Hi there,

The problem is that you determine affinity using two different approaches : Biacore which allows to determine kinetics parameters from association/dissociation kinetics curves whereas ELISA is based on analysis of a system at steady state equilibrium to eventually calculate affinity. The main encountered problem is that most of the time equilibrium is difficult to reach so one tends to underestimate affinity in ELISA compared to Biacore...

By the way, I dont understand your first sentence :"Higher affinities result in higher ELISA signals" as in ELISA the important thing for Kd determination is the transition of the curve fitted from the data (signal=f(ligand)). The amplitude of the signal has absolutely no connection with Kd value.

No, Im not trying to calculate affinity by ELISA. My ELISA was not designed for that purpose. But two observations are somehow interesting for me. And Idont know whether they are common findings.

1. That  two ligands with the same BIACore affinity (but differences in on and off rates) have a very different behavior in ELISA.

2. More strange- that two ligands with very similar affinities and also on- and off-rates have a different behaviour in ELISA.This is something difficult to explain in my opinion.

The sentence at the beginning means that for higher affinity molecules you normally get a high ELISA signal with lower concentrations than for a low affinity variant of the same molecule. This is a common finding.  

So to answer point 1: ideally system is set at equilibrium for ELISA experiment but the time required for the system to reach the equilibrium is variable from one system to an other. As the kobs for the reaction of association/dissociation during ELISA incubation is always kobs=kon*L0 + koff, the time needed to reach equilibrium will be directly dependent on kon and koff value if considering constant L0 (the higher the kon and koff, the higher the kobs) . So at constant Kd(let's say Kd=1), equilibrium will be reached much faster with kon=koff=1000 than with kon=koff=1. So if the incubation time at given L0 is not long enough, the first system will give higher ELISA signal than the second because the former will be closer to equilibrium than the latter. As a conclusion, in your experiment I guess the ligand giving less ELISA signal is the one with the lower k values.

Point 2 is hard to explain...

I do agree with you now when you say that at constant concentration of ligand (not saturating), high affinity system will give a higher ELISA signal (but it's different from what was in your first post where no ligand consideration was indicated!)

Point 1

The ligand with less signal has indeed a very low kon , while another ligand (with a very high koff)- equal dissociation equilibrium constants- produces more signal. So this is consistent with the equilibrium approach explanation. 

About Point 2, I guess that either the equilibrium constants are not correctly measured or there is some additional difference between the two samples (with equal KD, kon and koff) resulting in different ELISA results. But I dont know how to distinguish between both possibilities. Both are mutated variants of the same ligand, so the ELISA is the same.   

Hi, here is the basic problem.  Binding constants measured by ELISA are inherently flawed by two factors.  First, the immobilized binding partner, usually not the antibody, is stuck irreversibly to a matrix, usually the plastic well of a protein binding plate.  By definition, it is at least partially denatured, and you have little to no control over this process.  Keep in mind, that in general, antibodies recognize only short stretches of antigens, and this is determined by the fact that during antigen processing and presentation via the MHCII receptor, only stretches of 17-20 amino acids will actually fit into the MHCII binding cleft.  Secondly, and here is the biggest problem, is simple experimental manipulation.  During the wash cycles of an ELISA, there will be significant loss of bound antibody, because you take the concentration of antigen to 0 during each wash cycle.  If dealing with a polyclonal IgG, this can be very misleading, since you may be removing all the antibodies except those with the slowest dissociation constants.  In the case of a monoclonal, the loss is also there, but since only one IgG species is present, the loss will not be selective.  Now all of that said, none of this really matters for the day to day ELISA, where the assay is being used to determine relative amounts of antigen in a protein mixture.  All of this should be readily controlled for in the standard curves, using known amounts of antigen as stnadards.  Where this problem is serious, is when looking for functional antibodies that exhibit an in vivo activity, example would be toxin neutralization.  To safe-guard yourself, it is always best to take any monoclonal antibody candidates identified by ELISA screening, and immediately do parallel BIAcore measurements, to assure that you are not working with epitopes that are only exposed when the antigen is denatured by immobilization to a denaturing matrix like the plastic of a protein binding well.  If you care to look it up, there are several good publications about why high affinity antibodies against toxins isolated by ELISA screeningfail to offer bio-protection when injected into animals.  So it all really depends on what you are planning to do with any antibodies you generate.  As far as your current situation goes, I would only trust BIAcore measurements where the antibodies are immoblized to a ProteinA biacore chip and then challenged with analyte.  You can spend years chasing ELISA artifacts where you have little control over the "native" vs. "non-native" components of the binding partners. 

Im not working at all with antibodies, but with a receptor and mutated ligands which loose or not receptor affinities. And Im not measuring affinities by ELISA. What Im trying to understand is why two ligand variants with very similar affinities (and even kon and koff measured by BIAcore) are giving me very different ELISA results. In both cases the receptor is immobilized. The different ligands, which are very similar except for a few point mutations, are in solution. Since the different behavior in ELISA is a fact I observed several times, not  explainable by binding kinetics, Im trying to understand whether there is another difference between my samples....  

The simplest explanation I can think of: I would imagine that the mutations have cause a partial unfolding event, or at least has created a binding surface for your matrix, that is your plastic well in the ELISA. You can try to confirm this, to look at ELISA signal without the receptor attached. Have you done this? I can imagine that if you have a lot of mutants, you may not have tested all mutants in such a control setting.

Thanks for your answer. I actually tested all of them on BSA-coated plates, without evidences of non-specific background. But right now I think I found the explanation, which as it happens was rather simple. The affinities were not correctly calculated (the problem was in the evaluation, not in the experiment itself).There is actually a 7-fold difference between the two molecules (due to a kon change).