A Thesis

entitled

Ionically Crosslinked Chitosan Nanocarriers

by

Yuhang Cai

Submitted to the Graduate Faculty as partial fulfillment of the requirements for the

Doctor of Philosophy Degree in

Chemical Engineering

______Dr. Yakov Lapitsky, Committee Chair

______Dr. Constance A. Schall, Committee Member

______Dr. Sasidhar Varanasi, Committee Member

______Dr. Matthew W. Liberatore Committee Member

______Dr. Eda Yildirim-Ayan, Committee Member

______Dr. Amanda Bryant-Friedrich, Dean College of Graduate Studies

The University of Toledo

August 2017

Copyright 2017, Yuhang Cai

This document is copyrighted material. Under copyright law, no parts of this document may be reproduced without the expressed permission of the author. An Abstract of

Ionically Crosslinked Chitosan Nanocarriers

by

Yuhang Cai

Submitted to the Graduate Faculty as partial fulfillment of the requirements for the Doctor of Philosophy Degree in Chemical Engineering The University of Toledo

August 2017

Ionically crosslinked chitosan nanocarriers have attracted significant attention as potential drug delivery vehicles due to their biocompatibility, mucoadhesiveness, payload protection ability, and mild formation/payload encapsulation procedures. Despite these advantages, however, most studies on these materials have tuned their drug uptake and release properties by trial and error, and not infrequently reported conflicting results. This dissertation aimed to address some of these issues.

To better understand drug uptake properties, we have shown that, besides increasing with the drug/particle binding affinity, protein drug association efficiency (i.e., the fraction of the added protein that was taken up) increased almost linearly with the particle yield (the fraction of the added chitosan that self-assembled into particles). This scaling was explained via a predictive equilibrium binding model and suggested that many of the (often conflicting) variations in protein uptake reported in the literature might stem from the largely ignored variability in particle yield.

iii Because sustained drug release could be affected by particle dissolution stability, ionically crosslinked chitosan particle dissolution was also examined. This revealed hysteresis in the ionic crosslink formation/dissociation cycle (where particle dissolution occurred at lower ionic crosslinker concentrations than those required for particle formation). Also explored was whether drug/particle binding (where the drug molecules served as additional physical crosslinks between the chitosan chains) enhanced the dissolution stability of chitosan/tripolyphosphate (TPP) particles. This indicated that, while protein/chitosan binding was insufficiently strong to generate a stabilizing effect, chitosan/TPP particles could be stabilized against dissolution through the uptake of DNA.

Further, it has long been ignored that the in vitro drug release profiles obtained for chitosan particles via the common “solvent replacement” method may have been subject to strong experimental artifacts and not have reflected their true release behavior. To this end, we have explored the relevant experimental artifacts and showed that conflicting findings on drug release from chitosan/TPP particles can result from: (1) incomplete particle separation from the solvent; (2) irreversible particle coagulation; and (3) failure to maintain sink conditions. By analyzing these artifacts, this study provides guidelines for obtaining more-reliable release profiles for both chitosan/TPP particles and other colloidal drug carriers.

iv

Acknowledgements

My five years at University of Toledo like a rocket that was flying into the sky.

When it was just launched, it went slow but with powerful airflow, I have to try my best to prevent myself from blowing away by new things. The rocket speeded up until I cannot see it anymore, I know that my journey as a Ph.D. student is going to come to an end. The wonderful scenery it left for me, was fruitful and precious, and I have many people that I would like to acknowledge.

First and foremost, I thank my advisor, Professor Yakov Lapitsky, for the guidance with great patience. His enthusiasm in doing research inspired me with the logical thinking and troubleshooting, and I feel that he want to teach all knowledge to his students without hesitating. I was truly lucky to have him as a teacher during this critical period of my research career.

I want to thank all my dissertation committee members, Dr. Constance A. Schall,

Dr. Sasidhar Varanasi, Dr. Matthew W. Liberatore and Dr. Eda Yildirim-Ayan, that have given me insightful advises on my research and dissertation, and encourage me to think deeper about my research. I also would like to greatly thank Dr. Yan Huang, the first

Ph.D. student in our group. He taught me the operation of equipment in the lab, and

v greatly helped me with my live when I first come to U.S.A. Besides Yan, I want to thank all other the graduates and undergraduates worked in the lab who provided many helpful discussions.

Last but not the least, I acknowledge my parents and my wife, with their all support in love. I would like to express my deepest gratitude to my family in the language they can read:

感谢亲爱的爸爸妈妈，你们多年来对我所有决定的无条件支持，使我能够一

直做我想做的事，也让我感受到了无与伦比的爱与自由。在异国他乡读博的五年时

光里，我时常想象自己在家时候的感觉，却又无法时常陪伴在你们的身边。终于，

我博士毕业了，我希望能够尽快的找到一份好工作，可以使我有能力更好的孝敬你

们，让你们过上更好的日子。在我不在身边的日子里，你们一定要照顾好自己！同

样感谢我的老婆，在我博士最后一年半中一直陪伴照顾我，让我的生活充满了希望！

我爱你们！

vi

Table of Contents

Abstract ...... iii

Acknowledgements ...... v

Table of Contents ...... vii

List of Tables ...... xiv

List of Figures ...... xv

List of Abbreviations ...... xxviii

List of Symbols ...... xxix

1 Introduction and Objectives ...... 1

1.1. Introduction ...... 1

1.2. Objectives and Overview ...... 7

2 Background ...... 9

2.1. Introduction ...... 9

2.2. Preparation of Ionically Crosslinked Chitosan Particles ...... 9

2.3. Drug Loading into Ionically Crosslinked Chitosan Particles ...... 11

2.3.1. Effects of Drug Uptake Media ...... 11 vii 2.3.1.1. pH Effects ...... 11

2.3.1.2. Ionic Strength Effects ...... 12

2.3.2. Effects of Molecular Properties ...... 12

2.3.3. Effects of Composition ...... 13

2.4. Stability of Ionically Crosslinked Chitosan Particles ...... 14

2.4.1. Factors Affecting Particle Stability ...... 15

2.4.2. Methods of Increasing Particle Stability ...... 17

2.4.2.1. Covalent Crosslinking ...... 17

2.4.2.2. Further Ionic Crosslinking ...... 18

2.4.2.3. Other Methods ...... 20

2.5. Drug Release from Ionically Crosslinked Chitosan Particles ...... 21

2.5.1. Release Mechanisms ...... 23

2.5.2. Release Models ...... 25

2.5.2.1. Zero-Order Model ...... 25

2.5.2.2. First-Order Model ...... 27

2.5.2.3. Hixson-Crowell Model ...... 28

2.5.2.4. Hopfenberg Model ...... 28

2.5.2.5. Higuchi Model ...... 29

2.5.2.6. Korsmeyer-Peppas Model ...... 30

2.5.2.7. Weibull Model ...... 30

viii 2.5.2.8. Model Application to Ionically Crosslinked Chitosan Micro- and

Nanoparticles ...... 31

2.5.3. Methods for Conducting Drug Release Experiments and Their Potential

Pitfalls ...... 32

2.5.3.1. Solvent Replacement Method ...... 33

2.5.3.2. Dialysis Method ...... 34

2.5.3.3. In Situ Methods ...... 36

2.5.3.4. Conducting Drug Release Experiments to Ionically Crosslinked

Chitosan Micro- and Nanoparticles Methods ...... 36

3 Factors that Affect Drug-Uptake into Submicron Chitosan/TPP Particles ...... 38

3.1. Introduction ...... 38

3.2. Materials and Methods ...... 40

3.2.1. Materials ...... 40

3.2.2. Preparation and Analysis of FITC-Labeled Chitosan ...... 41

3.2.3. pH, Ionic Strength and TPP:Glucosamine Ratio Effects on Micro- and

Nanogel Yield ...... 42

3.2.4. Protein Uptake Effects on Micro- and Nanogel Yield ...... 43

3.2.5. Formulation and Uptake Method Effects on AE ...... 44

3.2.6. DLS and TEM Analysis of Protein/Dissolved Chitosan Binding ...... 45

3.3 Results and Discussion ...... 46

3.3.1 TPP:Glucosamine Ratio and pH Effects on the Micro- and Nanogel Yield 46

ix 3.3.2. Ionic Strength Effect on the Micro- and Nanogel Yield ...... 49

3.3.3. Protein Uptake Effects on the Chitosan/TPP Micro- and Nanogel Yield ... 51

3.3.4. Protein Concentration, Uptake Method and pH Effects on AE ...... 54

3.3.5. Relationship between AE and XAgg ...... 57

3.3.6. Model Analysis of Protein Uptake into Chitosan/TPP Particles ...... 60

3.3.7. Further Discussion ...... 70

3.4. Conclusions ...... 73

4 Formation and Dissolution of Ionically Crosslinked Chitosan Particles: Analysis of

Ionic Crosslink Reversibility ...... 76

4.1. Introduction ...... 76

4.2. Materials and Methods ...... 78

4.3. Results and Discussion ...... 80

4.3.1. Formation and Characterization of Chitosan/PPi Particles ...... 80

4.3.1.1. Formation of Chitosan/PPi Particles ...... 80

4.3.1.2. Particle Size and Composition ...... 82

4.3.1.3. Particle Colloidal Stability ...... 85

4.3.1.4. Particle Formation Kinetics ...... 86

4.3.2. Dissolution of Chitosan/PPi Particles ...... 88

4.3.3. Model Analysis of the Hysteresis Loops ...... 91

4.4. Conclusion ...... 98

x 5 Factors that Affect Drug-Loaded Submicron Chitosan/TPP Particle Stability... 100

5.1. Introduction ...... 100

5.2. Materials and Methods ...... 102

5.2.1. Materials ...... 102

5.2.2. Preparation of FITC-Chitosan/TPP Micro- and Nanoparticles ...... 102

5.2.3. Preparation of Drug-Loaded FITC-Chitosan/TPP Micro- and Nanoparticles

...... 103

5.2.4. Analysis of FITC-Chitosan/TPP Micro- and Nanoparticle Stability ...... 104

5.3. Protein Effect on the Chitosan/TPP Micro- and Nanoparticle Stability ...... 106

5.4. DNA Effect on the Chitosan/TPP Micro- and Nanoparticle Stability ...... 114

5.5. The Effect of Drug Loading Procedure on Particle Stability ...... 120

5.6. Conclusion ...... 121

6 Factors That Affect Drug Release from the Submicron Chitosan/TPP Particles 123

6.1. Introduction ...... 123

6.2. Materials and Methods ...... 126

6.2.1. Materials ...... 126

6.2.2. Preparation of FITC-labeled Chitosan ...... 127

6.2.3. Preparation of BSA-Loaded Chitosan/TPP Micro- and Nanoparticles .... 127

6.2.4. Particle Recovery via Centrifugation ...... 128

6.2.5. Drug Release from Gel Pellets ...... 129

xi 6.2.6. Drug Release from Micro- and Nanoparticles ...... 131

6.3 Results and Discussion ...... 133

6.3.1. Centrifugation Procedure Effects on Particle Recovery ...... 133

6.3.2. Coagulation Effects and Release Properties of Coagulated Particles ...... 139

6.3.2.1. Centrifugal Force Effect on the Release Profile ...... 141

6.3.2.2. Centrifugation Time Effect on the Release Profile ...... 142

6.3.2.3. Further Analysis of Protein Transport within the Chitosan/TPP Gel

Pellets ...... 143

6.3.3. Release from Dispersed Chitosan/TPP Particles: Ionic Strength and Solvent

Replacement Frequently Effects ...... 149

6.4. Conclusions ...... 156

7 Conclusions and Recommendations ...... 157

7.1. Conclusions ...... 157

7.2. Recommendations ...... 159

7.2.1 Further Studies on Drug Uptake into Ionically Crosslinked Particles ...... 160

7.2.2 Enhancing DNA Loading into Ionically Crosslinked Chitosan Particles .. 161

7.2.3 Improving the Solvent Replacement Method for Investigating Drug Release

from Colloids ...... 161

References ...... 163

Appendix A ...... 191

xii Appendix B ...... 205

B.1. Evolution in Particle ζ-Potential ...... 205

Appendix C ...... 206

C.1. The α-LA-Loaded Chitosan/TPP Micro- and Nanoparticle Stability ...... 206

Appendix D ...... 208

D.1. Chitosan/TPP Particle Effect on Protein Concentration Determination ...... 208

D.2. Determination of BSA-Loaded Chitosan/TPP Gel Pellet Density...... 209

D.3. Centrifugation Effect on Particle Size Distributions in the Presence of a

Glycerol Bed ...... 209

D.4. Centrifugation Procedure Effects on the BSA Recovery with FITC-

Chitosan/TPP Particles...... 211

D.5. Change in Chitosan/TPP Gel Pellet Mass during Release ...... 211

D.6. Effects of Non-Uniform Initial Protein Distributions within Chitosan/TPP Gel

Pellets ...... 213

D.7. Chitosan/TPP Gel Pellet Size Effect on the BSA Release ...... 216

xiii

List of Tables

3.1. Fitted equilibrium constants for BSA and α-LA binding to chitosan/TPP particles and soluble chitosan ...... 66

4.1. Chitosan/PPi dispersions analyzed by capillary viscometry. The final chitosan concentrations reflect the dilution of parent chitosan solutions during the titration of PPi.

The uncertainty in the η/η0-values represents the 95% confidence limits based on two replicate samples (each of which was measured thrice) ...... 85

6-1. Variations in the gel pellet thickness, BSA recovery and apparent BSA diffusivity within the gel with the centrifugal force (average ± standard deviation) ...... 142

6.2. Variations in the gel pellet thickness, BSA recovery and apparent BSA diffusivity within the gel with the centrifugation time (average ± standard deviation) ...... 143

6.3. The R2 (from Equation 6-8) and n-value (from Equation 2-6) of fitting the normalized release profile with model equations with a variation of centrifugation condition ...... 147

D.1. The n-value (obtained from Equation 2-6) fitted to the normalized release profiles ~ (from Equation 6-7) at variable Csurf values. In each case the initial protein concentration profiles within the pellets were assumed to be linear ...... 216

xiv

List of Figures

1-1 Scheme of ionically crosslinked polyelectrolyte nanoparticle formation and drug uptake processes ...... 2

1-2 Chemical structures of (a) TPP, (b) PPi and (c) chitosan ...... 3

2-1 Identification of compositions where chitosan/TPP nanoparticles form showing: (a) a coarse state diagram; and (b) a state diagram focusing on the region of particle formation. The figures were reproduced from Calvo et al. [1] with permission from John

Wiley & Sons ...... 10

2-2 Chitosan/TPP micro- and nanogels can be unstable due to their: (a) aggregation and subsequent precipitation, and (b) dissolution. These schemes were reproduced from

Huang et al. [2] with permission from the Royal Society of Chemistry ...... 15

2-3 TEM images of BSA-loaded chitosan/TPP nanoparticles taken during protein release in pH 7.0 PBS at 37 °C after (a) 3, (b) 6, (c) 12, (d) 24 and (e) 48 h of release time.

The figure was reproduced from Gan et al. [3] with permission from Elsevier ...... 19

2-4 Chemical structure of GM. The figure was reproduced from Alonso-Sande et al.

[4] with permission from the American Chemical Society ...... 20

2-5 Comparison of drug delivery though periodic dosage and controlled release. This figure was reproduced from Uhrich et al. [72] with permission from American Chemical

Society...... 22

xv 2-6 Scheme of the various drug release mechanism ...... 24

2-7 Representative scheme of zero, first-order release and t1/2 type release (where t is the release time). This figure was reproduced from Cussler [5] with permission from

Cambridge University Press ...... 26

2-8 Some methods can reach zero-order release: (a) using a permeable capsule containing saturated drug solution, (b) an impermeable capsule containing unsaturated drug solution with a swellable hole (c) an altered initial concentration profile in the solid carrier and (d) osmotic pumps. This figure was reproduced from Cussler [5] with permission from Cambridge University Press ...... 27

2-9 Scheme of the dialysis method of measuring drug release (adapted from

Washington [6] with permission from Elsevier ...... 35

3-1 Comparison of XAgg versus TPP:glucosamine molar ratio curves obtained by titrating TPP solutions at (■) pH 3.0, (●) pH 4.0 and (▲) pH 5.5 into 0.1 wt% chitosan solutions with matching pH-levels, and (▼) TPP solutions at their natural pH (pH 9.6) into 0.1 wt% chitosan solutions at pH 4.0. The error bars are standard deviations and the lines are guides to the eye ...... 47

3-2 Comparison of XAgg versus TPP:glucosamine molar ratio curves obtained from (a) pH 4.0 and (b) pH 5.5 chitosan and TPP solutions containing (■) 0 mM, (●) 150 mM and

(▲) 1000 mM NaCl. The error bars are standard deviations and the lines are guides to the eye ...... 50

3-3 Comparison of XAgg versus TPP:glucosamine molar ratio curves obtained: (a) at pH 5.5 in the presence of (■) 0, (●) 0.05, (▲) 0.10 and (▼) 0.15 wt% BSA loaded using the incubation method; and (b) at pH 6.0 in the presence of (■) 0, (●) 0.03, (▲) 0.06 wt%

xvi and (▼) 0.09 wt% α-LA loaded using the incubation method; (c) comparison of

XAgg-values obtained in the presence of BSA using the two uptake methods (incubation versus incorporation); and (d) comparison of XAgg versus TPP:glucosamine molar ratio curves obtained in the presence of (■) 0, (●) 0.05 and (▲) 0.15 wt% BSA at pH 4.0 and

(▼) 0, () 0.05 and () 0.15 wt% BSA at pH 5.5. The error bars are standard deviations and the lines are guides to the eye ...... 52

3-4 AE dependence on the TPP:glucosamine molar ratio for dispersions at: (a) pH 5.5 loaded with (■) 0.05, (●) 0.10 and (▲) 0.15 wt% BSA by the incubation method; and (b) pH 6.0 loaded with (■) 0.03, (●) 0.06, (▲) 0.09 wt% α-LA by the incubation method; (c) comparison of AE-values for BSA obtained using the two uptake methods (incubation and incorporation); and (d) BSA AE dependence on the TPP:glucosamine molar ratio at

(■) pH 4.0, (●) pH 5.5. The error bars are standard deviations and the lines are guides to the eye ...... 55

3-5 The scaling of AE with XAgg for particles loaded with (■) 0.05, (●) 0.10 and (▲)

0.15 wt% BSA via the (a) incubation and (b) incorporation method; and (c) for particles loaded with (■) 0.03, (●) 0.06 and (▲) 0.09 wt% α-LA via the incubation method. The error bars are standard deviations and the lines are linear regression fits (to data at all three protein concentrations) and the error bars are standard deviations ...... 59

3-6 Binding isotherm data showing: (a) BSA uptake into particles with

TPP:glucosamine molar ratios of (■) 0.026, (●) 0.053, (▲) 0.079, (▼) 0.106 and ()

0.132:1; (b) α-LA uptake into particles with TPP:glucosamine molar ratios of (■) 0.021,

(●) 0.042, (▲) 0.063 and (▼)0.084:1; and Keff versus (1 - XAgg) plots for (c) BSA and (d)

α-LA. In (a, b) the dashed lines are linear fits of each data set to Equation 3.3, while the

xvii solid lines are the theoretical predictions of all data sets based on Equations 3.3 and 3.6

~ (and fitted parameters K and K ). The lines in (c, d) are linear fits experimental data to

Equation 3.6. All error bars are standard deviations ...... 62

3-7 Representative volume-weighted size distributions obtained by DLS for (a) 0.15 wt% BSA and (b) 0.09 wt% α-LA: (■) protein, (●) 0.063 wt% chitosan and (▲) mixtures of protein with 0.063 wt% chitosan after ultracentrifugation ...... 64

3-8 The scaling AE with XAgg for particles prepared using (■) 0.31, (●) 0.63 and (▲)

0.94 mg/mL overall chitosan concentrations and loaded via the incubation method with (a)

0.05 wt% BSA and (b) 0.03 wt% α-LA. The error bars are standard deviations and the lines are the model predictions based on Equation 3.9 for mixtures containing (—) 0.31,

(‒ ‒) 0.63 and (····) 0.94 mg/mL chitosan, respectively ...... 69

3-9 Plots of protein uptake as function of total protein used during the encapsulation procedure for: (a) BSA loaded at (■) 0.026, (●) 0.053, (▲) 0.079, () 0.106 and (▼)

0.132:1 TPP:glucosamine molar ratios; and (b) α-LA loaded at (■) 0.021, (●) 0.042, (▲)

0.063 and ()0.084:1 TPP:glucosamine molar ratios. Both plots are obtained based on data in Figure 3-6. The lines are model predictions obtained using Equation 3.10 and the error bars are standard deviation ...... 70

4-1 The onset of chitosan/PPi particle formation illustrated by: (a) the evolutions in normalized light scattering intensity obtained during PPi titrations into () 0.03 wt% ()

0.10 wt% and () 0.20 wt% chitosan solutions; and (b) a phase map illustrating the transitions from (S) molecular solutions to (D) colloidal dispersions that occurred () during the PPi titrations and () after long-term equilibration. The lines are guides to the eye and the error bars indicate the uncertainty in the phase boundary ...... 81

xviii 4-2 The effect of the parent chitosan solution concentration on: (a) the z-average hydrodynamic diameter evolution during the titration of PPi into () 0.01 wt%, () 0.03 wt%, () 0.05 wt%, () 0.10 wt%, and () 0.20 wt% chitosan solutions (detected by

DLS); (b) the size distributions of particles prepared using (i) 0.03 wt% and (ii) 0.1wt% chitosan solutions (imaged by STEM; scale bar = 600 nm); (c) the () average particle volume and () particle volume fraction. The solid lines are power law fits while the dashed lines are guides to the eye. The drawing (d) shows the proposed structure of a solvent-swollen chitosan/PPi nanoparticle ...... 83

4-3 Evolution in the light scattering intensity analyzed via (a) DLS, where the derived light scattering intensity is plotted versus PPi concentration, and (b) stopped-flow turbidimetry, where the evolution in the turbidity is detected in 0.08 wt% (4.6 mM) chitosan solutions containing () 1.2, () 1.5, () 1.7 and () 2.3 mM PPi. The lines are guides to the eye ...... 87

4-4 Dilution effects on the light scattering intensity from preformed chitosan/PPi particles showing: (a) the derived count rate measured at various times after dilution plotted versus the PPi concentration; and (b) the light scattering intensity (normalized to that predicted by Beer’s Law) from particles diluted to PPi concentrations of () 0.2 mM,

() 0.5 mM, () 0.8 mM and () 1.1 mM plotted versus time. The solid lines are guides to the eye and the dashed lines indicate the hypothetical light scattering intensity in the absence of particle dissolution (or further aggregation). The error bars are standard deviations (n = 3) ...... 89

4-5 Hysteresis in the light scattering intensity obtained via () forward titration and

() backward dilution after (a) 10 min and (b) 2 weeks of equilibration (from the time of

xix dilution). The error bars are standard deviations (n = 3) and the lines are guides to the eye

...... 90

4-6 Effects of εc on the binding showing: (a) the dependence of μA on θ for various εc

-values (the shaded region indicates the spinodal envelope where the binding is unstable); and (b) model binding isotherms plotted for cases of weakly-cooperative (εc = -2 kBT) and strongly-cooperative (εc = -8kBT) binding. The solid binding isotherms indicate equilibrium binding, whereas the dashed and dotted curves indicate regions of (----) pseudo-equilibrium and (······) unstable binding. Points A, B, C and D are defined in the text 94

4-7 Model hysteresis loops for εc = -8kBT and KCP = 0.1 showing (a) the binding isotherm and (b) the fraction of the crosslinked polymer chains plotted versus the normalized total PPi concentration. The solid isotherms indicate equilibrium binding, whereas the dashed and dotted curves indicate regions of (----) pseudo-equilibrium and

(······) unstable binding. Points A, B, C and D are defined in the text ...... 95

5-1 (a) The normalized light scattering intensity from dispersions of drug-free chitosan/TPP particles; parity plots showing comparisons of normalized light scattering intensities from dispersion of particles containing (b) BSA and (c) α-LA and drug-free dispersion each diluted to 0.003, 0.013 and 0.031 wt% chitosan; (d) apparent size distributions obtained via NNLS fitting of DLS data for BSA-loaded chitosan/TPP particle dispersions (▼) before dilution (at 0.063 wt% chitosan) and after being diluted fivefold (to 0.013 wt% chitosan) in (■) salt-free water, (●) 150 mM NaCl solution and (▲)

PBS solution. The numbers in the parentheses in (b) and (c) are the chitosan concentrations (in units of wt%). The error bars are standard deviations and the dotted

xx lines are the parity lines ...... 108

5-2 Normalized FITC-chitosan UV-vis absorbance of ( ) FITC-chitosan solutions,

( ) FITC-chitosan/TPP particle dispersions and ( ) BSA-loaded FITC-chitosan/TPP particle dispersions after centrifugation in (a) NaCl-free water, (b) 150 mM NaCl solution and (c) PBS solution; and (d) normalized Micro BCA assay UV-vis absorbance of

BSA-loaded FITC-chitosan/TPP particle dispersions after centrifugation in (■) salt-free water, (●) 150 mM NaCl solution and (▲) PBS solution (where a normalized Micro BCA assay absorbance of 1.0 corresponding to complete protein release). The error bars are standard deviations and the lines are guides to the eye ...... 111

5-3 Comparison of normalized light scattering intensities for DNA-loaded and

DNA-free particles after dilution to 0.003, 0.013 and 0.031 wt% chitosan in (●) 150 mM

NaCl solution and (▲) PBS solution. The numbers in the parentheses are the chitosan concentrations (in units of wt%) The error bars are standard deviations and the dotted line is the parity line ...... 114

5-4 Normalized FITC-chitosan UV-vis absorbance of ( ) FITC-chitosan solutions,

( ) DNA/FITC-chitosan mixture, ( ) FITC-chitosan/TPP particle dispersions, and

( ) DNA-loaded FITC-chitosan/TPP particle dispersions after centrifugation in (a) 150 mM NaCl solution and (b) PBS solution; normalized DNA UV-vis absorbance obtained from (■) DNA/FITC-chitosan mixtures after centrifugation and (●) DNA-loaded

FITC-chitosan/TPP particle dispersions after centrifugation in (c) 150 mM NaCl solution and (d) PBS solution. The error bars are standard deviations and the lines are guides to the eye ...... 116

5-5 Dark field micrographs of chitosan/TPP particle dispersion prepared with (a) 2.5

xxi × 10-3 and (b) 5.0 × 10-3 wt% DNA via incorporation method, and (c) 1.3 × 10-3 and (d)

2.5 × 10-3 wt% DNA via incubation method. At the magnification used only macroscopic, fiber-liker precipitates are visible ...... 119

5-6 Normalized UV-vis absorbance signals obtained from (a) FITC-chitosan and (b)

Micro BCA assay of BSA-loaded FITC-chitosan/TPP particle dispersions prepared using the (●) incubation and (▲) incorporation drug loading methods (each obtained in pH 6.0

PBS after removing the undissolved particles through centrifugation). The error bars are standard deviations and the lines are guides to the eye ...... 121

6-1 Schematic representation of the solvent replacement method for measuring release kinetics ...... 124

6-2 Schematic representation of particle sedimentation in a centrifuge tube inside a swinging bucket centrifuge rotor ...... 134

6-3 The effects of (a, b) centrifugal force and (c, d) centrifugation time on the (a, c) chitosan recovery from (■) chitosan/TPP particles, (●) BSA-loaded chitosan/TPP particles, (▲) TPP-free chitosan and () TPP-free chitosan/BSA mixtures; and (b, d)

BSA recovery from (■) BSA-loaded chitosan/TPP particles, (●) chitosan- and TPP-free

BSA solutions and () TPP-free chitosan/BSA mixtures. The centrifugation force effects were investigated using a 30 min centrifugation time, while the centrifugation time effects were probed at a constant centrifugal force of 3.0 × 105 g. The error bars (which are largely obscured by the symbols) are standard deviations, while the lines are guides to the eye ...... 136

6-4 Model particle recovery effect on the apparent drug release profile calculated for conditions where (■) 99%, (●) 90%, (▲) 80% and (▼) 70% of the particles were

xxii recovered by centrifugation during each solvent replacement step. These model predictions were obtained under the simplifying assumption of no drug being released

(which means that the true release profile should be a flat line overlapping the abscissa)

...... 139

6-5 Photographs of (a) nanoparticle dispersion before centrifugation and (b) gel-like pellet after centrifugation ...... 140

6-6 Actual (a) and normalized (b) BSA release profiles obtained from chitosan/TPP particle pellets after 30 min of centrifugation at (■) 4.8 × 104, (●) 1.1 × 105 and (▲) 3.0 ×

105 g. The inset in (a) shows the first 3 h original BSA release profiles and the inset in (b) shows the first 60% normalized BSA release points and their model fit curves for pellets prepared by centrifugation at (black solid line) 4.8 × 104, (red dash line) 1.1 × 105 and

(blue dot line) 3.0 × 105 g using Equation 6-8. The lines in the other plots are guides to the eye. All error bars are standard deviations ...... 142

6-7 Actual (a) and normalized (b) BSA release profile obtained from chitosan/TPP particle pellets after (■) 30, (●) 60 and (▲) 90 min of centrifugation at 3.0 × 105 g. The inset in (a) shows the first 3 h original BSA release profiles and the inset in (b) shows the first 60% normalized BSA release points and their model fit curves for pellets prepared by centrifugation at (black solid line) 30, (red dash line) 60 and (blue dot line) 90 min using Equation 6-8. The lines in the other plots are a guides to the eye. All error bars are standard deviations ...... 143

6-8 Representative schematic of the chitosan/TPP gel pellet formed on the bottom of the centrifuge tube upon ultracentrifugation ...... 145

6-9 The ionic strength effect on the BSA release profile in the pH 6.0 5% trehalose

xxiii solution at with (■) 0 and (●) 150 mM NaCl. The error bars are standard deviations and the lines are guides to the eye ...... 151

6-10 BSA release into pH 6.0 5% trehalose solution, obtained using (■) 1, (●) 4, (▲)

12 and (▼) 24-h time intervals between the solvent replacement steps, plotted versus the:

(a, b) release time shown here (a) for the first day and (b) the whole experiment, (c) number of solvent replacement steps; and (d) photos showing particle coagulation after three centrifugation cycles with a glycerol bed taken (i) before and (ii) after vortexing.

The error bars are standard deviation and the lines are guides to the eye ...... 152

6-11 BSA release into 5% trehalose solution obtained (■) in this work using the 24-h time interval between the solvent replacement steps and (●) in Calvo’s work using a

2-day time interval between the solvent replacement steps. The error bars are standard deviation and the lines are guides to the eye ...... 155

A-1 Representative volume-weighted size distributions of (■) empty and (●)

BSA-loaded chitosan/TPP particles, which were loaded via incorporation using a 0.05 wt% overall BSA concentration. Each particle batch was prepared at a 0.13:1

TPP:glucosamine molar ratio from NaCl-free parent chitosan and TPP solutions at pH 5.5

...... 191

A-2 UV-Vis absorbance readings from the Bradford assay at various concentration of

(■) chitosan, (●) BSA without chitosan and (▲) BSA at a fixed chitosan concentration of

0.625 mg/mL (λ = 595 nm). The lines are guides to the eye ...... 192

A-3 pH drift that occurs with the addition of TPP: (a) during BSA uptake, where the overall BSA concentration is either (■) 0 wt%, (●) 0.05 wt% or (▲) 0.15 wt%; and (b) during α-LA uptake, where the overall α-LA concentration is either (■) 0 wt%, (●) 0.06

xxiv wt% or (▲) 0.09 wt%. The error bars are standard deviations while the lines are guides to the eye ...... 193

A-4 XAgg versus BSA concentration curves obtained at pH 5.5 and TPP:glucosamine

▼ ◆ molar ratios of ( ) 0, ( ) 0.026, (▲) 0.053, (▼) 0.079, ( ) 0.106, ( ) 0.132, ( )▼

■ ●

0.158:1 where the protein was loaded by the: (a) incubation and (b) incorporation methods. The error bars are standard deviations and the lines are guides to the eye ....194

A-5 The particle yield comparison of ( ) before and ( ) after adjusting pH back to

(a) 5.5 for BSA and (b) 6.0 for α-LA; The association efficiency comparison of ( ) before and ( ) after adjusting pH back to (c) 5.5 for BSA and (d) 6.0 for α-LA. AE vs.

XAgg plots obtained from (a - d) for (e) BSA and (f) α-LA (■) before and (●) after pH adjustment. The error bars are standard deviations and the lines are linear fits from Figure

3-5 ...... 197

A-6 Representative TEM images of dried: (a, b) chitosan/BSA complex dispersions; (c) chitosan solutions; and (d) BSA solutions. All of these samples were prepared at pH 6.0 and (like in the case of DLS analysis) subjected to ultracentrifugation prior to imaging

...... 198

~ T ot A-7 Plots showing the effects of K, K and CCS values on the linearity of the AE versus XAgg curves, showing: (a - c) AE versus XAgg curves predicted by Equation 3.8 for various parameter values; (d) the R2-values for the linear regressions to such curves as a

Tot function of K/ at varying KCCS -values (indicated by the numbers on the curves); and (e) a diagram that shows the combinations of K/ and -values where these R2-values exceed 0.97 ...... 201

A-8 Comparison of linear approximations of a moderately non-linear AE versus XAgg

xxv function showing: (black solid line) non-linear curve predicted by Equation 8 for when K

~ T ot = 10 K = 17 mL/mg and CCS = 0.1 mg/mL; (blue dashed line) linearization based on

Equation A.1; (green solid line) linearization based on Eqns. S3 and S6; and (red dotted line) linear regression to non-linear curve (R2 = 0.970) ...... 203

B-1 Evolution in particle ζ-potential with the addition of (■) PPi and (●) TPP. The dashed line indicates the ζ-potential at which the chitosan/TPP particles rapidly coagulate.

The chitosan/TPP particle data is adapted from Huang and Lapitsky [7] ...... 205

C-1 Normalized FITC-chitosan UV-vis absorbance of ( ) FITC-chitosan solutions,

( ) FITC-chitosan/TPP particle dispersions and ( ) α-LA-loaded FITC-chitosan/TPP particle dispersions after centrifugation in (a) 150 mM NaCl and (b) PBS; (c) Normalized

Micro BCA assay UV-vis absorbance of α-LA-loaded FITC-chitosan/TPP particle dispersions after centrifugation in (●) 150 mM NaCl and (▲) PBS. The error bars are standard deviations ...... 207

D-1 Effect of chitosan/TPP micro- and nanoparticles (added in a 0.06 wt% final chitosan concentration) on the Micro BCA absorbance obtained from 0.5 mg/mL BSA solutions in pH 5.5 water after 20× dilution and incubation at 37 °C for 2 h. The error bars are standard deviations ...... 208

D-2 Representative volume-weighted size distributions of BSA-loaded chitosan/TPP particles in (■) glycerol-free water before centrifugation, (●) glycerol solution before centrifugation and (▲) glycerol solution after a centrifugation and redispersion treatment.

The glycerol solutions were obtained by vortex-mixing the glycerol beds into the aqueous dispersion, which is done to redisperse the sedimented particles ...... 210

D-3 The effects of (a) centrifugal force and (b) centrifugation time on (■) BSA and (●)

xxvi chitosan recovery achieved with FITC-chitosan/TPP particles. The centrifugation force effects were investigated using a 30 min centrifugation time, while the centrifugation time effects were probed at a constant centrifugal force of 3.0 × 105 g. The error bars

(which are largely obscured by the symbols) are standard deviations, while the lines are guides to the eye...... 211

D-4 Changes in chitosan/TPP pellet mass during the release process obtained using 30 min of centrifugation at (■) 4.8 × 104, (●) 1.1 × 105 and (▲) 3.0 × 105 g, or 60 min of centrifugation at (▼) 3.0 × 105 g. The dashed line is a guide to the eye and the error bars are standard deviations ...... 212

D-5 Normalized protein concentration plotted as a function of normalized position, at normalized times of (black solid line) 0, (red dashed line) 0.1, (blue dotted line) 0.2 and ~ (green dash-dot line) 0.6 for the cases when Csurf equals to (a) 1 and (b) 0; (c) the release profile for the case when equals to (black solid line) 1 and (red dashed line)

0 ...... 214

D-6 BSA release profiles from chitosan/TPP particle pellets prepared by centrifuging for 30 min at 4.8 × 104 g obtained (■) from whole pellets and (●) pellets cut into 0.1 - 1 mm pieces. The inset focuses on the first 3 h of the release profiles. The lines are guides to the eye while the error bars are standard deviations...... 217

xxvii

List of Abbreviations

AE ...... Association Efficiency α-LA ...... α-lactalbumin

BSA ...... Bovine Serum Albumin

DD ...... Degree of Deacetylation DLS ...... Dynamic Light Scattering DMSO ...... Dimethyl Sulfoxide

FITC ...... Fluorescein Isothiocyanate

GM ...... Glucomannan

HCl ...... Hydrochloric Acid HPLC ...... High-performance Liquid Chromatography

NNLS ...... Non-negative Least Squares algorithm

MW ...... Molecular Weight

NaCl ...... Sodium Chloride NaOH ...... Sodium Hydroxide

PDI ...... Polydispersity Index PPi ...... Sodium Pyrophosphate

TEM ...... Transmission Electron Microscopy TPP ...... Sodium Tripolyphosphate

xxviii

List of Symbols

Cf ...... Final chitosan concentration after centrifugation Ci ...... Initial chitosan concentration before centrifugation I ...... Light scattering intensity mi ...... Mass of the drug released between (i - 1)th and the ith of solvent replacement Ci ...... Supernatant drug concentrations after ith solvent replacement steps R ...... Solvent replacement ratio V ...... Volume C0...... Initial concentration m0 ...... Initial amount of drug within the gel pellet

m∞ ...... Amount of drug released at the end of the experiment mTot ...... Total amount of drug in the particle dispersion Δρ ...... Density difference between the particles and solvent a...... Nanoparticle radius η...... Solvent viscosity ω ...... Centrifugation speed (rad/s) r0 ...... Radial position at the dispersion surface rf ...... Radial position at the bottom of the centrifuge tube XAgg ...... Particle Yield t ...... Time Q0 ...... Initial amount of drug in the drug carrier Qt ...... Amount of drug in the drug carrier at time t K0...... Zero-order release constant k ...... Mass transfer coefficient A ...... Total drug carrier surface area Ks ...... Rate constant k0 ...... Erosion rate constant n...... Diffusional exponent D ...... Diffusion coefficient Cs ...... Drug solutbility i CCS ...... Initial chitosan concentration f CCS ...... Free chitosan concentration i CP ...... Initial protein concentration f CP ...... Free protein concentration remaining in the supernatant after

xxix uptake and centrifugation. Tot CP ...... Total protein concentration Tot CCS ...... Total chitosan concentration q...... Extent of protein binding Keff ...... Effective binding constant C ...... Supernatant protein concentration ~P C ...... Free (unbound) protein concentration ~P K ...... Equilibrium constant for protein/dissolved chitosan binding K ...... Equilibrium constant for protein binding to the particulate chitosan  ...... Particle volume fraction η...... Dispersion viscosity η0 ...... Solvent viscosity N ...... Aggregation number Cnp ...... Particle concentration NA ...... Avogadro’s number Mn ...... Molecular weight Ĉc ...... Final chitosan concentration Ec ...... Cooperative interaction T ...... Absolute temperature εi ...... Non-cooperative binding energy kB ...... Boltzmann constant θ ...... Fractional coverage z ...... Distance from the bottom of the centrifuge tube A(z) ...... Cross-sectional area of the gel pellet at the axial position, z R ...... Inner radius of the centrifuge tube r(z) ...... z-dependent radius within the curved bottom of the centrifuge tube

xxx

Chapter 1

Introduction and Objectives

1.1. Introduction

Polyelectrolytes are polymers composed of ionic repeating units. Based on the functional groups on their backbones, they can be either positively charged (i.e., polycations) or negatively charged (i.e., polyanions). If multivalent counterions are introduced into a polyelectrolyte solution, they can (depending on the polymer and multivalent counterion used) crosslink the polyelectrolyte into 2- or 3-D networks, which form diverse structures, which include particles [8-10], fibers [11, 12], films [13-15] and macroscopic hydrogels [16-18]). This ionotropic gelation process depends strongly on the properties of the polyelectrolytes and counterions used, such as their functional groups and charge densities, as well as the mixture composition – i.e., the polyelectrolyte/counterions charge ratio and the pH, ionic strength and solute composition of the media in which they are placed [19, 20].

By carefully controlling the above formulation parameters, some polyelectrolyte and multivalent counterion mixtures can form colloidally stable micro- and nanoparticles,

1 ranging from tens of nanometers to several microns in diameter. These ionically crosslinked polyelectrolyte particles can be loaded with oppositely charged drugs (and other active payloads) through ionic interactions. This provides a very mild way to prepare drug-loaded micro- and nanocarriers without using organic solvents, toxic covalent crosslinking agents or extreme reaction conditions (see Figure 1-1). Moreover, by reacting to external stimuli (e.g., changes in ionic strength, pH, temperature or enzyme concentration [21]), these ionically crosslinked polyelectrolyte particles can swell, dissociate or degrade, which enables further control over their drug release properties [22,

23]. These ionically crosslinked polyelectrolyte particles can serve several functions, such as sustaining drug release [24], protecting the drug against degradation [25], enhancing drug bioavailability and reducing its side effects [26], and achieving targeted delivery to specific organs [27].

Multivalent Drug Polyelectrolyte Counterions Molecules

Figure 1-1. Scheme of ionically crosslinked polyelectrolyte nanoparticle formation and drug uptake processes.

Among various ionically crosslinked polyelectrolyte drug carriers, ionically crosslinked chitosan particles attract extensive interest as carriers for small molecules such as aspirin [28, 29], doxorubicin [30, 31] and tea catechins [32], and bioactive macromolecules such as proteins [33-36] and DNA [37, 38]. The multivalent ions such as

2 tripolyphosphate (TPP) and pyrophosphate (PPi) (see Figures 1-2a and b) can be used as ionic crosslinkers. The drug uptake ability of ionically crosslinked chitosan particles results from the amine groups on the chitosan backbone (see Figure 1-2c), which become positively charged in acidic media [1]. Therefore, anionic drugs can bind to these cationic binding sites through ionic interactions [3].

(a) (b)

(c)

Figure 1-2. Chemical structures of (a) TPP, (b) PPi and (c) chitosan

Additionally, ionically crosslinked chitosan particles are attractive for drug delivery due to their: (1) mild formation procedures (which typically involve simple mixing of multivalent counterion solutions with chitosan solutions at room temperature

[1, 2, 39]); (2) ability to facilitate drug penetration through mucous membranes (e.g., chitosan is bioadhesive and opens the tight junctions between epithelial cells [40]); and (3) biocompatibility (chitosan is non-toxic and biodegradable [41]). Whether a polymeric carrier is applicable for drug delivery or not can also depend on its size distribution, stability, and drug uptake/release properties [42]. Previous studies have shown that the size and stability of drug-free ionically crosslinked chitosan particles can be tuned by

3 varying the chitosan degree of deacetylation (DD) and molecular weight, chitosan/multivalent counterion concentrations and mixing procedures, and the environment to which the particles are exposed (e.g., the ambient pH and ionic strength)

[3, 43].

When it comes to the drug uptake, ionically crosslinked chitosan particles should typically be designed to optimize their drug content while maintaining control over their other physicochemical properties (i.e., size distribution, swelling and stability). Although factors that affect these physicochemical properties have been studied by many groups, the introduction of drug molecules into the ionically crosslinked chitosan particles can complicate the particle properties (since the pseudoternary chitosan/multivalent ion/water system becomes peudoquaternary). This is because drug molecules might affect ionically crosslinked chitosan particle size, swelling properties and stabilities (to both coagulation and dissolution) through electrostatic, hydrophobic, Van der Waals and hydrogen bonding interactions [44-46].

Furthermore, because some drugs can bind to molecular chitosan, we hypothesize that the existence of free (uncrosslinked) chitosan within ionically crosslinked chitosan particle dispersions can strongly affect particle properties, including their abilities to take up drugs. We therefore postulate that the particle yield (i.e., the fraction of the chitosan molecules aggregated into particles) could be a key determinant of drug uptake efficiency.

A high particle yield might not only provide more binding sites for the drug uptake, but also prevent soluble (non-particulate) chitosan from interfering with the drug uptake. This importance of particle yield, however, has long been ignored in previous drug uptake studies. Though it is generally accepted that the association efficiency (AE), defined as

4 the fraction of drug that is taken up by the drug carrier, increases with the drug/chitosan binding strength [47], variations in AE with the drug, chitosan and TPP concentrations still lack a convincing explanation and, indeed, often exhibit opposing trends. In the literature examining drug concentration effects on the AE of proteins, for instance, several groups [1, 33-35, 39] reported AE to decrease with the protein concentration used during drug uptake, while several others [3, 36, 47-49] reported AE to increase with the protein concentration. Providing more reliable guidelines for controlling these AE-values is one of the objectives of this dissertation.

Controllable drug release behavior, on the other hand, is the final goal of ionically crosslinked chitosan particle design. Drug release from ionically crosslinked chitosan particles can be considered as the reverse process of particle formation and drug uptake.

Parameters that influence their drug uptake (such as their ambient pH and ionic strength) can therefore also affect the drug release process [3, 35]. The drug is believed to slowly diffuse out of ionically crosslinked chitosan particles when placed in the release medium if the particles are stable to dissolution; however, a burst release is expected if the particles dissociate easily (i.e., due to the leaching of the crosslink-forming counterions)

[50]. Accordingly, the drug release profile is also likely to depend on the stability of chitosan/TPP particles, and stable particles should be designed in order to obtain sustained release.

Besides these stability effects, the release profiles reported in the literature are often conflicting, with some studies indicating immediate payload release [27, 51, 52] and others showing release that is sustained over timescales as long as weeks [24, 34, 53].

These opposing findings cloud the understanding of the release properties of

5 chitosan/TPP particles and make their safe and efficacious use difficult to achieve. Thus, a clear understanding of the drug release mechanism, which is needed to predictably control their drug release behavior is still lacking. Though previous studies have paid attention to either the stability of ionically crosslinked chitosan particles [2] or the drug release from ionically crosslinked chitosan particles individually, no attempt has been made to establish a quantitative link between their stability and release properties. Given this lack of insight into the factors underlying their drug release kinetics, the design of drug-loaded chitosan/TPP particles is performed by trial and error [42].

Further confusion regarding particle release properties stems from experimental artifacts which, when the release is analyzed via the commonly-used “solvent replacement” method, may be caused by: (1) irreversible, centrifugation-induced coagulation of chitosan/TPP nanoparticles into macroscopic gel pellets; (2) incomplete particle recovery during centrifugation; and (3) variable frequencies of release media replacement. The first problem stems from the centrifugation step that is used to separate the nanoparticles from surrounding solution, which is done either to collect the particles after their preparation, or to separate the particles from their release media while characterizing their release profiles (i.e., to quantify the the amount of released drug) [3,

34-36, 54]. Though this centrifugation can coalesce the submicron particles into macroscopic gel-like pellets (which increases the diffusion path length and undoubtedly affects the apparent release rates), the literature on chitosan/TPP particles ignores these structural changes.

The third likely artifact in the release measurements could stem from differences in the frequency of release media replacement. Most studies on drug release from

6 chitosan/TPP nanoparticles periodically replace the release media (either fully or partially) with fresh buffer. The selection of media replacement frequencies, however, has been largely arbitrary (with some groups choosing to replace the media every 15 minutes [27,

51, 52], others choosing to replace the media hourly [3, 32, 54], and others yet choosing to replace the media daily [24, 34, 53]). It is therefore possible that the release times reported by various groups might simply be artifacts of the frequency of media replacement, where (if drug transport is rapid relative to the media replacement frequency and drug/particle binding is strong) frequent media exchange might yield rapid release, while infrequent (daily) media exchange might yield slow release [55]. By investigating the effects of both centrifugation conditions and media replacement frequencies, this dissertation aims to reconcile the opposing literature findings and elucidate the true release properties of chitosan/TPP nanoparticles.

1.2. Objectives and Overview

The overarching goals of this dissertation are to develop more-robust guidelines for designing submicron chitosan/TPP particles for optimized drug uptake and controlled release. To address the problems described above, we hypothesized that by probing the fraction of the chitosan that self-assembles into chitosan/TPP particles (and remains within the particulate phase during the release process), one could ascertain the mechanisms of drug uptake and release, so as to reconcile previous conflicting results reported in the literature. To this end, the objectives of this dissertation are to determine the key factors that: (1) control protein encapsulation within submicron ionically crosslinked chitosan particles; (2) affect the stability of ionically crosslinked chitosan

7 particles in the presence of drugs; and (3) control drug release from submicron ionically crosslinked chitosan particles.

This dissertation is organized into eight chapters. To introduce the background and previous achievements in the field, Chapter 2 provides background on the drug uptake, stability and drug release of the ionically crosslinked chitosan particles. Chapter 3 examines the factors that affect chitosan/TPP particle drug uptake and explains the correlation between the AE and the particle yield by using a simple competitive binding model. Because drug release can depend strongly on the dissolution stability of the ionically crosslinked chitosan particles, Chapter 5 investigates the dissolution of ionically crosslinked chitosan particles and Chapter 6 explores the effect of drug loading on the particle dissolution stability and explores the mechanism by which the particles release their payloads. Chapter 7 probes some of the experimental artifacts that effect on the chitosan/TPP micro- and nanoparticles drug release profiles. Finally, Chapter 8 summarizes the key findings of this dissertation and provides recommendations for future work.

8

Chapter 2

Background

2.1. Introduction

This chapter summarizes the previous understanding of protein uptake, dissolution stability and drug release properties of ionically crosslinked chitosan particles. A brief description of the ionically crosslinked chitosan particle preparation is briefly described as the beginning. Afterwards, factors that affect the drug loading into ionically crosslinked chitosan particles are discussed. Then, an overview of factors affecting the stability of ionically crosslinked chitosan particles and methods used to enhance their stability is provided. Finally, we summarizes the hypothesized mechanisms of drug release from ionically crosslinked chitosan particles and reviews some commonly-used methods to conduct release experiments on these colloidal drug carries (as well as the mathematical models used to fit and/or interpret the release profiles).

2.2. Preparation of Ionically Crosslinked Chitosan Particles

Ionically crosslinked chitosan particles can be formed under very mild

9 conditions by mixing multivalent counterion solutions (tripolyphosphate, sulfate and citrate ions) with chitosan solutions [56-58]. By varying the chitosan and multivalent counterion concentration, the particle size can be easily tuned. Higher chitosan concentrations, for example, can lead to a larger particle size (due to more chitosan aggregating into each particle) [59]. A higher multivalent counterion-to-chitosan ratio, on the other hand, tends to generate smaller particles (due to a higher crosslink density) [60].

Furthermore, the chitosan and counterion concentration can also affect whether the particles are colloidally stable – e.g., Calvo et al. reported that chitosan/TPP nanoparticles can only form within limited chitosan and TPP concentration ranges [1].

(a) (b)

Figure 2-1. Identification of compositions where chitosan/TPP nanoparticles form showing: (a) a coarse state diagram; and (b) a state diagram focusing on the region of particle formation. The figures were reproduced from Calvo et al. [1] with permission from John Wiley & Sons.

Additionally, the strength of interaction between chitosan and multivalent counterions can affect the kinetics of particle formation and control the particle size uniformity. Lower chitosan degree of deacetylation (DD) and pH-value, and higher ionic strength can slow down the particle formation by reducing both the chitosan/multivalent counterion binding strength, and the particle formation speed [2, 61]. Huang et al., for

10 example, reported that monovalent salt can slow down the particle formation and (by enabling more-uniform aggregation) lead to a fairly-uniform size distribution with PDIs below 0.2 [7].

2.3. Drug Loading into Ionically Crosslinked Chitosan Particles

Due to the protonation of amine groups (effective pKa = 6.0 - 6.5 [62]), chitosan is positively charged in acidic solutions. When positively charged amine groups are not completely neutralized by multivalent counterions, they can bind anionic drugs (e.g., small molecule drugs [26], protein therapeutics and vaccines [3, 34, 43] and DNA [37]).

Here, a high drug loading in particles is desired because it reduces the number of particles that must be administered to achieve the desired effect. The drug loading process for chitosan-based particles, however, is usually dominated by ionic interactions. Thus, the drug loading efficiency (i.e., the drug fraction that becomes loaded into the particles) can be strongly affected by both the components that form the particles and the environment

(e.g., solvent pH and ionic strength) where the drug loading occurs [35, 63]. Factors that could affect the drug loading into ionically crosslinked chitosan particles are summarized in this section.

2.3.1. Effects of Drug Uptake Media

2.3.1.1. pH Effects

The charge on chitosan, ionic crosslinkers and drugs can be very sensitive to environmental pH changes. For chitosan, the protonation of its amine groups could change significantly with the pH at pH-values near its effective pKa [64]. The higher the pH, the fewer cationic binding sites there are [2]. The degree of ionic crosslinker (e.g.,

11 TPP) ionization, on the other hand, increases with the pH, which leads them to be more-negatively charged [65]. Similarly, the negative charges on the drug can also vary with the pH. For instance, protein drugs are amphoteric molecules whose net charge is controlled by their relative abundance of their cationic amine groups and anionic carboxyl groups [66]. Whether a protein is positively or negatively charged depends on how the pH of its media compares with its isoelectric point (pI) [54, 67]. If pH < pI, the protein has a net positive charge and typically has little affinity for the cationic chitosan-based particles. If pH > pI, however, the protein has a negative net charge and thus binds to the cationic particles more strongly [1, 39]. Consequently, the drug loading into ionically crosslinked chitosan particles can be strongly affected by the ambient pH.

2.3.1.2. Ionic Strength Effects

During their loading into ionically crosslinked chitosan particles, drug molecules have to compete with monovalent ions for the amine groups on the chitosan chain, which is similar to what happens during the binding of crosslinking ions to chitosan [7]. A higher ionic strength leads to more monovalent counterions competing with the drug molecules for the chitosan amine groups, and thus reduces the drug loading into the particles [68]. The ionic strength can therefore greatly affect the drug loading efficiency

[69] – i.e., to maximize the drug loading efficiency, a low ionic strength is preferred.

2.3.2. Effects of Molecular Properties

The interaction between drugs and chitosan can also depend on the chitosan DD and pKa (or pI) of the drug molecules. A high chitosan DD can contribute to a more-efficient drug loading due to a stronger drug/chitosan interaction [35]. Similarly, a

12 lower drug pKa (or pI) can enhance its negative charge, which favors its uptake into cationic chitosan particles.

The chitosan molecular weight (MW) and drug molecular structure can also affect the drug loading. The drug loading efficiency can slightly increase with the chitosan MW

[29, 35]. Moreover, drugs with multiple anionic binding sites usually bind more strongly to chitosan than those with only one binding site (though this may also depend on their degree of ionization and the net charge of the drug molecule [70, 71]). The drug molecule hydrophobicity can also affect the drug loading [72]. Hydrophobic drugs are difficult to dissolve in water and may require organic cosolvents to load into the chitosan particles

[73, 74]. Yet, moderate hydrophobicity enhances the drug loading due to its possible hydrophobic attraction to the chitosan backbone [46].

2.3.3. Effects of Composition

The drug loading efficiency decreases with chitosan concentration [3, 35]. This has been postulated to be a viscosity effect, where a chitosan concentration increase hindered the drug loading by inhibiting drug movement around the chitosan molecules

[3]. Moreover, ionic crosslinker:chitosan ratio can also affect drug loading by varying the abundance of unoccupied, cationic binding sites on the chitosan chains [26, 35]. Besides the chitosan and ionic crosslinker concentrations, the drug concentrations can also affect whether the binding sites within the particles become saturated [29, 75]. Similarly, drug loading can be affected by the addition of components such as surfactants [29], polyanions [76] and other co-encapsulated drugs [26, 35]. Yet, the possibility that variations in the particle yield might also be one of the key factors that influence the drug

13 loading efficiency remains largely ignored. Investigating this effect is one of the goals of this thesis.

2.4. Stability of Ionically Crosslinked Chitosan Particles

Dispersion stability is critical for the preparation, storage and administration of ionically crosslinked chitosan particles. Two aspects are involved in this stability (see in

Figure 2-2 [2]): (1) dissolution stability, which is the particle resistance against dissolution; and (2) colloidal stability, which is the stability of these particles against aggregation and macroscopic precipitation [77]. Micro- and nanoparticles developed for sustained drug release should be stable against dissolution, since a burst release (i.e., rapid, premature release) could occur if particles are rapidly dissolved when administered.

Consequently, this dissolution stability is one of the key themes of this thesis and, as background for this study, this section summarizes current understanding of particle dissolution stability and methods to improve it.

14

(a)

(b)

Figure 2-2. Chitosan/TPP micro- and nanogels can be unstable due to their: (a) aggregation and subsequent precipitation, and (b) dissolution. These schemes were reproduced from Huang et al. [2] with permission from the Royal Society of Chemistry.

2.4.1. Factors Affecting Particle Stability

Because chitosan and ionic crosslinker interactions can change when the particles are exposed to physiological conditions, the ionically crosslinked chitosan particles might not be stable to dissolution. This dissolution could result in rapid drug release. Thus, to make the sustained release possible, the particles should be stable to dissolution and previous work has identified several key factors that affect their stability [2, 78-80].

Like drug loading, the dissolution stability of ionically crosslinked chitosan particles depends strongly on the ambient pH and ionic strength. Since the particles are usually prepared in a mildly acidic and salt free environment, which differs from the physiological media (i.e., the pH ~ 7 in intranasal [81] and ocular environment [82]) they

15 encounter upon administration, both the pH and ionic strength could change drastically during their use [2]. An increased pH can deprotonate chitosan and reduce its positive charge. Moreover, when pH > 7, the chitosan solubility may significantly decrease, causing the chitosan to precipitate from the dispersion [83]. Similarly, an increased ionic strength would weaken the chitosan/ionic crosslinker binding, evidently due to the competitive binding of monovalent and multivalent ions to the chitosan amine groups [2,

7]. Thus, a combination of these effects can cause the ionic crosslinked chitosan particles to dissolve under physiological conditions.

Besides these environmental factors, the inherent properties of compounds that form the ionically crosslinked chitosan particles can also influence the particle stability.

Chitosan/TPP particles prepared from chitosan with higher DD-values, for instance, show a stronger resistance to dissolution under physiological conditions [2]. This is mainly due to a stronger chitosan/TPP binding exhibited by high-DD chitosan. Unfortunately, high-DD chitosan has a lower solubility at the near-neutral pH that is typically found at their application sites compared to that of its lower-DD counterparts [84]. Similarly, the chitosan MW is an additional factor that could affect particle stability. Higher chitosan

MW can enhance the dissolution stability of ionically crosslinked chitosan particles, but has a lower solubility at physiological pH than low MW chitosan [2].

Moreover, the dissolution stability of ionically crosslinked chitosan particles also depends on the chitosan and ionic crosslinker concentrations after particle dilution. When their concentrations are low (i.e., the dispersion are highly diluted), the particles tend to dissociate [2]. Thus, the final chitosan and crosslinker concentrations after dilution should be matched to their application conditions when probing the dissolution stability

16 of ionically crosslinked chitosan particles.

2.4.2. Methods of Increasing Particle Stability

Previous studies have shown that payload-free chitosan/TPP particles (which are the most widely studied ionically crosslinked chitosan particles) tend to have a low dissolution stability under physiological conditions [2, 85, 86], and several approaches were explored to increase the particle stability. The underlying principle of increasing the dissolution stability is to strengthen the intermolecular attraction between the particle components. These interactions can be enhanced by covalent crosslinking, further ionic crosslinking and hydrogen bond formation. Methods developed to increase the dissolution stability of ionically crosslinked chitosan particles are summarized in this section.

2.4.2.1. Covalent Crosslinking

Covalently crosslinked chitosan particles, which are synthesized using covalent crosslinkers such as glutaraldehyde [87], have enhanced dissolution stability in physiological environments [88]. Despite this advantage, however, drawbacks such as a drastic decrease of chitosan bioadhesiveness and increased toxicity prevent their clinical applications [89, 90]. Glutaraldehyde, for instance, can cause irritation to mucosal membranes [91]. Genipin, which is isolated from fruits, was proposed as an alternative to glutaraldehyde due to its lower toxicity [92]. The covalent crosslinking with genipin, however, also requires the use of an organic solvent. Thus, if drug molecules are loaded during particle formation (aside from possible side reactions between the genipin and the

17 drug), the washing out of organic solvent could also elute the drug molecules. Conversely, if drug molecules are loaded after the covalent crosslinking, drug penetration into the particles could be limited by the crosslinked polymer network and might cause a low drug loading efficiency. Therefore, the use of additional non-covalent crosslinkers to increase the stability of chitosan/TPP particles (via further ionic or hydrogen bonding) is likely more favorable.

2.4.2.2. Further Ionic Crosslinking

Compared to covalent crosslinking, ionic crosslinking occurs at much milder condition (e.g., no organic solvent and no heating) and avoids the need to remove unreacted crosslinking agents [39]. Therefore, using a second ionic crosslinker might be a good alternative approach to increasing the dissolution stability of ionically crosslinked chitosan particles. Giacalone et al. [93], for instance, have shown that chitosan/TPP particles can be stabilized in PBS by also crosslinking the particles with iron ions, which would associate with both chitosan and TPP. Since most macromolecular drugs such as proteins and DNA have multiple binding sites (e.g., phosphate and carboxyl groups) and are able to form insoluble complexes with molecular chitosan [83, 94-96], these macromolecular drugs might also increase the particle stability against dissolution. Based on TEM analysis, Gan et al. [3] has suggested that bovine serum albumin (BSA)-loaded chitosan/TPP particles persisted in PBS for over 48 h (see Figure 2-3), rather than rapidly dissociating, as reported for their payload-free counterpart by Huang et al. [2]. Similarly,

Chen et al. reported that β-lactoglobulin-loaded chitosan/TPP nanoparticles remained intact in simulated intestinal fluid (pH 7.5) for 10 h [67]. Similar to these proteins, DNA may also stabilize chitosan/TPP particles against dissolution. Csaba et al. loaded plasmid

18 DNA (pDNA) into chitosan/TPP particles, and the particles did not release any of their pDNA regardless of whether pH 4 or 7.4 acetate buffer was used as the release medium

[37]. This again suggested that, due to the strong chitosan/pDNA binding, the chitosan/TPP particles did not dissolve.

(a) (b)

(c) (d)

(e)

Figure 2-3. TEM images of BSA-loaded chitosan/TPP nanoparticles taken during protein release in pH 7.0 PBS at 37 °C after (a) 3, (b) 6, (c) 12, (d) 24 and (e) 48 h of release time. The figure was reproduced from Gan et al. [3] with permission from Elsevier.

19 2.4.2.3. Other Methods

Besides the ionic and covalent bonding, hydrogen bonding may also increase the ionically crosslinked chitosan particle stability against dissolution. Glucomannan (GM), for instance (see Figure 2-4), has been used to stabilize the chitosan/TPP particle network by forming hydrogen bonds with chitosan amine groups [4]. Other than adding another crosslinker, nonionic surfactants may also be able to stabilize ionically crosslinked chitosan particles through hydrophobic association and hydrogen bonding. For instance,

Keawchaoon et al. [97] reported that a polyoxyethylene sorbitan monostearate (Tween 60) assisted in the formation of carvacrol-loaded chitosan/TPP nanoparticles. The particles remained intact in both pH 3 acetate buffer and pH 7 and 11 phosphate buffers for two months.

Figure 2-4. Chemical structure of GM. The figure was reproduced from Alonso-Sande et al. [4] with permission from the American Chemical Society.

Introducing a stabilizer (e.g., another crosslinker or chitosan-binding polymer) to drug-loaded chitosan-based particles may stabilize the particle against dissolution, however, usually at the expense of a lower drug loading efficiency due to their occupation of chitosan binding sites [4, 98]. Thus, a situation where the drug itself stabilizes the particles is more advantageous.

20

2.5. Drug Release from Ionically Crosslinked Chitosan Particles

Previous studies have shown that the drug molecules can be stabilized against degradation during administration when loaded into ionically crosslinked polyelectrolyte particles [98-102]. Moreover, due to the sensitivity of ionically crosslinked polyelectrolyte particles to pH and ionic strength, the drug-loaded particles can be stimulus-responsive and release their payloads only upon reaching their targets [103]. For instance, because the environment in tumors is more acidic than in healthy tissues [104,

105], the pH change can trigger payload release from ionically crosslinked polyelectrolyte particles. Additionally, compared to the administration of periodic doses, such sustained release systems can maintain the drug concentrations at therapeutic levels while reducing their toxicity (see Figure 2-5 [5]). To make this possible for chitosan-based micro- and nanoparticles, a good understanding of their release mechanism should be provided for their design and optimization. However, conflicting release results have frequently been reported in the literature for ionically crosslinked chitosan particles, with some studies indicating immediate payload release [27, 51, 52], while others showing release that is sustained over timescales as long as weeks [24, 34,

53].

21

Figure 2-5. Comparison of drug delivery though periodic dosage and controlled release. This figure was reproduced from Uhrich et al. [106] with permission from American Chemical Society.

Experiments by Zhang et al., for example, have shown that 90% of BSA was released in pH 7.5 simulated intestinal fluid up to a week [36]. A similar trend was also indicated by Xu et al., who reported that the BSA released in pH 7.4 PBS for about a week [35]. A faster release was observed by Gan et al., where 30 - 75% of BSA was released from chitosan/TPP particles within 6 h in PBS, during which the particles maintained their morphology [3]. Conversely, other studies have shown a much faster release [51, 52, 98]. Li et al., for instance, studied the BSA release from chitosan/TPP particles in pH 7.4 PBS [98]. The release process ended within 1 h, whereupon the fraction of BSA released remained constant at 95%.

These opposing findings confound the understanding of the release properties of chitosan/TPP particles and make their safe and efficacious application difficult to achieve.

22 Though the abovementioned drug-loaded particles were prepared with similar preparation procedures, their release behavior may have been affected by the experimental methods used to obtain the release profiles [55]. To understand the effect of release experiment method on the particle release behavior, three most-commonly-used experimental techniques for characterizing drug release profiles are illustrated below and their potential limitations are discussed. As a prelude to this discussion, this section summarizes common mechanisms and mathematical models of drug release, which may have been widely misused in analyzing submicron ionically crosslinked chitosan particles, to provide a better understanding of the release behavior of ionically crosslinked chitosan micro- and nanoparticles.

2.5.1. Release Mechanisms

Depending on the particle preparation method and the affinity between all the particle components, the drug release from a particulate carrier is typically controlled by one or more of the following mechanisms (see Figure 2-6) [50, 107]: (1) drug desorption from the particle surface, (2) drug diffusion through the particle matrix, (3) bulk particle degradation, (4) surface erosion and (5) enhanced drug transport due to matrix swelling.

Drug desorption is usually a rapid process since the drug molecules that are attached to the particle surface are directly exposed to the release media during administration. Thus, this release mechanism typically gives rise to a burst release [108-110]. Conversely, in cases where the particles are stable to degradation, the release of the drug that is inside the particles is often controlled by a diffusion process [50]. This diffusion can be accelerated by matrix swelling, which is common in dry powder formulation [111, 112], and pH-sensitive gels (e.g., gels that are collapsed and poorly-permeable at low, gastric

23 pH, but swollen and highly-permeable at near-natural intestinal pH [21, 113]). Drug release can also be greatly accelerated when particles undergo matrix degradation or surface erosion, which results from either environmentally-triggered (e.g., pH and ionic strength-triggered) dissolution [114, 115], or hydrolytic and enzymatic degradation [114,

116, 117].

Desorption Swelling

Diffusion Surface Erosion

Figure 2-6. Scheme of the various drug release mechanism.

To prevent a rapid drug release, the particles should be stable against rapid dissolution (see Section 2.4). Whether the particles are stable or not during the release experiment has been qualitatively probed by transmission electron microscopy (TEM) [3,

118] and dynamic light scattering (DLS) [107]. Additionally, mathematical release models have been used in many studies on colloidal drug carriers, to either quantitatively characterize the drug release behavior or infer their release mechanisms. Some of the commonly-used release models are illustrated and discussed in Section 2.5.2.

24 2.5.2. Release Models

To more quantitatively analyze the release properties of drug carriers (and/or infer their release mechanisms), numerous release models have been developed

[119-122]. Some of these model equations are based on the release mechanisms described in the last section, while others are empirical. Efforts have also been made to develop models that describe release kinetics based on multiple simultaneous release mechanisms [50, 123]. Such model equations, however, tend to become very complex and hard to use [124]. Thus, though simple conventional models use many limiting assumptions, they are extensively utilized in analyzing drug release mechanisms. The most common of these simple models are summarized in this section.

2.5.2.1. Zero-Order Model

When cumulative drug release scales linearly with time regardless of the drug concentration remaining in the drug carrier, the drug release is called a zero-order release process (see Figure 2-7) [125]:

Q0 - Qt = K0t (2.1)

where Q0 is the initial amount of drug in the drug carrier, Qt is the amount of drug in the drug carrier at time t and K0 is the zero-order release constant.

25

Figure 2-7. Representative scheme of zero, first-order release and t1/2 type release (where t is the release time). This figure was reproduced from Cussler [5] with permission from Cambridge University Press.

Zero-order release is usually desirable, and can be achieved by controlling the drug diffusion (e.g., using a permeable capsule containing saturated drug solution, an impermeable capsule containing unsaturated drug solution with a swellable hole, and an altered initial concentration profile in the solid carrier) or solvent diffusion in the drug carrier (e.g., using swelling-controlled carriers and osmotic pumps) (see Figure 2-8) [5].

Additionally, the superposition of drug dissolution (e.g., when drug diffusion is much faster than the movement of glassy rubbery interface of dry drug carrier) and carrier matrix dissolution (e.g., when the loaded drug concentration is higher than its solubility and the drug dissolution is the limiting parameter of drug release) with drug diffusion and carrier swelling can also lead to a zero-order release [126-129].

26 (a) (b) (c)

(d)

Figure 2-8. Some methods can reach zero-order release: (a) using a permeable capsule containing saturated drug solution, (b) an impermeable capsule containing unsaturated drug solution with a swellable hole (c) an altered initial concentration profile in the solid carrier and (d) osmotic pumps. This figure was reproduced from Cussler [5] with permission from Cambridge University Press.

2.5.2.2. First-Order Model

Drug release that follows first order kinetics (see Figure 2-7) can be modeled by the equation [125, 130, 131]:

M t  kAt  1 exp   (2.2) M   V 

where Mt/M∞ is the fraction of drug released at time t, k is the mass transfer coefficient, A is the total drug carrier surface area and V is the volume of water. The model is based on the following assumptions: (1) the drug is highly water soluble, (2) the matrix is insoluble,

27 nonswellable and porous, and (3) the matrix maintains its integrity throughout the release process [132].

2.5.2.3. Hixson-Crowell Model

The Hixson-Crowell model describes drug release from carriers that undergo dissolution or surface erosion (see Figure 2-6). In this case, there is a shrinkage in particle surface area and volume, and the release is described by [133]:

1 3 1 3 M0  Mt  Kst (2.3)

where M0 is the initial amount of drug in the carrier, Mt is the remaining drug in the carrier at time t and Ks is a rate constant incorporating the surface-volume relation, drug solubility and solvent phase mass transfer rate. This model assumes that sink conditions

(where the drug concentration in the release medium kept low enough to not affect further drug release) are maintained and that the shape of the drug carrier remains constant as it dissolves or erodes.

2.5.2.4. Hopfenberg Model

The Hopfenberg model was developed for planar, cylindrical and spherical drug carriers undergoing surface erosion (see Figure 2-6) [121, 134]:

n M t  k0t   1 1  (2.4) M   C0a0 

28 where Mt/M∞ is the fraction of drug released at time t, k0 is the erosion rate constant, C0 is the initial drug concentration in the carrier and a0 is the initial radius for a sphere or cylinder, or the half-thickness for a slab. Further, n is a parameter related to the carrier shape, which equals to 1, 2 and 3 for a slab, cylinder and sphere, respectively. The model assumes that the drug release is surface-erosion controlled and that drug diffusion within the matrix is negligible [121]. When applied to spherical drug carriers, the Hopfenberg model reduces to the Hixson-Crowell model.

2.5.2.5. Higuchi Model

The Higuchi model has been developed for slab, spherical and cylindrical matrix systems, where the drug is dispersed in particulate or sparingly-soluble droplet form and release occurs through Fickian diffusion (see Figure 2-6) [135-137]. A simplified Higuchi equation describes drug release from a spherical matrix at short times is (which corresponds to t1/2 type release, see Figure 2-7) [135, 137]:

1 2 M 18DC t  t   s   2  (2.5) M   r0 C0 

where Mt/M∞ is the fraction of drug released at time t, D is the diffusion coefficient of the drug in the particle matrix, Cs is the drug solubility, C0 is the drug concentration in the particle matrix. t is the cumulative release time and r0 is the radius of particles. This model assumes that: (1) matrix swelling and dissolution/erosion are negligible; (2) the drug is initially uniformly dispersed within the drug carrier matrix and its concentration in the matrix greatly exceeds its solubility; (3) drug diffusivity remains constant; and (4) perfect sink conditions are maintained [120].

29 2.5.2.6. Korsmeyer-Peppas Model

The Korsmeyer-Peppas model describes drug release from slabs, spheres or cylinders with a semi-empirical power-law equation [119]:

M t  ktn (2.6) M 

where Mt/M∞ is the fraction of drug released at time t, k is a constant incorporating the characteristics of the macromolecular network and the drug, and n is the diffusional exponent, which can indicate the transport mechanism [119]. This equation only applies to the first 60% of fractional release, with the assumptions of one-dimensional diffusion and perfect sink conditions. Since the n-value depends on the drug carrier shape, when the drug release is controlled by Fickian diffusion, n = 0.5 for slabs, 0.45 for cylinders and 0.43 for spheres [119]. Conversely, when 0.43 < n < 1.00 for spherical carriers, the release is attributed to anomalous (non-Fickian) transport, where the drug transport may be affected by a swelling of the particle matrix (see Figure 2-6). Thus, fitting the release profile with the Korsmeyer-Peppas model can help infer the mechanism of drug release from micro- and nanoparticles from the n-value.

2.5.2.7. Weibull Model

The Weibull model is an empirical equation that describes the rate of drug release when the release occurs through matrix dissolution (see Figure 2-6) [138]:

M t 1 exp - at b  (2.7) M 

30 where Mt/M∞ is the fraction of drug released at time t, a, is the timescale of the release process, b is a parameter that describes the shape of the release curve. Although this empirical model can fit most release profiles very well, its parameters a and b do not have physical meaning [125, 131]. Thus, it is of limited use in ascertaining the release mechanism [139].

2.5.2.8. Model Application to Ionically Crosslinked Chitosan Micro- and Nanoparticles

Most of the above models are only strictly applicable to release processes that occur through specific release mechanisms. The Higuchi model, for example, can fit the release profile well when the drug release is diffusion-controlled and the initial drug content greatly exceeds its matrix solubility within the release system. When additional release mechanisms (such as surface erosion or matrix degradation) are involved, they can be reflected in the n-value when the release profile is fitted by the Korsmeyer-Peppas equation. However, this model cannot quantify the contribution of each release mechanism to the release profile. To analyze the mechanism of drug release from ionically crosslinked chitosan micro- and nanoparticles, many studies simply fitted the release data to various conventional model equations and selected the one with the highest R2-value, without any regard for whether the assumptions of these models corresponded to their release experiments [87, 140-142]. The use of this method to infer the drug release mechanism, however, is unconvincing [143]. This is because: (1) the

R2-value of different models can be very close which, given the uncertainty in experimental data, makes the “best model” difficult to truly determine; and (2) conditions

31 corresponding to the model assumptions – such as perfect sink conditions being maintained or surface erosion taking place – were often not achieved (or at least not proven to occur) in the release experiments. Thus, we postulate many of the above model equations to be inappropriate for modeling release from ionically crosslinked chitosan particles. A more reliable way to explore the release mechanism might be to first experimentally/qualitatively determine which of the assumption are valid (e.g., whether the release is erosion/degradation-controlled), and only then using applicable models

(from those described above) to quantitatively analyze their release profiles. Probing the role of degradation on drug release from ionically crosslinked chitosan particles is one of the goals of this thesis.

2.5.3. Methods for Conducting Drug Release Experiments and Their Potential Pitfalls

Before conducting in vivo drug release experiments, the drug release properties are usually explored by in vitro experiments. The methods used in the in vitro release experiments, however, may strongly affect the apparent drug release behavior [55].

Before using a release model to theoretically/quantitatively analyze the drug release kinetics, it is important to ensure that the release experimental method does not contradict the model assumptions. This section summarizes the three most commonly-used in vitro methods of measuring release profiles and their potential pitfalls.

32 2.5.3.1. Solvent Replacement Method

The solvent replacement method has been extensively used in characterizing drug release from micro- and nanoparticles [27, 51, 52, 144]. To start the experiment, the drug-loaded particle dispersion or powder is mixed with the release medium. After some time, the particles are separated from the release media by centrifugation. A fraction of the solvent is then collected and replaced with the same volume of fresh release media, whereupon the particles are redispersed to continue the release process. The cumulative drug release is determined from the collected supernatant using UV-vis spectroscopy

[145] (or other analytical methods [32, 146]) and this centrifugation/solvent replacement procedure is repeated until the release curve reaches a plateau.

This technique has advantages of enabling many release experiments to be easily conducted in parallel and, aside from a centrifuge and a spectrophotometer, not requiring specialized equipment. However, it can result in artifacts that can produce unreliable release profiles. Avoiding these artifacts requires: (1) a high solvent replacement frequency (especially at early release stages) to create “sink conditions;” (2) strong centrifugation conditions to ensure that all particles are sedimented from the supernatant;

(3) a centrifugation procedure that avoids irreversible coagulation of micro- and nanoparticles into macroscopic pellets.

Unfortunately, these artifacts have been largely ignored in the literature on ionically crosslinked colloids, which may be an underlying cause of the conflicting release behaviors that have been reported. The selection of media replacement frequencies, for instance, has been essentially random (with some groups replacing the media every 15 min [27, 51, 52, 144], others replacing the media hourly [3, 32, 54, 147],

33 and others yet replacing the media daily [24, 34, 53]). It is therefore possible that the release times reported by various groups might simply be artifacts of the media replacement frequencies, where (if drug transport is rapid relative to the media replacement frequency and drug/particle binding is strong [24]) sink conditions are not attained, and frequent media exchange yields rapid release, while infrequent (daily) media exchange yields slow release [55].

Moreover, the selection of centrifugation conditions (for separating the particles from the release media) may have a strong impact on the release profile. If the centrifugation is less-intense, some of the particles may remain in the supernatant and lead to erroneous measurements of supernatant drug concentration (and therefore erroneous release profiles). Thus, a complete particle separation from solution may require very high centrifugal forces and long centrifugation times. Unfortunately, despite this issue remaining largely ignored in ionically crosslinked particle literature, strong centrifugation can also cause the particles to agglomerate into gel-like pellets. Because redispersion of such pellets might be only partial, this nanoparticle agglomeration may drastically increase the diffusion distance and, in turn, retard drug release. Because each group uses a different centrifugation speed, which invariably leads to differences in nanoparticle agglomeration and quality of nanoparticle separation, we postulate that some of the conflicting reports are caused by “centrifugation artifacts.”

2.5.3.2. Dialysis Method

In the dialysis method, a particle dispersion is placed in a dialysis bag and dialyzed against a release medium, which is maintained at sink conditions [148]. The

34 dialysis membrane allows the permeation of drug molecules, while keeping the particles within the bag (see Figure 2-9). The cumulative drug release is determined by measuring the drug concentration outside the dialysis bag. Though this technique is very easy to use, it has two key limitations: (1) the release profile can be strongly affected by the rate of drug permeation through the dialysis membrane when drug release from the particles is fast relative to the time required for its release from the dialysis bag [6, 55, 149, 150]; (2) certain drugs, such as proteins, may adsorb on the dialysis membrane [151, 152]; and (3) the release profile obtained by this method is subject to equilibrium partitioning artifacts, where in the case where release from the particles is faster than permeation through the dialysis bag, release can be retarded by the non-sink condition inside the dialysis bag

[55].

Dialysis Bag Sink

Drug in Drug Diffusion Particle Across Membrane Drug Release

Stirring

Figure 2-9. Scheme of the dialysis method of measuring drug release (adapted from Washington [6] with permission from Elsevier).

35 2.5.3.3. In Situ Methods

An alternative method, which avoids the problems associated with the dialysis method, is to directly place the particles in the release media at a concentration that corresponds to sink condition [153]. Here, the release media directly receives the drugs released from the particles. Before analyzing the released drug concentration, the particles can simply be removed using filters with nanoscale pore size [153] or centrifugation [154, 155]. Since the released drug concentration remains very low with this method, accurate drug quantification was typically achieved with HPLC [153-155].

Compared to the other two methods used to obtain drug release profiles, this method is more likely to reveal the true release behavior of ionically crosslinked micro- and nanoparticles.

2.5.3.4. Conducting Drug Release Experiments to Ionically Crosslinked Chitosan Micro- and Nanoparticles Methods

Among the abovementioned methods of conducting drug release experiments, the solvent replacement method is the most commonly used. This may be because the solvent replacement method uses only a centrifuge, which is widely available, to separate the particles from the solvent, and typically uses UV-vis spectroscopy to quantify the drug concentration in the release media. The dialysis method and in situ method, however, may require using expensive equipment such as HPLC to determine the low drug concentrations maintained in the receiving solution throughout the release process. The use of HPLC is also time-consuming, which is unfavorable for drugs that rapidly degrade.

Additionally, the solvent replacement method enables the release media to be easily

36 changed during the release experiment (e.g., to create a pH gradient during the release experiment, which mimics the gastrointestinal tract [156]). Thus, numerous drug release experiments on ionically crosslinked chitosan micro- and nanoparticles were conducted using the solvent replacement method [30, 34, 39, 42]. Based on this previous literature, this dissertation will focus its analysis on the solvent replacement method and will systematically explore the consequences of its various artifacts.

37

Chapter 3

Factors that Affect Drug-Uptake into Submicron

Chitosan/TPP Particles

3.1. Introduction

Though many studies have been performed to elucidate factors that control chitosan/TPP micro- and nanogel size [1, 7, 32, 36, 63, 80, 157] and drug uptake [63, 158,

159], there remain significant challenges in optimizing drug association efficiency (AE) of the loaded drug, which is the percentage of the added drug that is successfully loaded into the drug carrier. This AE varies significantly with the formulation parameters (e.g., the chitosan, TPP and drug concentrations used), but these variations remain both poorly understood and difficult to reliably predict.

Though it is generally accepted that AE increases with the chitosan/payload binding strength, variations in AE with the payload, chitosan and TPP concentrations still lack a convincing explanation and, indeed, often show opposing findings. In the literature examining drug concentration effects on the AE of proteins, for instance, several groups

38 reported AE to increase with the protein concentration used during drug uptake [3, 36, 47,

48, 160], while several others reported AE to decrease with the protein concentration [1,

33-35, 39].

Further confusion exists regarding the effects of TPP:glucosamine molar ratio and chitosan concentration. Several groups reported the AE to increase with the

TPP:glucosamine ratio [3, 27, 48, 161], but provide no consensus for the reasons for this effect (with one group suggesting that the enhanced AE at higher TPP concentrations reflects denser crosslinking [27], while another attributing it to a shift in the pH that is speculated to enhance protein/chitosan binding [3, 27]). Moreover, a report by Zhang et al. suggested a nonmonotonic TPP effect on AE, where the uptake first increases with the

TPP concentration and then slightly decreases [33], which brings the generality of protein uptake increasing with the TPP:glucosamine ratio into question.

Similarly, Gan et al. and Zhang et al. have reported a counterintuitive result where increasing the chitosan concentration during protein uptake decreased the

AE-value [3, 33]. This was regardless of whether the chitosan concentration was varied at a constant TPP concentration [33] or at a constant TPP:glucosamine ratio [3], and both groups attributed the reduction in uptake efficiency to increased solution viscosity. The reasons for this speculation, however, are unclear and could be challenged. Indeed, a report by Yang et al. indicated opposing findings, where the AE first increased drastically with the chitosan concentration, and then decreased slightly with the chitosan concentration as the chitosan concentration became high [162]. The variability and poor understanding of these results make it difficult to draw generalized conclusions on the factors that determine protein uptake. To this end, we hypothesized that many of these

39 variations in AE with the protein/chitosan/TPP concentrations might reflect the largely-ignored variability in the particle yield.

To test this hypothesis, here we systematically explore the effects of

TPP:glucosamine molar ratio, pH, ionic strength and added protein on the chitosan/TPP particle yield. We then use bovine serum albumin (BSA; molecular weight ≈ 66.5 kDa; pI

≈ 4.7 - 4.9 [163]) and α-lactalbumin (α-LA; molecular weight ≈ 14.2 kDa; pI ≈ 4.2 - 4.5

[163]) as model proteins to investigate the effect of particle yield on the protein uptake.

The relationship between the AE and the particle yield that emerges from this analysis is then used to provide essential guidelines for choosing chitosan, TPP and protein concentrations for optimizing protein uptake into chitosan/TPP particles. Finally, we explain the correlation between the AE and particle yield using a simple competitive binding model. Most of the text and all or the data in this chapter was reproduced from:

Cai, Y.; Lapitsky, Y.; J Colloid Interface Sci. 2017, 494, 242.

3.2. Materials and Methods

3.2.1. Materials

Millipore Direct-Q 3 deionized water (DI water) with resistivity of 18.2 MΩ·cm was used in all experiments. Chitosan (viscosity-average molecular weight = 154 kDa and DD = 86% as determined by capillary viscometry [2] and pH titration [164], respectively), sodium tripolyphosphate (TPP), fluorescein isothiocyanate (FITC), BSA, dimethyl sulfoxide (DMSO) and Bradford protein reagent were all purchased from

Sigma-Aldrich (St. Louis, USA). α-LA was a kind gift from Davisco Foods International,

Inc. (Eden Prairie, MN). Hydrochloric acid (HCl), sodium chloride (NaCl) and sodium

40 hydroxide (NaOH) were purchased from Fisher Scientific (Fair Lawn, NJ). All materials were used as received.

3.2.2. Preparation and Analysis of FITC-Labeled Chitosan

To facilitate measurement of the soluble (i.e., unaggregated) chitosan concentration before and after particle formation, the chitosan was labeled with FITC.

Specifically, 5.7 mL of FITC solution in DMSO (1 mg/mL) were added to 150-mL batches of 0.2 wt% chitosan solution (adjusted to pH 4.0 using HCl) and allowed to react for 3 h in the dark at room temperature. The resulting FITC-labeled chitosan

(FITC-chitosan) was then dialyzed for 24 h against DI water three times. By quantifying the unreacted FITC concentration in the dialysate, it was determined that roughly 0.3% of the chitosan amine groups were labeled with FITC. Once dialyzed, the purified

FITC-chitosan solution was freeze dried and stored until use at -18 °C .

Because FITC absorbs UV and visible light, dissolved FITC-chitosan can be quantified by UV-Vis spectrometry. Here, since the UV-Vis spectra of FITC-chitosan solutions was pH-dependent, an isosbestic wavelength of λ = 463 nm was chosen to quantify FITC-chitosan concentrations at pH ≥ 4.0 and 0 mM NaCl (where ε = 0.944 mL mg-1 cm-1 remained constant with the pH). When parent solution pH was reduced to 3.0, however, ε at λ = 463 nm became pH-dependent and the FITC-chitosan concentration was quantified at λ = 440 nm (ε = 0.791 mL mg-1 cm-1).

The UV-Vis spectra of FITC-chitosan solutions also varied with the ionic strength.

Thus, at pH > 4.7 and NaCl concentrations of 150 and 1000 mM, the dissolved

FITC-chitosan was quantified at λ = 463 nm (where ε = 1.002 mL mg-1 cm-1 and 1.324

41 mL mg-1 cm-1, respectively). Conversely, when the parent solution pH was 4.0, the

FITC-chitosan in 150 and 1000 mM NaCl solutions was quantified at λ = 440 nm (where

ε = 0.776 mL mg-1 cm-1 in 150 mM NaCl and 1.025 mL mg-1 cm-1 in 1000 mM NaCl).

3.2.3. pH, Ionic Strength and TPP:Glucosamine Ratio Effects on Micro- and Nanogel Yield

To study the pH and TPP:glucosamine ratio effects on the fraction of

FITC-chitosan that assembled into particles, the pH of parent FITC-chitosan solutions was adjusted to either 4.0 or 5.5 using HCl and NaOH. This enabled the pH to be tuned with little (no more than a few mM) effect on the ionic strength. To prepare

FITC-chitosan/TPP micro- and nanogels at variable TPP:glucosamine ratios, 1 mL of variably concentrated TPP solutions, 0 - 0.30 wt% (0 - 8.2 mM) at pH 4.0 and 0 - 0.15 wt%

(0 - 4.1 mM) at pH 5.5, were added dropwise into 5 mL of pH-matched 0.1 wt%

FITC-chitosan solution (which was 5.2 mM in its glucosamine monomer units). The resulting chitosan/TPP particle dispersions were then equilibrated for 30 min. To make the final chitosan and TPP concentrations here the same as those used in the protein uptake experiments (vide infra), an additional 2 mL of protein-free water (which was the volume of protein solution added during protein uptake) with corresponding pH-values were then added dropwise to the FITC-chitosan/TPP mixtures. Moreover, to ensure proper mixing during each addition step, the chitosan/TPP mixture was stirred with a 12 mm × 4 mm magnetic stir bar at 800 rpm throughout this entire procedure.

After stirring for an additional 30 min, the dispersions (which were composed of polydisperse 40 - 1000 nm particles; Appendix A, Figure A-1) were centrifuged in a

Beckman Coulter ultracentrifuge (Ann Arbor, MI; Motor model: SW 55Ti) at 50,000 rpm 42 (303,000 g) and 20 °C for 1.5 h. This centrifugation force and time was sufficient to separate almost all of the particles from solution while keeping dissolved chitosan and unencapsulated BSA dissolved in the supernatant. The FITC-chitosan that remained dissolved in the supernatant (i.e., the FITC-chitosan that was not aggregated into particles) was then quantified by UV-Vis spectrometry at the wavelengths described above. Based on these measurements, the particle yield, defined here as the fraction of chitosan formed into chitosan/TPP particles (XAgg), was determined as:

i f CCS - CCS X Agg = i (3.1) CCS

i f where CCS was the initial chitosan concentration and CCS was the free chitosan concentration remaining in the supernatant after particle formation.

To explore the ionic strength effect on XAgg, chitosan/TPP particles were prepared by titrating TPP solutions into FITC-chitosan solutions (and then adding 2 mL of pH-matched water) as described in the preceding paragraphs. The ionic strength, however, was now varied by adding either 0, 150 or 1000 mM NaCl to all parent FITC-chitosan,

TPP and pH-matched water solutions (so that the particles were now formed in the presence of either 0, 150 or 1000 mM of added NaCl). To confirm that the salt effects were qualitatively similar regardless of the preparation of pH, two different pH-values

(4.0 and 5.5) were used in this experiment. Moreover, to improve reproducibility, all the

XAgg measurements were repeated three times.

3.2.4. Protein Uptake Effects on Micro- and Nanogel Yield

To probe the effect of protein addition on XAgg, 0.2 - 0.6 wt% BSA solutions at

43 pH-values of either 4.0 or 5.5, and 0.12 - 0.36 wt% α-LA solutions at pH 6.0 were prepared (where a higher α-LA solution pH was used due to the insufficient α-LA solubility at lower pH-levels). Besides the protein concentration effect, the effect of the uptake method used (i.e., incubation versus incorporation [3]) was investigated. In the incubation method, 1 mL of variably concentrated TPP solutions (0 - 0.30 wt% at pH 4.0,

0 - 0.15 wt% at pH 5.5 and 0 - 0.12 wt% at pH 6.0) were added dropwise into 5 mL of pH-matched 0.1 wt% FITC-chitosan solution, to produce particles with variable

TPP:glucosamine ratios. After stirring for 30 min, 2 mL of pH-matched protein (either

BSA or α-LA) solution were added dropwise, whereupon the dispersions were stirred for another 30 min (to allow uptake to occur) before being centrifuged. Thus, protein was taken up into preformed particles. Conversely, in the incorporation method, 2 mL of protein solution was added to the 5 mL of FITC-chitosan solution before the TPP addition.

After stirring the chitosan/protein mixtures for 30 min, 1 mL of variably concentrated

TPP solution (at the same concentration and pH-values used in the incubation method) was added dropwise into the mixtures, and reacted for 30 min (i.e., causing the protein uptake and particle formation to occur simultaneously). Once protein uptake was completed, the XAgg-values were determined from the dissolved FITC-labeled chitosan concentration remaining after centrifugation as described above. Each condition was tested in triplicate.

3.2.5. Formulation and Uptake Method Effects on AE

Protein-loaded chitosan/TPP particles were prepared as described in the preceding section, except without FITC-labeling the chitosan. The protein concentration remaining

44 in solution after uptake was then determined using the Bradford assay (the accuracy of which was insensitive to the presence of chitosan; Appendix A, Figure A-2) after centrifuging out the particles. The protein AE was then calculated as:

i f CP - CP AE = i ×100% (3.2) CP

i where CP is the initial (or overall) protein concentration achieved after mixing the

f parent protein, chitosan and TPP solution together, and CP is the free protein concentration remaining in the supernatant after uptake and centrifugation. All AE measurements were repeated three times.

3.2.6. DLS and TEM Analysis of Protein/Dissolved Chitosan Binding

To improve understanding of protein uptake into chitosan/TPP particles, protein complexation with dissolved (TPP-free) chitosan was also probed by dynamic light scattering (DLS), using a Zetasizer Nano ZS (Malvern, UK) dynamic and electrophoretic light scattering instrument. Two mL of either 0.60 wt% BSA or 0.36 wt% α-LA solution were mixed with 5 mL 0.1 wt% chitosan solution at pH 6.0 for 30 min, whereupon

(instead of the TPP solution) 1 mL of pH-matched water was added to the chitosan/protein mixture. After 30 min of stirring, the chitosan/protein mixtures were ultracentrifuged via the procedure used for the protein uptake experiments, and the apparent volume-weighted size distributions were then obtained by DLS (for both chitosan/protein mixtures, and uncomplexed chitosan and protein controls) using a non-negative least squares (NNLS) algorithm.

45 Further, the DLS data on protein complexation with dissolved (TPP-free) chitosan was confirmed by transmission electron microscopy (TEM), using a Hitachi HD-2300 scanning transmission electron microscope. Here, chitosan/BSA mixtures (and chitosan and BSA solution controls) were prepared and centrifuged as described above. The resulting solutions were then deposited onto carbon grids (carbon type A; Ted Pella,

Redding, CA), dried for 10 min and imaged using a 200 kV acceleration voltage. To prevent dust contamination, all parent protein and chitosan solutions used for the DLS and TEM analyses were passed through 0.8-μm (Sartorius Minisart NML) cellulose acetate syringe filters before being mixed to form protein/chitosan complexes; and, as with other experiments, at least three replicate samples were used at each condition.

3.3 Results and Discussion 3.3.1 TPP:Glucosamine Ratio and pH Effects on the Micro- and Nanogel Yield

Chitosan/TPP micro- and nanogel yield depended strongly on the

TPP:glucosamine molar ratio. The XAgg increased sigmoidally with the TPP:glucosamine molar ratio (see Figure 3-1). This increase was qualitatively consistent with previous reports, where the particle yield also increased with the TPP:chitosan ratio [27, 160, 161].

When XAgg reached 1.0, however (at the 0.32:1 TPP:glucosamine molar ratio at pH 3.0; black squares in Figure 3-1), further TPP addition led to rapid particle aggregation and precipitation. This precipitation likely reflected the presence of free TPP ions which, once all the chitosan molecules were aggregated into particles, became available to bridge existing micro- and nanogels [7, 32].

46 1.0

0.8

0.6

Agg

X 0.4

0.2

0.0 0.0 0.1 0.2 0.3 0.4 TPP:Glucosamine Molar Ratio

Figure 3-1. Comparison of XAgg versus TPP:glucosamine molar ratio curves obtained by titrating TPP solutions at (■) pH 3.0, (●) pH 4.0 and (▲) pH 5.5 into 0.1 wt% chitosan solutions with matching pH-levels, and (▼) TPP solutions at their natural pH (pH 9.6) into 0.1 wt% chitosan solutions at pH 4.0. The error bars are standard deviations and the lines are guides to the eye.

The rate at which XAgg increased with the TPP:glucosamine ratio depended on the solution pH. As the pH of the parent chitosan and TPP solutions was increased from 3.0 to 5.5, the TPP:glucosamine molar ratio where the particles became fully formed decreased from 0.32:1 to 0.16:1 (Figure 3-1). This reflected changes in the chitosan and

TPP protonation. The protonation of chitosan (whose effective pKa is roughly 6.0 - 6.5

[62]) decreased as the pH was raised and diminished the cationic chitosan charge.

Conversely, TPP is a polyprotic acid with multiple pKa’s (pKa,3 = 2.8, pKa,4 = 6.5 and pKa,5 = 9.2 [165]), and became more-negatively charged at higher pH-levels. This simultaneous increase in TPP negative charge and reduction in chitosan positive charge enhanced the TPP:chitosan charge stoichiometry and decreased the TPP requirement for achieving high chitosan/TPP micro- and nanogel yields.

The pH effect was most significant as the pH was varied between 4.0 to 5.5 and was weak when the pH was varied between 3.0 and 4.0 (Figure 3-1). Additionally,

47 because the pH of parent TPP solutions is not adjusted by many groups, we also probed the evolution in XAgg upon the addition of TPP (its sodium salt) at its natural/unadjusted pH (of roughly 9.6) to a pH 4.0 chitosan solution. Not surprisingly, this yielded

XAgg-values that were intermediate to those obtained when both chitosan and TPP solutions were at pH 4.0 and 5.5 (Figure 3-1). This was because the high pH of the natural TPP solution elevated the acidic chitosan solution to higher pH-values (which, when the initial chitosan solution pH was 4.0, rose to pH 7.1 at the highest

TPP:glucosamine molar ratio).

Furthermore, even when pH-matched TPP solution was added to the chitosan solution, a pH drift occurred (Appendix A, Figure A-3). This pH drift reflected changes in the acid-base equilibria of the chitosan and TPP upon their association, which have also previously been reported in poly(allylamine)/TPP mixtures [166]. When the chitosan was not fully protonated (e.g., at pH 5.5), the complexation of its ionized amine groups with

TPP favored its further protonation (as predicted by Le Chatelier’s principle). This led to a depletion of free protons in solution and an increase in pH. Conversely, when TPP was not fully deprotonated (e.g., at pH 4.0 most TPP ions bore two protons), its binding to chitosan favored its further deprotonation (i.e., its neutralization by chitosan shifted its acid-base equilibria towards more ionized/deprotonated forms). Thus, upon binding to chitosan, TPP released protons and lowered the pH.

The extents to which these competing effects altered the mixture pH depended on the initial protonation states of the chitosan and TPP. At pH 5.5, where the chitosan was not fully protonated, the proton uptake by chitosan was more dominant than their release by TPP. The pH therefore rose from 5.5 to 6.8 as the TPP:glucosamine ratio was

48 increased from 0 to 0.16:1 (Appendix A, Figure A-3). When particles were prepared from solutions at pH 4.0, on the other hand, the chitosan was more-fully protonated and the pH dropped by 0.1 - 0.2 units (data not shown), because proton release from TPP ions outweighed their uptake by the chitosan. These pH drifts (especially the larger one that occurred when the parent solutions were at pH 5.5) might have further affected

XAgg-values – e.g., they might have promoted further particle formation when the mixture pH drifted upwards.

3.3.2. Ionic Strength Effect on the Micro- and Nanogel Yield

Several studies have demonstrated that ionic strength (i.e., monovalent salt concentration) affects chitosan/TPP particle formation and stability [2, 7, 61, 78, 79, 167].

Due to the competition between monovalent and multivalent ions, a high monovalent salt concentration inhibits chitosan/TPP binding [7]. Nonetheless, adding moderate (150 mM)

NaCl concentrations to the parent chitosan and TPP solutions appeared to have no impact on the TPP:glucosamine ratio where XAgg first reached unity (see black squares and red circles in Figure 3-2). Moreover, at lower TPP:glucosamine ratios, XAgg was actually higher in 150 mM NaCl solutions than in solutions without added NaCl (regardless of whether the parent solution pH was 4.0 or 5.5). This surprising increase in particle yield might reflect a more-uniform TPP distribution between the chitosan chains. In the absence of added NaCl, the ionic crosslinking is very rapid (and occurs within milliseconds) [7, 61]. Thus, the TPP might be consumed before being thoroughly mixed, which would limit the ionic crosslinking to only a fraction of the chitosan chains.

Conversely, in the presence of NaCl the ionic crosslinking is much slower [7, 61]. This

49 might enable more uniform mixing of TPP with the receiving chitosan solution prior to its

binding and, consequently, may lead to a higher fraction of the chitosan chains being

ionically crosslinked.

(a) (b) 1.0 1.0

0.8 0.8

0.6 0.6

Agg Agg

X 0.4 X 0.4

0.2 0.2

0.0 0.0 0.0 0.1 0.2 0.3 0.0 0.1 0.2 0.3 TPP:Glucosamine Molar Ratio TPP:Glucosamine Molar Ratio

Figure 3-2. Comparison of XAgg versus TPP:glucosamine molar ratio curves obtained from (a) pH 4.0 and (b) pH 5.5 chitosan and TPP solutions containing (■) 0 mM, (●) 150 mM and (▲) 1000 mM NaCl. The error bars are standard deviations and the lines are guides to the eye.

Another notable NaCl effect was that, unlike the particles formed in the absence

of added salt, no precipitation occurred when the TPP:glucosamine ratio exceeded the

point where XAgg first reached unity. This was consistent with previous findings that NaCl

stabilizes chitosan/TPP dispersions in the presence of excess TPP [7] and indicated that

NaCl addition might lower the risk of aggregating chitosan/TPP particles when operating

at TPP:glucosamine molar ratios that optimize the particle yield. When the monovalent

salt concentration was increased to very high levels, however (e.g., 1000 mM NaCl), the

chitosan/TPP binding was inhibited and, consistent with light scattering observations

reported earlier [7], very little chitosan aggregation occurred (see blue triangles in Figure

3-2). This suggests that XAgg-values are optimized at intermediate ionic strengths.

50 3.3.3. Protein Uptake Effects on the Chitosan/TPP Micro- and Nanogel Yield

To study the effect of protein uptake on particles yield, BSA and α-LA were used as model protein drugs. At pH-values above their isoelectric points, these proteins were negatively charged and were able to electrostatically bind to the cationic chitosan/TPP particles. This binding has been reported to yield insoluble chitosan/protein particle formation even without TPP [83, 168] and increased XAgg at every TPP:glucosamine molar ratio (see Figures 3-3a and b). Consequently, as protein was added, the

TPP:glucosamine molar ratios where XAgg reached unity and the particles started to aggregate and precipitate diminished – i.e., the onset of precipitation shifted from 0.16 to

0.13:1 TPP:glucosamine molar ratio when BSA was added (where the parent solution pH was 5.5) and from 0.11 to 0.08:1 when α-LA was added (where the parent solution pH was 6.0).

51

(a) 1.0 (b) 1.0

0.8 0.8

0.6 0.6

Agg

Agg

X 0.4 X 0.4

0.2 0.2

0.0 0.0 0.00 0.03 0.06 0.09 0.12 0.15 0.18 0.00 0.03 0.06 0.09 0.12 0.15 TPP:Glucosamine Molar Ratio TPP:Glucosamine Molar Ratio

(c) (d)

1.0 1.0

0.8 0.8

0.6 0.6

Agg

0.4 X 0.4

(Incorporation) 0.2 0.2

Agg

X 0.0 0.0 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.1 0.2 0.3 0.4 X (Incubation) TPP:Glucosamine Molar Ratio Agg

Figure 3-3. Comparison of XAgg versus TPP:glucosamine molar ratio curves obtained: (a) at pH 5.5 in the presence of (■) 0, (●) 0.05, (▲) 0.10 and (▼) 0.15 wt% BSA loaded using the incubation method; and (b) at pH 6.0 in the presence of (■) 0, (●) 0.03, (▲) 0.06 wt% and (▼) 0.09 wt% α-LA loaded using the incubation method; (c) comparison of XAgg-values obtained in the presence of BSA using the two uptake methods (incubation versus incorporation); and (d) comparison of XAgg versus TPP:glucosamine molar ratio curves obtained in the presence of (■) 0, (●) 0.05 and (▲) 0.15 wt% BSA at pH 4.0 and (▼) 0, () 0.05 and () 0.15 wt% BSA at pH 5.5. The error bars are standard deviations and the lines are guides to the eye.

52

Proteins are large molecules and their diffusion in and out of chitosan/TPP particles could be inhibited. Due to this possibility, effects of two uptake methods were compared: (1) the incubation method, where the protein was loaded into preformed particles; and (2) the incorporation method, where the particles were formed in the presence of protein and the protein was loaded during their formation (such that potential problems with protein diffusion into the particles were minimized) [3]. Surprisingly, the uptake method used had virtually no impact on the XAgg-values (see Figure 3-3c and

Appendix A, Figure A-4).

Furthermore, to probe the pH effect on particle yield in the presence of protein drugs, XAgg versus TPP:glucosamine molar ratio plots were obtained at various BSA concentrations at pH 4.0 and 5.5 (Figure 3-3d). At pH 4.0, the XAgg-values were independent of the BSA concentration and indicated that BSA had little impact on particle formation when the pH was low. This insensitivity may have reflected the positive net charge of BSA (pI ≈ 4.8 [163]) at pH 4.0, which largely eliminated its electrostatic binding to the chitosan (vide infra). Conversely, at pH 5.5 (i.e., above the pI of BSA) the negatively charged protein bound to the chitosan and, consequently, increased the XAgg (Figure 3-3d). This increase in chitosan/TPP particle yield with the chitosan-binding protein uptake (shown here for both BSA and α-LA) has also been reported to occur upon the encapsulation of insulin [169], which suggests this increase in

XAgg to be a general effect.

53 3.3.4. Protein Concentration, Uptake Method and pH Effects on AE

Protein uptake into chitosan/TPP particles was investigated using parent chitosan,

TPP and protein solutions at pH 5.5 for BSA and 6.0 for α-LA (with the pH of α-LA measurements being slightly higher due to its limited solubility at lower pH-levels).

Despite there being fewer unoccupied cationic amine groups on the chitosan chains at higher TPP:glucosamine ratios, the AE of BSA within chitosan/TPP particles rose from

10 - 20% to roughly 80 - 90% as the TPP:glucosamine molar ratio was increased from

0.025 to 0.13:1 (i.e., to the point where XAgg reached near-unity and further TPP addition led to precipitation; Figure 3-4a). As the TPP:glucosamine ratio was increased beyond this point, however, the AE dropped sharply and was again below 20% when the

TPP:glucoamine ratio exceeded 0.2:1. These BSA uptake trends were qualitatively similar to those obtained for α-LA (cf. Figures 3-4a and b). α-LA uptake, however, was less efficient with its peak AE below 60% and (just like the points where XAgg in these samples first reached near-unity; see Figures 3-3a and 3b) this peak α-LA AE occurred at a slightly lower TPP:glucosamine ratio than that for BSA. Furthermore, up to the

TPP:glucosamine ratios where the AEs reached their maximum values, the AE-values were slightly higher when higher protein concentrations were used (Figures 3-4a and b).

54 (a) (b) 100 100

80 80

60 60 (%)

(%)

40 AE 40

AE

20 20

0 0 0.00 0.05 0.10 0.15 0.20 0.25 0.00 0.05 0.10 0.15 TPP:Glucosamine Molar Ratio TPP:Glucosamine Molar Ratio

(c) (d)

100 100

) 80 80

60 60

(%)

40 40 AE

Incorporation, %

( 20 20

AE 0 0 0 20 40 60 80 100 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 AE (Incubation, %) TPP:Glucosamine Molar Ratio

Figure 3-4. AE dependence on the TPP:glucosamine molar ratio for dispersions at: (a) pH 5.5 loaded with (■) 0.05, (●) 0.10 and (▲) 0.15 wt% BSA by the incubation method; and (b) pH 6.0 loaded with (■) 0.03, (●) 0.06, (▲) 0.09 wt% α-LA by the incubation method; (c) comparison of AE-values for BSA obtained using the two uptake methods (incubation and incorporation); and (d) BSA AE dependence on the TPP:glucosamine molar ratio at (■) pH 4.0, (●) pH 5.5. The error bars are standard deviations and the lines are guides to the eye.

The initial increase in AE with the TPP:glucosamine ratio was consistent with

previous reports by Gan et al. [3] and Hu et al. [32]. These authors speculated that the

enhanced protein uptake likely reflected an increase in the negative BSA charge, which

strengthened the electrostatic chitosan/TPP binding and resulted from the pH increase

55 that occurred when alkaline sodium TPP was added to the chitosan. This pH effect on protein/particle association was also confirmed by Mattu et al., who showed that the AE of BSA increased with the dispersion pH [54]. Similarly, as discussed previously, the pH-value increased with the TPP addition even when the parent chitosan, TPP and protein solutions were at the same pH (Appendix A, Figure A-3). Yet, though the pH/increasing binding strength effect proposed in the previous studies has served as a plausible explanation for the increasing AE, data in Figures 3-1 and 3-3 suggests that it might not be the only possible interpretation. This is because XAgg, which increases the mass of chitosan/TPP particles available for protein uptake, also increases with both the

TPP:glucosamine ratio and pH. Thus, the increase in AE might also reflect an increase in

XAgg. This alternative interpretation may also explain the increase in AE with the protein content (since XAgg also increases with added protein; see Figures 3-3a and b) and will be discussed in more detail in the following section.

Once the chitosan was fully aggregated (i.e., XAgg reached near-unity) and the particles started precipitating, the AE of both BSA and α-LA decreased with the

TPP:glucoamine ratio. This was despite the continued increase in pH (cf. Figures 3-4a and b and Figure A-3) and likely reflected the strong competitive binding of TPP ions to the chitosan which, instead of forming new particles (as they did at lower

TPP:glucoamine ratios), started displacing the bound proteins from the chitosan/TPP matrix. Furthermore, consistent with observation on the protein effects on XAgg (see

Figure 3-3c), the AE-values obtained at each BSA concentration and TPP:glucosamine ratio were independent of the protein loading method used (i.e., incubation versus incorporation; see Figures 3-4c and A-3). This insensitivity suggested that protein uptake

56 was governed by the equilibrium binding of proteins to chitosan/TPP particles and, interestingly, opposed the data of Calvo et al., which showed the AE of BSA to be much higher when loaded via incorporation [39].

To further probe the role of binding on protein uptake, AE-values obtained by loading BSA (via incubation) at pH 4.0 and 5.5 were compared (see Figure 3-4d). At pH

5.5, the BSA had a net negative charge and, not surprisingly, was able to bind to chitosan efficiently. At pH 4.0, however, BSA had a net positive charge. This largely eliminated its electrostatic attraction for the chitosan and its AE remained below 10% for every

TPP:glucosamine value examined (despite the apparent chitosan/TPP particle ζ-potential at pH 4.0 being higher than that at pH 5.5; Appendix A, Table A-1). This very low protein uptake was consistent with the pH effect on XAgg (Figure 3-3d), where XAgg increased with

BSA addition at pH 5.5 but was insensitive to BSA at pH 4.0. It also confirmed that protein/particle affinity (which can be tuned by varying the pH) is essential to obtaining a high AE.

3.3.5. Relationship between AE and XAgg

In addition to affecting the protein/particle affinity, we hypothesized that the effects of the various formulation parameters on AE were strongly linked to variations in

XAgg. To investigate this relationship, the AE-values obtained at TPP:glucosamine molar ratios up to the precipitation point were plotted versus XAgg. Regardless of whether the

BSA was loaded through incubation or incorporation (cf. Figures 3-5a and b), the AE versus XAgg data obtained at each protein concentration collapsed onto a single line. A

2 linear regression on this collapsed data yielded the correlation AE = 80.2XAgg (R = 0.965)

57 2 for the incubation method and AE = 83.9XAgg (R = 0.968) for the incorporation method, confirming that the uptake method had almost no impact on AE. To test whether this nearly linear scaling applies to other chitosan-binding proteins, the same correlation was plotted for α-LA loaded using the incubation method. (Figure 3-5c). This again yielded a

2 linear function, where AE = 60.0XAgg (R = 0.955). The linear relationships obtained in each of these experiments show that, for affinity-driven protein uptake, XAgg is a key determinant of AE, irrespective of the protein concentration or loading procedure.

58

(a) 100 )

80 , %

60

40

ncubation

I (

20 AE 0 0.0 0.2 0.4 0.6 0.8 1.0

XAgg (b)

100 )

% 80

60

40

Incorporation, (

20 AE 0 0.0 0.2 0.4 0.6 0.8 1.0

XAgg (c) 100

) 80 %

60

40

ncubation,

I (

20 AE 0 0.0 0.2 0.4 0.6 0.8 1.0 X Agg

Figure 3-4. The scaling of AE with XAgg for particles loaded with (■) 0.05, (●) 0.10 and (▲) 0.15 wt% BSA via the (a) incubation and (b) incorporation method; and (c) for particles loaded with (■) 0.03, (●) 0.06 and (▲) 0.09 wt% α-LA via the incubation method. The error bars are standard deviations and the lines are linear regression fits (to data at all three protein concentrations) and the error bars are standard deviations.

59 Moreover, to test whether the increase in AE with the TPP:glucosamine ratio (and the XAgg) might also be affected by the upward pH drift (Appendix A, Figure A-3), AE and XAgg data obtained as described above was compared with that measured after lowering the dispersion pH back to the initial values of the parent chitosan, protein and

TPP solutions (pH 5.5 for BSA uptake and 6.0 for α-LA uptake). This pH reduction had little impact the AE and XAgg, and the points obtained at both pH levels fell on the same

AE versus XAgg lines (Appendix A, Figure A-5). Given that protein uptake appears to be governed by equilibrium partitioning (i.e., its insensitivity to the uptake method used), this invariance of the AE versus XAgg lines with the dispersion pH (at least when the pH exceeds the protein pI) strongly suggests that the increase in AE with the

TPP:glucosamine ratio primarily reflects the increase in XAgg and not the upward drift in the pH.

3.3.6. Model Analysis of Protein Uptake into Chitosan/TPP Particles

The trends in Figure 3-5, where (for a given XAgg-value) the fraction of protein that partitions into the particles is essentially independent of the total protein content, suggest that the protein uptake data can be described by the linear binding isotherm:

q = Keff CP (3.3) where q is the extent of protein binding (mg of bound protein per mg of particulate

chitosan), Keff is the effective binding constant and C P is the supernatant protein concentration. To confirm this, q was plotted versus by converting the AE-values into q-values as:

60 Tot CP AE q = Tot (3.4) CCS X Agg

Tot Tot where CP is the total protein concentration and CCS is the total chitosan concentration in each chitosan/TPP/protein mixture. Repeating this analysis at each TPP:glucosamine ratio and protein concentration (and neglecting the small increases in XAgg with the protein content) revealed binding isotherms for both BSA and α-LA which, at least to a first-order approximation, could indeed be modeled as linear (Figure 3-6a and b).

61

(a) (b)

2.5 1.6

2.0

1.2 )

) 1.5

0.8

mg/mg (

mg/mg 1.0

(

q q 0.4 0.5

0.0 0.0 0.0 0.3 0.6 0.9 1.2 1.5 0.0 0.2 0.4 0.6 0.8 C (mg/mL) CP (mg/mL) P (c) (d) 15 3.0

12 2.4

9 1.8

(mL/mg)

(mL/mg) 6 1.2

eff eff

K K 3 0.6

0 0.0 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 1 - X 1 - X Agg Agg

Figure 3-6. Binding isotherm data showing: (a) BSA uptake into particles with TPP:glucosamine molar ratios of (■) 0.026, (●) 0.053, (▲) 0.079, (▼) 0.106 and () 0.132:1; (b) α-LA uptake into particles with TPP:glucosamine molar ratios of (■) 0.021, (●) 0.042, (▲) 0.063 and (▼)0.084:1; and Keff versus (1 - XAgg) plots for (c) BSA and (d) α-LA. In (a, b) the dashed lines are linear fits of each data set to Equation 3.3, while the solid lines are the theoretical predictions of all data sets based on Equations 3.3 and 3.6 ~ (and fitted parameters K and K ). The lines in (c, d) are linear fits experimental data to Equation 3.6. All error bars are standard deviations.

Interestingly, the slopes of these isotherms increased with the TPP content. In other words, even when corrected for differences in particulate chitosan mass at variable

TPP:glucosamine ratios (and, accordingly, variable XAgg-values), protein partitioning into the particles was more favorable at higher XAgg-values. Thus, higher AE-values at higher

62 TPP:glucosamine ratios (and higher XAgg-values) could not be explained by an increase in protein binding sites on the particles alone. This observation suggested that the presence of unassociated chitosan might interfere with protein uptake into the particles.

Consequently, we hypothesized that the dissolved (uncrosslinked) chitosan molecules might compete for the protein molecules with the TPP-crosslinked chitosan chains within the particles.

This view is consistent with previous work on protein/polyelectrolyte interactions, which showed that soluble protein/polyelectrolyte complexes (including those based on chitosan [170, 171]) can also form [172, 173] and might even coexist with larger insoluble complexes [174]. The presence of the soluble protein/chitosan complexes was confirmed through the DLS analysis of TPP-free chitosan/BSA and chitosan/α-LA mixtures at pH 6.0. Without chitosan, the volume-weighted size distribution showed a single peak at 4 - 5 nm for BSA and 3 nm for α-LA (see Figures 3-7a and b), which were on the same order as the literature hydrodynamic diameters of those proteins (which were roughly 7 nm and 3 nm, respectively [175]). Similarly, the apparent volume-weighted size distribution obtained for protein-free chitosan solution at pH 6.0 produced a single peak with an apparent diameter of below 10 nm.

63

(a) (b) 30 30

25 25

20 20

15

15

10 10

Volume (%) Volume (%) 5 5

0 0 1 10 100 1000 10000 1 10 100 1000 10000 Hydrodynamic Diameter (nm) Hydrodynamic Diameter (nm)

Figure 3-7. Representative volume-weighted size distributions obtained by DLS for (a) 0.15 wt% BSA and (b) 0.09 wt% α-LA: (■) protein, (●) 0.063 wt% chitosan and (▲) mixtures of protein with 0.063 wt% chitosan after ultracentrifugation.

When the chitosan and proteins were mixed, however, and ultracentrifuged to remove any insoluble chitosan/protein complexes, the apparent volume-weighted size distributions obtained by DLS differed from those of the dissociated chitosan and protein molecules (green triangles in Figures 3-7a and b). The apparent size distributions in these mixtures were typically bimodal and composed of: (1) a large peak at near 20 nm; and (2) a small peak in the roughly 100 - 1000 nm range (whose coexistence was also confirmed by TEM; see Supplementary Data, Section G). Though the origin of the latter, smaller peak remains uncertain, the presence of these peaks (at apparent diameters larger than those of chitosan and protein alone) strongly suggests the formation of “soluble” protein/chitosan aggregates that are not removed from the solvent through ultracentrifugation. Likewise, the dominant peak at dH ≈ 20 - 30 nm in the DLS data agrees well with the hydrodynamic diameters of soluble protein/polyelectrolyte complexes reported in the previous works [174, 176]. The presence of these small aggregates confirms that, when XAgg is smaller than 1.0, competitive protein binding to

64 uncrosslinked chitosan chains likely limits protein uptake.

The inference of this competitive protein binding enables further modeling of the protein uptake data, which might explain the near-linear scaling of AE with XAgg. First, it indicates that the supernatant protein concentration measured by UV-Vis spectroscopy reflects the combined concentrations of: (1) free unbound protein; and (2) protein bound to soluble chitosan chains (i.e., protein bound to chitosan chains that are not removed through ultracentrifugation). Accordingly, the linear partitioning model can be modified to account for the competitive binding effect by expressing the total supernatant protein concentration as:

~ ~~ Tot CP = CP + KCPCCS (1 - X Agg ) (3.5)

~ ~ where CP is the free (unbound) protein concentration, K is the equilibrium constant

~~ Tot for the protein/dissolved chitosan binding, and the product KCPCCS (1- X Agg ) quantifies the concentration of protein bound to the soluble chitosan. This analysis again assumes

~ that the mass of protein bound per mass of soluble chitosan, q , can be approximated by

~ ~~ the partitioning expression q = KCP , while the mass of protein bound per mass of

~ particulate chitosan can be approximated as q = KCP , where K is the equilibrium constant for the protein binding to the particulate chitosan. Combining the latter q definition with Equation 3.3 and 3.5 yields an expression for Keff:

K Keff = ~ Tot (3.6) 1+ KCCS (1- X Agg )

65 These Keff-values can be experimentally obtained by fitting the binding data obtained at each TPP:glucosamine ratio (dashed lines in Figures 3-6a and b) to Equation 3.3.

Furthermore, once the Keff variation with the XAgg (which depends on the ~ TPP:glucosamine ratio) is known, the magnitudes of K and K can be obtained.

This was easily achieved by plotting Keff versus (1- XAgg) as shown in Figures 3-6c and d, and fitting this data to Equation 3.6 (see Table 3.1). These fits revealed the magnitudes of the K- and - values to be very close to one another (i.e., within 30% of one another for both protein species) and indicated that, while XAgg remained below 1.0,

TPP did not hinder protein binding to the chitosan. Furthermore, the fitted values of K and for both proteins suggest that the linear partitioning model given by Equations

3.3 and 3.6 can (at least at typical protein loading conditions) describe the experimental binding isotherms quite well – i.e., protein uptake data at every TPP:glucosamine ratio can be fitted using K and as the only adjustable parameters (see solid lines in

Figures 3-6a and b).

Table 3.1. Fitted equilibrium constants for BSA and α-LA binding to chitosan/TPP particles and soluble chitosan.

~ ~ Protein K (mL/mg) K (mL/mg) K/K R2

BSA 13.6 18.0 0.76 1.000

α-LA 2.72 2.72 1.00 0.987

Based on the above analysis, it is tempting to derive a binding isotherm-based

66 equation that predicts the linear AE versus XAgg correlations in Figure 3-5. Since AE is defined as the protein fraction that becomes loaded into the particles, it can be rewritten as:

Tot Keff CCS X Agg AE = Tot (3.7) 1+ Keff CCS X Agg

Tot which is obtained by expressing the encapsulated protein concentration as Keff CCS X Agg CP

Tot (or q CCS X Agg ) and the unencapsulated protein concentration as CP . Since Keff is a function of XAgg, the linearity of this expression can be checked by substituting Equation

3.6 for Keff, which yields:

Tot KCCS X Agg AE = Tot ~ ~ (3.8) 1+CCS [K +(K - K)X Agg ]

~ Though this expression is not strictly linear, it becomes so when K and K are roughly ~ equal like they are in Table 3.1, and (for 0.6 ≤ K/K ≤ 1.5; see Appendix A) simplifies

Equation 3.8 to the linear expression:

Tot KCCS AE = Tot X Agg (3.9) 1+ K' CCS where K’ is the weighted average binding constant (defined as K’ = 0.75K + 0.25 ) and is used in the denominator instead of K to minimize errors caused by small differences between K and (see Appendix A.9 for the derivation). This simplified linear expression is consistent with the shapes of the AE versus XAgg correlations in Figure 3-5, where the slope obtained for the stronger-binding BSA was higher than for the weaker-binding α-LA. Moreover, the slope of the AE versus XAgg line defined by this

67 equation increases with the protein/chitosan binding strength and the total chitosan ~ concentration, and (for systems where K = K ) approaches the value of one (or 100%)

T ot when K' CCS >> 1 (Appendix A, Figure A-7).

Notably, the sensitivity of this slope to the total chitosan concentration (which is varied in many studies) depends on the protein/chitosan binding strength. When

Tot Tot K' CCS >> 1, this slope is predicted to be practically independent of CCS . Conversely, at

Tot lower K' CCS -values, the slope is predicted to become more sensitive to . These model predictions were confirmed experimentally by varying the parent chitosan solution concentration between 0.05 and 0.15 wt% (such that, upon the addition of TPP and protein solution ranged between 0.31 and 0.94 mg/mL). The AE versus XAgg curves obtained for the stronger-binding BSA (for which ranged from 4.59 to 13.8) showed almost no variation, just as predicted by Equation 3.9 (see Figure 3-8a). Those for the weaker-binding α-LA, however (for which ranged from 0.85 to 2.55), increased noticeably with the overall chitosan concentration (Figure 3-8b).

68

(a) (b)

100 100

80 80 )

) 60 60 %

% (

(

40 40 AE

AE

20 20

0 0 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 X X Agg Agg

Figure 3-8. The scaling AE with XAgg for particles prepared using (■) 0.31, (●) 0.63 and (▲) 0.94 mg/mL overall chitosan concentrations and loaded via the incubation method with (a) 0.05 wt% BSA and (b) 0.03 wt% α-LA. The error bars are standard deviations and the lines are the model predictions based on Equation 3.9 for mixtures containing (—) 0.31, (‒ ‒) 0.63 and (····) 0.94 mg/mL chitosan, respectively.

In addition to relating AE to XAgg, the above analysis allows q (i.e., the amount of

protein taken up per mg of particulate chitosan, or the particle “loading capacity”) to be

Tot Tot predicted as a function of CP and CCS . This can be achieved by combining Equation

~ 3.4 with either Equation 3.8 or 3.9. For instance, if K and K have similar values,

Equation 3.9 can be used to yield:

Tot KCP q = Tot (3. 10) 1+ K' CCS

which predicts the concentration of protein within the particles to increase with the total

protein concentration used in the encapsulation process and the protein/chitosan binding

strength, and to decrease with the total chitosan concentration used. Notably, because for

equal K and the protein distributes itself evenly between dissolved and particulate

69 chitosan (and q is already normalized by the chitosan mass within the particles), q is

Tot predicted to be nearly independent of XAgg when plotted as a function of CP . This linear

Tot dependence of q on CP is confirmed in Figures 3-9a and b, where the data for each protein from Figure 3-6 (BSA in Figure 3-9a and α-LA in Figure 3-9b) collapses onto the lines predicted by Equation 3.10.

(a) (b)

2.5 1.6

2.0

1.2 )

) 1.5

0.8

mg/mg ( mg/mg 1.0 (

q q 0.4 0.5

0.0 0.0 0.0 0.3 0.6 0.9 1.2 1.5 1.8 2.1 0.0 0.2 0.4 0.6 0.8 1.0 1.2 Tot Tot CP (mg/mL) CP (mg/mL) Figure 3-9. Plots of protein uptake as function of total protein used during the encapsulation procedure for: (a) BSA loaded at (■) 0.026, (●) 0.053, (▲) 0.079, () 0.106 and (▼) 0.132:1 TPP:glucosamine molar ratios; and (b) α-LA loaded at (■) 0.021, (●) 0.042, (▲) 0.063 and ()0.084:1 TPP:glucosamine molar ratios. Both plots are obtained based on data in Figure 3-6. The lines are model predictions obtained using Equation 3.10 and the error bars are standard deviations.

3.3.7. Further Discussion

The sensitivity of XAgg to the TPP:glucosamine ratio, pH, ionic strength and protein concentration (Figures 3-1 to 3-3) clearly shows chitosan/TPP micro- and nanogel yield depends strongly on the compositions used during their preparation. This indicates that, when these mixture compositions are adjusted (e.g., to optimize particle size distribution or stability), their effect on particle yield must be taken into account. Further,

70 consistent with previous studies on how particle yield varies with the TPP:glucosamine molar ratio [27, 160, 161], it confirms that the assumption that nearly all of the chitosan self-assembles into particles (which not infrequently appears in the literature) is often incorrect and should be used with caution.

More importantly, this study clearly shows that protein AE depends strongly on the chitosan/TPP micro- and nanogel yields, and that XAgg should be optimized to ensure high AE-values. It also provides possible explanations for many of the trends that have been reported previously. Contrary to the literature explanations discussed in the

Introduction [3, 27], the primary reason for the increase in AE with the TPP:glucosamine ratio [3, 27, 48, 161] appears to be the increase in XAgg. Similarly, the reduction in AE with the chitosan concentration [3, 33] likely reflects a reduction in XAgg, which occurs regardless of whether the chitosan concentration is increased at a constant TPP concentration [33] or a constant TPP:glucosamine ratio [3]. When the TPP concentration is kept constant, XAgg decreases because the TPP:glucosamine ratio decreases. Likewise, when the chitosan is increased at a constant TPP:glucosamine ratio, XAgg likely decreases due to a decrease in the BSA:chitosan ratio (i.e., as the number of BSA molecules binding to each chitosan chain decreases the protein-mediated increase in XAgg becomes less pronounced).

The observations made in this study also explain the opposing findings reported by Yang et al., where the AE first sharply increased with the chitosan concentration, and then decreased with the chitosan concentration when the chitosan concentration became high [162]. In Yang’s work, the BSA:chitosan ratio was kept constant and the BSA and chitosan concentrations both increased while the TPP concentration was fixed [162].

71 Thus, the rise in chitosan and BSA concentrations corresponded to a decrease in the

TPP:glucosamine molar ratio. At low chitosan/BSA concentrations, the TPP:glucosamine molar ratio was high and the chitosan was fully assembled into particles (i.e., XAgg was

1.0). Under these conditions, AE decreases with the TPP:glucosamine molar ratios (cf.

Figures 3-3a and 3-4a). Hence, not surprisingly, the AE increased as the TPP:glucosamine ratio decreased with the chitosan/BSA concentrations. At high chitosan/BSA concentrations, however, TPP:glucosamine ratios became much lower, such that XAgg was likely lower than 1.0. Under these conditions, both XAgg and AE increase with the

TPP:glucosamine ratio (Figures 3-3a and 3-4a). Thus, as TPP:glucosamine ratio decreased with the chitosan/BSA concentrations at high chitosan/BSA concentrations, AE also decreased.

The effect of added protein on XAgg (see Figure 3-3) also explains the increase in

AE that has been reported to occur at higher protein concentrations [3, 36, 47, 48, 160].

Yet, there are also several reports that indicate an opposite trend [1, 33-35, 39]. The easiest of these opposing results to explain is that by Jarudilokkul et al. [34], who used a

TPP:glucosamine molar ratio well above 0.2:1. Under this condition, XAgg had already reached unity and TPP ions were displacing protein from the particles (see Figure 3-4a and b). Thus, at this high TPP:glucosamine ratio, further protein addition did not enhance the particle yield and AE decreased with the protein concentration due to a saturation of available chitosan binding sites.

In contrast, the other four studies that showed AE to decrease with the protein concentration used at TPP:glucosamine ratios of around 0.1:1, where Figure 3-3 shows protein addition might still increase XAgg-values. Interpretation of these opposing

72 literature results, however, is unfortunately complicated by the fact that none of these studies provide complete information on the pH-values of their parent chitosan and TPP solutions, and final chitosan/TPP particle dispersions. Since pH is a critical determinant of XAgg (see Figure 3-1), this makes it difficult to postulate on how XAgg changed with the protein concentrations and to explain the trends reported in these works. We therefore emphasize that detailed pH information is essential to include in reports on chitosan/TPP particles. Another cautionary note on performing XAgg and AE analyses is on selecting proper centrifugations procedure for removing the particles from the supernatant. Though many groups report centrifuging the dispersions for 30 min at 20,000 to 40,000 g, XAgg measurements at variable centrifugation speeds and times (data not shown) revealed that nearly-quantitative particle removal requires much more extensive centrifugation (such as described in Section 2.3).

More broadly, the findings reported herein highlight the need to carefully analyze the effects of formulation parameters on chitosan/TPP micro- and nanogel yield and show

(both experimentally and through model analysis) that this particle yield is a key determinant of AE. The relationship between AE and XAgg revealed in this work provides essential guidelines for optimizing protein uptake in these colloidal drug carriers, explains hitherto poorly-understood literature results, and can likely be extended to protein encapsulation within other ionically crosslinked colloidal drug carriers (e.g., those based on alginate [177]).

3.4. Conclusions

This study has systematically examined the relationship between chitosan/TPP

73 micro- and nanogel yields and their protein uptake properties. Spectroscopic analysis of chitosan aggregation confirmed chitosan/TPP particle yields to be acutely sensitive to the formulation parameters used in their preparation; where, in addition to increasing with the TPP:glucosamine ratio [27, 160, 161] and added protein [169] as reported previously, particle yields (which were defined here as the fractions of aggregated chitosan, or XAgg-

-values) unambiguously increased with the pH, and varied (nonmonotonically) with the ionic strength. This variability indicated that, when formulation parameters are optimized

(e.g., to tune the particle size or stability), their effects on the particle yield must be carefully considered.

Besides affecting the overall particle preparation efficiency, chitosan/TPP particle yields were discovered to have a critical impact on their protein uptake performance. For chitosan-binding proteins, the AE increased almost linearly with the particle yield (up until the point where all chitosan aggregated into particles), indicating that, to maximize protein uptake, the particle yield must also be maximized. Such favorable particle yields and AE-values were easily achieved by raising the TPP:glucosamine molar ratio. When the TPP:glucosamine ratio exceeded that needed to aggregate all the chitosan into particles, however (such that XAgg = 1.0), the excess TPP displaced the loaded protein from the particles and (due to the strong, competitive chitosan/TPP binding) caused the

AE to decrease with further TPP addition. This revealed that there is an optimal

TPP:glucosamine molar ratio for protein uptake which, for the chitosan used for this work (i.e., 86% DD chitosan with a viscosity-average molecular weight of 155 kDa), is near the 0.13:1 TPP:glucosamine molar ratio when the particles are formed from parent chitosan, TPP and protein solutions at pH 5.5, and near the 0.08 - 0.09:1

74 TPP:glucosamine molar ratio when the particles are formed from solutions at pH 6.0.

The nearly linear relationship between the particle yields and AE-values (which occurred when XAgg < 1.0), on the other hand, apparently reflected two effects: (1) that higher particle yields provided more particulate binding sites for the protein uptake; and

(2) that the dissolved chitosan (which was not crosslinked into particles) competed with the chitosan/TPP particles for the protein. Remarkably, the AE-values obtained at variable

TPP:glucosamine ratios and protein concentrations collapsed onto a single AE versus XAgg curve, whose insensitivity to the protein loading procedure suggests that protein uptake is governed by simple equilibrium partitioning. As a first order approximation, these near-linear AE versus XAgg relationships can be modeled based on linear binding isotherms, which successfully predict variations in both AE and particle protein content.

Overall, these findings show that knowledge of the chitosan/TPP micro- and nanogel yield is essential to understanding and optimizing their protein uptake performance, explain many of the trends in the literature AE data, and can likely be extended to other ionically crosslinked colloids that are currently being explored.

75

Chapter 4

Formation and Dissolution of Ionically Crosslinked

Chitosan Particles: Analysis of Ionic Crosslink

Reversibility

4.1. Introduction

Chapter 3 has shown that drug uptake into ionically crosslinked chitosan particles can occur during the particle formation. In contrast, drug release from these particles during their administration may be accompanied by particle dissolution. A low particle dissolution stability is undesirable because it can lead to a burst drug release. Thus, a better understanding of ionically crosslinked chitosan particle dissolution is important to attain. However, while some literature studied the dissolution of the entire ionically crosslinked chitosan particle matrix [2, 78, 85], little has been done to investigate particle dissolution due to ionic crosslinker leaching from the particles, when ionically crosslinked particles are placed into an excess of crosslinker-free solution (e.g., during

76 their application).

To begin addressing these issues, this chapter: (1) provides systematic guidelines for the preparation of chitosan/PPi nanoparticles, which (as shown herein) behave dissimilarly from those prepared using the more-common TPP; and (2) investigates the dissolution of ionically crosslinked chitosan particles. This second objective was pursued using PPi instead of the more-common TPP because chitosan/PPi binding was much weaker than chitosan/TPP binding [178] (i.e., because the very strong chitosan/TPP binding made TPP elution challenging to achieve [179], the PPi-based particles provided a more-tractable experimental system). Based on these considerations, this study consisted of three parts. First, dynamic and electrophoretic light scattering, scanning transmission electron microscopy (STEM), stopped-flow turbidimetry and viscometry were used to investigate the compositional parameters affecting the formation, size distributions and colloidal stability of chitosan/PPi nanoparticles. Second, light scattering was used to investigate particle dissolution that occurred when PPi was leached from the complex. Third, the irreversible formation/dissolution behavior revealed by these experiments was qualitatively modeled using the Bragg-Williams theory. Most of the text and all of the data in this chapter was reproduced from: Cai, Y.; Lapitsky, Y. Colloids Surf

B.. 2014, 115, 100.

77 4.2. Materials and Methods

Low molecular weight chitosan (nominal MW = 50 - 190 kDa), sodium pyrophosphate and acetic acid were purchased from Sigma-Aldrich (St. Louis, MO) and used as received. The degree of chitosan deacetylation was 90% [7]. All experiments were performed using 18.2 MΩ·m Millipore Direct-Q 3 deionized water at pH 4.0 (using acetic acid to adjust the pH) and, unless otherwise indicated, at room temperature.

To prepare the particles, 10 mL of 0.01 - 0.20 wt% chitosan solution (0.55 mM -

11 mM in its cationic monomer units) were stirred at 800 rpm with a cylindrical magnetic stir bar (12 mm × 4 mm) and titrated with 0.37 wt% (14 mM) PPi solution. The PPi solution was added in 50 to 200-µL increments, at a rate of two drops per second. After each PPi addition the mixtures were equilibrated for 10 min, whereupon light scattering and ζ-potential analysis was performed to detect particle formation and characterize the evolution in particle size, charge and colloidal stability. These measurements were performed using a Zetasizer Nano ZS (Malvern, UK) dynamic and electrophoretic light scattering instrument, wherein the light scattering intensities were reported as derived count rates (i.e., scattering intensities that would be obtained if each measurement was conducted at full laser power [61]) to facilitate the comparison between strongly and weakly-scattering samples.

The particle sizes determined by dynamic light scattering (DLS) were confirmed by STEM, using a Hitachi HD-2300 microscope. Here, the particle dispersions prepared from 0.03 and 0.10 wt% chitosan solutions (titrated to 2.5 and 3.6 mM PPi, respectively) were deposited onto carbon grids (carbon type A; Ted Pella, Redding, CA). The samples were then dried for 10 min and imaged using a 200 kV acceleration voltage. Further

78 characterization of the particle dispersions was performed via capillary viscometry (at

25 °C), using a Cannon-Ubbelohde Dilution viscometer (Size 75; Cannon Instrument Co.,

State College, PA). Additionally, the effect of PPi concentration on the nanoparticle formation kinetics was probed by stopped-flow turbidimetry, using a DM 45K stopped-flow spectrophotometer (Olis Inc., Bogart, GA). Here, 0.10 wt% (5.5 mM) chitosan and 0.18 - 0.55 wt% (6.9 - 21 mM) PPi solutions were rapidly mixed in the stopped-flow sample cell in a 5:1 chitosan:PPi solution volume ratio. The formation kinetics were then inferred from the evolution in turbidity (where the formation of particles yielded more scattered light).

To examine particle dissolution (and the reversibility of the chitosan/PPi particle formation), a two-step experiment was performed. In the first, “forward titration” step,

10-mL samples of 0.10 wt% (5.5 mM) chitosan solution were titrated with 13.7 mM PPi solution to final PPi concentrations ranging between 0.2 and 2.0 mM. The titrations were performed as series of 200-µL additions (conducted at a rate of 2 drops per second) separated by 1-min time intervals. The light scattering from these samples was then tracked for 29 d. Each mixture was stirred for the first day after preparation, and was then stored at rest.

In the second, “backward dilution” step, samples containing 2.0 mM PPi were first prepared as described in the forward titration step. After stirring for 1 d, however, these samples were diluted with 0.10 wt% chitosan solution to final PPi concentrations ranging between 0.2 and 1.7 mM, so that the “forward titration” and “backward dilution” samples had matching chitosan and PPi concentrations. Because the particles dissolved at low PPi concentrations, this allowed the determination of whether the particle

79 formation/dissolution cycle was thermodynamically reversible. As in the forward titration, the diluted dispersions were stirred for the first day after preparation, and then stored at rest. To monitor their dissolution, the evolution in their light scattering intensity was tracked for 28 d, and characterized at the same times as the “forward titration” samples

(so that the time elapsed since the initial PPi addition to chitosan was the same for both the “forward titration” and “backward dilution” samples). The sample analysis was limited to the 4 weeks after dilution, to reduce the effects of chitosan and PPi hydrolysis and, to ensure reproducibility, this experiment was repeated thrice.

4.3. Results and Discussion 4.3.1. Formation and Characterization of Chitosan/PPi Particles 4.3.1.1. Formation of Chitosan/PPi Particles

The formation of chitosan/PPi particles was determined from the changes in light scattering intensity during the titration of PPi into chitosan solutions. At low PPi concentrations (where no particles formed) the light scattering intensity remained similar to that of PPi-free chitosan solutions. At the onset of particle formation, however, the scattering intensity began to suddenly increase (see Figure 4-1a). Finally, at even higher

PPi concentrations, where the particle formation process ended, the light scattering intensity plateaued. These trends occurred at each chitosan concentration and were consistent with those recently reported by Huang and Lapitsky for PPi mixtures with 0.1 wt% chitosan [178]. By extending these measurements to other chitosan concentrations, a phase map was constructed to indicate the compositions where submicron particles began forming during the titrations (white squares in Figure 4-1b). This revealed that at higher chitosan concentrations more PPi was needed to form particles.

80

(a) (b)

10 1.0

0.8 8 S 0.6 6

0.4 4 D

0.2 2 Chitosan Concentration (mM) Normalized Scattering Intensity Scattering Normalized 0.0 0 0 1 2 3 4 5 6 0.0 0.4 0.8 1.2 1.6 2.0 2.4 PPi Concentration (mM) PPi Concentration (mM)

Figure 4-1. The onset of chitosan/PPi particle formation illustrated by: (a) the evolutions in normalized light scattering intensity obtained during PPi titrations into () 0.03 wt% () 0.10 wt% and () 0.20 wt% chitosan solutions; and (b) a phase map illustrating the transitions from (S) molecular solutions to (D) colloidal dispersions that occurred () during the PPi titrations and () after long-term equilibration. The lines are guides to the eye and the error bars indicate the uncertainty in the phase boundary.

Over longer timescales, particle formation occurred at lower PPi concentrations.

Thus, despite its practical utility in guiding particle preparation, the phase boundary indicated by the titration experiment reflected only the short-term (kinetic) phase behavior. To obtain the pseudo-equilibrium phase boundary (the thermodynamics of which will be explained in Section 3.3), additional chitosan/PPi mixtures were prepared and equilibrated until the phase behavior ceased to change with time. This yielded the near-linear pseudo-equilibrium phase boundary shown by the blue circles in Figure 4-1b, which indicated the real minimum PPi concentrations required for particle formation. The substantial PPi concentration that was required to form chitosan/PPi micro- and nanoparticles was in stark contrast to the chitosan/TPP system, where colloidal particles formed even at very low TPP concentrations [61, 178].

81 4.3.1.2. Particle Size and Composition

The evolutions in the z-average hydrodynamic diameters of chitosan/PPi particles during the titration experiments were quantified through the cumulant analysis of DLS data [180]. At PPi concentrations near the phase boundary in Figure 4-1b, the particle size increased with the PPi concentration (see Figure 4-2a). Once the particles were fully formed, however (i.e., the light scattering intensity in Figure 4-1a plateaued), their z-average hydrodynamic diameters were insensitive to the PPi concentrations (Figure

4-2a). Conversely, the particle size increased significantly with the chitosan concentration, from roughly 60 to 220 nm as the parent chitosan solution concentration was raised from

0.01 to 0.20 wt% (Figure 4-2a). This increase in particle size was confirmed by STEM, where the particles prepared from 0.03 wt% chitosan solutions were significantly smaller than those prepared from 0.10 wt% chitosan solutions (compare Figure 4-2bi and 4-2bii) and both particle sizes were similar to those measured by DLS. When the average particle volumes were estimated from the z-average diameters obtained via DLS (blue diamonds in Figure 4-2c), they revealed a near-linear power law scaling with the chitosan

1.25 concentration (V ~ Cc ), indicating that the size of chitosan/PPi particles can be predictably tuned. Moreover, the polydispersity indices of these fully-formed particles – measured by DLS (data not shown) – ranged between approximately 0.10 and 0.20, thus suggesting that the size distributions of chitosan/PPi micro- and nanoparticles were fairly-narrow.

82

(a) (b)

240 (i) (ii) 200

160

120

80 0.01 wt% CS 0.03 wt% CS 40 0.05 wt% CS 0.10 wt% CS

Hydrodynamic Diameter (nm) Diameter Hydrodynamic 0.20 wt% CS 0 0 1 2 3 4 5 6 PPi Concentration (mM)

(c) (d)

7 0 10 10

)

3

nm 6 -1 ( 10 V ~ C1.25 10 C

1.46 5  ~ C -2 10 C 10

Particle Volume Volume Particle

Fraction Volume Particle Chitosan chain PPi 104 10-3 0.01 0.1

Chitosan Concentration (wt%)

Figure 4-2. The effect of the parent chitosan solution concentration on: (a) the z-average hydrodynamic diameter evolution during the titration of PPi into () 0.01 wt%, () 0.03 wt%, () 0.05 wt%, () 0.10 wt%, and () 0.20 wt% chitosan solutions (detected by DLS); (b) the size distributions of particles prepared using (i) 0.03 wt% and (ii) 0.1wt% chitosan solutions (imaged by STEM; scale bar = 600 nm); (c) the () average particle volume and () particle volume fraction. The solid lines are power law fits while the dashed lines are guides to the eye. The drawing (d) shows the proposed structure of a solvent-swollen chitosan/PPi nanoparticle.

In addition to estimating the particle size, the effect of chitosan concentration on the particle volume fraction,  (white squares in Figure 4-2c), was estimated for chitosan concentrations ranging between 0.05 and 0.20 wt% (where the particle concentration was high enough to noticeably affect the viscosity). This was achieved by capillary

83 viscometry using [180]:

 0  1 2.5 (4.1)

where η and η0 represent the dispersion and solvent viscosities. The dispersions were prepared using PPi concentrations where, according to the plateaus in the light scattering intensities and hydrodynamic diameters (Figure 4-1a and 4-2a), the particles were fully formed (see Table 1). Here, the variation in  exhibited a power law scaling with Cc ( ~

1.46 Cc ), which was only slightly higher than that of V. These similar scaling relationships indicated that: (1) the concentration of particles, Cnp, roughly estimated as Cnp = /V ~

O(1012 - 1013 particles/ml), was insensitive to the chitosan concentration; and (2) the number of chitosan chains in each particle – i.e., the aggregation number, N, where N =

(Ĉc NA) /( Cnp Mn) ~ O(100 - 1000) – increased considerably with the parent chitosan solution concentration (where NA was Avogadro’s number, Mn was the chitosan molecular weight, and Ĉc was the final chitosan concentrations in Table 4.1, obtained after the addition of PPi). Additionally, this viscometric analysis allowed the estimation of the

chitosan concentrations within the particles, Cc,np = Ĉc / ~ O(1 - 10 wt%), which were comparable to those recently estimated for chitosan/TPP particles [181] and indicated that the particles had water-swollen, gel-like structures as depicted in Figure 4-2d.

84 Table. 4.1. Chitosan/PPi dispersions analyzed by capillary viscometry. The final chitosan concentrations reflect the dilution of parent chitosan solutions during the titration of PPi. The uncertainty in the η/η0-values represents the 95% confidence limits based on two replicate samples (each of which was measured thrice).

[Initial Chitosan] [Final PPi] [Final Chitosan] Relative Viscosity

wt% (mM) wt% (mM) wt% (mM) (η / η0)

0.05 (2.8) 0.07 (2.8) 0.04 (2.2) 1.014 ± 0.007

0.10 (5.5) 0.09 (3.5) 0.07 (4.1) 1.034 ± 0.000

0.15 (8.3) 0.10 (3.8) 0.11 (6.0) 1.060 ± 0.011

0.20 (11.0) 0.10 (3.9) 0.14 (7.9) 1.112 ± 0.025

4.3.1.3. Particle Colloidal Stability

mportantly, unlike the frequently-studied chitosan/TPP particles, which rapidly

coagulated and precipitated when TPP:chitosan molar ratio exceeded 0.2:1 [1, 7], the

chitosan/PPi particles remained stably dispersed even when the PPi was in multifold

excess (see Figure 4-2a). This suggests that, unlike the TPP ions, which coagulate the

particles through ionic bridging [7], the PPi ions do not readily bridge chitosan particles.

This interpretation is supported by the ability of lyophilic chitosan/PPi particles to remain

stably dispersed even when their ζ-potentials drop below 25 mV (i.e., the ζ-potential at

which the chitosan/TPP particles rapidly coagulate; see Appendix B, Figure B-1).

Accordingly, it is surmised that the colloidal stability of chitosan/PPi particles stems from:

(1) their lyophilicity (which prevents their coagulation in the absence of ionic bridging

[24]); and (2) the apparent inability of PPi to bridge them. Like the chitosan/TPP particles,

however [182], these particles become colloidally unstable in phosphate buffered saline

85 (PBS) at pH 7.4. This stems from the insolubility of chitosan at the slightly-alkaline, physiological pH, which causes the lyophilic particles to become lyophobic (and leads to their rapid coagulation). The insolubility of molecular chitosan in PBS and rapid particle coagulation confound the analysis of the ionic crosslink stability under physiological conditions (i.e., precipitated chitosan aggregates form regardless of whether the ionic crosslinks are intact). Hence, the particle dissolution experiments in this work were performed under buffer-free, acidic conditions (at pH 4.0).

4.3.1.4. Particle Formation Kinetics

To examine the kinetics and (as shown later) the reversibility of chitosan/PPi particle formation, chitosan/PPi mixtures were prepared using 0.10 wt% (5.5 mM) chitosan solutions, titrated to various PPi concentrations (ranging between 0.2 and 2.0 mM). The temporal changes in their light scattering intensities were then monitored over

29 d. At PPi concentrations above 1.4 mM, where particles formed during the titration

(see Figure 4-1), there was an immediate increase in the light scattering intensity (white squares in Figure 4-3a). This increase persisted (albeit at a much slower rate) throughout the 29-d experiment. Conversely, at PPi concentrations below 1.0 mM (i.e., below the pseudo-equilibrium phase boundary in Figure 4-1b) no particles formed even after 29 d of equilibration (Figure 4-3a). Finally, at the PPi concentration of 1.1 mM (which was only slightly above the pseudo-equilibrium phase boundary), the particles became detectable only after 1 d, suggesting that the particle formation kinetics are sensitive to the PPi concentration and, at lower PPi concentrations, can be rather slow.

86

(a) (b)

1.2x105 0.8 10 min 1 h 1.0x105 1 d ) 0.6 2 d -1

8.0x104 8 d cm

15 d ( 29 d

6.0x104 0.4

4.0x104

0.2 Turbidity 2.0x104

Derived Count Rate (kcps) 0.0 0.0 0.0 0.4 0.8 1.2 1.6 2.0 0 3000 6000 9000 12000 15000 PPi Concentration (mM) Time (s)

Figure 4-3. Evolution in the light scattering intensity analyzed via (a) DLS, where the derived light scattering intensity is plotted versus PPi concentration, and (b) stopped-flow turbidimetry, where the evolution in the turbidity is detected in 0.08 wt% (4.6 mM) chitosan solutions containing () 1.2, () 1.5, () 1.7 and () 2.3 mM PPi. The lines are guides to the eye.

This kinetic sensitivity was confirmed by stopped-flow turbidimetry. At the lowest, 1.2 mM PPi concentration, the turbidity (which reflected the extent of particle formation) only started to rise after a 2000-s induction period, and slowly increased throughout the 14,000-s experiment (blue diamonds in Figure 4-3b). At higher PPi concentrations, however, the induction time decreased dramatically and disappeared completely at 2.3 mM PPi (red circles in Figure 4-3b). Likewise, as the PPi concentration was increased from 1.2 to 2.3 mM, the slope in the turbidity curve increased by an order of magnitude. This sensitivity to ionic crosslinker concentration was consistent with that observed for the chitosan/TPP system [61]. The kinetics here, however, were significantly slower than those for the TPP-based particles (which were mostly formed within milliseconds of mixing under similar conditions [61]).

87 4.3.2. Dissolution of Chitosan/PPi Particles

To examine the dissolution of chitosan/PPi nanoparticles, the particles were first prepared as described above – i.e., by titrating TPP into 0.1 wt% (5.5 mM) chitosan solutions – with final PPi concentrations of 2.0 mM. After 24 h of equilibration, however, these dispersions were diluted to various PPi concentrations (ranging between 0.2 and 1.7 mM) in PPi-free 0.10 wt% chitosan solution, so that the mixture compositions precisely matched those in Figure 4-2a. This enabled the analysis of particle dissolution and the reversibility of the particle formation/dissolution cycle.

If the nanoparticles remained stably intact upon their dilution, their light scattering intensity was expected to scale linearly with the PPi concentration (i.e., obeying Beer’s Law as shown by the dashed line in Figure 4-4a). This relationship was initially obeyed at PPi concentrations above 1 mM (red circles in Figure 4-4a) – i.e., above the pseudo-equilibrium phase boundary. At lower concentrations, however, the light scattering intensity fell below this line, thus indicating particle dissolution. When equilibrated over longer timescales, light scattering from the diluted samples with PPi concentrations below 1 mM diminished further (indicating further dissolution), and reached a roughly-constant value after 1 d (see Figure 4-4b). Conversely, the scattering from samples with PPi concentrations significantly above 1 mM continued to increase with time (Figure 4-3a), thus indicating additional aggregation. At 1.1 mM PPi, which was near the pseudo-equilibrium phase boundary, the scattering intensity remained roughly constant (see Figure 4-4a and 4-4b).

88

(b) (a) 1.0x105 10 min 1.0 1 h 4 1 d 8.0x10 7 d 0.8 14 d

4 28 d

6.0x10 0.6 0

4 I(t)/I 4.0x10 0.4

4 2.0x10 0.2

Derived Count Rate (kcps) Rate Count Derived 0.0 0.0 0.0 0.4 0.8 1.2 1.6 2.0 0 50 100 150 200 250 300 350 PPi Concentration (mM) Time (h)

Figure 4-4. Dilution effects on the light scattering intensity from preformed chitosan/PPi particles showing: (a) the derived count rate measured at various times after dilution plotted versus the PPi concentration; and (b) the light scattering intensity (normalized to that predicted by Beer’s Law) from particles diluted to PPi concentrations of () 0.2 mM, () 0.5 mM, () 0.8 mM and () 1.1 mM plotted versus time. The solid lines are guides to the eye and the dashed lines indicate the hypothetical light scattering intensity in the absence of particle dissolution (or further aggregation). The error bars are standard deviations (n = 3).

Interestingly, although the three lowest PPi concentrations were below the

particle formation phase boundary (see Figure 4-1b), particle dissolution at these

concentrations was incomplete. Indeed, the normalized light scattering intensities at PPi

concentrations of 0.2 - 0.8 mM persisted at roughly 25 - 70% of the values expected in

the absence of dissolution (Figure 4-4b). Comparison of the scattering intensities

obtained by diluting chitosan/PPi mixtures (Figure 4-4a) to those obtained through the

“forward titration” (Figure 4-3a) reveals hysteresis in the particle formation/dissolution

cycle. This is illustrated in Figure 4-5 where, despite having identical compositions, the

samples show different light scattering intensities. The hysteresis loop obtained through

the addition and dilution of PPi shrinks with time (cf. Figure 4-5a and 4-5b). Yet, as

89 suggested by the incomplete particle dissolution in Figure 4-4b, the scattering intensities from the diluted samples remain higher than those from the “forward titration” samples, at every PPi concentration. This indicates that, once formed, portions of the chitosan/PPi complexes remain thermodynamically stable even when the PPi concentration is below the pseudo-equilibrium phase boundary in Figure 4-1b. Indeed, when the samples in

Figure 4-5 are analyzed again after 4 weeks, their light scattering intensities are essentially identical to those in Figure 4-5b, signifying that the data in Figure 4-5b reflects near-equilibrium (or near-pseudo-equilibrium) conditions.

(a) 5 (b) 1.2x10 1.2x105

5 1.0x10 1.0x105

4 8.0x10 8.0x104

4 6.0x10 4

6.0x10

4 4.0x10 4.0x104

4 2.0x10 2.0x104

Derived Count Rate (kcps)

0.0 Derived Count Rate (kcps) 0.0 0.0 0.4 0.8 1.2 1.6 2.0 0.0 0.4 0.8 1.2 1.6 2.0 PPi Concentration (mM) PPi Concentration (mM)

Figure 4-5. Hysteresis in the light scattering intensity obtained via () forward titration and () backward dilution after (a) 10 min and (b) 2 weeks of equilibration (from the time of dilution). The error bars are standard deviations (n = 3) and the lines are guides to the eye.

Similar hysteresis in the formation/dissolution of ionically crosslinked chitosan complexes has previously been suggested in a study where their formation and dissolution were triggered by changes in pH [179]. In that experiment, chitosan mixtures with either PPi or TPP were prepared near pH 1 (at which the ionic crosslinker charge was insufficient to form particles), whereupon the pH was raised (to form particles) and then lowered again (to redissolve them) through the sequential titration of NaOH and HCl 90 [179]. For both PPi and TPP, the pH required to redissolve the particles was lower than that required to form them, thus suggesting irreversibility. However, the equilibration time between the HCl and NaOH additions was very short (only 2 min), which means that equilibrium (or peudo-equilibrium) conditions were not achieved in this experiment.

Conversely, the present study allowed the mixtures to extensively equilibrate, and therefore indicates the hysteresis in the particle formation/dissolution cycle to be a thermodynamic effect. Furthermore, it shows that the hysteresis in the particle formation/dissolution cycle can occur even at constant pH, where the particles dissolve through simple PPi elution.

4.3.3. Model Analysis of the Hysteresis Loops

The hysteresis in the formation and dissolution of chitosan/PPi particles can be qualitatively interpreted using the Bragg-Williams theory [183, 184]. This theory can model irreversible cooperative binding [183, 184] and, in the present context, provides a simplified theoretical treatment of ionic crosslinking. The cooperativity of chitosan/PPi binding has recently been demonstrated using isothermal titration calorimetry (ITC), where its onset coincided with the PPi concentration where ionically crosslinked particles started forming [178]. Based on this observation, we modeled the cooperative chitosan/PPi crosslinking using the Bragg-Williams approach. Specifically, the free energy of cooperative interaction (Ec) was estimated as the product of the number of polymer-bound multivalent ions (N), the cooperative interaction energy of a single ion

(εc), and the probability of the bound ion interacting cooperatively with its neighboring binding sites (which, to avoid the double-counting of cooperative interactions, was

91 divided by two [183]). In the Bragg-Williams model, the cooperative interaction probability is estimated as the fraction of the binding sites that are occupied (N/M), so that:

N 2 E  c (4.2) c 2M where M is the total number of binding sites. This definition assumes that the cooperative interaction probability scales linearly with the fractional coverage of the binding sites.

Thus, it ignores the fact that, during chitosan/PPi binding, the cooperative crosslinking is only activated after a critical coverage is achieved [178]. Nonetheless, as shown below, this simplified model still captures the main features of the hysteresis loop in the chitosan/PPi association.

Using the above estimate for Ec, a canonical partition function for multivalent ion binding is defined as:

2 M!   N i  N  c 2M  Q(N, M ,T )  exp  (4.3) N!(M  N)!  kBT 

where T is the absolute temperature, εi is the non-cooperative binding energy and kB is the

Boltzmann constant. Based on this model, the chemical potential of the bound multivalent ions is determined using [183]:

  lnQ(N, M ,T)   A  k BT  (4.4)  N  M ,T

This expression yields:

    A  i  c  kBT ln  (4.5) 1 

92 where the fractional coverage, θ, replaces N/M. At equilibrium, µA equals to the chemical

potential of the free ions in solution (i.e., S  S,0  kBT lnCA , where CA is the free PPi concentration). Accordingly, the equilibrium CA-value can be related to θ, binding strength (K), and εc as:

      c  (4.6) CA  exp  K1   kBT  where K is given by:

 S ,0  i  K  exp  (4.7)  kBT 

The above equations predict how µA varies with θ (using Equation 4.5), and how θ varies with CA (using Equation 4.6 and 4.7). Depending on the εc-value, these expressions can yield two qualitatively-different trends. If the cooperativity is low (i.e., εc > -4kBT),

µA increases monotonically with θ (see Figure 4-6a, where (µA – εi )/kBT is plotted versus

θ). In this case, the binding isotherm is reversible (see model isotherm for εc = -2kBT in

Figure 4-6b) [184]. Conversely, for highly-cooperative interactions (where εc < -4kBT), the µA-dependence on θ becomes nonmonotonic. At very low θ-values, where few cooperative interactions occur, µA increases with θ. Once more multivalent ions bind to the polymer, however (i.e., at intermediate θ-values), the increase in the number of favorable cooperative interactions lowers µA and causes it to decrease with θ (Figure

4-6a). Finally, at high θ-values, this increase in the number of cooperative interactions becomes less-significant (since nearly all of the binding is already cooperative), and μA increases again.

93

(a) (b) 6 4 1.0  = -8k T B  = -2k T  = 0k T 0.9 c B c B 2 c B C -2k T 0 B 0.8 T -4k T B -2 B 0.7

/k

)

i 0.6

 -4 -8k T B PPi PPi

-

 0.5 -6

A





( -8 0.4 -10 0.3 -16kBT -12 0.2 -32kBT -14 0.1 A D 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 0.0  1E-4 1E-3 0.01 0.1 1 10 100 KC A Figure 4-6. Effects of εc on the binding showing: (a) the dependence of μA on θ for various εc-values (the shaded region indicates the spinodal envelope where the binding is unstable); and (b) model binding isotherms plotted for cases of weakly-cooperative (εc = -2kBT) and strongly-cooperative (εc = -8kBT) binding. The solid binding isotherms indicate equilibrium binding, whereas the dashed and dotted curves indicate regions of (----) pseudo-equilibrium and (······) unstable binding. Points A, B, C and D are defined in the text.

At fractional coverages where μA decreases with θ, the binding is unstable. Thus, the polyelectrolyte chains continue binding PPi until dμA/dθ becomes positive again and

μA equals to µS. Mathematically, the εc- and θ-values for this unstable binding (i.e., the spinodal envelope; see shaded region in Figure 4-6a) are defined by the region below the curve where dμA/dθ = 0 [184]. When the overall fractional coverage falls within this spinodal envelope, the polyelectrolyte chains are bifurcated into two populations: one with low θ-values (corresponding to solubilized chains) and one with high θ-values

(corresponding to ionically crosslinked particles). This is shown in the binding isotherm for εc = -8kBT (in Figure 4-7b) where, once KCA exceeds 0.053, θ jumps from 0.13 to 0.99

(from Point A to Point B; thus indicating particle formation). Similarly, when KCA is

94 reduced below 6.4 × 10-3 (e.g., during the dilution of PPi), θ drops abruptly from 0.87 to

6.7 × 10-3 (from Point C to Point D; thus indicating particle dissolution).

(a) (b)

1.0 1.0 B C B 0.9 C 0.9 0.8 0.8 0.7 0.7 0.6 0.6 PPi PPi PPi PPi



0.5 Agg 0.5

0.4 X 0.4 0.3 0.3 0.2 0.2 0.1 A 0.1 D D A 0.0 0.0 0.00 0.05 0.10 0.15 0.20 0.00 0.05 0.10 0.15 0.20 KC KC A,Total A,Total

Figure 4-7. Model hysteresis loops for εc = -8kBT and KCP = 0.1 showing (a) the binding isotherm and (b) the fraction of the crosslinked polymer chains plotted versus the normalized total PPi concentration. The solid isotherms indicate equilibrium binding, whereas the dashed and dotted curves indicate regions of (----) pseudo-equilibrium and (······) unstable binding. Points A, B, C and D are defined in the text.

Here, the hysteresis occurs because: (1) as KCA increases, particles cannot form

until a critical θ-value is reached (Point A, at which the μA curve crosses the left side of

the spinodal envelope; see εc = -8kBT curve in Figure 4-6a); and (2) during the dilution of

PPi it is thermodynamically favorable for the polymer chains to remain in the

highly-occupied (crosslinked particle) state for as long as possible (until Point C) because

it reduces CA (and therefore minimizes both µS and µA). Because the free energy is

minimized when the polyelectrolyte chains are in the highly-occupied state, the isotherm

followed during the dilution step (solid curve in Figure 4-6b) indicates the equilibrium

binding, while the isotherm followed during the forward titration (dashed curve in Figure

4-6b) indicates the pseudo-equilibrium binding (which is reflected in the phase map in

95 Figure 4-1b).

Based on this analysis, it is also possible to qualitatively model the evolution in θ and the fraction of chitosan chains that are crosslinked into particles (XAgg) as functions of the total PPi concentration in the sample, CA,Total. Here, CA,Total is equal to the sum of CA

(which is predicted from Equation 4.5) and CPθ, where CP is the concentration of PPi binding sites and CPθ is the concentration of chitosan-bound PPi:

      c  (4.8) CA,Total  exp   CP K1   kBT 

Equation 4.8 enables the binding isotherm in Figure 4-6b to be replotted as function of the normalized total PPi concentration, KCA,Total (see Figure 4-7a), where the particle formation again begins at Point A and continues with increasing KCA,Total (at a constant KCA, as shown in Figure 4-6b) until all chitosan chains are crosslinked (at Point B). Thus, before KCA,Total reaches Point A, XAgg = 0, beyond Point B, XAgg = 1 and, between these points, XAgg can be approximated using the lever rule as:

  A X Agg  (4.9) B  A

where θA is the fractional coverage at Point A and θB is the fractional coverage at Point B

(see dashed line in Figure 4-7b). During the dilution step, however (where CA,Total decreases and CP stays roughly constant), as shown by the arrows in Figure 4-6b and 4-7a, the polymer chains remain in the highly-occupied state for as long as possible (i.e., until Point

C is reached and the particles start to dissolve). At this point, some of the polymer chains begin transitioning from the highly-occupied, aggregated state (at Point C) to the sparcely-occupied, dissociated state (at Point D). Accordingly, between Points B and C,

96 XAgg = 1, while between Points C and D XAgg can be estimated as:

 D X Agg  (4.10) C D

where θC is the fractional coverage at Point C and θD is the fractional coverage at Point D.

(see solid line in Figure 4-7b). Finally, in the limit of very low PPi concentrations (where

KCA,Total is below Point D), all particles are dissolved and XAgg is zero.

These model trends in XAgg, where complete redissolution (Point D) occurs at lower

PPi concentrations that the onset of particle formation (Point A), are qualitatively consistent with the experimental hysteresis in the particle formation/dissolution cycle shown in Figure

4-5. Like in Figure 4-6b, the isotherm curve in Figure 4-7b connecting Points B, C and D corresponds to lower free energies than the isotherm connecting Points D, A and B. Thus,

Point A in Figure 4-7b represents the pseudo-equilibrium phase boundary (which is shown in Figure 4-1b) while Point D represents the equilibrium phase boundary. In addition to capturing the hysteresis in this phase transition, the Bragg-Williams analysis qualitatively captures the fact that the increase in XAgg with the increasing PPi concentration is gradual

(cf. Figures 5b and 4-7b). Indeed, the only qualitative difference between the experimental and model results is that the plateau at XAgg = 1 (connecting Points B and C in Figure 4-7b) was never reached in the experiment. This was primarily because the experimental PPi concentration (2.0 mM) was not high enough to achieve full chitosan aggregation (at which the light scattering intensity would have plateaued; see Figure 4-1a). This experimental condition was selected because, if higher initial PPi concentrations were used, the dispersions would have required more dilution, leading to unreliably low light scattering intensities. Nonetheless, the combination of experiment and theory clearly shows that the

97 particle formation/dissolution cycle is not entirely reversible, and that a fraction of the ionically crosslinked complex can remain intact even when the overall PPi concentration drops below that needed for the particles to form.

The coexistence of two binding states proposed above has previously been reported in other self-assembling mixtures – e.g., during highly-cooperative surfactant/polyelectrolyte binding [185-188], surfactant binding to dyes [189] and the binding of various ligands to DNA [185, 190]. Moreover, some of these systems exhibited binding hysteresis [186, 187, 189, 190] which, in the case of surfactant/polyelectrolyte binding, has also been modeled using the Bragg-Williams theory [184]. The highly-occupied binding states in these systems corresponded to colloidal complexes (e.g., surfactant/polyelectrolyte complexes or dye/surfactant crystals), while the substrate molecules with few bound ligands remained solubilized [185, 188,

189].

Analogous to these earlier findings, the present study reveals hysteresis in the ionic crosslinking of chitosan, and provides mechanistic insight on the dissociation of ionically crosslinked chitosan particles. This insight might also extend to other types of ionically crosslinked colloids, such as those prepared from calcium alginate. Indeed, exposure of covalently crosslinked alginate gels to calcium ions yields hysteresis in the gel swelling isotherms [191], thereby suggesting hysteresis in the calcium/alginate binding (which, consistent with the Bragg-Williams theory, is also cooperative [192]).

4.4. Conclusion

The crosslinking of chitosan with PPi yields micro- and nanoparticles, whose

98 size can be predictably tuned by varying the concentration of the parent chitosan solutions. The formation of these particles requires a critical PPi concentration, below which no particles form. Above this critical concentration, chitosan chains begin to self-assemble into colloidal complexes. The kinetics of this process depend strongly on the ionic crosslinker concentration. Unlike the more common chitosan/TPP particles

(which coagulate when TPP is in excess), these particles remain colloidally stable even at high PPi concentrations, suggesting that PPi ions do not readily bridge chitosan-based colloids. When PPi is leached from the colloidal particles, the particles dissolve. If a fraction of the PPi remains, however (even if its concentration is below that needed for particle formation), this dissolution process is incomplete. This hysteresis in the particle formation/dissolution process can be ascribed to cooperative chitosan/PPi binding (which can be qualitatively modeled using the Bragg-Williams theory), and suggests that the ionic crosslinking of chitosan is not strictly reversible.

99

Chapter 5

Factors that Affect Drug-Loaded Submicron

Chitosan/TPP Particle Stability

5.1. Introduction

In Chapter 4, the formation/dissolution hysteresis loop for the chitosan/PPi particles has shown that the multivalent counterion can elute from the particles upon dilution. It is already known that chitosan/TPP particles are stable in salt-free water.

However, they become less stable and tend to dissociate when placed in physiological conditions, where the pH (> pKa of chitosan) and ionic strength are significantly increased [2]. This is because increasing the pH-value and ionic strength can dramatically diminish the chitosan/TPP binding strength.

To improve the dissolution stability of chitosan/TPP particles, some groups have used crosslinking agents (e.g., glutaraldehyde [87]) to covalently crosslink the chitosan chains. Although this method improved particle stability, these crosslinking agents were toxic and procedures required to completely remove unreacted crosslinking agents from

100 the particles are likely to also release preloaded drugs [87, 88, 193]. An alternative way to stabilize chitosan/TPP particles was to use a second multivalent counterion. Giacalone et al., for instance, have shown that chitosan/TPP particles could be stabilized in PBS by crosslinking the particles with iron ions [93]. Another study by Gan et al. has suggested that bovine serum albumin (BSA)-loaded chitosan/TPP particles persisted in PBS for over 48 h. This may suggest that macromolecular drugs such as proteins and DNA, which bear negatively charged groups (e.g., carboxyl groups and phosphate groups), can provide further physical crosslinking between the chitosan chains when loaded into chitosan/TPP particles. Thus, dissolution stability and sustained drug release capabilities of the chitosan/TPP particle might sometimes be achieved due to the payload itself stabilizing the particles against dissolution in physiological environments.

To test this hypothesis, α-lactalbumin (α-LA), BSA and DNA were used to study the effect of protein drugs and polynucleotides on the chitosan/TPP particle stability. To explore the ionic strength dependence of this effect, particle stability was studied in salt-free water and 150 mM NaCl solution. pH 6.0 PBS solution was also explored as dissolution medium, since it mimics the aqueous environment within the nasal mucosa

(pH 5.5-6.5) [194]. Additionally, two different drug loading methods, incorporation method (where the drug was loaded during the particle formation) and incubation method

(where the drug was loaded after the particles were formed), were compared in their effect on the particle stability. Finally, to begin probing how this particle stability might affect the drug release properties, the fraction of dissociated drug achieved after 1 h under each dissolution condition was also investigated.

101

5.2. Materials and Methods

5.2.1. Materials

Chitosan (viscosity average molecular weight = 154 kDa and DD = 86% (as determined by capillary viscometry [2] and pH titration [164], respectively), sodium tripolyphosphate (TPP), fluorescein isothiocyanate (FITC), BSA and dimethyl sulfoxide

(DMSO) were all purchased from Sigma-Aldrich (St. Louis, USA). Micro BCA protein assay, single strand DNA, hydrochloric acid (HCl), sodium chloride (NaCl) and sodium hydroxide (NaOH) were purchased from Fisher Scientific (Fair Lawn, NJ). α-LA was a kind gift from Davisco Foods International, Inc (Eden Prairie, MN). All materials were used as received. Millipore Direct-Q 3 deionized water (DI water) with an 18.2 MΩ·cm resistivity was used in all experiments.

5.2.2. Preparation of FITC-Chitosan/TPP Micro- and Nanoparticles

FITC-labeled chitosan with roughly 0.3% of the chitosan amine groups labeled with FITC was prepared with described in Section 3.2.2. To prepare 0.1 wt%

FITC-chitosan solutions, 0.04 g FITC-chitosan was placed in 40 mL deionized water after adding 20 μL of 6 M HCl. The solution was then vortexed until the FITC-chitosan completely dissolved. Aqueous 0.125 wt% TPP and salt-free water were then prepared, and all solutions were adjusted pH to 5.5 using 1 M NaOH or 0.2 M HCl solutions. To prepare the FITC-chitosan/TPP particle dispersions, 4 mL of salt-free water was added dropwise into 10 mL of 0.1 wt% FITC-chitosan solution (to ensure that the final chitosan and TPP concentration was consistent with the drug-loaded particle dispersion in Section

102 5.2.3), while stirring the FITC-chitosan solution with 12 mm × 4 mm magnetic stir bar at a speed of 800 rpm. Two mL of 0.125 wt% TPP solution was titrated into FITC-chitosan solution at a rate of 2 drops/s and stirred overnight.

5.2.3. Preparation of Drug-Loaded FITC-Chitosan/TPP Micro- and Nanoparticles

To prepare BSA-loaded FITC-chitosan/TPP particles via the incorporation method,

0.1 wt% FITC-chitosan, 0.2 wt% BSA and 0.125 wt% TPP were first prepared and adjusted to pH 5.5. Then, 4 mL of 0.2 wt% BSA solution was added to 10 mL of 0.1 wt% chitosan solution at a rate of 2 drops/s and stirred at 800 rpm for 30 min. Two mL of

0.125 wt% TPP solution was then titrated into the chitosan/BSA mixture at the same rate and the mixture was stirred overnight. To repeat the experiment with another protein,

α-LA-loaded FITC-chitosan/TPP particles, were also prepared using the same procedure as the BSA-loaded FITC-chitosan/TPP particle except using 0.24 wt% α-LA solution and adjusting all solutions to pH 6.0 before mixing (which was necessary due to poor α-LA solubility at lower pH-level). The DNA-loaded FITC-chitosan/TPP particles were prepared with a similar same procedure as BSA-loaded FITC-chitosan/TPP particles, except using a 0.01 wt% DNA solution (pH 5.5) instead of two protein solution and stirring for only 10 min after adding the TPP solution. The low DNA concentration was used here because higher DNA concentrations could led to macroscopic chitosan/DNA precipitation. The DNA-loaded FITC-chitosan/TPP particles were then equilibrated for overnight without further stirring to avoid shear-induced precipitation. For comparison,

TPP-free DNA/FITC-chitosan complexes were also prepared using the same method as

103 the DNA-loaded chitosan/TPP particle dispersions, except using 2 mL of pH-matched water instead of the TPP solution.

To explore the drug loading procedure effect on the particle stability, BSA-loaded

FITC-chitosan/TPP particles were also prepared via the incubation method. Here, 2 mL of 0.125 wt% TPP solution was first titrated into 0.1 wt% FITC-CS solution. After 30 min of stirring, 4 mL of 0.2 wt% BSA solution was added into the FITC-chitosan/TPP particle dispersion and the dispersion was stirred overnight.

5.2.4. Analysis of FITC-Chitosan/TPP Micro- and Nanoparticle Stability

To probe the drug loading effect on the chitosan/TPP particle stability in different environments, pH-matched water (at pH 5.5 for BSA and DNA, and pH 6.0 for α-LA), containing either 0 or 150 mM of added NaCl, and pH 6.0 PBS solution were chosen as dissolution media. The particle dispersions (each containing 0.060 wt% chitosan) were also diluted 2, 5 and 20× in each media yielded final chitosan concentrations of 0.031,

0.013 and 0.003 wt%. To do this, the particle dispersions were first diluted twofold in either 300 mM NaCl or 2× PBS solution (which produced 2× diluted dispersion in either

150 mM NaCl solution or 1× PBS), and then immediately diluted in either 150 mM NaCl or 1× PBS to the desired particle concentration. The diluted samples were shaken by hand for 10 s to ensure a complete mixing. Since the dispersion pH could slightly increase (by roughly 0.1 - 0.2 pH-units) upon dilution, the sample pH was adjusted to its initial value

(either 5.5 or 6.0) immediately after dilution and all samples were equilibrated for 1 h before estimating the extent of particle dissolution.

104 Upon equilibration, the samples were characterized by DLS to obtain the light scattering intensities and size distributions. The light scattering intensities were normalized against the value of their original light scattering intensity (i.e., light scattering intensity measured right before dilution) and the dilution factor, such that [2]:

Scattering Intensity Dilution Factor I I  (5.1) 0 Initial Scattering Intensity

where I/I0 is the normalized light scattering intensity, which (according to Beer’s law) should have a value of 1.0 if the particles preserve their structure. The apparent particle size distributions estimated via the DLS analysis of the particles after dilution in different media using the multiple narrow modes non-negative least squares (NNLS) analysis

[195]were compared.

To confirm the stability data obtained from the light scattering measurements, the samples were centrifuged at 50,000 rpm for 1.5 h (at 20 °C ) to remove the undissolved particles. The FITC-chitosan concentrations in the supernatant phases after centrifugation were quantified by UV-vis spectroscopy at 490 nm wavelength, where ε = 2.038 mL mg-1 cm-1 in pH 5.5 salt-free water, 1.589 and 1.949 mL mg-1 cm-1 in pH 5.5 and 6.0 150 mM

NaCl, respectively, and 1.897 mL mg-1 cm-1 in pH 6.0 PBS. As a comparison, the dissolved FITC-chitosan solution (with a concentration matched to that of chitosan/TPP particles) were treated the same way as that used for the chitosan/TPP particles dispersions. These supernatant FITC-chitosan absorbances were normalized to the

FITC-chitosan UV absorbance before centrifugation to determine the fractions of dissolved particles.

To further the understanding of particle stability, the extents of drug dissociation from drug-loaded chitosan/TPP particles in different dissolution media were also 105 measured. The supernatant BSA and α-LA concentrations were characterized by Micro

BCA protein assay by UV-vis spectroscopy at λ = 562 nm. Conversely, the supernatant

DNA concentration was determined via UV-vis spectroscopy at λ = 259 nm (where ε =

18.87 mL mg-1 cm-1 and 19.63 mL mg-1 cm-1 in pH 5.5 150 mM NaCl and pH 6.0 PBS, respectively) by subtracting the FITC-chitosan absorbance at the same wavelength

(determined from its concentration and ε-value, where ε = 1.455 mL mg-1 cm-1 and 1.240 mL mg-1 cm-1 in pH 5.5 150 mM NaCl and pH 6.0 PBS, respectively). The supernatant drug concentrations were normalized to the molecular (particle-free) drug concentrations remaining after centrifugation to determine the fractions of dissociated drug, which is proportional to their normalized absorbance. This was calculated as:

Ap Normalized Absorbance = (5.2) Am where Ap is the supernatant drug absorbance after the drug-loaded particles are removed via centrifugation and Am is the control drug absorbance obtained after centrifugation of particle free drug solutions prepared at the same overall drug concentrations. All stability experiments were reproduced three times.

5.3. Protein Effect on the Chitosan/TPP Micro- and Nanoparticle Stability

The dissolution stability of chitosan/TPP micro- and nanoparticles is affected by both the media in which they are placed and the extent of their dilution [2]. This dissolution stability can often be inferred from the normalized light scattering intensity,

I/I0, where a value lower than 1.0 can be indicative of particle dissolution. When drug-free chitosan/TPP particles (prepared at pH 5.5) were diluted in the salt-free water at

106 pH 5.5, their I/I0, remained almost constant at roughly 0.9 - 1.2 even when the dispersions were diluted twentyfold to a chitosan concentration of 0.003 wt% (black squares in

Figure 5-1a). Thus, drug-free chitosan/TPP particles were stable to dilution in salt-free pH 5.5 water. In other words, due to the strong chitosan/TPP binding, particle dilution in salt-free water did not lead to significant elution of TPP ions. When diluted in 150 mM

NaCl (pH 5.5) and PBS (pH 6.0), however, I/I0 dropped to 0.8 when diluted twofold to

0.031 wt% chitosan (red circles for 150 mM NaCl and blue triangles for PBS in Figure

5-1a). This suggested that the particles partially dissolved upon dilution likely due to the weakened chitosan/TPP binding at elevated (roughly 150 mM) ionic strength [2]. The partial dissolution was also be affected by the extent of dilution, where I/I0 dropped dramatically from 0.8 to 0.1 - 0.3 when the chitosan concentration decreased from 0.031 to 0.003 wt% (likely again reflecting a change in the equilibrium chitosan/TPP binding).

The similarity of the I/I0 upon dilution in 150 mM NaCl and PBS solutions may result from their similar ionic strengths and pH-values.

107

(a) (b)

2.0 1.2

1.0 1.5 0.8 (0.013)

0

I

/

I 1.0 0.6 (0.031) 0.4 (0.003)

0.5 with BSA Uptake

0 I

/ 0.2 I 0.0 0.0 0.00 0.01 0.02 0.03 0.04 0.0 0.2 0.4 0.6 0.8 1.0 1.2 [Chitosan] After Dilution (wt%) I/I without BSA Uptake 0

(c) (d) 1.2 20

1.0 16 0.8 12 0.6

(0.031) 8 0.4

(0.013) Volume (%)

with α-LA Uptake

0 4

I

/ 0.2 (0.013) I (0.003) (0.003) 0.0 0 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1 10 100 1000 10000 I/I without α-LA Uptake Hydrodynamic Diameter (nm) 0

Figure 5-1. (a) The normalized light scattering intensity from dispersions of drug-free chitosan/TPP particles; parity plots showing comparisons of normalized light scattering intensities from dispersion of particles containing (b) BSA and (c) α-LA and drug-free dispersion each diluted to 0.003, 0.013 and 0.031 wt% chitosan; (d) apparent size distributions obtained via NNLS fitting of DLS data for BSA-loaded chitosan/TPP particle dispersions (▼) before dilution (at 0.063 wt% chitosan) and after being diluted fivefold (to 0.013 wt% chitosan) in (■) salt-free water, (●) 150 mM NaCl solution and (▲) PBS solution. The numbers in the parentheses in (b) and (c) are the chitosan concentrations (in units of wt%). The error bars are standard deviations and the dotted lines are the parity lines.

Similar I/I0 trends occurred when BSA was loaded into the chitosan/TPP particles.

This is seen in the parity plot (Figure 5-1b), where the I/I0 of BSA-loaded particle dispersions are plotted versus those of drug-free particles were all near the parity line

(regardless of whether the particles were diluted in 150 mM NaCl or PBS solution;

108 Figure 5-1b), This indicated that BSA did not improve the chitosan/TPP particle stability.

There was also no evidence of protein-induced stabilization in the light scattering data on

α-LA loaded particles. For the α-LA loaded particle dispersions, the I/I0 values fell slightly below the parity line (Figure 5-1c). This may be because the loading of α-LA led to a larger particle size (due to protein-induced higher-order aggregation), which significantly increased the I0-value compared to that of α-LA-free particles. Upon dilution in saline solution, however, the protein linkages, which (vide infra) are much weaker than those formed by TPP, caused the protein-induced higher order aggregates to fall apart. This decrease in particle size causes the I/I0-value of α-LA-bearing dispersions to drop more sharply compared to those of α-LA-free particle dispersions. Consequently, the I/I0 for α-LA loaded particle dispersions become smaller than that of α-LA-free particle dispersions.

Additionally, the apparent particle size distributions after fivefold dilution in water,

150 mM NaCl and PBS solution were compared (Figure 5-1d). Regardless of whether the

BSA-loaded chitosan/TPP particles were stable or partially dissociated, the fitted intensity-weighted size distributions were (at least when the particle sizes were plotted on a logarithmic scale) very similar to those of particles before dilution. This happened no matter whether the particles were diluted in salt-free water or in saline solutions (150 mM

NaCl and PBS), or whether the particles were loaded with α-LA or DNA (data not shown). Thus, as shown previously for drug-free chitosan/TPP particles by Huang et al.

[2], apparent size distributions obtained by DLS cannot be used as evidence of dissociation stability. In other words, similar particle size distributions do not exclude the possibility that, upon partial particle dissolution, some of the particles dissolve while the

109 rest remain essentially intact.

To verify the stability trends inferred from light scattering data, the protein uptake effect on the chitosan/TPP particle stability was also investigated by UV-vis spectroscopy.

When the chitosan/TPP particles were diluted in the salt-free water, the UV absorbance of dissolved FITC-chitosan was 20 - 30% of the control value (Figure 5-2a), which was the same as that before dilution and likely reflected the fact that not all the chitosan aggregated into particles during particle formation; see Chapter 3). The fact that nearly all of the chitosan remained in the particulate form, however, meant that the chitosan/TPP particles remained stable upon their dilution in the salt-free water (which was consistent with the high I/I0-values obtained via light scattering; Figure 5-1a).

110 (a) (b)

1.0 1.0

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

Normalized Absorbance 0.0 Normalized Absorbance 0.0 0.003 wt% 0.013 wt% 0.031 wt% 0.003 wt% 0.013 wt% 0.031 wt% DI water (pH 5.5) 150 mM NaCl (pH 5.5) (c) (d)

1.0 1.0

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

Normalized Absorbance 0.0 Normalized Absorbance 0.0 0.003 wt% 0.013 wt% 0.031 wt% 0.00 0.01 0.02 0.03 1X PBS (pH 6.0) Overall Chitosan Conc. (wt%)

Figure 5-2. Normalized FITC-chitosan UV-vis absorbance of ( ) FITC-chitosan solutions, ( ) FITC-chitosan/TPP particle dispersions and ( ) BSA-loaded FITC-chitosan/TPP particle dispersions after centrifugation in (a) NaCl-free water, (b) 150 mM NaCl solution and (c) PBS solution; and (d) normalized Micro BCA assay UV-vis absorbance of BSA-loaded FITC-chitosan/TPP particle dispersions after centrifugation in (■) salt-free water, (●) 150 mM NaCl solution and (▲) PBS solution (where a normalized Micro BCA assay absorbance of 1.0 corresponding to complete protein release). The error bars are standard deviations and the lines are guides to the eye.

When the experiment was repeated in the 150 mM NaCl solution, however, the fraction of dissolved FITC-chitosan was greater than that in salt-free water, and this fraction increased from 33 to 57% upon further dilution from 0.031 to 0.003 wt% chitosan (Figure 5-2b). This increase in dissolved FITC-chitosan indicated a reduced stability of drug-free particles at this near-physiological ionic strength. Similar to the low stability in 150 mM NaCl solution, drug-free particles also partially dissociated in PBS

111 upon dilution. The similar particle stability in the two dissolution media was not surprising, because the pH 6.0 PBS solution had a fairly-similar pH-value and ionic strength to that of 150 mM NaCl solution (pH 5.5). This partial particle dissolution in saline solutions suggested that, at near-physiological ionic strengths, the drug-free chitosan/TPP particle stability is highly concentration-dependent, which is consistent with the conclusion by Huang et al. [2].

When BSA was loaded in chitosan/TPP particles, the fraction of dissolved

FITC-chitosan in the salt-free water became lower (11-14%) than those obtained from drug-free particles, regardless of the final (post-dilution) chitosan concentration (Figure

5-2a). This was because BSA addition increased the fraction of the chitosan that aggregated into chitosan/TPP particles (i.e., by acting as a second crosslinker) [196].

When BSA-loaded particles were diluted in 150 mM NaCl and PBS, however, the fractions of dissolved FITC-chitosan were close to (or at least not much lower than) those of drug-free particles (Figures 5-2b and c). This was because the chitosan/BSA binding was much weaker than the chitosan/TPP interaction [196], which caused the BSA to be easily eluted at higher (near 150 mM) ionic strengths. Thus, since most of the protein was released from the particles, chitosan/TPP particle stability was very similar to that of their

BSA-free counterparts.

The above view was supported by the fraction of BSA present in the dissolution media upon dilution (Figure 5-2d). In the salt-free water (black squares), only about 10% of the total BSA remained in the supernatant, regardless of the extent of dilution. This indicated that the chitosan/BSA binding was strong enough to limit BSA dissociation in salt-free water. When the BSA-loaded chitosan/TPP particles were diluted twofold in the

112 150 mM NaCl (red circles) and PBS (blue triangles), however, roughly 76 - 78% of BSA became dissociated from the particles within the 1 h equilibration period. This value increased further (to 88% in 150 mM NaCl and to 81% in PBS), when the particles were diluted twentyfold. The fact that most of the BSA dissociated from chitosan/TPP particles indicated that the chitosan/BSA binding in saline media was too weak to prevent rapid/burst BSA release, and explained the weak effect of BSA on the particle stability

(see Figures 5-2b and c).

The inability of protein to increase chitosan/TPP particle stability was also seen

(via the same spectroscopic method) with, α-LA, which has a lower molecular weight and pI-value (α-LA; molecular weight ≈ 14.2 kDa; pI ≈ 4.2 - 4.5 [163]) than BSA (molecular weight ≈ 66.5 kDa; pI ≈ 4.7 - 4.9 [163]; see Appendix C, Figure C-1). Thus, protein loading does not appear to be a promising way to increase the particle stability in saline media. More importantly, though the protein-loaded chitosan/TPP particles remained partially stable when diluted only twofold, most of the protein was released from the particles. The rapid release of protein from chitosan/TPP nanoparticles was also reported by Fernandez-Urrusuno et al. [51], where more than 90% of loaded insulin was released in 15 min when placed in pH 6.4 phosphate buffer. These results suggested that protein release is likely not controlled by particle dissolution stability alone. Rapid protein diffusion within the particle matrix and protein desorption from the particle surface may be factors that control the release kinetics. These release mechanisms and rapid release kinetics might limit the use of protein-loaded chitosan/TPP particles for sustained release within saline media.

113 5.4. DNA Effect on the Chitosan/TPP Micro- and Nanoparticle Stability

Compared to the weak protein/chitosan binding (where the association constant typically ranges between roughly 10-6 and 10-3 M) [197, 198], DNA/chitosan binding is stronger with association constants reaching (10-10 - 10-9 M) [199]. Thus, unlike the weaker-binding proteins, DNA might be expected to increase the dissolution stability of chitosan/TPP particles. Similar to the drug-free chitosan/TPP particles, the I/I0 of

DNA-loaded chitosan/TPP particles remained almost constant, at roughly 1.0 - 1.2, when diluted in the salt-free water (data not shown). This invariance in I/I0-values indicated the particles to remain intact upon dilution. Further, when the DNA-loaded particles were diluted in 150 mM NaCl and PBS, the I/I0 decreased with dilution. In this case, however, the points on the parity plot (comparing I/I0-values of DNA-loaded particles with their

DNA-free counterparts) consistently fell on or slightly above the parity line, which might imply an increased particle stability in saline solutions upon DNA loading (Figure 5-3).

1.2

1.0

0.8 (0.013) (0.031)

0.6

0.4 (0.003)

with DNA Uptake

0 I

/ 0.2 I 0.0 0.0 0.2 0.4 0.6 0.8 1.0 1.2 I/I without DNA Uptake 0 Figure 5-3. Comparison of normalized light scattering intensities for DNA-loaded and DNA-free particles after dilution to 0.003, 0.013 and 0.031 wt% chitosan in (●) 150 mM NaCl solution and (▲) PBS solution. The numbers in the parentheses are the chitosan concentrations (in units of wt%) The error bars are standard deviations and the dotted line is the parity line.

114 To confirm that DNA can increase the chitosan/TPP particle stability, as suggested by the light scattering data, the dissolution stability of DNA-loaded particles was also probed by UV-vis spectroscopy. When DNA-loaded FITC-chitosan/TPP particles were diluted twofold in 150 mM NaCl, only 10 - 15% of the FITC-chitosan was dissociated from the particles, which was much lower than the roughly 30% that dissociated from the

DNA-free control particles (see Figure 5-4a). When DNA-loaded particles were diluted further, however, the fractions of dissolved FTIC-chitosan increased and were no longer higher than those of the DNA-free control particles. When diluted in PBS, the

DNA-loaded particles also had lower (or similar) fractions of dissociated chitosan compared to those of DNA-free particles (see Figure 5-4b). Though the reasons for the variability in this stabilization effect remain unclear, this result supports the view that under certain circumstances, DNA can enhance the stability of chitosan/TPP particles.

115

(b) (a) 1.0 1.0

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2 Normalized Absorbance Normalized Absorbance 0.0 0.0 0.003 wt% 0.013 wt% 0.031 wt% 0.003 wt% 0.013 wt% 0.031 wt% 150 mM NaCl (pH 5.5) 1X PBS (pH 6.0) (c) (d)

1.0 1.0

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2 Normalized Absorbance Normalized Absorbance 0.0 0.0 0.00 0.01 0.02 0.03 0.00 0.01 0.02 0.03 Overall Chitosan Conc. (wt%) Overall Chitosan Conc. (wt%)

Figure 5-4. Normalized FITC-chitosan UV-vis absorbance of ( ) FITC-chitosan solutions, ( ) DNA/FITC-chitosan mixture, ( ) FITC-chitosan/TPP particle dispersions, and ( ) DNA-loaded FITC-chitosan/TPP particle dispersions after centrifugation in (a) 150 mM NaCl solution and (b) PBS solution; normalized DNA UV-vis absorbance obtained from (■) DNA/FITC-chitosan mixtures after centrifugation and (●) DNA-loaded FITC-chitosan/TPP particle dispersions after centrifugation in (c) 150 mM NaCl solution and (d) PBS solution. The error bars are standard deviations and the lines are guides to the eye.

It was well known that DNA can complex with chitosan to form nanoparticles

even without TPP [200-202]. To further understand the DNA effect on particle stability,

this UV-vis spectroscopy experiment was repeated using TPP-free chitosan/DNA

complexes (Figures 5-4a and b). When the chitosan/DNA complexes were placed in 150

116 mM NaCl solution (pH 5.5), 50 - 85 % of FITC-chitosan was in the dissociated state

(Figure 5-4a). When diluted in pH 6.0 PBS solution, however, nearly all of the

FITC-chitosan was in the dissociated state. This might stem from both the slightly higher ionic strength (166 mM) and pH-value (pH 6.0) of PBS compared to that of 150 mM

3- NaCl (pH 5.5) and the presence of multivalent PO4 ions in PBS (which has a stronger interaction with chitosan than monovalent ions [203]). These two factors can enhance the competitive condensation of non-TPP ions and reduce the chitosan ionization both of which, further weaken chitosan/DNA binding.

When it came to the fraction of DNA dissociated from DNA-loaded chitosan/TPP particles and chitosan/DNA complexes, however, only a small portion of the DNA was in dissociated form, regardless of the release media used (Figures 5-4c and d). This result was consistent with previous findings and indicated very strong chitosan/DNA binding

[37]. Although DNA-loaded chitosan/TPP particles and chitosan/DNA complexes showed higher fractions of dissociated DNA when diluted twentyfold (to 3.1 × 10-3 wt% chitosan) in PBS, these fractions of dissociated DNA were much lower than the fraction of dissolved FITC-chitosan (compare Figures 5-4c and d to Figures 5-4a and b). This may have stemmed from a nonuniform distribution of DNA among chitosan chains during particle formation, where only a small portion of the chitosan chains formed complexes with the DNA. This might be explained by: (1) the chitosan/DNA binding being very fast and strong (such that most of the DNA complexed with the first chitosan chains it encountered during mixing); and (2) the DNA concentration (2.5 × 10-3 wt%) in the particle dispersion being much lower than the chitosan concentration (0.063 wt%) – i.e., the chitosan concentration was in great excess, with a chitosan amine groups to DNA

117 phosphate groups ratio (N/P ratio) of about 40.

Although the chitosan/DNA complexes can remain fairly-stable in saline media, due to a strong chitosan/DNA binding, most of the chitosan in chitosan/TPP particles

(and chitosan/DNA-mixtures) did not bind DNA, and therefore was free to dissolve. A higher DNA concentration may further enhance particle stability to dissolution. From the point of view of colloidal stability, however, this was very hard to achieve, because high

DNA concentrations led to the formation of fiber-like precipitates instead of stably-dispersed colloidal dispersions (compare Figures 5-5a and b). Due to this complication, initial DNA concentrations below 0.01 wt% (100 μg/mL) are typically used to prepare chitosan/DNA complexes and DNA-loaded chitosan/TPP particles [37, 99, 200,

204, 205].

118 (a) (b)

1 mm 1 mm

(c) (d)

1 mm 1 mm

Figure 5-5. Dark field micrographs of chitosan/TPP particle dispersion prepared with (a) 2.5 × 10-3 and (b) 5.0 × 10-3 wt% DNA via incorporation method, and (c) 1.3 × 10-3 and (d) 2.5 × 10-3 wt% DNA via incubation method. At the magnification used only macroscopic, fiber-liker precipitates are visible.

The improved stability of chitosan/TPP particles caused by the DNA uptake and strong chitosan/DNA binding could enable sustained DNA release in saline media, which has been reported by several groups [23, 37, 206]. Since the stability analysis in this study only explored drug dissociation using a short equilibrium time (1 h), with no special care taken to maintain sink conditions, the experiments described in this chapter do not pinpoint the precise release rates. Nevertheless, the method used in this chapter provides a quick preliminary analysis of the release properties (i.e., whether the particles dump most of their payload within the first hour; as seen in Figure 5-2d), and could save

119 time and – by screening the release behavior using only a single time point – resources in selecting promising sustained drug release systems.

5.5. The Effect of Drug Loading Procedure on Particle Stability

Since large molecule diffusion in and out of chitosan/TPP particles could be inhibited, the method of introducing drug molecules into chitosan/TPP particles has been hypothesized to affect the drug distribution within the ionic network [39]. Because this variable drug distribution could affect the particle stability, effects of the drug uptake method (i.e., incubation versus incorporation) on the particle stability were also explored.

Contrary to the above hypothesis, however, the UV-vis absorbance due to the dissociated

FITC-chitosan (in pH 6.0 PBS) was virtually the same, regardless of whether the BSA was loaded by the incorporation or incubation method (cf. circles and triangles in Figure

5-6a). This indicated that the drug loading method affected neither the drug uptake efficiency (see Chapter 3) nor the particle stability, and likely reflected the weak chitosan/BSA interaction in saline media. The analysis of BSA dissociation from the particles upon dilution further supported this conclusion – i.e., where the amounts of dissociated BSA obtained using the two drug loading methods were identical (Figure

5-6b).

120

(a) (b) 1.0 1.0

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

Normalized Absorbance 0.0 Normalized Absorbance 0.0 0.00 0.01 0.02 0.03 0.00 0.01 0.02 0.03 Overall Chitosan Conc. (wt%) Overall Chitosan Conc. (wt%)

Figure 5-6. Normalized UV-vis absorbance signals obtained from (a) FITC-chitosan and (b) Micro BCA assay of BSA-loaded FITC-chitosan/TPP particle dispersions prepared using the (●) incubation and (▲) incorporation drug loading methods (each obtained in pH 6.0 PBS after removing the undissolved particles through centrifugation). The error bars are standard deviations and the lines are guides to the eye.

When the above experiment was attempted with DNA, however, the incubation method caused the chitosan/TPP particles to coagulate and precipitate. This was probably due to the large dimensions of the DNA chains, which enabled them to easily bridge the preformed chitosan/TPP particles. Thus, macroscopic precipitation occurred even with

0.0013 wt% DNA (see Figures 5-5c and d). This precipitation suggests the incorporation method may be the only way to prepare the DNA-loaded submicron chitosan/TPP particles.

5.6. Conclusion

The study described in this chapter has revealed that protein loading had little impact on chitosan/TPP micro- and nanoparticle stability in saline media (150 mM NaCl pH 5.5 and pH 6.0 1× PBS). This finding was supported by both light scattering and

121 spectroscopic analysis. Additionally, particle stability was independent of whether the protein was loaded during particle formation (i.e., via incorporation) or loaded into the particles after they were formed (via incubation). The failure of protein to increase the particle stability may have stemmed from the very weak chitosan/protein binding in saline media, and suggests that, when used at high dilutions, protein-loaded chitosan/TPP particles will only remain stable when applied in low ionic strength environments.

Conversely, DNA uptake can, to some extent, increase the chitosan/TPP particle stability in saline media, despite there being very little DNA inside the particles. This is because the chitosan/DNA binding remains strong in moderate ionic strength environments. Due to this stability (and strong chitosan/DNA interactions), the DNA-loaded chitosan/TPP particles may be promising for the sustained DNA release at physiological ionic strengths.

Overall, to increase the dissolution stability of chitosan/TPP particles in saline environments, the chitosan/drug interaction should at least be similar to chitosan/TPP binding in strength and most drugs are unlikely to provide such a strong interaction.

122

Chapter 6

Factors That Affect Drug Release from the Submicron

Chitosan/TPP Particles

6.1. Introduction

The ability to control release kinetics is essential to drug carrier design. Numerous studies have been conducted to control drug release profiles – e.g., to release drugs in response to specific stimuli (such as a change in pH or ionic strength), prolong in vivo drug retention, or maintain drug release at a desired constant rate [207-209]. To better understand drug release mechanisms, many mathematical models, such as the

Korsmeyer-Peppas equation and Higuchi equation, have been developed and applied to experimental drug release profiles [120, 129, 210-212].

In recent decades, chitosan/TPP micro- and nanoparticles have been investigated for the delivery of drugs (such as insulin), vaccines (such as tetanus toxoid), and genes

(such as DNA and RNA) [159, 213-216]. The reported in vitro release profiles obtained with chitosan/TPP particles, however, varied significantly, even when the same model drug and similar release media (e.g., PBS solution) were used. Some studies, for instance,

123 indicated immediate payload release [27, 51, 52], while others showed release that was sustained over timescales as long as weeks [24, 34, 53]. These opposing findings cloud the understanding of the release properties of chitosan/TPP particles and make their safe and efficacious application difficult to achieve. Since lots of release profiles were obtained by the solvent replacement method, where centrifugation was used to separate the particles from the supernatant (see Figure 6-1), we have postulated that some of the inconsistency in the reported drug release profiles could stem from the centrifugation-induced coagulation of chitosan/TPP micro- and nanoparticles into macroscopic gels. Such coagulation is typically irreversible, and therefore (by increasing the diffusion distance) likely extends the release times. This likely makes the release profiles sensitive to variations in centrifugation methods (i.e., variations in centrifugal force and time). More extensive centrifugation, for instance, could increase the extent of such coagulation and may thus reduce the release rates.

Measure drug concentration

Remove Fill with Centrifuge supernatant fresh media

Figure 6-1. Schematic representation of the solvent replacement method for measuring release kinetics.

124 One of the most commonly used ways to help the redispersion of drug-loaded particles is using a glycerol bed [27, 37, 48]. By placing 50 - 200 mL of neat glycerol at the bottom of the centrifuge tube, the particles can be suspended inside the glycerol layer, which limits their irreversible coagulation upon centrifugation. This is because (through its interdiffusion with the aqueous buffer) the dense glycerol generates a density gradient at the bottom of the tube. Because of this, and the density of pure glycerol exceeds that of chitosan/TPP particles, the particles remain dispersed near the bottom of the centrifuge tube rather being deposited and coagulated on its bottom surface.

Further, although many groups have used glycerol beds in their drug release experiments, it was commonly ignored that the solvent replacement frequency (i.e., the time interval between each solvent replacement step) could affect the drug release profile, especially when the actual drug release is fast [55]. This may result from the equilibrium partitioning of the drug between the liquid supernatant and solid particles where, once equilibrium partitioning is achieved, no further drug may be released until the next solvent replacement step. Consequently, we hypothesized that low solvent replacement frequencies may lead to much slower apparent release than a high solvent replacement frequencies and introduce significant experimental artifacts into the release rate measurements.

A third major issue that could strongly distort the release profiles in these experiments is the incomplete recovery of drug-loaded particles upon centrifugation. In other words, a weak centrifugal force and short centrifugation time might not remove all the particles from the supernatant. This persistence of drug-loaded particles in the supernatant phase may lead to an overestimation of the drug release rate (because the

125 drug detection techniques do not always distinguish between the free drug and the drug that is loaded into the particles; see Appendix D.1). Conversely, if the particle-loaded drug in the supernatant is not detected, this incomplete particle recovery may lead to an underestimation of the total amount of drug released from the chitosan/TPP particles

(since the true amount of drug that can be released becomes smaller than that assumed to be present). To prevent this, a high centrifugal force and long centrifugation time may be needed.

Based on these potential artifacts, we hypothesized that the conflicting drug release profiles that were obtained by analyzing chitosan/TPP micro- and nanoparticles via the solvent replacement method were likely caused by: (1) irreversible micro- and nanoparticle coagulation during centrifugation; (2) failure to completely separate the particles from the supernatant; and (3) use of varied (and inappropriately selected) solvent replacement frequencies. To test these hypotheses, we investigated these artifacts using BSA as the model drug. Because the effects investigated in this study on submicron chitosan/TPP particles also likely apply to other micro- and nanoscale drug delivery systems (e.g., micro- and nanogels prepared from other polyelectrolytes or even colloids prepared from amphiphilic block copolymers), this analysis may provide better guidelines for obtaining more-reliable release kinetics data for an array of colloidal drug carriers.

6.2. Materials and Methods 6.2.1. Materials

Chitosan (viscosity average molecular weight = 154 kDa and DD = 86% as determined by capillary viscometry [2] and pH titration [164], respectively), sodium 126 tripolyphosphate (TPP), fluorescein isothiocyanate (FITC), BSA and dimethyl sulfoxide

(DMSO) were all purchased from Sigma-Aldrich (St. Louis, USA). Micro BCA protein assay, hydrochloric acid (HCl), sodium chloride (NaCl) and sodium hydroxide (NaOH) were purchased from Fisher Scientific (Fair Lawn, NJ). All materials were used as received. Millipore Direct-Q 3 deionized water (DI water) with a resistivity of 18.2

MΩ·cm was used in all experiments.

6.2.2. Preparation of FITC-labeled Chitosan

To facilitate measurement of the soluble (i.e., non-particulate) chitosan concentration before and after particle formation, the chitosan was labeled with FITC.

Specifically, 5.7 mL of FITC solution in DMSO (1 mg/mL) were added to 150-mL batches of 0.2 wt% chitosan solution (adjusted to pH 4.0 using HCl) and allowed to react for 3 h in the dark at room temperature. The resulting FITC-labeled chitosan

(FITC-chitosan) was then dialyzed for 24 h against DI water three times. By quantifying the unreacted FITC concentration in the dialysate, it was determined that roughly 0.3% of the chitosan amine groups were labeled with FITC. Once dialyzed, the purified

FITC-chitosan solution was freeze dried and stored until use at -18 °C .

6.2.3. Preparation of BSA-Loaded Chitosan/TPP Micro- and Nanoparticles

To prepare 0.1 wt% chitosan solutions, 0.08 g chitosan was placed in 80 mL of deionized water, and stirred at 150 rpm overnight to dissolve after adding 63 μL of 6 M

HCl. Aqueous 0.2 wt% BSA and 0.125 wt% TPP were then prepared, and all solutions were adjusted to pH 5.5 using either 1 M NaOH or 0.2 M HCl solutions. To prepare the

127 BSA-loaded chitosan/TPP particle dispersions, 4 mL of 0.2 wt% BSA solution was added to 10 mL of 0.1 wt% chitosan solution at a rate of 2 drops/s and stirred at 800 rpm for 30 min. Two mL of 0.125 wt% TPP solution was then titrated into the chitosan/BSA mixture at the same rate, whereupon the mixture was stirred overnight. Similarly, BSA-loaded

FITC-chitosan/TPP micro- and nanoparticles were prepared by the same procedure except using the FITC-labeled chitosan instead of the commercial chitosan. Further, to explore the effect of BSA loading on the particle recovery, BSA-free FITC-chitosan/TPP particles were prepared. To keep the chitosan and TPP concentrations consistent with those in the BSA-loaded chitosan/TPP particle dispersions, 4 mL of pH matched water was added during the particle preparation instead of the BSA solution.

6.2.4. Particle Recovery via Centrifugation

BSA and chitosan recovery were studied together to explore the centrifugal force and centrifugation time effects on the particle recovery. To probe the centrifugation condition effect on the chitosan recovery, BSA-free and BSA-loaded FITC-chitosan/TPP particle dispersions were centrifuged with centrifugal forces ranging between 2.1 × 104 and 3.0 × 105 g (with the centrifugation time fixed at 30 min) and centrifugation time ranging between 15 min and 2 h (with the centrifugal force fixed at 3.0 × 105 g) using a

Beckman ultracentrifuge (Ann Arbor, MI; motor model: SW 55Ti). As control experiments, TPP-free FITC-chitosan solutions and FITC-chitosan/BSA mixtures with the matching chitosan/BSA concentrations matching those of the BSA-loaded chitosan/TPP particle dispersions were centrifuged under identical centrifugation conditions. The chitosan recovery was calculated as follows:

128 C f Chitosan recovery (%) = (1 - ) ×100% (6-1) Ci

where Cf was the final chitosan concentration after centrifugation and Ci was the initial chitosan concentration before centrifugation.

Similar to the chitosan recovery measurements, the BSA recovery measurements were performed at varying centrifugation conditions using BSA-loaded chitosan/TPP particle dispersions, chitosan/BSA mixtures and molecular BSA solutions with matching chitosan/BSA concentrations. The supernatant BSA concentrations were determined using the Micro BCA protein assay, where the presence of chitosan/TPP particles did not significantly affect the determination of the total supernatant BSA concentration (see

Appendix D.1). The BSA recovery was then calculated using an analogous expression to

Equation 6-1.

6.2.5. Drug Release from Gel Pellets

To examine the release kinetics from the gel pellets that formed from chitosan/TPP micro- and nanoparticles upon ultracentrifugation, 5-mL aliquots of

BSA-loaded chitosan/TPP particle dispersions were placed in the centrifuge tubes. After being centrifuged for 30 to 90 min at 10 °C, 80% of supernatant was collected from each sample and replaced with 1.25× PBS buffer, so that the final dispersions contained 1×

PBS (pH 7.4). The centrifuge tubes were then placed into an Eppendorf Thermomixer set to 37 °C and a 220 rpm mixing speed. After a specific time interval (which ranged between 5 min and 24 h), 80% of the supernatant was replaced with fresh 1× PBS (pH

7.4). Since the gel pellets adhered to the bottoms of centrifugation tubes, only the tops of

129 the pellets were exposed to the release media (which meant that drug release only occurred through the tops of the pellets). The amount of drug released during each interval between solvent replacement steps was calculated as:

mi = Ci - Ci-1 1- RV (6-2)

where mi was the mass of the drug released between (i - 1)th and the ith of solvent replacement, and Ci and Ci-1 were the supernatant drug concentrations after the ith and (i -

1)th solvent replacement steps. When i = 1, C0 equaled to the initial supernatant drug concentration (i.e., before the dispersion was mixed with the PBS). R was the solvent replacement ratio, which was the supernatant volume fraction replaced during each solvent replacement step (and was 0.8 for all experiments in this study). Finally, V was the total release media volume. Based on this, the percent release (i.e., the loaded drug fraction released) and normalized percent release (which was normalized to the plateau % release at the end of the release experiment) was determined as:

mi Ci - Ci-1  1- RV (6-3) % Release   100% = Tot 100% m0   m

mi Ci - Ci-1  1- RV Normalized % Release   100% = 100% (6-4) m m

where m0 was the initial amount of drug within the gel pellet, m∞ was the amount of drug released at the end of the experiment (which was taken to be the amount of releasable drug), mTot was the total amount of drug in the particle dispersion that added into the release media at the beginning of the experiment, Σmi was the cumulative mass of drug released after the ith solvent replacement step, ε was the drug recovery efficiency (i.e.,

130 the fraction of drug was recovered from the supernatant through ultracentrifugation), which varied with centrifugation conditions.

6.2.6. Drug Release from Micro- and Nanoparticles

To help redisperse the particles after centrifugation, a glycerol bed was used.

Because glycerol has a higher density (1.26 g/cm3) than water (the density of water at

25 °C is 1.00 g/cm3), the glycerol remains at the bottom of the centrifugation tube during centrifugation. Because glycerol is also denser than the chitosan/TPP complex

(chitosan/TPP complex density ~ 1.08 g/cm3; see Appendix D.2), it prevents chitosan/TPP particle deposition at the bottoms of the centrifugal tubes. Moreover, because of its high viscosity (906 cP [217]), glycerol might inhibit particle coagulation by preventing their collision. Furthermore, glycerol is a water-soluble, nonionic small molecule, which does not appear to strongly interact with the chitosan/TPP particles.

Thus, the glycerol-protected particles can be easily redispersed by vortexing the sediment.

This redispersion is readily demonstrated by DLS, which shows that the particle size does not change much after this centrifugation and redispersion process (see Appendix D.3)

[48].

Two seminal papers by Calvo et al. have shown that chitosan/TPP particles can release BSA for more than one week [1, 39]. We speculated that this long-term release could reflect the choice of salt-free trehalose solution as the release media, where (unlike in a saline environment) BSA could remain associated with the particles due to the strong protein/particle binding. To confirm this, we performed a BSA release experiment using trehalose solutions (to mimic Calvo’s release conditions) as the release media. To

131 compare the conditions used by Calvo et al. with typical physiological ionic strengths, two ionic strength levels (0 and 150 mM NaCl) were used. To do this, 2.5-g batches of chitosan/TPP particle dispersions (prepared from a 0.1 wt% chitosan solutions) were mixed with 2.5-g aliquots of 10 wt% trehalose and either 0 or 300 mM NaCl mixture inside a centrifuge tubes (thus generating release media that contained 5 wt% trehalose and either 0 or 150 mM NaCl).

A glycerol bed was also used to limit coagulation during centrifugation by carefully adding 150 mg of glycerol dropwise into each particle dispersion, where the glycerol drops quickly sedimented to the bottoms of centrifuge tubes. After the dispersion was centrifuged at 3.0 × 105 g and 37 °C for 1 h, 4 g of the supernatant was replaced with a fresh mixture of 5 wt% trehalose and either 0 or 150 mM NaCl. The dispersion was then vortex-mixed for 20 s, and another 150 mg of glycerol was immediately added to the dispersion, whereupon the particles were centrifuged again (because the drug had already released for 1 h during the centrifugation). This procedure was then repeated several more times until the release profile reached a plateau.

When the drug release is faster than the solvent replacement frequency (such that equilibrium partitioning between the particles and media is reached before the solvent can be replaced), the release profile could be predominantly controlled by the solvent replacement frequency [55]. To probe the solvent replacement frequency effect on the chitosan/TPP particle drug release profile, a similar procedure as the above release experiment studying the ionic strength effect was used. Salt-free trehalose solution was used as the release media, in place of salt solutions. Additionally, the interval time between solvent replacement steps was varied from 1 to 12 h (where the 1 h

132 centrifugation time was counted as part of the time interval). This was because the drug release process persisted during the 1 h centrifugation process and the ultracentrifuge chamber was set at 37 °C (i.e., the same temperature as that used for the rest of the release experiment). When the interval time was longer than 1 h, the dispersions were placed in a thermomixer at 37 °C with a vortex speed of 220 rpm and incubated for the desired interval time. For example, to have an interval time of 12 h, the dispersions were placed in the thermomixer for only 11 h (and centrifuged for 1h). The release profiles were then calculated using the method described in Section 6.2.5.

6.3 Results and Discussion 6.3.1. Centrifugation Procedure Effects on Particle Recovery

The centrifugation-assisted solvent replacement method is the most widely used method for measuring the in vitro drug release from micro- and nanoparticles [27, 51, 52,

144]. This reflects the wide availability of centrifuges in academic labs and the ability to achieve selective separation of colloidal dispersions by tuning the centrifugation force and time. A drug-loaded chitosan/TPP micro- and nanoparticle dispersion, for instance, is a mixture of soluble drug, chitosan and drug/chitosan complexes, and drug-loaded chitosan/TPP particles [196]. Because the particles are denser than water (and are comparatively large in size), one can sediment the chitosan/TPP particles while keeping the other soluble components in the supernatant.

To determine the centrifugation time and force needed to sediment all the particles from the supernatant, a rough estimation can be made using Equation 6-5 [218]. This equation was derived from the Stokes’ law (because the Reynolds number ~ 10-2 << 1 and the Stokes’ law applies). It also assumes that the particles are spherical and sediment at 133 their terminal velocity; and that the centrifuge tube rotates parallelly to the ground.

9ηln (rf /r0 ) t = (6-5) 2Δρa2ω2 where Δρ is the density difference between the particles and solvent, a is the nanoparticle radius, η is the solvent viscosity, ω is the centrifugation speed (rad/s), r0 is the radial position of the dispersion surface and rf is the radial position at the bottom of the centrifuge tube (see Figure 6-2). This equation ignores: (1) the diffusion term, by assuming that the particle diffusion is much slower than its sedimentation; and (2) hydrodynamic interactions between the particles, and therefore provides a low (i.e., optimistic) estimate of the centrifugation time.

Dispersion surface Centrifuge tube

rf ω r0

Rotation center Drug-loaded particle

Figure 6-2. Schematic representation of particle sedimentation in a centrifuge tube inside a swinging bucket centrifuge rotor.

When a = 50 nm, Δρ ≈ 0.05 g/mL, η = 1.3 cP, r0 = 6.1 cm, rf = 10.9 cm and ω =

2094 rad/s (20000 rpm), as is roughly the case in the present study, the time required for the particles from the surface of the dispersion to sediment to the bottom of the centrifuge tube is approximately 100 min. The above conditions correspond to a centrifugal force of

134 2.7 × 104 g, which is close to that typically reported in the literature. The centrifugation time estimated above, however, is much longer than that typically used (30 min) [1,

34-36, 219-221]. Further, the sedimentation efficiency is affected by both diffusion and interparticle hydrodynamic interactions. If the sedimentation speed is not high enough, for instance, the diffusion could have a big impact on the time required to sediment all the particles [222-224]. Similarly, hydrodynamic interactions can significantly retard sedimentation once the dispersion became concentrated near the bottom of the tube [222].

The possibility of these effects further indicates that, to efficiently separate particles from the supernatant, a very powerful centrifuge is needed.

To confirm the above-required centrifugation time and force estimate, the recovery of empty and BSA-loaded chitosan/TPP micro- and nanoparticles through ultracentrifugation was measured at various ultracentrifugation speeds and times. When the centrifugation time was fixed at 30 min, the chitosan recovery from chitosan/TPP particles increased from 60 to 85% as the centrifugal force was increased from 2.1 × 104 to 3.0 × 105 g (see Figure 6-3a). This indicated that, at this frequently-used short centrifugation time, a significant portion of the chitosan/TPP particles was not removed from the supernatant at typical 2 - 4 × 104 g centrifugal forces [1, 34-36, 219-221]. When

BSA was loaded into chitosan/TPP particles, the chitosan recovery increased (Figure 6-3a) because the negatively-charged BSA caused more of the dissolved chitosan to self-assemble into particles (see Chapter 3). This increase, however, did not greatly diminish the centrifugal force requirement for sedimenting the particles. As a comparison, the chitosan recoveries in TPP-free chitosan and chitosan/BSA mixtures were less than

5%, which indicated their insensitivity to the centrifugal force range explored in this

135 study.

(a) (b)

100 100

80 80

60 60

40 40

20 20 BSA Recovery (%) BSA Recovery 0 0 Chitosan Recovery (%) Chitosan Recovery 0 1x105 2x105 3x105 4x105 0 1x105 2x105 3x105 4x105 (c) Centrifugal Force (g) Centrifugal Force (g) (d) 100 100

80 80

60 60

40 40

20 20 BSA Recovery (%) BSA Recovery

Chitosan Recovery (%) Chitosan Recovery 0 0 0.0 0.5 1.0 1.5 2.0 2.5 0.0 0.5 1.0 1.5 2.0 2.5 Centrifugal Time (h) Centrifugal Time (h)

Figure 6-3. The effects of (a, b) centrifugal force and (c, d) centrifugation time on the (a, c) chitosan recovery from (■) chitosan/TPP particles, (●) BSA-loaded chitosan/TPP particles, (▲) TPP-free chitosan and () TPP-free chitosan/BSA mixtures; and (b, d) BSA recovery from (■) BSA-loaded chitosan/TPP particles, (●) chitosan- and TPP-free BSA solutions and () TPP-free chitosan/BSA mixtures. The centrifugation force effects were investigated using a 30 min centrifugation time, while the centrifugation time effects were probed at a constant centrifugal force of 3.0 × 105 g. The error bars (which are largely obscured by the symbols) are standard deviations, while the lines are guides to the eye.

The importance of centrifugal force was further supported by BSA recovery measurement (Figure 6-3b). Only 26.1% of BSA were recovered together with the chitosan/TPP particles at a centrifugal force of 2.1 × 104 g. When the centrifugal force increased to 3.0 × 105 g, however, the recovered BSA increased significantly to 63.8%. In

136 the absence of chitosan/TPP particles, however, the sedimented protein fraction remained consistently low (well below 10%), indicating that the enhanced protein recovery reflected the sedimentation of particle-bound protein, and not free protein in solution.

This indicated that a very strong centrifugal force was needed to fully separate chitosan/TPP particles from the supernatant.

Similar to the effect of centrifugal force, the particulate chitosan recovery was also noticeably enhanced by increasing the centrifugation time (Figure 6-3c). When the centrifugal force was fixed at 3.0 × 105 g, the chitosan recovery increased from 82.1% to

95.2% as the centrifugation time for the BSA-free chitosan/TPP particles was increased from 15 min to 2 h. The loading of BSA further raised the particulate chitosan recovery.

Similar to the recovery of chitosan, the recovery of BSA increased with the centrifugation time and reached a plateau at around 1.5 h, where all BSA-loaded particles were apparently sedimented (Figure 6-3d). In the absence of TPP, however, less than 7% of both chitosan and BSA being removed, even after 2 h at the highest centrifugation speed.

Interestingly, the plateau chitosan recovery was achieved sooner than the BSA recovery, which contradicted the understanding that the BSA recovery evolution should be roughly proportional to the chitosan recovery for chitosan/TPP micro- and nanoparticles [196]. This may stem from the chitosan/TPP particles prepared with

FITC-labeled chitosan in the chitosan recovery experiment being bigger (330 ± 14 nm) than those prepared from commercial chitosan for the BSA recovery experiment (279 ±

15 nm), and therefore easier to sediment. This was confirmed by repeating some of the

BSA recovery measurements with FITC-chitosan, where plateau BSA recovery was achieved after 1 h, just like the chitosan recovery in Figure 6-3c (see Appendix D.4).

137 Based on the above analysis, near-quantitative chitosan/TPP particle recovery requires centrifugation times of at least 1.0 - 1.5 h at a centrifugal force of 3.0 × 105 g.

Most literature procedures, however, use centrifugation time of 30 min and much-smaller centrifugal forces, which typically range between 2 × 104 and 4 × 104 g [1, 34-36,

219-221]. This weak centrifugation likely fails to separate all the particles from the supernatant, as shown in Figure 6-3. The seminal papers on chitosan/TPP by Calvo et al., for example, centrifuged their particles for only 30 min at 4 × 104 g [1, 39]. This less-extensive centrifugation leads to a lower particle recovery, which not only causes underestimation of the drug uptake (since this centrifugation is also used during uptake experiments), but also can cause overestimation of release rates. This is demonstrated by the model analysis (i.e., Equation 6-3) in Figure 6-4 which, for simplicity, assumes that:

(1) all the drug is loaded into particles; (2) that the entire supernatant phase is replaced during each solvent replacement step; and (3) that none of the drug is really being released from the particles. When particle recovery is 99%, about 10% of the drug is

“released” (i.e., detected in the supernatant due to the persistence of unsedimented particles) after 10 solvent replacement steps. When particle recovery decreases to 70%, however, 97% of the drug is “released.” Thus, a low centrifugal force and/or short centrifugation have the potential to produce significant artifacts in the drug release rates by leaving detectable, particle-loaded drug in the supernatant phase.

138

100

80

60

40

20

Apparent % Released 0 0 1 2 3 4 5 6 7 8 9 10 Solvent Replacement Number

Figure 6-4. Model particle recovery effect on the apparent drug release profile calculated for conditions where (■) 99%, (●) 90%, (▲) 80% and (▼) 70% of the particles were recovered by centrifugation during each solvent replacement step. These model predictions were obtained under the simplifying assumption of no drug being released (which means that the true release profile should be a flat line overlapping the abscissa).

6.3.2. Coagulation Effects and Release Properties of Coagulated Particles

It is also important that the centrifugation procedure does not strongly affect the particle structure and release properties. Yet, particle coagulation into gel-like pellets (see

Figure 6-5) occurred even at the lowest centrifugal force and shortest centrifugation time analyzed in Figure 6-3 (i.e., at a centrifugal force of 2.1 × 104 g applied over 30 min).

This suggests that the more-extreme centrifugation that is needed to fully separate the particles from the supernatant could lead to their irreversible coagulation into a macroscopic, gel-like pellet (which would not be redispersed even after extensive agitation or ultrasonic treatment).

139 Supernatant Solution Dispersed Particles Gel-like 5 mm 5 mm Pellet

(a) (b)

Figure 6-5. Photographs of (a) nanoparticle dispersion before centrifugation and (b) gel-like pellet after centrifugation.

This particle coagulation could have a drastic impact on drug release rates. For particles with a 500 nm radius and an effective drug diffusivity of 1 × 10-7cm2/s, for instance, the characteristic release time should be less than 1 s [50] (i.e., τ ~ R2/D, where τ is the characteristic release time, R is the particle radius and D is the effective diffusivity of the drug within the particle [50, 225]). If the colloidal particles coagulate into a pellet with a radius of 1 mm, however, the release time increases drastically to approximately 1 d. These D and τ estimates assume that the protein drug molecules have no appreciable binding interaction with the chitosan/TPP particles (since the positive chitosan charges are mostly neutralized in the pH 7.4 PBS release media) and that the particles do not degrade [24, 34, 53]. If surface erosion or bulk particle degradation occur during the release process, the drug release could be even faster. Therefore, some of the slower release rates reported in the literature suggest that the drug may have been released from macroscale pellets rather than nanoscale particles [24, 34, 53]. To confirm this hypothesis, the drug release from the gel-like pellets obtained after ultracentrifugation was characterized.

140 6.3.2.1. Centrifugal Force Effect on the Release Profile

When the centrifugal force increased from 4.8 × 104 to 3.0 × 105 g, it caused more

BSA to be collected at the bottom of the centrifugation tubes (i.e., the BSA recovery increased from 39.7 to 63.8%; Table 1). This accumulation of both BSA and chitosan, however, did not slow down the release rate (Figure 6-6a). This virtual invariance in the release rate may have reflected the fact that the gel pellets (which determined the diffusion distance) did not get thicker. Indeed, the pellet thickness, defined as the distance from the bottom of the centrifuge tube to the top of the gel layer, decreased slightly (from

0.16 to 0.13 cm) as the centrifugal force increased (Table 6-1). Additionally, the slow

BSA release after the initial burst indicated that the gel pellets were stable against dissolution while in pH 7.4 PBS. This stability was likely because the deprotonation of chitosan reduced the chitosan solubility in water [2], which (even if TPP was leached) prevented the gel from dissolving (see Appendix D.5). Another interesting feature of the release data was that, roughly 20 - 30% of the protein appeared retained in the pellet, even when the plateau in the release profile was reached. This retention was highest at the intermediate centrifugal force (1.1 × 105 g). Overall, however, despite the variation of centrifugal force, these release curves were very close, indicating that the centrifugation force did not have a big impact on the drug release profile once the gel pellets were formed.

141

Table 6-1. Variations in the gel pellet thickness, BSA recovery and apparent BSA diffusivity within the gel with the centrifugal force (average ± standard deviation).

Centrifugal force Gel thickness BSA recovery D ×107 (g) (cm) (%) (cm2/s) 4.8 × 104 0.16 ± 0.00 39.7 ± 0.3 3.8 ± 0.7 1.1 × 105 0.15 ± 0.02 49.4 ± 0.5 3.4 ± 0.5 3.0 × 105 0.13 ± 0.00 63.8 ± 1.1 3.8 ± 0.3

(a) (b) 100 100

80 80

100 100 60 60

75 75

% Released

40 50 40 50 % Released % 25 25

% Released

20 Released % 20 0 0 0 1 2 3 Norm. 0.0 0.5 1.0 1.5 Time (h) Time (h)

0 Normolized 0 0 10 20 30 40 50 0 10 20 30 40 50 Time (h) Time (h) Figure 6-6. Actual (a) and normalized (b) BSA release profiles obtained from chitosan/TPP particle pellets after 30 min of centrifugation at (■) 4.8 × 104, (●) 1.1 × 105 and (▲) 3.0 × 105 g. The inset in (a) shows the first 3 h original BSA release profiles and the inset in (b) shows the first 60% normalized BSA release points and their model fit curves for pellets prepared by centrifugation at (black solid line) 4.8 × 104, (red dash line) 1.1 × 105 and (blue dot line) 3.0 × 105 g using Equation 6-8. The lines in the other plots are guides to the eye. All error bars are standard deviations.

6.3.2.2. Centrifugation Time Effect on the Release Profile

Similar to the centrifugal force effect, when the centrifugation time increased from 30 to 90 min, an increasing amount of BSA (from 63.8% to 84.4%) was collected at the bottoms of the centrifugation tubes. The gel pellet thickness, however, decreased from

0.13 to 0.11 cm (see Table 6-2). The reduction in the gel pellet thickness (along with the increased chitosan and BSA recovery) again suggested that gel compaction occurred

142 within the pellets. The BSA-loaded chitosan/TPP gel pellets obtained through 30 min of centrifugation resulted in the fastest BSA release (see insets in Figures 6-7a and b), which may have reflected its less-compact structure. The BSA release was slower when the centrifugation time increased to 60 and 90 min. Though the release rates varied, the final drug retention percentages were close, regardless of the centrifugation times.

Table 6-2. Variations in the gel pellet thickness, BSA recovery and apparent BSA diffusivity within the gel with the centrifugation time (average ± standard deviation).

Centrifugation time Gel thickness BSA recovery D ×107 (min) (cm) (%) (cm2/s) 30 0.13 ± 0.00 63.8 ± 1.1 3.8 ± 0.3 60 0.10 ± 0.00 78.9 ± 0.2 1.1 ± 0.2 90 0.11 ± 0.01 84.4 ± 0.8 1.5 ± 0.0

(a) (b)

100 100

80 80 100 100 60 60

75 75

% Released

40 50 40 50 % Released % 25 25

% Released 20 Released % 20 0 0 0 1 2 3 Norm. 0.0 0.5 1.0 1.5 2.0 Time (h) Time (h)

0 Normolized 0 0 10 20 30 40 50 0 10 20 30 40 50 Time (h) Time (h)

Figure 6-7. Actual (a) and normalized (b) BSA release profile obtained from chitosan/TPP particle pellets after (■) 30, (●) 60 and (▲) 90 min of centrifugation at 3.0 × 105 g. The inset in (a) shows the first 3 h original BSA release profiles and the inset in (b) shows the first 60% normalized BSA release points and their model fit curves for pellets prepared by centrifugation at (black solid line) 30, (red dash line) 60 and (blue dot line) 90 min using Equation 6-8. The lines in the other plots are guides to the eye. All error bars are standard deviations.

6.3.2.3. Further Analysis of Protein Transport within the Chitosan/TPP Gel Pellets

To better understand the protein release mechanism from the gel pellets, we

143 analyzed changes in their mass during the release process which, based on the gel mass remained roughly constant (see Appendix D.5), suggested that the gel swelling and degradation were very weak. Thus, the drug release is likely diffusion-limited and can be described using a Fickian diffusion equation. For one-dimensional diffusion, the microscopic mass balance needed to determine the time-dependent concentration profile within the gel pellets was given by:

C D   C  = A(z)  (6-6) t A(z) z  z  where C was the local drug concentration within the gel pellet, t was the release time, D was the drug diffusivity within the gel, z was the distance from the bottom of the centrifuge tube and A(z) was the cross-sectional area of the gel pellet at the axial position, z. Assuming that the inner radius of the centrifuge tube was R (see Figure 6-8), the z-dependent radius within the curved bottom of the centrifuge tube could be defined as r(z). Since A(z) equaled to πr2(z) and r (z) = R2 - (R - z)2 , A(z) could be expressed as

A(z)= π(2Rz - z2 ) . Moreover, because R >> z (i.e., R = 0.56 cm and z ≤ 0.16 cm, for which A(z) vs. z scaling remain roughly linear), A(z) was approximated as 2πR z .

Substituting this into Equation 6-6 yielded:

C D   C  = z (6-7) t z z  z 

This equation had the same functional form as that for the one-dimensional radial release from a cylinder (although here the radical position r, was replaced with z) [119]. Equation

144 6-7 had an alternative solution for short-time release behavior [119]:

1 2 3 2 M  Dt   Dt    Dt  t = 4    (6-8)  2   2   2  M  z  z  3 z 

where Mt and M∞ were the cumulative amounts of drug released at time t and infinite time, respectively, D was the drug diffusivity within the drug carrier matrix, and z was the thickness of the gel pellet at the center of the tube. This solution is valid for the first 60% of the release process [119]. Thus, the apparent diffusivity can be determined by fitting the release profile to Equation 6-8 as shown in the insets to Figures 6-6b and 6-7b.

R R-z

r z

(0,0)

Figure 6-8 Representative schematic of the chitosan/TPP gel pellet formed on the bottom of the centrifuge tube upon ultracentrifugation.

The apparent protein diffusivities within the gel pellets prepared using different centrifugation conditions have been summarized in Tables 6-1 and 6-2. With the increase

145 of centrifugal force (from 4.8 × 104 to 3.0 × 105 g), the apparent BSA diffusivity did not change much and remained in the range of 3.4 - 3.8 × 107 cm2/s, despite the apparent gel compression (which was evidenced by the reduction in gel thickness). When the centrifugation time increased from 30 to 60 min, however, the apparent BSA diffusivity dropped from 3.8 × 10-6 to 1.1 × 10-7 cm2/s, whereupon it did not change much when the centrifugation time increased further. The BSA diffusivities inside the chitosan/TPP gels were on the same order of magnitude with that in water (9.3 × 10-7 cm2/s at 37 °C [226]).

The high BSA diffusivity suggested that the mesh size of the gel network remained larger than the BSA molecule, despite the apparent gel compression during centrifugation.

Although the apparent protein diffusivity within the gel pellet can be estimated by fitting model Equation 6-8 to the normalized release data, the fit quality varied greatly with the centrifugation conditions. Namely, the qualities and corresponding R2-values of the model fits tended to decrease when the centrifugation condition became less-extensive (see Table 6-3). Interestingly, the release profile was more linear for the chitosan/TPP gel pellets prepared by centrifuging at 4.8 × 104 g for 30 min, than for the pellets prepared by centrifuging 3.0 × 105 g for 90 min (cf. squares and triangles in the inset of Figure 6-6b). This approach to an almost zero-order release with weaker centrifugation conditions was further supported by the n-values obtained by fitting the

n power-law, Korsmeyer-Peppas equation (where Mt/M∞ ∝ t [119]) to the experimental data, where the n-value increased from roughly 0.5 to 1 as the time and speed of centrifugation were diminished. Thus, at weaker centrifugation conditions the scaling with time deviated significantly from that expected for a gel, with a constant diffusivity and initially-uniform protein distributions where, for diffusion-controlled release and the

146 gel shape shown in Figure 6-8, the n-value was expected to be roughly 0.45 [119].

Table 6-3. The R2 (from Equation 6-8) and n-values (from Equation 2-6) of fitting the normalized release profiles (with the Korsmeyer-Peppas equation at various) centrifugation conditions.

Centrifugal force Centrifugation time R2 n (g) (min) 4.8 × 104 30 0.877 0.95 1.1 × 105 30 0.836 1.09 3.0 × 105 30 0.922 0.93 3.0 × 105 60 0.949 0.68 3.0 × 105 90 0.993 0.51

Further model analysis via Equation 6-7 suggested that the high n-values obtained with less-extensive centrifugation might have, at least in part, resulted from a nonuniform initial protein concentration within the chitosan/TPP gel pellet at the beginning of the release experiment. Such nonuniformity could stem from the gel at the bottom of the pellet (which formed first) becoming more compressed than that near the surface. This could produce a nonuniformity in drug the chitosan concentration within the pellet and, thus, a nonuniformity in binding sites. Because prior to being immersed in PBS, the protein exhibits significant affinity to the chitosan. This nonuniform binding site distribution might initially localize a much-higher BSA concentration near the bottom of the gel pellet than near the top. When the protein concentration was assumed to decrease linearly from its maximum value to 0 from the bottom to the top of the pellet, for example, the n value increased from 0.46 to 0.71 (see Appendix D.6).

The release rates obtained using the chitosan/TPP gel pellets can help to understand what might happen during the release experiments on chitosan/TPP micro- 147 and nanoparticles. The fact that the apparent diffusion coefficients were on the same order of magnitude as those of BSA in water indicated the BSA/particle binding to be weak during the release process (otherwise, the apparent diffusivity of BSA inside the gel would have been significantly smaller [227]). The weak BSA/particle binding might reflect the high pH-value of PBS (pH 7.4), which deprotonated the chitosan and eliminated most of its ionic interaction with the BSA [83]. The high BSA diffusivities measured herein were also similar to those in loosely-crosslinked (and essential non-interacting) poly(ethylene glycol) hydrogels at 37 °C (where D ~ 8.9 × 10-7 cm2/s

[228]). This further supported the above interpretation of the high BSA diffusivities within the chitosan/TPP gel pellets.

The average water content of the chitosan/TPP gel pellets was estimated at about

93 wt% (determined based on their weight loss upon drying for 24 h in a 60 °C oven).

Additionally, compared to the compressed gel pellets formed upon ultracentrifugation, dispersed chitosan/TPP micro- and nanoparticles might have a sparser gel network structure. Therefore, BSA diffusivity within the particles should be no less than that in the gel pellet. On the other hand, due to a high surface-to-volume ratio of the micro- and nanoparticles, the particles may be less stable to dissolution when placed in PBS [2].

Because of this, and their much lower diffusion path length, the release of BSA from the chitosan/TPP micro- and nanoparticles should be much faster than that from the macroscopic gel pellets. Indeed, even when the gel pellet is cut into smaller pieces

(approximately 0.1 - 1 mm in size) BSA release becomes much faster and more complete

(see Appendix D.7).

A number of reports (including the widely cited BSA release study by Gan et al.

148 [3]) have shown BSA release into 37 °C PBS from chitosan/TPP particles to occur over several hours before reaching a plateau [3, 32, 54]. These reports, however, did not indicate the use of any techniques (such as using a glycerol bed) for redispersing particles after centrifugation. Because coagulated chitosan/TPP particles are very difficult to redisperse to their original micro- and nanoscale state (even after the ultrasonic treatment or vigorous agitation), it is possible that these multiple-hour release times resulted from the irreversible coagulation of chitosan/TPP particles. Thus, it is very important to prevent irreversible coagulation during the centrifugation step. To our knowledge, the only method that has (occasionally) been utilized to achieve this was the use of a glycerol bed [27, 37, 48].

6.3.3. Release from Dispersed Chitosan/TPP Particles: Ionic Strength and Solvent Replacement Frequently Effects

Calvo et al. pioneered the use of chitosan/TPP micro- and nanoparticles as drug carriers [1, 39]. Their two seminal papers showed that the chitosan/TPP particles can release BSA for more than one week [1, 39]. In their work, however, a salt-free trehalose solution instead of a physiological ionic strength buffer (e.g., PBS) was used as the release medium. It is well-known that salt can diminish the electrostatic binding between oppositely-charged proteins and polyelectrolytes [83, 229, 230]. Thus, one would expect proteins to dissociate from the chitosan/TPP particles at physiological ionic strengths and suggests that the slow, one-week release profile reported by Calvo et al. likely resulted from the very low (and physiologically irrelevant) ionic strength of their the release media. To confirm this, we performed a BSA release experiment using a trehalose solution as release media (i.e., to mimic the release conditions in the Calvo study), but

149 using two different ionic strength levels.

The BSA release profiles from chitosan/TPP micro- and nanoparticles obtained in 5 wt% trehalose solution with either 0 mM NaCl (as used by Calvo et al.) or 150 mM NaCl

(i.e., physiological ionic strength) were compared in Figure 6-9. When the BSA-loaded chitosan/TPP particles were placed in 0 mM NaCl trehalose solution, the release rates were much lower. Only about 30% of BSA was released after 1 h and less than 45% of

BSA was released after 6 h.

In 150 mM NaCl media, however, (which is more representative of physiological environments), more than 90% of the BSA was released from the particles within 1 h.

Since the pH of the release media was the same as where the particles were formed (i.e., pH 5.5), this near-complete BSA release mainly reflected by the increased ionic strength, which caused the dissociation of BSA from the chitosan/TPP micro- and nanoparticles.

The rapid release of the dissociated protein may likely also reflected the small size of the chitosan/TPP micro- and nanoparticles. Indeed, most papers that used glycerol beds to redisperse the chitosan/TPP particles, also showed a burst drug release into PBS (even at the first time point), whereupon the release curve reached a plateau [4, 27, 146, 231].

150

100

80

60

40

% Released 20

0 0 1 2 3 4 5 6 7 Time (h)

Figure 6-9. The ionic strength effect on the BSA release profile in the pH 6.0 5% trehalose solution at with (■) 0 and (●) 150 mM NaCl. The error bars are standard deviations and the lines are guides to the eye.

Given this rapid protein release, the apparent release profile could also be affected by the solvent replacement frequency – i.e., equilibrium protein partitioning between the particles and the release media might be reached sooner than the time interval between the solvent replacement steps. To investigate this possibility, Figures 6-10a and b compare the release profiles obtained using 1, 4, 12 and 24-h intervals between the solvent replacement steps. The release profile obtained using the time interval of 1 h showed a faster release rate compared to that with the 4-h interval time (Figure 6-10a).

This indicates that sink conditions are not maintained with the 4-h interval time and that the presence of BSA in the release medium slows down further release. Similar trends are seen when data obtained with the 4-h interval are compared with that with the 12-h interval. Interestingly, however, the release profiles achieved with the 12 and 24-h solvent replacement interval times were partially overlapped (Figures 6-10a and b).

151

(a) (b)

100 100

80 80

60 60

40 40 % Released % Released 20 20

0 0 0 5 10 15 20 25 0 30 60 90 120 150 Time (h) Time (h)

(c) (d)

100 (i) (ii) 80

60

40

% Released 20 Sediment 1 mm 0 Ring 0 1 2 3 4 5 6 7 Solvent Replace Number

Figure 6-10. BSA release into pH 6.0 5% trehalose solution, obtained using (■) 1, (●) 4, (▲) 12 and (▼) 24-h time intervals between the solvent replacement steps, plotted versus the: (a, b) release time shown here (a) for the first day and (b) the whole experiment, (c) number of solvent replacement steps; and (d) photos showing particle coagulation after three centrifugation cycles with a glycerol bed taken (i) before and (ii) after vortexing. The error bars are standard deviations and the lines are guides to the eye.

When the percent release was plotted versus the solvent replacement number rather than versus time, the release profiles became very close to each other (Figure

6-10c). This further demonstrated that the BSA release profile could be highly dependent on the solvent replacement frequency. Each curve started with a 30 - 45% burst release at

152 the first solvent replacement followed by a much slower the BSA release rate. Here, the initial burst release may have stemmed from the rapid drug desorption from the surface of the particles, whereupon the release from their interior regions occurred more slowly. In other words, the BSA binding to the particle surface was likely weaker compared to that entrapped inside the particles, where (due to a greater abundance of chitosan/protein association sites) the BSA and chitosan binding was strong enough to retard the protein release into this NaCl-free release medium.

The release curve with the 1-h interval time was slightly lower than that with longer interval times (Figure 6-10c). This suggests that, though there is significant BSA accumulation in the release medium after 1 h of contact time (which means that sink conditions are not maintained), equilibrium BSA partitioning is not yet achieved. At higher time intervals, however, the curves in Figure 6-10c significantly overlap, indicating that equilibrium partitioning is nearly reached and that the apparent release profile obtained using the solvent replacement method are artifacts of the solvent replacement frequency used.

Interestingly, after the fourth solvent replacement, for the release profile obtained using a 24-h solvent replacement interval time, the percent released became higher than that with shorter time intervals (Figure 6-10c). This may have been caused by TPP and/or chitosan degradation [232-234], which may have led to partial particle dissociation during the longer, 6-day duration of the release experiments with the 24-h interval time between the solvent replacement steps. Overall, the above findings suggested that, to avoid “equilibrium partitioning artifacts,” a more frequent solvent replacement is preferred if the drug releases fast (which it appears to be in the case of chitosan/TPP

153 micro- and nanoparticles). Conversely, if the drug release was slow, a less frequent solvent replacement is more appropriate to minimize the “extra drug release” caused by the incomplete recovery of micro- and nanoparticles during centrifugation (as shown in

Figure 6-4).

Compared to the BSA release profiles in Figure 6-10 (performed using particles with drug loading capacities of about 34%), Calvo et al. (who used particles with a drug loading capacity of 41%) reported a much slower release rate and longer release time up to 1 week by replacing solvent every 2 days [1, 39]. Plotting the BSA release profile versus the solvent replacement number, however, showed the data reported by Calvo et al. to be quite similar to ours (see Figure 6-11). In both cases, 25 - 45% of the drug was released by the first solvent replacement step and about 50% of the drug was released after the fourth solvent replacement in both studies (with the moderate quantitative differences between the two data sets likely reflecting differences between the particle-to-solvent mass ratios used in the two studies). This result could imply that the drug release profile reported by Calvo et al. did not truly reflect the drug release behavior, since the solvent replacement interval time (2 days) was much longer than the time needed to reach equilibrium drug partitioning between the particles and the receiving media.

154 100

80

60

40

% Released 20

0 0 1 2 3 4 Solvent Replace Number

Figure 6-11. BSA release into 5% trehalose solution obtained (■) in this work using the 24-h time interval between the solvent replacement steps and (●) in Calvo’s work using a 2-day time interval between the solvent replacement steps [1]. The error bars are standard deviation and the lines are guides to the eye.

The glycerol bed, which was used to obtain the release data in Figures 6-9 and

6-10, helps prevent particle coagulation during centrifugation. After 3 or 4 centrifugation/redispersion cycles, however, ring-shaped deposits form on the centrifuge tubes and these coagulated particles cannot be completely redispersed even by extensive vortexing or sonication (Figure 6-10d). This coagulation becomes more prevalent with further centrifugation/redispersion cycles, and may slow down drug release by increasing the diffusion path length for drug molecules. Therefore, particle coagulation upon centrifugation appears to be unavoidable even with the glycerol bed. To avoid this coagulation problem in the drug release experiment, it may be a good alternative to place the drug-loaded particle dispersion in a large volume of release media (to create a true sink condition) and collect small fractions of the dispersion for testing during sampling times. To speed up the particle removal from the supernatant (to enable more-frequent solvent replacements), a stronger centrifugation force can be used. The centrifugation time can also be potentially reduced by only collecting the supernatant solution from the top of the centrifuge tube, which becomes depleted of particles. A filter with appropriate

155 size could also be a good way to rapidly remove particles from the supernatant.

6.4. Conclusions

Based on the above analysis, it was confirmed that the drug release profile can be affected by: (1) the particle recovery efficiency; (2) particle size change due to their centrifugation-induced coagulation; and (3) sink conditions not being achieved due to an insufficient solvent replacement frequency. Accordingly, to ensure accurate drug release profile determination via the solvent replacement method, a high centrifugal force and long centrifugation time should be used to recover the drug-loaded particles. A glycerol bed can help prevent the particle coagulation; however, this method no longer appears to work for the chitosan/TPP particles after the first 3 - 4 centrifugation/redispersion cycles.

For drug-loaded particles with fast release rates (such as the colloidal chitosan/TPP particles explored in this work), a more-frequent solvent replacement must also be used to maintain sink conditions. Due to the long centrifugation times required to fully sediment particles, however, this ultracentrifugation-assisted solvent replacement method cannot determine the drug release profile accurately. As an alternative, sink conditions can likely be achieved by diluting the drug-loaded particle dispersions in large volumes of release media and collecting small dispersion portions for drug concentration determination (using a filter to remove the particles). Although this work specifically focused on was done with chitosan/TPP micro- and nanoparticles, the experimental artifacts highlighted by this study likely also occur for other micro- and nanocarrier systems, such as other types of micro- and nanogels [72, 235], and (in certain situations) even liposomes and polymer micelles [236-238].

156

Chapter 7

Conclusions and Recommendations

7.1. Conclusions

The overarching goal of this dissertation was to achieve a better understanding of the drug uptake and release properties of ionically crosslinked chitosan micro- and nanoparticles. This goal was pursued via three major paths: (1) relating protein drug uptake into micro- and nanoparticles to the particle yield; (2) developing a better understanding of the particle dissolution stability; and (3) exploring experimental artifact effects on the drug release profiles obtained through commonly used experimental procedures. The principal findings from these analyses are summarized in the paragraphs that follow.

Protein uptake efficiency (i.e., the fraction of particle-associated protein, or

AE-values) achieved with chitosan/TPP micro- and nanoparticles increases almost linearly with the particle yields (fractions of aggregated chitosan, or XAgg-values). These near-linear AE versus XAgg relationships can be modeled via a simplified linear binding isotherm, which also successfully predicts AE variations with the chitosan concentration.

157 The nearly linear relationship between the particle yields and AE-values (which occurs when XAgg < 1.0), apparently reflects two effects: (1) higher particle yields provides more particulate binding sites for the protein uptake; and (2) the dissolved chitosan (which is not crosslinked into particles) competes with the chitosan/TPP particles for the protein.

Remarkably, the AE-values obtained at variable TPP:glucosamine ratios and protein concentrations collapses onto a single AE versus XAgg curve, whose insensitivity to the protein loading procedure suggests that protein uptake is governed by simple equilibrium partitioning. Overall, these uptake results mean that, to maximize the protein uptake into ionically crosslinked chitosan particles, the particle yield should also be maximized.

Experimental analysis of the reversibility of ionically crosslinked micro- and nanoparticle crosslinking revealed that particle formation is only partially reversible when diluted in the crosslinker-free solution. In other words, particles dissolve at a lower crosslinker concentration than that required for their formation. Further, protein loading has little impact on the chitosan/TPP micro- and nanoparticle stability to dissolution in saline media (i.e., it does not stabilize the particles by providing additional association between the chitosan chains). The failure of proteins to increase particle stability might stem from the very weak chitosan/protein binding in saline media, which causes the loaded proteins to rapidly dissociate from the particles. DNA, however, has a much stronger affinity for chitosan, and can therefore enhance chitosan/TPP micro- and nanoparticle stability to dissolution in saline media (thought the extent of this stabilization is limited by the very low DNA content within the particles). Thus, the extent to which drug loading can improve chitosan/TPP particle stability is strongly affected by the drug/chitosan binding strength. This indicates that only strongly binding

158 payloads (such as DNA) can enhance chitosan/TPP particle stability to dissolution in saline media and prevent the rapid burst release.

Analyses of the commonly used “solvent replacement” method of performing release experiments have also shown that conflicting findings might result from three experimental artifacts: (1) incomplete particle recovery during centrifugation; (2) irreversible centrifugation-induced particle coagulation; and (3) a failure to maintain sink conditions due to an insufficient solvent replacement frequency. These problems, however, are difficult to avoid simultaneously. Near-quantitative particle recovery, for instance, is only achieved after extensive ultracentrifugation which, while providing efficient particle recovery, irreversibly coagulates the colloidal chitosan/TPP particles into macroscopic gel pellets. Though a glycerol bed can mitigate this coagulation effect, its efficacy is limited to a small number of centrifugation-redispersion cycles.

Additionally, the long centrifugation times needed to separate particles from the supernatant also appear to prevent the very frequent solvent replacement that is needed to maintain sink conditions. Thus, to avoid these artifacts, either improvements to the centrifuge-assisted solvent replacement method should be developed, or other alternative methods such as in situ methods (i.e., directly placing the particles in the release media at a concentration that corresponds to sink condition [153]) should be used.

7.2. Recommendations

Although many questions regarding the drug uptake, dissolution stability and release behavior of ionically crosslinked chitosan micro- and nanoparticles have been answered in this dissertation, there are still problems that remain to be solved.

159 7.2.1 Further Studies on Drug Uptake into Ionically Crosslinked Particles

This dissertation revealed that the protein uptake primarily depended on the particle yield and drug/particle affinity. Although the particle yield of chitosan/TPP particles increased with the TPP:glucosamine ratio and pH-value, however, the relationship between the particle yield and ionic strength did not follow a monotonic trend. Specifically, the particle yield increased with the salt addition at mild-to-moderate ionic strengths, and decreased when the ionic strength became very high (see Section

3.3.2). This trend suggests that it may be possible to increase the AE by adding a small-to-moderate amounts of monovalent salt during protein uptake (i.e., by making the particle yield higher than it is under salt-free conditions). To explore this possibility, the ionic strength effect on the particle yield and AE-value should be analyzed simultaneously, to determine the optimal ionic strength for optimizing AE.

Another question relates to whether the above relationship between AE and particle yield extends to the uptake of small molecule drugs into chitosan/TPP particles.

Unlike the multivalent protein drugs, small molecule drugs (especially those with a single binding site) are unlikely to provide additional crosslinking between the chitosan chains

(which might affect how they contribute to the particle yields). Further, the introduction of small molecule drugs into particle dispersions can increase the ionic strength, thereby improving the particle yields and resulting in higher AE-values. Therefore, in order to extend the results in Chapter 3 to other classes of drug molecules (with various structures, sizes and functional groups), these questions should be studied in the future work.

160 7.2.2 Enhancing DNA Loading into Ionically Crosslinked Chitosan Particles

Analysis of the drug effect on the dissolution stability of chitosan/TPP particles in

Chapter 5 suggests that the loading of DNA, as done in gene therapy applications, into chitosan/TPP particles can increase the particle dissolution stability in saline media.

However, the loading of DNA into chitosan/TPP particles is greatly limited by the strong chitosan/DNA binding, which (as shown in Section 5.4) causes the particles to precipitate when high DNA concentrations are used. This can also cause a non-uniform DNA distribution within the particles, where all of the DNA is localized on a small fraction of the chitosan chains (or within a small fraction of the particles). To solve this problem, the

DNA and chitosan can be mixed at a high ionic strength to inhibit their complexation [95,

239]. Afterward, the chitosan/DNA mixture can be dialyzed against DI water to remove the salt, which should trigger the formation of chitosan/DNA complexes with a more-uniform DNA distribution. We hypothesize that this preparation procedure can preserve colloidal stability at higher DNA loadings and, accordingly, could provide more-stable and efficacious chitosan/TPP particles in for gene delivery applications.

7.2.3 Improving the Solvent Replacement Method for Investigating Drug Release from Colloids

This dissertation showed that the commonly-used, centrifuge-assisted solvent replacement method can lead to artifacts that prevent to an accurate determination of release rates from colloidal particles. Because centrifuges are widely available in research laboratories, however, it would be useful to explore ways to reduce these centrifugation-induced artifacts. To avoid the issues caused by the incomplete particle separation and irreversible coagulation that occur during centrifugation, the particle-free

161 supernatant solution can be collected by adjusting the centrifugation procedure to only partially sediment the particles. Here, the supernatant would only be collected from the surface layer of the dispersion (from which the particles would be sedimented first, without waiting for the particles to be fully sedimented to the bottom of the tube. This method could both greatly reduce the time required for separating particles and minimize the need for redispersing the coagulated particles. The main challenge of this method, however, is that it is not easy to determine the boundary between the particle-free supernatant layer and the particle-rich layer (which may vary between samples). This boundary can be potentially determined using a spectrophotometer-equipped centrifuge

[222, 240]. This kind of centrifuge, however, is not available in most labs and therefore is unlikely to be practical in most studies. Thus, easier methods need to be developed to determine the sedimentation profile in order to improve the centrifuge-assisted solvent replacement method.

162

References

[1] Calvo P, Remunan-Lopez C, VilaJato JL, Alonso MJ. Novel hydrophilic chitosan-polyethylene oxide nanoparticles as protein carriers. J Appl Polym Sci.

1997;63:125-32.

[2] Huang Y, Cai Y, Lapitsky Y. Factors affecting the stability of chitosan/tripolyphosphate micro-and nanogels: resolving the opposing findings. J Mater

Chem B. 2015;3:5957-70.

[3] Gan Q, Wang T. Chitosan nanoparticle as protein delivery carrier - Systematic examination of fabrication conditions for efficient loading and release. Colloids Surf B.

2007;59:24-34.

[4] Alonso-Sande M, Cuna M, Remunan-Lopez C. Formation of new glucomannan-chitosan nanoparticles and study of their ability to associate and deliver proteins. Macromolecules. 2006;39:4152-8.

[5] Cussler EL. Diffusion: Mass transfer in fluid systems: Cambridge University Press;

2009.

[6] Washington C. Evaluation of non-sink dialysis methods for the measurement of drug release from colloids: effects of drug partition. Int J Pharm. 1989;56:71-4.

[7] Huang Y, Lapitsky Y. Monovalent salt enhances colloidal stability during the

163 formation of chitosan/tripolyphosphate microgels. Langmuir. 2011;27:10392-9.

[8] Du Y, Xu Y. Chitin tetra ammonium salt nano-particle, its preparation method and use.

Univ Wuhan.

[9] Xu YM. Ionic structure and ionically crosslinked chitosan nanoparticles as a protein carrier. Abstr Pap Am Chems. 2004;228:U449-U.

[10] Jayakumar R, Chennazhi K, Muzzarelli R, Tamura H, Nair S, Selvamurugan N.

Chitosan conjugated DNA nanoparticles in gene therapy. Carbohydr Polym. 2010;79:1-8.

[11] Amaike M, Senoo Y, Yamamoto H. Sphere, honeycomb, regularly spaced droplet and fiber structures of polyion complexes of chitosan and gellan. Macromol Rapid

Commun. 1998;19:287-9.

[12] Pati F, Adhikari B, Dhara S. Development of chitosan-tripolyphosphate fibers through pH dependent ionotropic gelation. Carbohydr Res. 2011;346:2582-8.

[13] Park S, Bhang SH, La W-G, Seo J, Kim B-S, Char K. Dual roles of hyaluronic acids in multilayer films capturing nanocarriers for drug-eluting coatings. Biomaterials.

2012;33:5468-77.

[14] Gu R, Yuan X, Wu R, Li H, Xu S, Wang J. Layer-by-Layer Assembled Hydrogel

Nanocomposite Film with a High Loading Capacity. J Appl Polym Sci. 2014;131.

[15] Zimnitsky D, Shevchenko VV, Tsukruk VV. Perforated, freely suspended layer-by-layer nanoscale membranes. Langmuir. 2008;24:5996-6006.

[16] Tsao C-T, Lin Y-Y, Chang C-H, Han J-L, Hsieh K-H. Biocompatibility of chitosan/g-poly(glutamic acid) polyelectrolyte hydrogels. Abstracts of Papers of the

American Chemical Society. 2010;239.

[17] Hua FJ, Qian MP. Synthesis of self-crosslinking sodium polyacrylate hydrogel and

164 water-absorbing mechanism. J Mater Sci. 2001;36:731-8.

[18] Wang YS, Shen YB, Zhang YW, Yue B, Wu CX. pH-sensitive polyacrylic acid (PAA) hydrogels trapped with polysodium-p-styrenesulfonate (PSS). J Macromol Sci B

2006;45:563-71.

[19] Patil JS, Kamalapur MV, Marapur SC, Kadam DV. Ionotropic gelation and polyelectrolyte complexation: the novel techniques to design hydrogel particulate sustained, modulated drug delivery system: a review. Dig J Nanomater Biostruct.

2010;5:241-8.

[20] Lankalapalli S, Kolapalli VRM. Polyelectrolyte complexes: A review of their applicability in drug delivery technology. Indian J Pharm Sci. 2009;71:481-7.

[21] Alvarez-Lorenzo C, Blanco-Fernandez B, Puga AM, Concheiro A. Crosslinked ionic polysaccharides for stimuli-sensitive drug delivery. Adv Drug Del Rev. 2013;65:1148-71.

[22] Vervoort S, Budtova T. Evidence of shear-induced polymer release from a swollen gel particle. Polym Int. 2003;52:553-8.

[23] Koping-Hoggard M, Varum KM, Issa M, Danielsen S, Christensen BE, Stokke BT, et al. Improved chitosan-mediated gene delivery based on easily dissociated chitosan polyplexes of highly defined chitosan oligomers. Gene Ther. 2004;11:1441-52.

[24] Hou Y, Hu J, Park H, Lee M. Chitosan-based nanoparticles as a sustained protein release carrier for tissue engineering applications. J Biomed Mater Res A.

2012;100A:939-47.

[25] Hasanovic A, Zehl M, Reznicek G, Valenta C. Chitosan-tripolyphosphate nanoparticles as a possible skin drug delivery system for aciclovir with enhanced stability.

J Pharm Pharmacol. 2009;61:1609-16.

165 [26] Ji J, Hao S, Wu D, Huang R, Xu Y. Preparation, characterization and in vitro release of chitosan nanoparticles loaded with gentamicin and salicylic acid. Carbohydr Polym.

2011;85:803-8.

[27] Grenha A, Seijo B, Remunan-Lopez C. Microencapsulated chitosan nanoparticles for lung protein delivery. Eur J Pharm Sci. 2005;25:427-37.

[28] Ajun W, Yan S, Li G, Huili L. Preparation of aspirin and probucol in combination loaded chitosan nanoparticles and in vitro release study. Carbohydr Polym.

2009;75:566-74.

[29] Shi YX, Wan A, Shi YF, Zhang YY, Chen YP. Experimental and mathematical studies on the drug release properties of aspirin loaded chitosan nanoparticles. Biomed

Research International. 2014;613619

[30] Janes KA, Fresneau MP, Marazuela A, Fabra A, Alonso MaJ. Chitosan nanoparticles as delivery systems for doxorubicin. J Control Release. 2001;73:255-67.

[31] Tan ML, Choong PFM, Dass CR. Review: doxorubicin delivery systems based on chitosan for cancer therapy. J Pharm Pharmacol. 2009;61:131-42.

[32] Hu B, Pan C, Sun Y, Hou Z, Ye H, Hu B, et al. Optimization of fabrication parameters to produce chitosan-tripolyphosphate nanoparticles for delivery of tea catechins. J Agric Food Chem. 2008;56:7451-8.

[33] Zhang HL, Wu SH, Tao Y, Zang LQ, Su ZQ. Preparation and characterization of water-soluble chitosan nanoparticles as protein delivery system. J Nanomater.

2010;2010:898910.

[34] Jarudilokkul S, Tongthammachat A, Boonamnuayvittaya V. Preparation of chitosan nanoparticles for encapsulation and release of protein. Korean J Chem Eng.

166 2011;28:1247-51.

[35] Xu YM, Du YM. Effect of molecular structure of chitosan on protein delivery properties of chitosan nanoparticles. Int J Pharm. 2003;250:215-26.

[36] Zhang H, Oh M, Allen C, Kumacheva E. Monodisperse chitosan nanoparticles for mucosal drug delivery. Biomacromolecules. 2004;5:2461-8.

[37] Csaba N, Koeping-Hoeggard M, Jose Alonso M. Ionically crosslinked chitosan/tripolyphosphate nanoparticles for oligonucleotide and plasmid DNA delivery.

Int J Pharm. 2009;382:205-14.

[38] Zheng H, Tang C, Yin C. The effect of crosslinking agents on the transfection efficiency, cellular and intracellular processing of DNA/polymer nanocomplexes.

Biomaterials. 2013;34:3479-88.

[39] Calvo P, RemunanLopez C, VilaJato JL, Alonso MJ. Chitosan and chitosan ethylene oxide propylene oxide block copolymer nanoparticles as novel carriers for proteins and vaccines. Pharm Res. 1997;14:1431-6.

[40] de Pinho Neves AL, Milioli CC, Mueller L, Riella HG, Kuhnen NC, Stulzer HK.

Factorial design as tool in chitosan nanoparticles development by ionic gelation technique. Colloids Surf A. 2014;445:34-9.

[41] Pillai CKS, Paul W, Sharma CP. Chitin and chitosan polymers: Chemistry, solubility and fiber formation. Prog Polym Sci. 2009;34:641-78.

[42] Agnihotri SA, Mallikarjuna NN, Aminabhavi TM. Recent advances on chitosan-based micro- and nanoparticles in drug delivery. J Control Release.

2004;100:5-28.

[43] Xu YM, Du YM, Huang RH, Gao LP. Preparation and modification of

167 N-(2-hydroxyl) propyl-3-trimethyl ammonium chitosan chloride nanoparticle as a protein carrier. Biomaterials. 2003;24:5015-22.

[44] Koukaras EN, Papadimitriou SA, Bikiaris DN, Froudakis GE. Insight on the formation of chitosan nanoparticles through ionotropic gelation with tripolyphosphate.

Mol Pharm. 2012;9:2856-62.

[45] Kaya-Celiker H, Mallikarjunan K. Better nutrients and therapeutics delivery in food through nanotechnology. Food Engineering Reviews. 2012;4:114-23.

[46] Chang Y-L, Liu TC, Tsai M-L. Selective isolation of trypsin inhibitor and lectin from soybean whey by chitosan/tripolyphosphate/genipin co-crosslinked beads. Int J Mol Sci.

2014;15:9979-90.

[47] Ma Z, Yeoh HH, Lim L-Y. Formulation pH modulates the interaction of insulin with chitosan nanoparticles. J Pharm Sci. 2002;91:1396-404.

[48] Bahreini E, Aghaiypour K, Abbasalipourkabir R, Mokarram AR, Goodarzi MT,

Saidijam M. Preparation and nanoencapsulation of L-asparaginase II in chitosan-tripolyphosphate nanoparticles and in vitro release study. Nanoscale Res Lett.

2014;9.

[49] Rampino A, Borgogna M, Blasi P, Bellich B, Cesaro A. Chitosan nanoparticles:

Preparation, size evolution and stability. Int J Pharm. 2013;455:219-28.

[50] Arifin DY, Lee LY, Wang C-H. Mathematical modeling and simulation of drug release from microspheres: Implications to drug delivery systems. Adv Drug Del Rev.

2006;58:1274-325.

[51] Fernandez-Urrusuno R, Calvo P, Remunan-Lopez C, Vila-Jato JL, Alonso MJ.

Enhancement of nasal absorption of insulin using chitosan nanoparticles. Pharm Res.

168 1999;16:1576-81.

[52] Janes KA, Alonso MJ. Depolymerized chitosan nanoparticles for protein delivery:

Preparation and characterization. J Appl Polym Sci. 2003;88:2769-76.

[53] Bulmer C, Margaritis A, Xenocostas A. Production and characterization of novel chitosan nanoparticles for controlled release of rHu-Erythropoietin. Biochem Eng J.

2012;68:61-9.

[54] Mattu C, Li R, Ciardelli G. Chitosan nanoparticles as therapeutic protein nanocarriers: The effect of pH on particle formation and encapsulation efficiency. Polym

Compos. 2013;34:1538-45.

[55] Lapitsky Y. Ionically crosslinked polyelectrolyte nanocarriers: Recent advances and open problems. Curr Opin Colloid Interface Sci. 2014;19:122-30.

[56] Kleine-Brueggeney H, Zorzi G, Fecker T, El Gueddari N, Moerschbacher B,

Goycoolea F. A rational approach towards the design of chitosan-based nanoparticles obtained by ionotropic gelation. Colloids Surf B. 2015;135:99-108.

[57] Shu XZ, Zhu KJ. Controlled drug release properties of ionically cross-linked chitosan beads: the influence of anion structure. Int J Pharm. 2002;233:217-25.

[58] Fernandes A, Morais W, Santos A, De Araújo A, Dos Santos D, Dos Santos D, et al.

The influence of oxidative degradation on the preparation of chitosan nanoparticles.

Colloid Polym Sci. 2005;284:1-9.

[59] Cai Y, Lapitsky Y. Formation and dissolution of chitosan/pyrophosphate nanoparticles: Is the ionic crosslinking of chitosan reversible? Colloids Surf B.

2014;115:100-8.

[60] Rodrigues S, da Costa AMR, Grenha A. Chitosan/carrageenan nanoparticles: Effect

169 of cross-linking with tripolyphosphate and charge ratios. Carbohydr Polym.

2012;89:282-9.

[61] Huang Y, Lapitsky Y. Salt-assisted mechanistic analysis of chitosan/tripolyphosphate micro- and nanogel formation. Biomacromolecules. 2012;13:3868-76.

[62] Sorlier P, Denuzière A, Viton C, Domard A. Relation between the degree of acetylation and the electrostatic properties of chitin and chitosan. Biomacromolecules.

2001;2:765-72.

[63] Gan Q, Wang T, Cochrane C, McCarron P. Modulation of surface charge, particle size and morphological properties of chitosan-TPP nanoparticles intended for gene delivery. Colloids Surf B. 2005;44:65-73.

[64] Yu C-Y, Yin B-C, Zhang W, Cheng S-X, Zhang X-Z, Zhuo R-X. Composite microparticle drug delivery systems based on chitosan, alginate and pectin with improved pH-sensitive drug release property. Colloids Surf B. 2009;68:245-9.

[65] Huang Y, Lawrence PG, Lapitsky Y. Self-assembly of stiff, adhesive and self-healing gels from common polyelectrolytes. Langmuir. 2014;30:7771-7.

[66] Sonaje K, Lin Y-H, Juang J-H, Wey S-P, Chen C-T, Sung H-W. In vivo evaluation of safety and efficacy of self-assembled nanoparticles for oral insulin delivery. Biomaterials.

2009;30:2329-39.

[67] Chen LY, Subirade M. Chitosan/beta-lactoglobulin core-shell nanoparticles as nutraceutical carriers. Biomaterials. 2005;26:6041-53.

[68] Alsarra IA, Neau SH, Howard MA. Effects of preparative parameters on the properties of chitosan hydrogel beads containing Candida rugosa lipase. Biomaterials.

2004;25:2645-55.

170 [69] Becker AL, Henzler K, Welsch N, Ballauff M, Borisov O. Proteins and polyelectrolytes: A charged relationship. Curr Opin Colloid Interface Sci. 2012;17:90-6.

[70] Ubrich N, Bouillot P, Pellerin C, Hoffman M, Maincent P. Preparation and characterization of propranolol hydrochloride nanoparticles: a comparative study. J

Control Release. 2004;97:291-300.

[71] Sant VP, Smith D, Leroux J-C. Novel pH-sensitive supramolecular assemblies for oral delivery of poorly water soluble drugs: preparation and characterization. J Control

Release. 2004;97:301-12.

[72] Goycoolea FM, Lollo G, Remunan-Lopez C, Quaglia F, Alonso MJ.

Chitosan-alginate bended nanoparticles as carriers for the transmucosal delivery of macromolecules. Biomacromolecules. 2009;10:1736-43.

[73] Barakat NS, Almurshedi AS. Design and development of gliclazide-loaded chitosan microparticles for oral sustained drug delivery: in-vitro/in-vivo evaluation. J Pharm

Pharmacol. 2011;63:169-78.

[74] Cho AR, Chun YG, Kim BK, Park DJ. Preparation of chitosan-TPP microspheres as resveratrol carriers. J Food Sci. 2014;79:E568-E76.

[75] Wu Y, Yang W, Wang C, Hu J, Fu S. Chitosan nanoparticles as a novel delivery system for ammonium glycyrrhizinate. Int J Pharm. 2005;295:235-45.

[76] Bagre AP, Jain K, Jain NK. Alginate coated chitosan core shell nanoparticles for oral delivery of enoxaparin: In vitro and in vivo assessment. Int J Pharm. 2013;456:31-40.

[77] Huang Y. Micro- and nanogel formation through the ionic crosslinking of polyelectrolytes: University of Toledo; 2014.

[78] Lopez-Leon T, Carvalho ELS, Seijo B, Ortega-Vinuesa JL, Bastos-Gonzalez D.

171 Physicochemical characterization of chitosan nanoparticles: electrokinetic and stability behavior. J Colloid Interface Sci. 2005;283:344-51.

[79] Jonassen H, Kjoniksen A-L, Hiorth M. Stability of chitosan nanoparticles cross-linked with tripolyphosphate. Biomacromolecules. 2012;13:3747-56.

[80] Nasti A, Zaki N, de Leonardis P, Ungphaiboon S, Sansongsak P, Rimoli M, et al.

Chitosan/TPP and chitosan/TPP-hyaluronic acid nanoparticles: systematic optimisation of the preparative process and preliminary biological evaluation. Pharm Res.

2009;26:1918-30.

[81] Chellat F, Tabrizian M, Dumitriu S, Chornet E, Rivard CH, Yahia L. Study of biodegradation behavior of chitosan-xanthan microspheres in simulated physiological media. J Biomed Mater Res. 2000;53:592-9.

[82] Gupta H, Jain S, Mathur R, Mishra P, Mishra AK, Velpandian T. Sustained ocular drug delivery from a temperature and pH triggered novel in situ gel system. Drug Deliv.

2007;14:507-15.

[83] Kayitmazer AB, Strand SP, Tribet C, Jaeger W, Dubin PL. Effect of polyelectrolyte structure on protein−polyelectrolyte coacervates: coacervates of bovine serum albumin with poly(diallyldimethylammonium chloride) versus chitosan. Biomacromolecules.

2007;8:3568-77.

[84] Rinaudo M. Chitin and chitosan: Properties and applications. Prog Polym Sci.

2006;31:603-32.

[85] Lin Y-H, Sonaje K, Lin KM, Juang J-H, Mi F-L, Yang H-W, et al.

Multi-ion-crosslinked nanoparticles with pH-responsive characteristics for oral delivery of protein drugs. J Control Release. 2008;132:141-9.

172 [86] Goycoolea FM, Brunel F, Gueddari NEE, Coggiola A, Lollo G, Moerschbacher BM, et al. Physical properties and stability of soft gelled chitosan‐based nanoparticles.

Macromol Biosci. 2016;16:1873-82.

[87] Harde H, Siddhapura K, Agrawal AK, Jain S. Divalent toxoids loaded stable chitosan-glucomannan nanoassemblies for efficient systemic, mucosal and cellular immunostimulatory response following oral administration. Int J Pharm.

2015;487:292-304.

[88] Sadighian S, Hosseini-Monfared H, Rostamizadeh K, Hamidi M. pH-triggered magnetic-chitosan nanogels (MCNs) for doxorubicin delivery: physically vs. chemically cross linking approach. Adv Pharm Bull. 2015;5:115-20.

[89] Bhattarai N, Gunn J, Zhang M. Chitosan-based hydrogels for controlled, localized drug delivery. Adv Drug Del Rev. 2010;62:83-99.

[90] Genta I, Costantini M, Asti A, Conti B, Montanari L. Influence of glutaraldehyde on drug release and mucoadhesive properties of chitosan microspheres. Carbohydr Polym.

1998;36:81-8.

[91] Mi F-L, Shyu S-S, Wong T-B, Jang S-F, Lee S-T, Lu K-T. Chitosan–polyelectrolyte complexation for the preparation of gel beads and controlled release of anticancer drug. II.

Effect of pH-dependent ionic crosslinking or interpolymer complex using tripolyphosphate or polyphosphate as reagent. J Appl Polym Sci. 1999;74:1093-107.

[92] Yuan Y, Chesnutt B, Utturkar G, Haggard W, Yang Y, Ong J, et al. The effect of cross-linking of chitosan microspheres with genipin on protein release. Carbohydr Polym.

2007;68:561-7.

[93] Giacalone G, Hillaireau H, Capiau P, Chacun H, Reynaud F, Fattal E. Stabilization

173 and cellular delivery of chitosan–polyphosphate nanoparticles by incorporation of iron. J

Control Release. 2014;194:211-9.

[94] Corsi K, Chellat F, Yahia LH, Fernandes JC. Mesenchymal stem cells, MG63 and

HEK293 transfection using chitosan-DNA nanoparticles. Biomaterials. 2003;24:1255-64.

[95] Mao H-Q, Roy K, Troung-Le VL, Janes KA, Lin KY, Wang Y, et al. Chitosan-DNA nanoparticles as gene carriers: synthesis, characterization and transfection efficiency. J

Control Release. 2001;70:399-421.

[96] Carlsson F, Malmsten M, Linse P. Protein-polyelectrolyte cluster formation and redissolution: A monte carlo study. J Am Chem Soc. 2003;125:3140-9.

[97] Keawchaoon L, Yoksan R. Preparation, characterization and in vitro release study of carvacrol-loaded chitosan nanoparticles. Colloids Surf B. 2011;84:163-71.

[98] Li X, Kong X, Shi S, Zheng X, Guo G, Wei Y, et al. Preparation of alginate coated chitosan microparticles for vaccine delivery. BMC Biotechnol. 2008;8:89.

[99] Dai H, Jiang X, Tan GC, Chen Y, Torbenson M, Leong KW, et al. Chitosan-DNA nanoparticles delivered by intrabiliary infusion enhance liver-targeted gene delivery. Int J

Nanomedicine. 2006;1:507.

[100] Yao Y, Xia M, Wang H, Li G, Shen H, Ji G, et al. Preparation and evaluation of chitosan-based nanogels/gels for oral delivery of myricetin. Eur J Pharm Sci.

2016;91:144-53.

[101] Sousa F, Castro P, Fonte P, Kennedy PJ, Neves-Petersen MT, Sarmento B.

Nanoparticles for the delivery of therapeutic antibodies: Dogma or promising strategy?

Expert Opinion on Drug Delivery. 2017.

[102] Li XW, Lee DKL, Chan ASC, Alpar HO. Sustained expression in mammalian cells

174 with DNA complexed with chitosan nanoparticles. Biochimica et Biophysica Acta

(BBA)-Gene Structure and Expression. 2003;1630:7-18.

[103] Ganta S, Devalapally H, Shahiwala A, Amiji M. A review of stimuli-responsive nanocarriers for drug and gene delivery. J Control Release. 2008;126:187-204.

[104] Vivek R, Babu VN, Thangam R, Subramanian K, Kannan S. pH-responsive drug delivery of chitosan nanoparticles as Tamoxifen carriers for effective anti-tumor activity in breast cancer cells. Colloids Surf B. 2013;111:117-23.

[105] Montcourrier P, Silver I, Farnoud R, Bird I, Rochefort H. Breast cancer cells have a high capacity to acidify extracellular milieu by a dual mechanism. Clin Exp Metastasis.

1997;15:382-92.

[106] Uhrich KE, Cannizzaro SM, Langer RS, Shakesheff KM. Polymeric systems for controlled drug release. Chem Rev. 1999;99:3181-98.

[107] Kumari A, Yadav SK, Yadav SC. Biodegradable polymeric nanoparticles based drug delivery systems. Colloids Surf B. 2010;75:1-18.

[108] Sun Y, Wan A. Preparation of nanoparticles composed of chitosan and its derivatives as delivery systems for macromolecules. J Appl Polym Sci. 2007;105:552-61.

[109] Peng P, Voelcker NH, Kumar S, Griesser HJ. Nanoscale eluting coatings based on alginate/chitosan hydrogels. Biointerphases. 2007;2:95-104.

[110] Mora-Huertas CE, Fessi H, Elaissari A. Polymer-based nanocapsules for drug delivery. Int J Pharm. 2010;385:113-42.

[111] Yeh MK, Cheng KM, Hu CS, Huang YC, Young JJ. Novel protein-loaded chondroitin sulfate-chitosan nanoparticles: Preparation and characterization. Acta

Biomater. 2011;7:3804-12.

175 [112] Papadimitriou SA, Achilias DS, Bikiaris DN. Chitosan-g-PEG nanoparticles ionically crosslinked with poly(glutamic acid) and tripolyphosphate as protein delivery systems. Int J Pharm. 2012;430:318-27.

[113] Rekha MR, Sharma CP. Phthalyl chitosan-poly(ethylene oxide) semi-interpenetrating polymer network microparticles for oral protein delivery: an in vitro characterization. J Appl Polym Sci. 2008;110:2787-95.

[114] Wohl BM, Engbersen JFJ. Responsive layer-by-layer materials for drug delivery. J

Control Release. 2012;158:2-14.

[115] Qiu Y, Park K. Environment-sensitive hydrogels for drug delivery. Adv Drug Del

Rev. 2001;53:321-39.

[116] Siepmann J, Siepmann F, Florence AT. Local controlled drug delivery to the brain:

Mathematical modeling of the underlying mass transport mechanisms. Int J Pharm.

2006;314:101-19.

[117] Kamath KR, Park K. Biodegradable hydrogels in drug delivery. Adv Drug Del Rev.

1993;11:59-84.

[118] Berkland C, Kim KK, Pack DW. PLG microsphere size controls drug release rate through several competing factors. Pharm Res. 2003;20:1055-62.

[119] Ritger PL, Peppas NA. A simple equation for description of solute release I. Fickian and non-fickian release from non-swellable devices in the form of slabs, spheres, cylinders or discs. J Control Release. 1987;5:23-36.

[120] Siepmann J, Peppas NA. Higuchi equation: derivation, applications, use and misuse.

Int J Pharm. 2011;418:6-12.

[121] Katzhendler I, Hoffman A, Goldberger A, Friedman M. Modeling of drug release

176 from erodible tablets. J Pharm Sci. 1997;86:110-5.

[122] Brazel CS, Peppas NA. Modeling of drug release from Swellable polymers. Eur J

Pharm Biopharm. 2000;49:47-58.

[123] Helbling IM, Cabrera MI, Luna JA. Mathematical modeling of drug delivery from one-layer and two-layer torus-shaped devices with external mass transfer resistance. Eur J

Pharm Sci. 2011;44:288-98.

[124] Siepmann J, Peppas NA. Mathematical modeling of controlled drug delivery. Adv

Drug Del Rev. 2001;48:137-8.

[125] Costa P, Lobo JMS. Modeling and comparison of dissolution profiles. Eur J Pharm

Sci. 2001;13:123-33.

[126] Sackett CK, Narasimhan B. Mathematical modeling of polymer erosion:

Consequences for drug delivery. Int J Pharm. 2011;418:104-14.

[127] Siepmann J, Gopferich A. Mathematical modeling of bioerodible, polymeric drug delivery systems. Adv Drug Del Rev. 2001;48:229-47.

[128] Lin C-C, Metters AT. Hydrogels in controlled release formulations: Network design and mathematical modeling. Adv Drug Del Rev. 2006;58:1379-408.

[129] Lee AJ, King JR, Hibberd S. Mathematical modelling of the release of drug from porous, nonswelling transdermal drug-delivery devices. IMA J Math Appl Med Biol.

1998;15:135-63.

[130] Gibaldi M, Feldman S. Establishment of sink conditions in dissolution rate determinations. Theoretical considerations and application to nondisintegrating dosage forms. J Pharm Sci. 1967;56:1238-42.

[131] Dokoumetzidis A, Papadopoulou V, Macheras P. Analysis of dissolution data using

177 modified versions of Noyes–Whitney equation and the Weibull function. Pharm Res.

2006;23:256-61.

[132] Mulye N, Turco S. An examination of assumptions underlying the first-order kinetic model for release of water-soluble drugs from dicalcium phosphate dihydrate matrices. Drug Dev Ind Pharm. 1996;22:673-9.

[133] Hixson A, Crowell J. Dependence of reaction velocity upon surface and agitation.

Ind Eng Chem. 1931;23:1160-8.

[134] Paul DR, Harris FW. Controlled release polymeric formulations: ACS Publications;

1976.

[135] Brophy MR, Deasy P. Application of the Higuchi model for drug release from dispersed matrices to particles of general shape. Int J Pharm. 1987;37:41-7.

[136] Baker RW. Controlled release of biologically active agents: John Wiley & Sons;

1987.

[137] Higuchi T. Mechanism of sustained‐action medication. Theoretical analysis of rate of release of solid drugs dispersed in solid matrices. J Pharm Sci. 1963;52:1145-9.

[138] Weibull W. A statistical distribution function of wide applicability. Journal of applied mechanics. 1951;18:293-7.

[139] Papadopoulou V, Kosmidis K, Vlachou M, Macheras P. On the use of the Weibull function for the discernment of drug release mechanisms. Int J Pharm. 2006;309:44-50.

[140] Ramasamy T, Tuan Hiep T, Cho HJ, Kim JH, Kim YI, Jeon JY, et al.

Chitosan-based polyelectrolyte complexes as potential nanoparticulate carriers: physicochemical and biological characterization. Pharm Res. 2014;31:1302-14.

[141] Alam S, Khan ZI, Mustafa G, Kumar M, Islam F, Bhatnagar A, et al. Development

178 and evaluation of thymoquinone-encapsulated chitosan nanoparticles for nose-to-brain targeting: a pharmacoscintigraphic study. Int J Nanomedicine. 2012;7:5705-18.

[142] Jeon S, Yoo CY, Park SN. Improved stability and skin permeability of sodium hyaluronate-chitosan multilayered liposomes by Layer-by-Layer electrostatic deposition for quercetin delivery. Colloids Surf B. 2015;129:7-14.

[143] Barzegar-Jalali M, Adibkia K, Valizadeh H, Shadbad MRS, Nokhodchi A, Omidi Y, et al. Kinetic analysis of drug release from nanoparticles. J Pharm Pharm Sci.

2008;11:167-77.

[144] Luo Y, Zhang B, Cheng W-H, Wang Q. Preparation, characterization and evaluation of selenite-loaded chitosan/TPP nanoparticles with or without zein coating. Carbohydr

Polym. 2010;82:942-51.

[145] Li P, Dai Y-N, Zhang J-P, Wang A-Q, Wei Q. Chitosan-alginate nanoparticles as a novel drug delivery system for nifedipine. Int J Biomed Sci. 2008;4:221-8.

[146] Grenha A, Remuñán-López C, Carvalho EL, Seijo B. Microspheres containing lipid/chitosan nanoparticles complexes for pulmonary delivery of therapeutic proteins.

Eur J Pharm Biopharm. 2008;69:83-93.

[147] Makhlof A, Tozuka Y, Takeuchi H. Design and evaluation of novel pH-sensitive chitosan nanoparticles for oral insulin delivery. Eur J Pharm Sci. 2011;42:445-51.

[148] Pramod PS, Deshpande NU, Jayakannan M. Real-time drug release analysis of enzyme and pH responsive polysaccharide nanovesicles. The Journal of Physical

Chemistry B. 2015.

[149] Zambito Y, Pedreschi E, Di Colo G. Is dialysis a reliable method for studying drug release from nanoparticulate systems?—A case study. Int J Pharm. 2012;434:28-34.

179 [150] Modi S, Anderson BD. Determination of drug release kinetics from nanoparticles: overcoming pitfalls of the dynamic dialysis method. Mol Pharm. 2013;10:3076-89.

[151] Clark WR, Macias WL, Molitoris BA, Wang NL. Plasma protein adsorption to highly permeable hemodialysis membranes. Kidney Int. 1995;48:481-8.

[152] Sun S, Yue Y, Huang X, Meng D. Protein adsorption on blood-contact membranes.

J Membr Sci. 2003;222:3-18.

[153] Sinsuebpol C, Chatchawalsaisin J, Kulvanich P. Preparation and in vivo absorption evaluation of spray dried powders containing salmon calcitonin loaded chitosan nanoparticles for pulmonary delivery. Drug Des Devel Ther. 2013;7:861-73.

[154] Fontana G, Pitarresi G, Tomarchio V, Carlisi B, San Biagio PL. Preparation, characterization and in vitro antimicrobial activity of ampicillin-loaded polyethylcyanoacrylate nanoparticles. Biomaterials. 1998;19:1009-17.

[155] Zhang J, Zhu X, Jin Y, Shan W, Huang Y. Mechanism study of cellular uptake and tight junction opening mediated by goblet cell-specific trimethyl chitosan nanoparticles.

Mol Pharm. 2014;11:1520-32.

[156] Zhang H, Alsarra IA, Neau SH. An in vitro evaluation of a chitosan-containing multiparticulate system for macromolecule delivery to the colon. Int J Pharm.

2002;239:197-205.

[157] Fan W, Yan W, Xu ZS, Ni H. Formation mechanism of monodisperse, low molecular weight chitosan nanoparticles by ionic gelation technique. Colloids Surf B.

2012;90:21-7.

[158] Azevedo JR, Sizilio RH, Brito MB, Costa AMB, Serafini MR, Araujo AAS, et al.

Physical and chemical characterization insulin-loaded chitosan-TPP nanoparticles. J

180 Therm Anal Calorim. 2011;106:685-9.

[159] Vimal S, Majeed SA, Taju G, Nambi KSN, Raj NS, Madan N, et al. Chitosan tripolyphosphate (CS/TPP) nanoparticles: Preparation, characterization and application for gene delivery in shrimp. Acta Trop. 2013;128:486-93.

[160] Rampino A, Borgogna M, Blasi P, Bellich B, Cesàro A. Chitosan nanoparticles:

Preparation, size evolution and stability. Int J Pharm. 2013;455:219-28.

[161] Fernandez-Urrusuno R, Romani D, Calvo P, Vila-Jato JL, Alonso MJ. Development of a freeze-dried formulation of insulin-loaded chitosan nanoparticles intended for nasal administration STP Pharm Sci. 1999;9:429-36.

[162] Yang W, Fu J, Wang T, He N. Chitosan/sodium tripolyphosphate nanoparticles:

Preparation, characterization and application as drug carrier. J Biomed Nanotechnol.

2009;5:591-5.

[163] Zydney AL. Protein separations using membrane filtration: new opportunities for whey fractionation. Int Dairy J. 1998;8:243-50.

[164] Lapitsky Y, Zahir T, Shoichet MS. Modular biodegradable biomaterials from surfactant and polyelectrolyte mixtures. Biomacromolecules. 2008;9:166-74.

[165] Beukenkamp J, Rieman W, III, Lindenbaum S. Behavior of the condensed phosphates in anion-exchange chromatography. Anal Chem. 1954;26:505-12.

[166] Lawrence PG, Lapitsky Y. Ionically crosslinked poly(allylamine) as a stimulus-responsive underwater adhesive: Ionic strength and pH effects. Langmuir.

2015;31:1564-74.

[167] Jonassen H, Kjoniksen A-L, Hiorth M. Effects of ionic strength on the size and compactness of chitosan nanoparticles. Colloid Polym Sci. 2012;290:919-29.

181 [168] Kasimova MR, Velázquez-Campoy A, Nielsen HM. On the temperature dependence of complex formation between chitosan and proteins. Biomacromolecules.

2011;12:2534-43.

[169] Grenha A, Remuñán-López C, Carvalho ELS, Seijo B. Microspheres containing lipid/chitosan nanoparticles complexes for pulmonary delivery of therapeutic proteins.

Eur J Pharm Biopharm. 2008;69:83-93.

[170] Elmer C, Karaca AC, Low NH, Nickerson MT. Complex coacervation in pea protein isolate–chitosan mixtures. Food Res Int. 2011;44:1441-6.

[171] Guzey D, McClements DJ. Characterization of β-lactoglobulin–chitosan interactions in aqueous solutions: a calorimetry, light scattering, electrophoretic mobility and solubility study. Food Hydrocolloids. 2006;20:124-31.

[172] Kaibara K, Okazaki T, Bohidar HB, Dubin PL. pH-induced coacervation in complexes of bovine serum albumin and cationic polyelectrolytes. Biomacromolecules.

2000;1:100-7.

[173] Kizilay E, Kayitmazer AB, Dubin PL. Complexation and coacervation of polyelectrolytes with oppositely charged colloids. Adv Colloid Interface Sci.

2011;167:24-37.

[174] Dubin PL, Murrell JM. Size distribution of complexes formed between poly

(dimethyldiallylammonium chloride) and bovine serum albumin. Macromolecules.

1988;21:2291-3.

[175] Branco MC, Pochan DJ, Wagner NJ, Schneider JP. The effect of protein structure on their controlled release from an injectable peptide hydrogel. Biomaterials.

2010;31:9527-34.

182 [176] Xia J, Dubin PL, Kim Y, Muhoberac BB, Klimkowski VJ. Electrophoretic and quasi-elastic light scattering of soluble protein-polyelectrolyte complexes. J Phys Chem.

1993;97:4528-34.

[177] Pistone S, Qoragllu D, Smistad G, Hiorth M. Formulation and preparation of stable cross-linked alginate–zinc nanoparticles in the presence of a monovalent salt. Soft Matter.

2015;11:5765-74.

[178] Huang Y, Lapitsky Y. Determining the colloidal behavior of ionically crosslinked polyelectrolytes with isothermal titration calorimetry. J Phys Chem B. 2013;117:9548-57.

[179] Shu XZ, Zhu KJ. The influence of multivalent phosphate structure on the properties of ionically cross-linked chitosan films for controlled drug release. Eur J Pharm

Biopharm. 2002;54:235-43.

[180] Hiemenz PC, Rajagoplan R. Principles of Colloid and Surface Chemistry. New

York: Marcel Dekker, Inc.; 1997.

[181] Jonassen H, Kjoniksen AL, Hiorth M. Effects of ionic strength on the size and compactness of chitosan nanoparticles. Colloid Polym Sci. 2012;290:919-29.

[182] Fan W, Yan W, Xu ZS, Ni H. Erythrocytes load of low molecular weight chitosan nanoparticles as a potential vascular drug delivery system. Colloids Surf B.

2012;95:258-65.

[183] Hill TL. An Introduction to Statistical Thermodynamics. Reading: Addison-Wesley;

1960.

[184] Shirahama K. The nature of polymer-surfactant interactions. In: Kwak JCT, editor.

Polymer-Surfactant Systems: Marcel Dekker; 1998.

[185] Mel'nikov SM, Sergeyev VG, Yoshikawa K, Takahashi H, Hatta I. Cooperativity or

183 phase transition? Unfolding transition of DNA cationic surfactant complex. J Chem Phys.

1997;107:6917-24.

[186] Murase Y, Onda T, Tsujii K, Tanaka T. Discontinuous binding of surfactants to a polymer gel resulting from a volume phase transition. Macromolecules. 1999;32:8589 -

94.

[187] Sasaki S, Koga S, Imabayashi R, Maeda H. Salt effects on the volume phase transition of ionic gel induced by the hydrophobic counterion biding. J Phys Chem B.

2001;105:5852-5.

[188] Shirahama K, Kameyama K, Takagi T. Bimodality in the dodecylpyridinium bromide-sodium dextran sulfate system as observed by an electrophoretic method. J Phys

Chem. 1992;96:6817-20.

[189] Murakami K. Complex formation between dodecylpyridinium chloride and multicharged anionic planar substances. Langmuir. 2004;20:8183-91.

[190] Muller W, Crothers DM, Waring MJ. A non-intercalating proflavine derivative. Eur

J Biochem. 1973;39:223-34.

[191] Moe ST, Skjak-Braek G, Elgsaeter A, Smidsrod O. Swelling of covalently crosslinked alginate gels: Influence of ionic solutes and nonpolar solvents.

Macromolecules. 1993;26:3589-97.

[192] Fang Y, Al-Assaf S, Phillips GO, Nishinari K, Funami T, Williams PA, et al.

Multiple steps and critical behaviors of the binding of calcium to alginate. J Phys Chem B.

2007;111:2456-62.

[193] Chang C, Wang Z-C, Quan C-Y, Cheng H, Cheng S-X, Zhang X-Z, et al.

Fabrication of a novel pH-sensitive glutaraldehyde cross-linked pectin nanogel for drug

184 delivery. J Biomater Sci Polym Ed. 2007;18:1591-9.

[194] England RJA, Homer JJ, Knight LC, Ell SR. Nasal pH measurement: a reliable and repeatable parameter. Clin Otolaryngol Allied Sci. 1999;24:67-8.

[195] Roig AR, Alessandrini JL. Particle size distributions from static light scattering with regularized non-negative least squares constraints. Part Part Syst Char. 2006;23:431.

[196] Cai Y, Lapitsky Y. Analysis of chitosan/tripolyphosphate micro-and nanogel yields is key to understanding their protein uptake performance. J Colloid Interface Sci.

2017;494:242-54.

[197] Koppolu Bp, Smith SG, Ravindranathan S, Jayanthi S, Suresh Kumar TK, Zaharoff

DA. Controlling chitosan-based encapsulation for protein and vaccine delivery.

Biomaterials. 2014;35:4382-9.

[198] Kasimova MR, Velázquez-Campoy A, Nielsen HM. On the temperature dependence of complex formation between chitosan and proteins. Biomacromolecules.

2011;12:2534-43.

[199] Ma PL, Lavertu M, Winnik FM, Buschmann MD. New Insights into chitosan−

DNA interactions using isothermal titration microcalorimetry. Biomacromolecules.

2009;10:1490-9.

[200] Morris VB, Neethu S, Abraham T, Pillai C, Sharma CP. Studies on the condensation of depolymerized chitosans with DNA for preparing chitosan‐DNA nanoparticles for gene delivery applications. J Biomed Mater Res B Appl Biomater.

2009;89:282-92.

[201] Turan K, Nagata K. Chitosan-DNA nanoparticles: the effect of cell type and hydrolysis of chitosan on in vitro DNA transfection. Pharm Dev Technol.

185 2006;11:503-12.

[202] Lavertu M, Methot S, Tran-Khanh N, Buschmann MD. High efficiency gene transfer using chitosan/DNA nanoparticles with specific combinations of molecular weight and degree of deacetylation. Biomaterials. 2006;27:4815-24.

[203] Kovačević D, Borković S, Požar J. The influence of ionic strength, electrolyte type and preparation procedure on formation of weak polyelectrolyte complexes. Colloids

Surf A. 2007;302:107-12.

[204] Liu W, Sun S, Cao Z, Zhang X, Yao K, Lu WW, et al. An investigation on the physicochemical properties of chitosan/DNA polyelectrolyte complexes. Biomaterials.

2005;26:2705-11.

[205] Dhadwar S, Kiernan J, Wen J, Hortelano G. Repeated oral administration of chitosan/DNA nanoparticles delivers functional FVIII with the absence of antibodies in hemophilia A mice. J Thromb Haemost. 2010;8:2743-50.

[206] Lee D, Zhang W, Shirley SA, Kong X, Hellermann GR, Lockey RF, et al. Thiolated chitosan/DNA nanocomplexes exhibit enhanced and sustained gene delivery. Pharm Res.

2007;24:157-67.

[207] Tiwari G, Tiwari R, Sriwastawa B, Bhati L, Pandey S, Pandey P, et al. Drug delivery systems: An updated review. Int J Pharm Investig. 2012;2:2-11.

[208] Kulkarni RV, Biswanath SA. Electrically responsive smart hydrogels in drug delivery: a review. J Appl Biomater Biomech. 2007;5:125-39.

[209] Siegel RA. Stimuli sensitive polymers and self regulated drug delivery systems: A very partial review. J Control Release. 2014;190:337-51.

[210] Siepmann J, Siepmann F. Mathematical modeling of drug delivery. Int J Pharm.

186 2008;364:328-43.

[211] Fernandes M, Simon L, Loney NW. Mathematical modeling of transdermal drug-delivery systems: Analysis and applications. J Membr Sci. 2005;256:184-92.

[212] Zargar A, Chang K, Taite LJ, Schork FJ. Mathematical modeling of hyperbranched water-soluble polymers with applications in drug delivery. Macromol React Eng.

2011;5:373-84.

[213] Nagpal K, Singh SK, Mishra DN. Chitosan nanoparticles: a promising system in novel drug delivery. Chem Pharm Bull (Tokyo). 2010;58:1423-30.

[214] Vimal S, Majeed SA, Nambi KSN, Madan N, Farook MA, Venkatesan C, et al.

Delivery of DNA vaccine using chitosan-tripolyphosphate (CS/TPP) nanoparticles in

Asian sea bass, Lates calcarifer (Bloch, 1790) for protection against nodavirus infection.

Aquaculture. 2014;420:240-6.

[215] Vila A, Sánchez A, Janes K, Behrens I, Kissel T, Jato JLV, et al. Low molecular weight chitosan nanoparticles as new carriers for nasal vaccine delivery in mice. Eur J

Pharm Biopharm. 2004;57:123-31.

[216] Katas H, Alpar HO. Development and characterisation of chitosan nanoparticles for siRNA delivery. J Control Release. 2006;115:216-25.

[217] Association GP. Physical properties of glycerine and its solutions: Glycerine

Producers' Association; 1963.

[218] Van Holde KE, Johnson WC, Ho PS. Principles of physical biochemistry. 2006.

[219] Wan Aj, Sun Y, Li W-t, Li H-l. Transmission electron microscopy and electron diffraction study of BSA-loaded quaternized chitosan nanoparticles. J Biomed Mater Res

B Appl Biomater. 2008;86B:197-207.

187 [220] Xu Q, Guo L, Gu X, Zhang B, Hu X, Zhang J, et al. Prevention of colorectal cancer liver metastasis by exploiting liver immunity via chitosan-TPP/nanoparticles formulated with IL-12. Biomaterials. 2012;33:3909-18.

[221] Liu Z, Jiao Y, Liu F, Zhang Z. Heparin/chitosan nanoparticle carriers prepared by polyelectrolyte complexation. J Biomed Mater Res A. 2007;83A:806-12.

[222] Schuck P. Size-distribution analysis of macromolecules by sedimentation velocity ultracentrifugation and lamm equation modeling. Biophys J. 2000;78:1606-19.

[223] Dam J, Schuck P. Calculating sedimentation coefficient distributions by direct modeling of sedimentation velocity concentration profiles. Methods Enzymol.

2004;384:185-212.

[224] Lauffer MA. The size and shape of tobacco mosaic virus particles. J Am Chem Soc.

1944;66:1188-94.

[225] Deen WM. Analysis of Transport Phenomena. New York: Oxford University Press;

1998.

[226] Fardet A, Hoebler C, Djelveh G, Barry J-L. Restricted bovine serum albumin diffusion through the protein network of pasta. J Agric Food Chem. 1998;46:4635-41.

[227] Lawrence PG, Patil PS, Leipzig ND, Lapitsky Y. Ionically cross-linked polymer networks for the multiple-month release of small molecules. ACS Appl Mater Interfaces.

2016;8:4323-35.

[228] Weber LM, Lopez CG, Anseth KS. Effects of PEG hydrogel crosslinking density on protein diffusion and encapsulated islet survival and function. J Biomed Mater Res A.

2009;90:720-9.

[229] Mattison KW, Dubin PL, Brittain IJ. Complex Formation between Bovine Serum

188 Albumin and Strong Polyelectrolytes: Effect of Polymer Charge Density. J Phys Chem B.

1998;102:3830-6.

[230] Kizilay E, Kayitmazer AB, Dubin PL. Complexation and coacervation of polyelectrolytes with oppositely charged colloids. Adv Colloid Interface Sci.

2011;167:24-37.

[231] Krauland AH, Alonso MJ. Chitosan/cyclodextrin nanoparticles as macromolecular drug delivery system. Int J Pharm. 2007;340:134-42.

[232] Chang C, Racz G. Effects of temperature and phosphate concentration on rate of sodium pyrophosphate and sodium tripolyphosphate hydrolysis in soil. Can J Soil Sci.

1977;57:271-8.

[233] Jeon Y-J, Shahidi F, KIM S-K. Preparation of chitin and chitosan oligomers and their applications in physiological functional foods. Food Rev Int. 2000;16:159-76.

[234] Vårum K, Ottøy M, Smidsrød O. Acid hydrolysis of chitosans. Carbohydr Polym.

2001;46:89-98.

[235] You J-O, Peng C-A. Calcium-alginate nanoparticles formed by reverse microemulsion as gene carriers. Macromol Symp. 2005;219:147-53.

[236] Dhoot NO, Wheatley MA. Microencapsulated liposomes in controlled drug delivery: strategies to modulate drug release and eliminate the burst effect. J Pharm Sci.

2003;92:679-89.

[237] Shabbits JA, Chiu GN, Mayer LD. Development of an in vitro drug release assay that accurately predicts in vivo drug retention for liposome-based delivery systems. J

Control Release. 2002;84:161-70.

[238] Zhang Z, Feng S-S. The drug encapsulation efficiency, in vitro drug release,

189 cellular uptake and cytotoxicity of paclitaxel-loaded poly (lactide)–tocopheryl polyethylene glycol succinate nanoparticles. Biomaterials. 2006;27:4025-33.

[239] Xu C, Cai H, Xu Q, He P, Fang Y. Characterization of single-stranded DNA on chitosan-modified electrode and its application to the sequence-specific DNA detection.

Fresenius J Anal Chem. 2001;369:428-32.

[240] Schuck P. Sedimentation analysis of noninteracting and self-associating solutes using numerical solutions to the lamm equation. Biophys J. 1998;75:1503-12.

190

Appendix A

A.1. Representative Chitosan/TPP Particle Size Distributions

Chitosan/TPP particle size distributions were analyzed by dynamic light scattering

(DLS) using a Zetasizer Nano ZS (Malvern, UK) dynamic and electrophoretic light scattering system. Because no special care was taken to minimize particle polydispersity, the particles were highly polydisperse and (as shown by the representative data in Figure

A-1) ranged between roughly 40 and 1000 nm in hydrodynamic diameter.

15

10

5

Volume (%)

0 1 10 100 1000 10000 Hydrodynamic Diameter (nm)

Figure A-1. Representative volume-weighted size distributions of (■) empty and (●) BSA-loaded chitosan/TPP particles, which were loaded via incorporation using a 0.05 wt% overall BSA concentration. Each particle batch was prepared at a 0.13:1 TPP:glucosamine molar ratio from NaCl-free parent chitosan and TPP solutions at pH 5.5.

191 A.2. Chitosan Effect on Bradford Assay Results

Figure A-2 shows that the presence of chitosan had negligible impact on the

Bradford protein assay reading. In the absence of BSA, the absorbance remained at baseline levels as the chitosan concentration was increased (black squares). Similarly, the increase in the Bradford assay absorbance with the BSA concentration was unaffected by chitosan addition (compare red circles and blue triangles). This confirmed that the presence of chitosan did not interfere with the assay.

1.0

0.8

0.6

0.4

Absorbance 0.2

0.0 0.0 0.2 0.4 0.6 0.8 Concentration (mg/ml)

Figure A-2. UV-Vis absorbance readings from the Bradford assay at various concentration of (■) chitosan, (●) BSA without chitosan and (▲) BSA at a fixed chitosan concentration of 0.625 mg/mL (λ = 595 nm). The lines are guides to the eye.

192 A.3. pH Drift Upon TPP Addition to Chitosan Solutions

Figure A-3 illustrates the effect of TPP:glucosamine molar ratios and protein

content on the pH drift that occurs upon TPP addition to chitosan. The pH increases by

over a pH unit with the TPP:glucosamine molar ratio, but (with the exception of one data

point in Figure A-2a) is insensitive to the protein content. Notably, the pH increased from

the parent chitosan solution pH (which was 5.5 for Figure A-3a and 6.0 for Figure A-3b)

even when TPP-free water was added to the chitosan mixture (which likely reflected the

dilution of the weakly acidic mixtures).

(a) (b)

7.5 7.5

7.0 7.0

6.5 6.5

pH pH

6.0 6.0

5.5 5.5 0.00 0.05 0.10 0.15 0.20 0.00 0.04 0.08 0.12 0.16 TPP:Glucosamine Molar Ratio TPP:Glucosamine Molar Ratio

Figure A-3. pH drift that occurs with the addition of TPP: (a) during BSA uptake, where the overall BSA concentration is either (■) 0 wt%, (●) 0.05 wt% or (▲) 0.15 wt%; and (b) during α-LA uptake, where the overall α-LA concentration is either (■) 0 wt%, (●) 0.06 wt% or (▲) 0.09 wt%. The error bars are standard deviations while the lines are guides to the eye.

193 A.4. BSA Concentration and Uptake Method Effects on XAgg

Figure A-4a and b show a comparison of the XAgg versus BSA concentration curves obtained at each TPP:glucosamine molar ratio using incubation (Figure A-4a) and incorporation (Figure A-4b) drug uptake methods.

(a) (b)

1.0 1.0

0.8 0.8

0.6 0.6

Agg

Agg

X 0.4 X 0.4 0.2 0.2

0.0 0.0 0.00 0.03 0.06 0.09 0.12 0.15 0.00 0.03 0.06 0.09 0.12 0.15 BSA Concentration ( wt%) BSA Concentration ( wt%) Figure A-4. XAgg versus BSA concentration curves obtained at pH 5.5 and TPP:glucosamine molar ratios of (■) 0, (●) 0.026, (▲) 0.053, (▼) 0.079, (◆) 0.106,

▼

( ) 0.132, ( ▼ ) 0.158:1 where the protein was loaded by the: (a) incubation and (b)

incorporation methods. The error bars are standard deviations and the lines are guides to the eye.

194

A.5. Electrophoretic Light Scattering Measurements

The electrophoretic mobilities and apparent ζ-potentials of the chitosan/TPP

micro- and nanogels prepared from NaCl-free pH 4.0, 5.5 and 6.0 solutions were

determined using the Zetasizer Nano ZS dynamic and electrophoretic light scattering

system (where the apparent ζ-potentials were estimated from the electrophoretic

mobilities using the Smoluchowski equation). To minimize the interference from the

unaggregated chitosan, these measurements were all performed at TPP:glucosamine

molar ratios where the XAgg-values approached 1.0 (i.e., where nearly all the chitosan was

aggregated). Though less anionic TPP needed to be added to aggregate the chitosan at

higher pH-values, both the electrophoretic mobilities and apparent ζ-potentials of the

chitosan/TPP particles decreased with the pH (see Table A.1). This reflected the

deprotonation of both chitosan amine groups (which diminished the positive chitosan

charge) and TPP (which increased the TPP negative charge) at higher pH-levels.

Table A.1. Electrophoretic light scattering data showing the effect of pH on the apparent micro- and nanogel ζ-potentials, each of which was obtained without any added NaCl at TPP:glucosamine ratios where the XAgg-values approached unity at each pH (mean ± standard deviation).

Electrophoretic Mobility Apparent ζ-Potential pH TPP:Glucosamine Molar Ratio XAgg (m2 V-1 s-1) (mV) 4.0 0.23:1 0.98 ± 0.00 (2.68 ± 0.00) × 10-8 36.9 ± 0.1 5.5 0.13:1 0.85 ± 0.07 (2.27 ± 0.03) × 10-8 31.2 ± 0.5 6.0 0.10:1 0.95 ± 0.00 (1.99 ± 0.08) × 10-8 27.5 ± 1.1

195 A.6. Decoupling pH Effects from TPP:Glucosamine Molar Ratio Effects

To confirm that the correlation between AE and XAgg did not simply reflect the upward pH drift with further TPP addition, XAgg and AE-values obtained at the final pH reached after TPP addition (open bars in Figures A-5a - d) were compared with those measured after the pH was restored to the initial values of the parent chitosan mixtures

(i.e., to pH 5.5 for the BSA-loaded particles and pH 6.0 for the α-LA-loaded particles).

This pH-change had no impact on the AE versus XAgg curves (see Figures A-5e and f).

196 (a) (b)

1.2 1.2

pH 6.3 pH 5.5 pH 6.6 pH 6.1 pH 5.5 pH 5.5 0.9 0.9 pH 6.4 pH 6.0 pH 5.5 pH 5.5 pH 6.3 pH 5.5

Agg 0.6 Agg 0.6

X X 0.3 0.3

0.0 0.0 0.079 0.106 0.132 0.042 0.063 0.084 TPP:Glucosamine Molar Ratio TPP:Glucosamine Molar Ratio (c) (d)

100 100 pH 5.5 pH 6.1 pH 6.3 80 pH 5.5 80

pH 6.6 pH 5.5 60 pH 6.0 60 pH 6.4

(%) pH 5.5

(%) pH 5.5

pH 6.3 pH 5.5 40

AE 40 AE

20 20

0 0 0.079 0.106 0.132 0.042 0.063 0.084 TPP:Glucosamine Molar Ratio TPP:Glucosamine Molar Ratio

(e) (f) 100 100

80 80

) ) 60 60

%

%

(

(

AE 40 AE 40

20 20

0 0 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0

XAgg XAgg

Figure A-5. The particle yield comparison of ( ) before and ( ) after adjusting pH back to (a) 5.5 for BSA and (b) 6.0 for α-LA; The association efficiency comparison of ( ) before and ( ) after adjusting pH back to (c) 5.5 for BSA and (d) 6.0 for α-LA. AE vs. XAgg plots obtained from (a - d) for (e) BSA and (f) α-LA (■) before and (●) after pH adjustment. The error bars are standard deviations and the lines are linear fits from Figure 3-5.

197 A.7. TEM Analysis of Chitosan/Protein Complexes The coexistence of small (20 - 30 nm) and large (100 - 1000 nm) chitosan/protein complexes seen in the DLS data (in Figure A-7) was supported by the TEM images of dried chitosan/BSA mixtures. These revealed the existence of many nanoscale chitosan/BSA complexes (Figure A-6a) along with larger chitosan/BSA aggregates, which exceeded 100 nm in size and are evident in the top portion of Figure A-6a and, even to a greater extent, in Figure A-6b. As controls, dried chitosan and BSA solutions were also imaged. The dried chitosan exhibited non-uniformly assembled structure.

Conversely, no structure was evident in obtained from dried BSA solutions (Figure A-6d).

The clear morphological differences between the chitosan/BSA mixtures and control solutions indicated that the structures in Figure A-6a and b were chitosan/BSA complexes which, in good agreement the DLS data, varied from less than 30 nm to hundreds or nanometers in size.

198

(a) (b)

(c) (d)

Figure A-6. Representative TEM images of dried: (a, b) chitosan/BSA complex dispersions; (c) chitosan solutions; and (d) BSA solutions. All of these samples were prepared at pH 6.0 and (like in the case of DLS analysis) subjected to ultracentrifugation prior to imaging.

199 A.8. Defining the Conditions for the Linearization of the AE vs. XAgg Equation (Equation 3.8) ~ To determine how close K and K must be for the AE versus XAgg curve to be adequately described by a linearized form of Equation 3.8, AE versus XAgg curves

Tot predicted by Equation 3.8 were plotted for various K, and CCS -values. When K = ,

Tot the predicted curves were perfectly linear and had slopes that increased with the KCCS

-value (see Figure A-6a). As K and became different, however, the function became progressively more curved, with a concave-up shape when K < (Figure A-6b) and a concave-down shape when K < (Figure A-6c). To propose limits on the K, and

-values where these curves can be described by a linear approximation, linear regressions were performed on curves predicted by Equation 3.8 for a wide range of K,

and -values. These revealed that the goodness of fit (or the R2-value) depended on two factors: (1) the similarity between K and ; and (2) the product . This is illustrated in Figure A-6d, which shows the R2-values to diminish as the quotient K/ moves further from the value of 1.0 and as increases.

200 (a) (b) (c)

(e) (d) 1.0 0.01 100 1 0.1 0.8 10 0.6

2 R 2 ≥ 0.97

~ R 1

0.4 K/K

10 0.1 0.2 100 0.0 0.01 0.01 0.1 1 10 100 0.01 0.1 1 10 100 ~ K/K KCTot CS ~ T ot Figure A-7. Plots showing the effects of K, K and CCS values on the linearity of the AE versus XAgg curves, showing: (a - c) AE versus XAgg curves predicted by Equation 3.8 for various parameter values; (d) the R2-values for the linear regressions to such curves as Tot a function of K/ at varying KCCS -values (indicated by the numbers on the curves); and (e) a diagram that shows the combinations of K/ and -values where these R2-values exceed 0.97.

Based on these trends in R2-values, a range of K, and -values where the

linearized form of Equation 3.8 might be appropriate may be proposed. Here, we do this

by setting the minimum acceptable R2-value to 0.97 (for which deviations from linearity

are only moderate) and, as shown by the shaded region in Figure A-6e, define a range of

K/ and -values for which this fit quality is achieved. This analysis clearly

201 ~ Tot indicates that the curves remain linear over a broader range of K/ K -values when KCCS

<< 1 than at higher -values (see Figure A-6e). To provide more-general recommendations on the similarity between K and for which the linearized form of

Equation 3.8 can be used, the limits on K/ -values are approximated based on the values shown in Figure A-6e at the highest -value. This suggests that the the AE versus

XAgg curve will be roughly linear when 0.6 ≥ K/ ≥ 1.5.

A.9. Deriving Linearizing Form of the AE vs. XAgg Equation (Equation 3-8) Now that the conditions for where linearization of Equation 3.8 is appropriate are identified, a linearized form of this expression must be derived. The simplest approach to doing this would be to simply set K equal to , which yields:

Tot KCCS AE = Tot X Agg (A.1) 1+ KCCS Despite its simplicity, however, this result does not provide the most accurate linear representation of the data when K and are not strictly equal. This can be seen by examining where the line defined by Figure A-1 crosses the non-linear Equation 3.8

(given below):

Tot KCCS X Agg AE = Tot ~ ~ (A.2) 1+CCS [K +(K - K)X Agg ]

As shown in Figure A-7, when K and are not equal, the curve defined by the above equation (black solid line) is intersected by the line defined by Equation A.1 (blue dashed line) only when: (1) XAgg = 0; and (2) XAgg = 1.0. In other words, unlike the linear regression to Equation 3.8 (red dotted line), the slope of this line is unaffected by the

202 curvature in the more-detailed expression and therefore provides an inferior estimate.

1.0

0.8

0.6

AE 0.4

0.2 Crossover w/ best fit line

occurs at XAgg ≈ 0.75 0.0 0.0 0.2 0.4 0.6 0.8 1.0 X Agg Figure A-8. Comparison of linear approximations of a moderately non-linear AE versus XAgg function showing: (black solid line) non-linear curve predicted by Equation 8 for ~ T ot when K = 10 K = 17 mL/mg and CCS = 0.1 mg/mL; (blue dashed line) linearization based on Equation A.1; (green solid line) linearization based on Eqns. S3 and S6; and (red dotted line) linear regression to non-linear curve (R2 = 0.970).

Observations on where the best fit line crosses the curve predicted the by

Equation 3.8, however, indicate that (for relevant K, and -values) the crossover occurs near XAgg ≈ 0.75 (see Figure A-7). Accordingly, we linearize Equation 3.8 by defining an alternative expression:

Tot KCCS AE = Tot X Agg (A.3) 1+ K' CCS where K’ is the weighted average of K and , and the weighting is selected to make the line defined by Equation A.3 (or Equation 3.9 in the manuscript) intersect the curve defined by Equation 3.8 at XAgg = 0.75. In other words, so that it becomes nearly identical to the line of best fit (see green solid line and red dotted lines in Figure A-7). To define K’ for such and expression, the above linear and non-linear expressions for AE (Eqns. A.2 and A.3) are set equal to each other and XAgg is set to 0.75:

Tot Tot KCCS (0.75) KCCS (0.75) AE =  Tot ~ ~ Tot (A.4) 1+CCS [K +(K - K)(0.75)] 1+ K' CCS

203 Comparing these two forms of the AE expression reveals that, in order for the crossover to occur at XAgg = 0.75: ~ ~ K'  K +(K - K)(0.75) (A.5) ~ Simplifying this expression yields the final weighted average of K and K : ~ K'  0.75K  0.25K (A.6)

204

Appendix B

B.1. Evolution in Particle ζ-Potential

Figure B-1 illustrates the evolution in the ζ-potential of chitosan/PPi and chitosan/TPP particles with the addition of either PPi or TPP. It shows that at higher PPi concentrations the ζ-potentials of chitosan/PPi particles drop below the ζ-potential where chitosan/TPP particles rapidly coagulate (shown by the dashed line). This suggests that the colloidal stability of chitosan/PPi particles is not due to greater electrostatic stabilization.

80

70

60

50

40

30

-Potential (mV) -Potential

 20

10 Onset of Coagulation 0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 PPi or TPP Concentration (mM)

Figure B-1. Evolution in particle ζ-potential with the addition of (■) PPi and (●) TPP. The dashed line indicates the ζ-potential at which the chitosan/TPP particles rapidly coagulate. The chitosan/TPP particle data is adapted from Huang and Lapitsky [7].

205

Appendix C

C.1. The α-LA-Loaded Chitosan/TPP Micro- and Nanoparticle Stability

Figure C-1 shows that when the α-LA-loaded chitosan/TPP particles were diluted in 150 mM NaCl and PBS, the portions of FITC-chitosan that dissolved were roughly comparable to those dissolved when drug-free chitosan/TPP particles were used. This occurred regardless of the final chitosan concentration and suggested that α-LA did not increase the chitosan/TPP particle stability to dissolves. Additionally, more than 75% of

α-LA were released from the chitosan/TPP particles within this 1-h experiment 150 mM

NaCl (pH5.5) and 1× PBS solution were used as the dissolution media. This suggested that the failure of α-LA to increase the chitosan/TPP particle stability likely reflected its weak binding to the chitosan and rapid elution.

206

(a) 1.0

0.8

0.6

0.4

0.2

Normalized Absorbance 0.0 0.003 wt% 0.013 wt% 0.031 wt% 150 mM NaCl (pH 6.0)

(b) 1.0

0.8

0.6

0.4

0.2

Normalized Absorbance 0.0 0.003 wt% 0.013 wt% 0.031 wt% 1X PBS (pH 6.0)

(c) 1.0

0.8

0.6

0.4

0.2

Normalized Absorbance 0.0 0.00 0.01 0.02 0.03 Overall Chitosan Conc. (wt%)

Figure C-1. Normalized FITC-chitosan UV-vis absorbance of ( ) FITC-chitosan solutions, ( ) FITC-chitosan/TPP particle dispersions and ( ) α-LA-loaded FITC-chitosan/TPP particle dispersions after centrifugation in (a) 150 mM NaCl and (b) PBS; (c) Normalized Micro BCA assay UV-vis absorbance of α-LA-loaded FITC-chitosan/TPP particle dispersions after centrifugation in (●) 150 mM NaCl and (▲) PBS. The error bars are standard deviations.

207 Appendix D

D.1. Chitosan/TPP Particle Effect on Protein Concentration Determination

Figure D-1 shows that the presence of chitosan/TPP particles has virtually no impact on the protein absorbance signal obtained using the Micro BCA protein assay.

These measurements were performed on a 0.5 mg/mL BSA solution at pH 5.5 either with:

(1) no chitosan/TPP particles present; and (2) chitosan/TPP particles prepared at the

0.13:1 TPP:glucosamine molar ratio and an overall chitosan concentration of 0.06 wt%.

Because a significant fraction of the BSA gets taken up by the particles when chitosan/TPP particles are present (see Chapter 3), this invariance in absorbance shows that the Micro BCA assay does not distinguish between the free protein and chitosan/TPP particle-loaded protein.

1.0

0.8

0.6

Absorbance 0.4

0.2

Micro BCA 0.0 With particles Without particles

Figure D-1. Effect of chitosan/TPP micro- and nanoparticles (added in a 0.06 wt% final chitosan concentration) on the Micro BCA absorbance obtained from 0.5 mg/mL BSA solutions in pH 5.5 water after 20× dilution and incubation at 37 °C for 2 h. The error bars are standard deviations.

208 D.2. Determination of BSA-Loaded Chitosan/TPP Gel Pellet Density

To test the density of BSA-loaded chitosan/TPP gel pellets that formed during the centrifugation process, a “suspending method” was used. Specifically, the gel pellet was put into a glycerol/water mixture. If the gel pellet density matched that of the glycerol solution, the gel pellet became suspended rather than floating or sinking. Once the density-matched water/glycerol mixture was identified, the density of gel pellet was determined by measuring the density of that glycerol solution. To do this, a 50 wt% glycerol solution was prepared, which at 25 °C corresponded to a density of 1.12 g/cm3.

The BSA-loaded chitosan/TPP particle gel pellet was prepared by centrifuging a particle dispersion (prepared at the 0.13:1 TPP: glucosamine molar ratio and an overall chitosan concentration of 0.06 wt% and BSA concentration of 0.5 mg/mL) at 3.0 × 105 g and

37 °C for 1.5 h. After putting the gel pellet in the 50 wt% glycerol solution, the gel pellet floated to the top, indicating that the gel pellet density was lower than 1.12 g/cm3. DI water was then added into glycerol solution until the gel became suspended (which occurred in a 34 wt% glycerol solution), whereupon the gel pellet density was determined from a glycerol solution density vs. glycerol concentration table [217]. This yielded a chitosan/TPP gel pellet density of 1.08 g/cm3.

D.3. Centrifugation Effect on Particle Size Distributions in the Presence of a Glycerol Bed

The BSA-loaded chitosan/TPP particles were first prepared using the method described in the Section 6.2.3. To test if the presence of glycerol in the particle dispersion could change the particle size distribution, 150 mg of glycerol was added into 5 mL particle dispersion (prepared at the 0.13:1 TPP: glucosamine molar ratio and an overall

209 chitosan concentration of 0.06 wt% and BSA concentration of 0.5 mg/mL). After being vortex-mixed for 20 s, the particle size distribution was characterized by DLS. Since the glycerol concentration after dispersion was about 3 wt%, it has only a marginal impact on the viscosity change (e.g., the viscosity of 10 wt% glycerol solution and water are 1.31 and 1.00 cP, respectively, at 20 °C [217]). Thus, the presence of glycerol does not have a strong impact on the estimation of particle size via the DLS. To test whether the centrifugation and redispersion treatment with the assistance of a glycerol bed could affect the particle size distribution, a 5-mL chitosan/TPP particle dispersion containing a glycerol bed was centrifuged at 3.0 × 105 g at 37 °C for 1 h. The resulting pellet was then redispersed by being vortex-mixed for 20 s, and characterized by DLS after being cooled down to room temperature. The volume-weighted size distributions for these samples were very close, showing that ultracentrifugation in the presence of a glycerol bed did not greatly affect the particle size (Figure D-2).

15

10

5 Volume (%)

0 1 10 100 1000 10000 Hydrodynamic Diameter (nm)

Figure D-2. Representative volume-weighted size distributions of BSA-loaded chitosan/TPP particles in (■) glycerol-free water before centrifugation, (●) glycerol solution before centrifugation and (▲) glycerol solution after a centrifugation and redispersion treatment. The glycerol solutions were obtained by vortex-mixing the glycerol beds into the aqueous dispersion, which is done to redisperse the sedimented particles.

210 D.4. Centrifugation Procedure Effects on the BSA Recovery with FITC- Chitosan/TPP Particles

Figure D-3 shows the BSA and chitosan recovery for FITC-chitosan particles

(prepared at the 0.13:1 TPP:glucosamine molar ratio, an overall FITC-chitosan concentration of 0.06 wt% and BSA concentration of 0.5 mg/mL) obtained at variable centrifugation forces and times. The similarity between BSA and chitosan recovery curves agrees with the understanding (developed in Chapter 3) that the BSA recovery within the chitosan/TPP particles should be proportional to the chitosan recovery.

(a) (b) 100 100

80 80

60 60

40 40

20 20

BSA Recovery (%) BSA Recovery BSA Recovery (%) BSA Recovery

Chitosan Recovery (%) Chitosan Recovery

0 (%) Chitosan Recovery 0 0 1x105 2x105 3x105 4x105 0.0 0.5 1.0 1.5 2.0 2.5 Centrifugal Force (g) Centrifugal Time (h)

Figure D-3. The effects of (a) centrifugal force and (b) centrifugation time on (■) BSA and (●) chitosan recovery achieved with FITC-chitosan/TPP particles. The centrifugation force effects were investigated using a 30 min centrifugation time, while the centrifugation time effects were probed at a constant centrifugal force of 3.0 × 105 g. The error bars (which are largely obscured by the symbols) are standard deviations, while the lines are guides to the eye.

D.5. Change in Chitosan/TPP Gel Pellet Mass during Release

To analyze whether swelling or degradation/dissolution affect drug release from macroscopic chitosan/TPP gel pellets, the gel pellet mass was tracked during the release process. This experiment was performed in parallel with the release experiment where, after the removal of the supernatant (at each time point used in the release experiment),

211 the gel mass was measured. The change in the gel mass was calculated as:

m Gel Mass (%)  t  100% (D-1) m0 where m0 was the initial gel pellet mass and mt was the gel pellet mass at time t. Since the model analysis of the release profile only analyzed the first 60% of the drug release profile, this analysis was only conducted for 2 h (which covered all the time points used in the model analysis). These measurements revealed that the gel mass did not change much during the release experiments, regardless of the centrifugation conditions used.

Thus, swelling and degradation did not appear to strongly affect the release process, and protein release from the gel pellet was likely diffusion-controlled.

120

100

80

60

40

Gel Mass (%) Gel Mass 20

0 0.0 0.5 1.0 1.5 2.0 2.5 Time (h)

Figure D-4. Changes in chitosan/TPP pellet mass during the release process obtained using 30 min of centrifugation at (■) 4.8 × 104, (●) 1.1 × 105 and (▲) 3.0 × 105 g, or 60 min of centrifugation at (▼) 3.0 × 105 g. The dashed line is a guide to the eye and the error bars are standard deviations.

212 D.6. Effects of Non-Uniform Initial Protein Distributions within Chitosan/TPP Gel Pellets

To explore a possible reason for the zero-order release profiles from pellets produced using less-extensive centrifugation, the protein release process was modeled by converting Equation 6-7 into its dimensionless form: ~ ~ ~ C  2C 1 C ~ = 2  (D-2) t ~z ~z ~z

~ ~ 2 ~ where t is the normalized time ( t  Dt/z ),C is normalized protein concentration (where

~ ~ C  C / C0 and C0 is original protein concertation), and z is normalized diffusion length

~ ~ ~ (where z  z / z gel and z ge l is the gel thickness). For = 0, C / z = 0 and for = 1, =

0. By numerically solving the above differential equation via finite difference approximation (FDA) using Matlab, the release profile can be obtained as:

1 ~ ~ ~ ~ ~ C(t , z )Ac (z )dz M t 0 =1- (D-3) 1 ~ M ∞ ~ ~ ~ C(0, z )Ac (z )dz 0

~ ~ ~ ~ ~ where C(0, z ) is the initial protein concentration profile, C(t , z ) is the protein

~ concentration profile at time, and Ac (z) is the z-dependent circular sectional area of the

gel (see Figure 6-8), which, as shown in Section 6.3.2.3, equals to 2πRz .

Assuming that the protein diffusivity and initial protein concentration within the gel pellet are constant, the n-value (i.e., the time scaling) obtained by fitting the model release profile (obtained from Equation D-3) with the Korsmeyer-Peppas equation

(Equation 2-6) should equal to 0.45 for diffusion-controlled release. This case is illustrated in Figure D-5, which shows the evolution in the normalized protein

213 concentration profile with time (Figure D-5a) and its corresponding release profile (solid line in Figure D-5c).

(a)

1.0

0.8

0.6

~ C 0.4

0.2

0.0 0.0 0.2 0.4 0.6 0.8 1.0 (b) z~

1.0

0.8

0.6

~ C 0.4

0.2

0.0 0.0 0.2 0.4 0.6 0.8 1.0 (c) ~z

1.0

0.8

0.6

∞

M

/

t 0.4 M

0.2

0.0 0.0 0.2 0.4 0.6 0.8 1.0 ~t

Figure D-5. Normalized protein concentration plotted as a function of normalized position, at normalized times of (black solid line) 0, (red dashed line) 0.1, (blue dotted ~ line) 0.2 and (green dash-dot line) 0.6 for the cases when Csurf equals to (a) 1 and (b) 0; (c) the release profile for the case when equals to (black solid line) 1 and (red dashed line) 0.

214 Conversely, when the initial protein concentration was nonuniform and decreased

~ ~ ~ with z (as C(0, z ) = 1- ), the n-value increased. When the protein diffusivity was held

~ constant while the normalized protein concentration at the top surface of the pellet ( Cs ur f ) decreased to 0.75, such that decreased linearly from 1 to 0.75 from the bottom to the top of the gel, for example, the n-value increased to 0.49 (see Table D-1). The n-value further increased to 0.71 when dropped to 0. This case was illustrated by Figures

D-5b and c, where the release was slower and closer to a zero-order release (for the first

60% of the release profile) compared to the case where the protein distributed homogeneously ( = 1). This indicated that the increased n-values of the

Korsmeyer-Peppas equation that fit to experimental release profiles can be a result of the non-uniform protein distributions within the gel pellets. This non-uniform protein distribution might be caused by a non-uniformed gel compression in the centrifugation process, where the protein concentrations at the bottoms of the gel pellets were high because of more-compact structures and the protein concentration at the surfaces of the gel pellets were low because the tops of the pellets were composed of more-loosely coagulated micro- and nanoparticles coagulations.

215 Table D-1. The n-values (obtained from Equation 2-6) fitted to the normalized release ~ profiles (from Equation 6-7) at variable Cs ur f -values. In each case, the initial protein concentration profiles within the pellets were assumed to be linear.

0.75 0.50 0.25 0

n 0.49 0.52 0.60 0.71

D.7. Chitosan/TPP Gel Pellet Size Effect on the BSA Release

To confirm that the BSA release kinetics were sensitive to the diffusion path length, chitosan/TPP particle gel pellets prepared by centrifuging chitosan dispersions at

4.8 × 104 g for 30 min were cut into smaller, 0.1 - 1 mm pieces. Release from these gel pellet fragments into PBS was then then characterized by the same procedure as that used for whole gel pellets, whereupon the release profiles obtained using fragmented and whole gel pellets were compared (see Figure D-7). As expected, the BSA release from pellet pieces was faster than from the whole gel pellet because of the reduced diffusion path length. This further demonstrated that the multiple-hour protein release observed in some studies on chitosan/TPP micro- and nanoparticles could have been due to irreversible particle coagulation during centrifugation.

216

100

80

100 60

75

40 50

25 % Released

20 Released % 0 0 1 2 3 0 Time (h) 0 10 20 30 40 50 Time (h) Figure D-6. BSA release profiles from chitosan/TPP particle pellets prepared by centrifuging for 30 min at 4.8 × 104 g obtained (■) from whole pellets and (●) pellets cut into 0.1 - 1 mm pieces. The inset focuses on the first 3 h of the release profiles. The lines are guides to the eye while the error bars are standard deviations.

217