Back to the Top
The following message was posted to: PharmPK
I'm frequently interested in predicting in vivo binding from in vitro data.
Typically we have a Ki for a compound in vitro. Usually, this is done in a
low protein concentration (~5% BSA or similar). Then we measure the protein
binding, and get 99% or so. If the in vitro work was done in 0% protein, it
would be easy, just multiply the predicted concentration in vivo by the free
fraction and calculate the occupancy based on standard equations. But, I'm
not sure it is that easy when the in vitro work was in 5% albumin. We
usually refer to albumin as high capacity system. If this is the case, then
we may have very high binding even with 5% albumin, so the free
concentration in the in vitro work might be very different from the total
concentration. Presumably this would be described by Michaelis-Menten
Kinetics. But it likely is even more complicated, since there is likely
both a variety of proteins that bind any drug, including high-affinity, low
capacity proteins, as well as a spectrum of affinity on albumin. We might
ask the assay people to assay the free fraction (and give us the protein
binding in 5% albumin), but we never do this. Any insights on how to
address this?
Mark Sale
>
Back to the Top
The following message was posted to: PharmPK
Dear Mark,
You have described what for me is a critical issue in relating in vitro
work to reality in vivo. In all our predictive work, we must strive to get
comparable conditions for the in vitro experiment and the in vivo
situation. This is not often done, but could be done in two ways.
The first is to run the in vitro system under conditions which are as close
as possible to the in vivo physiologic situation. For the purposes of
calculating a Ki, Km, etc. this would mean adding the full complement of
protein and lipid to the system. As you intimated, the concentration of
drug added to the test-tube does not reflect the free concentration in the
system. Even without the BSA, the enzyme system, which will often have
lipidic membranes included, will offer substantial non-specific binding,
even if it is simply by partition into the hydrophobic phase. With the
estimation of Ki done under physiologic conditions, then it is relatively
straightforward (not necessarily simple) to relate this to the total local
organ concentrations (e.g. liver in a physiology based PK model), but not
the blood or plasma concentrations which can be different (especially
portal blood during an absorption phase).
Secondly, one could go to the other extreme and relate the two situations
via the unbound drug concentration. From the in vivo data, we can estimate
this from the total plasma drug and fu (assuming validity of fu for the
observed concentration range; fu may be non-linear with concentration).
However, the issue is then to estimate Ki, Km, etc relative to the unbound
drug. This in my experience is rarely done, as it requires a correction for
fu in the in vitro system. This can be done either by analysing for the
unbound drug in the test-tube (which takes time for the bioanalysts), or by
running the Ki estimation at varying protein concentrations and
extrapolating back to zero. I have seen this done for a Km estimation at
Brian Houston's lab in Manchester, where fu correction gave the same
results as back extrapolation. We then have one more assumption to rest
upon, which is that the unbound blood concentration is equal to the unbound
target site concentration. At equilibrium this should be the case for a
non-clearing organ containing no efflux transporters (e.g. PGP). Otherwise
a model will need to be used to relate the two unbound concentrations for
non-equilibrium conditions. This again falls into the realm of physiology
based PK modelling.
I hope this offers some insights.
Best regards, Phil.
PharmPK Discussion List Archive Index page
Copyright 1995-2010 David W. A. Bourne (david@boomer.org)