Absorption and Drug Development: Solubility, Permeability, and Charge State

PREFACE

In the nine years since the first edition of Absorption and Drug Development, a number of advances have been made, especially in the permeability methods. Several PAMPA models based on targeted lipid formulations have been described by pharmaceutical researchers. New data processing procedures were introduced to interpret permeability–pH dependence (gradient- and iso-pH) in PAMPA, as well as in cultured epithelial cell lines (e.g., Caco-2, MDCK), in primary endothelial cultured cells [e.g., porcine brain microcapillary endothelial cells (BMEC) and human BMEC], and in the rodent in situ brain perfusion model. The first PAMPA models specifically directed at modeling the blood–brain barrier (BBB) permeability have been reported. PAMPA models for skin penetration have been described. Even areas of solubility data analysis have seen some progress.

In the first edition, the pKa and solubility sections were sketchy, more like reviews than book chapters. The original permeability chapter was long and focused on the early stages of the evolution of what came to be known as the Double-Sink PAMPA method. Upon reflection, the need for a more balanced coverage was evident.

In this second edition, most of the original PAMPA material has been scrapped and has been replaced by descriptions and applications of models based on the more recent research described in literature, drawing on over 30 PAMPA-related papers published from the group at pION INC that the author headed. Also, two new chapters have been added: Chapter 8 (Permeability: Caco-2/MDCK) and Chapter 9 (Permeability: Blood–Brain Barrier). The pKa chapter has been vastly expanded. The potentiometric technique is covered comprehensively, but the treatment is still slim on UV and other methods. The new origin-shifted Yasuda–Shedlovsky (OSYS) method revealed some novel insights about how to treat insoluble acids and bases differently. The solubility chapter has been brought up to date with many examples of the treatment of practically insoluble test compounds. It was tempting to add a new chapter on dissolution, but the size of the book would have exceeded the planned limit. It was thought that a separate treatment of solubility–dissolution would best be left to a future project.

All of the database tables have been reviewed and updated with more values. The pKa table now has more than 900 entries, with many determined at 37°C. New tables have been added to each of the permeability chapters, with extensive listings of Double-Sink PAMPA, PAMPA–BBB, Caco-2/MDCK, multispecies BMEC, and in situ brain perfusion (PS) values. The introductory chapter, Chapter 1, has been updated, since the R&D paradigm of pharmaceutical research has undergone important changes since the first edition.

Based on the content of the first edition, the author twice taught an informal 10-week course at King’s College, London. There were other, smaller, teaching presentations at the University of Helsinki on two separate occasions. The notion of the book serving an educational purpose was recurrent. Several pharmacy and pharmaceutical sciences university departments have courses in physical pharmacy and pharmaceutics based on Martin’s classical textbook, Physical Pharmacy and Pharmaceutical Sciences (now in its sixth edition). This is an excellent and comprehensive text for a two-semester introductory graduate course. The author taught selected topics from it as a guest lecturer at Northeastern University, Boston, on a couple of occasions. However, one cannot learn how to do physicochemical measurements (e.g., pKa, solubility, and permeability) from Martin’s book alone. Therefore, a more advanced treatment of the physicochemical methods related to drug absorption is needed for pharmaceutics graduate students, especially those headed for careers in the pharmaceutical industry. The author has received comments from several professors who have used parts of the first edition of Absorption and Drug Development to supplement advanced pharmaceutics courses. Slanting the second edition toward an educational textbook was very tempting, but due to time constraints it was decided to leave that for a future separate add-on booklet to accompany the main text. Preparing useful questions and answers is not a minor project. The second edition still can be used to augment advanced graduate courses in pharmaceutics and as a reference for researchers in pharmaceutical R&D (and in some instances in agrochemical, environmental, and related industries). The author welcomes more feedback from academics and other readers about how the book can be improved, both as a teaching guide and as a reference.

The second edition is organized into 10 chapters. Chapter 1 describes the physicochemical measurement needs of pharmaceutical R&D, in a quickly changing environment. Chapter 2 defines the flux model, based on Fick’s laws of diffusion, in terms of solubility, permeability, and charge state (pKa), and lays the foundation for the rest of the book. Chapter 3 covers the topic of ionization constants: how to measure pKa constants well and quickly, and which strategies to use. It has been completely rewritten from the short previous version. Chapter 4 is about experimental methods of measuring partition coefficients, log P and log D. It contains a description of the Dyrssen dual-phase potentiometric method, which remains the gold standard technique for measuring log P of ionizable molecules, having the unique 10-orders-of-magnitude range (log P from 2 to +8). Chapter 5 considers the special topic of partition coefficients where the lipid phase is made of liposomes formed from bilayers of phospholipids. This chapter remains largely the same. Chapter 6 covers solubility measurements and has been broadly expanded. Chapter 7 describes PAMPA, the high-throughput artificial membrane permeability method originally introduced by Manfred Kansy and co-workers from Hoffmann–La Roche. The chapter has been substantially revised and remains a deep account of the rapidly developed important topic. Many hundreds of original measurements are tabulated in the chapter. Chapter 8 considers permeability measurements using epithelial cell models, such as Caco-2 and MDCK. Chapter 9 does so with endothelial cultured cell models, and it attempts to correlate these to animal in situ brain perfusion measurements of luminal permeability. Chapter 10 concludes with simple physicochemical property approximations. More than 1350 references and well over 200 drawings and 200 pages of tables substantiate the book as an extensively documented reference work.

I have many colleagues to thank for their thoughts, criticism, guidance, and opportunities for collaborations: Joan Abbott, Mike Abraham, Per Artursson, David Begley, Stephanie Bendels, Christel Bergström, Marival Bermejo, Li Di, Jennifer Dressman, Beate Escher, Bernard Faller, Holger Fischer, Norman Ho, Pranas Japertas, Paulius Jurgutis, Manfred Kansy, Ed Kerns, Stefanie Krämer, Chris Lipinski, Sibylle Neuhoff, Alanas Petrauskas, Tom Raub, Jean-Michel Scherrmann, Abu Serajuddin, Kiyohiko Sugano, Krisztina Takács-Novák, Bernard Testa, Björn Wagner, Han van de Waterbeemd, and Shinji Yamashita. I owe gratitude to many others, including my former colleagues at pION INC and Sirius Analytical Instruments Ltd. I left pION at the beginning of the year, to start in-ADME Research (ADME software and consulting) and to finish writing this book.

Salvatore Cisternino, Markus Fridén, Margareta Hammarlund-Udanaes, Krisztina Takács-Novák, and Kin Tam were most kind to read various chapters as the book was being written and offered many helpful suggestions, for which I am especially grateful.

Joan Abbott is a dear friend and has been a generous host on a number of occasions that I spent time writing and recharging in her group at King’s College, London.

I am especially privileged and grateful to have known Manfred Kansy as a friend for the last 20 years.

I would also like to thank Joyce Saltalamachia for her love and support, as she put up with a lot during my 12 months of writing and other things.

ALEX AVDEEF

Cambridge, Massachusetts

September 2011

PREFACE TO THE FIRST EDITION

This book is written for the practicing pharmaceutical scientist involved in ADME measurements, who needs to communicate with medicinal chemists persuasively, so that new synthesized molecules will be more drug-like. ADME is all about a day in the life of a drug molecule (absorption, distribution, metabolism, excretion). Specifically, this book attempts to describe the state of the art in measurement of ionization constants (pKa), oil–water partition coefficients (log P/log D), solubility, and permeability (artificial phospholipid membrane barriers). Permeability is covered in considerable detail, based on a newly developed methodology known as PAMPA (Parallel Artificial Membrane Permeability Assay).

These physical parameters form the major components of physico­chemical profiling (the Absorption in ADME) in the pharmaceutical industry, from drug discovery through drug development. However, there are opportunities to apply the methodologies in other fields, particularly the agrochemical and environmental industries. In addition, new applications to augment animal-based models in the cosmetics industry may be interesting to explore.

It has been the author’s observation that graduate programs in pharmaceutical sciences often neglect adequately to train students in these classical solution chemistry topics. Often young scientists in pharmaceutical companies are assigned the task of measuring some of these properties in their projects. Most find the learning curve somewhat steep. In addition, experienced scientists in mid careers come upon the topic of physicochemical profiling for the first time, and they find few resources to draw upon outside of the primary literature.

The idea for a book on the topic has morphed through various forms, beginning with focus on the subject of metal-binding to biological ligands, when the author was a postdoc in Professor Ken Raymond’s group at University of California, Berkeley. When the author was an Assistant Professor of Chemistry at Syracuse University, every time the special topics course on speciation analysis was taught, more notes were added to the book. After five years, more than 300 pages of hand-scribbled notes and derivations accumulated, but no book emerged. Some years later, a subsection of the original notes acquired a binding and saw light in the form of Applications and Theory Guide to pH-Metric pKa and log P Measurement, out of the early effort in the start-up of Sirius Analytical Instruments Ltd., in Forest Row, a charming four-pub village at the edge of Ashdown Forest, south of London. At Sirius, the author was involved in teaching a comprehensive three-day training course to advanced users of pKa and log P measurement equipment manufactured by Sirius. The trainees were from pharmaceutical and agrochemical companies, and they shared many new ideas during the courses. Over the last decade, Sirius has standardized the measurement of pKa values in the pharmaceutical and agrochemical industries. Some 50 courses later, the practice continues at another young company, pION, located along high-tech highway 128, north of Boston, Massachusetts. The list of topics has expanded over the last 12 years, to cover solubility, dissolution, and permeability, as new instruments were developed. Last year, an opportunity to write a review article came up, and a bulky piece appeared in Current Topics in Medicinal Chemistry, entitled Physicochemical Profiling (Solubility, Permeability and Charge State). In reviewing that manuscript, Cynthia Berger (pION) said that with a little extra effort, this could be a book. Further encouragement came from Bob Esposito of John Wiley & Sons. My colleagues at pION were kind about my taking a sabbatical in England, to focus on the writing. I was privileged to join Professor Joan Abbott’s neuroscience laboratory at King’s College London for three months, where I conducted an informal 10-week graduate short course on the topics of this book, as the material was freshly written. After hours, it was my pleasure to jog with my West London Hash House Harrier friends. As the chapter on permeability was being written, my very capable colleagues at pION were quickly measuring permeability of membrane models freshly inspired by the book writing. It is due to their efforts that Chapter 7 is loaded with so much original data, out of which emerged the "Double-Sink" PAMPA model for predicting human intestinal permeability. Per Nielsen (pION) reviewed the manuscript as it slowly emerged, with a keen eye. Many late-evening discussions with him led to freshly inspired insights, now imbedded in various parts of the book.

The book is organized into eight chapters. Chapter 1 describes the physicochemical needs of pharmaceutical research and development. Chapter 2 defines the flux model, based on Fick’s laws of diffusion, in terms of solubility, permeability, and charge state (pKa), and lays the foundation for the rest of the book. Chapter 3 covers the topic of ionization constants: how to measure pKa values well and quickly, and which methods to use. Bjerrum analysis is revealed as the secret weapon behind the most effective approaches. Chapter 4 is about experimental methods of measuring partition coefficients, log P and log D. It contains a description of the Dyrssen dual-phase potentiometric method that truly is the gold standard method for measuring log P of ionizable molecules, having the unique 10 orders of magnitude range (log P from 2 to +8). High-throughput methods are also described. Chapter 5 considers the special topic of partition coefficients where the lipid phase is made of liposomes formed from vesicles made of bilayers of phospholipids. Chapter 6 dives into solubility measurements. A unique approach, based on the Dissolution Template Titration method, has demonstrated capabilities to measure solubilities as low as one nanogram per milliliter. In addition, high-throughput microtiter plate UV methods for determining thermodynamic solubility constants are described. At the ends of Chapters 3–6, an effort has been made to collect tables of critically selected values of the constants of drug molecules, the best available values. Chapter 7 describes PAMPA, the high-throughput method recently introduced by Manfred Kansy et al. of Hoffmann–La Roche. Chapter 7 is the first thorough account of the topic and takes up almost half of the book. Nearly 4000 original measurements are tabulated in the chapter. Chapter 8 concludes with simple rules. Over 600 references and well over 100 drawings substantiate the book.

Professor Norman Ho (University of Utah) was very kind to critically read the permeability chapter and comment on the various derivations and concepts. His unique expertise on the topic spans many decades. His thoughts and advice (15 pages of handwritten notes) inspired me to rewrite some of the sections in that chapter. I am very grateful to him. I am grateful to other colleagues at pION who expertly performed many of the measurements of solubility and permeability, which are presented in the book: Chau Du, Jeffrey Ruell, Melissa Strafford, Suzanne Tilton, and Oksana Tsinman. In addition, I thank Dmytro Voloboy and Konstantin Tsinman for their help in database, computational, and theoretical matters. The helpful discussion with many colleagues, particularly Manfred Kansy and Holger Fischer at Hoffmann La-Roche, Ed Kerns and Li Di at Wyeth Pharmaceuticals, and those at Sirius Analytical Instruments, especially John Comer and Karl Box, are gratefully acknowledged. Helpful comments from Professors John Dearden (Liverpool John Moores University) and Hugo Kubinyi (Heidelberg University) are greatly appreciated. I also thank Professor Anatoly Belyustin (St. Peterburgh University) for pointing out some very relevant Russian literature. Chris Lipinski (Pfizer) has given me a lot of good advice over the last 10 years about instrumentation and pharmaceutical research, for which I am grateful. Collaborations with Professors Krisztina Takács-Novák (Semmelweis University, Budapest) and Per Artursson (Uppsala University) have been very rewarding. James McFarland (Reckon.Dat) and Alanas Petrauskas (Pharma Algorithms) have been my teachers of in silico methods. I am in debt to Professor Joan Abbott and Dr. David Begley for allowing me to spend three months in their laboratory at King’s College London, where I learned a lot about the blood–brain barrier. Omar at Cafe Minon, Warwick Street in Pimlico, London, was kind to let me spend many hours in his small sandwich shop, as I wrote several papers and drank a lot of coffee. Lasting thanks go to David Dyrssen and the late Jannik Bjerrum for planting the seeds of most interesting and resilient pH-metric methodologies, as well as to Professor Bernard Testa of Lausanne University for tirelessly fostering the white light of physicochemical profiling. My congratulations to him on the occasion of his retirement.

ALEX AVDEEF

Boston, Massachusetts

September 2002

ABBREVIATIONS

NOMENCLATURE

COMMERCIAL TRADEMARKS

pCEL-X™ and μDISS-X™ are trademarks of in-ADME Research. Double-Sink™, Prisma™, PAMPA Evolution™, μSOL Evolution™, and STIRWELL™ are commercial trademarks of pION INC. Transwell®, Freedom Evo®, Biomek-FX®, and Excel® are the registered trademarks of Corning, Tecan, Beckman Coulter, and Microsoft, respectively.

1

INTRODUCTION

The search for new drugs is a long process. Attrition is high and the costs keep escalating (now perhaps as high as $2 billion per marketed drug). The traditional discovery–development models are undergoing change, as many pharmaceutical companies reign in the R&D costs, by consolidating research sites, downsizing research staff, engaging in more outside collaborations, and outsourcing.

1.1 BULLDOZER SEARCHING FOR A NEEDLE IN A HAYSTACK?

Although the last decade has led to improvements in attrition due to poor pharmacokinetic profiles of discovery compounds, drug absorption continues to be an important issue in modern pharmaceutical research and development. The search for new drugs is daunting, expensive, and highly risky, but potentially highly rewarding.

If chemicals were confined to molecular weights of less than 600 Da and consisted of common atoms, the chemistry space is estimated to contain 10⁴⁰ to 10¹⁰⁰ molecules, an impossibly large space to search for potential drugs [1]. To address this limitation of vastness, maximal chemical diversity [2] was applied in constructing large experimental screening libraries. It’s now widely accepted that the quality of leads is more important than the quantity. Traditionally, large compound libraries have been directed at biological targets to identify active molecules, with the hope that some of these hits may someday become drugs. The pre-genomic era target space was relatively small: Less than 500 targets had been used to discover the known drugs [3]. This number may expand to several thousand in the next few years as genomics-based technologies and better understanding of protein–protein interactions uncover new target opportunities [4, 5]. Of the estimated 3000 new targets, only about 20% are commercially exploited [5]. Due to unforeseen complexities of the genome and biologic systems, it is taking a lot longer and is more expensive to exploit the new opportunities than originally thought [5–8].

Although screening throughputs have massively increased over the past 20 years (at great cost in set up and run), lead discovery productivity has not necessarily increased accordingly [5–8]. C. Lipinski has suggested that maximal chemical diversity is an inefficient library design strategy, given the enormous size of the chemistry space, and especially that clinically useful drugs appear to exist as small tight clusters in chemistry space: … one can make the argument that screening truly diverse libraries for drug activity is the fastest way for a company to go bankrupt because the screening yield will be so low [1]. Hits are made in pharmaceutical companies, but this is because the most effective (not necessarily the largest) screening libraries are highly focused, to reflect the putative tight clustering. Looking for ways to reduce the number of tests, to make the screens smarter, has an enormous cost reduction implication.

Figure 1.1 sketches out the process of drug exploration, discovery, and development followed at several pharmaceutical companies in the early 2000s [9–12]. A large pharmaceutical company may screen 100,000 to 1,000,000 molecules for biological activity each year. Some 3000–10,000 hits are made. Most of these molecules, however potent, do not have the right physicochemical, stability, and safety properties. Large pharmaceutical companies promote about 12 molecules into preclinical development each year. Only about 5 in 12 candidates survive after Phase I (Figure 1.1). A good year sees perhaps just one molecule reach the product stage after 9 molecules enter first-in-man clinical testing [6]. For that molecule, the start-to-finish may have taken 14 years (Figure 1.1).

Figure 1.1 Chart summarizing the various stages in the evolution of a drug product. Included are estimated times at each stage, the cost of each stage, and the relative attrition rates.

Based on studies taken from multiple sources [7–10].

The molecules that fail have off-target activity or poor side effects profiles. Unfortunately, animal models have been weak predictors of efficacy and/or safety in humans [7]. The adverse reactions in humans are sometimes not discovered until the drug is on the market in large-scale use in humans.

In 2001, a drug product cost about $880 million to bring out to market—which included the costs of numerous failures (Figure 1.1). In 2010, the cost was closer to $2 billion/approval [7]. It has been estimated that about 33% of the molecules that reach preclinical development are eventually rejected due to ADME (absorption, distribution, metabolism, excretion) problems. Other attrition causes are lack of efficacy (33%) and toxicity (34%). Much more money is spent on compounds that fail than on those that succeed. The industry has started to respond by attempting to screen out those molecules with poor ADME properties during discovery, before the molecules reach development. However, that has led to another challenge: how to do the additional screening quickly enough [13]. An undesirable consequence of cheap and quick assays it that their quality is low [5].

Combinatorial chemistry programs have tended to select for higher-molecular-weight molecules, predictably low in solubility. Early warning tools, such as Lipinski’s Rule of Five [1] and simple computer programs that predict solubility and other properties from 2-D structure [14, 15], attempt to weed out such molecules early in discovery programs. Still, many solubility-problematic molecules remain unrecognized in early studies, due to the overly simplistic methods used to measure solubility in discovery [16]. More accurate (but still fast) solubility [16–19] (Chapter 6) and artificial membrane permeability [20–24] (Chapter 7) methods in the candidate selection stage in pharmaceutical R&D have proven to be particularly helpful for recognizing at a much earlier time the truly problematic molecules. It had even been suggested that screening for future formulation efficacy (pH and excipient effects on solubility and permeability) of candidates could be justified, if the methods were fast, compound-sparing, cost effective, and reasonably accurate [16, 18].

1.2 AS THE PARADIGM TURNS

As a consequence of the increased and unsustainable cost of bringing out a therapeutic product, many pharmaceutical companies have begun to change the way discovery and development are done [5]:

Size and scope of internal research capabilities are decreasing, as more outsourcing is considered, not only in discovery, but also in development.

Several companies have rearranged internal structures to be smaller biotech-like units.

External collaborations with small biotech companies and academia have increased.

Many in the industry predict that more biologic therapies will emerge (which have lower Phase II attrition [6]), and the emphasis on small molecules may decrease.

Strategies of discovery are changing [7]:

Development of multitargeted therapeutics will increase.

Whole pathway approaches, drawing on increasing understanding of protein–protein interactions, will be increasingly explored.

Biology-driven drug discovery, starting with a specific disease model and a pathway, benefitting from external collaborations with academic groups.

Analysis of multigenic complex diseases.

Network pharmacology.

Obtaining early proof of concepts, with small clinical studies and/or applying microdosing.

The open innovation model (OIM) [8] involves the progression of discovery and development that’s different from that depicted in Figure 1.1. An attrition funnel will start with many test compounds. Even at the early stage, ideas and technologies may be either in-licensed or out-licensed. At later optimization stages, two-way collaborations with academic labs will play an increasing role. Product in-licenses will be considered. Near the product launch stage, line extensions via partners and joint ventures will become increasingly popular. In the OIM, intellectual property would be selectively distributed and proactively managed and shared to create value that otherwise would not surface.

1.3 SCREEN FOR THE TARGET OR ADME FIRST?

Most commercial combinatorial libraries, some of which are very large and may be diverse, have a very small proportion of drug-like molecules [1]. Should only the small drug-like fraction be used to test against the targets? The existing practice is to screen for the receptor activity before drug-likeness. The reasoning is that structural features in molecules rejected for poor ADME properties may be critical to biological activity related to the target. It is believed that active molecules with liabilities can be modified later by medicinal chemists, with minimal compromise to potency. Lipinski [1] suggested that the order of testing may change in the near future, for economic reasons. He adds that looking at data already available from previous successes and failures may help to derive a set of guidelines to apply to new compounds. When a truly new biological therapeutic target is examined, nothing may be known about the structural requirements for ligand binding to the target. Screening may start as more or less a random process. A library of compounds is tested for activity. Then computational models are constructed based on the results, and the process is repeated with newly synthesized molecules, perhaps many times, before adequately promising compounds are revealed. With large numbers of molecules, the process can be costly. If the company’s library is first screened for ADME properties, that screening is done only once. The same molecules may be recycled against existing or future targets many times, with knowledge of drug-likeness to fine-tune the optimization process. If some of the molecules with very poor ADME properties are judiciously filtered out, the biological activity testing process would be less costly. But the order of testing (activity versus ADME) is likely to continue to be the subject of future debates [1].

1.4 ADME AND MULTIMECHANISM SCREENS

In silico property prediction is needed more than ever to cope with the screening overload [14, 15]. Improved prediction technologies are continuing to emerge. However, reliably measured physicochemical properties to use as training sets for new target applications have not kept pace with the in silico methodologies.

Prediction of ADME properties should be simple, since the number of descriptors underlying the properties is relatively small, compared to the number associated with effective drug-receptor binding space. In fact, prediction of ADME is difficult. The current ADME experimental data reflects a multiplicity of mechanisms, making prediction uncertain. Screening systems for biological activity are typically single mechanisms, where computational models are easier to develop [1].

For example, aqueous solubility is a multimechanism system. It is affected by lipophilicity, H-bonding between solute and solvent, intra- and intermolecular H-bonding, electrostatic bonding (crystal lattice forces), and charge state of the molecule. When the molecule is charged, the counterions in solution may affect the measured solubility of the compound. Solution microequilibria occur in parallel, affecting the solubility. Many of these physicochemical factors are not well understood by medicinal chemists, who are charged with making new molecules that overcome ADME liabilities without losing potency.

Another example of a multimechanistic probe is the Caco-2 permeability assay (Chapter 8). Molecules can be transported across the Caco-2 monolayer by several mechanisms operating simultaneously, but to varying degrees: transcellular passive diffusion, paracellular passive diffusion, lateral passive diffusion, active influx or/and efflux mediated by transporters, passive transport mediated by membrane-bound proteins, receptor-mediated endocytosis, pH-gradient- and electrostatic-gradient-driven mechanisms, and so on (Chapter 2). The P-glycoprotein (Pgp) efflux transporter can be saturated if the solute concentration is high enough during the assay. If the substance concentration is very low (perhaps because not enough of the compound is available during discovery, or due to low solubility), the importance of efflux transporters in gastrointestinal tract (GIT) absorption can be overestimated, providing the medicinal chemist with an overly pessimistic prediction of intestinal permeability [1, 25]. Drug metabolism in some in vitro cellular systems can further complicate the assay outcome.

Compounds from traditional drug space (common drugs—readily available from chemical suppliers), often chosen for studies by academic laboratories for assay validation and computational model-building purposes, can lead to misleading conclusions when the results of such models are applied to real [12] discovery compounds, which most often have extremely low solubilities [25].

Computational models for single-mechanism assays (e.g., biological receptor affinity) get better as more data are accumulated [1]. Computational models for multimechanism assays (e.g., solubility, permeability, charge state), in contrast, get worse as more measurements are accumulated [1]. Predictions of human oral absorption using Caco-2 permeability values can look very impressive when only a small number of molecules is considered. However, good correlations deteriorate as more molecules are included in the plot, and predictivity soon becomes tenuous. The solution to this dilemma is to carry out single-mechanism ADME experimental assays and to construct single-mechanism ADME computational models. The ADME area is at least 5 or more years behind the biology therapeutic target area in this respect [1].

1.5 ADME AND THE MEDICINAL CHEMIST

Although ADME assays are usually performed by analytical chemists, medicinal chemists—the molecule makers—need to have some understanding of the physicochemical processes in which the molecules participate.

It is now almost a century since Overton and Meyer first demonstrated the existence of a relationship between the biological activity of a series of compounds and some simple physical property common to its members. In the intervening years the germ of their discovery has grown into an understanding whose ramifications extend into medicinal chemistry, agrochemical and pesticide research, environmental pollution, and even, by a curious reinvention of familiar territory, some areas basic to the science of chemistry itself. Yet its further exploitation was long delayed. It was 40 years later that Ferguson at ICI [AstraZeneca] applied similar principles to a rationalization of the comparative activity of gaseous anaesthetics, and 20 more were to pass before the next crucial step was formulated in the mind of Hansch. … Without any doubt, one major factor [for delay] was compartmentalism. The various branches of science were much more separate then than now. It has become almost trite to claim that the major advances in science take place along the borders between its disciplines, but in truth this happened in the case of what we now call Hansch analysis, combining as it did aspects of pharmacy, pharmacology, statistics, and physical organic chemistry. Yet there was another feature that is not so often remarked, and one with a much more direct contemporary implication. The physical and physical organic chemistry of equilibrium processes—solubility, partitioning, hydrogen bonding, etc.—is not a glamorous subject. It seems too simple. Even though the specialist may detect an enormous information content in an assemblage of such numbers, to synthetic chemists used to thinking in three-dimensional terms they appear structureless, with no immediate meaning that they can visually grasp. Fifty years ago it was the siren call of Ehrlich’s lock-and-key theory that deflected medicinal chemists from a physical understanding that might otherwise have been attained much earlier. Today it is glamour of the television screen. No matter that what is on display may sometimes possess all the profundity of a five-finger exercise. It is visual and therefore more comfortable and easier to assimilate. Similarly, MO theory in its resurgent phase combines the exotic appeal of a mystery religion with a new-found instinct for three-dimensional colour projection which really can give the ingénue the impression that he understands what it is all about. There are great advances and great opportunities in all this, but nevertheless [there is] a concomitant danger that medicinal chemists may forget or pay insufficient attention to hurdles the drug molecule will face if it is actually to perform the clever docking routine they have just tried out: hurdles of solubilization, penetration, distribution, metabolism and finally of its nonspecific interactions in the vicinity of the active site, all of them the result of physical principles on which computer graphics has nothing to say. Such a tendency has been sharply exacerbated by the recent trend, for reasons of cost as much as of humanity, to throw the emphasis upon in vitro testing. All too often, chemists are disconcerted to discover that the activity they are so pleased with in vitro entirely fails to translate to the in vivo situation. Very often, a simple appreciation of basic physical principles would have spared them this disappointment; better, [it] could have suggested in advance how they might avoid it. We are still not so far down the path of this enlightenment as we ought to be. What is more, there seems a risk that some of it may fade if the balance between a burgeoning receptor science and these more down-to-earth physical principles is not properly kept.—Peter Taylor [26].*

In 1990, Taylor [26] described physicochemical profiling in a comprehensive and richly descriptive way, but much has happened since then. Then, instrument companies took no visible interest in making pKa (Chapter 3), log P (Chapters 4 and 5), or solubility (Chapter 6) analyzers; it did not occur to anyone to do PAMPA (Chapter 7). Combinatorial chemistry, HTS, Caco-2 (Chapter 8), IAM, and CE were largely unknown. Thus it is a good time to take stock of what can be learned from the work of the last two decades.

1.6 THE ABSORPTION IN ADME

This book focuses on physicochemical profiling in support of improved prediction methods for the absorption in ADME. Metabolism and other components of ADME will be beyond the scope of this book. Further­more, properties related to passive absorption will be the focus, and active transport mechanisms will be considered only indirectly. The most important physicochemical parameters associated with passive absorption are acid–base character (which determines the charge state of a molecule in a solution of a particular pH), lipophilicity (which determines distribution of a molecule between the aqueous and the lipid environments), solubility (which limits the concentration that a dosage form of a molecule can present to the solution and the rate at which the molecule dissolves from the solid form), and membrane permeability (which determines how quickly molecules can cross membrane barriers). Current state of the art in measurement of these properties, as the ever important function of pH, will be discussed in depth in this book.

1.7 IT IS NOT JUST A NUMBER, IT IS A MULTIMECHANISM

Drugs exert their therapeutic effects through reactions with specific receptors. Drug-receptor binding depends on the concentration of the drug near the receptor. Its form and concentration near the receptor depend on its physical properties. Orally administered drugs need to be dissolved at the site of absorption in the GIT, and need to traverse several membrane barriers before receptor interactions can commence. As the drug distributes into the various compartments of the body, a certain (small) portion finds itself at the receptor site. Transport and distribution of most drugs are affected by passive diffusion, which depends on lipophilicity, since lipid barriers need to be crossed [27]. Passive transport is well described by the principles of physical chemistry [27–29].

The goal of this book is to examine the components of the multimechanistic processes related to charge state: the pKa of molecules (Chapter 3), lipophilicity (Chapters 4 and 5), solubility (Chapter 6), and permeability (Chapters 7–9), with the aim of advancing improved strategies for in vitro assays related to drug absorption. In high-throughput screening (HTS) these parameters are sometimes viewed simply as numbers, quickly and roughly determined, to be used to rank molecules into good and bad classes. An attempt will be made to examine this important aspect. In addition, how fundamental, molecular-level interpretations of the physical measurements can help to improve the design of the profiling assays will be examined, with the aim of promoting the data fodder of HTS to a higher level of quality, without compromising the need for high speed [16–24]. Quality measurements in large quantities will lead to improved in silico methods. Simple rules (presented in visually appealing ways), in the spirit of Lipinski’s rule of fives, will be sought, of use not only to medicinal chemists but also to preformulators. This book attempts to make easier the dialog between the medicinal chemists charged with modifying test compounds and the pharmaceutical scientists charged with physicochemical profiling, who need to communicate assay results in an optimally effective manner.

Note

* This excerpt was published in Comprehensive Medicinal Chemistry, Vol. 4, Peter J. Taylor, Hydrophobic Properties of Drugs, pp. 241–294, Copyright Elsevier (1990). Reproduced with permission from Elsevier.

REFERENCES

 1. Lipinski, C. A. Drug-like properties and the causes of poor solubility and poor permeability. J. Pharmacol. Toxicol. Methods 44, 235–249 (2000).

 2. Martin, E. J.; Blaney, J. M.; Siani, M. A.; Spellmeyer, D. C.; Wong, A. K.; Moos, W. H. Measuring diversity: Experimental design of combinatorial libraries for drug discovery. J. Med. Chem. 38, 1431–1436 (1995).

 3. Drews, J. Drug discovery: A historical perspective. Science 287, 1960–1963 (2000).

 4. Pickering, L. Developing drugs to counter disease. Drug Discov. Dev. Feb., 44–47 (2001).

 5. Perrior, T. Overcoming bottlenecks in drug discovery. Drug Discov. World 29–33 (Fall 2010).

 6. Kola, I.; Landis, J. Can the pharmaceutical industry reduce attrition rates? Nature Rev. Drug Discov. 3, 711–715 (2004).

 7. Haberman, A. B. Overcoming phase II attrition problem. Gen. Eng. Biotech. News 29, 63–67 (2009).

 8. Hunter, J. Is the pharmaceutical industry open for innovation? Drug Discov. World Fall, 9–14 (2010).

 9. Allan, E.-L. Balancing quantity and quality in drug discovery. Drug Discov. World Winter, 71–75 (2002/2003).

10. Browne, L. J.; Taylor, L. L. Drug Discov. World Fall, 71–77 (2002).

11. Kerns, E. H.; Di, L. Drug-like Properties: Concepts, Structure Design and Methods, Academic Press, Amsterdam, 2008.

12. Rydzewski, R. M. Real World Drug Discovery—A Chemist Is Guide to Biotech and Pharmaceutical Research, Elsevier, Amsterdam, 2008.

13. Lipinski, C. A.; Lombardo, F.; Dominy, B. W.; Feeney, P. J. Experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings. Adv. Drug Deliv. Rev. 23, 3–25 (1997).

14. Algorithm Builder v1.8; ADME Boxes v4.9; ACD/pKa Database in ACD/ChemSketch v3.0; ACD/Solubility DB. Advanced Chemistry Development Inc., Toronto, Canada (www.ACD/Labs.com).

15. MarvinSketch v5.3.7. ChemAxon, Budapest, Hungary (www.chemaxon.com).

16. Glomme, A.; März, J.; Dressman, J. B. Comparison of a miniaturized shake-flask solubility method with automated potentiometric acid/base titrations and calculated solubilities. J. Pharm. Sci. 94, 1–16 (2005).

17. Bergström, C. A. S.; Luthman, K.; Artursson, P. Accuracy of calculated pH-dependent aqueous drug solubility. Eur. J. Pharm. Sci. 22, 387–398 (2004).

18. Avdeef, A.; Bendels, S.; Tsinman, O.; Kansy, M. Solubility—Excipient classification gradient maps. Pharm. Res. 24, 530–545 (2007).

19. Avdeef, A. Solubility of sparingly-soluble drugs. [Dressman, J; Reppas, C. (eds.). Special issue: The Importance of Drug Solubility]. Adv. Drug Deliv. Rev. 59, 568–590 (2007).

20. Kansy, M.; Avdeef, A.; Fischer, H. Advances in screening for membrane permeability: High-resolution PAMPA for medicinal chemists. Drug Discov. Today: Technologies 1, 349–355 (2005).

21. Avdeef, A.; Artursson, P.; Neuhoff, S.; Lazarova, L.; Gräsjö, J.; Tavelin, S. Caco-2 permeability of weakly basic drugs predicted with the double-sink PAMPA c01ue001 method. Eur. J. Pharm. Sci. 24, 333–349 (2005).

22. Avdeef, A. The rise of PAMPA. Expert Opinion Drug Metab. Toxicol. 1, 325–342 (2005).

23. Avdeef, A.; Bendels, S.; Di, L.; Faller, B.; Kansy, M.; Sugano, K.; Yamauchi, Y. PAMPA—A useful tool in drug discovery. J. Pharm. Sci. 96, 2893–2909 (2007).

24. Sugano, K.; Kansy, M.; Artursson, P.; Avdeef, A.; Bendels, S.; Di, L.; Ecker, G. F.; Faller, B.; Fischer, H.; Gerebtzoff, G.; Lennernäs, H.; Senner, F. Coexistence of passive and active carrier-mediated uptake processes in drug transport: A more balanced view. Nature Rev. Drug Discov. 9, 597–614 (2010).

25. Lipinski, C. A. Avoiding investment in doomed drugs—Is solubility an industry wide problem? Curr. Drug Discov. Apr, 17–19 (2001).

26. Taylor, P. J. Hydrophobic properties of drugs. In: Hansch, C.; Sammes, P. G.; Taylor, J. B. (eds.). Comprehensive Medicinal Chemistry, Vol. 4, Pergamon, Oxford, 1990, pp. 241–294.

27. Kubinyi, H. Lipophilicity and biological activity. Arzneim.-Forsch./Drug Res. 29, 1067–1080 (1979).

28. van de Waterbeemd, H.; Smith, D. A.; Jones, B. C. Lipophilicity in PK design: Methyl, ethyl, futile. J. Comp.-Aided Molec. Design 15, 273–286 (2001).

29. van de Waterbeemd, H.; Smith, D. A.; Beaumont, K.; Walker, D. K. Property-based design: Optimization of drug absorption and pharmacokinetics. J. Med. Chem. 44, 1313–1333 (2001).

2

TRANSPORT MODEL

This chapter defines the relationship between ionization constants, solubility, and permeability as a function of pH, using the framework of Fick’s laws of diffusion. In simplest terms, the flux across a membrane barrier is a product function of solubility and permeability in saturated solutions. The contrast between the flux profile and the pH-Partition Hypothesis is addressed. Following the theoretical backdrop, a brief summary of the properties of the gastrointestinal tract is presented. The intestinal structure (folds, villi, microvilli, mucus layer) of the epithelial cell monolayer is considered. Transcellular and paracellular permeation is defined. The pH microclimate at the intestinal surface is discussed briefly. The Biopharmaceutics Classification System from the FDA integrates concepts related to solubility, permeability, and pH.

2.1 PERMEABILITY–SOLUBILITY–CHARGE STATE AND PH-PARTITION HYPOTHESIS

Fick’s first law applied to a membrane [1–3] shows that passive diffusion of a solute is the product of the diffusivity and the concentration gradient of the solute inside the membrane. The membrane/water apparent partition coefficient, DMEM/W, relates the latter internal gradient to the external bulk-water concentration difference between the two solutions separated by the membrane. For an ionizable molecule to permeate by passive diffusion most efficiently, the molecule needs to be in its uncharged form at the membrane surface. This is the essence of the pH-Partition Hypothesis [4]. The amount of the uncharged form present at a given pH, which directly contributes to the flux, depends on several important factors, such as pH, binding to indigenous carriers (proteins and bile acids), self-binding (aggregate or micelle formation), and solubility (a solid-state form of self-binding). Low solubility enters the transport consideration as a thermodynamic speed attenuator, as a condition that lowers the opportunity for transport. In this way, permeability and solubility are the linked kinetic and thermodynamic parts of transport across a membrane.

Consider a vessel divided into two chambers, separated by a homogeneous lipid membrane. Figure 2.1 is a cartoon of such an arrangement. The left-hand side is the donor compartment, where the sample molecules are first introduced; the right-hand side is the receiver compartment, which at the start has no sample molecules. Fick’s first law applied to homogeneous membranes at steady state is a transport equation,

c02e001  (2.1)

where J is the flux, in units of mol·cm−2·s−1, where c02ue001 and c02ue002 are the concentrations, in mol·cm−3 units, of the uncharged form of the solute within the membrane at the two water-membrane boundaries (at positions x = 0 and x = h in Figure 2.1, where h is the thickness of the membrane in cm units) and where Dm is the diffusivity of the solute within the membrane, in units of cm²·s−1. At steady state, the concentration gradient, dCm/dx, within the homogeneous membrane is linear, so the difference may be used on the right-hand side of Eq. (2.1). Steady state takes about 3 min to be established in a membrane of thickness 125 µm [3], assuming the solution is very well stirred.

Figure 2.1 Transport model diagram, depicting two aqueous cells separated by a membrane barrier. The drug molecules are introduced in the donor cell. The concentration gradient in the membrane drives the molecules in the direction of the receiver compartment. The apparent partition coefficient, Kd, is 2.

[Reprinted from Avdeef, A. Curr. Topics Med. Chem. 1, 277–351 (2001). Reproduced with permission from Bentham Science Publishers, Ltd.]

The limitation of Eq. (2.1) is that measurement of concentrations of solute within different parts of the membrane is very inconvenient. However, since the distribution coefficients between bulk water and the membrane, log DMEM/W (the pH-dependent apparent partition coefficient), can be estimated (or possibly measured), Eq. (2.1) may be converted into a more accessible form,

c02e002  (2.2)

where the substitution of DMEM/W allows for the use of bulk water concentrations in the donor and receiver compartments, CD and CR, respectively. (With ionizable molecules, CD and CR refer to the concentrations of the solute summed over all forms of charge state.) These concentrations may be readily measured by standard techniques. Equation (2.2) is still not sufficiently convenient, since one needs to estimate Dm and DMEM/W. It is a common practice to lump these parameters and the thickness of the membrane into a composite parameter, called membrane permeability, Pm,

c02e003  (2.3)

The relevance of Eq. (2.2) (which predicts how quickly molecules pass through simple membranes) to solubility comes in the concentration terms. Consider sink conditions, where CR is essentially zero. Equation (2.2) reduces to the following flux equation:

c02e004  (2.4)

Flux depends on the product of membrane permeability of the solute times the concentration of the solute (summed over all charge state forms) at the water side of the donor surface of the membrane. This concentration ideally may be equal to the dose of the drug, unless the dose exceeds the solubility limit at the pH considered, in which case it is equal to the solubility. Since the uncharged molecular species is the permeant, Eq. (2.4) may be restated as

c02e005  (2.5)

where P0 and C0 are the intrinsic permeability and concentration of the uncharged species, respectively. The intrinsic permeability does not depend on pH, but its cofactor in the flux equation, C0, does. The concentration of the uncharged species is always equal to or less than the intrinsic solubility of the species, S0, which also does not depend on pH.

Note that for the uncharged species, Eq. (2.3) takes on the form

c02e006  (2.6)

where c02ue003 ; also, c02ue004 ; c02ue005 and c02ue006 are the aqueous solution concentrations of the uncharged species in the donor and receiver sides, respectively.

In solutions saturated at some pH (i.e., excess solid present), the plot of log C0. versus pH for an ionizable molecule is simple in form: It is a combination of straight segments, joined at points of discontinuity indicating the boundary between the saturated state and the state of complete dissolution. The pH of these junction points is dependent on the dose used in the calculation, and the maximum value of log C0 is always equal to log S0 in a saturated solution [5].

Figure 2.2 illustrates this idea using ketoprofen as an example of an acid, verapamil as a base, and piroxicam as an ampholyte. In the three cases, the assumed concentrations in the calculation were set to the respective doses [5]. For an acid, log C0 (dashed curve in Figure 2.2a) is a horizontal line (log C0 = log S0) in the saturated solution (at low pH), and it decreases with a slope of −1 in the pH domain where the solute is dissolved completely. For a base (Figure 2.2b) the plot of log C0 versus pH is also a horizontal line at high pH in a saturated solution and is a line with a slope of +1 for pH values less than the pH of the onset of precipitation.

Figure 2.2 Log flux–pH profiles at dosing concentrations: (a) Ketoprofen (acid, pKa 3.99), (b) verapamil (base, pKa 9.06), (c) piroxicam (ampholyte, pKa 5.17, 2.21). The permeability and the concentration of the uncharged species are denoted P0 and C0.

[Reprinted from Avdeef, A. Curr. Topics Med. Chem. 1, 277–351 (2001). Reproduced with permission from Bentham Science Publishers, Ltd.]

The plot of log C0 versus pH has been called the flux factor profile [5], with the idea that such a plot, when combined with intrinsic permeability, can be the basis of an in vitro classification scheme to predict passive oral absorption as a function of pH. This will be discussed in later sections.

Figures 2.1 and 2.2 represent the basic model that will be used to frame the discussion of the measurement of the physicochemical parameters and the interpretation of their role in the oral absorption process [6–17].

2.2 PROPERTIES OF THE GASTROINTESTINAL TRACT (GIT)

The properties of the human GIT that are relevant to drug absorption have been collected from several sources [17–22]. Figure 2.3 shows a cartoon of the GIT, indicating surface area and pH (fasted and fed state) in various segments. The surface area available for absorption is highest in the jejunum and the ileum, accounting for more than 99% of the total. In the fasted state, the pH in the stomach is about 1.7. The acidified contents of the stomach are neutralized in the duodenum by the infusion of bicarbonate ions from the pancreatic duct. Past the pyloric sphincter separating the stomach and the duodenum, the pH steeply rises to about 4.6. Between the proximal jejunum and the distal ileum, the pH gradually rises from about 6 to 8. The pH can drop to values as low as 5 in the colon, due to the microbial digestion of certain carbohydrates, producing short-chain fatty acids (SCFA) in concentration as high as 60–120 mM [23]. The GIT exhibits a considerable pH gradient, and the pH-Partition Hypothesis predicts that the absorption of ionizable drugs may be location-specific.

Figure 2.3 Physical properties of the GIT, with approximate values compiled from several sources [18–22]. Mostly, the pH values refer to median quantities and the range in parentheses refers to interquartile values [21, 22]. The quoted surface areas are taken from reference 20.

[Reprinted from Avdeef, A. Curr. Topics Med. Chem. 1, 277–351 (2001). Reproduced with permission from Bentham Science Publishers, Ltd.]

When food is ingested, the pH in the stomach can rise briefly to 7, but after 0.1 h it drops to pH 5, after 1 h to pH