This item is only available as the following downloads:
1 APPL ICATION OF PHARMACOKINETIC AND PHARMACODYNAMIC CONCEPTS I N PRECLINICAL AND CLINICAL STUDIES FOR DRUG DEVELOPMENT By LI LI A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL F UL L FILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2013
2 2013 Li Li
3 To my family
4 ACKNOWLEDGMENTS First, I would like to thank my advisor, Dr. Hartmut Derendorf, for his great support in my research and study. His insight and good advice helped me successfully analyze my experimental results and move into the next stage of my career. I also would like to express great appreciation to my co advisor Dr. Veronika Butterweck for introducin g me into the pharmaceutics study and guiding me in my research work. She allowed me think freely on research, supported my ideas, and broadened my research experience. Special thanks also go to my supervisory committee members, Dr. Reginald Frye and Dr. S aunjoo Yoon for their valuable advice throughout my doctoral research. I would like to specially thank Dr. Reginald Frye for his help and guidance on LC MS/MS analysis in my preclinical study. Meanwhile, I would like to thank the postdoctoral fellows Dr. C hethan Sampath, Dr. Ravi Singh and all my labmates. I would also like to take this opportunity to express my thanks to Karin Haug, Xuan Liu and all interns for their nice help in my extensive sampling work. Extended special thanks go to all my Chinese frie nds for their help and support over the years. Finally, I deeply thank my parents. Their support and unconditional love have been accompanying me for all my life. I would like to thank my husband, whose continuing support and encouragement helped me overco me all the obstacles in my Ph.D. study.
5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 7 LIST OF FIGURE S ................................ ................................ ................................ .......... 8 ABSTRACT ................................ ................................ ................................ ................... 10 CHAPTER 1 I NTRODUCTION ................................ ................................ ................................ .... 12 Pharmacokinetics (PK) and PK Models ................................ ................................ .. 12 Bioavailability ................................ ................................ ................................ .... 12 Physicochemical properties ................................ ................................ ....... 13 Physiologic factors ................................ ................................ ..................... 13 Pharmacokinetic Models ................................ ................................ .................. 14 Pharmacodynamics (PD) and PK/PD Models ................................ ......................... 14 Population Pharmacokinetic/Pharmacodynamic Modeling ................................ ..... 16 Summary ................................ ................................ ................................ ................ 18 2 PHARMACOKINETIC PR OPERTIES OF MANGOSTEEN PURE COMPOUNDS OR EXTRACT IN RATS ................................ ................................ 20 Background ................................ ................................ ................................ ............. 20 Materials and Methods ................................ ................................ ............................ 22 Chemic als and Reagents ................................ ................................ ................. 22 Instrumentation and Chromatographic Condition ................................ ............. 22 Preparation of Standards Samples for Validation ................................ ............. 23 Method Validation ................................ ................................ ............................. 24 Pharmacoki netic Study ................................ ................................ ..................... 25 Free and Total Concentration Measurements ................................ .................. 26 Data Analysis ................................ ................................ ................................ ... 27 Results ................................ ................................ ................................ .................... 28 Method Development and Validation ................................ ................................ 28 P harmacokinetics of Pure Compounds after I.V. Administration ...................... 31 P harmacokinetics of Pure Compounds or Extract after Oral Administration .... 32 Discussion ................................ ................................ ................................ .............. 34 3 PHARMACOKINETIC/PHARMACODYNAMIC MODELING OF METHYLPREDNISOLONE AND PRE DNISOLONE IN HUMAN ............................ 54 Background ................................ ................................ ................................ ............. 54 Materials and Meth ods ................................ ................................ ............................ 55
6 Historic Data ................................ ................................ ................................ ..... 55 MP PKPD Models ................................ ................................ ............................. 56 PNL PKPD Models ................................ ................................ ........................... 58 PKPD Analysis ................................ ................................ ................................ 59 Results ................................ ................................ ................................ .................... 61 Pharmacokinetics of MP and PNL ................................ ................................ .... 61 Cortisol Suppression ................................ ................................ ........................ 62 Lymphocy tes Trafficking ................................ ................................ ................... 64 Discussion ................................ ................................ ................................ .............. 65 4 CONCLUSION AND DISCUSSION ................................ ................................ ........ 92 APPENDIX A NONMEM CODE FOR PNL PK MODEL PNL COMP ................................ ............ 95 B NONMEM CODE FOR PNL PK MODEL PNL SIMP ................................ .............. 98 C NONMEM CODE FOR VPC SIMULATION ON CORTISOL SUPPRESSION WITH MODEL COMP ................................ ................................ ............................. 99 D NONMEM CODE FOR PKPD MODEL COMP ON LYMPHOCYTES TRAFFICKING ................................ ................................ ................................ ...... 102 E INDIVIDUAL FITTING PLOTS FOR MP PK DATA ................................ ............... 105 F INDIVIDUAL FITTING PLOTS FOR PNL PK DATA WITH MODEL SIMP ............ 108 G INDIVIDUAL FITTING PLOTS FOR MP CORTISOL DATA WITH MODEL SIMP 111 H INDIVIDUAL FITTING PLOTS FOR PNL CORTISOL DATA WITH MODEL SIMP ................................ ................................ ................................ ..................... 114 I INDIVIDUAL FITTING PLOTS FOR MP LYMPHOCYTE DATA WITH MODEL SIMP ................................ ................................ ................................ ..................... 117 J INDIVIDUAL FITTING PLOTS FOR PNL LYMPHOCYTE DATA WITH MODEL SIMP ................................ ................................ ................................ ..................... 120 LIST OF REFERENCES ................................ ................................ ............................. 123 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 131
7 LIST OF TABLES Table page 2 1 Preparation of working solution of and mangostin ................................ ...... 50 2 2 Preparation of calibration standard curve and QC plasma samples of and mangostin ................................ ................................ ................................ ....... 50 2 3 Preparation of recovery samples of and mangostin ................................ ... 51 2 4 Preparation of matrix samples of and mangostin ................................ ....... 51 2 5 Recovery and matrix validation results of and m angostin .......................... 51 2 6 Calibration curve of and m angostin : constructed by plotting peak area ratio of pure compound to bergamottin to the concentration of the pure compounds. ................................ ................................ ................................ ........ 52 2 7 Intra run (n = 8) and inter run (n = 24) precision for analysis of and m angostin in plasma. ................................ ................................ .......................... 52 2 8 Stability validation results of and m an gostin ................................ ............... 52 2 9 Pharmacokinetic parameters of free, unconjugated and mangostin concentration after a single intravenous dose of 2 mg/kg pure compounds in rats. ................................ ................................ ................................ .................... 53 3 1 Pharmacokinetic parameters o f methylprednisolone (MP) after oral administration. ................................ ................................ ................................ .... 89 3 2 Pharmacokinetic parameters of prednisolone (PNL) after oral administration obtained by Model COMP and Model SIMP. ................................ ...................... 89 3 3 Cortisol suppression pharmacodynamic parameters of methylprednisolon e (MP) and prednisolone (PNL) obtained by Model COMP and Model SIMP. ....... 90 3 4 Predicted cumulative cortisol and lymphocyte suppressi on percentage in 24 hr at steady state from Model SIMP and Model COMP. ................................ ..... 90 3 5 Lymphocytes trafficking pharmacodynamic parameters o f methylprednisolone (MP) and prednisolone (PNL) obtained by Model COMP and Model SIMP. ................................ ................................ ................................ 91
8 LIST OF FIGURES Figure page 1 1 Examples of efficacy and potency comparison by effect concentration profiles ................................ ................................ ................................ ............... 19 2 1 Mangosteen plant pictures from "Fleurs, Fruits et Feuillages Choisis de l'Ile de Java" ................................ ................................ ................................ .............. 42 2 2 Chemical structure s of (A) mangostin ; (B) mangostin ; (C) b ergamottin. ...... 43 2 3 S chematic diagram of t wo compartment body model after i.v. dosing. ............... 44 2 4 The extracted LC MS/MS chromatograms of mangostin, mangostin and bergamottin. ................................ ................................ ................................ ........ 44 2 5 Accuracy validation results of mangostin ................................ ....................... 45 2 6 mangostin. ................................ ........................ 46 2 7 Free, unconjugated concentration time profiles after i.v. administration of 2 mangostin. ....................... 47 2 8 Observed free, unconjugated compounds concentration time profiles after oral administration of mangosteen compounds/extract. ................................ ..... 48 2 9 Observed total compounds concentration time profiles after oral administration of mangosteen compounds/extract. ................................ ............ 49 3 1 Chemical structures of (A) methylprednisolone (MP); (B) prednisolone (PNL); (C) cortisol. ................................ ................................ ................................ ......... 69 3 2 PK/PD model of exogenous methylprednisolone and prednisolone. .................. 70 3 3 Cortisol baseline concentration and production rate (K in ) changes during every 24 hr period. ................................ ................................ .............................. 71 3 4 Diagnostic plots for MP population PK. ................................ .............................. 72 3 5 Visual predictive check (VPC) results for MP population PK model. .................. 73 3 6 Diagnostic plots for PNL population PK. ................................ ............................. 74 3 7 VPC results for PNL PK data from Mod el SIMP. ................................ ................ 75 3 8 VPC results for PNL PK data from Model COMP. ................................ .............. 76
9 3 9 Diagnostic plots for cortisol suppression fitting after administering methylprednisolone or prednisolone ................................ ................................ .. 77 3 10 Individual predictions of each time point from Model SIMP and Model COMP for cortisol concentration (ng/mL) and lymphocyte count (%) ............................ 78 3 11 VPC results from Model SIMP for cortisol suppression after MP treatments. ..... 79 3 12 VPC results from Model COMP for cortisol suppression after MP treatments. ... 80 3 13 VPC results from Model SIMP for cortisol suppression after PNL treatments. ... 81 3 14 VPC results from Model COMP for cortisol suppression after PNL treatments. 82 3 15 Diagnostic plots for lymphocytes fitting after administering methylprednisolone and prednisolone ................................ ............................... 83 3 16 VPC results from Model SIMP for lym phocytes trafficking after MP treatments. ................................ ................................ ................................ ........................... 84 3 17 VPC results from Model COMP for lymphocytes trafficking after MP treatments. ................................ ................................ ................................ .......... 85 3 18 VPC results from Model SIMP for lymphocytes trafficking after PNL treatments. ................................ ................................ ................................ .......... 86 3 19 VPC results from Model COMP for lymphocytes trafficking after PNL treatments. ................................ ................................ ................................ .......... 87 3 20 Lymphocytes placebo data and baseline models. ................................ .............. 88
10 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial F ul fillment of the Requirements for the Degree of Doctor of Philosophy APPL ICATION OF PHARMACOKINETIC AND PHARMACODYNAMIC CONCEPTS I N PRECLINICAL AND CLINICAL STUDIES FOR DRUG DEVELOPMENT By Li Li May 2013 Chair: Hartmut Derendorf Cochair: Veronika Butterweck Major: Pharmaceutic al Sciences Pharmacokinetic (PK) and p harmaco dynamic (PD) concepts have been applied to drug development for decades. This study gives two examples of these applications. Pharmacokinetics is important as it connects in vitro physiological activity with in vivo activity by studying absorption, distrib ution, metabolism and elimination of compounds. Caution must be used when claiming medicinal benefits for active compounds with high in vitro activity but low bioavailability as these compounds may have low in vivo activity. The first project was a mangosteen preclinical PK study and showed that after oral administration, a majority of the pure mangostin) were conjugated by glucuronidation/sulfation. This conjugation process also occurred after a n oral administration of a mangosteen extract. However, when mangosteen was administered as an extract rather than as pure compounds, the time to reach maximum conjugation level was extended from 1 hour to 2 hours. The faster conjugation of the pure compou nds may cause loss of biological activity and may explain the limited reports of in vivo activity. Thus, it is proposed that a high dose and a long treatment period are required to show benefits from mangosteen consumption. Furthermore, PK/PD
11 concepts are not only important for new drug candidates (in this case mangosteen compounds) under exploration for their potential usage, but are also strong tools for understanding drugs that have been in clinical practice for decades and improving their dose regimens. For example, long term use of methylprednisolone and prednis olone will cause side effects. However, most published data are from single dose studies. Thus, the objective of the second project is to understand PK/PD properties of glucocorticoids (methylpre dnisolone and prednisolone) by analyzing data collected from a previously conducted multiple dose study. Twenty four healthy subjects were recruited for the study and treated with twelve different dose regimens for two days in separate blocks with proper w ashout period. The profiles of two biomarkers, cortisol and lym phocyte counts, were obtained. Since the same subjects separately underwent treatment with both methylprednisolone and prednisolone, in separate blocks, direct comparison between methylpredniso lone and prednisolone was feasible. The results showed that PK/PD modeling/simulation approach is successfully applied to predict drug effects of methylprednisolone and prednisolone after multiple dosing in human subjects. M ethylprednisolone is more potent than prednisolone in cortisol suppression and lymphocyte trafficking enhancement. Moreover, although the endogenous glucocorticoid (cortisol) can compete with prednisolone to bind tra n scortin and regulate lymphocyte trafficking, the effects from endogenou s cortisol can be ignored for modeling purposes.
12 CHAPTER 1 I NTRODUCTION Pharmacokinetics (PK) is defined as the study of (liberation), absorption, distribution, metabolism and excretion of drugs. PK can be used to predict drug concentration at any time point after administ rat ion of a certain dose of drug. Pharmacodynamics (PD) is defined as the study of biochemical and physiological effec ts of drugs. PD describes the relationship between drug concentration at the site of action and drug effects corresponding to that concentration. Thus, the combination of PK and PD concepts can predict drug effects at any time point following administrati on of a given dose of drugs. PK/PD can be used as a tool to evaluate the efficacy and safety of compounds, to help in developing new drug candidates or improve drugs that are already on the market. P harmacokinetic s (PK) and PK Models Bioavailability Many n ew drug candidates have shown good in vitro activity but failed to show good in vivo activity. Pharmacokinetics can explain this dis correlation between in vitro activity and in vivo activity via poor bioavailability and/.or short duration of action (AUC, t1/2, etc.). The absolute oral bioavailability is defined as the fraction of an oral dose that reaches the systemic circulation (blood) in comparison to an intravenous (i.v.) dose. A orally, since the compound has to pass through gastrointestinal tract and liver before it reaches systemic blood stream.
13 Physicochemical properties The physicochemical properties of a drug affect how much and how fast the drug is released/absorbed. For ex ample, it has been suggested that poor lipophilicity and large polar surface area are correlated with low oral bioavailability [ 1 ] Solubility is another important factor, as a drug needs to have reasonable aqueous solubility to be absorbed [ 2 ] Another good exam ple showing the relationship between physicochemical properties and bioavailability is the Biopharmaceutic Classification System (BCS). The BCS was developed based on the theory that dissolution, solubility and intestinal permeability are the three major f actors that affect the rate and extent of absorption of a drug [ 3 ] Based on the BCS sys tem, drugs are classified into four categories: Class I with high solubility and high permeability; Class II with low solubility and high permeability; Class III with high solubility and low permeability; and Class IV with low solubility and low permeabili ty. The BCS Class I based waivers in searching for bioequivalent drug candidates for immediate release solid oral dosage forms, as suggested by FDA, was proposed based on the physicochemical PK relationship behind this system [ 4 ] Physiologic factors Gastrointestinal pH may affect how drug is ionized and eventually affects the solubility and permeability of the drug. Also, metabolism enzymes in the intestine and liver play an important role in determining the extent of first pass effect. Extensive first pass effect corresponds to low bioavailability. Thus, factors that change hepatic metabolism will change bioavailability. For example, with l iver disease a dysfunctional liver may lead to impaired drug metabolism and reduced first pass extent [ 5 ] Also,
14 newborn children and the elderly generally have weaker liver functions than adults, so age may play an important role in determine pharmacokinetic profiles [ 6 7 ] Food or other drugs may modify the physiological environment or interact with metabolism enzymes (drug drug inter action and drug food interaction). For example, grapefruit juice increases bioavailability of statins (Lipitor, Zocor, Mevacor), as it can inhibit the cytochrome P450 3A4 [ 8 ] Milk products which contain calcium may chelate with the drug and decr ease drug absorption [ 9 ] O ther examples include high protein meal s that will increase propranolol bioavailability [ 10 ] and high fat meal s may decrease tacrolimus bioavailability [ 11 ] In summary, both physicochemical and physiologic factors can affect the bioavailability of drugs. Pharmacokinetic Models Understanding both the overall bioavailability of drugs as well as the ability to predict drug concentration are equally important. Drug concentration time profiles can be described/predicted by compartment pharmacokineti c models. Generally linear pharmacokinetic model s are used in models, which mean that rate of elimination is proportional to amount of drug present in the body. However, if saturable protein binding and/ or saturable metabolism are present, then nonlinear k inetics may be need to be in troduced into model development Pharmacodynamic s (PD) and PK/PD M odels Pharmacodynamics is the study of the biochemical and physiologic processes underlying drug action, or the relationship between drug and responses. Generally response increases as the drug dose increases, until it reaches its maximum. Based on the occupation theory of drug receptor interactions, an E max model (Equation 1 1) or
15 Sigmoid E max model (Equation 1 2) has been introduced to describe this drug respons e relationship : (1 1) (1 2 ) E max is the maximal response produced by a drug, EC 50 is the concentration of a drug at which 50% of the maximal response will be produced, and r is the hill factor. In the E max model, efficacy is represented by E max and a drug with high E max is more efficacious than a drug with lower E max Potency is represented by EC 50 and a drug with low EC 50 is more potent than a drug with high EC 50 The effect concentration curve descri bed by E max model is shown in Figure 1 1. T h erefore, potency differences can be shown as the relative positions of the curve on the x axis. The curve of Drug A is further left than the curve of Drug C, suggesting that Drug A is more potent than Drug C. Efficacy differences are shown as the height of the plateau of the curve on the y axis. Thus, the plateau of Drug A is higher than Drug B, suggesting Drug A is more efficaci ous. As discussed before, pharmacokinetic/ pharmacodynamics (PK/PD) modeling connects the concentration time profile with concentration effect profile, and therefore can describe/predict the extent of effect with time after a certain dose regimen. T his con nection between PK and PD can be direct or indirect. If there is a time delay between concentration and effect, then the indirect link PK/PD modeling is more likely to be used. This time delay might be caused by drug transference from central compartment t o effect site (distribution process), or by an indirect response mechanism (indirect response PK/PD model).
16 The indirect response modeling has been developed for decades [ 12 14 ] commonly include s four basic models : (1 3 ) (1 4 ) (1 5 ) (1 6 ) The four basic models are models for drugs whose mechanism of action are inhibition or stimulation of a production (production rate K in ) or elimination (elimination rate K out ) physiologic process. I max or S max represents the maximal inhibition or simulation response produced by a drug. IC 50 or SC 50 represents the drug concentration that produce 50% of the maximal inhibition or simulation responses. Population Pharmacokinetic/Pharmacodynamic Modeling The old str ategy of PK/PD modeling analysis begins with individual parameter estimation based on individual data, and then seeks for population values by calculating mean and variability (standard deviation or standard error) of those individual parameters. Another t raditional strategy estimates population parameter values based on mean values of observed data. However, generally both approaches include multiple analysis steps and are hardly to explain the sources of variability among subjects. Population pharmacokine tic/ pharmacodynamics analysis using Nonlinear Mixed Effect Model (NONMEM) was introduced into drug research by Sheiner and Beal in early 1980s [ 15 17 ] to explain data variability Nonlinear Mixed Effect Model (NONMEM) means to use non linear function to estimate fixed effects and random effects
17 simultaneously. Non linear function is very common in drug modeling, as the partial of the dependent variable (Y, in PK would be concentration, in PD would be effect) is based on more than one unknown parameters. For example, the concentration calculation equation for two compartment body model after intravenous administration would be non linear function: (1 7 ) where time t is the independent variable and C p is the dependent variable. The partial of the dependent variable C p ( or ) are based on two unknown parameters (A and or B and ). Generally subjects given the same dose regimen have widely varied drug responses (concentration or effect). This response variability could attribute to the subjects physiological differences caused by age, weight, gender, genetics, etc. NONMEM divides the respons e variability to two categories: fixed effect (measurable variables such as age, weight, gender) and random effects (random variability such as between subject variability (BSV) and residual variability (RV) ) Thus, the advantage of NONMEM is that it estim ates fixed and random effects in one stage and helps explain the sources of variability. During NONMEM analysis, the typical value of a parameter in a population is represented as (Theta). BSV, represented as (Eta), describes the relationship between p opulation parameter ( ) and individual parameter ( i ). Eta ( ) has a distribution with an expected mean as zero and variance as 2 RV, represented as (Epsilon), describes the relationship between individual prediction ( ) and observed response ( ). Epsilon ( ) has a distribution with an expected mean as zero and
18 variance as 2 Thus, an observed response (drug concentration or effect for i th individual at time t j ) is calculated by following equations in NONMEM: (1 8 ) (1 9 ) (1 10 ) Summary In summary, PK describes a drug concentration time profile at a certain dose regimen, and PD describes the drug effect vs. drug concentration profile. The combination of PK/PD concepts can describe a drug effect time profile at a certain dose regimen. Population PK/PD modeling with NONMEM not only predicts drug effects in general as population prediction, but also predicts individual drug responses and helps explain sources of v ariability.
19 Figure 1 1 Examples of e fficacy and p otency comparison by effect concentration profiles
20 CHAPTER 2 PHARMACOKINETIC PROP ERTIES OF MANGOSTEEN PURE COMPOUNDS OR EXTRACT IN RATS Background Garcinia mangostana L. (Clusiaceae) is a tropical evergreen fruit tree also known as mangosteen. The mature fruit has an edible white flesh inside, and an inedible hard purple pericarp outside (Figure 2 1). Mangosteen is believed to originate from Southeast Asia and is mainl y found in Thailand, Burma, Malaya, Indonesia, and Singapore [ 18 ] Traditionally, d ifferent parts of mangosteen, mostly the pericarp, the leaves and the bark have been used for a variety of medical conditions, such as diarrhea, arthritis, dysentery, inflammation, skin disorders, and wound hea ling properties [ 19 ] The major bioactive compounds in mangosteen are xanthone derivatives [ 20 22 ] Up to date, more than 50 xanthones have be en isolated from the pericarp of mangosteen [ 23 27 ] Alpha mangostin are the two major xanthone derivatives in mangosteen extracts, although the percentages of these substances in such extracts depend on the extraction procedure used [ 28 30 ] Like other xanthone mangostin have a tricyclic aromatic ring system [ 31 ] The only structure differen ce mangostin is that a hydrogen atom mangostin is substituted by mangostin (Figure 2 2) Recently, extracts obtained from the pericarp of mangosteen exhibited a variety of medicinal properties in in vitro assays such as antioxidant [ 32 34 ] anti inflammatory [ 35 36 ] anti tumor [ 37 38 ] antibacterial [ 39 ] antiviral [ 40 41 ] mangostin are the two major bioactive compounds in mangosteen pericarp, it is not surprising that most of those in vitro biological activities also presented in pure compounds. For example, b oth mangostin and mangostin have high anti
21 oxidant [ 42 ] and anti inflammatory activity [ 43 45 ] among isolated xanthones. Nakagawa et al. and Chang et al. showed in vitro mangostin and mangostin, respectively [ 46 47 ] lpha mangostin also showed chemoprevention effects [ 48 31 ] Because of the promising in vitro data on biological activities extracts and pure compounds derived from mangosteen have been commercially used as botanical supplement s The consumption of those supplements has been increased dramatically in the United States and mangosteen fruit juice was ranked as one of the top five selling [ 49 ] Not only the juice but also tablets made from extracts or pure compounds can be purchased online and in herbal s hops. For assessment of quality, safety, and efficacy of mangosteen products, there is a need for quantitative data regarding absorption, metabolism, and other pharmacokinetic (PK) properties of mangosteen compounds and extracts Up to now, only a single report on bioavailability of mangostin in humans after ingestion of a xanthone rich liquid preparation has been published [ 50 ] The liquid administered contained mangosteen juice, aloe vera, green tea, compounds labeled as epigallocatechin gallate (EGCG) and multivitami ns. The total amount of xanthones in the product was 94.2 mg/dose. However, the exact amounts of mangostin consumed were not reported. The maximum concentration of a mangostin was 3.12 1.47 ng/mL with a T max of 1 h r after consumption of the supplement. S ince the exact dose of mangostins before administration was not determined and only four time points were chosen for blood collection, the information that can be obtained on the bioavailability of mangostin from this study is limited. In addition, it has to be
22 considered that the supplement contained reasonable amounts of the polyphenol, epigallocatechin galla te, as well as other vitamins. I t can be rationalized that the bioavailability of mangostin might also be influenced by other components in the complex mixture that may act to improve the stability, solubility, or the half life time of the active compounds. Thus, the objective of the present study is to identify the absolute bioavailability of mangostin and compare bioavailability differences when administered as pure compounds or an extract Materials and Methods Chemicals and R eagents Mangostin mangostin and a MeOH extract of the freeze dried powder of the dried and milled pericarp of mangosteen fruits were prepared as described earlie r [ 42 ] A reference standard of mangostin (purity 96.5%) was purchased f rom ChromaDex (Irvine, CA, USA) Bergamottin (purity 98%), which was used as internal standard, was purchased from Indofine Chemical Company, Inc. (Hillsborough, NJ, USA). Acetonitrile (HPLC grade), methanol, formic acid, acetic acid, ascorbic acid and sodium chloride were purchased from Fisher Scientific (Pittsburgh, PA, USA). Ethanol (HPLC grade), Tween 80, glucose, heparin sodium salt (202 unit/mg) and glucuronidase/s ul fatase (Type H 2, 85,000 units /mL ) were o btained from Sigma (St. Louis, MO, USA). Distilled water was purified before use with a Barnstead Nanopure Diamond UV ul tra pure water system (Dubuque, IA, USA). Instrumentation and Chromatographic C ondition The LC MS/MS system consisted of a Surveyor HPLC autosampler, Surveyor MS quaternary pump and a TSQ Quantum Discovery triple quadrupole mass spectrometer (ThermoFisher Scientific, San Jose, CA, USA). The instrument was controlled by
23 Xcalibur software (Vers ion 1.4). The samples were analyzed at room temperature. The TSQ Quantum was equipped with an electrospray (ESI) ionization source and operated in the positive ion mode. The instrument parameters were as follows: spray voltage set to 4.0 k V and source CID 12 V. The instrument parameters included an ion transfer tube temperature of 325 spray voltage of 5.0 kV, and source CID set to 5 V. Nitrogen was used as the sheath and auxiliary gas and set to 35 and 10 (arbitrary units), respectively. The collision en ergy was 27 eV for m angostin and mangostin and 22 eV for bergamottin. The instrument was operated in resolution mode with the peak width (f ul l width at half maximum, FWHM) set to 0.7 m/z both at Q1 and Q3. The selected reaction monitoring scheme follow ed transitions of the precursor to selected product ions with the following values: m/z 411 mangostin, m/z 397 mangostin and m/z 339 202 for bergamottin. Chromatography was perform ed on a Symmetry C18, 4.6 mm x 5 0 mm, 3.5 um analytical column (Waters, Milford, MA, USA). The mobile phase consisted of 0.05 % formic acid in acetonitrile: water (80:20, v/v) delivered at a flow rate of 0.5 m L /min (injection volume 50 mL). Preparation of Standards S amples for Validation The stock solu mangostin mangostin and bergamottin were prepared in acetonitrile at a concentration of 4 mg /mL The internal standard (IS) solution was prepared by diluting the bergamo ttin stock solution in methanol: acetonitrile (25:75, v/v) to produce the final concentration of 6 0 ng /mL Bergamottin was chosen as an IS since it is not present in mangosteen plant material; it is commercially available and has a similar molecular weight as and mangostin From the 20 tested IS candidates, best results were obtained with bergamottin. Working solutions of and mangostin with concentration in the range of 4 1000 g/mL were obtained by diluting the sto ck solution
24 with acetonitrile: water (50:50, v/v) (Table 2 1) The calibration standards were prepared by spiking blank rat plasma (1m L ) with the corresponding working solution (5 10 L ) to yield seven concentrations ranging from 20 to 2000 ng /mL (Table 2 2) Thus, 20 ng/mL concentration samples were the lowest limit of quantification (LLOQ) samples. For validation, quality control (QC) samples were prepared in the same way as the calibration standards at three concentrations (low concentration (LC) = 70, medium concentration (MC) = 850, and high conce ntration (HC) = 1800 ng/ mL) (Table 2 2) Also, dilution QC samples at 10000 ng /mL were prepared and then diluted with blank plasma at 1:10 ratio (called Dil QC samples). Method V alidation The method was f ul ly validated for recovery, matrix effect, linear ity, selectivity precision, and accuracy according to the United States Food and Drug Administration (FDA) Bioanalytical Method Validation Guidance [ 51 ] Three sets of standard samples were prepared to assess linearity, accuracy and precision Each set included two replicates per concentration of calibration standards and eight replicates per concentration of QC ( LLOQ, LC, MC HC Dil QC ) samples. Linearity was determined by replicate analysis of calibration standard samples and a ccuracy and precision were determined by replicate analysis of QC samples For validating matrix and recovery, standard p lasma samples (curve and QCs) ( 50 L) were mixed with 200 L IS solution, and centrifuged at 13.2 x 1000 rpm for 15 min at room temperature. The supernatant was then transferred to autosampler vials and 50 L of sample was injected into the LC MS system for analysis. To prepare recovery samples, 50 L blank plasma samples were mixed with 200 L IS solution, and centrifuged at 13.2 x 1000 rpm for 15 min at room temperature. Supernatant was
25 transferred to new tubes, and then the co rresponding working solution s ( 5 L ) were added to yield three concentrations (LC MC and HC) (Table 2 3) The matrix samples were prepared by spiking organic solvent ( 1m L acetonitrile: water 50:50 ) with the co rresponding working solution ( 10 L ) to yield three QC concentr ations (Table 2 4) After that 50 L samples were mixed with 200 L IS solution to produce matrix samples. Extraction recoveries were determined by comparing the peak areas of QC samples with those of the recovery samples Matrix effects were determined by comparing the peak areas of the recovery samples with those of the matrix samples Six determinations per concentration were applied in assessing recovery and matrix effects. The f reeze thaw stability was determined by thawing QC (LC and HC) samples at room temperature and re freezing at 20 for three cycles, and co mpound concentrations were measured during each cycle Also, QC samples were stored at room temperature for 6 hr and 24 hr before processing to assess plasma stability. Post preparative samples were stored in room temperature and refrigerator for 24 hr bef ore measurement The stability was assessed by comparing the peak area ratio of mangostins to bergamottin from above stability sample groups to the ratio from fresh process and analyzed group. Pharmacokinetic S tudy The male Sprague Dawley rats with jug ul ar vein catheters (weighing between 320 350g) were purchased from Charles River (Wilmington, MA, USA). The animals were single housed in plastic cages and received a standard chow and water ad libitum during the experiments. All the animals were maintained o n a 12hr/12hr light/dark cycle. Non fasted animals were used for the study. All animal experiments were performed according to the policies and guidelines of the Institutional Animal Care and Use
26 Committee (IACUC) of the University of Florida, Gainesville, USA (NIH publication # 85 23), study protocol # 200802291. The same rats were used for the intravenous (i.v.) and oral study for same compound mangostin or m angostin pure comound at 2 mg/kg i.v. and after a washout period of one week 20 mg/kg of same compound orally. In another group rats ( n = 8) 160 mg/kg of the mangosteen extract was administered orally. M angostin m angostin and the mangosteen extract w ere dissolved in an aqueous solution containing 2% ethanol and 2% Tween 80. Blood samples (500 L ) were collected from the sublingual vein into Vaccuette heparinized tubes 0 (prior to dosing), 2, 5, 10, 20, 30 min, 1, 2, 4, 6 9 hr. The loss of blood volume was replaced with 1 m L normal saline. Blood samples were centrifuged at 4000 rpm for 15 min at 4 Plasma samples (PK samples) were then transferred into 1.5 m L tubes and stored at 20 until analysis. Free and Total Concentration Measurements To measure free concentration, 50 L plasma samples were mixed with 200 L IS solution, and centrifuged at 13.2 x 1000 rpm for 15 min at room temperature. The supernatant was then transferred to autosampler vials and 50 L of sample was injected into the LC MS system for a nalysis. Enzyme buffer was prepared with water, ascorbic acid (0.5% m/v), and acetic acid glucuronidase/sulfatase or water was added. The mixture was incubated at 37 standard (IS) solution (bergamottin 60 ng/mL in acetonitrile:methanol 3:1), and
27 centrifuged at 13.2 x 1000 rpm for 15 min at room temperature. The supernatant was then LC MS/MS system for analysis. Data A nalysis Mean plasma concentrations versus time curve were generated in Graphpad Prim (version 5.01, San Diego, CA, USA). The pharmacokinetic (PK) parameters were determined by non compartmental and compartmental analysis using WinNonlin software (version 5.2.1, Pharsight Corporation, USA). P harmacokinetic p arameters both from non compartmental and compartmental were under descriptive statistical analysis using WinNonlin software to calc ul ate the mean and standard error (SE ). Non compartmental PK: The PK parameters determined were the concen tration at time zero (C 0 ), the terminal elimination rate constant ( K e ) the terminal elimination half life (t 1/2 ), the area under the curve (AUC), the MRT, the volume of distribution at terminal phase (V z ), the volume of distribution at steady state (V ss ), the central compartment volume of distribution (V c ), the peripheral compartment volume of distribution (V p ), and the clearance (CL). AUC 0 l ast was calc ul ated using linear/log trapezoidal method from time zero to last sampling point 6 hr after administration. The terminal phase to calc ul ate K e was specified to the range from 0.5 hr to 6 hr after administration, and K e was calc ul ated by linear regression of the terminal log linear phase. AUC 0 was calculated by adding C last /K e to AUC 0 l as t (C last : concentration at last time point). Compartmental PK: The free compound concentration time profile after i.v. administration was teste d in two compartment body model (Figure 2 3) and three compartment model with different weighting factor s and the go odness of fit was
28 determined by the AIC (Akaike Criteria), SC (Schwartz Criteria) and WSSR (Weighted sum of square of residuals). The lower the AIC, SC, and WSSR the more appropriate is the selected model. Pharmacokinetic parameters were calc ul ated using t he selected two compartment body model with the Gauss Newton minimization method and 2 The equation for two compartment model is as followed: C = A e + B e where C is the concentration of drug in plasma at time t; A and B are constant; and t is the time. Res ul ts Method Development and V alidation The LC MS/MS method was developed to determine mangostin and m angostin concentration in rat plasma samples. Standard compounds ( mangostin and m angostin ) were spiked into b lank rat plasma for validation and the corresponding chromatogram is shown in Fi gure 2 4 In Fig ure 2 4 C plasma samples spiked with bergamottin as inte rnal standard ( IS ) are displayed. The retention times were approximately 4. 4 min, 3.2 min, and 4.9 min for mangostin m angostin and bergamottin, respectively. The method was validated by evaluating recovery, matrix effect, linearity, precision, accuracy, and stability [ 51 ] The extraction recovery of mangostin and m angostin using protein precipitation was expressed as the ratio of the mean peak area of extracted plasma QC samples (LC, MC, and HC) to the mean peak area of the blank plasma samples spiked with compound after extraction. The recovery percentages for mangostin were 90.8 % 92.1 % and 96.6% at concentrations of 70, 850, and 1800 ng/mL, respectively The recovery percentages for mangostin were 87.7% 78.1%,
29 and 83.8% at concentrations of 70, 850, and 1800 ng/mL, respectively (Table 2 5 ). The matrix effect of mangostin and m angostin in plasma was expressed as the ratio of the mean peak area of blank plasma sa mples spiked with compound at three level QC concentrations after extraction with the mean peak area of the injected solution spiked with compound at the same concentration. No significant matrix effects were observed for mangostin (123.1%, 118.6% and 11 2.3% for LC, MC, and HC ) but these were evident for bergamottin in rat plasma (171.3%) (Table 2 5 ). The matrix effects for mangostin were about half fold of the matrix effects for mangostin (95.7%) and same fold of reduction was found for bergamottin when validating for mangostin (46.6%, 41.1% and 49.7% for LC, MC, and HC ) (Table 2 5 ). The calibration curves of mangostin and mangostin were constructed by plotting peak area ratio of pure compound to bergamottin to the concentratio n of the pure compound Calibration curves for mangostin and were linear within the concentration range of 20 2000 ng/mL with a mean SD correlation coefficient (R 2 ) of 0.989 0.003 5 Calibration curves for mangostin were linear within the concentration range of 20 2000 ng/mL with a mean SD correlation coefficient (R 2 ) of 0.9 92 0.00 55 (Table 2 6 ) The LLOQ for mangostin and mangostin in plasma was defined as 20 ng/mL, as it was the lowest concentration on the calibration curve. To extend the highest limit of the calibration curve, Dil QC samples were prepared by spiking blank plasma with pure compounds at 10 000 ng/mL and then diluted with blank plasma at 1:10 ratio. The accuracy described the closeness of the mean analyzed results obtained by the method to the true value (concentration) of the compounds [ 51 ] The mean value
30 should be within 15% of the actual value except at LLOQ, where it should not deviate by more than 20%. Accuracy was determined by eight replicate s of five different concentrations ( LLOQ, LC, MC, HC, and Dil QC) for 3 sets As shown in Figure 2 5 and Figure 2 6, minimum of 67% of the eight determinations were in the desired range for mangostin and m angostin for all five QC concentrations The precision describes the close ness of individual measures to its mean results when measuring multiple aliquots of the same concentration samples [ 51 ] The precision determined at each concentration level should not exceed 15% of the relative standard deviation ( % RSD) except for the LLOQ, where i t s %RSD should not exceed 20%. Precision is further subdivided into i ntra run precision which assesses precision during a single analytical run (calculate %RSD within the set) and inter run precision, which measures precision with time (calculate %RSD among dif ferent set) At the LLOQ, the i ntra and inter run precision s (%RSD) for mangostin were 2.0%, 14.4%, and m angostin 15.4% and 8.7%, respectively. The in tra and int er run precision for mangostin and m angostin at LC, MC, HC, and Dil QC concentrations are summarized in Table 2 7 and they were all less than 15% The freeze t haw stabilit y, postpreparative stability and plasma stability were assessed by performing eight replicate analyses of LC and HC samples for mangostin and m angostin The peak area ratio of mangosteen compounds to bergamottin of those stability samples had no significant differences comparing to the plasma samples analyzed freshly ( stored in the freezer and extracted prior to analysis ) ( Table 2 8 ). Overall, the stabilit y of mangostin and m angostin w ere found to be reasonable.
31 P harmacokinetics of Pure Compounds after I.V. Administration The validated method was us ed successfully to support the p harmacokinetic (P K ) stud ies of mangostin and m angostin in rats after a single i.v. dose of 2 mg/kg pure compounds The free, unconjugated plasma concentration versus time profile s after i.v. administration of and m angostin were biphasic, subdivided into a distribution phase and a slow elimination phase, as shown in Fig ure 2 7 Non compartmental and compartmental analyses have been performed for mangostin and m angostin using WinNonlin software (version 5.2.1, Pharsight C orporation, St. Louis, MO, USA) The two compartment body model with the weighting factor 1/Y 2 described the i.v. data of mangostin and m angostin wel l, as the AIC, SC, and WSSR values of two compartme ntal body model were smaller than the ones of three compartmental body model Also, the goodness of fit can be seen in Fig ure 2 7 whic h shows the concentrations simulated by the two compartment body model fit the observed concentrations Furthermore, the PK parameters obtained by compartmental analysis showed a good agreement with the PK parameters obtained by non compartmental analysis (Table 2 9). Following i.v. administration, the maximum concentration was 1 4 08 mg/mL for mangostin and 5.87 mg/mL for m angostin at time 0 calculated from non compartmental analysis and 10.35 mg/mL for mangostin and 5.83 mg/mL for m angostin from compartmental analysis respectively Two compartment body model has two half lives, one for distribution ( t 1/2 ) and one for terminal elimination ( t 1/2 ) T he values of t 1/2 from two compartment analysis were similar to the one from non compartmenta l analysis. Also, the values of AUC 0 MRT, c learance, volume of distribution from two compartment body model were similar to the values obtained from
32 non compartmental analysis. Additional proof of the presence of peripheral compartment is that both mangostin and m angostin had the following relationship on volume of distribution : V z > V ss > V c Although mangostin and mangostin have very similar structure, their pharmacokinetic properties are different in certain points. Alpha m angostin had hig her C 0 larger AUC 0 longer MRT, larger clearance and longer t 1/2 than m angostin It had same t 1/2 same K 10 and same K 12 /K 21 ratio as mangostin. Moreover, m angostin had similar volume of distribution in V ss V p V z but slightly different V c as mangostin (Table 2 9) P harmacokinetics of Pure Compounds or Extract after Oral Administration A single oral dose of 20 mg/kg pure compound was also given to the same rats after a washout period of 1 w ee k. Compared to the free unconjugated mangostin level after i.v. administration, the free mangostin level after oral administration of 20 mg/kg was very low as it was below LLOQ 10 min after administration and remained at this low level at least until 12 h after administration (Fig ure 2 8A ). Also, the free mangostin level after oral administration was relatively low, as the maximum observed free concentration was below 100 ng/mL and below LLOQ after 2 hr (Fig ure 2 8B ). Thus, it was not possible to obtain full unconjugated concentration time profile s for or mangostin after oral administration to calculate the value of absolute bioavailability. To equalize the dose of a mangosteen extract to the pure compounds, the mangostin in the extract were quantified. The per centages (Mean mangostin in the extract were 12.6 0.14% and 2.8 0.02%, respectively. Therefore, a 160 mg/kg dose of the mangosteen extract was administered
33 mangostin an d 4.5 mangostin level was still very low in rat plasma, but it was possible to detect the free unconjugated concentration peak for 4 hr after adm inistration of the extract (Fig ure 2 8C ). mangosti n could not be measured at 10 min after oral administration of 20 mg/kg pure compound, after treating the same plasma samples mangostin concentration was higher than 100 ng/mL at 10 min and continued to increase until it reached the maximum level at 1 hr (Fig ure 2 9 A). This finding indicates that after oral administration only very low amounts man gostin seems to be rapidly conjugated b y UDP glucuronosyltransferases and/or sulfotransferases. Considering the structural similarity between mangostin and mangostin, it can be assumed that the low bioavailability of mangostin after oral administrati on is also caused by rapid conjugation. As shown in Fig ure 2 9 mangostin concentration after 20 mg/kg pure compound was higher than 100 ng/mL at 5 min and had the same T max mangostin. Fig ure 2 9 C and 2 9 D show the mangostin after oral administration of 160 mg/kg mangosteen fruit extract, respectively. It can be seen that the time to reach the maximum concentration (T max ) was delayed for the extract as the T max changed from 1 hr after pure compound administration (Fig ure 2 9 A and 2 9 B) to 2 hr after extr act administration ( Fig ure 2 9 C and 2 9 mangostin. Also, the elimination patterns of the total concentration were changed from two phases (Fig ure 2 9 A and 2 9 B ) after pure compound administration to one phas e (Fig ure 2 9 C and 2 9 D) after
34 mangostin. The terminal phase half lives of the total concentration when given as the pure compounds were significantly larger than the ones when given as the extract for both (18.5 mangostin (5.33 hr vs 2.49 hr). Furthermore, area under curve (AUC) values of the total concentration time profiles were calculated for the pure compounds and the extract by linear/log linear mangostin was the same when administered as a pure compound or when present in the extract (20 mg/kg when given as pure compound, 160 mg/kg as extract, equivalent to 20 mg/kg of the pure mangostin was different (20 mg/kg when give n as pure compound but 4.5 mg/kg when given as the extract). Therefore, to equalize the dose differences, the mean of total AUC/DOSE was calculated. The mean of total mangostin when given as pure compound (240 g*hr/mL) was the same as that w hen given as the extract (240 g*hr/mL). Also, the mean of the total mangostin when given as pure compound (471 g*hr/mL) was not significantly different from the one when given as the extract (452 g*hr/mL). Since the clearance is equal to D mangostin were not significantly different for the pure compounds and the extract. Discussion Although in recent years the number of in vitro studies investigating the pharmacodynamic effects of prepa rations obtained from G. mangostana as well as selected xanthones such as mangostin and mangostin has increased rapidly, there is still limited information available regarding their PKs. The lack of PK data might be due to the following reason: the study of herbal PKs in general is extraordinarily complex because herbal preparations are multicomponent mixtures that contain
35 numerous phytochemicals. Therefore, concentrations of single compounds in the final product are in the lower mg range per dose. T he resulting plasma concentrations are often in the mg to pg per liter range. As a consequence, analytical methods determining bioavailability and PKs of herbal preparations have to be sufficiently sensitive In the present study, a simple, rapid, and ful ly validated LC MS/MS method has been developed for the determination of mangostin and mangostin in rat plasma. To our knowledge, this is the first time that a concentration time profile of mangostin and mangostin after i.v. administration has been obtained. Our data show that the PK profile s of mangostin and mangostin are biphasic, including a rapid distribution phase and a slow elimination phase. Both compounds fitted the two compartmental body model after i.v. administration. However, there a re some differences in their pharmacokinetic parameters. After a 2 mg/kg i.v. dose, the values of C 0 and AUC 0 for mangostin. While the distribution half life ( t 1/2 ) mangostin (0.04 hr) was not significantly different from the one mangostin (0.05 hr), the terminal half life ( t 1/2 ) mangostin (1.52 hr) was less than half of the terminal half mangostin (3.46 hr). This could be explained by higher clearance mangostin (2.87 L/hr/kg vs 1.70 L/hr/kg) and a similar total volume of distribution (V c + V p mangostin (3.99 L vs. 4.15 L). mangostin seems to be distributed to the peripheral compartmen t in the same mangostin. Chen et al. [ 52 ] mangostin but mangostin significantly inhibited the paw edema in mice. The authors also mangostin had a higher anti mangostin
36 in vitro On e possible explanation for this in vivo and in vitro activity difference could be mangostin with a higher clearance was eliminated faster, and therefore had lower in vivo anti mangostin even with higher in vitro activity mangostin occurred in relatively lo w concentration levels in the f r ee unconjugated mangostin, mangostin concentration could be mangostin after oral administration of a mangosteen product has been reported by Chitchumroonchokchai et al. [ 53 ] The present results confirmed that after oral administration as pure compoun ds, mangostin occurs in the body as glucuronide and/or sulfate conjugates. Rapid and extensive conjugation has been reported also for other polyphenolic compounds, such as quercetin and luteolin [ 54 56 ] Since the free, their total concentrations administere d, the analysis of the total concentrations would more properly represent the pharmacokinetic properties of these conjugates. The half life for the conjugates after pure compound oral administration was 18.5 hr and 5.33 hr mangostin conjugates respectively. These half lives of the conjugates were longer than half lives of each free compound from the non compartmental analysis mangostin). This is not surprising since most glucuronide/sulfate co njugates will undergo enterohepatic recirculation and thus they will be re absorbed from the bile to the intestine and back into the bloodstream again, causing half life prolongation. mangostin had a higher maximum total concentration and mangostin after oral
37 administration of identical doses. A possible explanation for this observation could be mangostin has one more functional hydroxyl group that can react with UDP glucuronosyltransferase, thus it is easie mangostin. For most cases, glucuronidation and sulfation will result in loss of biological activity [ 57 ] and the resulting conjugate molecule is substantially less lipophilic than its precursor so it is more readily excreted in the bile or urine. However, in past decades it also has been shown that glucuronidation increased/enhanced biological act ivity of morphine and steroids [ 58 ] Thus, it is questionable if and how biological activities of mangosteen xanthones will change after conjugation. So far, more than 30 studies hav e been reported discussing a variety of in vitro biological activities of mangosteen compounds or related xanthones [ 19 23 ] However only two in vivo studies have been reported for anti inflammatory properties and one in vivo studies for anti cancer activity mangostin). As discussed before, Chen et al. mangostin significantly prevented carrageenan (CG) induced inflammation by reducing induced foot pad thickness in mice [ 52 ] On the other hand, it has been shown that intraperitoneal injection of low doses o f mangostin (up to 20 mg/kg) was inactive against xenograft models [ 59 ] In the same year, Sampath et al. [ 60 ] sh owed that after 200 mg/kg/day oral administration for 6 days mangostin can protect against ISO induced myocardial infarction in rats. Shibata et al. [ 61 ] demon mangostin reduced tumor growth and lymph node metastasis in a metastatic mammary cancer mouse mangostin via subcutaneously implanted mini osmotic pumps at 20 mg/kg/day for 7 weeks. No other in vivo act ivity has been
38 mangostin. Therefore, although there is no direct activity test on xanthone conjugates were not tested directly in pharmacodynamics mangostin may lose their activity after in vivo conjugation, causing a lack of in vivo activity reports unless high doses and long treatment periods are used. Although there has been only two stud ies reported to date regarding the anti tumor in vivo activity of a mangosteen pure xanthone, there hav e been five reports describing the in vivo activity of mangosteen mixtures regarding their putative anti tumor or chemopreventive effects. I n particular, Nabandith et al. [ 62 ] suggested that a crude mangostin) has a cancer chemopreventive effect in rats. They administered a crude G. mangostana fruit extract as (0.05% diet continuously for 5 weeks), which led to significant inhibition of the dimethylhydrazine (DMH) induction of several proliferation factors: aberrant crypt foci (ACF), dysplastic foci (DF) and beta catenin accumulated crypts (BCAC). More recentl y, Watanapokasin et al. [ 63 ] showed the cytotoxic eff ect of another mangosteen ethyl mangostin) after intratumoral administration using a mouse subcutaneous tumor model. Doi et al. [ 64 ] also demonstrated the antitumor growth activity of a xanthone mixture (75 mangostin and 5 mangostin) as a diet in a mouse metastatic mammary cancer model. Furthermore, in a recent study Kosem et al. [ 65 ] reported that a crude methanolic mangostin) significantly reduced tumor size as well as increased median survival time and life span of tumor bearing mice after intraperitoneal injection for 14 days. Moreover, a recently published paper suggested that long term feeding of
39 mangostin per kg diet significantly reduced tumor volume and tumor mass in a xenograft mouse model [ 66 ] One possible explanation for the increased in vivo activity reports from mangosteen fruit extracts is that the low bioavailability of active compounds after pure compound administration is improved by being administered with other co effectors when given as an extract. Our data compared the pharmacokinetic differences between the pure compounds and a G. mangostana fruit extract. The results suggest that, when given as a mangosteen extrac mangostin absorbed into the body is not increased, as the AUC does not change significantly. However, the mangostin after given in an extract were different from th ose of the pure compounds, as the elimination phases of the conjugates changed from two phases for pure compounds to one phase for the extract. Also, the half lives of the total concentration after extract oral administration were reduced to 2.51 hr and 2. mangostin, respectively (18.5 hr and mangostin, respectively). Since conjugates have longer half lives than free compounds, one hypothesis is that other components in the extract may inhib it/reduce glucuronidation/sulfation of the mangostins by increasing the free compound exposure and decreasing conjugates exposure in the total concentration profile. Another hypothesis is that the enterohepatic recirculation of the conjugates is inhibited by other components, leading to a short half life when given as an extract. We consider that the first hypothesis, inhibition of the conjugation, is more likely, because of the following reasons: i mangostin were higher
40 after an oral extract dose (detectable until 4 hr) than those after an oral pure compound dose (not detectable after 10 min); ii The T max of the total concentration (mainly conjugates) was delayed from 1 hr to 2 hr when given as the extract; iii The clearance of total concentration (mainly conjugates) was not significantly different between each pure compound and extract. Moreover, the inhibition of the UDP glucuronosyl transferase (UGT) enzymes by phytochemicals in herbal supplements has been reported [ 67 ] Therefore, it may be speculated that the higher number of in vivo anti tumor activity reports when administered as mixtures ( five for extracts /mixtures vs one for mangostin) may come from the slower conjugation of the free compound at the beginning and subsequen tly a longer exposure of the free active compound when given in extract form. However, no direct comparison between the pure compounds and the extract has been carried out regarding the in vivo activity. In conclusion, this study represents the first eval uation of the pharmacokinetics of and mangostin in rats after i.v. and oral administration. It further provides the first comparison of the pharmacokinetic profile of pure mangostin derivatives in comparison to an extract. Mangostin and m angostin show the two compartmental i.v. pharmacokinetic profile, indicating that both compounds will distribute into the peripheral compartment. In addition, both mangostin undergo an intensive first pass metabolism and the majority of compounds are conju gated rapidly after oral administration. Other components in the mangosteen extract did not increase the mangostin, but affect the metabolism of these compounds by slowing down the conjugation at the beginning and therefore increasin g the free
41 compound exposure in vivo Since beneficial biological activities of mangostins seem to be based on the free compound, it is recommended that food supplements made from mangosteen extracts should be preferred to pure compound products, as they p rovide longer free compound exposure and subsequently higher in vivo activities. However, further investigations are required to directly compare the in viv o activity between pure compounds and extracts to make the final conclusion.
42 Figure 2 1. Mangosteen plant pictures from "Fleurs, Fruits et Feuillages Choisis de l'Ile de Java" 1863 1864 by Berthe Hoola van Nooten (Pieter De Pannemaeker lithographer) (http://en.wikipedia.org/wiki/Purple_mangosteen)
43 Figure 2 2 Chemical structure s of (A) mangostin ; (B) mangostin ; (C) b ergamottin.
44 Figure 2 3 S chematic diagram of t wo compartment body model after i.v. dosing. Compound was disposed directly into central compartment. K 10 is the elimination rate constant from central compartment. K 12 is the transfer rate constant from central compartment to peripheral compartment. K 21 is the transfer rate constant from peripheral compartment to central compartment. Figure 2 4. The extr acted LC MS/MS chromatograms of mangostin, mangostin and bergamottin. ( A ) standard plasma sample of mangostin (concentration = 2000 ng/mL) ( B ) plasma sample from a rat obtained 5 min after mangostin (concentration = 12 15 ng/mL) and ( C ) bergamottin (60 ng/mL) as the internal standard.
45 Figure 2 5 A ccuracy validation results of mangostin Black line is the desired concentration and the block between blue lines is the desired range that QC samples should be in (80% 120% for LLOQ and 85% 115% for all other QCs). Circles are QC samples falling into the range and red stars are QC samples out of the range.
46 Figure 2 6 A ccuracy validation results of mangostin. Black lin e is the desired concentration and the block between blue lines is the desired range that QC samples should be in (80% 120% for LLOQ and 85% 115% for all other QCs). Circles are QC samples falling into the range and red stars are QC samples out of the range.
47 Figure 2 7 Free, unconjugated concentration time profiles after i.v. administration of 2 mg/kg pure compound of (A) m mangostin Lines represent mean concentration p redicted by two compartmental body model Rectangle symbols represent observed con centration (Mean and SE)
48 Figure 2 8 Observed f ree, unconjugated compounds concentration time profiles after oral administration of mangosteen compounds/extract (A) Mangostin and mangostin free, unconjugated plasma co ncentration time profile (Mean and SE) in rats after a single oral dose of 20 mg/kg pure compound. ( C Mangostin free, unconjugated plasma concentration time profile (Mean and mangostin dose = 20 m g/kg).
49 Figure 2 9 Observed t otal compounds concentration time profiles after oral administration of mangosteen compounds/extract mangostin total plasma concentration time profile (Mean and SE) in rats after a singl mangostin total plasma concentration mangostin dose mangostin dose = 4.5 mg/kg).
50 Table 2 1. Preparation of working solution of and mangostin Start conc. Add volume ( L ) ACN:H2O(50:50) volume ( L ) Working solution conc. ( g/mL ) Compound s tock conc. 1mg /mL Add 300 ul 4mg /mL + 900 ul ACN:H2O = 1.2m L 1mg /mL To add 1mg/mL solution 1000 0 1000 100 400 200 90 410 180 100 900 100 90 910 90 85 915 85 42.5 957.5 42.5 10 990 10 To add 1 00 g/mL solution 70 930 7 40 960 4 35 965 3.5 Table 2 2. Preparation of calibration standard curve and QC plasma samples of and mangostin Final conc. (ng /mL ) Plasma volume (m L ) Working solution conc. ( g/mL ) Add volume ( L ) ) Standard c urve 20 1 4 5 50 1 10 5 100 1 10 10 500 1 100 5 1000 1 100 10 1500 1 200 7.5 2000 1 200 10 QCs 70 1 7 10 850 1 85 10 1800 1 180 10 100 00 1 1000 10
51 Table 2 3. Preparation of recovery samples of and mangostin Final conc. (ng/m L ) Supernatant volume (m L ) Working solution conc. ( g /m L ) Add volume ( L ) 7 0 about 0.25 3.5 5 850 about 0.25 42 .5 5 1800 about 0.25 90 5 Table 2 4. Preparation of matrix samples of and mangostin Final conc. (ng/m L ) Organic s olvent volume (m L ) Working solution conc. ( g /m L ) Add volume ( L ) 70 1 7 10 850 1 85 10 1800 1 180 10 Table 2 5. Recovery and matrix validation results of and m angostin Nominal concentration (ng/mL) Extraction recovery (n = 6) Mean SD Matrix effect (n = 6) Mean SD M angostin 70 90.8% 8.8% 123.1% 15.6% 850 92.1% 8.4% 118.6% 16.7% 1800 96.6% 5.1% 112.3% 13.1% Bergamottin 200 99.6% 8.0% 171.3% 16.6% M angostin 70 87.7% 10.0% 46.6% 5.1% 850 78.1% 5.8% 41.1% 4.1% 1800 83.8% 2.5% 49.7% 6.6% Bergamottin 200 93.8% 10.6% 95.7% 10.2%
52 Table 2 6. Calibration curve of and m angostin : constructed by plotting peak area ratio of pure compound to bergamottin to the concentration of the pure compound s. Equation R 2 M angostin Y = 0.000437+0.000099*X 0.991 Y = 0.000412+0.000094*X 0.985 Y = 0.000467+0.000103*X 0.991 M angostin Y = 0.000772+0.000463*X 0.992 Y = 0.000693+0.000493*X 0.987 Y = 0.000892+0.000698*X 0.998 Table 2 7. Intra run (n = 8) and inter run (n = 24) precision for analysis of and m angostin in plasma Data were presented as the relative standard deviation (%RSD) to sample mean. Quality control samples (ng/m L ) (n=8) 20 70 850 1800 10000 M angostin Intra run precision (%RSD) 2.0% 12.8% 8.9% 7.9% 7.6% Inter run precision (%RSD) 14.4% 13.2% 9.1% 8.6% 8.6% M angostin Intra run precision (%RSD) 15.4% 6.6% 12.4% 5.8% 8.0% Inter run precision (%RSD) 8.7% 8.8% 9.9% 10.3% 8.3% Table 2 8. Stability validation results of and m angostin Data were ratio of stability samples to fresh made/analyzed samples and presented as Mean (Standard deviation, SD) M angostin Mean (SD) M angostin Mean (SD) 20 ng/m L 1800 ng/m L 20 ng/m L 1800 ng/m L F reeze/thaw 1 105% (7.8%) 86.6% (1.4%) 107.5% (7.5%) 97.5% (7.6%) F reeze/thaw 2 119.1% (11.9%) 100.8% (6.3%) 107.2% (9.8%) 89.3% (4.7%) F reeze/thaw 3 98.2% (10.7%) 86.8% (4.7%) 106.5% (9.5%) 85.4% (3.2%) Plasma 6 hr room 122.3% (10.7%) 100.2% (10.5%) 88.6% (5.9%) 91.6% (4.8%) Plasma 24 hr room 95.9% (3.4%) 110.1% (7.0%) 87.5% (9.7%) 80.9% (5.0%) Post preparative 24 hr room 102% (6.4%) 87.8% (10.2%) 107.4% (2.4%) 104.4% (3.8%) Post preparative 24 hr fridge 97.8% (5.8%) 122.7% (16.9%) 111.2% (11.5%) 108.8% (3.8%)
53 Table 2 9. Pharmacokinetic parameters of free, unconjugated and mangostin concentration after a single intravenous dose of 2 mg/kg pure compounds in rats. Non compartmental and compartmental analyses were performed using WinNonlin and presented as Mean ( S tandard Error S E). For compartmental analysis, data were fitted to a two compartment body model. Parameters M angostin Mean (SE) M angostin Mean (SE) Non compartmental Analysis Compartmental Analysis Non compartmental Analysis Compartmental Analysis C 0 (ng/mL) 14080 (2635) 10353 (1879) 5872 (856) 5836 (1082) AUC 0 (ng*hr/mL) 1185 (71) 1237 (111) 720 (45.5) 708 (36.4) MRT (hr) 2.12 (0.42) 2.36 (0.62) 1.48 (0.06) 1.44 (0.03) CL (L/hr/kg) 1.72 (0.03) 1.70 (0.13) 2.85 (0.17) 2.87 (0.13) V ss (L/kg) 3.69 (0.28) 3.99 (0.33) 4.27 (0.20) 4.15 (0.11) V c (L/kg) 0.17 (0.01) 0.23 (0.03) 0.39 (0.02) 0.41 (0.06) V p (L/kg) 3.52 (0.24) 3.76 (0.91) 3.89 (0.19) 3.74 (0.22) V z (L/kg) 7.85 (0.39) 8.40 (0.57) 8.10 (0.69) 6.29 (0.12) t 1/2 (hr) 0.05 (0.01) 0.04 (0.01) t 1/2 (hr) 3.15 (0.39) 3.46 (0.74) 1.95 (0.06) 1.52 (0.04) K e (1/hr) 0.24 (0.03) 0. 2 6 (0.05) 0.36 (0.01) 0. 46 (0.01) K 10 (1/hr) 8.24 (0.97) 8.13 (1.35) K 12 (1/hr) 5.81 (0.66) 15.6 (5.12) K 21 (1/hr) 0.45 (0.07) 1.33 (0.14) C 0 = the concentration at time zero; AUC 0 = area under the curve with extrapolation; MRT = Mean residence time; CL = clearance; V ss = volume of distribution at steady state; V c = volume of distribution of central compartment; V p = volume of distribution of peripheral compartment; V z = volume of d istribution at terminal phase; t = half life of distribution phase; t = half life of elimination phase; K e = elimination rate constant; K 10 = elimination rate constant of central compartment; K 12 = transfer rate constant from central compartment to peripheral compartment; K 21 = transfer rate constant from peripheral compartment to central compartment
54 CHAPTER 3 PHARMACOKINETIC/PHAR MACODYNAMIC MODELING OF METHYLPREDNISOLONE A ND PREDNISOLONE IN HUMAN Background Methylprednisolone (MP) and prednisolone (PNL) are synthetic glucocorticoids. Cortisol is the endogenous gl ucocorticoid in humans (Figure 3 1 ). Exogenous glucocorticoids have been used as treatments in a wide variety of inflammatory and allergic conditions, such as rheumatoid diseases (rheumatoid arthritis, systemic lupus erythematosus), lung disease (asthma), inflammatory bowel disease, and skin diseases [ 68 ] The first sign of glucocorticoid administration is decrease of cortisol concentration, as exogenous glucocorticoids have negative feedback regulation on the hypothalamic pituitary adrenal (HPA) axis, which eventually suppres ses cortisol production [ 69 ] Another glucocorticoid drug effect is en hancement of lymphocyte trafficking and decrease of lymphocyte number in blood [ 70 ] Long term treatment with glucocorticoids causes a series of adverse effects, such as adrenal insufficiency [ 71 ] It has also been suggested that exogenous glucocorticoids lead to osteoporosis via both direct and indirect effects [ 72 ] Moreover, although short term treatment with glucocorticoids increases in sulin synthesis and improves glucose sensitivity, long term treatment has the opposite effects, decreasing glucose stimulated insulin secretion [ 73 ] Thus, dosage regimen optimization for exogenous glucocorticoids (such as methylprednisolone and prednisolone) is desired. Pharmacokinetic pharmacodynamics (PK/PD) modeling is an efficient approach to optim ize dosage regimen, as it can predict the drug effect time profile by connecting the drug concentration time (PK) profile with the concentration effect (PD) profile. PK/PD properties of methylprednisolone and prednisolone after single doses have been
55 studi ed thoroughly. After oral administration, both methylprednisolone and prednisolone have high oral bioavailability [ 74 78 ] The protein binding properties of methylprednisolone and prednisolone played an important role in PK m odeling, especially for prednisolone, as nonlinearity is caused when PNL bind s to saturable transcortin [ 79 81 ] Although total drug concentration was measured in most cases, the free drug concentration profile can be described by the one compart ment body model [ 80 82 87 ] For PK/PD modeling, the indirect response model has been used for glucocorticoid drug effects on cortisol suppression and lymphocyte trafficking, and parameters have been reported [ 70 79 83 85 94 ] However, very few multiple dose studies of methylprednisolone have been done using modeling approaches to estimate parameters [ 70 82 95 96 ] The present study showed that PK/PD models from single dose studies can also be applied to multiple dose studies, based on previously obtained clinical data. It further compared the potency and efficacy between methylprednisolone and pred nisolone, as both drugs were given to same subject in separate blocks with proper washout periods. Also, it suggested that, for modeling purposes, it is unnecessary to consider the interaction from endogenous cortisol when exogenous glucocorticoids were gi ven orally. Materials and Methods Historic Data Data analyzed in this study were collected during a previously conducted study, in which twelve different dose regimens for methylprednisolone (MP) and prednisolone (PNL) were examined [ 82 ] Twenty four healthy males, ranging in age from 20 40 years,and ranging in weight from 50 80 kg, were recruited for this open, randomized cross over study. The volunteers were screened before the study, and the inclusion and
56 exclusion criteria were discussed in [ 82 ] During this multiple dose study, drugs were given either eight times, six times, four times, three times, twice, or once daily for two days in separate blocks, with proper washout period, with corresponding MP doses of 1 mg, 2 mg, 4 mg, 8 mg, 24 mg, and 80 mg, respectively. Also, in separate blocks, PNL was given at the same dosing frequency, and the corresponding doses were 1.25 mg, 2.5 mg, 5 mg, 10 mg, 30 mg, and 100 mg, respectively. Blood samples were collected for three days, and tot al drug concentration, total cortisol concentration, and differential white blood cell counts (lymphocytes) were obtained. The detail of data collection was described in [ 82 ] MP PKPD M odel s Total methylprednisolone (MP) concentration was described using a one compartment body model with first order absorption and first order elimination (Figure 3 2). Because MP binds to albumin with a constant binding percentage of 77% [ 81 ] free MP concentration was calcul ated as total concentration multiplied by 23% Total cortisol baseline production was described using a turnover model with a production rate K in and a first order elimination rate K out ( E quation 3 1). ( 3 1 ) K in was replaced by a dual linear function to mimic the circadian rhythm of cortisol baseline production (Figure 3 3) This dual linear model divided 24 hr baseline production into two phases: a linear increase of the production rate (K in ) from T min (time at minimum production rate) to T max (time at maximum production rate), and a linear decrease from T max to T min [ 89 ] : From 0 T min : ( 3 2 )
57 From T min T max : ( 3 3 ) From T max 24 hr: (3 4) The maximum production rate was assumed to be R max and the minimum production rate was assumed to be 0. T gap is the time gap between T min and T max Cortisol suppression by methylprednisolone was described by an indirect response model (Equation 3 5) with K in replaced by a dual linear model (Figure 3 2 Effect ) : ( 3 5 ) is the maximum suppression effect of the drug, and is the free drug concentration which produces 50% of maximum suppression effect. Also, an indirect response model was used to describe lymphocyte trafficking effects from exogenous MP (Figure 3 2 Effect ): ( 3 6) K ily is the influx rate constant of cells from the extravascular space and K oly is the efflux rate constant of cells export to the vascular space. is the maximum inhibition effect for the drug. is the free MP concentration w hich produces 50% of the maximum effect. Without administration of exogenous glucocorticoids, endogenous cortisol is the driving force of lymphocyte trafficking, and this effect can be described by an indirect response model (Figure 3 2 Effect ) :
58 ( 3 7 ) is the maximum inhibition effect for cortisol. is the free cortisol concentration which produces 50% of the maximum effect. Since glucocorticoids must bind to a glucocorticoid receptor to be effective, it is reasonable to assume that the interaction between methylprednisolone and cortisol on lymphocyte trafficking is a competitive mechanism. Thus, lymphocyte trafficking effects from methylprednisolone and cortisol can be added t ogether, as suggested for a competitive drug drug interaction [ 97 ] (Equation 3 8) ( 3 8 ) is a theoretical drug concentration that was converted from real cortisol concentration (Equation 3 9) with the assumpt ion that : t he theoretical drug concentration and the real cortisol concentration have the same lymphocyte trafficking effects (Equation 3 10) with the E max model parameters corresponding to drug or cortisol, respectively. ( 3 9 ) ( 3 10 ) P NL PKPD M odel s Free prednisolone (PNL) concentration was described using a one compartment body model with first order abs orption and first order elimination (Figure 3 2 ). B ecause free PNL can bind to saturable transcortin and cause nonlinearity the observed total PNL concentration ( ) was fitted with following equation:
59 (3 11 ) where and are the total number of binding sites of transcortin and albumin in human plasma. and are the binding affinities of PNL to transcortin and albumin, respectively. As with MP, an indirect response model was used to describe cortisol suppression (Figure 3 2 Effect ) and lymphocyte trafficking (Figure 3 2 Effect ) by PNL. For PNL as with MP, there is a possible drug drug interaction between exogenous PNL and endogenou s cortisol on lymphocyte trafficking (Figure 3 2 Effect ) Furthermore, because cortisol also can bind to transcortin, this protein binding competiti on (Figure 3 2 Effect ) from endogenous cortisol may need to be added into Equation 3 11 for total prednisolone concentration fitting [ 91 ] : ( 3 1 2 ) (3 1 3 ) where and are the binding affinities of cortisol to transcortin and albumin, respectively. PKPD Analysis Model SIMP was constructed from PKPD models for MP and PNL without considering lymphocyte trafficking and competitive protein binding from endogenous cortisol. Model COMP was more complicated, as it was constructed from PKPD models for MP and PNL including those considerations (Figure 3 2). Model SIMP: Population PK analyses were separately per formed for MP and PNL concentration. After that, PK parameters for MP and PNL were combined with
60 cortisol data or lymphocyte data to do a sequential PK/PD modeling in NONMEM to obtain PD parameters for cortisol suppression or lymphocyte trafficking, respec tively. Model COMP: Population PK analysis was performed for MP concentration. Simultaneous PKPD modeling was performed by combining PNL concentration time profile and cortisol time profile under PNL treatment. PK parameters for MP and PNL were then combi ned with cortisol data to do a sequential PK/PD modeling in NONMEM for cortisol suppression parameters. Then, PK parameters for MP and PNL, PD parameters for cortisol suppression, and observed lymphocyte data were combined for sequential PK/PD modeling in NONMEM to obtain lymphocyte trafficking parameters. For all model components, the between subject variability (BSV) was modeled with an exponential error model. The residual error was modeled with an additive plus proportional error model for PK data, an e xponential error model for cortisol data, and an additive error model for lymphocyte data. The first order conditional method with interaction was used for parameter estimation. During sequential PKPD modeling, c ortisol baseline parameters and lymphocyte t rafficking parameters (K out R max K ily K oly etc.) were assumed to be consistent for same subject no matter which drug was administered. and w ere set differently correspond ing to drugs B oth normal scale data and log transformed data were tested for cortisol suppression modeling Final models for Model SIMP and Model COMP were validated by visual predictive check (VPC) as d rug concentration, cortisol concentration and lymphocytes number were simulated 1000 times for each dose regimen Mean of prediction and 95% prediction interval were calculated based on simulation data Moreover, predicted cumulative cortisol suppression (CCS) and lymphocyte suppression (CLS) at steady
61 state was calculated by subtracting area under curve (AUC) of simu lated PD profiles (cortisol concentration or lymphocyte count versus time) under exogenous drug treatments from the AUC of simulated baseline profiles (n=1000). Res ul ts Pharmacokinetic s of MP and PNL The measured total methylprednisolone (MP) concentration can be described by a one compartment body model with first order absorption and first order elimination. PK parameters are summarized in Table 3 1 Between subject variability (BSV) was estimated for absorption rate (KA), apparent clearance (CL/F) and apparent volume of distribution (V d /F). Residual variability (RV) was estimated by a proportional plus additive error model. The correlation coefficient between KA and CL/F was 0.21, KA and V d /F was 0.86, and CL/F and V d /F was 0.58. A good fit is shown by diagnostic plot in Figure 3 4 Furthermore, visual predictive check (VPC) results shows that most of the observed data fall into the 95% prediction interval for all six dose regimens (Figure 3 5 ). MP concentration drops to almost zero 24 hr after the last treatment Unsurprisingly, the highest dose regimen (80 mg once a day) had the largest fluctuation of drug concentration, with the highest maximum concentration and lowest minimum concentration during the treatment period. Free prednisolone (PNL) concentr ation was also described by a one compartment body model. To calculate total PNL concentration, Model SIMP used a nonlinear Langmuir equation without considering the protein binding competition from endogenous cortisol. Model COMP calculated total PNL conc entration with a nonlinear Langmuir equation that did take into consideration the protein binding competition from endogenous cortisol. The fitting results for total PNL concentration from Model SIMP
62 and Model COMP were compared, and PK parameters were sum marized in Table 3 2 During PK modeling, total transcortin and albumin binding sites in plasma were fixed as 0.7*10 6 M and 500*10 6 M, respectively. Cortisol binding affinity to transcortin and albumin were fixed as 30 M 1 *10 6 and 0.05 M 1 *10 6 respectiv ely. PNL binding affinity to albumin was fixed as 0.002 M 1 *10 6 [ 79 91 98 100 ] B etween subject variability (B SV ) was estimated for apparent clearance (CL/F), apparent volume of distribution (V d /F) and prednisolone affinity to transcortin (K t PNL ). R esidual v ariability (R V ) was estimated by proportional plus additive error model. The results of Model SIMP suggested the correlation coefficient between CL/F and V d /F was 0.91, CL/F and K t PNL was 0.69, and V d /F and K t PNL was 0.46. The results of Model COMP showed the correlation coefficient between CL/F and V d /F ( 0.97 ) CL/F and K t PNL ( 0.84 ) and V d /F and K t PNL ( 0.78 ) A good fit is shown in d iagnostic plots for both Model SIMP and Model COMP. Model SIMP results are shown in blue and Model COMP results are shown in red. There was no significant difference between two models in terms of tot al PNL concentration prediction (Figure 3 6) Furthermore, VPC results show that most of the observed data fall into the 95% prediction interval mad e by Model SI MP (Figure 3 7 ) and Model CO MP (Figure 3 8 ) for all six dose regimens. PNL concentration also drops to a very low level a fter the last treatment in the 3 rd day, except for the highest dose regimen (100 mg once a day). Similar to the MP pharmac okinetic profile, the highest dose regimen (100 mg once a day) had the largest fluctuation of the total PNL concentration with the highest maximum concentration and lowest minimum concentration during the treatment period. Cortisol Suppression PK paramete rs from previous pharmacokinetic analyses for MP and PNL were combined with cortisol d ata t o assess drug effect differences by sequential PKPD
63 modeling. Log transformation of cortisol data was used to stabilize PKPD modeling on glucocorticoid cortisol supp ression. C ortisol baseline parameters were set up identically for the same healthy subject but I max and IC 50 w ere set up corresponding to different drugs Again, Model SIMP did not consider the competitive binding to transcortin from endogenous cortisol, and Model COMP did T he estimated cortisol suppression PD parameters were summarized in Table 3 3 For Model SIMP, BSV was estimated for cortisol elimination rate constant (K out ) and initial cort isol concentration (BSL), and the correlation coefficient between K out and BSL was 0.20. For Model COMP, BSV was estimated for maximum cortisol synthesis rate (R max ), cortisol elimination rate constant (K out ), and initial cortisol concentration (BSL). The correlation coefficient between R max and K out was 0.96, and K out and BSL was 0.30. RV was estimated by exponential model. Modeling results suggested that the maximum cortisol suppression effect was different between MP and PNL, as I max for MP was 0.6 and I max for PNL was 0.72. MP had a lower IC 50 than PNL, and thus was more potent (Table 3 3 ). Diagnostic plots showed that the overall fit of cortisol concentration was relatively good ( Figure 3 9 ) and there were no significant differences for individual pre dictions for each time point from Model SIMP and Model COMP (Figure 3 10 A ) VPC has been performed for both models and both models showed a similar cortisol suppression pattern, with the observed cortisol data falling into a 95% prediction interval (Figur e 3 11 vs. Figure 3 12 for MP and Figure 3 13 vs. Figure 3 14 for PNL) Cortisol level reduced dramatically after treatment began, and was suppressed to low level during the drug treatment In the 3rd day of the study, as drug concentration dropped to zero cortisol levels gradually rose, returning to normal levels 24 hrs after the last treatment. Except
64 for the highest dose regimen (80 mg MP once a day or 100 mg PNL once a day), all other dose regimens were predicted to have similar cortisol suppression profile s at steady state (day 2) These similarities are shown in Table 3 4 as the values of predicted cumulative cortisol suppression (CCS) were similar among different dose regimens. At steady state, about 50% to 60% of cortisol was suppressed after met hylprednisolone treatment, and 60% to 70% suppressed after prednisolone treatment (Table 3 4 ). Lymphocytes Trafficking Similarly, l ymphocyte numbers (represented as percentage of white blood cells) after MP or PNL treatment were combined for sequential PKPD modeling, to assess efficacy differences between the two drugs. Besides the transcortin competitive binding between PNL and cortisol, Model COMP also considered trafficking effects both from exogenous glucocorticoids and endogenous cortiso l. Model SIMP assumed that during the drug treatment period, the trafficking effect only came from exogenous drugs. Previously estimated results for PK parameters and cortisol suppression parameters were used in the sequential modeling, and lymphocyte traf ficking parameters were summarized in Table 3 5 BSV was estimated for lymphocyte efflux rate (K oly ), IC 50 of free MP and free PNL. MP and PNL had the same maximum effects on lymphocyte trafficking (I max = 0.8). MP was more potent than PNL on lymphocyte tr afficking, as MP's IC 50 was smaller than PNL's IC 50 Again, diagnostic plots showed that the overall fit of lymphocyte count was relatively good ( Figure 3 15 ) and there were no significant differences for individual predictions of each time point from Mode l SIMP and Model COMP (Figure 3 10 B) VPC has been performed for both models and both models showed a similar lymphocyte
65 suppression pattern, with the observed lymphocyte count falling into 95% prediction interval (Figure 3 16 vs Figure 3 17 for MP and Figure 3 18 vs Figure 3 19 for PNL) Lymphocyte numbers in plasma were reduced during the treatment period and went back to normal level gradually after treatment stopped. Unlike the similarity of cortisol suppression profiles among different d ose regimens, t he extent of fluctuation of lymphocyte count during a 24 hr period changed according to dose frequency. The dose regimen with higher dose and lower dose frequency caused higher fluctuation in lymphocyte number during a 24 hr period Moreover t he differences of overall predicted cumulative lymphocyte suppression (CLS) were bigger among different dose regiments comparing to CCS (Table 3 4 ). At steady state, about 30% to 60% of lymphocytes were suppressed after administering exogenous glucocort icoids (Table 3 4 ) Discussion Methylprednisolone (MP) and prednisolone (PNL) have been used as systemic anti inflammatory drugs in clinical practice for decades Long term use of the drugs may cause side effects such as diabetes, osteoporosis, hypertensi on and mood disturbances. To enhance the safety of the drugs without affecting the efficacy, pharmaceutical scientists have been studying optimization of the dose regimen using PK/PD modeling. Pr otein binding properties of glucocorticoids, especially PNL and cortisol, play an important role in PKPD modeling, as only free compounds have drug effects. It has been shown that MP binds to albumin with an average 77% protein binding percentage [ 81 ] and its pharmacokinetic profile after oral administration can be described by a one co mpartment body model [ 82 83 ] The present study confirmed that a one compartment body model is sufficient to describe the concentration time profile of MP for the six multiple d ose regimens ranging from 1 mg eight times a day to 80 mg once a day. The
66 PK parameters obtained in this study agree with the literature values. For example, V d /F has been reported ranging from 74 134 L, and CL/F ranging from 19 37 L/kg for total MP [ 68 ] and the present study reported V d /F as 84.7 L and CL/F 25.2 L/kg. Both PNL and cortisol bind to albumin and saturable transcortin, which leads to nonlinear pharmacokinetic properties [ 79 80 ] The present study applied a one compartment body model to describe free PNL concentration, as suggested by other literature [ 80 84 87 91 99 ] Total PNL concentrations were calculated with the bi nding parameter values from literature with or without the consideration of competitive binding between cortisol and PNL by Model COMP or Model SIMP, respectively [ 79 98 ] A s suggested by Legler et al [ 100 ] a lthough the transcortin binding competition between PNL an d cortisol has been suggested, the protein binding parameters were not significantly different between Model SIMP and Model COMP More o ver the population PK parameters obtained in the present study agree with the l iterature values, and are not significant ly different between Model SIMP and Model COMP Cortisol baseline is circadian during a 24 hr period, with the maximum level in the evening and the minimum level in the afternoon [ 90 101 ] Thus, cortisol baseline profile is non symmetrical and can be described by different functions, in this case two linear functions [ 90 91 ] The present study successfully used this dual linear model to describe the cortisol baseline profile and an indirect response model to describe the cortisol suppression profile when given exogenous glucocorticoids. Without considering competitive binding between PNL and cor tisol, Model SIMP produced similar fitting results as Model COMP, suggesting that protein binding completion from endogenous
67 cortisol can be ignored after administrating exogenous glucocorticoids in terms of cortisol suppression modeling. Endogenous corti sol does not only affect protein binding of PNL, but also regulates the lymphocyte trafficking in humans [ 102 ] L ymphocytes travel between blood and extravascular pools, and also exhibit a circadian rhythm, as does cortisol but in the opposite direction (high level in the night and low level in the morning) (Figure 3 20) [ 103 ] Thus, lymphocyte number in plasma was assumed to be a constant by a turnover model, and its circadian rhythm, caused by regulation from endogenous circadian cortisol, can be described by an indirect response model. The dramatic reduction of lymphocyte number in plasma when given exogenous glucocorticoids is caused by t he induction of lymphocyte trafficking. Since both exogenous and endogenous glucocorticoids need to bind to glucocorticoid receptors to regulate lymphocyte trafficking, an a dditive relationship has been adapted into the indirect response model to describe the cortisol drug interaction [ 83 92 94 97 ] In the present study, Model COMP considers the endogenous lymphocyte trafficking effect from cortisol by converting concentration of endogenous cortisol to a fictitious drug concentration and then adding this fictitious drug concentratio n to the real drug concentration in the indirect response model. Model SIMP did not consider the regulation from endogenous cortisol and the simple indirect response model was used. Thus, the lymphocyte baseline described by Model SIMP would be a horizonta l line (Figure 3 20) The present study showed that, although Model COMP was better in describing the circadian rhythm of lymphocyte placebo data, the two models had no
68 significant differences in terms of modeling lymphocyte suppression when given exogenou s glucocorticoid treatments. Not only are there no significant differences between Model SIMP and Model COMP on c ortisol and lymphocyte suppression parameters, they also had a good agreement with literature values [ 68 ] W ithout cortisol baseline data, the BSV from cortisol baseline (R max K out BSL) and the BSV from drug effects (I max and IC 50 ) cannot be distinguished. Thus, only the BSV of cortisol baseline parameters were estimate d. For lymphocyte trafficking modeling, both the BSV from lymphocyte baseline (K oly ) and the BSV from drug effects (IC 50 ) were estimated. T he present study first suggested that PNL is more efficient than MP, as it has a bigger I max MP had a lower IC 50 th an free PNL, suggesting that MP is more potent than PNL in suppressing cortisol production (0.21 ng/mL vs 0.46 ng/mL) and enhancing lymphocyte trafficking (3.45 ng/mL vs 9.8 ng/mL). As the glucocorticoid receptor binding affinity of PNL is also about 2 to 3 fold comparing to MP [ 104 ] the same fold differences in IC 50 and receptor binding affinity confirmed th at the mechanism of action of glucocorticoids is receptor mediated. Overall, this study showed that after multiple doses, drug concentration stayed at steady state during treatment, and cortisol concentration and lymphocyte count in plasma were suppressed to low levels. MP is more potent than PNL in suppressing cortisol production and inducing lymphocyte trafficking after oral administration. Although cortisol and PNL compete to bind to transcortin in plasma, this competition can be ignored when modeling PNL concentration. Furthermore, if subject is under drug treatment, only the lymphocyte trafficking effect from exogenous drugs needs to be considered for modeling purposes
69 Figure 3 1 Chemical structures of (A) met hylprednisolone (MP) ; (B) p rednisolone (PNL) ; (C) cortisol.
70 Figure 3 2. PK/PD model of exogenous methylprednisolone and prednisolone The blue block represents the PK model for free drug concentration, a one compartment body model with Ka as the first order absorption rate constant and Ke as the first order elimination rate constant. The orange block represents the cortisol production m odel, with Kin as the synthesis rate and Kout as the first order elimination rate constant. Kin is replaced with dual linear functions in the present study. The green block represents the lymphocyte trafficking turn over model, with Kily as the influx rate constant and Koly as the efflux rate constant. Free MP or PNL suppresses cortisol production through effect and induces lymphocyte trafficking through effect both described by the indirect response model. Free cortisol induces lymphocyte trafficking through effect which is described by the indirect response model. Free PNL and free cortisol compete to bind to transcortin, which is represented by line Model COMP considers PD effects and competitive binding Model SIMP only considers PD effects and
71 F igure 3 3 Cortisol baseline concentration and production rate (K in ) changes during every 24 hr period Cortisol concentration is circadian during 24 hr period as sh own by black line. Cortisol production rate ( K in ) is described by dua l linear function as shown by blue line.
72 Figure 3 4 Diagnostic plots for MP population PK: p opulation prediction versus o bserved concentrations in plasma ( upper left) ; Individual prediction versus o bserved concentrations in plasma ( upper right ) ; weighted residuals versus population prediction in plasma ( lower left ) ; weighted residuals versus time ( lower right ). Circle s represent each time point data. T he closer the trend (red line) is to the blue line, the fitting is better.
73 Figure 3 5 Visual predictive check (VPC) results for MP population PK model Blocks represent 95% prediction interval after 1000 time simulation. Lines represent mean of prediction. Circles are observed data.
74 Figure 3 6 Diagnostic plots for PNL population PK : p opulation prediction versus o bserved concentrations in plasma ( upper left) ; Individual prediction versus o bserved concentrations in plasma ( upper right ) ; weighted residuals versus population prediction in plasma ( lower left ) ; weighted residuals versus time ( lower right ). Circles represent each time point data. Red color represents results from Model COMP. Blue color represents results from Model SIMP. T he closer the trend (red or blue line) is to the black line, the fitting is better.
75 Figure 3 7 VPC results for PNL PK data from Model SIMP. Blocks represent 95% prediction interval after 1000 time simulation. Lines represent mean of prediction. Circles are observed data.
76 Figure 3 8. VPC results for PNL PK data from Model COMP. Blocks repres ent 95% prediction interval after 1000 time simulation. Lines represent mean of prediction. Circles are observed data.
77 Figure 3 9 Diagnostic plots for cortisol suppression fitting after administering methylprednisolone or prednisolone D ata from Model COMP are represented by red color D ata from Model SIMP are represented by blue color T he closer the trend (red or blue line) is to the black line, the fitting is better.
78 Figure 3 10. Individual predictions of each time point from Model SIMP and Model COMP for cortisol concentration (ng/mL) and lymphocyte count (%) A. B
79 Figure 3 1 1 VPC results from Model SIMP for cortisol suppression after MP treatments. Blocks represent 95% prediction interval after 1000 time simulation. Lines repr esent mean of prediction. Circles are observed data.
80 Figure 3 12. VPC results from Model COMP for cortisol suppression after MP treatments. Blocks represent 95% prediction interval after 1000 time simulation. Lines represent mean of prediction. Circles are observed data.
81 Figure 3 1 3 VPC results from Model SIMP for cortisol suppression after PNL treatments. Blocks represent 95% prediction interval after 1000 time simulation. Lines represent mean of prediction. Circles are observed data.
82 Figure 3 1 4 VPC results from Model COMP for cortisol suppression after PNL treatments. Blocks represent 95% prediction interval after 1000 time simulation. Lines represent me an of prediction. Circles are observed data.
83 Figure 3 15. Diagnostic plots for lymphocytes fitting after administering methylprednisolone and prednisolone Results from Model COMP are represented by red color Results from Model SIMP are represented by blue color T he closer the trend (red or blue line) is to the black line, the fitting is better.
84 Figure 3 1 6 VPC results from Model SIMP for lymphocytes trafficking after MP treatments. Blocks represent 95% prediction interval after 1000 time simulation. Lines represent mean of prediction. Circles are observed data.
85 Figure 3 17. VPC results from Model COMP for lymphocytes trafficking after MP treatments. Blocks represent 95% prediction interval after 1000 time simulation. Lines represent mean of prediction. Circles are observed data.
86 Fi gure 3 1 8 VPC results from Model SI MP for lymphocytes trafficking after PNL treatments. Blocks represent 95% prediction interval after 1000 time simulation. Lines represent mean of prediction. Circles are observed data.
87 Figure 3 19. VPC results from Model COMP for lymphocytes trafficking after PNL treatments. Blocks represent 95% prediction interval after 1000 time simulation. Lines represent mean of prediction. Circles are observed data.
88 Figure 3 20. Lymphocytes placebo data and ba seline models. Red color represents simulation data from Model COMP ; Blue color represents simulation data from Model SIMP; Black color represents observed placebo data. Solid line represents me an of the data; Dash line represents 2.5% quantile and 97.5% q uantile of the data.
89 Table 3 1 P harmacokinetic parameters of methylprednisolone (MP) after oral administration. Parameters Estimates (%RSE) BSV (%RSE) KA (1/hr) 0.94 (16.9%) 46.3% (43.9%) CL/F (L/hr) 25.2 (3.2%) 12.2% (24.7%) V d /F (L) 84.7 (9.1%) 29.2% (42.4%) RV (prop ortional ) 34.5% (5.1%) RV (add itive ) (ng/mL) 5.23 (15.0%) KA = absorption rate constant ; CL /F = apparent clearance ; V d /F = apparent volume of distribution; RV = residual variability; %RSE = relative standard error; BSV = between subject variability Table 3 2 P harmacokinetic parameters of prednisolone (PNL) after oral administration obtained by Model COMP and Model SIMP. Parameters Model COMP Model SIMP Estimates (%RSE) BSV (%RSE) Estimates (%RSE) BSV (%RSE) KA (1/hr) 1.8 (0.7%) 1.49 (14.2%) CL/F (L/hr) 61 (4.9%) 70.2% (62.5%) 55 (13.7%) 51.8% (26.2%) V d /F (L) 216 (4.5%) 58.1% (57.9%) 191 (12.7%) 48.2% (36.3%) K t PNL (M 1 *10 6 ) 16 (9.7%) 106% (67.2%) 15.7 (18.3%) 50.5% (36.3%) RV (prop ortional ) 38.4% (1.8%) 229% (9.9%) RV (add itive )(ng/mL) 5.21 (16.9%) 12.1 (20.6%) KA = absorption rate constant ; CL /F = apparent clearance ; V d /F = apparent volume of distribution; K t PNL = prednisolone binding affinity to transcortin ; RV = residual variability; %RSE = rel ative standard error; BSV = between subject variability
90 Table 3 3 Cortisol suppression pharmacodynamic parameters of methylprednisolone (MP) and prednisolone (PNL) obtained by Model COMP and Model SIMP. Parameters Model COMP Model SIMP Estimates (%RSE) BSV (%RSE) Estimates (%RSE) BSV (%RSE) R max (ng/mL) 68 (6.6%) 80.9% (42.1%) 68 (14.5%) K out (1/hr) 0.53 56.3% (39.7%) 0.53 21.5% (39.0%) BSL (ng/mL) 160 (7.9%) 21.7% (67.4%) 160 (7.6%) 31.4% (36.7%) MP I max 0.6 0.6 PNL I max 0.72 0.72 Free MP IC 50 (ng/mL) 0.23 (0.4%) 0.21 (1.2%) Free PNL IC 50 (ng/mL) 0.37 (2.7%) 0.46 (5.7%) RV(exp onential ) 0.63 (1.6%) 0.65 (4.3%) R max = cortisol maximum synthesis rate ; K out = cortisol elimination rate constant; BSL = cortisol concentration at time zero; I max = maximum suppression effect; IC 50 = drug concentration when reach 50% of maximum effect; RV = residual variability; %RSE = relative standard error; BSV = between subject variability Table 3 4. Predicted c umulative cortisol and lymphocyte suppression percentage in 24 hr at steady state from Model SIMP and Model COMP Dose Regimen Cumulative Cortisol Suppression Cumulative Lymphocyte Suppression Model SIMP Mean (SD) Model COMP Mean (SD) Model SIMP Mean (SD) Model COMP Mean (SD) 1 mg MP 55.9% (5.4%) 55.6% (0.3%) 36 .0 % (4.0%) 32.1% (3.8%) 2 mg MP 57.1% (5.2%) 56.9% (0.2%) 43.2% (3.9%) 40 .0 % (3.8%) 4 mg MP 57.7% (5.2%) 57.5% (0.2%) 47.5% (3.7%) 44.9% (3.6%) 8mg MP 58.3% (5.1%) 58.2% (0.2%) 52.8% (3.3%) 50.9% (3.2%) 24 mg MP 58.5% (5.1%) 58.4% (0.2%) 58.4% (2.7%) 57.2% (2.6%) 80 mg MP 52.2% (6.1%) 51.8% (1.8%) 48.3% (2.8%) 46.7% (2.8%) 1.25 mg PNL 67.3% (1.0 % ) 67.3% (1.5%) 33.6% (5.9%) 32.4% (7.6%) 2.5 mg PNL 68.6% (3.9%) 68.7% (1.1%) 40.1% (6.0%) 39.4% (7.5%) 5 mg PNL 69.2% (3.8%) 69.2% (1.0%) 43.9% (5.8%) 43.5% (7.2%) 10 mg PNL 69.7% (3.7%) 69.7% (0.9%) 48.7% (5.5%) 48.6% (6.8%) 30 mg PNL 69.7% (3.7%) 69.7% (1.1%) 53.9% (4.8%) 54.3% (6.0%) 100 mg PNL 61.0% (5.8%) 61.7% (3.9%) 44.8% (4.7%) 45.4% (5.9%)
91 Table 3 5 L ymphocytes trafficking pharmacodynamic parameters of methylprednisolone (MP) and prednisolone (PNL) obtained by Model COMP and Model SIMP Parameters Model COMP Model SIMP Estimates (%RSE) BSV (%RSE) Estimates (%RSE) BSV (%RSE) K ily (%) 18.1 (9.6%) 20.6 (10.8%) K oly (1/hr) 0.43 (9.8%) 19.5% (23.1%) 0.52 (10.8%) 17.9% (23.4%) BSLY (%) 36.3 (3.7%) 36.5 (3.7%) MP I max 0.8 0.76 PNL I max 0.8 0.76 Cortisol I max 1 Free MP IC 50 (ng/mL) 3.50 (12.7%) 44.8% (34.1%) 3.45 (11.4%) 43.9% (33.0%) Free PNL IC 50 (ng/mL) 7.48 (13.1%) 59.6% (41.7%) 9.8 (13.8%) 60.2% (25.9%) Free Cortisol IC 50 (ng/mL) 12.1 RV(add itive ) (%) 6.25 (5.3%) 6.19 (5.3%) K ily = lymphocyte influx rate ; K oly = lymphocyte efflux rate; BSLY = lymphocyte number at time zero; I max = maximum suppression effect; IC 50 = drug concentration when reach 50% of maximum effect; RV = residual variability; %RSE = relative standard error; BSV = between subject variability
92 CHAPTER 4 CONCLUSION AND DISCUSSION The h erbal supplement market has been growing for eight consecutive years in the United States. Sales of herbal supplements increased 4.5% and reached 5.278 billion in 2011 [ 105 ] However, concerns on herbal supplements are rising align ing with the ir sales growth. The U.S. Department of Health & Human Services published a report regarding the issues of misleading claims and missing reliable scientific data during supplement marketing [ 10 6 ] For example, man gosteen products have been globally available claiming making your respiratory health, immune health, intestinal health, joint health and just overall health better every day [ 10 7 ] or Mangosteen has been used for more than 130 diseases successfully [ 10 8 ] However, information about the PK properties of man gosteen is limited. Although mangosteen extracts showed a variety of pharmacological activities, very few of those activities were confirmed by in vivo studies. PK data are important in herb al pharmaceutical research as they link in vitro pharmacological activities to in vivo activities by answering questions such as bioavailability and metabolic pathways. Thus, it is important to understand the PK properties of mangosteen. Before performing PK studies, the active compounds in herbs need t o be identified. Xanthone derivatives are the major bioactive compounds in mangosteen and and mangostin are the most presented xanthones. Two studies showed that the oral bioavailability of mangostin is very low after taking mangosteen juice product s [ 50 53 ] Th e first project here confirmed the low bioavailability of and mangostin after oral administration of pure forms and mixtures. This low bioavailability of the active xanthones should attract public attention on mangosteen product consumption as
93 generally low bioavailability means high dose is required to achieve effective thre shold Furthermore, results of the first project demonstrated that fast conjugation was the reason for the low bioavailability. Thus, t he first project not only put forward concerns on dose requirement s for mangosteen products, but also proposed possible o ptimization directions by administering cofactors that can inhibit conjugation. Obtaining PK/PD information of newly drug candidates (such as biological active mangosteen compounds) is not only important for new drug research. Modeling analyses of PK/PD in formation are also important for drugs that have been used for de cades (such as glucocorticoids) Glucocorticoids have been used as treatments for chronic inflammatory diseases for decades. However, long term use of glucocorticoids causes side effects such as hyperglycemia, osteoporosis, adrenal insufficiency, negative calcium balance, etc. PK/PD modeling is a strong tool to describe the relationship between dose regimen and drug effects, so it can be used to evaluate efficacy and safety of drugs and explor e dosage optimization possibilities. Th e second project in the present study successfully estimated population PK/PD parameters for methylprednisolone (MP) and prednisolone (PNL) after multiple dos es which are not significantly different from single dose data. Furthermore, it distinguished the random variability among observations (BSV and RV) and the fixed response differences caused by different drugs by population analyses with NONMEM. Th is direct comparison between drugs was made based on efficacy (E ma x ) and potency (EC 5 0 ) after modeling and the results suggested that MP is more potent than PNL on cortisol suppression and lymphocyte trafficking. In addition, although in a real physiological situation there will be interactions between endogenous
94 cortisol and exogenous methylprednisolone and prednisolone, these interactions can be ignored for modeling purpose. Thus, this second project is a good example showing that it is not always true that a more complicated model will have better data fitting a nd prediction s In conclusion, the present study gave two examples of how PK/PD concepts can be applied in drug research. Chapter 2 discusses the importance of understanding bioavailability for new drug candidates. Chapter 3 shows how population PK/PD mod eling can be applied to predict drug responses and compare drug differences. Overall, PK/PD concepts have been applied in drug research for decades and they will continue to shine brightly.
95 APPENDIX A NONMEM CODE FOR P NL PK MODEL PNL COMP $PROBLEM PREDNISOLONE SIMULTANEOUS PKPD MODEL COMPETITION $INPUT C ID TIME DV DOSE DOSF AMT ADDL II EVID MDV CMT COH SID TYP $DATA .. \ P PK CORT nm Compte m1 all log sep.csv IGNORE=C $SUBROUTINE ADVAN6 TRANS1 TOL=3 $MO DEL NCOMP = 3 COMP = (PKDEPOT,DEFDOSE) COMP = (PKCNTRL,DEFOBS) COMP = (PDCORT) $PK "FIRST COMMON /PRCOMG/ IDUM1,IDUM2,IMAX,IDUM4,IDUM5 INTEGER IDUM1,IDUM2,IMAX,IDUM4,IDUM5 IMAX=100000000 ;PK MODEL K12 = THETA(1)*EXP(ETA(1)) CL = THETA(2)*EXP(ETA(2)) V2 = THETA(3)*EXP(ETA(3)) S2 = V2/1000 K20 = CL/V2 PT = 0.7 ; N t *Pt transcortin binding sites number PA = 500 ; N a *P a albumin binding sites number KTP = THETA(4)*EXP(ETA(4)) ; PNL transcortin binding affinity KTC = THETA(5)*EXP(ETA(5)) ; Cortisol transcortin binding affinity KAP = 0.002 ; PNL albumin binding affinity KAC = 0.05 ; cortisol albumin binding affinity ;PD MODEL RMAX = THETA(6)*EXP(ETA(6)) cortisol maximum synthesis rate KOUT = THETA(7)*EXP(ETA(7)) ;cortisol elimination rate BSL = THET A(8)*EXP(ETA(8)) ;cortisol conc. at time zero IMX = THETA(9)*EXP(ETA(9)) ;PNL maximum effect IC50 = THETA(10)*EXP(ETA(10)) ;PNL suppression IC50 HI = THETA(11)*EXP(ETA(11)) ;hill factor TMIN = THETA(12)+ETA(12) ; time at maximum cortisol synthesis rate TGAP = THETA(13)+ETA(13 ) TMAX = TMIN+TGAP ; time at minimum cortisol synthesis rate A_0(3) = BSL SD1 = THETA(14) ;PNL concentration proportional residual variability SD2 = THETA(15) ;PNL concentration additive residual variability SD3 = THETA(16) ;c ortisol concentration exponential residual variability $DES DADT(1)= K12*A(1) ;drug disposition compartment DADT(2)= K12*A(1) K20*A(2) ;central compartment, plasma CP=A(2)/S2
96 EFF=1 (IMX*CP**HI)/(IC50**HI+CP**HI) ;Emax model T1 = T 24*INT(T/24) ; conver t time to 0 24 hr period IF (T1.LE.TMIN) THEN ; convert time to start from Tmin T2=T1+24 ELSE T2=T1 ENDIF IF (T2.LE.TMAX) THEN ;calculate Kin (cortisol synthesis rate) KIN=RMAX*(T2 TMIN)/TGAP ELSE KIN=RMAX*(TMIN+24 T2)/(24 TGAP) ENDIF DADT(3)= KIN*EFF KOUT*A(3) ;cortisol compartment $ERROR CTA=KTC*(1+PA*KAC) CTB=KTC*(PT A(3)/362.46)+(1+PA*KAC)*(1+KTP*F/360.444) CTC= A(3)*(1+KTP*F/360.444)/362.46 CTF=(SQRT(CTB*CTB 4*CTA*CTC) CTB)/2/CTA CT=PT*KTP*F/(1+KTP*F/360.444+KTC*CTF)+PA*KAP*F+F C2=A(3) IF (TYP.EQ.0) THEN W=SQRT(SD1*SD1*CT*CT+SD2*SD2) IPRE=CT Y=IPRE+W*EPS(1) ; total PNL concentration IRES=DV IPRE ELSE W=SD3 IPRE=LOG(C2) Y=LOG(C2)+W*EPS(1) ;total cortisol concentration IRES=DV IPRE ENDIF $THETA (0, 2, 10) ;KA (0, 30, 1000) ;CL (0, 130, 1000) ;V (0, 40, 100) ;KTP 30 FIX ;KTC (0, 60, 500 ) ;RMAX 0.56 FIX ;KOUT (1/HR) (0, 120, 1000) ;BSL (NG/ML) 0.83 FIX ; IMX (0, 2, 50) ;IC50 (NG/ML) 1 FIX ;HILL 16 FIX ;TMIN (HR) 4 FIX ;TGAP (HR) (0, 0.5, 100) ; SD1
97 (0, 0.5, 100) ; SD2 (0, 0.5 10) ;SD 3 $OMEGA 0 FIX ;KA $OMEGA BLOCK(3) 0.3 ;CL 0.01 0.3 ;V 0.01 0.01 0.3 ;KTP $OMEGA 0 FIX ;KTC $OMEGA BLOCK(3) 0.3 ;RMAX 0.1 0.3 ;KOUT 0.01 0.01 0.1 ;BSL $OMEGA 0 FIX ;IMX 0 FIX ;IC50 0 FIX ;HILL 0 FIX ;TMIN 0 FIX ;TGAP $SIGMA 1 FIXED $EST MAX=9999 PRINT=10 METHOD=1 INT NOABORT $COV PRINT=E $TABLE ID DOSE TIME EVID IPRE CMT MDV SID ONEHEADER NOPRINT FILE=SDTAB024 $TABLE ID DOSE K12 CL V2 KTP RMAX BSL KOUT IC50 SID ONEHEADER NOPRINT N OAPPEND FILE=PATAB024
98 APPENDIX B NONMEM CODE FOR PNL PK MODEL PNL SIMP $PROBLEM PREDNISOLONE PK MODEL NO COMPETITION $INPUT C ID TIME DAY DV DOSE DOSF AMT ADDL II EVID MDV CMT COH SID $DATA Prednisolone_PK_nm m1.csv IGNORE=C $SUBROUTINES ADVAN2 TRANS2 $PK TVKA = THETA(1) ;population absorption rate TVCL = THETA(2) ;population clearance TVV = THETA(3) ;population volume of distribution KA = TVKA*EXP(ETA(1)) ;individual absorption rate CL = TVCL*EXP(ETA(2)) ;individual clearance V = TVV*EXP(ETA(3)) ;individual volume of distribution S2=V/1000 A = 0.7 ;Nt*Pt transcortin binding sites number B = 1.04 ;Na*Ka*Pa PNL that binds to albumin KT = THETA(4)*EXP(ETA(4)) ; PNL transcortin binding affinity SD1 = T HETA(5) ;PNL conc. proportional residual variability SD2 = THETA(6) ;PNL conc. additive residual variability $ERROR W=SQRT(SD1*SD1*F*F+SD2*SD2) CT=A*KT*F/(1+KT*F/360.444)+B*F+F ;total PNL concentration IPRED=CT Y=CT+W*EPS(1) IRES=DV IPRED IWRES=IRES/W $T HETA (0, 5, 10) ;KA (0, 20, 1000) ;CL (0, 100, 1000) ;V (0, 20, 1000) ;KT (0, 0.5, 100) ;Prop (0, 0.5, 100) ;Add $OMEGA 0 FIX ;KA $OMEGA BLOCK(3) 0.5 ;CL 0.01 0.5 ;V 0.01 0.01 0.5 ;KT $SIGMA 1 FIX $EST MAXEVALS=9999 PRINT=5 METHOD=1 INT NOABORT $COV PRINT=E $TABLE ID DOSE DOSF EVID TIME IPRED IWRES MDV SID ONEHEADER NOPRINT FILE=sdtab012 $TABLE ID DOSE DOSF KA CL V KT SID ONEHEADER NOPRINT FILE=patab012
99 APPENDIX C NONMEM CODE FOR VPC SIMULATION ON CORTISOL SUPPRESSION WITH MODEL COMP $PROBLEM model simulation for MP and PNL $INPUT C ID TIME DAY DV DOSE DOSF AMT ADDL II EVID MDV CMT TYP FLG $DATA sim drug .csv IGNORE=C $SUBROUTINE ADVAN6 TRANS1 TOL=3 $MODEL NCOMP = 3 COMP=(PKDEPOT,DEF DOSE) COMP=(PKCNTRL,DEFOBS) COMP=(PDCORT) $ABBREVIATED DERIV2=NO $PK "FIRST COMMON /PRCOMG/ IDUM1,IDUM2,IMAX,IDUM4,IDUM5 INTEGER IDUM1,IDUM2,IMAX,IDUM4,IDUM5 IMAX=100000000 ; PNL PK Model KA2 = THETA(1)*EXP(ETA(1)) CL2 = THETA(2)*EXP(ETA(2)) V2 = THETA(3)*EXP(ETA(3)) S2=V2/1000 KE2=CL2/V2 PT = 0.7 ; N t *Pt transcortin binding sites number PA = 500 ; N a *P a albumin binding sites number KTP = THETA(4)*EXP(ETA(4)) ; PNL transcortin binding affinity KTC = 30 ; Cortisol transcortin binding affinity KAP = 0.002 ; PNL albumin binding affinity KAC = 0.05 ; cortisol albumin binding affinity ;PD Model RMAX = THETA( 5 )*EXP(ETA( 5 )) cortisol maximum synthesis rate KOUT = THETA( 6 )*EXP(ETA( 6 )) ;cortisol elimination rate BSL = THETA( 7 )*EXP(ETA( 7 )) ;cortisol conc. at time zero IC1 = THETA(8)*EXP(ETA(8)) ;Total MP IC50 IC2 = THETA(9)*EXP(ETA(9)) ;Free PNL IC50 IMX1 = 0.6 ;Total MP Imax IMX2 = 0.72 ;Free PNL Imax TMIN = 16 ;time at maximum cortisol synthesis rate T MAX = 20 ; time at minimum cortisol synthesis rate T GAP =4 A_0(3) = BSL ; MP PK Model KA1 = THETA(10)*EXP(ETA(10)) CL1 = THETA(11)*EXP(ETA(11)) V1 = THETA(12)*EXP(ETA(12))
100 S1 =V1/1000 KE1=CL1/V1 SD1 = THETA(13) ;RV SD2 = THETA(14) ;RV SD3 = THETA(15) ;RV SD4 = THETA( 16) ;RV SD5 = THETA(17) ;RV $DES DADT(1)= (1 FLG)*KA1*A(1) FLG*KA2*A(1) DADT(2)= (1 FLG)*KA1*A(1) (1 FLG)*KE1*A(2)+FLG*KA2*A(1) FLG*KE2*A(2) C1=A(2)/S1 C2=A(2)/S2 EFF=(1 FLG)*(1 (IMX1*C1)/(IC1+C1))+FLG*(1 (IMX2*C2)/(IC2+C2)) T1 = T 24*INT(T/24) IF (T1.LE .TMIN) THEN T2=T1+24 ELSE T2=T1 ENDIF IF (T2.LE.TMAX) THEN KIN=RMAX*(T2 TMIN)/TGAP ELSE KIN=RMAX*(TMIN+24 T2)/(24 TGAP) ENDIF DADT(3)= KIN*EFF KOUT*A(3) $ERROR REP=IREP CO1=A(2)/V1*1000 CO2=A(2)/V2*1000 IF (FLG.EQ.0.AND.TYP.EQ.0) THEN ; Predict MP total c oncentration W=SQRT(SD1*SD1*CO1*CO1+SD2*SD2) IPRE=CO1 Y=IPRE+W*EPS(1) IRES=DV IPRE ENDIF IF (FLG.EQ.1.AND.TYP.EQ.0) THEN ; Predict PNL total concentration CTA=KTC*(1+PA*KAC) CTB=KTC*(PT A(3)/362.46)+(1+PA*KAC)*(1+KTP*CO2/360.444) CTC= A( 3)*(1+KTP*CO2/360.444)/362.46 CTF=(SQRT(CTB*CTB 4*CTA*CTC) CTB)/2/CTA CT=PT*KTP*CO2/(1+KTP*CO2/360.444+KTC*CTF)+PA*KAP*CO2+CO2 W=SQRT(SD3*SD3*CT*CT+SD4*SD4) IPRE=CT Y=IPRE+W*EPS(1) IRES=DV IPRE ENDIF
101 IF (TYP.EQ.1) THEN ; Predict total cortisol concentrat ion W=SD5 IPRE=A(3) Y=IPRE*EXP(SD5*EPS(1)) ENDIF $THETA (0, 1.8, 10) ;KA2 (0, 61, 1000) ;CL2 (0, 216, 1000) ;V2 (0, 16, 100) ;KTP (0, 68, 800) ;RMAX (ng/ml/hr) 0.53 FIX ;KOUT (1/hr) (0, 160, 1000) ;BSL (ng/ml) (0, 1.02, 100) ;IC1 (ng/m l) (0, 0.37, 100) ;IC2 (ng/ml) (0, 0.94, 10) ;KA1 (0, 25.2, 500) ;CL1 (0, 84.7, 1000) ;V1 (0, 0.345, 10) ;Prop (0, 5.23, 100) ;Add (0, 0.384, 10) ;Prop (0, 5.21, 100) ;Add (0, 0.628, 10) ;SD5 $OMEGA 0 FIX ;KA2 $OMEGA BLOCK(3) 0.493 ;CL2 0.394 0.337 ;V2 0.625 0.48 1.13 ;KTP $OMEGA BLOCK(3) 0.655 ;RMAX 0.435 0.317 ;KOUT 0 0.0361 0.0472 ;BSL $OMEGA 0 FIX ;IC1 0 FIX ;IC2 $OMEGA BLOCK(3) 0.214 ;KA1 0.0123 0.0153 ;CL1 0.116 0.0209 0.0848 ;V1 $ SIGMA 1 FIXED $SIMULATION (214748364) ONLYSIMULATION SUBPROBLEMS=200 $TABLE ID DOSE EVID TIME IPRE PRED REP CMT TYP FLG ONEHEADER NOPRINT FILE=sdtab009
102 APPENDIX D NONMEM CODE FOR PKPD MODEL COMP ON LYMPHO CYTES TRAFFICKING $PROBLEM lymphocytes modeling for MP and PNL with cortisol interaction $INPUT C ID TIME DV DOSE AMT ADDL II EVID MDV CMT IKA ICL IV IKTP IRMX IKOC BSLC SID FLG $DATA PMP base CORT LYN nm m1.csv IGNORE=C $SUBROUTINE ADVAN6 TRANS1 TOL=3 $MODEL NCOMP = 4 COMP=(PKDEPOT,DEFDOSE) COMP=(PKCNTRL) COMP=(PDCORT) COMP=(PDLYN,DEFOBS) $PK "FIRST COMMON /PRCOMG/ IDUM1,IDUM2,IMAX,IDUM4,IDUM5 INTEGER IDUM1,IDUM2,IMAX,IDUM4,IDUM5 IMAX=100000000 ;PK MODEL K12 = IKA CL = ICL V2 = IV S2=V2/1000 K20=CL/V2 PT = 0.7 ; N t *Pt transcortin binding sites number PA = 500 ; N a *P a albumin binding sites number KTP = IKTP ; PNL transcortin binding affinity KTC = 30 ; Cortisol transcortin binding affinity KAP = 0.002 ; PNL albumin binding affinity KAC = 0.05 ; cortiso l albumin binding affinity ;CORT PD MODEL RMAX = IRMX ; maximum cortisol release rate TMIN = 16 ; time at minimum release rate TGAP = 4 ; time from minimum to maximum release rate KOCT = IKOC ; cortisol elimination rate constant T MAX = TMIN+TGAP CIM1= 0.6 ;methylprednisolone cortisol IMX CIM2 = 0.72 ;prednisolone cortisol IMX CIC1 = 1.02 ; total methylprednisolone IC50 CIC2 = 0.37 ;free prednisolone cortisol IC50 A_0(3) = BSLC ;LYM PD MODEL KILY = THETA(1)*EXP(ETA(1)) ; influx rate constant KOLY = THETA(2)*EXP(ETA(2)) ; efflux rate constant BSLY = THETA(3)*EXP(ETA(3)) ; Initial lymphocyte number ICD1 = THETA(4)*EXP(ETA(4)) ; total methylprednisolone LYN IC50
103 ICD2 = THETA(5)*EXP(ETA(5)) ;free prednisolone LYN IC50 ICC = THETA(6 )*EXP(ETA(6)) ; free cortisol LYN IC50 IMD1 = THETA(7) IMD2 = THETA(8) IMC = THETA(9) SD1 = THETA(10) A_0(4) = BSLY $DES DADT(1)= K12*A(1) DADT(2)= K12*A(1) K20*A(2) CP=A(2)/S2 T1 = T 24*INT(T/24) IF (T1.LE.TMIN) THEN T2=T1+24 ELSE T2=T1 ENDIF IF (T2.LE.TMAX) THEN KICT=RMAX*(T2 TMIN)/TGAP ELSE KICT=RMAX*(TMIN+24 T2)/(24 TGAP) ENDIF EFF1=(1 FLG)*(1 (CIM1*CP)/(CIC1+CP))+FLG*(1 (CIM2*CP)/(CIC2+CP)) DADT(3)= KICT*EFF1 KOCT*A(3) CCT=A(3) ;total cortisol conc. CTA=KTC/362.46*(1+PA*KAC) CTB=(PA*KAC+1)*(K TP*CP/360.444+1)+KTC*(PT CCT/362.46) CTC= CCT*(1+KTP*CP/360.444) CCF=(SQRT(CTB*CTB 4*CTA*CTC) CTB)/2/CTA ;free cortisol conc. IF (FLG.EQ.3) THEN EFF2=1 ENDIF IF (FLG.EQ.0) THEN ;MP drug effect EFF2=1 IMD1*(IMC*ICD1/(IMD1 IMC+IMD1*ICC/CCF)+CP)/(ICD1+(IM C*ICD1/(IMD1 IMC+IMD1*ICC/CCF)+CP)) ENDIF IF (FLG.EQ.1) THEN ;PNL drug effect EFF2=1 IMD2*(IMC*ICD2/(IMD2 IMC+IMD2*ICC/CCF)+CP)/(ICD2+(IMC*ICD2/(IMD2 IMC+IMD2*ICC/CCF)+CP)) ENDIF DADT(4)= KILY*EFF2 KOLY*A(4) $ERROR W=SD1 IPRE=F Y=F+W*EPS(1) IRES=DV IPRE
104 IWRE=IRES/W $THETA (0, 10, 500) ;KILY (0, 0.7, 10) ;KOLY (0, 30, 400) ;BSL Y (0, 10, 1000) ;ICD1 (0, 10, 1000) ;ICD2 12.1 FIX ;ICC 0.8 FIX ;IMD1 0.8 FIX ;IMD2 1 FIX ;IMC (0, 0.5, 100) ;SD 1 $OMEGA 0 FIX ;KILY 0.1 ;KOLY 0 FIX ;BSLY 0.1 ; ICD1 0.1 ;ICD2 0 FIX ;ICC $SIGMA 1 FIXED $EST MAX=9999 PRINT=10 METHOD=1 INT NOABORT $COV PRINT=E $TABLE ID DOSE TIME EVID IPRE IWRE MDV SID ONEHEADER NOPRINT FILE=SDTAB0 22 $TABLE ID DOSE KILY BSLY KOLY ICD1 ICD2 ICC SID ONEHEADER NOPRINT NOAPPEND FILE=PATAB0 22
105 APPENDIX E INDIVIDUAL FITTING P LOTS FOR MP PK DATA
108 APPENDIX F INDIVIDUAL FITTING P LOTS FOR PNL PK DATA WITH MODEL SIMP
111 APPENDIX G INDIVIDUAL FITTING PLOTS FOR MP CORTISOL DATA WITH M ODEL SIMP
114 APPENDIX H INDIVIDUAL FITTING P LOTS FOR PNL CORTISO L DATA WITH MODEL SIMP
117 APPENDIX I INDIVIDUAL FITTING P LOTS FOR MP LYMPHOCY TE DATA WITH MODEL SIM P
120 APPENDIX J INDIVIDUAL FITTING P LOTS FOR PNL LYMPHOC YTE DATA WITH MODEL SIMP
123 LIST OF REFERENCES 1. Veber DF, Johnson SR, Cheng HY, Smith BR, Ward KW, Kopple KD (2002) Molecular properties that influence the oral bioavailability of drug candidates. J Med Chem 45 (12):2615 2623 2. Lin JH, Lu AY (1997) Role of pharmacokinetics and metabolism in drug discovery and development. Pharmacol Rev 49 (4):403 449 3. Amidon GL, Lennernas H, Shah VP, Crison JR (1995) A theoreti cal basis for a biopharmaceutic drug classification: the correlation of in vitro drug product dissolution and in vivo bioavailability. Pharm Res 12 (3):413 420 4. FDA US (1995) Guidance for industry, immediate release solid oral dosage forms: scale up and post approval changes 5. Blaschke TF, Rubin PC (1979) Hepatic first pass metabolism in liver disease. Clin Pharmacokinet 4 (6):423 432 6. Alcorn J, McNamara PJ (2003) Pharmacokinetics in the newborn. Adv Drug Deliv Rev 55 (5):667 686 7. Hilmer SN, Shenfie ld GM, Le Couteur DG (2005) Clinical implications of changes in hepatic drug metabolism in older people. Ther Clin Risk Manag 1 (2):151 156 8. Mertens Talcott SU, Zadezensky I, De Castro WV, Derendorf H, Butterweck V (2006) Grapefruit drug interactions: can interactions with drugs be avoided? J Clin Pharmacol 46 (12):1390 1416 9. Wallace AW, Amsden GW (2002) Is it really OK to take this with food? Old interactions with a new twist. J Clin Pharmacol 42 (4):437 443 10. Liedholm H, Wahlin Boll E, Melander A (1990) Mechanisms and variations in the food effect on propranolol bioavailability. Eur J Clin Pharmacol 38 (5):469 475 11. Bekersky I, Dressler D, Mekki QA (2001) Effect of low and high fat meals on tacrolimus absorption following 5 mg single oral doses to healthy human subjects. J Clin Pharmacol 41 (2):176 182 12. Nagashima R, O'Reilly RA, Levy G (1969) Kinetics of pharmacologic effects in man: the anticoagulant action of warfarin. Clin Pharmacol Ther 10 (1):22 35 13. Dayneka NL, Garg V, Jusko WJ (1993) Comparison of four basic models of indirect pharmacodynamic responses. J Pharmacokinet Biopharm 21 (4):457 478 14. Sharma A, Jusko WJ (1998) Characteristics of indirect pharmacodynamic models and applications to clinical drug responses. Br J Clin Pharmacol 45 (3):229 239 15. Sheiner LB, Beal SL (1983) Evaluation of methods for estimating population pharmacokinetic parameters. III. Monoexponential model: routine clinical pharmacokinetic data. J Pharmacokinet Biopharm 11 (3):303 319 16. Sheiner LB, Beal SL (1 981) Evaluation of methods for estimating population pharmacokinetic parameters. II. Biexponential model and experimental pharmacokinetic data. J Pharmacokinet Biopharm 9 (5):635 651
124 17. Sheiner LB, Beal SL (1980) Evaluation of methods for estimating popul ation pharmacokinetics parameters. I. Michaelis Menten model: routine clinical pharmacokinetic data. J Pharmacokinet Biopharm 8 (6):553 571 18. http://www.mangosteen.com/ Accessed March 2 4, 2013 19. Obolskiy D, P ischel I, Siriwatanametanon N, Heinrich M (2009) Garcinia mangostana L.: a phytochemical and pharmacological review. Phytother Res 23 (8):1047 1065 20. Schmid W (1855) Ueber das Mangostin. Justus Liebigs Annalen der Chemie 93 (1):83 88 21. Nilar, Harrison LJ (2002) Xanthones from the heartwood of Garcinia mangostana Phytochemistry 60 (5):541 548 22. Suksamrarn S, Suwannapoch N, Ratananukul P, Aroonlerk N, Suksamrarn A (2002) Xanthones from the green fruit hulls of Garcinia mangostana J Nat Prod 65 (5):761 763 23. Pedraza Chaverri J, Cardenas Rodriguez N, Orozco Ibarra M, Perez Rojas JM (2008) Medicinal properties of mangosteen ( Garcinia mangostana ). Food Chem Toxicol 46 (10):3227 3239 24. Zhang Y, Song Z, Hao J, Qiu S, Xu Z (2010) Two new prenylated xantho nes and a new prenylated tetrahydroxanthone from the pericarp of Garcinia mangostana Fitoterapia 81 (6):595 599 25. Zhao Y, Liu JP, Lu D, Li PY, Zhang LX (2010) A new antioxidant xanthone from the pericarp of Garcinia mangostana Linn. Nat Prod Res 24 (17):1664 1670 26. Zhao Y, Liu JP, Lu D, Li PY, Zhang LX (2012) Two new xanthones from the pericarp of Garcinia mangostana Nat Prod Res 26 (1):61 65 27. Jefferson AQA, Scheimann F, Sim KY (1970) Isolation of gamma mangostin from Garcinia mangostana an d preparation of the natural mangostins by selective demethylation. Aust J Chem 23:2539 2543 28. Shan Y, Zhang W (2010) Preparative separation of major xanthones from mangosteen pericarp using high performance centrifugal partition chromatography. J Sep Sc i 33 (9):1274 1278 29. Pothitirat W, Gritsanapan W (2008) Thin layer chromatography densitometric analysis of alpha mangostin content in Garcinia mangostana fruit rind extracts. J AOAC Int 91 (5):1145 1148 30. Pothitirat W, Chomnawang MT, Supabphol R, Grit sanapan W (2010) Free radical scavenging and anti acne activities of mangosteen fruit rind extracts prepared by different extraction methods. Pharm Biol 48 (2):182 186 31. Chin YW, Kinghorn AD (2008) Structural c haracterization, b iological e ffects, and s yn thetic s tudies on x anthones from m angosteen ( Garcinia mangostana ), a p opular b otanical d ietary s upplement. Mini Rev Org Chem 5 (4):355 364 32. Ngawhirunpat T, Opanasopi P, Sukma M, Sittisombut C, Kat A, Adachi I Antioxidant, free radical scavenging activity and cytotoxicity of different solvent extracts and their
125 phenolic constituents from the fruit hull of mangosteen ( Garcinia mangostana ). P harm Biol 48 (1):55 62 33. Weecharangsan W, Opanasopit P, Sukma M, Ngawhirunpat T, Sotanaphun U, Siripong P (2006) Antioxidative and neuroprotective activities of extracts from the fruit hull of mangosteen ( Garcinia mangostana Linn.). Med Princ Pract 15 (4 ):281 287 34. Sun D, Zhang S, Wei Y, Yin L (2009) Antioxidant activity of mangostin in cell free system and its effect on K562 leukemia cell line in photodynamic therapy. Acta biochimica et biophysica Sinica 41 (12):1033 1043 35. Chairungsrilerd N, Furukaw a K, Ohta T, Nozoe S, Ohizumi Y (1996) Pharmacological properties of alpha mangostin, a novel histamine H1 receptor antagonist. Eur J Pharmacol 314 (3):351 356 36. Romier Crouzet B, Van De Walle J, During A, Joly A, Rousseau C, Henry O, Larondelle Y, Schne ider YJ (2009) Inhibition of inflammatory mediators by polyphenolic plant extracts in human intestinal Caco 2 cells. Food Chem Toxicol 47 (6):1221 1230 37. Moongkarndi P, Kosem N, Kaslungka S, Luanratana O, Pongpan N, Neungton N (2004) Antiproliferation, a ntioxidation and induction of apoptosis by Garcinia mangostana (mangosteen) on SKBR3 human breast cancer cell line. J Ethnopharmacol 90 (1):161 166 38. Moongkarndi P, Kosem N, Luanratana O, Jongsomboonkusol S, Pongpan N (2004) Antiproliferative activity of Thai medicinal plant extracts on human breast adenocarcinoma cell line. Fitoterapia 75 (3 4):375 377 39. Chomnawang MT, Surassmo S, Wongsariya K, Bunyapraphatsara N (2009) Antibacterial activity of Thai medicinal plants against methicillin resistant Staphylococcus aureus Fitoterapia 80 (2):102 104 40. Chen SX, Wan M, Loh BN (1996) Active constituents against HIV 1 protease from Garcinia mangostana Planta medica 62 (4):381 382 41. Vlietinck AJ, De Bruyne T, Apers S, Pieters LA (1998) Plant derived le ading compounds for chemotherapy of human immunodeficiency virus (HIV) infection. Planta Med 64 (2):97 109 42. Jung HA, Su BN, Keller WJ, Mehta RG, Kinghorn AD (2006) Antioxidant xanthones from the pericarp of Garcinia mangostana (Mangosteen). J Agric Food Chem 54 (6):2077 2082 43. Bumrungpert A, Kalpravidh RW, Chitchumroonchokchai C, Chuang CC, West T, Kennedy A, McIntosh M (2009) Xanthones from mangosteen prevent lipopolysaccharide mediated inflammation and insulin resistance in primary cultures of human adipocytes. J Nutr 139 (6):1185 1191 44. Bumrungpert A, Kalpravidh RW, Chuang CC, Overman A, Martinez K, Kennedy A, McIntosh M (2009) Xanthones from mangosteen inhibit inflammation in human macrophages and in human adipocytes exposed to macrophage conditio ned media. J Nutr 140 (4):842 847
126 45. Tewtrakul S, Wattanapiromsakul C, Mahabusarakam W (2009) Effects of compounds from Garcinia mangostana on inflammatory mediators in RAW264.7 macrophage cells. J E thnopharmacol 121 (3):379 382 46. Nakagawa Y, Iinuma M, Naoe T, Nozawa Y, Akao Y (2007) Characterized mechanism of alpha mangostin induced cell death: caspase independent apoptosis with release of endonuclease G from mitochondria and increased miR 143 expression in human colorectal cancer DLD 1 cells. Bioorg Me d Chem 15 (16):5620 5628 47. Chang HF, Huang WT, Chen HJ, Yang LL (2010) Apoptotic effects of gamma mangostin from the fruit hull of Garcinia mangostana on human malignant glioma cells. Molecules 15 (12):8953 8966 48. Balunas MJ, Su B, Brueggemeier RW, Kinghorn AD (2008) Xanthones from the botanical dietary supplement mangosteen ( Garcinia mangostana ) with aromatase inhibitory activity. J Nat Prod 71 (7):1161 1166 49. NBJ's Supplement Business Report (2009). Penton Media Inc., Nutrition Business Journal 5 0. Kondo M, Zhang L, Ji H, Kou Y, Ou B (2009) Bioavailability and antioxidant effects of a xanthone rich Mangosteen ( Garcinia mangostana ) product in humans. J Agric Food Chem 57 (19):8788 8792 51. FDA US (2001) Bioanalytical Method Validation Guidance for Industry. Available at http://www.fda.gov/downloads/Drugs/GuidanceComplianceRegulatoryInformation/Guida nces/ucm070107.pdf Accessed March 2 4 ,2013 52. Chen LG, Yang LL, Wang CC (2008) Anti inflammatory activity of mangostins from Garcinia mangostana Food Chem Toxicol 46 (2):688 693 53. Chitchumroonchokchai C, Riedl KM, Suksumrarn S, Clinton SK, Kinghorn AD, Failla ML (2012) Xanthones in mang osteen juice are absorbed and partially conjugated by healthy adults. J Nutr 142 (4):675 680 54. Terao J, Murota K, Kawai Y (2011) Conjugated quercetin glucuronides as bioactive metabolites and precursors of aglycone in vivo Food Funct 2 (1):11 17 55. Shi moi K, Okada H, Furugori M, Goda T, Takase S, Suzuki M, Hara Y, Yamamoto H, Kinae N (1998) Intestinal absorption of luteolin and luteolin 7 O beta glucoside in rats and humans. FEBS Lett 438 (3):220 224 56. Manach C, Williamson G, Morand C, Scalbert A, Remesy C (2005) Bioavailability and bioefficacy of polyphenols in humans. I. Review of 97 bioavailability studies. Am J Clin Nutr 81 (1 Suppl):230S 242S 57. Gibson G, Skett P (2001) Introduction to Drug Met abolism. 58. Mulder GJ (1992) Glucuronidation and its role in regulation of biological activity of drugs. Annu Rev Pharmacol Toxicol 32:25 49 59. Han AR, Kim JA, Lantvit DD, Kardono LB, Riswan S, Chai H, Carcache de Blanco EJ, Farnsworth NR, Swanson SM, K inghorn AD (2009) Cytotoxic xanthone constituents of the stem bark of Garcinia mangostana (mangosteen). J Nat Prod 72 (11):2028 2031
127 60. Sampath PD, Vijayaragavan K (2008) Ameliorative prospective of alpha mangostin, a xanthone derivative from Garcinia man gostana against beta adrenergic cathecolamine induced myocardial toxicity and anomalous cardiac TNF alpha and COX 2 expressions in rats. Exp Toxicol Pathol 60 (4 5):357 364 61. Shibata MA, Iinuma M, Morimoto J, Kurose H, Akamatsu K, Okuno Y, Akao Y, Otsuki Y (2011) A lpha m angostin extracted from the pericarp of the mangosteen ( Garcinia mangostana Linn) reduces tumor growth and lymph node metastasis in an immunocompetent xenograft model of metastatic mammary cancer carrying a p53 mutation. BMC Med 9 (1):69 6 2. Nabandith V, Suzui M, Morioka T, Kaneshiro T, Kinjo T, Matsumoto K, Akao Y, Iinuma M, Yoshimi N (2004) Inhibitory effects of crude alpha mangostin, a xanthone derivative, on two different categories of colon preneoplastic lesions induced by 1, 2 dimethy lhydrazine in the rat. Asian Pac J Cancer Prev 5 (4):433 438 63. Watanapokasin R, Jarinthanan F, Jerusalmi A, Suksamrarn S, Nakamura Y, Sukseree S, Uthaisang Tanethpongtamb W, Ratananukul P, Sano T (2010) Potential of xanthones from tropical fruit mangoste en as anti cancer agents: caspase dependent apoptosis induction in vitro and in mice. Appl Biochem Biotechnol 162 (4):1080 1094 64. Doi H, Shibata MA, Shibata E, Morimoto J, Akao Y, Iinuma M, Tanigawa N, Otsuki Y (2009) Panaxanthone isolated from pericarp of Garcinia mangostana L. suppresses tumor growth and metastasis of a mouse model of mammary cancer. Anticancer Res 29 (7):2485 2495 65. Kosem N, Ichikawa K, Utsumi H, Moongkarndi P (2012) In vivo toxicity and antitumor activity of mangosteen extract. J Nat Med 66. Chitchumroonchokchai C, Thomas Ahner JM, Li J, Riedl KM, Nontakham J, Suksumrarn S, Clinton SK, Kinghorn AD, Failla ML (2012) Anti tumorigenicity of dietary alpha mangostin in an HT 29 colon cell xenograft model and the tissue distribution of xanthones and their phase II metabolites. Molecular nutrition & food research 67. Mohamed ME, Frye RF (2011) Effects of herbal supplements on drug glucuronidation. Review of clinical, animal, and in vitro studies. Planta Med 77 (4):311 321 68. Czock D, Keller F, Rasche FM, Haussler U (2005) Pharmacokinetics and pharmacodynamics of systemically administered glucocorticoids. Clin Pharmacokinet 44 (1):61 98 69. Gordijn MS, Gemke RJ, van Dalen EC, Rottev eel J, Kaspers GJ (2012) Hypothalamic pituitary adrenal (HPA) axis suppression after treatment with glucocorticoid therapy for childhood acute lymphoblastic leukaemia. Cochrane Database Syst Rev 5:CD008727 70. Chow FS, Sharma A, Jusko WJ (1999) Modeling in teractions between adrenal suppression and T helper lymphocyte trafficking during multiple dosing of methylprednisolone. J Pharmacokinet Biopharm 27 (6):559 575 71. Krasner AS (1999) Glucocorticoid induced adrenal insufficiency. JAMA 282 (7):671 676
128 72. Ca nalis E, Mazziotti G, Giustina A, Bilezikian JP (2007) Glucocorticoid induced osteoporosis: pathophysiology and therapy. Osteoporos Int 18 (10):1319 1328 73. Beaudry JL, Riddell MC (2012) Effects of glucocorticoids and exercise on pancreatic beta cell func tion and diabetes development. Diabetes Metab Res Rev 28 (7):560 573 74. Alhabet SMH, Rogers HJ (1989) Methylprednisolone p harmacokinetics after i ntravenous and o ral a dministration. Brit J Clin Pharmaco 27 (3):285 290 75. Garg V, Blum R, Wilner KD, Jusko W J (1992) Effect of the anti inflammatory agent tenidap on the pharmacokinetics and pharmacodynamics of prednisolone. J Clin Pharmacol 32 (3):222 230 76. Derendorf H, Mollmann H, Rohdewald P, Rehder J, Schmidt EW (1985) Kinetics of methylprednisolone and it s hemisuccinate ester. Clin Pharmacol Ther 37 (5):502 507 77. Rose JQ, Yurchak AM, Jusko WJ (1981) Dose dependent pharmacokinetics of prednisone and prednisolone in man. J Pharmacokinet Biopharm 9 (4):389 417 78. Patel PM, Selby PJ, Graham MA, Viner C, New ell DR, Mcelwain TJ (1993) Pharmacokinetics of h igh d ose m ethylprednisolone and u se in h ematological m alignancies. Hematol Oncol 11 (2):89 96 79. Wald JA, Law RM, Ludwig EA, Sloan RR, Middleton E, Jr., Jusko WJ (1992) Evaluation of dose related pharmacokin etics and pharmacodynamics of prednisolone in man. J Pharmacokinet Biopharm 20 (6):567 589 80. Rohatagi S, Hochhaus G, Mollmann H, Barth J, Derendorf H (1995) Pharmacokinetic interaction between endogenous cortisol and exogenous corticosteroids. Pharmazie 50 (9):610 613 81. Szefler SJ, Ebling WF, Georgitis JW, Jusko WJ (1986) Methylprednisolone versus prednisolone pharmacokinetics in relation to dose in adults. Eur J Clin Pharmacol 30 (3):323 329 82. Rohatagi S, Barth J, Mollmann H, Hochhaus G, Soldner A, M ollmann C, Derendorf H (1997) Pharmacokinetics of methylprednisolone and prednisolone after single and multiple oral administration. J Clin Pharmacol 37 (10):916 925 83. Hong Y, Mager DE, Blum RA, Jusko WJ (2007) Population pharmacokinetic/pharmacodynamic modeling of systemic corticosteroid inhibition of whole blood lymphocytes: modeling interoccasion pharmacodynamic variability. Pharm Res 24 (6):1088 1097 84. Magee MH, Blum RA, Lates CD, Jusko WJ (2001) Prednisolone pharmacokinetics and pharmacodynamics in relation to sex and race. J Clin Pharmacol 41 (11):1180 1194 85. Mager DE, Lin SX, Blum RA, Lates CD, Jusko WJ (2003) Dose equivalency evaluation of major corticosteroids: pharmacokinetics and cell trafficking and cortisol dynamics. J Clin Pharmacol 43 (1 1):1216 1227 86. Chakraborty A, Blum RA, Mis SM, Cutler DL, Jusko WJ (1999) Pharmacokinetic and adrenal interactions of IL 10 and prednisone in healthy volunteers. J Clin Pharmacol 39 (6):624 635
129 87. Magee MH, Blum RA, Lates CD, Jusko WJ (2002) Pharmacokinetic/pharmacodynamic model for prednisolone inhibition of whole blood lymphocyte proliferation. Br J Clin Pharmacol 53 (5):474 484 88. Booker BM, Magee MH, Blum RA, Lates CD, Jusko WJ (2002) Pharmacokinetic and pharmacodynamic interactions betwe en diltiazem and methylprednisolone in healthy volunteers. Clin Pharmacol Ther 72 (4):370 382 89. Chakraborty A, Krzyzanski W, Jusko WJ (1999) Mathematical modeling of circadian cortisol concentrations using indirect response models: comparison of several methods. J Pharmacokinet Biopharm 27 (1):23 43 90. Rohatagi S, Bye A, Mackie AE, Derendorf H (1996) Mathematical modeling of cortisol circadian rhythm and cortisol suppression. Eur J Pharm Sci 4 (6):341 350 91. Xu J, Winkler J, Derendorf H (2007) A pharmac okinetic/pharmacodynamic approach to predict total prednisolone concentrations in human plasma. J Pharmacokinet Pharmacodyn 34 (3):355 372 92. Wu K, Goyal N, Stark JG, Hochhaus G (2008) Evaluation of the administration time effect on the cumulative cortiso l suppression and cumulative lymphocytes suppression for once daily inhaled corticosteroids: a population modeling/simulation approach. J Clin Pharmacol 48 (9):1069 1080 93. Stark JG, Werner S, Homrighausen S, Tang Y, Krieg M, Derendorf H, Moellmann H, Hochhaus G (2006) Pharmacokinetic/pharmacodynamic modeling of total lymphocytes and selected subtypes after oral budesonide. J Pharmacokinet Pharmacodyn 33 (4):441 459 94. Xu J, Winkler J, Sabarinath SN, Derendorf H (2008) Assessment of the impact of dosin g time on the pharmacokinetics/pharmacodynamics of prednisolone. AAPS J 10 (2):331 341 95. Ferry JJ, Della Coletta AA, Weber DJ, VanderLugt JT (1994) Pilot study of the pharmacokinetics of methylprednisolone after single and multiple intravenous doses of m ethylprednisolone sodium succinate and methylprednisolone suleptanate to healthy volunteers. J Clin Pharmacol 34 (11):1109 1115 96. Meibohm B, Hochhaus G, Mollmann H, Barth J, Wagner M, Krieg M, Stockmann R, Derendorf H (1999) A pharmacokinetic/pharmacodyn amic approach to predict the cumulative cortisol suppression of inhaled corticosteroids. J Pharmacokinet Biopharm 27 (2):127 147 97. Earp J, Krzyzanski W, Chakraborty A, Zamacona MK, Jusko WJ (2004) Assessment of drug interactions relevant to pharmacodynam ic indirect response models. J Pharmacokinet Pharmacodyn 31 (5):345 380 98. Boudinot FD, Jusko WJ (1984) Plasma protein binding interaction of prednisone and prednisolone. J Steroid Biochem 21 (3):337 339 99. Rocci ML, Jr., D'Ambrosio R, Johnson NF, Jusko WJ (1982) Prednisolone binding to albumin and transcortin in the presence of cortisol. Biochemical pharmacology 31 (3):289 292
130 100. Legler UF, Benet LZ (1986) The effect of prednisone and hydrocortisone on the plasma protein binding of prednisolone in man. Eur J Clin Pharmacol 30 (1):51 55 101. Liddle GW (1966) An analysis of circadian rhythms in human adrenocortical secretory activity. Trans Am Clin Climatol Assoc 77:151 160 102. Levi FA, Canon C, Touitou Y, Sulon J, Mechkouri M, Ponsart ED, Touboul JP, Va nnetzel JM, Mowzowicz I, Reinberg A, et al. (1988) Circadian rhythms in circulating T lymphocyte subtypes and plasma testosterone, total and free cortisol in five healthy men. Clin Exp Immunol 71 (2):329 335 103. Tavadia HB, Fleming KA, Hume PD, Simpson HW (1975) Circadian rhythmicity of human plasma cortisol and PHA induced lymphocyte transformation. Clin Exp Immunol 22 (1):190 193 104. Rohdewald P, Mllmann H, Muller M, Hochhaus G (1984) Glukokortikoid Rezeptoren in den Atemwegen. Bochumer Treff:223 242 1 05. Blumenthal M, Lindstrom A, Ooyen C, Lynch ME (2012) Herb Supplement Sales Increase 4.5% in 2011. HerbalGram (95):60 64 10 6 Levinson DR (2012) Dietary Supplements: Structure/Function Claims Fail To Meet Federal Requirements. Department of Health and Hu man Services Available at https://oig.hhs.gov/oei/reports/oei 01 11 00210.asp Accessed March 24, 2013 10 7 http://www.xango.co m/products/xango juice Accessed March 24, 2013 10 8 http://www.vemmajuice.eu/ Accessed March 24, 2013
131 BIOGRAPHICAL SKETCH Li Li was born in December 1982 in China. In 200 3 degree in Biology from Wuhan University Hubei China. Then s he stayed as Research Assistant in State Key Laboratory of Virology at Wuhan University In 200 6 s he came to the University of Florida studying for h er master s degree in Horticultural Sciences. Her Multiple Non redundant Roles for Plastidic 6 Phosphogluconate dehydrogenase (6PGDH) in Maize Li Li was enrolled as a Ph.D student in Pharmaceutics Department at University of Florida in 2009 She has finished a four month internship during her Ph.D. study in the department of Biotherapeutics Clinical Pharmacology at Pfizer Cambridge, MA. Li completed her Ph.D. in May 201 3