Magnet-Directed Bioadhesive Nanoparticles for Localized Oral Delivery

Home , Bolus (medicine), Lottia limatula

Magnet-Directed Bioadhesive Nanoparticles

for Localized Oral Delivery

Bryan Laulicht

B.A. Columbia University, 2005

Submitted in partial fulfillment of the

Requirements for the degree of Doctor of Philosophy

In the Division of Biology and Medicine at Brown University

Providence, Rhode Island

May 2010

This dissertation by Bryan Laulicht is accepted in its present form by

the Division of Biology and Medicine as satisfying the dissertation requirement

for the degree of Doctor of Philosophy.

Date ______

Edith Mathiowitz, Ph.D., Director

Recommended to the Graduate Council

Date ______Anubhav Tripathi, Ph.D., Co-Advior

Date ______Diane Hoffman-Kim, Ph.D., Reader

Date ______Jeffrey Morgan, Ph.D., Reader

Date ______Solomon Steiner, Ph.D., External Reader

Approved by the Graduate Council

Date ______Sheila Bonde Dean of the Graduate School

iii

Curriculum Vitae

Bryan Laulicht

Date of Birth: August 26, 1981

Place of Birth: New York, NY

EDUCATION

Brown University, Providence, RI Expected May, 2010

PhD, Medical Science . Program in Artificial Organs, Biomaterials, and

Cellular Technology, Department of Molecular Pharmacology,

Physiology, and Biotechnology

Advisors: Edith Mathiowitz, PhD and Anubhav Tripathi, PhD

Thesis: Magnet-Directed Bioadhesive Nanoparticles for Localized Oral Delivery

Columbia University, New York, NY May, 2005

BA, Biophysics.

Isidor Isaac Rabi Science Research Scholar

iv RESEARCH EXPERIENCES

Perosphere Biopharmaceuticals , Providence, RI 2009-Present

Optimize bioadhesive polymer nanosphere formulations to maximize uptake

Formulate oral protein delivery systems

Biodel Inc. , Providence, RI 2008-Present

Analyze and optimize long-acting insulin formulations

Brown University , Providence, RI 2005-Present

Develop magnetically retentive drug delivery systems

Characterize the safety of magnetically-retained pills in vivo using

biplanar fluoroscopy

Determine the effectiveness of magnetic retention in the small

intestines in a small animal model

Develop nanosphere formulations to improve uptake of narrow

absorption window therapeutics (e.g. Lasix for congestive heart

failure)

Engineered Release Systems Inc. , Rensselaer, NY 2003-Present

Develop bioinspired surgical materials

Design and produce MEMS systems for single-cell microencapsulation

HID International , Providence, RI 2008

Test the effects of mechanically induced birefringence on the

mechanical properties of biodegradable smart cards

v Freedom-2 Inc. , Providence, RI 2007-2008

Development of encapsulated inks for use in permanent removable tattoos

Scale-up process to pilot scale

Scale-up and manufacturing of dermal filler

Center for Nanoscale Science and Engineering , Albany, NY 2004-2005

Collaborative Effort with Columbia University Microelectronics

Sciences Lab to investigate the mechanism of implantation-based lift-

off in Helium implanted Lithium Niobate thin films

Columbia University , New York, NY 2002-2004

Investigate the mechanism of giant uni-lamellar vesicle formation

Design and construct a high pressure nano-perfusion bioreactor for articular

cartilage permeability testing

Weizmann Institute of Science , Rehovot, Israel 1999

Synthesized siderophores for use as an antibiotic delivery system to combat

drug resistant infections

Michigan State University , East Lansing, MI 1998-9

Investigated the behavior of vortex rings propagating through tubes of varying

diameters using laser induced fluorescence

Cold Spring Harbor Laboratories , Cold Spring Harbor, NY 1996-8

Identified and genotypically characterized a female sterile mutant Arabidopsis

thaliana that had been induced by DS tag insertion for the Arabidopsis Genome

Project

vi PUBLICATIONS

1. Bryan Laulicht, Anubhav Tripathi, Vincent Schlageter, Pavel Kucera, Edith

Mathiowitz, “Understanding Gastric Forces Calculated from High Resolution Pill

Tracking Data” PNAS April 19, 2010, 10.1073/pnas.1002292107.

2. Bryan Laulicht, Peter M. Cheifetz, Anubhav Tripathi, Edith Mathiowitz, “Are in vivo

gastric bioadhesive forces accurately reflected by in vitro experiments?” Journal of

Controlled Release 2009, 134(2), 103-110.

3. Bryan Laulicht, Peter Cheifetz, Edith Mathiowitz, Anubhav Tripathi, “Evaluation of

Continuous Flow Nanosphere Formation by Controlled Microfluidic Transport,”

Langmuir 2008, 24 (17), 9717-26.

4. Ryan M. Roth, Djordje Djukic, Yoo Seung Lee, Richard M. Osgood, Sasha Bakhru,

Bryan Laulicht, Kathleen Dunn, Hassaram Bakhru, Liqi Wu, Mengbing Huang,

“Compositional and structural changes in LiNbO3 following deep He+ ion

implantation for film exfoliation,” Applied Physics Letters 2006, 89(11)112906,1-3.

5. Joshua Reineke, Yu-Ting Liu, Daniel Cho, A. Peter Morello III, Bryan Laulicht, Edith

Mathiowitz, “Mechanisms of Polymer Microsphere Uptake Following Oral Delivery”

in process.

6. Bryan Laulicht, Nicholas J Gidmark, Anubhav Tripathi, Edith Mathiowitz, “A New

Method for Improving Localized Delivery from Magnetic Pills” in process.

7. Bryan Laulicht, Alexis Mancini, Nathanael Geman, Anubhav Tripathi, Edith

Mathiowitz, “Bioinspired Synthetic Bioadhesive Polymers,” in process.

vii 8. Bryan Laulicht, Anubhav Tripathi, Edith Mathiowitz, “Bioactivity Optimization of

Furosemide Nanospheres” in process.

9. Peter M. Cheifetz, Sarah Rose, Elaine Kim, Jill Javier, Bryan E. Laulicht, Haitao Qian,

Jules Jacob, Avinash Nangia, Anubhav Tripathi, Edith Mathiowitz, “Development of

New Artificial Tissue Substrate for Bioadhesion Testing,” in process.

PATENTS

1. Bryan Laulicht, Edith Mathiowitz. “Magnet-Retained Localized Oral Drug Delivery

Systems,” invention disclosure submitted.

2. Edith Mathiowitz, Arthur Peter Morello, Joshua Reineke, Bryan Laulicht, Peter

Cheifetz. “Drug Delivery Formulations for Targeted Delivery,” Patent application

US2008193543 , WO2006125074, filed May 17, 2006.

3. Bryan Laulicht, Sasha Bakhru. “Chemically Cross-linked Elastomeric Microcapsules,”

US2006/027163, WO/2007/009023.

4. Bryan Laulicht, Sasha Bakhru. “Polymer-Based Microstructures,” Patent application

US2004/036158, WO/2005/041884, filed October 29, 2004.

TEACHING EXPERIENCE

Teaching Assistantships, Brown University

Principles of Experimental Surgery Spring 2007

Biotechnology in Medicine Fall 2006

viii Invited Lectures, Brown University

Bioadhesion, Drug and Gene Delivery Fall 2009

Co-Instructor of Drug and Gene Delivery with Professor Edith Mathiowitz Fall 2008

Bioadhesive and Bioerodible Drug Delivery, Polymer Science for Biomaterials Fall 2008

Graduate Education in the Biomedical Sciences, Introduction to Biotechnology Fall 2008

PRESENTATIONS/ABSTRACTS

1. Bryan Laulicht, Peter Cheifetz, Edith Mathiowitz, Anubhav Tripathi, “Are in vivo

gastric bioadhesive forces accurately reflected by in vitro experiments?” Controlled

Release Society Annual Meeting, Copenhagen, Denmark, July 18-22, 2009.

2. Bryan Laulicht, Peter Cheifetz, Edith Mathiowitz, Anubhav Tripathi, “Evaluation of

Continuous Flow Nanosphere Formation by Controlled Microfluidic Transport,”

AIChE Annual Meeting, Philadelphia, PA, November 16-21, 2008.

3. Cartney Smith, Peter Cheifetz, Bryan Laulicht, Edith Mathiowitz, Anubhav Tripathi,

“Targeting Precise Nanoencapsulation by Controlled Microfluidic Transport,”

Showcase of Nanomedicine, Providence, RI, May 24, 2006.

4. Bryan Laulicht, Doglas Bohl, Manoochehr Koochesfahani, “Vortex Ring in a Tube”

American Physical Society Division of Fluid Dynamics Annual Meeting, Philadelphia,

PA, November 22-4, 1998.

ix RESEARCH AWARDS

Columbia University Isidor Isaac Rabi Science Research Scholar 1999-2005

First Step to the Nobel Prize in Physics, Grand Prize Recipient 1999

Intel Science Talent Search Semifinalist 1999

International Science and Engineering Fair, 2 nd Place in Engineering 1999

University of Pennsylvania Roy and Diana Vagelos Scholar (declined) 1999

Hunter R. Rawlings III Cornell Presidential Research Scholar (declined) 1999

CONSULTING

Franz Cell Optimization for Transdermal Drug Delivery, Isis Biopolymer Spring 2010

Characterization of Coated Microparticles, Panacos Pharmaceuticals Fall 2009

Cellulose-Derivative Gel Viscosity Determination, GelMed Sciences Fall 2008

PROFESSIONAL ASSOCIATIONS

American Association for the Advancement of Science

Controlled Release Society

American Institute of Chemical Engineers

Society for Bioengineers

Sigma Xi

Acknowledgements

I wish to express my deepest gratitude to my thesis advisors, Professors Edith Mathiowitz and Anubhav Tripathi, for their constant encouragement, invaluable guidance, and unending enthusiasm about our work. I also thank my thesis committee – Professors

Diane Hoffman-Kim, Jeffrey Morgan, and Solomon Steiner – for their time and most helpful advice.

I similarly extend heartfelt thanks to my collaborators and former labmates at Brown; their friendship and support made my graduate experience exciting and fruitful, and for this I will be forever grateful. In particular, I thank Dr. Peter Cheifetz for acclimating me to the Mathiowitz laboratory and for his collaborative efforts on characterizing bioadhesives and fabricating nanospheres. I also thank Dr. A. Peter Morello III for teaching me about polymer characterization and for all of his insightful and exciting discussions. I thank Professor Joshua Reineke for all of his work and discussions regarding the links among particle size, bioadhesion, and uptake, and thank Dr. Ana

Jaklenec and Dr. Michael Harrison for their thoughtful discussions about polymer chemistry and drug delivery. I would like to thank Dr. Haitao Qian for his dedication to

xi creating new polymers, and would like to thank Dr. Stacia Furtado for all of her support and help, especially with NMR, first as a fellow student and then as an investigator. I would also like to thank Dr. Jinkee Lee and Dr. Matthew Kerby for all of their help and patient support working with glass chip microfluidics.

I would like to thank my current labmates, Christopher Baker, Daniel Cho, Danya

Decoteau, Roshni Patel, Stephanie Angione, Glareh Azadi, Elejdis Kulla, Stephanie

McCalla, Kenneth Morabito, and Leah Seward for their friendship and creating a wonderful work environment. In particular I would like to thank Chris for his collaborative work on characterizing induced birefringence in thermoplastic polymers and Dan for his collaborative work involving uptake of bioadhesive nanospheres.

I would like to thank Dr. Vincent Schlageter and Professor Pavel Kucrea for their collaborative work on modeling the forces experienced by pills. I would also like to thank

Professor Elizabeth Brainerd and Nicholas Gidmark for teaching me about biplanar videofluoroscopy and collaborating with me on visualizing and characterizing localization of magnetic oral dosages I would also like to thank Dr. Timothy Murphy and

Dr. Felix Shvartsman for initiating our collaborations with Sentient and HID respectively.

Many individuals contributed significantly to the experimental work described in this dissertation; to these individuals I am most grateful. In particular, I thank Roxanne Burrill for all of her support and help with everything ranging from animal care to laboratory techniques. I would also like to thank Veronica Budz and Anne Beauregard-Young from

xii animal care facilities for all of their support. I also thank Paula Weston for her help with histology and Geoffrey Williams for his support and help with microscopy. I would like to thank Kenneth Talbot and Tim Pimentel for their help with and advice on machining, and thank Dr. Russell Hopson for all of his help with NMR characterization. I would also like to thank Dr. Edward Walsh, Dr. Michael Worden, Lynn Fanella and Erika Nixon for all of their invaluable help and guidance with MRI. I thank Dr. James Clifton for his help with circular dicrhoism spectroscopy, and Marc Johnson for his help with Texture

Exponent programming.

I was also incredibly fortunate to work with a fantastic group of Brown undergraduates who both contributed significantly to the work in our lab and inspired me to pursue a career in academia. These students include Cartney Smith for his work with microfluidics; Sarah Rose, Elaine Kim, Stephen Ting, Alisha Ranadive, Alexis Mancini, and Nathanael Geman for their work on characterizing bioadhesives; and the many others with whom I interacted in the lab and the classroom over the past years.

I would like to thank my partners in Experimental Surgery and Human Anatomy, Andrew

Kim, Robert Kambic, Henry Astley, and Jorn Cheney for their patience, work, and friendship.

I would like to thank all of my professors, especially Moses Goddard, Jim Harper, and

Michael Lysaght, for whom I served as a teaching assistant, for all they have taught me in the classroom and by example. Additionally, I would like to thank Professor Jay Tang for

xiii all of his discussions about biophysics and Professors Dale Ritter, Stephen Gatesy, and

Thomas Roberts for their instruction and patient guidance in Anatomy.

I would like to thank Carol Folan for her friendship and support throughout my graduate work and Loretta Burns, Cheryl Parisseau, and Monique Victor for all of their support and help.

I would like to thank Brown University, Freedom-2, Human Interface Devices, Biodel, and Sentient for their generous support and inspiring learning experiences that they have provided. At Biodel I would like to thank Nandini Kashyap for most interesting discussions about protein science and product formulation. I would also like to thank Kay

Balun for her help with coordinating my work with Biodel.

Lastly, I wish to express my most sincere, heartfelt thanks to my family. I thank my sister

Freda for always being there for me. And I would like to express my deepest gratitude to my parents Joyce and Don, and grandparents Rose, Phil, and Mildred for always supporting, encouraging, and enabling me to follow my dreams – without all of you, none of this would have been possible.

xiv

Dedication

I dedicate this dissertation to my family.

Contents

LIST OF TABLES ...... xix

LIST OF FIGURES ...... xx

1. INTRODUCTION AND MOTIVATION...... 1

1.1 THE ROLE OF PARTICLE SIZE IN NANOSPHERE UPTAKE ...... 1

1.2 THE ROLE OF BIOADHESION IN NANOSPHERE UPTAKE ...... 4

1.3 THE ROLE OF ADMINISTRATION LOCATION IN NANOSPHERE

UPTAKE ...... 5

1.4 METHODS AND MATERIALS SUMMARY ...... 9

1.5 HYPOTHESIS AND SPECIFIC AIMS ...... 10

1.6 REFERENCES ...... 17

2. EVALUATION OF CONTINUOUS FLOW NANOSPHERE FORMATION BY

CONTROLLED MICROFLUIDIC TRANSPORT ...... 22

2.1 BACKGROUND AND MOTIVATION ...... 23

2.2 EXPERIMENTAL METHODS...... 27

2.3 RESULTS AND DISCUSSION ...... 32

2.4 CONCLUSIONS ...... 40

2.5 REFERENCES ...... 52

xvi 3. DIURETIC BIOACTIVITY OPTIMIZATION OF FUROSEMIDE...... 56

3.1 BACKGROUND AND INTRODUCTION ...... 57

3.2 MATERIALS AND METHODS ...... 60

3.3 RESULTS AND DISCUSSION ...... 63

3.4 CONCLUSIONS ...... 70

3.5 REFERENCES ...... 77

4. ARE IN VIVO GASTRIC BIOADHESIVE FORCES ACCURATELY

REFLECTED BY IN VITRO EXPERIMENTS? ...... 81

4.1 INTRODUCTION ...... 82

4.2 MATERIALS AND METHODS ...... 85

4.3 RESULTS AND DISCUSSION ...... 94

4.4 CONCLUSIONS ...... 102

4.5 REFERENCES ...... 111

5. BIOINSPIRED SYNTHETIC BIOADHESIVE POLYMERS...... 116

5.1 INTRODUCTION ...... 117

5.2 MATERIALS AND METHODS ...... 119

5.3 RESULTS AND DISCUSSION ...... 123

5.4 CONCLUSIONS ...... 132

5.5 REFERENCES ...... 140

xvii 6. UNDERSTANDING GASTRIC FORCES CALCULATED FROM HIGH

RESOLUTION PILL TRACKING...... 144

6.1 BACKGROUND AND INTRODUCTION ...... 145

6.2 RESULTS AND DISCUSSION ...... 149

6.3 CONCLUSIONS AND PERSPECTIVES ...... 159

6.4 PATIENTS AND METHODS ...... 160

6.5 REFERENCES ...... 168

7. NOVEL METHOD FOR LOCALIZED DELIVERY FROM MAGNETIC PILLS

...... 172

7.1 INTRODUCTION ...... 173

7.2 RESULTS AND DISCUSSION ...... 174

7.3 CONCLUSIONS ...... 178

7.4 MATERIALS AND METHODS ...... 179

7.5 REFERENCES ...... 188

8. CONCLUSIONS AND FUTURE DIRECTIONS...... 191

8.1 REFERENCES ...... 196

xviii

List of Tables

1.1 Organ disposition kinetics...... 16

2.1 Population characteristics of continuous flow produced nanospheres ...... 51

5.1 Bioinspired bioadhesive side chain attachment efficiencies...... 136

6.1 Area normalized inter-speices comparison of gastric emptying forces and torques 167

xix

List of Figures

Introduction

1.1 Nanosphere uptake as a function of diameter...... 12

1.2 Nanosphere uptake as a function of bioadhesiveness...... 13

1.3 Bioadhesive fracture strength comparison...... 14

1.4 Nanosphere uptake as a function of location of administration...... 15

Chapter 2

2.1 Continuous flow phase inversion nanospheres...... 43

2.2 Sizing of polymer nanospheres...... 44

2.3 off chip Batch produced phase inversion nanospheres...... 45

2.4 Continuous flow phase inversion nanospheres...... 46

2.5 Size histograms of flow pinching continuous flow phase inversion nanospheres...... 47

2.6 Size histograms of continuous flow phase inversion nanospheres without flow pinching...... 48

2.7 Nanosphere mean diameter as a function of polymer concentration...... 49

xx 2.8 Size histograms and mean diameter comparison of continuous flow phase inversion nanospheres produced using glass capillary tubes...... 50

Chapter 3

3.1 Characterization of oral furosemide doses...... 71

3.2 Metabolic cage setup...... 72

3.3 Furosemide 2.5 mg/kg oral dose comparison ...... 73

3.4 Furosemide 5 mg/kg oral dose comparison ...... 74

3.5 Furosemide 10 mg/kg oral dose comparison ...... 75

3.6 Bioactivity profiles of optimized furosemide oral formulations...... 76

Chapter 4

4.1 Bioadhesion setup, in vivo and in vitro ...... 105

4.2 Calculation of cross-sectional contact area...... 106

4.3 Hold time optimization ...... 107

4.4 in vivo bioadhesive fracture strength comparison amongst bioerodible polymers... 108

4.5 in vitro bioadhesive fracture strength comparison amongst bioerodible polymers .. 109

4.6 in vitro/in vivo comparison of bioadhesive fracture strengths of bioerodible polymers

...... 110

Chapter 5

5.1 Chemical structures of bioinspired polymers tested for bioadhesive properties ...... 133

5.2 Chemical analysis of synthetic bioinspired bioadhesives...... 134

xxi 5.3 HSQC NMR phase analysis of bioinspired bioadhesives...... 135

5.4 Bioadhesion testing setup ...... 137

5.5 Bioadhesive properties of PBMA-derivative polymers...... 138

5.6 Bioadhesive properties of PEMA-derivative polymers ...... 139

Chapter 6

6.1 Net force and trajectory plots of pills in human stomachs...... 163

6.2 Histograms and mean gastric emptying forces experienced by pills in the human

stomach ...... 164

6.3 Histograms and mean gastric emptying forces experienced by pills in the canine

stomach ...... 165

6.4 Histograms and mean gastric emptying forces experienced by pills in the rat stomach

...... 166

Chapter 7

7.1 Biplanar videofluoroscopic tracking of magnetically retained model pills in vivo .. 184

7.2 Confirmation of magnetic capture by x-ray, of in vitro force measurement in vivo, and of the force exerted by the internal magnet on underlying tissue ...... 185

7.3 Photograph of the magnetic oral dosage...... 186

7.4 Histological analysis of intestinal tissue before and after magnetic localization ..... 187

xxii

Chapter 1

Introduction and Motivation

Nanosphere-based drug delivery systems demonstrate great potential for oral protein administration. However, nanosphere uptake remains the primary obstacle. Size greatly influences the uptake and resultant organ distribution of polymer nanospheres, as shown in Figure 1.1. Additionally, we observe a correlation between bioadhesiveness and uptake indicating that utilizing bioadhesive polymers can increase microsphere uptake by absorptive epithelia from 5.8 ± 1.9% to 66.9 ± 12.9%, as shown in Figure 1.2. The

location of administration within the gastrointestinal tract also influences the uptake and

biodistribution or polymer nanospheres, particularly the presence or absence of Peyer’s

Patches in the dosing region, as seen in Figure 1.4.

1.1 The role of particle size in nanosphere uptake

Oral delivery of proteins is the Holy Grail of drug delivery. Our group approaches oral

protein delivery using Phase Inversion Nanoencapsulation (PIN), an emulsion-free

nanoencapsulation technique [1]. When delivering a therapeutic protein or biologic

orally, every incremental increase in uptake advances the technology towards

1 implementation in widespread clinical practice. Since oral administration of therapeutic proteins and biologics alone typically exhibit negligible gastrointestinal (GI) absorption, numerous strategies have been devised to aid in uptake and reduce denaturation by the GI tract including permeation enhancers, enteric coatings, retentive devices, and polymer encapsulation [1-4]. Polymer nanoencapsulation both protects and promotes the uptake of the encapsulated therapeutics when the spheres are small enough to achieve cellular uptake [1].

Phase inversion is commonly used in the production of polymer dialysis membranes [5].

Under the appropriate conditions, phase inversion can be used to form discrete polymer nanospheres rather than an interconnected polymer network, termed Phase Inversion

Nanoencapsulation (PIN) [6]. PIN does not require impeller mixing and is performed entirely in organic solvents enabling protein encapsulation with a highly bioactive yield by the avoidance of high sheer and oil/water interfaces that can denature proteins [6-10].

To investigate PIN under highly controlled laminar flow conditions, we performed continuous flow PIN on both microfluidic and glass capillary tube platforms, as presented in Chapter 2 [11]. Previous microfluidic methods of nanosphere production rely on emulsification or Rayleigh instability to generate discrete droplets that can then be chemically or physically cross-linked. We developed an emulsion-free, continuous flow

PIN process for producing monodisperse populations of polymer nanospheres in the size range desirable for achieving uptake via the oral route, described in Chapter 2,

“Evaluation of Continuous Flow Nanosphere Formation by Controlled Microfluidic

Transport”.

2 Chapter 2 presents new modes of continuously producing polymer nanospheres within the optimal size range for uptake-based oral drug delivery, 500 nm [1,9,12,13].

Moreover, nanosphere uptake is primarily governed by the properties of the nanosphere rather than the encapsulated therapeutic. Therefore nanosphere-based delivery systems potentiate the development of platform technologies that can deliver a wide range of therapeutics including those that exhibit poor absorption, limited water solubility, or are degraded by GI secretions.

We also employed phase inversion to micronize furosemide, a hydrophobic diuretic commonly used to treat congestive heart failure (CHF) [14]. Clinical trials using intra venous administration to compare bolus against controlled introduction of furosemide found improved diuretic efficiency and decreased hospitalization incidence when furosemide was introduced in a controlled release formulation [14,15]. More commonly, in less severe CHF cases, furosemide is administered in pill form [16]; however, furosemide is absorbed only in the proximal gastrointestinal tract where pH conditions are favorable challenging the development of controlled release oral formulations [17].

Numerous strategies have been developed and tested to prolong retention of furosemide in the stomach to enable controlled release, without clinical success [18]. By both micronizing furosemide using phase inversion and adding pH-altering bioadhesive poly(acrylic acid)-derived polymers, we have optimized the bioactivity response of orally administered furosemide in a rat model, described in Chapter 3, “Diuretic Bioactivity

Optimization of Furosemide”.

3 1.2 The role of bioadhesion in nanosphere uptake

Polymers with a high degree of carboxylation have demonstrated the ability to bind mucus-lined tissues, termed mucoadhesion or more generally bioadhesion, via hydrogen bonding [19,20]. Poly(acrylic acid) is a prime example of a polymeric bioadhesive with a high degree of carboxylation that is currently used in commercial formulations [19].

However, poly(acrylic acid) is ultimately water soluble and upon dissolution loses its bioadhesive strength [19]. Mathiowitz et al . have employed polyanhydride polymers that are initially hydrophobic and through hydrolysis of the anhydride bonds expose carboxylic acid groups capable of hydrogen bonding [20,21]. Combining hydrophobicity and carboxylic acid groups has proven very effective at prolonging small intestinal retention of oral doses [22]. However, bioadhesion has demonstrated less success in the gastric environment than in the small intestines [23]. To quantitatively investigate bioadhesion in the gastric environment, we designed a surgical procedure that placed a large-bore gastric tube capable of enabling tensile bioadhesive testing in live, anesthetized rats for the first time in Chapter 4, “Are in vivo gastric bioadhesive forces

accurately reflected by in vitro experiments?”.

Our lab has previously shown that nanospheres formulated with bioadhesive materials

enhanced oral bioavailability of therapeutic molecules in vivo [1]. Additionally, Behrens,

et al. have shown a correlation of bioadhesion and nanoparticle uptake in cell culture and

qualitatively with imaging of intestinal tissue from in vivo [24]. To create novel

bioadhesives, we looked to a naturally occurring example of extremely strong

bioadhesion, the byssal threads of gastropods [25,26]. The amino acid DOPA has been

4 implicated in the bioadhesiveness of mussel associated proteins that confer the bioadhesive nature to byssal threads in a marine environment [25,26]. Therefore, we synthesized bioinspired bioadhesives with DOPA functionality (e.g. poly(butadiene-co- maleic anhydride-graft-DOPA) or “PBMAD”) grafted to hydrophobic backbone anhydride polymers presented in Chapter 5, “Bioinspired Synthetic Bioadhesive

Polymers” [27].

To test if bioinspired bioadhesive coated nanospheres experience increased quantitative internalization from the rat jejunum, 500 nm PMMA nanospheres (weakly bioadhesive) and 500 nm PMMA-core, pBMAD-shell nanospheres (strongly bioadhesive) were administered to rat jejunum isolated loops and uptake was quantified. Percent uptake for both the srongly and weakly bioadhesive formulations are shown in Figure 1.2.

Bioadhesive fracture strength of both polymers is plotted in Figure 1.3. The strongly bioadhesive PBMAD coating of weakly bioadhesive PMMA microspheres greatly increased microsphere uptake to 66.9 ± 12.9 % from 5.8 ± 1.9 % for PMMA microspheres of the same size. Testing indicates with the polymers tested that the 5x increase in mean bioadhesive fracture strength yielded 11.5x increase in nanosphere uptake. As indicated by these results, bioadhesive materials may greatly enhance intestinal uptake of nanospheres in non-phagocytic intestinal regions.

1.3 The role of administration location in nanosphere uptake

Given that specific regions within the small intestines exhibit different nanosphere uptake capacities and resultant biodistributions as seen in Figure 1.4, a magnet-based system was

5 designed to non-invasively retain orally administered pills within any region of the gastrointestinal tract. Bioadhesion and nanoencapsulation combined show tremendous promise for oral drug delivery, especially when administered to particular regions within the small intestines. We set out to develop a device to retain oral doses within the small intestines using magnetic force. In order to design a magnet-based retentive system, first we investigated the net forces experienced by ingested pills in Chapter 6, “Understanding

Gastric Forces Calculated from High Resolution Pill Tracking” [28]. Although, intraluminal pressure measurements can be made within the gastrointestinal tract using balloon catheters with pressure gauges, termed manometry, pressure measurements do not yield directional data or relate to the local net force experienced by pills [29]. We applied Newton’s laws to high resolution pill tracking data to make the first quantitative calculations of the net forces experienced by pills. The gastric region was targeted because it is the region of greatest propulsive forces and therefore serves to provide a high estimate of the forces as a useful starting point in the magnetic oral drug delivery system design. Given the net forces experienced by pills, we set out to design a novel method for retaining pills using inter-magnetic forces.

Previous magnet-based retention systems used off-line secondary bio-marker evaluation to determine if magnetic capture and prolonged retention were effective at the culmination of retention experiments, such as quantifying the bioavailability of a model therapeutic with and without the application of an external magnetic field [30-36]. As proof of concept, Chen and Langer encapsulated ferromagnetic iron oxide in polymerized liposomes and administered them to restrained mice positioned above neodymium iron

6 boron magnets [30]. Polymerized liposome bioavailability increased in the presence of an external magnetic field [30]. Similarly, Ito et al . co-encapsulated ultrafine ferrite particles and brilliant blue dye into cellulose-derivative microspheres to create magnetic microparticles [33]. Microspheres were then gavaged to physically restrained rabbits with magnets positioned around the base of their necks to retain the microspheres at the distal esophagus as a model for locally treating esophageal cancer [33]. External neodymium- iron-boron magnets were fixed in place and at the completion of experimentation it was determined that brilliant blue dye did achieve increased bioavailability in the presence of an external magnetic field [33]. Other investigators taking similar approaches achieved increased bioavailability of acyclovir [32] and doxorubicin [36]. To ensure uptake and increase bioavailability all of the studies positioned magnets as close to the area of interest as possible to maximize the magnetic attractive force on the dose and consequently the underlying tissue [30-36]. Additionally, none of the studies provide any inter-magnetic force measurement or means of directly determining if the external magnet is retaining the oral dose at the site of interest during the period of application

[30-36]. We present the first system designed to continuously monitor inter-magnetic force between an externally applied magnet and an orally administered dose. Moreover, the external magnet is placed on a moveable arm that adjusts its position with respect to the internal magnet to minimize the force necessary to retain the internal dose and therefore the force applied to the underlying tissue.

Chapter 7 describes the safety and efficacy testing of the magnet-based localization of model oral doses in rats. With the ability to prolong and control the duration of an oral

7 dosage form in a particular region of the GI, we hope to provide a system for investigating the region-dependence of uptake in live, awake, subjects as compared with an isolated loop model. Moreover, pill localization provides a platform for testing the oral administration of therapeutics that exhibit optimal absorption in a single region of the GI

(e.g. narrow absorption window therapeutics and vaccines), and to address GI diseases by delivering directly to the site of the pathophysiology (e.g. inflammatory bowel disease and colon cancer).

Additionally, the duration of localized polymer nanosphere administration affects uptake and biodistribution as shown in Table 1.1. Very few nanospheres are internalized within the first hour; however after 5 hours of administration within a particular region of the gastrointestinal tract significant uptake occurred. Therefore the duration of localized delivery within the small intestines also greatly affects nanosphere uptake and biodistribution. Since the residence time of standard oral doses in the regions of the gastrointestinal tract are controlled by physiological mechanisms, we developed a magnet-based localization system to prolong and control residence time as described in

Chapter 7, “Novel Method for Localized Delivery from Magnetic Pills.”

Background and introductory material addressing prior work and literature relevant to each topic is presented in the first section of each chapter.

8 1.4 Methods and materials summary

Polymer Nanospheres

Polystyrene micro- and nano-spheres were purchased from Polysciences, Inc.

(Warrington, PA). Polymethylmethacrylate (PMMA) and poly(butadiene malaec anhydride-co-L-dopamine) (pBMAD)-PMMA microspheres were fabricated by phase inversion nanoencapsulation and confirmed to be 0.5 µm in diameter by Coulter particle size analysis (Beckman Coulter, Brea, CA). Chlorpromazine and cytochalasin B were purchased from Sigma-Aldrich (St. Louis, MO).

Isolated loop

An isolated loop procedure was performed on 6cm intestinal sections as previously described [22] in male, Sprague-Dawley rats weighing 200-250g with a 1 ml (25 mg) dosage of PM (n=4). Following specified incubation periods, tissue samples were collected and stored at -18 oC until further processing.

Uptake Quantification

Poly(styrene) microsphere concentration was quantified in each sample via size exclusion chromotagraphy as described by Jani et al. [6] on a Shimadzu GPC equipped with Waters

Styragel HR5E and HR4E columns and a Shimadzu RID-10A refractive index detector resulting in an R 2 value of 0.9993 for the calibration curve. Positive controls (doped

organ samples) resulted in 99.8 ± 1.3 % recovery and negative controls resulted in 0.0 % recovery.

Percent of uptake was calculated by taking the sum of all amounts detected in tissues

(excluding isolated loop and loop rinse samples) divided by the total dose administered and multiplied by 100. Additionally, taking the sum of all amounts detected in every tissue sample and dividing by the total administered dose resulted in a mass balance calculation. For all study groups the mass balance results were within 11.3% of the total dose.

Statistical analysis

Standard errors were calculated and a 1-way ANOVA was performed with Microcal

Origin Graphical Software (Northampton, MA). Significance was determined at p ≤0.05.

1.5 Hypothesis and specific aims

The primary hypothesis is that narrowing size distribution of polymer nanospheres

produced by phase inversion, increasing the bioadhesiveness of the nanospheres, and

retaining the nanospheres at particular sites within the GI tract will greatly increase

uptake and bioavailability. To address the hypothesis, the following specific aims were

addressed:

1) Investigate the capability of phase inversion to produce polymer and drug

microparticles within desired size ranges

a) Develop and test novel methods for producing polymer nanospheres by phase

inversion under continuous flow conditions – addressed in Chapter 2

10 b) Test the ability of phase inversion to alter the bioactivity of a model hydrophobic

drug, furosemide – addressed in Chapter 3

2) Investigate the ability of bioadhesive polymers to adhere to the gastric mucosa and

synthesize polymers that exhibit exceptionally strong bioadhesive fracture strength

and tensile work

a) Determine if bioadhesives act similarly in the stomach in vivo as compared with

in vitro – addressed in Chapter 4

b) Synthesize bioinspired bioadhesives that incorporate DOPA and test bioadhesive

properties on small intestinal tissue – addressed in Chapter 5

3) Design and test a novel method for magnetically retaining pills in the GI

a) Model the forces experienced by a model magnetic pill in the area of greatest

peristaltic activity, the stomach – addressed in Chapter 6

b) Create and test a system for retaining magnetic pills in the GI with the minimum

necessary inter-magnetic force - addressed in Chapter 7

*p = 0.0294

0.51 2 Diameter / Microns

Figure 1.1: Uptake of poly(styrene) polymer nanospheres of varying diameter in the jejunum of rats (N=4) demonstrating the dependence of uptake on particle size.

*p < 0.01

Strongly Weakly Bioadhesive Bioadhesive PBMAD PMMA

Figure 1.2: Uptake of strongly bioadhesive poly(butadiene-co-maleic anhydride-graft-DOPA) (PBMAD) coated, as compared to that of weakly bioadhesive poly(methyl methacrylate) (PMMA) 500 nm in the jejunum of rats (N=4) demonstrating the dependence of uptake on bioadhesiveness.

750 *

500

250

0 BioadhesiveStrength Fracture[mN/sq cm] Strongly Bioadhesive Weakly Bioadhesive PBMAD PMMA

Figure 1.3: Bioadhesive fracture strength of the strongly bioadhesive, poly(butadiene-co-maleic anhydride-graft-DOPA) (PBMAD), as compared to that of weakly bioadhesive poly(methyl methacrylate) (PMMA) as tested on rat intestinal tissue in vitro (*p<0.05).

*p = 0.03 *p = 0.02

0.51 2 0.5 1 2 Jejunum Ileum

Figure 1.4: Uptake of polystyrene spheres as a function of size and location of administration demonstrating the intestinal location dependence on percent uptake.

Time Organ Disposition (PTD) (hrs) Liver Kidneys Lungs 1 0.1 ± 0.1 --- 0.2 ± 0.2 Jejunum 3 19.0 ± 12.8 --- 0.6 ± 0.6 (0.5 µm) 5 36.7 ± 9.9 3.6 ± 2.4 0.5 ± 0.3 1 --- 0.2 ± 0.2 4.0 ± 3.6 Ileum 3 24.8 ± 14.8 3.5 ± 3.5 1.0 ± 1.0 (1 µm) 5 26.3 ± 14.2 0.4 ± 0.4 0.1 ± 0.1

Table 1.1: Organ disposition kinetics of polymer nanospheres in the rat jejunum and ileum, PTD = percent total dose.

16 1.6 References

1. Mathiowitz, E., et al., Biologically erodable microspheres as potential oral drug delivery

systems . Nature , 1997. 386(6623): p. 410-4.

2. Bernkop-Schnurch, A, Kast, CE, and Guggi, D. Permeation enhancing polymers in oral

delivery of hydrophilic macromolecules: thiomer/GSH systems . J Control Release 2003.

93(2): 95-103.

3. Park, H, Park, K, and Kim, D, Preparation and swelling behavior of chitosan-based superporous hydrogels for gastric retention application . J Biomed Mater Res A , 2006. 76(1):

144-50.

4. Toorisaka, E, et al., An enteric-coated dry emulsion formulation for oral insulin delivery .

J Control Release , 2005. 107(1): 91-6.

5. Fissell, WH, Humes, HD, Fleischman, AJ, Roy, S. Dialysis and nanotechnology: Now,

10 years, or never? Blood Purification. 2007. 25(1): 12-17.

6. Mathiowitz, E, Jacob, JS. 2002. Novel mechanism for spontaneous encapsulation of active

agents: Phase inversion nanoencapsulation. Abstracts of Papers of the American Chemical

Society 223 : 374-COLL, Part 1.

7. Furtado, S, Abramson, D, Burrill, R, Olivier, G, Gourd, C, Bubbers, E, Mathiowitz, E.

2008. Oral delivery of insulin loaded poly(fumaric-co-sebacic) anhydride microspheres.

International Journal of Pharmaceutics 347(1-2): 149-155.

8. Furtado, S, Abramson, D, Simhkay, L, Wobbekind, D, Mathlowitz, E. 2006. Subcutaneous

delivery of insulin loaded poly(fumaric-co-sebacic anhydride) microspheres to type 1

diabetic rats. European Jornal of Pharmaceutics and Biopharmaceutics 63 (2): 229-236.

17 9. Carino, GP, Jacob, JS, Mathiowitz, E. 2000. Nanosphere based oral insulin delivery.

Journal of Controlled Release 65 (1-2): 261-269.

10. Morello, AP, Forbes, N, Mathiowitz, E. 2007. Investigating the effects of surfactants on

the size and hydrolytic stability of poly(adipic anhydride) particles. Journal of

Microencapsulation 24 (1): 40-56.

11. Laulicht, B, Cheifetz, P, Mathiowitz, E, Tripathi, A. 2008. Evaluation of continuous flow nanosphere formation by controlled microfluidic transport. Langmuir 24 (17): 9717-9726.

12. Jani, P, et al., The uptake and translocation of latex nanospheres and microspheres after oral administration to rats . J Pharm Pharmacol , 1989. 41(12): 809-12.

13. Jani, P, et al., Nanoparticle uptake by the rat gastrointestinal mucosa: quantitation and particle size dependency . J Pharm Pharmacol , 1990. 42(12): 821-6.

14. Dormans, TPJ, vanMeyel, JJM, Gerlag, PGG, Tan, Y, Russel, FGM, Smits, P. Diuretic efficacy of high dose furosemide in severe heart failure: Bolus injection versus continuous infusion. J. Am. College Cardio., 1996. 28 (2): 376-382.

15. Salvador, DRK, Rey, NR, Ramos, GC, Punzalan, FER. Continuous infusion versus bolus

injection of loop diuretics in congestive heart failure. Cochrane Database of Systematic

Reviews 2005. (3).

16. Murray, MD, Haag, KM, Black, PK, Hall, SD, Brater, DC. Variable furosemide

absorption and poor predictability of response in elderly patients. Pharmacotheraphy, 1997.

17(1): 98-106.

17. Davis, SS. Formulation strategies for absorption windows. Drug Disc. Today, 2005.

10(4): 249-257.

18 18. Bardonnet, PL, Faivre, V, Pugh, WJ, Piffaretti, JC, Falson, F. Gastroretentive dosage forms: Overview and special case of Helicobacter pylori. J. Controlled Release, 2006. 111

(1-2): 1-18.

19. Achar, L and Peppas, N. Preparation, characterization and mucoadhesive interactions of poly(methacrylic acid) copolymers with rat mucosa, J. Control. Release, 1994. 31 (3): 271–

276.

20. Chickering, D, Jacob, J, Mathiowitz, E. Bioadhesive microspheres. 2. Characterization and evaluation of bioadhesion involving hard bioerodible polymers and soft-tissue, React.

Polym., 1995. 25 (2–3): 189–206.

21. Santos, C, Freedman, B, Leach, K, Press, D, Scarpulla, M, Mathiowitz, E. Poly(fumaric- co-sebacic anhydride) — a degradation study as evaluated by FTIR, DSC, GPC and X-ray diffraction, J. Control. Release, 1999. 60(1): 11–22.

22. Chickering, DE, Jacob, JS, Desai, TA, Harrison, M, Harris, WP, Morrell, CN,

Chaturvedi, P, Mathiowitz, E. Bioadhesive microspheres .3. An in vivo transit and bioavailability study of drug-loaded alginate and poly(fumaric-co-sebacic anhydride) microspheres. J. of Control. Release, 1997. 48(1): 35-46.

23. Laulicht, B, Cheifetz, P, Tripathi, A, Mathiowitz, E. Are in vivo gastric bioadhesive forces accurately reflected by in vitro experiments? J. Controlled Release, 2009. 134(2): 103-

110.

24. Behrens, I., et al., Comparative uptake studies of bioadhesive and non-bioadhesive nanoparticles in human intestinal cell lines and rats: the effect of mucus on particle adsorption and transport. Pharm Res , 2002. 19(8): 1185-93.

19 25. Lee, H, Scherer, NF, and Messersmith, PB. Single-molecule mechanics of mussel adhesion. Proc. Nat. Acad. Sci., 2006. 103: 12999-13003.

26. Waite, JH. The DOPA Ephemera - A Recurrent Motif in Invertebrates. Bio. Bull., 1992.

183: 178-84.

27. Schestopol, MA, Jacob, JS, Donnely, R, Ricketts, TL, Nangia, A, Mathiowitz, E, Shaked,

Z. Bioadhesive Polymers with Catechol Functionality, WO2005/056708.

28. Laulicht, B; Tripathi, A; Shlageter, V; Kucera, P; Mathiowitz, E. April 19, 2010.

Understanding Gastric Forces Calculated from High Resolution Pill Tracking. PNAS

10.1073/pnas.1002292107.

29. Hveem, K, Sun, WM, Hebbard, GS, Horowitz, M, Dent, J. Insights into Stomach

Mechanics from Concurrent Gastric Ultrasound and Manometry. Gastroenterololgy, 1994.

107(4): 1236-1236.

30. Chen, HM, Langer, R. Magnetically-responsive polymerized liposomes as potential oral delivery vehicles. Pharm. Res., 1997. 14: 537-540.

31. Arruebo, M, Fernandez-Pacheco, R, Ibarra, MR, Santamaria, J. Magnetic nanoparticles for drug delivery. Nano Today , 2007. 2: 22-32.

32. Groning, R, Berntgen, M, Georgarakis, M. Acyclovir serum concentrations following peroral administration of magnetic depot tablets and the influence of extracorporal magnets to control gastrointestinal transit. Eur. J. Pharm. Biopharm., 1998. 46: 285-91.

33. Ito, R, Machida, Y, Sannan, T, Nagai, T. Magnetic Granules – A Novel System for

Specific Drug Delviery to Esophageal Mucosa in Oral-Administration. Int. J. Pharm. 1990.

61, 109-17.

20 34. Polyak, B, Friedman, G. Magnetic targeting for site-specific drug delivery: applications

and clinical potential. Expert Opin. Drug Delivery, 2009. 6, 53-70.

35. Teply, BA. et al. The use of charge-coupled polymeric microparticles and micromagnets

for modulating the bioavailability of orally delivered macromolecules. Biomaterials, 2008.

29: 1216-1223.

36. Widder, K.J. et. al. Tumor Remission in Yoshida Sarcoma-Bearing Rats by Selective

Targeting of Magnetic Albumin Microspheres Containing Doxorubucin. Proc. Natl. Acad.

Sci. USA, 1981. 78: 579-81.

37. Brainerd, EL et al. X-ray Reconstruction of Moving Morphology (XROMM): Precision,

Accuracy and Applications in Comparative Biomechanics Research. Journal of Experimental

Zoology A , 2010. 313A.

Chapter 2

Evaluation of Continuous Flow Nanosphere

Formation by Controlled Microfluidic

Transport

Abstract

Improved size monodispersity of populations of polymer nanospheres is of enormous interest in the fields of nanotechnology and nanomedicine. As such, the exact experimental conditions precisely producing polymer nanospheres are needed for nonaqueous systems. This work presents the use of controlled microfluidic transport methods to study the experimental parameters for fabricating nanoparticles utilizing phase inversion. We report two microfluidic methods for forming polymer nanospheres in small batches to determine the formation conditions. These conditions were then implemented to perform higher throughput formation of polymer nanospheres of the desired size. The controlled microfluidic environment, operating in the laminar flow

22 regime, produces improved size monodispersity, decreased average diameter, and affords a greater degree of control over the nanosphere size distribution without adding surfactants or additional solvents. Experiments show a nonlinear trend toward decreasing particle size with decreasing polymer concentration and a linear trend toward decreasing size with increasing flow rate indicating a time-course-dependent nucleation and growth mechanism of formation within the range of conditions tested.

2.1 Background and Motivation

Nanospheres having diameters less than 1 µm offer significant advantages over larger, more conventional microsphere formulations for oral drug delivery [1]. Delivery of the entire drug-polymer complex, made possible by producing nanospheres on the order of the size of lipid rafts or clathrin coated pits has been shown to greatly improve cellular uptake over larger microspheres that are only uptaken by phagocytotic M-cells in the

Peyer’s patches. Studies by the groups of Rejman [2] and Florence [3-5] indicate that polymer nanospheres differing by mere hundreds of nanometers in diameter experience widely different bioavailabilities and biodistributions within mammalian cells and among various tissues. Polymer nanospheres of approximately 250 nm in diameter remain in early stage endocytotic vesicles in cell culture; whereas larger particles of approximately

500 nm in diameter make their way to late-stage, enzyme containing vesicles. Jani et al.

[3-5] demonstrated a high degree of submicron diameter polymer sphere uptake in rats. In particular, spheres less than 1000 nm in diameter achieved gastrointestinal uptake of up to 34%. [5] In the Jani et al . study, nanospheres with a mean diameter of 100 nm showed decreased uptake compared to 500 nm spheres indicating an active transcellular uptake

23 mechanism for polymer nanospheres [5]. Therefore the size range of greatest potential therapeutic benefit for polymer nanospheres in oral drug delivery is 200−1000 nm [3-5].

The implications of these studies for nanomedical drug delivery technologies will be immense, necessitating the development of methods to produce monodisperse populations of nanospheres within the various size ranges. However, methods for entrapping sensitive therapeutics including proteins and biologics within polymer nanospheres are currently lacking in the literature and in commercial practice.

Numerous microfluidic devices and techniques have been developed to produce polymer microspheres and microfibers [6] that take advantage of emulsion formation [7,8],

Rayleigh instability [7,9], photochemical cross-linking [10-12], and/or chemical synthesis

[13]. The size of the droplets and their corresponding microspheres produced by microfluidic emulsions utilize droplet breakup strategies to create monodisperse droplets on the micron-scale. Monomers or prepolymers are introduced into catalysts and cross- linking agents in a controlled geometry yielding microspheres or microcapsules. In some setups reactions occur at the interface between two flowing streams without the formation of an emulsion to produce microfibers by interfacial polymerization [6]. Thermally or photoinitiated cross-linking of emulsified droplets can also be used to polymerize microspheres [10,11,14,15]. Another photoinitiated cross-linking technique avoids the formation of droplets by using a photolithography-like setup for directly photopolymerizing a flowing polymer solution allowing for control over microparticle shape [12]. Similarly, other aqueous microfluidic techniques [8,9,16-21] have produced

24 polymer microspheres; however, the production methods are limited to using an aqueous continuous phase.

One promising method is phase inversion nanosphere formation (PIN) method [22], in which nanosized particles of a chosen polymer can be prepared by pouring the polymeric organic solution (solvent) into another organic phase (nonsolvent) without any mechanical stirring. In contrast to previous polymer microsphere microfluidic studies,

PIN does not involve the formation of droplets or cross-links. Instead, PIN utilizes the miscibility of the organic solvent and nonsolvent pair to enable production of thermoplastic polymer nanospheres on the nanoscale, at least an order of magnitude smaller than the microfluidic channels in which they are produced. As a result, PIN is a very promising method for producing polymer nanospheres from water-insoluble thermoplastic polymers, including those possessing desirable oral drug delivery properties including bioadhesive and bioerodible polymers.

On the benchtop scale the PIN process produces nanosized particles of a chosen polymer prepared by pouring the polymeric organic solution (solvent) into another organic phase

(nonsolvent) without any mechanical stirring or the creation of an oil/water interface in conventional glassware. For a microfluidic approach, due to the swelling of molded poly(dimethyl siloxane) (PDMS) and other polymer-involving systems the microfluidic platform had to be designed to withstand organic solvent usage. Materials including silicon, stainless steel, and glass were the most attractive options and glass was chosen due to the transparency, which allowed us to observe the flow pinching process, to

25 confirm the lack of visible droplet formation, and to identify the nature of clogged channels when investigating the range of suitable polymer concentrations.

Etched glass microchannels allowed for the introduction of chlorinated organic polymer solutions (e.g., PMMA in methylene chloride) into miscible organic polymer nonsolvents

(e.g., pentane) that induce phase separation of the polymer from the solvent-nonsolvent system to produce nanospheres. By choosing the appropriate solvent-nonsolvent pair that causes partitioning of the therapeutic agent with the polymer excipient, polymer encapsulated drug nanospheres are formed [1]. On the macro-scale the formation of polymer nanoparticles by phase inversion leads to highly size polydisperse nanosphere populations when no excipients are added [1]. Also, on the macro-scale, tuning of experimental parameters is very cost-intensive because it requires liters of organic solvents. Hence, the knowledge of the fundamental mechanism and exact experimental conditions for precise production of nanoparticles are still missing for organic solvent- based systems.

In this paper, we describe a microfluidic phase inversion method for producing nanoparticles under highly controlled transport conditions using organic solvent-based system. The method requires only tens of microliters producing a tremendous cost savings. Although the microfluidic PIN procedure has a lower throughput than batch PIN, the methods presented provide increased control over production parameters such as flow rate, polymer concentration and dilution, greatly accelerating the pace of investigation into the mechanism of formation. Establishing the conditions leading to production of

26 nanospheres in the desired size range for oral drug delivery on a microfluidic platform enables the rapid, low-cost investigation of processing parameters that lead to changes in the size distribution of nanosphere populations.

2.2 Experimental Methods

Off-Chip Phase Inversion Nanosphere Formation (PIN)

100 µL of 0.01 weight per volume percent 50 kDa PMMA (M w/M n = 1.06) in methylene chloride (solvent phase) was ejected from a solvent-friendly pipet tip into 10 mL of pentane (nonsolvent). A 30 µL aliquot was then withdrawn from the bottom of the nonsolvent vessel by a fresh solvent-friendly micropipette tip and placed into an aluminum sample pan (Perkin-Elmer, Waltham MA). The liquid phase is allowed to evaporate and the resultant nanospheres were imaged by scanning electron microscopy

(SEM). Experiments were repeated using starting concentrations of 0.001 and 0.0001 weight per volume percent PMMA. Laboratory grade pentane and methylene chloride were supplied by Sigma-Aldrich and PMMA was supplied by Polymer Source (Montreal,

Canada).

Microfluidic Nanosphere Production with Flow Pinching

Microfluidic chips were fabricated in borosilicate glass substrate at the Brown University

Microelectronics Facilities using a protocol based on standard microlithographic techniques. Briefly, the glass substrate was coated with chrome ( 800 Å thickness) and gold ( 400 Å thickness) using chemical vapor deposition after which a layer of Shipley

1818 photoresist was spin coated. After exposing the photoresist layer to UV through a negative mask, microchannels were etched in 49% hydrofluoric acid using calibrated

27 etching versus time curves. A computer numeric control (CNC) lathe drilled holes in a second glass capping wafer, which later serve as reagent reservoirs. The etched glass wafer with the microchannel geometry is then bonded to the capping wafer using a controlled thermal bonding procedure [23-25].

The dilution and flow of the solvent and nonsolvent phases are regulated with a custom- designed two-component programmable control system. First, four independent pressure ports impose and measure the air pressure over any microchannel network containing fluids. Each pressure port has accuracy to within 0.01 psi of the assigned value over a

±15 psi range and response time of 5 psi/s. Second, a custom LABVIEW software program and data acquisition boards (National Instruments Corporation) provide an automatic calibration and programming interface for the user. We have developed a simplified protocol for controlling the microfluidic chip dilutions by solving a system of momentum and continuity equations. In each channel i, the pressure drop can be related according to

where η is the viscosity of liquid and Qi is the flow rate. The hydrodynamic resistance Ri of an isotropically etched microchannel is given by

28 where wi, di and Li are the width, depth and length of the microchannel i, respectively.

The correction factor αi, which is multiplied by the hydrodynamic resistance of rectangular channel, accounts for the isotropic shape of the channel.

In the flow pinching experiment, reservoir 1 was filled with the polymer in solvent solution and the remaining three were filled with nonsolvent. The glass microchip is shown in schematic form and in a photograph in Figure 2.1a and b. A negative pressure of 1 psi was applied to reservoir 4 pulling the two nonsolvent and one solvent channels into the mixing channel inducing flow pinching. The solvent-nonsolvent interface has a sufficient difference in refractive index to allow viewing of the pinching flow using light microscopy shown diagrammatically in Figure 2.1c and in a light micrograph in Figure

2.1d. However, due to the miscibility of the solvent and nonsolvent, the interface is maintained only in the portion of the channel closest to the channel junction, beyond which mixing yields a single visually distinguishable liquid phase. Using the solvent trap setup, the viscosity of a 0.01% PMMA in methylene chloride solution was measured to be 0.4 cP, negligibly different from pure methylene chloride, on a TA Instruments

AR2000 Rheometer. The viscosity of n-pentane used in calculations was measured to be

0.2 cP. The solvent phase flow rate was calculated using eqs 1 and 2 to be 0.093 nL/s.

Flow pinching conditions produce a 30:70 dilution ratio of the solvent to nonsolvent phases at the junction. After ten minutes of run time, the contents of well 4 were collected for size analysis.

29 Microfluidic PIN Nanosphere Production without Flow Pinching

Next we tested the second solvent/nonsolvent configuration. In this configuration, reservoirs 1, 2, and 3 (Figure 2.1a) were filled with the dilute polymer solutions and reservoir 4 was filled with the nonsolvent pentane. Negative pressure of 1 psi was applied to the nonsolvent reservoir causing flow of the dilute polymer solution into reservoir 4 via the cross chip microchannels avoiding the flow-pinching phenomenon. In this configuration the solvent phase flow rate is calculated to be 0.31 nL/s.

Capillary Tube PIN Nanosphere Production

In an effort to increase the scale of nanosphere production, glass capillary tubes (Labcraft

100 µL disposable glass micropipette tubes) were press-fit into male luer to 1/16 in. tube

fittings (McMaster Carr) heated to 120 °C. The glass capillary tubes were interfaced with

solvent-friendly syringes that were filled with dilute 0.001 wt % PMMA in methylene

chloride (Figure 2.1f). Each solvent-friendly syringe containing dilute polymer in organic

solvent solutions was placed into a Harvard Apparatus Pump 11 Pico Plus syringe pump

(Hamden, CT). The syringe pump was set to flow organic solvent through the glass

capillary tube (inner diameter = 1 mm) at rates ranging from 1−100 nL/s into a 10 mL

Pyrex beaker containing 10 mL of nonsolvent, pentane. After formation, 50 µL of each sample are collected for size analysis.

Nanosphere Sizing

For the microfluidic nanospheres production methods, at the end of each 10 min run, the entire contents of the collection reservoir were transferred by micropipette into an

30 aluminum sample pan (Perkin-Elmer, Waltham, MA). Aluminum sample pans were used because of their geometry and conductivity, providing an ideal vessel for evaporating volatile organics leaving behind polymer nanospheres. Sample pans were placed on SEM stubs with double-sided carbon tape and sputter-coated with 50−100 Å of gold−palladium

(Emitech K550, Kent, UK). The stub was inserted into the Hitachi S-2700 scanning electron microscope (Tokyo, Japan) with an accelerating voltage of 8 kV. The microscope was aligned and then digital pictures were obtained via the Quartz PCI digital imaging system and software (Quartz Imaging Corporation, Vancouver, BC). Resultant images were analyzed for Ferret’s mean diameter of fitted ellipsoids using NIH ImageJ

(Bethesda, MD) as depicted in Figure 2.2 to determine Ferret’s mean diameter of the nanospheres.

Statistical Analysis

Statistical analysis was performed using SPSS software (Chicago, IL). In populations of nanospheres produced that had inhomogeneous variances, the Welch and Brown-

Forsythe robust tests of equality of means were run followed by a Dunnett T3 posthoc test. The Student’s t test was used to compare nanosphere populations produced by

0.001% PMMA solutions on both microfluidic methods since the variances were homogeneous. From the SPSS calculations we calculated p-values. The p-value measures consistency between nanosphere populations by calculating the probability of observing the same results between experimental data sets.

31 2.3 Results and Discussions

We seek to understand the formation of poly(methyl methacrylate) (PMMA) chains into spheres. PMMA chains of molecular weight Mw = 50 kDa ( Mw/Mn = 1.06) consisting of

N statistically independent-segments each of length b (the “Kuhn length”). The number of segments and the “Kuhn length” in the equivalent freely jointed Kuhn chain are

2 computed as N = 3 Mw sin θ/MoC∞ = 130, and b = C∞l/sin θ = 1.72 nm, where C∞ (=9.1) for polyethylene oxide chains) is the characteristic ratio, l (=0.154 nm) is the

carbon−carbon bond length, Mo (=100 g/mol) is the molecular mass of the repeat unit and

θ (=54.5°) is the half angle between carbon−carbon bonds in a polymer chain. The root- mean-square end-to-end distance of the equivalent Kuhn length is

. A solution in MeCl 2 of such polymer chains of narrow molecular weight distribution with a concentration c per unit volume is injected. The number density of the chains can be computed as n = cN A/Mw, where NA is Avogadro’s

number. The average distances between chains for 0.01%, 0.001% and 0.0001% solutions

are approximately 58 nm, 125 nm, and 271 nm, respectively. Hence, the distance between

the nucleation sites grows nonlinearly with concentration. Since the distances between

chains are greater than the radius of gyration, the polymer solutions are in the perfectly

dilute regime; therefore, polymer chains must diffuse, agglomerate, and collapse to result

in nanoparticles of sizes observed in the experimental results.

Results of four phase inversion nanosphere (PIN) production methods are reported. In the

first method the PIN are formed in a conventional way. Here, the polymer solution is put

into a beaker full of nonsolvent. In the second method, PIN spheres are formed by a

32 microfluidic flow pinching (Figure 2.1a). Here, the polymer molecules flow in contact with nonsolvent molecules and diffusion occurs across the microchannel laminar flow. In the third method, the solvent phase flows directly into the stagnant nonsolvent well without first pinching the flow. Here, the polymer molecules flow from the microchannel into the pool of non solvent molecules. The diffusion mixing is similar to a radial

“source” flow mixing. In the final method, the polymer molecules are injected into the pool of nonsolvent molecules using a capillary flow. The diffusional mixing is similar to a “fluid jet” mixing. In all cases the nanospheres generated were imaged and sizes were analyzed on NIH ImageJ software, an example of which is shown in Figure 2.2.

Off Chip PIN Experiments

Our experimental runs using 0.0001 and 0.001 weight per volume percent PMMA solutions showed no visible formation of particles in the bulk phase. It appears that particles were either not formed or smaller than lowest detectable diameter by SEM ( 50 nm). At these very low concentrations, the polymers collapse as isolated coils that are very far apart. Experiments were then repeated using starting concentration of 0.01 weight per volume percent PMMA and synthesized spheres that were observed under the

SEM are shown in Figure 2.3. Here, polymer chains are in close enough proximity to attract each other while phase inverting to form the cores of polymer nanospheres, but polymer chains at the surface of the nanospheres contact pure solvent as it is driven from the core of the polymer due to the phase inversion coupled with diffusion into the nonsolvent. Owing to the high cost of their surface energy, nanospheres would like to stick together, forming larger clusters with lower surface energy per molecule due to

33 reduced surface area of contact with the nonsolvent. This tendency results in formation of bigger nanoparticles and agglomerates. The figure clearly shows signs of agglomerated nanoparticles of spherical and nonspherical shapes of different sizes. A size histogram is shown in Figure 2.3b. The data shows 660 ± 48 nm effective diameter nanoparticles including agglomerates. The data shows huge scatter in size. 53% of the nanosphere population is larger than 500 nm and 12% is larger than 1 µm. Moreover, the population

of nanospheres produced by PIN not only is limited to the highest polymer concentration

used on the glass microfluidic chips, but also has one of the largest coefficients of

variance measured in testing, 52%, in which coefficient of variance is defined as the

percent that the standard deviation is of the mean diameter. Hence, in the previous PIN

investigations [1,22], as with the off-chip production, the uncontrolled introduction of the

solvent into the nonsolvent phase perhaps led to nonhomogenous nucleation sites

resulting in increased polydispersivity and coefficient of variance. It should be noted that

the off-chip experiments require large amounts of solvent and polymer solutions and

hence it would require large number of experiments to obtain a desired relationship

between PMMA concentration and average particle size.

Microfluidic Nanosphere Production with and without Flow Pinching

We first investigated the effect of polymer concentration on particle production. Dilute

polymer (PMMA) in methylene chloride (MeCl 2) solutions were run in the above chip

conditions at concentrations ranging from 0.0001 to 1 wt % in orders of magnitude.

Above 0.01 wt % polymer formed network structures in the microchannels (75 µm wide

and 12 µm deep) along the solvent/nonsolvent interface indicating that the polymer

34 chains precipitated in close enough proximity to join into a bulk microstructure rather than discrete nanospheres as shown in Figure 2.1e. This experimental result suggests the rapid collapse of polymer chains across the entire cross-section of the channel. The diffusion time for a 50 kDa polymer in a good solvent to travel across a 75 µm wide

2 2 2 channel is td w /2 D = 75 /2 · 67 = 42 s. Here, D = 67 µm /s is the molecular diffusivity of PMMA [26]. Since the diffusion of solvent and polymer molecules across the width of the channel was rapid, the time scale of this particle growth was almost instantaneous.

The residence time of the polymer molecules while traversing that microchannel is tr la /Q = 400 s in the flow-pinching and tr lA /Q = 120 s in the nonflow pinching

configurations. Therefore the residence time in the channel far exceeds the diffusion time.

Finally, it is noted that the overlap concentration (c*) for PMMA in the solvent is c* ≈

2.5/[ η] = 0.012 g/mL ≡ 1.2% Here, [ η] is the measured intrinsic viscosity of 50 kDa

PMMA in a MeCl 2 solution.

Microfluidic nanosphere populations (Figure 2.4) produced in glass chip microfluidics are significantly different from the off-chip (Figure 2.1), uncontrolled nanosphere formation ( p < 0.01). In the microfluidic flow pinching configuration, 0.01 wt % 50 kDa

PMMA ( Mw/Mn = 1.06) solution formed 731 ± 32.7 nm diameter nanospheres on average

(Figure 2.5). At 0.001 and 0.0001 weight percent 533 ± 18.5 nm and 517.4 ± 19.6 nm

diameter spheres respectively were produced (Figure 2.5). In the nonflow pinching

conditions on the same microfluidic chips, similar results were achieved: 846 ± 19.3 nm

at 0.01, 529 ± 23.7 nm at 0.001, and 525 ± 17.4 nm at 0.0001 PMMA weight percent

(Figure 2.6). On the whole, in both cases mean nanosphere size decreased significantly as

35 polymer concentration decreased (Figure 2.7). However, in the flow pinching case the mean sphere volume decreased by a factor of 2.6 and without flow pinching the mean sphere volume decreased by a factor of 4.1. Moreover, the 10-fold differences in initial polymer concentration do not linearly correlate with the mean nanospheres volume for either microfluidic technique possibly indicating a difference in degree of polymer chain collapse and/or that solvent is trapped within the nanospheres. Note that the nanosphere sizes are much smaller than microchannel width (75 µm) or depth (12 µm). The size depends only on the local concentration and diffusion time of polymer chains in the microchannel.

Populations of nanospheres formed by the microfluidic PIN methods can be statistically grouped into two homogeneous subsets of 0.01% PMMA and less than 0.01% PMMA.

Additionally, the populations of nanospheres formed at 0.01% PMMA vary significantly between the two microfluidic methods ( p < 0.01), while at lower concentrations they are negligibly different (not statistically significant). The variation seen between the two microfluidic methods at higher concentrations evidences the extreme sensitivity to manufacturing conditions that have substantially complicated previous PIN-based processes.

The mean volume decrease is negligible in both microfluidic testing conditions further evidencing a nonlinear relationship between polymer concentration and resultant particle size. On the whole, the flow pinching method tended to produce smaller spheres at the same polymer concentrations. While the flow rate of the solvent phase is slower in the

36 flow pinching case than the nonflow pinching case, pinching the flow causes thinning of the solvent phase stream effectively increasing the interfacial area to volume ratio and allowing for more rapid diffusion-driven exchange between the solvent and nonsolvent.

The increase in transport kinetics from the nonflow pinching to the flow pinching case would tend to accelerate polymer chain collapse leading to less trapped solvent, which may account for the observed size differences.

The nanosphere populations at concentrations below 0.01% produced by both microfluidic methods under the same conditions were very similar. Since the flow pinching method introduces the polymer solution into the nonsolvent rapidly it appears that the kinetics of nanosphere formation is faster still as the mean diameter and distribution of particles is similar in both microfluidic setups.

Results indicate that the conditions for formation at 0.001 and 0.0001 wt % PMMA in methylene chloride are similar as evidenced by the similarities of resultant nanosphere populations formed by both microfluidic methods. Between 0.001 and 0.01 wt % there is a significant increase in mean nanosphere diameter. The overlap concentration ( c*) for

PMMA in a good solvent was calculated to be c* ≈ 1.2 weight per volume percent.

Additionally the viscosities of all of the polymer solutions used was negligibly different from pure methylene chloride indicating that all of the studies were performed in perfectly dilute solution conditions. Since only the highest concentration formed nanospheres off-chip perhaps the polymer−solvent and nonsolvent phases are miscible at the lower polymer concentrations when mixed rapidly. At low flow rates in the laminar

37 regime the solvent phase is introduced slowly and so the volume of the solvent phase is greatly reduced relative to that of the nonsolvent phase at any given time. In this case the nonsolvent to solvent ratio is effectively very high throughout the phase inversion process producing smaller and more uniform spheres due to the controlled mixing conditions.

Capillary Tube Nanosphere Production

To further investigate the effect of solvent phase flow rate, the flow was varied and polymer concentration was held constant in the glass capillary tube experiments. The results show a statistically significant ( p < 0.01) decreasing trend in mean nanosphere diameter with increasing flow rate (Figure 2.8). While the bulk flow rate was increased when moving from a flow-pinching to a nonflow pinching setup in the microfluidic case, the effective interfacial area involved in diffusion decreased indicating that both area for solvent nonsolvent exchange and flow rate can be tuned to control the resultant size distribution of the nanosphere population. The population of nanospheres produced by the glass capillary tube method at 100 nL/s flow rate produced a population of nanospheres with the greatest percentage in the 200−500 nm diameter range, 78.5%, of all the methods tested. Moreover, the population of nanospheres formed by the glass capillary tube method flowing the solvent phase at 1 nL/s has no statistically significant different ( P < 0.01) from that formed by the microfluidic methods.

For both microfluidic methods decreasing polymer concentration from 0.01 to 0.0001 wt% demonstrated an increase in size monodispersity with decreasing polymer concentration. However, with the decrease in concentration comes an increase in the

38 required amount of organic solvent to yield the same final weight of product. If the cost of the therapeutic agent outweighs the cost of the organic solvent, as it typically does, then the increase in production within the size range of interest for improving bioavailability greatly outweighs the cost.

Based on above results, we hypothesize two mechanisms (1) spinoidal decomposition and

(2) nucleation and growth could explain the PIN phenomenon. In the spinoidal decomposition process polymer molecules instantaneously collapse upon phase inversion

(exchange of solvent and nonsolvent) to form nanoparticles. In the nucleation and growth process, nucleation sites are created by inhomogeneous mixture sites around which polymer chains collapse to form polymer-rich and polymer-poor regions that after phase inversion yield solid polymer nanospheres. To test above hypothesizes of nanoparticle formation we varied the polymer phase flow rate within the laminar regime given that the mixing time is independent of flow rate. The experiments show a trend toward decreasing size with increasing flow rate indicating time-course-dependent nucleation and growth mechanism of formation for the resultant nanosphere population within the range of conditions tested. If the mechanism were instantaneous, flow rate is not expected to affect the resultant population. Additionally the formation of polymer network structures at higher concentrations morphologically suggests crowded nucleation sites that join as polymer chain collapse occurs as water freezes around nucleation sites to form snow flakes.

39 Two competing theories have been used to explain PIN: nucleation and growth and spinoidal decomposition. The linear dependence of mean diameter on flow rate supports the time course dependent mechanism, nucleation and growth, as does the formation of the polymer network within the microchannels observed at higher polymer concentrations.

2.4 Conclusions

Polymer nanospheres formulated in organic solvents were produced for the first time on a microfluidic platform by the PIN method. All but two of the described experimental conditions produced populations of nanospheres with at least 90% in the desired size range for uptake by nonphagocytitic cells, 200−1000 nm as shown in Table 2.1.

Microfluidic investigations show a nonlinear dependence of population mean diameter on polymer concentration. In the lower range of concentrations tested particle size appeared to be insensitive to concentration evidencing an optimal concentration that will minimize solvent usage and maximize throughput for nanospheres in the desired size range. Mean nanosphere diameter varied linearly with flow rate in the glass capillary tube studies within the range of flow rates from 1−100 nL/s, making solvent phase flow rate a very useful control parameter for further tuning resultant size distribution given an optimal concentration within the range of flow rates examined. Moreover, the experimental setups have and will continue to shed light on the mechanism of nanosphere production by PIN.

40 In conclusion, PIN performed in laminar flow conditions yields polymer nanospheres of the optimal size for oral drug delivery to absorptive GI epithelial cells.[1,27] Glass chip microfluidics provides a platform in which conditions for polymer nanosphere formation can be studied quickly and inexpensively. The glass microfluidic platform will prove useful for rapidly investigating the experimental parameters necessary for producing polymer nanospheres containing therapeutic agents. Production scale-up is easily achieved both in series (using larger channels) and in parallel (using many chips or glass capillary tube setups at once) allowing for continuous-flow production. The population of nanospheres produced by the glass capillary tube method using 0.001 wt % PMMA at a flow rate of 1 nL/s is statistically similar ( P < 0.05) to that produced by the flow pinching microchip method at the same concentration with a solvent flow rate of 0.0932 nL/s flow rate indicating a link between the two platforms that could be used to enable testing production parameters in small batches and then rapidly scaling production. From a mechanistic standpoint that flow pinching produces a nanosphere population more similar to the glass capillary tube method than the no flow pinching setup indicates that the availability of nonsolvent influences the resultant size distribution, which also supports a nucleation and growth mechanism of formation.

Additionally, the glass microfluidics platform and simple chip design will allow for integration of other applications (e.g., dissolution profiling) on the same chip yielding laboratory-on-a-chip technologies that will greatly accelerate data acquisition regarding nanosphere formation by phase inversion. Data obtained from varying flow rate in the glass capillary tube experiments suggests that PIN formation of nanospheres under the

41 conditions investigated is mediated by nucleation and growth. As the mechanisms and kinetics of cellular nanosphere uptake are elucidated, controlled transport PIN will offer a technique for drug delivery scientists to improve cellular uptake of the increasing number of hydrophobic and biologic new chemical entities.

Controlling the introduction of the solvent phase into the nonsolvent improves the reproducibility of manufacturing conditions. Through introducing microfluidic-controlled transport conditions we aim to determine the effect of three flow configurations of polymer and non solvent on the mean particle diameter and to explore mechanistic hypotheses of PIN polymer nanosphere formation.

Figure 2.1: (a) Full microchip assembly described from the top piece down: (i) pressure manifold with 8 threaded side ports for pressure lines. (A viton gasket to seal with the caddy is not shown.) (ii) Teflon caddy with through holes and O-ring seat. When assembled, the Teflon well has a capacity of 30 µL, while 6 µL is a functional minimum volume. Experiments were conducted with 20 µL volumes. (iii) Viton o-rings (iv) double layered, thermally bonded, glass microchip (wells and channels not shown). (v) threaded compression plate (vi) threaded microscope mounting plate. (b) Photograph of glass microfluidic cross chip containing a network of microchannels paired to microwells. Fluids move from reservoirs 1, 2, and 3 to reservoir 4, which serves as a collection reservoir for the resultant nanospheres. (c) Schematic of microfluidic flow pinching setup for nanosphere production. (d) A photograph of the microchannel flow showing 1:3 pinching. (e) phase contrast image of the branched polymer microstructure that forms upon phase inversion when the polymer concentration of the solvent is large (>0.01 wt %). (f) Schematic of the glass capillary tube setup for nanosphere production.

Figure 2.2: (a) Exemplary scanning electron micrograph (SEM) of PIN nanospheres produced (b) NIH ImageJ generated ellipsoid outlines generated during nanosphere population size measurements.

Figure 2.3: (a) Scanning electron micrograph (SEM) of nanospheres produced off chip. 0.01 wt % PMMA was used. (b) Size histogram of nanospheres.

Figure 2.4: SEM of a population of nanospheres produced by glass chip microfluidics.

Figure 2.5: Size histograms of nanosphere populations produced by the flow pinching glass chip microfluidics methods.

Figure 2.6: Size histograms of nanosphere populations produced by glass chip microfluidics method without flow pinching.

Figure 2.7: Mean nanosphere diameter plotted as a function of polymer concentration for both microfluidic nanosphere production methods. ** = Statistical difference between methods, p < 0.01; † = statistical difference between concentrations, p < 0.001.

Figure 2.8: (a) Nanosphere population size histograms produced by the glass capillary tube method compared with both microchip production methods formed using the same polymer concentration. (b) Mean diameter of nanospheres produced by the glass capillary tube method plotted as a function of flow rate. A linear best fit line is shown. ** = Statistical difference between 1 and 10 nL/s flow rates, p < 0.01; † = 100 nL/s is statistically different from 1 and 10 nL/s flow rates, p < 0.001.

Polymer Solvent Phase Flow Rate Mean Diameter 90% > 90% < 200-1000 nm Method Concentration (nL/s) (nm) x(nm) x(nm) (%) (wt %) off chip 0.01 uncontrolled/high 660 419 1047 88 flow pinching microchip 0.0001 0.0932 517 417 603 100 flow pinching microchip 0.001 0.0932 533 370 695 100 flow pinching microchip 0.01 0.0932 731 415 987 90.9 no flow pinching microchip 0.0001 0.3141 525 365 652 97.6 no flow pinching microchip 0.001 0.3141 529 328 791 98.6 no flow pinching microchip 0.01 0.3141 846 353 1628 69.9 glass capillary tube 0.001 1 508 243 761 91.8 glass capillary tube 0.001 10 431 234 647 96.7 glass capillary tube 0.001 100 334 207 526 89.2

Table 2.1: Nanosphere population characteristics including mean diameter, the diameter of a sphere that 90% of the population is larger than (90% > x), the diameter of a sphere that 90% of the population is smaller than (90% < x), and the percentage of the population of nanospheres that are within the size range ideal for non-phagocytotic cellular uptake (200−1000 nm in diameter).

51 2.5 References

1. Mathiowitz, E., Jacob, J. S., Jong, Y. S., Carino, G. P., Chickering, D. E.,

Chaturvedi, P., Santos, C. A., Vijayaraghavan, K., Montgomery, S., Bassett, M.,

and Morrell, C. Biologically erodable microsphere as potential oral drug delivery

system. Nature 1997, 386 (6623), 410−414.

2. Rejman, J., Oberle, V., Zuhorn, I. S., and Hoekstra, D.Size-dependent

internalization of particles via the pathways of clathrin-and caveolae-mediated

endocytosis. Biochem. J. 2004, 377, 159−169.

3. Florence, A. T., Hillery, A. M., Hussain, N., and Jani, P. U.Nanoparticles as

Carriers for Oral Peptide Absorption - Studies on Particle Uptake and Fate. J.

Controlled Release 1995, 36(1−2), 39− 46.

4. Jani, P., Halbert, G. W., Langridge, J., and Florence, A. T. The Uptake and

Translocation of Latex Nanospheres and Microspheres after Oral-Administration

to Rats. J. Pharm. Pharmacol. 1989, 41(12), 809.

5. Jani, P., Halbert, G. W., Langridge, J., and Florence, A. T. Nanoparticle Uptake

by the Rat Gastrointestinal Mucosa-Quantitation and Particle-Size Dependency. J.

Pharm. Pharmacol. 1990, 42(12), 821−826.

6. Steinbacher, J. L., and McQuade, D. T. Polymer chemistry in flow: New

polymers, beads, capsules, and fibers. J. Polym. Sci., Part A 2006, 44(22),

6505−6533.

7. Ganan-Calvo, A. M. Polyphonic microfluidics. Nat. Phys. 2005, 1(3), 139−140.

8. Whitesides, G. M.The origins and the future of microfluidics. Nature 2006, 442

(7101), 368− 373.

52 9. Utada, A. S., Lorenceau, E., Link, D. R., Kaplan, P. D., Stone, H. A., and Weitz,

D. A. Monodisperse double emulsions generated from a microcapillary device.

Science 2005, 308(5721), 537−541.

10. Dendukuri, D., Pregibon, D.C., Collins, J., Hatton, T. A., and Doyle, P. S.

Continuous-flow lithography for high-throughput microparticle synthesis. Nat.

Mater. 2006, 5(5), 365−369.

11. Dendukuri, D., Tsoi, K., Hatton, T. A., and Doyle, P. S.Controlled synthesis of

nonspherical microparticles using microfluidics. Langmuir 2005, 21(6), 2113−

2116.

12. Nisisako, T., Torii, T., Takahashi, T., and Takizawa, Y.Synthesis of monodisperse

bicolored janus particles with electrical anisotropy using a microfluidic co-flow

system. Adv. Mater. 2006, 18(9), 1152.

13. Sung, K. E., Vanapalli, S. A., Mukhija, D., Mckay, H. A., Millunchick, J. M.,

Burns, M. A., and Solomon, M. J. Programmable fluidic production of

microparticles with configurable anisotropy. J. Am. Chem. Soc. 2008, 130(4),

1335−1340.

14. Dendukuri, D., Gu, S. S., Pregibon, D. C., Hatton, T. A., and Doyle, P. S. Stop-

flow lithography in a microfluidic device. Lab Chip 2007, 7(7), 818−828.

15. Dendukuri, D., Hatton, T. A., and Doyle, P. S. Synthesis and self-assembly of

amphiphilic polymeric microparticles. Langmuir 2007, 23(8), 4669−4674.

16. Abraham, S., Jeong, E. H., Arakawa, T., Shoji, S., Kim, K. C., Kim, I., and Go, J.

S. Microfluidics assisted synthesis of well-defined spherical polymeric

53 microcapsules and their utilization as potential encapsulants. Lab Chip 2006, 6(6),

752−756.

17. Liu, K., Ding, H. J., Chen, Y., and Zhao, X. Z. Droplet-based synthetic method

using microflow focusing and droplet fusion. Microfluidics Nanofluidics 2007,

3(2), 239−243.

18. Liu, K., Ding, H. J., Liu, J., Chen, Y., and Zhao, X. Z. Shape-controlled

production of biodegradable calcium alginate gel microparticles using a novel

microfluidic device. Langmuir 2006, 22(22), 9453−9457.

19. Martin-Banderas, L., Flores-Mosquera, M., Riesco-Chueca, P., Rodriguez-Gil, A.,

Cebolla, A., Chavez, S., and Ganan-Calvo, A. M. Flow focusing: A versatile

technology to produce size-controlled and specific-morphology microparticles.

Small 2005, 1(7), 688-692.

20. Okushima, S., Nisisako, T., Torii, T., and Higuchi, T. Controlled production of

monodisperse double emulsions by two-step droplet breakup in microfluidic

devices. Langmuir 2004, 20(23), 9905−9908.

21. Seo, M., Nie, Z. H., Xu, S. Q., Mok, M., Lewis, P. C., Graham, R., and

Kumacheva, E. Continuous microfluidic reactors for polymer particles. Langmuir

2005, 21(25), 11614−1622.

22. Carino, G. P., Jacob, J. S., and Mathiowitz, E.Nanosphere based oral insulin

delivery. J. Controlled Release 2000, 65(1−2), 261−269.

23. Kerby, M. B., Lee, J., Ziperstein, J., and Tripathi, A. Kinetic measurements of

protein conformation in a microchip. Biotechnol. Prog. 2006, 22(5), 1416−1425.

54 24. Kerby, M. B., Legge, R. S., and Tripathi, A. Measurements of kinetic parameters

in a microfluidic reactor. Anal. Chem. 2006, 78(24), 8273−8280.

25. Lee, J., and Tripathi, A. Intrinsic viscosity of polymers and biopolymers measured

by microchip. Anal. Chem. 2005, 77(22), 7137−7147.

26. Larson, R. G. The Structure and Rheology of Complex Fluids. Oxford University

Press: New York 1999

27. Norris, D. A., Puri, N., and Sinko, P. J. The effect of physical barriers and

properties on the oral absorption of particulates. Adv. Drug Delivery Rev. 1998,

34(2−3), 135−154.

Chapter 3

Diuretic Bioactivity Optimization of Furosemide

Abstract

Furosemide is a loop diuretic widely used by congestive heart failure (CHF) patients to rid excess body water, reducing blood pressure, and mobilizing edemas. However, due to the narrow window of furosemide absorption, occurring only in the proximal gastrointestinal tract, only immediate release oral formulations are clinically available.

Comparisons of bolus and continuous administration of furosemide in intravenous settings demonstrate that continuous administration at lower concentrations produced greater diuretic efficiency and reduced subsequent hospitalization rates in patients experiencing severe CHF. We report a systematic investigation of the diuretic bioactivity profiles of phase inversion micronized furosemide and furosemide co-precipitated with

Eudragit L100, as well as their blends with stock furosemide, targeted at reducing the rapid spike in diuresis associated with immediate release formulations while maintaining cumulative urine output. Of the formulations tested, blends of micronized furosemide and

56 Eudragit L100 polymer with stock furosemide demonstrated optimal diuretic bioactivity profiles in a rat model.

3.1 Background and Introduction

In its current pharmaceutical formulation Lasix or Furosemide, the most commonly used diuretic for treating congestive heart failure (CHF), is only absorbed in the proximal small intestines [1]. Because it is a weak acid (pKa 3.9), furosemide is protonated only in the acidic lumen of the stomach and proximal small intestines [2]. In the more distal gastrointestinal (GI) tract, furosemide becomes deprotonated and carries a negative charge that significantly reduces its ability to cross biological membranes [1].

Compounding the site specificity of absorption, furosemide has very low water solubility leading to its classification as a class IV narrow absorption window therapeutic [3,4].

In keeping with its absorptive properties, furosemide bioactivity is characterized by a sharp onset of diuresis, sometimes referred to as the “Niagara effect” in Depomed press releases, that occurs when furosemide blocks the Na-K-2Cl co-transporter (NKCC) in the thick ascending limb of the kidneys causing diuresis [5]. CHF patients experience rapid diuresis daily from Lasix followed by an increase in water intake over the course of the day that leads to peaks and troughs in blood pressure and often leads to patient tolerance of Lasix that requires increased dosing over time [6]. The bioactivity profile of current formulations of furosemide is inconvenient for patients, produces inefficient diuresis, and causes increased renal stress with dose escalation.

57 Given the magnitude of the problem, more than 5.8 million Americans are living with

CHF, numerous investigators have created controlled release formulations of furosemide with varying degrees of success [7]. Further evidencing the need for a controlled release formulation, continuous release intravenous administration of furosemide has demonstrated benefit when compared to bolus intravenous control patients in the form of reduced hospitalization to human patients [8,9]. The primary obstacle to creating a controlled release furosemide formulation is the limited residence in the proximal GI. In

1997, Santus et al. studied GR delivery of furosemide to validate a generalized

mucoadhesive gastroretentive drug delivery system that relies on the mucoadhesion of a

blend of carbomer and hydroxypropyl methyl cellulose first in a rabbit model and then in

six healthy human volunteers [10]. Santus et al. observed an insignificant increase in

gastric residence time with bioadhesive formulations [10]. In more recent studies,

Sakkinen et al. in 2003 and again in 2005 studied the ability of bioadhesive

microcrystalline chitosan to increase gastric residence time and reached similar

conclusions to Santus et al . [10-12].

In 2000 Ozdemir et al. prepared a floating dosage form of furosemide to enhance

bioavailability in six healthy human volunteers [13]. Residence time of the floating pill

was determined by radiography [13]. Each subject imbibed 100ml of water hourly and

although 100ml of water hourly may not be excessive during diuretic testing, it may

dramatically effect the residence time of the floating pill within the stomach [13]. Davis

asserts that in the fasted state, very little fluid remains within the stomach, which has led

to the irreproducibility and failure of previous floating-pill studies [1]. While Ozdemir et

58 al. present convincing data estimating gastric retention of up to six hours, Davis indicates

that the results are of questionable clinical relevance because of the prescribed drinking

schedule [13]. Depomed formulated furosemide into a swelling gel that once ingested

became too large to pass through the pyloric sphincter, leading to gastric retention [1].

Phase II clinical results reported in press releases showed inconsistent reduction in

urinary urgency in CHF patients. In addition to gastroretentive strategies, Terao et al.

tested the ability of methacrylate-derivative, pH altering polymers to widen the

absorption window of furosemide by lowering the pH in the distal GI [14]. And Shin et

al. increased the water solubility of furosemide by cogrinding and coprecipitation with

the hydrophilic polymer, crospovidone [15].

While gastric retention of furosemide has proven challenging, spherical crystallization

and co-precipitation with pH-altering polymers show great potential to alter the diuretic

profile of furosemide without requiring prolonged gastric residence. We focused our

efforts on utilizing phase inversion-based precipitation techniques in combination with GI

pH-altering polymers, and blends with stock furosemide to reduce the Niagara effect

without reducing the total urinary output over a 10 hour period in a rat model. Ideally, a

linear urine mass output as a function of time profile could reduce the Niagara effect,

increase diuretic efficiency, and reduce renal stress without requiring gastric retention.

59 3.2 Materials and Methods

Furosemide micronization by phase inversion

Furosemide (Sigma Aldrich, St Louis, MO) was dissolved in ethyl acetate (Fluka, St

Louis, MO) at a concentration of 4mg/ml, near its maximum solubility. The furosemide solution was poured into an excess of miscible non-solvent, petroleum ether (Sigma

Aldrich, St Louis, MO), at a volume ratio of 1:20 ethyl acetate to petroleum ether causing phase inversion of the furosemide [16]. Once phase inverted, the petroleum ether suspension is filtered using a Millipore stainless steel filter column (Billerica, MA) fitted with a 0.2 micron mixed cellulose ester filter membrane (Millipore, Billerica, MA). The furosemide retentate is then transferred on the membrane into 50ml conical tubes topped with Kimwipes (Kimberly-Clark, Mississaua, Ontario) held in place by rubber bands and wrapped in aluminum foil to reduce light exposure. The phase inverted furosemide is then placed inside a lyophilization jar (VirTis, Gardiner, NY) and lyophilized for 24 hours until the powder is fully dried. Dried powder is then separated from the filter membrane and stored in amber glass containers to minimize light-induced degradation

[17].

Co-precipitation with Eudragit L100

Furosemide was dissolved in ethyl acetate as in the precipitation procedure and then mixed with 1w/v% Eudragit L100 (Rohm GmbH, Darmstadt, Germany) pH-sensitive, acrylic acid derived polymer in ethanol (Sigma Aldrich, St Louis, MO) to create a 1:1 dissolved mass ratio of furosemide to Eudragit. The furosemide and Eudragit solution was poured into an excess of petroleum ether (Sigma Aldrich, St Louis, MO), a non-

60 solvent for both furosemide and Eudragit, leading to co-precipitation of the polymer and drug. After co-precipitation, the polymer and drug are filtered and lyophilized as with phase inversion micronized furosemide. Because Eudragit L100 has a pKa of 6.0, at small intestinal pH >6.0, protonated carboxylic acid residues liberate hydrogen ions locally reducing pH temporarily as the polymer dissolves [18-21].

Scanning electron micrograph (SEM) analysis of furosemide doses

Conductive, double-sided carbon tape was overlaid on top of aluminum SEM stubs. Dry powder samples of stock furosemide, phase inversion micronized furosemide, and co- precipitated furosemide and Eudragit L100 were transferred onto the carbon tape. The

SEM stubs were then sputter coated with 50-100 Ǻ of gold-palladium (Emitech K550,

Kent, UK). Each stub was imaged by SEM (Hitachi S-2700, Tokoyo, Japan) with an

accelerating voltage of 8kV. The electron beam was aligned and digital images were

obtained at 1,000x and 5,000x (Quartz Imaging Corporation, Vancouver, BC).

Differential scanning calorimetry (DSC) analysis of furosemide doses

3-5 mg of each furosemide powder dosage form was weighed in aluminum sample pans

(Perkin-Elmer, Waltham, MA). Each sample was covered with an aluminum lid (Perkin-

Elmer, Waltham, MA) and crimped to seal the sample within the pan (Perkin-Elmer,

Waltham, MA). Sealed pans were then placed into a DSC7 (Perkin-Elmer, Waltham,

MA), controlled by Pyris software (Perkin-Elmer, Waltham, MA). Samples were cooled

to -25°C then heated to 250°C and compared to an empty reference pan to quantify heat

flow as a function of sample temperature during thermal transitions.

61 Fourier transform infrared spectroscopy (FTIR) analysis of furosemide doses

Infrared transmittance of the powdered furosemide samples was measured using total internal reflectance FTIR (Spectrum One, Perkin Elmer, Waltham, MA). Absorption peaks were labeled using Spectrum software (Perkin Elmer, Waltham, MA). IR spectra were analyzed to evaluate the presence of functional groups.

Dose preparation

At the start of each experiment, the rats were weighed. Each formulation was prepared and then loaded into size 9 gelatin capsules using the gelatin capsule filler (Torpac,

Fairfield, NJ). Each gelatin capsule was weighed on a microbalance (AD-4 Autobalance,

Perkin Elmer, Waltham, MA) prior to and after drug loading to create oral doses of 2.5,

5, or 10 milligrams of drug per kilogram of body mass within 0.5mg of dose mass.

Oral administration

Each rat was induced in an induction chamber with 3.5% isoflurane (Novation, Irving,

TX) for 5-10 minutes. Once anesthetized, the rat was removed from the induction chamber and dosed with a size 9 gelatin capsule containing one of the furosemide formulations using the gelatin capsule dosing syringe (Torpac, Fairfield, NJ). As a negative control the same rats were administered empty gelatin capsules without furosemide to quantify the basal urine mass output for comparison. Upon recovery from anesthesia, each rat was then transferred to a metabolic cage for urine collection over a 10 hour period.

62 Diuretic bioactivity analysis

Twelve albino, male, Sprague-Dawley rats (450-750g) were employed in the bioactivity analysis study. Between dosing with furosemide, rats were housed in standard bedded cages in accordance with NIH and IACUC guidelines. Immediately after oral gavage with a furosemide formulation, rats were housed individually in a metabolic cage rack

(Unifab Cages, Kalamazoo, MI). Subjects had access to food and water ad libitum throughout the study. Metabolic cages were equipped with wire grating floors that allowed for free passage and collection of excreted material while containing the rat.

Feces were caught beneath the large opening wire grating floor by smaller opening wire mesh screen and urine continued through the screen into a funnel for collection in pre- weighed glass scintillation vials (Cole-Parmer, Vernon Hills, IL). At two hour intervals after dosing, the glass scintillation vials were weighed (Mettler Toledo, Columbus, OH) to quantify urine output as a function of time and then replaced. Increased urine output above baseline values is used as a non-invasive measure of diuretic activity of the various furosemide formulations examined. After each 10 hour study, rats were housed in bedded cages for a recovery period of at least 48 hours between metabolic cage studies.

3.3 Results and Discussion

Furosemide dose physiochemical analysis

Scanning electron microscopy (SEM) shows the angular, crystalline nature of stock, pharmaceutically available furosemide having individual crystals with length of ~5 µm

(Figure 3.1a,i). Image analysis shows that phase inversion of furosemide alone decreases the majority of crystal lengths, or micronized the furosemide, from ~5 µm in the stock

63 formulation to <1 µm (Figure 3.1a,ii). Co-precipitation of furosemide with L100 in equal

masses results in more needle-like crystal formation than stock furosemide (Figure 3.1a,

iii). SEM reported by Aceves and Hernandez of furosemide dissolved in methanol and

phase inverted by evaporation also resulted in reduced crystal size [22].

Differential scanning calorimetry (DSC) shows that the thermal decomposition

temperature of stock furosemide, 217°C, is unchanged by phase inversion micronization

(Figure 3.1b) [23]. Eudragit L100, as a thermoplastic polymer, exhibits a glass transition

temperature at 65°C and a melting temperature at 220°C. Due to the similarity of the

thermal decomposition temperature of furosemide and the melting temperature of L100,

mixed formulations show the beginning of an endothermic melting transition interrupted

by the exothermic decomposition. Both the co-precipitated and physically mixed solid

dispersion of furosemide and Eudragit L100 demonstrate very similar thermal behavior

indicating that neither co-precipitation nor physical mixing yields covalent bonding.

FTIR analysis of phase inversion micronized furosemide shows no substantial difference

from stock furosemide (Figure 3.1c). L100 has a broad peak at 1705 cm -1, corresponding to carbonyl stretching that is not present in either the co-precipitated or physically mixed formulations. Disappearance of the carbonyl peak indicates possible interaction and stabilization by the amine group of the furosemide. Aceves and Hernandez report that both solid dispersion and precipitation of furosemide with Eudragit R/L-100 showed a loss of the amine peak at 3400cm -1 and translation of the 1900cm -1 carbonyl peak

64 indicating a secondary interaction between the quaternary ammonium groups of R/L-100 and the carbonyl groups of furosemide [22].

Each of the furosemide formulations analyzed were then administered at varying doses alone and mixed with stock furosemide, and were compared to basal urine output and stock furosemide to quantify changes in the bioactivity profiles resulting from the physiochemical alterations induced by phase inversion.

Oral furosemide dose escalation

In order to quantify how phase inversion micronized and L100 mixtures with furosemide altered urine output as compared to stock furosemide, rats were orally dosed gelatin capsules containing furosemide doses and housed in metabolic cages enabling urine isolation and collection every 2 hours for a total of 10 hours. Each dose was administered to 3 rats and the average urine mass produced in the two hour time periods is reported as compared to the average output of the same rats that received a sham dosage without any furosemide, referred to in the figures as the basal output. To determine the minimum necessary dose required to sufficiently emulate the Niagara effect, a dose escalation study was performed. The bioactivity profile of stock furosemide was compared to that of micronized furosemide and an equal parts mixture of stock and micronized furosemide.

Additionally, the bioactivity response of stock furosemide was compared to that of furosemide co-precipitated with Eudragit L100, an equal parts mixture of stock furosemide and co-precipitated furosemide with L100, and a physically mixed solid dispersion of equal parts stock furosemide and L100. All furosemide formulations

65 produced greater cumulative urine output 10 hours after dosing than basal output at 2.5 mg/kg (Figure 3.3) and 5 mg/kg (Figure 3.4). Additionally, 5 mg/kg doses produced greater urine output than 2.5 mg/kg doses. However, the trends were statistically insignificantly different from basal output (p>0.05) indicating that higher furosemide dosing was required to mimic the bioactivity profile observed in humans.

When the dose was increased to 10mg/kg (N=3), the mean urine mass output 2 hours after dosing was statistically significantly 12.2x greater than that of basal urine output and 8.6x greater than the output produced after administration of an equivalent dose of micronized furosemide (p<0.01). The sharp increase in urine production observed 2 hours following administration of 10mg/kg furosemide mimics the Niagara effect reported in clinical use. Yet by hour 10, the cumulative urine output caused by the administration of stock and micronized doses were statistically insignificantly different with the stock furosemide producing 1.02x the total urine output of the micronized dose (Figure 3.5a).

Therefore, the micronized dose demonstrated similar diuretic activity without the Niagara effect in hour 2. However, the increase in cumulative urine output in hour 4 of the micronized dose indicates that micronization alone may merely delay the Niagara effect.

Given the delay in the Niagara effect produced by micronization, combinations of stock furosemide and micronized were mixed at lower doses to produce a combined effect of maintaining diuresis while reducing the Niagara effect. Towards that end, the equal parts mixture of stock and micronized furosemide demonstrates 47% less urine output at hour

2 than stock furosemide and maintains a more linear cumulative urine output profile

(r 2=0.86) than the micronized dose alone (r 2=0.83).

66 While phase inversion micronization reduces crystal size, it also causes re-crystallization of dissolved furosemide in the hydrophobic non-solvent, petroleum ether. Crystallization of furosemide in hydrophobic media may yield increased hydrophobicity to minimize interfacial energy with the non-solvent. Therefore, although phase inversion reduces particle size it may also increase hydrophobicity leading to delayed water dissolution that corresponds to the delayed onset of pharmacological action.

With the administration of co-precipitated and solid dispersion furosemide and L100 doses at 10mg/kg, the co-precipitated dose produces 1.28x the diuresis at 2 hours than the stock furosemide (Figure 3.5b). The addition of L100 may temporarily, locally reduce pH increasing the amount of time that the furosemide spends in the protonated state, which is more apt to cross biological membranes than the anionic, deporotnated form. L100 may also act as a bioadhesive promoting prolonged intimate contact of furosemide with the GI mucosa as the crystals hydrate and dissolve. Administering 10mg/kg of L100 alone did not significantly increase urine output above basal levels indicating that the polymer alone has little or no effect on diuresis. The physical mixture of equal parts stock furosemide and L100 produced 0.15x the mean urine output of stock furosemide at 2 hours and 0.96x the mean cumulative diuresis after 10 hours of stock furosemide. The diuretic activity profile is even more linear (r 2=0.94) than that of the mixture of stock and micronized furosemide.

67 Optimization of furosemide bioactivity

The two lead candidate formulations, 10mg/kg of an equal parts mixture of stock and micronized furosemide and a solid dispersion of equal parts stock furosemide and

Eudragit L100, were administered to 6 additional rats for a total of N=9 to directly compare the doses with a larger cohort (Figure 3.6). Both doses continued to demonstrate reduced diuresis compared to stock furosemide at hour 2 and similar diuresis to stock furosemide at hour 10. The mixture of stock and micronized furosemide produced less of a Niagara effect than the stock and L100 mixture with 0.56x the mean urine output at hour 2. Additionally at hour 10, the mixture of stock and micronized furosemide produced 0.95x the cumulative diuresis of the same dose of stock alone. Finally, the diuretic profile of the stock and micronized mixture (r 2=0.85) was more linear than the

mixture with L100 (r 2=0.78). By the established parameters for the optimal bioactivity

profile, the equal parts mixture of stock and micronized furosemide performed the best of

all formulations tested followed by the equal parts mixture of stock furosemide and

Eudragit L100.

Clinical potential of bioactivity optimized oral furosemide

Physiochemical analysis by FTIR and DSC indicates that furosemide does not undergo

any chemical change in response to phase inversion micronization, co-precipitation with

L100, or physical mixing with L100. Therefore, the safety master file from the widely

clinically used furosemide should apply to the described doses. The lack of a spike in

urine output within 2 hours of oral administration observed in the equal parts mixture of

stock and micronized, as well as the equal parts mixture of L100 and stock, has excellent

68 clinical potential. Unlike a previous study by Terao et al. that reports the bioavailability

of 15mg/kg aqueous furosemide solution co-administered with 400mg/kg aqueous

Eudragit L100-55 solution, the orally administered doses contain at most 10mg/kg

Eudragit polymer [14]. While Eudragit polymers are well-tolerated in clinical practice,

minimizing Eudragit incorporation is important in a clinical, daily dosing regimen.

Additionally, the Terao et al. study utilizes an isolated loop such that the pH of the entire loop is significantly altered by the incorporation of Eudragit L100-55 [14]. In this study, furosemide is delivered orally in a gelatin capsule with little L100 polymer and therefore is unlikely to significantly alter the pH of a large segment of the intestines. Instead, we hypothesize that physical bonding occurs between the carbonyl groups of the L100 and the amine groups of furosemide promoting physical proximity within the intestines. The

L100 may locally reduce pH and supply hydrogen ions to protonate furosemide at higher pH than the drug alone [21]. Additionally, the temporary bioadhesiveness of Eudragit

L100 prior to its dissolution as an acrylic acid derived polymer may serve to promote intimate contact of the furosemide with the absorptive epithelium [10,24].

Al Gohary and El Gamal administered furosemide and Eudragit R/L-100 to humans and it reduced the Niagara effect at a dose of constant mass dose of 40mg of furosemide in healthy human volunteers [25]. However, the cumulative urine output 10 hours after oral administration was only ~53% of the stock furosemide dose. Therefore, higher furosemide doses would be required to achieve the same diuretic efficiency. If the bioactivity profile translates from the small animal trials conducted in this study to the

69 clinic, the described doses have the potential to reduce the Niagara effect and maintain diuretic efficiency obviating the need for higher doses to achieve similar diuresis.

3.4 Conclusions

Mixtures of phase inversion micronized and stock furosemide, as well as Eudragit L100 and stock furosemide, demonstrated the ability to reduce the Niagara effect while maintaining diuretic efficiency in rats. The reduced Niagara effect coupled with the similar cumulative urine output 10 hours after oral administration is promising. If the optimal bioactivity profile observed in rats translates to larger animals and humans, it may reduce the risk of ototoxicity and acute tolerance currently associated with clinical use of furosemide.

70 A i ii iii

B 150 Furosemide Micronized furosemide

Eudragit L-100 )

) L-100/furosemide co-precipitation W

W L-100/furosemide solid dispersion

m m (

( 100

p p

u u

o o

d d

n n

e e

w w

o o

l l f

f 50

t t

a a

e e

H H Heat Flow Heat EndoUpFlow (mW)

0 50 100 150 200 250 Temperature (Celsius) C

3399.47 3123.50 1493.38 1353.06 983.52 823.29 3350.91 1451.05 1072.53 944.96 3282.48 1591.54 1408.45 1261.08 1052.44 922.04 787.99 1668.84 1318.16 1015.25 882.41 743.39 1561.75 1240.19 908.99 707.73

1140.02 847.34 684.03

cece n n 3401.24 2871.81 1353.91 1072.47 823.59 1493.53 944.18

3351.30 1591.84 1447.10 1053.17 847.22 684.42 itta itta 3286.66 1669.55 1409.73 1260.68 1014.94 882.69 708.13 1562.84 1319.05 980.82 806.28

1242.45 921.88 788.78 sm

sm 1140.70 743.31

n n 2952.66 1482.17 965.05

1448.77 752.39 ra

ra 1389.06 1258.96

T T t t 1705.13

1151.68

n n

1445.91 rce rce 3350.14

1592.02 1072.36 944.05 836.11 e e 3286.98 1488.14 1260.47 1014.05 882.86 684.85

1669.60 1409.89 980.80 789.15

P P 1563.56 1319.02 743.32

PercentTransmittance 1242.60 708.53707.78 2843.35 1449.90 1141.37 909.12 3399.31 1353.01 982.84 823.41 3350.84 1484.01 1204.55 1072.411015.14 847.27 683.99 3282.05 1591.41 1408.48 1260.70 1052.41 922.09 787.95 1669.34 1562.09 1317.81 1240.21 944.62 882.35 743.24 4000 2800 2000 1600 1200 800 Wavelength (1/cm)

Figure 3.1: Characterization of oral furosemide doses. (a) Scanning electron micrographs of stock furosemide (i) , micronized furosemide (ii) , and furosemide co-precipitated with Eudragit L-100 (iii) at 1,000x (top) and 5,000x (bottom) magnification. Rod-like crystals shown in (i) are characteristic of stock furosemide [22]. The crystal size appears smaller in the micronized doses (ii) and more needle-like when co-precipitated with Eudragit L- 100 (iii). (b) Plot of heat flow as a function of temperature acquired by differential scanning calorimetry of oral furosemide doses. Stock and micronized furosemide thermally decompose ~220°C. Eudragit L-100 undergoes glass-to-rubber transition ~65°C and melts ~220°C. When Eudragit and furosemide are co-precipitated or physically mixed into a solid dispersion, the glass transition and melting temperature of L-100 are still apparent, and the polymer melt is overtaken by the thermal decomposition of furosemide both occuring at ~220°C. (c) Stacked line plot of percent transmittance as a function of wavelength acquired by Fourier transform infrared spectroscopy of the same oral doses analyzed in b. Stock and micronized formulations show nearly the same absorption pattern. Eudragit L-100 shows similar absorption to furosemide with the most notable exceptions of the additional peak at 1705cm-1 and lack of peaks ~3400cm -1. Both the co-precipitated and physically mixed formulations show similar absorption patterns indicating that neither formulation yields additional chemical bonds, although physical bonding between furosemide and L-100 remains plausible.

Figure 3.2: Rats housed in a metabolic cage rack enabling non-invasive quantification of urine output without anesthesia or handling.

72 A 15 Basal 2.5 mg/kg Stock Furosemide

] 2.5 mg/kg Micronized Furosemide 2.5 mg/kg: 50% Stock/50% Micronized Furosemide 10

5 Cumulative Urine Output [g] Cumulative UrineOutput

0 2 4 6 8 10 Hours After Dosing B 15 Basal 2.5 mg/kg Stock Furosemide 2.5 mg/kg L100 Co-precipitated Fursoemide ] 2.5 mg/kg: 50% L100/50% Stock Fursoemide 10

5 Cumulative Urine Output [g] Cumulative UrineOutput

0 2 4 6 8 10 Hours After Dosing

Figure 3.3: Cumulative bioactivity response to 2.5mg/kg micronized and Eudragit L100- incorporated oral furosemide doses as compared to stock (N=3). (a) Bioactivity profiles of micronized and physically mixed solid dispersion of equal parts stock and micronized furosemide as compared to basal and stock furosemide. Micronized shows a statistically insignificant trend towards increased urine output at all timepoints as compared to stock furosemide. (b) Bioactivity profiles of Eudragit L100 incorporated into oral furosemide doses by co-precipitation and physical mixing to form a solid dispersion. L100 co- precipitated furosemide doses show a statistically insignificant trend towards increased cumulative urine output at hours 4-10, and decreased urine output at hour 2 as compared to stock. Error bars depict s.e.m.

73 A 15 Basal 5 mg/kg Stock Furosemide 5 mg/kg Micronized Furosemide 5 mg/kg: 50% Stock/50% Micronized Furosemide ]

5 Cumulative Urine Output [g] Cumulative UrineOutput

0 2 4 6 8 10 Hours After Dosing B 15 Basal 5 mg/kg Stock Furosemide 5 mg/kg L100 Co-precipitated Fursoemide 5 mg/kg: 50% L100/50% Stock Fursoemide ]

5 Cumulative Urine Output [g] Cumulative UrineOutput

0 2 4 6 8 10 Hours After Dosing

Figure 3.4: Cumulative bioactivity response to 5mg/kg micronized and Eudragit L100- incorporated oral furosemide doses as compared to stock (N=3). (a) Equal parts mixture of stock and micronized shows a statistically insignificant trend towards increased urine output at all timepoints as compared to stock furosemide. (b) L100 co-precipitated furosemide doses also show a statistically insignificant trend towards increased cumulative urine output at all timepoints compared to stock. All 5mg/kg doses produce greater urine output than 2.5mg/kg doses. Error bars depict s.e.m.

74 A 25 Basal 10 mg/kg Stock Furosemide

] 20 10 mg/kg Micronized Furosemide 10 mg/kg: 50% Stock/50% Micronized Furosemide

** **

Cumulative Urine Output [g] CumulativeUrineOutput 5

0 2 4 6 8 10 Hours After Dosing B Basal 25 10 mg/kg Stock Furosemide 10 mg/kg L100 Co-precipitated Fursoemide

] 20 10 mg/kg Eudragit L100 10 mg/kg: 50% L100/50% Stock Fursoemide

10 * *

Cumulative Urine Output [g] CumulativeUrineOutput 5

0 2 4 6 8 10 Hours After Dosing

Figure 3.5: Comparison of bioactivity in response to 10mg/kg oral doses of micronized and L-100 mixed formulations. (a) Stock furosemide induces a statistically significantly higher urine output 2 hours after oral dosing than in response to micronized and basal output with 10mg/kg doses (p<0.01, N=3). Increased urine output within the first 2 hours mimics the Niagara effect experienced clinically indicating that 10mg/kg is the appropriate dose for testing the effectiveness of formulations to reduce the Niagara effect without reducing cumulative urine output 10 hours after administration. Both micronized and an equal parts mixture of stock and micronized cause less urine output than stock furosemide at 2 hours and both produce similar cumulative diuresis by hour 10. The linearity of the bioactivity profile is greater for the mixture of stock and micronized furosemide (R 2=0.86) than micronized alone (R 2=0.83), therefore it was chosen as the lead micronized formulation used in further testing. (b) Although co-precipitated furosemide and L100 produce the greatest cumulative diuresis, it also produces the greatest diuresis at hour 2, statistically significantly greater than basal and that produced by a physically mixed solid dispersion of L100 and stock furosemide (p<0.05, N=3). Therefore, the co-precipitated dose produces a greater Niagara effect than stock furosemide. An equal parts solid dispersion of L100 and stock furosemide produces similar cumulative urine output at hour 10 and very little urine output by hour 2. Combined with the linearity of the bioactivity profile (R 2=0.94), the solid dispersion is the lead L100-incorporated formulation used in further testing. Error bars depict s.e.m, *p<0.05, **p<0.01.

20 Basal 10mg/kg Stock Furosemide 10 mg/kg: 50% Stock/50% Micronized Furosemide 10 mg/kg: 50% L-100/50% Stock Furosemide 15 * ** * ** ** ** 10

Grams of Urine of Grams * ** 5 Cumulative Urine Output [g] Output Cumulative Urine

0 2 4 6 8 10 Hours After Dosing

Figure 3.6: Bioactivity response to the leading micronized and L100-incorporated oral furosemide dose candidates tested in an increased size cohort (N=9) to determine the optimal oral fuorsemide formulation that reduces the Niagara effect, while maintaining cumulative diuresis at 10 hours. Both the equal parts mixture of micronized and stock furosemide, and L100 and stock furosemide produce less urine output than stock furosemide alone at hour 2 and similar cumulative urine output at hour 10. The mixture of stock and micronized produces less urine output at hour 2 and has a more linear bioactivity profile (R 2=0.85) than the L100/stock mixture (R 2=0.78) and therefore presents the optimal diuretic bioactivity observed in this study. Error bars depict s.e.m, *p<0.05, **p<0.01.

76 3.5 References

1. Davis, SS. 2005. Formulation strategies for absorption windows. Drug Disc.

Today 10 (4): 249-257.

2. Sistovaris, N; Hamachi, Y; Kuriki, T. 1991. Multifunctional Substances –

Determination of pKA Values By Various Methods. Fresinius J. Analytical

Chem. 340 (6): 345-349.

3. Murray, MD; Haag, KM; Black, PK; Hall, SD; Brater, DC. 1997. Variable

furosemide absorption and poor predictability of response in elderly patients.

Pharmacotheraphy 17 (1): 98-106.

4. Vanderwatt, JG; Devilliers, MM. 1995. The Effect of Mixing Variables on the

Dissolution Properties of Direct Compression Formulations of Furosemide. Drug

Dev. Ind. Pharm. (18): 2047-2056.

5. Gimenez, I. 2006. Molecular mechanisms and regulation of furosemide-sensitive

Na-K-Cl cotransporters. Curr. Op. Nephrology Hypertension 15 (5): 517-523.

6. Hammarlund, MM; Odlind, B; Paalzow, LK. 1985. Acute Tolerance to

Furosemide Diuresis in Humans – Pharmacokinetic-Pharmacodynamic Modeling.

J. Pharm. And Exp. Thera. 233 (2): 447-453.

7. American Heart Association. 2000 heart and stroke statistical update. Dallas

(TX): American Heart Association, 1999.

8. Dormans, TPJ; vanMeyel, JJM; Gerlag, PGG; Tan, Y; Russel, FGM; Smits, P.

1996. Diuretic efficacy of high dose furosemide in severe heart failure: Bolus

injection versus continuous infusion. J. Am. College Cardio. 28 (2): 376-382.

77 9. Salvador, DRK; Rey, NR; Ramos, GC; Punzalan, FER. 2005. Continuous

infusion versus bolus injection of loop diuretics in congestive heart failure.

Cochrane Database of Systematic Reviews (3).

10. Santus, G; Lazzarini, C; Bottoni, G; Sandefer, EP; Page, RC; Doll, WJ; Ryo, UY;

Digenis, GA. 1997. An in vitro in vivo investigation of oral bioadhesive

controlled release furosemide formulations. Eur. J. Pharm. Biopharm. 44 (1): 39-

52.

11. Sakkinen, M; Marvola, J; Kanerva, H; Lindevall, K; Ahonen, A; Marvola, M.

2006. Are chitosan formulations mucoadhesive in the human small intestine? An

evaluation based on gamma scintigraphy. Int. J. Pharm. 307 (2): 285-291.

12. Sakkinen, M; Tuononen, T; Jurjenson, H; Veski, P; Marvola, M. 2003. Evaluation

of microcrystalline chitosans for gastro-retentive drug delivery. Eur. J. Pharm.

Sci. 19 (5): 345-353. S34, Suppl. 1.

13. Ozdemir, N; Ordu, S; Ozkan, Y. 2000. Studies of floating dosage forms of

furosemide: In vitro and in vivo evaluations of bilayer tablet formulations. Drug

Dev. Ind. Pharm. 26 (8): 857-866.

14. Terao, T; Matsuda, K; Shouji, H. 2001. Improvement in site-specific intestinal

absorption of furosemide by Eudragit L100-55. J. Pharm. Pharma. 53 (4): 433-

440.

15. Shin, SC; Oh, IJ; Lee, YB; Choi, HK; Choi, JS. 1998. Enhanced dissolution of

furosemide by coprecipitating or cogrinding with crospovidone . Int. J. Pharm.

175 (1): 17-24.

78 16. Carino, GP; Jacob, JS; Mathiowitz, E. 2000. Nanosphere based oral insulin

delivery. J. Controlled Release 65 (1-2): 261-269.

17. Thoma, K; Klimek, R. 1991. Photostabilization of Drugs in Forms without

Protection from Packaging Materials. Int. J. Pharm. 67 (2): 169-175.

18. Cilurzo, F; Minghetti, P; Selmin, F; Casiraghi, A; Montanari, L. 2003.

Polymethacrylate salts as new low-swellable mucoadhesive materials. J.

Controlled Release 88 (1): 43-53.

19. Di Colo, G; Falchi, S; Zambito, Y. 2002. In vitro evaluation of a system for pH-

controlled peroral delivery of metformin. J. Controlled Release 80 (1-3): 119-128.

20. Kislalioglu, MS; Khan, MA; Blount, C; Goettsch, RW; Bolton, S. 1991. Physical

Characterization and Dissolution Properties of Ibuprofen – Eudragit

Coprecipitates. J. Pharm. Sci. 80 (8): 799-804.

21. Moustafine, RI; Kabanova, TV; Kemenova, VA; Van den Mooter, G. 2005.

Characteristics of interpolyelectrolyte complexes of Eudragit E100 with Eudragit

L100. J. Controlled Release 103 (1): 191-198.

22. Aceves, JM; Cruz, R; Hernandez, E. 2000. Preparation and characterization of

Furosemide-Eudragit controlled release systems. Int. J. Pharm. 195 (1-2): 45-53.

23. Beyers, H; Malan, SF; van der Watt, JG; de Villiers, MM. 2000. Structure-

solubility relationship and thermal decomposition of furosemide. Drug Dev. Ind.

Pham. 26 (10): 1077-1083.

24. Laulicht, B; Cheifetz, P; Tripathi, A; Mathiowitz, E. 2009. Are in vivo gastric

bioadhesive forces accurately reflected by in vitro experiments? J. Controlled

Release 134 (2): 103-110.

79 25. Al Gohary, O; El Gamal, S. 1991. Release of Furosemide from Sustained-Release

Microcapsules Prepared by Phase-Separation Technique. Drug Dev. Ind. Pharm.

17 (3): 443-450.

Chapter 4

Are in vivo gastric bioadhesive forces accurately reflected by in vitro experiments?

Abstract

Bioadhesive polymers have been used in oral drug delivery to prolong the contact of dosage forms with the site of drug absorption. Previous investigators have coated oral dosage forms in polymers that demonstrated bioadhesive properties during in vitro screens in efforts to prolong the gastric residence of drugs absorbed only in the stomach and proximal duodenum without clinical success. To further investigate the bioadhesive properties of the gastric environment, an in vivo quantitative bioadhesive fracture strength test was developed. Bioadhesive and non-bioadhesive bioerodible polymers with potential for use in oral drug delivery were tested for bioadhesive fracture strength both in vivo and in vitro . Surprisingly, no statistically significant difference was found between the bioadhesive fracture strength of fast eroding polyanhydride and slowly

81 eroding hydrophobic polymers in vivo . When the same polymers were tested in vitro , the

expected difference was observed. The lack of IVIVC ( in vitro /in vivo correlation) among bioadhesive fracture strengths reflects the clinical finding that polymers that produced strong bioadhesive forces in vitro may not achieve prolonged gastric retention in vivo due

to differences between the in vitro screening conditions and the in vivo bioadhesive

environment.

4.1 Introduction

In the late 1960s and early 1970s, investigators reported the first in vivo quantitative tensile bioadhesion measurements of marine invertebrates, particularly limpets, to rocks

[1-5]. In most experiments a linear translating motor in series with a load cell separated each limpet, by the shell, from the substrate to which it was anchored [2] and [4]. The mucus adhering the feet of the limpets to various substrates is extremely bioadhesive,

1.95–5.8 kg/cm 2 [2]. Due to the unique anatomical features of marine invertebrates,

namely the shell and the external secretion of strongly bioadhesive mucus, tensile

bioadhesion measurements were readily obtained [2].

Mammalian bioadhesion tensile testing results were first reported in 1982 when Marvola

et al . made measurements on excised intestines from freshly slaughtered sheep [6]. Martti measured the “detachment force” necessary to separate a pill from various sections of the esophagus and intestines [6]. Force was measured by adding water into a beaker until the weight of the water exceeded the bioadhesive fracture strength [6]. Since that time investigators have employed various materials testing apparatus including tensiometers

82 and microbalances to measure the fracture strength of freshly excised tissues in various states of simulated physiological conditions [2,6-16]. Investigators including Mathiowitz et al. have shown a strong correlation between in vitro fracture strength and in vivo transit time results [17-22].

Numerous in vitro material testing methods exist for quantifying bioadhesive forces that correlate with the overall goals of bioadhesive drug delivery: to promote intimate contact of a dosage with the gastrointestinal mucosa and extend gastrointestinal residence yielding increased bioavailability of a therapeutic agent [21,23-5]. However, most in vivo bioadhesion testing involves quantifying parameters associated with the goals of bioadhesion such as residence time or relative bioavailability [17-19,21,22,26]. We believe that the following work provides the first in vivo bioadhesive force measurements and the first direct comparison of bioadhesive forces in vivo and in vitro using a single testing method.

Medical bioadhesives include any of a class of biomaterials that adhere to biological substrates [25,27]. Polymer bioadhesives are used in many medical devices and drug delivery systems including transdermal patches and Gliadel wafers [21]. To date bioadhesive polymers have not achieved clinically significantly improved gastric retention time [28-30]. Numerous therapeutic agents, especially polar and anionic small molecules, would greatly benefit from improved gastric retention time [28-30].

In vivo bioadhesion measurements have consisted of transit time or relative bioavailability assays [17-19,21,22]. Prevalent methods for monitoring gastrointestinal transit time of radio-opaque or radiation emitting doses include X-ray and gamma

83 scintigraphy [17-19,21,22]. Relative bioavailability measurements are made by comparing the plasma level concentrations of drugs administered in bioadhesive per oral dosage forms compared to standard per oral dosage forms and intravenous infusions [21].

Each of these methods provides data that support or reject the bioadhesiveness of a material, which can be correlated indirectly to parameters measured in vitro .

One major obstacle in screening bioadhesives is the lack of in vivo quantitative methodologies that are directly comparable to in vitro testing data [17-20]. We report a novel means of obtaining in vivo bioadhesive fracture strength by testing through a surgically implanted, re-closable gastric cannula. Investigating the link between in vitro and in vivo bioadhesion experiments will lead to improved screening methods for bioadhesive materials and improved translational research outcomes when transitioning from bench top to preclinical trials. Quantitative in vivo bioadhesion measurements are useful in establishing if the results obtained in vitro reflect the in vivo environment. The new technique for comparing in vivo to in vitro bioadhesion measurements quantitatively provides a means for analyzing the correlation between in vitro and in vivo bioadhesive performance indicator, fracture strength.

Establishing the criteria that yield an effective bioadhesive in vivo and then linking it to in vitro data will yield improved understanding of how to design bioadhesive materials for gastroretentive oral drug delivery systems. In this paper we optimized the testing parameters (contact force, contact time, presence of PBS, and testing speed) for poly(fumaric-co-sebacic anhydride), which has demonstrated strong bioadhesive fracture strength in previous studies performed on small intestinal tissue [14,15,21,31,32]. We

84 then applied the optimized conditions to measure the bioadhesive fracture strengths of five bioerodible polymers in vivo and in vitro . Within the course of in vivo testing a gastric cannula confining apparatus was machined to limit motion of the stomach during testing and to more closely approximate the in vitro settings. Based on the results, the bioadhesive fracture strength of the loosely adherent gastric mucus layer was measured in vitro to test the hypothesis that the in vivo bioadhesive environment is governed primarily by the properties of the loosely adherent mucus.

4.2 Materials and methods

Materials selection: bioerodible polymers

Five low melting temperature, bioerodible, thermoplastic polymers that have proven orally acceptable in small animal trials were used throughout the in vitro and in vivo experiments [14,15,21,31-34]. Each polymer, upon introduction into the gastric environment presents a hydrophobic surface. In the presence of water, the polymer chains undergo hydrolysis at the water-labile bonds at varying rates, which increased the hydrogen bonding capacity of the polymers and increasing bioadhesion [21]. Three of the polymers were synthesized in-house, poly(fumaric-co-sebacic anhydride) 20:80

(PFASA2080) Mw = 12.5 kDa, poly(Adipic Anhydride) (PAA) Mw =7.5 kDa, and poly(carboxyphenoxy-co-sebacic anhydride) 20:80 (PCPHSA2080) Mw = 10 kDa. The other two polymers tested, poly(caprolactone) (PCL) (Sigma Aldrich Saint Louis, MO)

Mw = 65 kDa and poly(lactic-co-glycolic acid) 50:50 (PLGA5050) Mw = 25 kDa

(Resomer 503H, Boehringer-Ingelheim Ingelheim, Germany) were purchased.

85 PFASA2080 and PAA are fast-eroding anhydride polymers that undergo hydrolysis rapidly to expose carboxylic acid residues rapidly enough to produce hydrogen bonding to mucus during gastrointestinal transit indicating that they would be good bioadhesives

[14,15,21,31,32]. In previous studies FASA2080 has demonstrated strong bioadhesion to intestinal mucus compared to slow eroding hydrophobic polymers (e.g. PCL) by numerous techniques including everted sac, CAHN microbalance, and X-ray transit time

[14,15,21,31,32].

PFASA2080 has been one of the most successful polymers for increasing total gastrointestinal transit time [33]. In a previous investigation by our lab 90% of a population of PFASA2080 microspheres was eliminated after 34 h, while the hydrogel alginate took 20 h [33]. However, the amount of time the microspheres remain in the stomach was not studied.

PCPHSA2080 is an aromatic anhydride polymer and therefore degrades more slowly than aliphatic PAA and PFASA2080 [21]. PCL and PLGA5050 are the slowest eroding polymers of the panel and bond to mucus primarily through hydrophobic–hydrophobic interactions shown in previous studies to be significantly lower in magnitude than more rapidly eroding polyanhydride polymers [21]. We believe the in vitro and in vivo results

are the first reported rat gastric bioadhesion on all of the tested polymers.

Probe preparation

Each polymer was heated to 90 °C, at least 5 °C above the melting temperature. Stainless

steel pins were dipped into the polymer and then allowed to cool suspended head-down to

86 form polymer beads for testing. The diameter of each probe is measured by calipers

(Mitutoyo Kawasaki, Japan) and the diameter is used in projected cross-sectional area of probe–tissue contact calculations. Probes range from 1.5–2.5 mm in diameter chosen to ensure they will easily fit through the lumen of the gastric cannula (4.5 mm) during in vivo testing. Probe size was chosen based on previous studies in our lab that indicated probes on the order of a millimeter in diameter produce bioadhesive tensile forces detectable by the Texture Analyzer load cell. While not in use probes were stored at

− 20 °C in vacuum-sealed bags under nitrogen gas in the presence of Drierite (W.A.

Hammond Drierite Xenia, OH) desiccant to minimize degradation between manufacture and testing. Each probe was tested only once since contact with the testing buffer accelerates polymer degradation.

Gastric cannula surgical procedure

The cannula consists of the barrel of a polypropylene 1 cm 3 syringe (Becton Dickinson

Franklin Lakes, NJ) that has been machined to remove the dispensing tip and reduce the length to 3/4 in. The inner diameter of the gastric cannula was chosen to easily fit the polymer probes. As a result of the relatively large diameter of the gastric cannula, direct gastric cannulation was required, rather than transesophageal or nasogastric tube placement.

The modified syringe barrel is then tapped to interface with a 10–32 knurled, unslotted stainless steel thumb screw. Two tightly fitting silicone bands, 1 mm thick sections of

1/4 inch OD × 1/8 inch ID Silastic tubing (Cole Palmer Vernon Hills, IL), were fitted

87 tightly around the cannula for anchoring to the stomach serosa and dermis as diagrammed in Figure 4.1a.

Each 400–500 g albino Spague–Dawley rat was fasted overnight in a metabolic cage and then induced on 3.5% and maintained at 2.5% isoflurane adjusted to effect. Hair was clipped from the ventral rib cage to the pelvis and from the left shoulder to the left hip and prepared with iodophor to sterilize the skin. The rat was covered in a fenestrated drape and body temperature was maintained on a heating pad set to low. A 3–5 cm incision was made in the skin and ventral mid-line fascia caudal to the xiphoid process.

Upon entering the peritoneal cavity, the least vascularized portion of the greater curvature of the fundus was identified. Using 7-0 prolene a purse-string suture was made at the site of least vasculature to minimize blood loss as reported by Pare et al. [35]. Once the purse- string suture was in place, a scalpel armed with a number 11 blade punctured the full thickness of the stomach mucosa within the middle of the purse-string suture. Pressure was applied immediately using sterile gauze to achieve hemostasis.

Afterwards, the flanged finger holds of the syringe that form the base of the cannula was inserted through the puncture site into the stomach. The purse-string was pulled tightly around the cannula and secured. Then a series of 3–5 simple interrupted seromuscular sutures affixed the suture cuff to the stomach to minimize movement of the cannula with respect to the stomach.

Once the flanged portion of the cannula has been placed within the stomach, an exit point for the tube portion of the cannula is chosen in the left lateral abdominal oblique muscles and overlying skin. With another sterile number 11 blade, used because the one that

88 punctured the stomach has contacted unsterile stomach contents, the muscles and overlying skin are incised to create an opening for the barrel of the syringe that comprises the body of the cannula. The opening was widened with the scalpel until it tightly fit the cannula and the tube portion of the cannula was pushed through the opening in the muscle and skin. A silicone bumper was placed around the outside to help limit translation of the cannula with respect to the skin. The stainless steel, knurled capping thumb screw was then tightened to close the cannula between bioadhesion tests. To complete the surgical procedure, the ventral mid-line abdominal muscle fascia was closed with simple interrupted 4-0 Vicryl (Ethicon Somerville, NJ) sutures and original skin incision with 5-0 Vicryl (Ethicon Somerville, NJ) running subcuticular stitches. At the completion of the procedure Rimadyl (Pfizer New York, NY), a non-steroidal anti- inflammatory agent is administered daily for 3 days and the rats are allowed to recover for at least 10 days before beginning optimization testing.

All animals were allowed at least a two week recovery period in accordance with IACUC guidelines prior to testing protocol optimization conditions. After optimizing the in vivo bioadhesion testing parameters, at least another two weeks of recovery time was allowed and the panel of bioerodible polymers was tested. From the results of the “unconfined” bioadhesion testing in the first two rats, a device was constructed to fix the cannula tube in place during testing, referred to as “confined” in vivo testing. All animals at Brown

University were cared for according to NIH and IACUC guidelines.

89 In vivo tensile bioadhesion testing protocol

After recovery from gastric cannulation, each rat was fasted overnight in a metabolic cage to minimize stomach contents and then induced on 3.5% isoflurane anesthesia. Once induced, the rat was transferred to a nose cone for maintenance at 2% isoflurane. While anesthetized each rat was positioned supine on a heating pad set to low placed at the base of the Texture Analyzer (TA) with the gastric cannula tube facing upwards. The capping screw was removed and the stomach was flushed with phosphate buffered saline (PBS) removing any remaining chyme. In the confined testing cases a custom-machined aluminum brace was placed around the external portion of the gastric cannula to immobilize it during testing, shown in Figure 4.1b. Once the gastric cannula was secured, the lumen of the stomach was filled with PBS until the fluid level reached the top of the tube to allow observation of the fluid level, adding PBS when necessary to ensure tissue hydration.

For all experiments each probe is held by a pin vice attached to the load cell of the

Texture Analyzer and the movement of the probe is controlled by Texture Exponent software. Test speed (5 mm/s), contact force (5gf), and contact time (64 s) were determined by systematic pilot testing. Each round of testing was conducted for no longer than 4 h and a recovery period of at least 5 days elapsed between testing in the same rat in accordance with IACUC guidelines.

90 In vitro tensile bioadhesion testing

Male Sprague–Dawley rats were euthanized in a carbon dioxide chamber. A midline incision was made to access the peritoneum. The stomach was isolated and excised. The lumen was rinsed with pH 7.4 PBS and the tissue was stored in plastic bags on ice for no more than 4 h.

Tissue samples were allowed to reach ambient temperature. To prepare sections for testing the stomach was cut longitudinally along the mesenteric and anti-mesenteric borders, to expose the mucosa. Each half of the stomach tissue was placed mucus side up in a rigid, acrylic chamber and clamped in place. The chamber was filled with PBS buffer that was heated to physiological temperature (37 °C) via a circulating water bath built into the chamber, shown in Figure 4.1c.

Texture Exponent software was used to control the TA.XT plus Texture Analyzer

(Texture Technologies Corp., Scarsdale, NY/Stable Micro Systems, Godalming, Surrey,

UK) equipped with a 1 kg load cell with a sensitivity of 0.2 g. During a standard tensile

test, a pin coated in one of the panel of five polymers was brought towards rat stomach

tissue at a constant velocity of 5 mm/s. The stomach tissue was mechanically affixed to a

rigid support and the probe is brought into contact with the mucosa at a constant velocity.

When the polymer contacted the mucus, a load cell in series with the probe registers a

compressive load that increases with the depth of penetration until it achieves a specified

depth or contact force generated. For bioadhesion testing, the Texture Analyzer arm

descended at 5 mm/s until a specified target force of 5 g between the probe and the

substrate was reached. The arm held its position for 64 s. During the contact time the

91 mucosa underwent stress relaxation, in which the load-bearing mucosal glycoprotein macromolecules realign releasing the tensile load during the contact time, reducing the measured compressive load due to the viscoelastic properties of the tissue. The arm of the

Texture Analyzer then returned to its original height at the same speed while measuring tensile load.

Fracture strength determination

Before testing, the diameters ( Ф = 2R) of the probes were measured using Mituyo digital calipers. After the Texture Analyzer finished each testing cycle, the peak tensile load

(PTL) of each run was identified and recorded. The contact area ( A0) between the spherical probe and the stomach tissue mucosa is the surface area of the spherical cap interacting with the substrate. The projected surface area (PSA) was estimated to be

2 2 Area = A0 = πR − π(R − a) where R is the radius of the sphere and “ a” is the probe

penetration depth as described previously [21]. The penetration depth “ a” was determined

after each tensile test on the Texture Analyzer by measuring the distance traveled from

the point at which the probe began interacting with the mucosa until it reached the

contact force. The projected surface area, rather than the surface area of the contact is

used in calculating tensile fracture strength since load is only measured along the axis of

motion. Therefore the measured loads act only through the horizontal cross-section of the

portion of the spherical probe contacting the mucosa. After each set of tests, the normal

stress at the de-adherence point (Fracture Strength = PTL / A0) was calculated as a

quantitative measure of bioadhesion. The projected surface area A0 was measured and

calculated as shown in Figure 4.2.

92 In vitro mucus–mucus bond strength testing

Rat stomach tissue was excised and a portion of gastric tissue was then removed and the serosal side was dried with a Kimwipe (Kimberly-Clark Roswel, GA). The serosal side of the gastric tissue punch was then affixed to a hemispherical 10-32 nylon cap nut (Small

Parts, Inc. Mirimar, FL) by cyanoacrylate glue (Henkel Consumer Adhesives, Inc. Avon,

OH) giving the mucus in a hemispherical morphology for testing as shown in Figure 4.1c.

The cap nut was then placed at the end of a 10-32 threaded nylon rod (McMaster-Carr

Princeton, NJ), which was held by a vise grip attached to the TA. The mucus probe was then tested for bioadhesive fracture strength. Since the mucus probe was cut from the same tissue as the testing sample it has the same compressive modulus. Therefore, unlike when testing hard thermoplastic probes, “ a” is divided in half to reflect that compression

will equally compress the probe and the tissue being tested in the calculation of

2 2 A0 = πR − π[R − (a / 2)] , in which “ a” and “R” are measured.

Statistical analysis

Since variances were not found to be homogenous in most cases, the Welch and Brown–

Forsythe robust tests of equality of means were run followed by a Dunnett T3 post-hoc

test. In cases where the variances were found to be statistically equal a Student's t-test or

an analysis of variance (ANOVA) was used depending upon how many groups were

compared. Tukey's Honestly Significantly Different post hoc test was then applied.

93 4.3 Results and discussion in vivo optimization of bioadhesion testing parameters

The force with which the probe contacts the mucosa (contact force), duration of contact

(contact time), testing speed, and presence of PBS have all been shown to affect the tensile bioadhesion results [16]. As a result comparisons between in vitro tests performed

in different labs using different conditions are difficult to compare, which has

confounded drawing meaningful comparisons between reported values. To prepare for

comparison between in vitro and in vivo measurements testing conditions were designed

to be identical and in vivo testing parameters were optimized to minimize distress of the live subjects while maximizing the reproducibility of bioadhesive fracture strength.

Contact force was chosen to minimize distress and the potential for tissue damage of the animals while still making reproducible measurements of tensile bioadhesive fracture strength. Therefore all of the reported in vivo and in vitro measurements were made at 5gf contact force. Contact time for in vivo bioadhesion measurements was optimized to be the minimum contact time at which measurements are made reproducibly, minimizing the testing time under anesthesia necessary for the animals. To determine the optimal contact time tests with PFASA2080, bioadhesion testing was conducted at 5gf contact force for durations between 32 and 256 s (Figure 4.3).

In the absence of PBS mean fracture strength increases and then plateaus above 128 s.

Mean fracture strength measured at 32 s contact time is statistically significantly lower than all of the other tested values ( p < 0.05). The mean fracture strength at 32 s is 41% lower than the mean fracture strength at 128 s hold time in the absence of PBS and 11%

94 lower in the presence of PBS. Previous studies have shown that the hydrolysis of anhydride bonds leading to increased carboxylic acid content within the erosion zone correlates with the increased bioadhesive fracture strength observed with increasing hold time [14].

At the longest contact time (256 s) the fracture strength decreases by 32% compared to

128 s ( p < 0.05) in the presence of PBS whereas it only decreases by 3% in the absence of

PBS. Previous studies report that after considerable hydrolysis of the anhydride bonds, some of the PFASA2080 has reduced in molecular weight significantly leaving monomers and oligomers. While the monomers and oligomers are highly bioadhesive due to their increased relative carboxylic acid content, due to their low molecular weight they can diffuse into the surroundings [14,32]. PBS raises the pH and increases the available water for hydrolysis, both of which will increase the degradation rate of PFASA2080

[14]. PBS also provides a low viscosity medium (relative to mucus) for rapid diffusion of monomers and oligomers away from the mucus–probe interface, which in sum could account for the observed earlier onset in decreasing fracture strength in the presence of

PBS between 128 s and 256 s (Figure 4.3).

Although filling the stomach with simulated gastric fluid would be more physiologically accurate than PBS, the acidic conditions are not suitable for a cannulated stomach or excised stomach tissue, in which the muscularis and serosa are also exposed to the media.

Adding PBS to the stomach during testing also maintains hydration of the mucosa. All subsequent testing was performed in the presence of PBS with 64 s contact time.

95 The testing speed used throughout the in vivo and in vitro experiments was chosen to be fast relative to the speed of the stomach wall as it moves during the respiratory cycle in vivo . Rats respire at a rate of 80–120 cycles per minute [36]. During each inhalation and exhalation the momentum of the diaphragm is transferred in part to the stomach wall due to intimate contact with a portion of the serosa. As a result the stomach could move with a velocity as high as a few millimeters per second. In two fasted rats the average stomach movement as recorded by Guignet et al. was 1.3 mm/s in the plane perpendicular to the stomach wall [36]. To exceed the speed of this motion so as to minimize its contribution to the overall in vivo bioadhesion measurements, all tests were performed at 5 mm/s testing speed.

Unconfined in vivo bioadhesion testing

After the testing conditions were chosen based on the described optimization, bioadhesion testing of a panel of polymers was performed in two rats. Testing conditions are referred to as “unconfined” since no external support is provided for the gastric cannula.

The ranking of polymers by mean fracture strength is: 1) PCL, 2) PFASA2080, 3)

PLGA5050, 4) PAA, and 5) PCPHSA2080. However, there is no statistically significant difference in mean fracture strength and the mean fracture strength of PCL is only 25% greater than PCPHSA2080 (Figure 4.4). Since previous in vitro tests led to statistically significant differences in bioadhesive fracture strength the differences in testing conditions were identified. The main differences are the lack of rigid confinement of the tissue sample and the vital rhythms that cyclically move the stomach with respect to the

96 testing probe. Since the latter is inherent in the in vivo design due to the location of the

tissue and species tested, confinement conditions were altered for the bioadhesion testing

in two additional rats to reduce gross stomach motion during testing and to more closely

mimic in vitro testing conditions without added distress to the subject.

Confined in vivo bioadhesion testing

An adjustable aluminum cannula holder consisting of two slotted vertical supports and a horizontal cross-piece that has a hole with a set screw for the cannula tube was machined.

After each rat was induced, the cross-piece of the holder was adjusted to the height of the cannula and leveled. With the set screw in place the cannula is fixed in position limiting travel of the stomach more than in the unconfined conditions (Figure 4.1b). In the confined conditions the ranking of polymers by mean fracture strength is: 1)

PCPHSA2080, 2) PFASA2080, 3) PLGA5050, 4) PAA, and 5) PCL (Figure 4.4). The mean fracture strength of PCPHSA2080 is 37% greater than PCL. In comparison with the unconfined testing results, PCL changed the most significantly in order from second in the unconfined case to fifth in the confined case. Additionally, PFASA2080 shows statistically higher bioadhesive tensile fracture strength than PAA ( p < 0.05) in the

confined setup.

Decreasing the mobility of the stomach during testing by anchoring the gastric cannula to

a rigid support is meant to more closely reproduce the in vitro testing conditions that employ a rigid tissue holder. Additionally, each of the polymers tested except for

PCPHSA2080, shows statistically significantly greater bioadhesive tensile fracture strength in the unconfined case ( p < 0.05 for PFASA2080 and PLGA5050 and p < 0.001

97 for PAA and PCL) as shown in Figure 4.4. The mean fracture strength of PFASA2080 is

17.5% higher than PAA, which is roughly half of the percent difference observed between the two groupings of bioadhesive polymers discussed in the following section. in vitro bioadhesion testing

Mean fracture strength in all cases is high compared to reported excised rat and porcine small intestinal tissue results using similar tensile bioadhesion testing methods in our lab and others [13-16,21,35]. The order of adhesiveness is consistent among the in vitro test

subjects. The in vitro ranking by mean fracture strength is: 1) PFASA2080, 2) PAA, 3)

PCPHSA2080, 4) PCL, and 5) PLGA5050 (Figure 4.5). The mean fracture strength of

PFASA2080 is 62% greater than PCL. Additionally, each rat individually and

collectively demonstrates two statistical groupings of polymer bioadhesivenesses.

PFASA2080 demonstrates significantly higher bioadhesive fracture strength than each of

the other polymers tested ( p < 0.001) except for PAA, from which there is no statistically significant difference. PAA bioadhesive fracture strength is statistically significantly higher than each of the other polymers tested ( p < 0.001 in comparison with PCL and

PLGA5050 and p < 0.01 in comparison with PCPHSA2080). On average, the bioadhesive fracture strength of the fast eroding polyanhydrides is 31% greater than the slow eroding polymers. The more bioadhesive polymers, PFASA2080 and PAA, both contain a high density of water-labile anhydride bonds. As a result both degrade rapidly via hydrolysis yielding numerous carboxylic acid groups that have been demonstrated to hydrogen bond strongly to mucin glycoproteins [6,12,14,15,32]. The other three polymers (PCL, PCPSHSA2080, and PLGA5050) that have lower mean fracture

98 strengths still have high values of adhesion compared to previous studies on other tissues, but are statistically significantly lower than the polyanhydride polymers tested (Figure

4.4). Each of the slowly degrading polymers presents a hydrophobic surface to the mucosa upon testing primarily enabling hydrophobic bonding to the hydrophobic amino acids present in the protein core of the mucus glycoproteins. The in vitro results are expected based on previous studies, unlike the in vivo results.

Comparison between in vitro and in vivo bioadhesion testing

Although attempts were made to confine the stomach during in vivo testing, ultimately for the safety of the animal no rigid support of the musculoserosal layers of tissue could be applied. As a result, the stomach tissue is expected to deform during the application of both compressive and tensile load more in vivo testing cases than in vitro . The tissue moves with the probe during testing, which would indicate a splaying of the tensile load versus distance curve yielding reduced fracture strength compared to in vitro (Figure 4.6

— in vitro ).

Under the in vivo testing conditions, the mucus moves relative to the testing probe with

the biorhythms of breathing and heart beating unlike in vitro. As a result it is more likely that the probe will reside in the luminal, less adherent mucus lining than in the adherent mucus compared to the in vitro setup. Increased mean coefficient of variance (1.9% in vitro versus 16% confined in vivo ) and decreased mean fracture strength

(1015 ± 20 mN/cm 2 in vitro versus 513 ± 82 mN/cm 2 confined in vivo ) is consistent with the hypothesis that the position of the probe during adhesion testing is more variable, as well as more likely to be in the less adherent mucus layer (Figure 4.5— confined in vivo ).

99 The confined in vivo mean fracture strengths of all polymers tested are significantly

lower than in vitro (PFASA2080, PAA, and PCL p < 0.001; PCPHSA2080 and PLGA

p < 0.01) and the mean fracture strength is 49% lower confined in vivo than in vitro .

Additionally, the ranking of polymer bioadhesiveness by mean fracture strength changes

from 1) PFASA2080, 2) PAA, 3) PCPHSA2080, 4) PCL, and 5) PLGA5050 in vitro to 1)

PCPHSA2080, 2) PFASA2080, 3) PLGA5050, 4) PAA, and 5) PCL in vivo . The in vitro tests more closely follow the trend predicted by previous in vitro tests that indicate polyanhydrides are more bioadhesive than hydrophobic polymers [14,15,21,32,34].

Surprisingly, the in vivo results may reflect the marginal improvements in gastroretentive bioadhesive oral drug delivery systems observed in preclinical and clinical trials [28-30].

Mucus–mucus bioadhesive bond strength

Lack of correlation between the in vitro and in vivo results motivated the hypothesis that

during in vivo testing the mucus–mucus bond strength of loosely adherent gastric mucus

is the limiting factor in gastric bioadhesion testing. Previous investigations utilizing

intravital microscopy report that the rat gastric mucosa can be divided into two layers

[27]. The luminal layer of the antrum of rats (120 ± 38 µm thick) is weaker mechanically and is therefore referred to as the “loosely adherent layer” [27]. The underlying mucus that is in direct contact with the gastric enterocytes (154 ± 16 µm thick) is referred to as the “firmly adherent layer” [27]. Based upon the in vivo results, which we believe are closer to the conditions achieved during per oral dosing of a bioadhesive pill, we believe that the probes are encountering primarily the loosely adherent mucus. To test the hypothesis the fracture strength of the mucus–mucus bond strength was measured in

100 vitro . Probes were designed to bring two loosely adherent gastric mucus surfaces into

contact with one another to test the bioadhesive fracture strength of the loosely adherent

mucus–mucus bond.

Mucus probes were prepared and tested in vitro . The fracture strength of the mucus–

mucus bioadhesive bonds is measured to be 446 ± 59 mN/cm 2 ( n = 9). The solid line in

Figure 4.5 is the mean fracture strength of the mucus–mucus bioadhesive bonds and the

error bars are represented by the dotted lines above and below. The mean fracture

strength of the mucus–mucus bond is statistically equivalent to all of the confined in vivo results and statistically significantly lower than all of the in vitro measurements

(PFASA2080, PAA, and PCL p < 0.001; PCPHSA2080 and PLGA p < 0.01) supporting the hypothesis that in vivo results reflect the properties of gastric mucus more than the bioadhesive nature of the polymer probe.

Hypothetically if a polymer bonds to mucus more strongly than mucus binds to itself, the mucus–mucus bond strength would be the limiting factor during tensile bioadhesion testing. Since all previous quantitative tensile testing has been performed on explanted tissues anchored by rigid tissue holders, the probes contact firmly adherent portions of the mucosa. Since the mechanical properties of the loosely adherent mucus dominate the in vivo and the firmly adherent properties dominate the in vitro data there is little correlation in the stomach. Therefore results indicate that macro-sized doses including standard tablets and gelatin capsules coated with a bioadhesive polymer may only contact the loosely adherent gastric mucosa in vivo marginalizing the effectiveness of the

bioadhesive coating.

101

4.4 Conclusions

The first objective of this investigation was to identify the feasibility and reproducibility

of quantifying tensile bioadhesive properties in live rats through implanting re-closable

gastric cannulae. Using the surgical cannulation procedure, we were able to make

measurements of fracture strength in live, anesthetized rats. The data presented are the

first report of mammalian bioadhesive fracture strength measured in vivo . More broadly,

the study aims to set up a direct quantitative comparison between in vivo and in vitro

mean bioadhesive fracture strengths. Additionally, the results are the first report of

polyanhydride bioadhesion in rat stomach both in vitro and in vivo .

The standard in vitro tensile bioadhesion assay results were more reproducible than in vivo for all of the polymers tested, as the mean coefficient of variance is 16% in confined

in vivo testing and 1.9% in vitro . Additionally in vitro , the panel of bioerodible polymers divides into two statistically significantly different groups: PFASA2080 and PAA were more adhesive than PCL, PCPHSA2080, and PLGA5050 ( p < 0.05). The more adhesive

groups consisting of PFASA2080 and PAA are both rapidly bioerodible polyanhydrides,

confirming the correlation between surface carboxylic acid content and tensile

bioadhesive properties reported in previous publications [14,15,21,31,32]. However, in

vivo in the confined testing conditions the difference between the mean fracture strengths

of the fast eroding polyanhydrides and the hydrophobic, slow eroding polymers

disappears indicating that perhaps standard in vitro bioadhesion testing conditions do not adequately reflect in vivo environment. The lack of a clear-cut difference between

102 traditionally bioadhesive and non-bioadhesive thermoplastic bioerodible polymers in the

stomach measured in vivo may be of great importance to the design of bioadhesive gastric retentive oral drug delivery systems. When designing a bioadhesive oral drug delivery system aiming for prolonged gastric residence time, factors like size, density, and shape should be considered to maximize the likelihood of contacting the firmly adherent gastric mucus.

Testing of the mucus–mucus bond fracture strength indicates that in vivo data reflect the mechanical properties of the loosely adherent gastric mucus [27] rather than the bioadhesiveness of the polymer probe (Figure 4.5). The rigid tissue holder used during in vitro testing is not physiological and provides rigidity to the tissue not found in vivo . As a

result, given the same contact force, polymer probes will penetrate deeper into the gastric

mucus in vitro than in vivo . The lack of increased fracture strength of polyanhydrides

over other polymers in vivo reflects the preclinical and clinical findings reported in the

literature [28-30]. It is of note that polymers allowing polymer-mucin chain

entanglement, including traditional bioadhesives such as chitosan and polyacrylic acids,

may penetrate the mucus more deeply and deviate from the behaviors observed in the

thermoplastic bioerodibles tested. Additionally, factors that were not tested in this study

such as shear stresses, polymer molecular weight, and dose geometry can all still be

utilized to achieve improved gastric retention.

Overall the methods employed to yield direct, quantitative comparison between in vitro

and in vivo data have improved the understanding of the in vivo gastric mucosal

environment encountered by thermoplastic bioerodible polymers, which appears to be

103 dominated by the loosely adherent gastric mucus. Our improved understanding may lead to new research efforts that focus on utilizing gastroretentive strategies in addition to bioadhesion to enable bioadhesive doses to contact the underlying firmly adherent gastric mucus achieving prolonged gastric retention.

104

Figure 4.1: (a) Photograph and schematic of the reclosable gastric cannula. (b) Photographs and cross-sectional schematic representation of the in vivo bioadhesion confined experimental setup during testing in which quantitative tensile bioadhesion measurements are made on the stomachs of anesthetized live rats through the reclosable gastric cannula. (c) Photograph of in vitro tensile bioadhesion testing setup for polymer probes (left) and mucosa probes (right).

105

Figure 4.2: (a) Schematic diagram of a spherical polymer probe contacting mucus while approaching the mucosa and reaching its contact force. (b) From the schematic diagram of a spherical polymer probe retracting from the mucosa during tensile adhesion testing it is clear that only the forces parallel to the direction of motion ( F║) are measured by the uniaxial load cell. As a result the projected surface area about which the tensile and compressive forces act is the horizontal projection of the mucus covered surface contacting the probe. (c) Time points T0 and T1 refer to the probe positions depicted in (a) demonstrating how penetration depth “ a” is measured on a load versus distance curve. (d) Once penetration depth “ a” and probe radius “ R” have been measured the listed equations are used to calculate projected surface area (PSA).

106

Figure 4.3: Bioadhesive fracture strength of quick-eroding, bioadhesive polyanhydride PFASA2080 as a function of hold time at 5gf contact force in vivo in the absence (“Native”) and presence of PBS (“PBS”) filling the stomach lumen ( n ≥ 3). In the absence of PBS the 32 second hold time fracture strength is significantly smaller than the longer hold times in the absence of PBS ( p < 0.05). In the presence of PBS, there are no statistical differences among fracture strengths with varying hold times. At the longest hold time (256 s) there is a statistically significant difference between tests performed in the absence and presence of PBS ( p < 0.05). Error bars are SEM. ( p < 0.05).

107

Figure 4.4: Bioadhesive fracture strength measured in vivo “unconfined” ( n ≥ 9) and “confined” ( n = 18). No statistical significance was found among the fracture strengths of polymers tested in the unconfined conditions. In the confined conditions PFASA2080 has a statistically significantly higher mean fracture strength than PAA ( p < 0.05). Comparing the unconfined and confined testing conditions PFASA2080 ( p < 0.05), PAA ( p < 0.001), PLGA5050 ( p < 0.05), and PCL ( p < 0.001) all have higher mean fracture strengths unconfined than confined. PCPHSA2080 mean fracture strength shows no statistical significance between the unconfined and confined testing conditions. Error bars are SEM. ( p < 0.05, p < 0.01, p < 0.001).

108

Fig. 4.5: Bioadhesive fracture strength of five biodegradable polymers tested on three excised rat stomachs plotted in vitro ( n = 27 per polymer). Results show a greater degree of statistical significance overall among polymers than in vivo . In particular PFASA2080 demonstrates statistically significantly higher fracture strength than PCL ( p < 0.001), PCPHSA2080 ( p < 0.001), and PLGA5050 ( p < 0.001) and shows no statistical difference from PAA. PAA also demonstrates statistically significantly higher fracture strength than PCL ( p < 0.001), PCPHSA2080 ( p < 0.01), and PLGA5050 ( p < 0.001). Statistical analysis indicates two groupings of polymers tested, divided by the grey vertical dashed line. The two classes of bioadhesives consist of fast eroding polyanhydrides (PFASA2080 and PAA), which are more adhesive than the slower eroding polymers (PCL, PCPHSA2080, and PLGA5050). Error bars are SEM. ( p < 0.01, p < 0.001).

109

Figure 4.6: Bioadhesive fracture strength comparison among confined in vivo , in vitro , and in vitro stomach mucus–mucus bond strength. The horizontal solid line denotes the mean fracture strength of rat stomach mucus–mucus adhesive strength measured in vitro (n = 9) and the dotted lines above and below the solid line indicate the SEM. There is no statistical significance among any of the polymers tested under confined in vivo conditions and the stomach mucus bioadhesive bond strength tested. All polymers tested in vitro have statistically higher fracture strengths than both confined in vivo and stomach mucus tests (PFASA2080 p < 0.001, PAA p < 0.001, PCL p < 0.001, PCPHSA2080 p < 0.01, and PCL p < 0.01). The similarity between the confined in vivo and stomach mucus fracture strengths indicates that the mechanical properties of loosely adherent gastric mucus may set an upper limit to bioadhesive fracture strength in vivo and could explain why in vivo tests showed little or no difference between rapidly eroding polyanhydrides and slow eroding polymers. Error bars are SEM. ( p < 0.01, p < 0.001).

110 4.5 References

1. P. Aubin, The limpet's power of adhesion, Nature 45 (1892), pp. 464–465.

2. G. Branch and A. Marsh, Tenacity and shell shape in six Patella species: adaptive

features, J. Exp. Mar. Biol. Ecol. 34 (2) (1978), pp. 111–130.

3. M. Denny, Limits to optimization: fluid dynamics, adhesive strength and the

evolution of shape in limpet shells, J. Exp. Biol. 203 (17) (2007), pp. 2603–2622.

4. J. Grenon, J. Elias, J. Moorcroft and D. Crisp, A new apparatus for force

measurement in marine bioadhesion, Mar. Biol. 53 (4) (1979), pp. 381–388.

5. A. Smith, T. Quick and R. Peter, Differences in the composition of adhesive and non-

adhesive mucus from the limpet Lottia limatula , Biol. Bull. 196 (1) (1999), pp. 34–44.

6. M. Marvola, K. Vahervuo, A. Sothmann, E. Marttila and M. Rajaniemi, Development

of a method for study of the tendency of drug products to adhere to the oesophagus, J.

Pharm. Sci. 71 (1982), pp. 975–977.

7. L. Achar and N. Peppas, Preparation, characterization and mucoadhesive interactions

of poly(methacrylic acid) copolymers with rat mucosa, J. Control. Release 31 (3)

(1994), pp. 271–276.

8. A. Deascentiis, P. Colombo and N. Peppas, Screening of potentially mucoadhesive

polymer microparticles in contact with rat intestinal-mucosa, Eur. J. Pharm.

BioPharm. 41 (4) (1995), pp. 229–234.

9. J. Haas and C. Lehr, Developments in the area of bioadhesive drug delivery systems,

Exp. Opin. Biol. Therapy 2 (3) (2002), pp. 287–298.

10. C. Lehr, Bioadhesive drug delivery systems for oral application, Pharm. Weekbl. Sci.

Ed. 14 (3) (1992), pp. 95–96.

111 11. F. Lejoyeux, G. Ponchel, D. Wouessidjewe, N. Peppas and D. Duchene, Bioadhesive

tablets influence of the testing medium composition on bioadhesion, Drug Dev. Ind.

Pharm. 15 (12) (1989), pp. 2037–2048.

12. S. Leung and J. Robinson, Polymer structure features contributing to mucoadhesion.

2, J. Control. Release 12 (3) (1990), pp. 187–194.

13. M. Liebau, A. Hildebrand, G. Bendas, U. Rothe and R. Neubert, Development of a

bioadhesion model based on quartz crystal microbalance for the characterization of

drug delivery systems part 1: introduction and methods, Pharm. Ind. 61 (5) (1999),

pp. 459–462.

14. C. Santos, B. Freedman, S. Ghosn, J. Jacob, M. Scarpulla and E. Mathiowitz,

Evaluation of anhydride oligomers within polymer microsphere blends and their

impact on bioadhesion and drug delivery in vitro, Biomaterials 24 (20) (2003), pp.

3571–3583.

15. C. Santos, J. Jacob, B. Hertzog, B. Freedman, D. Press, P. Harnpicharnchai and E.

Mathiowitz, Correlation of two bioadhesion assays: the everted sac technique and the

CAHN microbalance, J. Control. Release 61 (1–2) (1999), pp. 113–122.

16. M. Tobyn, J. Johnson, P. Dettmar, Factors affecting in-vitro gastric mucoadhesion. 1.

Test conditions and instrumental parameters. Eur. J. Pharm. Biopharm. 41 (4)(2995)

235–241.

17. K. Albrecht, M. Greindl, C. Kremser, C. Wolf, P. Debbage and A. Bernkop-

Schnurch, Comparative in vivo mucoadhesion studies of thiomer formulations using

magnetic resonance imaging and fluorescence detection, J. Control. Release 115 (1)

(2006), pp. 78–84.

112 18. D. Ameye, J. Voorspoels, P. Foreman, J. Tsai, P. Richardson, S. Geresh and J.

Remon, Ex vivo bioadhesion and in vivo testosterone bioavailability study of

different bioadhesive formulations based on starch-g-poly(acrylic acid) copolymers

and starch/poly(acrylic acid) mixtures, J. Control. Release 79 (1–3) (2002), pp. 173–

182.

19. R. Chary, G. Vani and Y. Rao, In vitro and in vivo adhesion testing of mucoadhesive

drug delivery systems, Drug Dev. Ind. Pharm. 25 (5) (1999), pp. 685–690.

20. T. Goto, M. Morishita, N. Kavimandan, K. Takayama and N. Peppas, Gastrointestinal

transit and mucoadhesive characteristics of complexation hydrogels in rats, J. Pharm.

Sci. 95 (2) (2006), pp. 462–469.

21. In: E. Mathiowitz, Editor, Encyclopedia of Controlled Drug Delivery , John Wiley &

Sons, Inc., New York, NY (1999).

22. M. Sakkinen, J. Marvola, H. Kanerva, K. Lindevall, A. Ahonen and A.M. Marvola,

Are chitosan formulations mucoadhesive in the human small intestine? An evaluation

based on gamma scintigraphy, Int. J. Pharm. 307 (2) (2006), pp. 285–291.

23. A. Bernkop-Schnurch, D. Guggi and Y. Pinter, Thiolated chitosans: development and

in vitro evaluation of a mucoadhesive, permeation enhancing oral drug delivery

system, J. Control. Release 94 (1) (2004), pp. 177–186.

24. S. Kockisch, G. Rees, S. Young, J. Tsibouklis and J. Smart, A direct-staining method

to evaluate the mucoadhesion of polymers from aqueous dispersion, J. Control.

Release 77 (1–2) (2001), pp. 1–6.

113 25. A. Shojaei and X. Li, Mechanisms of buccal mucoadhesion of novel copolymers of

acrylic acid and polyethylene glycol monomethylether monomethacrylate, J. Control.

Release 47 (2) (1997), pp. 151–161.

26. P. Marshall, J. Snaar, Y. Ng, R. Bowtell, F. Hampson, P. Dettmar, E. Onsoyen and C.

Melia, Localised mapping of water movement and hydration inside a developing

bioadhesive bond, J. Control. Release 95 (3) (2004), pp. 435–446.

27. F. Varum, E. McConnell, J. Sousa, F. Veiga and A. Basit, Mucoadhesion and the

gastrointestinal tract, Crit. Rev. Ther. Drug Carrier Sys. 25 (3) (2008), pp. 207–258.

28. P. Bardonnet, V. Faivre, W. Pugh, J. Piffaretti and F. Falson, Gastroretentive dosage

forms: overview and special case of Helicobacter pylori , J. Control. Release 111 (1–

2) (2006), pp. 1–18.

29. S. Davis, Formulation strategies for absorption windows, Drug Discov. Today 10 (4)

(2005), pp. 249–257.

30. R. Talukder and R. Fassihi, Gastroretentive delivery systems: a mini review, Drug

Dev. Ind. Pharm. 30 (10) (2004), pp. 1019–1028.

31. E. Mathiowitz and R. Langer, Polyanhydride microspheres as drug delivery systems,

Abst. Pap. Am. Chem. Soc. 200 (1999) 72-PMSE, Part 2.

32. C. Santos, B. Freedman, K. Leach, D. Press, M. Scarpulla and E. Mathiowitz,

Poly(fumaric-co-sebacic anhydride) — a degradation study as evaluated by FTIR,

DSC, GPC and X-ray diffraction, J. Control. Release 60 (1) (1999), pp. 11–22.

33. D. Chickering, J. Jacob, T. Desai, M. Harrison, W. Harris, C. Morrell, P. Chaturvedi

and E. Mathiowitz, Bioadhesive microspheres. 3. An in vivo transit and

114 bioavailability study of drug-loaded alginate and poly(fumaric-co-sebacic anhydride)

microspheres, J. Control. Release 48 (1) (1997), pp. 35–46.

34. D. Chickering, J. Jacob and E. Mathiowitz, Bioadhesive microspheres. 2.

Characterization and evaluation of bioadhesion involving hard bioerodible polymers

and soft-tissue, React. Polym . 25 (2–3) (1995), pp. 189–206.

35. W. Pare, K. Isom, G. Vincent Jr and G. Glavin, Preparation of a chronic gastric fistula

in the rat, Lab. Anim. Sci. 27 (2) (1977), pp. 244–247.

36. R. Guignet, G. Bergonzelli, V. Schlageter, M. Turini and P. Kucera, Magnet

Tracking: a new tool for in vivo studies of the rat gastrointestinal motility,

Neurogastroenterol. Motil. 18 (6) (2006), pp. 472–478.

115

Chapter 5

Bioinspired Synthetic Bioadhesive Polymers

Abstract

Bioadhesive polymers promote and prolong intimate contact between dosage forms and the mucosal surfaces to which they are administered. Increased duration of close contact has yielded increased bioavailability of numerous therapeutics. We report the synthesis and bioadhesive strength characterization of novel bioinspired bioadhesive polymers that contain L-3,4-dihydroxyphenylalanine (DOPA), implicated in the extremely adhesive byssal fibers of certain gastropods, and its biochemical precursor amino acids in their side chains. The novel bioinspired bioadhesive polymers consist of combinations of either of two backbones, poly(butadiene-co-maleic anhydride) 1:1 or poly(ethylene-co-maleic anhydride) 1:1, with any of three amino acids, phenylalanine, tyrosine, or DOPA, grafted as side chains. DOPA-grafted hydrophobic backbone polymers exhibit excellent bioadhesive properties, demonstrating as much as 2.5x the fracture strength and 2.8x the tensile work of bioadhesion of a commercially available polyacrylic acid derivative as tested on live, excised, rat intestinal tissue.

116 5.1 Introduction

Bioadhesion, adherence to or of a biological material, has fascinated scientists for over a century beginning most notably with the strong adherence of mollusks and barnacles [1].

Chemical composition studies of mollusk adhesive mucus determined that DOPA (L-3,4- dihyroxyphenylalanine) was present in large quantities and may account for the high adhesive strength [2-5]. In mollusk mucus, DOPA chelates and complexes with metal ions, most notably ferric iron, on rocks enabling glue-like adhesion [2-6]. Additionally, the ability of multiple DOPA residues to bind metal ions in aqueous conditions contributes to byssal fiber, insoluble, highly adhesive silk-like fibers produced by various mollusks, formation [4,5]. Taking a cue from nature, polymer chemists began incorporating DOPA into hydrogel polymers to promote bioadhesive properties [3,7,8].

Lee et al. polymerized DOPA with the surfactant, Pluronic F127, to create self- assembling micelles with bioadhesive end groups [9]. Schnurrer and Lehr measured the tensile bioadhesive fracture strength of mussel adhesive protein, containing 10-20%

DOPA, and found it to be 1-3x as adhesive to porcine intestinal mucosa as the commercial bioadhesive polycarbophil, a poly(acrylic acid) derivative, depending on the oxidation conditions [10].

As byssal threads promote mollusk adhesion to rocks in the ocean, the goal of polymeric bioadhesives for oral drug delivery is to promote adhesion of oral dosage forms to the gastrointestinal (GI) mucosa. Bioadhesives can prolong GI residence and promote contact between an oral dose and the GI mucosa leading to increased bioavailability and duration

117 of activity of therapeutic agents [11-14]. Many strategies have been employed for

creating bioadhesives including adding functional groups to promote mucus bonding and

bioinspired approaches such as lectins [15]. Some of the most successful bioadhesives

have been synthesized to contain high densities of carboxylic acid or hydroxyl groups

that in water lead to sharing of hydrogen ions between the polymer and mucus

glycoproteins yielding hydrogen bonds [10,16]. Due to their polar nature, bioadhesion

promoting functional groups such as carboxylic acids and hydoxyls increase

hydrophilicity leading to water solubility [13,16]. Examples of hydrophilic bioadhesives

include hydrogels such as alginate and polyacrylic acid derivatives such as polycarbophil

[9,14]. Ultimately if the polymer hydrates, the bioadhesive strength is lost [18,19].

Mathiowitz et al . developed polyanhydride-based bioadhesives that are hydrophobic initially, and then expose carboxylic acid groups as they hydrolyze [10,16].

This manuscript presents a combined approach to synthesizing novel bioinspired bioadhesive polymers that have a hydrophobic backbone with DOPA functionality, similar to mussel adhesive protein [8,20]. Since the amino acids phenylalanine (Phe) and tyrosine (Tyr) are metabolic precursors to DOPA, we also investigated the bioadhesive properties of novel polymers with the same hydrophobic backbone with Phe and Tyr functionality [4,5]. Bioadhesive fracture strength and tensile work were quantified in vitro on freshly excised rat intestinal tissue as compared to a commercial acrylic acid- derived bioadhesive polymer.

118 5.2 Materials and Methods

Bioinspired bioadhesive polymer synthesis

500mg of each polymer backbone, poly(butadiene-co-maleic anhydride) 1:1 (PBMA) and poly(ethylene-co-maleic anhydride) 1:1 (PEMA) (Polysciences, Warrington, PA), was dissolved at a concentration of 1w/v% with one of three amino acid derivatives, phenylalanine, tyrosine, or DOPA (Sigma Aldrich, St Louis, MO) in dimethyl sulfoxide

(DMSO) (Mallinckrodt, Hazelwood, MO) (Figure 5.1) [21]. The molar ratio of side chain to backbone was determined by assuming side addition to each site of attachment, the maleic anhydride residues (e.g. for reacting PBMA with DOPA, (side chain molar mass/monomer molar mass)*polymer mass = side chain mass, (197amu/152amu)*500mg

= 650mg). The DMSO solution was stirred and headed on a thermostat controlled stirring hot plate (Fisher Scientific, Pittsburgh, PA) set to 70 degrees Centigrade and 500 revolutions per minute [21]. The flasks in which the reaction took place were sealed by rubber stoppers to minimize any atmospheric water vapor ingress, and the reaction was run for 12 hours.

At the completion of the side chain addition reaction, the solution was allowed to cool to room temperature then twice the volume of room temperature distilled water was added to dilute the DMSO prior to dialyzing. Dialysis was performed in 4L stainless steel vessels using 1cm of 10kDa cut-off SnakeSkin tubing (Thermo Scientific, Rockford, IL) for every 3ml of added liquid with room for increased water absorption so that the polymer remained and any un-reacted side chain as well as DMSO was dialyzed and

119 discarded. At least 5 water changes were performed over the course of 3 days to ensure

minimal residual organic solvent and un-reacted side chain [21]. After dialysis, the

remaining aqueous polymer solution was lyophilized (VirTis, Gardiner, NY) yielding dry

powders. Each batch produced approximately 600mg of side chain grafted polymer with

a yield of ~50-60%.

Nuclear magnetic resonance (NMR) analysis of bioinspired bioadhesives

Both polymer backbones, PBMA and PEMA, along with their bioinspired derivatives

were dissolved in deuterated DMSO (D6-DMSO, Cambridge Isotope Laboratories,

Andover, MA) at a concentration of 25mg/ml. Each polymer solution was loaded into a

5mm thin wall 300MHz NMR sample tube (Wilmad Lab Glass, Vineland, NJ) and an

average of sixteen scans was acquired for analysis. 1H NMR analysis was performed on a

Bruker DPX 300MHz spectrometer equipped with a BBO probe and processed using

TopSpin 1.3 software (Bruker, Billerica, MA). Peak assignment of PEMA-derived polymers was confirmed by multiplicity edited hetero-nuclear single quantum coherence

1H NMR, performed on a Bruker Ultraspin 400MHz spectrometer (Bruker, Billerica,

MA).

Polymer probe preparation

As reported previously, each of the bioadhesive polymers tested was solvent cast onto the

heads of glass-headed pins ( Φ=2-3mm) [18]. To prepare 5w/v% solutions for dip coating,

acetone was the solvent for PBMA, PEMA, and their derivative polymers and ethyl

acetate was used for Polycarbophil AA-1 (Noveon, Cleveland, OH). Glass-headed pins

120 were dipped and dried three times to ensure a continuous polymer coating prior to

bioadhesion testing.

Tensile bioadhesion testing

Bioadhesive tensile fracture strength and tensile work were performed on a Texture

Analyzer TA.XT Plus (TA) (Texture Technologies, Scarsdale, NY) as reported previously in our lab and others (Figure 5.3) [18,22,23]. In short, intestinal tissue is excised from

200-300g albino, male, Sprague-Dawley rats immediately post mortem . Tissue is sectioned into 3cm lengths and stored in phosphate buffered saline (PBS) on ice until bioadhesion testing for a maximum of 4 hours. The tissue lumen is rinsed with 10ml of

PBS then cut along the anti-mesenteric boarder and placed mucus-side up in PBS on a water heated tissue holder set to 37 degrees Centigrade to mimic physiological conditions.

Bioadhesion testing begins with the polymer probe approaching the intestinal mucus at

0.5mm/s until a contact force of 5gF is reached. Once the contact force is reached, the probe ceases motion and remains for a predetermined period of time to allow for polymer hydration and adhesive bond formation. PBMA and its derivative polymers were tested using a contact time of 7 minutes, as previously reported in our lab. PEMA and its derivatives had not sufficiently hydrated after a period of 7 minutes and so a contact time of 14 minutes was used. In order to compare the bioinspired bioadhesive polymers to the commercial bioadhesive, Polycarbophil AA-1, Polycarbophil-coated probes were tested both at 7 and 14 minutes contact time. After the contact time elapses, the probe retracts

121 from the mucus at 0.5mm/s while measuring tensile load caused by bioadhesion. The

peak tensile load normalized by the cross-sectional contact area yields a measure of

bioadhesive fracture strength and the area under the tensile force-distance curve measures

tensile work. Both fracture strength and tensile work have shown strong correlations with

the in vivo performance of bioadhesive polymers. All polymers were tested six times and each intestinal tissue segment was used for a maximum of 30 minutes. Tissue explants from a total of 5 rats were used and all animal procedures were performed in accordance with NIH and IACUC guidelines.

Contact area determination and validation

Contact area was calculated by measuring the diameter of each polymer probe and quantifying probe penetration depth, or compressive deformation of the intestinal tissue, during each test. Given probe radius (R) and penetration depth (a), the radius of the cross sectional area of contact (r) can be calculated, r=(R 2-(R-a) 2)½, using the Pythagorean

Theorem. Assuming spherical polymer probes, the circular cross-sectional contact area

(A) is calculated as A= πr2=π(R 2-(R-a) 2) [17].

To experimentally validate the cross-sectional area calculation, glass-headed pins were dry powder coated in 80-mesh carbon black (Sigma Aldrich, St Louis, MO) and then loaded into the TA bioadhesion testing setup. In place of tissue, double-sided foam tape

(3M, St Paul, MN) was used so that the carbon black powder would transfer to the tape in the area contacted by the probe. The cross-sectional area of contact was calculated using

122 the above described method and then compared to the area of carbon black left on the tape as determined by NIH ImageJ analysis of digital photographs.

Statistical analysis

All data were analyzed for statistical significance by one-way ANOVA in OriginLab

(Origin, Elk Grove Village, IL).

5.3 Results and Discussion

Polymer synthesis efficiency

Each polymer backbone, poly(butadiene-co-maleic anhydride) 1:1 (PBMA) and poly(ethylene-co-maleic anhydride) 1:1 (PEMA), was reacted with each of three side chains, phenylalanine, tyrosine, or DOPA to create poly(butadiene-co-maleic anhydride- graft-phenylalanine) (PBMAP), poly(butadiene-co-maleic anhydride-graft-tyrosine)

(PBMAT), poly(butadiene-co-maleic anhydride-graft-DOPA) (PBMAD), poly(Ethylene- co-maleic anhydride-graft-phenylalanine) (PEMAP), poly(butadiene-co-maleic anhydride-graft-tyrosine) (PEMAT), and poly(butadiene-co-maleic anhydride-graft-

DOPA) (PEMAD). The chemical structures of each are shown in Figure 5.1. During the side chain grafting reaction, the polymer solution exhibits a color change. The color change is most apparent in the DOPA grafted polymers, PBMAD and PEMAD, where the solution turns from pale amber to brown (Figure 5.2a). The addition of tyrosine to

PEMA to form PEMAT also produces a notable color change from pale amber to pink

(Figure 5.2b). All of the polymers are synthesized in DMSO and therefore DMSO is a

123 good solvent. Ethanol is also a good solvent for all of the bioinspired bioadhesives.

Chlorinated organic solvents such as chloroform and dichloromethane are non-solvents

for the polymers.

To determine the side chain attachment efficiency, 1H nuclear magnetic resonance

(NMR) spectra were acquired of each of the bioinspired bioadhesives and their constituent backbone polymers. For the PBMA-derived polymers, the peaks corresponding to the olefinic protons present in the backbone ( δ~4.4-5.5) was used as a basis of comparison with the hydrogen atoms bound to the aromatic carbons present in the side chains ( δ~6-7) (Figure 5.2a). Since each monomeric unit of the PBMA-derived polymers should contain two olefinic backbone protons, the area under the associated peaks was assigned a value of 2 and all other peak areas are measured with respect to it.

Each of the grafted side chains contains an aromatic ring not present in PBMA or PEMA.

Assuming 100% attachment to all maleic anhydride residues, the peak area associated with the hydrogens in the aromatic ring of phenylalanine would have an area ratio of 5:2, tyrosine would have an area ratio of 4:2, and DOPA would have an area ratio of 3:2, as compared to the backbone hydrogens bound to the doubly bonded carbons in PBMA or to the carbons bound to three other carbon atoms in PEMA. By comparing the measured peak area ratio to the theoretical peak area ratio, a measure of side chain attachment efficiency is provided in Table 5.1. PBMA-derived bioinspired bioadhesive polymers exhibit approximately 70-90% side chain attachment efficiency.

124 A similar analysis was performed for the PEMA-derived polymers; however, since

PEMA does not contain any olefinic protons, the two methine protons of maleic anhydride were used were used in the analysis (highlighted in yellow in Figure 5.1d).

PEMAP, PEMAT, and PEMAD exhibit side chain attachment efficiencies of 98%, 73%, and 91%, respectively. Differences in attachment efficiency may have resulted from differing confirmations of the polymer during side chain attachment that could promote or hinder the reaction based upon steric constraints.

Peak assignment was confirmed by multiplicity edited hetero-nuclear single quantum coherence (HSQC) 1H NMR. Confirmation was based on the phase of the carbon atoms, which is dependent on the number of bound hydrogens. Using HSQC, methyl and methine carbons appear in phase and methylene carbons in opposite phases, shown for the bioinspired bioadhesive polymers in Figure 5.3.

Experimental validation of cross-sectional contact area

We compared the projected cross-sectional area calculated based upon the probe radius and penetration depth measured by the TA against the area of the carbon black residue left on double-sided foam tape as analyzed by imageJ (Figures 5.3b and 5.3c). The two mean values of cross-sectional contact area were statistically insignificantly different

(p>0.05) as analyzed by one-way analysis of variance (ANOVA) (N=6), experimentally validating the compressive deformation-based cross-sectional contact area calculation.

125 PBMA-derived polymer bioadhesion testing

Polyacrylic acids consist of a polyethylene backbone that has a high density of hydrophilic carboxylic acid side-groups (Figure 5.1a) [13,17]. Carboxylic acid residues confer strong bioadhesive properties achieved by a high degree of hydrogen bonding and promote water solubility [13,17]. Once dissolved, polyacrylic acids no longer provide any substantive bioadhesive linkage between an oral dosage form and the GI mucosa [13,17].

By comparison, as an anhydride polymer, PBMA is initially hydrophobic and water insoluble. As the maleic anhydride sidegroups hydrolyze to form dicarboxylic acids, the polymer increases its carboxylic acid content over time in aqueous media and therefore increases both bioadhesiveness via hydrogen bonding and as water solubility [10,16].

Through the process of adding aromatic amino acid side chains to PBMA, the maleic anhydride is converted into a carboxylic acid and creates an amide bond to the amino acid forming a side chain [21]. In the case of phenylalanine addition (PBMAP), the carboxylic acid and the aromatic ring of the amino acid are added as a side-group presenting both hydrophilic and hydrophobic moieties. Tyrosine and DOPA addition

(PBMAT and PBMAD) include a singly and doubly hydroxyl-substituted aromatic side group increasing the hydrophilicity and hydrogen bonding capacity of the polymers

(chemical structures provided in Figure 5.1). Additionally, the physiochemical properties of the DOPA side chains on PBMAD may provide appropriate spacing and partial charge distribution for hydroxyl groups to form bonds with any multivalent cations found within the GI mucosa [6,7,17]. Free DOPA has been shown to chelate iron in both in vitro and in vivo settings [2-5]. Proteins bearing DOPA functionality, such as mussel adhesive

126 proteins, have demonstrated strong iron binding capabilities [8]. We hypothesize that the hydroxyl groups on the DOPA side chains bind ferric ions and other multivalent cations present in the GI mucosa [6,17,24].

Mean fracture strength of Polycarbophil AA-1 (a commercially utilized polyacrylic acid- derived bioadhesive), PBMA, and its derivatives are plotted in Figure 5.4a. The bioadhesive bond between Polycarbophil and freshly excised rat intestinal mucosal tissue exhibits 245.7±65.3 mN/cm 2 peak strength prior to mucoadhesive bond failure, or fracture strength (N=6). PBMA, PBMAP, PBMAT, and PBMAD demonstrate 1.54x,

0.86x, 1.46x, and 2.12x the mean fracture strength of Polycarbophil respectively.

Although there is no statistically significant difference among the fracture strengths

(p>0.05), there is a linear trend towards increased fracture strength from phenylalanine to tyrosine to DOPA functional polymers in order of their biochemical synthetic pathway in humans [4,9].

With respect to the area under the bioadhesive force-distance curve, or tensile work,

Polycarbophil exhibits 4093±177 nJ. PBMA, PBMAP, PBMAT, and PBMAD demonstrate 3.08x, 1.53x, 1.64x, and 4.83x the mean bioadhesive tensile work of

Polycarbophil respectively (Figure 5.5b). While the linear trend of increasing bioadhesion along the biochemical synthetic pathway is not present in the tensile work comparison, the overall order of adhesiveness is preserved with the exception of Polycarbophil and

PBMAP reversing order as last and next to last. PBMAD exhibits a statistically significantly higher mean bioadhesive tensile work than PBMAT and PBMAP (p<0.01),

127 as well as Polycarbophil (p<0.001). We hypothesize that the high tensile work and fracture strength demonstrated by PBMAD may result in part from the exceptional ability to bind multivalent cations present in mucus, in addition to standard hydrogen bonding due to carboxylic acid groups and other potential bioadhesive mechanisms [12,25].

PEMA-derived polymer bioadhesion testing

PEMA has a polyethylene-based backbone that lends itself to a greater degree of crystallinity than the polybutadiene-rubber-based backbone of PBMA (Figure 5.1).

Additionally, PEMA has a significantly higher molecular weight (M w=400kDa) and smaller repeat unit (M R=126Da) than PBMA (M w=10-15kDa, M R=151Da). The increased molecular weight and anhydride density of PEMA as compared to PBMA are indicative of greater bioadhesive properties based on previous studies with other polymers.

However, the increased crystallinity of the polyethylene backbone reduces the hydration rate and therefore bioadhesion testing performed with 7 minutes of contact time between the polymer probes and intestinal mucosa, as with the PBMA-derived polymers, demonstrated negligible bioadhesion. Doubling the contact time to 14 minutes allowed for sufficient hydration of PEMA and yielded measureable bioadhesive properties. The difference in contact time obviates direct comparison between PBMA- and PEMA- derived polymers. Therefore, the bioadhesive properties of Polycarbophil were tested with 14 minutes contact time to provide a common basis for comparison.

Under the 14 minutes contact time testing conditions, Polycarbophil produced a mean bioadhesive fracture strength of 334.9±33.4 mN with freshly excised rat intestinal tissue,

128 statistically similar to the fracture strength measured using 7 minutes hold time (p>0.05).

PEMA, PEMAP, PEMAT, and PEMAD produced 0.42x, 0.61x, 0.50x, and 2.5x the mean bioadhesive fracture strength of Polycarbophil tested under the same conditions (Figure

5.6a). PEMAD produced the greatest bioadhesive fracture strength of any polymer tested in this study, statistically significantly higher than each of the other polymers PEMA derivatives (p<0.01) and Polycarbophil (p<0.05) tested under the same conditions, indicating that it is a very strong bioadhesive. The linear trend of increasing bioadhesive fracture strength of PEMA-derived polymers coinciding with the biochemical synthetic pathway of DOPA is not present as with the PBMA derivatives. Instead there is a sharp increase in bioadhesiveness from PEMAP and PEMAT to PEMAD. The increase may be due in part to the ability of DOPA-functionalized bioadhesives to bind multivalent cationic species in mucus and also in part to the increased hydrophilicity afforded by the two hydroxyls present on the aromatic rings of the DOPA side chains, in addition to other bioadhesive mechanisms.

Mean tensile work of the PEMA-derived polymers is compared with Polycarbophil in

Figure 5.6b. After 14 minutes of contact time, Polycarbophil produces significantly less tensile work of bioadhesion than after 7 minutes with a mean of 2,299±575 nJ. The statistically significant, 44% reduction in mean tensile work (p<0.05) of Polycarbophil tested with 14 minutes hold time as compared to 7 minutes may be due to the increased hydration of the polymer leading to decreased mechanical. PEMA, PEMAP, PEMAT, and PEMAD produce 0.67x, 0.37x, 1.9x, and 2.8x the mean tensile work of

Polycarbophil respectively. PEMAD produced a statistically significantly higher mean

129 bioadhesive tensile work than Polycarbophil (p<0.05), PEMA (p<0.05), and PEMAP

(p<0.01). In both PBMA- and PEMA-derivative polymer testing the DOPA

functionalized polymer produced the highest mean bioadhesive fracture strength and

tensile work.

Comparison of bioadhesive fracture strength measurements with intraluminal

pressure recordings

Having different contact times due to different hydration rates complicates direct

comparison between the PEMA- and PBMA-derived bioinspired bioadhesives. Yet, the

strongest adhesion was observed in DOPA-functionalized polymers in all tested

conditions. In particular the increase in mean bioadhesive fracture strength of PEMAD

compared to the other PEMA-derived polymers and Polycarbophil (p<0.01) strongly

implicates the catechol functionality in promoting bioadhesion. The catechol functional

groups may play a role in multivalent cationic binding as they do in other species [4,5].

At the 7 minute hold time testing condition, all of the polymers excepting PBMAP

exceeded the mean maximum recorded manometric pressure of 213 mN/cm2 in rat small intestines [26]. While in the 14 minute hold time testing condition only Polycarbophil and

PEMAD exceeded 213 mN/cm 2 [26]. Of all the polymers tested, only PBMAD and

PEMAD exceed the 440 mN/cm 2 manometric pressures recorded in the human proximal

small intestines during phase 3 of digetsion [19]. While manometric pressure is not a

direct measure of the force exerted by the GI on an oral dosage form, it provides a

guideline for predicting success of bioadhesive dosage forms. Mucus turnover and

130 cohesive failure strength of the mucus lining also play significant roles in the in vivo performance of bioadhesives, as discussed in Chapter 4 [14,18,27]. In light of the fracture strength data presented in Figures 5.4a and 5.5a, the DOPA-derived polymers show tremendous promise for use as bioadhesives in oral drug delivery.

Discussion of the role of hydration time in bioadhesion testing

Creating adhesive forces in the fully hydrated state is very challenging [12,19]. Therefore the rate of hydration and water solubility of bioadhesive polymers strongly affects its adhesive properties [12,19]. For that reason the PBMA and PEMA polymers required different tissue contact times for measuring bioadhesive properties. Bioadhesives with varying hydration times and durations of bioadhesiveness in aqueous media could directly impact the performance of oral formulations. Bioadhesives have demonstrated the ability to promote intimate contact with the GI mucosa for prolonged periods of time leading to increased bioavailability of small molecule drugs [12,19]. Additionally, investigators have reported a relationship between increasing bioadhesiveness and increasing nanoparticle uptake [5,25,28]. Given the therapeutic aims of the oral formulation, taking into account the pharmacokinetics of the release and mucus turnover, choosing a polymer that will remain bioadhesive for the desired duration is of great importance to the field of oral drug delivery.

For example to achieve prolonged release in the intestines of a small molecule over the period of hours a bioadhesive with a low rate of hydration might be ideal, e.g. poly(fumaric-co-sebacic anhydride). However, as a carrier to enhance nanoparticle

131 uptake the bioadhesive polymer may function to promote contact between the nanoparticle and the GI mucosa for a short time until the nanoparticle can achieve mucus permeation and then dissolve prior to nanoparticle uptake. The bioinspired bioadhesives presented in this manuscript can be used to begin to test how the duration of bioadhesiveness affects uptake of small and large molecules delivered orally to the GI, as well as via other means to mucus-coated membranes including inhalational, vaginal, and ophthalmic routes.

5.4 Conclusions

DOPA-functionalized polymers with hydrophobic backbones represent a new class of excellent bioadhesives. PBMAD and PEMAD showed higher fracture strengths and tensile works than Polycarbophil, exemplary of the most common commercially used bioadhesives – acrylic acid derivatives. DOPA binding and chelation of multivalent cationic metal ions in other biological systems introduce the potential for utilization by synthetic bioadhesives to improve upon above standard hydrogen bonding models.

PBMA- and PEMA-derived, bioinspired polymers also exhibit different rates of hydration affecting the onset and duration of bioadhesiveness opening a window to test how therapeutics respond to bioadhesives with varying time courses of efficacy.

Synthetic biomaterials inspired by the strong adhesion of mollusks in salt water environments present a new class of bioadhesives that enable the design and testing of new drug delivery systems and device coatings.

132 a) Polycarbophil

1 2 b) 1 R R 2 R R ∆ O O + O O O DMSO OH HN

R 3

c) PBMA PBMAP PBMAT PBMAD

n n n n

d) PEMA PEMAP PEMAT PEMAD OH OH OH

O O O

n n n n O O O NH NH NH HO HO HO

O O O

OH OH

Figure 5.1: Chemical structures of polymers tested for bioadhesive properties. (a) Chemical structure of a generalized acrylic acid polymer such as Polycarbophil. (b) To synthesize bioinspired bioadhesive polymers with hydrophobic backbones and amine- functional side chains, maleic anhydride based polymers were heated in anhydrous conditions in dimethyl sulfoxide (DMSO) with a molar excess of the amine-functional agent (i.e. phenylalanine, tyrosine, or DOPA) to cause covalent attachment of the side chain. (c) Chemical structures of Polycarbophil AA-1 (Polycarbophil) and novel bioinspired, bioadhesive poly(butadiene maleic anhydride) (PBMA) and its derivatives: poly(butadiene maleic acid-phyenylalanine) (PBMAP), poly(butadiene maleic acid- tyrosine) (PBMAT), and poly(butadiene maleic acid-DOPA) (PBMAD). (d) Chemical structures of novel bioinspired, bioadhesive poly(ethylene maleic anhydride) (PEMA) and its derivatives: poly(ethylene maleic acid-phyenylalanine) (PEMAP), poly(ethylene maleic acid-tyrosine) (PEMAT), and poly(ethylene maleic acid-DOPA) (PEMAD). In (c) and (d) the backbone carbons and aromatic carbons for which the bound hydrogens highlighted in yellow and blue respectively are used in the calculation of side chain attachment efficiency.

133 a)

PBMA PBMAP PBMAT PBMAD

PBMA PBMAP

PBMAT PBMAD

PEMA PEMAP PEMAT PEMAD

PEMA PEMAP

PEMAT PEMAD

Figure 5.2: Chemical analysis of synthetic bioinspired bioadhesives. Photograph and 1H NMR spectra of PBMA-derived (a) and PEMA-derived (b) , bioinspired bioadhesive polymer solutions in deuterated dimethyl sulfoxide (DMSO). The peaks assigned relative area values of 2.000 correspond to either the olefinic protons in the PBMA monomer (a) or to the backbone methine protons in the PEMA monomer (b) . Peak area is shown in red text on each spectrum and the peak areas of the backbone carbons are highlighted in yellow and that of the aromatic carbon bound hydrogens is highlighted in light blue. Phenylalanine, tyrosine, and DOPA derived polymers have a maximum of 5, 4, and 3 hydrogens ( δ~6-7) per monomeric unit of the backbone assuming 100% attachment efficiency. By comparing the ratio of the areas under the backbone and side chain associated peaks to the maximum theoretical attachment, the side chain attachment efficiency is quantified.

134

PBMAP PBMAT PBMAD

PEMAP PEMAT PEMAD

Figure 5.3: Multiplicity edited hetero-nuclear single quantum coherence (HSQC) nuclear magnetic resonance (NMR) spectra used to confirm the peak assignment of bioinspired bioadhesive polymers. Green indicates in phase and blue indicates opposite phase carbons.

135

Side Chain Polymer Attachment Efficiency PBMAP 79% PBMAT 86% PBMAD 75% PEMAP 98% PEMAT 73% PEMAD 91%

Table 5.1: Bioinspired bioadhesive polymer side chain attachment efficiencies.

136

a) b) Probe

R R-a

R r r Polymer Probe a Intestinal Segment Tissue Segment c) i) ii) Heated Tissue Holder

Texture Analyzer

Figure 5.4: (a) Schematic of tensile bioadhesion testing setup. (b) Diagram of compressive deformation-based determination of contact area, A= πr2= π(R2-(R-a))2. (c) (i) Photograph of contact area between a carbon black coated probe and adhesive foam tape. (ii) NIH ImageJ outline of the contact area that experimentally validates the compressive deformation based contact area calculation approach.

137

a) 750 b) 25,000 *** **

20,000

500

15,000

10,000

250 [nJ] Work Tensile Fracture Strength [mN/sq cm] Strength [mN/sq Fracture 5,000

0 0 Polycarbophil PBMA PBMAP PBMAT PBMAD Polycarbophil PBMA PBMAP PBMAT PBMAD

Figure 5.5: (a) Bioadhesive properties of PBMA-derivative polymers as compared to the commercial bioadhesive, Polycarbophil AA-1, with respect to mean fracture strength . (b) PBMA derivatives compared to Polycarbophil with respect to mean tensile work. PBMAD demonstrates statistically significantly higher mean bioadhesive tensile work than Polycarbophil (p<0.001), PBMAP (p<0.01), and PBMAT (p<0.01). Error bars represent the standard error of mean. **p<0.01, ***p<0.001

138

a) 1,250 * b) 10,000 ** **

1,000 *

750

5,000

500 Tensile Work[nJ] Tensile Fracture Strength[mN/sq cm] 250

0 0 Polycarbophil PEMA PEMAP PEMAT PEMAD Polycarbophil PEMA PEMAP PEMAT PEMAD

Figure 5.6: Bioadhesive properties of PEMA-derivative polymers as compared to the commercial bioadhesive, Polycarbophil AA-1, showing that PEMAD has higher bioadhesive fracture strength than Polycarbophil (p<0.05), as well as the other PEMA- derivatives tested (p<0.01) (a) and that PEMAD has a higher mean tensile work value than Polycarbophil (p<0.05), PEMA (p<0.05), and PEMAP (p<0.01) (b) . Error bars represent the standard error of mean. *p<0.05, **p<0.01

139 5.5 References

1. P.A. Aubin, The limpet's power of adhesion. Nature 45 (1892), 464-5.

2. A.M. Smith, The Structure and Function of Adhesive Gels from Invertebrates.

Integ. And Comp. Bio. 42 (2002), pp. 1164-1171.

3. A.M. Smith, The Role of Suction in the Adhesion of Limpets. J. Exp. Bio.

161(1991), pp. 151-69.

4. J.H. Waite, Reverse engineering of bioadhesion in marine mussels. Bioart.

Organs II: Tech., Med., and Mat. 875 (1999), pp. 301-309. edited by D. Hunkeler,

D, A. Prokop, A.D. Cherrington, R.V. Rajotte, RV, and M. Sefton.

5. J.H. Waite, The DOPA Ephemera - A Recurrent Motif in Invertebrates. Bio. Bull.

183 (1992), pp. 178-84.

6. H. Lee, N.F. Scherer, and P.B. Messersmith, Single-molecule mechanics of

mussel adhesion. Proc. Nat. Acad. Sci. 103 (2006), pp. 12999-13003.

7. D.L. Dalsin, B.H. Hu, B.P. Lee, P.B. Messersmith, Mussel adhesive protein

mimetic polymers for the preparation of nonfouling surfaces. J. Am. Chem. Soc.

125 (2003), pp. 4253–8.

8. J. Schnurrer, and C.M. Lehr, Mucoadhesive properties of the mussel adhesive

protein. Int. J. Pharm. 141 (1996), pp. 251-256.

9. B.P. Lee, J.L. Dalsin, and P.B. Messersmith, Synthesis and gelation of DOPA-

Modified poly(ethylene glycol) hydrogels. Biomacromolecules 3 (2002), pp.

1038-1047.

10. C.A. Santos, B.D. Freedman, K.J. Leach, D.L. Press, M. Scarpulla, E.

Mathiowitz, Poly(fumaric-co-sebacic anhydride) - A degradation study as

140 evaluated by FTIR, DSC, GPC and X-ray diffraction. J. Control. Rel. 60 (1999),

pp. 11-22.

11. D. Duchene, and G. Ponchel, Bioadhesion of solid oral dosage forms, why and

how? Eur. J. Pharm. and Biopharm. 44 (1997), pp. 15-23.

12. D. Duchene, and G. Ponchel, Principle and Investigation of the Bioadhesion

Mechanism of Solid Dosage Forms. Biomaterials 13 (1992), pp. 709-14.

13. E. Jabbari, N. Wisniewski, and N.A. Peppas, Evidence of Mucoadheison by Chain

Interpenetration at a Poly(acrylic acid) Mucin Interface Using ATIR-FTIR

Spectroscopy. J. Control. Rel. 26 (1993), pp. 99-108.

14. N.A. Peppas, and J.J. Sahlin, Hydrogels as mucoadhesive and bioadhesive

materials: A review. Biomaterials 17 (1996), pp. 1553-1561.

15. J.M. Irache, C. Durrer, D. Duchene, and G. Ponchel, Preparation and

Characterization of Lectin-Latex Conjugates for Specific Bioadhesion.

Biomaterials 15 (1996), pp. 899-904.

16. C.A. Santos, B.D. Freedman, S. Ghosn, J.S. Jacob, M. Scarpulla, and E.

Mathiowitz, Evaluation of anhydride oligomers within polymer microsphere

blends and their impact on bioadhesion and drug delivery in vitro. Biomaterials

24 (2003), pp. 3571-3583.

17. B. Kriwet, and T. Kissel, Interactions between bioadhesive poly(acrylic acid) and

calcium ions. Int. J. Pharm. 127 (1996), pp. 135-145.

18. B. Laulicht, P. Cheifetz, A. Tripathi, and E. Mathiowitz, Are in vivo gastric

bioadhesive forces accurately reflected by in vitro experiments? J. Control. Rel.

134 (2009), pp. 103-10

141 19. A. Mellander, K. Järbur, and H. Sjövall, Pressure and frequency dependent

linkage between motility and epithelial secretion in human proximal small

intestine. Gut 46 (1999), pp. 376-84.

20. Q. Lin, D. Gourdon, C.J. Sun, N. Holten-Andersen, T.H. Anderson, J.H. Waite,

and J.N. Israelachvili, Adhesion mechanisms of the mussel foot proteins mfp-1

and mfp-3. Proc. Nat. Acad. Sci. 104 (2007), pp. 3782-6.

21. M.A. Schestopol, J.S. Jacob, R. Donnely, T.L. Ricketts, A. Nangia, E.

Mathiowitz, Z. Shaked, Bioadhesive Polymers with Catechol Functionality,

WO2005/056708.

22. Y. Jacques, and P. Buri, Optimization of an ex vivo Method for Bioadhesion

Quantification. Eur. J. Pharm. and Biopharm. 38 (1992), p. 195-8.

23. M.J. Tobyn, J.R. Robinson, and P.W. Dettmar, Factors Affecting in vitro Gastric

Mucoadesion I. Test Conditions and Instrumental Parameters. Eur. J. Biopharm.

41 (1995), pp. 235-41.

24. E.M. Wein and D.R. Van Campen, Mucus and iron absorption regulation in rats

fed various levels of dietary iron. J Nutrition , 121 (1991), pp. 92-100.

25. E. Mathiowitz, J.S. Jacob, Y.S. Jong, G.P. Carino, D.E. Chickering, P.

Chaturvedi, C.A. Santos, K. Vijayaraghavan, S. Montgomery, M. Bassett, C.

Morrell, Biologically erodable microsphere as potential oral drug delivery system.

Nature 386 (1997), pp. 410-414.

26. D.M. Ferens, E.C. Chang, G. Bogeski, A.D. Shafton, P.D. Kitchener, and J.B.

Furness, Motor patterns and propulsion in the rat intestine in vivo recorded by

spatio-temporal maps. Neurogastro. and Mot. 17 (2005), pp. 714-20.

142 27. S.A. Mortazavi, and S.A. Smart, Factors Influencing Gel-Strengthening at the

Mucoadhesive-Mucus Interface. J. Pharm, and Pharma. 46 (1994), pp. 86-90.

28. P. Decuzzi, and M. Ferrari, The role of specific and non-specific interactions in

receptor-mediated endocytosis of nanoparticles. Biomaterials 28 (2007), pp.

2915-22.

143

Chapter 6

Understanding Gastric Forces Calculated from

High Resolution Pill Tracking

Abstract

While other methods exist for monitoring gastrointestinal motility and contractility, this study exclusively provides direct and quantitative measurements of the forces experienced by an orally ingested pill. We report motive forces and torques calculated from real-time, in vivo measurements of the movement of a magnetic pill in the stomachs of fasted and fed humans. Three dimensional net force and two dimensional net torque vectors as a function of time data during gastric residence are evaluated using instantaneous translational and rotational position data. Additionally, the net force calculations described can be applied to high resolution pill tracking acquired by any modality. The fraction of time pills experience ranges of forces and torques are analyzed and correlate with the physiological phases of gastric digestion. We also report the maximum forces and torques experienced in vivo by pills as a quantitative measure of the

144 amount of force pills experience during the muscular contractions leading to gastric

emptying. Results calculated from human data are compared with small and large animal models with a translational research focus. The reported magnitude and direction of gastric forces experienced by pills in healthy stomachs serves as a baseline for comparison with pathophysiological states. Of clinical significance, the directionality associated with force vector data may be useful in determining the muscle groups associated with gastrointestinal dysmotility. Additionally, the quantitative comparison between human and animal models improves insight into comparative gastric contractility that will aid rational pill design and provide a quantitative framework for interpreting gastroretentive oral formulation test results.

6.1 Background and Introduction

The question of what forces a pill experiences in the gastric environment is of great importance to the rational design of pills that target drug delivery to the proximal gastrointestinal (GI) tract. Many therapeutic agents would benefit from increased residence time in the stomach and a number of gastroretentive techniques for prolonging gastric residence time have been devised and tested [1-3]. The most prominent gastroretentive methods are density mismatching, geometry-based, and bioadhesive doses

[1-3]. By calculating the forces a pill experiences in human stomachs, in vitro models can

be designed that will aid in predicting clinical success. The net force results in this study

can also serve as baseline measurements for comparison with pathophysiological GI

dysmotilities. The calculations can also be used to analyze differences in strength of

gastric emptying forces amongst age groups and patient populations useful for disease-

145 specific oral dosage design. Additionally, understanding the quantitative relationships among small animal, large animal, and human gastric forces provided by this work will aid in the selection of animal models and in interpreting translational research results.

Previous methods of measuring gastric forces include manometry measurements, which can be made by placing a balloon catheter in the stomachs of patients and monitoring pressure experienced by the balloon. Vassallo et al . added a load cell to a balloon catheter to monitor load experienced in the antegrade direction [4]. By placing balloon catheters within the antral portions of the stomachs of human volunteers, linear force measurements were made and correlated to muscular contractions [4]. Measurements were reported as the cumulative load experienced by the balloon over a period of 30 minutes and are the most closely related to the measurements made in our study [4].

However, the measurements are made on a 2 centimeter diameter balloon, which is far larger than a typical pill [4]. Additionally, the balloon catheter is tethered and therefore unrepresentative of motive forces experienced by a pill [4]. Moreover, the force measurements are limited to a single axis, whereas the measurements reported in this study are made in three translational and two rotational axes [4]. The measurements made by Vassallo et al. would also be difficult to make in smaller animals due to the size requirements of the experimental apparatus [4].

Studies by Kamba et al. investigated the relationship between gastrointestinal contractility and pill crushing force by assessing the destruction of pills with varying moduli in healthy male subjects [5,6]. Magnetic tracking methods can also be used to

146 assess crushing force, disintegration, and tablet breakability [7]. The crushing force of the

gastrointestinal tract is very useful for designing oral dosage forms. However, crushing

force studies do not elucidate the motive forces experienced by a pill in the stomach that most directly relate to aboral propulsion into the small intestines [5,6].

Parkman et al. have produced a miniaturized pressure sensor that they have incorporated

into a pill, which communicates real-time manometry and other data via radio telemetry

during gastrointestinal transit [8]. While the manometry data alone does not yield force or

torque data, forces and torques could be calculated and paired with the pill data given

detailed position as a function of time measurements. While manometry measures the

total contractile action of the entire muscularis mucosae, the directional data associated

with force measurements in the appropriate coordinate system could be used to evaluate

the contractility of individually oriented muscle layers (e.g. circumferential or

longitudinal).

Biomechanics testing and manometry have been employed to investigate the contractility

and motility of stomach muscle [9-11]. Biomechanical measurements give great insight

into the operation of stomach muscle and can measure how much force muscle exerts for

a given morphology and manometry techniques yield quantitative information regarding

local pressure changes during gastrointestinal contractions [9]. However, forces

experienced by a pill result from the pressure differences across the surface of a pill, its

interaction with the mucosal lining, and the gastrointestinal contents. All of these factors

affect the motion producing forces experienced by pills in the gastrointestinal tract.

147 Therefore monitoring the motion of pills in real time is important to the accurate determination of gastric motive forces.

The method we employed for calculating instantaneous net forces experienced by model pills in the stomach began by obtaining high-resolution pill tracking data.

Superconducting Quantum Interference Device (SQUID), radiotelemetry, ultrasound, fluoroscopy, and gamma scintigraphy are all capable of providing high resolution pill location as a function of time data [8,11-15]. All of the methods are high cost and all except SQUID require image analysis to extract position data [15,16]. Additionally, pill tracking methods have primarily been used to determine gastrointestinal tract residence time. Recently our group and others have reported measurements using inexpensive, highly accurate real-time magnetic tracking [17-20]. We employed a Hall array sensor technology to inexpensively and non-invasively track the position and orientation of magnetic model pills within the stomachs of humans, rats, and dogs without anesthesia.

Force calculations were made using the acquired position data. The same force calculation technique could be applied to similar data obtained by any methodology, including the ones listed above. By evaluating the force from the transient position vectors of pill motion enables us to answer the question of what forces a pill experiences in the stomach. The data provides a pill’s eye view of the stomach that serves as a platform for investigating basic gastroenterological questions of motility and contractility, as well as for establishing quantitative design criteria for gastroretentive dosage forms.

148 6.2 Results and Discussion

Pill tracking analysis

The motion of the model oral dosage forms while in the stomachs of humans, dogs, and

rats is governed by the net forces and torque it experiences. Gastric transit data provided

the Lagrangian position of the magnetic model pill in three translational [x( t), y( t), z( t)] and

two rotational [θ (t), ϕ(t)] coordinate planes at a rate of 10 Hz. In the human studies the zˆ corresponds to the cephalo-caudal, yˆ to the dorsal-ventral, and xˆ to the lateral

directions. While in the animal studies zˆ corresponds to the dorsal-ventral, yˆ to the cephalo-caudal, and xˆ to the lateral directions as plotted in Figure 6.1a. The myoelectric slow wave that governs the frequency of gastric contractions occurs at 0.05-0.08 Hz in humans, dogs, and rats indicating that 10Hz data collection rate can be assumed to provide sufficient resolution for instantaneous velocity and acceleration calculations

[6,11,17]. We evaluated the instantaneous translational velocity components as

Vx = dx (t) dt ,Vy = dy (t) dt ,Vz = dz (t) dt and the angular velocity components as

νθ = dθ (t) dt ,νϕ = dϕ(t) dt . Similarly, we evaluated the instantaneous acceleration (a x, a y, a z,

αθ, αΦ) of pill motion by evaluating corresponding instantaneous time derivatives of

velocity components. Given the acceleration, magnitude of net force ( v ) at each time Fnet

v point was calculated using the equation 2 2 2 , in which m denotes the Fnet ()t = m pill ax + a y + az pill

v mass of the pill. The magnitude of net torque ( τ net ) is calculated using the rotational

equivalent of the force equation, in which the pill is approximated as a cylinder,

2 v mpill l τ ()t = I α 2 + α 2 , in which I = . net cyl θ φ cyl 12

149 Force and Torque Analysis v v v The translational components of force Fx , Fy , and Fz from the data acquired during one of the fasted human trials are plotted as a function of time in Figure 6.1b. Although the v gravitational field points in the − zˆ direction, the Fz component follows the other force

components with no significant differences in magnitude or timing. Since that holds true

for all cases studied the gravitational component, the weight of the pill, was not factored

into calculations. For studies in which gravity pointed in a direction other than − zˆ during

testing, a Cartesian coordinate transformation was performed so that gravity was pointing

in the − zˆ direction for data analysis. Data analyzed in this study were from healthy subjects, establishing the force experienced by pills under physiological conditions. In the event of gastrointestinal dysmotility, unusual force patterns along a particular axis could indicate pathophysiological neuromuscular activity of a particular layer of the muscularis mucosae. Specific knowledge of the muscle fibers implicated in a case of gastrointestinal dysmotility provided by force and torque modeling data may aid in diagnosis or in deciding the course of treatment. Additionally, the directionality of the force vector data is useful in analyzing the contributions of different layers of the muscularis mucosae to gastric emptying.

Three components of force and two components of torque constitute force and torque vectors. Small fluctuations in the magnitude of force and torque (<250 dynes and <100 dynes*cm respectively) reflect small stomach wall movements and noise, as shown in

Figure 6.1b and 6.1d. In previous studies, Stathopoulos et al . [20] demonstrated that the dominant frequencies of pill movement unrelated to measurement noise match the

150 respiration and heart rate. Large spikes in the magnitude of force and torque appearing

approximately 1,100 seconds after dosing this particular fasted human correspond to the

phasic contractions associated with gastric emptying that occur just before the pill exits

the stomach through the pyloric sphincter into the duodenum, as shown in Figure 6.1c.

The maximum force and torque experienced by the model pills prior to exiting the v v stomach is the gastric-emptying force Fmax and torque τ max are reported for fasted and

fed humans, dogs, and rats. Plotting the magnitude of the force and torque vectors v v ( Fnet and τ net ) as a function of time gives a pill’s eye view of the digestive forces

experienced in the gastric environment, Figure 6.1.

It is important to note that in the fed state, the stomach is filled with partially digested

food mixed with gastric secretions, i.e. chyme [21]. In this case, the stomach contents

have the properties of water. In the fed state the presence of food increases the chyme

viscosity [21-23]. Experiments to determine the exact chyme viscosity are invasive and

often involve aspiration of contents or excision of the region of the gastrointestinal tract

in question. Such experiments would detract from the non-invasive nature of the study. v Hence, in this study, the contribution from viscous force on the net force Fnet was not evaluated.

Human Trials

In the fasted state, pills are the only ingested solid stomach contents therefore minimal muscular contractions occur until the migrating myoelectric complex (MMC) or

“housekeeping” wave, which occurs approximately every 90 minutes in fasted humans

151 and causes large spikes in force until the stomach is empty [24]. Figures 6.1b and 6.1c

depict the components and magnitude of forces experienced by the magnetic pill over v v

time. The highest magnitude gastric motive forces: F1 =2481 dynes, F2 =3014 dynes, and v

F3 =1236 dynes, respectively (Note that the gravitational force in the z direction is always mg=667 dynes) occurring approximately 90 minutes after ingestion that directly precede gastric emptying are attributed to the MMC (seen in Figure 6.1b). Based on the force profile observed in Figure 6.1b, the motive forces were overlaid onto the 3D trajectory of the pill onto a lateral projection of the human stomach during the time period that corresponds to the MMC (1100-1200s). While the majority of the low magnitude forces v v v have randomly distributed orientations, the highest magnitude forces F1 , F2 , and F3 are oriented aborally while the pill is in the vicinity of the pyloric sphincter. In accordance v with GI transit, the final large magnitude force F3 prior to gastric emptying is aligned with the projected opening of the pyloric sphincter. In the exemplary fasted human data plotted in Figure 6.1c, the largest magnitude force vectors point primarily along the cephalo-caudal and dorso-ventral axes. Motive forces in the cephalo-caudal direction are likely associated with contractions of the circumferentially oriented muscularis mucosae.

In the fasted state, because motion of the dense magnetic pills is inertially dominated the motive forces are assumed to occur from solid-body motion rather than motion of the gastric contents. Therefore, in the fasted state analysis of the 3D force vectors calculated from data collected by any high resolution pill-tracking data in a gastric dysmotility patient could help to distinguish if the force profile along a particular axis is diminished or if all forces are diminished compared to healthy subjects to differentiate between weakening of muscle fibers in a particular orientation or of the muscularis as a whole.

152

When fed, the ingested food undergoes muscular contractions for a greater portion of the

gastric residence time of the pill, while the food grinds and mixes until it has been

sufficiently digested to pass through the pyloric sphincter. In the human gastric v environment the time fraction histogram of the magnitude of force ( Fnet ) suggests that the MMC wave accounts for a small fraction of the total gastric residence time in the v fasted state [24]. When fed, the time fraction Fnet histogram indicates a greater

percentage of time spent in digestive contractions that produce forces higher than

biorhythms in the fasted state. The motive forces experienced by the model pills are in

the low range (<3% of the weight of the pill, where the gravitational force in the z

direction is mg=667 dynes) during 94.5±1.9% of the gastric residence time in the fasted v state and 68.3±12.1% in the fed (Figure 6.2a). Torque ( τ ) histograms are statistically

similar in the fasted and fed states indicating that the presence of food minimally affects

the rotational forces experienced by the pills (Figure 6.2b).

v v While the time fraction Fnet and τ histograms describe the distribution of forces experienced by pills during the quiescent stages of digestion, quantifying the peristaltic forces and torques that lead to the ejection of the model pills from the stomach are of great utility to the assessment of gastric function and for the rational design of gatroretentive pills. The maximum force and torque experience during gastric residence is plotted in both the fasted and fed states in Figure 6.2c. The average human gastric- v average emptying force ( Fnet ) is 414±194 dynes (62% of the weight of the pill) (N=3) in the

153 fasted state, which is statistically insignificantly lower than in the fed state, 657±84 dynes

(99% of the weight of the pill) (N=3). Average human gastric-emptying torque shows a

similar trend in that fasted and fed show little difference, 2,525±5 and 2,582±39

dynes*cm, respectively. The insignificant differences in fed and fasted gastric-emptying

forces and torques in humans indicates that the motive forces exerted upon pills are

similar in both the fasted and fed states. In light of the emptying force similarities and

since the pills are too large to pass through the pyloric sphincter during the grinding and

mixing of ingested food, the MMC wave is likely responsible for gastric emptying of the

pills in both fasted and fed states.

For comparison with the uniaxial cumulative force results reported by Camilleri and

Prather, calculated forces were normalized by the cross-sectional area of a sphere with

equivalent volume to the pill (in the units of mechanical stress) [25]. The average

cumulative stress summed over the 30 minute period prior to gastric emptying is

160,000±70,000 dynes/cm 2 fasted and 520,000±270,000 dynes/cm 2 fed, 42% and 1.5 times the cumulative uniaxial area normalized forces (or stresses) measured by the force traction catheter respectively. The area-normalized forces (or stresses) results from both studies demonstrate no statistically significant difference (p>0.05) and are therefore are in good agreement [24].

Dog Trials

In the fasted and fed, force and torque histograms, no statistically significant difference is observed between the fasted and fed state (Figures 6.3a and 6.3b). One possible

154 explanation for the observed lack of differentiation may be that the canine stomach is

very muscular and the contractions are similarly forceful independent of the gastric

contents [6,25]. In keeping with the supposition that the canine stomach undergoes very

forceful digestive contractions, the average gastric-emptying force in the fasted state is

2,633±78.7 dynes (3.95 times the weight of the pill) and in the fed state is 2,483±161.5

dynes (3.73 times the weight of the pill). Average gastric-emptying force is 6% higher

fasted than fed, and there is no statistically significant difference between the values.

Also, no statistically significant difference is observed between the fasted and fed gastric-

emptying torques, 2,610±41.27 and 2,518±12.77 dynes*cm, respectively. The lack of

dependence of gastric emptying force on feed state may be attributed to the non-erodible

pill being large enough to require emptying by the MMC wave independent of feed state.

Rat Trials

The mass of the model pills dosed to rats is much smaller than those dosed to humans and dogs due to differences in size. The presence of food had no significant effect upon the average fraction of gastric residence time spent in specific ranges of force or torque

(Figure 6.4). The average gastric-emptying force and torque experienced show a trend towards increasing in the fed state values, which are 7.4 and 2.1 times the fasted averages, respectively. The average gastric-emptying force is 0.43±0.05 fasted (6% of the weight of the pill, mg=6.8 dynes) and 3.2±1.9 dynes (47% of weight of the pill) fed. The average gastric-emptying torque is 0.05±0.05 fasted and 0.12±0.01 dynes/cm fed (N=2 fasted and N=2 fed). While there is a trend towards increasing gastric-emptying force and torque, it is not statistically significant.

155 Interspecies Comparison and Implications

Results indicate that fed dogs and humans produce statistically similar gastric emptying forces and as such would be a significantly better preclinical model for gastroretentive dosage forms than rats. However, the fasted average canine gastric-emptying forces are roughly 5 times greater than those of human subjects indicating that in the fasted state canine stomachs may not provide a good gastric-emptying model for humans.

Since the dog and human model pills are identical, the comparison of forces and torques does not require size-normalization. However, the rat model pills scale with the size of the animal and so to compare the gastric environments, results were normalized by the cross-sectional area of spheres with equivalent volumes to the sphero-cylindrical pills to yield units of mechanical stress. Size normalized gastric-emptying stresses in the fed rats are on the order of fed and fasted humans and an order of magnitude lower than dogs

(Table 6.1). Fasted rats exhibit average normalized gastric-emptying forces an order of magnitude lower than humans and two orders of magnitude lower than dogs. Therefore if rats are used as a preclinical gastric-emptying model for humans, it is likely best to perform experiments in the fed state. Fasted and fed dogs exhibit more similar gastric- emptying forces and torques to humans and dogs can accept human-size pills, therefore although dogs exhibit higher gastric-emptying forces in general than humans they are better preclinical models of human gastric emptying than rats.

Researchers and clinicians can utilize the force and stress calculations presented in this paper as quantitative guidelines for approaching the active research topic of achieving

156 prolonged gastric residence time to improve the therapeutic benefit of pill-based

therapies. It is important to note that the pills used in this study are models in that their

size and mass approximate those typically dosed to humans, dogs, and rats. In the event

that dosage forms or feeding conditions are altered from the standard, the method of

using pill tracking data to monitor force can be employed to calculate more accurate

forces and stresses for those doses.

Discussion

Monitoring inertial net forces in real time has been made possible by the advent of increasingly high-resolution pill tracking methods including Hall array magnet position tracking and radiotelemetry. This manuscript presents the first quantitative analysis of inertial net forces experienced by magnetic model pills during gastric residence in humans and two pre-clinical animal models both fasted and fed as tracked by Hall array sensors. Calculations of maximal force experienced in the stomach, or gastric emptying forces, can serve as guidelines for the rational design of standard pill dosage forms as a guide for minimum hardness, frangibility, and crushing strength. The standard tablet parameters of crush strength and breakability can be easity tested against the in-vivo force and torque data. The method and calculations could readily be applied to yield inertial force data for any oral dosage form in nearly any species even if the pill differs significantly in size or mass from those reported in the study.

In particular, for pills designed to achieve increased gastric residence time, quantifying inertial forces is essential to the rational design of pills that will overcome gastric

157 emptying forces to remain in the stomach for extended periods of time. Gastroretentive

pills have been investigated for decades because increasing the residence time of pills in

the stomach would greatly benefit the numerous narrow absorption window therapeutics

that are primarily absorbed in the proximal small intestines [1-3]. Therefore by retaining

the pill in the stomach, proximal to the site of absorption, more of the drug could achieve

uptake and time-release formulations could be developed.

The most prevalent strategies for achieving gastric retention are density mismatching,

geometry-based, and bioadhesive pills [1-3]. The dense pill used in this study is an

excellent model for dense pills that are meant to reside on the greater curvature of the

stomach avoiding the pylorus for longer than pills that are of similar density to ingested

food. Floating pills, another density mismatching technique, could be analyzed with the

same technique by using less dense materials in pill fabrication and calculations and the v vv gravitational component ( Fbouyancy = ∆ρVg ) would be designed to be greater than the gastric emptying force.

Numerous swelling or unfolding pills have been developed with the intention of being easily swallowed and then changing shape upon introduction into the stomach such that they are too large to pass through the pyloric sphincter [1-3]. For swelling tablets, the forces calculated in this manuscript could be used as a guideline in designing tablets with a sufficiently high compressive strength to overcome the gastric emptying forces when swollen. Similarly, for unfolding films such as the Accordian Pill, gastric emptying

158 forces can be used to inform the minimum bending modulus of the unfolded film

necessary for gastric retention [27].

Finally, for bioadhesive dosage forms, the gastric emptying forces calculated indicate

how strong the tensile bioadhesive bond strength must be to achieve gastric retention [1-

3]. In bioadhesive doses, the cohesive strength of the loosely adherent mucus must also

be taken into account [28].

In addition to pharmaceutical design, gastric net force calculations set forth in this

manuscript can be used to assess gastrointestinal motility pathophysiologies and

differences in gastric forces experienced by pills in different age groups or patient

populations quantitatively. Patients with gastrointestinal dymostility, for example, may

benefit from the forces calculated based on non-invasive high resolution pill tracking to

help pinpoint exactly where a loss of motility has occurred yielding atypically low forces

in a particular region. GI net force calculations can also be used to assess the physiology

of aging patients to investigate any correlation between age and the strength of gastric

emptying force, which could be of use to physiologists, physicians, and pharmaceutical

scientists designing pills for diseases that afflict patient populations of a particular age

group.

6.3 Conclusions and Perspectives

The net forces and torques, experienced by model pills in the stomach, can be calculated from position as a function of time data acquired by any modality. Forces and torques

159 derived from the Motilis Magnet Tracking System data provide tremendous insight into

the motive forces experienced by standard oral dosage forms. Analysis of the calculated

forces yields a measure of gastric-emptying force experienced by pills that provides

researchers and clinicians in the fields of gastric retention and gastroenterology a

quantitative framework for designing gastroretentive pills and understanding their

behavior in preclinical and clinical trials. From a clinical perspective, force and torque

vectors calculated from high resolution pill tracking data provide directional data that

relate to physiology or a suspected pathophysiology useful in diagnosing gastric

dysmotility disorders and diseases. In the healthy fasted human subjects studied, the

direction of the forces with the greatest magnitude all originate in the distal greater

curvature of the antrum towards the pyloric sphincter implicating primarily

circumferential muscle fibers in gastric emptying. From a pharmaceutical research

perspective the gastric emptying force and torque data can be used to estimate what

bioadhesive or bouyancy force would be necessary to retain a dose within the stomach.

Additionally, from a clinical perspective force and torque monitoring provides magnitude

and directional data that serve as a baseline for comparing with pathophysiologal

digestive states.

6.4 Patients and Methods

Magnet Tracking System

The tracking system consists of a pill containing a permanent magnet, a detection matrix

(4 • 4 magnetic field sensors) and dedicated software implanted in a laptop computer

160 (MTS-1, Motilis, Lausanne, Switzerland) [18]. The tracking algorithm calculates the

position and the orientation of the pill at 10Hz, except the rotation around the

magnetization axis (i.e. 5 degrees of freedom, three translations and two rotations). The

trajectory of the magnet was monitored and stored. Force data was calculated from

position as a function of time data and the temporal distribution of force was analyzed.

Patients

The size of the pills can be adapted to the size of the subject, which allows using the same approach for human, large animals and rodents [17,18]. Pills containing magnets were either taken voluntarily by humans and dogs or by oral gavage for rats. The size and mass of the pills were 6.0 mm in diameter and 16 mm long (740 mg, density of 1.75 g cm-3), 5.3 mm in diameter and 15 mm long (530 mg, 1.8 g cm-3), and 0.85 mm in diameter and 1.1 mm long (4.5 mg, 7.0 g cm-3), for human, dogs and rats respectively.

The coating of the pill was Palaseal® for humans and dogs and gold for rats. Experiments in fasted and fed states were carried out. Rat and canine subjects confined to limit movement and pill tracking was performed without anesthesia. Human test subjects were seated in a semi-reclining position.

Three fed and three fasted human subjects ingested model non-erodible, rigid pills containing magnets per os . Four fed and four fasted Beagle dog subjects ingested the same model non-erodible, rigid pills as the human subjects. Additionally, two fed and two fasted Hooded Long Evans rat subjects were administered small cylindrical magnets that served as model oral dosages. For all subjects, the position of the magnet was

161 monitored by the Motilis Magnet Tracking System. Force data was calculated from

position as a function of time data and the temporal distribution of force was analyzed.

All testing was performed in accordance with IACUC and IRB guidelines.

We report maximum forces and torques and analyzed as measures of the forces

experienced during the peristaltic contractions that led to gastric emptying of a typical

oral dosage form, termed gastric-emptying force and torque.

Statistical Analysis

Comparisons between mean values of forces, torques, and stresses between feed states and amongst animals were analyzed by one-way analysis of variance (ANOVA).

162

Figure 6.1: (a) The solid red line shows a three dimensional trajectory plot of the model pill in an exemplary fasted human subject. A dashed, dark blue outline of the frontal projection of the stomach is superimposed to correlate the pill motion with the anatomical position. The pill resides predominantly along the greater curvature of the stomach. At first the pill resides more orally (position “1”) and at later times (position “2”) more aborally as expected. Ultimately the pill is emptied through the pylorus into the small intestines. (b)v Exemplaryv v fasted human subject axial force components in three dimensions ( Fx , Fy , and Fz ) showing no preference for the − zˆ gravitational direction v 2 2 2 plotted with the magnitude of the net force vector ( Fnet ()t = mpill ax + ay + az ) as a function v of time.v Thev three largest force vectors associated with gastric emptying are labeled F1 , F2 , and F3 (2481, 3014, and 1236 dynes respectively). (c) Three dimensional force vectors plotted with their origins at the position of the pill, corresponding with position 2 in the trajectoryv v plot, asv it moves during phase III of digestion (1100-1200s).

Propulsive forces F1 , F2 , and F3 generated by phasic contractions of the antrum are labeled and overlaid on a frontal projection of the stomach marked with a longitudinalv v (L)v

and circumferential (C) curvilinear coordinate system. The directions of F1 , F2 , and F3 indicate that circumferentially oriented muscle fibers play a large role in the gastric emptying of the magnetic model pill. (d) Exemplary fasted human inclination ( θ), v 2 2 orientation ( Φ), and magnitude of torque components ( τ net ()t = Icyl αθ +αφ ) as a function of time plot. As with force, torque remains low during the initial phases of digestion and then spike during the housekeeping wave resulting in gastric emptying of the pill.

163

a) Human Mean Force Histograms

on 0.75 i t c a r F

e 0.5 * m i T

0.25

0 0 10 20 30 40 50 60 70 80 90 100 |F| [dynes] Fasted Fed

b) Human Mean Torque Histograms

0.75 tion

0.5

Time Frac 0.25

0 0 100 200 300 400 500 |T| [dynes*cm] Fasted Fed

c) Human Gastric Emptying Force and Torque v. Feed State

1000 3000

750

2000

500 |F| [dynes] | [dynes*cm] τ 1000 |

250

0 0 Fasted Fed |F| |τ| v Figure 6.2: (a) Frequency histogram of the magnitude of force ( F ) experienced by pills in the stomachs of fasted and fed human subjects (N=3 each). In the fed state, the histogram curve shifts to the right reflecting greater time fraction of gastric residence that the pill experiences forceful contractions associated with the gastric grinding and mixing of ingested food (phase II and III of digestion). * p<0.05 between the fasted and fed cases v (b) Frequency histogram of the magnitude of torque ( τ ) experienced by pills in the stomachs of fasted and fed human subjects (N=3 each). No difference in torque distribution is observed between the fasted and fed states. (c) Plot shows the average maximum magnitude of force and torque experienced by the pills during gastric residence. The values correspond to the gastric emptying forces experienced in phase III of digestion that lead to the passage of the pills from the stomach to the small intestines. No statistically significant difference appears between feed states in either force or torque although there is a trend towards increased mean gastric emptying force and torque in the fed state. All error bars depict the standard error of the mean (SEM).

164 a) Dog Mean Force Histograms

0.75

0.5 Time Fraction

0.25

0 0 10 20 30 40 50 60 70 80 90 100 |F| [dynes] Fasted Fed

b) Dog Mean Torque Histograms

0.75

0.5

Time Fraction 0.25

0 0 100 200 300 400 500 |τ| [dynes*cm]

Fasted Fed

c) Dog Gastric Emptying Force and Torque v. Feed State

5000 3000

4000

2000 3000

2000 | [dynes*cm] |F| [dynes] τ

1000 |

1000

0 0 Fasted Fed |F| |τ|

Figure 6.3: (a) Frequency histogram of the magnitude of force experienced by pills in the stomachs of fasted and fed canine subjects (N=3 each). There is no appreciable difference in the magnitude of force distribution between the fasted and fed states. (b) Frequency histogram of the magnitude of torque experienced by pills in the stomachs of fasted and fed canine subjects (N=3 each). No significant difference in torque distribution is observed between the fasted and fed states. (c) While there is no statistically significant difference between feed states in either gastric emptying force or torque, canine gastric emptying forces are considerably higher than in the human in the fed state (p<0.05). Additionally, the average coefficient of variance in the canine gastric emptying forces is small (5%) compared to humans (30%). Error bars depict the SEM.

165 a) Rat Mean Force Histograms

0.75

0.5 Time Fraction 0.25

0 0 0.25 0.5 0.75 1 |F| [dynes]

Fasted Fed

b) Rat Mean Torque Histograms

1.25

0.75

0.5 Time Fraction

0.25

0 0 0.025 0.05 |T| (dynes*cm)

Fasted Fed

c) Rat Gastric Emptying Force and Torque v. Feed State

10 0.15

7.5

0.1

5 | [dynes*cm] |F| [dynes] τ

0.05 |

2.5

0 0 Fasted Fed |F| |τ|

Figure 6.4: (a) Frequency histogram of the magnitude of force experienced by pills in the stomachs of fasted and fed rats (N=2 each). Feed state does not significantly affect the time fraction of force distribution. (b) Frequency histogram of the magnitude of torque experienced by pills in the stomachs of fasted and fed rats (N=2 each). No difference in torque distribution is observed between the fasted and fed states. (c) Mean gastric emptying force in rats is 2 orders of magnitude smaller than in humans and 3 orders of magnitude smaller than in dogs. However, pills dosed to rats are significantly smaller than those dosed to dogs and humans due to the relatively small size of rats. Trends towards increased gastric emptying force and torque in the fed state are pronounced, but not statistically significant due to inter-subject variability.

166

Table 6.1: Gastric emptying forces and torques are directly comparable between canine and human studies because the pills used were identical. However, pills administered to rats are significantly smaller than those administered to dogs and humans. Therefore, gastric emptying forces and torques were normalized by the cross-sectional area of a sphere with the same volume as the pill in the units of mechanical stress. In comparing the area normalized gastric emptying force (or stress) between rats and humans, although the dosage forms administered to rats are substantially smaller and have significantly lower mass, fed rats exhibit similar area-normalized forces to fasted humans.

167 6.5 References

1. Bardonnet, PL; Favre, V; Pugh, WJ; Piffaretti, JC; Falson, F (2006) Gastroretentive

dosage forms: Overview and special case of Helicobacter Pylori. J Contr Rel 111 (1-2):

1-18.

2. Davis,SS (2005) Formulation strategies for absorption windows. Drug Disc Tod

10(4):249-57.

3. Talukder, R; Fassihi, R (2004) Gastroretentive delivery systems: A mini review. Drug

Dev and Ind Pharm 30 (10): 1019-1028.

4. Vassallo, MJ; Camilleri, M; Prather, CM; Hanson,RB; Thomforde, GM (1992)

Measurement of Axial Forces During Emptying from the Human Stomach. Am J Phys

263(2):G230-9, Part 1.

5. Kamba, M; Seta, Y; Kusai, A; Ikeda, M; Nishimura, K (2000) A unique dosage form to evaluate the mechanical destructive force in the gastrointestinal tract. Int J Pharm 208

(1-2): 61-70.

6. Kamba, M; Seta, Y; Kusai, A; Nishimura, K (2001) Evaluation of the mechanical

destructive force in the stomach of dog. Int J Pharm 228 (1-2): 209-217.

7. Goodman, K et al. (2010) Assessing Gastrointestinal Motility and Disintegration

Profiles of Magnetic Tablets by a Novel Magnetic Imaging Device and Gamma

Scintigraphy. J Contr Rel 74 (1) 84-92.

8. Parkman, HP; Jones, MP (2009) Test of Gastric Neuromuscular Function.

Gastroenterology 136: 1526-43.

168 9. Gregersen, H; Hausken, T; Yang, J; Odegaard, S; Gilja, OH (2006) Mechanosensory

properties in the human gastric antrum evaluated using B-mode ultrasonography during

volume-controlled antral distension. Am J Phys-GI and Liver Phys 290 (5): G876-G882.

10. Gregersen, H; Kassab, G (1996) Biomechanics of the gastrointestinal tract.

Neurogastro and Mot 8 (4): 277-297.

11. Hveem, K; Sun, WM, Hebbard, GS; Horowitz, M; Dent, J (1994) Insights into

Stomach Mechanics from Concurrent Gastric Ultrasound and Manometry.

Gastroenterololgy 107 (4): 1236-1236.

12. Christensen, FN et al. (1985) The Use of Gamma-Scintigraphy to Follow the

Gastrointestinal Transit of Pharmaceutical Formulations. J Pharm and Pharma 37 (2):

91-95.

13. Johannessen, EA; Wang, L; Reid, SWJ; Cumming, DRS; Cooper, JM (2006)

Implementation of radiotelemetry in a lab-in-a-pill format. Lab on a Chip 6 (1): 39-45.

14. Rao, SSC; Lu, C; SchulzeDelrieu, K (1996) Duodenum as an immediate brake to

gastric outflow: A videofluoroscopic and manometric assessment. Gatroenterology 110

(3): 740-747.

15. Weitschies, W; Kosch, O; Monnikes, H; Trahms, L (2005) Magnetic Marker

Monitoring: An application of biomagnetic measurement instrumentation and principles for the determination of the gastrointestinal behavior of magnetically marked solid dosage forms. Adv Drug Del Rev 57 (8): 1210-2.

16. Andra, W et al. (2000) A novel method for real-time magnetic marker monitoring in

the gastrointestinal tract. Physics in Medicine and Biology 45 (10): 3081-3093.

169 17. Guignet, R; Bergonzelli, G; Schlageter, V; Turini, M; and Kucera, P (2006) Magnet

Tracking: a new tool for in vivo studies of the rat gastrointestinal motility. Neurogastro

and Mot 18: 472-8.

18. Hiroz, P; Schlageter, V; Givel, J-C; and Kucera, P (2009) Colonic movements in

healthy subjects as monitored by a Magnet Tracking System, Neurogastro and Mot 21

(8): 837-8.

19. Stathopoulos, E; Schlageter, V; Meyrat, B; de Ribaupierre, Y; and Kucera, P (2005)

Magnetic pill tracking: a novel non-invasive tool for investigation of human digestive

motility. Neurogastro and Mot 17: 148-54.

20. Weitschies, W; Kotitz, R; Trahms, L; Cordini, D (1997) Gastrointestinal transit of a

magnetically marked capsule monitored using a 37-channel SQUID-magnetometer.

Journal de Physique IV 7 (C1): 667-668.

21. Miller, LJ; Go, VLW; Malagelada, JR (1978) Effect of Individual Chyme

Components on Gastric-Secretion and Emptying After a Meal. Gastroenterology 74 (5):

1068-1068.

22. Amidon, GL; Debrincat, GA; Najib, N (1991) Effects of Gravity on Gastric-

Emptying, Intestinal Transit, and Drug Absorption. J Clin Pharma 31 (10): 968-973.

23. Dillard, S; Krishnan, S; Udaykumar, HS (2007) Mechanics of flow and mixing at

antroduodenal junction. World J of Gastroenterology 13 (9): 1365-1371.

24. Bortolotti, M et al. (1983) The Interdigestive Migrating Motor Complex (IMMC) in

Idiopathic Chronic Gastric Retention. Gastroenterology 84 (5): 1112-1112.

25. Camilleri, M; Prather, CM (1994) Axial Forces During Gastric-Emptying in Health

and Models of Disease. Digest Dis and Sci 39 (12): S14-S17, Suppl. S.

170 26. Kararli, TT (1995) Comparison of the Gastrointestinal Anatomy, Physiology, and

Biochemistry of Humans and Commonly Used Laboratory-Animals. Biopharm and Drug

Disp 16 (5): 351-380.

27. Kagan, L et al. (2006) Gastroretentive Accordian Pill: Encancement of Riboflavin

Bioavailaility in Humans. J Cont Rel 113 (3): 208-15.

28. Laulicht, B; Cheifetz, P; Tripathi, A; Mathiowitz, E. (2009) Are in vivo gastric

bioadhesive forces accurately reflected by in vitro experiments? J Cont Rel 134 (2): 103-

10.

171

Chapter 7

Novel Method for Localized Delivery from

Magnetic Pills

Abstract

Numerous therapeutics demonstrate optimal absorption or activity at specific sites within the gastrointestinal (GI) tract. By monitoring attractive force between an orally administered magnetic dose and an external magnet, we developed an effective method for prolonging (>12 hours) localization of therapeutics within the rat GI. We simultaneously visualized internal dose motion in real time using biplanar videofluoroscopy. Combining the two data streams, we quantified tissue elasticity as a measure of tissue health during magnetic localization. Our technique improves safety, efficacy, and monitoring capacity of magnetically localized doses and provides a platform for testing the benefits of GI site-specific drug delivery.

172 7.1 Introduction

For many orally administered pharmaceuticals, increased residence time in a particular region of the gastrointestinal (GI) tract would greatly improve their therapeutic benefit

[1]. Controlling GI residence may even enable oral administration of therapeutics administered by injection [2-6]. We have developed a magnet-based delivery system visualized by biplanar videofluoroscopy in vivo that yields real-time monitoring and control over the duration of residence of model magnetic pills in the small intestines of rats. Our system can safely and reliably retain drugs for up to 12 hours in any region of the GI with the ability to control the force applied by the orally ingested magnet to the intestinal wall. What our method of GI retention adds to previous systems is the ability to visually confirm the anatomical location of the oral dose and to constantly monitor and control the inter-magnetic force ensuring safe capture of the oral dosage in the appropriate region of the GI [1,7].

Previous studies have used external magnets to improve the bioavailability of orally administered proteins including insulin 4, narrow absorption window (NAW) therapeutics

including acyclovir [1,8] and therapeutics for site-specific pathologies including

bleomycin for esophageal cancer [9]. In all previous studies, the magnet was applied in

fixed position without monitoring inter-magnetic force or visually verifying the capture

of the oral dosage [1,4,8,9]. This study presents the first method for monitoring the force

applied by an orally administered magnetic dosage to the GI tissue to ensure safety and

efficacy of prolonged retention at a site of therapeutic interest. Though not previously

tractable, localized oral drug delivery would be extremely useful for delivery of

173 therapeutics for inflammatory bowel disease to the colon, of orally administered

chemotherapeutics to GI cancers, and of oral vaccines to the ileum. 1-12 In particular, our platform method enables investigation of the benefits of localized as compared to systemic administration of therapeutics.

7.2 Results and Discussion

Early GI magnetic retentive efforts for oral administration focused on creating the maximal attractive force between a dosage and an external magnet to either retain a large dosage form at the site of interest [8] or to increase the uptake of magnetic nanoparticle formulations [2] at the site of interest. We utilized a computer-controlled material testing device equipped with a load cell that has a programmed feedback loop, adjusting the position of the external magnet to constantly cycle between upper and lower inter- magnetic force bounds defined by the user in real time. As a result, we ensure that the GI tissue experiences the smallest force possible that still retains the magnetic oral dose.

Magnetic localization ensures and prolongs intimate contact between the dose and the absorptive GI epithelium promoting uptake and bioavailability without damaging intestinal tissue.

The orally administered model dose consists of a cylindrical neodymium iron boron

(NIB) magnet ( Φ=1.6mm, length=1.6mm) coated in chrome and a non-erodible polymer

to ensure that the magnet is not damaged within the GI. A freeze dried calcium alginate

sphere containing magnetic, radiopaque iron oxide microparticles was placed at either

end of the magnet, held in place by the attractive force of the internal magnet (Figure

174 7.3). Any more than one sphere on each end will dissociate from the internal magnet due to GI motility when held by the magnetic attractive force alone. We used alginate spheres because they can readily be loaded with therapeutics [13]; however, any drug delivery device that can be affixed to the internal magnet could alternatively be used [1].

Each dose was administered by oral gavage to rats prior to physical restraint. After the dosed magnet entered the small intestines, the restrained rat is placed on a modified materials testing device without anesthesia (Figure 7.1a). A cylindrical NIB magnet

(Φ=25mm, length=25mm) is brought towards the subject until a maximal force of 4mN is achieved. Upon reaching the maximum desired force, the external magnet retreats from the subject until a minimal force of 1mN is reached. The cycle repeats with the external magnet moving at 0.5mm/s for a period of 12 hours. Periodically releasing the inter- magnetic force approximately every 10 seconds, by force cycling, allows the tissue to recover from intestinal vasculature compression and mesenteric stretching in between periods of maximal force. Biplanar fluoroscopic videos were recorded at prescribed timepoints (first instance of retention, and 1, 2, 4, 8 and 12 hours thereafter) to quantify internal dose motion. Exemplary x-ray images from fluoroscopy videos showing the magnetic dosage in the small intestines, with co-administration of aqueous barium sulfate for contrast, are provided in Figures 7.1b and 7.1c.

To visualize the anatomical location and magnet-induced motion of the internal magnet, we use biplanar x-ray videofluoroscopy [14]. From the orthogonal biplanar fluoroscope videos, we tracked the three dimensional position of the internal magnet over time [15].

175 Motion of the internal magnet over the course of an exemplary force cycle is plotted in

three dimensions in Figure 7.1b. From proper orthogonal decomposition, we calculate

that 98.2±1.8% of the three dimensional motion of the pill can be described by a single

axis (N=5), termed mode 1 [16]. By taking the slope of the inter-magnetic force as a

function of position along mode 1, we calculate the Hookean elasticity of the tissue

(Figure 7.1c). Due to hysteresis in the force-distance curve caused by the viscoelastic

nature of the intestinal tissue, the slope of the ascending, descending, and whole force

cycle was measured over time to determine if the tissue retains its mechanical integrity

(Figure 7.1d). The elastic constants measured at the start of the retention were

insignificantly different from those measured at later timepoints as analyzed by one-way

ANOVA (P=0.52, P=0.68, and P=0.48 with respect to the ascending, descending, and

whole force cycle respectively). Therefore, the intestinal tissue maintained its mechanical

integrity throughout 12 hours of retention (N=5), suggesting that the structural integrity

of the tissue was maintained. Histological analysis post mortem confirmed that no signs of damage were caused by retention (Figure 7.4). Additionally, the FDA classifies magnets with field strength <2T, such as the neodymium iron boron permanent magnets used in this study, as nonsignificant risk devices [7].

Effectiveness and reproducibility of real-time force monitoring magnetic localization for

12 hours was confirmed in 6 additional rats by standard x-ray (Figure 7.2a). In all cases, the magnet remained within the small intestines for 12 hours, while without an external magnet none of the rats had dosages within the small intestines after 12 hours.

Additionally, when the entire rat is removed from the materials testing device, the inter-

176 magnetic force measurement drops instantaneously, indicating the loss of magnetic

retention to the investigator in real time. Due to the slow mean velocity of pill intestinal

transit (2 cm/min in the rat jejunum) and the length of the rat jejunum (~100 cm) the

external magnet can be removed for up to ~1 hour before the dose has progressed into the

next segment of the GI [17].

Inter-magnetic force as a function of inter-magnetic distance is minimally affected by the

presence of a live subject (Figure 7.2b), which demonstrates that magnet size and

strength selection in vitro will translate well into in vivo studies in any species. Therefore given readily quantifiable parameters including the lateral dimensions of the experimental subject, the inter-magnetic force, and the GI propulsive force in the region of retention, capture efficacy can be evaluated in vitro prior to live subject studies. If the inter- magnetic force measured at the physiologically relevant distance between the nearest external surface of the subject and the internal magnet is greater than the maximal propulsive force in the region, estimated by analyzing the inertial force from high resolution magnetic pill tracking data as described in Chapter 6 [18], magnetic capture can be expected for any species including humans. Duration of capture is commensurate with the application of external magnetic force cycling.

To quantify the net inertial force experienced by the internal magnet, we tracked the three dimensional position as a function of time and its instantaneous acceleration (Figure 7.2c)

[18]. Since the inertial net force is only 0.0005±0.0005% of the measured intermagetic

force, the inter-magnetic force closely approximates the force the magnet applies to the

177 GI tissue. Normalizing the force experienced by the tissue by the cross-sectional area of contact with the internal magnet yields a measure of the stress experienced by the tissue that ranges from 4-15 mmHg. Manometric jejunal pressures recorded in rats range from

4-16mmHg indicating that that retention stresses are within the normal physiological range for rat jejunal tissue [19].

7.3 Conclusions

Monitoring and controlling a cyclical inter-magnetic force between a magnetic oral dose and an external magnet enables the safe and effective localization of a model drug delivery system. Magnetic force monitoring can report the inter-magnetic force and distance in real time. Biplanar x-ray fluoroscopy with contrast enables visualization and quantification of the three dimensional position of the internal magnet in vivo . Co-

administration of a radiopaque contrast agent can elucidate more precisely the anatomical

position of the magnet at the cost of magnet localization precision. Together, inter-

magnetic force monitoring and biplanar fluoroscopic visualization provide the first

localized oral drug delivery system with quantitative means of assessing safety and

efficacy, in terms of both intestinal tissue damage and localization. Magnetically

localized oral drug delivery will be readily applicable to investigating the therapeutic

benefit of prolonged local delivery of NAW therapeutics within their therapeutic

windows, of chemotherapeutics to GI tumors to avoid side effects caused by systemic

administration, of nanoencapsulated proteins postulated to achieve increase uptake in

certain regions of the small intestines, and of therapeutics for GI diseases enabling

administration directly at the site of action.

178 7.4 Materials and Methods

Magnetic Pill Preparation

Orally administered doses consist of two freeze dried alginate spheres on either side of a

NIB magnet ( Φ=1.6mm, length=1.6mm, KJ Magnetics, Jamison, PA). The alginate

spheres were created by introducing 30w/v% iron microparticles (Sigma-Aldrich, Saint

Lois, MO) suspended in an aqueous 2w/v% low viscosity sodium alginate (Sigma-

Aldrich, Saint Lois, MO) into an aqueous 1w/v% calcium chloride (Sigma-Aldrich, Saint

Lois, MO) receiving bath. The sodium alginate solution was extruded through a 21 gauge

syringe needle at 3ml/min by a vertically oriented syringe pump (Harvard Apparatus,

Holliston, MA). Upon entering the receiving bath, the divalent cationic calcium ionically

cross-links the polyanionic alginate as described previously. 13 The alginate spheres are collected, rinsed with distilled water and then freeze dried overnight. Once the doses are assembled, they are loaded into size 9 gelatin capsules (Torpac, Fairfield, NJ) for oral gavage. While magnetic alginate spheres are suitable for delivering numerous therapeutics in a controlled fashion, any dosage form that is small enough for gastric emptying and that has intermagnetic strength for retention could be used with any species.

Force Monitoring of Magnetic Localization

A texture analyzer XT-plus (Texture Technologies, Scarsdale, NY) was modified to hold a rat in an acrylic restraint tube on its base while the load cell containing arm is oriented to move horizontally. The texture analyzer was programmed using Texture Exponent

179 Software (Texture Technologies, Scarsdale, NY) to begin its cycle 50mm from the outer

surface of the acrylic restraint tube and to approach the tube at 0.5mm/s until a force of

4mN is reached. Upon reaching an intermagnetic tensile force of 4mN, the arm moves

away from the rat at constant speed until a minimal force of 1mN is reached. The cycle,

which takes approximately 30 seconds, repeats for a user-defined period. Force cycling

allows for the intermittent release of retaining force to ensure minimal tissue damage.

Real time intermagnetic force monitoring the force ensures that the internal magnet does

not apply any undue stress to the GI tissue.

Biplanar Videofluoroscopic Spatial Calibration, Visualization, Tracking, and

Analysis of Magnet Motion

C-arm fluoroscopes (OEC Model 9400) were retrofitted with 30cm Image Intensifiers

(Dunlee model TH9432HX, Dunlee Inc., Aurora, IL) and Photron video cameras

(Fastcam 1024 PCI, Photron, inc., San Diego, CA). Algorithms to account for distortion introduced by the fluoroscopes and to determine their 3D positions were executed in

MATLAB (The Mathworks, Natick, MA) using custom software and a 64-point calibration cube. 14 We used MATLAB scripts embedded in XrayProject version 2.0.7,

available for download at http://www.xromm.org. Marker tracking scripts, embedded

within XrayProject version 2.0.7 were used in calculating 3D positions of the internal

magnet and the external arm. 14,15 We used a 30Hz low-pass filter to remove breathing artifact in the internal magnet’s 3D coordinates.

180 Motion of the internal magnet was primarily along a single line (the line of motion of the

TA arm). We tracked movement in world space, but this line of action of the TA arm

was not precisely contained within either the x, y or z dimension of our calibration cube.

To reduce this dimensionality, we used Proper Orthogonal Decomposition (POD).

Mathematically, this technique is identical to Principal Components Analysis (PCA) or

Singular Value Decomposition (SVD), transforming 3D coordinate space such that one

axis (herein termed mode 1) explains the greatest possible amount of variation in the

data. 16 Given three parameters (x, y, and z coordinates, in cube space, over time), our dataset has three modes. Mode 1 explained 98.2±1.8% of variation in position over time, so we used the position along mode 1 as an approximation of movement of the internal magnet. Similar analyses were performed on the position of the TA arm, where mode 1 explained 98.6±1.3% of the variation.

Force recordings from the Texture Analyzer were synchronized with recordings from the videofluoroscopy by comparing the TA arm position (as measured by the Texture

Analyzer) with the position of the arm along mode 1 (according to the dimensionally- reduced videofluoroscope analysis). We used a cross-correlation algorithm in MATLAB to correlate timing of these two waves and synchronize the TA and fluoroscopy data sets.

X-ray Verification of Magnetic Localization

Six 600-800g, male, albino Sprague-Dawley rats underwent localization of a model magnetic pill for a period of 12 hours. All rats had access to food and water ad libitum

within their acrylic restraint tubes and were handled in accordance with NIH and IACUC

181 guidelines. X-rays were taken prior to the start of and after 12 hours of magnetic

localization to test the efficacy of magnetic capture. All subjects showed magnetic

intestinal retention for 12 hours.

in vitro Magnetic Force Testing

A cylindrical NIB magnet identical to the orally dosed magnets was affixed by

cyanoacrylate glue to a non-magnetic aluminum pedestal ( Φ=1.6mm, length=1.6mm, KJ

Magnetics, Jamison, PA). The cylindrical external NIB magnet ( Φ=25mm, length=25mm, KJ Magnetics, Jamison, PA) was then brought towards the immobilized magnet while monitoring inter-magnetic force and separation distance. The in vitro force

as a function of distance curve was compared with the in vivo experiments, in which the inter-magnetic distance is calculated from tracking the location of the internal and external magnets from biplanar fluoroscopic videos. Inter-magnetic force as a function of distance was shown to be negligibly different between the in vitro and in vivo cases.

Therefore in vitro inter-magnetic force as a function of magnet separation testing can be

used to predict if a pair of internal and external magnets will retain an orally administered

magnetic pill given the dimensions of the subject and an estimate of the local, propulsive

GI forces experienced during digestion.

Histology

Intestinal tissue samples from 3 rats were recovered post mortem in the region of

magnetic retention and 2cm distal to the region of magnetic retention. Sections were

fixed in paraformaldehyde, imbedded in paraffin, sectioned, and stained with

182 hemotoxylin and eosin. Sections were imaged on a Zeiss Axiovert 200M (Oberkochen,

Germany) motorized inverted microscope equipped with an AxioCam MRc5 color camera (Zeiss, Oberkochen, Germany). Intestinal tissue at the site of localization showed no difference in mechanical integrity or signs of inflammation from distal control samples indicating that the method of magnetic retention has no immediate untoward effects on the intestinal tissue under the reported testing conditions.

183

Figure 7.1: Biplanar videofluorscopic tracking of magnetically retained model pills in vivo . (a) Schematic of setup for retaining magnetic pills within rat intestines visualized by biplanar fluoroscopy. (b-c) Still images acquired from biplanar fluoroscopic videos showing magnetic model pill retention due to the cyclic application of an external magnetic force. Insets highlight magnified views of the magnetic model pill, a cylindrical NIB magnet with an iron-loaded alginate bead on either end, localized in the small intestines. (d) Exemplary 3D trajectory plot of a model magnetic pill moving in response to a single force cycle of the external magnet. Arrows point along the trajectory in the direction of increasing time. (e) Exemplary inter-magnetic force plotted as a function of travel along mode 1. Slopes of the best fit lines to the ascending, descending, and whole cycle of the internal magnet are the effective elastic constants of the intestinal tissue in response to force cycling. (f) Tissue elastic constant plotted as a function of time after the start of magnetic retention (N=5) showing that there is no significant change (P asc =0.52, Pdes =0.68, P wc =0.48) in tissue elasticity during 12 hours of magnetic retention, indicating negligible change in the intestinal tissue mechanical integrity. Error bars represent s.e.m.

184

Figure 7.2: Confirmation of magnetic capture by x-ray, of in vitro force measurement in vivo, and of the force exerted by the internal magnet on underlying tissue. (a) X-ray confirmation of magnetic pill retention in the small intestines of rats (N=6) demonstrating the efficacy of magnetic retention. Orally administered magnets are circled in white. (b) Plot of inter-magnetic force as a function of inter-magnetic distance in vitro and in vivo showing minimal differences enabling the accurate prediction of magnetic pill capture in vitro for use in choosing the appropriate magnets for achieving localized drug delivery in any species. (c) Comparison of median peak inter-magnetic, net inertial, and tissue forces (N=5) showing that a negligible fraction (0.0005±0.0005%) of inter-magnetic force translates into net inertial force demonstrating that the measured inter-magnetic force is a good approximation of the force imparted by the internal magnet upon the underlying intestinal tissue. Error bars represent s.e.m.

185

Figure 7.3: Photograph of the magnetic oral dosage next to a U.S. quarter for size comparison.

186

Rat 1 Rat 2 Rat 3

Control

Experimental

Figure 7.4: Bright field micrographs of hemotoxylin and eosin stained segments of the small intestines in the region of magnetic localization for 12 hours (Experimental) and distal to that region (Control). All images were acquired at 100x magnification. There is no observable difference among the intestinal sections demonstrating that magnetic retention does not cause inflammation or necrosis.

187 7.5 References

1. Davis, S.S. Formulation strategies for absorption windows. Drug Discovery Today

10, 249-57 (2005).

2. Chen, H.M., Langer, R. Magnetically-responsive polymerized liposomes as potential

oral delivery vehicles. Pharm. Res. 14, 537-540 (1997).

3. Goldberg, M., Gomez-Orellana, I. Challenges for the oral delivery of macromolecules

Nat. Rev. Drug Discovery 2, 289-95 (2003).

4. Langer, R. Drug delivery and targeting. Nature 392, 5-10 Suppl. S (1998).

5. Mathiowitz, E. et. al. Biologically erodable microsphere as potential oral drug

delivery system. Nature 386, 410-14 (1997).

6. Whitehead, K., Shen, Z.C., Mitragotri, S. Oral delivery of macromolecules using

intestinal patches: applications for insulin delivery. J. Controlled Release 98, 37-45

(2004).

7. Arruebo, M., Fernandez-Pacheco, R., Ibarra, M.R., Santamaria, J. Magnetic

nanoparticles for drug delivery. Nano Today 2, 22-32 (2007).

8. Groning, R., Berntgen, M., Georgarakis, M. Acyclovir serum concentrations

following peroral administration of magnetic depot tablets and the influence of

extracorporal magnets to control gastrointestinal transit. Eur. J. Pharm. Biopharm.

46, 285-91 (1998).

9. Ito, R., Machida, Y., Sannan, T., Nagai, T. Magnetic Granules – A Novel System for

Specific Drug Delviery to Esophageal Mucosa in Oral-Administration. Int. J. Pharm.

61, 109-17 (1990).

188 10. Polyak, B., Friedman, G. Magnetic targeting for site-specific drug delivery:

applications and clinical potential. Expert Opin. Drug Delivery 6, 53-70 (2009).

11. Teply, B.A. et al. The use of charge-coupled polymeric microparticles and

micromagnets for modulating the bioavailability of orally delivered macromolecules.

Biomaterials 29, 1216-1223 (2008).

12. Widder, K.J. et. al. Tumor Remission in Yoshida Sarcoma-Bearing Rats by Selective

Targeting of Magnetic Albumin Microspheres Containing Doxorubucin. Proc. Natl.

Acad. Sci. USA 78, 579-81 (1981).

13. Edelman, E.R., Mathiowitz, E., Langer, R., Klagsbrun, M. Controlled Release of

Basic Fibroblast Growth-Factor. Biomaterials 12, 619-26 (1991).

14. Brainerd, E.L. et al. X-ray Reconstruction of Moving Morphology (XROMM):

Precision, Accuracy and Applications in Comparative Biomechanics Research.

Journal of Experimental Zoology A , 313A (2010).

15. Hedrick, T.L. Software Techniques for Two- and Three-Dimensional Kinematic

Measurements of Biological and Biomimetic Systems. Bioinspir. Biomim. 3, 6

(2008).

16. Riskin, D.K., et al. Quantifying the complexity of bat wing kinematics. J. Theor. Biol.

254, 604-615 (2008).

17. Guignet, R. et al. Magnet Tracking: A New Tool for in vivo Studies of the Rat

Gastrointestinal Motility. Neurogastro. and Mot. 18, 472-478 (2006).

18. Laulicht, B. et. al. Understanding Gastric Forces Calculated from High Resolution

Pill Tracking. Proc. Natl. Acad. Sci. USA in press (2010) .

189 19. Ferens, D.M. et al. A Quantitative Approach to Recording Peristaltic Activity From

Segments of Rat Small Intestine in vivo. Neurogastro. and Mot. 17, 714-20 (2005).

190

Chapter 8

Conclusions and Future Directions

In summary, continuous flow phase inversion nanoencapsulation is a novel, controllable method for producing size-monodisperse polymer nanospheres in the optimal range for uptake-based oral drug delivery, 500nm. The linear dependence between flow rate and mean nanosphere diameter indicates a time-course-dependent mechanism of phase inversion that supports nucleation and growth. Additionally, the particle size dependence on flow rate provides a useful parameter for adjusting the size distribution of a population of polymer nanospheres within limits, while keeping all other production conditions constant (Chapter 2) [1]. Additionally, phase inversion micronization of the hydrophobic drug, furosemide, and its mixture with the stock formulation can lead to a greatly improved bioactivity profile (Chapter 3). Phase inversion is a very useful process for not only encapsulating sensitive therapeutics in thermoplastic polymers for oral administration, but also for micronizing drug formulations themselves. In the future, I would like to investigate the scale-up potential of continuous flow phase inversion nanoencapsulation and to test phase inversion micronized formulations of furosemide on a large animal model to see if the improved bioactivity profile is maintained.

191

Our in vivo investigation of bioadhesion in the gastric environment indicates that

bioadhesion in the stomach is confounded by a thick, loosely adherent mucus layer the

strength of which dominates bioadhesive forces in a rat model (Chapter 4) [2]. The

inability of bioadhesives to yield significantly improved bioadhesive forces over standard

bioerodible polymers evidences the need for methods of gastric retention that do not rely

on bioadhesion alone. Towards that end, magnet-based localization could be of great use

for prolonging gastric residence time. Moreover, the in vitro results indicate that with sufficient contact force bioadhesives can penetrate the loosely adherent gastric mucus enabling classically strong bioadhesives to outperform weak bioadhesives. Therefore, coating magnetic oral doses in bioadhesive polymers is a promising future direction for the field of gastric retention.

Bioinspired polymers with DOPA grafted onto hydrophobic backbones, most notably poly(butadiene-co-maleic anhydride-graft-DOPA) (PBMAD) and poly(ethylene-co- maleic anhydride-graft-DOPA) (PEMAD), possess the strongest bioadhesive properties measured in our laboratory to date, yielding 2.5-3x the fracture strength and tensile work of the commercially available acrylic acid-based bioadhesive, Polycarbophil (Chapter 5).

When weakly bioadhesive non-erodible, 500nm polymer nanospheres were encapsulated in PBMAD, their uptake in the jejunum of rats was increased by a factor of 11.5x.

Combining strong bioadhesives with appropriately sized nanospheres has tremendous potential to enhance the uptake of nanoencapsulated therapeutics (e.g. proteins) bringing it into a clinically effective range. Designing a continuous flow phase inversion

192 nanoencapsulation system to create polymer nanospheres suspended in a bioadhesive

polymer continuous phase that feeds into a spray drier is a very promising way to create

bioadhesive nanospheres for oral drug delivery.

Given that the location of administration within the gastrointestinal tract is of great

importance to the oral delivery of poorly water soluble, poorly absorbed therapeutics

using bioadhesive nanospheres, creating a system that enables non-invasive prolonged

localization of oral doses is of great utility. Before creating a novel method of achieving

prolonged localization, quantifying the net force experienced by pills in the

gastrointestinal tract calculated from high resolution tracking data was essential (Chapter

6) [3]. The net force model was successful in quantifying gastric emptying forces in rats,

dogs, and humans and could readily be expanded to any high resolution pill tracking data.

In the future, I would like to expand the force based study from the gastric environment

to the entire gastrointestinal tract. Another potential future direction for the research is to

coat pills in bioadhesive polymers and quantitatively assess how the adhesion changes the

force profile as a new method of quantitative in vivo bioadhesion measurements. In addition to quantifying the gastrointestinal forces of various pills, there may be clinical utility to assessing the force profiles of the same pills in patients exhibiting gastrointestinal pathophysiologies. Comparing baseline force measurements in healthy and pathophysiological patients may elucidate differences in net forces experienced by the pill that could ultimately be used as a non-invasive diagnostic for diseases such as inflammatory bowel disease and gastrointestinal dysmotility.

193 Quantifying the net forces experienced by model magnetic pills was also used as a guideline in creating a safe and effective means of retaining pills within specific gastrointestinal locations (Chapter 7). The creation of a system for monitoring the inter- magnetic force and location of an orally administered magnetic model pill in real time using a combination of a modified materials testing device and biplanar videofluroscopy presents a platform for safely and effectively capturing oral doses within the gastrointestinal tract for up to 12 hours. The safety of prolonged magnetic retention was assessed by monitoring the effective spring constant of the intestinal tissue at specified time points throughout testing, as well as by histology. The efficacy of magnetic capture was significantly improved over previous methodologies made possible by the real time force monitoring. Maintaining inter-magnetic force within a desired range has demonstrated safe and effective capture of model magnetic pills.

The novel method of localizing pills using constantly monitored inter-magnetic attractive forces provides a platform for testing the administration site specific properties of many therapeutics including bioadhesive nanosphere encapsulated proteins and genes, as well as enabling the study of localized administration of therapeutics directly at the site of gastrointestinal pathophysiologies such as inflammatory bowel disease and gastrointestinal cancers. A promising future direction is to investigate the effects of retaining orally administered controlled release chemotherapeutic formulations at the sites of tumors in a small animal colon cancer model. Prolonged, direct application of chemotherapeutics may achieve or even outperform the effectiveness of systemically administered chemotherapeutics with reduced systemic side effects for colon cancer

194 patients. As colon cancer is the second deadliest form of cancer in the United States, improvements in treatment would greatly impact the quality of life of millions of

Americans.

195 8.1 References

1. Laulicht, B; Cheifetz, P; Mathiowitz, E; Tripathi, A. 2008. Evaluation of continuous flow nanosphere formation by controlled microfluidic transport. Langmuir 24 (17): 9717-

9726.

2. Laulicht, B; Cheifetz, P; Tripathi, A; Mathiowitz, E. 2009. Are in vivo gastric

bioadhesive forces accurately reflected by in vitro experiments? Journal of Controlled

Release 134 (2): 103-110.

3. Laulicht, B; Tripathi, A; Shlageter, V; Kucera, P; Mathiowitz, E. April 19, 2010.

Understanding Gastric Forces Calculated from High Resolution Pill Tracking. PNAS

10.1073/pnas.1002292107.

196