Did you also try to titrate CaM into buffer alone to exclude any "heat of dilution" effects? Maybe CaM is slightly aggregated or oligomerizes in the syringe and dissociates upon dilution into the cell.

Yes I have done that, inclusively at much higher concentrations than the ones I use for this set of proteins. It looks normal, no aggregation/dissociation, but thanks for the input. What puzzles us are the stoichiometries and the fact that the Kds are not that far apart, although consistent throughout duplicate titrations and controls.

Because in the direct titration a higher affinity event (Kd 10 nM) is observed after a lower affinity event (Kd 30 nM), together with the fractional stoichiometries, that assay may reflect some kind of binding cooperativity or monomer-oligomer equilibrium coupled to binding.

In proteins with two non-identical and/or cooperative binding sites direct and reverse titrations look very different.

Do you have calcium in the CaMBD solution? if not, then there's the effect of the stock of CaM calcium being diluted throughout the titration, while it's constant in the other. But I'm assuming you have both proteins in the same buffer, otherwise the mixing effects would also be very large.

Reverse titrations are VERY tricky because you have at the same time two opposite effects: increasing the number of binding sites and also reversing the proportion of ligand vs macromolecule. This is about DNA binding, but perhaps qualitatively could be of some help in your case.

S. C. Kowalczykowski, L. S. Paul, N. Lonberg, J. W. Newport, J. A. McSwiggen, P. H. von Hippel, Biochemistry 1986, 25, 1226–1240.

by the way, I forgot, try repeating the reverse titration at a different (higher?) concentration, and see what happens.

Thank you for the quick replies! I will read the suggested paper.

I extensively dialized both proteins against a Ca2+-containing buffer. We have tried to do SEC-DLS to study potential oligomerization but failed to get any result. We might attempt AUC in a near future. Could this 2-step binding be due to a conformational change of calmodulin, for example, that renders it with a higher affinity for the CaMBD? I have some evidence that the N-lobe of CaM is not involved in binding and might be without calcium, but I don't know if this could be reflected in an isotherm that looks like that...

whatever the mechanism it should operate in both direct and reverse titrations. I feel that the difference might come from an effect of the concentration. you might not be exactly under the same conditions in both titrations.

I use 6 micromolar on the cell and 60 on the syringe, on both direct and reverse titrations.

Hi Maria I think with a concentration of 6000nM on the cell you are far away from the optimal conditions for a Kd smaller than 100nM interaction measurement. If possible you should go down below the Kd with your concentration inside the cell. This will significantly enhance your Kd measurement resolution. Lowering the concentration will also allow you to get rid or at least to quantify/detect dimerization effects.

If material or sensitivity of your ITC device are a problem: you could use MicroScale Thermophoresis (MST). MST can measure affinities between Kd < 1pM up to Kd>10mM with a very low material consumption.

http://www.sciencedirect.com/science/article/pii/S0022286014002750

@ Philipp,

You may have been gone a bit far with your suggestion about reducing the concentration in the cell below the Kd.

As most of ITC users know, in order to get a reliable result (that is, a good estimate for the affinity, the enthalpy and the stoichiometry of binding, with no dependency or correlation between parameters and good convergence in the NLSF analysis), the well-known c-parameter (c = [P]cell / Kd) must be between 1 and 10000, with an optimal range of 10

@Adrian,

thank you for your statement. Yes you are right, regarding the measurement sensitivity of ITC my dilution factor was far too high.

I was just looking at the law of mass action for a bi-molecular interaction, rearranged this equation for the fraction of bound molecules with taking into account the ligand and protein depletion (free ligand = titrated ligand - ligand bound to target protein) and (free protein = protein put into cell - protein bound to ligand) and got a quadratic equation which showed me that I'll get the best resolution when my target protein concentration is in the range or below of the Kd (please see attached plot).

Regardless of the method we are using, we are always taking into account the depletion of molecules and work with the full nonlinear Kd-equation for the law of mass action.

However, if it is measurable, the "Kd-resolution"="the discrimination between Kd-values" should always be best if the target concentration is within the range of the Kd.

I was thinking about this optimal discrimination as we see these both events at Kd=10nM and Kd=30nM and 6000 nM target concentration are really far away from it..

I'am still afraid that we can't fully exclude that we have dimerized molecules at such high micromolar concentrations?

I'am asking because I know this issue from John Ladbury (MD Anderson Cancer Center, Houston Tx) he used MicroScale Thermophoresis in his CELL paper "Inhibition of Basal FGF Receptor Signaling by Dimeric Grb2, DOI 10.1016/j.cell.2012.04.033" to quantify the "dimerization Kd" to be sure that he works with dimers in his ITC experiments.

@Philipp,

I have made calculations with the values you have employed: Kd= 25-250 nM, [P]=0.1-10 micromolar.

As you can see in the attached file, the titrations with [P]=0.1 micromolar (c=0.4 and 4)do not give sufficient signal (almost a flat line; if you expand the scale you may observe a series of peaks, but well below the usual noise level in ITC instruments) compared to the ones with [P]=10 micromolar. I have considered that the enthalpy of binding is 5 kcal/mol, a reasonable value for biological interactions.

In particular, the titration with c=0.4 cannot be reliably analyzed unless the stoichiometry is fixed to n=1. The titration with c=4 is in the limit of the practical range for independent affinity, enthalpy and stoichiometry estimation.

Furthermore, the titrations with [P]=10 micromolar (c=40 and 400), corresponding to a 10-fold change in Kd, can be well analyzed and discriminated, even though the Kd is well below the protein concentration in the cell.

If anyone is interested the data treatment I referred to was published last year in Structure, using the software Affinimeter.