Exclude experimental artefacts
Instead of applying different models in search of the best fit through the data, first, try to optimize the experimental conditions. There are three main groups of experimental artefacts to consider:
- the system used, including the instrument, the measuring cell, the matrix on which the ligand is bound, and the buffers.
- the ligand. Is it pure and homogeneous? Does the immobilization destroy epitopes? How does it perform after regeneration?
- the analyte. Is it reacting only with the ligand? Is it a monomer?
Several part of the system can give rise to some artefacts in the sensorgrams. Knowing them can help you to avoid these.
SPR is temperature dependent because the refractive index of a liquid is temperature dependent. Since the detection unit is temperature controlled, any influence of temperature is usually negligible. Even injecting a liquid that has a different temperature than the detector, will not affect the data because the small volume of liquid will have the proper temperature before reaching the detector.
The temperature dependency can be exploited. With measurements at different temperatures, it is possible to investigate the influence of the temperature on the thermodynamics of the binding. When changing the detector temperature, the detector should be reset with the command NORMALIZE to ensure proper functioning (1).
Injecting a volume of flow buffer over a non derivatised surface, ideally should give a flat baseline with low drift.
In general, a slight positive bulk effect is seen (5-10 RU) during injection. If the response between the flow channels is not comparable, it may be an indication that the IFC needs replacement (2).
Matrix effects are changes in the extension of the dextran matrix on the sensor chip surface due to variations in pH or ionic strength (3). The effect of these variations is usually small in terms of absolute response, but may be significant in experiments where the highest sensitivity is required. Matrix effects may have time constants ranging from seconds (for e.g. small changes in pH or ionic strength) to minutes or even hours (for example after regeneration with extreme conditions). Normally, matrix effects are directly related to the mass concentration of the immobilized ligand, with greater effects at higher concentrations (4), (5). When the matrix interferes with the experiments, other dextran based or sensor chips without a dextran layer can be used.
Matrix effects can be minimized by matching the injected sample to the flow buffer used, either by dialysing the analyte or by adjusting the flow buffer.
|Solute||dR/dC (RU .(mg/ml)-1|
|proteins and DNA||~180|
Antibodies, like IgG, have two identical binding sites. Using antibodies as an analyte, will give rise to a bivalent analyte model. In principle, this model is useful, assuming that the ligand density is not too low and the dextran matrix is flexible enough. Nevertheless, it is better to avoid this kind of multiple binding site situations by immobilizing the antibody.
However, there are situations where dimerization or multivalence binding is beneficial (2). Receptors often have to dimerize before they activate a signal (6), (7), (8). By immobilizing the single binding site receptors and using a multivalent analyte, multimerization can occur. More information regarding choosing the ligand can be found in chapter 4.
It is important that the ligand is pure and does not contain other forms. This is easily checked by electrophoresis. Care must be taken though, to ensure that the protein is properly handled and remains active. Ligand heterogeneity may give rise to non-Langmuirian kinetics, which is more difficult to solve.
Also, make sure that heterogeneity caused by immobilization is minimized. Amine coupling results in random immobilization, with different orientations of the ligand. To overcome ligand heterogeneity other immobilization chemistries or ligand capturing strategies can be used.
For most applications, it is important to use the lowest possible ligand density, as this will provide good quality data (see chapter 4.5). This will prevent multiphasic behaviour of the system by reducing steric hindrance and mass transport (see chapter 9). However, with low ligand densities and low response levels, noise and drift become important factors (9). Equilibrate the sensor surface thoroughly and match flow and analyte buffer carefully. The noise of the system can be measured by setting a report point of 60 seconds on a stable baseline. The noise should be < 0.3 RU on BIACORE type of systems.
The technique of signal averaging can lead to a reduction of the noise level, assuming that all data sets are identically generated (10). The identical data sets are generated by injecting each analyte concentration several times and averaging the sensorgrams to one curve for each analyte concentration. This will reduce the noise in the resulting curves, making them easier to analyse.
Immobilization with the amine coupling protocol can affect the ligand stability and activity due to the sudden pH drop and low salt conditions (11). After immobilization of the ligand, the sensor surface must be checked for integrity. This is done by injecting a known analyte concentration until steady state is reached and calculating the Rmax. The calculated Rmax indicates how active the ligand is. To reach Rmax during injection, the analyte concentration must be at least 50 - 100 times the KD
For reliable kinetics, it is not necessary to know exactly how much ligand is still biologically active. Nor is it necessary to saturate the ligand.
However, knowing the actual amount of active ligand can sometimes be useful when the kinetics deviate from the 1:1 interaction. For instance, when the Rmax is lower than the expected value, consider if the active ligand sites could be blocked. When the Rmax is higher than the expected value, consider if there are more binding sites on ligand or if the analyte becomes dimerized. Sometimes the analyte is binding promiscuously to the ligand and the sensor surface (12),(13). In the case of a promiscuous interaction, higher analyte concentrations will not saturate the surface.
In addition, the ligand integrity must be preserved even after multiple regeneration cycles. The ligand integrity can be verified after regeneration by repeatedly injecting analyte and regeneration solutions and monitoring whether the ligand binding characteristics remain comparable.
Minimize mass transfer by immobilizing a small amount of ligand (see ligand density). Check for mass transfer limitation by using different flow rates. When the association and dissociation rate constants are independent of the flow rate, no mass transfer limitation is to be expected.
It becomes clear from the next table and the ‘Detecting mass transfer’ figure that for a high ligand density a flow rate of at least 40 µ/min is necessary to avoid most of the mass transport limitation. When analysing the sensorgram it is tempting to add mass transfer to the equations (5)to improve the fit. However, it is better to use the simple 1:1 model if there are no indications of mass transfer.
|High ligand||Low ligand|
|Flow rate||Injection volume||Curve slope||Curve slope|
Use a reference surface (9), (14)to compensate for matrix effects, refractive index effects and non-specific binding of the analyte. It is important to match the reference surface with the other surfaces as close as possible. Depending on the nature of the ligand and analyte there are three types of reference cells.
- The unmodified sensor surface
- The deactivated sensor surface
- The matched sensor surface
Check the suitability of a reference sensor surface by injecting the analyte at the highest concentration that you expect to use. If the non-specific binding is low, use this surface.
It is possible due to differences in ligand density and immobilization that both reference and active surface react differently to changes in ionic strength or organic solvents like DMSO in the analyte solution. The difference in channel behaviour is caused by the different displaced volumes and ligand properties. This type of artefact can be detected by injecting a control solution with the same refractive index as the analyte solution (15). This will provide essential information about the reference surface versus the specific surfaces. When differences between the reference and specific surfaces are observed, a calibration plot has to be made (16).
Especially with low ligand densities it is important to match the reference cell as close as possible with the other cells. Try to use an inactive ligand or a similar protein like a non-related IgG or use BSA to mimic a protein surface.
To measure low molecular mass analytes, a system of four flow cells, each with a higher ligand concentration, can be used. By proper bulk subtraction even small response changes can be measured accurately (17).
Even with a reference surface, it is better to eliminate bulk and drift effects. Carefully designed procedures and matched buffers will enhance the data fitting process. In any case, always use pure, filtered and degassed solutions (18). More detailed information can be found in (9).
Just like the ligand, the analyte can give some real artefacts. It is good to be aware of the possible problems you can encounter.
Some proteins have a tendency to multimerize. This may affect the determination of affinity, since the reaction is then multimeric instead of monomeric. Even proteins that were purified as monomeric may multi¬merize in time (19). In case of doubt, check by using a non-reducing gel or by size-exclusion chromatography (15). It has been demonstrated that glutathione S-transferase (GST) that has fused to a protein of interest, thereby facilitating purification, dimerizes in solution (20).
Check if the analyte has more binding sites per molecule. Use specific interaction inhibitors and stoichiometric measurements to investigate if the observed interaction is specific and 1:1.
The analyte concentration has a direct influence on the association phase because the equation contains a concentration term. With an actual analyte concentration that is half of the expected value, the ka will be reduced by half and the KD twice as high (as should it be). Dilution errors, evaporation of the solution and adsorption of the analyte to the vial wall can cause higher or lower concentrations than expected. The first runs after DESORB or cleaning can suffer from adsorption of the analyte to the tubing and IFC-walls. A pre-run with a high protein solution (for instance BSA) can reduce this effect. When analyte and flow buffers are not matched, drift and bulk effects may cause large residuals (28).
With a real concentration half of the expected the ka will be half lower and the KD two times higher:
|expected||50 nM||ka= 2.75 105M-1s-1||KD= 4.42 10-9M|
|real injected||25 nM||ka= 5.45 105M-1s-1||KD= 2.21 10-9M|
|No influence on kd, Rmax, R,sub>eqand Chi2|
Use the correct analyte concentration:
|affinity||0.1 - 10 x KD|
|kinetics||0.1 - 10 x KD||for kd > 10-2s-1|
|1000 x KD||for kd < 10-4s-1|
Steric hindrance (steric crowding) occurs when the initial binding causes physical occlusion of binding sites within the matrix (21)and thereby reduces the association. O'Shannessey (22)refers to this situation as the "parking problem". One way to avoid steric hindrance is to immobilize less ligand. An alternative - though second best approach - is to use initial rate binding analysis (21).
The masking of more than one ligand site by a single analyte molecule leads to a decrease in available binding sites and this decrease occurs faster than the decrease described in the 1:1 Langmuirian model (22). Adding a steric hindrance factor can compensate for this (23). On the other hand, the masking of potential ligand sites is essentially analyte concentration dependent, with higher concentrations leading to a more pronounced “parking problem”.
Analyte concentration and injection time should be chosen in that at least some of the sensorgrams reach steady state during injection and approach complete dissociation during the dissociation phase. Analysis based on incomplete association and dissociation data run the risk of missing heterogeneous binding behaviour (16).
However, when reaching equilibrium the signal decreases to the end of the injection, a smaller volume should be used (3). This is because with long injection times the sample plug will disperse a little with the running buffer lowering the initial analyte concentration. To check for a uniform sample plug, inject running buffer with 0.25 mM NaCl. A non-uniform sample plug may indicate that the system needs cleaning.
Independent versus linked reactions
When two proteins bind, a complex is formed. After a while the complex breaks apart and the two proteins are as the where before. However, when two proteins bind, the interaction can induce a rearrangement of the individual components in the complex. This can make the overall interaction stronger (or weaker). The binding characteristics may have been affected by the ligand, the analyte or both. This so called conformational change is a first order kinetics.
When the complex components not change the kinetics upon interaction, it is called an independent reaction. In case of an independent reaction, the dissociation curves should be identical for short and long analyte injections (association) (5).
A linked reaction is where the interaction induces a change in the binding strength. Although linked reactions cannot be directly measured with SPR, it is possible to investigate. By varying the analyte injection period during full ligand saturation and monitoring subsequent dissociation it can be seen if the dissociation is altered with the injection time. It is imperative that with each injection the ligand is saturated. Without full saturation the method cannot distinguish between a conformational change or something else like surface heterogeneity (25).
On the other hand, a competitive reaction could also give rise to a different dissociation rate. One form of the analyte may replace the other and thereby shifting the dissociation rate. To identify a competing reaction, use fraction collection of the dissociation phase and re-inject the concentrated fractions. When the dissociation constant of the re-injected sample is largely different from the original sample there is evidence for competing reactions (5).
Even when it is found that there is a linked reaction or a competing reaction, this must be confirmed by other types of measurements such as circular dichroisme or fluorescence spectroscopy. Fitting alone cannot prove a model is correct.
When higher flow rates give higher dissociation rates, this could indicate rebinding effects (26). Rebinding effects may also be suspected when the dissociation does not follow a single exponential curve and baseline levels are not reached.
Low-affinity interactions are characterized by continuous rebinding of the analyte. When only a fraction of the ligand sites are occupied, the binding seems to be more stable than at high concentrations. High concentrations of analyte can give multiphasic binding curves and a drop in binding level during injection phase indicating a less stable binding. In addition, at high analyte concentrations there are fewer free binding sites and thus higher dissociation constants.
To test if rebinding occurs during dissociation, an injection with ligand can be done. When the dissociation rate increases rebinding can be assumed (27).
Non specific binding
Non-specific binding can occur between the analyte and the negatively charged dextran-matrix (unreacted carboxyl groups) (28)or between the analyte and ligand. The source of the non-specific binding can be the analyte, a contaminant in the sample or the sample matrix. Non-specific binding is often hydrophobic in character and strong enough to withstand common regeneration conditions (29). SPR cannot differentiate between the binding of a specific analyte and the binding of a non-specific analyte. Therefore, take special care to recognize and avoid non-specific binding.
To investigate the influence of the surface matrix, use a native and a deactivated sensor chip surface. By simply injecting the analyte at the highest concentration over the two sensor chip surfaces, non-specific matrix binding becomes apparent. A minor binding by the analyte, may be compensated for using a reference surface (2). The use of smaller or less charged matrices may reduce non-specific binding. Deactivating the surface with other molecules like ethylenediamine, PEG amine, aspartate or glutamate can also help to reduce non-specific binding.
A change in buffer conditions, for instance, adding more salt (up to 0.5 M NaCl), a detergent (0.005% P20) or chemicals (3 mM EDTA) can minimize non-specific binding. The addition of carboxyl methyl dextran (0.1 – 10 mg/ml) to the sample can reduce non-specific binding even further.
Although CM5 dextran is the most commonly used sensor chip surface, it can suffer from a relative high level of fouling in bio fluids like plasma. Therefore, it can be beneficial to use alternative surface chemistries when using complex solutions (30), (31). Proposed coatings are alkanethiolates terminated with diethylene glycol and carboxylic groups, poly(ethylene glycol) grafted onto the SAMs and zwitterionic polymer brushes of poly(carboxybetaine methacrylate), poly(sulfobetaine methacrylate), and poly(phosphorylcholine methacrylate) (32).
Regeneration is complete and non-damaging
Regeneration of the sensor chip surface must totally remove the analyte without damaging the immobilized ligand. Conditions must be stringent enough to remove the analyte but mild enough to keep the ligand intact. Only a few solutions will fulfil these requirements. Therefore, during the experiments build-up of residual analyte and/or degrading ligand must be monitored by checking the baseline after regeneration and performing control injections with standard analyte concentrations. See also the regeneration section.
|(1)||Zeder-Lutz, G., E. Zuber, J. Witz, et al.Thermodynamic analysis of antigen-antibody binding using biosensor measurements at different temperatures.Analytical Biochemistry246: 123-132; (1997).|
|(2)||Myszka, D. G.Improving biosensor analysis.J.Mol.Recognit.12: 279-284; (1999). Goto reference|
|(3)||Paynter, S. and D. A. RussellSurface plasmon resonance measurement of pH-induced responses of immobiized biomolecules: conformational changes or electrostatic interaction effects?Anal.Biochem309: 85-95; (2002).|
|(4)||Biacore ABBIACORE Instrument Handbook. (1998).|
|(5)||Karlsson, R. and A. FaltExperimental design for kinetic analysis of protein-protein interactions with surface plasmon resonance biosensors.Journal of Immunological Methods200: 121-133; (1997). Goto reference|
|(6)||Cheskis, B. and L. P. FreedmanModulation of nuclear receptor interactions by ligands: kinetic analysis using surface plasmon resonance.Biochemistry35: 3309-3318; (1996).|
|(7)||Ibrahimi, O. A., F. Zhang, S. C. Lang Hrstka, et al.Kinetic Model for FGF, FGFR, and Proteoglycan Signal Transduction Complex Assembly.Biochemistry43: 4724-4730; (2004).|
|(8)||Klein, P., D. Mattoon, M. A. Lemmon, et al.A structure-based model for ligand binding and dimerization of EGF receptors.Proc.Natl.Acad.Sci.U.S.A.101: 929-934; (2004).|
|(9)||Ober, R. J. and E. S. WardThe Choice of Reference Cell in the Analysis of Kinetic Data Using BIAcore.Analytical Biochemistry271: 70-80; (1999).|
|(10)||Ober, R. J. and E. S. WardThe influence of signal noise on the accuracy of kinetic constants measured by surface plasmon resonance experiments.Analytical Biochemistry273: 49-59; (1999). Goto reference|
|(11)||Rich, R. L. and D. G. MyszkaThe Revolution of Real-Time, Label-Free Biosensor Applications. (2011). Goto reference|
|(12)||Giannetti, A. M., B. D. Koch and M. F. BrownerSurface Plasmon Resonance Based Assay for the Detection and Characterization of Promiscuous Inhibitors.Journal of Medicinal Chemistry51: 574-580; (2008). Goto reference|
|(13)||Ryan, A. J., N. M. Gray, P. N. Lowe, et al.Effect of detergent on "promiscuous" inhibitors.J.Med.Chem.46: 3448-3451; (2003).|
|(14)||Dorn, I. T., K. Pawlitschko, S. C. Pettinger, et al.Orientation and two-dimensional organization of proteins at chelator lipid interfaces.Biological Chemistry Hoppe-Seyler379: 1151-1159; (1998).|
|(15)||van der Merwe, P. A.Surface Plasmon Resonance. (2003).|
|(16)||Roos, H., R. Karlsson and K. AnderssonA calibration routine to improve the interpretation of low signal levels and low affinity interactions.|
|(17)||Karlsson, R. and R. StahlbergSurface plasmon resonance detection and multispot sensing for direct monitoring of interactions involving low-molecular-weight analytes and for determination of low affinities.Analytical Biochemistry228: 274-280; (1995).|
|(18)||Biacore ABKinetic and Affinity analysis using BIA - Level 2. (1998).|
|(19)||van der Merwe, P. A., M. H. Brown, S. J. Davis, et al.Affinity and kinetic analysis of the interaction of the cell adhesion molecules rat CD2 and CD48.EMBO J.12: 4945-4954; (1993).|
|(20)||Myszka, D. G.Kinetic analysis of macromoleculair interactions using surface plasmon resonance biosensors. (1997).|
|(21)||Edwards, P. R. and R. J. LeatherbarrowDetermination of association rate constants by an optical biosensor using initial rate analysis.Analytical Biochemistry246: 1-6; (1997).|
|(22)||O'Shannessy, D. J. and D. J. WinzorInterpretation of deviations from pseudo-first-order kinetic behavior in the characterization of ligand binding by biosensor technology.Analytical Biochemistry236: 275-283; (1996). Goto reference|
|(23)||Edwards, D. A.Steric hindrance effects on surface reactions: applications to BIAcore.J.Math.Biol.. (2007).|
|(24)||Edwards, P. R., A. Gill, K. D. V. Pollard, et al.Kinetics of protein-protein interactions at the surface of an optical biosensor.Analytical Biochemistry231: 210-217; (1995).|
|(25)||Rich, R. L. and D. G. MyszkaGrading the commercial optical biosensor literature-Class of 2008: 'The Mighty Binders'.J.Mol.Recognit.23: 1-64; (2010). Goto reference|
|(26)||Nieba, L., A. S. E. Nieba, A. Persson, et al.BIACORE analysis of histidine-tagged proteins using a chelating NTA sensor chip.Analytical Biochemistry252: 217-228; (1997).|
|(27)||De Crescenzo, G., S. Grothe, R. Lortie, et al.Real-Time Kinetic Studies on the Interaction of Transforming Growth Factor alpha with the Epidermal Growth Factor Receptor Extracellular Domain Reveal a Conformational Change Model.Biochemistry39: 9466-9476; (2000).|
|(28)||Webster, C. I., M. A. Cooper, L. C. Packman, et al.Kinetic analysis of high-mobility-group proteins HMG-1 and HMG-I/Y binding to cholesterol-tagged DNA on a supported lipid monolayer.Nucleic Acids Res.28: 1618-1624; (2000).|
|(29)||Biacore ABBIACORE Application Handbook. (1998).|
|(30)||Masson, J. F., T. M. Battaglia, M. J. Davidson, et al.Biocompatible polymers for antibody support on gold surfaces.Talanta67: 918-925; (2005). Goto reference|
|(31)||Masson, J. F., T. M. Battaglia, Y. C. Kim, et al.Preparation of analyte-sensitive polymeric supports for biochemical sensors.Talanta64: 716-725; (2004). Goto reference|
|(32)||Rodriguez Emmenegger, C., E. Brynda, T. Riedel, et al.Interaction of Blood Plasma with Antifouling Surfaces.Langmuir25: 6328-6333; (2009). Goto reference|