Citation
Topical delivery of a model phenolic compound

Material Information

Title:
Topical delivery of a model phenolic compound
Abbreviated Title:
: alkyoxycarbonyl prodrugs of acetaminophen
Creator:
Wasdo, Scott C. ( Dissertant )
Perrin, John H. ( Thesis advisor )
Place of Publication:
Gainesville, Fla.
Publisher:
University of Florida
Publication Date:
Copyright Date:
2005
Language:
English

Subjects

Subjects / Keywords:
Database models ( jstor )
Databases ( jstor )
Lipids ( jstor )
Melting points ( jstor )
Modeling ( jstor )
Molecular weight ( jstor )
Molecules ( jstor )
Prodrugs ( jstor )
Skin ( jstor )
Solubility ( jstor )
Dissertations, Academic -- UF -- Medicinal Chemistry
Medicinal Chemistry thesis, Ph.D
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Abstract:
Topical delivery is an attractive route of administration for a number of therapeutic agents. However, many drugs possess physical and chemical properties that limit their ability to permeate the skin. By masking select functional groups on a drug with proper moieties, it is possible to create prodrugs with physical and chemical properties that greatly improve topical delivery. Efficient development of these prodrugs requires knowledge of how the physical and chemical characteristics of a drug influence dermal absorption. In response to this need, solubility and diffusion experiments were performed on prodrugs of 5-flurouracil, 6- mercaptopurine and theophylline to develop a model that predicts maximum flux of these compounds through hairless mouse skin when delivered from isopropyl myristate (IPM). The purpose of this work was to expand this database to include phenol containing compounds and refine this model by synthesizing and characterizing a series of alkyloxycarbonyl derivatives and a set of methoxyalkyloxycarbonyl derivatives of a model phenolic compound, acetaminophen (APAP). In addition, two new predictive flux models, a solubility based model to predict flux from water and a new model based upon solubility and flux through polydimethylsiloxane (PDMS) membrane as an additional parameter, were developed. As with the other prodrug series studied, the acetaminophen prodrugs with the best combination of lipid (SIPM) and water (SAQ) solubility, the methyloxycarbonyl and the methoxyethyloxy carbonyl derivatives, had the highest flux through mouse skin from both IPM (JMAQ) and water (JMIPM). The addition of APAP and its prodrugs to the IPM and aqueous databases produced Roberts-Sloan equations of log JMIPM = -0.501 + 0.517 log SIPM + (1 - 0.517) log SAQ -0. 00266 MW and log JMAQ = -1.665+ 0.657 log SIPM + (1 - 0.657) log SAQ -0.00409 MW for flux through hairless mouse skin from IPM and water, respectively. These models can predict flux of a drug through hairless mouse skin from IPM with an average error of prediction of 0.16 log units and from water with an average error of 0.17 log units. A simple model was found to relate flux through PDMS membrane (JPAQ) and flux through hairless mouse skin from water. Using the equation, log JMAQ = -1.156 + 0.245 log SAQ + 0.409 log JPAQ, flux from water through hairless mouse skin was predicted with an average absolute error of 0.13 log units. ( , )
Subject:
acetaminophen, alkyloxycarbonyl, modeling, mouse, pdms, polymer, predictive, prodrugs, topical
General Note:
Title from title page of source document.
General Note:
Document formatted into pages; contains 137 pages.
General Note:
Includes vita.
Thesis:
Thesis (Ph.D.)--University of Florida, 2005.
Bibliography:
Includes bibliographical references.
General Note:
Text (Electronic thesis) in PDF format.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright Wasdo, Scott C.. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Embargo Date:
7/30/2007
Resource Identifier:
003398417 ( aleph )

Downloads

This item has the following downloads:


Full Text












TOPICAL DELIVERY OF A MODEL PHENOLIC COMPOUND:
ALKYLOXYCARBONYL PRODRUGS OF ACETAMINOPHEN















By

SCOTT C. WASDO


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


2005

































Copyright 2005

by

Scott C Wasdo

































This document is dedicated to my parents Charles and Barbara, my brother Shaun and my
sister Christine.















ACKNOWLEDGMENTS

Looking back, it is clear that my success has depended upon the kindness and

support of more friends and coworkers then I could possibly name here. Even so, a few

individuals were so profoundly important that they deserve a special mention. First, I

would like to thank my parents, Charles and Barbara for fostering my love of science and

for providing me with a lifetime of unwavering support. In addition, I would like to

thank Charles Schmidt and Ian Tebbett, for their advise and the early opportunities they

provided, Nancy Szabo for years of encouragement and many campaigns on my behalf,

Carolynn Diaz, for always carrying more than her share of the burden, my committee

members Margaret O. James and Stephen M. Roberts, who have been helpful and

accommodating of my many idiosyncrasies and John Perrin, who selflessly agreed to

accept the responsibility of chairing my committee with the fore knowledge that my work

would do little to advance his own research.

Most of all, I would like to thank Kenneth B. Sloan who, above all others, has been

instrumental in this accomplishment. He has shown tremendous patience and dedication

to my progress and has been invaluable. I could not have found a better mentor.
















TABLE OF CONTENTS

page

A C K N O W L E D G M E N T S ................................................................................................. iv

LIST OF TABLES ............... ................. ........ ................ .................. vii

LIST OF FIGURES ......... ........................................... ............ ix

ABSTRACT ........ .............. ............. ...... ...................... xi

CHAPTER

1 BACKGROUND AND SPECIFIC AIMS ..................................... ................1

The C ase for Topical D delivery .................................. ..........................................
Approaches to Increasing Topical Delivery ...................... ............................... 3
Basic Theoretical Considerations of Prodrug Design for Topical Delivery and
Previously Synthesized Prodrugs....................................... .......................... 6
Specific Objectives and Preliminary W ork ............... .......................................... 13
F first O bjectiv e ......... ............................................................ ............................13
S econ d O bjectiv e ........... ... .................................................. .. .................... 16
T h ird O bj ectiv e ................................................. ................ 18

2 BASIC ANATOM Y OF THE SKIN ................................... .................................... 22

H ypoderm is ..................................................................................................... ........22
D erm is ................................................................................ 2 3
Epiderm is .................... .................................... ........ ..... ..... ......... 26

3 DEVELOPMENT OF THE PREDICTIVE SOLUBILITY AND FLUX MODELS 33

Derivation of the Series Specific Organic and Aqueous Solubility Equations ..........33
Derivation of the Potts-Guy Equation ....... ....................................39
Derivation of the Roberts-Sloan Equation ............................................ ......... .......43
Modification of the Roberts-Sloan Equation to Include Synthetic Membrane Data..44

4 EXPERIM ENTAL DESIGN ......................................................... .............. 49

Section I: Synthesis and Characterization of the 4-Alkyloxycarbonyl and 4-
Methyloxyalkyloxy Prodrugs of Acetaminophen.....................................49









Synthesis .............................................. 49
C haracterization .................................. ................................ 52
Section II: Determination of IPM and Aqueous Solubilities..........................53
Section III: Determination of Flux through Hairless Mouse Skin and
Polydimethylsiloxane M embranes......... ...................... ..... ................. 55
Preparation of the Membranes and Assembly of the Diffusion cells.................55
Preparation and Application of Donor Phases............................................... 58
Sampling of the Difussion cells for Flux and Residual Skin Samples ..............58
Evaluation of M em brane Integrity ............................................. ............... 60
Determination of Analyte Concentration and Extent of Hydrolysis ................60
Determination of M aximum Flux (JM) ................... ..... ............................ 61

5 R E S U L T S .......................................................................... 6 3

Melting Point Behavior of the APAP Carbonates .................................................63
D direct IPM and A queous Solubility ........................................ ....................... 64
Partition Coefficients and Solubility Ratios .................................... ............... 66
Perm ability Coefficient B behavior .................................. .............................. ....... 68
Conversion to the Parent Drug ........................... ......... ..... ..................70
Flux of APAP and its AOC and MOAOC Prodrugs through Hairless Mouse Skin...71
Effect of the Vehicle on Flux through Hairless Mouse Skin................ .......... 75
Residual M embrane Am ounts .............................................................................76

6 PREDICTIVE MODELS OF SOLUBILITY AND FLUX.................. ......... 79

Determination of the Coefficients of the General Solubility Equations...................79
Solubility Behavior of the 4-AOC and 4-MOAOC-APAP Prodrugs .......................84
Modeling the Flux of the 4-AOC and 4-MOAOC APAP Prodrugs through
Hairless M house Skin from IPM and W ater...................... ....................... ......89
Modeling the Flux of the 4-AOC and 4-MOAOC APAP Prodrugs through PDMS
Polymer M embrane from W ater ............. .. ... ............. ........... ........ ...... 100
Prediction of Flux through Hairless Mouse Skin from Flux through PDMS...........109

7 CONCLUSIONS AND FUTURE WORK.................. .......... ..................112

L IST O F R E F E R E N C E S ......... .. ............... ................. .............................................. 118

B IO G R A PH ICA L SK ETCH ......... ................. ...................................... .....................125















LIST OF TABLES


Table pge

4-1 1H NMR Data for the AOC and MOAOC-APAP prodrugs .................................... 53

4-2 Melting point and absorptivity values for APAP, the AOC-APAP and MOAOC
A PA P prodrugs. ...............................................................53

4-3 Sampling times for APAP and the AOCA prodrug diffusion experiments ............59

5-1 Melting points (C), log partition coefficients (log KIPM:AQ), log solubility ratio
(log SRIPM:AQ), log solubilities in IPM (log SIPM ) and log solubilities in water
(log SAQ) for APAP and the 4-AOC and 4-MOAOC prodrugs.............................. 63

5-2 Log permeability values for the APAP prodrugs from IPM through hairless
mouse skin (log PMIPM), from water through hairless mouse skin (log PMAQ) and
from water through PDMS membrane (log PPAQ) ........... .......................68

5-3 Maximum steady-state flux and second application flux of the AOC-APAP and
MOAOC-APAP prodrugs through hairless mouse skin and PDMS membrane......72

5-4 Average residual amounts (+ Std. Dev.) of APAP and its prodrugs remaining in
hairless mouse skin (HMS) and PDMS membrane after the flux experiments........76

6-1 Series specific best fit coefficients to equation 3.10 using IPM solubility data
and physical properties from the 5-FU and 6-MP prodrugs ...............................79

6-2 Series specific best-fit coefficients to equation 3.14 using aqueous solubility
data and physical properties from the 5-FU, 6-MP and phenytoin (PhT)
p ro d ru g s ...................................................................... 8 0

6-3 Predicted IPM and water solubilities for compounds 3-6 .......................................87

6-4 Estimated AIPM and AAQ using data from compounds 7 and 8...............................89

6-5 Predicted and experimental flux values for compounds 1-8 through hairless
m house skin from IPM ........................... .................... ... ........ .. .... ...... ...... 90

6-6 Predicted and experimental flux values for compounds 1-8 through hairless
m ou se skin from w after. .................................................................. ....................94









6-7 Molecular weights, log IPM solubilities, log aqueous solubilities, log maximum
flux values through hairless mouse (log JMAQ) and log maximum flux values
through PDMS from water for the chemically stable prodrug series.....................101

6-8 Calculated maximum log flux values through PDMS from water and errors of
calculation for the chemically stable prodrug series. ...........................................105

6-9 Solubility, molecular weight and flux data for the PABA esters...........................107
















LIST OF FIGURES


Figure pge

1-1 Structures of the database parent molecules and the pKa values for potential
prom oiety attachm ent sites .......................................................................... ......... 8

1-2 Probable hydrolysis mechanisms of the 1-N-acyl 5-FU prodrugs at pH 7.4. ............9

1-3 Probable hydrolysis mechanisms of the soft alkyl promoieties. ............................10

1-4 Structure of naringenin ......................................................................... 15

2-2 Structures of the principal ceramides comprising the lipid bilayers of the
lam ellar bodies. .......................................................................29

3-1 Simplified diagram of an experimental diffusion apparatus. ..................................40

4-1 Regions of the 4-AOCO-ACA prodrug corresponding to the letters given in
ta b le 4 1 ........................................................................................5 2

4-2 Fully assem bled diffusion cell.......................................... ............................ 57

4-3 Plot of cumulative amount of prodrug delivered versus time for compound 3
from IPM through hairless mouse skin. ...................................... ............... 62

5-1 Plots of log KIPM:4.0 versus log P for compounds 2 through 8 using flux data from
the three m em brane /vehicle system s ............................................ ............... 69

5-2 Correlation between solubility and flux for APAP, its' AOC derivatives and its'
M O A O C derivatives. ...................................................................... ................... 72

6-1 Calculated versus experimental log IPM solubility using equation 3.10 and A
coefficients determined from the smallest series members............................... 82

6-2 Calculated versus experimental log SAQ using equation 3.14 and A coefficients
determined from the smallest series members. .............................. ............... .83

6-3 Predicted versus experimental IPM solubilities for the 4-AOC-APAP prodrugs
(A IpM = 2 .63). ...................................................... ................. 88

6-4 Predicted versus experimental aqueous solubilities for the 4-AOC-APAP
prodrugs (A AQ = 11.76) .................................................................. ................ .. 88









6-5 Experimental versus calculated log maximum flux values through hairless
mouse skin from IPM using equation 6.5. .................................... .................91

6-6 Calculated versus experimental log maximum flux values through hairless
mouse skin from water using equation 6-7. .................................. .................94

6-7 Calculated versus experimental log maximum flux values through PDMS from
w ater by E quation 6.13................................................ ............................... 102

6-8 Log IPM solubility versus log maximum flux through PDMS membrane from
w ater for the aqueous database........................................ ........................... 104

6-6 Calculated versus experimental log maximum flux values through hairless
m house skin from w ater by equation 6.23. ........... ................ ..... .......... ..........111















Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

TOPICAL DELIVERY OF A MODEL PHENOLIC COMPOUND:
ALKYLOXYCARBONYL PRODRUGS OF ACETAMINOPHEN

By

Scott C. Wasdo

August 2005

Chair: John Perrin
Major Department: Medicinal Chemistry

Topical delivery is an attractive route of administration for a number of therapeutic

agents. However, many drugs possess physical and chemical properties that limit their

ability to permeate the skin. By masking select functional groups on a drug with proper

moieties, it is possible to create prodrugs with physical and chemical properties that

greatly improve topical delivery. Efficient development of these prodrugs requires

knowledge of how the physical and chemical characteristics of a drug influence dermal

absorption. In response to this need, solubility and diffusion experiments were performed

on prodrugs of 5-flurouracil, 6-mercaptopurine and theophylline to develop a model that

predicts maximum flux of these compounds through hairless mouse skin when delivered

from isopropyl myristate (IPM). The purpose of this work was to expand this database to

include phenol containing compounds and refine this model by synthesizing and

characterizing a series of alkyloxycarbonyl derivatives and a set of

methoxyalkyloxycarbonyl derivatives of a model phenolic compound, acetaminophen









(APAP). In addition, two new predictive flux models, a solubility based model to predict

flux from water and a new model based upon solubility and flux through

polydimethylsiloxane (PDMS) membrane as an additional parameter, were developed.

As with the other prodrug series studied, the acetaminophen prodrugs with the best

combination of lipid (SIPM) and water (SAQ) solubility, the methyloxycarbonyl and the

methoxyethyloxy carbonyl derivatives, had the highest flux through mouse skin from

both IPM (JMAQ) and water (JMIPM). The addition of APAP and its prodrugs to the IPM

and aqueous databases produced Roberts-Sloan equations of log JMIPM = -0.501 + 0.517

log SIPM + (1 0.517) log SAQ-0.00266 MW and log JMAQ = -1.665+ 0.657 log SIPM + (1 -

0.657) log SAQ -0.00409 MW for flux through hairless mouse skin from IPM and water,

respectively. These models can predict flux of a drug through hairless mouse skin from

IPM with an average error of prediction of 0.16 log units and from water with an average

error of 0.17 log units. A simple model was found to relate flux through PDMS

membrane (JPAQ) and flux through hairless mouse skin from water. Using the equation,

log JMAQ = -1.156 + 0.245 log SAQ + 0.409 log JPAQ, flux from water through hairless

mouse skin was predicted with an average absolute error of 0.13 log units.














CHAPTER 1
BACKGROUND AND SPECIFIC AIMS

The Case for Topical Delivery

The route of administration can have a profound effect upon the utility of a drug.

For a given drug, the preferred method of administration will be determined by that

method's influence on a number of factors including bioavailability, rate of drug delivery,

ability to reach the physiological target and patient compliance. Historically, oral

delivery has been the most common means of drug administration and it continues to be

the preferred target route when developing new pharmaceutical formulations. However,

despite this prevalence, there are numerous situations where an alternative route, namely

topical delivery, has advantages.

Topical delivery is an obvious consideration when formulating therapies to treat

skin diseases or disorders. The accessibility of the skin makes site-specific delivery

relatively easy compared to internal organ systems. If a drug has sufficient ability to

permeate into the skin, it is possible to achieve a high local drug concentration in target

areas with a minimum amount of drug. In addition to this greater efficiency, limiting

exposure to the surrounding and distant tissues in this way reduces systemic toxic effects.

In some cases, even agents that cause significant complications when given orally or by

injection can potentially be used topically without causing significant adverse effects.

For example, the topical application of 5-FU is effective in treating actinic keratosis

(Jorizzo et al, 2004) and psoriasis (Ljundggren and Moller, 1972) in amounts that spare

the body from most systemic effects.









Orally delivered drugs encounter numerous physiological processes that can

potentially limit their bioavailibilty. The acidity of the stomach and the enzymes of the

intestinal lumen can degrade a drug before it has an opportunity to reach the intestinal

epithelium. The intestinal epithelium can prevent absorbed drugs from reaching the

intestinal vasculature by returning them into the gut immediately following absorption

through p-gylcoprotein efflux transporters (Kunta and Sinko, 2004). As a final hurdle,

drugs entering the intestinal vasculature from the epithelium are passed via the portal

hepatic vein to the liver before reaching systemic circulation. While passing through the

liver, they are subjected to the body's highest concentration of biotransforming enzymes.

In particular, the lack of substrate specificity in the hepatic CYP enzymes makes

numerous drugs potential targets for premature metabolism and is responsible for many,

if not most, instances of low oral bioavailability (Wrighton and Stevens, 1992).

For some drugs, solving the problem of low bioavailability is no more difficult than

merely increasing the dose accordingly. However, for other drugs, especially compounds

that require complex synthesis or purification, this is simply not an option. A great deal

of effort and ingenuity has been invested to develop delivery systems to circumvent the

limitations inherent to oral delivery. However, despite this effort there are many

examples of therapies that require such extensive or expensive formulation before they

are amenable to oral dosing that oral delivery is not cost effective. Peptide based drugs

(e.g. insulin) and some steroid based drugs (e.g., estradiol) are such instances. In these

cases, delivery by an alternative route becomes attractive.

Topical delivery is one alternative that potentially can circumvent these difficulties.

Topically absorbed compounds enter systemic circulation without undergoing an initial









metabolism by the liver. There is enzymatic activity in the skin, but with regard to the

transformation of foreign compounds, it is primarily limited to nonspecific esterase

activity. In addition, while some epidermal cells do express p-glycoprotein transporters

(Laupeze et al, 2001), there is no systematic arrangement of these cells or the transporters

at the skin surface, which would allow them to actively clear drugs in an efficient

manner. Finally, the reported pH of the skin surface does vary widely (pH 3 to 6,

Hemmingway and Molokhia, 1987), but in general it is close to 4.5, far less acidic than

the interior of the stomach. Given this relatively gentle environment, dermally absorbed

drugs are more likely to reach systemic circulation intact than are intestinally absorbed

drugs.

Instead of high enzymatic activity and active clearance, the skin possesses a

specialized barrier to prevent the permeation of foreign materials. While the skin's

permeation barrier does limit the rate with which drugs enter the skin, this same barrier

can be used in a beneficial way. Since movement through the skin is slower than

absorption through the epithelial cells of the intestine, dermally delivered drugs often

show a more sustained and consistent serum level than orally delivered drugs. This is an

obvious advantage in a number of conditions (arthritis, hypertension, chronic pain, etc.)

where maintaining a constant or near constant serum drug level for an extended time is

the most effective dosage regimen.

Approaches to Increasing Topical Delivery

Formulation, or manipulation of the delivery vehicle, is the most popular method

for increasing topical delivery. These approaches can be further divided into two

categories: formulations that interact with or alter the skin and those that do not. With

non-interacting formulations, the basic principle is to adjust the polarity of the vehicle by









altering its composition until optimum partitioning of the drug into the skin is achieved.

However, there is an inherent limitation associated with a partitioning driven system.

According to calculations based in regular solution theory, a 10-fold increase in flux is

the maximum improvement accessible by altering vehicle polarity before damage to the

skin occurs (Sloan, 1992). With interactive formulations, or penetration enhancers, the

vehicle components are intended to reduce the skin barrier. The disruption of the barrier

can be caused by the movement of solvent into the skin and/or by the leaching of

components from the skin that are essential for maintaining the barrier. This is a

common effect and it is well established that prolonged contact with many different

solvents, both polar and non-polar, disrupts the skin and increases the flux of drugs in a

reproducible manner. For example, the flux of theophylline through hairless mouse skin

that has been in contact with isopropyl myristate (IPM) is approximately 50 times higher

than through mouse skin that has been exposed to water (Sloan et al, 2003).

More recently, several physical methods of circumventing the stratum corneum

have been designed as an alternative to interactive chemical modification. Electrical

potential (Riviere and Heit, 1997) has been applied across the skin to provide an

electromotive force capable of driving charged molecules though the stratum corneum.

Both ultrasonic (Mitragotri and Kost, 2004) and laser (Doukas and Kollias, 2004) energy

can reportedly induce temporary defects in the lipid barrier large enough to allow the

transdermal delivery of macromolecules, including therapeutically significant amounts of

insulin. Using techniques developed for the fabrication of computer chips, micron-sized

needles have been coupled to standard transdermal patches (Prausnitz, 2004). When

these patches are affixed to the skin surface, the needles pierce the skin to a depth just









beneath the stratum corneum and provide a conduit from the drug reservoir directly to the

viable epidermis. The drawback of increasing flux by any disruptive method is that

higher skin flux is achieved at the cost of greater damage or perturbation to the skin's

permeation barrier.

The preceding methods attempt to overcome the poor physical properties of a

drug by changing the environment from which it is delivered or by changing the nature of

the skin. While these methods do increase delivery under the right circumstances and can

be used to improve stability, no external modification can adequately overcome limited

intrinsic water or lipid solubility of the drug which limits its solubility in the skin. The

most efficient method to increase solubility is to chemically modify the drug.

Unfortunately, such modification often reduces or eliminates the drug's beneficial

activity. In addition, even if the modified drug remained active, it is a new entity and its

pharmacokinetics and toxicity may be quite different from those of the original drug. A

modified drug that persists during delivery into the body, but reverts to the original drug

after absorption, would reduce the potential for such complications. Such an entity is

referred to as a prodrug after the term used by Alfred nearly 50 years ago to describe a

pharmacologically inactive molecule that becomes active following some biological

transformation. For the purpose of drug delivery, the ideal prodrug has favorable

physiochemical properties, is at least 1000 times less potent than the parent drug, is stable

enough to resist premature conversion, has no more toxicity than that attributed to the

parent drug and will completely revert to the parent in vivo.









Basic Theoretical Considerations of Prodrug Design for Topical Delivery and
Previously Synthesized Prodrugs

There are many strategies to prodrug design, but the most direct involves

conjugation of the parent drug to a moiety using a linkage that is susceptible to chemical

or enzymatic hydrolysis. In such cases, the linkage and its associated side chain are

collectively referred to as the promoiety. Hydroxyl, amine, amide, carboxylate and

cabonyl groups are the usual functional group targets for conjugation though any

sufficiently nucleophilic or electrophilic site can be used. A great number of potential

promoieties have been investigated ranging from relatively simple ester type derivatives

to multi-functional groups requiring a number of sequential activation and hydrolysis

steps before regenerating the parent drug. As one may expect, much of the fundamental

research regarding topical delivery has been accomplished using less complex

promoieties. When evaluating the utility of a promoiety, it is a common practice to study

a promoiety by preparing a series of produgs that share the same parent drug, attachment

point to the drug and promoiety linkage, but differ by sequential addition of methylene

units to the promoiety's alkyl side chains. These homologous series of compounds have

been useful for they exhibit systematic changes in some key physical properties (i.e.,

partition coefficient and polarity) and therefore make correlations between these

properties and therapeutic behavior easier to identify.

Before an effective prodrug can be synthesized, one must first have an

understanding of why the parent drug behaves poorly. Initial chemical analysis of skin

components had determined that the dermal barrier was composed primarily of lipophilic

compounds (Downing, 1992). This agreed with empirical observations that highly

hydrophilic drugs with low lipid solubility exhibited poor topical delivery.









Consequently, attempts to improve topical delivery have centered upon preparing

prodrugs that show greater lipid solubility than the parent molecules.

The ability of a drug to form strong intermolecular bonds with itself is the major

obstacle to solubility in non-polar media. The strength of these intermolecular forces is

reflected in the compound's heat of fusion (AHf). Assuming that AHf remains constant

from the temperature of the solution (T) to the compounds melting point (TM), AHf can be

correlated to the mole fraction solubility (X) of non-electrolytes through the relationship:


lnX= (TM -T)-lny (1-1)
RTMT

where y is the activity coefficient. From equation 1, it is clear that increases in heat of

fusion, melting point and activity coefficient will reduce solubility. Heat of fusion is not

a routinely measured property during characterization of a new compound, but it can be

eliminated from (1) by using Gibb's relationship, AHf = ASf TM


InX= (TM T)- Iny (1-2)
RT

Therefore, melting point is a conveniently determined measure of the heat of fusion and,

to a first approximation, any change to the molecule that decreases melting point without

disproportionately increasing the activity coefficient will improve solubility.

Hydrogen-bonding functional groups are the most common structural features that

contribute to high lattice energy and the resultant low lipid solubility of non-ionic drugs.

If the masking of such a group were capable of disrupting its ability to hydrogen bond,

then it would be a reasonable first choice for modification. This choice is reinforced by

two fortunate attributes of hydrogen bonding functional groups. First, as they are often

reasonably nucleophilic and, given the number of commercially available electrophilic









reagents, they lend themselves to simple synthetic schemes that can be used to produce a

variety of promieties. Second, masking even a single hydrogen-bonding site can have a

profound effect on solubility. A good example of this is the difference in isopropyl

myristate (IPM) solubility of 6-mercaptopurine (6-MP) and its 6-S-methyl

carbonyloxymethyl derivative, where masking the SH group alone reduces the melting

point from 320C to 1240C and increases IPM solubility from 0.022 to 1.05 mM (Waranis

and Sloan, 1988).

Until recently, work in this lab has focused upon improving the topical delivery of

the heterocyclic drugs 5-FU, 6-MP and theophylline by masking amide, imide and

thioamide groups on the respective parent molecules.

pKa = 7.5
0 SH 0 pKa = 8.8
pKa= 8.0 H
F N H3C N
HN N N

0 N N N O0 N N
H H
pKa = 8.5 pKa = 11.0 CH3

5-Flurouracil 6-Mercaptopurine Theophylline
(5-FU) (6-MP) (Th)

Figure 1-1. Structures of the database parent molecules and the pKa values for potential
promoiety attachment sites.

The hydrogen-bonding moieties on these three compounds are bonded directly or

through resonance to multiple electron withdrawing groups, which significantly lowers

their pKa values. Being approximately as acidic as a phenol, these moieties are readily

converted to their anionic form. This allows them to function readily as nucleophiles and

makes their derivatization relatively simple. In addition, the stability of these anions










makes them good leaving groups, which aides the subsequent regeneration of the parent

molecule from the prodrug.

By 1999, seven homologous 5-FU, 6-MP and theophylline prodrug series had been

prepared and characterized. The promoieties, which were used to prepare these series,

could be classified as one of two general types. In the first type, the nucleophilic site on

the parent molecule was bonded directly to a carbonyl group resulting in an acyl type

promoiety. In the second group, the nucleophilic site on the parent molecule was

separated from the carbonyl by a methyloxy spacer to form a soft alkyl promoiety. As

separate entities, these promoieties are fairly simple chemical systems. However, when

they are coupled to the parent molecule, the resulting prodrugs displayed a wide range in

their chemical behavior and, subsequently have different stabilities to chemical and

enzymatic hydrolysis.

Alkylaminocarbonyl (AAC) 5-FU prodrugs
O O
-" F .. F H20 H F

O N HO N
O N HO N NH
N 0 +
R RN=C=O

Alkyloxycarbonyl (AOC) 5-FU prodrugs
O 0 0
O F O- F H2 FOH


O OR : H + OOH
OH2 o-- OH o HN

: OR

Alkylcarbonyl (AC) 5-FU prodrugs
0 0 0
O O O
:N F N F H20 HN F
O N N0 N
R 0:
R--CO:

Figure 1-2. Probable hydrolysis mechanisms of the 1-N-acyl 5-FU prodrugs at pH 7.4.










Parent Drug Soft Alkyl Prodrug
O
Drug XH Drug X O R

X = acidic SH, NH or OH
Mechanism 1

Drug-~- .?- Drug -R -X + CH20
Drug +
OH OH H20 RCOOH

Drug XH
Mechanism 2
0 0 H20
DrugX R Drug CH2 Drug H + CH20


Figure 1-3. Probable hydrolysis mechanisms of the soft alkyl promoieties.

Figure 1-2 illustrates likely hydrolysis mechanisms of the N-acyl prodrugs of 5-FU

at physiological pH and demonstrates how functional groups on the parent drug can

interact with the promoiety to result in unexpected behavior in the prodrug. If each 5-FU

prodrug hydrolyzed via a typical addition elimination mechanism, then the expected

order of stability for the N-acyl promoieties would be AAC > AOC > AC. Empirically,

however, the AAC 5-FU prodrugs exhibit lower chemical stability than the AOC type

prodrugs and possess half-lives of only 8 to 11 minutes, which is inconsistent with such a

mechanism (Sloan et al, 1993). As an alternative, it has been suggested that once the

acidic N3 H of the substituted 5-FU (pKa = 6.6) (Burr and Buungard, 1985) becomes

ionized, it is capable of acting through resonance with the C2 oxygen as a general base in

an intra-molecular Elcb type hydrolysis mechanism (Sloan et al, 1993). This is

significant since it is only after the addition of the promoiety to the N1 that the N3

becomes acidic enough to be predominantly ionized at physiological pH.









The AOC 5-FU compounds are believed to hydrolyze through an addition

elimination mechanism with water acting as the nucleophile as shown in Figure 1-2

(Buur and Bundgaard, 1986). Since the AC 5-FU prodrugs have a more electrophilic

carbonyl than the AOC prodrugs and are therefore better targets for nucleophilic attack, a

similar mechanism would be expected for the hydrolysis of the AC promoiety. However,

increasing steric bulk in the alkyl portion of the AC promoiety increases the rate of their

hydrolysis which is more consistent with an SN1 mechanism (Buur and Bundgaard,

1984).

The soft N and S alkyl prodrugs of theophylline and 6-MP are believed to

regenerate their respective parent drugs by the initial hydrolysis of the ester portion of the

promoiety followed by the decomposition of the resulting hydroxymethyl compound

shown as mechanism 1 in Figure 1-3 (Sloan and Wasdo, 2003). This mechanism tends to

be found in those ACOM derivatives where the masked functional group on the parent

molecule is reasonably acidic (Bundgaard et al, 1985). It is of interest to note that for

amide prodrugs wherein the masked amide functionality has a pKa of approximately 15,

the ACOM promoiety hydrolysis mechanism changes to the SN1 process shown in

mechanism 2 of Figure 1-3 (Bundgaard et al, 1991). Since both of these mechanisms

occur relatively slowly at neutral pH, the ACOM prodrugs are stable for several hours in

aqueous solution.

In addition to the wide range of chemical properties, the heterocycle prodrugs

demonstrated a wide range of physical properties. Melting points for these compounds

range from 57.5-212 C, IPM solubility ranged from 0.3-174 mM and aqueous

solubilities ranged from 0.001-182 mM (Roberts and Sloan, 1999).









Along with their physical properties, the abilities of these prodrugs to permeate

hairless mouse skin from saturated solutions of prodrug in IPM were also measured. The

seven homologous series contained a total of 39 prodrugs. To these compounds,

solubility and IPM flux data for 5-FU, 6-MP, theophylline and the pivaloyl ACOM 5-FU

were added to produce the largest set of compounds for which physical properties and in

vitro flux data had been measured under the same experimental conditions by the same

laboratory.

The purpose of compiling this database was to provide empirical data for new

predictive flux models. Flux and permeation models for topical delivery had been

developed by a number of researchers and their development will be discussed in detail in

chapter 3. For now, it is sufficient to note that each model was based upon lipid

solubility being the principle predictor of absorption into the skin. Whether lipid

solubility was determined from solubility in a model lipophilic solvent or from a partition

coefficient, these models shared a common deficiency in predicting the behavior of

homologous series. For each series in the 43 compound database, the more water-soluble

member of the series had the highest flux through mouse skin (Roberts and Sloan, 1999).

The previously published flux models failed to predict this qualitative observation. Using

the permeability model of Potts and Guy as a foundation, Roberts and Sloan were able to

produce a flux model that overcame this deficiency and quantified the positive influence

of aqueous solubility in determining flux (Roberts and Sloan, 1999). In subsequent

work, this same model was applied to data obtained from drugs delivered from mineral

oil through human skin in vivo (Roberts and Sloan, 2001). As with the in vitro mouse

skin data, aqueous solubility was shown to exert a positive influence.









Specific Objectives and Preliminary Work

First Objective

The first objective of this work is to expand the 43 compound database with an

additional series of prodrugs based upon a new type of parent molecule. Given the

benefits and limitations inherent to topical delivery, it seemed reasonable to examine

situations for which sustained long-term delivery is the preferred regimen. Of course, the

treatment of disease is not the only instance were it is desirable to administer compounds

in this way. It is well established that the two leading causes of death in the United States

in 2001 according to the CDC's National Center for Health Statistics, cancer and heart

disease, are conditions that develop over many years. Once either of these diseases reach

an advanced state, they are often impossible to treat. Given the difficulty of treatment for

advanced conditions, a prophylactic approach to prevent or slow their onset is desirable.

In the last decade, there has been an increased interest in the use of natural products to

provide such prevention and to increase general health. While the beneficial attributes of

many natural preparations are undoubtedly over-stated, the benefits of some naturally

occurring compounds merit further research.

A number of epidemiological studies report a lower cancer rate among individuals

whose diets are rich in certain fruits and vegetables (Lui, 2004). While non-dietary

influences in these studies are hard to control and they often suffer from poor data

collection (Michels, 2005), several compounds are suspected of being responsible for

these benefits. Specifically, the polyphenolic flavanoids, or more commonly

polyphenols, are thought to play an important role in chemoprevention (Yang et al,

2001). Polyphenols are a large class of plant-derived compounds that contain either a

flavone or flavanone structure and multiple hydroxyl groups. The phenolic groups are









often glycosylated or alkylated by the plant (presumably to protect them from oxidation)

and it is these forms that they are principally found in the diet. During digestion, the

hydroxyl groups are uncovered. This is an important process, as many of the beneficial

effects attributed to polyphenols are likely to require the presence of at least one free

phenolic hydroxyl group.

Polyphenols demonstrate an influence on numerous physiological processes in vitro

and in vivo. They can function as antioxidants and they have been shown to protect

against oxidative stress in vitro and in vivo (Vinson, 1998), although the clinical

significance of this has not been determined. Polyphenols reportedly have the potential

to modulate radical mediated signaling pathways. Of these, the ability to inhibit nitric

oxide synthase may be significant due to its participation in inflammatory processes (Wu

and Meninger, 2002). Specific polyphenols have been reported to modulate enzymes

involved in carcinogenesis, namely tyrosine kinase (Lin, 2004) and NF kappa B (Kundu

and Surh, 2004). Perhaps more importantly, polyphenols inhibit CYP-450 enzymes and

may reduce the bioactivation of procarcinogens. This inhibition is presumably the

mechanism by which green tea phenols prevent methylcholanthrene and phorbol acetate

induced tumor genesis in mouse skin (Wang et al, 1989). While it is still unclear which

effect or combination of effects is responsible for the perceived benefit of polyphenols, if

they are efficacious, then having a chronic supplementation based upon this class of

compounds and a method to dermally deliver them would be useful.

The triphenolic flavanone naringenin (4', 5, 7-trihydroxyflavanone), a well-studied

inhibitor of hepatic CYP-450 3A4 (Geungerich and Kim, 1990), was investigated for our

initial experiments.










OH

HO 0O



OH 0


Figure 1-4. Structure of naringenin.

The melting point of naringenin is 247-250C. When it is reacted with excess

acetyl chloride, the resulting triacetate product has a melting point of 72-85C

(unpublished results), indicating a substantial drop in lattice energy. However, even with

the substantial drop in melting point, the aqueous and IPM solubility of the triacetate as

well as its flux through mouse skin remained very low (unpublished results). When

attempts to prepare and purify diesterified derivatives of naringenin and other flavanoids

proved especially difficult, it was decided to seek a simpler phenol containing model

compound. Hydroquinone was considered as a model compound, but it was discarded

after the mono-acetate was found to disproportionate into the diacetate and hydroquinone

during purification. Acetaminophen (APAP) was eventually chosen as a replacement.

Though APAP itself is not polyphenolic, it was considered relevant because it possesses

structural features similar to polyphenols. In particular, APAP contains one phenol, has a

predominantly planar shape, aromatic character and an amide group, which, along with

the phenol, allows it to form multiple hydrogen bonds. More importantly, it was

anticipated that the difference in pKa values between the phenol and amide groups

(approximately 10 versus 13 respectively) would allow selective ionization and









consequently derivatization of the phenolic oxygen without a significant amount of

competing derivatization at the amide nitrogen.

Second Objective

The choice of promoiety was guided by the second objective of this work.

Although the Roberts-Sloan equation was developed specifically for delivery from IPM

and was initially applied to a relatively specialized group of compounds delivered from a

single vehicle, its theoretical foundation suggested that it should have wider applicability.

If an aqueous vehicle is assumed in the derivation of the Roberts-Sloan model, the final

form of the equation remains unchanged. As a result, the model could be used to predict

flux from IPM or water depending upon which values are used for the x, y and z

coefficients. To estimate the coefficients for the aqueous model, Sloan et al (2003)

collected the eighteen most chemically stable ACOM prodrugs present in the 43

compound database and measured their flux through hairless mouse skin from unbuffered

water. To make the coefficients of the model more significant, more compounds would

be needed in the dataset. In order to add the APAP prodrugs to this group, they would

need to contain promoieties that were resistant to hydrolysis in aqueous solution.

Despite their prevalence among commonly prescribed drugs, very few phenolic

compounds have been converted to prodrugs in an effort to improve their topical

delivery. Phenol masking prodrugs of only four related compounds, morphine (Drustrup

et al, 1991), bupenorphine (Stinchcomb et al, 1996), nalbuphine (Sung et al, 1998) and

naltrexone (Stinchcomb et al, 2002) have been prepared and studied in flux experiments.

The promoieties for most of these studies have been limited to simple alkyl esters.

Although they improved topical delivery of these narcotic analgetics, simple alkyl esters,

especially esters of phenols, are too unstable for prolonged exposure to aqueous solution.









Swintowski (in Dittert et al 1963) synthesized some simple alkyl esters of acetaminophen

along with the corresponding alkyl carbonate esters and studied their stability to chemical

and enzymatic hydrolysis. As expected, he found that the less electrophilic carbonate

esters hydrolyzed times more slowly in pH 7 aqueous buffer than the more electrophilic

alkyl esters and both groups were susceptible to enzymatic hydrolysis. Even though

Swintowski did not synthesize all of the alkyl homologs, he did synthesize the methyl

carbonate, which is expected to undergo chemical hydrolysis more rapidly than any of the

longer chain carbonates. Given the long half-life of the methyl carbonate, we concluded

that the carbonate functional group had adequate stability to serve as a promoiety for the

APAP prodrugs.

Since the importance of increasing biphasic solubility when optimizing topical

delivery is now well established, it may seem counterproductive to only study

promoieties that contain alkyl side chains of ever increasing length. While these

promoieties can produce compounds with reduced crystal lattice energy, their other

properties are counter productive to good water solubility. When a compound dissolves

in water, each solute molecule unavoidably disrupts a number of water-water hydrogen

bonds by displacing water molecules from the volume it must occupy.

Thermodynamically, this process has a substantial negative enthalpy. Once in solution,

any functional groups on the solute molecule that are capable of forming new interactions

with surrounding water molecules can reduce this enthalpy cost. Alkyl groups are only

capable of interacting through relatively weak Van der Waals forces and, therefore, do

little to recover the enthalpy that was lost during dissolution. It can be conjectured that









replacing the alkyl side chains with entities that are capable of forming beneficial

interactions would increase aqueous solubility.

The challenge behind this approach is to find a functional group that has a large

enough dipole moment to interact favorably with surrounding water molecules without

unduly increasing crystal lattice energy. Polyether functional groups have these

properties. Straight chain alcohols larger than propanol are immiscible with water. By

contrast, polyethyleneglycols (PEGs) continue to be soluble in water even when their

molecular weight grows over several thousand amu. Greenwald has used conjugation to

large PEG molecules to improve the water solubility of a number of poorly soluble

amine, imide and hydroxyl containing compounds (Greenwald et al, 2000). However,

Greenwald favors the use of high molecular weight PEG conjugates (>20,000 amu), with

the promoiety comprising the overwhelming majority of the prodrug's total mass. These

large PEG prodrugs remain water-soluble primarily because their characteristics remain

close to those of the unconjugated PEGs. For promoieties containing only a few or even

one ether promoiety, the question becomes to what extent will the replacement of a

methylene group by an oxygen improve water solubility? To address this question, two

additional promoieties will be synthesized that have such a modification; a

methoxyethoxycarbonyl APAP and a methoxyisopropoxy carbonyl APAP. It is

hypothesized that for both types of prodrugs the best performing compounds will be

those that possess the best combination of lipid and water solubility.

Third Objective

Mammalian skin continues to be the membrane of choice for assessing drug

permability in vitro. Although there are substantial differences in the thickness of the

various skin layers between species, the general histology, chemical composition and









physiochemical properties of the skin barrier are reasonably similar for all mammals.

Therefore, while absolute permeability varies with skin thickness, the relative flux of

compounds through mammalian skin remains consistent. In other words, those

compounds that diffuse most readily though human or pig skin are likely be the same

compounds which diffuse most readily though mouse or rat skin. In particular, if the

compounds of a given homologous series are ranked according to their flux through

hairless mouse skin, it has been well established that this order will match the rank order

of their flux through human skin (Scheuplein and Blank, 1973, and Sloan et al, 1997).

This ability of hairless mouse skin to predict the best performing members of a

homologous series is one of the attributes that has made it a popular choice for flux

experiments. In contrast to human skin, mouse skin has much less variability and data

collected from mouse skin experiments can be used without normalization. This obviates

the multiple additional control diffusion cells, which must be run with in vitro human

skin experiments. This consistency also allows even small differences in fluxes to be

discerned. Unlike human skin, which is usually dermatomed or exposed to heat or

enzymatic digestion to isolate the outermost skin layers prior to use, hairless mouse skin

can be used full thickness and requires little more preparation than removal from the

mouse.

Despite these advantages, the use of mouse skin, or any mammalian skin, does

have drawbacks. The care of the animal prior to use is expensive and special care must

be taken to maintain the integrity of the skin for the duration of the experiment.

Increasing regulatory requirements and ethical concerns about the use of animals in

research makes finding an artificial membrane that can be used in the place of animal









tissue desirable. Most attempts to predict skin permeation using an artificial membrane

have focused upon finding a surrogate material whose properties are sufficiently similar

to human skin to allow a direct comparison. Untreated lipophilic polymers such as

polydimethyl silioxane (PDMS) (Geinoz et al, 2002, and Cronin et al, 1998) and solvent

modified lipophilic polymers (Twist and Zatz, 1990, Maitani, 1996, and DuPlessis, 2001)

have been popular systems for this purpose. However, none of these systems have

proven particularly useful at predicting flux through skin.

The complexity of the skin makes it unlikely that simply knowing the flux of a

drug through a simple polymer, modified or not, will be sufficient to predict topical

delivery. Despite this limitation, there are aspects of any compound's diffusion behavior

which hold true regardless of the nature of the membrane. For example, the kinetics of a

compound's dissolution, as well as its propensity to cluster or stack in solution, can affect

its flux in a manner that is difficult to determine from its solubility alone. Therefore,

despite its inability to predict flux through skin directly, a compound's flux through a

polymer membrane may still contain information that is useful for predicting its flux

through skin. However, extracting and utilizing this information will likely require a

more sophisticated treatment then has yet been reported.

Thus, it is the final objective of this research to produce a predictive model for

flux through hairless mouse skin that is based upon its flux through PDMS membrane in

addition to its solubility properties. PDMS membrane is a rubbery polymer composed of

cross-linked chains of repeating Me2SiO units. As a rubbery polymer, the chains that

comprise the membrane are flexible and return quickly to an equilibrium position after

being disturbed. Like diffusion through the skin, diffusion in such a material is believed






21


to be Fickian in nature (Crank, 1975). It is expected that the additional information

provided by PDMS flux data will improve the predictability of the solubility-based

models and this will serve as first step in creating an artificial membrane system that will

be useful in screening for topical delivery.














CHAPTER 2
BASIC ANATOMY OF THE SKIN

Taken in its entirety, the skin is the largest organ in the human body. On average,

it accounts for approximately 10% of the total adult body weight and covers a total area

of 1.5 to 2 m2 (Schaefer and Redelmeier, 1996). As the principle interface with our

environment, the skin must be capable of simultaneously providing sensory input and

functioning as a barrier to the transfer of materials into and out of the body. Specifically,

the skin allows the body to effectively control fluid loss, regulate body temperature,

protect against physical trauma, defend against microbial infection and provide tactile

sensation. To perform these functions, the skin contains numerous specialized cells and

tissues arranged with a specific architecture. While modern imaging techniques have

revealed that the skin is composed of a complex microanatomy, to the naked eye, the skin

appears as three superimposed layers. From deepest to most shallow, these layers are

known as the hypodermis, the dermis and the epidermis.

While the skin contains a wide variety of cells and microstructures, many of them

do not significantly affect the bioavailability of topically applied substances. In the

following discussion, greater attention will be paid to what defines the environment of the

skin and to those elements that are most likely to impact topical absorption

Hypodermis

The hypodermis, or subcutaneous tissue, contains loose connective fibers and

associated adipose tissue. This connective tissue serves to anchor the overlying layers of

the skin to the body and the adipose tissue functions as a carbohydrate source and thermal









insulator. It varies greatly in thickness throughout the body and between individuals. It is

usually thickest around the trunk of the body and thinnest on the backs of the hands and

feet. The hypodermis has a less extensive vasculature than the outer layers of the skin but

it does house the larger blood vessels that ultimately feed the rest of the integument.

Since the hypodermis does not contain the same specialized tissues as the other skin

layers and since it does not actively participate in the regulatory functions of the skin, it is

usually no longer classified as properly belonging to the skin.

Dermis

The dermis is responsible for the skin's structural integrity and, with the exception

of controlling water loss, performs most of its critical functions.

The tensile strength and flexibility of the dermis are the result of a tightly knit mesh

of collagen and elastin fibers (Ushiki, 2002). This collagen lattice defines the dermis and

provides a relatively fixed scaffolding onto which the other structures of the dermis are

anchored. Both collagen and elastin are synthesized from water-soluble precursors that

are secreted into the intracellular space by the most common cellular component of the

dermis, the fibroblast. After leaving the fibroblast, these precursors undergo modification

and are assembled into thin fibrils. In the deeper region of the dermis (the reticular

dermis), primary collagen fibrils are further assembled into thick, closely grouped, and

extensively cross-linked strands that run roughly parallel to the skin surface. Nearer to

the epidermis (the papillary dermis), the strands become thinner and less heavily cross-

linked to allow space for the proliferation of capillaries and nerve endings.

Surrounding the collagen lattice and filling most of the dermal intracellular space

are extremely hydrophilic macromolecules known as proteoglycans. Proteoglycans are a

group of fibroblast derived compounds that contain numerous unbranched polysaccharide









chains covalently bound to a common polypeptide backbone (Iozzo, 1998). The

polysaccharide side chains are composed of several hundred repeating

glycosaminoglycan disaccharides that have been sulfated to a high degree. The large

number of sulfate and carboxylate groups gives a strong anionic charge to the

polysaccharide side chains and causes them to repel one another. In response to this

repulsion, the side chains are forced to nearly full extension from the peptide backbone

and the molecule, in turn, resists compression. In addition, a large amount of water

remains associated around the sulfate and carboxylate groups in hydration spheres. The

ordering of water around the ionized proteoglycans gives the environment of the dermis

properties similar to those of a hydrophilic gel. Small water-soluble molecules move

with relative ease but highly lipophilic or high molecular weight molecules are impeded.

A multitude of anatomical structures is found within the dermal matrix (Thibodeau

and Patton, 2003). Corpuscular nerve endings that are sensitive to vibration and pressure

(Meissner and Pacinian corpuscles) are present as well as free nerve endings that are

sensitive to pain and temperature. Smooth muscles cells control the dilation of blood

vessels and follicular arrector pili cause the familiar appearance of goose bumps.

Resident immunologically active Mast cells and phagocytic macrophages provide a

second line of defense against foreign microorganisms as do dermal dendrocytes and

circulating t-lymphocytes. However, of the many structures of the dermis, the three that

have the highest potential impact on topical delivery are the hair follicles, the sweat

glands and the capillary vasculature.

Hair follicles and exocrine sweat glands are rooted in the reticular dermis, but

they both have pores that extend to the skin surface. These pores provide openings in and









through the epidermis and its permeation barrier. Bypassing this barrier allows free

diffusion down the sweat duct or hair follicle and allows direct access to the dermis.

However, this access is balanced by the physical dimensions of the pores. While there

are several million follicles and exocrine glands distributed over the skin surface, the

individual pore size is so small that the total cross sectional area of all pores is less then

0.1% of skin surface area (Schaefer and Redelmeier, 1996). Therefore, the pore-

mediated pathway should only be important for compounds whose ability to penetrate the

epidermal barrier is very low.

The epidermis contains no vasculature of its own and must depend upon dermal

blood vessels to supply essential compounds and remove waste material (Thibodeau and

Patton, 2003). To facilitate this process, the border between the dermis and the epidermis

is uneven with papillae from each layer interlocking with one another. Inside the dermal

papillae are capillary plexuses from which nutrients and oxygen move by diffusion. The

papillary structure of the border assures that epidermal cells surround each plexus. This

arrangement increases the likelihood of dermal nutrients reaching epidermal cells and

increases the efficiency with which epidermal waste products reach the dermal

capillaries. Similarly, this design increases access to the capillaries for any exogenous

compound that reaches the lowest level of the epidermis. With several thousand dermal

plexuses per each square centimeter of skin, this route is the principle entry into systemic

circulation for topical agents.

In addition to being the most readily available path to systemic circulation, the

capillary plexus is also an effective sink. In a static system, the volume of blood

contained in the vessels of an individual plexus is small and only a nominal mass of









absorbed material is necessary to produce a high local concentration. However, in the

living system, the blood flow through the plexus prevents stagnant accumulation and

dilutes absorbed compounds into systemic circulation. For poorly water-soluble

molecules, serum proteins such as albumin should facilitate this process in a manner

similar to what is observed with in vitro experiments (Cross et al, 2003). Many soluble

proteins have non polar pockets into which lipophilic compounds can partition.

Therefore, soluble proteins represent an additional phase that reduces the thermodynamic

activity exhibited by blood borne non-polar molecules and, concurrently, increases the

blood's capacity to carry them.

Epidermis

Unlike the multi-functional dermis, the epidermis is primarily designed to perform

one function; provide a barrier that prevents the loss of water and essentially seals the

skin against the entrance of foreign materials. The epidermal barrier covers over 99% of

the body surface and it is this barrier that must be circumvented to effectively deliver

drugs topically. Ironically, most of the body the permeation barrier is contained in only

the outermost 10%-20% of the epidermis (Elias and Friend, 1975). The formation of

these few essential im of integument represents the final stage in a well-orchestrated

transformation of viable epidermal cells into an inert and physiologically unique matrix.

Keratin-producing epithelial cells, keratinocytes, account for over 95% of the

epidermal cell volume and are ultimately responsible for the formation of the barrier

(Steinert et al, 1991). They are formed from resident stem cells attached to the basement

membrane, which is a specialized collagen structure that defines the lower limit of the

epidermis and connects it to the dermis. When a keratinocyte stem cell divides, it

produces a daughter stem cell, which remains attached to the basement membrane, and a









transit-amplifying cell that begins to differentiate once it detaches from the membrane

(Watt, 1989). As the daughter stem cells repeat this process, the formation and migration

of new transit-amplifying cells forces older cells closer to the skin surface. Since the rates

with which both stem cells divide and transit cells mature are coordinated, keratinocytes

at similar stages of development are found at roughly equivalent depths of the epidermis.

The difference in morphology between the various stages of keratinocyte development

has resulted in the epidermis being divided into the five visually distinct regions

The first region, the stratum basale or stratum germinativum, is comprised of the

epithelial stem cells and recently formed transit-amplifying cells. These cells appear

columnar and have large nucleii. In the stratum spinosum (the next 2 to 7 cell layers), the

keratinocytes have lost the columnar organization of the basal cells and have begun to

elongate. The name of this layer refers to the many surface protein complexes

(desmosomes) attaching one keratinocyte to another, which give the cells serrated edges

(Burge, 1994). The stratum granulosum contains the uppermost two or three viable cell

layers and is the region in which keratinocytes make the most rapid and dramatic

transformation. The cells flatten markedly and a simultaneous disintegration of the

nucleus occurs. Stratum granulosum cells contain a large number of small granular

deposits that give them a distinctive speckled appearance. The outermost layer, the

stratum comeum, begins with the appearance of closely packed layers of corneocytes.

Corneocytes are thin, proteinaceous and roughly polygonal with a homogeneous internal

structure. They remain essentially unchanged throughout the stratum corneum and are

intact when released from the skin surface. Several in vitro and in vivo experiments, and

the study of diseases that impair stratum corneum formation, have consistently indicated









that this layer provides most of the permation barrier (Scheuplein and Blank, 1971 and

Lavrijsen et al, 1993). On the palms of the hands, soles of the feet, and wherever the

skin is subjected to a large amount of mechanical stress, the stratum comeum can contain

30 or more comeocyte layers and be thicker than the viable epidermis. However, for

most of the body, the stratum comeum is much thinner and is composed of 5-20

corneocyte layers.

When the epidermis is studied on the submicron scale, it is clear that the

formation of the stratum corneum is a process that begins several cell layers beneath the

first discernable corneocytes in a light microscope image. At this scale, the subcellular

granules, that are numerous in the stratum granulosum, can be resolved into two separate

functional entities; keratohyalin granules and lamellar bodies.

Keratohyalin granules (KG) produce the structural proteins that are used in the

assembly of the comeocytes. Some of these proteins (principally loricrin and involucrin)

are used to create a cornified envelope that forms the corneocyte's outer surface. The

envelope begins to form as a thin layer just inside the apical membranes of keratinocytes

in the upper stratum spinosum but thickens rapidly throughout the straum granulosum as

transaminase enzymes attach additional layers (Stevens et al, 1990). Other KG generated

proteins, principally the keratins K1 and K10, and the keratin aggregating protein

filaggrin, are crosslinked across the inside of the cornified envelope to form the

corneocyte core (Eckert, 1989). When fully formed, the comeocyte has a sufficient

number of internal protein filaments to resist swelling upon contact with water and to

resist permeation by other exogenous compounds.













Lamellar bodies (LBs) are the second most prevalent granular organelle found in


viable epidermal tissue. Like KGs, LBs are first observable in the upper layers of the


stratum spinosum and they increase in number within the stratum granulosum. LBs


process and store a complex mixture of lipids and lipid-like materials which ultimately


control the loss of water from the skin (Roberts et al, 1978). Along with free fatty acids


and cholesterol, this lipid matrix contains a number of hydroxylated amide compounds


known as ceramides.



0 0


OH

OOH

Ceramide 1







OOH
Ceramide 2 Ceramide 3









Ceramide 4


OH


HN


Ceramide 5


OH


HN OH


OH
Ceramide 6


OH

A ~HN
OH




Ceramide 7 Ceramide 8



Figure 2-2. Structures of the principal ceramides comprising the lipid bilayers of the
lamellar bodies.


HOH
OH









Electron micrographs reveal that the lipid components in LBs are often arranged in

an ordered pattern that resembles compressed and stacked micelles (Fartasch et al, 1993).

On the border between the stratum granluosum and the stratum corneum, LBs migrate to

the keratinocyte membrane and their lipid components are expelled into the intercellular

space where they are reformed into larger stacked sheets. Concurrently, a single layer of

lipid components, primarily ceramides 1 and 2, is bound covalently to the surface of the

corneocyte (Wertz et al, 1989). In the fully formed stratum corneum, the inter-

keratinocyte lipid components produce a characteristic electron micrograph pattern of

alternating light and dark bands that run parallel to the surfaces of the keratinocytes. This

pattern is generally thought to indicate that final arrangement of the lipid components in

the stratum corneum is a nearly continuous series of stacked and planar lipid bilayers.

It is important to note that the composition of the inter-corneocyte lipids is different

from the lipid composition of the lamellar bodies. Phospholipids and glucosylated

ceramides constitute a significant proportion of the lamellar body lipids but they are

essentially absent in the stratum corneum. In fact, while it contains some ionizable

components such as free fatty acids, amino acids and a small amount of cholesterol

sulfate, the stratum corneum lipid matrix is predominantly composed of neutral

molecules (Lampe et al, 1983). The principle polar functional groups of the matrix are

the ceramide hydroxyl and amide groups. They are capable of hydrogen bonding but

there are only a small number of these groups per molecule. The polar functional groups

of the ceramides do associate to form hydrophilic planes within the lipid lamella but these

planes are thin compared to the alkyl regions. It is a reasonable speculation that removal

of the highly acidic phosphate and the poly-hydroxylated sugar moieties is necessary to









limit the water permeability of the stratum corneum and to remove groups that would

disrupt the cohesiveness of the lipid lamella.

Studies of model systems suggest that the specific molecular arrangement or

phase of the lipids in the lamellae influences permeability as much as the chemical

composition (Xiang et al, 1998). As such, there has been a fair amount of effort

expended in the attempt to experimentally determine the arrangement of lipids in the

stratum corneum. X-ray diffraction, NMR, FTIR and electron microscopy have all been

used to gather data from the stratum corneum and compare it to the data from model

systems, but the results have been difficult to interpret and sometimes contradictory

(Hsueh et al, 2004 and Pilgram and Bouwstra, 2004). Many variations of the bilayer lipid

arrangement have been observed in the model systems, but they can generally be

classified into one of three broad phases (Sparr and Engstrom, 2004). Lipids in a solid

crystalline phase (characterized by an orthorhombic or triclinic packing) have the highest

cohesive forces, the least freedom of movement and the lowest permeability. Lipids in a

gel phase (characterized by a hexagonal packing) have greater rotational freedom and a

higher permeability. Lipids in a liquid crystal phase lack a discrete arrangement and have

the highest permeability. Many researchers have reported that the stratum corneum

matrix contains or can adopt many phases depending upon environmental factors such as

temperature, humidity or pH. Others have suggested that the high amount of cholesterol

precludes the existence of the more ordered phases and that a single-phase model is more

appropriate (Norlen, 2001). Given the number of components contained in the lipid

matrix and the changes in water content and temperature between the viable epidermis

and the skin surface, it seems unlikely that a single phase can exist throughout the stratum






32


corneum. For the purpose of predicting dermal delivery, such discussions, while

interesting, are perhaps less germane. Ultimately, the most useful description of the skin

will be a functional description derived from the skin's interaction with other compounds

rather than a physical description based upon the skin alone.














CHAPTER 3
DEVELOPMENT OF THE PREDICTIVE SOLUBILITY AND FLUX MODELS

Derivation of the Series Specific Organic and Aqueous Solubility Equations

To a large extent, a drug's solubility properties determine its transdermal flux and

much of its other physiological behavior. Unfortunately, since a compound's solubility is

impossible to predict a priori, the only certain means of identifying those prodrugs with

optimal solubilities is to synthesize them. In keeping with this limitation, understanding

the behavior of a given promoiety has traditionally required the synthesis of many

compounds in which the parent drug is joined with many alkyl homologs of the

promoiety. Comparing the behavior of two promoieties has only been done once data

from several examples of each promoiety has been collected. This approach presents a

potential difficulty in estimating the relative performance of the 4-AOC and 4-MOAOC

APAP compounds given that only two members of the latter series were synthesized.

With such limited data, the relevant question becomes can the behavior of a series be

estimated from the performance of two, or even one, of its members? In other words, are

there parameters or descriptors, common to each member of a series, which can be used

as a means of comparison? The following discussion identifies these parameters and

illustrates their use.

The dissolution of a crystalline nonelectrolyte is a complex process that is

inaccessible to direct mathematical treatment. However, the free energy of this

transformation can be described. Free energy is a state function. Therefore any path or

combination of paths (actual or hypothetical) that begins with the crystalline state and









end with the solvated state can be used to calculate free energy for dissolution. With this

in mind, it is convenient to separate the dissolution process into two sequential events;

the removal of a molecule of solid from the crystal lattice (decrystallization) and the

movement of this freed molecule into solution. Both of these events have free energies

associated with them and the sum of these energies equals the free energy of the overall

process:

AGD + AGMx = AGDis (3.1)

where AGD is the free energy of decrystallization, AGMix is the free energy of mixing and

AGD is the free energy of dissolution. In a saturated solution, the crystalline form of the

drug is in equilibrium with the solvated form and AGD is equal to 0.

AGD + AGMx = 0

AGD = -AGixx (3.2)

Each side of this equation can be expanded using the Gibbs relationship:

AH TASD = -AHM1X + TASMix

The enthalpy (AHD) and entropy (ASD) of decrystallization specifically refers to the

energy required to remove a molecule of the drug from the crystal lattice at the solution

temperature (T). These are not readily obtained quantities. However, the enthalpy (AHF)

and entropy (ASF) of fusion can be can be used in their place if the heat capacities of the

crystalline and molten forms of the drug are equal to one another and reasonably constant

over from the solution temperature to the melting point. With this substitution,

AHF -TASF = -AH +ix + TASuix









Since AHF = TMASF (where TM is the melting point of the drug), this can be rewritten as

ASF(TM T)= -AHix + TASuix (3.3)

According to ideal solution theory, ASMIx is related to the mole fraction of the

components in solution by the equation:

ASix = -(nR In x, + n2R In x2) (3.4)

where nl and xl are the amount and mole fraction of the drug in solution and n2 and x2 are

the amount and mole fraction of solvent. If there is no substantial loss of entropy upon

mixing due to factors such as solvent ordering, then the ideal expression for ASMIx can be

substituted into equation 3.3.

ASF(T -T)= -AHix -T(nR In x, + n2Rlnx2) (3.5)
This expression can be further simplified if the solubility of the drug in the solvent

is low (< 1%). In this case, the mole fraction of the solvent is approximately equal to 1

and equation 3.5 reduces to

ASF(TM -T)= -AHMix -RTnln x

ASF(TM T)+ AHix
F= In x (3.6)
RTn

where n and x now refer only to the amount and mole fraction of drug respectively. If the

ASF and AHMix are presumed to be molar quantities, then the amount of drug in solution,

n, disappears from the equation.

ASF (T -T)+AHMix =lnx (3.7)
SRT RT )

To convert equation 3.7 into an expression for solubility as amount per unit

volume, a dilute solution assumption is again used:










n1 n,
X=--
n, +n n2


With this assumption, the volume of the solution (Vs) is equal to the molar volume of the

solvent (V2) multiplied by the amount of solvent (n2). Therefore:

n n1 x
S=
Vs n2V2 V2

nS= Inx -InV2

Substituting this into equation 3.7yields:


nV2 AH ASF (T -T) =lnS
RT RT


If both sides of the equation are divided by 2.303, the equation is converted from the

natural log to the more common base ten log:


RTln V2 AH'-x ASF (T T)=logS (3.8)
2.303RT 2.303RT

Equation 3.8 is an expression for solubility in what Hildebrand referred to as a

regular solution; i.e. a solution that possesses a non-zero enthalpy of mixing and a nearly

ideal entropy of mixing (Hildebrand et al, 1970). This behavior is more likely to occur in

organic solutions and it should be followed by saturated IPM solutions of our prodrugs.

According to regular solution theory, the enthalpy of mixing arises from the breaking and

reforming of intermolecular bonds that occur when the drug molecule enters solution.

Typically, the strength of these intermolecular bonds is determined by the presence of

functional groups on the drug that possess unpaired electrons, significant dipole moments

or polarizable electron clouds. In a homologous alkyl series, the number and type of









these interacting functional groups remain the same in each series member. Therefore, to

a good approximation, AHMix is a constant that is characteristic to a given series. If the

solubility of each series member is taken at the same temperature, then all quantities

preceding the melting point term can be grouped into a single constant that is also

characteristic to the series:


As ASF (T -T)=logS (3.9)
2.303RT

Walden's rule states that the entropy of fusion for most rigid small molecules is

approximately equal to 56 J mol-lK-1. If the database compounds follow this rule, and if

the temperature at which solubility is determined is the same for all compounds, then the

terms of the melting point coefficient can be grouped into a second constant that should

be independent of series and solvent.

Ao B(TM T)= logSo (3.10)

This is the general equation that relates melting point to solubility for homologs in

a regular (organic) solution.

An analogous equation for the aqueous solubility across a homologous series was

derived using a method that parallels the derivation of Yalkowsky's general solubility

equation (Yalkowsky and Valvani, 1980). Yalkowsky suggested that the partition

coefficient was a simple means of transforming an organic solubility equation to an

aqueous solubility if the partition coefficient is assumed to approximately equal to the

ratio of the drug's organic solubility to its aqueous solubility.

So
KO:AQ Q
SAQ










log KOAQ = log S -logSAQ

log SAQ = log So -log KOAQ (3.11)

Substituting equation 3.10 into equation 3.11 yields:

logSAQ = Ao -Bo(T -T)- log KAQ (3.12)

The partition coefficients of alkyl homologs are related to one another through the

equation:

logK:AQ = logK:AQ + C (3.13)

where log KO:AQ is the log of the partition coefficient for a given homolog, log KOO:AQ is a

series specific constant, and CMW is a constant (C) multiplied by the homolog's

molecular weight (MW) (Hansch and Leo, 1971). The left side of this equation can be

used to replace log KO:AQ in equation 3-12.

log SAQ = Ao Bo (TM T) log K:AQ CMW

By combining Ao and log KO0:AQ into a new constant, AAQ, the final form of the general

aqueous solubility equation is obtained.

log SAQ = AAQ Bo(TM T) CMW (3.14)

In the final form, the aqueous solubility equation is essentially the organic

solubility equation with an additional correction term for molecular weight. This is a

reasonable result given that the dissolution of a non-electrolyte in aqueous solution is

similar in many aspects to its' dissolution in an organic solution with one principle

difference. Creating an aqueous cavity large enough to accommodate a solute molecule

disrupts a number of hydrogen bonds between water molecules; an energy expenditure

that has no counterpart in most organic solvents. The amount of energy required to open

the cavity is proportional to the number of hydrogen bonds that must be broken in its'









formation, which, in turn, is proportional to the cavity's size. Therefore, the amount of

energy required by this process should be related directly to the size of the solute

molecule and reasonably proportional to its' molecular weight. In addition, since the

only parameter that affects the number of hydrogen bonds lost during solvation is the size

of the solute, the molecular weight coefficient should be independent of the solute.

The parameters that appear in the solubility equations can be characterized as

independent or universal (Bo and C, respectively), compound specific (TM and MW), or

series specific (Ao or AAQ). Since there is only one series specific parameter appearing in

each equation, all the coefficients for a given series can be determined once data from

any one series member is obtained. As the only series specific parameter, A is a measure

of the intrinsic influence on solubility conveyed to the prodrug by the promoiety. In

other words, when two different prodrugs posses similar melting points and molecular

weights, that compound belonging to the series with the higher A coefficient will have

the higher solubility. Of course, to make a proper estimate of A, proper values for Bo

and C must first be determined.

Derivation of the Potts-Guy Equation

As with many fundamental flux models, the Potts-Guy equation is developed

from Fick's 1st Law of Diffusion. Fick discovered this law in the 1800's by observing the

movement of dissolved compounds through permeable membranes separating

compartments containing solutions of differing concentration (figure 3.1).



























Figure 3-1. Simplified diagram of an experimental diffusion apparatus.

Using heat flow equations as a guide, Fick proposed that the flux per unit area (J)

through each section of the membrane was proportional to the local concentration

gradient:

J = D C/OL (3.15)

where D is the proportionality constant.

In general, the differential 8C/8L need not be a constant across the entire

membrane. However, if the concentrations in each compartment (CD and CR

respectively) are held constant for a sufficient amount of time, equilibrium

concentrations, C1 and C2, are established just inside each exposed face of the membrane

and the flux through the membrane reaches steady state. Under these conditions, a

homogeneous membrane (or a membrane that behaves in a homogeneous manner) will

have a constant concentration gradient across the membrane. For a linear concentration

gradient, the partial differential will equal the difference in concentration just inside each


Donor Receptor
Reservoir C C1 C Reservoir


Membrane









face divided by the thickness of the membrane. Substituting this into equation 3.15

yields:

J = D (Ci C2)/L (3.16)

where L is the thickness of the membrane. For a given membrane, the highest flux (JM) is

obtained when the difference between Ci and C2 is a maximum. This occurs when C1 is

equal to the compound's solubility in the membrane (SM) and when C2 is close to zero

(sink conditions). Experimentally, these conditions are obtained by keeping a saturated

solution of the test compound in the donor compartment while periodically replenishing

the solution in the receptor compartment with clean solvent to keep its concentration

under 10% of its solubility in the receptor phase. Under these conditions, (Ci- C2) = (SM

- 0) and equation 3.16 becomes:

JM = DSM/L (3.17)

From a practical point of view, solubility in the skin is a difficult quantity to

measure directly. However since the membrane is in equilibrium with a saturated

solution, SM should equal the solubility of the compound in donor compartment (SD)

multiplied by the partition coefficient between the membrane and the donor solution

(KM:D). Substituting this relationship into equation 3.17 yields:

JM = D SD KM:D/L (3.18)

From equation 3.18, the maximum steady state flux of a compound is

proportional to the solubility of that compound in the donor phase. If flux is divided by

this solubility, a new quantity, permeability (P), is obtained

P = D K:D/L (3.19)









Potts and Guy recognized that each term in equation 3.19 could be

estimated from readily available empirical data. Since they were interested in delivery

from water and the skin (or at least the stratum corneum) is a lipid rich membrane, they

asserted that the skin:water partition coefficient, KM:D, could be related to the

octanol:water partition coefficient, KOCT, through the equation:

log KM:D = flog KOCT + b (3.20)

Furthermore, previous work had suggested that the diffusion coefficient (D) should be

related exponentially to the molecular volume of the compound (Cohen and Tumbull,

1959). Since molecular weight correlates well to molecular volume for organic

molecules, it is convenient to substitute molecular weight into this equation:

D = D e-aMW

log D = log Do (1/ln 10)aMW (3.21)

Taking the logarithmic form of equation 3.18 and making substitutions using equations

3.20 and 3.21 yields the following result:

log P = log KM:D + log D + log (1/L)

log P = flog KOCT + (1/ln 10)aMW + log (D/L) + b (3.22)

The final terms of equation 3.22, log (Do/L) and b, can be collected into a single term for

convenience. Similarly, the conversion factor of 1/ln 10 can be combined with the a

term to create a single coefficient, P. Doing these two substitutions results in the

standard form of the Potts-Guy equation.

log P = flog KOCT PMW + c (3.23)









Derivation of the Roberts-Sloan Equation

The Potts-Guy equation is useful in certain circumstances, but it does have

significant limitations. Flux, not permeability, is the most relevant clinical parameter.

Since flux equals permeability multiplied by concentration in the donor phase, there is a

tendency to equate increasing permeability with increasing flux. For a homologous series

of prodrugs delivered from an aqueous vehicle, log KOCT and log P rise predictably and

consistently with increasing alkyl chain length but flux does not. Since the Potts-Guy

equation is only concerned with permeability, it does not predict the important

experimental observation that the most water-soluble members of homologous series tend

to have the highest fluxes (Sloan and Wasdo, 2003). The use of KOCT to estimate KM:D S

not applicable to vehicles that are miscible with octanol. This is an important limitation

when studying prodrugs, since an aqueous donor is often incompatible with commonly

employed labile derivatives. Roberts and Sloan (1999) used the following mathematical

manipulation to estimate a partitioning coefficient, KSKIN:IPM, between the skin (where

SSKIN is solubility in the skin) and isopropyl myristate (IPM):

KSKIN:IPM = SSKIN/SIPM

KSKIN:IPM = (SSKIN/SAQ) / (SIPM/SAQ)

KSKIN:IPM = KSKIN:AQ / KIPM:AQ

log KSKIN:IPM = log KSKIN:AQ log KIPM:AQ (3.24)

KIPM:AQ can be used in the same manner as KOCT to estimate KSKIN:AQ,

log KSKIN:AQ = flog KIPM:AQ + b

log KSKIN:IPM = flog KIPM:AQ log KIPM:AQ + b

log KSKIN:IPM = (f-1) log KIPM:AQ + b (3.25)









For an IPM vehicle, (f-1) log KIPM:AQ + b takes the place of flog KOCT in the

Potts-Guy equation:

log P = (f-1) log KIPM:AQ + b PMW + c

log P = (f-1) log SIPM + (1-f) log SAQ PMW + b + c (3.26)

Adding log SIPM to both sides transforms the equation to flux (JM) instead of P:

log JM = flog SIPM + (1-f) log SAQ PMW + b + c

In the standard representation z, y and x are used respectively to replace P, f and the

combination of b +c:

log JM = x + y log SIPM + (1-y) log SAQ z MW (3.27)

This equation is the Roberts-Sloan (RS) model and it represents our primary method of

predicting flux through hairless mouse skin from IPM.

Modification of the Roberts-Sloan Equation to Include Synthetic Membrane Data

We also wish to construct a predictive transdermal model that includes the

compound's flux through a surrogate membrane as one of the descriptive parameters.

Although there are several possible forms for this model (depending principally upon the

degree of similarity between the surrogate membrane and skin), they are all derived from

the general flux equation (eq. 3.16). The logarithmic form of this equation for flux

through skin is shown below.

log Js = log Ss + log Ds log Ls (3.28)

Flux through an artificial membrane (X) should follow an analogous equation:

log Jx = log Sx + log Dx log Lx (3.29)

To incorporate a log Jx term into the skin equation, we must first decide which

parameter in equation 3.29 best estimates its counterpart in equation 3.28; the solubility









term or the diffusivity term. Between any membrane and skin, one or both of the

following relationships will be valid:

log Ss = As log Sx + Bs (3.30)

log Ds = AD log Dx + BD (3.31)

Relationship 3.30 is expected to be valid only in those situations where the

surrogate membrane is very similar to skin itself. In general, this is expected only to be

true when comparing skin to a similar biological membrane such as mammalian skin

from another species. In contrast, equation 3.31 is a more general relationship and should

be more widely applicable. Specifically, equation 3.31 will hold for all membranes in

which molecular weight and the diffusion coefficient are related as described in equation

3.21. From this we conclude that two models should be investigated; one for comparing

fluxes through related biological membranes and one for comparing synthetic to

biological membranes.

To derive a model for chemically similar membranes, both equations 3.30 and

3.31 are assumed to be valid. Equation 3.29 can be rewritten to show solubility in the

membrane in terms of the other variables.

log Sx = log Jx log Dx + log Lx

This relationship can be substituted into equation 3.30:

log Ss = As log Sx + Bs (3.30)

log Ss = As log Jx As log Dx + As log Lx + Bs

This expression for log Ss can now be substituted into equation 3.29:

log Js = As log Jx As log Dx + As log Lx + Bs + log Ds log Ls
(3.32)
Substituting equation 3.31 into 3.32 removes log Ds:









log Js = As log Jx As log Dx + As log Lx + Bs + ADlog Dx + BD- log Ls

log Js = As log Jx + (AD-As) log Dx + [BD + Bs + As log Lx log Ls]

All the terms appearing between the brackets can be collapsed into a single coefficient,

Ki:

log Js = As log Jx + (AD-As) log Dx + K1 (3.33)

From equation 3.21, log Dx = log Dox (1/ln 10)ax MW. If this relationship is

substituted into equation 3.32:

log Js = As log Jx (AD-As) (1/ln 10)ax MW + (AD-As) log Dox + Ki

If the constants are grouped renamed in the following manner:

(AD-As) log Dox + K1 = a

As = b

(AD-As) (l/In 10)ax = c

the result is the general form of similar membrane model.

log Js = a + b log Jx + c MW (3.34)

To develop a model to relate the fluxes through two dissimilar membranes,

equation 3.29 is rewritten in terms of log Dx and then substituted into equation 3.31:

log Dx = log Jx log Sx + log Lx (3.29)

log Ds = AD log Dx + BD (3.31)

log Ds = AD log Jx AD log Sx + AD log Lx + BD

log Js = AD log Jx AD log Sx + AD log Lx + BD + log Ss log Ls

log Js = AD log Jx + log Ss AD log Sx + AD log Lx log Ls (3.35)









In order to proceed past equation 3.34, some assumptions must be made about the

log solubility terms. In the RS equation, log Ss and log Sx are estimated from log SIPM

and log SAQ through the following equations:

log Ss = ys log SIPM + (1-ys) log SAQ (3.36)

log Sx = yx log SIPM + (1-yx) log SAQ (3.37)

where yx and ys are the corresponding y coefficients for skin (S) and the synthetic

membrane (X). If these two relationships are substituted into equation 3.35, the result is

an expression for log Js in terms of experimentally determined values:

log Js = AD log Jx + (ys ADyx) logSIPM + (1-ys AD + ADyx ) logSAQ + ADlog Lx -logLs

This expression can be greatly simplified by grouping and renaming constants:

ADlog Lx -logLs = a

AD= b

ys ADyx = c

log Js = a + b log Jx + c log SIPM+ (1 b c) log SAQ (3.38)

Equation 3.38 will only be predictive when the assumptions of the RS equation

are valid. In other words, only when log KM:IPM is related to log SIPM and log SAQ in the

following manner:

log KM:IPM = (f-1) log SIPM (f-1) log SAQ + b

The success of the RS equation in describing flux though hairless mouse skin from IPM

indicates that the above expression is reasonably followed for that system. However, it is

not yet known whether or not this can be applied as a general rule. When comparing two

widely different systems, it may be necessary to allow unconstrained coefficients for log

SIPM and log SAQ to estimate partitioning or solubility. In this case, the coefficients in






48


equation 25 would no longer be related and a more general form of model 3.38 is

produced:

log Js = a + b log Jx + c log SIPM + d log SAQ (3.39)

It will require empirical data to determine which of these two models is most applicable.













CHAPTER 4
EXPERIMENTAL DESIGN

Section I: Synthesis and Characterization of the 4-Alkyloxycarbonyl and 4-
Methyloxyalkyloxy Prodrugs of Acetaminophen

Synthesis

Five homologous n-alkyl and two methyloxyalkyl carbonates of 4-

hydroxyacetanilide (APAP) have been prepared. The general synthetic method for each

carbonate entailed reacting APAP with equimolar amounts of the proper

alkylchloroformate in the presence of a poorly nucleophilic base.

N O ROH 0 Base 0 O OR
A + RO C A 0
H H

A well-stirred suspension of APAP (-1.51 g, -0.01 mol) was prepared in 30 mL of

CH2C12 containing an equimolar amount of pyridine or triethylamine. To this mixture, a

solution of the desired alkylchloroformate (0.01mol) in -10 mL CH2C12 was added in a

drop-wise manner, and the reaction mixture was allowed to react for two hours. The

reaction solution was then diluted to approximately 200 mL and extracted sequentially

with 10 mL of -0.6 N HC1 and 10 mL deionized water. After the water wash, the CH2C12

solution was dried over Na2SO4 for two hours, filtered and concentrated under vacuum

until solvent free. The resulting material was purified by recrystallization and, if

necessary, column chromatography until a sharp melting point was observed, only one

component was discernable by TLC and a clean 1H-NMR was obtained. In this fashion,









the methyl, ethyl, propyl and butyl carbonates of APAP were prepared from

commercially available alkylchloroformates.

For the hexyl and the two methyloxyalkyl carbonates, it was necessary to first

synthesize the desired alkylchloroformate. This was accomplished by reacting the proper

primary alcohol with triphosgene, a synthetic phosgene equivalent, and a poorly

nucleophilic base (triethylamine or pyridine).

0 0
Ao Base O
3 ROH + CI3CO OCC3 B- 3 R CI


One molecule of triphosgene rearranges to ultimately yield three molecules of

phosgene during the reaction. Therefore a 3:1 molar ratio of alcohol to triphosgene was

used to maintain equimolar amounts of phosgene and the alcohol. A solution of

triphosgene (-0.0033 mol) in 20 mL CH2C2 was first prepared. To this solution, a

mixture of triethylamine or pyridine (-0.01 moL) and the alcohol (-0.01 mol) in 10 mL

CH2C12 was added at a rate of-1 mL/min in a drop-wise manner. The ensuing

exothermic reaction was allowed to proceed until the solution had once again returned to

room temperature. A suspension of APAP and triethylamine or pyridine (0.01 mol each)

in 20 mL of CH2C12 was then added to the alkyl chloroformate solution at a rate of- 2

mL /min. After being allowed to react for at least two hours, the mixture was washed

with aqueous acid and dried over Na2SO4 as described above. The crude product was

purified by recrystallization and column chromatography as necessary to achieve the

previously stated criteria of purity. The results and specific conditions for each prodrug's

synthesis are listed below.









2; 4-methyloxycarbonyloxyacetanilide This compound was prepared in 49 % yield from

methyl chloroformate and pyridine in CH2C12 after recrystallization from diethyl

ether/hexane.

3; 4-ethyloxycarbonyloxyacetanilide This compound was prepared in 82 % yield from

ethyl chloroformate and triethylamine in CH2C2 after recrystallization from ethyl

acetate/hexane.

4; 4 -propyloxycarbonyloxyacetanilide This compound was prepared in 59% yield from

propyl chloroformate and triethylamine in CH2C12 after recrystallization from ethyl

acetate/hexane.

5; 4 -butyloxycarbonyloxyacetanilide This compound was prepared in 63% yield from

butyl chloroformate and triethylamine in CH2C2 after recrystallization from ethyl

acetate/hexane.

6; 4-hexyloxycarbonyloxyacetanilide This compound was prepared in 51% yield from

hexanol, triphosgene and pyridine in CH2C2 after silica gel chromatography in ethyl

acetate and recrystallization from ethyl acetate/hexane.

7; 4-(2 '-methyloxyethyloxycarbonyloxy)acetanilide This compound was prepared in

44% yield from 2-methoxyethanol, triphosgene and triethylamine after recrystallization

from ethyl acetate/hexane.

8; 4-(1 '-methyl-2 '-methyloxethyloxycarbonyloxy)acetanilide This compound was

prepared in 29 % yield from 1-methyl-2-methoxyethanol, triphosgene and triethylamine

after recrystallization from diethyl ether/hexane.









Characterization

Melting point determination and 1H-NMR analysis comprised the initial

characterization of the 4-AOC and 4-MOAC APAP prodrugs. Melting points were

determined using a Meltemp capillary melting point apparatus and were uncorrected. 90

MHz 1H-NMR spectra were obtained in CDC13 using a Varian EM-390 spectrometer.

In addition to the initial characterization, molar absorptivities (s) in acetonitrile

(ACN) and pH 7.1 phosphate buffer containing 0.11% formaldehyde were measured to

facilitate quantitation in subsequent solubility and flux experiments. For each compound,

three replicate stock solutions were prepared by diluting 10 +1 mg portions of purified

material to 25 mL in ACN. Aliquots of these stock solutions were further diluted with

either ACN or buffer and the background-corrected absorbance of these diluted solutions

were measured from 400 to 200 nm using a Shimadzu UV 265 Spectrophotometer.

Maximum absorbance was observed at 240 nm in both solvents for APAP and each

prodrug. In addition, APAP displayed a pronounced shoulder at 280 nm in buffer that

was not evident in the carbonate derivatives. The molar absorptivity at these wavelengths

for each compound was determined by taking the average of the individual molar

absorptivities of the replicate solutions.

H (E)
(A) H3C N 0 R2
| O 0 R2

OB 0 0 H R1 (D)
(B)
(C)

Figure 4-1. Regions of the 4-AOCO-ACA prodrug corresponding to the letters given in
table 4-1.










Table 4-1. H NMR Data for the AOC and MOAOC-APAP prodrugs.

1H-NMR (6)a
Compound R R2 A B C D E
1, APAP -
2 -H -H s2.15 d 7.06, d 7.45 s3.95 --
3 -CH3 -H s2.12 d 7.03, d 7.42 q4.28 t 1.47 -
4 -CH2CH3 H s2.08 d 7.04, d 7.42 t 4.20 m 1.80, t 1.02 -
5 -(CH2)2CH3 -H s2.14 d 7.08, d 7.45 t4.28 t 0.95 -
6 -(CH2)4CH3 -H s2.14 d 7.04, d 7.42 t4.20 Unresolved m -
7b -CH2OCH3 -H s2.14 d 7.06, d 7.46 t4.38 t 3.68, s 3.42 ---
8 -CH2OCH3 -CH3 s 2.08 d 7.04, d 7.42 m 5.00 d 3.50, s 3.40 d 1.34
a Obtained in CDCl3 with TMS as an internal standard
bElemental analysis for C12H15NO5, Found (Expected): C= 56.84 (56.91), H = 5.98 (5.97), N = 5.53 (5.53).
"Elemental analysis for C13H18NOs, Found (Expected): C = 58.43 (58.42), H = 6.43 (6.41), N = 5.27 (5.24).

Table 4-2. Melting point and absorptivity values for APAP, the AOC-APAP and
MOAOC APAP prodrugs.
a a a
Melting S240a -240a -280
Comp. Point in in in
(C) ACN Bufferb Bufferb
1, APAP 167-170 1.36 0.981 0.191
2 112-115 (115.5-116.5)c 1.67 1.25 0.0560
3 120-122 (121-122)c 1.64 1.24 0.0483
4 104-106 (105-108)d 1.63 1.28 0.0560
5 118-120 (119-121)c 1.75 1.24 0.0376
6 108-110 (112.5-113.5)c 1.79 1.12 0.0623


7 78-81 1.74
8 120-123 1.60
units of M- x 104.
b pH 7.1 phosphate buffer with 0.11% formaldehyde.
Literature values in parentheses from Dittert et al, 1963
d Literature values in parentheses from Merck, 1897.


1.26
1.28


0.0623
0.0520


Section II: Determination of IPM and Aqueous Solubilities

IPM solubilities (SIpM) were determined directly by UV analysis on filtered IPM

suspensions of prodrugs that were subsequently diluted with ACN. To prepare a

suspension, a three mL volume of IPM was initially combined in a 15 mL test tube with

approximately twice the amount of APAP or prodrug needed to just saturate the IPM.









The resulting mixture was then stirred continuously at room temperature (23+1 C) for 72

hours. After stirring, the suspension was rapidly filtered through a 0.25 pm nylon syringe

filter and an aliquot of the filtrate (- 0.100 mL) was diluted to at least 10.0 mL with

ACN. The final solution was analyzed by UV to determine its absorbance at 240 nm.

The molar absorptivity of each compound in the diluted filtrate was assumed to be

equivalent to its' corresponding molar absorptivity in pure ACN. This assumption

allowed each compound's IPM solubility to be determined from the relationship:

SIPM = (Vfinal / Valiquot) A240/F240 (4.1)

where Valiquot is the volume of the saturated filtrate aliquot, Vfinal is the final diluted

sample volume, A is the sample absorbance at 240 nm and 8240 is the compound's molar

absorptivity in ACN at 240 nm. Three replicate suspensions were prepared and analyzed

for each compound. The average of the three values was reported as SIPM.

Aqueous solubilities (SAQ) were estimated using two methods: a direct dissolution

in water and a calculation using the compound's IPM:water partition coefficient. For the

direct measurements, suspensions of each compound in unbuffered deionized water were

prepared in the same manner as the IPM suspensions. However, to be consistent with the

preparation of the suspensions used in the diffusion cell experiments and to limit the

extent of hydrolysis during analysis, aqueous suspensions were stirred at room

temperature for only one hour before filtration and dilution in ACN. As with the IPM

solutions, the ACN diluted filtrates were analyzed by UV and the concentration of each

compound was calculated using equation 4.1.

To measure each compound's IPM:water partition coefficient (KIPM:4.0), a measured

volume of the prodrug suspension in IPM was placed in a 10 mL test tube along with









four milliliters of 0.01 M acetate buffer (pH 4.0). The test tube was capped, shaken

vigorously for 10 seconds and allowed to stand until the two phases separated. An

aliquot of the IPM layer was removed, diluted in ACN and analyzed by UV using the

same protocol followed for the saturated IPM solutions. KIPM:4.0 was calculated from the

concentration of prodrug remaining in the IPM phase after partitioning (CF), the initial

saturated concentration (SIPM) and the ratio of IPM and buffer volume used in the

partitioning.

KIPM:4.0 = (VAQ/VIPM) CF / (SIPM-CF) (4.2)

Within reasonable accuracy, KIPM:AQ is equal to the ratio of SIPM to the prodrug's

solubility in water (S4.0). Therefore, aqueous solubility is estimated from the relationship:

S4.0 = SIPM / KIPM:AQ (4.3)

It is important to note that, while this process does not provide a rigorously

measured solubility, it does provide a value that has been shown to be consistent with the

directly measured solubility (Taylor and Sloan, 1998). More importantly, it is a method

that can be used to estimate the solubility of prodrugs that are too unstable to permit a

direct solubility and, as such, can be used to compare compounds regardless of their

intrinsic stability.

Section III: Determination of Flux through Hairless Mouse Skin and
Polydimethylsiloxane Membranes

Preparation of the Membranes and Assembly of the Diffusion cells

Adult female hairless mice were rendered unconscious by CO2 and quickly

sacrificed by cervical dislocation. Whole thickness skin from the entire region was

immediately removed from each mouse by blunt dissection. Sections of this separated

skin were cut to a proper size and immediately mounted on the diffusion cell. For the









PDMS membrane, properly sized and shaped sections were trimmed from larger sheets

using the diffusion cell donor compartment as a guide. Just prior to mounting, the

trimmed sections were rinsed with water and MeOH and then blotted dry to remove

accumulated dust.

Shown in vertical profile by the figure below, a Franz cell consists of two separate

compartments. The upper or donor compartment (A) is essentially a glass cylinder with a

flared and grooved lower edge. The lower or receptor compartment (B) is cylindrical

reservoir with a grooved upper edge that mirrors the donor compartment. In addition, the

receptor reservoir is equipped with a temperature controlling water jacket and a side arm

that allows filling and access to the receptor fluid.

To assemble the cell for analysis, the donor side compartment (A) was inverted and

a section of membrane was placed over the opening. Hairless mouse skin sections were

placed with the epidermal side facing the donor compartment and gently stretched into

place until they completely covered the entire lower opening and edge of the donor

compartment without sagging. PDMS membrane sections were placed over the donor

compartment opening without further adjustment. Once the membrane section was in

position, a rubber o-ring was placed over it and aligned with the groove on the donor

cylinder. The receptor section (B) was then inverted and carefully aligned with the donor

section. The two sections were then clamped together with a screw locked spring clamp

and the assembled cell was returned to an upright position. After assembly, the receptor

compartment was completely filled with 0.5 M, pH 7.1 phosphate buffer containing 0.1%

formaldehyde by weight as an anti-microbial agent. This concentration of formaldehyde

i













Donor Phase
z


Membrane
Receptor
Phase

\sr ooB

Water ->| --







Figure 4-2. Fully assembled diffusion cell.

is essential for maintaining membrane integrity and will prevent membrane degradation

for more then 144 hours (Sloan et al, 1991). After filling the receptor compartment, care

was taken to ensure that no air bubbles remained adhered to the bottom of the membrane

and the fluid level in the side arm was adjusted to the same height as the membrane to

prevent any increased hydrostatic pressure. A magnetic stir bar was added through the

side arm and the cell suspended over a magnetic stir plate to continuously stir the

receptor buffer throughout the experiment. The water circulating through the insulating

jacket was set to 32 C. The cells were kept in contact with the receptor fluid for 48

hours prior to the application of the donor phase to leech out any UV active components

and to allow the membranes to equilibrate with the receptor phase. Twice during this 48-

hour conditioning period, the receptor phase was completely withdrawn and replaced

with fresh receptor buffer.









Preparation and Application of Donor Phases

IPM and aqueous suspensions of each compound were prepared. IPM suspensions

were prepared by stirring 0.6 to 1.0 mmol of test compound with 3 mL of IPM for 24

hours at room temperature. For each compound, the final suspension concentration

exceeded the compound's solubility in IPM by at least threefold. For aqueous

suspensions, similar mmol quantities of material were stirred with 4 mL of deionized

water, but the suspensions were stirred for only one hour prior to application to limit

hydrolysis of the prodrugs. In addition, new aqueous suspensions were prepared every

24 hours. For either vehicle, application of the donor suspensions occurred immediately

after preparation.

Just prior to application of the donor suspension, the entire receptor phase was

removed and replaced with fresh buffer solution. Either a 0.5 mL aliquot of well-stirred

IPM suspension or a 1.0 mL aliquot of well-stirred aqueous suspension was evenly

applied to the conditioned membrane surface as shown in figure 4-2. When working

with the aqueous suspensions, the donor compartments were covered with parafilm to

prevent excessive loss of water to evaporation. After removal from the diffusion cells, all

prodrug donor suspensions were analyzed by 1H-NMR to ensure the prodrug had not

hydrolyzed in the vehicle.

Sampling of the Difussion cells for Flux and Residual Skin Samples

To obtain a sample of the receptor phase, a 3 to 4 mL aliquot of receptor buffer was

removed from the sidearm by Pasteur pipette and placed in a test tube for subsequent

quantitation. To maintain sink conditions during the experiment, the remaining receptor

fluid was removed after each sampling and the entire receptor compartment refilled with

fresh buffer. The frequency with which samples were taken was determined by the rate









at which analyte accumulated in the receptor phase and it differed for each membrane-

vehicle combination. The table below outlines the sampling times for APAP and each

prodrug.

Table 4-3: Sampling times for APAP and the AOCA prodrug diffusion experiments


Compound
APAP


Cl AOCA


C2 AOCA


C3 AOCA


C4 AOCA


C6 AOCA


C2 MOACA


Ci3 MOACA


Membrane/Vehicle
HMS/IPM
HMS/Water
PMDS / Water
HMS/IPM
HMS /Water
PMDS / Water
HMS/IPM
HMS/Water
PMDS / Water
HMS /IPM
HMS /Water
PMDS / Water
HMS/IPM
HMS/Water
PMDS / Water
HMS/IPM
HMS /Water
PMDS / Water
HMS /IPM
HMS/Water
PMDS / Water
HMS /IPM
HMS /Water
PMDS / Water


Sampling Timesa
8.5, 19,22,25,28,31, 34,
11,24,36,52
24,36, 48
8, 19,22,25,28,31, 34,43,
10,24,27,30,33,36,39,48
24,36, 48
8, 19,22,25,28,31, 34,44,
14.5, 24, 27, 30, 33, 36, 39,
24,30, 36, 48
8, 19, 22, 25, 28, 31, 34, 43,
14.5, 24, 27, 30, 33, 36, 39,
24,30, 36, 48
7,20, 22, 25, 28, 31, 34, 43,
10,24,27,30,33,36,39,48
24,36, 48
7,20, 22, 25, 28, 31, 34, 43,
11,24,36,52
24,36, 48
8.5, 19, 22, 25, 28, 31, 34,
12,24,27,30,33,36,39,48
24,36, 48
8, 19, 22, 25, 28, 31, 34, 44,
12,24, 36, 48
24,36, 48


a Time in hours after initial application of donor suspension


Receptor phase samples were analyzed by UV analysis within an hour of

collection. Just prior to analysis, a background correction from 350 to 200 nm was

performed on the instrument using matched quartz cuvettes containing blank receptor

buffer. After this correction, the spectrum of each recently collected receptor sample was

taken over the same wavelength range.









Following the last sampling and refilling of the receptor compartment, the cells

were allowed to sit for 24 hours. For IPM treated hairless mouse skin, this period was

sufficient to leach any test compound remaining within the skin. As such, the

concentration of compound found in the receptor phase after these 24 hours was used to

determine the total amount of material that had been absorbed by the skin (the residual

skin amount). For compounds delivered from water through hairless mouse skin and

PDMS membrane, an additional 24-hour leeching period was necessary and the residual

skin amount was determined by totaling the amount of material leeched after the first and

second 24 hour periods.

Evaluation of Membrane Integrity

After the leaching period, a standard suspension of theophylline in propylene

glycol (400 mg/mL) was applied to the membranes to determine their general integrity.

Following the final residual membrane sample, the remaining receptor buffer was

removed, replaced with fresh solution and a 0.5 mL aliquot of the theophylline

suspension was applied evenly to the membrane surface. Samples of the receptor

compartment were taken at regular intervals using the same protocol as for a test

compound. A sufficient number of samples were taken over the subsequent 24 hours to

determine the flux of theophylline through the membrane.

Determination of Analyte Concentration and Extent of Hydrolysis

For any wavelength, the absorbance of each sample was assumed to result solely

from a combination of the absorbances of APAP and any intact prodrug. Using Beer's

law, this assumption can be stated mathematically as:

AX = Cpsp D + CDEDX (4.4)









AX represents the absorbance at wavelength X, Cp and CD are the respective

concentrations of the prodrug and drug, and Spx and fSD are the respective molar

absorptivities for the prodrug and APAP, respectively, at wavelength X. By measuring

the absorbance of a sample at two wavelengths (240 nm and 280 nm for APAP and its

prodrugs), it was possible to determine the values of Cp and CD by simultaneously solving

the resulting two Beer's Law equations.

A280 = CpP2800 + CDED280

A240 = CPSP240 + CDED240

Cp = (A240SD280 A280SD240)/(SP240D280 SP280SD240) (4.5)

CD = (A240 CpFP240) / ED240 (4.6)

Cp and CD were added to determine the total concentration of APAP species present in

the sample. The total amount of APAP species present was then determined by

multiplying total concentration by the volume of the receptor phase.

Determination of Maximum Flux (JM)

Maximum flux was determined from a plot of the cumulative amount of APAP

species delivered through the membrane versus time. Since the receptor fluid was

replaced after every sampling with clean buffer, the amount of compound found in each

sample was indicative of only that amount which had passed through the membrane

between sampling times. As such, the cumulative amount delivered was calculated by

sequentially adding these sampled amounts. A typical graph of cumulative amount

plotted against sample time is shown below.














y = 1.7012x 26.715
R2 = 0.9995


80
70
" 60
0 50
x 40
LL 30
5 20
I-
10
0


30 40 50 60 70
Time (h)


Figure 4-3. Plot of cumulative amount of prodrug delivered versus time for compound 3
from IPM through hairless mouse skin.

Linear regression was performed on data in the steady-state region of the plot


(usually data collected after 24 hours). The slope of this best-fit line was divided by the


cross sectional area of the diffusion cell to yield the compound's maximum flux, JM.


10 20
0 10 20














CHAPTER 5
RESULTS

Melting Point Behavior of the APAP Carbonates

As noted in chapter 1, nonionic compounds, in general, display a negative

correlation between melting point and solubility. As such, compounds with lower

melting points tend to have higher solubilities in lipid and water which favor higher

transdermal flux. It was also noted that the relatively high melting points of APAP and

of other phenol containing compounds are likely a consequence of their ability to form

strong intermolecular hydrogen bonds. It was anticipated that masking the phenolic

moiety of APAP would disrupt this ability and produce derivatives with consistently

lower melting points. Table 5.1 shows that this is indeed the case. All of the carbonate

prodrugs have melting points that are substantially lower then that of APAP. In

particular, the two highest melting prodrugs (3 and 8) melt 50C lower then APAP while

the lowest melting compound (7) melts approximately 90C lower.

Table 5-1: Melting points (C), log partition coefficients (log KIPM:AQ), log solubility ratio
(log SRIPM:AQ), log solubilities in IPM (log SIPM ) and log solubilities in water
(log SAQ) for APAP and the 4-AOC and 4-MOAOC prodrugs.
Melting
Point log Direct log Estimated log
Compound (oC) SIPMa log SAQab KIPM:4.b log S4.0a,b SRIPM:AQ
APAP, 1 167-170 0.278 2.000 -0.174
2 112-115 1.076 1.314 -0.156 1.232 -0.238
3 120-122 0.968 0.578 0.315 0.653 0.390
4 104-106 1.374 0.427 0.897 0.477 0.947
5 118-120 1.143 -0.372 1.497 -0.354 1.515
6 108-110 1.219 -1.324 2.706 -1.487 2.543
7 78-81 1.014 1.536 -0.300 1.314 -0.522
8 120-123 0.529 0.516 0.132 0.386 0.013
a In units of mM at 230C
b Using pH 4.0 buffer at 230C.









However, the drop in melting point does not follow any simple trend among the

prodrugs. The 4-AOC-APAP compounds demonstrate an odd-even alteration in their

melting points as the number of methylene groups in the promoiety increases. This

effect, where the melting points of homologs will alternately rise and fall with the

addition of each methylene group to the alkyl chain of the promoiety, has been observed

in several series of homologous series (Chikos and Nichols, 2001), but has not been seen

in the compounds of the 5-FU and 6-MP series (Roberts and Sloan, 1999).

Since the addition of each CH2 has a proportionally larger effect upon the entire

molecule when the promoiety is small, this behavior is generally more pronounced with

the first few members of a series. In the 4-AOC-APAP series, there is an average

alternating fluctuation of 12 C between the first four members. Perhaps more surprising,

the addition of a single methyl side chain to the ethoxymethyl promoiety of compound 7

resulted in a 40C increase in melting point and produced the highest melting prodrug of

either type (i.e. 8). Such erratic behavior highlights the difficulty in predicting melting

points apriori without empirical evidence.

Direct IPM and Aqueous Solubility

With the exception of compound 7, the melting points of the APAP carbonates fall

within a relatively small range (-20C). Given the relationship between melting point

and IPM solubility, a correspondingly small range in IPM solubility was expected and

consequently observed. The most IPM soluble derivative (4) is only eight times more

soluble than the least soluble compound (8). However, all of the prodrugs were more

soluble in IPM then APAP itself. Compound 2 showed a 6-fold improvement in IPM

solubility over APAP and even compound 8 was 1.8 times as soluble. This range of

solubility is similar to that of the 3-AC 5-FU series and the maximum IPM solubility of









this series ranks as forth among the nine homologous series so far studied (Sloan and

Wasdo, 2003).

Unlike the IPM solubility values, the direct aqueous solubilities of compounds 2

through 8 were substantially less than that of APAP. As previously noted, all of the

APAP prodrugs possessed lower melting points than APAP, which indicates a decrease

in the lattice energy of these compounds relative to APAP. This lower lattice energy is

most likely the result of masking the phenolic moiety. However, masking this moiety in

the prodrugs also removes some of their ability to form beneficial hydrogen bonds when

in aqueous solution. It is apparent from the consistent decrease in aqueous solubility

exhibited by the APAP prodrugs that the loss of hydrogen bond stabilization in solution

caused by masking the phenol, combined with the extra energy required to create a larger

cavity in solution to accommodate their larger molecular size, outweighs the benefits of

lower lattice energy for these compounds.

In terms of relative solubility, compound 2 had an aqueous solubility of

approximately 21% that of APAP, whereas compound 6 possessed only 0.04% of

APAP's water solubility. Compound 7 had highest water solubility of all prodrugs and

was 34% as soluble in water as APAP, while the other MOAOC derivative, compound 8,

had a water solubility of approximately 3% that of APAP. When compared to the entire

database (Sloan and Wasdo, 2003), compounds 2 and 7 have water solubilities greater

than two thirds of the characterized compounds. However, five out of the other nine

series have at least one member with water solubility as good or better then either

compound 2 or 7. The remaining APAP produgs rank only within the lower 50% of the

database's values. Therefore, the performance of these prodrugs is somewhat mediocre.









Partition Coefficients and Solubility Ratios

As discussed in chapter 3, there is a linear relationship in a homologous alkyl series

between the log of the partition coefficient and the molecular weight of the homolog (eq.

3-13). Since molecular weight increases proportionally with the number of methylene

units in the promoiety, a plot of log K versus methylene number will also yield a linear

relationship. The slope of this line, the 7t value, should be independent of the series'

parent molecule and, therefore, should be the same for all homologous series.

Empirically, this prediction has been supported. For the series comprising our database,

K7 values are very consistent with an average value of 0.58 and a standard deviation of

0.03. This consistency makes the K7 value a robust indicator of homologous series

behavior.

Linear regression on compounds 2 through 6 yields a 7 value of 0.58, which closely

agrees with the database average (Sloan and Wasdo, 2003). This is evidence that the

physical properties of this series are both internally consistent and behaving in a

theoretically predicted manner. Since the other series have displayed similar properties

and since this consistency is fundamental in determining comparable estimated water

solubility (S4.0), we can anticipate that the estimated water solubility for these compounds

should be compatible with the other estimated database values.

The two MOAOC derivatives differ by the addition of a methyl group as a side

chain rather then the insertion of a CH2 group into the existing alkyl chain. However, due

to its similar size, chemical composition and polarity, the addition of the methyl group

should have a similar effect upon partitioning as a methylene group. If the basic theory

of partitioning holds true, the slightly larger size of the methyl group should have resulted









in larger drop in log K between compounds 7 and 8 than what was measured for the

AOC-APAP compounds.

However, this was not the case. Using the difference in log KIPM:4.0 as an

approximation, a n value of 0.44 was obtained. This is significantly smaller then

expected and only 75 % of the n value for the AOC series. To develop a possible

explanation for these results, we must first examine the behavior of the actual IPM:water

solubility ratios (SRIPM:AQ = SIPM / directly measured SAQ) as a point of comparison.

By design, many members of the current database are unstable to hydrolysis even

at neutral pH. In contrast, the APAP carbonate prodrugs are much more stable and

afforded an opportunity to directly measure aqueous solubility and subsequently

determine SRIPM:AQ. If log SRIPM:AQ is plotted against methylene number instead of log

KIPM:4.0, a n value of 0.55 is obtained for the AOC compounds. As predicted, this is in

close agreement with the n value obtained from the log KIPM:AQ plot. The agreement is

so close that a difference of less then 0.1 log units was measured between estimated and

direct solubility for compounds 2 through 5. Even for compound 6, whose low solubility

values have the highest associated uncertainty, a discrepancy of less then 0.2 log units

was measured.

For the 4-MOAOC-APAP derivatives, the change in log SRIPM:AQ between

compound 7 and 8 was 0.53. While still lower then expected, this is 0.1 log units larger

then the corresponding change in log KIPM:4.0 and closer to the log KIPM:4.0 7t value for the

AOC-APAP prodrugs. There is too little experimental evidence to ascertain the reason

for this difference, however the higher log SR relative to log K does suggest that

unexpected partitioning behavior is responsible rather then an intrinsic attribute of the

MOAOC promoiety.









Permeability Coefficient Behavior

The permeability coefficient is the most often used parameter appearing in models

of topical delivery. As discussed in chapter 3, many authors have reported that the log of

the permeability coefficient correlates positively to the log of octanol:water partition

coefficient (log Kow) and this relationship was used by Potts and Guy (1992) in the

development of their permeability model.

log P = flog KocT + PMW + c (3.23)

It is this positive association with log Kow that has led in part to the misleading principle

that greater lipophilicty is absolutely beneficial to topical delivery.

Table 5-2: Log permeability values for the APAP prodrugs from IPM through hairless
mouse skin (log PMIPM), from water through hairless mouse skin (log PMAQ)
and from water through PDMS membrane (log PPAQ).

Compound log PMIPMa log PMAQa log PpAQa
2 -1.08 -2.69 -2.74
3 -1.73 -2.27 -2.39
4 -1.82 -2.04 -1.92
5 -2.15 -1.82 -1.43
6 -2.71 -0.79 -0.68
7 -1.12 -2.77 -3.16
8 -1.59 -2.77 -2.79
a In units of cm h-


For the AOC-APAP prodrugs, log PMAQ increases linearly with increasing size of

the promoeity in a manner similar to log KIPM:4.0 and if log KIPM:AQ is plotted against log

PMAQ, the following best-fit linear relationship is found:

log PMAQ = 0.652 log KIPM:4.0 -2.64 (r2 = 0.963) (5.1)

A plot of log PPAQ versus log KIPM:AQ yields a similar correlation.

log PPAQ = 0.805 log KIPM:4.0 -2.45 (r2 = 0.974) (5.2)







69



Mouse Skin/ IPM
-0.5
o Mouse Skin/ Water
A
A PDMS/ Water o
-1


-1.5 -

O 0




-2.5
02
A
-2.5


-3
A

-3.5
-0.5 0 0.5 1 1.5 2 2.5 3
log KAQ PM



Figure 5-1. Plots of log KIPM:4.0 versus log P for compounds 2 through 8 using flux data
from the three membrane /vehicle systems.

As expected, the permeability coefficient has a positive correlation to the IPM-water


partition coefficient for both membranes. However, a quite different picture emerges if


log PMIPM is plotted against log KIPM:AQ:


log PMIPM = -0.517 KIPM:AQ 1.36 (r2 = 0.940) (5.3)

When the delivery vehicle is changed from water to IPM, the correlation between


permeability and partition coefficient becomes negative. These are unusual findings for


they suggest that mouse skin behaves as a lipophilic membrane for one system and as a


hydrophilic membrane in another. One can reconcile this behavior by realizing that IPM


and water lie at opposing ends of the polarity spectrum. Therefore, the data more


accurately shows that, relative to water, mouse skin is a lipophilic membrane and, relative


to IPM, mouse skin is a hydrophilic membrane.









Conversion to the Parent Drug

As mentioned earlier, simple carbonates undergo chemical hydrolysis at a much

slower rate then do simple esters and are expected to be relatively stable in neutral

aqueous solutions. Dittert et al (1963) studied the hydrolysis of compounds 2, 3, 5 and 6

in pH 7.4 buffer and found them to have half-lives in excess of 150 hours. Hydrolysis

studies were not performed on the novel members of the APAP series (4, 7 and 8), but

aqueous donor phases from all diffusion cell experiments were collected after use,

allowed to evaporate overnight and were subsequently analyzed by 1H-NMR and melting

point. Even after more then 48 hours of exposure to an aqueous environment and

recrystallization from water, no evidence of hydrolysis was detectable for any member of

the series. This data indicates that chemical hydrolysis would not be responsible for a

significant release of parent drug during the diffusion cell experiments.

By comparison to chemical hydrolysis, carbonates are far more susceptible to

enzymatic hydrolysis. When Dittert and Swintoski (1968) exposed carbonates 2, 3, 5 and

6 to a 2% solution of human plasma, the most stable compound, 2, had a half life of three

hours; a 50 fold increase in the rate of hydrolysis over that in buffer. They also found the

rate of enzymatic hydrolysis increased with the size of the alkyl side chain. By

compound 6 (C6), the half-life in human plasma had fallen to only 11 minutes compared

to its half-life in buffer of 22,800 minutes; a 200 fold increase.

The percentage of intact prodrug appearing in the receptor phase the carbonate

series is consistent with these findings and indicative of enzymatic conversion back to

APAP, especially for the IPM data. Compounds 5 and 6, the most enzymatically labile

derivatives, were completely hydrolyzed during the experiment. In contrast, the most

enzymatically stable compound, 2, showed the highest percentage of intact prodrug when









delivered from IPM. There is also a general correlation between higher flux and a higher

percentage of intact prodrug. Two mechanisms could explain this. A higher flux may

result in a shorter residence time in the membrane and reduce the chance for interaction

between a prodrug molecule and hydrolytic enzymes in the skin. Alternatively, higher

flux also correlates with a higher concentration of prodrug in the skin, which could

overwhelm the enzyme system and allow a higher percentage of intact molecules through

the membrane. Regardless, the best evidence for the enzymatic role in hydrolysis comes

from the high percentage of intact prodrug appearing in the receptor phase of the enzyme-

free polymer membrane system. With the exception of compound 2, which was 70%

hydrolyzed, greater then 90% of each alkyl carbonate was recovered intact from the

receptor phase after passing through the PDMS membrane.

Flux of APAP and its AOC and MOAOC Prodrugs through Hairless Mouse Skin

The maximum flux, JM, of each compound through hairless mouse skin and PDMS

membrane is presented in table 5-3. The flux through hairless mouse skin from a

saturated solution of IPM is designated JMIPM, the flux through hairless mouse skin from a

saturated aqueous solution is designated JMAQ, and the flux through PDMS membrane

from a saturated aqueous solution is designated JPAQ. For convenience, these values are

shown in their logarithmic form. The table also contains the second application flux of

theophylline (Jj) corresponding to each compound and experimental condition.

When delivered from IPM, APAP permeates mouse skin more rapidly than 5-FU,

6-MP or theophylline. It also demonstrates a maximum flux through mouse skin higher

than half the prodrugs in the IPM database. In contrast, the carbonate derivatives of

APAP perform worse than this on average. Only two of the carbonate prodrugs, 2 and










Table 5-3. Maximum steady-state flux and second application flux of the AOC-APAP
and MOAOC-APAP prodrugs through hairless mouse skin and PDMS
membrane.
Compound log JMIPMa JJMIPMa log JMAQa JJMAQa log JpAQa JJPAQa
APAP, 1 -0.29 0.74 -1.73 0.015 -2.68 0.0013
2 -0.00 1.12 -1.46 0.034 -1.51 0.0013
3 -0.76 0.64 -1.62 0.078 -1.74 0.0017
4 -0.45 1.14 -1.57 0.072 -1.44 0.0018
5 -1.01 0.85 -2.17 0.051 -1.79 0.0013
6 -1.49 0.76 -2.28 0.018 -2.16 0.0017
7 -0.11 0.98 -1.45 0.033 -1.85 NA
8 -1.06 0.94 -2.38 0.022 -2.41 NA
an units of [mol cm-2 h1


3 -


2-


S- i -*-log SIPM
| -U-log SAQ
0 -A-log JMIPM
-*K log JMAQ
1 i -log JPAQ


-2 -


-3
1 2 3 4 5 6 7 8 9
Compound


Figure 5-2. Correlation between solubility and flux for APAP, its' AOC derivatives and
its' MOAOC derivatives.

7, give maximum flux values from IPM that are higher then APAP itself. Compound 2

delivered 1.95 times the amount of APAP through mouse skin as the parent drug and

compound 7 delivered 1.5 times as much APAP as the parent. Of the remaining series

members, compound 4 performed the best (maximum flux 70% that of APAP) and

compound 6 performed the worst (maximum flux 6% that of APAP).









From water, the flux through mouse skin of all compounds is substantially reduced.

APAP does not perform quite as well from water and has a lower aqueous flux than four

of the carbonate prodrugs. As with flux from IPM, compounds 2 and 7 deliver the

highest amounts of APAP, but compound 7 out performs compound 2 when delivered

from water (1.90 and 1.68 times APAP flux respectively). Compounds 3 and 4 produced

slightly higher aqueous flux values than APAP but, with flux values of only 1.16 and

1.09 times that of APAP, these differences lack significance. Compounds 6 and 8 are the

worst performing carbonates with compound 8 giving the lowest aqueous flux. It is

worth noting, that while the relative performance among APAP, the AOC compounds

and the MOAOC compounds changes depending upon the nature of the vehicle, the order

of performance for each type remains nearly the same.

The fluxes of the APAP prodrugs through PDMS membrane from water were

similar in magnitude to their fluxes from water through hairless mouse skin, but the order

of their performance was not the same. Through PDMS, compound 4 was the best

performing prodrug and it possessed a flux similar to that of compound 2. This is a

significant change from the flux through mouse skin where the flux of compound 2 was

triple that of compound 4 when delivered from IPM and 50 % higher when delivered

from water. Compound 5, the third most lipophilic prodrug, has a PDMS flux nearly

equal to that of compound 3 and a PDMS flux higher then that of compound 7. In

contrast, compound 5 has only one third the flux of compound 3 and one fifth the flux of

compound 5 through mouse skin from the same vehicle. An equally profound change is

in the performance of the parent APAP. APAP has the third highest flux through hairless

mouse skin when delivered from IPM and fifth highest when delivered from water, but it

has the lowest flux through PDMS.









Even though the stratum corneum is widely considered to be primarily a lipophilic

membrane, IPM solubility alone does not correlate well with flux through mouse skin for

either vehicle. The most IPM soluble prodrug, 5, has only the third highest flux from

both vehicles and the second most lipid soluble compound, 6, has the lowest flux from

both vehicles. In contrast, the most water soluble prodrug, 2, has the highest flux from

IPM and the second most water soluble prodrug, 7, has the highest flux from water.

Comparing compounds 3 and 4, which have nearly equal IPM solubility, a decrease in

water solubility correlates to a drop in flux. Conversely, when aqueous solubility

remains nearly constant and IPM solubility increases (such as between compounds 4 and

5) flux also rises. Even without a strict mathematical treatment of this data, it is apparent

that flux through mouse skin is benefited by a increasing both lipid and aqueous

solubility, regardless of the vehicle used.

The situation is very different for the purely lipophillic PDMS membrane.

Although the mouse skin and PDMS flux plots shown in figure 5-3 appear similar in

many respects, flux through PDMS shows a much greater dependence on lipophilicity

and a decreased dependence on hydrophilicity than flux through mouse skin. As stated

previously, the most hydrophilic prodrugs, 2 and 7, have the best flux through mouse skin

from either vehicle whereas the most lipophilic compound, 4, has only the fourth highest

flux. Through PDMS, however, it is compound 4, which has the highest flux and APAP,

the compound with the lowest IPM solubility, which has the lowest flux. Compounds 5

and 7 possess similar molecular weights and similar IPM solubilities but differ by an

order of magnitude in their water solubilities. Through PDMS membrane, their fluxes

are similar with the slightly more lipophilic compound 5 possessing the higher flux.

Through hairless mouse skin, it is the more hydrophilic compound 7 which has 7.9 times









the flux of compound 5 when delivered from IPM and 5 times the flux of compound 5

when delivered from water.

Application of the Roberts-Sloan equation to this and other data in the next chapter

will allow us to quantify the relative effects of IPM solubility, water solubility and

molecular weight on flux through PDMS membrane. However, even without a strict

analysis, it is obvious that, without some modification, flux through PDMS does not

correlate to flux through mouse skin. The physical basis for this effect is still a matter of

debate, but the favorable influence of aqueous and IPM solubility has been observed in

all series comprising the current databases. Determining the relative effect of log SIPM

and log SAQ will require a more rigorous model and will be addressed in chapter 6 when

this and the other mouse skin data are fit to the Roberts-Sloan equation.

Effect of the Vehicle on Flux through Hairless Mouse Skin

It is clear that the vehicle has a profound effect on flux through hairless mouse

skin. Flux values from IPM were nearly 10 times higher then flux values from water no

matter which compound was present in the vehicle. A similarly consistent 10-fold

increase was seen for the second application fluxes through IPM exposed mouse skin

compared to those through water exposed mouse skin. This is especially significant

because the same vehicle, propylene glycol, was always used to deliver theophylline.

The persistence of the increased flux through the IPM exposed mouse skin, even after

removal of the initial vehicle, supports the conclusion that exposure to IPM alters hairless

mouse skin in an irreversible way and permanently decreases its resistance to penetration.

The relative effects of IPM and water upon the permeation barrier of mouse skin

have been well characterized. In 2003, we compared the delivery of chemically stable 5-

FU and 6-MP prodrugs through hairless mouse skin from water and IPM (Sloan et al









2003). There were a sufficient number of compounds included in this study to prepare a

predictive flux model for each vehicle. The conclusions of these experiments will be

covered in more detail in the next chapter; however, an important finding was a nearly

constant ten-fold increase in the flux of compounds delivered from IPM. From this and

from the APAP prodrug data, it would appear that the influence of IPM on the

permeability of hairless mouse skin is consistent and predictable.

Residual Membrane Amounts

If the assumptions of the Roberts-Sloan model hold true and if mouse skin is

reasonably consistent between individuals, the amount of drug or prodrug remaining in

the skin after the diffusion cell experiment should be proportional to the solubility of the

prodrug in the skin. Furthermore, since higher solubility in the skin corresponds to

higher flux, a correlation should be observed between higher flux and higher residual skin

amount. This should be especially true for the lower weight prodrugs whose smaller

molecular size has a lesser effect on flux.

Table 5-4. Average residual amounts ( Std. Dev.) of APAP and its prodrugs remaining
in hairless mouse skin (HMS) and PDMS membrane after the flux
experiments.
Compound HMS/IPM',b HMS/Aqab PDMS/Aqab
APAP, 1 2.74 (0.70) 0.90(0.30) 0.18 (0.04)
2 5.45(+1.57) 0.95(0.15) 0.88(0.31)
3 1.08(0.13) 0.76(0.13) 0.38(0.06)
4 2.84(1.44) 0.95(0.22) 0.63(0.05)
5 1.91(+0.08) 0.25(0.05) 0.13(0.03)
6 1.79(0.43) 0.40(0.14) 0.12(0.02)
7 3.75(0.74) 1.56(0.22) NAC
8 0.64(0.12) 0.34(0.07) NAC
a Membrane/Vehicle.
b In units of pmols.
C Not measured.









The data does show a reasonable agreement correlation between the residual

membrane amount and the measured flux values for hairless mouse skin. From both

water and IPM, those compounds with the highest flux and next to highest flux (I and VI)

also delivered the highest amounts of APAP to the skin. For the remaining APAP

carbonates, the relative position of each compound, when ranked by flux, is within one,

or at most two positions, of its value when ranked by residual skin amount.

This correlation is acceptable but it does not take into account the effect of

molecular weight on flux. Flux and residual skin data can be log transformed and fit to a

simplified version of the Roberts-Sloan model with the residual skin amount being used

as a surrogate for skin solubility:

log JMV = x + log (Residual Skin Amount) z MW

The best fit equations for flux from water and IPM through hairless mouse skin and from

water through PDMS are:

log JMIPM = 0.650 + log (Residual Skin Amount) 0.00748 MW (r2 = 0.797) (5.4)

log JMAQ = -1.38 + log (Residual Skin Amount) 0.00113 MW (r2 = 0.820) (5.5)

log JPAQ = -2.86 + log (Residual Skin Amount) + 0.0069 MW (r2 = 0.851) (5.6)

Although the correlation coefficients are not particularly high, it must be

remembered that residual skin amount is more sensitive to minute differences in the

membrane then is maximum flux. Even for the highly homogenenous PDMS membrane,

which showed very consistent flux values, residual membrane amounts had an average

relative standard deviation of 19% which is only slightly smaller then the same values for

IPM and water treated mouse skin (23% and 22% respectively). Given the inherent

variability in the amount of compound absorbed into the membrane, equations 5-4, 5-5







78


and 5-6 are consistent with the homogenous membrane assumption for hairless mouse

skin.














CHAPTER 6
PREDICTIVE MODELS OF SOLUBILITY AND FLUX

Determination of the Coefficients of the General Solubility Equations

In theory, the melting point coefficient found in equations 3-10 and 3-14, Bo, can

be calculated directly by using the Walden's Rule, which states that ASF for most small

nonelectrolytes is approximately 56.6 J oK-1 mol-1. However, an empirical approach

based primarily upon our own solubility data was expected to give a better representative

value. It was also necessary to use our solubility data to determine a value for the

molecular weight coefficient, C. Using our 5-FU and 6-MP prodrugs and an additional

set of ACOM phenytoin (ACOM-PhT) prodrugs from Stella et al. (1999), aqueous and

IPM solubility, melting point and molecular weight data was analyzed using nonlinear

multiple regression to generate best-fit coefficients to equations 3.10 and 3.14 for each

series. The results of this analysis are summarized in tables 6-1 and 6-2.

Table 6-1: Series specific best fit coefficients to equation 3.10 using IPM solubility data
and physical properties from the 5-FU and 6-MP prodrugs.
Series ID AIPM Bo

1-ACOM-5-FU 2.41 0.0188
1-AOC-5-FU 2.98 0.0190
1-AC-5-FU 2.99 0.0144
1-AAC-5-FU 3.65 0.0217
6-ACOM-6-MP 2.19 0.0129
6,9-ACOM-6-MP 2.86 0.0225
3-ACOM-5-FU 2.37 0.0165
Average (Std. Dev) 2.74 (0.50) 0.0177 (0.0034)









Table 6-2: Series specific best-fit coefficients to equation 3.14 using aqueous solubility
data and physical properties from the 5-FU, 6-MP and phenytoin (PhT)
prodrugs.
Series ID AAQ Bo C

1-ACOM-5-FU 12.74 0.0167 0.0433
1-AOC-5-FU 12.16 0.0160 0.0416
1-AC-5-FU 10.58 0.0125 0.0412
1-AAC-5-FU 13.65 0.0238 0.0458
6-ACOM-6-MP 10.44 0.0106 0.0348
6,9ACOM-6-MP 14.23 0.0218 0.0395
3-ACOM-5-FU 12.00 0.0154 0.0403
3-ACOM-PhT 20.54 0.0230 0.0568
Average (Std. Dev) 13.29 (3.21) 0.0175 (0.0049) 0.0429 (0.0064)


The estimate for the molecular weight coefficient, C, appears to be the most

robust. The individual series values for C possess a lower variability then the

corresponding series values used to estimate the other coefficients and have a relative

standard deviation of 14%. The average C value of 0.0429 agrees very well with the

molecular volume effect on aqueous solubility of 0.0437 MV reported by Huibers and

Katritzky (1998) in their model for predicting the aqueous solubility of hydrocarbons and

halogenated hydrocarbons. Given that their model (which is based upon a dataset of

241compounds) and the prodrug series model are both attempting to measure the effect

of what is essentially pure steric bulk on aqueous solubility, it is encouraging that they

are in such close agreement.

When derived from IPM solubility data, the best-fit Bo values have a relative

standard deviation that is close to the relative standard deviation for the best-fit A values

(19% compared to 18%, respectively). When derived from water solubility data, the

relative standard deviations of the best-fit Bo and A coefficients increase to 28% and









24%, respectively. For both water and IPM data, the A coefficient estimates have less

relative variability than the Bo estimates. This somewhat undermines the supposition

that Bo should be more conserved between different series then A. While it is possible

that for a given solvent both A and Bo may be equally independent of the nature of the

solute, it is also possible that some of the consistency in the A value estimates is a

consequence of the homogeneity in the series used to generate the estimates. Of the

seven series used to estimate coefficients from IPM solubility (table 6-1), five of them

have 5-FU as the parent compound. Similarly, five of the eight series used to determine

the A and Bo coefficients from water solubility are based on 5-FU (table 6-2). Since the

A coefficient is, in theory, determined by a molecule's specific structure and since the

same parent moiety appears in all the members of the various 5-FU series, it is reasonable

that the small variation in the A coefficient estimates is a result of the large proportion of

5-FU compounds comprising the two data sets.

In addition to the preceding arguments, there are two observations, which lend

support to Bo being considered a universal value for prediction. First, despite the

somewhat large variability among the individual series values, the average IPM derived

Bo value is nearly identical to the water derived value. This is consistent with the

expectation that the Bo would be the same for all solvents. Second, when the average A

value for a given solvent is used to calculate potential Bo values for the members of a

series, these calculated Bo values are erratic across the series. Conversely, if the average

Bo value for a given solvent is used to calculate A values for series members, the A

values are more consistent across the series. Therefore, the predictive strength of

equations 3.10 and 3.14 was tested using the average values of C and Bo.










The intent of developing the solubility equations was to enable data from a single

series member to predict the behavior of the other series members. To establish the

ability of these equations to accomplish this, solubility data from the first member of each

series appearing in tables 6-1 and 6-2 were used to determine AIPM and AAQ coefficient

values for their respective series. The one exception to this was the ACOM-Th in which

the first series member appears to be an outlier so the second series member was used as

a replacement. In turn, these AIPM and AAQ values were used to calculate IPM and

aqueous solubility values for all remaining compounds covered in the tables. Figures 6-2

and 6-3 show the correlations between predicted and experimental solubilities in IPM and

water.


2.5


2-


1.5 -*
2 *
*



40

0.5


0
0 0.5 1 1.5 2 2.5
Calculated log SIPM



Figure 6-1. Calculated versus experimental log IPM solubility using equation 3.10 and A
coefficients determined from the smallest series members.











2.5

1.5

0 0.5

-0.5

2 -1.5

-2.5-

-3.5

-4.5
-4.5 -3.5 -2.5 -1.5 -0.5 0.5 1.5 2.5

Calculatedlog SAQ



Figure 6-2. Calculated versus experimental log SAQ using equation 3.14 and A
coefficients determined from the smallest series members.

The average absolute error for predicting solubility in both solvents is

approximately the same with 0.18 log units for the IPM model and 0.21 log units for the

aqueous model. This is approximately half that of the 0.41 log unit average absolute

error reported by Ran, Jain and Yalkowsky (2001) with their modified form of the

general solubility equation. The largest outlier is the C1-ACOM-Theopyilline compound

with an absolute average error of approximately one log unit for both models; twice the

magnitude of the next largest errors in both models. Since the IPM and aqueous

solubility of this compound are consistent with its' flux and with its' expected

partitioning coefficient, the most likely explanation for its' poor adherence to the

solubility models lies with its' melting point. It is quite possible that the original method

for purifying the C1-ACOM theophylline isolated a low melting point polymorph that

was converted to a more stable (and less soluble) form during solubility determination.









The C 1-ACOM-Theophylline notwithstanding, the low average absolute errors of both

models indicate that they are capable of making useful predictions and some useful

generalizations

Solubility Behavior of the 4-AOC and 4-MOAOC-APAP Prodrugs

The 4-AOC-APAP prodrugs exhibit solubility behavior that is typical of the other

homologous prodrug series. When the solubility data from compounds 2-6 were fit to the

general organic and aqueous solubility models (eq. 3.10 and 3.14), the following best-fit

equations were obtained:

log SIPM = 2.921 0.0200 (TM T) (r2 = 0.765) (6.1)

log SAQ = 11.183 0.0169 (TM T) 0.0400 MW (r2 = 0.990) (6.2)

The agreement between calculated and experimental solubility values generated

from these equations is very close for both IPM and aqueous solubility. For IPM

solubility, the average absolute error is only 0.059 log units and, for aqueous solubility, it

is 0.085 log units. The accuracy of the calculated IPM solubility values indicates that the

relatively poor correlation coefficient for the IPM solubility equation should not be

interpreted as a failure of the model. The poor correlation coefficient of equation 6.1 is

actually a consequence of the small range in log SIPM values possessed by the AOC-

APAP prodrugs. Since the difference in IPM solubility between the most and least

soluble prodrug in this series is only 0.406 log units, even the small observed deviation

from the from the model is sufficient to produce a poor r2.

The series independent coefficients, Bo and C, determined from the AOC-APAP

prodrug data are close to the corresponding average values obtained from the eight other

prodrug series. The melting point coefficient obtained from AOC-APAP IPM solubility

data is only 0.0023 from the combined database average Bo value, while the coefficient









obtained from aqueous solubility data is only 0.0008 from the database average.

Similarly, the molecular weight coefficient of the AOC-APAP series differs from the

database average C value by 0.0029. As with the other series, these observations support

the relative independence of these coefficients from series-specific influences.

The series specific solubility parameter for the IPM solubility, AIPM, ranged from

2.19 to 3.65 for the series comprising the database. The AIPM value for compounds 2-6

lies precisely in the center of this range. In direct comparison, this AIPM value is very

close to the corresponding values of the 1-AOC-5-FU, 1-AC-5-FU and 6,9-ACOM-6-MP

series. Of these three series, the 1-AOC-5-FU series also has a Bo value of 0.0190,

which is close to the 0.0200 Bo value of the AOC APAP series. Therefore, the AOC-

APAP series has an inherent organic solubility equivalent to that of 1-AOC-5-FU series.

In other words, for a given molecular weight and melting point, a 4-AOC-APAP prodrug

will have the approximately same IPM solubility as a 1-AOC-5-FU prodrug.

The situation is quite different for aqueous solubility. The AOC-APAP

compounds have the third smallest AAQ parameter of the homologous series yet

examined. Compared to the 1-AOC-5-FU series, AAQ for the 4-AOC-APAP series is

smaller by 0.98. With such a decrease in AAQ, a 4-AOC-APAP prodrug will therefore be

approximately one-tenth as water soluble as a 1-AOC-5-FU prodrug with similar

molecular weight and melting point. Such a comparison can be made between compound

2 and the C2 member of the 1-AOC-5-FU series. Both compounds have similar

molecular weights (209 and 202 amu respectively) but a 15C difference in melting point.

As predicted, despite having the higher melting point, the C2-AOC-5-FU compound has

a ten-fold higher water solubility. It is this systematically weak ability of the 4-AOC-









APAP prodrugs to interact favorably in aqueous solution, which, along with their only

average IPM solubility, ultimately results in their modest flux.

In Chapter 3, it was theorized that AIPM and AAQ for a given homologous series

could be estimated from a single series member so long as Bo and C were sufficiently

constant for all series. Using solubility data from the first member of each homologous

series in the database, IPM and water solubility for the remaining series members were

predicted from melting point and molecular weight with reasonable accuracy. This same

process was applied to the members of the 4-AOC APAP series. Using the IPM and

water solubility of compound 2, and using the average Bo and C parameters obtained

from the IPM flux database, the following AIPM and AAQ were obtained.

log SIPM = AIPM 0.0176 (TM T)

AIPM = log SIPM + 0.0176 (TM T)

AIPM = 1.076 0.0176 (113.5-25) = 2.63

log SAQ = AAQ 0.0176 (TM T) 0.0429 MW

AAQ = log SAQ+ 0.0176 (TM T) + 0.0429 MW

AAQ = 1.232 + 0.0176 (113.5-25)+ 0.0429 (209) = 11.76

Using these values for AIPM and AAQ, calculated IPM and water solubilities were

predicted for compounds 3-6.

Using an AIPM of 2.63, IPM solubility was very well predicted for the 4-AOC series

yielding an average prediction error of 0.10 log units. Using an AAQ value of 11.76,

water solubility is less well predicted, although, with an average prediction error of 0.24









Table 6-3. Predicted IPM and water solubilities for compounds 3-6.
Exp. Pred.
Log Log Exp. Pred.
Compound SIPMa SIPMa Log SAQa Log SAQa
3 0.968 0.944 0.653 0.499
4 1.37 1.23 0.477 0.180
5 1.14 0.979 -0.354 -0.666
6 1.22 1.16 -1.49 -1.69
a In Units of mM.


log units, the model is still reasonably accurate. It is important to note that the predicted

water solubilities are consistently lower then the experimental values. This suggests that

the data from compound 2 did not estimate a value for AAQ, which truly reflected the

series value. If compound 3 had been used to estimate AAQ instead of compound 2, then

the average error of prediction for log SAQ would have been 0.08 log units. This is a

problem intrinsic to this type of analysis and it highlights how care must be taken in the

interpretation of predicted results.

In addition to predicting solubility behavior across a series, the series specific

IPM and water solubility equations were developed to quantify the effects of various

physical parameters on solubility. Accounting for melting point and molecular weight

influences on solubility allows the inherent solubilizing effect of different promoieties to

be compared. Compounds 7 and 8 can be used to estimate the series specific AIPM and

AAQ values for ether containing carbonate promoities by following the same procedure

that was used to estimate these parameters from compound 2. Since the 4-AOC-APAP

prodrugs and the 4-MOAOC-APAP prodrugs differ only in the composition of the

promoiety, any difference in the AIPM and AAQ values must be due the promoiety alone.

The results of this analysis are summarized in Table 6-4.






























o Database Compounds
Compounds 3-6





1.5 2 2.5


Predicted log SIPM

Figure 6-3. Predicted versus experimental IPM solubilities for the 4-AOC-APAP
prodrugs (AIPM = 2.63).




2.5


1.5


0.5


-- -0.5


-1.5


-2.5


0
00

+c^c~


0
0 0o
00.


0 Database Compounds
* Compounds 3-6


-4.5 -3.5 -2.5 -1.5 -0.5


0.5 1.5 2.5


Predicted log SAQ

Figure 6-4. Predicted versus experimental aqueous solubilities for the 4-AOC-APAP
prodrugs (AAQ = 11.76).


X
l 1.5
0


"
* 1


Iz 1
&
K!


0 8
o

o




Full Text

PAGE 1

TOPICAL DELIVERY OF A MODEL PHENOLIC COMPOUND: ALKYLOXYCARBONYL PRODRUGS OF ACETAMINOPHEN By SCOTT C. WASDO A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2005

PAGE 2

Copyright 2005 by Scott C Wasdo

PAGE 3

This document is dedicated to my parents Ch arles and Barbara, my brother Shaun and my sister Christine.

PAGE 4

iv ACKNOWLEDGMENTS Looking back, it is clear that my success has depended upon the kindness and support of more friends and coworkers then I could possibly name here. Even so, a few individuals were so profoundly important that they deserve a special mention. First, I would like to thank my parents, Charles and Ba rbara for fostering my love of science and for providing me with a lifetime of unwave ring support. In addition, I would like to thank Charles Schmidt and Ian Tebbett, for th eir advise and the ear ly opportunities they provided, Nancy Szabo for years of encouragement and many campaigns on my behalf, Carolynn Diaz, for always carrying more th an her share of the burden, my committee members Margaret O. James and Stephen M. Roberts, who have been helpful and accommodating of my many idiosyncrasies and John Perrin, who selflessly agreed to accept the responsibility of chai ring my committee with the fore knowledge that my work would do little to advance his own research. Most of all, I would like to thank Kenneth B. Sloan who, above all others, has been instrumental in this accomplishment. He ha s shown tremendous patience and dedication to my progress and has been invaluable. I could not have found a better mentor.

PAGE 5

v TABLE OF CONTENTS page ACKNOWLEDGMENTS.................................................................................................iv LIST OF TABLES............................................................................................................vii LIST OF FIGURES...........................................................................................................ix ABSTRACT....................................................................................................................... xi CHAPTER 1 BACKGROUND AND SPECIFIC AIMS...................................................................1 The Case for Topical Delivery.....................................................................................1 Approaches to Increasing Topical Delivery.................................................................3 Basic Theoretical Considerations of Prodrug Design for Topical Delivery and Previously Synthesized Prodrugs.............................................................................6 Specific Objectives and Preliminary Work................................................................13 First Objective.....................................................................................................13 Second Objective.................................................................................................16 Third Objective....................................................................................................18 2 BASIC ANATOMY OF THE SKIN..........................................................................22 Hypodermis.................................................................................................................22 Dermis......................................................................................................................... 23 Epidermis....................................................................................................................26 3 DEVELOPMENT OF THE PREDICTIVE SOLUBILITY AND FLUX MODELS33 Derivation of the Series Specific Organic and Aqueous Solubility Equations..........33 Derivation of the Potts-Guy Equation........................................................................39 Derivation of the Roberts-Sloan Equation..................................................................43 Modification of the Roberts-Sloan Equati on to Include Synthetic Membrane Data..44 4 EXPERIMENTAL DESIGN......................................................................................49 Section I: Synthesis and Characteriz ation of the 4-Alkyloxycarbonyl and 4Methyloxyalkyloxy Prodrugs of Acetaminophen..................................................49

PAGE 6

vi Synthesis..............................................................................................................49 Characterization...................................................................................................52 Section II: Determination of IPM and Aqueous Solubilities......................................53 Section III: Determination of Fl ux through Hairless Mouse Skin and Polydimethylsiloxane Membranes.........................................................................55 Preparation of the Membranes and Assembly of the Diffusion cells..................55 Preparation and Application of Donor Phases.....................................................58 Sampling of the Difussion cells for Flux and Residual Skin Samples................58 Evaluation of Membrane Integrity......................................................................60 Determination of Analyte Concentr ation and Extent of Hydrolysis...................60 Determination of Maximum Flux (JM)................................................................61 5 RESULTS...................................................................................................................63 Melting Point Behavior of the APAP Carbonates......................................................63 Direct IPM and Aqueous Solubility...........................................................................64 Partition Coefficients and Solubility Ratios...............................................................66 Permeability Coefficient Behavior.............................................................................68 Conversion to the Parent Drug...................................................................................70 Flux of APAP and its AOC and MOAOC Pr odrugs through Hairless Mouse Skin...71 Effect of the Vehicle on Flux through Hairless Mouse Skin......................................75 Residual Membrane Amounts....................................................................................76 6 PREDICTIVE MODELS OF SOLUBILITY AND FLUX........................................79 Determination of the Coefficients of the General Solubility Equations.....................79 Solubility Behavior of the 4AOC and 4-MOAOC-APAP Prodrugs.........................84 Modeling the Flux of the 4-AOC an d 4-MOAOC APAP Prodrugs through Hairless Mouse Skin from IPM and Water............................................................89 Modeling the Flux of the 4-AOC and 4-MOAOC APAP Prodrugs through PDMS Polymer Membrane from Water..........................................................................100 Prediction of Flux through Hairless Mouse Skin from Flux through PDMS...........109 7 CONCLUSIONS AND FUTURE WORK...............................................................112 LIST OF REFERENCES.................................................................................................118 BIOGRAPHICAL SKETCH...........................................................................................125

PAGE 7

vii LIST OF TABLES Table page 4-1 1H NMR Data for the AOC and MOAOC-APAP prodrugs.....................................53 4-2 Melting point and absorptivity valu es for APAP, the AOC-APAP and MOAOC APAP prodrugs........................................................................................................53 4-3 Sampling times for APAP and th e AOCA prodrug diffusion experiments.............59 5-1 Melting points (C), log pa rtition coefficients (log KIPM:AQ), log solubility ratio (log SRIPM:AQ), log solubilities in IPM (log SIPM ) and log solubilities in water (log SAQ) for APAP and the 4-AOC and 4-MOAOC prodrugs................................63 5-2 Log permeability values for the APAP prodrugs from IPM through hairless mouse skin (log PMIPM), from water through hair less mouse skin (log PMAQ) and from water through PDMS membrane (log PPAQ)....................................................68 5-3 Maximum steady-state flux and second application flux of the AOC-APAP and MOAOC-APAP prodrugs through hairless mouse skin and PDMS membrane......72 5-4 Average residual amounts ( Std. Dev.) of APAP and its prodrugs remaining in hairless mouse skin (HMS) and PDMS me mbrane after the flux experiments........76 6-1 Series specific best fit coefficients to equation 3.10 using IPM solubility data and physical properties from the 5-FU and 6-MP prodrugs.....................................79 6-2 Series specific best-fit coefficients to equation 3.14 using aqueous solubility data and physical properties from the 5-FU, 6-MP and phenytoin (PhT) prodrugs....................................................................................................................80 6-3 Predicted IPM and water solubilities for compounds 3 6 ........................................87 6-4 Estimated AIPM and AAQ using data from compounds 7 and 8. ................................89 6-5 Predicted and experimental flux values for compounds 1 8 through hairless mouse skin from IPM...............................................................................................90 6-6 Predicted and experimental flux values for compounds 1 8 through hairless mouse skin from water.............................................................................................94

PAGE 8

viii 6-7 Molecular weights, log IPM solubilities, log aqueous solubilities, log maximum flux values through hairless mouse (log JMAQ) and log maximum flux values through PDMS from water for the chemically stable prodrug series.....................101 6-8 Calculated maximum log flux values through PDMS from water and errors of calculation for the chemically stable prodrug series..............................................105 6-9 Solubility, molecular weight a nd flux data for the PABA esters...........................107

PAGE 9

ix LIST OF FIGURES Figure page 1-1 Structures of the database parent mo lecules and the pKa values for potential promoiety attachment sites.........................................................................................8 1-2 Probable hydrolysis mechanisms of the 1-N-acyl 5-FU prodrugs at pH 7.4.............9 1-3 Probable hydrolysis mechanisms of the soft alkyl promoieties...............................10 1-4 Structure of naringenin.............................................................................................15 2-2 Structures of the principal ceramides comprising the lipid bilayers of the lamellar bodies.........................................................................................................29 3-1 Simplified diagram of an e xperimental diffusion apparatus....................................40 4-1 Regions of the 4-AOCO-ACA prodrug co rresponding to the letters given in table 4-1....................................................................................................................52 4-2 Fully assembled diffusion cell..................................................................................57 4-3 Plot of cumulative amount of pr odrug delivered versus time for compound 3 from IPM through hairless mouse skin....................................................................62 5-1 Plots of log KIPM:4.0 versus log P for compounds 2 through 8 using flux data from the three membrane /vehicle systems.......................................................................69 5-2 Correlation between solubility and fl ux for APAP, its AOC derivatives and its MOAOC derivatives................................................................................................72 6-1 Calculated versus experimental l og IPM solubility using equation 3.10 and A coefficients determined from the smallest series members......................................82 6-2 Calculated versus experimental log SAQ using equation 3.14 and A coefficients determined from the smallest series members.........................................................83 6-3 Predicted versus experimental IPM so lubilities for the 4-AOC-APAP prodrugs (AIPM = 2.63)............................................................................................................88 6-4 Predicted versus experimental a queous solubilities for the 4-AOC-APAP prodrugs (AAQ = 11.76)............................................................................................88

PAGE 10

x 6-5 Experimental versus calculated log maximum flux values through hairless mouse skin from IPM using equation 6.5................................................................91 6-6 Calculated versus experimental log maximum flux values through hairless mouse skin from water using equation 6-7..............................................................94 6-7 Calculated versus experimental l og maximum flux values through PDMS from water by Equation 6.13...........................................................................................102 6-8 Log IPM solubility versus log ma ximum flux through PDMS membrane from water for the aqueous database...............................................................................104 6-6 Calculated versus experimental log maximum flux values through hairless mouse skin from water by equation 6.23...............................................................111

PAGE 11

xi Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy TOPICAL DELIVERY OF A MODEL PHENOLIC COMPOUND: ALKYLOXYCARBONYL PRODRUGS OF ACETAMINOPHEN By Scott C. Wasdo August 2005 Chair: John Perrin Major Department: Medicinal Chemistry Topical delivery is an attr active route of administration for a number of therapeutic agents. However, many drugs possess physical and chemical properties that limit their ability to permeate the skin. By masking se lect functional groups on a drug with proper moieties, it is possible to crea te prodrugs with physical and chemical properties that greatly improve topical delivery. Efficien t development of these prodrugs requires knowledge of how the physical and chemical ch aracteristics of a dr ug influence dermal absorption. In response to this need, solubi lity and diffusion experiments were performed on prodrugs of 5-flurouracil, 6-mercaptopurine and theophylline to develop a model that predicts maximum flux of these compounds th rough hairless mouse skin when delivered from isopropyl myristate (IPM). The purpose of th is work was to expand this database to include phenol containing compounds and refine this model by synthesizing and characterizing a series of alkyloxyc arbonyl derivatives and a set of methoxyalkyloxycarbonyl derivatives of a model phenolic compound, acetaminophen

PAGE 12

xii (APAP). In addition, two new predictive flux m odels, a solubility ba sed model to predict flux from water and a new model ba sed upon solubility and flux through polydimethylsiloxane (PDMS) membrane as an additional parameter, were developed. As with the other prodrug series studie d, the acetaminophen prodrugs with the best combination of lipid (SIPM) and water (SAQ) solubility, the methyloxycarbonyl and the methoxyethyloxy carbonyl derivatives, had th e highest flux through mouse skin from both IPM (JMAQ) and water (JMIPM). The addition of APAP and its prodrugs to the IPM and aqueous databases produced R oberts-Sloan equations of log JMIPM = -0.501 + 0.517 log SIPM + (1 0.517) log SAQ -0.00266 MW and log JMAQ = -1.665+ 0.657 log SIPM + (1 0.657) log SAQ -0.00409 MW for flux through hairless mouse skin from IPM and water, respectively. These models can predict fl ux of a drug through hairless mouse skin from IPM with an average error of prediction of 0.16 log units and from water with an average error of 0.17 log units. A simple mode l was found to relate flux through PDMS membrane (JPAQ) and flux through hairless mouse skin from water. Using the equation, log JMAQ = -1.156 + 0.245 log SAQ + 0.409 log JPAQ, flux from water through hairless mouse skin was predicted with an aver age absolute error of 0.13 log units.

PAGE 13

1 CHAPTER 1 BACKGROUND AND SPECIFIC AIMS The Case for Topical Delivery The route of administration can have a profound effect upon the utility of a drug. For a given drug, the preferred method of ad ministration will be determined by that methodÂ’s influence on a number of factors including bioavailab ility, rate of drug delivery, ability to reach the physiologi cal target and patient complia nce. Historically, oral delivery has been the most common means of drug administration and it continues to be the preferred target route when developing new pharmaceutical formulations. However, despite this prevalence, there are numerous s ituations where an alternative route, namely topical delivery, has advantages. Topical delivery is an obvious considerat ion when formulating therapies to treat skin diseases or disorders. The accessibil ity of the skin makes site-specific delivery relatively easy compared to internal organ sy stems. If a drug ha s sufficient ability to permeate into the skin, it is possible to achie ve a high local drug conc entration in target areas with a minimum amount of drug. In a ddition to this greater efficiency, limiting exposure to the surrounding and distant tissues in this way reduces systemic toxic effects. In some cases, even agents that cause signifi cant complications when given orally or by injection can potentially be used topically without causing significant adverse effects. For example, the topical application of 5-FU is effective in treating actinic keratosis (Jorizzo et al, 2004) and psoriasis (Ljundggren and Moller, 1972) in amounts that spare the body from most systemic effects.

PAGE 14

2 Orally delivered drugs encounter nume rous physiological processes that can potentially limit their bioavailibilty. The aci dity of the stomach and the enzymes of the intestinal lumen can degrade a drug before it has an opportunity to reach the intestinal epithelium. The intestinal epithelium can prevent absorbed drugs from reaching the intestinal vasculature by retu rning them into the gut immediately following absorption through p-gylcoprotein efflux tr ansporters (Kunta and Sinko, 2004). As a final hurdle, drugs entering the intestinal vasculature from the epitheliu m are passed via the portal hepatic vein to the liver before reaching sy stemic circulation. While passing through the liver, they are subjected to the bodyÂ’s highest concentration of biotransforming enzymes. In particular, the lack of substrate speci ficity in the hepatic CYP enzymes makes numerous drugs potential targets for premat ure metabolism and is responsible for many, if not most, instances of low oral bioa vailability (Wrighton and Stevens, 1992). For some drugs, solving the problem of low bioavailability is no more difficult than merely increasing the dose accordingly. However, for other drugs, especially compounds that require complex synthesis or purification, this is simply not an option. A great deal of effort and ingenuity has been invested to develop delivery systems to circumvent the limitations inherent to oral delivery. However, despite this effort there are many examples of therapies that require such exte nsive or expensive formulation before they are amenable to oral dosing that oral delivery is not cost e ffective. Peptide based drugs (e.g. insulin) and some steroid based drugs (e .g., estradiol) are such instances. In these cases, delivery by an alternative route becomes attractive. Topical delivery is one alternative that pot entially can circumvent these difficulties. Topically absorbed compounds enter systemic circulation without undergoing an initial

PAGE 15

3 metabolism by the liver. There is enzymatic ac tivity in the skin, but with regard to the transformation of foreign compounds, it is primarily limited to nonspecific esterase activity. In addition, while some epidermal cells do express p-glyc oprotein transporters (Laupeze et al, 2001), there is no systematic arra ngement of these cells or the transporters at the skin surface, which would allow them to actively clear drugs in an efficient manner. Finally, the reported pH of the skin surface does vary widely (pH 3 to 6, Hemmingway and Molokhia, 1987), but in genera l it is close to 4.5, far less acidic than the interior of the stomach. Given this relati vely gentle environment, dermally absorbed drugs are more likely to reach systemic circul ation intact than are intestinally absorbed drugs. Instead of high enzymatic activity and active clearance, the skin possesses a specialized barrier to preven t the permeation of foreign materials. While the skinÂ’s permeation barrier does limit the rate with whic h drugs enter the ski n, this same barrier can be used in a beneficial way. Since movement through the skin is slower than absorption through the epithelial cells of the intestine, dermally delivered drugs often show a more sustained and consistent serum leve l than orally delivered drugs. This is an obvious advantage in a number of conditions (arthritis, hypertension, chronic pain, etc.) where maintaining a constant or near consta nt serum drug level for an extended time is the most effective dosage regimen. Approaches to Increasing Topical Delivery Formulation, or manipulation of the deliv ery vehicle, is th e most popular method for increasing topical delivery. These a pproaches can be further divided into two categories: formulations that interact with or alter the skin and those that do not. With non-interacting formulations, the basic principle is to adjust the polarity of the vehicle by

PAGE 16

4 altering its composition until optimum partitioning of the drug into the skin is achieved. However, there is an inherent limitation asso ciated with a partitioning driven system. According to calculations based in regular so lution theory, a 10-fold increase in flux is the maximum improvement accessible by altering vehicle polarity before damage to the skin occurs (Sloan, 1992). With interactive formulations, or penetration enhancers, the vehicle components are intended to reduce the sk in barrier. The disr uption of the barrier can be caused by the movement of solvent into the skin and/or by the leaching of components from the skin that are essential for maintaining the barrier. This is a common effect and it is well established that prolonged contact with many different solvents, both polar and non-polar, disrupts the skin and increases the flux of drugs in a reproducible manner. For example, the flux of theophylline through hairless mouse skin that has been in contact with isopropyl myri state (IPM) is approximately 50 times higher than through mouse skin th at has been exposed to wa ter (Sloan et al, 2003). More recently, several physical methods of circumventing the stratum corneum have been designed as an alternative to in teractive chemical modification. Electrical potential (Riviere and Heit, 1997) has been applied acro ss the skin to provide an electromotive force capable of driving char ged molecules though the stratum corneum. Both ultrasonic (Mitragotri and Kost, 2004) and laser (Doukas and Kollias, 2004) energy can reportedly induce temporary defects in the lipid barrier larg e enough to allow the transdermal delivery of macromolecules, incl uding therapeutically si gnificant amounts of insulin. Using techniques developed for the fabrication of computer chips, micron-sized needles have been coupled to standard transdermal patches (Prausnitz, 2004). When these patches are affixed to the skin surface, the needles pierce the skin to a depth just

PAGE 17

5 beneath the stratum corneum and provide a condu it from the drug reser voir directly to the viable epidermis. The drawback of incr easing flux by any disruptive method is that higher skin flux is achieved at the cost of greater damage or pert urbation to the skinÂ’s permeation barrier. The preceding methods attempt to over come the poor physical properties of a drug by changing the environment from which it is delivered or by changing the nature of the skin. While these methods do increase de livery under the right circumstances and can be used to improve stability, no external modification can adequately overcome limited intrinsic water or lipid solubil ity of the drug which limits its solubility in the skin. The most efficient method to increase solubili ty is to chemically modify the drug. Unfortunately, such modification often reduc es or eliminates the drugÂ’s beneficial activity. In addition, even if the modified dr ug remained active, it is a new entity and its pharmacokinetics and toxicity may be quite diffe rent from those of the original drug. A modified drug that persists during delivery in to the body, but reverts to the original drug after absorption, would reduce th e potential for such complica tions. Such an entity is referred to as a prodrug after the term used by Alfred nearly 50 years ago to describe a pharmacologically inactive molecule that becomes active follow ing some biological transformation. For the purpose of drug de livery, the ideal prodrug has favorable physiochemical properties, is at least 1000 times less potent than the parent drug, is stable enough to resist premature conversion, has no more toxicity than that attributed to the parent drug and will completely revert to the parent in vivo

PAGE 18

6 Basic Theoretical Considerations of Prod rug Design for Topical Delivery and Previously Synthesized Prodrugs There are many strategies to prodrug de sign, but the most direct involves conjugation of the parent drug to a moiety usi ng a linkage that is susceptible to chemical or enzymatic hydrolysis. In such cases, the linkage and its associated side chain are collectively referred to as the promoiety. Hydroxyl, amine, amide, carboxylate and cabonyl groups are the usual functional group targets for conjugation though any sufficiently nucleophilic or elec trophilic site can be used. A great number of potential promoieties have been investig ated ranging from relatively s imple ester type derivatives to multi-functional groups requiring a numbe r of sequential activation and hydrolysis steps before regenerating the parent drug. As one may expect, much of the fundamental research regarding topica l delivery has been accomplished using less complex promoieties. When evaluating the utility of a promoiety, it is a common practice to study a promoiety by preparing a series of produgs th at share the same parent drug, attachment point to the drug and promoi ety linkage, but differ by seque ntial addition of methylene units to the promoietyÂ’s alkyl side chains. These homol ogous series of compounds have been useful for they exhibit systematic cha nges in some key physic al properties (i.e., partition coefficient and pol arity) and therefore make correlations between these properties and therapeutic be havior easier to identify. Before an effective prodrug can be synthesized, one must first have an understanding of why the parent drug behaves p oorly. Initial chemical analysis of skin components had determined that the dermal barrier was composed primarily of lipophilic compounds (Downing, 1992). This agreed with empirical observati ons that highly hydrophilic drugs with low lipid solubili ty exhibited poor topical delivery.

PAGE 19

7 Consequently, attempts to improve topi cal delivery have centered upon preparing prodrugs that show greater lipid solub ility than the parent molecules. The ability of a drug to form strong intermolecular bonds with itself is the major obstacle to solubility in non-pol ar media. The strength of these intermolecular forces is reflected in the compoundÂ’s heat of fusion ( Hf). Assuming that Hf remains constant from the temperature of the solution (T) to the compounds melting point (TM), Hf can be correlated to the mole fraction solubility (X) of non-electrol ytes through the relationship: ln T) (T T RT H lnXM M f (1-1) where is the activity coefficient. From equatio n 1, it is clear that increases in heat of fusion, melting point and activity coefficient will reduce solubility. Heat of fusion is not a routinely measured property during charac terization of a new co mpound, but it can be eliminated from (1) by using GibbÂ’s relationship, Hf = Sf TM ln T) (T RT S lnXM f (1-2) Therefore, melting point is a conveniently determined measure of the heat of fusion and, to a first approximation, any change to the molecule that decreases melting point without disproportionately increasi ng the activity coefficient will improve solubility. Hydrogen-bonding functional groups are the mo st common structural features that contribute to high lattice energy a nd the resultant low lipid sol ubility of non-ionic drugs. If the masking of such a group were capable of disrupting its ab ility to hydrogen bond, then it would be a reasonable first choice fo r modification. This c hoice is reinforced by two fortunate attributes of hydrogen bonding functional groups. First, as they are often reasonably nucleophilic and, given the number of commercially available electrophilic

PAGE 20

8 reagents, they lend themselves to simple synt hetic schemes that can be used to produce a variety of promieties. Second, masking ev en a single hydrogen-bonding site can have a profound effect on solubility. A good example of this is the difference in isopropyl myristate (IPM) solubility of 6-mer captopurine (6-MP) and its 6-S-methyl carbonyloxymethyl derivative, where maski ng the SH group alone reduces the melting point from 320C to 124C and increases IP M solubility from 0.022 to 1.05 mM (Waranis and Sloan, 1988). Until recently, work in this lab has focused upon improving the topical delivery of the heterocyclic drugs 5-FU, 6-MP and theophylline by masking amide, imide and thioamide groups on the respective parent molecules. N H N H O F O N N H N N SH N H N N N C H3 O O CH3 5-Flurouracil (5-FU) pKa = 8.0 pKa = 8.5 pKa = 7.5 pKa = 11.0 6-Mercaptopurine (6-MP) Theophylline (Th) pKa = 8.8 Figure 1-1. Structures of the database parent molecules and the pKa values for potential promoiety attachment sites. The hydrogen-bonding moieties on these thr ee compounds are bonded directly or through resonance to multiple electron wit hdrawing groups, which significantly lowers their pKa values. Being approximately as ac idic as a phenol, these moieties are readily converted to their anionic form. This allows them to function readily as nucleophiles and makes their derivatization relatively simple. In addition, the stability of these anions

PAGE 21

9 makes them good leaving groups, which aides th e subsequent regeneration of the parent molecule from the prodrug. By 1999, seven homologous 5-FU, 6-MP a nd theophylline prodrug series had been prepared and characterized. The promoieties, which were used to prepare these series, could be classified as one of two general type s. In the first type, the nucleophilic site on the parent molecule was bonded directly to a carbonyl group resulti ng in an acyl type promoiety. In the second group, the nuc leophilic site on the parent molecule was separated from the carbonyl by a methyloxy spacer to form a soft alkyl promoiety. As separate entities, these promoieties are fairly simple chemical systems. However, when they are coupled to the parent molecule, the resulting prodrugs displayed a wide range in their chemical behavior and, subsequently have different stabilities to chemical and enzymatic hydrolysis. N N O O F N O H R . N N O O H F X RN O N N O O F O OR .N N O O F R O .N N O O F .O R . . N N O O F O OR OH H .-. . .. .RO OH O .N H N H O O F .. .. .. N H N H O O F .. .. .. N H N H O O F .. .. .. Alkylaminocarbonyl (AAC) 5-FU prodrugs Alkyloxycarbonyl (AOC) 5-FU prodrugs Alkylcarbonyl (AC) 5-FU prodrugs.. .. .. .. C +.. .. .. .. .. .. .. .. .. + C + OH2.. .. .. .. .. ..+ H2O + H2O H2O Figure 1-2. Probable hydrolysis mechanisms of the 1-N-acyl 5-FU prodrugs at pH 7.4.

PAGE 22

10 XH X O R O :OH XH X O R O O R O X XH O R O X : : Drug X Drug.. O O OH X R : : Drug .. Soft Alkyl Prodrug X = acidic SH, NH or OH Parent Drug Drug Mechanism 1 + CH2O RCOO H + H2O Drug Mechanism 2 Drug.. .. .. .. .. : + Drug CH2+ H2O Drug + CH2O Drug Figure 1-3. Probable hydrolysis mechan isms of the soft alkyl promoieties. Figure 1-2 illustrates likely hydrolysis mechanisms of the N-acyl prodrugs of 5-FU at physiological pH and demonstrates how functional groups on the parent drug can interact with the promoiety to result in unexp ected behavior in the prodrug. If each 5-FU prodrug hydrolyzed via a typica l addition elimination mechanism, then the expected order of stability for the Nacyl promoieties would be AAC > AOC > AC. Empirically, however, the AAC 5-FU prodrugs exhibit lowe r chemical stability than the AOC type prodrugs and possess half-lives of only 8 to 11 minutes, which is inconsistent with such a mechanism (Sloan et al, 1993). As an altern ative, it has been s uggested that once the acidic N3 H of the substituted 5-FU (pKa 6.6) (Burr and Buungard, 1985) becomes ionized, it is capable of acting through resonance with the C2 oxygen as a general base in an intra-molecular E1cb type hydrolysis mechanism (Sloan et al, 1993). This is significant since it is only after th e addition of the promoiety to the N1 that the N3 becomes acidic enough to be predominan tly ionized at physiological pH.

PAGE 23

11 The AOC 5-FU compounds are believe d to hydrolyze through an addition elimination mechanism with water acting as the nucleophile as shown in Figure 1-2 (Buur and Bundgaard, 1986). Since the AC 5-FU prodrugs have a more electrophilic carbonyl than the AOC prodrugs and are therefor e better targets for nuc leophilic attack, a similar mechanism would be expected for the hy drolysis of the AC promoiety. However, increasing steric bulk in the al kyl portion of the AC promoiet y increases the rate of their hydrolysis which is more consistent with an SN1 mechanism (Buur and Bundgaard, 1984). The soft N and S alkyl prodrugs of th eophylline and 6-MP are believed to regenerate their respective parent drugs by the initial hydrolysis of th e ester portion of the promoiety followed by the decomposition of the resulting hydroxymethyl compound shown as mechanism 1 in Figure 1-3 (Sloan and Wasdo, 2003). This mechanism tends to be found in those ACOM derivatives where the masked functional group on the parent molecule is reasonably acidic (Bundgaard et al, 1985). It is of interest to note that for amide prodrugs wherein the masked amide func tionality has a pKa of approximately 15, the ACOM promoiety hydrolysis mechanism changes to the SN1 process shown in mechanism 2 of Figure 1-3 (Bundgaard et al, 1991). Since both of these mechanisms occur relatively slowly at neut ral pH, the ACOM prodrugs are stable for several hours in aqueous solution. In addition to the wide range of chemi cal properties, the he terocycle prodrugs demonstrated a wide range of physical propert ies. Melting points for these compounds range from 57.5-212 C, IPM solubility ranged from 0.3-174 mM and aqueous solubilities ranged fr om 0.001-182 mM (Roberts and Sloan, 1999).

PAGE 24

12 Along with their physical properties, the ab ilities of these prodrugs to permeate hairless mouse skin from saturated solutions of prodrug in IPM were also measured. The seven homologous series contained a tota l of 39 prodrugs. To these compounds, solubility and IPM flux data for 5-FU, 6-MP theophylline and the pivaloyl ACOM 5-FU were added to produce the largest set of compounds for which physical properties and in vitro flux data had been measured under the sa me experimental conditions by the same laboratory. The purpose of compiling this database wa s to provide empirical data for new predictive flux models. Flux and permeati on models for topical delivery had been developed by a number of researchers and their development will be discussed in detail in chapter 3. For now, it is sufficient to note that each model was based upon lipid solubility being the principle predictor of absorption into the skin. Whether lipid solubility was determined from solubility in a model lipophilic solvent or from a partition coefficient, these models shared a common deficiency in predicting the behavior of homologous series. For each series in th e 43 compound database, the more water-soluble member of the series had the highest flux th rough mouse skin (Roberts and Sloan, 1999). The previously published flux models failed to predict this qualitativ e observation. Using the permeability model of Potts and Guy as a foundation, Roberts and Sloan were able to produce a flux model that overcame this deficiency and quantified the positive influence of aqueous solubility in determining flux (Roberts and Sloan, 1999). In subsequent work, this same model was applied to data obtained from drugs delivered from mineral oil through human skin in vivo (Roberts and Sloan, 2001). As with the in vitro mouse skin data, aqueous solubility was shown to exert a positive influence.

PAGE 25

13 Specific Objectives and Preliminary Work First Objective The first objective of this work is to expand the 43 compound database with an additional series of prodrugs based upon a ne w type of parent molecule. Given the benefits and limitations inherent to topical delivery, it seemed reasonable to examine situations for which sustained long-term delivery is the preferred regimen. Of course, the treatment of disease is not the only instance were it is desirable to administer compounds in this way. It is well established that the two leading causes of death in the United States in 2001 according to the CDCÂ’s National Center for Health Statistics, cancer and heart disease, are conditions that de velop over many years. Once eith er of these diseases reach an advanced state, they are often impossible to treat. Given the diffi culty of treatment for advanced conditions, a prophylactic approach to prevent or slow their onset is desirable. In the last decade, there has been an increase d interest in the use of natural products to provide such prevention and to increase genera l health. While the beneficial attributes of many natural preparations are undoubtedly over-s tated, the benefits of some naturally occurring compounds meri t further research. A number of epidemiological studies re port a lower cancer rate among individuals whose diets are rich in cert ain fruits and vegetables (L ui, 2004). While non-dietary influences in these studies are hard to c ontrol and they often suffer from poor data collection (Michels, 2005), several compounds are suspected of being responsible for these benefits. Specifically, the poly phenolic flavanoids, or more commonly polyphenols, are thought to play an important role in chemoprevention (Yang et al, 2001). Polyphenols are a large cl ass of plant-derived compo unds that contain either a flavone or flavanone structure and multip le hydroxyl groups. The phenolic groups are

PAGE 26

14 often glycosylated or alkylated by the plant (presumably to protect them from oxidation) and it is these forms that they are principally found in the diet. During digestion, the hydroxyl groups are uncovered. This is an im portant process, as many of the beneficial effects attributed to polyphenols are likely to require the presence of at least one free phenolic hydroxyl group. Polyphenols demonstrate an influen ce on numerous physiological processes in vitro and in vivo. They can function as antioxidants an d they have been shown to protect against oxidative stress in vitro and in vivo (Vinson, 1998), although the clinical significance of this has not been determined. Polyphenols reportedly have the potential to modulate radical mediated signaling pathways Of these, the ability to inhibit nitric oxide synthase may be significant due to its participation in inflammatory processes (Wu and Meninger, 2002). Specific polyphenols ha ve been reported to modulate enzymes involved in carcinogenesis, namely tyrosine kinase (Lin, 2004) and NF kappa B (Kundu and Surh, 2004). Perhaps more importantly polyphenols inhibit CYP-450 enzymes and may reduce the bioactivation of procarcinogens. This i nhibition is presumably the mechanism by which green tea phenols prevent methylcholan threne and phorbol acetate induced tumor genesis in mouse skin (Wang et al, 1989). While it is still unclear which effect or combination of eff ects is responsible for the perc eived benefit of polyphenols, if they are efficacious, then having a chronic supplementation based upon this class of compounds and a method to dermally deliver them would be useful. The triphenolic flavanone naringenin (4Â’, 5, 7-trihydroxyflava none), a well-studied inhibitor of hepatic CYP-450 3A 4 (Geungerich and Kim, 1990) was investigated for our initial experiments.

PAGE 27

15 Figure 1-4. Structur e of naringenin. The melting point of naringenin is 247-250 C. When it is reacted with excess acetyl chloride, the resulting triacetat e product has a melting point of 72-85C (unpublished results), indicating a substantial drop in lattice energy. However, even with the substantial drop in melting point, the aqueous and IPM solubility of the triacetate as well as its flux through mouse skin remain ed very low (unpublished results). When attempts to prepare and purify diesterified derivatives of na ringenin and other flavanoids proved especially difficult, it was decided to seek a simpler phenol containing model compound. Hydroquinone was considered as a model compound, but it was discarded after the mono-acetate was found to dispropor tionate into the diace tate and hydroquinone during purification. Acetaminophen (APAP) wa s eventually chosen as a replacement. Though APAP itself is not polyphe nolic, it was considered rele vant because it possesses structural features similar to polyphenols. In particular, APAP contains one phenol, has a predominantly planar shape, aromatic char acter and an amide group, which, along with the phenol, allows it to form multiple hydrogen bonds. More importantly, it was anticipated that the difference in pKa va lues between the phenol and amide groups (approximately 10 versus 13 respectively ) would allow selective ionization and O OH O O H OH

PAGE 28

16 consequently derivatization of the pheno lic oxygen without a significant amount of competing derivatization at the amide nitrogen. Second Objective The choice of promoiety was guided by the second objective of this work. Although the Roberts-Sloan equation was deve loped specifically for delivery from IPM and was initially applied to a relatively sp ecialized group of compounds delivered from a single vehicle, its theoretical foundation suggested that it should have wider applicability. If an aqueous vehicle is assumed in the deri vation of the Roberts-Sloan model, the final form of the equation remains unchanged. As a result, the model could be used to predict flux from IPM or water depending upon whic h values are used for the x, y and z coefficients. To estimate the coefficients for the aqueous model, Sloan et al (2003) collected the eighteen most chemically stable ACOM prodrug s present in the 43 compound database and measured their flux through hairless mouse skin from unbuffered water. To make the coefficients of the model more significant, more compounds would be needed in the dataset. In order to add the APAP prodrugs to this group, they would need to contain promoieties that were resi stant to hydrolysis in aqueous solution. Despite their prevalence among commonly prescribed drugs, very few phenolic compounds have been converted to prodrugs in an effort to improve their topical delivery. Phenol masking prodrugs of only four related compounds, morphine (Drustrup et al, 1991), bupenorphine (Stin chcomb et al, 1996), nal buphine (Sung et al, 1998) and naltrexone (Stinchcomb et al, 2002) have been prepared and studied in flux experiments. The promoieties for most of these studies have been limited to simple alkyl esters. Although they improved topical de livery of these narcotic anal getics, simple alkyl esters, especially esters of phenols, are too unstabl e for prolonged exposure to aqueous solution.

PAGE 29

17 Swintowski (in Dittert et al 1963) synthesized some simple alkyl esters of acetaminophen along with the corresponding alky l carbonate esters and studied their stability to chemical and enzymatic hydrolysis As expected, he found that the less el ectrophilic carbonate esters hydrolyzed times more slowly in pH 7 aqueous buffer than the more electrophilic alkyl esters and both groups were suscep tible to enzymatic hydrolysis. Even though Swintowski did not synthesize all of the al kyl homologs, he did synthesize the methyl carbonate, which is expected to undergo chemical hydrolysis more rapidly than any of the longer chain carbonates. Given the long halflife of the methyl carbonate, we concluded that the carbonate functional group had adequate stability to serve as a promoiety for the APAP prodrugs. Since the importance of increasing biphasic solubility when optimizing topical delivery is now well established, it ma y seem counterproductive to only study promoieties that contain alkyl side chains of ever incr easing length. While these promoieties can produce compounds with redu ced crystal lattice energy, their other properties are counter productive to good water solubility. When a compound dissolves in water, each solute molecule unavoidably disrupts a number of water-water hydrogen bonds by displacing water molecules from the volume it must occupy. Thermodynamically, this process has a substan tial negative enthalpy. Once in solution, any functional groups on the solute molecule th at are capable of forming new interactions with surrounding water molecules can reduce th is enthalpy cost. Alkyl groups are only capable of interacting through relatively weak Van der Waals forces and, therefore, do little to recover the enthalpy that was lost duri ng dissolution. It can be conjectured that

PAGE 30

18 replacing the alkyl side chai ns with entities that are cap able of forming beneficial interactions would increa se aqueous solubility. The challenge behind this approach is to find a functional group that has a large enough dipole moment to interact favorably with surrounding wate r molecules without unduly increasing crystal lattice energy. Polyether functional groups have these properties. Straight chain alcohols larger th an propanol are immiscible with water. By contrast, polyethyleneglycols (PEGs) continue to be soluble in water even when their molecular weight grows over several thousand amu. Greenwald has used conjugation to large PEG molecules to improve the water solubility of a number of poorly soluble amine, imide and hydroxyl containing compoun ds (Greenwald et al, 2000). However, Greenwald favors the use of high molecular weight PEG conjugates (>20,000 amu), with the promoiety comprising the overwhelming majo rity of the prodrugÂ’s total mass. These large PEG prodrugs remain water-soluble primarily because their characteristics remain close to those of the unconjugated PEGs. Fo r promoieties containing only a few or even one ether promoiety, the question becomes to what extent will the replacement of a methylene group by an oxygen improve water sol ubility? To address this question, two additional promoieties will be synthesized that have such a modification; a methoxyethoxycarbonyl APAP and a met hoxyisopropoxy carbonyl APAP. It is hypothesized that for both types of prodrugs the best performing compounds will be those that possess the best combination of lipid and water solubility. Third Objective Mammalian skin continues to be the membrane of choice for assessing drug permability in vitro. Although there are substantial di fferences in the thickness of the various skin layers between species, the general histology, chemical composition and

PAGE 31

19 physiochemical properties of the skin barri er are reasonably similar for all mammals. Therefore, while absolute perm eability varies with skin thickness, the relative flux of compounds through mammalian skin remains consistent. In other words, those compounds that diffuse most readily though human or pig skin are likely be the same compounds which diffuse most readily though mous e or rat skin. In particular, if the compounds of a given homologous series ar e ranked according to their flux through hairless mouse skin, it has been well establis hed that this order will match the rank order of their flux through human skin (Scheuplein and Blank, 1973, and Sloan et al, 1997). This ability of hairless mouse skin to predict the best performing members of a homologous series is one of the attributes that has made it a popular choice for flux experiments. In contrast to human skin, m ouse skin has much less variability and data collected from mouse skin experiments can be used without normalization. This obviates the multiple additional control diffusion cells, which must be run with in vitro human skin experiments. This consistency also a llows even small differences in fluxes to be discerned. Unlike human skin, which is us ually dermatomed or exposed to heat or enzymatic digestion to isolate the outermost skin layers prior to use, hairless mouse skin can be used full thickness and requires little more preparation than removal from the mouse. Despite these advantages, the use of m ouse skin, or any mammalian skin, does have drawbacks. The care of the animal prio r to use is expensive and special care must be taken to maintain the integrity of the skin for the duration of the experiment. Increasing regulatory requireme nts and ethical concerns about the use of animals in research makes finding an artifi cial membrane that can be used in the place of animal

PAGE 32

20 tissue desirable. Most attempts to predict skin permeation using an artificial membrane have focused upon finding a surrogate material whose properties are sufficiently similar to human skin to allow a direct comparis on. Untreated lipophilic polymers such as polydimethyl silioxane (PDMS) (Geinoz et al, 2002, and Cronin et al, 1998) and solvent modified lipophilic polymers (Twist and Zatz, 1990, Maitani, 1996, and DuPlessis, 2001) have been popular systems for this purpose. However, none of these systems have proven particularly useful at predicting flux through skin. The complexity of the skin makes it unlikely that simply knowing the flux of a drug through a simple polymer, modified or not, will be sufficient to predict topical delivery. Despite this limitation, there ar e aspects of any compoundÂ’s diffusion behavior which hold true regardless of the nature of th e membrane. For example, the kinetics of a compoundÂ’s dissolution, as well as its propensity to cluster or st ack in solution, can affect its flux in a manner that is difficult to determ ine from its solubility alone. Therefore, despite its inability to predict flux thr ough skin directly, a compoundÂ’s flux through a polymer membrane may still contain informa tion that is useful for predicting its flux through skin. However, extracting and utiliz ing this information will likely require a more sophisticated treatment then has yet been reported. Thus, it is the final objective of this re search to produce a predictive model for flux through hairless mouse skin that is ba sed upon its flux through PDMS membrane in addition to its solubility prope rties. PDMS membrane is a rubbery polymer composed of cross-linked chains of repeating Me2SiO units. As a rubbery polymer, the chains that comprise the membrane are flexible and retu rn quickly to an equilibrium position after being disturbed. Like diffusion through the skin diffusion in such a material is believed

PAGE 33

21 to be Fickian in nature (Crank, 1975). It is expected that the additional information provided by PDMS flux data will improve th e predictability of the solubility-based models and this will serve as first step in cr eating an artificial membrane system that will be useful in screeni ng for topical delivery.

PAGE 34

22 CHAPTER 2 BASIC ANATOMY OF THE SKIN Taken in its entirety, the sk in is the largest organ in the human body. On average, it accounts for approximately 10 % of the total adult body weight and covers a total area of 1.5 to 2 m2 (Schaefer and Redelmeier, 1996). As the principle interface with our environment, the skin must be capable of simultaneously providing sensory input and functioning as a barrier to the transfer of mate rials into and out of the body. Specifically, the skin allows the body to effectively cont rol fluid loss, regula te body temperature, protect against physical trauma, defend agains t microbial infection and provide tactile sensation. To perform these functions, the skin contains numerous specialized cells and tissues arranged with a specific architecture. While modern imaging techniques have revealed that the skin is composed of a complex microanatomy, to the naked eye, the skin appears as three superimposed layers. From deepest to most shallow, these layers are known as the hypodermis, the dermis and the epidermis. While the skin contains a wide variety of cells and microstructures, many of them do not significantly affect the bioavailability of topically applied substances. In the following discussion, greater attention will be pa id to what defines the environment of the skin and to those elements that are mo st likely to impact topical absorption Hypodermis The hypodermis, or subcutaneous tissue, contains loose connective fibers and associated adipose tissue. This connective tissu e serves to anchor th e overlying layers of the skin to the body and the adipose tissue f unctions as a carbohydrat e source and thermal

PAGE 35

23 insulator. It varies greatly in thickness throughout the body and between individuals. It is usually thickest around the trunk of the body and thinnest on the backs of the hands and feet. The hypodermis has a less extensive vasculat ure than the outer laye rs of the skin but it does house the larger blood vessels that ulti mately feed the rest of the integument. Since the hypodermis does not contain the same specialized tissues as the other skin layers and since it does not activ ely participate in the regulatory functions of the skin, it is usually no longer classified as pr operly belonging to the skin. Dermis The dermis is responsible for the skinÂ’s st ructural integrity and, with the exception of controlling water loss, performs most of its critical functions. The tensile strength and flexibility of the dermis are the result of a tightly knit mesh of collagen and elastin fibers (Ushiki, 2002). This collagen lattice defines the dermis and provides a relatively fixed sca ffolding onto which the other structures of the dermis are anchored. Both collagen and elastin are s ynthesized from water-soluble precursors that are secreted into the intra cellular space by the most comm on cellular component of the dermis, the fibroblast. After leaving the fi broblast, these precurs ors undergo modification and are assembled into thin fibrils. In th e deeper region of the dermis (the reticular dermis), primary collagen fibrils are further assembled into thick, closely grouped, and extensively cross-linked strands that run roughly parallel to th e skin surface. Nearer to the epidermis (the papillary dermis), the strands become thinner and less heavily crosslinked to allow space for the proliferati on of capillaries and nerve endings. Surrounding the collagen lattice and filli ng most of the dermal intracellular space are extremely hydrophilic macromolecules known as proteoglycans. Proteoglycans are a group of fibroblast derived compounds that contain numerous unbranched polysaccharide

PAGE 36

24 chains covalently bound to a common polypeptide backbone (Iozzo, 1998). The polysaccharide side chains are com posed of several hundred repeating glycosaminoglycan disaccharides that have b een sulfated to a high degree. The large number of sulfate and carboxylate groups gives a strong anionic charge to the polysaccharide side chains and causes them to repel one another. In response to this repulsion, the side chains are forced to n early full extension from the peptide backbone and the molecule, in turn, resists compression. In addition, a la rge amount of water remains associated around the sulfate and car boxylate groups in hydration spheres. The ordering of water around the io nized proteoglycans gives the environment of the dermis properties similar to those of a hydrophilic gel. Small water-soluble molecules move with relative ease but highly lipophilic or high molecular weight molecules are impeded. A multitude of anatomical structures is found within the dermal matrix (Thibodeau and Patton, 2003). Corpuscular ne rve endings that ar e sensitive to vibration and pressure (Meissner and Pacinian corpus cles) are present as well as free nerve endings that are sensitive to pain and temperature. Smooth muscles cells control the dilation of blood vessels and follicular arrector pili caus e the familiar appearance of goose bumps. Resident immunologically active Mast ce lls and phagocytic macrophages provide a second line of defense against foreign mi croorganisms as do dermal dendrocytes and circulating t-lymphocytes. However, of the ma ny structures of the dermis, the three that have the highest potential impact on topical delivery are the hair follicles, the sweat glands and the capilla ry vasculature. Hair follicles and exocrine sweat glands are rooted in the reticular dermis, but they both have pores that exte nd to the skin surface. Thes e pores provide openings in and

PAGE 37

25 through the epidermis and its pe rmeation barrier. Bypassing this barrier allows free diffusion down the sweat duct or hair follicle and allows direct access to the dermis. However, this access is balanced by the physi cal dimensions of the pores. While there are several million follicles and exocrine gla nds distributed over the skin surface, the individual pore size is so small that the tota l cross sectional area of all pores is less then 0.1% of skin surface area (Schaefer a nd Redelmeier, 1996). Therefore, the poremediated pathway should only be important fo r compounds whose ability to penetrate the epidermal barrier is very low. The epidermis contains no vasculature of its own and must depend upon dermal blood vessels to supply essential compounds a nd remove waste material (Thibodeau and Patton, 2003). To facilitate this process, th e border between the de rmis and the epidermis is uneven with papillae from each layer interloc king with one another. Inside the dermal papillae are capillary plexuses from which nutrients and oxygen move by diffusion. The papillary structure of the border assures that epidermal cells surround each plexus. This arrangement increases the likelihood of derm al nutrients reaching epidermal cells and increases the efficiency with which ep idermal waste products reach the dermal capillaries. Similarly, this design increas es access to the capillaries for any exogenous compound that reaches the lowest level of the epidermis. With several thousand dermal plexuses per each square centimeter of skin, this route is the principle entry into systemic circulation for topical agents. In addition to being the most readily avai lable path to systemic circulation, the capillary plexus is also an effective sink. In a static system, the volume of blood contained in the vessels of an individual plexus is small and only a nominal mass of

PAGE 38

26 absorbed material is necessary to produce a high local concentration. However, in the living system, the blood flow through the pl exus prevents stagnant accumulation and dilutes absorbed compounds into systemic circulation. For poorly water-soluble molecules, serum proteins such as albumin should facilitate this process in a manner similar to what is observed with in vitro experiments (Cross et al, 2003). Many soluble proteins have non polar pocke ts into which lipophilic co mpounds can partition. Therefore, soluble proteins represent an a dditional phase that reduces the thermodynamic activity exhibited by blood borne non-polar mo lecules and, concurrently, increases the bloodÂ’s capacity to carry them. Epidermis Unlike the multi-functional dermis, the epid ermis is primarily designed to perform one function; provide a barrier that prevents the loss of water and essentially seals the skin against the entrance of foreign materials. The epidermal barrier covers over 99% of the body surface and it is this ba rrier that must be circumvented to effectively deliver drugs topically. Ironically, most of the body th e permeation barrier is contained in only the outermost 10%-20% of the epidermis (E lias and Friend, 1975). The formation of these few essential m of integument represen ts the final stage in a well-orchestrated transformation of viable epidermal cells into an inert and physiologically unique matrix. Keratin-producing epithelial cells, kera tinocytes, account for over 95% of the epidermal cell volume and are ultimately res ponsible for the forma tion of the barrier (Steinert et al, 1991). They are formed from re sident stem cells attached to the basement membrane, which is a specialized collagen structure that defines the lower limit of the epidermis and connects it to the dermis. When a keratinocyte stem cell divides, it produces a daughter stem cell, which remains attached to the basement membrane, and a

PAGE 39

27 transit-amplifying cell that begins to differentiate once it detaches from the membrane (Watt, 1989). As the daughter stem cells repe at this process, the formation and migration of new transit-amplifying cells forces older cell s closer to the skin surface. Since the rates with which both stem cells divide and transit cells mature are coordinated, keratinocytes at similar stages of development are found at r oughly equivalent depths of the epidermis. The difference in morphology between the vari ous stages of keratinocyte development has resulted in the epidermis being divided into the five visually distinct regions The first region, the stratum basale or stratum germinativum, is comprised of the epithelial stem cells and recen tly formed transit-amplifying cells. These cells appear columnar and have large nucleii. In the stratu m spinosum (the next 2 to 7 cell layers), the keratinocytes have lost the columnar organi zation of the basal cells and have begun to elongate. The name of this layer refe rs to the many surface protein complexes (desmosomes) attaching one ke ratinocyte to another, which give the cells serrated edges (Burge, 1994). The stratum granulosum contai ns the uppermost two or three viable cell layers and is the region in which keratinoc ytes make the most rapid and dramatic transformation. The cells flatten markedly and a simultaneous disintegration of the nucleus occurs. Stratum granulosum cells contain a large number of small granular deposits that give them a distinctive speck led appearance. The outermost layer, the stratum corneum, begins with the appearance of closely packed layers of corneocytes. Corneocytes are thin, proteinaceous and r oughly polygonal with a homogeneous internal structure. They remain essentially unch anged throughout the stratum corneum and are intact when released from the skin surface. Several in vitro and in vivo experiments, and the study of diseases that impair stratum corn eum formation, have consistently indicated

PAGE 40

28 that this layer provides most of the permation barrier (S cheuplein and Blank, 1971 and Lavrijsen et al 1993). On the palms of th e hands, soles of the f eet, and wherever the skin is subjected to a large amount of mechan ical stress, the stratum corneum can contain 30 or more corneocyte layers and be thicker than the viable epidermis. However, for most of the body, the stratum corneum is much thinner and is composed of 5-20 corneocyte layers. When the epidermis is studied on the submicron scale, it is clear that the formation of the stratum corneum is a process that begins several ce ll layers beneath the first discernable corneocytes in a light microsc ope image. At this scale, the subcellular granules, that are numerous in the stratum gra nulosum, can be resolved into two separate functional entities; keratohyalin granules and lamellar bodies. Keratohyalin granules (KG) produce the stru ctural proteins that are used in the assembly of the corneocytes. Some of these proteins (princip ally loricrin and involucrin) are used to create a cornified envelope that forms the corneocyteÂ’s outer surface. The envelope begins to form as a thin layer just inside the apical membranes of keratinocytes in the upper stratum spinosum but thickens rapidly throughout the straum granulosum as transaminase enzymes attach additional laye rs (Stevens et al, 1990). Other KG generated proteins, principally the kera tins K1 and K10, and the keratin aggregating protein filaggrin, are crosslinked acr oss the inside of the cornified envelope to form the corneocyte core (Eckert, 1989). When fully formed, the corneocyte has a sufficient number of internal protein filaments to re sist swelling upon contact with water and to resist permeation by other exogenous compounds.

PAGE 41

29 Lamellar bodies (LBs) are the second most prevalent granular organelle found in viable epidermal tissue. Like KGs, LBs are first observable in the upper layers of the stratum spinosum and they increase in numb er within the stratum granulosum. LBs process and store a complex mixture of lipid s and lipid-like materials which ultimately control the loss of water from the skin (Roberts et al, 1978). Along with free fatty acids and cholesterol, this lipid matrix cont ains a number of hydroxyl ated amide compounds known as ceramides. O O N H O OH OH N H O OH OH N H O OH OH OH O O N H O OH OH OH N H O OH OH OH N H O OH OH OH OH N H O OH OH OH OH N H O OH OH OH Ceramide 1 Ceramide 2 Ceramide 3 Ceramide 4 Ceramide 5 Ceramide 6 Ceramide 7 Ceramide 8 Figure 2-2. Structures of the principal cera mides comprising the lipid bilayers of the lamellar bodies.

PAGE 42

30 Electron micrographs reveal that the lipid components in LBs are often arranged in an ordered pattern that resembles compressed and stacked micelles (Fartasch et al, 1993). On the border between the stratum granluosum and the stratum corneum, LBs migrate to the keratinocyte membrane and their lipid components are expelled into the intercellular space where they are reformed into larger stack ed sheets. Concurrently, a single layer of lipid components, primarily ceramides 1 and 2, is bound covalently to the surface of the corneocyte (Wertz et al, 1989). In the fully formed stratum corneum, the interkeratinocyte lipid components produce a char acteristic electron mi crograph pattern of alternating light and dark bands that run parall el to the surfaces of th e keratinocytes. This pattern is generally thought to indicate that final arrangement of the lipid components in the stratum corneum is a nearly continuous seri es of stacked and plan ar lipid bilayers. It is important to note that the composition of the inter-corneocyte lipids is different from the lipid composition of the lamellar bodies. Phospholipids and glucosylated ceramides constitute a significant proportion of the lamellar body lipids but they are essentially absent in the stratum corneum. In fact, while it contains some ionizable components such as free fatty acids, amino acids and a small amount of cholesterol sulfate, the stratum corneum lipid matrix is predominantly composed of neutral molecules (Lampe et al, 1983). The principl e polar functional groups of the matrix are the ceramide hydroxyl and amide groups. They are capable of hydrogen bonding but there are only a small number of these groups per molecule. The polar functional groups of the ceramides do associate to form hydrophili c planes within the lipid lamella but these planes are thin compared to the alkyl regions. It is a reasonable speculation that removal of the highly acidic phosphate and the poly-hydr oxylated sugar moieties is necessary to

PAGE 43

31 limit the water permeability of the stratum corneum and to remove groups that would disrupt the cohesiveness of the lipid lamella. Studies of model systems suggest that the specific molecular arrangement or phase of the lipids in the lamellae influen ces permeability as much as the chemical composition (Xiang et al, 1998). As such, th ere has been a fair amount of effort expended in the attempt to experimentally determine the arrangement of lipids in the stratum corneum. X-ray diffraction, NMR, FT IR and electron microscopy have all been used to gather data from the stratum corn eum and compare it to the data from model systems, but the results have been difficu lt to interpret and sometimes contradictory (Hsueh et al, 2004 and Pilgram and Bouwstra, 20 04). Many variations of the bilayer lipid arrangement have been observed in the model systems, but they can generally be classified into one of three broad phases (S parr and Engstrom, 2004). Lipids in a solid crystalline phase (characterized by an orthorhom bic or triclinic packi ng) have the highest cohesive forces, the least freedom of movement and the lowest permeability. Lipids in a gel phase (characterized by a hexagonal packing) have grea ter rotational freedom and a higher permeability. Lipids in a liquid crystal phase lack a di screte arrangement and have the highest permeability. Many researchers have reported that the stratum corneum matrix contains or can adopt many phases depending upon environmental factors such as temperature, humidity or pH. Others have suggested that the high amount of cholesterol precludes the existence of the more ordered phases and that a single-phase model is more appropriate (Norln, 2001). Given the num ber of components contained in the lipid matrix and the changes in water content and temperature between the viable epidermis and the skin surface, it seems unlikely that a single phase can exist throughout the stratum

PAGE 44

32 corneum. For the purpose of predicting dermal delivery, such discussions, while interesting, are perhaps less germane. Ultimatel y, the most useful description of the skin will be a functional description derived from the skinÂ’s interaction with other compounds rather than a physical descrip tion based upon the skin alone.

PAGE 45

33 CHAPTER 3 DEVELOPMENT OF THE PREDICTIVE SOLUBILITY AND FLUX MODELS Derivation of the Series Specific Organ ic and Aqueous Solubility Equations To a large extent, a drugÂ’s solubility pr operties determine its transdermal flux and much of its other physiological behavior. Unfortunately, sinc e a compoundÂ’s solubility is impossible to predict a priori, the only certa in means of identifying those prodrugs with optimal solubilities is to synt hesize them. In keeping with this limitation, understanding the behavior of a given promoiety has tr aditionally required the synthesis of many compounds in which the parent drug is jo ined with many alkyl homologs of the promoiety. Comparing the be havior of two promoieties ha s only been done once data from several examples of each promoiety has been collected. This approach presents a potential difficulty in estimating the relati ve performance of the 4-AOC and 4-MOAOC APAP compounds given that only two members of the latter series were synthesized. With such limited data, the relevant question becomes can the behavior of a series be estimated from the performance of two, or even one, of its members? In other words, are there parameters or descriptors, common to each member of a series, which can be used as a means of comparison? The following discussion identifies these parameters and illustrates their use. The dissolution of a crysta lline nonelectrolyte is a complex process that is inaccessible to direct mathematical treatme nt. However, the free energy of this transformation can be described. Free energy is a state function. Therefore any path or combination of paths (actual or hypothetical) th at begins with the crystalline state and

PAGE 46

34 Dis Mix DG G G end with the solvated state can be used to calculate free energy for dissolution. With this in mind, it is convenient to separate the di ssolution process into two sequential events; the removal of a molecule of solid from th e crystal lattice (decry stallization) and the movement of this freed molecule into solution. Both of these events have free energies associated with them and the sum of these en ergies equals the free energy of the overall process: where GD is the free energy of decrystallization, GMix is the free energy of mixing and GDis is the free energy of dissolution. In a saturated solution, the crystalline form of the drug is in equilibrium with the solvated form and GDis is equal to 0. Each side of this equation can be ex panded using the Gibbs relationship: The enthalpy ( HD) and entropy ( SD) of decrystallization sp ecifically refers to the energy required to remove a molecule of the drug from the crystal lattice at the solution temperature (T). These are not readily obt ained quantities. However, the enthalpy ( HF) and entropy ( SF) of fusion can be can be used in thei r place if the heat capacities of the crystalline and molten forms of the drug are equal to one an other and reasonably constant over from the solution temperature to th e melting point. With this substitution, Mix Mix D DS T H S T H Mix Mix F FS T H S T H ) 1 3 ( ) 2 3 (Mix D G G 0 G Gx Mi D

PAGE 47

35 Since HF = TM SF (where TM is the melting point of the drug ), this can be rewritten as (3.3) According to ideal solution theory, SMIX is related to the mole fraction of the components in solution by the equation: (3.4) where n1 and x1 are the amount and mole fraction of the drug in solution and n2 and x2 are the amount and mole fraction of solvent. If there is no substantial loss of entropy upon mixing due to factors such as solvent ordering, then the ideal expression for SMIX can be substituted into equation 3.3. This expression can be further simplified if the solubility of the drug in the solvent is low (< 1%). In this case, the mole fract ion of the solvent is approximately equal to 1 and equation 3.5 reduces to where n and x now refer only to the amount and mole fraction of drug respectively. If the SF and HMix are presumed to be molar quantities, then the amount of drug in solution, n, disappears from the equation. To convert equation 3.7 into an expre ssion for solubility as amount per unit volume, a dilute solution assumption is again used: Mix Mix M FS T H T T S 2 2 1 1 Mixx ln R n x ln R n S 2 2 1 1 Mix M Fx ln R n x ln R n T H T T S x ln RTn H T T SMix M F x ln RTn H T T SMix M F ) 6 3 ( ) 5 3 ( x ln RT H T T RT SMix M F ) 7 3 (

PAGE 48

36 With this assumption, the volume of the solution (VS) is equal to the molar volume of the solvent (V2) multiplied by the amount of solvent (n2). Therefore: Substituting this into equation 3.7yields: If both sides of the equation are divided by 2.303, the equation is converted from the natural log to the more common base ten log: Equation 3.8 is an expression for solubility in what Hildebrand referred to as a regular solution; i.e. a solution that possesse s a non-zero enthalpy of mixing and a nearly ideal entropy of mixing (Hildebrand et al, 1970). This behavior is more likely to occur in organic solutions and it should be followed by saturated IPM solutions of our prodrugs. According to regular solution theory, the enth alpy of mixing arises from the breaking and reforming of intermolecular bonds that occur when the drug mo lecule enters solution. Typically, the strength of th ese intermolecular bonds is de termined by the presence of functional groups on the drug that possess unpa ired electrons, signi ficant dipole moments or polarizable electron clouds. In a homol ogous alkyl series, the number and type of 2 1 2 1 1n n n n n x 2 2 2 1 s 1V x V n n V n S 2V ln x ln S ln lnS T T RT S RT H lnVM F Mix 2 S T T RT S RT H V RTM F Mixlog 303 2 303 2 ln2 ) 8 3 (

PAGE 49

37 these interacting functional groups remain the same in each series member. Therefore, to a good approximation, HMix is a constant that is character istic to a given series. If the solubility of each series member is taken at the same temperature, then all quantities preceding the melting point term can be grouped into a single constant that is also characteristic to the series: WaldenÂ’s rule states that the entropy of fusion for most rigid small molecules is approximately equal to 56 J mol-1K-1. If the database compounds follow this rule, and if the temperature at which solubility is determ ined is the same for all compounds, then the terms of the melting point coefficient can be grouped into a second constant that should be independent of se ries and solvent. This is the general equation that relates melting point to solubility for homologs in a regular (organic) solution. An analogous equation for the aqueous sol ubility across a homologous series was derived using a method that parallels the deri vation of YalkowskyÂ’s general solubility equation (Yalkowsky and Valvani, 1980). Yalkowsky suggested that the partition coefficient was a simple means of transformi ng an organic solubility equation to an aqueous solubility if the part ition coefficient is assumed to approximately equal to the ratio of the drugÂ’s organic solubi lity to its aqueous solubility. S log T T RT 303 2 S AM F s O M O OS log T T B A ) 9 3 ( ) 10 3 ( AQ O AQ : OS S K

PAGE 50

38 Substituting equation 3.10 into equation 3.11 yields: The partition coefficients of alkyl homol ogs are related to one another through the equation: where log KO:AQ is the log of the partition coefficient for a given homolog, log K0 O:AQ is a series specific constant, and CMW is a constant (C) multiplied by the homologÂ’s molecular weight (MW) (Hansch and Leo, 1971). The left side of this equation can be used to replace log KO:AQ in equation 3-12. By combining AO and log K0 O:AQ into a new constant, AAQ, the final form of the general aqueous solubility equation is obtained. In the final form, the aqueous solubili ty equation is essentially the organic solubility equation with an a dditional correction term for mo lecular weight. This is a reasonable result given that the dissolution of a non-electrolyte in aqueous solution is similar in many aspects to itsÂ’ dissolution in an organic solution with one principle difference. Creating an aqueous cavity larg e enough to accommodate a solute molecule disrupts a number of hydrogen bonds between water molecules; an energy expenditure that has no counterpart in most organic solvents. The amount of energy required to open the cavity is proportional to the number of hydrogen bonds that must be broken in itsÂ’ AQ O AQ : OS log S log K log AQ : O O AQK log S log S log ) 11 3 (AQ : O M O O AQK log ) T T ( B A S log ) 12 3 (CMW logK logK0 :AQ O :AQ O CMW K log ) T T ( B A S log0 AQ : O M O O AQ CMW ) T T ( B A S logM O AQ AQ ) 14 3 ( ) 13 3 (

PAGE 51

39 formation, which, in turn, is proportional to the cavityÂ’s size. Ther efore, the amount of energy required by this process should be re lated directly to the size of the solute molecule and reasonably proportional to itsÂ’ molecular weight. In addition, since the only parameter that affects the number of hydr ogen bonds lost during solvation is the size of the solute, the molecular weight coeffi cient should be independent of the solute. The parameters that appear in the solub ility equations can be characterized as independent or universal (BO and C, respectively), compound specific (TM and MW), or series specific (AO or AAQ). Since there is only one series specific parameter appearing in each equation, all the coefficients for a give n series can be determined once data from any one series member is obtained. As the onl y series specific parameter, A is a measure of the intrinsic influence on solubility conveyed to the pr odrug by the promoiety. In other words, when two different prodrugs posses similar melting points and molecular weights, that compound belonging to the series with the higher A coefficient will have the higher solubility. Of course, to make a proper estimate of A, proper values for BO and C must first be determined. Derivation of the Potts-Guy Equation As with many fundamental flux models the Potts-Guy equation is developed from Fick's 1st Law of Diffusion. Fick discovered this law in the 1800's by observing the movement of dissolved compounds th rough permeable membranes separating compartments containing solutions of differing concentration (figure 3.1).

PAGE 52

40 L Donor Reservoir CDC1 C2 CR Receptor Reservoir Membrane Figure 3-1. Simplified diagram of an experimental diffusion apparatus. Using heat flow equations as a guide, Fick proposed that the flux per unit area (J) through each section of the membrane was proportional to the local concentration gradient: J = D C/L where D is the proportionality constant. In general, the differential C/L need not be a consta nt across the entire membrane. However, if the con centrations in each compartment (CD and CR respectively) are held constant for a sufficient amount of time, equilibrium concentrations, C1 and C2, are established just inside each exposed face of the membrane and the flux through the membrane reaches steady state. Under these conditions, a homogeneous membrane (or a membrane that behaves in a homogeneous manner) will have a constant concentration gradient across the membrane. For a linear concentration gradient, the partial differential will equal the difference in concentration just inside each ) 15 3 (

PAGE 53

41 face divided by the thickness of the memb rane. Substituting this into equation 3.15 yields: J = D (C1 – C2)/ L where L is the thickness of the membrane. For a given membrane, the highest flux (JM) is obtained when the difference between C1 and C2 is a maximum. This occurs when C1 is equal to the compound’s solubi lity in the membrane (SM) and when C2 is close to zero (sink conditions). Experimentally, these conditions are obtained by keeping a saturated solution of the test compound in the donor co mpartment while periodically replenishing the solution in the receptor compartment with clean solvent to keep its concentration under 10% of its solubility in the rece ptor phase. Under these conditions, (C1C2) = (SM 0) and equation 3.16 becomes: JM = DSM/L From a practical point of view, solubility in the skin is a difficult quantity to measure directly. However since the membra ne is in equilibrium with a saturated solution, SM should equal the solubility of the compound in donor compartment (SD) multiplied by the partition coefficient between the membrane and the donor solution (KM:D). Substituting this relationship into equation 3.17 yields: JM = D SD KM:D/L From equation 3.18, the maximum steady state flux of a compound is proportional to the solubility of that compound in the donor pha se. If flux is divided by this solubility, a new quantity, permeability (P), is obtained P = D KM:D/L ) 16 3 ( ) 17 3 ( ) 18 3 ( ) 19 3 (

PAGE 54

42 Potts and Guy recogni zed that each term in equation 3.19 could be estimated from readily available empirical data Since they were interested in delivery from water and the skin (or at least the stra tum corneum) is a lipid rich membrane, they asserted that the skin:water partition coefficient, KM:D, could be related to the octanol:water partition coefficient, KOCT, through the equation: log KM:D = f log KOCT + b Furthermore, previous work had suggested that the diffusion coefficient (D) should be related exponentially to the molecular vol ume of the compound (Cohen and Turnbull, 1959). Since molecular weight correlates well to molecular volume for organic molecules, it is convenient to substitute molecular weight into this equation: D = D0 e-MW log D = log D0 (1/ln 10)MW Taking the logarithmic form of equation 3. 18 and making substitutions using equations 3.20 and 3.21 yields the following result: log P = log KM:D + log D + log (1/L) log P = f log KOCT + (1/ln 10)MW + log (D0/L) + b The final terms of equation 3.22, log (D0/L) and b, can be collected into a single term for convenience. Similarly, the conversion fact or of 1/ln 10 can be combined with the term to create a single coefficient, Doing these two substitutions results in the standard form of the Potts-Guy equation. log P = f log KOCT MW + c ) 20 3 ( ) 21 3 ( ) 22 3 ( ) 23 3 (

PAGE 55

43 Derivation of the Roberts-Sloan Equation The Potts–Guy equation is useful in ce rtain circumstances, but it does have significant limitations. Flux, not permeability, is the most relevant clinical parameter. Since flux equals permeability multiplied by co ncentration in the donor phase, there is a tendency to equate increasing permeability w ith increasing flux. For a homologous series of prodrugs delivered from an aqueous vehicle, log KOCT and log P rise predictably and consistently with increasing alkyl chain length but flux does not Since the Potts-Guy equation is only concerned with permeability, it does not predict the important experimental observation that the most wate r-soluble members of homologous series tend to have the highest fluxes (Sloan and Wasdo, 2003). The use of KOCT to estimate KM:D is not applicable to vehicles that are miscible with octanol. This is an important limitation when studying prodrugs, since an aqueous donor is often incompatible with commonly employed labile derivatives. Roberts and Sl oan (1999) used the following mathematical manipulation to estimate a partitioning coefficient, KSKIN:IPM, between the skin (where SSKIN is solubility in the skin) and isopropyl myristate (IPM): KSKIN:IPM = SSKIN/SIPM KSKIN:IPM = (SSKIN/SAQ) / (SIPM/SAQ) KSKIN:IPM = KSKIN:AQ / KIPM:AQ log KSKIN:IPM = log KSKIN:AQ log KIPM:AQ KIPM:AQ can be used in the same manner as KOCT to estimate KSKIN:AQ, log KSKIN:AQ = f log KIPM:AQ + b log KSKIN:IPM = f log KIPM:AQ log KIPM:AQ + b log KSKIN:IPM = (f-1) log KIPM:AQ + b ) 24 3 ( ) 25 3 (

PAGE 56

44 For an IPM vehicle, (f-1) log KIPM:AQ + b takes the place of f log KOCT in the Potts-Guy equation: log P = (f-1) log KIPM:AQ + b MW + c log P = (f-1) log SIPM + (1-f) log SAQ MW + b + c Adding log SIPM to both sides transforms the equation to flux (JM) instead of P: log JM = f log SIPM + (1-f) log SAQ MW + b + c In the standard representation z, y and x are used respectively to replace f and the combination of b +c: log JM = x + y log SIPM + (1-y) log SAQ z MW This equation is the Robert s-Sloan (RS) model and it repr esents our primary method of predicting flux through hairless mouse skin from IPM. Modification of the Roberts-Sloan Equation to Include Synthetic Membrane Data We also wish to construct a predictive transdermal model that includes the compound's flux through a surroga te membrane as one of th e descriptive parameters. Although there are several possible forms for this model (depending principally upon the degree of similarity between the surrogate memb rane and skin), they are all derived from the general flux equation (eq. 3.16). The logarithmic form of this equation for flux through skin is shown below. log JS = log SS + log DS log LS Flux through an artificial membrane (X) should follow an analogous equation: log JX = log SX + log DX log LX To incorporate a log JX term into the skin equation, we must first decide which parameter in equation 3.29 best estimates its counterpart in equation 3.28; the solubility ) 26 3 ( ) 27 3 ( ) 28 3 ( ) 29 3 (

PAGE 57

45 term or the diffusivity term. Between a ny membrane and skin, one or both of the following relationships will be valid: log SS = AS log SX + BS log DS = AD log DX + BD Relationship 3.30 is expected to be va lid only in those situations where the surrogate membrane is very similar to skin itse lf. In general, this is expected only to be true when comparing skin to a similar biological membrane such as mammalian skin from another species. In contrast, equation 3. 31 is a more general relationship and should be more widely applicable. Specifically, equation 3.31 will hold for all membranes in which molecular weight and the diffusion coeffi cient are related as described in equation 3.21. From this we conclude that two mode ls should be investigated; one for comparing fluxes through related biological membrane s and one for comparing synthetic to biological membranes. To derive a model for chemically sim ilar membranes, both equations 3.30 and 3.31 are assumed to be valid. Equation 3.29 can be rewritten to show solubility in the membrane in terms of the other variables. log SX = log JX log DX + log LX This relationship can be subs tituted into equation 3.30: log SS = AS log SX + BS log SS = AS log JX AS log DX + AS log LX + BS This expression for log SS can now be substituted into equation 3.29: log JS = AS log JX AS log DX + AS log LX + BS + log DS log LS Substituting equation 3.31 into 3.32 removes log DS: ) 30 3 ( ) 31 3 ( ) 30 3 ( ) 32 3 (

PAGE 58

46 log JS = AS log JX AS log DX + AS log LX + BS + ADlog DX + BDlog LS log JS = AS log JX + (AD-AS) log DX + [BD + BS + AS log LX – log LS] All the terms appearing between the brackets can be collapsed into a single coefficient, K1: log JS = AS log JX + (AD-AS) log DX + K1 From equation 3.21, log DX = log D0 X – (1/ln 10)X MW. If this relationship is substituted into equation 3.32: log JS = AS log JX (AD-AS) (1/ln 10)X MW + (AD-AS) log D0 X + K1 If the constants are grouped re named in the following manner: (AD-AS) log D0 X + K1 = a AS = b (AD-AS) (1/ln 10)X = c the result is the general form of similar membrane model. log JS = a + b log JX + c MW To develop a model to relate the fl uxes through two dissimilar membranes, equation 3.29 is rewritten in terms of log DX and then substituted into equation 3.31: log DX = log JX log SX + log LX log DS = AD log DX + BD log DS = AD log JX AD log SX + AD log LX + BD log JS = AD log JX AD log SX + AD log LX + BD + log SS log LS log JS = AD log JX + log SS AD log SX + AD log LX log LS ) 33 3 ( ) 34 3 ( ) 35 3 ( ) 29 3 ( ) 31 3 (

PAGE 59

47 In order to proceed past equation 3.34, so me assumptions must be made about the log solubility terms. In the RS equation, log SS and log SX are estimated from log SIPM and log SAQ through the following equations: log SS = yS log SIPM + (1-yS) log SAQ log SX = yX log SIPM + (1-yX) log SAQ where yX and yS are the corresponding y coefficients for skin (S) and the synthetic membrane (X). If these two relationships ar e substituted into equation 3.35, the result is an expression for log JS in terms of experimentally determined values: log JS = AD log JX + (yS ADyX) logSIPM + (1-yS AD + ADyX ) logSAQ + ADlog LX -logLS This expression can be greatly simpli fied by grouping and renaming constants: ADlog LX -logLS = a AD = b yS ADyX = c log JS = a + b log JX + c log SIPM + (1 b c) log SAQ Equation 3.38 will only be predictive when the assumptions of the RS equation are valid. In other words, only when log KM:IPM is related to log SIPM and log SAQ in the following manner: log KM:IPM = (f-1) log SIPM (f-1) log SAQ + b The success of the RS equation in describi ng flux though hairless mouse skin from IPM indicates that the above expre ssion is reasonably followed for th at system. However, it is not yet known whether or not this can be appl ied as a general rule. When comparing two widely different systems, it may be necessary to allow unconstrained coefficients for log SIPM and log SAQ to estimate partitioning or solubility In this case, the coefficients in ) 36 3 ( ) 37 3 ( ) 38 3 (

PAGE 60

48 equation 25 would no longer be related and a more general form of model 3.38 is produced: log JS = a + b log JX + c log SIPM + d log SAQ It will require empirical data to determine wh ich of these two models is most applicable. ) 39 3 (

PAGE 61

49 CHAPTER 4 EXPERIMENTAL DESIGN Section I: Synthesis and Characteriza tion of the 4-Alkyloxycarbonyl and 4Methyloxyalkyloxy Prodrugs of Acetaminophen Synthesis Five homologous n-alkyl and tw o methyloxyalkyl carbonates of 4hydroxyacetanilide (APAP) have been prepared The general synthetic method for each carbonate entailed reacting APAP w ith equimolar amounts of the proper alkylchloroformate in the presen ce of a poorly nucleophilic base. A well-stirred suspension of APAP (~1.51 g, ~0.01 mol) was prepared in 30 mL of CH2Cl2 containing an equimolar amount of pyridine or triethylamine. To this mixture, a solution of the desired alkylchloro formate (0.01mol) in ~10 mL CH2Cl2 was added in a drop-wise manner, and the reaction mixture was allowed to react for two hours. The reaction solution was then d iluted to approximately 200 mL and extracted sequentially with 10 mL of ~0.6 N HCl and 10 mL deioni zed water. After the water wash, the CH2Cl2 solution was dried over Na2SO4 for two hours, filtered and concentrated under vacuum until solvent free. The resulting material was purified by recrystallization and, if necessary, column chromatography until a sh arp melting point was observed, only one component was discernable by TLC and a clean 1H-NMR was obtained. In this fashion, N H O OH N H O O O OR RO Cl O + Base

PAGE 62

50 the methyl, ethyl, propyl and butyl car bonates of APAP were prepared from commercially available alkylchloroformates. For the hexyl and the two methyloxyalkyl carbonates, it was necessary to first synthesize the desired alkylchloroformate. This was accomplished by reacting the proper primary alcohol with triphosgene, a synt hetic phosgene equivalent, and a poorly nucleophilic base (triethylamine or pyridine). One molecule of triphosgene rearranges to ultimately yield three molecules of phosgene during the reaction. Therefore a 3: 1 molar ratio of alcohol to triphosgene was used to maintain equimolar amounts of phosgene and the alcohol. A solution of triphosgene (~0.0033 mol) in 20 mL CH2Cl2 was first prepared. To this solution, a mixture of triethylamine or pyridine (~0.01 moL) and the alcohol (~0.01 mol) in 10 mL CH2Cl2 was added at a rate of ~1 mL/min in a drop-wise manner. The ensuing exothermic reaction was allowed to proceed until the solution had once again returned to room temperature. A suspension of APAP and triethylamine or pyridine (0.01 mol each) in 20 mL of CH2Cl2 was then added to the alkyl chloroformate solution at a rate of ~ 2 mL /min. After being allowed to react for at least two hours, the mixture was washed with aqueous acid and dried over Na2SO4 as described above. The crude product was purified by recrystallization and column ch romatography as necessary to achieve the previously stated criteria of purity. The re sults and specific conditions for each prodrugÂ’s synthesis are listed below. RO Cl O OCCl3 O Cl3CO ROH Base + 3 3

PAGE 63

51 2; 4-methyloxycarbonyloxyacetanilide – This compound was prepared in 49 % yield from methyl chloroformate and pyridine in CH2Cl2 after recrystallization from diethyl ether/hexane. 3; 4-ethyloxycarbonyloxyacetanilide – This compound was prepared in 82 % yield from ethyl chloroformate and triethylamine in CH2Cl2 after recrystallization from ethyl acetate/hexane. 4; 4 -propyloxycarbonyloxyacetanilide – This compound was prepared in 59% yield from propyl chloroformate and triethylamine in CH2Cl2 after recrystallization from ethyl acetate/hexane. 5; 4 -butyloxycarbonyloxyacetanilide – This compound was prepared in 63% yield from butyl chloroformate and triethylamine in CH2Cl2 after recrystallization from ethyl acetate/hexane. 6; 4-hexyloxycarbonyloxyacetanilide – This compound was prepared in 51% yield from hexanol, triphosgene and pyridine in CH2Cl2 after silica gel chro matography in ethyl acetate and recrystallization from ethyl acetate/hexane. 7; 4-(2’-methyloxyethylo xycarbonyloxy)acetanilide This compound was prepared in 44% yield from 2-methoxyethanol, triphosgene and triethylamine after recrystallization from ethyl acetate/hexane. 8; 4-(1’-methyl-2’-methyloxethy loxycarbonyloxy)acetanilide This compound was prepared in 29 % yield from 1-methyl-2-m ethoxyethanol, triphosgene and triethylamine after recrystallization fro m diethyl ether/hexane.

PAGE 64

52 Characterization Melting point determination and 1H-NMR analysis comprised the initial characterization of the 4-AOC and 4-MOAC APAP prodrugs. Melting points were determined using a Meltemp capillary melting point apparatus and were uncorrected. 90 MHz 1H-NMR spectra were obtained in CDCl3 using a Varian EM-390 spectrometer. In addition to the initial charac terization, molar absorptivities () in acetonitrile (ACN) and pH 7.1 phosphate buffer containi ng 0.11% formaldehyde were measured to facilitate quantitation in subsequent solubil ity and flux experiments. For each compound, three replicate stock solutions were prepared by diluting 10 1 mg portions of purified material to 25 mL in ACN. Aliquots of th ese stock solutions were further diluted with either ACN or buffer and the background-correct ed absorbance of thes e diluted solutions were measured from 400 to 200 nm using a Shimadzu UV 265 Spectrophotometer. Maximum absorbance was observed at 240 nm in both solvents for APAP and each prodrug. In addition, APAP displayed a pronounced shoulder at 280 nm in buffer that was not evident in the carbonate derivatives. The molar absorptivity at these wavelengths for each compound was determined by taking the average of the individual molar absorptivities of the replicate solutions. O N H C H3 O O O C H R2 R1 (A) (B) (C) (D) (E) Figure 4-1. Regions of the 4-AOCO-ACA prodrug corresponding to the letters given in table 4-1.

PAGE 65

53 Table 4-1. 1H NMR Data for the AOC and MOAOC-APAP prodrugs. 1H-NMR ()a Compound R1 R2 A B C D E 1 APAP -------------------2 H H s 2.15 d 7.06, d 7.45 s 3.95 ------3 -CH3 H s 2.12 d 7.03, d 7.42 q 4.28 t 1.47 ---4 -CH2CH3 H s 2.08 d 7.04, d 7.42 t 4.20 m 1.80, t 1.02 ---5 -(CH2)2CH3 H s 2.14 d 7.08, d 7.45 t 4.28 t 0.95 ---6 -(CH2)4CH3 H s 2.14 d 7.04, d 7.42 t 4.20 Unresolved m ---7b -CH2OCH3 H s 2.14 d 7.06, d 7.46 t 4.38 t 3.68, s 3.42 --8c -CH2OCH3 -CH3 s 2.08 d 7.04, d 7.42 m 5.00 d 3.50, s 3.40 d 1.34 a Obtained in CDCl3 with TMS as an internal standard b Elemental analysis for C12H15NO5, Found (Expected): C = 56.84 (56.91), H = 5.98 (5.97), N = 5.53 (5.53). c Elemental analysis for C13H18NO5, Found (Expected): C = 58.43 (58.42), H = 6.43 (6.41), N = 5.27 (5.24). Table 4-2. Melting point and absorptiv ity values for APAP, the AOC-APAP and MOAOC APAP prodrugs. Comp. Melting Point ( C ) 240 a in ACN 240 ain Buffe r b 280 a in Buffe r b 1 APAP 167-170 1.36 0.981 0.191 2 112-115 (115.5-116.5)c 1.67 1.25 0.0560 3 120-122 (121-122)c 1.64 1.24 0.0483 4 104-106 (105-108)d 1.63 1.28 0.0560 5 118-120 (119-121)c 1.75 1.24 0.0376 6 108-110 (112.5-113.5)c 1.79 1.12 0.0623 7 78-81 1.74 1.26 0.0623 8 120-123 1.60 1.28 0.0520 a units of M-1 x 104. b pH 7.1 phosphate buffer with 0.11 % formaldehyde. c Literature values in parentheses from Dittert et al, 1963 d Literature values in parentheses from Merck, 1897. Section II: Determination of IPM and Aqueous Solubilities IPM solubilities (SIPM) were determined directly by UV analysis on filtered IPM suspensions of prodrugs that were subseque ntly diluted with ACN. To prepare a suspension, a three mL volume of IPM was init ially combined in a 15 mL test tube with approximately twice the amount of APAP or pr odrug needed to just saturate the IPM.

PAGE 66

54 The resulting mixture was then stirred continuously at room temperature (231 C) for 72 hours. After stirring, the suspension was ra pidly filtered through a 0.25 m nylon syringe filter and an aliquot of the filtrate (~ 0.100 mL) was dilute d to at least 10.0 mL with ACN. The final solution was analyzed by UV to determine its absorbance at 240 nm. The molar absorptivity of each compound in th e diluted filtrate was assumed to be equivalent to itsÂ’ corresponding molar abso rptivity in pure ACN. This assumption allowed each compoundÂ’s IPM solubility to be determined from the relationship: SIPM = (Vfinal / Valiquot) A240/240 where Valiquot is the volume of the sa turated filtrate aliquot, Vfinal is the final diluted sample volume, A is the sample absorbance at 240 nm and 240 is the compoundÂ’s molar absorptivity in ACN at 240 nm. Three replic ate suspensions were prepared and analyzed for each compound. The average of the three values was reported as SIPM. Aqueous solubilities (SAQ) were estimated using two methods: a direct dissolution in water and a calculation us ing the compound's IPM:water pa rtition coefficient. For the direct measurements, suspensions of each compound in unbuffered deionized water were prepared in the same manner as the IPM suspen sions. However, to be consistent with the preparation of the suspensions used in th e diffusion cell experiments and to limit the extent of hydrolysis during analysis, a queous suspensions were stirred at room temperature for only one hour before filtrati on and dilution in ACN. As with the IPM solutions, the ACN diluted filtrates were analyzed by UV and the concentration of each compound was calculated using equation 4.1. To measure each compoundÂ’s IPM: water partition coefficient (KIPM:4.0), a measured volume of the prodrug suspension in IPM wa s placed in a 10 mL test tube along with ) 1 4 (

PAGE 67

55 four milliliters of 0.01 M acetate buffer (pH 4. 0). The test tube was capped, shaken vigorously for 10 seconds and allowed to stand until the two phases separated. An aliquot of the IPM layer was removed, diluted in ACN and analyzed by UV using the same protocol followed for the saturated IPM solutions. KIPM:4.0 was calculated from the concentration of prodrug remaining in the IPM phase after partitioning (CF), the initial saturated concentration (SIPM) and the ratio of IPM and buffer volume used in the partitioning. KIPM:4.0 = (VAQ/VIPM) CF / (SIPM-CF) Within reasonable accuracy, KIPM:AQ is equal to the ratio of SIPM to the prodrugÂ’s solubility in water (S4.0). Therefore, aqueous solubility is estimated from the relationship: S4.0 = SIPM / KIPM:AQ It is important to note that, while this process does not provide a rigorously measured solubility, it does provide a value that has been shown to be consistent with the directly measured solubility (Taylor and Sloa n, 1998). More importantly, it is a method that can be used to estimate the solubility of prodrugs that are too unstable to permit a direct solubility and, as such, can be used to compare compounds regardless of their intrinsic stability. Section III: Determination of Flux through Hairless Mouse Skin and Polydimethylsiloxane Membranes Preparation of the Membranes and Assembly of the Diffusion cells Adult female hairless mice were rendered unconscious by CO2 and quickly sacrificed by cervical dislocation. Whole thickness skin from the entire region was immediately removed from each mouse by blunt di ssection. Sections of this separated skin were cut to a proper size and immediat ely mounted on the diffusion cell. For the ) 2 4 ( ) 3 4 (

PAGE 68

56 PDMS membrane, properly sized and shaped se ctions were trimmed from larger sheets using the diffusion cell donor compartment as a guide. Just prior to mounting, the trimmed sections were rinsed with water and MeOH and then blotted dry to remove accumulated dust. Shown in vertical profile by the figure belo w, a Franz cell consists of two separate compartments. The upper or donor compartment (A) is essentially a glass cylinder with a flared and grooved lower edge. The lower or receptor compartment (B) is cylindrical reservoir with a grooved upper edge that mirro rs the donor compartment. In addition, the receptor reservoir is equipped with a temperat ure controlling water jacket and a side arm that allows filling and ac cess to the receptor fluid. To assemble the cell for analysis, the donor side compartment (A) was inverted and a section of membrane was placed over the open ing. Hairless mouse skin sections were placed with the epidermal side facing the donor compartment and gently stretched into place until they completely covered the enti re lower opening and edge of the donor compartment without sagging. PDMS membra ne sections were placed over the donor compartment opening without further adjustme nt. Once the membrane section was in position, a rubber o-ring was placed over it and aligned with the groove on the donor cylinder. The receptor section (B) was then i nverted and carefully aligned with the donor section. The two sections were then clampe d together with a screw locked spring clamp and the assembled cell was returned to an uprig ht position. After assembly, the receptor compartment was completely filled with 0.5 M, pH 7.1 phosphate buffer containing 0.1% formaldehyde by weight as an anti-microbial agent. This concentration of formaldehyde i

PAGE 69

57 Figure 4-2. Fully assembled diffusion cell. is essential for maintaining membrane integrity and will prevent membrane degradation for more then 144 hours (Sloan et al, 1991). Af ter filling the receptor compartment, care was taken to ensure that no air bubbles remained adhered to the bottom of the membrane and the fluid level in the side arm was adjust ed to the same height as the membrane to prevent any increased hydrosta tic pressure. A magnetic st ir bar was added through the side arm and the cell suspended over a magne tic stir plate to continuously stir the receptor buffer throughout the experiment. Th e water circulating through the insulating jacket was set to 32 C. The cells were ke pt in contact with the receptor fluid for 48 hours prior to the application of the donor phase to leech out any UV active components and to allow the membranes to equilibrate wi th the receptor phase. Twice during this 48hour conditioning period, the re ceptor phase was completely withdrawn and replaced with fresh receptor buffer.

PAGE 70

58 Preparation and Application of Donor Phases IPM and aqueous suspensions of each co mpound were prepared. IPM suspensions were prepared by stirring 0.6 to 1.0 mmol of test compound with 3 mL of IPM for 24 hours at room temperature. For each co mpound, the final suspension concentration exceeded the compoundÂ’s solubility in IPM by at least threefold. For aqueous suspensions, similar mmol quantities of materi al were stirred with 4 mL of deionized water, but the suspensions we re stirred for only one hour pr ior to application to limit hydrolysis of the prodrugs. In addition, ne w aqueous suspensions were prepared every 24 hours. For either vehicle, application of the donor suspensions occurred immediately after preparation. Just prior to application of the donor suspension, the entire receptor phase was removed and replaced with fresh buffer solution. Either a 0.5 mL a liquot of well-stirred IPM suspension or a 1.0 mL aliquot of we ll-stirred aqueous suspension was evenly applied to the conditioned membrane surface as shown in figure 4-2. When working with the aqueous suspensions, the donor comp artments were covered with parafilm to prevent excessive loss of water to evaporation. After remova l from the diffusion cells, all prodrug donor suspensions were analyzed by 1H-NMR to ensure the prodrug had not hydrolyzed in the vehicle. Sampling of the Difussion cells fo r Flux and Residual Skin Samples To obtain a sample of the receptor phase, a 3 to 4 mL aliquot of receptor buffer was removed from the sidearm by Pasteur pipette and placed in a test tube for subsequent quantitation. To maintain sink conditions during the experiment, the remaining receptor fluid was removed after each sampling and the entire receptor compartment refilled with fresh buffer. The frequency with which samples were taken was determined by the rate

PAGE 71

59 at which analyte accumulated in the recep tor phase and it differed for each membranevehicle combination. The table below ou tlines the sampling times for APAP and each prodrug. Table 4-3: Sampling times for APAP a nd the AOCA prodrug diffusion experiments Compound Membrane/Vehicle Sampling Timesa APAP HMS / IPM8.5, 19, 22, 25, 28, 31, 34, HMS / Water 11, 24, 36, 52 PMDS / Water 24, 36, 48 C1 AOCA HMS / IPM 8, 19, 22, 25, 28, 31, 34, 43, HMS / Water 10, 24, 27, 30, 33, 36, 39, 48 PMDS / Water 24, 36, 48 C2 AOCA HMS / IPM 8, 19, 22, 25, 28, 31, 34, 44, HMS / Water 14.5, 24, 27, 30, 33, 36, 39, PMDS / Water 24, 30, 36, 48 C3 AOCA HMS / IPM 8, 19, 22, 25, 28, 31, 34, 43, HMS / Water 14.5, 24, 27, 30, 33, 36, 39, PMDS / Water 24, 30, 36, 48 C4 AOCA HMS / IPM 7, 20, 22, 25, 28, 31, 34, 43, HMS / Water 10, 24, 27, 30, 33, 36, 39, 48 PMDS / Water 24, 36, 48 C6 AOCA HMS / IPM 7, 20, 22, 25, 28, 31, 34, 43, HMS / Water 11, 24, 36, 52 PMDS / Water 24, 36, 48 C2 MOACA HMS / IPM 8.5, 19, 22, 25, 28, 31, 34, HMS / Water 12, 24, 27, 30, 33, 36, 39, 48 PMDS / Water 24, 36, 48 Ci3 MOACA HMS / IPM 8, 19, 22, 25, 28, 31, 34, 44, HMS / Water 12, 24, 36, 48 PMDS / Water 24, 36, 48 a Time in hours after initial application of donor suspension Receptor phase samples were analyzed by UV analysis within an hour of collection. Just prior to analysis, a background correction from 350 to 200 nm was performed on the instrument using matche d quartz cuvettes containing blank receptor buffer. After this correction, the spectrum of each recently collected receptor sample was taken over the same wavelength range.

PAGE 72

60 Following the last sampling and refilling of the receptor compartment, the cells were allowed to sit for 24 hours. For IPM treated hairless mouse skin, this period was sufficient to leach any test compound rema ining within the skin. As such, the concentration of compound found in the recep tor phase after these 24 hours was used to determine the total amount of material that had been absorbed by the skin (the residual skin amount). For compounds delivered fr om water through hairless mouse skin and PDMS membrane, an additional 24-hour leechi ng period was necessary and the residual skin amount was determined by totaling the amou nt of material leeched after the first and second 24 hour periods. Evaluation of Membrane Integrity After the leaching period, a standard suspension of theophylline in propylene glycol (400 mg/mL) was applied to the membra nes to determine their general integrity. Following the final residual membrane sample, the remaining receptor buffer was removed, replaced with fresh solution a nd a 0.5 mL aliquot of the theophylline suspension was applied evenly to the memb rane surface. Samples of the receptor compartment were taken at regular interval s using the same protocol as for a test compound. A sufficient number of samples were taken over the subsequent 24 hours to determine the flux of theophylline through the membrane. Determination of Analyte Concentr ation and Extent of Hydrolysis For any wavelength, the absorbance of each sample was assumed to result solely from a combination of the absorbances of AP AP and any intact pr odrug. Using BeerÂ’s law, this assumption can be stated mathematically as: A = CPP + CDD ) 4 4 (

PAGE 73

61 A represents the absorbance at wavelength CP and CD are the respective concentrations of the prodrug and drug, and P and D are the respective molar absorptivities for the prodrug and APAP, respectively, at wavelength By measuring the absorbance of a sample at two wavelengths (240 nm and 280 nm for APAP and its prodrugs), it was possible to determine the values of CP and CD by simultaneously solving the resulting two BeerÂ’s Law equations. A280 = CPP280 + CDD280 A240 = CPP240 + CDD240 CP = (A240D280 A280D240)/(P240D280 P280D240) CD = (A240 CPP240) / D240 CP and CD were added to determine the total concentration of APAP species present in the sample. The total amount of APAP species present was then determined by multiplying total concentration by the volume of the receptor phase. Determination of Maximum Flux (JM) Maximum flux was determined from a pl ot of the cumulative amount of APAP species delivered through the membrane versus time. Since the receptor fluid was replaced after every sampling with clean buffer, the amount of compound found in each sample was indicative of only that amount which had passed through the membrane between sampling times. As such, the cumulative amount delivered was calculated by sequentially adding these sampled amounts. A typical graph of cumulative amount plotted against sample time is shown below. ) 5 4 ( ) 6 4 (

PAGE 74

62 Figure 4-3. Plot of cumulati ve amount of prodrug delivered versus time for compound 3 from IPM through hairless mouse skin. Linear regression was performed on data in the steady-state region of the plot (usually data collected after 24 hours). The slope of this best-fit line was divided by the cross sectional area of the diffusion cell to yield the compoundÂ’s maximum flux, JM. y = 1.7012x 26.715 R2 = 0.9995 0 10 20 30 40 50 60 70 80 010203040506070 Time (h)Total Flux (moles)

PAGE 75

63 CHAPTER 5 RESULTS Melting Point Behavior of the APAP Carbonates As noted in chapter 1, nonionic com pounds, in general, display a negative correlation between melting point and solubi lity. As such, compounds with lower melting points tend to have higher solubiliti es in lipid and water which favor higher transdermal flux. It was also noted that the relatively high melti ng points of APAP and of other phenol containing compounds are like ly a consequence of their ability to form strong intermolecular hydrogen bonds. It wa s anticipated that masking the phenolic moiety of APAP would disrupt this ability and produce derivatives with consistently lower melting points. Table 5.1 shows that this is indeed the case. All of the carbonate prodrugs have melting points that are substa ntially lower then that of APAP. In particular, the two highest melting prodrugs (3 and 8) melt 50C lower then APAP while the lowest melting compound (7) melts approximately 90C lower. Table 5-1: Melting poi nts (C), log partition coefficients (log KIPM:AQ), log solubility ratio (log SRIPM:AQ), log solubilities in IPM (log SIPM ) and log solubilities in water (log SAQ) for APAP and the 4-AOC and 4-MOAOC prodrugs. Compound Melting Point (C) log SIPM a Direct log SAQ a,b log KIPM:4.0 b Estimated log S4.0 a,b log SRIPM:AQ APAP, 1 167-170 0.2782.000-0.174 2 112-115 1.076 1.314 -0.156 1.232 -0.238 3 120-122 0.968 0.578 0.315 0.653 0.390 4 104-106 1.374 0.427 0.897 0.477 0.947 5 118-120 1.143 -0.372 1.497 -0.354 1.515 6 108-110 1.219 -1.324 2.706 -1.487 2.543 7 78-81 1.014 1.536 -0.300 1.314 -0.522 8 120-123 0.529 0.516 0.132 0.386 0.013 a In units of mM at 23C b Using pH 4.0 buffer at 23C.

PAGE 76

64 However, the drop in melting point does not follow any simple trend among the prodrugs. The 4-AOC-APAP compounds demonstr ate an odd-even alteration in their melting points as the number of methylene gr oups in the promoiety increases. This effect, where the melting points of homologs will alternately rise and fall with the addition of each methylene group to the alkyl chain of the promoiety, has been observed in several series of homologous series (Chikos and Nichols, 2001), but has not been seen in the compounds of the 5-FU and 6-MP series (Roberts and Sloan, 1999). Since the addition of each CH2 has a proportionally larger effect upon the entire molecule when the promoiety is small, this behavior is generally more pronounced with the first few members of a series. In th e 4-AOC-APAP series, there is an average alternating fluctuation of 12 C between the fi rst four members. Perhaps more surprising, the addition of a single methyl side chai n to the ethoxymethyl pr omoiety of compound 7 resulted in a 40C increase in melting point and produced the highest melting prodrug of either type (i.e. 8). Such erratic behavior highlights the difficulty in predicting melting points a priori without empirical evidence. Direct IPM and Aqueous Solubility With the exception of compound 7, the melting points of the APAP carbonates fall within a relatively small range (~20C). Given the relationship between melting point and IPM solubility, a correspondingly small ra nge in IPM solubility was expected and consequently observed. The most IPM soluble derivative (4) is only eight times more soluble than the leas t soluble compound (8). However, all of the prodrugs were more soluble in IPM then APAP itself. Compound 2 showed a 6-fold improvement in IPM solubility over APAP and even compound 8 was 1.8 times as soluble. This range of solubility is similar to that of the 3-AC 5FU series and the maxi mum IPM solubility of

PAGE 77

65 this series ranks as forth among the nine homologous series so far studied (Sloan and Wasdo, 2003). Unlike the IPM solubility values, the di rect aqueous solubilities of compounds 2 through 8 were substantially less than that of APAP. As previously noted, all of the APAP prodrugs possessed lower melting points than APAP, which indicates a decrease in the lattice energy of these compounds rela tive to APAP. This lower lattice energy is most likely the result of masking the phenolic moiety. However, masking this moiety in the prodrugs also removes some of their abil ity to form beneficial hydrogen bonds when in aqueous solution. It is a pparent from the consistent de crease in aqueous solubility exhibited by the APAP prodrugs that the loss of hydrogen bon d stabilization in solution caused by masking the phenol, combined with th e extra energy required to create a larger cavity in solution to accommodate their larger molecular size, outweighs the benefits of lower lattice energy for these compounds. In terms of relative solubility, compound 2 had an aqueous solubility of approximately 21% that of APAP, whereas compound 6 possessed only 0.04% of APAPÂ’s water solubility. Compound 7 had highest water solub ility of all prodrugs and was 34% as soluble in water as APAP, wh ile the other MOAOC derivative, compound 8, had a water solubility of approximately 3% th at of APAP. When compared to the entire database (Sloan and Wasdo, 2003), compounds 2 and 7 have water solubilities greater than two thirds of the char acterized compounds. However, five out of the other nine series have at least one member with wate r solubility as good or better then either compound 2 or 7. The remaining APAP produgs rank only within the lo wer 50% of the databaseÂ’s values. Therefore, the performa nce of these prodrugs is somewhat mediocre.

PAGE 78

66 Partition Coefficients and Solubility Ratios As discussed in chapter 3, there is a linea r relationship in a homologous alkyl series between the log of the partition coefficient and the molecular weight of the homolog (eq. 3-13). Since molecular weight increases pr oportionally with the number of methylene units in the promoiety, a plot of log K versus methylene num ber will also yield a linear relationship. The slope of this line, the value, should be independent of the seriesÂ’ parent molecule and, therefore, should be the same for all homologous series. Empirically, this prediction has been supported. For the series comprising our database, values are very consistent with an averag e value of 0.58 and a st andard deviation of 0.03. This consistency makes the value a robust indicator of homologous series behavior. Linear regression on compounds 2 through 6 yields a value of 0.58, which closely agrees with the database av erage (Sloan and Wasdo, 2003). This is evidence that the physical properties of this series are both internally consistent and behaving in a theoretically predicted manner. Since the ot her series have displayed similar properties and since this consistency is fundamental in determining comparable estimated water solubility (S4.0), we can anticipate that the estimate d water solubility for these compounds should be compatible with the ot her estimated database values. The two MOAOC derivatives differ by th e addition of a methyl group as a side chain rather then the insertion of a CH2 group into the existing alkyl chain. However, due to its similar size, chemical composition and polarity, the addition of the methyl group should have a similar effect upon partitioning as a methylene group. If the basic theory of partitioning holds true, the slightly larger size of the methyl group should have resulted

PAGE 79

67 in larger drop in log K between compounds 7 and 8 than what was measured for the AOC-APAP compounds. However, this was not the case. Using the difference in log KIPM:4.0 as an approximation, a value of 0.44 was obtained. This is significantly smaller then expected and only 75 % of the value for the AOC series. To develop a possible explanation for these results, we must first ex amine the behavior of the actual IPM:water solubility ratios (SRIPM:AQ = SIPM / directly measured SAQ) as a point of comparison. By design, many members of the current database are unstable to hydrolysis even at neutral pH. In contrast, the APAP car bonate prodrugs are much more stable and afforded an opportunity to directly meas ure aqueous solubility and subsequently determine SRIPM:AQ. If log SRIPM:AQ is plotted against methylene number instead of log KIPM:4.0, a value of 0.55 is obtained for the AOC co mpounds. As predicted, this is in close agreement with the value obtained from the log KIPM:AQ plot. The agreement is so close that a difference of less then 0.1 log units was measured between estimated and direct solubility for compounds 2 through 5. Even for compound 6, whose low solubility values have the highest associated uncertain ty, a discrepancy of le ss then 0.2 log units was measured. For the 4-MOAOC-APAP deriva tives, the change in log SRIPM:AQ between compound 7 and 8 was 0.53. While still lower then e xpected, this is 0.1 log units larger then the corresponding change in log KIPM:4.0 and closer to the log KIPM:4.0 value for the AOC-APAP prodrugs. There is too little experimental evid ence to ascertain the reason for this difference, however the higher log SR relative to log K does suggest that unexpected partitioning behavior is responsible rather then an intrin sic attribute of the MOAOC promoiety.

PAGE 80

68 Permeability Coefficient Behavior The permeability coefficient is the most of ten used parameter appearing in models of topical delivery. As discussed in chapter 3, many authors have repo rted that the log of the permeability coefficient correlates positiv ely to the log of oc tanol:water partition coefficient (log KOW) and this relationship was used by Potts and Guy (1992) in the development of their permeability model. log P = f log KOCT + MW + c It is this positive association with log KOW that has led in part to the misleading principle that greater lipophilicty is absolutely beneficial to topical delivery. Table 5-2: Log permeability values for the APAP prodrugs from IPM through hairless mouse skin (log PMIPM), from water through hair less mouse skin (log PMAQ) and from water through PDMS membrane (log PPAQ). Compound log PMIPM a log PMAQ a log PPAQ a 2 -1.08-2.69-2.74 3 -1.73 -2.27 -2.39 4 -1.82 -2.04 -1.92 5 -2.15 -1.82 -1.43 6 -2.71 -0.79 -0.68 7 -1.12 -2.77 -3.16 8 -1.59 -2.77 -2.79 a In units of cm h-1. For the AOC-APAP prodrugs, log PMAQ increases linearly with increasing size of the promoeity in a manner similar to log KIPM:4.0 and if log KIPM:AQ is plotted against log PMAQ, the following best-fit linear relationship is found: log PMAQ = 0.652 log KIPM:4.0 –2.64 (r2 = 0.963) (5.1) A plot of log PPAQ versus log KIPM:AQ yields a similar correlation. log PPAQ = 0.805 log KIPM:4.0 –2.45 (r2 = 0.974) (5.2) ) 23 3 (

PAGE 81

69 -3.5 -3 -2.5 -2 -1.5 -1 -0.5 -0.500.511.522.53 log KAQ:IPMlog P Mouse Skin/ IPM Mouse Skin/ Water PDMS/ Water Figure 5-1. Plots of log KIPM:4.0 versus log P for compounds 2 through 8 using flux data from the three membrane /vehicle systems. As expected, the permeability coefficient ha s a positive correlation to the IPM-water partition coefficient for both membranes. Howe ver, a quite different picture emerges if log PMIPM is plotted against log KIPM:AQ: log PMIPM = -0.517 KIPM:AQ – 1.36 (r2 = 0.940) (5.3) When the delivery vehicle is changed from water to IPM, the correlation between permeability and partition coefficient becomes negative. These are unusual findings for they suggest that mouse skin behaves as a lipophilic membrane for one system and as a hydrophilic membrane in another. One can rec oncile this behavior by realizing that IPM and water lie at opposing ends of the polarity spectrum. Therefore, the data more accurately shows that, relative to water, mous e skin is a lipophilic membrane and, relative to IPM, mouse skin is a hydrophilic membrane.

PAGE 82

70 Conversion to the Parent Drug As mentioned earlier, simple carbonates undergo chemical hydrolysis at a much slower rate then do simple esters and are e xpected to be relatively stable in neutral aqueous solutions. Dittert et al (196 3) studied the hydrol ysis of compounds 2, 3, 5 and 6 in pH 7.4 buffer and found them to have half-lives in excess of 150 hours. Hydrolysis studies were not performed on the no vel members of the APAP series (4, 7 and 8), but aqueous donor phases from all diffusion cell experiments were collected after use, allowed to evaporate overnight a nd were subsequently analyzed by 1H-NMR and melting point. Even after more then 48 hours of exposure to an aqueous environment and recrystallization from water, no evidence of hydrolysis was detectable for any member of the series. This data indicates that chemi cal hydrolysis would not be responsible for a significant release of parent drug dur ing the diffusion cell experiments. By comparison to chemical hydrolysis, carbonates are far more susceptible to enzymatic hydrolysis. When Dittert a nd Swintoski (1968) exposed carbonates 2, 3, 5 and 6 to a 2% solution of human plasma, the most stable compound, 2, had a half life of three hours; a 50 fold increase in the rate of hydrolys is over that in buffer. They also found the rate of enzymatic hydrolysis increased with the size of the alkyl side chain. By compound 6 (C6), the half-life in human plasma had fallen to only 11 minutes compared to its half-life in buffer of 22, 800 minutes; a 200 fold increase. The percentage of intact prodrug appear ing in the receptor phase the carbonate series is consistent with these findings a nd indicative of enzymatic conversion back to APAP, especially for the IPM data. Compounds 5 and 6, the most enzymatically labile derivatives, were completely hydrolyzed duri ng the experiment. In contrast, the most enzymatically stable compound, 2, showed the highest percentage of intact prodrug when

PAGE 83

71 delivered from IPM. There is also a gene ral correlation between higher flux and a higher percentage of intact prodrug. Two mechanisms could expl ain this. A higher flux may result in a shorter residence time in the me mbrane and reduce the chance for interaction between a prodrug molecule and hydrolytic enzy mes in the skin. Alternatively, higher flux also correlates with a higher concentr ation of prodrug in the skin, which could overwhelm the enzyme system and allow a highe r percentage of intact molecules through the membrane. Regardless, the best evidence for the enzymatic role in hydrolysis comes from the high percentage of intact prodrug a ppearing in the receptor phase of the enzymefree polymer membrane system. With the exception of compound 2, which was 70% hydrolyzed, greater then 90% of each alkyl carbonate was recovered intact from the receptor phase after passing through the PDMS membrane. Flux of APAP and its AOC and MOAOC Prodrugs through Hairless Mouse Skin The maximum flux, JM, of each compound through hair less mouse skin and PDMS membrane is presented in table 5-3. Th e flux through hairless mouse skin from a saturated solution of IPM is designated JMIPM, the flux through hairless mouse skin from a saturated aqueous solution is designated JMAQ, and the flux through PDMS membrane from a saturated aqueous solution is designated JPAQ. For convenience, these values are shown in their logarithmic form. The table al so contains the second application flux of theophylline (JJ) corresponding to each compound and experimental condition. When delivered from IPM, APAP permeates mouse skin more rapidly than 5-FU, 6-MP or theophylline. It also demonstrat es a maximum flux through mouse skin higher than half the prodrugs in the IPM database. In contrast, the car bonate derivatives of APAP perform worse than th is on average. Only two of the carbonate prodrugs, 2 and

PAGE 84

72 Table 5-3. Maximum steady-state flux and second application fl ux of the AOC-APAP and MOAOC-APAP prodrugs through hairless mouse skin and PDMS membrane. Compound log JMIPM a JJMIPM a log JMAQ aJJMAQ a log JPAQ a JJPAQ a APAP, 1 -0.29 0.74 -1.73 0.015 -2.68 0.0013 2 -0.00 1.12 -1.46 0.034 -1.51 0.0013 3 -0.76 0.64 -1.62 0.078 -1.74 0.0017 4 -0.45 1.14 -1.57 0.072 -1.44 0.0018 5 -1.01 0.85 -2.17 0.051 -1.79 0.0013 6 -1.49 0.76 -2.28 0.018 -2.16 0.0017 7 -0.11 0.98 -1.45 0.033 -1.85 NA 8 -1.06 0.94 -2.38 0.022 -2.41 NA aIn units of mol cm-2 h-1 -3 -2 -1 0 1 2 3 123456789Compound log Parameter log SIPM log SAQ log JMIPM log JMAQ log JPAQ Figure 5-2. Correlation between solubility and flux for APAP, itsÂ’ AOC derivatives and itsÂ’ MOAOC derivatives. 7, give maximum flux values from IPM that are higher then APAP itself. Compound 2 delivered 1.95 times the amount of APAP th rough mouse skin as the parent drug and compound 7 delivered 1.5 times as much APAP as the parent. Of the remaining series members, compound 4 performed the best (maximum flux 70% that of APAP) and compound 6 performed the worst (maximum flux 6% that of APAP).

PAGE 85

73 From water, the flux through mouse skin of all compounds is substantially reduced. APAP does not perform quite as well from wa ter and has a lower aqueous flux than four of the carbonate prodrugs. As with flux from IPM, compounds 2 and 7 deliver the highest amounts of APAP, but compound 7 out performs compound 2 when delivered from water (1.90 and 1.68 times APAP flux respectively). Compounds 3 and 4 produced slightly higher aqueous flux values than AP AP but, with flux values of only 1.16 and 1.09 times that of APAP, these diffe rences lack significance. Compounds 6 and 8 are the worst performing carbonates with compound 8 giving the lowest aqueous flux. It is worth noting, that while the relative performance among APAP the AOC compounds and the MOAOC compounds changes depending upon the nature of the vehicle, the order of performance for each type remains nearly the same. The fluxes of the APAP prodrugs thr ough PDMS membrane from water were similar in magnitude to their fluxes from wa ter through hairless mouse skin, but the order of their performance was not th e same. Through PDMS, compound 4 was the best performing prodrug and it possessed a flux similar to that of compound 2. This is a significant change from the flux through mouse skin where the flux of compound 2 was triple that of compound 4 when delivered from IPM a nd 50 % higher when delivered from water. Compound 5, the third most lipophilic prodr ug, has a PDMS flux nearly equal to that of compound 3 and a PDMS flux higher then that of compound 7. In contrast, compound 5 has only one third the flux of compound 3 and one fifth the flux of compound 5 through mouse skin from the same vehicle. An equally profound change is in the performance of the parent APAP. AP AP has the third highest flux through hairless mouse skin when delivered from IPM and fift h highest when delivered from water, but it has the lowest flux through PDMS.

PAGE 86

74 Even though the stratum corneum is widely considered to be primarily a lipophilic membrane, IPM solubility alone does not corr elate well with flux through mouse skin for either vehicle. The most IPM soluble prodrug, 5, has only the third highest flux from both vehicles and the second most lipid soluble compound, 6, has the lowest flux from both vehicles. In contrast, the most water soluble prodrug, 2, has the highest flux from IPM and the second most water soluble prodrug, 7, has the highest flux from water. Comparing compounds 3 and 4, which have nearly equal IP M solubility, a decrease in water solubility correlates to a drop in fl ux. Conversely, when aqueous solubility remains nearly constant and IPM solubili ty increases (such as between compounds 4 and 5) flux also rises. Even without a strict mathematical treatmen t of this data, it is apparent that flux through mouse skin is benefite d by a increasing both lipid and aqueous solubility, regardless of the vehicle used. The situation is very different for the purely lipophillic PDMS membrane. Although the mouse skin and PDMS flux plots shown in figure 5-3 appear similar in many respects, flux through PDMS shows a much greater depende nce on lipophilicity and a decreased dependence on hydrophilicity than flux through mouse skin. As stated previously, the most hydrophilic prodrugs, 2 and 7, have the best flux through mouse skin from either vehicle whereas the most lipophilic compound, 4, has only the fourth highest flux. Through PDMS, however, it is compound 4, which has the highest flux and APAP, the compound with the lowest IPM solubil ity, which has the lowest flux. Compounds 5 and 7 possess similar molecular weights and similar IPM solubilities but differ by an order of magnitude in their water solubilities. Through PDMS membrane, their fluxes are similar with the slig htly more lipophilic compound 5 possessing the higher flux. Through hairless mouse skin, it is the more hydrophilic compound 7 which has 7.9 times

PAGE 87

75 the flux of compound 5 when delivered from IPM and 5 times the flux of compound 5 when delivered from water. Application of the Roberts-Sloan equation to this and other data in the next chapter will allow us to quantify the relative effect s of IPM solubility, water solubility and molecular weight on flux through PDMS memb rane. However, even without a strict analysis, it is obvious that, without some modification, flux through PDMS does not correlate to flux through mouse skin. The physical basis for this effect is still a matter of debate, but the favorable influence of aqueous and IPM solubility has been observed in all series comprising the current databases. Determining the relative effect of log SIPM and log SAQ will require a more rigorous model and will be addressed in chapter 6 when this and the other mouse skin data ar e fit to the Roberts-Sloan equation. Effect of the Vehicle on Flux through Hairless Mouse Skin It is clear that the vehicle has a pr ofound effect on flux through hairless mouse skin. Flux values from IPM were nearly 10 times higher then flux values from water no matter which compound was present in the ve hicle. A similarly consistent 10-fold increase was seen for the second applicat ion fluxes through IPM exposed mouse skin compared to those through water exposed mouse skin. This is especially significant because the same vehicle, propylene glycol, was always used to deliver theophylline. The persistence of the incr eased flux through the IPM expos ed mouse skin, even after removal of the initial vehicle, supports the conclusion that ex posure to IPM alters hairless mouse skin in an irreversible way and permanently decreases its resistance to penetration. The relative effects of IPM and water upon the permeation barrier of mouse skin have been well characterized. In 2003, we co mpared the delivery of chemically stable 5FU and 6-MP prodrugs through hairless mouse skin from water and IPM (Sloan et al

PAGE 88

76 2003). There were a sufficient number of compo unds included in this study to prepare a predictive flux model for each vehicle. The conclusions of these experiments will be covered in more detail in th e next chapter; however, an important finding was a nearly constant ten-fold increase in the flux of com pounds delivered from IPM. From this and from the APAP prodrug data, it would a ppear that the influence of IPM on the permeability of hairless mouse skin is consistent and predictable. Residual Membrane Amounts If the assumptions of the Roberts-Sloan model hold true and if mouse skin is reasonably consistent between individuals, th e amount of drug or prodrug remaining in the skin after the diffusion cell experiment should be proportiona l to the solubility of the prodrug in the skin. Furthermore, since hi gher solubility in the skin corresponds to higher flux, a correlation should be observed between higher flux and higher residual skin amount. This should be especi ally true for the lower we ight prodrugs whose smaller molecular size has a lesser effect on flux. Table 5-4. Average residual amounts ( Std. Dev.) of APAP and its prodrugs remaining in hairless mouse skin (HMS) a nd PDMS membrane after the flux experiments. Compound HMS/IPMa,bHMS/Aqa,b PDMS/Aqa,b APAP, 1 2.74 (0.70)0.90(0.30) 0.18 (0.04) 2 5.45(1.57) 0.95(0.15) 0.88(0.31) 3 1.08(0.13) 0.76(0.13) 0.38(0.06) 4 2.84(1.44) 0.95(0.22) 0.63(0.05) 5 1.91(0.08) 0.25(0.05) 0.13(0.03) 6 1.79(0.43) 0.40(0.14) 0.12(0.02) 7 3.75(0.74) 1.56(0.22) NAc 8 0.64(0.12) 0.34(0.07) NAc a Membrane/Vehicle. b In units of mols. c Not measured.

PAGE 89

77 The data does show a reasonable agreem ent correlation between the residual membrane amount and the measured flux valu es for hairless mouse skin. From both water and IPM, those compounds with the highe st flux and next to highest flux (I and VI) also delivered the highest amounts of APAP to the skin. For the remaining APAP carbonates, the relative position of each com pound, when ranked by flux, is within one, or at most two positions, of its value when ranked by residual skin amount. This correlation is acceptable but it does not take into account the effect of molecular weight on flux. Flux and residual skin data can be log transformed and fit to a simplified version of the Roberts-Sloan model with the residual skin amount being used as a surrogate for skin solubility: log JMV = x + log (Residual Skin Amount) – z MW The best fit equations for flux from water and IPM through hairless mouse skin and from water through PDMS are: log JMIPM = 0.650 + log (Residual Skin Amount) – 0.00748 MW (r2 = 0.797) (5.4) log JMAQ = -1.38 + log (Residual Skin Amount) – 0.00113 MW (r2 = 0.820) (5.5) log JPAQ = -2.86 + log (Residual Skin Amount) + 0.0069 MW (r2 = 0.851) (5.6) Although the correlation coefficients ar e not particularly high, it must be remembered that residual skin amount is more sensitive to minute differences in the membrane then is maximum flux. Even for the highly homogenenous PDMS membrane, which showed very consistent flux values, residual membrane amounts had an average relative standard deviat ion of 19% which is only slightly smaller then the same values for IPM and water treated mouse skin (23% and 22% respectively). Given the inherent variability in the amount of compound absorb ed into the membrane, equations 5-4, 5-5

PAGE 90

78 and 5-6 are consistent with the homogenous membrane assumption for hairless mouse skin.

PAGE 91

79 CHAPTER 6 PREDICTIVE MODELS OF SOLUBILITY AND FLUX Determination of the Coefficients of the General Solubility Equations In theory, the melting point coefficien t found in equations 3-10 and 3-14, BO, can be calculated directly by using the WaldenÂ’s Rule, which states that SF for most small nonelectrolytes is approximately 56.6 J K-1 mol-1. However, an empirical approach based primarily upon our own solubility data was expected to give a better representative value. It was also necessary to use our solubility data to determine a value for the molecular weight coefficient, C. Using our 5-FU and 6-MP prodrugs and an additional set of ACOM phenytoin (ACOM-PhT) prodrugs from Stella et al. (1999), aqueous and IPM solubility, melting point and molecular weight data was analyzed using nonlinear multiple regression to generate best-fit coe fficients to equations 3.10 and 3.14 for each series. The results of this analysis are summarized in tables 6-1 and 6-2. Table 6-1: Series specific best fit coefficients to equation 3. 10 using IPM solubility data and physical properties from the 5-FU and 6-MP prodrugs. Series ID AIPM BO 1-ACOM-5-FU 2.41 0.0188 1-AOC-5-FU 2.98 0.0190 1-AC-5-FU 2.99 0.0144 1-AAC-5-FU 3.65 0.0217 6-ACOM-6-MP 2.19 0.0129 6,9-ACOM-6-MP 2.86 0.0225 3-ACOM-5-FU 2.37 0.0165 Average (Std. Dev) 2.74 (0.50) 0.0177 (0.0034)

PAGE 92

80 Table 6-2: Series specific best-fit coefficien ts to equation 3.14 using aqueous solubility data and physical properties from the 5-FU, 6-MP and phenytoin (PhT) prodrugs. Series ID AAQ BO C 1-ACOM-5-FU 12.74 0.0167 0.0433 1-AOC-5-FU 12.16 0.0160 0.0416 1-AC-5-FU 10.58 0.0125 0.0412 1-AAC-5-FU 13.65 0.0238 0.0458 6-ACOM-6-MP 10.44 0.0106 0.0348 6,9ACOM-6-MP 14.23 0.0218 0.0395 3-ACOM-5-FU 12.00 0.0154 0.0403 3-ACOM-PhT 20.54 0.0230 0.0568 Average (Std. Dev) 13.29 (3.21) 0.0175 (0.0049) 0.0429 (0.0064) The estimate for the molecular weight coefficient, C, appears to be the most robust. The individual se ries values for C possess a lower variability then the corresponding series values used to estimate the other coefficients and have a relative standard deviation of 14%. The average C value of 0.0429 agrees very well with the molecular volume effect on aqueous solubi lity of 0.0437 MV reported by Huibers and Katritzky (1998) in their model for predicti ng the aqueous solubility of hydrocarbons and halogenated hydrocarbons. Given that thei r model (which is based upon a dataset of 241compounds) and the prodrug series model ar e both attempting to measure the effect of what is essentially pure st eric bulk on aqueous solubility it is encouraging that they are in such close agreement. When derived from IPM solubility data, the best-fit BO values have a relative standard deviation that is close to the relativ e standard deviation fo r the best-fit A values (19% compared to 18%, respectively). When derived from water solubility data, the relative standard deviati ons of the best-fit BO and A coefficients increase to 28% and

PAGE 93

81 24%, respectively. For both water and IPM da ta, the A coefficient estimates have less relative variability than the BO estimates. This somewhat undermines the supposition that BO should be more conserved between differe nt series then A. While it is possible that for a given solvent both A and BO may be equally independent of the nature of the solute, it is also possible that some of th e consistency in the A value estimates is a consequence of the homogeneity in the series used to generate the estimates. Of the seven series used to estimate coefficients fr om IPM solubility (table 6-1), five of them have 5-FU as the parent compound. Similarly, five of the eight series used to determine the A and BO coefficients from water solubility ar e based on 5-FU (table 6-2). Since the A coefficient is, in theory, determined by a molecule's specific structure and since the same parent moiety appears in all the member s of the various 5-FU series, it is reasonable that the small variation in the A coefficient estimates is a result of the large proportion of 5-FU compounds comprising the two data sets. In addition to the preceding arguments there are two observations, which lend support to BO being considered a universal value for prediction. First, despite the somewhat large variability among the individua l series values, the average IPM derived BO value is nearly identical to the water deri ved value. This is consistent with the expectation that the BO would be the same for all solven ts. Second, when the average A value for a given solvent is used to calculate potential BO values for the members of a series, these calculated BO values are erratic acr oss the series. Convers ely, if the average BO value for a given solvent is used to calculate A values for series members, the A values are more consistent across the seri es. Therefore, the predictive strength of equations 3.10 and 3.14 was tested using the average values of C and BO.

PAGE 94

82 The intent of developing the solubility e quations was to enable data from a single series member to predict the behavior of th e other series members. To establish the ability of these equations to accomplish this, solu bility data from the first member of each series appearing in tables 6-1 a nd 6-2 were used to determine AIPM and AAQ coefficient values for their respective series. The one exception to this was the ACOM-Th in which the first series member appears to be an outli er so the second series member was used as a replacement. In turn, these AIPM and AAQ values were used to calculate IPM and aqueous solubility values for all remaining compounds covered in the tables. Figures 6-2 and 6-3 show the correlations between predic ted and experimental solubilities in IPM and water. 0 0.5 1 1.5 2 2.5 00.511.522.5Calculated log SIPMExperimantal log SIPM Figure 6-1. Calculated versus experimental log IPM solubi lity using equation 3.10 and A coefficients determined from the smallest series members.

PAGE 95

83 -4.5 -3.5 -2.5 -1.5 -0.5 0.5 1.5 2.5 -4.5-3.5-2.5-1.5-0.50.51.52.5Calculated log SAQExperimantal log SAQ Figure 6-2. Calculated ve rsus experimental log SAQ using equation 3.14 and A coefficients determined from the smallest series members. The average absolute error for predicting solubility in both solvents is approximately the same with 0.18 log units for the IPM model and 0.21 log units for the aqueous model. This is approximately half that of the 0.41 log unit average absolute error reported by Ran, Jain and Yalkowsky (2001) with their modified form of the general solubility equation. The largest outlier is th e C1-ACOM-Theopyilline compound with an absolute average error of approxima tely one log unit for both models; twice the magnitude of the next largest errors in both models. Since the IPM and aqueous solubility of this compound are consistent with itsÂ’ flux and w ith itsÂ’ expected partitioning coefficient, the most likely explanation for itsÂ’ poor adherence to the solubility models lies with itsÂ’ melting point. It is quite possible that the original method for purifying the C1-ACOM theophylline isolated a low melting point polymorph that was converted to a more stable (and less solubl e) form during solubility determination.

PAGE 96

84 The C1-ACOM-Theophylline notwithstanding, th e low average absolute errors of both models indicate that they are capable of making useful predicti ons and some useful generalizations Solubility Behavior of the 4-AO C and 4-MOAOC-APAP Prodrugs The 4-AOC-APAP prodrugs exhibit solubility behavior that is t ypical of the other homologous prodrug series. When th e solubility data from compounds 2-6 were fit to the general organic and aqueous solubility mode ls (eq. 3.10 and 3.14), the following best-fit equations were obtained: log SIPM = 2.921 – 0.0200 (TM – T) (r2 = 0.765) (6.1) log SAQ = 11.183 – 0.0169 (TM – T) – 0.0400 MW (r2 = 0.990) (6.2) The agreement between calculated and expe rimental solubility values generated from these equations is very close for both IPM and aqueous solubility. For IPM solubility, the average absolute error is only 0.059 log units and, fo r aqueous solubility, it is 0.085 log units. The accuracy of the calcula ted IPM solubility valu es indicates that the relatively poor correlation coefficient for the IPM solubility equation should not be interpreted as a failure of the model. The poor correlation coefficient of equation 6.1 is actually a consequence of the small range in log SIPM values possessed by the AOCAPAP prodrugs. Since the difference in IP M solubility between the most and least soluble prodrug in this series is only 0.406 log units, even the small observed deviation from the from the model is sufficient to produce a poor r2. The series independent coefficients, BO and C, determined from the AOC-APAP prodrug data are close to the corresponding aver age values obtained from the eight other prodrug series. The melting point coefficien t obtained from AOC-APAP IPM solubility data is only 0.0023 from the combined database average BO value, while the coefficient

PAGE 97

85 obtained from aqueous solubility data is only 0.0008 from the database average. Similarly, the molecular weight coefficien t of the AOC-APAP series differs from the database average C value by 0.0029. As with the other series, these observations support the relative independence of these coefficients from series-specific influences. The series specific solubility parameter for the IPM solubility, AIPM, ranged from 2.19 to 3.65 for the series comprising the database. The AIPM value for compounds 2-6 lies precisely in the center of this range. In direct comparison, this AIPM value is very close to the corresponding va lues of the 1-AOC-5-FU, 1-AC-5-FU and 6,9-ACOM-6-MP series. Of these three series, the 1-AOC-5-FU series also has a BO value of 0.0190, which is close to the 0.0200 BO value of the AOC APAP series. Therefore, the AOCAPAP series has an inherent organic solubility equivalent to that of 1-AOC-5-FU series. In other words, for a given molecular we ight and melting point, a 4-AOC-APAP prodrug will have the approximately same IPM so lubility as a 1-AOC-5-FU prodrug. The situation is quite different for aqueous solubility. The AOC-APAP compounds have the third smallest AAQ parameter of the homologous series yet examined. Compared to the 1-AOC-5-FU series, AAQ for the 4-AOC-APAP series is smaller by 0.98. With such a decrease in AAQ, a 4-AOC-APAP prodrug will therefore be approximately one-tenth as water solubl e as a 1-AOC-5-FU prodrug with similar molecular weight and melting point. Such a comparison can be made between compound 2 and the C2 member of the 1-AOC-5-FU series. Both compounds have similar molecular weights (209 and 202 amu respectively) but a 15C difference in melting point. As predicted, despite having the higher melting point, the C2-AOC-5-FU compound has a ten-fold higher water solubility. It is th is systematically weak ability of the 4-AOC-

PAGE 98

86 APAP prodrugs to interact favorably in a queous solution, which, along with their only average IPM solubility, ultimately results in their modest flux. In Chapter 3, it was theorized that AIPM and AAQ for a given homologous series could be estimated from a single series member so long as BO and C were sufficiently constant for all series. Using solubility da ta from the first member of each homologous series in the database, IPM and water solubility for the remaining series members were predicted from melting point and molecular weight with reasonable accuracy. This same process was applied to the members of th e 4-AOC APAP series. Using the IPM and water solubility of compound 2, and using the average Bo and C parameters obtained from the IPM flux database, the following AIPM and AAQ were obtained. log SIPM = AIPM – 0.0176 (TM – T) AIPM = log SIPM + 0.0176 (TM – T) AIPM = 1.076 – 0.0176 (113.5-25) = 2.63 log SAQ = AAQ – 0.0176 (TM – T) – 0.0429 MW AAQ = log SAQ + 0.0176 (TM – T) + 0.0429 MW AAQ = 1.232 + 0.0176 (113.5-25) + 0.0429 (209) = 11.76 Using these values for AIPM and AAQ, calculated IPM and water solubilities were predicted for compounds 3-6. Using an AIPM of 2.63, IPM solubility was very well predicted for the 4-AOC series yielding an average prediction erro r of 0.10 log units. Using an AAQ value of 11.76, water solubility is less well predicted, although, with an average prediction error of 0.24

PAGE 99

87 Table 6-3. Predicted IPM and wa ter solubilities for compounds 3-6. Compound Exp. Log SIPM a Pred. Log SIPM a Exp. Log SAQ a Pred. Log SAQ a 3 0.968 0.944 0.653 0.499 4 1.37 1.23 0.477 0.180 5 1.14 0.979 -0.354 -0.666 6 1.22 1.16 -1.49 -1.69 a In Units of mM. log units, the model is still reasonably accurate. It is important to note that the predicted water solubilities are consistently lower then the experimental values This suggests that the data from compound 2 did not estimate a value for AAQ, which truly reflected the series value. If compound 3 had been used to estimate AAQ instead of compound 2, then the average error of prediction for log SAQ would have been 0.08 log units. This is a problem intrinsic to this type of analysis a nd it highlights how care must be taken in the interpretation of predicted results. In addition to predicting solubility beha vior across a series, the series specific IPM and water solubility equations were developed to quantify the effects of various physical parameters on solubility. Accoun ting for melting point and molecular weight influences on solubility allows the inherent solubilizing effect of di fferent promoieties to be compared. Compounds 7 and 8 can be used to estimate the series specific AIPM and AAQ values for ether containing carbonate pr omoities by following the same procedure that was used to estimate these parameters from compound 2. Since the 4-AOC-APAP prodrugs and the 4-MOAOC-APAP prodrugs differ only in the composition of the promoiety, any difference in the AIPM and AAQ values must be due the promoiety alone. The results of this analysis are summarized in Table 6-4.

PAGE 100

88 0 0.5 1 1.5 2 2.5 00.511.522.5Predicted log SIPMExperimantal log SIPM Database Compounds Compounds 3-6 Figure 6-3. Predicted versus experimental IPM solubilities for the 4-AOC-APAP prodrugs (AIPM = 2.63). -4.5 -3.5 -2.5 -1.5 -0.5 0.5 1.5 2.5 -4.5-3.5-2.5-1.5-0.50.51.52.5Predicted log SAQExperimantal log SAQ Database Compounds Compounds 3-6 Figure 6-4. Predicted versus experimental aqueous solubilities for the 4-AOC-APAP prodrugs (AAQ = 11.76).

PAGE 101

89 Table 6-4. Estimated AIPM and AAQ using data from compounds 7 and 8. Data Source Est. AIPM Est. AAQ Comp. 7 1.96 13.12 Comp. 8 2.24 13.55 4-AOC-APAP Averagea 2.71 11.95 a Average of estimated parameters using data from compounds 1-5. The two 4-MOAOC APAP compounds genera te series specific parameters that are consistent with one another. Both compounds indicate th at ether containing carbonate promoieties are less effective at increasing IPM sol ubility than alkyl carbonates, but are better at improving water solubility. When matched for melting point and molecular weight, a prodrug of the MOAOC type will have approximately one fourth the IPM solubility of an AOC type prodrug (log SIPM lower by 0.6 log units), but approximately 24 times the water solubility (log SAQ higher by 1.38 log units). Given the roughly equal effect of IPM and aqueous sol ubility on flux, ether-c ontaining promoieties have a higher inherent potential to improve topica l delivery then do n-alky l promoietes. Modeling the Flux of the 4-AOC and 4MOAOC APAP Prodrugs through Hairless Mouse Skin from IPM and Water To examine the flux behavior of the APAP prodrugs relative to the flux behavior of the heterocyclic prodrugs, the Roberts-Sloan equation was applied to several data sets. The first attempt to predict the fluxes of the APAP prodrugs was performed using the Roberts-Sloan model derived from the or iginal 42-compound database Roberts and Sloan, 1999).

PAGE 102

90 log JMIPM = -0.216 + 0.534 log SIPM + (1-0.534) log SAQ – 0.00361 MW (r2 = 0.951) (6.3) Following publication of the first RobertsSloan equation, three additional series of 5-FU prodrugs and the C5 member of the 6,9-bis-ACOM 6-MP series were synthesized, characterized and added to the IPM/Hairless mouse skin database. The inclusion of these thirteen compounds revised the coefficients of the model as shown in equation 6.4. log JMIPM = -0.287 + 0.535 log SIPM + (1-0.535) log SAQ – 0.00341 MW (r2 = 0.937) (6.4) Finally, compounds 1 through 8 were added to the database to give the final form of the model. Log JMIPM = -0.501 + 0.517 log SIPM + (1-0.517) log SAQ – 0.00266 MW (r2 = 0.912) (6.5) Each model was used to calculate expect ed IPM flux values for the APAP prodrugs and these values are presented and compared to the experimental flux values in table 6-5. Table 6-5. Predicted and experime ntal flux values for compounds 1-8 through hairless mouse skin from IPM. Compound Experimental log JMPIM a Predicted log JMIPM a Eq. 6.3 log JMIPM Eq. 6.3 Predicted log JMIPM a Eq. 6.4 log JMIPM Eq. 6.4 Predicted log JMIPM a Eq. 6.5 log JMIPM Eq. 6.5 1, APAP -0.29 0.32 -0.61 0.28 -0.57 0.21 -0.50 2 0.00 0.18 -0.18 0.15 -0.15 0.09 -0.10 3 -0.76 -0.20 -0.56 -0.23 -0.53 -0.28 -0.48 4 -0.45 -0.12 -0.33 -0.14 -0.31 -0.19 -0.26 5 -1.01 -0.68 -0.33 -0.70 -0.31 -0.75 -0.26 6 -1.49 -1.27 -0.22 -1.28 -0.21 -1.33 -0.16 7 -0.11 0.02 -0.13 0.00 -0.11 -0.02 -0.09 8 -1.06 -0.72 -0.34 -0.73 -0.33 -0.75 -0.31 a In units of mol cm-2 h-1.

PAGE 103

91 -2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 -2.50-2.00-1.50-1.00-0.500.000.501.001.50log JMIPM = -0.501 + 0.517 log SIPM + (1-0.517) log SAQ 0.00266 MWExperimental log JMIPM 5-FU, 6-MP and Th Prodrugs APAP Prodrugs Figure 6-5. Experimental ve rsus calculated log maximum fl ux values through hairless mouse skin from IPM using equation 6.5. The revised coefficients of equation 6.5 are close to those obtained from the smaller database and are not significantly di fferent from the original coefficients of equation 6.3. In addition, the average difference between the calculated and the experimental log JMIPM values for members of both databases is identical (~0.16 log units). These observations indicate that the behavior of the additional heterocyclic produgs is similar to that of the original database members and their addition to the database does not signif icantly alter the model. APAP and all of its prodrugs gave lower flux then anticipated, regardless of the model used. Compounds 1 though 8 under performed by an average of 0.34 log units according to equation 6.3 and by an average of 0.32 log units according to equation 6.4.

PAGE 104

92 Even when the APAP prodrugs are included in the model (eq. 6.5), they still under perform by an average of 0.27 log units according to equati on 6.4; approximately twice the average difference of the remaining members of the database. In addition, three of the series members, compounds 1, 3 and 8, have the largest estimation error of the entire database. With no other APAP series or seri es based upon molecules similar to APAP in the database to serve as a comparison, it is impossible to determine whether the lower than expected performance is due to a unpredicted aspect of APAP or due to the model being influenced disproportionately by the ot her heterocyclic com pounds. In other words, it is not clear how the addition of other types of prodrugs to the database will ultimately affect the IPM model and it is possible that their addition will mitigate the apparent underperformance of the APAP prodrugs. In any case, the performance of the APAP series is not likely a result of the carbonate promoiety since APAP itself also underperforms. Although the IPM model does not predict the behavior of the APAP compounds as well as the other prodrug series, it still provides a more then adequate estimation of their flux. Along with predicting topical de livery, flux models are used to determine those members of a series that are most likel y to perform best. Therefore, while accurate flux prediction is the target, it is also important for a flux model to identify the relative order of performance for the members of a given series. All the models correctly identified the best performing prodrugs of the 4-AOC and 4-MOAOC APAP series and they correctly ranked the performanc e of all the remaining prodrugs. The Roberts-Sloan equation was initially derived to predict flux from IPM. However, its final form is independent of th e vehicle used. Therefore, it can be used

PAGE 105

93 without modification to predict the deliver y of compounds from water, or any other vehicle, in a manner analogous to the predicti on of flux from IPM. In the heterocylic database, only three series, the 3-ACOM 5-FU, 6-ACOM 6-MP and bis-6,9-ACOM 6MP prodrugs, were sufficiently stable to hydrol ysis to allow their delivery from water in mouse skin diffusion cell experiments. The flux of these compounds through hairless mouse skin from saturated aqueous solutions was measured and, when combined, the members of these series produced a database of 18 compounds (Sloan et al, 2003). These data were used to generate a R oberts-Sloan model for aqueous flux: log JMAQ = -1.497 + 0.660 log SIPM + (1-0.660) log SAQ 0.00468 MW (r2 = 0.765) (6.6) Although the aqueous Roberts-Sloan mode l does not have as high a regression coefficient as the IPM model, its average erro r of prediction is only slightly larger, 0.18 log units for flux from water as opposed to 0. 14 log units for flux from IPM. In contrast to their delivery from IPM, the APAP pr odrugs conform to the aqueous model more closely then any other database series. When equation 6.6 was used to predict the flux for compounds 1 through 8, the average error of predicti on was 0.15 log units; close to the error for the IPM model. It is consistent with this good fit to model 6.6 that inclusion of the APAP prodrugs into the database only sl ightly affects the coefficients of the model and gives a slight improvement to the modelÂ’s overall fit (equation 6.7). log JMAQ = -1.665 + 0.657 log SIPM + (1-0.657) log SAQ 0.00409 MW (r2 = 0.774) (6.7) Equation 6.7 generates an average absolute error of prediction of 0.16 log units for the entire database and 0.13 l og units for the APAP prodrugs.

PAGE 106

94 Table 6-6. Predicted and experime ntal flux values for compounds 1-8 through hairless mouse skin from water. Compound Experimental log JMAQ a Predicted log JMAQ a Eq. 6-6 log JMAQ Eq. 6-6 Predicted log JMAQ a Eq. 6-7 log JMAQ Eq. 6-7 1, APAP -1.73 -1.34 -0.39 -1.41 -0.32 2 -1.46 -1.35 -0.12 -1.39 -0.07 3 -1.62 -1.68 0.06 -1.72 0.10 4 -1.57 -1.54 -0.03 -1.57 0.00 5 -2.17 -2.04 -0.13 -2.06 -0.11 6 -2.28 -2.50 0.23 -2.52 0.24 7 -1.45 -1.57 0.11 -1.58 0.13 8 -2.38 -2.27 -0.11 -2.28 -0.10 a In units of mol cm-2 h-1. -3.50 -3.00 -2.50 -2.00 -1.50 -1.00 -0.50 0.00 -3.50-3.00-2.50-2.00-1.50-1.00-0.500.00 log JMAQ = -1.665 + 0.657 log SIPM + (1-0.657) log SAQ 0.00409 MWExperimental log JMAQ 5-FU and 6-MP Prodrugs APAP Prodrugs Figure 6-6. Calculated vers us experimental log maximu m flux values through hairless mouse skin from water using equation 6-7.

PAGE 107

95 The flux of 6-MP from IPM demonstrates the largest deviation from the RobertsSloan model of any compound yet measured ( log JMIPM = 0.69). A similarly large deviation from the model is seen when 6-MP is delivered from water. Since the low flux of 6-MP is independent of the vehicle, it is reasonable to speculate that 6-MP exhibits unusual behavior in solution. Sloan hypothesized that the ab ility of 6-MP molecules to stack efficiently with one another might lead to a strong self-associati on in solution. This association would limit the amount of monomeri c 6-MP available to enter the skin and reduce the observed flux. Such a situation does not exist with the APAP prodrugs. The close agreement between the predicted and experimental flux values for the water delivered APAP prodrugs suggests that, unl ike 6-MP, these compounds do not possess inherent properties that cause them to deviat e from behavior predicted by the model. Prediction of flux is not the only informa tion that can be gained from the various Roberts-Sloan models. The theoretical develo pment of the Roberts-Sloan model, like the Potts-Guy model, makes as one of its fundament al assumptions that the skin behaves like a homogeneous membrane with regard to diffu sion. Recalling the equation for maximum flux through a homogeneous membrane and comparing this to the Roberts-Sloan equation, it is possible for the coefficients a nd solubility terms of the flux models can be correlated to the properties of the skin Therefore given the two equations, log JM = log D0 – log L + log SM – MW log JM = x + y log SIPM + (1-y) log SAQ – z MW the following correlations can be made: x = log D0 – log L (6.8) log SM = y log SIPM + (1-y) log SAQ (6.9)

PAGE 108

96 z = (6.10) Equation 6.8 indicates that the x coeffici ent is a combination of the membraneÂ’s intrinsic resistance to diffusion, D0, and the effective thickness or path length through the membrane (L). Whether the skin is expos ed to IPM or an aqueous vehicle, the conditioning process ensures that the skin is near maximum hydration before the application of the drug suspensions and ha s undergone a similar amount of swelling. Therefore, the thickness of the skin is appr oximately the same regardless of the vehicle used. Given a near constant thickness, the x coefficient becomes a reflection of the intrinsic diffusivity of the skin and the di fference between the x coefficients the two models is equal to the log of the D0 ratio. Recalling the theoretical development from chapter 2, the factor D0 has units of h-1 and it indicates the apparent average velocity with which a hypothetical entity of negligible volume moves thro ugh the skin. Since a higher apparent velocity corresponds to a lowe r energy requirement for movement, D0 is an inverse indicator of the work require d to move through the membrane. The difference between x terms in the IP M model and the aqueous model is 1.15, which converts to an anti-log value of approximately 14 for the ratio of D0 IPM/D0 AQ. In other words, for a given compound, the flux th rough IPM damaged skin is expected to be 14 times higher then through water damaged skin as a result of the d ecrease in the energy cost required to move through the skin. This is consistent with the observed approximate tenfold lower flux of compounds delivered from water as compared to those delivered from IPM (Sloan et al, 2003). Unfortunately, the models are unable to a ssign an underlying physical cause to the decrease in membrane integrity. However, it is well established that the high degree of

PAGE 109

97 organization in the components of the stratum corneum is responsible for the barrier of the skin. The substantial amount of these comp onents is kept in this organization without covalent bonding to relatively immobile structures such as the corneocytes. While IPM does not readily enter undamaged stratum corneu m, it is possible that IPM is capable of leaching from the stratum corneum some of its lipophilic components. For example, octanol has been shown to remove enzymes fr om the stratum corneum that are capable of hydrolyzing prodrugs in the ve hicle (Waranis and Sloan, 1987). The loss of these components may disrupt inter-corneocyte lipid organization and lead to a loss in barrier function. The inability of water to simila rly leach these components may account for its less damaging effect. Although the intercorneocyte lipids have a lamellar structure of alternating polar and non-polar regions, the results of the Robe rts-Sloan treatments demonstrate that a series description is not required for accurate prediction of flux. From a functional point of view, the barrier of the skin can be viewed as a single phase and the properties of this phase can be inferred from the y coefficient. As indicated by equation 6.9, a compoundÂ’s solubility in the skin can be estimated from its solubility in IPM and water. Beginning with the fundamental work of Hildebrand ( 1936), many researchers have made attempts to predict or explain solubili ty behavior from the physical properties of the solute and solvent. Just as solubility can be predicte d from physical parameters, physical parameters can be conversely predicted from solubility. Of the solubility models available, the amended solvation energy relationship of Abraham (Abraham and Le, 1999) provides the most convenient method of examining the physic al properties of skin from solubility.

PAGE 110

98 The general form of AbrahamÂ’s model st ates that any physical property (ie. solubility, partition coefficient or permeability ) can be predicted from parameters that reflect the various molecular interac tions of the solute and solvent: log SP = c + rR2 + s 2 H + a 2 H + b 2 H + vVx (6.11) Each term of the equation contains a parameter derived from the solute (R2, 2, and Vx) and a corresponding coefficient characteristi c of the solvent (r, s, a, b and v). R is the excess molar refraction, 2 is the polarizability, is the hydrogen bond acidity, is hydrogen bond basicity and Vx is the molecular volume. An equation corresponding to 6.9 may be written for both IPM and water solubility: log SIPM = c + rIPMR2 + sIPM2 H + aIPM2 H + bIPM2 H + vIPMVx log SAQ = c + rAQR2 + sAQ2 H + aAQ2 H + bAQ2 H + vAQVx Substituting the above two equations into equation 6.11 results in an Abraham equation for solubility in the skin: log SMS = y log SIPM + (1-y) log SAQ y log SIPM = y cIPM + y rIPMR2 + y sIPM2 H + y aIPM2 H + y bIPM2 H + y vIPMVx (1-y) log SAQ = (1-y) cAQ + (1-y) rAQ R2 + (1-y) sAQ2 H + (1-y) aAQ2 H + (1-y)bAQ2 H + (1-y) vAQVx log SMS = [y cIPM + (1-y) cAQ] + [y rIPM + (1-y) rAQ]R2 + [y sIPM + (1-y) sAQ] 2 H + [y aIPM + (1-y) aAQ] 2 H + [y bIPM + (1-y) bAQ] 2 H (6.12) From equation 6.12, each coefficient that is characteristic to the skin is equivalent to a sum of the corresponding Abraham terms for water and IPM scaled by y coefficient from the Roberts-Sloan model.

PAGE 111

99 Water solubility is greatly influenced by the ability to form hydrogen bonds. The hydrogen bond acidity and basicity terms for water solubility (aAQ and bAQ) were reported by Abraham and Le to be 0.646 and 3. 279, respectively. Corresponding model coefficients for IPM have not been yet report ed, but Abraham has used the same model to predict partitioning between water and various organic solvents. Using a and b values from the partitioning models for organic solvents similar to IPM, we can infer that aIPM and bIPM are either small compared to aAQ and bAQ or somewhat negative. Assuming that aIPM and bIPM are small compared to and aAQ and bAQ, the hydrogen bond values for mouse skin, aMS and bMS, can be approximated as: aMS (1-y) aAQ 0.483 aAQ bMS (1-y) bAQ 0.483 bAQ With these assumptions, IPM-treated mous e skin has approximately one half the ability to interact favorably with a dr ug through hydrogen bonding as does water. However, even if the assumptions are not true, hydrogen-bonding cap acity of mouse skin still increases linearly with decreasing y and this factor can be used to compare skin sections exposed to different vehicles. It follows that compounds, which take advantage of these interactions, will more readily pene trate into the skin and the parallel between water solubility and the ability to form hydr ogen bonds underlies the positive correlation between water solubility and flux. The y coefficient of mouse skin exposed to IPM is 0.14 lower then the y value for mouse skin exposed to water. This indica tes that mouse skin damaged by IPM has a higher capacity to form hydrogen bonds then does water damaged skin. Given that the matrix of the stratum corneum is a complex, multi-component mixtur e, a relative increase

PAGE 112

100 in the capacity to hydrogen bond is most likely the result of an increase in the concentration of hydrogen bonding components. Th is result is consiste nt with the theory that IPM is capable of leach ing lipophilic components from the stratum corneum. As lipophilic components are removed from the inter-corneocyte matrix and the organization of the matrix disrupted, water from the dermis will diffuse into the stratum corneum. Since the prodrug saturated IPM solution covering the epidermis prevents the loss of this water through evaporatio n, the resulting equilibrium concen tration of water in the stratum corneum is increased. In turn, the greater percentage of water in the stratum corneum is observed as a decrease in the y coefficient of the Roberts-Sloan model. Ultimately, the observed effect on flux caused by IPM result s from a simultaneous change in the lipid organization and composition of the stratum corneum. Modeling the Flux of the 4-AOC and 4MOAOC APAP Prodrugs through PDMS Polymer Membrane from Water The initial attempts to m odel the flux of the water stable prodrugs through PDMS membrane were unexpectedly disappointing. Since PDMS membrane is very homogeneous and given the highly consiste nt results from the diffusion cells using polymer membranes, it was anticipated that the flux through the polymer would show the best correlation to the Robe rts-Sloan model. Unfortunate ly, the model generated from the entire dataset displayed an unusually poor correlation: log JPAQ = -2.131 + 0.841 log SIPM + (1-0.841) log SAQ – 0.00364 MW (r2 = 0.653) (6.13) In spite of the poor fit, eq uation 6.13 still possessed co efficients that were in agreement with known properties of PDMS membrane. The higher y coefficient,

PAGE 113

101 Table 6-7. Molecular weights, log IPM so lubilities, log aqueous solubilities, log maximum flux values through hairless mouse (log JMAQ) and log maximum flux values through PDMS from water for the chemically stable prodrug series. Compound ID MW log SIPM a log SAQ a log JMAQ b log JPAQ b C1-ACOM-5-FU 202 0.09 1.8 -1.77 -2.64 C2 216 1.2 2.25 -1.41 -1.78 C3 230 1.42 1.93 -1.13 -1.6 C4 244 1.47 1.32 -1.43 -1.7 C5 258 1.63 0.92 -1.41 -1.58 C7 286 1.6 -0.25 -1.85 -1.82 C1-6-ACOM-6-MP 224 0.02 0.86 -2.55 -3.32 C2 238 0.36 0.61 -2.19 -2.82 C3 252 0.52 0.31 -2.00 -2.67 C4 266 0.62 -0.1 -2.18 -2.65 C5 280 0.57 -0.63 -2.37 -2.73 C1-6,9ACOM-6-MP 296 0.72 0.46 -1.98 -1.92 C2 324 1.53 0.22 -1.89 -1.36 C3 352 1.96 -0.71 -2.27 -1.68 C4 380 2.24 -1.33 -2.48 -2.4 C5 408 1.7 -2.98 -3.07 -3.29 APAP,1 151 0.28 2 -1.73 -2.68 C1, 2 209 1.08 1.31 -1.46 -1.51 C2, 3 223 0.97 0.58 -1.62 -1.74 C3, 4 237 1.37 0.43 -1.57 -1.44 C4, 5 251 1.14 -0.37 -2.17 -1.79 C6, 6 279 1.22 -1.32 -2.28 -2.16 MeO-C2, 7 253 1.01 1.54 -1.45 -1.85 MeO-C3i, 8 267 0.53 0.52 -2.38 -2.41 a In units of mM b In units of mol cm-2 h-1 indicating a greater dependence of flux on IPM solubility, was consistent with the more lipophilic nature of the polymer membrane re lative to mouse skin. The lower intrinsic diffusivity of the polymer membrane was also reflected by a relative decrease in x. The

PAGE 114

102 reasonable coefficients of the model suggested that the poor correlation may be due to the unusual behavior of select database member s, but an examination of the delta log JPAQ values failed to identify any specific outliers. To further investigate the behavior of the polymer membrane, alternate models for predicting flux were used. -3.50 -3.00 -2.50 -2.00 -1.50 -1.00 -3.00-2.50-2.00-1.50-1.00log JPAQ = -2.131 + 0.841 log SIPM + (1-0.841) log SAQ -0.00364 MWExperimental log JPAQ Figure 6-7. Calculated vers us experimental log maximu m flux values through PDMS from water by Equation 6.13. In the development of their flux model, Kastings, Smith and Cooper (1987) assumed that the stratum corneum was lipophil ic enough to use octanol solubility as a surrogate to estimate stratu m corneum solubility. While this assumption may be debatable for stratum corneum, it is likely to apply to the hi ghly lipophilic PDMS membrane. By assuming that the log of IP M solubility is proportional to the log of polymer solubility (y log SIPM log SPDMS) and by keeping the other assumptions

PAGE 115

103 inherent in the Roberts-Sloan model, a si mplified model for flux through PDMS can be derived: log JPAQ = x + y log SIPM – z MW (6.14) Performing multiple linear regression on the flux through PDMS data using equation 6.14 resulted in the following best fit coefficients: log JPAQ = -1.206 + 1.00 log SIPM – 0.00758 MW (r2 = 0.617) (6.15) Although the fit of this mode l is no better, equation 615 justifies the assumption that IPM solubility and polyme r solubility mirror one another. A y coefficient of unity indicates that the best estima te of polymer flux is obtained if it is presumed that SIPM SPDMS. Therefore, if molecular weight effect s are reasonably small, a plot of log SIPM versus log JPAQ should produce a linear relationship. Figure 6-8 demonstrates the clear linear co rrelation between flux and solubility in IPM. Even without a correction for molecu lar weight, all but two of the prodrugs lie within +/-0.5 log units of the best-fit regression line shown. This is in spite of the fact that the molecular weight s of these compounds range from 151 to 352 amu. In comparison, the two outliers, the C4 and C5 bis-6,9-6 MP prodrugs, have molecular weights of 380 and 408, an increase of only 28 and 56 amu over the largest remaining database member, and yet they under perfor m by over 1.5 log units. Excluding the C4 and C5 bis 6,9-6-MP prodrugs from the database, improves the fit of the IPM based polymer model (6.15) and the Roberts-Sloan pol ymer model to approximately that of the aqueous mouse skin model (6-7). log JPAQ = -2.323 + 1.013 log SIPM – 0.00295 MW (r2 = 0.794) (6.16) log JPAQ = -2.855 + 0.902 log SIPM + (1-0.902) log SAQ – 0.00064 MW (r2 = 0.814) (6.17)

PAGE 116

104 -3.5 -3 -2.5 -2 -1.5 -1 -0.5 0 00.511.522.5 log SIPMExperimental log JPAQ Figure 6-8. Log IPM solubility versus l og maximum flux through PDMS membrane from water for the aqueous database -3.50 -3.00 -2.50 -2.00 -1.50 -1.00 -0.50 -3.50-3.00-2.50-2.00-1.50-1.00-0.50log JPAQ = -2.878 + 0.897 log SIPM + (1-0.897) log S A Q 0.00055 MWExperimental log JPAQ 5-FU, 6-MP and Th Prodrugs APAP Prodrugs Figure 6-9. Calculated vers us experimental log maximu m flux values through PDMS from water by Equation 6.17.

PAGE 117

105 Table 6-8. Calculated maximum log flux valu es through PDMS from water and errors of calculation for the chemically stable prodrug series. Compound ID Exp. log JPAQ a Calc. log JPAQ a (eq. 6.16) log JPAQ a Calc. log JPAQ (eq 6.17) log JPAQ a C1-ACOM-5-FU -2.64 -2.83 -0.19 -2.73 -0.09 C2 -1.78 -1.74 0.04 -1.69 0.09 C3 -1.6 -1.56 0.04 -1.53 0.07 C4 -1.7 -1.55 0.15 -1.55 0.15 C5 -1.58 -1.43 0.15 -1.46 0.12 C7 -1.82 -1.54 0.28 -1.62 0.21 C1-6-ACOM-6-MP -3.32 -2.96 0.36 -2.89 0.42 C2 -2.82 -2.66 0.17 -2.62 0.20 C3 -2.67 -2.54 0.13 -2.52 0.15 C4 -2.65 -2.47 0.18 -2.47 0.18 C5 -2.73 -2.57 0.16 -2.59 0.14 C1-6,9ACOM-6-MP -1.92 -2.46 -0.54 -2.35 -0.42 C2 -1.36 -1.73 -0.37 -1.66 -0.30 C3 -1.68 -1.38 0.30 -1.38 0.29 C4 -2.4 -1.17 1.23 -1.21 1.19 C5 -3.29 -1.80 1.49 -1.87 1.42 APAP,1 -2.69 -2.48 0.20 -2.50 0.18 C1, 2 -1.5 -1.85 -0.34 -1.90 -0.39 C2, 3 -1.74 -2.00 -0.26 -2.06 -0.32 C3, 4 -1.43 -1.63 -0.19 -1.72 -0.28 C4, 5 -1.79 -1.91 -0.12 -2.02 -0.23 C6, 6 -2.18 -1.91 0.25 -2.08 0.08 MeO-C2, 7 -1.85 -2.04 -0.19 -1.97 -0.13 MeO-C3i, 8 -2.38 -2.57 -0.17 -2.51 -0.10 a In units of mol cm-2 h-1 If the two largest outliers are excluded from database, the average absolute error of calculation for the remain ing compounds is 0.28 log units and 0.22 log units for the APAP prodrugs when calculated by equation 6.16 If equation 6.17 is used instead, the

PAGE 118

106 average absolute error of cal culation for the whole database and the APAP prodrug series is 0.21 log units. It is important to note th at while the Roberts-Sloan equation is capable of adequately predicting flux through polym er membrane, a reasonable prediction is possible without considering wate r solubility; even though water was used as the vehicle. This is quite different from flux thr ough mouse and human skin where accurate predictions require adequate c onsideration of water solubility. This is further evidence that the influence of aqueous solubility on fl ux is not a fictitious effect derived from an aqueous vehicle, but an a ttribute of mammalian skin. Unfortunately, it is still uncertain why a flux through homogeneous membrane like PDMS, which in theory should conform to the simplifying assumptions of the model far more closely than mouse skin, is not more accurately predicted. Central to this uncertainty is why the two largest prodrugs displayed such divergent flux behavior, although there are numerous mechanisms that would account for these observations. One possibility is that the dissolu tion rate of these compounds is much slower then the rate with which they diffuse into the polymer (dissolution-rate control). It is also possible (as suggested by Sloan to explain the unusual behavior of 6-MP) that these compounds may associate so strongly in solu tion that only a small the amount of monomer is present for diffusion. Yet another mechanism is that of static diffusion layer control. Flynn and Yalkowsky discussed the possibi lity of such a mechanism wh en attempting to explain the fluxes of large (>C5) n-alkyl-para-aminobenz oic acid (PABA) esters through silicone membrane. They reported that the time required by these compounds to reach steady state flux in diffusion cells wa s disproportionately high rela tive to the smaller series members. In addition, for the C6 PABA ester, flux through silicone membrane was

PAGE 119

107 independent of the membrane thickness. Given the low aqueous solubility of large PABA esters and their relatively high so lubility in silicone membrane, Flynn and Yalkowky argued that it was po ssible for the region of satura ted solution nearest to the silicone membrane to become depleted of PA BA ester. They hypothesized that this layer of depleted solution acted effectively as a second “membrane” which controlled flux through the silicone polymer. Although Flynn and Yalkowsky developed seve ral equations to relate flux to partition coefficient and alkyl chain length, they failed to investigate the most fundamental relationships between solubil ity, molecular weight and flux. Flynn and Yalkowski measured and reported the solubilities of the PABA esters in both silicone oil and hexane and considered th ese solubilities to be indica tive of solubility in PDMS polymer. But, they did not adequately accoun t for the effect of molecular weight on flux and in doing so may have drawn an erroneous conclusion. Table 6-9. Solubility, molecular weight and flux data for the PABA esters PABA Ester MW log SSO log SHEX log JPDMS C1 202 0.72 0.75 -2.61 C2 216 0.92 1.11 -2.20 C3 230 1.11 1.39 -2.11 C4 244 1.25 1.66 -2.14 C5 258 1.25 1.80 -2.50 C6 286 1.05 1.60 -2.96 C7 224 1.00 1.58 -3.59 a In units of mM. b In units of mol cm-2 h-1. By using this data, a Kastings type flux model can be developed for the first seven PABA esters.

PAGE 120

108 log JPDMS = -1.196 + 2.27 log SSO – 0.0195 MW (r2 = 0.958) (6.18) log JPDMS = 0.648 + 1.73 log SHEX – 0.0295 MW (r2 = 0.953) (6.19) Regardless of the whether silicone oil or he xane is used in the model, the model can accurately calculate the flux of the PABA esters. Such close agreement to a single membrane model and the strong dependen ce of flux through PDMS upon organic solubility indicates that a static diffusion laye r is not needed to desc ribe the behavior of this series. All of these previously described mechanis ms limit diffusion independently of the membrane. Therefore, if any of these mechan isms are responsible fo r the poor fit of the largest 6-MP prodrugs to the polymer model, then a similar deviation in their flux through hairless mouse skin should also be observed. However, the C4 and C5 bis-6,9ACOM-6-MP compounds only unde r perform from the aqueous mouse skin model by 0.23 log units. This is larger then the averag e absolute error for the database but smaller or equal to five other series members. Hence, it seems likely that an alternative mechanism is required. The most direct membrane based mechanism involves a disproportional increase in the energy requirement for the opening of accessible volumes of space within the polymer for larger molecules. The struct ure of PDMS is generally accepted as an amorphous polymer that has no regular lattic e or ordered spacing between the polymer chains. Unlike cavity formation in a small mo lecular liquid, which generally requires the translational movement of several molecules, diffusion in a polymer occurs when rotation of the polymer chains opens holes into whic h the drug molecules can move. The rate at which these rotations occur govern s the rate of diffusion. Th ere is a certainly a limit to

PAGE 121

109 the cavity size that can result from the rotation of a single polymer chain. It may be that molecules above ~360 amu exceed this limit and require the movement of more then one chain or require flexing as well as rotation to achieve a sufficiently large volume of open space. Prediction of Flux through Hairless Mouse Skin from Flux through PDMS The traditional view of skin permeation presumes that the principle barrier of the skin is purely lipophilic. This view has led some researchers to expect a direct correlation between flux through hydrophobic polymers and flux through skin. The success of these attempts was usually disa ppointing but a similar attempt with our database gave results that were unexpectedly good. Using the equation 3.34 for relating flux through similar membranes, the following model was obtained: log JMAQ = 0.396 + 0.526 log JPAQ – 0.00459 MW (r2 = 0.825) (6.20) Even though this is an overly si mple model, the average absolute log JMAQ was 0.17. This is a slight improvement over the results of the Roberts-Sloan model, but the improvement is too small to be considered significant. However, even though the database error is comparable, the average erro r in calculation of the APAP series is 0.25; more then twice that of the Roberts-Sloan model. In Chapter 3, it was asserted that a m odel which uses the flux through a synthetic membrane to estimate the diffusivity term (eq. 3.38) would be the most useful means to incorporate flux data from a synthetic memb rane into a predictive model for biological membranes. The equation of this form which best fits the aqueous database is shown in equation 6.21.

PAGE 122

110 log JMAQ = -1.841 + 0.305 log SIPM + 0.367 log JPAQ + (10.305 – 0.367) log SAQ (r2 = 0.830) (6.21) The correlation improves slightly over th at of the equation 6.20, but again the improvement is minimal. With an average prediction error of 0.16, model 6.21 performs only as well as model 6.20, though its ability to predict the behavior of the APAP compounds is much better. Unfortunately this improvement in pred iction for compounds 1 through 8 is balanced by poor prediction for both 6-MP based series, which have average prediction errors of 0.21. It is only when unconstrained coeffici ents are allowed into the model that the predictability improves above mediocre: log JMAQ = -1.471 + 0.115log SIPM + 0.323 log JPAQ + 0.289 log SAQ (r2 = 0.894) (6.22) With equation 6.22, the average absolute error of prediction drops to 0.13; which translates to an absolute prediction error of about 35% Given the inherent 30 % variability in mouse skin, this level of uncertainty is as low as is realistically obtainable. All series in the database are well predicte d with the bis-6,9-ACOM-6 MP series having the largest error (0.16) and 3-ACOM-5 FU se ries having the smallest error (0.10). The model even predicts the fluxes of C4 and C5 bis-6,9-ACOM-6 MP with errors of 0.12 and 0.11, respectively. As stated in chapter 3, the effects of solubility and molecular weight on flux, while related, will in general vary by di fferent proportions. Th erefore, flux through a surrogate membrane can typically be used to substitute for either solubility or molecular weight effects, but not both. However, with the current database it is possible for IPM solubility to be removed from the model withou t substantially affecting its performance:

PAGE 123

111 log JMAQ = -1.156 + 0.409 log JPAQ + 0.245 log SAQ (r2 = 0.887) (6.23) Equation 6.23 is the most efficient mo del for predicting flux from water through hairless mouse skin from flux through a polymer membrane. The average absolute prediction error for the database is 0.13 and the individual series average errors range from 0.11 to 0.16. -3.50 -3.00 -2.50 -2.00 -1.50 -1.00 -0.50 0.00 -3.50-3.00-2.50-2.00-1.50-1.00-0.500.00log JMAQ = -1.156 + 0.245 log SAQ + 0.409 log JPAQ Experimental log JMAQ 5-FU and 6-MP Prodrugs APAP Prodrugs Figure 6-6. Calculated vers us experimental log maximu m flux values through hairless mouse skin from water by equation 6.23. Yet again, the common result of equati ons 6.20 through 6.23 is that good biphasic solubility is the best indicator of high transdermal flux. Even with equation 6.23, which is well removed from the Roberts-Sloan e quation, flux is nearly equally improved by water and lipid solubility. While a molecular we ight term is not essential, it is clear that aqueous solubility cannot be excluded.

PAGE 124

112 CHAPTER 7 CONCLUSIONS AND FUTURE WORK There were three major objectives to th is work. The first objective was to synthesize and characterize a series of 4-alkyloxycarbonyl prodrugs of acetaminophen and examine the effectiveness of these compounds at improving the delivery of acetaminophen through hairless mouse skin from water and IPM. The second objective was to combine this new data with similar data collected from previously synthesized prodrug series of 5-FU, 6-MP and theophylline and use the resulting expanded databases to modify predictive models for topical de livery from both IPM and water. The final objective was to create a new database from flux measurements through silicone membrane of the hydrolytically stable prodrug s and use this data to develop a model to predict flux through mouse skin from flux th rough silicon membrane. Each of these objectives was achieved. Only one of the alkyloxycarbonyl pr odrugs of acetaminophen produced higher transdermal flux then acetaminophen itself. Howe ver, this is still encouraging. As with other hydrogen bonding functional groups, mask ing the phenolic moiety on a drug with the proper promoiety can result in a compound whose solubility char acteristics produce greater topical delivery. The fact that this was accomplished with acetaminophen, a compound whose relatively high water and lip id solubility are already conducive to topical delivery and therefore more difficu lt to improve, suggests that applying this approach to a polyphenolic compound will result in a greater relative improvement in flux.

PAGE 125

113 What remains to be demonstrated is whether masking only selected phenolic moieties on the polyphenol will be more benefici al then masking all phenolic groups. It has been well established that compounds, wh ich have good water and lipid solubility permeate the skin most effectively. In ge neral, masking hydroge n-bonding groups on the parent drug reduces the crystal lattice energy and results in higher lipid solubility for the prodrug. However, for water solubility to si multaneously increase, this drop in lattice energy must be higher then the increase in aqueous solvation energy resulting from any loss of favorable hydrogen bonding potential and the prodrug’s larger molecular volume. Using the general aqueous solubility equa tion presented in Chapter 3 (eq. 3.13), one can estimate the minimum loss in crystal la ttice energy that is required to compensate for the increase in molecular size in order to maintain aqueous solubility. For two consecutive members of a hypothetical homologous series (X and Y), separate aqueous solubility equations can be written: log SAQX = A – B0 ( TMX – T) – C MWX (7.1) log SAQY = A – B0 ( TMY – T) – C MWY (7.2) If X and Y have equal aqueous solubilitie s, then the following rearrangement of equations 7.1 from 7.2 can be performed. A – B0 (TMX – T) – C MWX = A – B0 (TMY – T) – C MWY -B0 (TMX – T) – C MWX = -B0 (TMY – T) – C MWY B0 (TMY – T) – B0 (TMX – T) = C MWX – C MWY B0 (TMY – TMX) = C (MWX – MWY) = -C(MWY – MWX) B0 TM = C MW TM = (C/ B0) MW

PAGE 126

114 If the values for C and B0, 0.0429 and 0.0176 respectively, are inserted then the result is TM = 2.44 MW (7.3) Since X and Y are sequential homologs, DMW equals 14. TM = 2.44 (14) = -34C In other words, for any pair of sequential alkyl homologs, the larger member will possess greater water solubility then the smaller member onl y if the melting point of the larger member is lower then that of the smaller member by at least 34C. This prediction is supported by an inves tigation of the database compounds. In the database, there are numerous examples of homologs having increased water solubility after the addition of a methylene group to th e promoiety. In each case, the larger homolog has a melting point at least 32 lower then the smaller homolog. For a tri-substituted prodrug, such as naringenin triacetate, the addition of only one methylene group to each promoiety would re sult in an overall increase of 42 amu in molecular weight. By equation 7.3, such an increase in molecular weight would require a drop in melting point of over 100C for water so lubility to increase. It is therefore not surprising that even though the triacetate of naringenin has a melting point of ~78, which is ~170C lower then the melting point of nari ngenin, it still has poor water solubility. Producing polyphenolic prodrugs with the highest topical de livery will most likely be accomplished by masking as few phenolic groups as possible to re duce crystal lattice energy. It was hypothesized that re placing an alkyl functional group with an ether in the promoiety would create more water-soluble prodrugs. Even though only a small amount

PAGE 127

115 of data was gathered from these types of pr omoieties, the experimental results do support this hypothesis. The methoxyethyl carbonate of APAP outperformed all other APAP prodrugs when delivered from water and a ll but the methyl carbonate when delivered from IPM. After accounting for the influe nce of molecular weight and melting point using eq. 3.14, the 4-MOAOC-APAP compounds demonstrated a greater intrinsic capacity for water solubility and a roughly equiva lent capacity for lipid solubility than the AOC-APAP compounds, which should result in hi gher flux. While it is still not certain whether these promoieties will consistently produce better performing prodrugs, they have performed well enough to warrant further study. The addition of APAP and its prodrugs to the IPM database reinforced the conclusions of the previous Roberts-Sloan mode l. After final regression of the expanded IPM database, flux from IPM remained equally dependent upon IPM and water solubilities. The flux of the APAP compounds were not predic ted as well as other series in the database, but well enough to demonstrat e that they do conform to the model. A similar result was obtained from the e xpanded aqueous database. The relative importance of water and IPM solubility on flux from water remained essentially unchanged after the addition of the APAP comp ounds. As with flux from IPM, flux from water was positively influenced by both aqueous and IPM solubility. However a slightly greater positive influence of lipid solubility and a slightly reduced positive influence of water solubility were observed for aqueous delivery. In contrast to IPM model, the aqueous model predicted the flux of the APAP compounds more accurately than any other series in the database.

PAGE 128

116 For a compound to have optimal transder mal flux, it must also have good biphasic solubility. The data of the expanded IP M and aqueous databases have given this principle increased experimental support. In addition, it is now certain that the importance of biphasic solubility is independent of the vehicle used and, therefore, is an attribute of the skin itself. As such, the coe fficients can be interpreted as representing the chemical nature of the skin barrier. It is noteworthy that in spite of the increased damage caused by IPM to the skin and the resultant > 10-fold increase in flux when IPM is used a vehicle compared to water, the corresponding co efficients of the two models remain close to one another. This is an indication that IPM alters the or ganization of the skin barrier more so then its composition. It would be interesting to see how other vehicles alter these coefficients. The Roberts-Sloan equation has been succe ssful in predicting the behavior of homologous series. The next challenge will be to apply this treatment to a database of heterogeneous compounds. This work is currently underway. By itself, PDMS membrane proved to be a poor experimental surrogate for mouse skin, which suggests it would be a poor surro gate for mammalian skin in general. Although polymer flux can yield a fair approx imation of the mechanical aspects of diffusion and estimate the effect of molecular we ight on skin flux, solubility in the purely lipophilic polymer does not conform to the solubility in the skin The simple addition of an aqueous solubility term, however, overcam e this deficiency and greatly improved the modelÂ’s predictability. For this limited databa se, the predictability of the polymer model was better then the aqueous Roberts-Sloan m odel. However, it will require a larger

PAGE 129

117 database to determine whether this improve ment actually indicates a better performing model. Both the polymer model and the Robert s-Sloan model provide evidence that, despite the skinÂ’s complexity, homogeneous membrane models can be effective at predicting flux through skin. The next stage in model development w ill be to investigate more complicated models that mimic the true construction of skin more closely. Roberts and Sloan have already publis hed a series/parallel model, which describes skin permeation as occurring through multiple paths that mirror the anatomical structures found in the skin. In this model, diffusi on can occur through lipid paths, aqueous paths or across alternating lipid and aqueous layers This model was developed from 41 of the IPM database compounds and it successfully predicted flux through mouse skin. Future modeling of the current IPM database and, when expanded, the aqueous database will involve applying these more sophisticated models.

PAGE 130

118 LIST OF REFERENCES Abraham MH, Le J. The correlation and pred iction of the solubility of compounds in water using an amended solvation en ergy relationship. J Pharm Sci. 1999; 88(9):868-80. Bundgaard H. Design of prodrugs: Bioreversi ble derivatives for various functional groups and chemical entities. In: Bundg aard H, editor. Design of prodrugs. New York: Elsevier; 1985:1-92 Bundgaard H, Rasmussen GJ. Prodrugs of peptides. 11. Chemical and enzymatic hydrolysis kinetics of N-acyloxymethy deri vatives of a peptide-like bond. Pharm Res 1991; 8:1238. .Burge S. Cohesion in the epidermis. Br J Dermatol. 1994; 131(2):153-9. Buur A, Bundgaard H. Prodrugs of 5-fl uorouracil I. Hydrolysis kinetics and physicochemical properties of various N-acy l derivatives of 5fluorouracil. Int J Pharm 1984; 21:349-64. Buur A, Bundgaard H. Prodrugs of 5-fluorour acil: V. 1-Alkyloxycarbonyl derivatives as potential prodrugforms for improved rectal or oral delivery of 5-fluorouracil. J Pharm Sci 1986; 75:522-27. Chikos JS, Nichols G. Simple Relationships for the estimation of melting temperatures of homologous series. J Chem Eng Data. 2001; 46:562-73. Cohen MH, Turnbull D, Molecu lar transport in liquids a nd gases. J Chem Phys. 1959; 31:1164-69. Crank J. The mathematics of diffusion. Ne w York: University Oxford Press; 1975.. Cronin MT, Dearden JC, Gupta R, Moss GP. An investigation of the mechanism of flux across polydimethylsiloxane membranes by use of quantitative structurepermeability relationships. J Pharm Pharmacol. 1998; 50(2):143-52. Cross SE, Anissimov YG, Magnusson BM, Roberts MS. Bovine-serum-albumincontaining receptor phase better predicts transdermal absorption parameters for lipophilic compounds. J Invest Derm atol. 2003; 120(4):589-91.

PAGE 131

119 Cross SE, Pugh WJ, Hadgraft J, Roberts MS. Probing the effect of vehicles on topical delivery: understanding the basic rela tionship between solvent and solute penetration using silicone membrane s. Pharm Res. 2001; 18(7):999-1005. Dittert LW, Caldwell HC, Adams HJ, Irw in GM, Swintosky JV. Acetaminophen prodrugs I. Synthesis, physicochemicalpr operties and analgesic activity. J Pharm Sci. 1963; 57:774–80. Dittert LW, Irwin GM, Chong CW Swintosky JV. Acetaminophe n prodrugs. II. Effect of structure and enzyme source on enzyma tic and nonenzymatic hydrolysis of carbonate esters. J Pharm Sci. 1968; 57(5):780-3. Doukas AG, Kollias N. Transdermal drug delive ry with a pressure wave. Adv Drug Deliv Rev. 2004; 56(5):559-79. Downing DT. Lipid and protein structures in the permeability barrier of mammalian epidermis. J Lipid Res. 1992; 33(3):301-13. Drustrup J, Fullerton A, Christrup L, Bundg aard H. Utilization of prodrugs to enhance the transdermal absorption of mo rphine. Int J Pharm 1991; 71:105–16. Du Plessis J, Pugh WJ, Judefeind A, Ha dgraft J. The effect of hydrogen bonding on diffusion across model membranes: cons ideration of the number of H-bonding groups. Eur J Pharm Sci. 2001; 13(2):135-41. E. Merck. Zur kenntnis der einwirkung von phosgen bxw chlorkohlensaure ester auf pacetylaminophenole und poxyphenylurethane. Chem Zentralbl. 1897; I:468–9. Eckert RL. Structure, function, and differentiation of the ke ratinocyte. Physiol Rev. 1989; 69(4):1316-46. Elias PM, Friend DS. The permeability barr ier in mammalian epidermis. J Cell Biol. 1975; 65(1):180-91. Fartasch M, Bassukas ID, Diepgen TL. Struct ural relationship between epidermal lipid lamellae, lamellar bodies and desmosomes in human epidermis: an ultrastructural study. Br J Dermatol. 1993;128(1):1-9. Geinoz S, Rey S, Boss G, Bunge AL, G uy RH, Carrupt PA, Reist M, Testa B. Quantitative structure-permeation relationships for solute transport across silicone membranes. Pharm Res. 2002; 19(11):1622-9. Greenwald RB, Conover CD, Choe YH. Poly (ethylene glycol) conjugated drugs and prodrugs: a comprehensive review. Cr it Rev Ther Drug Carrier Syst. 2000; 17(2):101-61.

PAGE 132

120 Guengerich FP, Kim DH. In vitro inhibiti on of dihydropyridine oxida tion and aflatoxin B1 activation in human liver microsom es by naringenin and other flavonoids. Carcinogenesis. 1990; 11(12):2275-9. Hansch L, Leo AJ, Substituent constants for correlation analysis in chemistry and biology. New York: Wiley; 1979. Hemingway JD, Molokhia MM. The dissolution of metallic nickel in artificial sweat. Contact Derm. 1987; 16(2):99-105. Hildebrand JH, Prausnitz JM, Scott RL. Regul ar and related solutions. New York :Van Nostrand-Reinhold;. 1970. Hildebrand, JH. The Solubility of NonElectrolytes. New York: Reinhold; 1936. Hseuh Y, Thewait J, Kitson N, Moore DJ. St ratum corneum lipid structure: insights from NMR and FTIR spectroscopic studies. In: Skin, hair and nails; Structure and function. New York. Marcel Decker Inc; 2004. Huibers PD, Katritzky AR. Correlation of the aqueous solubility of hydrocarbons and halogenated hydrocarbons with molecular structure. J Chem Inform. 1998; 38(2): 283-92. Iozzo RV. Matrix proteoglycans: from mol ecular design to cellula r function. Annu Rev Biochem. 1998; 67:609–52. Jorizzo JL, Carney PS, Ko WT, Robins P, Weinkle SH, Werschler WP. Fluorouracil 5% and 0.5% creams for the treatment of actinic keratosis: equivalent efficacy with a lower concentration and more conveni ent dosing schedule. Cutis. 2004; 74(6 Suppl):18-23. Kastings GB, Smith RL, Cooper ER. Effect of lipid solubility and molecular size on percutaneous absorption. Pharmacol Skin. 1987; 1:138-53. Kundu JK, Surh YJ. Molecular basis of chem oprevention by resver atrol: NF-kappaB and AP-1 as potential targets. Mutat Res. 2004; 555(1-2):65-80. Kunta JR, Sinko PJ. Intestinal drug transporte rs: in vivo function a nd clinical importance. Curr Drug Metab. 2004; 5(1):109-24. Lampe MA, Burlingame AL, Whitney J, Willia ms ML, Brown BE, Roitman E, Elias PM. Human stratum corneum lipids: characteriz ation and regional variations. J Lipid Res. 1983; 24(2):120-30. Laupeze B, Amiot L, Bertho N, Grosset JM, Lehn e G, Fauchet R, Fardel O. Differential expression of the efflux pumps P-glycoprot ein and multidrug resistance-associated protein in human monocyte-derived dendritic cells. Hum Immunol. 2001; 62(10):1073-80.

PAGE 133

121 Lavrijsen AP, Oestmann E, Hermans J, Bodde HE, Vermeer BJ, Ponec M. Barrier function parameters in various keratinizat ion disorders: transepidermal water loss and vascular response to he xyl nicotinate. Br J Dermatol. 1993; 129(5):547-53. Lin JK. Suppression of protei n kinase C and nuclear oncoge ne expression as possible action mechanisms of cancer chemoprevention by Curcumin. Arch Pharm Res. 2004; 27(7):683-92. Liu RH. Potential synergy of phytochemicals in cancer prevention: mechanism of action. J Nutr. 2004; 134(12 Suppl):3479S-3485S. Ljunggren B, Moller H., Topical use of fluorouracil in the treatment of psoriasis. Arch Dermatol. 1972; 106(2):263. Maitani Y, Shimada K, Nagai T. l-Menthol, oleic acid and lauricidin in absorption enhancement of free and sodium salt of di clofenac using ethanol treated silicone membrane as model for skin. Chem Pharm Bull (Tokyo). 1996; 44(2):403-8. Michels KB. The role of nutri tion in cancer development and prevention. Int J Cancer. 2005; 114(2):163-5. Mitragotri S, Kost J. Low-frequency sonophor esis: a review. Adv Drug Deliv Rev. 2004; 56(5):589-601. Norln L. Skin barrier structure and function: the single gel phase model. J Invest Derm. 2001; 118(5):899-901. Pauletti GM, Gangwar S, Siahaan TJ, Aub J, Borchardt RT. Improvement of oral peptide bioavailability: Peptidomimetics and prodrug strategies. Adv Drug Deliv Rev. 1997; 27(2-3):235-56 Pilgram GSK, Bouwstra JA. Stratum corneum lipid organization in vitro and in vivo as assessed by diffraction methods. In: Skin, ha ir and nails: struct ure and function. New York: Marcel Decker Inc; 2004. Prausnitz MR. Microneedles for transdermal drug delivery. Adv Drug Deliv Rev. 2004; 56(5):581-7. Ran Y, Jain N, Yalkowsky SH. Prediction of aqueous solubility of organic compounds by the general solubility equation (GSE). Ch em Inf Comput Sci. 200; 41(5):1208-17. Riviere JE, Heit MC. Electrically-assisted tr ansdermal drug delivery. Pharm Res. 1997; 14(6):687-97. Roberts MS, Anderson RA, Swarbrick J, M oore DE. The percutaneous absorption of phenolic compounds: the mechan ism of diffusion across the stratum corneum. J Pharm Pharmacol. 1978; 30(8):486-90.

PAGE 134

122 Roberts WJ, Sloan KB. Application of the transformed Potts-Guy equation to in vivo human skin data. J Pharm Sci. 2001; 90(9):1318-23. Roberts WJ, Sloan KB. Correlation of aque ous and lipid solubilities with flux for prodrugs of 5-fluorouracil, theophylline, and 6-mercaptopurine: a Potts-Guy approach. J Pharm Sci. 1999: 88(5):515-22. Schaefer H, Redelmeier TE. Skin barrier: principles of percutaneous absorption. New York: S. Karger AG Publishers; 1996. Scheuplein RJ, Blank IH. Mechanism of perc utaneous absorption IV: Penetration of nonelectrolytes (alcohols) from aqueous solu tions and from pure liquid. J Invest Dermatol 1973; 60:286-296. Scheuplein RJ, Blank IH. Permeability of th e skin. Physiol Rev. 1971; 51(4):702-47. Sloan KB. Functional group considerations in the development of prodrug approaches to solving topical delivery problems. In: Sloan KB, editor. Prodrugs: topical and ocular drug delivery. New York: Marcel Dekker; 1992. Sloan KB, Beall HD, Weimar WR, Villane uva R. The effect of receptor phase composition on the permeability of hairless mouse skin in diffusion cell experiments. Int J Pharm. 1991; 73(2); 97-104. Sloan KB, Getz JJ, Beall HD, Prankerd RJ. Transdermal delivery of 5-fluorouracil (5FU) through hairless mouse skin by 1-alkylaminocarbonyl-5-FU prodrugs: physicochemical characterization of prodr ugs and correlations with transdermal delivery. Int J Pharm.1993; 93:27-36. Sloan KB, Taylor HE, Hamilton JC. Alcohol flux and effect on the delivery of theophylline from propylene glycol. Int J Pharm 1997; 156:17-26. Sloan KB, Wasdo SC. Designing for topical deliv ery: prodrugs can make the difference. Med Res Rev. 2003; 23(6):763-93. Sloan KB, Wasdo SC, Ezike-Mkparu U, Murray T, Nickels D, Singh S, Shanks T, Tovar J, Ulmer K, Waranis R. Topical delivery of 5-fluorouracil and 6-mercaptopurine by their alkylcarbonyloxymethyl prodrugs from water: vehicle effects on design of prodrugs. Pharm Res. 2003; 20(4):639-45. Sparr E, Engstrom S. Lipid phase behavior; a basis for the unders tanding of membrane structure and function. In: Skin, hair and nails: structure and function. New York: Marcel Decker Inc. 2004. Steintert PM, Freedberg IM. Epidermal structur al proteins. In: Goldsmith L.A., editor Physiology, biochemistry and mo lecular biology of the skin, 2nd Ed. New York: New York Unversity Press; 1991.

PAGE 135

123 Stella VJ, Martodihardjo S, Terada K, Rao VM. Some relationships between the physical properties of various 3-acyloxymethyl prodrugs of phenytoin to st ructure: potential in vivo performance implications J Pharm Sci. 1999; 88(10):1100-4. Steven AC, Bisher ME, Roop DR, Steinert PM Biosynthetic pathways of filaggrin and loricrin—two major proteins expressed by terminally differentiated epidermal keratinocytes. J Struct Biol, 1990; 104 (1-3):150-162. Stinchcomb AL, Paliwal A, Dua R, Imot o H, Woodard RW, Flynn GL. Permeation of buprenorphine and its 3-alkylester prodr ugs through human skin. Pharm. Res. 1996; 13:1519–1523. Stinchcomb AL, Swaan P, Ekabo O, Harri s KK, Browe J, Hammell PD, Cooperman PA, Pearsall M. Straight-chain naltrexone ester prodrugs: diffusion and concurrent esterase biotransformation in human skin. J Pharm Sci. 2002; 91:2571–2578. Sung KC, Han HY, Hu OYP, Hsu LR. Controlle d release of nalbuphine prodrugs from biodegradable polymeric ma trices: influences of prodrug hydrophilicity and polymer composition. Int J Pharm. 1998; 172:17–25. Taylor HE, Sloan KB. 1-Alkylcarbonyloxym ethyl prodrugs of 5fluorouracil (5-FU): synthesis, physicochemical properties, and topical delivery of 5-FU. J Pharm Sci. 1998; 87(1):15-20. Thibodeau GA, Patton KT. Anatomy and Physiology 5th Ed. Philadelphia: Mosby Publishing; 2003. Twist JN, Zatz JL. A model for alcohol-enhanced permeation through polydimethylsiloxane membranes. J Pharm Sci. 1990; 79(1):28-31. Ushiki T. Collagen fibers, reticular fibe rs and elastic fibers. A comprehensive understanding from a morphological vi ewpoint. Arch Histol Cytol. 2002; 65(2):109-26. Vinson JA. Flavonoids in foods as in vitro and in vivo antioxidants. Adv Exp Med Biol. 1998; 439:151-64. Wang ZY, Khan WA, Bickers DR Mukhtar H. Protection agai nst polycyclic aromatic hydrocarbon-induced skin tumor initia tion in mice by green tea polyphenols. Carcinogenesis. 1989; 10(2):411-5. Waranis RP, Sloan KB. Effects of vehicles and prodrug properties a nd their interactions on the delivery of 6-mercaptopurin e through skin: S6acyloxymethyl-6mercaptopurine prodrugs. J Pharm Sci. 1988; 77(3):210-5. Watt FM. Terminal differentiation of epider mal keratinocytes. Curr Opin Cell Biol. 1989; 1(6):1107-15.

PAGE 136

124 Wertz PW, Madison KC, Downing DT. C ovalently bound lipids of human stratum corneum. J Invest Dermatol. 1989; 92(1):109-11. Wrighton SA, Stevens JC. The human hepa tic cytochromes P450 involved in drug metabolism. Crit Rev Toxi col. 1992; 22(1):1-21. Wu G, Meininger CJ. Regulation of nitric oxide synthesis by diet ary factors. Annu Rev Nutr. 2002; 22:61-86. Xiang T, Xu Y, Anderson BD. The barrier domai n for solute permeation varies with lipid bilayer phase structure. J Membr Biol. 1998; 165(1):77-90. Yalkowsky SH, Valvani SC. Solubility and partit ioning I: Solubility of nonelectrolytes in water. J Pharm Sci. 1980; 69(8):912-22. Yang CS, Landau JM, Huang MT, Newmark HL. Inhibition of carcinogenesis by dietary polyphenolic compounds. Annu Rev Nutr. 2001; 21:381-406.

PAGE 137

125 BIOGRAPHICAL SKETCH Scott C. Wasdo was born in Waterbury, Connecticut, on July 4, 1965. At the age of five, he and his family moved to Roches ter, New York, where he lived until his graduation from Penfield High School in June of 1983. He attended the University of Hawaii at Manoa until January of 1985, when he transferred to the Univ ersity of Florida. After graduating with a B.S. degree in chemis try in 1987, he worked as an analytical chemist until 1998, when he returned to gra duate school in the Department of Medicinal Chemistry.