Surface Plasmon Resonance
How it works
An SPR sensor chip carries a thin gold film; shining polarized light on it from the back at a specific angle excites a surface plasmon whose resonance angle is exquisitely sensitive to the refractive index within roughly 100–300 nm of the surface. One binding partner (the ligand) is immobilized on the chip — commonly via amine coupling to a carboxymethylated dextran matrix on a CM5 chip, or via a capture antibody, streptavidin-biotin or His-tag/NTA system when a more defined or reversible attachment is preferred — while the second partner (the analyte) is flowed over the surface in a microfluidic channel. As analyte binds, the local refractive index rises and the resonance angle shifts, plotted continuously as response (in resonance units, RU) against time — a sensorgram whose rising phase during injection is the association, and whose falling phase after the injection ends is the dissociation.
What you measure
Because the sensorgram is recorded continuously through both phases, fitting it to a binding model (usually 1:1 Langmuir binding, though more complex models exist for avidity or conformational-change interactions) resolves the association rate constant kon and dissociation rate constant koff separately, not just their ratio. The equilibrium dissociation constant KD = koff/kon follows, but two interactions with the same KD can behave very differently in practice — a slow kon/slow koff pair stays bound far longer once formed than a fast/fast pair with the same overall affinity, which matters for a therapeutic antibody's dosing interval as much as for its nominal affinity. The maximum response (Rmax) also confirms that the immobilized surface is behaving as expected and gives a check on active concentration.
A typical experiment
A run starts with chip selection and immobilization: ligand density is kept as low as practical, since a high density increases the risk of mass-transport limitation, where diffusion of analyte to the surface — rather than the binding reaction itself — becomes rate-limiting and distorts the measured kinetics. The analyte is then injected in a dilution series across several concentrations, each followed by a dissociation phase and a regeneration step that strips bound analyte without damaging the immobilized ligand, so the same surface can be reused for the next concentration. A full kinetic experiment — several concentrations, each with association, dissociation and regeneration — typically runs in a few hours and consumes only micrograms of material, since the flow cell volume is tiny.
Applications
Antibody engineering & affinity maturation. SPR kinetics guide affinity maturation campaigns by showing whether a mutation improves affinity through a faster on-rate, a slower off-rate, or both — information a single-point affinity measurement cannot provide.
Epitope mapping. Pairwise binding experiments — capturing an antigen with one antibody and testing whether a second antibody can still bind — build a competition matrix that groups antibodies by epitope bin without any labelling or structural data.
Target engagement in drug discovery. SPR confirms and quantifies that a small molecule or biologic engages its intended target, and is used alongside cell-based assays to link biochemical affinity to functional activity.
Therapeutic antibody characterization. Kinetic screening of candidate monoclonal antibodies — for example against inflammatory targets such as TNF-α in rheumatoid arthritis or Crohn's disease programmes — is a standard step before a lead candidate advances to further development.
Strengths & limitations
SPR's kinetic window is roughly kon up to about 106–107 M-1s-1 and KD down to low picomolar under good conditions, but very fast on-rates run into the same mass-transport limitation that low ligand density is meant to avoid, and very slow off-rates (very tight, long-lived complexes) can be hard to distinguish from a flat, non-dissociating baseline within a practical assay time. Because one partner must be immobilized, results can in principle differ from a free-solution measurement (e.g. ITC) if immobilization perturbs the binding site — a risk minimized by keeping ligand density low and, where possible, immobilizing the smaller or less functionally sensitive partner. Against that, SPR needs far less material than ITC, works over a very wide affinity range, and — uniquely among the platform's binding techniques — separates on- and off-rate rather than reporting affinity alone.
Frequently asked questions
How much material do I need to send?
Two very different amounts, because the two partners play different roles.
The immobilised partner, the ligand, is used sparingly: 20-100 µg at 10-50 µg/ml in coupling buffer covers immobilisation plus the pH scouting that precedes it. For capture-based surfaces even less.
The injected partner, the analyte, is the expensive one. You need a dilution series of five to eight concentrations spanning roughly 0.1 to 10 times the KD, and each injection consumes 50-150 µl depending on flow rate and contact time. A kinetic run in duplicate with a 100 nM top concentration and a 50 kDa analyte uses on the order of 100 µg. Push the KD to the micromolar range and the requirement climbs steeply — that is the calculation to do before ordering peptide synthesis.
Which of my two partners goes on the chip?
Response is proportional to the mass arriving at the surface, so the maximum signal is Rmax = (MW analyte / MW ligand) × immobilisation level × stoichiometry. Immobilising the small partner and injecting the large one gives a strong signal; the reverse gives a weak one. A 300 Da compound injected over a 50 kDa protein at 5000 RU yields an Rmax of about 30 RU, which is workable but leaves no room for a drifting baseline.
The second consideration is robustness. The immobilised partner sits on the chip through dozens of injection and regeneration cycles, so it should be the one that tolerates that. Antibodies, tagged proteins and biotinylated oligonucleotides are usually the better choice for the surface.
Where possible we run both orientations during method development; discrepancies between them are informative rather than annoying.
Does my protein have to be modified?
It depends on the coupling strategy, and there are several.
Amine coupling attaches the protein through its lysines onto the carboxymethyl dextran of a CM5 chip. No modification needed, but the attachment is random and can block the binding site or damage activity. It is the default and works for most proteins.
Capture avoids that. A His-tag on an NTA or anti-His surface, a biotin on streptavidin, an Fc on a protein A or anti-Fc surface, a GST on anti-GST — the protein is held in a defined orientation and the surface is rebuilt between cycles. This costs you a tag but generally gives cleaner kinetics.
So: no modification is strictly required, but a tag you already have will usually improve the experiment. Tell us what tags are on the construct before we design the assay.
My compound is dissolved in DMSO. Is that a problem?
DMSO is manageable up to about 5% and is routine at 1-2%, but it demands discipline. The bulk refractive index of DMSO is far higher than that of buffer, so a 0.1% mismatch between sample and running buffer produces a jump larger than the binding of a small molecule.
Two rules follow. The running buffer must contain exactly the same DMSO percentage as every sample, which means diluting your compound into DMSO-containing buffer rather than into buffer. And a solvent correction series — typically eight buffer standards bracketing the nominal percentage — is injected with each run so the residual bulk contribution can be subtracted.
Same logic applies to any additive: glycerol, high salt, detergent above what the running buffer contains. Whatever is in the sample must be in the running buffer.
What affinities and rates can you actually measure?
Association rates from about 10^3 to 10^7 M-1s-1 and dissociation rates from about 10^-5 to 0.5 s-1 are within reach on a T200, which puts KD anywhere from pM to high µM.
The limits are practical rather than theoretical. A kd faster than roughly 0.1 s-1 means the complex falls apart within the dissociation phase and only a steady-state affinity can be extracted, not separate rate constants. A kd slower than 10^-5 s-1 means nothing measurable comes off during the run, so the off-rate is a lower bound and regeneration becomes the difficult part. Very fast on-rates run into mass transport limitation, where the measured rate reflects diffusion to the surface rather than binding; low ligand density and high flow rate push that back, and the fit includes a mass transport term when it cannot be avoided.
What is in the report?
Referenced sensorgrams — sample surface minus reference surface minus blank injections — the fitted ka, kd and KD with their standard errors, Rmax, the fit residuals and chi², and where relevant the steady-state affinity as an independent estimate.
Two numbers are worth checking yourself. Rmax from the fit should be close to the Rmax predicted from the immobilisation level and molecular weights; a fitted Rmax far below prediction means much of the immobilised protein is inactive. And chi² should be small compared with the response range — a beautiful-looking curve with a systematic residual is a sign that a 1:1 model is being forced onto something more complicated, such as avidity from a bivalent analyte or a heterogeneous surface.
My analyte sticks to everything. Can that be fixed?
Often, and it is a normal part of method development rather than a failure.
The first tools are referencing: an activated-and-blocked reference flow cell, blank buffer injections subtracted from every cycle, and a control surface carrying an irrelevant protein of similar character. Non-specific binding that survives all three is real surface chemistry, and the levers are ionic strength (150 mM to 300-500 mM NaCl), a non-ionic surfactant at 0.05%, carrier proteins such as BSA, soluble carboxymethyl dextran to compete with the chip matrix, and pH. Very basic proteins bind the negatively charged dextran electrostatically; a low-charge chip such as C1 or a capture surface avoids the problem instead of fighting it.
Regeneration is the other half. The condition has to strip the analyte while leaving the ligand active, and finding it is a scouting exercise with acidic, basic and high-salt candidates. If nothing works, a single-cycle kinetics format skips regeneration entirely.
Can you work with membrane proteins, DNA, whole cells or serum?
Biotinylated DNA and RNA on a streptavidin chip are among the cleanest SPR systems there are, and small-molecule work on protein surfaces is routine.
Membrane proteins are done either as detergent-solubilised protein on a standard chip, with the detergent present in the running buffer above its CMC, or as liposomes and nanodiscs captured on an L1 chip. Both need method development time.
Crude supernatants and serum can be injected over a capture surface, provided the capture is specific and you accept a higher bulk-refractive-index correction; the instrument tolerates it better than most people expect. Whole cells are out of scope — the sensing field extends only a few hundred nanometres from the gold, and a cell is far larger than that.
How long does an SPR project take?
Machine time is short; method development is what fills the calendar.
Immobilisation with its pH scouting is half a day. A full kinetic titration, five to eight concentrations in duplicate with regeneration between cycles, runs three to six hours and can go overnight unattended. Once a method exists, a follow-up sample is a same-day measurement.
The unpredictable part is getting to a working surface: pH scouting, choice of coupling chemistry, regeneration scouting, and the checks that the surface is still active after ten cycles. For a well-behaved protein-antibody pair that is a day or two; for a fragile membrane protein with a sticky analyte it can be a couple of weeks. We normally split the quotation into a development phase and a measurement phase for exactly this reason.
What do you need from me before we start?
A short description of the system is worth more than the samples at this stage: what binds what, the expected affinity even as an order of magnitude, molecular weights, tags, the isoelectric point of the protein you want immobilised, and known stability limits.
The pI matters because amine coupling relies on pre-concentrating the protein electrostatically into the dextran, which requires a coupling buffer at a pH below the pI and low ionic strength. A protein with a pI of 4.5 cannot be pre-concentrated in the usual way and needs a different route.
With the samples: both partners, the running buffer composition you would like, and any specific requirement such as a divalent cation or a reducing agent. If the analyte is in DMSO, send the compound stock separately rather than pre-diluted.
Instruments
Biacore T200