Assessing the Intestinal Absorption of New Pharmaceuticals

The advent of more efficient methods to synthesize and screen new chemical compounds is increasing the number of chemical leads identified in the drug discovery phase. Compounds with good biological activity may fail to become drugs due to insufficient oral absorption. Selection of drug development candidates with adequate absorption characteristics should increase the probability of success in the development phase. To assess the absorption potential of new chemical entities numerous in vitro and in vivo model systems have been used. Many laboratories rely on cell culture models of intestinal permeability such as, Caco-2, HT-29 and MDCK. To attempt to increase the throughput of permeability measurements, several physicochemical methods such as, immobilized artificial membrane (IAM) columns and parallel artificial membrane permeation assay (PAMPA) have been used. More recently, much attention has been given to the development of computational methods to predict drug absorption. However, it is clear that no single method will sufficient for studying drug absorption, but most likely a combination of systems will be needed. Higher throughput, less reliable methods could be used to discover ‘loser'compounds, whereas lower throughput, more accurate methods could be used to optimize the absorption properties of lead compounds. Finally, accurate methods are needed to understand absorption mechanisms (efflux -limited absorption,

carrier-mediated, intestinal metabolism) that may limit intestinal drug absorption. This information could be extremely valuable to medicinal chemists in the selection of favorable chemo-types. This article describes different techniques used for evaluating drug absorption and indicates their advantages and disadvantages.

 Models of Intestinal Drug Absorption Prediction

The ability of an orally administered compound to permeate the intestinal mucosa may be limited by the physical and/or the biochemical component of the intestinal mucosal barrier. Thus, in vitro intestinal permeability models should not only predict intestinal drug absorption potential but also provide some understanding of the absorption mechanism(s). This can be accomplished to the extent in which the permeability model incorporates the functionality of the physical and biochemical barrier components.

A. In Vivo Drug Absorption

In general, drug absorption in animals is believed to be a good predictor of absorption in humans. Animals integrate all the biological factors that may affect drug absorption.

Unlike in vitro systems, in which a correlation to in vivo data must be established, this step is unnecessary when animals are used. An important advantage of whole animals is that the species used in absorption studies could be the same one used in pharmacology and/or toxicology evaluations. They also can be used to evaluate complex formulations, which would be very difficult to test in vitro. Some of the disadvantages of studies with whole animals include the need for relatively large amounts of material, the complexity

of the analytical methods needed for plasma analysis, the time-consuming and labor-intensive nature of experiments, and the fact that they provide little mechanistic information on drug absorption.

B. In Vitro Models of Intestinal Drug Absorption

The successful application of in vitro models of intestinal drug absorption depends on the extent to which the model comprises the relevant characteristics of the in vivo biological barrier. Despite the obvious difficulties associated with trying to reproduce in vitro all the characteristics of the intestinal mucosa, various systems have been developed which mimic, to varying degrees, the relevant barrier properties of the intestinal mucosa. These systems include excised tissue (e.g. isolated intestinal segments, everted sacs, intestinal rings, stripped and unstripped mucosal sheets), cultured cells (e.g. Caco-2, HT-29, T84, and MDCK), physicochemical methods (e.g. Log D/P, immobilized artificial membrane, parallel artificial membrane permeation assay), mucosal cell membrane vesicles, isolated mucosal cells, and computational (in silico) methods. Below is description of the most commonly used methods used to evaluate study intestinal drug absorption potential.

1. Excised Tissue

Excised intestinal tissues have been used to study intestinal drug and nutrient absorption. The solution containing the drug is applied to one side (mucosa or serosa) of the mucosa and the rate of drug absorption is determined by measuring either the disappearance of drug from the dosing solution or appearance of drug in the serosal side. Although they vary in complexity and versatility, excised tissue preparations share the two important advantages: a) preservation of the architectural integrity and b) ability to determine absorption across different gastrointestinal segments. A common disadvantage is the limited viability of this type of preparations.

i. Perfused Intestinal Segments

Isolated intestinal segments comprise the absorptive cells and the underlying muscle layers. As it is commonly used, this technique only allows sampling from the mucosal side; drug disappearance is assumed to be equal to drug absorption. This assumption is valid when apical uptake is the rate-limiting step in drug absorption. But, considering

that drug absorption is not the only factor responsible for luminal drug disappearance, this assumption could be misleading. For example, mucosal metabolism and mucosal accumulation could lead to an overestimation of true drug absorption. Indeed, an earlier study found that, for several b-lactam antibiotics, absorption based on luminal disappearance was roughly twice the true amount transported across the tissue.Perfused intestinal loops offer a few advantages over other drug absorption models. Unlike whole animals, perfused intestinal loops can be used to study segmental differences in drug absorption and metabolism without the interference from physiological factors such as, gastric emptying, surface area of the segment and/or small intestinal transit time. With respect to throughput and complexity, this technique offers a few advantages over whole animals, but not over other in vitro techniques. It also has numerous disadvantages. First, the determination of absorption based on luminal disappearance is potentially misleading. Second, it requires large amounts of compound, relative to other in vitro systems. Third, the number of intestinal segments that can be obtained from one animal is limited. Fourth, as is the case with other excised tissue preparations, the viability of perfused intestinal segments is limited. As a result of these limitations, this technique is not likely to be useful as a screening tool. It has greater value in the elucidation of transport mechanisms. It may also be useful to evaluate the absorption of drugs whose poor solubility requires the use of complex dosing vehicles, which could not be presented to other in vitro systems such as, culture cells.

ii. Everted Sacs

The everted sac was one of the first in vitro techniques used to study intestinal drug absorption. It is prepared by inverting a piece of intestine using a glass rod. The two ends

of the intestinal segment are tied and the inside of the resulting sac is filled with an oxygenated buffer. The sac is then placed in a container that has the test compound. Drug

absorption is measured by sampling the solution inside and outside the sac. It includes the mucosa plus underlying muscle layers. This differs from the in vivo situation and could lead to biased values for transport. The presence of this muscle layer could accumulate drugs, and thus lead to poor recovery. This technique was popular a few decades ago, but its utilization in recent years has been greatly reduced. It is unlikely that it will constitute an important absorption-screening tool in the future.

iii. Intestinal Mucosa (Stripped or Unstripped) Mounted in Ussing Chambers

To prepare intestinal mucosal sheets suitable for mounting in Ussing chambers a longitudinal cut of an intestinal segment is made to produce a long mucosal sheet.

This sheet is cut as needed to produce mucosal strips of adequate size to fit in the opening of the diffusion (Ussing) chamber. Although mucosal sheets are used with or without

the underlying muscle layer, in vivo, the muscle layer is not a barrier to absorption. Thus, the removal of this muscle layer, a process known as stripping, is advantageous for two

reasons. First, it removes an artificial permeability barrier, and second, stripped tissues can be oxygenated more efficiently. Usually, Ussing chambers are connected to voltage clamps, which are used to make electrical measurements during the course of experiments. These measurements can be used to monitor tissue integrity and viability. This technique is very valuable. It allows the determination of transport polarity, which is indicative of the involvement of carrier or efflux mechanisms. This technique also allows determination of segmental-dependence of transport. It has been used to evaluate the in vitro permeability with varying degrees of success. For instance, one study found that rat intestinal mucosa mounted in Ussing chambers were useful in describing the regional variability in gastrointestinal absorption of mixed series of compounds . Also, a comparison between the Ussing chamber technique and human jejunal permeability found a good correlation for a discrete small series of 12 compounds . In general, the in vitro technique generally underestimated the transport of compounds that undergo carrier-mediated absorption . While this small discrepancy may be reflect the limitation of the in vitro-in vivo correlation between these systems, it is also possible that the intrinsic level of expression of the carriers involved, may differ between rat and human. It has also been argued that the presence of the muscularis mucosa in the Ussing technique constitutes an artificial barrier not present in vivo, which can bias the permeability data. And because in vivo the rate limiting barrier is the epithelial cell layer, in vitro cell culture models may better reflect the absorption potential of a compound. However, the complexity of this technique and the  amount of compound needed in studies, are likely to limit the application of this model system to the study of transport

mechanisms.

2. Isolated Enterocytes

Isolated intestinal cells have been used to study intestinal drug absorption. To prepare these cells, first, an intestinal segment is cut and then the mucosal surface is treated with

mechanical forces, enzymes or chelating agents to dissociate the cells from the underlying tissues. This technique is not commonly used because of its limited utility. The process of cell isolation destroys many cells and greatly diminishes cell

viability. Because cells must be used as a suspension, they lack the polarity that characterizes intestinal mucosal cells in vivo. Thus, isolated enterocytes can be used to study drug uptake, but not transepithelial transport or transport polarity.

3. Membrane Vesicles (BBMV, BLMV)

Cell membrane preparations include, brush-border membrane vesicles and basolateral membrane vesicles. The membrane vesicles are suspended in a physiologic buffer

that contains the test compound. Uptake into these vesicles is supposed to mimic transport across cell membranes. The vesicles are useful to study drug uptake or metabolism by plasma membrane-bound enzymes. These preparations constitute simplified models to study mucosal absorption. They have been proved useful to study specific membrane processes such as, binding of organic cations and uptake of PEPT1 substrates by small intestine.Advantages of this preparation include, the short duration of experiments, the small amount of material needed, and the ease with which vesicles can be made. However, they also possess many disadvantages. Their lack of cellular metabolism makes it difficult to study important aspects of ATPdependent

transport. In addition, uptake into membrane vesicles does not provide any insight into paracellular transport.

i. Caco-2 Cells

Numerous cultured cells have been used to study intestinal permeability. Caco-2, HT-29, and MDCK cells have been used with varying degrees of success. Caco-2 cells, first characterized as an intestinal permeability model in 1989 have been used to study various aspects of intestinal permeability. They exhibit morphological features of small intestinal cells (e.g., tight intercellular junctions and microvilli) and express intestinal enzymes

(e.g., aminopeptidases, esterases, sulfatases, and cytochrome P450 enzymes) and transporters (e.g., bile acid carrier, large neutral amino acid carrier, biotin carrier, monocarboxylic acid carrier, PEPT1, and p-glycoprotein). Results from many laboratories suggest a correlation between in vitro permeability coefficient values in Caco-2 monolayers and indicators of in vivo absorption (e.g., bioavailability,

absorption, pharmacological effect) Encouraging in vitro -in vivo correlations have led to the widespread use of this model. The ease with which Caco-2 cells can be cultured has permitted their utilization in many laboratories. In many instances, little, if any, effort has been made to implement control measurements for monitoring the performance of the cells. Thus, the heterogeneity of wild type Caco-2 cells and the different culturing conditions used in various laboratories can give rise to the selection of different cell populations. This gradual selection, referred to as phenotypic drift, could contribute to the large inter-laboratory variability in Caco-2 permeability measurements. For example,

transepithelial electrical resistance (TEER) and the transepithelial permeability of a paracellular flux marker such as, mannitol are commonly used to monitor monolayer

integrity or cell damage.

ii. Madin-Darby Canine Kidney (MDCK)

MDCK cells have received attention as an alternative to Caco-2 cells for permeability measurements. When grown under standard culture conditions, MDCK develop tight

junctions and form monolayers of polarized cells. Cultured on filters they can be used to study not only cellular uptake but also vectorial fluxes. The main advantage over Caco-2

cells is the shorter culture times, which in some laboratories equals 24 hours. The transepithelial electrical resistance of  MDCK cells is lower than that of Caco-2 cells and thus closer to the TEER of the small intestine. Several studies have found a good correlation between permeability coefficient values in Caco-2 and MDCK cells [85, 104]. The permeability coefficients of hydrophilic compounds are usually lower in Caco-2 cells than in MDCK cells.This is consistent with the higher transepithelial electrical

resistance of Caco-2 cells. The canine kidney origin of MDCK cells is occasionally considered a disadvantage. In addition, the expression of intestinal transporters in MDCK

cells has not been well characterized. The true utility of MDCK as a model system of intestinal permeability will be better understood when more data become available.

iii. HT29

Several clones of HT29 cells have been used to study different aspects of intestinal drug absorption. Wild type HT29 cells form multilayers of undifferentiated cells, which

are not useful for intestinal permeability studies. Culturing wild-type HT29 cells in media containing galactose instead of glucose leads to the selection of a subclone of HT29 that

form monolayers of polarized cells. Several HT29 clones differentiate into enterocytic cells or goblet cells, which secrete mucus. An enterocytic HT29 clone, HT29-18-C1,

was proposed as a model of intestinal permeability ; however, these cells grow very slowly and a large number of cultures failed to develop acceptable barrier characteristics

(e.g. transepithelial electrical resistance and mannitol permeability). Two HT29 clones that secrete mucus are HT29-H and HT29- MTX. The HT29-H clone was used to demonstrate the role of mucus on the permeability of testosterone. HT29-H monolayers had TEER values equal to 159 Wcm2, whereas the corresponding TEER value for Caco-

2 monolayers was 401 Wcm2. The presence of a mucus layer in HT29-H monolayers resulted in lower permeability coefficients in HT29-H (6.8 x 10-6 cm/s) than Caco-2

monolayers (31.5 x 10-6 cm/s). This indicates that the mucus layer accounts for most of the permeability resistance to testosterone. With the exception of a few studies the role of

mucus on drug transport has been largely ignored as most pharmaceutical scientist rely only on Caco-2 monolayers to predict intestinal permeability. In an attempt to incorporate

into Caco-2 cell monolayers characteristics more representative of the intestinal mucosa, Caco-2 cells were cocultured with HT29-MTX or HT29-H cells

V. ARTIFICIAL MEMBRANE METHODS OF DRUG ABSORPTION

Several methods, which do not involve biological preparations, have been used to model intestinal permeability. The advantages of this type of techniques are their potential for higher throughput and reproducibility. The disadvantages include the lack of enzymes, influx and efflux transporters, and paracellular pathways. In addition, most artificial membrane systems such as, immobilized artificial membrane (IAM) and parallel artificial membrane permeation assay (PAMPA) do not consist of a lipid bilayer. Despite these limitations, artificial membrane assays can be useful to study the relationship between structures and lipophilicity.

A. Immobilized Artificial Membrane (IAM) Columns

IAM columns are essentially reverse-phase liquid chromatographic columns where the usual hydrocarbon phase that coats the solid support is replaced with lipids. These lipids are supposed to mimic the lipid environment of the cell membrane. IAM columns have been developed under the premise that permeation of solute across the cell membrane is limited by its ability to partition into the lipid domain. Compounds that interact with the lipid phase of IAM have longer retention in the column. Presumably, a compound that has a long retention in the column (i.e. large capacity factor, k') should have a good

permeability across lipid bilayers. This assumption may not be accurate because for a compound to cross the epithelial cell layer it must also diffuse across the outer and inner

leaflets of the lipid bilayer and back into the aqueous environment of the cytoplasm. One of the perceived advantages of this technique is that it can be conducted in a higher throughput compared with cell-based permeability assays. Because it eliminates the need to use a biological system, it can be faster; however, in most cases, the bottleneck in the evaluation of permeability is drug analysis, which is not eliminated with this technique. Unlike usual LC/MS and LC/MS/MS which require limited chromatographic separation, this technique requires greater separation because the absorption potential is based on k',

the capacity factor, which requires separation of a hydrophilic, which has poor retention.

The system has been validated by showing correlation between k' and permeability across Caco-2 cell monolayers . Closer examination shows potential problems with this type of validation. This validation method has a problem. Because k' is a physicochemical measurement, it is reasonable to expect reproducible results from different laboratories. However, considering the dramatic interlaboratory variability in Caco-2 permeability measurements, it is unrealistic to expect good correlations between k' and Caco-2 permeability in different laboratories. The conclusions drawn from this study were that, as is the case for log P or log D, IAM columns may provide a reasonable permeability ranking for homologous, but not for

diverse series of compounds. Some of the general disadvantages of this technique are that it does not consider the potential role of paracellular transport, carrier-mediated

transport, drug metabolism, and efflux transporters on permeability. Overlooking the possible influence of all these factors on intestinal permeability seems dangerous and may provide a very limited view of the true intestinal absorption potential of a compound.

B. Parallel Artificial Membrane Permeation Assay (PAMPA)

PAMPA is run in a 96-well plate that consists of two parts. The bottom is a standard 96-well plate. All the wells at the bottom are filled with buffer. The top part contains a series of filters, which match with the wells in the bottom part. One half of the filters on the top part are impregnated with an organic solvent, which supposedly mimics the cell membrane, and the other half are wetted with methanol/buffer. The drug solution is applied to the top filters and the rate of appearance in the bottom wells should reflect the diffusion across the lipid layer. A recent study found a good correlation between flux in the PAMPA system and % absorption in humans for a selected series of compounds PAMPA was portrayed as a high throughput assay because, unlike the IAM, it does not

require HPLC analysis. However, it has important disadvantages. For example, it requires UV absorbance, which many compounds do not exhibit. The limited sensitivity of UV detection and the small diffusional surface area may require long experimental time. The incubation time a recent study was 15 hours . In a screening program, such a long incubation time may present problems with unstable compounds. Like IAM columns, it ignores the role of enzymes, influx and efflux transporters, and the paracellular pathway in intestinal drug absorption.

VI. IN SILICO METHODS

Assessment of absorption potential using computational methods instead of experimental data has promoted in recent years. The attractiveness of this strategy is undisputable. The

suitability of chemical structures could be determined even  before the compound is synthesized. This could make it possible to save money, time and effort spent in bad

compounds. It also has an ethical component in that it would avoid the need to use either animals or animal tissues to test absorption potential. Within the in silico screening area there have been a number of approaches.

For instance, Lipinski's "rule of five" predicts poor absorption for compounds exhibiting the following characteristics:

1) molecular weight > 500,

2) H-bond donors >5,

3) H-bond acceptors > 5,

4) Mlog P> 4.15, and

5) cLOgP > 5 

This technique appears to have some success for closely related analog series. Other groups have used more traditional quantitative structure-transport relationship (QSTR) models to predict permeability. A recent QSTR study, found a reasonable correlation between observed and predicted Caco-2 permeability for small series of compounds. The fact that different equations had to be used is consistent with physicochemical methods such as, Log P, in that it may be useful only for small series of compounds. QSTR models tend to be deterministic, but usually provide very little understanding of the molecular determinants of permeability. A different approach has consisted in generating numerous molecular descriptors from structural information on the compounds and to assess the importance of these descriptors in predicting permeability. These studies have shown that absorption is influenced by H-bonding capacity and molecular size, but that the dynamic polar surface area (PSAd) was the stronger predictor of absorption.

While work in this area continues and several software packages claim to be able to predict absorption and/or Caco- 2 permeability, the common theme voiced by several

researchers is the lack of reliable data on which to base the development of the models. Unless the models are derived from diverse and reliable databases, their utility will

probable remain limited to compounds closely related to the training sets used to develop the models.