Bad fits for protein protein binding

Dear All,

I am trying to fit some data, but it does not appear to be right. The data and the model diverge at the very beginning of the injection and at the very end of the dissociation, but mostly at this second part. According to the 1:1 Model, the dissociation is very fast and drops down within a few seconds. However, at the very end, the data shows a slowed down dissociation, whereas according to the model, it is supposed to be faster.

I have immobilized a homodecamer that was biotinylated, and trying to measure the affinities of small nanobodies.
The running buffer is 20 mM Hepes pH=7.4. 150 mM NaCl, 2 mM CaCl2, 0.1% LMNG (detergent) and 0.1% BSA. All analytes were dilluted more than 100-fold in the running buffer, the system after immobilization was well equilibrated over night, and measurements occur at 10°C.

Heterogenous ligand or bivalent analyte fits the data perfectly, but we cannot explain why that should be the case. We expect a 1:1 binding, and the samples are very pure.

Any tips would be greatly appreciated.

Hi,
The first and third fitting seem to suffer from large RI values. The second looks reasonable well but the dissociation at the end seems not to fit well. You can try to use a higher flow rate to see if mass transport is an issue. I also noticed that the start of the sensorgram before the first injection is not horizontal. Is there drift in the system?
The first equilibrium fit is better than the second. However, higher analyte concentrations are necessary to do a proper equilibrium fit. I suggest to look at www.sprpages.nl/sensorgram-tutorial/a-curve and compare your equilibrium fitting with the kinetic fitting.

As a general comment (also to those who read along): diluting an analyte sample is not an guarantee that the bulk effect is low. A 1 mM NaCl difference is inducing 20 RU of bulk effect. Best way to reduce bulk effect is to dialyse or change buffer by gel filtration.

Kind regards
Arnoud

Thank you very much, Arnoud!
I would have a follow up question, if you dont mind. We expect a 1:1 interaction between our ligand and analyte. But it is the heterogenous ligand models that fit the data best. Is it possible to tell apart the background signal from the real binding if we were to fit the data with the heterogenous ligand model?
In my opinion it is wrong, and incorrect to do so.

There was no drift in the system. We will try higher flowrates with the next set-up.

Thank you for the advices.
Marton

Dear Marton,

You are right. A heterogeneous model will always fit better and that is because there are more parameters in the equation. For an indication of the background you can look at the Rmax. If the Rmax of one of the kinetic sets is low compared with the other you can assume this is the background. However, it is very bad practise to report the other kinetic set as your result. You will notice when you start the fitting with other initial values the kinetic parameters will vary and therefore are not well defined.
Besides the bulk effect, non-specific binding must be low or properly compensated by a reference before you can report reliable kinetic values.

Kind regards
Arnoud

Thank you Arnoud.
We tried higher flow rates, unfortunately the fit for the 1:1 binding model did not improve. The same issue appeared: the measured dissociation slows down at the very end of the sensogram, while the model predicts a faster dissociation. All in all, for about 80-90% the model fits the dissociation of the measured values, just the very last bit deviates.

Hi Morton,

I was thinking about the dissociation which did not go to baseline. If possible, try a different type of sensor chip to investigate if the matrix is binding the analyte. If you used a dextran type of chip you can try a low carboxylated type or one with a short destran matrix. In addition you can look at the CAP system which is different in matrix.

Kind regards
Arnoud