UC San Francisco Electronic Theses and Dissertations

UCSF UC San Francisco Electronic Theses and Dissertations

Title Resolving the Conflict between BCS and BDDCS for the Advancement of the Drug Discovery, Development, and Regulatory Processes

Permalink https://escholarship.org/uc/item/9wc7754r

Author Larregieu, Caroline A.

Publication Date 2014

Peer reviewed|Thesis/dissertation

eScholarship.org Powered by the California Digital Library University of California

I dedicate this to my parents, Joyce and Habeeb Zayne, who pushed me to strive to

be the most independent, self-sufficient, and intellectual version of myself,

and my husband, Vibs, without whom I simply could not do this.

iii ACKNOWLEDGMENTS

I feel extremely honored and privileged to be the 52nd doctoral student to have graduated under Dr. Leslie Z. Benet. For someone as accomplished as Les, it never ceases to amaze me how humble, approachable, and open-minded he is. I admire his ability to recognize raw talents and polish them to their full potential. Even though he can be extremely busy with his sea of commitments, he has always happily made himself available whenever I needed his guidance. I cannot express enough my deepest gratitude to Les for his nurturing mentorship, his unflailing support, and his unshakable confidence in me.

I am grateful to have worked with numerous current and past students and fellows of the Benet lab, including Dr. Sarah B. Shugarts and Dr. Maribel Reyes for easing my transition into the lab; Dr. Hideaki Okochi for his tremendous breadth of knowledge, guidance, and help; Dr. Fabio Broccatelli for his brilliant collaborative work and stimulating discussions; Dr. Jason Baik for his insightful, intellectual, and philosophical balance; Alan R. Wolfe for voluntarily helping to keep lab operations smooth and in compliance; and Frances M. Peterson for being a fantastic support system and confidante.

I would like to express my sincere appreciation to the entire Pharmaceutical

Sciences and Pharmacogenomics (PSPG) family, including Dr. Kathleen M. Giacomini and Dr. Xin Chen for serving on my Thesis and Oral Qualifying Examination

Committees; Dr. Deanna L. Kroetz and Dr. Laura Bull for serving on my Oral Qualifying

Examination Committee; Dr. Kareen Riviere for being my senior mentor when I first joined the program; Dr. Lindsay M. Reynolds for being my favorite classmate and

iv kindred spirit; and Debbie Acoba-Idlebi for always being available to counsel on graduate school and personal issues.

A special mention must be made of my “sisters” of over 20 years – Sanola A.

Daley, Rebekah L. Francis, Conniel A. Malek, Shushanna C. Mignott, and Dr. Cheryl S.

Stewart – who never ceased to inspire and motivate me during this endeavor.

It has meant the world to me to have the unconditional love and support of my family throughout this journey. I am indebted to my parents, Joyce and Habeeb Zayne, who have sacrificed beyond words can describe to provide me the best opportunities in life. I must also express my heartfelt love and appreciation for the rest of my family – notably my aunt, Jasmine Burke; my mother-in-law, Dr. Ranjna Jindal; my father-in-law,

Dr. Vinod K. Jindal; my aunt-in-law, Asha Mansinghka; and my uncle-in-law, Dr.

Surendra K. Mansinghka.

Finally, I simply could not have arrived at this milestone without the unwavering support and relentless cheerleading of my anchor, my rock, my sanity, my best friend, my soul mate, and my husband – Vibhav “Vibs” Jindal – who I proudly share this achievement with.

v ABSTRACT

Drug permeability is accepted as a screening tool for determining intestinal absorption via the Biopharmaceutics Classification System (BCS) during the drug development and regulatory approval processes. Currently, predicting clinically significant drug interactions during drug development is a known challenge for the pharmaceutical industry and regulatory agencies. The Biopharmaceutics Drug Disposition Classification

System (BDDCS), a modification of BCS utilizing drug metabolism instead of intestinal permeability, predicts drug disposition and potential drug-drug interactions in the intestine, the liver, and most recently the brain. While correlations between BCS and

BDDCS have been observed with drug permeability, discrepancies in drug classification between the two systems using different permeability models, accepted as surrogate models for demonstrating human intestinal permeability by the FDA, have been noted.

This project examines the role of drug permeability in drug absorption and drug metabolism and recommends the suitability of these models for predicting BDDCS and

BCS classifications. This project evaluates methodologies that can lead to recommendations for facilitating the drug discovery, development, and regulatory approval processes.

vi TABLE OF CONTENTS

ACKNOWLEDGMENTS...... iv

ABSTRACT...... vi

LIST OF TABLES...... xi

LIST OF FIGURES...... xii

CHAPTER 1: UNDERSTANDING THE DIFFERENCES BETWEEN BCS AND BDDCS...... 1

1.1 THE BIOPHARMACEUTICS CLASSIFICATION SYSTEM (BCS)...... 1

1.2 METHODS FOR ASSESSING BCS CLASSIFICATION...... 3

1.3 THE BIOPHARMACEUTICS DRUG DISPOSITION CLASSIFICATION SYSTEM (BDDCS)...... 4

1.4 DISCREPANCIES BETWEEN CLASSIFYING DRUGS UNDER BCS AND BDDCS...8

1.5 GOALS AND HYPOTHESES...... 9

1.6 REFERENCES...... 11

CHAPTER 2: INVESTIGATING AND IMPROVING IN SILICO AND IN VITRO PREDICTION MODELS THAT USE CACO-2 AS A SURROGATE FOR HUMAN INTESTINAL PERMEABILITY...... 14

2.1 INTRODUCTION...... 14

2.2 IDENTIFYING COMPOUNDS FOR WHICH CACO-2 MAY POORLY PREDICT HUMAN INTESTINAL PERMEABILITY RATE MEASUREMENTS...... 17

2.2.1 Substrates of Highly Expressed Human Small Intestinal Transporters: Peptide, Amino Acid, and Nucleoside Transporters...... 17

vii TABLE OF CONTENTS (continued)

2.2.2 Differences in the Paracellular Junctions between Caco-2 and the Human Intestine do not Explain the Inaccurate In Vitro–In Vivo Permeability Predictions by Caco-2 for Hydrophilic Compounds...... 25

2.3 THE NEED FOR LARGER, SINGLE-SOURCE CACO-2 PERMEABILITY RATE DATASETS...... 28

2.4 IN SILICO WORK CONSIDERATIONS...... 32

2.5 IN VITRO SCREENING CONSIDERATIONS...... 41

2.6 REGULATORY CONSIDERATIONS...... 43

2.7 CONCLUSIONS...... 45

2.8 ACKNOWLEDGMENT...... 47

2.9 REFERENCES...... 48

CHAPTER 3: DISTINGUISHING BETWEEN THE PERMEABILITY RELATIONSHIPS WITH ABSORPTION AND METABOLISM TO IMPROVE BCS AND BDDCS PREDICTIONS...... 67

3.1 INTRODUCTION...... 67

3.2 METHODS...... 70

3.2.1 Compilation of Permeability Rate Datasets...... 70

3.2.2 Correlations of BCS and BDDCS with Drug Permeability Rate Measures...... 74

3.3 RESULTS...... 81

viii TABLE OF CONTENTS (continued)

3.3.1 Comparison of BCS and BDDCS Classifications Using Human Intestinal Permeability Rate Measures...... 81

3.3.2 Comparison of BCS and BDDCS Classifications Using Caco-2 Permeability Rate Measures...... 85

3.3.3 Comparison of BCS and BDDCS Classifications Using PAMPA Permeability Rate Measures...... 91

3.4 DISCUSSION...... 98

3.4.1 Use of BDDCS in the BCS FDA Guidance for Industry...... 98

3.4.2 Drug Discovery Considerations when Making BCS and BDDCS Predictions...... 100

3.5 REFERENCES...... 103

CHAPTER 4: EXTENDING THE APPLICATION OF BDDCS TO THE PREDICTION OF BRAIN DISPOSITION OF ORALLY ADMINISTERED DRUGS...... 110

4.1 INTRODUCTION...... 110

4.2 CRITERIA AND RATIONALE FOR DATA COLLECTION...... 112

4.2.1 Efflux Ratio...... 112

4.2.2 BDDCS Classification...... 114

4.2.3 BBB Penetration Data...... 114

4.2.4 In Silico Calculations...... 116

ix TABLE OF CONTENTS (continued)

4.3 ELABORATION, ANALYSIS, AND APPLICATION OF RULES TO ESTIMATE BRAIN DISPOSITION...... 118

4.3.1 P-gp and Brain Disposition...... 118

4.3.2 In Silico Permeability and Brain Disposition...... 119

4.3.3 Generation of BDDCS-Based BBB Rules...... 121

4.3.4 Outliers Analysis...... 126

4.3.5 Molecular Properties and Brain Disposition...... 135

4.3.6 Case Study: Predicting the Brain Disposition of Antihistamines with BDDCS...... 137

4.4 INTEGRATION OF FINDINGS AND RECOMMENDATIONS...... 138

4.5 FUTURE PROSPECTS OF BDDCS IN DRUG DISCOVERY...... 145

4.6 ACKNOWLEDGMENT...... 147

4.7 REFERENCES...... 147

CHAPTER 5: CONCLUSIONS AND PERSPECTIVES...... 167

5.1 THE BCS AND BDDCS IN DRUG DISCOVERY, DEVELOPMENT, AND REGULATORY PROGRAMS...... 167

5.2 RECOMMENDATIONS FOR FUTURE BCS AND BDDCS SCREENING...... 168

5.3 REFERENCES...... 171

APPENDIX...... 173

x LIST OF TABLES

TABLE 2.1: Highly expressed small intestinal transporters with expression differences of ≥10-fold between the human intestine and the Caco-2 cell line...... 20

TABLE 2.2: Inaccurate predictions of human intestinal absorption by Caco-2 for substrates of highly expressed transporters with expression differences of ≥10-fold between the human intestine and Caco-2 cell line...... 24

TABLE 2.3: Cases of intestinal transport by paracellular diffusion and carrier-mediated processes for hydrophilic compounds with their Caco-2 permeabilities and human extents of absorption...... 27

TABLE 2.4: Variability in Caco-2 permeability measurements from different investigators...... 31

TABLE 2.5: Published in silico predictive models using Caco-2 permeability measurements...... 34

TABLE 3.1 Comparison of the human in vivo effective permeability rate coefficients

(Peff) of 30 drugs with their extents of absorption and metabolism...... 71

-6 TABLE 3.2: Comparison of the Caco-2 permeability rate coefficients, Papp (x 10 cm/s), from 4 different laboratories with their extents of absorption and metabolism...... 72

-6 TABLE 3.3: Comparison of the PAMPA permeability rate coefficients, Papp (x 10 cm/s), from 4 different PAMPA models with their extents of absorption and metabolism...... 75

TABLE 3.4: Comparison of all analyzed drugs with their measures of lipophilicity....78

TABLE 4.1: List of 153 oral drugs and data used for brain penetration prediction....127

xi LIST OF FIGURES

FIGURE 1.1: Overview of the Biopharmaceutics Classification System (BCS)...... 2

FIGURE 1.2: Predominant routes of elimination for each BCS class...... 6

FIGURE 1.3: Overview of the Biopharmaceutics Drug Disposition Classification System (BDDCS)...... 7

FIGURE 2.1: Effect of including amino acid and peptide transporter substrates in the in vitro–in vivo correlations of Caco-2 and human intestinal permeability measurements where Caco-2 permeability measurements were available for at least 6 drugs, including metoprolol, of the 30 drugs with available human intestinal permeability measurements...... 22

FIGURE 2.2: The apparent Caco-2 permeability coefficients selected from 11 studies that included metoprolol, for 20 of the 30 drugs where human intestinal permeability data are available...... 29

FIGURE 3.1: Relationships between the extent of absorption, extent of metabolism, and human intestinal permeability rates for 28 drugs...... 82

FIGURE 3.2: Comparison of BCS and BDDCS classifications with Caco-2 permeability rate measures...... 86

FIGURE 3.3: Comparison of BCS and BDDCS classifications with PAMPA permeability rate measures...... 93

FIGURE 4.1: Pie and bar charts showing the relative percentage of the drugs in the data set classified based on their ability to cross the blood brain barrier, P-gp profile, and BDDCS class...... 120

xii LIST OF FIGURES (continued)

FIGURE 4.2: Receiver-operating characteristic (ROC) curves for CACO2 permeability and cLogP used as classifiers to discriminate BBB+ and BBB- classes, as well as VolSurf+ CACO2 descriptor and cLogP distributions of BBB+ and BBB- drugs...... 122

FIGURE 4.3: Pipeline of rules for predicting brain disposition...... 124

FIGURE 4.4: Accuracy of blood-brain barrier crossing prediction for the different methods...... 125

FIGURE 4.5: Average and standard deviation for the following drug properties: nitrogen, hydrogen bond donor, oxygen and hydrogen bond acceptor counts, molecular weight, and lipophilicity (cLogP)...... 136

FIGURE 4.6: Predicted blood-brain barrier penetration for 28 drugs with affinity for H1 receptor and associated severity of sedative effect. BBB- drugs do not exert CNS side-effects...... 139

FIGURE 4.7: BDDCS in drug discovery...... 146

xiii Chapter 1: Understanding the Differences between BCS and BDDCS

1.1 The Biopharmaceutics Classification System (BCS)

Many potential drug candidates fail due to undesirable efficacy and safety issues.

Pharmacokinetic optimization with respect to ADME (absorption, distribution, metabolism and elimination) at the drug discovery phase is extremely valuable in reducing compounds failing in late preclinical and clinical development, since efficacy and safety issues are related in part to pharmacokinetic profiles. Predicting intestinal absorption is also particularly important since it is the preferred route of drug administration with increased patient convenience, lower cost, and reduced invasiveness.

The Biopharmaceutics Classification System (BCS) is one of the most signiﬁcant prognostic tools created to facilitate oral drug product development. The human permeability studies of 30 drugs by Amidon and co-workers1 showed that an excellent correlation existed between the human jejunal permeability rate (Peff) measured using intestinal perfusion and the fraction of dose absorbed obtained from pharmacokinetic or mass balance studies in humans. The predictability of highly absorbed, highly soluble

(dosed in solution or in dosage forms that dissolve very rapidly) drugs is consistently correlated with a human in vivo intestinal permeability greater than or equal to that of metoprolol.

Based on their solubility and intestinal membrane permeability characteristics, drug substances have been classiﬁed into one of four categories according to the BCS

(Figure 1.1). Its principles are extensively used by the pharmaceutical industry throughout drug discovery and development.2, 3 Over the years, the use of BCS in drug

Figure 1.1 Overview of the Biopharmaceutics Classification System (BCS).1

2 development has been expanded to regulatory practices in determining whether an in vivo bioequivalence study may be waived for IR solid oral dosage forms. It has been adopted by the US Food and Drug Administration (FDA),4 the European Medicines Agency

(EMA),5 and the World Health Organization (WHO)6 for setting bioavailability/bioequivalence (BA/BE) standards for immediate-release (IR) oral drug product approval.

1.2 Methods for Assessing BCS Classification

Specifically, the FDA has implemented a BCS guidance4 to allow a waiver of in vitro BE testing for BCS class 1, high-solubility, high-permeability drugs in IR fast- dissolving solid dosage forms. A drug substance is deemed highly soluble if the highest dose strength is soluble in 250 mL or less aqueous media over the pH range of 1–7.5. In the absence of evidence suggesting instability in the gastrointestinal tract, a drug substance is considered to be highly permeable when the extent of absorption in humans is 90% or more of an administered dose. Accordingly, the FDA guidance indicates that permeability classiﬁcation can be determined directly by measuring the rate of mass transfer across human intestinal membrane, or indirectly by estimating the extent of drug absorption in human pharmacokinetic studies in comparison to an IV reference dose.

The BCS guidance does allow the possibility that biowaivers may be based on in vivo or in vitro permeability measures. In its guidance for industry,4 the FDA recommends that if an alternative method to human extent of absorption is utilized, a list of 20 model drugs are suggested to be used in establishing the suitability of an alternative

3 method using metoprolol as a high-permeability reference cut-off. The following methods can be used to determine the permeability of a drug substance from the gastrointestinal tract: (1) in vivo intestinal perfusion studies in humans; (2) in vivo or in situ intestinal perfusion studies using suitable animal models; (3) in vitro permeation studies using excised human or animal intestinal tissues; or (4) in vitro permeation studies across a monolayer of cultured epithelial cells. Even though permeability measures across the human and rat intestine have been reported to most accurately predict human intestinal absorption, these methods are not optimal for high-volume screening during early drug discovery since they are costly, difficult, and time- consuming.

Hence, the most commonly employed alternate approach is the use of permeability measurements across the human colon carcinoma cell line, Caco-2, where high-throughput permeability methods have been developed. Countless studies have been carried out and published in the literature reporting permeability measurements for this cell line. Compared to the intact human or rat intestine, this cultured cell line has a lower permeation surface area,7 differential expression of intestinal transporters,8 and fewer paracellular pores,9 which can impact its predictability for intestinal absorption and accurate BCS classification. Many programs throughout industry utilize other in vitro and in silico methods, which share similar limitations with this cell line.

1.3 The Biopharmaceutics Drug Disposition Classification Systems (BDDCS)

While the BCS is using for predicting oral absorption, Benet and co-workers were interested in creating a classification system that could better predict drug disposition.

4 Examining more than 130 compounds classified according to BCS, Wu and Benet10 observed an apparent relationship between a drug’s intestinal permeability rate and its extent of metabolism. They observed that high-permeability drugs (classes 1 and 2) were primarily eliminated via metabolism in humans, while low-permeability drugs (classes 3 and 4) were primarily eliminated via urinary and/or biliary excretion (Figure 1.2). They subsequently developed the Biopharmaceutics Drug Disposition Classification System

(BDDCS), a modification of BCS utilizing drug metabolism by Phase 1 and Phase 2 enzymes instead of intestinal permeability. Based on their solubility and extent of drug metabolism characteristics, drug substances have been classified into one of four categories according to the BDDCS (Figure 1.3). The classification system is useful in predicting routes of elimination, effects of efflux and absorptive transporters on oral absorption, when transporter-enzyme interplay will yield clinically significant effects such as low bioavailability and drug-drug interactions, the direction and importance of food effects, and transporter effects on post-absorption systemic levels following oral and intravenous dosing.10-12

Because highly metabolized drugs are also highly absorbed and highly permeable,

Benet and colleagues proposed that extensive metabolism can be used as a surrogate to demonstrate high permeability under BCS.13 They recommended that regulatory agencies add the extent of drug metabolism (i.e., ≥90% metabolized) as an alternative method for the extent of drug absorption (i.e., ≥90% absorbed) in defining Class 1 drugs suitable for a waiver of in vivo studies of bioequivalence. They proposed the following criteria be used to define ≥90% metabolized for marketed drugs: “Following a single oral dose to humans, administered at the highest dose strength, mass balance of the Phase 1 oxidative

Figure 1.2 Predominant routes of elimination for each BCS class observed by Wu and Benet.10

Figure 1.3 Overview of the Biopharmaceutics Drug Disposition Classification System (BDDCS).10

7 and Phase 2 conjugative drug metabolites in the urine and feces, measured either as unlabeled, radioactive labeled or non-radioactive labeled substances, account for ≥90% of the drug dosed. This is the strictest definition for a waiver based on metabolism. For an orally administered drug to be ≥90% metabolized by Phase 1 oxidative and Phase 2 conjugative processes, it is obvious that the drug must be absorbed.”13 These metabolism measures have been adopted by the EMA5 and most recently FDA scientists have concurred with them14 as a basis for an in vivo bioequivalence study biowaiver.

1.4 Discrepancies between Classifying Drug under BCS and BDDCS

Even though both BCS and BDDCS relate to drug permeability, discrepancies in classifying drugs between the two systems using different permeability models have been noted. When regulatory scientists, Chen and Yu,15 examined the role of drug metabolism in predicting human intestinal permeability, they found 14 high-permeability drugs based on mass balance measurements (absolute bioavailability or amount of drug excreted unchanged in the urine) that exhibit poor metabolism. Upon close examination, 11 of these 14 drugs have published in vitro permeation studies using Caco-2 cells available for which 8 (73%) drugs, at their highest dose strength concentration in 250 ml of water, have low permeability in vitro, which contradicts their high permeability rates in vivo and results in incorrect BCS classifications. In fact, one of the drugs, sotalol, has a Caco-2 permeability determined to be 4% of the high-permeability reference standard, metoprolol, as determined by Yu and FDA coworkers.16 Interestingly, these drugs with

8 low permeability rates in vitro accurately predict their poor metabolism in vivo and lead to accurate BDDCS classifications.

The current practice of the FDA is that their policy during regulatory reviews is to allow in vivo human data to triumph over any contradictory animal and/or in vitro data.14

And while this may suffice for regulatory programs that have access to thorough submissions of in vivo, in vitro, and in silico data, this is not beneficial for drug discovery and development programs that utilize the guidance for industry and potentially could lead to discarding many new molecular entities (NMEs) that could be viable drug candidates based on implemented in vitro screening processes during the early drug discovery phase. Hence, better guidelines are necessary for efficient and accurate BCS and BDDCS predictions for the drug discovery, development, and regulatory programs.

1.5 Goals and Hypotheses

All of the currently used in vitro and in vivo permeability models, accepted by the

FDA for demonstrating human intestinal permeability, share overlapping, but also different characteristics, which can lead to discrepant permeability results among the models. My thesis aims to explain why these discrepancies occur and recommend the appropriate permeability models for improving the prediction of BCS and BDDCS classifications. Cephalexin is an example that has been shown to be highly permeable in humans. This finding correctly predicts its high absorption yet incorrectly predicts its poor metabolism. However, cephalexin exhibits a poor permeability rate in vitro, which accurately predicts its poor metabolism yet incorrectly predicts its high extent of

9 absorption. Intestinal permeability relating to absorption may involve passive and transporter-mediated mechanisms through paracellular and transcellular pathways.

However, extensive drug metabolism may not occur if the drug is predominantly absorbed via the paracellular pathway or by an active process, suggesting that drug metabolism may be better correlated with passive transcellular permeability. Hence, I hypothesize that passive transcellular permeability correlates with drug metabolism, where high versus low passive transcellular permeability can predict extensively versus poorly metabolized (BDDCS) drugs. Moreover, permeability models that are deficient in transporter expression and paracellular junctions will most accurately predict BDDCS classifications, while these latter systems will inaccurately predict BCS classifications that are primarily absorbed by the paracellular pathway or a highly expressed intestinal transporter.

I examine the role of drug permeability in drug absorption and drug metabolism using different permeability models and recommend the suitability of these models for predicting BDDCS and BCS drug classifications. Based on the limitations of the each system (in terms of their paracellular pathway and transporter expression), I demonstrate through my analyses which permeability systems are most suitable for predicting BCS and BDDCS classifications. Until recently, BDDCS has been particularly useful in predicting drug disposition and potential clinically significant drug–drug interactions in the human body, particularly in the liver and intestine. I present an extension of applying

BDDCS for predicting drug disposition and the potential for drug–drug interactions for the brain. This work provides results that can lead to recommendations for facilitating the drug development and drug regulatory processes.

10 1.6 References

1. Amidon, G. L.; Lennernas, H.; Shah, V. P.; Crison, J. R. A theoretical basis for a biopharmaceutic drug classification: the correlation of in vitro drug product dissolution and in vivo bioavailability. Pharm. Res. 1995, 12, 413-420.

2. Cook, J.; Addicks, W.; Wu, Y. H. Application of the biopharmaceutical classification system in clinical drug development--an industrial view. AAPS J. 2008, 10,

306-310.

3. Ku, M. S. Use of the Biopharmaceutical Classification System in early drug development. AAPS J. 2008, 10, 208-212.

4. Waiver of In Vivo Bioavailability and Bioequivalence Studies for Immediate-

Release Solid Oral Dosage Forms Based on a Biopharmaceutics Classification System.

FDA Guidance for Industry; Food and Drug Administration: Rockville, MD, 2000.

5. Guideline on the Investigation of Bioequivalence. European Medicines Agency:

London, 2010.

6. Takagi, T.; Ramachandran, C.; Bermejo, M.; Yamashita, S.; Yu, L. X.; Amidon,

G. L. A provisional biopharmaceutical classification of the top 200 oral drug products in the United States, Great Britain, Spain, and Japan. Mol. Pharmaceutics 2006, 3, 631-643.

7. Artursson, P. Epithelial transport of drugs in cell culture. I: A model for studying the passive diffusion of drugs over intestinal absorptive (Caco-2) cells. J. Pharm. Sci.

1990, 79, 476-482.

8. Sun, D.; Lennernas, H.; Welage, L. S.; Barnett, J. L.; Landowski, C. P.; Foster,

D.; Fleisher, D.; Lee, K. D.; Amidon, G. L. Comparison of human duodenum and Caco-

11 2 gene expression profiles for 12,000 gene sequences tags and correlation with permeability of 26 drugs. Pharm. Res. 2002, 19, 1400-1416.

9. Linnankoski, J.; Makela, J.; Palmgren, J.; Mauriala, T.; Vedin, C.; Ungell, A. L.;

Lazorova, L.; Artursson, P.; Urtti, A.; Yliperttula, M. Paracellular porosity and pore size of the human intestinal epithelium in tissue and cell culture models. J. Pharm. Sci. 2010,

99, 2166-2175.

10. Wu, C. Y.; Benet, L. Z. Predicting drug disposition via application of BCS: transport/absorption/elimination interplay and development of a biopharmaceutics drug disposition classification system. Pharm. Res. 2005, 22, 11-23.

11. Custodio, J. M.; Wu, C. Y.; Benet, L. Z. Predicting drug disposition, absorption/elimination/transporter interplay and the role of food on drug absorption. Adv.

Drug Deliv. Rev. 2008, 60, 717-733.

12. Shugarts, S.; Benet, L. Z. The role of transporters in the pharmacokinetics of orally administered drugs. Pharm. Res. 2009, 26, 2039-2054.

13. Benet, L. Z.; Amidon, G. L.; Barends, D. M.; Lennernas, H.; Polli, J. E.; Shah, V.

P.; Stavchansky, S. A.; Yu, L. X. The use of BDDCS in classifying the permeability of marketed drugs. Pharm. Res. 2008, 25, 483-488.

14. Chen, M. L.; Amidon, G. L.; Benet, L. Z.; Lennernas, H.; Yu, L. X. The BCS,

BDDCS, and regulatory guidances. Pharm. Res. 2012, 28, 1774-1778.

15. Chen, M. L.; Yu, L. The use of drug metabolism for prediction of intestinal permeability. Mol. Pharmaceutics 2009, 6, 74-81.

12 16. Yang, Y.; Faustino, P. J.; Volpe, D. A.; Ellison, C. D.; Lyon, R. C.; Yu, L. X.

Biopharmaceutics classification of selected beta-blockers: solubility and permeability class membership. Mol. Pharmaceutics 2007, 4, 608-614.

13 Chapter 2: Investigating and Improving In Silico and In Vitro Prediction Models that Use Caco-2 as a Surrogate for Human Intestinal Permeability

2.1 Introduction

In the drug discovery process, the ability of a new molecular entity (NME) to cross biological membranes, particularly the human intestinal mucosa, is crucial in establishing it as a viable drug candidate for clinical development and eventual regulatory approval. The prevalence of oral formulations on the market indicates a strong industry- wide preference for drugs that can permeate the human intestine and be absorbed, further highlighting that poor intestinal permeability is viewed as an unfavorable biopharmaceutical property.1 Since the inception of the Biopharmaceutics Classification

System (BCS), which originally demonstrated a relationship between the human jejunal permeability rate of a drug and its extent of absorption,2 the U.S. FDA has supported waivers of in vivo bioavailability and bioequivalence studies to facilitate the drug approval process of BCS Class 1 drugs that are highly permeable, highly soluble, and rapidly dissolvable.3 Because human intestinal permeability studies are both costly and difficult,4 studies in the human colon carcinoma cell line, Caco-2, are the most commonly used biological tool for screening the intestinal permeability of NMEs during drug discovery and development.5, 6 As such, measurements in the Caco-2 cell line are accepted by the FDA as a surrogate for human intestinal permeability measurements as part of their BCS guidance.3

Caco-2 was originally proposed as an in vitro model system for studying human intestinal permeability based on its shared characteristics with the small intestinal

14 epithelium of cell morphology, polarity, enterocytic differentiation, and high TEER values.7 Its suitability for intestinal permeability screening was initially confirmed when strong correlations between in vitro Caco-2 and in vivo human intestinal permeability measurements were observed for some passively absorbed drugs.5, 8, 9 Since then, Caco-2 has been used to screen the intestinal permeabilities of an increasingly diverse population of compounds. Yee10 had reported that Caco-2 can accurately predict the human intestinal permeability or absorption of compounds regardless of transcellular, paracellular, and carrier-mediated transport mechanisms. However, there are Caco-2 permeability results that disagree with measured human intestinal permeability rates or measured extents of human intestinal absorption.11, 12

Most expert reviews address drug transport processes across Caco-2 monolayers, factors influencing their permeability measurements, and some of the limitations associated with using Caco-2 for human intestinal or absorption predictions.5, 6, 13-15 Since

Caco-2 has been widely used for screening the intestinal permeabilities of passively and actively absorbed compounds, these expert reviews recommend limiting the application of Caco-2 permeability screening to passive compounds. However, merely using this approach may be unsuccessful based on a recent report showing that permeability measures for the majority of drugs on the market include both passive diffusion and active transport.16

Because discrepancies have only been observed for cases where some highly absorbed drugs have been predicted to be poorly permeable by Caco-2, a consensus of

FDA and academic scientists prefer that permeability determinations be based on a pharmacokinetic approach in humans for justifying biowaivers in drug applications rather

15 than the in vitro approach, when discordant results between the in vitro and in vivo methods arise.17 Although the FDA states that human in vivo information will “trump” or hold precedence over discordant Caco-2 permeability results in regulatory decisions,17 the lack of available human in vivo information during drug discovery to identify when an inaccurate in vitro–in vivo permeability prediction occurs with Caco-2 can hinder the drug discovery process in one of two ways.

First, the existence of these variants questions the reliability of using the Caco-2 cell line as a primary source of screening the intestinal permeabilities of NMEs for lead selection, optimization, and candidate prioritization. Second, Caco-2 permeability data generated during drug discovery are used to develop in silico predictive models based on correlations with calculated molecular descriptors that describe permeability requirements. The inclusion of discrepant Caco-2 permeability measurements during in silico model development can produce misleading information when using the models for drug design. For these reasons, it is important to identify the types of compounds for which Caco-2 may be deficient in accurately predicting their human intestinal permeabilities, thereby ensuring that the selection of potential drug candidates is not tarnished by inaccurate information during the drug discovery process.

Hence, the aim of this work was to identify the cases where using Caco-2 permeability measurements for in vitro screening and in silico model development purposes can lead to discordant human intestinal permeability predictions. Based on these findings, recommendations on how to use Caco-2 as a primary in vitro permeability screening tool and for developing in silico predictive models, as well as for improving the current regulatory guidance on how Caco-2 can be used for human intestinal permeability

16 screening during drug discovery and development, are proposed.

2.2 Identifying Compounds for which Caco-2 may Poorly Predict Human Intestinal Permeability Rate Measurements

2.2.1 Substrates of Highly Expressed Human Small Intestinal Transporters: Peptide, Amino Acid, and Nucleoside Transporters

In vitro Caco-2 permeability measurements have been found to correlate well with in vivo human intestinal permeability measurements for passively absorbed drugs, but much less so for compounds that are in part transported by carrier-mediated mechanisms.9, 18 In those cases, permeability measurements were found to be much higher in humans than Caco-2 for carrier-mediated compounds, compared to passively absorbed compounds. These workers9, 18 also observed that discrepancies generally correlated with the expression differences of transporters between the two membranes for which the carrier-mediated compounds are substrates. Hence, differences in the expression of a transporter between Caco-2 and the human intestine may explain differences in the permeability of a substrate between the two membranes, resulting in an incorrect in vitro–in vivo permeability prediction.

To determine which transporter substrates’ permeabilities may be impacted by transporter expression differences between Caco-2 and the human intestine, the relative expression of transporters in the human small intestine versus transporters in the Caco-2 cell line and the ability of Caco-2 to make an accurate in vitro–in vivo permeability

17 prediction for known transporter substrates were examined. In an analysis of 36 transporters, Hilgendorf and colleagues19 found PEPT1 to be a highly expressed transporter in the human jejunum showing significantly larger (≥10-fold) difference in transporter expression than any other transporters that they analyzed. The permeabilities of several PEPT1 transporter substrates – cephalexin,20 cephradine,21 and Gly-Sar22 – have been shown to correlate with the expression levels of human PEPT1 in Caco-2 cells.

Additionally, cephalexin’s permeability has been shown to correlate with the expression of rat Pept1 in the rat jejunum. Chong et al.23 were the first to illustrate that Caco-2 is deficient in predicting the high human intestinal absorption of PEPT1 substrates, amoxicillin and cephalexin, where intestinal absorption is directly related to intestinal permeability as defined by the BCS2 and FDA.3 Hilgendorf et al.19 also found HPT1 to be highly expressed in the small intestine, however, its expression level was comparable between Caco-2 and the human jejunum. Other SLC transporters such as OATPs, OATs, and OCTs were found to be moderately to poorly expressed in the small intestine and their expression differences from Caco-2 were <10-fold. Of the efflux transporters they analyzed, BCRP and MRP2 were found to be highly expressed in the small intestine, yet only BCRP had a ≥10-fold difference in expression between Caco-2 and the human jejunum.

Englund and colleagues24 also found PEPT1 and BCRP to be highly expressed in the small intestine, with variability being ≥10-fold between their human intestinal segments and Caco-2. Seithel and coworkers25 only found PEPT1 to be highly expressed in the human jejunum with a ≥10-fold difference between the human jejunum and Caco-

2. While they measured high expression levels of MRP2 and lower levels of BCRP in the

18 small intestine, the variability in their MRP2 and BCRP expression levels between the small intestine and Caco-2 was <10-fold. In a separate analysis of 183 transporters, Sun and colleagues18 found 87 transporters expressed in the human duodenum, of which at least five transporters that had a high expression in the small intestine and a relative expression of ≥10-fold difference between the human duodenum and Caco-2 cell line.

These not only included the peptide transporter, PEPT1, but also members of the family of amino acid, concentrative nucleoside, and equilibrative nucleoside transporters (Table

2.1).

I investigated how transporter expression differences between the Caco-2 and the human intestine affected the in vitro–in vivo permeability correlation of Caco-2 and human intestinal permeability data for 30 drugs reported in the literature.4 Unlike Sun et al.,18 correlations with Caco-2 permeability measurements from datasets taken from the same laboratory were conducted so as to avoid interlaboratory variation and selection bias. In Figure 2.1, the correlation coefficients of human intestinal permeability measurements versus those of Caco-2 from eight independent studies where metoprolol was included in the analysis are presented. Of the 30 drugs with available human intestinal permeability data, four (amoxicillin, cephalexin, levodopa, and lisinopril) are known substrates of one or more of the transporters with a ≥10-fold difference in expression, particularly amino acid or peptide transporters. Including all of the drugs in each dataset, the correlation coefficients (r2) ranged from 0.409 to 0.917. When the original datasets included any of the amino acid or peptide substrates, the r2 values ranged from 0.409 to 0.651; when they did not, the r2 values ranged from 0.835 to 0.917. For the

5 datasets that included the amino acid or peptide substrates, the correlations improved

Table 2.1 Highly expressed small intestinal transporters18, 19, 26, 27 with expression differences of ≥10-fold between the human small intestine and the Caco-2 cell line.

Relative Expression

Transporter Differences between Reference Transporter Family/System the Human Intestine and Name (Gene Name) Type Caco-2

>10-fold higher in the ileum PEPT1 (SLC15A1) Peptide Englund than in Caco-2

et al.24a >10-fold higher in the ileum BCRP (ABCG2)b Efflux than in Caco-2

>10-fold higher in the jejunum PEPT1 (SLC15A1) Peptide Hilgendorf than in Caco-2

et al.19a >10-fold higher in the jejunum BCRP (ABCG2)b Efflux than in Caco-2

Seithel 10-fold higher in the jejunum PEPT1 (SLC15A1) Peptide et al.25a than in Caco-2

>15-fold higher level in the Peptide PEPT1 (SLC15A1) duodenum than in Caco-2

>12-fold higher level in Caco- Amino Acid ATB0 (SLC1A5) 2 than in the duodenum

Sun >20-fold higher level in the Amino Acid et al.18 GABA/noradrenaline duodenum than in Caco-2

(SLC3A2) >110-fold higher level in the Nucleoside CNT2 (SLC28A2) duodenum than in Caco-2

>25-fold higher level in Caco- Nucleoside ENT1 (SLC29A1) 2 than in the duodenum

aThese studies did not analyze Amino Acid and Nucleoside transporters. bOther sources have reported BCRP to be moderately expressed in the small intestine25-27 and have <10-fold variability between the small intestine and Caco-2 cell line.25

when excluding these substrates one by one or completely (Figure 2.1). When all of the amino acid and peptide substrates were absent from all of the datasets, 7 of the 8 datasets of Caco-2 permeability measurements correlated well (r2>0.800) with human intestinal permeability rate measurements. The in vitro–in vivo permeability correlations are strong even when including substrates of other transporters that are <10-fold difference in expression between the two membranes, suggesting that Caco-2 can predict the human intestinal permeability of substrates of transporters that differ no more than

10-fold between Caco-2 and the human intestine.

Due to the limited availability of human intestinal permeability rate data, the ability of Caco-2 measurements to predict the closest surrogate for such data, namely the human extent of absorption, where I could find much more information available in the literature, was analyzed. Of more than 50 compounds for which in vitro–in vivo correlations agreed, I found more than 30 that are known substrates of transporters – such as OAT, OATP, OCT, and P-gp – where expression differences between Caco-2 and the human intestine are <10-fold. This suggests that there may be a likely threshold for transporter expression differences between Caco-2 and the human intestine for Caco-2 measurements to be able to accurately predict the intestinal permeability of transporter substrates. When the lipophilicity of 66 compounds with available Caco-2 permeability and human extent of absorption measurements was examined, these lipophilic compounds (where LogP ≥1) were found to have high Caco-2 permeability measurements that correctly predicted their high extents of absorption. Hence, it is possible that the lipophilic properties of the compounds (and consequently higher passive permeabilities) may explain the better correlations of OAT, OATP, OCT and P-gp

Figure 2.1 Effect of including amino acid and peptide transporter substrates in the in vitro–in vivo correlations of Caco-2 and human intestinal permeability measurements for 8 studies,15, 28-34 where Caco-2 permeability measurements were available for at least 6 drugs, including metoprolol, of the 30 drugs with available human intestinal permeability measurements. The total number of compounds in each correlation range from 6 to 22, with the average being 11.

substrates rather than their <10-fold expression differences between the two membranes.

However, studies with Caco-2 permeability measurements for other lipophilic compounds like BCRP substrates and a high-permeability reference standard could not be found to evaluate if their human extents of intestinal absorption could be correctly predicted by their physicochemical properties rather than reported ≥10-fold differences in transporter expression between Caco-2 and the human small intestine.

In Table 2.2, 11 hydrophilic drugs where low Caco-2 permeability measurements in the literature do not correctly predict the high extents of absorption of these compounds are identified. These drugs were found to be substrates of members of the amino acid, nucleoside, and peptide transporter families, where the expression difference between Caco-2 and the human intestine is ≥10-fold. Three hydrophilic drugs (lisinopril, gabapentin, and lamivudine) are additionally included in Table 2.2 that are also substrates of these transporters but have low Caco-2 permeability measurements and low extents of absorption, to demonstrate that Caco-2 cannot discriminate between high and low extents of absorption for these hydrophilic compounds. The passive permeabilities of these compounds are expected to be low, due to the compounds being poorly lipophilic in nature.35 Hence, their ability to exhibit a high intestinal permeability or a high extent of absorption will likely be dependent on the degree of carrier-mediated transport. Caco-2 permeability measurements between pH 5.0 and 7.5 in the literature were examined to see if pH optimization could possibly improve predictability. Even though pH slightly increased the permeability of some substrates, no published evidence of changing the permeability from low to high relative to a high-permeability reference standard could be found. Hence, it may not be possible to optimize Caco-2 assays to correctly predict the

Table 2.2 Inaccurate predictions of human intestinal absorption by Caco-2 for substrates of highly expressed transporters with expression differences of ≥10-fold between the human intestine and Caco-2 cell line.

cLogP Transporter Caco-2 Extent of Compound a Substrate Permeability Absorption

Lisinopril -1.69 Peptide 36, 37 Low 23, 29, 37 Low 23, 29 Amoxicillin -1.87 Peptide 23, 38 Low 23, 29, 30 High 23, 29, 30 Cefadroxil -2.51 Peptide 37, 39 Low 37 High 46 Ceftibuten -1.21 Peptide 16 Low 44 High 44 Cephalexin -1.84 Peptide 23, 37, 40 Low 23, 29, 30, 37, 44 High 4, 23, 29, 30, 44, 46 Cephradine -1.73 Peptide 37, 41 Low 37 High 46 Enalapril 0.67 Peptide 42 Low 45 High 4 Loracarbef -0.47 Peptide 43 Low 29 High 29, 46

Gabapentin -0.66 Amino Acid 47, 48 Low 29 Low 29 Levodopa -2.82 Amino Acid 47 Low 32 High 4, 32 Ofloxacin -0.51 Amino Acid 49 Low 50 High 46 Pregabalin -0.92 Amino Acid 48 Low 51 High 46

Lamivudine -1.46 Nucleoside 52 Low 53 Low 54 Zidovudine 0.04 Nucleoside 52 Low 29, 33 High 29, 33

acLogP values taken from Reference 55

intestinal permeability rates of these transporter substrates. This may be due to the inherent expression of these transporters in the colonic region, from which Caco-2 is derived, which is much lower than regions of the small intestine.24 Because of the significant difference in expression levels (≥10-fold) for these transporters between Caco-

2 and the human small intestine, Caco-2 measurements will be unable to predict which of these compounds have high intestinal absorption.

2.2.2 Differences in the Paracellular Junctions between Caco-2 and the Human Intestine do not Explain the Inaccurate In Vitro–In Vivo Permeability Predictions by Caco-2 for Hydrophilic Compounds

Hydrophilic compounds are generally believed to use the paracellular rather than transcellular pathway through intestinal membranes because these compounds lack the necessary lipophilic properties to penetrate the cell membrane.8, 44 When a hydrophilic compound is observed to have a high intestinal permeability or be highly absorbed, the first hypothesis considered is that this occurs due to paracellular diffusion. Caco-2 has been shown to have a significantly lower number of paracellular pores than the human intestine,56 which may explain why Caco-2 measurements underpredict the intestinal permeabilities of several hydrophilic compounds that are absorbed paracellularly.57

Hence, I sought to understand whether the differences between the paracellular pathway in Caco-2 and the human intestine could explain the poor in vitro–in vivo permeability predictions of hydrophilic compounds by Caco-2.

In Table 2.3, three different cases for the intestinal transport of hydrophilic compounds via paracellular passive and transcellular carrier-mediated mechanisms are

listed. In Case 1, four drugs where greater than 80% of their intestinal absorption has been estimated to result from paracellular passive diffusion are identified. For these drugs, the intestinal absorption is characterized to be low (<90% according to BCS).

No cases in the literature could be found when the absorption of a highly absorbed drug

(having an extent measure ≥90%) definitively occurs more than 80% via the paracellular pathway. This may not occur because the paracellular pathway can be saturated (or more colloquially “clogged up”) since it only occupies 0.01–0.1% of the total surface area of the intestine,71, 72 and the size-restricting gate function of the paracellular pathway limits permeability.73 Because of this, a high extent of intestinal absorption will not be solely a consequence of paracellular diffusion and an inaccurate in vitro–in vivo permeability prediction cannot be explained solely from differences in the paracellular pathways between Caco-2 and the human intestine. Thus, compounds whose intestinal absorption occur >80% via the paracellular pathway will exhibit a low Caco-2 permeability and a low intestinal permeability or extent of absorption.

In Case 2, three hydrophilic drugs that are in part (14 to 60%) reported to be paracellularly absorbed and that are also substrates of transporters having a <10-fold difference in expression between Caco-2 and the human intestine, where the transporters have been reported to be poorly to moderately expressed in the human intestine by Sun et al.18 are identified (Table 2.3). These three drugs exhibit low Caco-2 permeabilities and low extents of absorption due to their intestinal transport being limited by their carrier- mediated and paracellular permeabilities.

In the third case, two hydrophilic drugs (ofloxacin and pregabalin) are identified that are substrates of transporters having a ≥10-fold difference in expression between

Table 2.3 Cases of intestinal transport by paracellular diffusion and carrier-mediated processes for hydrophilic compounds with their Caco-2 permeabilities and extents of absorption.

cLogP Paracellular Transporter Caco-2 Extent of Case Compound a Contribution Substrate Permeability Absorption

OAT,60 OCT60 Acyclovir -2.42 99%58 Low 33 Low63 P-gp61 Atenolol -0.11 80-100%59 Low33, 34 Low4, 63 1 OAT,62 OCT,62 Famotidine -1.17 98%58 Low15 Low64 P-gp15 Ganciclovir -2.73 99%58 Low33 Low63 OAT,60 OCT60 OAT,62 OCT,62 Cimetidine 0.19 14-30%65 P-gp67 Low33 Low4 2 Furosemide 1.90b 55%65 OAT,68 P-gp69 Low34 Low4 Ranitidine 0.67 45-60%66 OAT,62 OCT,62 Low33 Low4 P-gp67

Ofloxacin -0.51 8%58 Amino Acid49 Low50 High46 3 Pregabalin -0.92 ND51d Amino Acid48 Low51 High46 Sotalolc 0.23b ND70d NDd Low12 High46

acLogP values taken from Reference55 b 55 Measured LogD7.4 values of furosemide and sotalol are -1.54 and -0.79, respectively cHypothesized to belong to this set dAbbreviation: ND, not determined

Caco-2 and the human intestine, and these transporters are also among the most highly expressed transporters characterized by Sun et al.18 in the human intestine (Table 2.3).

While these hydrophilic compounds can be in part paracellularly absorbed, it is likely that their affinity for a highly expressed intestinal transporter – such as the amino acid, nucleoside, or peptide transporter – is responsible for their high intestinal absorption as demonstrated in Table 2.2. Pregabalin is an example of a hydrophilic compound whose high extent of absorption was long believed to result from paracellular diffusion until a highly expressed amino acid transporter was implicated in its intestinal transport.51

Similarly, the high extent of absorption and low Caco-2 permeability of sotalol likely involves a highly expressed intestinal transporter in contrast to some authors who hypothesize that sotalol’s high extent of absorption is due to paracellular diffusion.12, 46, 70

For these Case 3 drugs, Caco-2 will not be able to predict the high intestinal absorption of these hydrophilic compounds due to the cellular membrane’s significant underexpression of relevant transporters compared to the human intestine.

2.3 The Need for Larger, Single-Source Caco-2 Permeability Rate Datasets

The existence of variability in Caco-2 permeability measurements between laboratories has been widely recognized.22, 74 To help understand the impact of selecting

Caco-2 permeability data from multiple sources to develop an in silico model, Figure 2.2 illustrates the degree of variability in Caco-2 permeability measurements for 20 out of 30 drugs, that is those drugs for which both human intestinal permeability rate measurements are available4 and multiple sources report their Caco-2 permeability rate

Figure 2.2 The apparent Caco-2 permeability coefficients selected from 11 studies that included metoprolol8, 15, 19, 28-34, 75 for 20 of the 30 drugs where human intestinal permeability data are available.4 The number of observations (filled circles) per compound is 2 to 11. The lines represent arithmetic mean values.

measurements.8, 15, 19, 28-34, 75 Permeability values for each compound were included if metoprolol had also been studied in the Caco-2 evaluation to determine if the observed interlaboratory variability could be mitigated when the measurements were normalized to a reference standard. The ratios of the maximum to minimum Caco-2 permeability measurements found for each compound in the literature are presented in Table 2.4 with and without normalization to metoprolol’s permeability.

Permeability values varied greatly for lipophilic compounds and much more so for hydrophilic compounds. The largest interlaboratory variability was observed for substrates with a high affinity for amino acid and peptide transporters, e.g., amoxicillin

(180-fold) and levodopa (100-fold), followed by compounds that are paracellularly absorbed, e.g., terbutaline (27-fold) and atenolol (25-fold). With normalization to the reference standard metoprolol, the interlaboratory variability in permeability measurements for hydrophilic compounds worsened for the transporter substrates and for some of the paracellularly absorbed compounds. Even though the variability decreased for some paracellular compounds, the resulting normalized variability was still high.

Hence, for hydrophilic compounds, transporter expression and paracellular pathway differences will be a major source of variability across Caco-2 cell lines from different laboratories. Normalization to a reference standard will not successfully reduce the variability in Caco-2 permeability measurements from different laboratories. Lipophilic compounds (such as metoprolol, naproxen, propranolol, and verapamil) that are mostly absorbed by passive transcellular processes across the intestinal mucosa18 also had marked variability in Caco-2 permeability data from different laboratories (Table 2.4).

However, as seen in Table 2.4, these variances between maximum and

Table 2.4 Variability in Caco-2 permeability measurements from different investigators.

Ratio of maximum to Ratio of maximum to minimum Caco-2 minimum (with Drug cLogPa permeability normalization to measurementb metoprolol)b

Amoxicillin -1.87 180 508 Cephalexin -1.84 9.96 77.5 Levodopa -2.82 100 34.5 Lisinopril -1.69 5.77 25.4 Atenolol -0.11 25.4 9.88 Cimetidine 0.19 4.58 4.38 Furosemide 1.90c 9.29 39.2 Hydrochlorothiazide -0.37 11.9 32.2 Ranitidine 0.67 7.00 4.98 Terbutaline 0.48 27.4 133 Antipyrine 0.20 7.56 2.74 Carbamazepine 2.38 1.74 1.53 Desipramine 4.47 1.76 1.31 Enalaprilat 0.88 2.98 1.69 Ketoprofen 2.76 5.50 1.57 Metoprolol 1.49 7.78 1.00 Naproxen 2.82 8.47 2.24 Piroxicam 1.89 1.32 1.65 Propranolol 2.75 7.75 3.18 Verapamil 4.47 3.91 2.84 acLogP values taken from Reference 55 bThe maximum and minimum Caco-2 permeability measurements were determined from References 8, 15, 19, 28-34, 75. c 55 Measured LogD7.4 values of furosemide is -1.54 Bold italic compounds are known substrates of amino acid and peptide transporters Bold compounds are compounds that are paracellularly absorbed

minimum permeability measurements are generally lower, and normalization to metoprolol generally decreases the variability.

These findings suggest that the interlaboratory differences are not just a function of Caco-2 cell passage number changing inherent permeability, but instead reflect the marked variability and unpredictability of using Caco-2 permeability measurements from multiple studies for in silico correlations. Thus, the selection of particular permeability values from multiple studies could significantly skew correlations using these measurements and produce misleading information. Despite interlaboratory differences, good correlations between Caco-2 and human intestinal permeability measurements can be achieved provided that the data used for the correlations are from the same investigators, as demonstrated earlier with datasets from 7 different investigators. In order to ensure that Caco-2 permeability measurements are a reliable surrogate for human intestinal permeability measurements for in silico work, larger datasets of Caco-2 permeability measurements from the same sources are needed for creating accurate in silico predictive models.

2.4 In Silico Work Considerations

In silico predictive models of intestinal permeability have been proposed to replace a large part of conventional Caco-2 screening in drug discovery.6 For these models to be reliable and accurate, their predictions will be greatly dependent on the quality of Caco-2 permeability measurements that are used as a substitute for human intestinal permeability measurements in their development and validation.

Table 2.5 presents some common practices of selecting Caco-2 permeability measurements for datasets used to generate in silico models. Many of these datasets used absolute Caco-2 permeability measurements to determine the high versus low permeability classification of the compounds. However, Table 2.4 shows that the absolute

Caco-2 permeability measurements for compounds (e.g., the high-permeability reference standard, metoprolol) vary for different studies. Hence, rather than using a standardized permeability cutoff for determining Caco-2 permeability, it is most accurate to determine the permeability of a compound relative to a permeability reference standard as recommended by the FDA.3 Out of 55 published in silico models, more than 34% have

Caco-2 permeability measurements in their datasets from more than one source. Table 2.4 shows the variability associated with selecting compounds from more than one source.

The variability of Caco-2 permeability measurements for lipophilic compounds could be minimized with normalization to a reference standard. However, the variability in Caco-2 permeability measurements for hydrophilic compounds remained high regardless of normalization to a reference standard. Hence, it is most appropriate to select Caco-2 permeability measurements for developing in silico models from the same source. Good in vitro–in vivo permeability correlations could only be achieved when datasets of Caco-2 permeability measurements originated from the same laboratory. Selecting particular permeability measurements from different laboratories for in silico datasets not only introduces bias or variability in correlations, but can also result in generating models that produce misleading information for drug design.

As discussed earlier, paracellular pathway and transporter expression differences between Caco-2 and the human intestine can affect the in vitro predictions of human

Table 2.5 Published in silico predictive models using Caco-2 permeability measurements.

Data taken Known Substrates of Prediction from Amino Acid, Peptide, Suggested Paracellular Model multiple and/or Nucleoside Compounds Included* sources? Transporters Included* Atenolol, Cimetidine, Amoxicillin, Levodopa, Creatinine, Furosemide, Alsenz and No Lisinopril, Hydrochlorothiazide, Haenel28 Phenoxymethylpenicillin PEG400, Ranitidine, Terbutaline Atenolol, Cimetidine, Amoxicillin, Cephalexin, Creatinine, Furosemide, Avdeef and Yes Enalapril, Levodopa, Hydrochlorothiazide, Tam76 Lisinopril, Valacyclovir multiple PEG compounds, Ranitidine, Terbutaline Avdeef et Atenolol, Cimetidine, Yes No known substrates al.77 Nadolol, Terbutaline Amoxicillin, Cefatrizine, Akamatsu et Yes Cephalexin, Fosinopril, Nadolol, Ranitidine al.78** Loracarbef Bergström et No Amoxicillin, Zidovudine Acyclovir al.79 Acebutolol, Acyclovir, Atenolol, Cimetidine, Amoxicillin, Cephalexin, Furosemide, Castillo-Garit Enalapril, Gly-Pro, Yes Hydrochlorothiazide, et al.80 Lisinopril, L-Phenylalanine, Mannitol, Nadolol, Penicillin, Zidovudine Ranitidine, Scopolamine, Terbutaline

Collett et al.81 No No known substrates No suggested compounds

Acyclovir, Atenolol, Furosemide, Corti et al.82 Yes No known substrates Hydrochlorothiazide, Nadolol, Ranitidine Amoxicillin, AZT Acebutolol, Atenolol, Cruciani et (Zidovudine), Bretylium, No Hydrochlorothiazide, al.83 Cephalexin, Gabapentin, Sotalol, Terbutaline Lisinopril, Loracarbef

Table 2.5 Published in silico predictive models using Caco-2 permeability measurements. (continued)

Data taken Known Substrates of Prediction from Amino Acid, Peptide, Suggested Paracellular Model multiple and/or Nucleoside Compounds Included* sources? Transporters Included* Acebutolol, Acyclovir, Atenolol, Cimetidine, Di Fenza et Hydrochlorothiazide, No Zidovudine al.84 Mannitol, Nadolol, Ranitidine, Terbutaline, Urea Acebutolol, Acyclovir, Atenolol, Cimetidine, Furosemide,

Du-Cuny et Amoxicillin, Enalapril, Hydrochlorothiazide, Yes al.85 Zidovudine Mannitol, Nadolol, Ranitidine, Scopolamine, Terbutaline

Ekins et al.86 No No known substrates No suggested compounds

Atenolol, Mannitol, Ertl et al.87 No No known substrates Terbutaline

Föger et al.88 Yes No known substrates No suggested compounds

Ampicillin, Cefmetazole, Furosemide, Nadolol, Fossati et al.89 No Cefoperazone, Cefoxitin, Ranitidine, Terbutaline Enalapril, Norfloxacin

Acebutolol, Acyclovir, Atenolol, Cimetidine, Fujiwara et Yes Zidovudine Hydrochlorothiazide, al.90 Mannitol, Ranitidine, Terbutaline Goodwin et No No known substrates No suggested compounds al.91

Atenolol, Mannitol, PEG- Gres et al.30 No Amoxicillin, Cephalexin 400, PEG-4000, Terbutaline

Table 2.5 Published in silico predictive models using Caco-2 permeability measurements. (continued)

Data taken Known Substrates of Prediction from Amino Acid, Peptide, Suggested Paracellular Model multiple and/or Nucleoside Compounds Included* sources? Transporters Included* Amoxicillin, Bretylium Acebutolol, Acyclovir, Tosylate, Captopril, Cimetidine, Furosemide, Cefadroxil, Cephalexin, Grice et al.92 Yes Hydrochlorothiazide, Enalapril, Gabapentin, Lucifer Yellow, Nadolol, Norfloxacin, Penicillin V, Terbutaline Zidovudine Acebutolol, Acyclovir, Atenolol, Cimetidine, Amoxicillin, Enalapril, Furosemide, Guangli and Yes Ceftriaxone, Ganciclovir, Hydrochlorothiazide, Yiyu93 Zidovudine Mannitol, Nadolol, Ranitidine, Scopolamine, Terbutaline, Urea Hilgers et No No known substrates No suggested compounds al.94 Acebutolol, Acyclovir, Atenolol, Cimetidine, Amoxicillin, Enalapril, Furosemide, Hou et al.31 Yes Ceftriaxone, Ganciclovir, Hydrochlorothiazide, Zidovudine Mannitol, Nadolol, Ranitidine, Scopolamine, Terbutaline, Urea

Amoxicillin, AZT, Acebutolol, Acyclovir, Bretylium, Cefatrizine, Furosemide, Cefuroxime, Cephalexin Hydrochlorothiazide, Irvine et al.29 No Monohydrate, Gabapentin, Mannitol, Nadolol, Lisinopril, Loracarbef, Ranitidine HCl, Sotalol HCl, Penicillin V Terbutaline Hemisulfate

Jung et al.31 No No known substrates No suggested compounds Krarup et al.95 No No known substrates Acebutolol Acyclovir, Atenolol, Kulkarni et No Zidovudine Mannitol, Nadolol, al.96 Terbutaline

Table 2.5 Published in silico predictive models using Caco-2 permeability measurements. (continued)

Data taken Known Substrates of Prediction from Amino Acid, Peptide, Suggested Paracellular Model multiple and/or Nucleoside Compounds Included* sources? Transporters Included* Acyclovir, Atenolol, Cimetidine, Furosemide, Hydrochlorothiazide, Li et al.15 No No known substrates Mannitol, Nadolol, PEG- 4000, Ranitidine, Terbutaline Acebutolol, Acyclovir, Cimetidine, Liang et al.97 No No known substrates Hydrochlorothiazide, Mannitol, Ranitidine, Scopolamine, Urea

Atenolol, Creatinine, Linnankoski et No Ganciclovir Mannitol, Nadolol, al.98 Terbutaline

Lundquist et Mannitol, Terbutaline, No No known substrates al.99 Urea

Marrero Ponce Atenolol, Mannitol, No No known substrates et al.100 Terbutaline

Acebutolol, Acyclovir, Cimetidine, Marrero Ponce No No known substrates Hydrochlorothiazide, et al.101 Mannitol, Ranitidine, Scopolamine, Urea Amoxicillin, Cephalexin, D–Phe–L–Pro, Gabapentin, Acebutolol, Atenolol, Marrero Ponce Yes Gly-Pro, Lisinopril, L- Cimetidine, Furosemide, et al.102 Phenylalanine, Penicillin Mannitol, Nadolol, Urea G, Zidovudine Acyclovir, Atenolol, Matsson et Glycylsarcosine, Cimetidine, Creatinine, No al.103 Lobucavir, Valacyclovir Nadolol, Ranitidine, Scopolamine, Terbutaline

Table 2.5 Published in silico predictive models using Caco-2 permeability measurements. (continued)

Data taken Known Substrates of Prediction from Amino Acid, Peptide, Suggested Paracellular Model multiple and/or Nucleoside Compounds Included* sources? Transporters Included* Acebutolol, Atenolol, AZT, Benzylpenicillin, L- Nordquist et Cimetidine, Inulin, No Alanine, L-Dopa, al.104 Mannitol, Nadolol, Gabapentin, Methyldopa Ranitidine, Urea

Norinder et Atenolol, Mannitol, PEG, No No known substrates al.105 Terbutaline

Norinder and Atenolol, Mannitol, No No known substrates Osterberg106 Terbutaline Osterberg and No No known substrates Acebutolol Norinder107 Osterberg and Atenolol, Mannitol, PEG, Yes No known substrates Norinder108 Terbutaline Amoxicillin, Ampicillin, Acebutolol, Acyclovir, Ceftriaxone, Cephalexin, Atenolol, Cimetidine, Cephradine, Enalapril, Famotidine, Furosemide, Paixão et al.109 Yes Gly-Pro, Levodopa, Hydrochlorothiazide, Lisinopril, Loracarbef, Mannitol, multiple PEG Norfloxacin, Penicillin V, compounds, Ranitidine, Valacyclovir, Zidovudine Scopolamine, Terbutaline Palm et al.110 No No known substrates Atenolol Palm et al.111 No No known substrates Atenolol Acyclovir, Atenolol, Parrott and No Ganciclovir, Penicillin V Hydrochlorothiazide, Lavé 112 Ranitidine, Terbutaline Pickett et al. Yes No known substrates No suggested compounds 113** Refsgaard et No No known substrates Furosemide, Ranitidine al.114** Acebutolol, Acyclovir, Atenolol, Cimetidine, Ren and Yes Zidovudine Hydrochlorothiazide, Lien115 Mannitol, Nadolol, Terbutaline

Table 2.5 Published in silico predictive models using Caco-2 permeability measurements. (continued)

Data taken Known Substrates of Prediction from Amino Acid, Peptide, Suggested Paracellular Model multiple and/or Nucleoside Compounds Included* sources? Transporters Included* Acebutolol, Furosemide, Cefuroxime, Cephalexin, Skolnik et Hydrochlorothiazide, No Ganciclovir, Lisinopril, al.116 Nadolol, Ranitidine, Norfloxacin Terbutaline Stenberg et Atenolol, Cimetidine, No No known substrates al.117** Mannitol Benazepril, Acebutolol, Acyclovir, Benzylpenicillin, Atenolol, Famotidine, Thomas et No Bretylium, Enalapril, Hydrochlorothiazide, al.118 Gabapentin, Norfloxacin, Nadolol, Ranitidine, Ofloxacin Scopolamine, Terbutaline

Tronde et al.119 No No known substrates Terbutaline

Waterbeemd et Atenolol, Mannitol, No No known substrates al.35 Terbutaline

Acebutolol, Acyclovir, Hydrochlorothiazide, Yamashita et Yes Ganciclovir, Zidovudine Mannitol, Nadolol, al.44 Ranitidine, Scopolamine, Terbutaline Acebutolol, Acyclovir, Yazdanian et Hydrochlorothiazide, No Zidovudine al.33 Mannitol, Nadolol, Ranitidine, Terbutaline Acebutolol, Atenolol, Zhang et al.113 Yes No known substrates Mannitol, Terbutaline Amoxicillin, Bretylium Acebutolol, Acyclovir, Tosylate, Captopril, Cimetidine, Furosemide, Cefadroxil, Cephalexin, Zhu et al.120 Yes Hydrochlorothiazide, Enalapril, Gabapentin, Lucifer Yellow, Nadolol, Norfloxacin, Penicillin V, Terbutaline Zidovudine *Likely underestimated as these compounds have not been thoroughly characterized **Entire dataset not published to examine all of its compounds

intestinal permeabilities for hydrophilic drugs. At least 87% of the in silico models in

Table 2.5 use datasets containing Caco-2 permeability measurements for hydrophilic

116 drugs, defined here as compounds having LogP (or LogD7.4) values <1. Estimations of the human intestinal permeability or absorption of these compounds were generally observed to be underpredicted. Complete exclusion of all hydrophilic compounds from training and validation datasets produces models that are restricted to a confined chemical space. Hence, for the 87% of in silico models that also include implied paracellularly absorbed compounds, rather than eliminating these compounds from datasets, it is suggested that correction factors should be in place to adjust for the pore radius of tight junctions in Caco-2 monolayers to improve their paracellular permeability prediction.16

While it is generally accepted that substrates of carrier-mediated transporters have higher in vivo permeabilities than those measured in Caco-2, there are many examples of compounds where a low permeability in Caco-2 is reflective of a low permeability in vivo. However, I show that these compounds are likely substrates of transporters that have <10-fold variability in expression between Caco-2 and the human small intestine.

For compounds that are particularly substrates of highly expressed intestinal transporters that have a ≥10-fold variability in expression between Caco-2 and the human small intestine such as the amino acid, nucleoside, and peptide transporters, it is shown that not only are their human intestinal permeability predictions incorrect but that when Caco-2 permeability measurements are included in in vitro–in vivo correlations, poorer correlations are obtained. Similarly, it is very possible that 50% of the published models that include Caco-2 permeability measurements for these substrates as part of their training and/or validation datasets during in silico model development can give skewed,

misleading correlations and therefore these compounds should be excluded from correlations. In fact, these models could not predict the high intestinal permeability or absorption of these substrates. The only scenario where including these substrates may not impact an in silico correlation is perhaps when predicting passive intestinal permeability. Consequently, these in silico models will not accurately predict the high intestinal permeability of hydrophilic compounds in drug discovery, which may only be seen through their affinity for carrier-mediated transporters that are significantly underexpressed in the Caco-2 cell line.

2.5 In Vitro Screening Considerations

It is very important to accurately determine whether a drug candidate has desirable biopharmaceutical properties, such as high intestinal permeability, during drug discovery. With a standardized, validated permeability method, Caco-2 can predict the high intestinal permeability of lipophilic compounds relative to a permeability standard such as metoprolol. This same case does not apply to certain hydrophilic compounds.

Because hydrophilic drugs represent about 27% of oral drugs presently on the market, and approximately the same percentage of NMEs being investigated by the industry,121 it is crucial to recognize the limitations Caco-2 may have as the primary screening source of NMEs for lead selection, lead optimization, and candidate prioritization.

It was discussed earlier that the intestinal permeability of hydrophilic compounds will likely be dependent on either passive paracellular diffusion and/or carrier-mediated transport mechanisms. Strategies for improving the oral absorption of hydrophilic

compounds have either attempted to enhance drug absorption via the paracellular pathway,71, 122, 123 or to create prodrug forms of the NME that target highly expressed intestinal transporters, such as the amino acid,124, 125 nucleoside,126 and peptide transporters.127, 128 There is no published evidence to support that manipulating the paracellular pathway increased the intestinal permeability or absorption of a hydrophilic compound from low to high (i.e., changing a drug from BCS Class 3 to Class 1). Hence, the latter strategy of targeting a highly expressed intestinal transporter may be more successful in improving the intestinal permeability or BCS absorption of a hydrophilic

NME.

For compounds that are paracellularly absorbed, Caco-2 will underpredict their human intestinal permeabilities due to the lesser number of paracellular pores in Caco-2 compared to the human intestine.56 Correcting for their paracellular differences can improve their paracellular permeability prediction.16 However, Caco-2 is not suitable for estimating the contribution of the paracellular pathway in the intestinal transport of

NMEs because it lacks the sensitivity to detect changes in the paracellular pathway due to the presence of a lesser number of pores. As such, other systems have been proposed as better methodologies than Caco-2 for evaluating paracellular intestinal permeability.56, 57,

129

Several hydrophilic drugs may be highly absorbed because of their carrier- mediated transport. As such, assessing the BCS classification of hydrophilic compounds using the cLogP method, where hydrophilic compounds are generally classified as being poorly absorbed due to their low distribution coefficients31 is not recommended. For example, this method cannot differentiate between BCS Class 1 zidovudine and BCS

Class 3 lamivudine, thereby classifying both drugs as BCS Class 3 compounds. The better affinity for a nucleoside transporter may be the reason for zidovudine being highly absorbed and lamivudine being moderately absorbed and ineligible to be a BCS Class 1 drug.

It is likely that Caco-2 permeability measurements will incorrectly predict the high intestinal permeabilities of many hydrophilic compounds because of significant transporter expression differences between Caco-2 and the human intestine, such as for the amino acid, nucleoside, and peptide transporters.18 No published evidence of an optimized Caco-2 permeability method to accurately predict the high intestinal permeability of these hydrophilic substrates could be found. As a result, it is recommended that alternate or additional methods to Caco-2 for not only determining the intestinal permeability of hydrophilic compounds, but also for evaluating the substrate specificity of highly expressed intestinal transporters on the intestinal permeability of hydrophilic compounds are used. While alternate in vitro systems37, 41 may be suitable for evaluating their substrate specificity, these cell lines may under- or overexpress these transporters relative to levels in the human intestine, which in turn can under- or overpredict the permeability of these compounds. Hence, the most suitable methodology to characterize overall carrier-mediated transport and the intestinal permeability of these compounds may be using the rat intestine.130

2.6 Regulatory Considerations

With current regulatory standards conveying the importance of compounds that

have high intestinal permeability,3 drug discovery and development programs that use

Caco-2 as a primary permeability screening tool may be discarding viable drug candidates that are incorrectly predicted to be poorly permeable. For example, some hydrophilic compounds that are substrates of members of the amino acid, nucleoside, and peptide transporter families can have high intestinal permeabilities and be highly absorbed. However, due to the significant difference in the expression of these transporters between Caco-2 and the human intestine, the high intestinal permeability of these compounds may be incorrectly predicted to be poorly permeable by Caco-2.

The current FDA BCS guidance (Guidance) can be viewed as perpetuating the misconception that methods for evaluating intestinal permeability should be limited to studying passively transported compounds. Not only have I discussed earlier why this is this extremely difficult, but Simon and colleagues131 have shown that four of the nine high-permeability model drugs, and all four low-permeability model drugs, that are recommended for evaluating permeability in the Guidance, exhibit significant active intestinal transport. Whenever the regulatory agency has found discrepancies between

Caco-2 and an in vivo method17, where they use the in vivo information to “trump” in vitro data, it is likely that the difference in permeability observed is due to a carrier- mediated mechanism that is lacking in the Caco-2 cell line. Thus, it is important for the

Guidance to clarify that Caco-2 will likely exhibit passive permeability, while the other human or animal methods for evaluating intestinal permeability may involve carrier- mediated transport. When intestinal permeability measured across the Caco-2 cell line is high (i.e., the passive permeability of the compound is high), then the total permeability will still be high according to the BCS, even if there is a carrier-mediated component not

being characterized in Caco-2. However, when Caco-2 (or passive) permeability is low, hydrophilic compounds that are substrates for transporters (e.g., amino acid, nucleoside, and peptide transporters) can still exhibit a BCS Class 1 high permeability. Hence, the

FDA should update its Guidance for Industry to caution drug discovery programs that are evaluating hydrophilic compounds from using Caco-2 as a primary screening source.

Moreover, they should recommend using methods to classify permeability (such as those in humans or rats) that may be more representative of the absorptive processes for these compounds, thereby enabling programs to correctly assess the human intestinal permeability of these compounds.

Not all of the model drugs selected by the FDA are appropriate for establishing the suitability of a permeability method. For example, the list of model drugs includes amoxicillin, a peptide transporter substrate, which, as discussed earlier, can yield discrepancies in permeability between in vitro and in vivo methods. Therefore, it is recommended that amoxicillin be deleted from the list of model drugs and possibly be substituted by another lowly permeable drug that is more passively absorbed.

2.7 Conclusions

Several recent articles have extensively elaborated on the importance, concepts, and problems of drug permeation across biological membranes in drug discovery, development, and regulatory evaluation.16, 132, 133 Many of these expert reviews have concluding remarks stating that better systems need to be implemented in order to achieve good in vitro–in vivo correlations that are crucial for progress in the field. While

it seems ideal to work towards developing new in vitro systems that can better predict the in vivo performance of NMEs, it is not practical to continue using current in vitro permeability systems in drug discovery, development, and regulatory settings, all the while neglecting the known limitations of these systems.

Industry and regulatory scientists recognize the need for in silico and in vitro models that can accurately forecast in vivo information and serve as a reliable surrogate for animal and human studies, thereby facilitating the discovery and development pipeline, as well as the regulatory approval processes.16, 134 The Caco-2 cell system is an example of such an in vitro methodology that can produce discordant human intestinal permeability predictions. With a standardized, validated permeability method, Caco-2 can predict the human intestinal permeability of lipophilic compounds. However, it is not suitable for predicting the human intestinal permeability of hydrophilic compounds, particularly those that are substrates of highly expressed intestinal transporters such as members of the amino acid, nucleoside, and peptide transporter families. To date, rat intestinal tissues have been the most successful at predicting the human intestinal permeability of these compounds.

Apart from cautioning drug discovery programs that are interested in these compounds and urging them to use alternate or additional permeability methods during their permeability screening process, investigators developing in silico predictive models using such Caco-2 permeability measurements as a surrogate for human intestinal permeability values should be cautioned when developing their models. The inclusion of such data can distort in silico correlations, and should be excluded altogether. Because the absolute Caco-2 permeability measurements of compounds vary among different

Caco-2 studies, permeability classification should be made relative to a reference standard rather than an absolute permeability cutoff. Significant amounts of variability can be introduced into in silico models when their training and validation datasets consist of Caco-2 permeability measurements taken from more than one source. Normalization to the same reference standard in each dataset may minimize the interlaboratory variability for lipophilic compounds. However, this is unsuccessful for hydrophilic compounds. As such, in silico modelers should not use Caco-2 permeability measurements in their in silico model development from more than one source. Lastly, regulatory agencies should recommend that drug discovery and development programs use alternative permeability methods to Caco-2, such as those involving human or animal intestinal tissues, for assessing the intestinal permeability of these NMEs and supporting their new drug applications.

2.8 Acknowledgment

This work is a reprint of “Drug discovery and regulatory considerations for improving in silico and in vitro predictions that use Caco-2 as a surrogate for human intestinal permeability measurements. AAPS J. 2013, 15, 483-497”.121 Refer to the Appendix section for documentation of permission to republish this material as part of my thesis dissertation.

2.9 References

1. Testa, B.; Caldwell, J. Prodrugs revisited: the "ad hoc" approach as a complement to ligand design. Med. Res. Rev. 1996, 16, 233-241.

2. Amidon, G. L.; Lennernas, H.; Shah, V. P.; Crison, J. R. A Theoretical Basis for a Biopharmaceutic Drug Classification: The Correlation of In Vitro Drug Product

Dissolution and In Vivo Bioavailability. Pharm. Res. 1995, 12, 413-420.

3. Waiver of In Vivo Bioavailability and Bioequivalence Studies for Immediate-

Release Solid Oral Dosage Forms Based on a Biopharmaceutics Classification System.

FDA Guidance for Industry; Food and Drug Administration: Rockville, MD, 2000.

4. Lennernas, H. Intestinal Permeability and its Relevance for Absorption and

Elimination. Xenobiotica 2007, 37, 1015-1051.

5. Artursson, P.; Palm, K.; Luthman, K. Caco-2 monolayers in experimental and theoretical predictions of drug transport. Adv. Drug Deliv. Rev. 2001, 46, 27-43.

6. Ungell, A. L. Caco-2 Replace or Refine? Drug Discov. Today Tech. 2004, 1,

423-430.

7. Hidalgo, I. J.; Raub, T. J.; Borchardt, R. T. Characterization of the human colon carcinoma cell line (Caco-2) as a model system for intestinal epithelial permeability.

Gastroenterology 1989, 96, 736-749.

8. Artursson, P.; Karlsson, J. Correlation between oral drug absorption in humans and apparent drug permeability coefficients in human intestinal epithelial (Caco-2) cells.

Biochem. Biophys. Res. Commun. 1991, 175, 880-885.

9. Lennernas, H.; Palm, K.; Fagerholm, U.; Artursson, P. Comparison between active and passive drug transport in human intestinal epithelial (caco-2) cells in vitro and human jejunum in vivo. Int. J. Pharm. 1996, 127, 103-107.

10. Yee, S. In vitro permeability across Caco-2 cells (colonic) can predict in vivo

(small intestinal) absorption in man--fact or myth. Pharm. Res. 1997, 14, 763-766.

11. Benet, L. Z.; Larregieu, C. A. The FDA Should Eliminate the Ambiguities in the

Current BCS Biowaiver Guidance and Make Public the Drugs for Which BCS

Biowaivers Have Been Granted. Clin. Pharmacol. Ther. 2010, 88, 405-407.

12. Yang, Y.; Faustino, P. J.; Volpe, D. A.; Ellison, C. D.; Lyon, R. C.; Yu, L. X.

Biopharmaceutics Classification of Selected Beta-Blockers: Solubility and Permeability

Class Membership. Mol. Pharmaceutics 2007, 4, 608-614.

13. Sun, H.; Chow, E. C.; Liu, S.; Du, Y.; Pang, K. S. The Caco-2 cell monolayer: usefulness and limitations. Expert Opin. Drug Metab. Toxicol. 2008, 4, 395-411.

14. Press, B.; Di Grandi, D. Permeability for intestinal absorption: Caco-2 assay and related issues. Curr. Drug Metab. 2008, 9, 893-900.

15. Li, C.; Liu, T.; Cui, X.; Uss, A. S.; Cheng, K. C. Development of in vitro pharmacokinetic screens using Caco-2, human hepatocyte, and Caco-2/human hepatocyte hybrid systems for the prediction of oral bioavailability in humans. J. Biomol. Screen.

2007, 12, 1084-1091.

16. Sugano, K.; Takata, N.; Machida, M.; Saitoh, K.; Terada, K. Prediction of

Passive Intestinal Absorption Using Bio-Mimetic Artificial Membrane Permeation Assay and the Paracellular Pathway Model. Int. J. Pharm. 2002, 241, 241-251.

17. Chen, M. L.; Amidon, G. L.; Benet, L. Z.; Lennernas, H.; Yu, L. X. The BCS,

BDDCS, and regulatory guidances. Pharm. Res. 2011, 28, 1774-1778.

18. Sun, D.; Lennernas, H.; Welage, L. S.; Barnett, J. L.; Landowski, C. P.; Foster,

D.; Fleisher, D.; Lee, K. D.; Amidon, G. L. Comparison of human duodenum and Caco-

2 gene expression profiles for 12,000 gene sequences tags and correlation with permeability of 26 drugs. Pharm. Res. 2002, 19, 1400-1416.

19. Hilgendorf, C.; Ahlin, G.; Seithel, A.; Artursson, P.; Ungell, A. L.; Karlsson, J.

Expression of thirty-six drug transporter genes in human intestine, liver, kidney, and organotypic cell lines. Drug Metab. Dispos. 2007, 35, 1333-1340.

20. Chu, X. Y.; Sanchez-Castano, G. P.; Higaki, K.; Oh, D. M.; Hsu, C. P.; Amidon,

G. L. Correlation between epithelial cell permeability of cephalexin and expression of intestinal oligopeptide transporter. J. Pharmacol. Exp. Ther. 2001, 299, 575-582.

21. Behrens, I.; Kamm, W.; Dantzig, A. H.; Kissel, T. Variation of peptide transporter (PepT1 and HPT1) expression in Caco-2 cells as a function of cell origin. J.

Pharm. Sci. 2004, 93, 1743-1754.

22. Hayeshi, R.; Hilgendorf, C.; Artursson, P.; Augustijns, P.; Brodin, B.; Dehertogh,

P.; Fisher, K.; Fossati, L.; Hovenkamp, E.; Korjamo, T.; Masungi, C.; Maubon, N.; Mols,

R.; Mullertz, A.; Monkkonen, J.; O'Driscoll, C.; Oppers-Tiemissen, H. M.; Ragnarsson,

E. G.; Rooseboom, M.; Ungell, A. L. Comparison of drug transporter gene expression and functionality in Caco-2 cells from 10 different laboratories. Eur. J. Pharm. Sci. 2008,

35, 383-396.

23. Chong, S.; Dando, S. A.; Soucek, K. M.; Morrison, R. A. In vitro permeability through Caco-2 cells is not quantitatively predictive of in vivo absorption for peptide-like drugs absorbed via the dipeptide transporter system. Pharm. Res. 1996, 13, 120-123.

24. Englund, G.; Rorsman, F.; Ronnblom, A.; Karlbom, U.; Lazorova, L.; Grasjo, J.;

Kindmark, A.; Artursson, P. Regional levels of drug transporters along the human intestinal tract: co-expression of ABC and SLC transporters and comparison with Caco-2 cells. Eur. J. Pharm. Sci. 2006, 29, 269-277.

25. Seithel, A.; Karlsson, J.; Hilgendorf, C.; Bjorquist, A.; Ungell, A. L. Variability in mRNA expression of ABC- and SLC-transporters in human intestinal cells: comparison between human segments and Caco-2 cells. Eur. J. Pharm. Sci. 2006, 28,

291-299.

26. Bleasby, K.; Castle, J. C.; Roberts, C. J.; Cheng, C.; Bailey, W. J.; Sina, J. F.;

Kulkarni, A. V.; Hafey, M. J.; Evers, R.; Johnson, J. M.; Ulrich, R. G.; Slatter, J. G.

Expression profiles of 50 xenobiotic transporter genes in humans and pre-clinical species: a resource for investigations into drug disposition. Xenobiotica 2006, 36, 963-988.

27. Nishimura, M.; Naito, S. Tissue-specific mRNA expression profiles of human

ATP-binding cassette and solute carrier transporter superfamilies. Drug Metab.

Pharmacokinet. 2005, 20, 452-477.

28. Alsenz, J.; Haenel, E. Development of a 7-day, 96-Well Caco-2 Permeability

Assay with High-Throughput Direct UV Compound Analysis. Pharm. Res. 2003, 20,

1961-1969.

29. Irvine, J. D.; Takahashi, L.; Lockhart, K.; Cheong, J.; Tolan, J. W.; Selick, H. E.;

Grove, J. R. MDCK (Madin-Darby Canine Kidney) Cells: A Tool for Membrane

Permeability Screening. J. Pharm. Sci. 1999, 88, 28-33.

30. Gres, M. C.; Julian, B.; Bourrie, M.; Meunier, V.; Roques, C.; Berger, M.;

Boulenc, X.; Berger, Y.; Fabre, G. Correlation between oral drug absorption in humans, and apparent drug permeability in TC-7 cells, a human epithelial intestinal cell line: comparison with the parental Caco-2 cell line. Pharm. Res. 1998, 15, 726-733.

31. Kasim, N. A.; Whitehouse, M.; Ramachandran, C.; Bermejo, M.; Lennernas, H.;

Hussain, A. S.; Junginger, H. E.; Stavchansky, S. A.; Midha, K. K.; Shah, V. P.; Amidon,

G. L. Molecular Properties of WHO Essential Drugs and Provisional Biopharmaceutical

Classification. Mol. Pharmaceutics 2004, 1, 85-96.

32. Lennernas, H. Human intestinal permeability. J. Pharm. Sci. 1998, 87, 403-410.

33. Yazdanian, M.; Glynn, S. L.; Wright, J. L.; Hawi, A. Correlating Partitioning and

Caco-2 Cell Permeability of Structurally Diverse Small Molecular Weight Compounds.

Pharm. Res. 1998, 15, 1490-1494.

34. Markowska, M.; Oberle, R.; Juzwin, S.; Hsu, C. P.; Gryszkiewicz, M.; Streeter,

A. J. Optimizing Caco-2 cell monolayers to increase throughput in drug intestinal absorption analysis. J. Pharmacol. Toxicol. Methods 2001, 46, 51-55.

35. Camenisch, G.; Alsenz, J.; van de Waterbeemd, H.; Folkers, G. Estimation of

Permeability by Passive Diffusion through Caco-2 Cell Monolayers Using the Drugs'

Lipophilicity and Molecular Weight. Eur. J. Pharm. Sci. 1998, 6, 317-324.

36. Tsuji, A. Transporter-mediated drug interactions. Drug Metab. Pharmacokinet.

2002, 17, 253-274.

37. Chen, X.; Murawski, A.; Patel, K.; Crespi, C. L.; Balimane, P. V. A Novel

Design of Artificial Membrane for Improving the PAMPA Model. Pharm. Res. 2008, 25,

1511-1520.

38. Li, M.; Anderson, G. D.; Phillips, B. R.; Kong, W.; Shen, D. D.; Wang, J.

Interactions of amoxicillin and cefaclor with human renal organic anion and peptide transporters. Drug Metab. Dispos. 2006, 34, 547-555.

39. Knutter, I.; Kottra, G.; Fischer, W.; Daniel, H.; Brandsch, M. High-affinity interaction of sartans with H+/peptide transporters. Drug Metab. Dispos. 2009, 37, 143-

149.

40. Watanabe, K.; Jinriki, T.; Sato, J. Effects of progesterone and norethisterone on cephalexin uptake in the human intestinal cell line Caco-2. Biol. Pharm. Bull. 2004, 27,

559-563.

41. Han, H. K.; Rhie, J. K.; Oh, D. M.; Saito, G.; Hsu, C. P.; Stewart, B. H.; Amidon,

G. L. CHO/hPEPT1 cells overexpressing the human peptide transporter (hPEPT1) as an alternative in vitro model for peptidomimetic drugs. J. Pharm. Sci. 1999, 88, 347-350.

42. Friedman, D. I.; Amidon, G. L. Passive and carrier-mediated intestinal absorption components of two angiotensin converting enzyme (ACE) inhibitor prodrugs in rats: enalapril and fosinopril. Pharm. Res. 1989, 6, 1043-1047.

43. Dantzig, A. H.; Bergin, L. Uptake of the Cephalosporin, Cephalexin, by a

Dipeptide Transport Carrier in the Human Intestinal Cell Line, Caco-2. Biochim.

Biophys. Acta 1990, 1027, 211-217.

44. Takagi, T.; Ramachandran, C.; Bermejo, M.; Yamashita, S.; Yu, L. X.; Amidon,

G. L. A Provisional Biopharmaceutical Classification of the Top 200 Oral Drug Products

in the United States, Great Britain, Spain, and Japan. Mol. Pharmaceutics 2006, 3, 631-

643.

45. Fossati, L.; Dechaume, R.; Hardillier, E.; Chevillon, D.; Prevost, C.; Bolze, S.;

Maubon, N. Use of simulated intestinal fluid for Caco-2 permeability assay of lipophilic drugs. Int J Pharm 2008, 360, 148-55.

46. Chen, M. L.; Yu, L. The Use of Drug Metabolism for Prediction of Intestinal

Permeability. Mol. Pharmaceutics 2009, 6, 74-81.

47. Uchino, H.; Kanai, Y.; Kim, D. K.; Wempe, M. F.; Chairoungdua, A.; Morimoto,

E.; Anders, M. W.; Endou, H. Transport of amino acid-related compounds mediated by

L-type amino acid transporter 1 (LAT1): insights into the mechanisms of substrate recognition. Mol. Pharmacol. 2002, 61, 729-737.

48. Su, T. Z.; Feng, M. R.; Weber, M. L. Mediation of highly concentrative uptake of pregabalin by L-type amino acid transport in Chinese hamster ovary and Caco-2 cells. J.

Pharmacol. Exp. Ther. 2005, 313, 1406-1415.

49. Rabbaa, L.; Dautrey, S.; Colas-Linhart, N.; Carbon, C.; Farinotti, R. Absorption of ofloxacin isomers in the rat small intestine. Antimicrob. Agents Chemother. 1997, 41,

2274-2277.

50. Mensch, J.; Melis, A.; Mackie, C.; Verreck, G.; Brewster, M. E.; Augustijns, P.

Evaluation of various PAMPA models to identify the most discriminating method for the prediction of BBB permeability. Eur. J. Pharm. Biopharm. 2010, 74, 495-502.

51. Jezyk, N.; Li, C.; Stewart, B. H.; Wu, X.; Bockbrader, H. N.; Fleisher, D.

Transport of pregabalin in rat intestine and Caco-2 monolayers. Pharm. Res. 1999, 16,

519-526.

52. Gray, J. H.; Owen, R. P.; Giacomini, K. M. The concentrative nucleoside transporter family, SLC28. Pflugers Arch. 2004, 447, 728-734.

53. de Souza, J.; Benet, L. Z.; Huang, Y.; Storpirtis, S. Comparison of bidirectional lamivudine and zidovudine transport using MDCK, MDCK-MDR1, and Caco-2 cell monolayers. J. Pharm. Sci. 2009, 98, 4413-4419.

54. Mueller, B. U.; Lewis, L. L.; Yuen, G. J.; Farley, M.; Keller, A.; Church, J. A.;

Goldsmith, J. C.; Venzon, D. J.; Rubin, M.; Pizzo, P. A.; Balis, F. M. Serum and cerebrospinal fluid pharmacokinetics of intravenous and oral lamivudine in human immunodeficiency virus-infected children. Antimicrob. Agents Chemother. 1998, 42,

3187-3192.

55. Benet, L. Z.; Broccatelli, F.; Oprea, T. I. BDDCS applied to over 900 drugs.

AAPS J. 2011, 13, 519-547.

56. Linnankoski, J.; Makela, J.; Palmgren, J.; Mauriala, T.; Vedin, C.; Ungell, A. L.;

Lazorova, L.; Artursson, P.; Urtti, A.; Yliperttula, M. Paracellular porosity and pore size of the human intestinal epithelium in tissue and cell culture models. J. Pharm. Sci. 2010,

99, 2166-2175.

57. Hilgendorf, C.; Spahn-Langguth, H.; Regardh, C. G.; Lipka, E.; Amidon, G. L.;

Langguth, P. Caco-2 versus Caco-2/HT29-MTX co-cultured cell lines: permeabilities via diffusion, inside- and outside-directed carrier-mediated transport. J. Pharm. Sci. 2000,

89, 63-75.

58. Tam, K. Y.; Avdeef, A.; Tsinman, O.; Sun, N. The permeation of amphoteric drugs through artificial membranes--an in combo absorption model based on paracellular and transmembrane permeability. J. Med. Chem. 2010, 53, 392-401.

59. Adson, A.; Burton, P.; Raub, T.; Barsuhn, C.; Audus, K.; Ho, N. Passive diffusion of weak organic electrolytes across Caco-2 cell monolayers: uncoupling the contributions of hydrodynamic, transcellular, and paracellular barriers. J. Pharm. Sci.

1995, 84, 1197-1204.

60. Takeda, M.; Khamdang, S.; Narikawa, S.; Kimura, H.; Kobayashi, Y.;

Yamamoto, T.; Cha, S. H.; Sekine, T.; Endou, H. Human organic anion transporters and human organic cation transporters mediate renal antiviral transport. J. Pharmacol. Exp.

Ther. 2002, 300, 918-924.

61. Neuhoff, S.; Ungell, A. L.; Zamora, I.; Artursson, P. pH-dependent bidirectional transport of weakly basic drugs across Caco-2 monolayers: implications for drug-drug interactions. Pharm. Res. 2003, 20, 1141-1148.

62. Tahara, H.; Kusuhara, H.; Endou, H.; Koepsell, H.; Imaoka, T.; Fuse, E.;

Sugiyama, Y. A species difference in the transport activities of H2 receptor antagonists by rat and human renal organic anion and cation transporters. J. Pharmacol. Exp. Ther.

2005, 315, 337-345.

63. Tavelin, S.; Taipalensuu, J.; Soderberg, L.; Morrison, R.; Chong, S.; Artursson, P.

Prediction of the oral absorption of low-permeability drugs using small intestine-like

2/4/A1 cell monolayers. Pharm. Res. 2003, 20, 397-405.

64. Dahan, A.; Miller, J. M.; Amidon, G. L. Prediction of Solubility and Permeability

Class Membership: Provisional BCS Classification of the World's Top Oral Drugs. AAPS

J. 2009, 11, 740-6.

65. Pade, V.; Stavchansky, S. Estimation of the relative contribution of the transcellular and paracellular pathway to the transport of passively absorbed drugs in the

Caco-2 cell culture model. Pharm. Res. 1997, 14, 1210-1215.

66. Bourdet, D. L.; Pollack, G. M.; Thakker, D. R. Intestinal absorptive transport of the hydrophilic cation ranitidine: a kinetic modeling approach to elucidate the role of uptake and efflux transporters and paracellular vs. transcellular transport in Caco-2 cells.

Pharm. Res. 2006, 23, 1178-1187.

67. Collett, A.; Higgs, N. B.; Sims, E.; Rowland, M.; Warhurst, G. Modulation of the permeability of H2 receptor antagonists cimetidine and ranitidine by P-glycoprotein in rat intestine and the human colonic cell line Caco-2. J. Pharmacol. Exp. Ther. 1999, 288,

171-178.

68. Hasannejad, H.; Takeda, M.; Taki, K.; Shin, H. J.; Babu, E.; Jutabha, P.;

Khamdang, S.; Aleboyeh, M.; Onozato, M. L.; Tojo, A.; Enomoto, A.; Anzai, N.;

Narikawa, S.; Huang, X. L.; Niwa, T.; Endou, H. Interactions of human organic anion transporters with diuretics. J. Pharmacol. Exp. Ther. 2004, 308, 1021-1029.

69. Chiou, W. L.; Chung, S. M.; Wu, T. C.; Ma, C. A comprehensive account on the role of efflux transporters in the gastrointestinal absorption of 13 commonly used substrate drugs in humans. Int. J. Clin. Pharmacol. Ther. 2001, 39, 93-101.

70. Alt, A.; Potthast, H.; Moessinger, J.; Sickmuller, B.; Oeser, H.

Biopharmaceutical characterization of sotalol-containing oral immediate release drug products. Eur. J. Pharm. Biopharm. 2004, 58, 145-150.

71. Nellans, N. Paracellular intestinal transport: modulation of absorption. Adv. Drug

Deliv. Rev. 1991, 7, 339-364.

72. Pappenheimer, J. R.; Reiss, K. Z. Contribution of solvent drag through intercellular junctions to absorption of nutrients by the small intestine of the rat. J.

Membr. Biol. 1987, 100, 123-136.

73. Madara, J. L. Regulation of the movement of solutes across tight junctions. Annu.

Rev. Physiol. 1998, 60, 143-159.

74. Volpe, D. A. Variability in Caco-2 and MDCK cell-based intestinal permeability assays. J. Pharm. Sci. 2008, 97, 712-725.

75. Caldwell, G. W.; Easlick, S. M.; Gunnet, J.; Masucci, J. A.; Demarest, K. In vitro permeability of eight beta-blockers through Caco-2 monolayers utilizing liquid chromatography/electrospray ionization mass spectrometry. J. Mass Spectrom. 1998, 33,

607-614.

76. Avdeef, A.; Tam, K. Y. How well can the Caco-2/Madin-Darby canine kidney models predict effective human jejunal permeability? J. Med. Chem. 2010, 53, 3566-

3584.

77. Avdeef, A.; Artursson, P.; Neuhoff, S.; Lazorova, L.; Grasjo, J.; Tavelin, S.

Caco-2 permeability of weakly basic drugs predicted with the double-sink PAMPA pKa(flux) method. Eur. J. Pharm. Sci. 2005, 24, 333-349.

78. Akamatsu, M.; Fujikawa, M.; Nakao, K.; Shimizu, R. In silico prediction of human oral absorption based on QSAR analyses of PAMPA permeability. Chem.

Biodivers. 2009, 6, 1845-1866.

79. Bergstrom, C. A.; Strafford, M.; Lazorova, L.; Avdeef, A.; Luthman, K.;

Artursson, P. Absorption classification of oral drugs based on molecular surface properties. J. Med. Chem. 2003, 46, 558-570.

80. Castillo-Garit, J. A.; Marrero-Ponce, Y.; Torrens, F.; Garcia-Domenech, R.

Estimation of ADME properties in drug discovery: predicting Caco-2 cell permeability using atom-based stochastic and non-stochastic linear indices. J. Pharm. Sci. 2008, 97,

1946-1976.

81. Collett, A.; Tanianis-Hughes, J.; Hallifax, D.; Warhurst, G. Predicting P- glycoprotein effects on oral absorption: correlation of transport in Caco-2 with drug pharmacokinetics in wild-type and mdr1a(-/-) mice in vivo. Pharm. Res. 2004, 21, 819-

826.

82. Corti, G.; Maestrelli, F.; Cirri, M.; Zerrouk, N.; Mura, P. Development and evaluation of an in vitro method for prediction of human drug absorption II.

Demonstration of the method suitability. Eur. J. Pharm. Sci. 2006, 27, 354-362.

83. Broccatelli, F.; Larregieu, C. A.; Cruciani, G.; Oprea, T. I.; Benet, L. Z.

Improving the Prediction of the Brain Disposition for Orally Administered Drugs Using

BDDCS. Adv. Drug Deliv. Rev. 2012, 64, 95-109.

84. Di Fenza, A.; Alagona, G.; Ghio, C.; Leonardi, R.; Giolitti, A.; Madami, A.

Caco-2 cell permeability modelling: a neural network coupled genetic algorithm approach. J. Comput. Aided Mol. Des. 2007, 21, 207-221.

85. Du-Cuny, L.; Song, Z.; Moses, S.; Powis, G.; Mash, E. A.; Meuillet, E. J.; Zhang,

S. Computational modeling of novel inhibitors targeting the Akt pleckstrin homology domain. Bioorg. Med. Chem. 2009, 17, 6983-6992.

86. Ekins, S.; Durst, G. L.; Stratford, R. E.; Thorner, D. A.; Lewis, R.; Loncharich, R.

J.; Wikel, J. H. Three-dimensional quantitative structure-permeability relationship

analysis for a series of inhibitors of rhinovirus replication. J. Chem. Inf. Comput. Sci.

2001, 41, 1578-1586.

87. Ertl, P.; Rohde, B.; Selzer, P. Fast calculation of molecular polar surface area as a sum of fragment-based contributions and its application to the prediction of drug transport properties. J. Med. Chem. 2000, 43, 3714-3717.

88. Foger, F.; Kopf, A.; Loretz, B.; Albrecht, K.; Bernkop-Schnurch, A. Correlation of in vitro and in vivo models for the oral absorption of peptide drugs. Amino Acids 2008,

35, 233-241.

89. Fossati, L.; Dechaume, R.; Hardillier, E.; Chevillon, D.; Prevost, C.; Bolze, S.;

Maubon, N. Use of simulated intestinal fluid for Caco-2 permeability assay of lipophilic drugs. Int. J. Pharm. 2008, 360, 148-155.

90. Fujiwara, S.; Yamashita, F.; Hashida, M. Prediction of Caco-2 cell permeability using a combination of MO-calculation and neural network. Int. J. Pharm. 2002, 237, 95-

105.

91. Goodwin, J. T.; Mao, B.; Vidmar, T. J.; Conradi, R. A.; Burton, P. S. Strategies toward predicting peptide cellular permeability from computed molecular descriptors. J.

Pept. Res. 1999, 53, 355-369.

92. Grice, J. E.; Cross, S. E.; Brownlie, C.; Roberts, M. S. The application of molecular structural predictors of intestinal absorption to screening of compounds for transdermal penetration. J. Pharm. Pharmacol. 2010, 62, 750-755.

93. Guangli, M.; Yiyu, C. Predicting Caco-2 permeability using support vector machine and chemistry development kit. J. Pharm. Pharm. Sci. 2006, 9, 210-221.

94. Hilgers, A. R.; Smith, D. P.; Biermacher, J. J.; Day, J. S.; Jensen, J. L.; Sims, S.

M.; Adams, W. J.; Friis, J. M.; Palandra, J.; Hosley, J. D.; Shobe, E. M.; Burton, P. S.

Predicting oral absorption of drugs: a case study with a novel class of antimicrobial agents. Pharm. Res. 2003, 20, 1149-1155.

95. Krarup, L. H.; Christensen, I. T.; Hovgaard, L.; Frokjaer, S. Predicting drug absorption from molecular surface properties based on molecular dynamics simulations.

Pharm. Res. 1998, 15, 972-978.

96. Kulkarni, A.; Han, Y.; Hopfinger, A. J. Predicting Caco-2 cell permeation coefficients of organic molecules using membrane-interaction QSAR analysis. J. Chem.

Inf. Comput. Sci. 2002, 42, 331-342.

97. Liang, E.; Chessic, K.; Yazdanian, M. Evaluation of an accelerated Caco-2 cell permeability model. J. Pharm. Sci. 2000, 89, 336-345.

98. Linnankoski, J.; Ranta, V. P.; Yliperttula, M.; Urtti, A. Passive oral drug absorption can be predicted more reliably by experimental than computational models-- fact or myth. Eur. J. Pharm. Sci. 2008, 34, 129-139.

99. Lundquist, S.; Renftel, M.; Brillault, J.; Fenart, L.; Cecchelli, R.; Dehouck, M. P.

Prediction of drug transport through the blood-brain barrier in vivo: a comparison between two in vitro cell models. Pharm. Res. 2002, 19, 976-981.

100. Marrero Ponce Y; Cabrera MA; Romero V; Ofori E; LA, M. Total and local quadratic indices of the ‘‘Molecular Pseudograph’s Atom Adjacency Matrix’’.

Application to prediction of Caco-2 permeability of drugs. Int J Mol Sci 2003, 4, 512–

536.

101. Marrero Ponce, Y.; Cabrera Pérez, M. A.; Romero Zaldivar, V.; Gonzalez Diaz,

H.; Torrens, F. A new topological descriptors based model for predicting intestinal epithelial transport of drugs in Caco-2 cell culture. J. Pharm. Pharm. Sci. 2004, 7, 186-

199.

102. Marrero Ponce, Y.; Cabrera Pérez, M. A.; Romero Zaldivar, V.; Bermejo Sanz,

M.; Siverio Mota, D.; Torrens, F. Prediction of intestinal epithelial transport of drug in

(Caco–2) cell culture from molecular structure using in silico approaches during early drug discovery. Internet Electron. J. Mol. Des. 2005, 4, 124–150.

103. Matsson, P.; Bergstrom, C. A.; Nagahara, N.; Tavelin, S.; Norinder, U.;

Artursson, P. Exploring the role of different drug transport routes in permeability screening. J. Med. Chem. 2005, 48, 604-613.

104. Nordqvist, A.; Nilsson, J.; Lindmark, T.; Eriksson, A.; Garberg, P.; Kihlen, M. A general model for prediction of Caco-2 cell permeability. QSAR Comb. Sci. 2004, 23,

303-310.

105. Norinder, U.; Osterberg, T.; Artursson, P. Theoretical calculation and prediction of Caco-2 cell permeability using MolSurf parametrization and PLS statistics. Pharm.

Res. 1997, 14, 1786-1791.

106. Norinder, U.; Osterberg, T. Theoretical calculation and prediction of drug transport processes using simple parameters and partial least squares projections to latent structures (PLS) statistics. The use of electrotopological state indices. J. Pharm. Sci.

2001, 90, 1076-1085.

107. Osterberg, T.; Norinder, U. Prediction of polar surface area and drug transport processes using simple parameters and PLS statistics. J. Chem. Inf. Comput. Sci. 2000,

40, 1408-1411.

108. Osterberg, T.; Norinder, U. Prediction of drug transport processes using simple parameters and PLS statistics. The use of ACD/logP and ACD/ChemSketch descriptors.

Eur. J. Pharm. Sci. 2001, 12, 327-337.

109. Paixao, P.; Gouveia, L. F.; Morais, J. A. Prediction of the in vitro permeability determined in Caco-2 cells by using artificial neural networks. Eur. J. Pharm. Sci. 2010,

41, 107-117.

110. Palm, K.; Luthman, K.; Ungell, A. L.; Strandlund, G.; Artursson, P. Correlation of drug absorption with molecular surface properties. J. Pharm. Sci. 1996, 85, 32-39.

111. Palm, K.; Luthman, K.; Ungell, A. L.; Strandlund, G.; Beigi, F.; Lundahl, P.;

Artursson, P. Evaluation of dynamic polar molecular surface area as predictor of drug absorption: comparison with other computational and experimental predictors. J. Med.

Chem. 1998, 41, 5382-5392.

112. Parrott, N.; Lave, T. Prediction of intestinal absorption: comparative assessment of GASTROPLUS and IDEA. Eur. J. Pharm. Sci. 2002, 17, 51-61.

113. Varma, M. V.; Gardner, I.; Steyn, S. J.; Nkansah, P.; Rotter, C. J.; Whitney-

Pickett, C.; Zhang, H.; Di, L.; Cram, M.; Fenner, K. S.; El-Kattan, A. F. pH-Dependent

Solubility and Permeability Criteria for Provisional Biopharmaceutics Classification

(BCS and BDDCS) in Early Drug Discovery. Mol. Pharmaceutics 2012, 9, 1199-1212.

114. Refsgaard, H. H.; Jensen, B. F.; Brockhoff, P. B.; Padkjaer, S. B.; Guldbrandt,

M.; Christensen, M. S. In silico prediction of membrane permeability from calculated molecular parameters. J. Med. Chem. 2005, 48, 805-811.

115. Benet, L. Z.; Amidon, G. L.; Barends, D. M.; Lennernas, H.; Polli, J. E.; Shah, V.

P.; Stavchansky, S. A.; Yu, L. X. The Use of BDDCS in Classifying the Permeability of

Marketed Drugs. Pharm. Res. 2008, 25, 483-488.

116. Skolnik, S.; Lin, X.; Wang, J.; Chen, X. H.; He, T.; Zhang, B. Towards prediction of in vivo intestinal absorption using a 96-well Caco-2 assay. J. Pharm. Sci.

2010, 99, 3246-3265.

117. Stenberg, P.; Norinder, U.; Luthman, K.; Artursson, P. Experimental and computational screening models for the prediction of intestinal drug absorption. J. Med.

Chem. 2001, 44, 1927-1937.

118. Varma, M. V.; Khandavilli, S.; Ashokraj, Y.; Jain, A.; Dhanikula, A.; Sood, A.;

Thomas, N. S.; Pillai, O.; Sharma, P.; Gandhi, R.; Agrawal, S.; Nair, V.; Panchagnula, R.

Biopharmaceutic Classification System: a Scientific Framework for Pharmacokinetic

Optimization in Drug Research. Curr. Drug Metab. 2004, 5, 375-388.

119. Tronde, A.; Norden, B.; Marchner, H.; Wendel, A. K.; Lennernas, H.; Bengtsson,

U. H. Pulmonary absorption rate and bioavailability of drugs in vivo in rats: structure- absorption relationships and physicochemical profiling of inhaled drugs. J. Pharm. Sci.

2003, 92, 1216-1233.

120. Zhu, C.; Jiang, L.; Chen, T. M.; Hwang, K. K. A Comparative Study of Artificial

Membrane Permeability Assay for High Throughput Profiling of Drug Absorption

Potential. Eur. J. Med. Chem. 2002, 37, 399-407.

121. Larregieu, C. A.; Benet, L. Z. Drug Discovery and Regulatory Considerations for

Improving In Silico and In Vitro Predictions that Use Caco-2 as a Surrogate for Human

Intestinal Permeability Measurements. AAPS J. 2013, 15, 483-497.

122. Kotze, A. F.; Thanou, M. M.; Luebetaen, H. L.; de Boer, A. G.; Verhoef, J. C.;

Junginger, H. E. Enhancement of paracellular drug transport with highly quaternized N- trimethyl chitosan chloride in neutral environments: in vitro evaluation in intestinal epithelial cells (Caco-2). J. Pharm. Sci. 1999, 88, 253-257.

123. Zheng, K.; Trivedi, M.; Siahaan, T. J. Structure and function of the intercellular junctions: barrier of paracellular drug delivery. Curr. Pharm. Des. 2006, 12, 2813-2824.

124. Umapathy, N. S.; Ganapathy, V.; Ganapathy, M. E. Transport of amino acid esters and the amino-acid-based prodrug valganciclovir by the amino acid transporter

ATB(0,+). Pharm. Res. 2004, 21, 1303-1310.

125. Hatanaka, T.; Haramura, M.; Fei, Y. J.; Miyauchi, S.; Bridges, C. C.; Ganapathy,

P. S.; Smith, S. B.; Ganapathy, V.; Ganapathy, M. E. Transport of amino acid-based prodrugs by the Na+- and Cl(-) -coupled amino acid transporter ATB0,+ and expression of the transporter in tissues amenable for drug delivery. J. Pharmacol. Exp. Ther. 2004,

308, 1138-1147.

126. Li, F.; Maag, H.; Alfredson, T. Prodrugs of nucleoside analogues for improved oral absorption and tissue targeting. J. Pharm. Sci. 2008, 97, 1109-1134.

127. Han, H.; de Vrueh, R. L.; Rhie, J. K.; Covitz, K. M.; Smith, P. L.; Lee, C. P.; Oh,

D. M.; Sadee, W.; Amidon, G. L. 5'-Amino acid esters of antiviral nucleosides, acyclovir, and AZT are absorbed by the intestinal PEPT1 peptide transporter. Pharm.

Res. 1998, 15, 1154-1159.

128. Han, H. K.; Amidon, G. L. Targeted prodrug design to optimize drug delivery.

AAPS Pharm. Sci. 2000, 2, E6.

129. Collett, A.; Sims, E.; Walker, D.; He, Y. L.; Ayrton, J.; Rowland, M.; Warhurst,

G. Comparison of HT29-18-C1 and Caco-2 cell lines as models for studying intestinal paracellular drug absorption. Pharm. Res. 1996, 13, 216-221.

130. Fagerholm, U.; Johansson, M.; Lennernas, H. Comparison between permeability coefficients in rat and human jejunum. Pharm. Res. 1996, 13, 1336-1342.

131. Simon, Z.; Katja, B.; Robert, R.; Bor, K.; Albin, K. Do the recommended standards for in vitro biopharmaceutic classification of drug permeability meet the

"passive transport" criterion for biowaivers? Curr. Drug Metab. 2012, 14, 21-27.

132. Dobson, P. D.; Kell, D. B. Carrier-mediated cellular uptake of pharmaceutical drugs: an exception or the rule? Nat. Rev. Drug Discov. 2008, 7, 205-220.

133. Giacomini, K. M.; Huang, S. M.; Tweedie, D. J.; Benet, L. Z.; Brouwer, K. L.;

Chu, X.; Dahlin, A.; Evers, R.; Fischer, V.; Hillgren, K. M.; Hoffmaster, K. A.; Ishikawa,

T.; Keppler, D.; Kim, R. B.; Lee, C. A.; Niemi, M.; Polli, J. W.; Sugiyama, Y.; Swaan, P.

W.; Ware, J. A.; Wright, S. H.; Yee, S. W.; Zamek-Gliszczynski, M. J.; Zhang, L.

Membrane transporters in drug development. Nat. Rev. Drug Discov. 2010, 9, 215-236.

134. US Department of Health and Human Services, Food and Drug Administration,

Critical Path Initiative Challenges and Opportunities Report. Innovation or Stagnation:

Challenge and Opportunity on the Critical Path to New Medical Products. US FDA website 2004, http://www.fda.gov/ScienceResearch/SpecialTopics/CriticalPathInitiative/CriticalPathOp portunitiesReports/ucm077262.htm.

Chapter 3: Distinguishing between the Permeability Relationships with Drug Absorption and Metabolism to Improve BCS and BDDCS Predictions

3.1 Introduction

Many promising drug candidates fail during drug discovery and development due to unacceptable toxicity and inefficacies caused by unfavorable absorption, distribution, metabolism, and excretion (ADME) properties.1 It is exceedingly desirable for these compounds to be disqualified early in the drug discovery phase when they are new molecular entities (NMEs), rather than later during the much more costly drug development phases. Hence, it is of vital importance to implement strategies that identify toxicity and ADME properties of NMEs during early screening for candidate prioritization and elimination, thereby benefiting the drug discovery and development process immensely.

Ever since its inception in 1995, the Biopharmaceutics Classification System

(BCS) has been an invaluable tool for predicting intestinal drug absorption following oral administration. The BCS framework classifies compounds into four groups according to their aqueous solubility and their intestinal permeability: Class I (high solubility, high permeability), Class II (low solubility, high permeability), Class III (high solubility, low permeability), and Class IV (low solubility, low permeability). Amidon and co-workers2 found a strong correlation between human jejunal permeability rate (Peff) measures determined from intestinal perfusion studies in humans and the fraction of dose absorbed obtained from pharmacokinetic or mass balance studies in humans. They observed that a drug substance had a high intestinal permeability rate when the extent of absorption (Fa)

67 was ≥90% of the oral dose.

Based on these findings, the U.S. Food and Drug Administration (FDA) implemented a BCS Guidance3 that supports waivers of bioequivalence clinical studies of highly permeable, highly soluble BCS Class 1 drugs. The FDA BCS Guidance3 describes several permeability methodologies for determining BCS classification and demonstrating bioequivalence. The BCS class can be determined by measuring human effective permeability rates (Peff) across the jejunal membrane, or alternatively, apparent permeability rates (Papp) across in vitro epithelial cell monolayers, such as the human intestinal cell line Caco-2, relative to a high-permeability reference standard (e.g. metoprolol). Only limited jejunal permeability studies in humans have been conducted with published information available for about 30 drugs, likely due to the complexity and high costs of each procedure.4 Subsequently, the most popular human intestinal permeability screening method has used in vitro Caco-2 cell monolayers.5 Even though the BCS is beneficial for obtaining waivers for bioequivalence studies in humans, it can also be very useful in predicting the absorption of NMEs for candidate lead selection in early drug discovery programs.6

The Biopharmaceutics Drug Disposition Classification System (BDDCS) is advantageous to early drug discovery programs for predicting an NME’s drug disposition characteristics and potentially clinically significant drug interactions that may arise in the intestine, liver, and brain.7-9 The BDDCS framework is a modification of the BCS that utilizes drug metabolism (by Phase 1 oxidative and Phase 2 conjugative processes) rather than intestinal permeability.7 As such, it classifies compounds into four groups according to their aqueous solubility and their extent of metabolism: Class I (high solubility,

68 extensive metabolism), Class II (low solubility, extensive metabolism), Class III (high solubility, poor metabolism), and Class IV (low solubility, poor metabolism). Benet and colleagues10 recognized that orally administered drugs that were ≥90% metabolized by

Phase 1 and Phase 2 processes had to be ≥90% absorbed. Thus, they recommended that the extent of drug metabolism can be an alternative method for supporting a biowaiver for BCS Class 1 drugs.

As a result of the good correlation between high intestinal permeability rates, high intestinal absorption, and extensive metabolism, the permeability methods for determining BCS classification have also been used for BDDCS classification.11-14

However, discrepancies between BCS and BDDCS classification among the various permeability methods have been observed whereby assignment by one method was in accordance with BDDCS but not with BCS and vice versa.11-17 Of 14 drugs with reported high intestinal permeability extent and poor metabolism,13 published Caco-2 permeability rate measurements could be found for 11 drugs. Of these 11 drugs, 8 (73%) were poorly permeable based on their in vitro data, thereby agreeing with their BDDCS classification and disagreeing with their BCS classification.15, 16 The Caco-2 cellular system is known to be deficient in carrier-mediated and paracellular mechanisms.18 Therefore, I hypothesize that the extent of drug metabolism (BDDCS-permeability) is particularly correlated with passive transcellular permeability rate, while the extent of drug absorption

(BCS-permeability) is correlated with complete human intestinal permeability. To test this hypothesis, I evaluated correlations of BCS and BDDCS with permeability rate data from studies across the human jejunum in vivo, Caco-2 cell monolayers, and the parallel artificial membrane permeation assay (PAMPA) membrane, which is devoid of carrier-

69 mediated and paracellular processes.19 Consequently, recommendations on the most appropriate permeability models for accurately predicting BCS and BDDCS classifications in early drug discovery programs are presented.

3.2 Methods

3.2.1 Compilation of Permeability Rate Datasets

A literature search was performed to compile all the available human intestinal

(jejunal) permeability rate measures of drugs from published human studies (Table 3.1).

Metoprolol was chosen as the high-permeability reference standard cutoff as it has been previously used by the BCS in the same role.2 Thus, drugs exhibiting permeability rates greater than or equal to the corresponding value of metoprolol are considered high- permeability rate drugs. Conversely, drugs with permeability rates less than the corresponding value of metoprolol are classified as low-permeability rate drugs.

Another literature search was performed to compile multiple datasets of in vitro permeability rate measures. It has been previously shown how using Caco-2 permeability rate values measured from different laboratories can skew correlations.18 So in order to avoid interlaboratory variation and selection bias, each dataset contained Caco-2 permeability rate values of at least 23 drugs, including metoprolol, that were measured by the same laboratory (Table 3.2). Upon extensive evaluation of the literature, permeability rate datasets could be found for multiple variations of the PAMPA model. For this study, permeability rate datasets of measurements for at least 35 drugs, including metoprolol,

Table 3.1 Comparison of the human in vivo effective permeability rate coefficients (Peff) of 30 drugs with their extents of absorption and metabolism.

Peff Extent of Extent of Drug (x 10-4 cm/s) Absorption (%) Metabolism (%)

Ketoprofen 8.70 100 90 Naproxen 8.50 100 95 Verapamil 6.80 100 96 Piroxicam 6.65 100 95 Antipyrine 5.60 100 95 Desipramine 4.50 100 98 Carbamazepine 4.30 100 98 Levodopa 3.40 100 95 Propranolol 2.91 100 99 Fluvastatin 2.40 100 98 Valacyclovir* 1.66* Cyclosporine 1.61 100 99 Amiloride 1.60 90 0 Enalapril maleate* 1.57* Cephalexin 1.56 90 10 Metoprolol 1.34 95 95 Losartan 1.15 100 97 Isotretinoin 0.99 95 95 Lisinopril 0.33 35 0 Terbutaline 0.30 40 30 Amoxicillin 0.30 67 30 Ranitidine 0.27 50 30 Cimetidine 0.26 75 25 Enalaprilat 0.20 8 10 Atenolol 0.20 62 10 α-Methyldopa 0.10 65 50 Fexofenadine 0.07 10 5 Furosemide 0.05 40 10 Hydrochlorothiazide 0.04 60 0 Inogatran 0.03 10 0 *Excluded from analyses since these are prodrugs whose extents of absorption and permeability rate measures were made for the active species rather than the dosed prodrug.

-6 Table 3.2 Comparison of the Caco-2 permeability rate coefficients, Papp (x 10 cm/s), from 4 different laboratories with their extents of absorption and metabolism.

Alsenz Extent of Extent of Irvine Li Yazdanian and Drug Absorption Metabolism et al24 et al25 et al26 Haenel23 (%) (%) Papp Papp Papp Papp

Acebutolol 90 88 0.51 Acetaminophen 95 95 100 Acetylsalicylic Acid 85 85 2.2 9.09 Acyclovir 30 25 0 1.3 0.25 Alprenolol 100 99 170 36.6 25.3 Amiloride 90 0 1.13 Amoxicillin 67 30 0.01 0.021 Antipyrine 100 95 54.3 150 35.7 Atenolol 62 10 1.73 3.3 1.6 0.53 Benserazide 100 100 1.85 Bupropion 97 99 150 Caffeine 100 99 34.5 30.8 Carbamazepine 100 98 62.23 42.5 Cefuroxime 5 1 0.38 Cephalexin 90 10 0.27 Chlorothiazide 20 4 0.32 0.19 Chlorpromazine 100 99 19.9 Cimetidine 75 25 0.59 2.7 1.37 Desipramine 100 98 43 24.4 Dexamethasone 97 97 40 12.5 12.2 Diazepam 98 98 33.4 Enalaprilat 8 10 1.85 Estradiol 100 92 16.9 Etoposide 75 55 1.5 Furosemide 40 10 0.31 0.14 1.3 Ganciclovir 50 10 0.38 Griseofulvin 99 99 36.6 Hydrochlorothiazide 60 0 0.42 1.5 0.51 Hydrocortisone 99 99 56 20.7 14 Indomethacin 90 90 20.4 Ketoprofen 100 90 24.36 93 34.7

-6 Table 3.2 Comparison of the Caco-2 permeability rate coefficients, Papp (x 10 cm/s), from 4 different laboratories with their extents of absorption and metabolism. (continued)

Alsenz Extent of Extent of Irvine Li Yazdanian and Drug Absorption Metabolism et al24 et al25 et al26 Haenel23 (%) (%) Papp Papp Papp Papp

Labetolol 100 95 76 9.31 Lamotrigine 100 90 110 Levodopa 100 95 0.01 Lisinopril 35 0 1.27 0.22 Loracarbef 90 0 0.24 Meloxicam 100 99 19.5 Methyldopa 65 50 0.15 Methylprednisolone 98 98 25 14.6 Metoprolol 95 95 31.77 140 33.2 23.7 Nadolol 30 0 0.39 0.6 3.88 Naproxen 100 95 53.07 33.8 Nevirapine 97 97 30.1 Nicotine 90 90 19.4 Ondansetron 95 95 110 Penicillin V 60 10 1.9 0.17 Phenytoin 95 95 160 26.7 Pirenzepine 30 0 0.44 Piroxicam 100 95 28.85 35.6 Progesterone 95 95 98 23.7 Propranolol 100 99 47.2 110 39.4 21.8 Ranitidine 50 30 0.67 0 2.1 0.49 Salicylic Acid 90 90 13 22 Scopolamine 100 99 11.8 Sotalol 95 15 4.2 Sumatriptan 80 80 0 Terbutaline 40 30 1.71 0.41 0.8 0.47 Testosterone 90 90 100 24.9 Theophylline 98 90 22.6 Timolol 90 90 100 12.8 Verapamil 100 96 44.67 45.7 Warfarin 100 99 96 21.1 Zidovudine 95 86 28 6.93

73 were selected from 4 different PAMPA models: (i) traditional,20 (ii) a lipid/oil/lipid tri- layer,20 (iii) a bio-mimetic layer,21 and (iv) a hydrophilic filter membrane PAMPA assay22 (Table 3.3).

3.2.2 Correlations of BCS and BDDCS with Drug Permeability Rate Measures

To facilitate a comparison between BCS and BDDCS classifications using drug permeability rate data, information on the extents of absorption and metabolism measures were compiled from standard references27-30 and verified with their original references

(Table 3.4). Drugs for which absorption data could not be found, such as those administered non-orally, were disqualified and excluded from the study. Each qualifying drug was evaluated for its permeability rate in relation to its extent of absorption and its extent of metabolism. Here, a drug is classified as having a high extent of absorption or metabolism in humans when the extent is ≥90%, and a low extent of absorption or metabolism when the extent is <90%. Comparisons were then made between the class based on permeability rate, versus the classes based on the extent of absorption and the extent of metabolism. Non-linear regression analyses were performed using GraphPad

Prism software version 4.03 (GraphPad Software, Inc., San Diego, CA) for the datasets of human absorption or metabolism with their respective permeability rate measures. Drugs that exhibit permeability rates greater than that of metoprolol, but are <90% absorbed or

<90% metabolized in humans, are termed false positives. False negatives, on the other hand, are drugs that are ≥90% absorbed or ≥90% metabolized, but have experimental human intestinal permeability rates that are lower than that of metoprolol.

-6 Table 3.3 Comparison of the PAMPA permeability rate coefficients, Papp (x 10 cm/s), from 4 different PAMPA models with their extents of absorption and metabolism.

Trad Lipid/Oil Extent of Extent of Biomimetic Hydrophilic Model /Lipid Drug Absorption Metabolism Model21 Model22 20 Model20 (%) (%) Papp Papp Papp Papp

Acebutolol 90 88 0.03 0.21 2.95 3.3 Acetaminophen 95 95 2.39 3.5 Acetylsalicylic Acid 85 85 3.8 Acyclovir 30 25 0.04 0.1 0.04 0 Allopurinol 90 70 0.51 Alprenolol 100 99 11.81 9.71 11.1 15.1 Amiloride 90 0 0 0.08 0.93 0 Amoxicillin 67 30 1.5 Antipyrine 100 95 0.74 7.51 9.12 13.2 Atenolol 62 10 0.06 0.1 0.56 0 Bromocriptine 95 94 0.99 0 Bumetanide 95 35 0.3 Bupropion 97 99 12.1 14.1 Caffeine 100 99 1.2 9.89 10.8 Captopril 75 55 19.1 Carbamazepine 100 98 6.4 9.44 11.3 Cefadroxil 90 10 1 Cefuroxime 5 1 0.07 Cephalexin 90 10 0.1 Chloramphenicol 100 90 22.1 1.7 Chlorothiazide 20 4 0.31 1.3 Chlorpromazine 100 99 4 Cimetidine 75 25 0.86 0 Ciprofloxacin 60 60 7.39 Clofibrate 95 90 0.3 Cloxacillin 75 60 1.92 Clozapine 95 97 28.1 Cyclosporine 100 99 0.3 Desipramine 100 98 16.59 8.67 14.6 Dexamethasone 97 97 33.9 8.1 Diclofenac 99 99 1.37 6.95 53.3 12.5 Dicloxacillin 76 10 4.16 Diltiazem 99 96 17.09 10.61 39.3 18.5

-6 Table 3.3 Comparison of the PAMPA permeability rate coefficients, Papp (x 10 cm/s), from 4 different PAMPA models with their extents of absorption and metabolism. (continued) Trad Lipid/Oil Extent of Extent of Biomimetic Hydrophilic Model /Lipid Drug Absorption Metabolism Model21 Model22 20 Model20 (%) (%) Papp Papp Papp Papp

Ethambutol 80 20 0.53 Famotidine 60 45 0.06 0.04 0.05 Flecainide 90 75 9.44 Flucytosine 90 1 0.18 Fluoxetine 95 90 14.1 Furosemide 40 10 0.01 0.46 3.6 0.6 Gabapentin 60 2 1.2 Griseofulvin 99 99 5.3 Guanabenz 99 99 8.52 17.5 Hydrochlorothiazide 60 0 0.02 0.09 1.71 0 Hydrocortisone 99 99 18.8 3.4 Ibuprofen 99 99 2.4 10.73 6.8 Imipramine 96 96 14.04 10.11 42.1 Indomethacin 90 90 0.3 6.24 2.4 Ketoprofen 100 90 0.05 4.13 33.8 16.7 Ketorolac 100 40 1.4 Labetalol 100 95 4.5 Lansoprazole 100 99 3.39 Melphalan 93 90 5.7 Metformin 60 0 0.27 Methotrexate 90 20 0.1 Methylprednisolone 98 98 30.7 5.9 Metoprolol 95 95 0.41 4.29 5.67 3.5 Nadolol 30 0 0.28 0.16 0.8 0 Naltrexone 96 98 11.3 Naproxen 100 95 0.34 6.03 49.5 10.6 Nicotine 90 90 10.47 4.28 21.2 Norfloxacin 40 30 5.43 0.9 Penicillin V 60 10 0.1 Phenytoin 95 95 0.38 5.73 5.1 Piroxicam 100 95 2.64 4.96 8.2 Pravastatin 34 50 1.52

-6 Table 3.3 Comparison of the PAMPA permeability rate coefficients, Papp (x 10 cm/s), from 4 different PAMPA models with their extents of absorption and metabolism (continued). Trad Lipid/Oil Extent of Extent of Biomimetic Hydrophilic Model /Lipid Drug Absorption Metabolism Model21 Model22 20 Model20 (%) (%) Papp Papp Papp Papp

Prazosin 95 94 18.2 13.5 Prednisolone 99 99 5.7 Probenecid 100 99 2.4 Procainamide 95 50 6.97 Progesterone 95 95 4 Propranolol 100 99 14.3 8.64 13.7 23.5 Propylthiouracil 95 90 10.4 Quinidine 100 90 10.9 Ranitidine 50 30 0.01 0.45 1.63 0.5 Salicylic Acid 90 90 3.3 Sotalol 95 15 0.3 1.1 Sulindac 90 90 9.2 Sulpiride 35 0 0.03 0.18 1.3 0.1 Sumatriptan 80 80 2.36 Terazosin 90 90 8.8 Terbutaline 40 30 0.05 0.46 0.23 0.1 Tetracycline 60 5 4.7 Theophylline 98 90 0.04 3.53 2.78 4.8 Timolol 90 90 0.61 4.45 10.2 5.1 Tolbutamide 96 80 51.5 Valsartan 25 10 2.1 Verapamil 100 96 39.4 8.75 38.4 7.4 Warfarin 100 99 1.58 4.96 72 12.3 Zidovudine 95 86 3.65 4.9 Zopiclone 96 96 8.9

Table 3.4 Comparison of all analyzed drugs with their measures of lipophilicity.

Extent of Extent of Measured Measured Drug cLogP Absorption Metabolism LogP LogD 7.4 (%) (%)

Acebutolol 1.71 0.19 1.71 90 88 Acetaminophen 0.20 0.40 0.49 95 95 Acetylsalicylic Acid 1.19 -2.57 1.02 85 85 Acyclovir -1.56 -1.76 -2.42 30 25 Allopurinol -0.55 -0.55 0.63 90 70 Alprenolol 3.10 1.34 2.65 100 99 Amiloride 0.10 -1.25 0.11 90 0 Amoxicillin 0.87 -1.52 -1.87 67 30 Antipyrine 0.28 0.20 100 95

Atenolol 0.16 -1.03 -0.11 62 10 Benserazide -2.90 100 100

Bromocriptine 4.59 6.58 95 94

Bumetanide -0.45 3.37 95 35

Bupropion 3.27 3.21 97 99

Caffeine -0.07 -0.07 -0.04 100 99 Captopril 0.34 -1.98 0.89 75 55 Carbamazepine 2.45 2.45 2.38 100 98 Cefadroxil -3.40 -2.51 90 10

Cefuroxime -0.16 -1.91 0.23 5 1 Cephalexin -0.67 -2.40 -1.84 90 10 Chloramphenicol 1.14 1.00 1.28 100 90 Chlorothiazide -0.24 -1.10 -1.00 20 4 Chlorpromazine 5.41 2.82 5.30 100 99 Cimetidine 0.40 0.33 0.19 75 25 Ciprofloxacin 0.28 -1.21 -0.73 60 60 Clofibrate 3.60 3.60 3.02 95 90 Cloxacillin 2.48 -1.82 2.52 75 60 Clozapine 3.23 2.99 3.71 95 97 Cyclosporine 2.95 2.92 14.36 100 99 Desipramine 4.90 1.40 4.47 100 98 Dexamethasone 1.94 1.83 1.79 97 97 Diazepam 2.82 2.99 2.96 98 98 Diclofenac 4.51 1.13 4.73 99 99 Dicloxacillin 2.91 2.98 76 10 Diltiazem 2.70 2.22 3.65 99 96 Enalaprilat -0.74 -0.28 0.88 8 10 Estradiol 4.01 3.78 100 92

Table 3.4 Comparison of all analyzed drugs with their measures of lipophilicity. (continued)

Extent of Extent of Measured Measured Drug cLogP Absorption Metabolism LogP LogD 7.4 (%) (%)

Ethambutol -2.79 0.12 80 20

Etoposide 0.60 0.60 0.03 75 55 Famotidine -0.64 -1.50 -1.17 60 45 Fexofenadine 2.68 1.96 10 5 Flecainide 3.78 1.14 3.66 90 75 Flucytosine -1.97 -1.64 90 1

Fluoxetine 4.05 1.95 4.57 95 90 Fluvastatin 3.80 4.05 100 98 Furosemide 2.03 -1.54 1.90 40 10 Gabapentin -1.10 -1.31 -0.66 60 2 Ganciclovir -1.66 -4.25 -2.73 50 10 Griseofulvin 2.18 2.18 1.91 99 99 Guanabenz 0.12 2.98 99 99

Hydrochlorothiazide -0.07 -0.07 -0.37 60 0 Hydrocortisone 1.61 1.37 1.70 99 99 Ibuprofen 3.97 0.81 3.68 99 99 Imipramine 4.80 2.20 5.04 96 96 Indomethacin 4.27 0.77 4.18 90 90 Isotretinoin 6.30 4.23 6.74 95 95 Ketoprofen 3.12 -0.01 2.76 100 90 Ketorolac 1.04 1.62 100 40 Labetolol 100 95

Lamotrigine -0.19 2.53 100 90

Lansoprazole 2.36 2.60 100 99

Levodopa -2.74 -2.39 -2.82 100 95 Lisinopril -1.22 -3.40 -1.69 35 0 Loracarbef -0.47 90 0

Losartan 4.10 100 97

Meloxicam 3.02 0.10 2.29 100 99 Melphalan -0.11 -0.21 93 90

Metformin -5.41 -1.63 60 0

Methotrexate -1.85 -2.52 -0.53 90 20 Methyldopa 0.39 -2.39 -2.26 65 50 Methylprednisolone 2.18 1.74 98 98

Metoprolol 1.88 0.16 1.49 95 95 Nadolol 0.81 0.93 0.38 30 0

Table 3.4 Comparison of all analyzed drugs with their measures of lipophilicity. (continued)

Extent of Extent of Measured Measured Drug cLogP Absorption Metabolism LogP LogD7.4 (%) (%)

Naltrexone 1.92 0.90 0.36 96 98 Naproxen 3.18 1.70 2.82 100 95 Nevirapine 1.81 2.65 97 97 Nicotine 1.17 0.43 0.88 90 90 Norfloxacin -1.03 -2.00 -0.78 40 30 Ondansetron 2.12 2.72 95 95

Penicillin V 2.09 -1.54 1.94 60 10 Phenytoin 2.47 2.47 2.09 95 95 Pirenzepine 0.10 -0.39 -0.35 30 0 Piroxicam 3.06 0.20 1.89 100 95 Pravastatin 2.18 -0.23 2.05 34 50 Prednisolone 1.62 1.62 1.42 99 99 Probenecid 3.21 -0.26 3.37 100 99 Procainamide 0.88 -1.15 1.42 95 50 Progesterone 3.87 3.78 95 95 Propranolol 3.48 1.20 2.75 100 99 Propylthiouracil 1.09 0.97 95 90

Quinidine 3.44 1.82 2.79 100 90 Ranitidine 0.27 0.54 0.67 50 30 Salicylic Acid 2.26 -1.51 2.19 90 90 Scopolamine 0.98 0.62 0.29 100 99 Sotalol 0.24 -0.79 0.23 95 15 Sulindac 3.42 -0.66 3.16 90 90 Sulpiride 0.57 -1.15 1.11 35 0 Sumatriptan 0.93 -1.17 0.74 80 80 Terbutaline 0.90 0.48 40 30 Testosterone 3.32 3.22 90 90 Tetracycline -1.30 -1.41 -0.91 60 5 Theophylline -0.04 -0.02 -0.03 98 90 Timolol 1.83 1.91 1.21 90 90 Tolbutamide 2.34 2.52 2.50 96 80 Valsartan 4.86 25 10

Verapamil 3.79 4.47 100 96 Warfarin 2.60 1.12 2.90 100 99 Zidovudine 0.05 0.08 0.04 95 86 Zopiclone 1.50 -0.90 1.25 96 96

3.3 Results

3.3.1 Comparison of BCS and BDDCS Classifications Using Human Intestinal Permeability Rate Measures

The available human intestinal (jejunal) permeability rates for 30 drugs in the literature4 are summarized in Table 3.1, with estimates of their extent of absorption and extent of metabolism values in Table 3.4. Enalapril and valacyclovir are prodrugs whose extents of absorption and permeability rate measures were made for the active species rather than the dosed prodrug. As a result, these drugs were excluded from these analyses. The list contains 14 high-permeability rate drugs (that have permeability rates greater than or equal to the corresponding value for metoprolol) and 14 low-permeability rate drugs (that have permeability rates less than the corresponding value for metoprolol).

Figure 3.1 allows for a side-by-side comparison of BCS and BDDCS for the 28 drugs with published human intestinal permeability rate measures. Figure 3.1(a) depicts a graph of the human intestinal permeability rates versus the extent of absorption, while Figure

3.1(b) shows these permeability numbers versus the extent of metabolism.

When using the BCS criterion of an extent of absorption ≥90%, the human intestinal permeability rates accurately predicted the extent of absorption for 26 out of 28

(93%) drugs. The ≥90% extent of absorption criterion accurately correlated with all of the high-permeability rate drugs, and the <90% extent of absorption criterion accurately correlated with 12 out of 14 (86%) low-permeability rate drugs. Isotretinoin and losartan are 2 false negatives. Both drugs are considered low-permeability rate drugs since their human intestinal permeability rates are lower than that of metoprolol. However, the

Figure 3.1 Relationships between the extent of absorption, extent of metabolism, and human intestinal permeability rates for 28 drugs. (a) Correlation plot of the extent of absorption with the human jejunal permeability rate. (b) Correlation plot of the extent of metabolism with the human intestinal permeability rate. (c) Correlation plot of the extent of metabolism with the extent of absorption.

(a)

(b)

(c)

83 complete absorption of losartan and 90% absorption of isotretinoin would meet the ≥90%

BCS absorption criterion.

When using the BDDCS criterion of an extent of metabolism ≥90%, the human intestinal permeability rates accurately predicted the extent of metabolism for 24 out of

28 (86%) drugs. The ≥90% extent of metabolism criterion accurately correlated with 12 out of 14 (86%) high-permeability rate drugs, and the <90% extent of metabolism criterion accurately correlated with 12 out of 14 (86%) low-permeability rate drugs.

Isotretinoin and losartan are the same false negatives found in both the BCS absorption and the BDDCS metabolism predictions. Both drugs are considered low-permeability rate drugs since their human intestinal permeability rates are lower than that of metoprolol.

However, their low permeability rates would incorrectly predict their extent of metabolism measures to be low. The BDDCS metabolism predictions resulted in 2 false positives. Amiloride and cephalexin are both poorly metabolized, even though they exhibit high human intestinal permeability rates relative to metoprolol. Their high- permeability rates accurately predicted their high extents of absorption. It should be noted that when a drug is both ≥90% absorbed and ≥90% metabolized, it always has a high intestinal permeability rate.

The false negatives and false positives from the BCS and BDDCS predictions can be seen in Figures 3.1(a) and 3.1(b). The two plots are very similar in appearance. From

Figure 3.1(b), one can see few drugs within the 30% to 70% extent of metabolism range, a general conclusion noted by Wu and Benet.7 However, as seen in Figure 3.1(a), a number of drugs fall in the 30% to 70% absorption range. Figure 3.1(c) depicts the relationship between the extent of metabolism and the extent of absorption for the 30

84 drugs listed in Table 3.1. The correlation coefficient (r2) for the correlation in Figure

3.1(c) is 0.630. If the 10 compounds with 100% absorption are omitted, the r2 for the remaining 18 compounds decreases to 0.390. Hence, there is generally a poor correlation between the extent of absorption and the extent of metabolism. However, drugs that are

≥90% metabolized accurately predict an extent of absorption ≥90%.

3.3.2 Comparison of BCS and BDDCS Classifications Using Caco-2 Permeability Rate Measures

Table 3.2 shows a summary of 4 datasets each comprising at least 23 drugs, including metoprolol, and their Caco-2 permeability rate measures, taken from 4 different laboratories.23-26 Figure 3.2 depicts the correlation plots of the Caco-2 permeability rate measures of each dataset in relation to their extents of absorption and extents of metabolism. For each dataset, the r2 of the correlation of Caco-2 permeability rate is higher with the extent of metabolism than with the extent of absorption: 0.490 in Figure

3.2(b) vs. -0.693 in 3.2(a), 0.738 in Figure 3.2(d) vs. 0.0757 in 3.2(c), 0.918 in Figure

3.2(f) vs. 0.912 in 3.2(e), and 0.764 in Figure 3.2(h) vs. 0.702 in 3.2(g).

When using the BCS criterion of an extent of absorption ≥90%, the Caco-2 permeability rates accurately predicted the extent of absorption for a mean of 68  14% of the drugs. The ≥90% extent of absorption criterion accurately correlated with all of the high-permeability rate drugs, and the <90% extent of absorption criterion accurately correlated with a mean of 56  15% of the low-permeability rate drugs. When using the

BDDCS criterion of an extent of metabolism ≥90%, the Caco-2 permeability rates accurately predicted the extent of metabolism for a mean of 75  9% of the drugs. The

Figure 3.2 Comparison of BCS and BDDCS classifications with Caco-2 permeability rate measures. Correlation plots of the extents of absorption (a) and metabolism (b) with Caco-2 permeability rate data from Alsenz and Haenel.23 Correlation plots of the extents of absorption (c) and metabolism (d) with Caco-2 permeability rate data from Irvine et al.24 Correlation plots of the extents of absorption (e) and metabolism (f) with Caco-2 permeability rate data from Li et al.25 Correlation plots of the extents of absorption (g) and metabolism (h) with Caco-2 permeability rate data from Yazdanian et al.26 Solid line represents the best fit of a non-linear regression.

(a)

(b)

(c)

(d)

(e)

(f)

(g)

(h)

≥90% extent of metabolism criterion accurately correlated with all of the high- permeability rate drugs, and the <90% extent of absorption criterion accurately correlated with a mean of 65  9% of the low-permeability rate drugs. In order to evaluate if the high number of false positives in the BCS and BDDCS predictions was due to metoprolol being too strict of a high-permeability reference standard, BCS and

BDDCS predictions were also conducted using a high permeability cut-off that was 30% lower than metoprolol’s permeability rate. Using this adjusted cut-off, the predictions improved with the <90% extent of absorption criterion accurately correlated with a mean of 73  6% of the low-permeability rate drugs and the <90% extent of metabolism criterion accurately correlated with a mean of 78  8% of the low-permeability rate drugs. There were no false negatives found in the BCS absorption and BDDCS metabolism predictions. Hence, all high-permeability rate drugs (that have permeability rates greater than or equal to the corresponding value for metoprolol) in Caco-2 accurately predict a ≥90% extent of absorption and metabolism in humans.

3.3.3 Comparison of BCS and BDDCS Classifications Using PAMPA Permeability Rate Measures

The permeability rate measures for at least 35 drugs, including metoprolol, measured by 4 different PAMPA models: (i) traditional,20 (ii) a lipid/oil/lipid tri-layer,20

(iii) a bio-mimetic layer,21 and (iv) a hydrophilic filter membrane PAMPA assay,22 are summarized in Table 3.3. Figure 3.3 displays the correlation plots of the permeability rate measures for each PAMPA model in relation to their extents of absorption and extents of metabolism. In all 4 models, the PAMPA permeability rate measures had a stronger

91 correlation with the extent of metabolism than with the extent of absorption: 0.549 in

Figure 3.3(b) vs. -0.0580 in 3.3(a), 0.823 in Figure 3.3(d) vs. 0.262 in 3.3(c), 0.552 in

Figure 3.3(f) vs. 0.210 in 3.3(e), and 0.559 in Figure 3.3(h) vs. -0.664 in 3.3(g).

Out of the 4 models, the BCS absorption and BDDCS metabolism criteria were best predicted via the lipid/oil/lipid tri-layer model.20 When using the BCS criterion of an extent of absorption ≥90%, the PAMPA permeability rates in this model accurately predicted the extent of absorption for 84% of the drugs. The ≥90% extent of absorption criterion accurately correlated with all of the high-permeability rate drugs, and the <90% extent of absorption criterion accurately correlated with 64% of the low-permeability rate drugs. When using the BDDCS criterion of an extent of metabolism ≥90%, the PAMPA permeability rates in this model accurately predicted the extent of metabolism for 88% of the drugs. The ≥90% extent of metabolism criterion accurately correlated with all of the high-permeability rate drugs, and the <90% extent of metabolism criterion accurately correlated with 71% of the low-permeability rate drugs. The number of false positives in

BCS absorption predictions with the PAMPA permeability rate data was higher relative to the BDDCS metabolism predictions (5 compared to 4). Because some of the false positives are due to metoprolol being too strict of a high-permeability rate reference standard, a high-permeability rate cut-off value that was 30% lower than that of metoprolol’s was estimated. Using this adjusted high-permeability rate measure, there was 1 false positive in the BCS prediction and no false positives in the BDDCS prediction. There were no false negatives found in the BCS absorption and BDDCS metabolism predictions for this model. Hence, using this PAMPA model, all high permeability rate drugs (that have permeability rates greater than or equal to that of

Figure 3.3 Comparison of BCS and BDDCS classifications with PAMPA permeability rate measures. Correlation plots of the extent of absorption (a) and metabolism (b) with PAMPA permeability rate data from a traditional model.20 Correlation plots of the extents of absorption (c) and metabolism (d) with PAMPA permeability rate data from a lipid/oil/lipid tri-layer model.20 Correlation plots of the extents of absorption (e) and metabolism (f) with PAMPA permeability rate data from a bio-mimetic layer model.21 Correlation plots of the extents of absorption (g) and metabolism (h) with PAMPA permeability rate data from a hydrophilic filter membrane assay.22 Solid line represents best fit of a non-linear regression.

(a)

(b)

(c)

(d)

(e)

(f)

(g)

(h)

97 metoprolol) accurately predict a ≥90% extent of absorption and metabolism in humans.

3.4 Discussion

3.4.1 Use of BDDCS in the BCS FDA Guidance for Industry

For this study, I revisited the original human intestinal (jejunal) permeability rate measures that were used to illustrate a correlation with human intestinal absorption, and demonstrate for the first time a correlation with human drug metabolism. The intestinal permeability rate accurately predicted the BCS absorption (high versus low) for 93% of the 28 drugs, all of the high-permeability rate drugs and 86% of the low-permeability rate drugs. The intestinal permeability rate accurately predicted the BDDCS metabolism (high versus low) for 86% of the 28 drugs, 86% of the high-permeability rate drugs and 86% of the low-permeability rate drugs. It is evident that when both criteria are used, a high intestinal permeability rate is always accurately predicted.

For such simple categorizations in BCS and BDDCS, the extent of absorption and the extent of metabolism both do a remarkable job in predicting the human intestinal permeability rates of drugs. Both criteria result in similar outcomes when predicting high intestinal permeability rates. They both share two false negatives, losartan and isotretinoin. Here, it may be that metoprolol is too strict a permeability standard with a

95% extent of absorption, and a compound with a 90% extent of absorption may be more suitable as the cut-off. Labetalol has been proposed as an alternative,31 but it may not be ideal since its permeability has been recently shown to be concentration-dependent.32

Thus, work should be underway to find a more suitable alternative permeability reference marker to metoprolol that is about 90% absorbed primarily by passive diffusion.

Amiloride and cephalexin are false positives of the BDDCS metabolism prediction, where the BDDCS metabolism could not accurately predict amiloride’s and cephalexin’s high intestinal permeability rates, since neither are >10% metabolized.33, 34

Both of their permeabilities across the intestinal membrane have been characterized to be primarily due more to carrier-mediated than passive processes.35, 36 These are two examples where we see how different permeability rates are needed to accurately predict

BCS absorption and BDDCS metabolism. Amiloride and cephalexin’s high intestinal permeability rates (inclusive of both carrier-mediated and passive processes) accurately predict their high extents of absorption, whereas their low passive transcellular permeability rates accurately predict their low extents of metabolism.

It is shown that generally drug metabolism above does not predict drug absorption. However, there are cases when drug metabolism can be helpful in predicting the BCS criterion of a high extent of absorption. An extent of metabolism ≥90% accurately predicted the extent of absorption ≥90% in all cases. It has been proposed by

Benet and coworkers10 that the following criteria be used to define ≥90% metabolized for marketed drugs: “Following a single oral dose to humans, administered at the highest dose strength, mass balance of the Phase 1 oxidative and Phase 2 conjugative drug metabolites in the urine and feces, measured either as unlabeled, radioactive labeled, or non-radioactive labeled substances, account for ≥90% of the drug dosed. This is the strictest definition for a waiver based on metabolism. For an orally administered drug to be ≥90% metabolized by Phase 1 oxidative and Phase 2 conjugative processes, it is

99 obvious that the drug must be absorbed.” Since there is a good correlation in extensive metabolism predicting extensive absorption (although not necessarily vice versa), the recommendation for regulatory agencies to add the extent of drug metabolism (i.e., ≥90% metabolized) as an alternative method for the extent of drug absorption (i.e., ≥90% absorbed) seems appropriate in defining Class 1 drugs suitable for a waiver of in vivo studies of bioequivalence.

3.4.2 Drug Discovery Considerations when Making BCS and BDDCS Predictions

Academic, industry, and regulatory scientists have attempted to predict either

BCS and BDDCS classifications,11-13 or a provisional biopharmaceutics classification system14 that combines the two using the same permeability method. These predictions can successfully achieve accuracies when their classes agree (i.e. both are high or low), and be unsuccessful when their classes disagree (i.e. one is high and the other is low).

Since cases of high metabolism (i.e. ≥90% metabolized) will only exist with high absorption (i.e. ≥90% absorbed), discontinuity will only occur when absorption is high

(i.e. ≥90% absorbed) and metabolism is low (i.e. <90% metabolized). In the latter case, only using the most suitable permeability method will likely lead to the correct class prediction.

The PAMPA and Caco-2 systems are attractive in vitro systems for high throughput permeability screening, known to be deficient in transporter expression and paracellular mechanisms (if not completely devoid of them as in the case of PAMPA).

Hence, they are excellent for characterizing the passive transcellular permeability rates of

100 compounds in early drug discovery programs. The PAMPA and Caco-2 permeability rate measures have a stronger correlation and higher prediction accuracy with BDDCS metabolism than with BCS absorption. This agrees with my hypothesis that passive transcellular permeability rates accurately predict BDDCS metabolism, while complete human intestinal permeability rates accurately predict BCS absorption. This may not be surprising since previous correlations of lipophilicity have been observed with passive transcellular permeability37 and with drug metabolism.38

When the passive transcellular permeability rate is high (i.e. PAMPA or Caco-2 permeability rate is high), then the human intestinal permeability rate (or extent of absorption) should also be high, regardless of whether carrier-mediated transport is occurring. This is consistently true for the Caco-2 datasets independent of their laboratories of origin. However, only a single PAMPA model was found to be consistent in this finding. Although the differences between each PAMPA model cannot be confirmed due to the information being proprietary, the lipid/oil/lipid PAMPA membrane20 may be more representative of transcellular membrane transport than other models that include hydrophilic membrane transport.21, 22 Thus, the Caco-2 cell line, and only some PAMPA models, may be appropriate for screening and supporting drugs that are potentially eligible for a biowaiver or for predicting their biopharmaceutics (whether

BCS or BDDCS) classifications.

A number of outliers were observed when using low PAMPA and Caco-2 permeability rate measures to predict a low extent (<90%) of absorption and slightly fewer when predicting a low (<90%) extent of metabolism. It has been previously noted that metoprolol may be too strict of a high-permeability rate measures. I also observed

101 that there were many drugs that were predicted as false positives that were slightly lower than metoprolol’s permeability rate but had high extents of absorption and metabolism.

When an analysis was conducted using a permeability rate value that was 30% lower than that of metoprolol’s, many of these previous false positives were found to be correctly predicted using this adjusted high-permeability rate cut-off. In the original BCS paper,2 it was suggested that metoprolol was suitable as a high-permeability reference standard since its extent of absorption is close to 90%. However, there are numerous sources indicating that its extent of absorption (as well as extent of metabolism) is closer to

95%,30 supporting that a more suitable high-permeability reference standard with an extent measure closer to 90% should be investigated.

Even with a less strict permeability rate cut-off, there are still false positives that can arise in the BCS and BDDCS predictions when using PAMPA and Caco-2 systems.

The remaining false positives in the BCS predictions were also observed to be very hydrophilic in nature (Table 3.4). Many of these hydrophilic false positives (e.g., amiloride, cephalexin, levodopa, and zidovudine) have been identified to be substrates of highly expressed intestinal transporters (e.g., amino acid, peptide, and nucleoside transporters) that are deficient in these in vitro systems.18 It should be cautioned though that these systems may be deficient in accurately predicting the high intestinal permeability rates of all NMEs that are particularly substrates of these highly expressed intestinal transporters. I have previously recommended that drug discovery programs utilize other permeability methods that are more representative of carrier-mediated and passive processes for screening these NMEs, such as the human or rat intestine.18

In evaluating the remaining false positives in the BDDCS predictions using

102

PAMPA and Caco-2 systems, we found these compounds also to be very hydrophilic in nature (Table 3.4). Many of these false positives are extensively metabolized either primarily or initially by non-CYP metabolism, for example acebutolol,39 acetylsalicylic acid,40 bromocriptine,41 ketoprofen,42 labetolol,43 levodopa,44 salicylic acid,45 and zidovudine.46 This indicates that the metabolism of some very hydrophilic compounds may not be accurately predicted by these systems and alternative methods should be used to verify their BDDCS classification.

3.5 References

1. Li, A. P. Screening for human ADME/Tox drug properties in drug discovery.

Drug Discov. Today 2001, 6, 357-366.

2. Amidon, G. L.; Lennernas, H.; Shah, V. P.; Crison, J. R. A theoretical basis for a biopharmaceutic drug classification: the correlation of in vitro drug product dissolution and in vivo bioavailability. Pharm. Res. 1995, 12, 413-420.

3. Waiver of In Vivo Bioavailability and Bioequivalence Studies for Immediate-

Release Solid Oral Dosage Forms Based on a Biopharmaceutics Classification System.

FDA Guidance for Industry; Food and Drug Administration: Rockville, MD, 2000.

4. Lennernas, H. Intestinal permeability and its relevance for absorption and elimination. Xenobiotica 2007, 37, 1015-1051.

5. Ungell, A. L. Caco-2 replace or refine? Drug Discov. Today Tech. 2004, 1, 423-

430.

103

6. Varma, M. V.; Khandavilli, S.; Ashokraj, Y.; Jain, A.; Dhanikula, A.; Sood, A.;

Thomas, N. S.; Pillai, O.; Sharma, P.; Gandhi, R.; Agrawal, S.; Nair, V.; Panchagnula, R.

Biopharmaceutic classification system: a scientific framework for pharmacokinetic optimization in drug research. Curr. Drug Metab. 2004, 5, 375-388.

7. Wu, C. Y.; Benet, L. Z. Predicting drug disposition via application of BCS: transport/absorption/elimination interplay and development of a biopharmaceutics drug disposition classification system. Pharm. Res. 2005, 22, 11-23.

8. Shugarts, S.; Benet, L. Z. The role of transporters in the pharmacokinetics of orally administered drugs. Pharm. Res. 2009, 26, 2039-2054.

9. Broccatelli, F.; Larregieu, C. A.; Cruciani, G.; Oprea, T. I.; Benet, L. Z.

Improving the prediction of the brain disposition for orally administered drugs using

BDDCS. Adv. Drug Deliv. Rev. 2012, 64, 95-109.

10. Benet, L. Z.; Amidon, G. L.; Barends, D. M.; Lennernas, H.; Polli, J. E.; Shah, V.

P.; Stavchansky, S. A.; Yu, L. X. The use of BDDCS in classifying the permeability of marketed drugs. Pharm. Res. 2008, 25, 483-488.

11. Takagi, T.; Ramachandran, C.; Bermejo, M.; Yamashita, S.; Yu, L. X.; Amidon,

G. L. A provisional biopharmaceutical classification of the top 200 oral drug products in the United States, Great Britain, Spain, and Japan. Mol. Pharmaceutics 2006, 3, 631-643.

12. Dahan, A.; Miller, J. M.; Amidon, G. L. Prediction of solubility and permeability class membership: provisional BCS classification of the world's top oral drugs. AAPS J.

2009, 11, 740-6.

13. Chen, M. L.; Yu, L. The use of drug metabolism for prediction of intestinal permeability. Mol. Pharmaceutics 2009, 6, 74-81.

104

14. Varma, M. V.; Gardner, I.; Steyn, S. J.; Nkansah, P.; Rotter, C. J.; Whitney-

Pickett, C.; Zhang, H.; Di, L.; Cram, M.; Fenner, K. S.; El-Kattan, A. F. pH-dependent solubility and permeability criteria for provisional biopharmaceutics classification (BCS and BDDCS) in early drug discovery. Mol. Pharmaceutics 2012, 9, 1199-1212.

15. Larregieu, C. A.; Benet, L. Z. The confounding FDA redefinition of intestinal permeability in terms of extent of absorption. AAPS J. 2009, 11, Abstract W1840.

16. Benet, L. Z.; Larregieu, C. A. The FDA should eliminate the ambiguities in the current BCS biowaiver guidance and make public the drugs for which BCS biowaivers have been granted. Clin. Pharmacol. Ther. 2010, 88, 405-407.

17. Chen, M. L.; Amidon, G. L.; Benet, L. Z.; Lennernas, H.; Yu, L. X. The BCS,

BDDCS, and regulatory guidances. Pharm. Res. 2012, 28, 1774-1778.

18. Larregieu, C. A.; Benet, L. Z. Drug discovery and regulatory considerations for improving in silico and in vitro predictions that use Caco-2 as a surrogate for human intestinal permeability measurements. AAPS J. 2013, 15, 483-497.

19. Kansy, M.; Senner, F.; Gubernator, K. Physicochemical high throughput screening: parallel artificial membrane permeation assay in the description of passive absorption processes. J. Med. Chem. 1998, 41, 1007-1010.

20. Chen, X.; Murawski, A.; Patel, K.; Crespi, C. L.; Balimane, P. V. A novel design of artificial membrane for improving the PAMPA model. Pharm. Res. 2008, 25, 1511-

1520.

21. Sugano, K.; Takata, N.; Machida, M.; Saitoh, K.; Terada, K. Prediction of passive intestinal absorption using bio-mimetic artificial membrane permeation assay and the paracellular pathway model. Int. J. Pharm. 2002, 241, 241-251.

105

22. Zhu, C.; Jiang, L.; Chen, T. M.; Hwang, K. K. A comparative study of artificial membrane permeability assay for high throughput profiling of drug absorption potential.

Eur. J. Med. Chem. 2002, 37, 399-407.

23. Alsenz, J.; Haenel, E. Development of a 7-day, 96-well Caco-2 permeability assay with high-throughput direct UV compound analysis. Pharm. Res. 2003, 20, 1961-

1969.

24. Irvine, J. D.; Takahashi, L.; Lockhart, K.; Cheong, J.; Tolan, J. W.; Selick, H. E.;

Grove, J. R. MDCK (Madin-Darby canine kidney) cells: a tool for membrane permeability screening. J. Pharm. Sci. 1999, 88, 28-33.

25. Li, C.; Liu, T.; Cui, X.; Uss, A. S.; Cheng, K. C. Development of in vitro pharmacokinetic screens using Caco-2, human hepatocyte, and Caco-2/human hepatocyte hybrid systems for the prediction of oral bioavailability in humans. J. Biomol. Screen.

2007, 12, 1084-1091.

26. Yazdanian, M.; Glynn, S. L.; Wright, J. L.; Hawi, A. Correlating partitioning and

Caco-2 cell permeability of structurally diverse small molecular weight compounds.

Pharm. Res. 1998, 15, 1490-1494.

27. L. Z. Benet; S. Øie; Schwarz., J. B. Design and Optimization of Dosage

Regimens: Pharmacokinetic Data. In Goodman and Gilman’s The Pharmacological Basis of Therapeutics, 9th ed.; J. G. Hardman; L. E. Limbird; P. B. Molinoff; R. W. Ruddon;

Gilman., A. G.; McGraw Hill: New York, 1996; pp 1707–1792.

28. Thummel, K. E.; Shen, D. D. Design and Optimization of Dosage Regimens:

Pharmacokinetic Data. . In Goodman and Gilman’s The Pharmacological Basis of

106

Therapeutics, 10th ed.; Hardman, J. G., Limbird, L. E., Eds.; McGraw Hill: New York,

2001; pp 1924–2023.

29. Lacy, C. F.; Armstrong, L. L.; Goldman, M. P.; Lance, L. L. Drug Information

Handbook: A Comprehensive Resource for All Clinicians and Healthcare Professionals.

19th ed.; Lexi-Comp, Inc.: Hudson, Ohio, 2010; p 16-1655.

30. DRUGDEX® System [Internet database]. Greenwood Village, Colo: Thomson

Reuters (Healthcare) Inc. Updated periodically.

31. Volpe, D. A.; Faustino, P. J.; Ciavarella, A. B.; Asafu-Adjaye, E. B.; Ellison, C.

D.; Yu, L. X.; Hussain, A. S. Classification of drug permeability with a Caco-2 cell monolayer assay. Clin. Res. Reg. Affairs 2007, 24, 39-47.

32. Incecayir, T.; Tsume, Y.; Amidon, G. L. Comparison of the permeability of metoprolol and labetalol in rat, mouse, and Caco-2 cells: use as a reference standard for

BCS classification. Mol. Pharmaceutics 2013, 10, 958-966.

33. Weiss, P.; Hersey, R. M.; Dujovne, C. A.; Bianchine, J. R. The metabolism of amiloride hydrochloride in man. Clin. Pharmacol. Ther. 1969, 10, 401-406.

34. Bergan, T.; Midtvedt, T.; Erikssen, J. Human pharmacokinetics of cephalexin.

Pharmacology 1970, 4, 264-472.

35. Wright, S. H.; Wunz, T. M. Amiloride transport in rabbit renal brush-border membrane vesicles. Am. J. Physiol. 1989, 256, F462-F468.

36. Dantzig, A. H.; Bergin, L. Uptake of the cephalosporin, cephalexin, by a dipeptide transport carrier in the human intestinal cell line, Caco-2. Biochim. Biophys.

Acta 1990, 1027, 211-217.

107

37. Camenisch, G.; Alsenz, J.; van de Waterbeemd, H.; Folkers, G. Estimation of permeability by passive diffusion through Caco-2 cell monolayers using the drugs' lipophilicity and molecular weight. Eur. J. Pharm. Sci. 1998, 6, 317-324.

38. Lewis, D. F. Human cytochromes P450 associated with the phase 1 metabolism of drugs and other xenobiotics: a compilation of substrates and inhibitors of the CYP1,

CYP2 and CYP3 families. Curr. Med. Chem. 2003, 10, 1955-1972.

39. Lilja, J. J.; Raaska, K.; Neuvonen, P. J. Effects of grapefruit juice on the pharmacokinetics of acebutolol. Br. J. Clin. Pharmacol. 2005, 60, 659-663.

40. Cummings, A. J.; King, M. L. Urinary excretion of acetylsalicylic acid in man.

Nature 1966, 209, 620-621.

41. Matsubayashi, K.; Matsumoto, H.; Fukui, Y. Contribution of cytochrome P450

3A pathway to bromocriptine metabolism and effects of ferrous iron and hypoxia-re- oxygenation on its elimination in the perfused rat liver. J. Pharm. Pharmacol. 1997, 49,

551-557.

42. Jamali, F.; Brocks, D. R. Clinical pharmacokinetics of ketoprofen and its enantiomers. Clin. Pharmacokinet. 1990, 19, 197-217.

43. McNeil, J. J.; Louis, W. J. Clinical pharmacokinetics of labetalol. Clin.

Pharmacokinet. 1984, 9, 157-167.

44. Kumakura, Y.; Cumming, P. PET studies of cerebral levodopa metabolism: a review of clinical findings and modeling approaches. Neuroscientist 2009, 15, 635-650.

45. Alpen, E. L.; Mandel, H. G.; Rodwell, V. W.; Smith, P. K. The metabolism of

C14 carboxyl salicylic acid in the dog and in man. J. Pharmacol. Exp. Ther. 1951, 102,

150-155.

108

46. Veal, G. J.; Back, D. J. Metabolism of zidovudine. Gen. Pharmacol. 1995, 26,

1469-1475.

109

Chapter 4: Extending the Application of BDDCS to the Prediction of Brain Disposition of Orally Administered Drugs

4.1 Introduction

Delivery of new drug candidates to the central nervous system (CNS) is a challenging problem in drug development. It is important to design drugs that target the

CNS when it is the site of action. It is also important to prevent peripherally acting drugs from accessing the CNS where they could exert undesirable and potentially harmful side effects. Being one of the oldest issues in drug discovery and QSAR (quantitative structure-activity relationship), significant efforts are made in academia and industry to develop screens for optimization and prioritization, which include in vitro assays and computational models to evaluate CNS penetration.

QSAR studies have been instrumental in early phases of drug discovery, serving as fast and inexpensive tools to prioritize new molecular entities (NMEs) as candidate

CNS agents. BBB disposition has been linked to molecular properties such as drug

1, 2 1-3 lipophilicity (LogP and LogD7.4), molecular weight, and the tendency to form hydrogen bonds quantified as polar surface area (PSA)4 or as a function of nitrogen and oxygen counts.5 These approaches led to simple rules of thumb, or to more elaborate in silico models for predicting CNS disposition yielding ~80% accuracies for early phase screening.5-7 However, compound optimization based simply on molecular properties restricts the portion of chemical space explored. Therefore, anchoring brain penetration directly to biological factors may result in a more flexible selection process for drug discovery, which is desirable when designing compound libraries or optimizing

110 interesting candidates.

After characterizing the expression of P-gp in the mouse brain, Cordon-Cardo et al. first suggested that P-gp might play a role in vivo in limiting brain penetration of xenobiotics.8 The importance of P-gp in CNS penetration has been established using knockout mice.9 Mahar Doan and coworkers experimentally tested the efflux ratio (ER) of 48 CNS and 45 non-CNS drugs, reporting that CNS agents are less likely to be P-gp substrates and more likely to be highly permeable drugs.10 By analyzing 119 marketed

CNS drugs and 108 proprietary CNS drug candidates, Wager and colleagues 11 confirmed the latter observations. Going from candidates to marketed CNS drugs, these investigators observed a marked increase in the number of highly permeable compounds

(from 51% to 75%) and P-gp non-substrates (from 55% to 75%). Moreover, only about

40% of CNS drug candidates were both highly permeable and P-gp non-substrates, as opposed to 70% of successfully CNS marketed drugs. This indicates that both high permeability and lack of P-gp efflux are desirable for CNS drug candidates to become marketed CNS drugs. Surprisingly, 20% of marketed CNS drugs analyzed exhibited both unfavorable permeability and P-gp efflux. Until now, no explanations have been proposed for these observations.

The Biopharmaceutics Drug Disposition Classification System (BDDCS) has been successful in predicting in vivo absorption, disposition, and drug-drug interactions of marketed drugs.12-14 BDDCS is a four class system based on extent of metabolism and solubility measures that has been used to explain the role of transporters in pharmacokinetics and their interplay with metabolizing enzymes in the liver and intestine. Based on observations in the intestine and liver, BDDCS class 1 drugs appear to

111 be the most biopharmaceutically desirable class due to their high intestinal absorption and lack of clinically relevant uptake and efflux transporter effects.12, 14 Nine of the 14 drugs with unfavorable in vitro permeability and P-gp profile reported by Wager and coworkers have BDDCS classification data available in the dataset of oral drugs that has been recently published;15 of these, 7 (78%) are BDDCS class 1 drugs. This suggests that, analogous to what has been observed in the intestine and liver, class 1 drugs may not be significantly influenced by transporters in the brain.

Here, the use of BDDCS in predicting the brain disposition of 153 orally administered drugs for which P-gp efflux data are available was investigated. Using

BDDCS, in vitro P-gp efflux ratio values, and in silico calculated permeability (VolSurf+ descriptor CACO2), three rules of thumb have been defined to predict brain penetration, which for the present data set are accurate over 90% of the time. The model outperforms in silico strategies for predicted BBB permeability,5, 16, 17 which for the 153 drugs were accurate around 80% of the time. Simple molecular properties such as those used in the

Norinder-Haeberlein5 and Lipinski BBB17 rules of thumb for brain disposition were analyzed with respect to BDDCS classification, P-gp impact, and in silico predicted permeability. The use of this framework in predicting the sedative effect of antihistaminic drugs is also discussed.

4.2 Criteria and Rationale for Data Collection

4.2.1 Efflux Ratio

The P-gp impact on drug permeability, evaluated in terms of efflux ratio (ER), is

112 defined as follows:

ER = Papp(ba)/Papp(ab) where Papp(ab) is permeability from the apical to the basolateral side (absorption) and

Papp(ba) is permeability from the basolateral to the apical side (secretion). For P-gp substrates, the absorption should be decreased and the secretion increased in cell lines over-expressing P-gp, as opposed to wild type cell lines, yielding ER values greater than one. In 2001, Polli and colleagues18 published a collection of over 60 molecules tested for

P-gp transport and inhibition, suggesting that an efflux ratio value ≥2 should be adopted to identify P-gp substrates. However, in confirmatory studies using a P-gp inhibitor for non-substrates, the authors noted that a few drugs having ER <2 (loratadine, diltiazem, and neostigmine) were P-gp substrates. For all of the oral drugs included in the previously analyzed BDDCS dataset,15 an extensive literature search for ER values in

MDCK-MDR1 cell lines was conducted. At least one ER value for 160 oral drugs from

16 different articles10, 11, 18-31 were found, resulting in a database of 318 ER values. About

77% of the ERs were derived from MDCK-MDR1 cell lines originating from the

Netherlands Cancer Institute (Borst cell lines), while the remaining 23% were derived from MDCK-MDR1 cell lines originating from the National Institute of Health (NIH cell lines). It was observed that ER values in the dataset from NIH cell lines tended to be higher than those from Borst cell lines, in agreement with the reported difference in P-gp expression levels for the two cell lines.32 In the Results section of this chapter, the criteria for defining P-gp substrates and non-substrates in these cell lines are addressed.

113

4.2.2 BDDCS Classification

BDDCS class assignments were based on extent of metabolism and FDA solubility. Highly metabolized drugs (equal or above 70% in humans) were assigned to classes 1 and 2, while highly soluble drugs were assigned to classes 1 and 3. Solubility class is defined based on the dose number (Do):

Do= (HDS/250mL)/LWS where HDS is the highest dose strength approved for commercial use, and LWS is the lowest water solubility at 37ºC in the pH range from 1 to 7.5. When Do 1, the drug is considered highly soluble.

4.2.3 BBB Penetration Data

An exhaustive search for blood-brain barrier (BBB) penetration data for the 160 oral drugs where BDDCS class and P-gp ER data were available was conducted to classify the drugs as either brain-penetrating (BBB+) or having little if any ability to cross the BBB (BBB-). Marketed CNS agents were directly assigned to the BBB+ class without further investigation. When the unbound drug brain/plasma ratios in humans were available, these values were used for defining the BBB class; drugs with values greater than or equal to 0.1 were assigned as BBB+, otherwise, they were assigned as

BBB-.

When the unbound brain/plasma ratio of a drug in humans was unavailable, then either the unbound CSF/plasma ratio in humans or brain penetration data from other

114 species was used. Interspecies differences and using CSF/plasma ratios in place of brain/plasma ratios may affect the BBB assignment outcome.33, 34 Therefore, a prioritization scheme of how to define the BBB class was created when more than one of these values was available. From an analysis of rat unbound brain/plasma and

CSF/plasma ratio data for 39 drugs,34 even when up to a five-fold difference was found between these data, the BBB assignment was in agreement for 35 out of 39 drugs. By comparing human and rat brain penetration data, the human CSF/plasma ratio was found to be a better surrogate for the human unbound brain/plasma ratio than any available data from other species. For example, verapamil is a BBB+ drug (human brain/plasma ratio =

0.55);35 its human CSF/plasma ratio of 1.1334 consistently classifies it as a BBB+ drug, while its rat brain/plasma ratio of 0.05334 misclassifies it as a BBB- drug. Hence, when human brain/plasma ratio data were unavailable, the human CSF/plasma ratio information for assigning the BBB class over any other species data (including brain/plasma ratios) was used. Due to the maximum five-fold numerical difference observed between CSF/plasma and brain/plasma ratios for the same drugs, no significant incongruence in the BBB assignment for drugs with CSF/plasma ratios greater than or equal to 0.5 (or at least 33% distributed in the CSF) was expected. When numerical data were unavailable in humans and other species, qualitative phrases like “distribution in brain negligible” were used to assign the drug as BBB-. Qualitative phrases like

“extensively distributed in brain tissues” or evidence of CNS side-effects were used to assign the drug as BBB+.

Diltiazem was excluded from our analysis since it was found to be controversial; it is reported to exhibit a “decreasing concentration gradient in the following organs:

115 liver, kidney, lung, spleen, heart and brain”,36 yet Naito et al.37 reported that the

CSF/plasma ratio for diltiazem was from 0.05 to 0.2. The prodrug famciclovir was also excluded since the BDDCS classification was relative to the dosed compound, not to the active species (penciclovir) that penetrates the BBB. Five other drugs had to be excluded because there was no BBB penetration data available that matched any of these criteria.

The original BBB penetration data for the remaining 153 oral drugs can be found in these references.10, 11, 22, 34-106

4.2.4 In Silico Calculations

Even though it was preferred to relate in vitro properties with human data, an in silico parameter for passive permeability had to be adopted due to the unavailability of consistent in vitro data for all the 153 drugs. The molecular descriptor CACO2 generated using VolSurf+,107 a computational procedure specifically designed to produce descriptors related to pharmacokinetic properties,16 was used as a surrogate for in vitro passive permeability data. Previous work in this laboratory has suggested that the

VolSurf+ CACO2 descriptor derived from the Caco-2 cell permeation studies was a suitable surrogate for passive permeability.15, 108 The CACO2 model is a qualitative model trained using a thousand compound in-house dataset. 109, 110

To compare how the proposed framework improved the brain penetration prediction of current state-of-the-art in silico rules of thumb and models, several other descriptors were included. In analogy with CACO2, the VolSurf+ descriptor LgBB is trained on a five hundred compound in-house data set, and it is based on the methodology

116 presented by Crivori et al.109, 111 The parameters necessary to predict brain penetration following the Norinder-Haeberlein5 and Lipinski BBB17 rules of thumb were also calculated. These included calculated LogP (cLogP), molecular weight (MW), hydrogen bond acceptor counts (HBA), hydrogen bond donor count (HBD), and number of violation to Lipinski Rule of 5 (Ro5) that were taken from previous work,108 and counts of oxygen and nitrogen that were extrapolated from the SMILES structure of the drugs as reported in WOMBAT-PK.112, 113 Predicted brain penetration values according to

Norinder-Haeberlein5 and Lipinski BBB17 rules of thumbs were also reported. Norinder and Haeberlein provided two rules for brain penetration: (i) when the sum of nitrogens and oxygens in the molecule (N+O) is five or less, the drug is considered BBB+

(brain/plasma ratio > 0.1), and (ii) when cLogP minus N+O is positive, brain/plasma ratio is greater than unity (brain/plasma ratio > 1). Lipinski BBB, as reported by Pajouhesh and Lenz,17 proposed that a drug is unlikely to be BBB+ if:

-Molecular weight > 400

-LogP > 5

-Hydrogen bond acceptor count > 7

-Hydrogen bond donor count > 3

To evaluate any possible influence of pKa on brain penetration, the lowest in silico pKa calculations for the acid and basic centers of the 153 drugs using the software MoKa114,

115 was performed and reported.

117

4.3 Elaboration, Analysis, and Application of Rules to Estimate Brain Disposition

4.3.1 P-gp and Brain Disposition

Of the 160 oral drugs for which P-gp ER data were available, 63 had more than one reported ER value in Borst cell lines. Adopting the ER threshold of 2 as suggested by

Polli and coworkers,18 87% of these 63 drugs were classified coherently as either P-gp substrates or non-substrates. Only 4 drugs had more than one reported ER in NIH cell lines. Of these 4 drugs, 3 of them were inconsistently classified using the threshold of 2.

For 53 drugs where values were available in both Borst and NIH cell lines, the agreement in assignment was only 67%. This suggests that another threshold should be adopted for the NIH cell lines. For the 53 drugs where at least one reported ER value was available from both the Borst and NIH cell lines, ERs in NIH cell lines were on average 4.26-fold higher than those in the Borst cell lines. Therefore, when assigning the P-gp class, a threshold of 2 for Borst cell lines, and 8.5 for NIH cell lines was adopted. When more than one ER value was available for the same drug in the same cell line, the threshold values were then applied to the averages. Using these criteria, the two systems agreed about 81% of the time. For the 10 drugs where there was no agreement (atomoxetine, chlorpromazine, citalopram, clozapine, metoclopramide, olanzapine, ranitidine, risperidone, sertraline, and tiagabine), assignment was based on the Borst cell line data.

Even though an ER value of 8.5 was adopted as the threshold for defining P-gp impact on brain penetration for data from the NIH cell lines, the appropriateness of this threshold should be further investigated.

118

The assumption that P-gp efflux is the only factor limiting brain penetration, was found to be correct 78% of the time for this dataset (Figure 4.1): 63% of the data set were

P-gp non-substrates that are BBB+, while 15% were P-gp substrates that are BBB-

(Figure 4.1A). In particular, 80% of BBB+ molecules were P-gp non-substrates (Figure

4.1C), and 72% of BBB- were P-gp substrates (Figure 4.1D). However, it must be noted that around 51% of P-gp substrates were BBB+ (Figure 4.1E), thus being a P-gp substrate does not imply being BBB-, while 92% of P-gp non-substrates were BBB+ (Figure 4.1F).

Interestingly, the absence of P-gp effect on brain penetration was mainly associated with BDDCS class 1 drugs (Figure 4.1B), where 98% of the time class 1 drugs were BBB+, even if 23% of them were P-gp substrates, therefore 88% of P-gp substrates in class 1 were BBB+. In contrast, 75% of the P-gp substrates in BDDCS classes 2, 3, and 4, were BBB-.

4.3.2 In Silico Permeability and Brain Disposition

Due to a lack of in vitro permeability data, the relationship between in silico permeability and brain disposition was evaluated using the calculated octanol/water partition coefficient cLogP116 and the VolSurf+ descriptor CACO2. LogP has been highly used in medicinal chemistry as a surrogate for permeability. It has previously been demonstrated that the VolSurf+ descriptor CACO2 can be used to distinguish highly versus poorly metabolized/permeable compounds.15 Therefore, the suitability of this descriptor was evaluated as a filter for poorly permeable compounds. In Figure 4.2, box plots117 and receiver-operating characteristic (ROC) curves118 are used to show the distribution of BBB+ and BBB- classes with respect to cLogP and calculated CACO2

119

Figure 4.1 Pie and bar charts showing the relative percentage of the drugs in the data set classified based on their ability to cross the blood brain barrier, P-gp profile, and BDDCS class.

120 permeability. In agreement with Wager and coworkers,11 no significant difference in cLogP distribution for the two sets (Figure 4.2A and C) was found. In contrast, a marked difference between BBB+ and BBB- with respect to CACO2 permeability was recognized (Figure 4.2A and B). If a value of CACO2 descriptor equal to -0.3 is adopted,

40.6% of BBB- molecules are filtered with a loss of only 3.3% of BBB+ (Figure 4.2A).

Others have linked brain penetration to pKa due to its effect on passive permeability; that is, neutral species are expected to be favored in crossing biological membranes, as opposed to charged species.119 In particular, this effect appears to be marked for acid centers having low pKa (under 5.5), since these drugs (e.g., Class 2

NSAIDs) are mostly negatively charged at physiological pH. The lowest acid pKa for drugs in this dataset using MoKa was estimated. Six (cefuroxime, cetirizine, fexofenadine, methotrexate, pravastatin, and sulfasalazine) of the 32 (18.7%) BBB- drugs had a calculated acid pKa under 5.5, while this was true for only 5 (gabapentin, indomethacin, rosuvastatin, tiagabine, and warfarin) out of 121 BBB+ drugs (4.1%).

However, of the 6 BBB- drugs with low acid pKa, all but cetirizine were predicted to have low permeability (VolSurf+ CACO2 descriptor below -0.3) while this was true for only 2 (gabapentin and rosuvastatin) BBB+ drugs with low acid pKa. Hence, even if based on these data an influence of pKa on brain penetration can be recognized, it appears to be taken into account by the VolSurf+ CACO2 descriptor.

4.3.3 Generation of BDDCS-Based BBB Rules

The data collected were used to elaborate a classification tree for predicting CNS

121

Figure 4.2 A) Receiver-operating characteristic (ROC) curves for CACO2 permeability and cLogP used as classifiers to discriminate BBB+ and BBB- classes, where BBB- is the targeted class (class to be predicted). Each point in the ROC curve represents a descriptor value (either CACO2 or cLogP) and the fraction of true positives (correctly classified BBB- drugs) and false positives (incorrectly classified BBB- drugs) associated, if that descriptor value is used as the cut-off to discriminate the two classes. B) VolSurf+ CACO2 descriptor distributions of BBB+ and BBB- drugs. The box is delimited by the 1st and the 3rd quartiles, the line inside the box is the median, the whiskers correspond to 5th and 95th percentiles, and the circles are outliers. C) cLogP distributions of BBB+ and BBB- drugs. The box is delimited by the 1st and the 3rd quartiles, the line inside the box is the median, the whiskers correspond to 5th and 95th percentiles, and the circles are outliers.

122

disposition (Figure 4.3). The classification tree was based on 153 oral drugs and was created using the data-mining software Orange Canvas.120 In order to have a straightforward framework, three different two-level variables were chosen: BDDCS class (either Class 1 or Classes 2, 3, and 4), P-gp Impact (“+” for substrates, “-” for non- substrates), and calculated permeability (“+” if the VolSurf+ CACO2 descriptor was above -0.3, otherwise “-”). The rationale for the choice of the threshold for the CACO2 descriptor is presented above. The following options were used for building the classification tree: attribute selection based on gain ratio, no binarization, stop splitting nodes with a majority class of 90%, post-pruning (recursively merge leaves with the same majority class, m=3).

Being based on a sequence of three two-level variables, it can be also simply considered as a pipeline of rules of thumb. By classifying drugs first based on calculated permeability, then P-gp efflux, and finally BDDCS Classification, the model is correct for 138 out of 153 drugs, gaining an accuracy of 90.2% for this data set (Figures 4.3 and

4.4). We address 15 incorrectly classified drugs in following Outliers Analysis section.

As previously discussed, the use of a filter for low permeability allows correct classification of a significant portion of BBB- drugs with only few BBB+ drugs being misclassified. Drugs that do not seem to be limited by passive permeability predominantly (93.7%) cross the BBB if they are P-gp non-substrates. Evaluating only permeable P-gp substrate drugs (the last level in Figure 4.3), BDDCS Class 1 drugs were generally capable of penetrating the BBB (94.4%), while the remaining classes of drugs were predominantly BBB- (68.1%). In Figure 4.4, the accuracy in predicting brain penetration for this dataset is shown using the classification tree (Table 4.1) as well as the

123

Figure 4.3 Pipeline of rules for predicting brain disposition. Rules are presented as a classification tree, and are based on the drug BDDCS class, P-gp profile, and the VolSurf+ descriptor CACO2 value (cPermeability is + if it is above -0.3, “-” if it is below). Pie chart red represents BBB- fraction of drug population evaluated; pie chart green represents BBB+ fraction.

124

Figure 4.4 Accuracy of blood-brain barrier crossing prediction for the different methods. The sensitivity is the accuracy in predicting BBB+ drugs, while specificity is the accuracy in predicting BBB- drugs.

125

VolSurf+ descriptor LgBB, and Lipinski BBB and Norinder-Haeberlein rules of thumb.

Notably, in silico methods solely based on the molecular structure have accuracy of

~80%, with Lipinski BBB rules and LgBB being reasonably accurate and well-balanced.

These methods appear to be well-suited for early phase drug discovery compound prioritization. The rule-based system derived from our classification tree outperforms these other models. However, this approach cannot be directly compared with the other models, as it requires several in vitro measures. Therefore, it is recommended to adopt it as guidelines during advanced stage drug discovery.

4.3.4 Outliers Analysis

Fifteen drugs (10% of the data set) were incorrectly predicted. Eleven of these: amisulpride, aprepitant, chloroquine, clozapine, enoxacin, gabapentin, paliperidone, rosuvastatin, trimethoprim, ziprasidone, and zidovudine were false negatives (incorrectly predicted as BBB-). Four drugs (enoxacin, gabapentin, rosuvastatin, and zidovudine) had false negative predictions due to poor passive permeability calculations. The remaining seven drugs were false negatives since they were not BDDCS class 1. Saturable uptake processes have been implicated in the transport of amisulpride121 and gabapentin122, 123 into the brain. Four other false negatives have been reported to be substrates for uptake transporters expressed in the brain124: rosuvastatin (OAT3, OATP1A2, OATP1B1, and

OATP2B1),125-127 zidovudine (OAT1, OAT2, OAT3, and OAT4),128 trimethoprim

(OCT2),129 and chloroquine (OCT2).129 Paliperidone is a false negative since its BBB assignment is BBB+ class based on its category as a neurodrug, 11 even though

126

Table 4.1 List of 153 oral drugs and data used for brain penetration prediction.