgram
A sensorgram plots response (RU) on the y-axis against time on the x-axis. A well-behaved 1:1 interaction produces a curve with three recognizable phases: a rising association curve that approaches a plateau while analyte is flowing, a step down in slope the instant buffer replaces analyte (the start of dissociation), and a decaying dissociation curve that falls back toward baseline. The steepness of the rise scales with analyte concentration; the shape of the decay does not (it should be the same curve shape, just present or absent, regardless of how much analyte was bound) — that concentration-independence of the dissociation phase’s shape is itself a useful sanity check on whether the model fits.
Deviations from that clean shape are diagnostic. A curve that fails to reach a plateau and instead keeps climbing roughly linearly suggests either the surface isn’t saturating within the contact time (slow kinetics or high KD) or the run is mass-transport limited. A dissociation phase that levels off above baseline instead of returning to it suggests either incomplete regeneration between cycles or a genuinely very slow off-rate. See the troubleshooting table further down for the fuller set of common failure patterns.
Worked calculation: deriving ka, kd, and KD from sensorgram data
The following example uses representative, illustrative rate constants to walk through the arithmetic — it is not drawn from a specific published interaction. Most kinetic evaluation software fits a 1:1 Langmuir binding model directly to the full sensorgram, but the two-step approach below (fit the dissociation phase first, then the association phase) is worth doing by hand at least once because it shows where each rate constant actually comes from, and gives you an independent cross-check between the two.
Step 1: fit the dissociation phase to get kd
During dissociation (no analyte flowing), bound complex decays as a single exponential: Rt = R0 × e−kdt, which rearranges to ln(R0/Rt) = kd × t. Suppose the response at the start of dissociation is R0 = 400 RU, and 300 seconds later it has fallen to Rt = 296 RU:
- ln(400 / 296) = ln(1.351) = 0.301
- kd = 0.301 / 300 s = 1.0 × 10−3 s−1
Step 2: fit the association phase across the concentration series to get ka
Each association curve at concentration [A] fits its own single exponential with an observed rate constant kobs, where kobs = ka[A] + kd. Fitting kobs at each concentration and plotting kobs against [A] gives a straight line whose slope is ka and whose y-intercept is kd — which should agree with the value from Step 1 as an internal consistency check.
| Analyte concentration | Observed kobs (s−1) |
|---|---|
| 6.25 nM | 1.63 × 10−3 |
| 12.5 nM | 2.25 × 10−3 |
| 25 nM | 3.50 × 10−3 |
| 50 nM | 6.00 × 10−3 |
| 100 nM | 1.10 × 10−2 |
Taking the slope between the two extreme points: (1.10 × 10−2 − 1.63 × 10−3) / (100 × 10−9 − 6.25 × 10−9 M) = 9.48 × 10−3 / 9.375 × 10−8 M ≈ 1.0 × 105 M−1s−1 = ka. Reading the intercept (kobs at [A] → 0) gives kd ≈ 1.0 × 10−3 s−1, matching Step 1.
Step 3: calculate KD
The equilibrium dissociation constant is the ratio of the two rate constants: KD = kd / ka = (1.0 × 10−3 s−1) / (1.0 × 105 M−1s−1) = 1.0 × 10−8 M = 10 nM. A smaller KD means tighter binding; a KD in the low-to-mid nanomolar range, as in this example, is a common result for a well-behaved protein-protein interaction and sits comfortably within the concentration range SPR resolves well kinetically.
Typical KD ranges by interaction class
These are broad, commonly cited ranges, not fixed thresholds — the achievable precision in any given assay depends on the actual ka/kd combination, not KD alone. Two interactions with the same KD but very different absolute rate constants can behave very differently on an SPR instrument.
| Interaction class | Typical KD range | Kinetic note |
|---|---|---|
| High-affinity, affinity-matured antibody-antigen (IgG) | ~100 pM – 10 nM | Often dissociation-limited; needs a long dissociation phase to resolve a slow kd accurately. |
| Typical protein-protein complex (receptor-ligand, enzyme-substrate mimic) | ~10 nM – 1 µM | The range SPR kinetic fitting resolves most reliably. |
| Moderate/weak protein or peptide-domain interactions | ~1 – 100 µM | Fast on- and off-rates approach the instrument’s temporal resolution; steady-state affinity analysis is often used instead of kinetic fitting. |
| Fragment-based screening hits, small-molecule leads | ~100 µM – low mM | Kinetics are usually too fast to resolve; KD is typically read from the equilibrium (steady-state) response instead. |
Selecting experimental parameters
| Parameter | What it trades off |
|---|---|
| Immobilization level (target Rmax) | Higher density gives a stronger signal but increases mass-transport limitation and analyte rebinding, both of which distort the fitted rate constants; kinetic experiments generally use the lowest density that still gives an adequate signal-to-noise ratio. |
| Flow rate | Higher flow rates (commonly in the tens of µL/min for kinetics) reduce mass-transport limitation by replenishing analyte at the surface faster than binding consumes it, at the cost of using more analyte per run. |
| Analyte concentration range | A series that doesn’t bracket the true KD — too far above or below it — produces a poorly constrained fit; a series spanning roughly 10-fold above and below the expected KD is the general starting point. |
| Association (contact) time | Too short and the curve doesn’t reach a fittable plateau for slow interactions; too long wastes analyte and instrument time once the curve has clearly plateaued. |
| Dissociation time | The most common source of an unreliable kd. Slow off-rates need proportionally longer monitoring — a dissociation phase that’s cut short simply because the curve “looks flat” will systematically overestimate kd. |
| Regeneration condition | Too mild leaves residual bound analyte, which shows up as a rising baseline across cycles; too harsh damages the immobilized ligand, which shows up as declining binding capacity across cycles. Both are visible by comparing Rmax across the run. |
Troubleshooting the sensorgram
| What you see | Likely cause | Fix |
|---|---|---|
| Association curve is concave/bowed rather than a clean rising exponential | Mass-transport limitation — analyte is being depleted at the surface faster than it diffuses in | Lower the immobilization density, increase flow rate, or both. |
| Baseline drifts upward across successive cycles | Incomplete regeneration, leaving residual bound analyte | Optimize regeneration buffer strength/pH or contact time; confirm Rmax returns to its original level after regeneration. |
| Binding capacity (Rmax) declines across successive cycles | Regeneration conditions are too harsh and are damaging the immobilized ligand | Use a milder regeneration condition, or switch to a capture-based (regenerable-surface) immobilization strategy. |
| Sharp spikes exactly at injection start/stop | Bulk refractive index mismatch between analyte and running buffer | Match the analyte dilution buffer to the running buffer as closely as possible; rely on reference-channel subtraction to remove the remainder. |
| Poor fit (high χ²) to a 1:1 Langmuir model despite a clean-looking curve | Heterogeneous ligand population, bivalent/avidity binding, or analyte rebinding at high surface density | Lower immobilization density to reduce rebinding and avidity effects; consider a bivalent-analyte or heterogeneous-ligand model if the mismatch persists. |
| Signal in the reference (blank) channel tracks the active channel | Nonspecific binding to the underlying dextran matrix rather than to the immobilized ligand | Include a buffer-only blank cycle for double referencing; consider a lower-nonspecific-binding chip surface if the problem persists across ligands. |
SPR compared with other binding assays
SPR’s distinguishing feature is that it is label-free and reports a continuous time-course, which is what makes true ka/kd kinetics possible rather than just an endpoint affinity value. An ELISA reports an endpoint colorimetric or fluorescent signal after a wash step, which is well suited to high-throughput screening and quantification but does not directly yield on- and off-rates. A western blot confirms that an interaction or a protein is present at all, but is not a quantitative binding assay in the SPR/ELISA sense. Where a study needs binding kinetics specifically — not just whether two molecules bind, but how fast they associate and how long the complex persists — SPR (or a comparable label-free kinetic method) is generally the more direct route than adapting an endpoint assay to estimate kinetics indirectly.
Frequently asked questions
What’s a “good” KD value?
There isn’t a universal threshold — what counts as tight or weak binding depends entirely on the biological question. A screening campaign looking for fragment hits may consider a KD in the hundreds of micromolar a meaningful starting point, while a therapeutic antibody program is typically optimizing well into the low nanomolar to picomolar range. Judge KD against the affinity that’s biologically relevant for the system being studied, not against a fixed number.
What’s the difference between kd and KD?
These are easy to confuse because they look almost identical in print. kd (lowercase, units of s−1) is the dissociation rate constant — how fast the complex falls apart. KD (uppercase subscript, units of molar concentration) is the equilibrium dissociation constant — the analyte concentration at which half the ligand sites are occupied at equilibrium, calculated as kd / ka.
How much protein does an SPR run need?
Far less than most solution-based binding assays, because the ligand is immobilized once on a reusable chip surface and the analyte flows past in microliter-scale injections rather than being consumed in bulk. Exact quantities depend heavily on the instrument, chip format and protein’s molecular weight, so check the specific platform’s documentation rather than assuming a fixed figure.
Can SPR distinguish two interactions with the same KD?
Yes, and this is one of the technique’s main advantages over an endpoint affinity measurement. Two interactions can share an identical KD while having very different absolute ka and kd values (fast-on/fast-off vs. slow-on/slow-off) — a distinction with real biological and pharmacological consequences (residence time in particular) that an equilibrium-only measurement cannot see.
Why does the dissociation phase matter so much for accuracy?
Because kd is usually the harder rate constant to pin down precisely, and KD is directly proportional to it. A dissociation phase that’s monitored for too short a time relative to the true off-rate will systematically overestimate kd, which propagates directly into an overestimated (too weak) KD even if the association-phase data and the ka fit are both good.
For related bench techniques on this site, see the guides on HPLC, LC-MS, gas chromatography, thin-layer chromatography, and flow cytometry, plus the practical basics of pipetting technique and tracking instrument data in a LIMS.
