Several people have already pointed out that a "pull-down" is not
a rigorous, quantitative method.  Mackay et al (TIBS 32, 530-531, 2007)
have some interesting comments to the effect that many reported
interactions found by pull-down or immunoprecipitation are simply
artefacts. Pull-downs are also a function of off-rate, unlike your ITC  
data.

These ITC data do seem to show a clear case of E-E compensation, but
it is simply not possible to infer much in the way of structural  
detail from
these numbers.  A particular delta S value or delta Cp value is not  
really
good evidence that a binding site is/is not hydrophobic or large. Small
changes in pH can affect deltaS and deltaH a lot. Tris has a very large
enthalpy of protonation and is not therefore always a suitable buffer  
for
ITC. The measured dS and dH values can change with buffer if there is a
protonation involved in binding of your protein partners. Changes in the
water structure around the binding site can also change binding  
thermodynamics
appreciably. You can see very large compensating changes in dS and dH
with very minor structural changes (see Davies et al Prot. Sci. 8,  
1432-44, 1999).

When you say the crystal structure suggests that the mutant would
bind the peptide weakly, can you give more detail? Are you convinced the
mode of binding is the same? Since "there is still a large interaction  
interface"
left intact, it seems that the ITC data are pretty solid, and the pull- 
downs
more suspect.

As for the pH difference between your pull-downs and ITC, remember that
the blood pH of healthy adults doesn't change much from about pH 7.2 -  
7.4,
but those changes are very important for helping Hb bind and release  
oxygen.






On Sep 3, 2008, at 5:37 AM, Brett Collins wrote:

Dear CCP4 Community,

My apologies for the non-crystallography biochemical question but it  
occurred to me that there are many people on this list who are also  
very good biochemists.

We have just performed an ITC experiment with two proteins and  
measured a Kd of 150 nM, deltaH of -15 kCal/mol, deltaS of -15 Cal/mol/ 
K and deltaCp of -2000 J/Mol/K.

We also measured the binding of a mutant of one of these proteins  
predicted from crystal structure to inhibit binding of a small  
fragment of peptide (this is predicted to reduce binding slightly but  
not to affect total binding as there is still a large interaction  
interface that is left intact).

This mutant has a Kd of 150 nM as well, but deltaH is -10 kCal/mol,  
delta S is essentially zero, and deltaCp reduces in magnitude to about  
-1500 J/Mol/K as we would predict from the change of buried surface  
area. The ITC data looks good and we have repeated the experiments a  
number of times so they are statistically significant. The experiments  
were performed within reasonable concentration limits (~10uM protein  
in the cell so the C-value is about 50-100)

Now the puzzle is that the mutant binds less strongly in pulldowns  
(about 50% reduction after several washes) but we see an almost  
identical Kd by ITC despite major changes in enthalpy/entropy  
contributions to binding. The mutant and wildtype appear to have  
identical fold by CD but of course there may be small differences.  
Everything makes sense except the lack of Kd change by ITC.

Does anyone have any experience of similar results, or more  
importantly have a possible explanation for them?

Any thoughts greatly appreciated.

Brett Collins