Validation of a PAMPA permeability assay


Over recent weeks I have been working to establish a suite of assays using my own UHPLC instrumentation.  LogD, solubility and PAMPA permeability measurement are all now available to clients via out-sourcing, in-sourcing or consultancy to set up the methods in your laboratory which, can also include bespoke software.

Here I am presenting some validation data to support the use of my PAMPA method.

The role of PAMPA in drug discovery

Screening Cascade

A typical drug discovery screening cascade might involve running the following assays in parallel with the primary potency screen:

QC (LC/MS): assure the integrity of the screening sample

Kinetic solubility: assure the validity of all in vitro data

Microsomal or hepatocyte stability: predict clearance

Hopefully, this will yield a number of compounds with adequate potency, solubility and metabolic stability for the stage of the programme that could advance into early PK assessment.  In order to prioritise within this group, some form of in vitro permeability measurement can be used to predict absorption from the gut and, where relevant, potential access to the brain.

The PK in turn informs selection of molecules with sufficient predicted exposure for in vivo pharmacological testing.

in vitro permeability measurement techniques

Transcellular permeability, measured across cell lines such as Caco-2 cells in vitro, is often used for assessing oral drug absorption potential. However, such techniques involve cell culture which is time consuming with a 21 day culture and inevitably less reproducible than the far simpler chemical PAMPA system which is also more cost effective and higher throughput making it an ideal tool for use at this stage of drug discovery.

It must be acknowledged that, owing to the availability of the paracellular absorption route, small molecules can still get onboard (Fabs is good) even with low Papp (Caco-2 or PAMPA). This manifests itself in a much wider spread in values obtained via in vitro permeability experiments than is found in absorption data.

When does permeability become an issue?

For polar compounds with low transcellular permeability, the issues are significantly more important for CNS penetration (due to the tight junctions of the BBB allowing no paracellular route through and its greater expression of efflux transporters) and for large molecules that can not access the paracellular route due to their size. Note also observations, such as by Gleeson, showing that acids and zwitterions have lower average permeability than bases and neutral molecules.

If a compound has an efflux liability, low passive permeability makes it a problem, in other words efflux can not be tolerated with low permeability compounds. For this reason, CNS drugs frequently need higher permeability (to overcome efflux).

As a simple rule of thumb, it appears that a Papp >20nm/s should not significantly limit Fabs and for CNS we should be aiming for >100nm/s. Bear in mind that the absolute numbers will depend on the particular system and these refer to the Corning/Gentest PAMPA system under evaluation here.

Of course, as is often the case in ADME optimization, there may be a trade-off to be made.  For instance, can Papp be sacrificed to improve metabolic stability and/or solubility?


PAMPA: Parallel Artificial Membrane Permeability Assay

Papp: apparent permeability

Fabs: fraction absorbed

BBB: Blood Brain Barrier

The PAMPA method

The diagram below describes the experimental set-up used in a PAMPA experiment.

After the incubation period, 5 hours in this system, the plates are separated and the compound concentrations determined in both the donor and acceptor wells.  If the initial concentration of the donor solution is also measured then it is possible to calculate the overall recovery in the system (and anything below 100% represents the amount of membrane retention of the compound). High membrane retention is to be expected for some highly lipophilic compounds. The first step of crossing the membrane is for the molecule to effectively dissolve in it and it is not surprising that in some instances the molecules remain in the membrane due to their high affinity for it.  Binding to plastic surfaces may also contribute.

The Corning Gentest Pre-coated PAMPA Plate system

BD Gentest commercialised a pre-coated PAMPA plate system and this has been continued following their acquisition by Corning.  There is a lot of literature regarding the advantages of the system on the Corning website which is available here.  Having had good previous experience with this system over several years, and knowing it to be both convenient and robust, I had no hesitation in selecting this system for my work at DAC.

Structure of the membrane

The Corning Gentest membranes consist of a lipid-oil-lipid tri-layer structure which is constructed in the pores of the porous filter using three consecutive coating steps.  The oil layer mimics the hydrophobic interior of the biological membrane; and the amphiphilic lipids anchoring on the oil / water interface mimic the exterior of the biological membrane.  Compared to traditional membranes, excessive amounts of solvent are not present, the layers are structured and there is a short permeation pathway which is closer to the situation in biological membranes.

As the system measures passive permeability, with no concern over saturation of transporters, a high compound concentration can be used, typically 100 or 200uM.  Using a high concentration is beneficial for accurately measuring permeability of low permeability compounds as it will lead to greater acceptor concentration and hence improved quantification by LC-UV and it is at this end of the scale that greatest accuracy is needed.  Using a wavelength close to the λmax for detection will also help and software that automates signal selection can save a lot of time and may be critical for high throughput.  For solubility, 5% DMSO can be included in both donor and acceptor buffers with no membrane stability issue. Ideally, samples will be filtered before being dispensed into the donor wells.  Although MS detection is not required, it is useful for positive compound ID in the acceptor wells, especially for low permeability compounds (small peaks).  To this end, UV spectra from a Diode Array Detector are also extremely useful.

Study Protocol

Donor wells: 2mM DMSO compound stocks were diluted 20x in buffer to give an initial compound concentration of 100uM and a DMSO concentration of 5%.  25mM sodium phosphate buffer was used at pH 6.5 and 7.4.

Acceptor wells: 25mM sodium phosphate pH 7.4 containing 5% DMSO was used

All experiments were performed in duplicate

Incubation: 5hours at ambient temperature

Analysis was by rapid gradient UHPLC-DAD 


The following set of known drugs was analysed.  The table below shows a comparison of the measured permeabilities and mass retention. All are mean values of duplicate measurements with no data having been excluded.


*The permeability classification used above was:

<2nm/s LOW

2-20nm/s MEDIUM

>20nm/s HIGH


Where there are significant differences in permeability between the two donor pH values studied these are all consistent with the expected acid/base properties of the compound and the principle that compounds will have lower permeability when they are in an ionised form due to their reduced lipophilicity.

It is interesting to see that metoprolol is border-line medium/high permeability as this compound has been suggested as a low/high BCS permeability class marker.

Saquinavir was the only compound to show high mass retention (about 25%).

The table below shows a comparison between the PAMPA permeability classification obtained in this study and literature Caco-2 and human fraction absorbed (Fabs) data.

Ref 1 = HH Usansky and PJ Sinko JPET 314:391-399, 2005

Where no reference is cited the data was found via multiple internet sources

There is clearly excellent agreement between PAMPA and Caco-2 in this data set.  All 4 compounds categorized as low by PAMPA also had low Caco-2 permeability (<5 x 10-6 cm/s) and, where Caco-2 data exists, all compounds categorized as medium or high by PAMPA also had high Caco-2 permeability (>20 x 10-6 cm/s) thus supporting the use of PAMPA as a cost-effective method for assessing compound in vitro permeability.

Note the much wider spread in values obtained via Caco-2 experiments than is found in human absorption data. As mentioned previously, compounds with low in vitro permeability can show a high overall absorption if absorbed paracellularly.  Indeed, this is known to be the case for both atenolol and hydrochlorothiazide which is the reason why their human Fabs are not expected to be well predicted by PAMPA. Caco-2 experiments also do not predict paracellular absorption well because the paracellular route is tighter in Caco-2 cells than that in the small intestine in vivo. While the average pore radius of the tight junctions in the human small intestine is around 8–13Å, the corresponding radius in Caco-2 cells is about 4Å. When the paracellular pathway is narrower, the intrinsic permeability will be lower than in the in vivo situation.

Of the two compounds with low Fabs, saquinavir is a zwitterion and sulfasalazine is predicted to be doubly negatively charged at physiological pH.  In contrast, atenolol is a base and hydrochlorothiazide is a neutral compound.


I would like to thank Xenogesis for providing the compounds used in this work.  In particular, I would like to thank Dr Richard Weaver, MD and Founder of Xenogesis, for his helpful comments and suggestions.

Further Information

If any of the issues discussed in these articles are of interest to you, please feel free to contact me directly on for further information.

Mark PortsmouthComment