Polymer Hydrogels, Aerogels, and Foams As Biomimetics of the Extracellular Matrix

Polymer Hydrogels, Aerogels, and Foams as Biomimetics of the Extracellular Matrix

Mo Kit Chau

A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy Department of Chemistry University of Toronto

Polymer Hydrogels, Aerogels, and Foams as Biomimetics of the Extracellular Matrix

Mo Kit Chau

Doctor of Philosophy

Department of Chemistry University of Toronto

2016 Abstract

Three-dimensional (3D) scaffolds that recapitulate the mammalian extracellular matrix

(ECM) have potential applications as tissue engineering scaffolds and as in vitro cell culture material. This thesis describes the fabrication and characterization of microgels, aerogels, hydrogels, and foams that are biomimetics of the natural ECM. The ability to tune the mechanical property and structure of the resulting scaffolds is of particular importance. In

Chapter 3, we explore a microfluidic platform for making biopolymer composite microgels in which the rigidity and structure can be tuned in high-throughput manner. Chapter 4 describes how cellulose nanocrystals can be used to make nanofibrillar hydrogels, in which the rheological properties and pore sizes can be controlled by the addition of inorganic salts. These properties are highly dependent on the size, charge, and concentration of the cations added.

Chapter 5 describes the freeze-casting of aldehyde-functionalized cellulose nanocrystals and hydrazide-functionalized poly(oligoethylene glycol methacrylate) into anisotropic composite aerogels and their corresponding hydrogels. The aerogel structure could be tuned from fibrillar to columnar and lamellar by varying the freeze-cast composition. The Young’s modulus and swelling of the hydrogels differed parallel and perpendicular to the freeze-cast direction.

Chapter 6 describes the freeze-casting of polyurethanes into anisotropic open-cell foams, which

ii have anisotropic mechanical and thermal properties. The thermal conductivity of these foams could also be switched between anisotropic and isotropic modes.

Key words: Hydrogel, microgel, foam, aerogel, extracellular matrix, thermal management, freeze-casting, polyurethane, biopolymer, polymer, anisotropic, cellulose nanocrystals, colloidal dispersion, agarose, gelatin, rheology, micropipette aspiration, mechanical properties, structure

iii

Acknowledgments

I want to thank my supervisor, Professor Eugenia Kumacheva for her enthusiasm, motivation, and guidance throughout the years. You are an excellent researcher and teacher. I am grateful for the lessons you have given me in science and in life. I am always amazed by how you can manage so many diverse projects with such vision and success. You work hard to move projects forward, leading by example. As I said when we first met, I am a fan of your work and I always will be. It is an honor for me to have been in your group.

I want to show my upmost appreciation for my committee members and teachers: Professor Mitch Winnik and Professor Dwight Seferos.

Professor Winnik, you provided me the mental support during the most critical time of my PhD giving me the strength to carry on. Your scientific insight particularly with regards to the polyurethane project, was imperative for its completion. I want to say a heartfelt thank you. Your kindness will always be remembered.

Anna Liza Villavelez, who helped me navigation through this program, I appreciate your years of diligence and your care for the students you work with.

A special thanks to Professor Markus Retsch and his group for hosting me in Bayreuth Universität. I appreciate your hospitality and openness. To Professor Retsch, I really enjoyed the discussions we had about science and the intricacies of manuscript writing. I have benefited much from your guidance. I really like that you encourage discussion even/especially when the views are disparate. I think that highlights the spirit of how science should be. Thanks to Alexandra Phillips who trained me on the XFA, and Fabian Nutz and Patrick Hummel who performing experiments for us. Thanks to Dr. Sabine Rosenfeldt for her expertise in. I’d like to thank Bernd Kopera for his critical scientific contributions to this thesis. Your high IQ, wide knowledge base, and lack of fear for complicated equations gives me confidence to believe that you will excel in whatever you chose to do in the future.

I want to thank NSERC CREATE IDEM for their years of funding and the NSERC CREATE IDEM group for encouraging a creative and collaborative. Thanks to Professor Harald Stover in particular for arranging this. I have enjoyed the wonderful NSERC CREATE meeting we had. They were fun and educational. I want to thank Professor Emily Cranston and Professor Todd iv

Hoare for collaborating with us on the anisotropic hydrogel project. Kevin de France. Thank you for being an excellent partner to work with. Our tag-team approach was efficient and successful, allowing us to complete one task after another.

Big thanks to the Ramachandran group, in particular Professor Arun Ramachandran, Suraj Borkar, Shashi Malladi, Rohit Sonthalia, and Ali Hussain Motagamwala, for their help on the pipette aspiration experiments.

I would also like to thank everyone in the Kumacheva group for their continued support. It was a pleasure working with you all. Thanks to Diego Velasco and Ethan Tumarkin for guiding in me during my initial time in the group. A special thanks to Dr. Héloï se Thérien-Aubin, mentor, desk mate, and friend. You have been generous in sharing with me your vast knowledge. You have really been there for me over last five years. A special thanks Shivanthi Sriskandha. You are a hard-working and reliable individual. I think back to our time together, doing chemistry and occasionally being silly, with smiles and laughs (Oh, Heathcliff!). Vanessa Machado, thank you for working with me during my last year. You were immensely helpful.

Big thanks to Sepehr Tehrani who took time to help us investigate polyurethanes by DSC.

I would like to thank Ilya Gourevich, Dr. Neil Coombs, Dr. Battista Calvieri, and Dr. Steven Doyle for their technical support on electron microscopy. Also a thank you to the machine shop staff: Johnny Lo, John Ford, David Heath, Ahmed Bobat for their masterful creations for the freeze-casting projects. I’d like to thank staff in the NMR lab, Dima Pichugin and Darcy Burns, who are extremely competent at their job.

I want to thank my original mentors, Dr. Robin Stoodley and Dr. Guillaume Bussiere who gave me my first shot at chemistry, and Professor Michael Wolf who initiated my interest in materials chemistry. To Mike: as I age, I appreciate more and more what you have done for me as a mentor and how you put the student’s well-being first.

I would like to thank my friends from the department and friends back home for the years of encouragement and moral support.

Finally and very importantly, I want to thank my family for raising me and providing me their unconditional love. To my mother, Ada Yip, who trained my mental discipline and stamina; my

v father, Chak Chau, who endured hard labor to keep food on our table; my brother, Yu-Hang Chau, who aided my transition to Toronto in every way he could; and Aunty Lisa, who support us financially through the most difficult times of my childhood: I write this thesis in your honor.

Preface

This thesis has been organized as a series of manuscripts (see the list below) which have either been published in peer-reviewed scientific journals or is in the process of submission. As identified by primary authorship, all manuscripts were written by Mo Kit Chau with critical comments and revision by Eugenia Kumacheva and corresponding collaborators. The contributions of other authors are provided in detail below.

Chapter 1 Polymer scaffolds as mimetics of the extracellular matrixes

The results in this chapter are partly from manuscripts published in Supramolecular Nanofibrillar Polymer Hydrogels in Supramolecular Polymer Networks and Gels, Springer 2015.

Authors: Mokit Chau, Shivanthi Sriskandha, Héloï se Thérien-Aubin, Eugenia Kumacheva

Contributions: M. Chau contributed to the article writing and figure design. S. Sriskandha contributed to the article writing and figure design. H. Thérien-Aubin contributed to the article writing.

Chapter 3 Microfluidic Generation of Composite Biopolymer Microgels with Tunable Compositions and Mechanical Properties

The results in this chapter are mainly from manuscripts published in Biomacromolecules, 15, 2013.

Authors: Mokit Chau, Milad Abolhasani, Héloï se Thérien-Aubin, Yang Li, Yihe Wang, Diego Velasco, Ethan Tumarkin, Arun Ramachandran, Eugenia Kumacheva

Contributions: M. Chau contributed to the carrying out experiments, data analysis and interpretation, and article writing. M. Abolhasani, H. Thérien- Aubin, Y. Li, and Y. Wang helped with microfluidic experiments. D. Velasco and E. Tumarkin provided guidance.

vii

Chapter 4 Ion-Mediated Gelation of Aqueous Suspensions of Cellulose Nanocrystals

The results in this chapter are mainly from manuscripts published in Biomacromolecules, 16, 2013.

Authors: Mokit Chau, Shivanthi Sriskandha, Dmitry Pichugin, Héloï se Thérien-Aubin, Dmitro Nykypanchuk, Greǵory Chauve, Myriam Méthot, Jean Bouchard, Oleg Gang, Eugenia Kumacheva

Contributions: M. Chau contributed to the carrying out experiments, data analysis and interpretation, and article writing. S. Sriskandha contributed to carrying out experiments and data analysis. D. Pichugin and H. Thérien-Aubin contributed to the NMR experiment and interpretation. D. Nykypanchuk and O. Gang contributed to the SAXS experiments and interpretation. G. Chauve, M. Méthot, J. Bouchard provided the raw materials.

Chapter 5 Anisotropic Hydrogels Derived from Cellulose Nanocrystals

The results in this chapter are mainly from a manuscript in preparation.

Authors: Mokit Chau, Kevin J. De France, Bernd Kopera, Vanessa R. Machado, Sabine Rosenfeldt, Laura Reyes, Katelyn J. W. Chan, Stephan Förster, Emily Cranston, Todd Hoare, Eugenia Kumacheva

Contributions: M. Chau contributed to the manuscript by designing and carrying out experiments, data analysis and interpretation, and article writing. B. Kopera helped with the freeze-casting setup design. K. J. De France and K. J. W. Chan prepared POEGMA solutions and CNC suspensions. K. J. De France performed the compression tests. S. Rosenfeldt and B. Kopera performed the SAXS experiments. V. Machado prepared some of the freeze-casted samples and performed the swelling experiments. E. Cranston, T. Hoare, and S. Förster

viii

provided guidance and suggestions on experimental design, data interpretation, and article writing.

Chapter 6 Elastomeric polyurethane foams with anisotropic structure and direction- dependent thermal conductivity

The results in this chapter are mainly from a manuscript in preparation.

Authors: Mokit Chau, Bernd Kopera, Vanessa R. Machado, Sepehr Tehrani, Mitchell A. Winnik, Eugenia Kumacheva, Markus Retsch

Contributions: M. Chau contributed to the manuscript by designing and carrying out experiments, data analysis and interpretation, and article writing. B. Kopera contributed to the thermal measurements and polyurethane synthesis development. XRD experiments were performed by W. Milius. V. Machado contributed to some of the sample preparation, HRSEM and TEM imaging. XPS was performed by R. Sodhi. Cyro- TEM imaging was performed by M. Drechsler. F. Nutz performed the DSC experiments. P. Hummel performed the TGA experiments. M. Retsch, E. Kumacheva, M. A. Winnik provided guidance and suggestions on experimental design, data interpretation, and article writing.

Publications during Ph.D. Studies

Peer Review Publications

1. M. Chau, B. F. A. Kopera, V. R. Machado, S. Tehrani, M. A. Winnik, E. Kumacheva, M. Retsch, Reversible transition between anisotropic and isotropic thermal conductivity in elastic polyurethane foams, manuscript in preparation.

2. M. Chau*, K. J. D. France*, B. Kopera, V. R. Machado, S. Rosenfeldt, L. Reyes, K. J. W. Chan, S. Förster, Emily Cranston, T. Hoare, E. Kumacheva, Anisotropic Hydrogels Derived from Cellulose Nanocrystals. Chemistry of Materials, 2016, 28, 3406–3415.

3. M. Chau, S. E. Sriskandha, D. Pichugin, H. Thérien-Aubin, D. Nykypanchuk, G. Chauve, M. Méthot, J. Bouchard, O. Gang, E. Kumacheva, Ion-mediated gelation of aqueous suspensions of cellulose nanocrystals. Biomacromolecules, 2015, 16, 2455- 2462.

4. M. Chau*, S. Sriskandha*, E. Kumacheva, Supramolecular nanofibrillar polymer hydrogels, In Supramolecular Polymer Networks and Gels. S. Seiffert, Ed., Adv. Polym. Sci., Springer International Publishing, Switzerland, 2015, 167-208.

5. M. Chau, M. Abolhasani, H. Thérien-Aubin, Y. Li, Y. Wang, D. Velasco, E. Tumarkin, A. Ramchandran, E. Kumacheva, Microfluidic generation of composite biopolymer microgels with tunable compositions and mechanical properties. Biomacromolecules, 2014, 15, 2419-2425.

6. D. Velasco, M. Chau, H. Thérien-Aubin, A. Kumachev, E. Tumarkin, Z. Jia, G. C. Walker, M. J. Monteiro, E. Kumacheva, Nanofibrillar thermoreversible micellar microgels. Soft Matter, 2013, 9, 2380-2383.

Table of Contents

Acknowledgments ...... iv

Preface ...... vii

Publications during Ph.D. Studies ...... x

Table of Contents ...... xi

List of Tables ...... xv

List of Figures ...... xvi

List of Abbreviations ...... xxvi

Chapter 1 Polymer Scaffolds as Artificial Extracellular Matrices ...... 1

Introduction ...... 1

1.1 The natural ECMs ...... 1

1.2 Mimicking natural ECMs ...... 4

1.3 Nanofibrils as building blocks of the ECM ...... 5

1.4 Artificial nanofibrils as building blocks of artificial ECMs ...... 7

1.4.1 Driving forces for formation of nanofibril building blocks from polymer molecules ...... 8

1.4.2 Driving forces for the formation of a nanofibrillar network...... 9

1.4.3 Artificial nanofibrillar ECM made from synthetic and natural building blocks ...... 10

1.4.4 Nanofibrillar hydrogels of agarose ...... 11

1.4.5 Cellulose nanocrystals as a building blocks for nanofibrillar hydrogels ...... 12

1.5 Microgels for tissue engineering ...... 15

1.6 Freeze-casting method to make anisotropic monolithic scaffolds ...... 16

1.7 Polyurethane scaffolds as mimics of the ECM ...... 19

Chapter 2 Materials and Methods ...... 22

Experimental ...... 22

2.1 Materials ...... 22 xi

2.1.1 Materials for the microfluidic generation of agarose and gelatin composites microgels...... 22

2.1.2 Materials for the investigation of the ionic gelation of cellulose nanocrystals ...... 23

2.1.3 Materials for the formation of anisotropic hydrogels ...... 23

2.1.4 Materials for the formation of anisotropic polyurethane foams ...... 23

2.2 Methods ...... 24

2.2.1 Methods for the microfluidic generation of agarose and gelatin composites microgels...... 24

2.2.2 Methods for the investigation of the ionic gelation of cellulose nanocrystals28

2.2.3 Methods for the formation of anisotropic freeze-cast aerogels and hydrogels...... 32

2.2.4 Preparation and characterization of anisotropic polyurethane foams ...... 36

Chapter 3 Microfluidic generation of composite biopolymer microgels with tunable compositions and mechanical properties ...... 44

Introduction ...... 44

3.1 Results and Discussion ...... 47

3.1.1 Design of the Microfluidic Device ...... 47

3.1.2 Gelation Time of Gelatin-Ph ...... 49

3.1.3 Microfluidic Generation of Composite Microgels ...... 49

3.1.4 Composition of the Droplets ...... 52

3.1.5 Microgel Morphology ...... 55

3.1.6 Microstructure of Composite Gels...... 58

3.1.7 Mechanical Properties of Composite Microgels...... 59

3.2 Conclusions ...... 62

Chapter 4 Ion-Mediated Gelation of Aqueous Suspensions of Cellulose Nanocrystals ...... 64

Introduction ...... 64

4.1 Results ...... 66

xii

4.1.1 Ionically Mediated Gelation of CNC Suspensions ...... 66

4.1.2 Rheological Properties of CNC Gels ...... 72

4.1.3 Characterization of Hydrogel Structure ...... 78

4.1.4 Characterization of the Gel Structure by Small-Angle X-ray Scattering ...... 84

4.2 Discussion ...... 86

4.3 Conclusions ...... 87

Chapter 5 Anisotropic Hydrogels Derived from Cellulose Nanocryals ...... 88

Introduction ...... 88

5.1 Fabrication and microstructure of anisotropic aerogels...... 90

5.2 Examination of the surface area of aerogels ...... 94

5.3 Small-angle X-ray scattering in aerogels ...... 95

5.4 Swelling behavior of anisotropic hydrogels ...... 100

5.5 Mechanical properties of anisotropic hydrogels ...... 102

5.6 Conclusions ...... 106

Chapter 6 Reversible transition between anisotropic and isotropic thermal conductivity in elastic polyurethane foams ...... 107

Introduction ...... 107

6.1 Synthesis and characterization of polyurethane ...... 108

6.2 Characterization of CB and CNF additives ...... 113

6.3 Fabrication and morphology of freeze-cast PU foams ...... 117

6.4 Mechanical properties of PU foams ...... 125

6.5 Thermal properties of PU foams ...... 129

6.6 Conclusion ...... 134

Chapter 7 Conclusion, Summary, and Future Works ...... 136

Conclusion ...... 136

7.1 Microfluidically generated biocomposite microgels. Summary and future works ... 136

xiii

7.2 Nanofibrillar hydrogels. Summary and future works ...... 137

7.3 Anisotropic hydrogels. Summary and future works ...... 137

7.4 Anisotropic polyurethane foams. Summary and future works ...... 138

References...... 140

xiv

List of Tables

Table 4.1 Nomenclature used for the CNC samples and characteristics of the cations...... 68

Table 4.2 Rheological properties and mesh size of CNC gels* ...... 74

Table 4.3 Diffusion coefficients of dextran in D2O and in CNC hydrogels ...... 81

Table 5.1 Recipes of freeze-cast aerogels and hydrogels and Young’s moduli of the rehydrated hydrogels...... 92

Table 6.1 Peak fits and assignments from the C1s spectra of CB...... 115

Table 6.2 Peak fits and assignments from the C1s spectra of CNF...... 115

Table 6.3 Thicknesses of the lamellae, inter-lamellar distances, and cp (at 25 ºC) for various PU foams...... 119

List of Figures

Figure 1.1 Simplified cartoon of a cell (green color) interacting with its ECM. Fibrous matrix proteins, such as collagen, fibrin, and elastin, provide structural support, as well as mechanical and biochemical cues, to direct cell behavior. Hydrophilic proteoglycans (proteins functionalized with polysaccharide) provide a gel-like matrix. Both the protein and proteoglycan components contain arginylglycylaspartic acid (RGD) peptides to which integrins (transmembrane receptors on the cell) can bind to for cell adhesion. The ECM also contains degradation enzymes which cleave the matrix components during cell motility and matrix remodeling. Reproduced with permission from Public Library of Science.4 ...... 2

Figure 1.2 Structures of the ECMs of ventricular myocardium, endomysium, and adipose tissue. SEM image of the ventricular myocardium containing lamellae of organized myocytes in a collagen matrix (scale bar is 50 μm). Adapted from American Physiological Society.6 b) SEM image of the columnar structure of endomysium of intramuscular connective tissue in semitendinosus muscle (scale bar 10 μm). Adapted with permission from Karger Publishers.11 c) SEM image of decellularized adipose tissue ECM with a collagen network architectures (scale bar 1.25μm). Adapted from reference 12 ...... 3

Figure 1.3 The biosynthetic route to collagen fibers. Reproduced with permission from Annual Reviews.20 ...... 6

Figure 1.4 Representative stress–strain curve of the collagen matrix hydrogel (polymer content 2 g L-1, pH 7.4) tested at a strain rate of 38.5 % per min. The stress–strain curve is separated into three distinct regions designated as the “toe”, “linear”, and “failure” regions. Adapted with permission from American Society of Mechanical Engineers.26...... 7

Figure 1.5 Hierarchical assembly of nanofibrillar hydrogels. Individual molecules organize into nanofibrils, which subsequently associate and/or entangle to form a 3D water-swollen network. Biopolymer molecules are often oriented parallel to the long axis of the nanofibril, whereas synthetic polymer chains are oriented perpendicularly to the main axis. Gelation can be triggered by changes in temperature and pH, an increase in polymer concentration, or an increase in ionic strength of the polymer solution, to name a few. Reproduced with permission from Springer.33 ...... 8

xvi

Figure 1.6 Structure of nanofibrils and mechanical properties of agarose hydrogels. a) Cartoon of agarose gel network made by association of double helices and amorphous domains. Reprinted (adapted) with permission from Reference 51. Copyright 1989 American Chemical Society. b) Variation in the elastic modulus over time of 2 % (w/w) aqueous agarose gels that were cured isothermally at temperatures Tf, as indicated. Samples were cooled from 90 to 10 ºC at a rate of 1 ºC min-1. Reproduced with permission from Wiley-VCH.49 ...... 12

Figure 1.7 a) Simplified schematic of the acid hydrolysis of cellulose microfibrils to cellulose nanocrystals. b—e) TEM micrographs of dispersion of cellulose nanocrystals derived from different sources: b) microcrystalline cellulose (Avicel), c) tunicate, d) green algae, and e) ramie. Reprinted (adapted) with permission from 57. Copyright 2014 American Chemical Society. Adapted with permission from American Chemistry Society...... 13

Figure 1.8 a) Schematic of a MF device for generating agarose microgels with tunable elasticity. The width of the horizontal channel supplying the mineral oil phase was 150 μm, and the width of the serpentine channel at the T-junction was 150 μm. Fluorescence optical microscopy images of the agarose microgels encapsulating murine embryonic stem cells in buffer. The scale bar is 100 mm. Adapted with permission from Elsevier.78 ...... 16

Figure 1.9 Directional freeze-casting of ceramics involving slurry preparation, solidficiation, sublimation and sintering. This procedure can be extended to other particles, monomer, and polymers as well. The freeze-casting for the production of aerogels follow a similar process except instead of sintering, the aerogels are rehydrated with water. Reproduced with permission from Elsevier.92 ...... 18

Figure 1.10 Parameters that affect freeze-cast structures. Reproduced with permission from The Royal Society of Chemsitry.93 ...... 19

Figure 3.1 a) Schematic of the MF device used for the generation of composite agarose-gelatin- Ph microgels. b) Enlargement of the green boxed area shown in a). The width of the microchannel carrying aqueous solutions and the continuous phase were 80 μm prior to the 80 μm-wide orifice. The width of the main channel downstream of the orifice was 640 μm. The height of the channels in the MF device was 130 μm. The labels 1, 2, 3, and 4 refer the the inlets

xvii and channels dedicated to fluorinated oil, gelatin-Ph solution, agarose solution, and crosslinker solution, respectively...... 48

Figure 3.2 Variation of gelation times of gelatin-Ph solutions, plotted as a function of the concentration of gelatin-Ph containing 1 (diamonds) and 5 (squares) units/mL of HRP at 37 C. The concentration of hydrogen peroxide was maintained at 1 mM...... 49

Figure 3.3 Emulsification of the liquid stream containing agarose, gelatin−Ph mixed with hydrogen peroxidase, and a cross-linker. Qcont = 0.6 mL/h, and Qcross = Qag = Qgel = 0.02 mL/h. The scale bar is 250 μm...... 50

Figure 3.4 (a) Optical microscopy image of the composite microgels generated from droplets produced at Qag = Qgel = Qcross = 0.2 mL/h and Qcont = 2.5 mL/h. The scale bar is 500 μm. (b) Distribution of the diameters of microgels shown in panel (a)...... 51

Figure 3.5 Viscosity of agarose-gelatin mixtures in HBSS were measured at 37 C as a function of ratio of gelatin to agarose by weight. Total concentration biopolymer was kept at 4% w/w. The shear rates used were 7.34 ( ), 14.68 ( ), 36.69 ( ), and 73.38 ( ) s-1...... 53

Figure 3.6 Calibration curve used for the determination of dye concentrations in droplets...... 54

Figure 3.7 Variation in the concentration of agarose in the precursor droplets (square symbols), plotted as a function of the relative flow rate of the agarose solution. The Qag/Qgel ratio changed from 0.25 to 4 at constant Qag + Qgel + Qcross = 0.6 mL/h, Qcross = 0.2 mL/h, and Qcont = 0.6 mL/h. The dashed line represents the theoretical concentration of agarose (determined using Equation 3.2). The solid line is the best linear fit for the experimental data...... 55

Figure 3.8 Representative confocal fluorescence microscopy images of microgels with different compositions, Cag‐FITC/Cgel‐Ph: (a) 1.43/0, (b) 0.95/0.72, (c) 0.79/0.93, and (d) 0.63/1.15. The images were taken at the equatorial plane of 110 μm diameter microgels. Scale bars are 25 μm...... 56

Figure 3.9 Optical fluorescence microscopy images of the composite droplets containing agarose-FITC, gelatin-Ph, and crosslinker, taken immediately after the orifice (a) and 15.5 mm

xviii

downstream of the orifice (b). Qag,= Qgel = Qcross = 0.1 mL/hr. Qcont = 2 mL/hr. The scale bars are 500 μm...... 57

Figure 3.10 Optical image of a) droplet formation and b) 15.5 mm downstream on chip. Basic blue is added to the agarose component for visualization. Scale bars are 500 μm...... 58

Figure 3.11 Optical image of a) droplet formation and b) 15.5mm downstream on chip. Basic blue is added to the gelatin component for visualization. Scale bars are 500 μm...... 58

Figure 3.12 SEM images of agarose−gelatin−Ph gels. The Cag/Cgel‐Ph ratios in the gels were (a) 2/0, (b) 1.5/0.5, (c) 1/1, (d) 0.5/1.5, and (e) 0/2. The scale bar is 250 nm...... 59

Figure 3.13 Aspiration of 150 μm-diameter agarose microgel into a glass micropipette with the inner diameter of 53 μm. Scale bar is 50 μm. Total concentration of polymer was 4 % (w/w) while the weight ratio of agarose:gelatin was 1:1...... 61

Figure 3.14 Variation of the dewatering propensity, D, as a function of the fraction of the agarose concentration, Cag/ (Cag +Cgel)...... 61

Figure 3.15 (a) Stress−strain curve for microgels with Cag/Cgel‐Ph ratios of 2.25 (■), 1.34 (●), 0.85 (▲), 0.55 (□), and 0.35 (○). Cag + Cgel-Ph was maintained constant at 4% (w/w). (b) Dependence of the stiffness of the composite microgels on their composition, measured at room temperature...... 62

Figure 4.1 Transmission electron microscopy image of CNCs after dialysis against water. The scale bar is 500 nm...... 67

Figure 4.2 Effect of CNC and salt concentrations on gelation of CNC suspension at 25 °C. State diagrams of CNC suspensions of various concentrations with the addition of various concentrations (a) NaCl, (b) MgCl2, and (c) AlCl3 solutions of various concentrations. (a′−c′) State diagrams as in (a−c), respectively, plotted for the corresponding Debye lengths of CNCs. The sol and gel states are indicated as triangles and squares, respectively. The solid lines represent the boundaries between the sol and gel states...... 69

Figure 4.3 State diagrams of aqueous CNC suspensions in the prescence of cations. The sol () and gel () states are observed following the addition of metal salts solutions of a) NaCl, b) xix

MgCl2, c) AlCl3, d) CaCl2, and e) SrCl2 at 25 °C (a-e) at 25 °C; and f) NaCl, g) MgCl2, h)

AlCl3, i) CaCl2, and j) SrCl2 at 37 °C...... 71

Figure 4.4 Strain amplitude sweep for the Ca50-4 gel at 25 oC...... 73

Figure 4.5 Dynamic frequency sweeps for (a) Na50−4 (circles), Mg50−4 (squares), and Al50−4 (triangles), and (b) Mg50−4 (circles), Ca50−4 (squares), and Sr50−4 (triangles). The variations in the storage moduli, G′, and loss moduli, G′′, are shown with closed and open symbols, respectively. The dynamic frequency sweeps were performed at 0.5% strain...... 73

Figure 4.6 Dynamic frequency sweeps for Ca50−2 (circles), Ca5−4 (squares), and Ca50−4 (triangles). The storage moduli, G′, and loss moduli, G′′, are shown in closed and open symbols, respectively. The experiments were performed at 0.5% strain...... 76

Figure 4.7 Hysteresis in shear properties of ionically crosslinked CNC gels. The blue and red symbols correspond to the first and the second consecutive dynamic frequency sweeps, respectively. The solid and open symbols correspond to G′ and G′′, respectively...... 77

Figure 4.8 Hysteresis in shear properties of the Ca5-4 and Ca50-2 gels. The red and blue curves correspond to the first and the second consecutive dynamic frequency sweeps, respectively. The solid and open symbols correspond to G′ and G′′, respectively...... 77

Figure 4.9 Scanning electron microscopy images of CNC hydrogels: (a) Na50−4, (b) Mg50−4, (c) Al50- 4, (d) Ca50−4, and (e) Sr50−4. The scale bars are 1 μm...... 78

Figure 4.10 1H NMR spectra of free dextran molecules at varying gradient strengths...... 79

Figure 4.11 1H NMR spectra for dextran molecules embedded in the Ca50-4 gel at varying gradient strengths...... 80

Figure 4.12 Experimental and fitted values of the normalized echo attenuation for 150 kDa dextran in the solution D2O and in a Ca50-4 gel, plotted against Z...... 80

Figure 4.13 Polarization optical microscopy images of (a) the CNC suspension (sample CNC0−4) and CNC gels of (b) Na50−4, (c) Mg50−4, (d) Al50−4, (e) Ca50−4, and (f) Sr50−4. The scale bars are 100 μm...... 83

Figure 4.14 Polarization optical microscopy images of (a) Ca5−4 gel, (b) CNC0−2 suspension, and (c) Ca50−2 gel. The scale bars are 100 μm...... 83

Figure 4.15 SAXS intensity profiles for the CNC0−4 suspension and CNC gels. The fitting for CNC0−4 is shown by the solid line. The SAXS intensity profiles were arbitrarily shifted for easier visualization...... 85

Figure 4.16 Variations in (a) the complex modulus, |G*|, and (b) the mesh size, both plotted as a function of Debye length for various gel samples containing 50 mm metal chloride and 4% w/w CNCs. The values of |G*| are obtained from dynamic frequency sweeps at an oscillatory frequency of 1 rad/s with 0.5% strain...... 86

Figure 5.1 Atomic force microscopy height image of aldehyde-functionalized CNCs...... 90

Figure 5.2(a) SEM images of aerogels cross-section (the XY-plane perpendicular to the ice- growth direction) with morphologies ranging from fibrillar (F) to columnar (C) to lamellar (L) and their combinations, dependent on A-CNC:H-POEGMA weight ratio and CA-CNC+H-POEGMA. Scale bars are 20 μm. (b) Photographs of aerogels cast in cylindrical molds. Scale bars are 0.5 cm...... 93

Figure 5.3 SEM images of aerogel 1:5-4 directionally freeze-cast from A-CNC + H-POEGMA dispersions at various temperatures. Top and bottom rows of images show the structure of the freeze-fractured planes that are perpendicular (cross-section) and parallel (side view), respectively, to the ice-growth direction, as shown in the corresponding cartoons. Scale bars are 50 μm...... 94

Figure 5.4 Surface area of (a) aerogels at a constant CA-CNC+H-POEGMA and varying A-CNC:H-

POEGMA ratio and (b) aerogels at constant A-CNC-to-H-POEMGA ratio and varying CA-

CNC+H-POEGMA...... 95

Figure 5.5 (a, right) Experimental 2D SAXS pattern from irradiating the 1:5-4 aerogel in the Z- direction. (a, left) Simulated 2D scattering pattern for an isotropic distribution of discs. (b, right) Experimental 2D SAXS pattern from irradiating the 1:5-4 aerogel in the XY-direction. (b, left) Simulated 2D scattering pattern for discs preferentially aligned in the Z-direction. (c) 1D radial- averaged SAXS plots of the 1:5-4 aerogel irradiated in the Z- (blue) and XY- (red) directions. A

xxi line with q-4 scaling is also shown as a visual aid. (d) Theoretical 1D radial-averaged SAXS plots for small cylinders (blue squares) with a radius of 3 nm (± 10 %) and a length of 150 nm; large cylinders (red triangles) with a radius of 65 nm (± 10 %) and a length of 50 μm; and discs (black dots) with a radius of 5 μm and a diameter of 260 nm (± 10 %). The dotted line marks the lower limit of the experimentally reachable q range...... 96

Figure 5.6 SEM images and corresponding 2D SAXS patterns of fibrillar (F), columnar (C), and lamellae (L) aerogels. The aerogels were irradiated in the Z- or XY-direction in SAXS experiments...... 97

Figure 5.7 2D SAXS patterns of a 1:5-4 aerogel irradiated in the (a) Z- and (b) XY-directions. 99

Figure 5.8 2D SAXS patterns of a 1:1-4 aerogel irradiated in the (a) Z- and (b) XY-directions. (c) 2D SAXS pattern for the same aerogel as in (b) irradiated in the same direction, except the sample was rotate by 90° about the axis parallel to the direction of irradiation...... 100

Figure 5.9 Swelling kinetics for A-CNC-H-POEGMA samples prepared at (a) CA-CNC+H-POEGMA = 4 wt% and varying A-CNC:H-POEGMA ratios and (b) at the weight ratio A-CNC:H-

POEGMA of 1:5 ratio and varying CA-CNC+H-POEGMA...... 101

Figure 5.10 (a) Coordinate system used for swelling and compression tests. (b, c) Degree of swelling in the XY (QXY,eq) and Z (QZ,eq) directions of aerogels cast from suspensions at varying

CNC:POEGMA weight ratio and CA-CNC+H-POEGMA=4 wt% (b) and varying CA-CNC+H-POEGMA and CNC:POEGMA weight ratio of 1:5 (c). (d, e) Young’s moduli of hydrogels prepared at varying

CNC:POEGMA weight ratio and CA-CNC+H-POEGMA=4 wt% (d) and varying CA-CNC+H-POEGMA and CNC:POEGMA weight ratio of 1:5 (e). *p >0.05, ** p <0.05, *** p <0.01, **** p <0.001, Student’s t-test. The error bars represent one standard deviation. Blue and red colored bars correspond to the XY-plane and Z-directions, respectively ...... 102

Figure 5.11 Stress strain curves for A-CNC-H-POEGMA hydrogels compressed parallel (Z) and perpendicular (XY) to the direction of ice growth (50 compression cycles)...... 104

Figure 5.12 Mechanical properties of anisotropic hydrogels. (a) Stress-strain curves for a representative hydrogel sample (1:5-4) over 50 compression cycles. (b) The first compression cycle for hydrogels cast from suspensions prepared at varying A-CNC:H-POEGMA ratios and

xxii

CA-CNC+H-POEGMA = 4 wt%. (c) The first compression cycle for hydrogels cast from suspensions prepared at varying CA-CNC+H-POEGMA and A-CNC:H-POEGMA ratio of 1:5. (d) The first compression cycle for 1:5-4 hydrogels freeze-cast at -196, -80 and -20 °C. All samples were subjected to a pre-compression and then strained for 50 compression cycles. Pre-compressions were performed at 10% and 50% strains for Z- and XY-directions, respectively...... 105

Figure 6.1 IR spectra of the PCL-b-PTHF-b-PCL oligomer and the PU polymer...... 110

Figure 6.2 GPC trace for the PU polymer in NMP eluent...... 110

Figure 6.3 a) Cyro-TEM image of PU particles. The scale bar is 100 nm. b) TEM image of CNCs. The scale bar is 250 nm. c) HRSEM image of CB particles. The scale bar is 250 nm. d)

TEM image of CNFs. Scale bar is 1 m. e—h) Photographs of PUpure, PUCNC, PUCNC-CB, and

PUCNC-CNF, respectively. The scale bars are 2.5 mm. (i—p) The corresponding SEM images show the cross-section of foams in the planes normal to the (i—l) parallel and (m—p) perpendicular direction of ice-growth. The scale bars are 50 m. q) Microtomography image of a PUCNC foam. The length of the cube is 930 µm. The lamellae are oriented along the ice-growth direction, without preferential orientation within the perpendicular planes...... 112

Figure 6.4 a) X-ray photoemission and Auger spectra for CB and CNFs. b) Derivatives of the C KLL spectra for CB and CNFs. The dotted lines are visual guides to the maxima and minima of the derivative spectra. c—f ) C1s and O1s spectra for CB and CNFs...... 114

Figure 6.6 a—d) Photographs and e—h) TEM images of CB and CNF with and without CNCs. The final concentration of CB or CNF was 5 wt%. The final concentration of CNCs, if present, was 2.5 wt%. Scale bars for the TEM images are 500nm...... 117

Figure 6.7 Low-magnification SEM images of various PU foams cut in the plane normal to the (a—d) parallel and (e—h) perpendicular direction of ice-growth. Scale bars are 400 μm. The red lines emphasize regular spacing in the projections of PUpure. The red arrows mark the strut-like bridging features in PUCNC-CB and PUCNC-CNF...... 120

Figure 6.8 SEM images of the same PUCNC-CNF foam show various structural features including a) CNFs embedded within the lamellae, inter-lamellar CNFs bridges, strut-like PU bridges, b)

xxiii blocked pores, c) pores bridged by a high concentration of CNFs, and d) isolated bundle of CNFs...... 121

-1 Figure 6.9 DSC curves measured at a heating rate of 5 K min for a) PUpure, b) PUCNC, c)

PUCNC-CB, and d) PUCNC-CNF before (red curves) and after (blue cruves) annealing at 90 °C for 8 h...... 123

Figure 6.10 First (red curve) and second (blue curve) heating cycle for PUpure at a heating rate of 5 K min-1...... 124

Figure 6.11 X-ray diffractograms for PUpure, PUCNC, PUCNC-CB, PUCNC-CNF, CNC, and CNF. .. 125

Figure 6.12 Representative stress-strain curves for samples a) PUpure, b) PUCNC, c) PUCNC-CB, and d) PUCNC-CNF compressed up to 50 % strain in the parallel (blue) and perpendicular (red) directions, after pre-compression to 20 % strain. e) The value of E of each sample was calculated using the linear elastic region of each compression curve. f) The recovery after 50 % strain for each sample compressed in the parallel and perpendicular directions...... 126

Figure 6.13 Stress-strain curves from the compression-decompression cycles on PUpure, PUCNC,

PUCNC-CB, and PUCNC-CNF. The scales of the compressive stress for the perpendicular cases are significantly lower than those in the parallel plots...... 128

Figure 6.14 (a—c) Thermal conductivity of PU foams in the parallel (blue) and perpendicular (red) directions measured in a) vacuum (pressure < 1 mbar), b) air (ambient pressure, 990 mbar), and c) helium (pressure = 1000 mbar). (d, f) Thermal conductivity of PUpure in the parallel (blue) and perpendicular (red) directions upon d) cycling between vacuum (pressure < 1 mbar) and helium (pressure = 1000 mbar) and f) cycling between air and helium at a constant pressure of

980 mbar. (e, g) parallel (blue) and perpendicular (red) κ of PUpure at e) various pressures of helium, and g) various volume fractions of helium in an helium/air mixture, while the total pressure was kept at 980 mbar. h) Infrared thermograms of PUCNC-CNF irradiated in the parallel and perpendicular directions. i) Ashby plot of κ versus E for the anisotropic foams reported herein and various polymeric bulk (green) and cellular materials (brown). The Ashby plot was generated from the Granta CES Selector 2015 software.275 ...... 131

xxiv

Figure 6.14 Thermal switching for various PU foams between helium (pressure = 1000 mbar) and vacuum (pressure < 1 mbar) atmospheres...... 133

xxv

List of Abbreviations

2D two-dimensional

3D three-dimensional

 thermal diffusivity

AA acrylic acid

A-CNC aldehdye-functionalized CNC

ADH adipic acid dihydrazide

AFM atomic force microscopy

AIBMe 2,2-azobisisobutyric acid dimethyl ester

ATR attenuated total reflectance

BE binding energy

BET Brunauer-Emmet-Teller

CA-CNC+H-POEGMA total concentration of A-CNCs and H-POEGMA in the mixture used for freeze-casting

Cag concentration of agarose

Cag,est estimated concentration of agarose

CB carbon black

CCNC concentration of cellulose nanocrystals

Cgel concentration of gelatin

xxvi

CHES N-cyclohexyl-2-aminoethanesulfonic acid

CNC cellulose nanocrystal

CNF carbon nanofiber

cp specific heat capacity cryo-TEM cryogentic TEM

D diffusion coefficient of dextran molecules in the gel

D proportionality constant related to the volume change of a microgel during pipette aspiration

D' fitting parameter for diffusion

D0 diffusion coefficient of dextran molecules in solution

DIW deionized water

DLS dynamic light scattering

DMPA dimethylolpropionic acid

DMSO dimethyl sulfoxide

DSC differential scanning calorimetry

E Young's modulus

ECM extracellular matrix

EDC 1-ethyl-3-[3-(dimethylamino)propyl]carbodiimide hydrochloride

Epara Young's modulus in the parallel direction

Eperp Young's modulus in the perpendicular direction

xxvii

Exy Young’s modulus in the X- or Y-direction

Ez Young’s modulus in the Z-direction

FITC fluorescein isothiocyanate isomer I

FPL Forest Products Laboratory

FTIR Fourier transform infrared g gradient strength

G' storage modulus

G" loss modulus

G* complex shear modulus

Gelatin-Ph phenol-functionalized gelatin

GPC gel permeation chromatography h(t) height of hydrogel at time, t

h0 height of a cuboidal aerogel

HBSS Hank’s Balanced Salt Solution

H-POEGMA hydradize functionalized POEGMA

HRP horseradish peroxidase

HRSEM high-resolution SEM

HSAB hard-soft acid-base

I intensity of dextran signal at different gradient amplitudes

ICP-AES inductively-coupled plasma atomic emission spectroscopy

xxviii

IDPI isophoronediisocyante

Io intensity of dextran signal at the lowest gradient

IR infrared

 thermal conductivity

 Debye length

kB Boltzmann constant l(t) length of hydrogel at time, t

l0 length of a cuboidal aerogel

LCST lower-critical solution temperature

M(EO)2MA di(ethylene glycol) methyl ether methacrylate

MF microfluidics

Mi molar concentration of species, i microCT a micro-computed tomography

Mn Number average molecular weight

Mw Weight average molecular weight

NA Avagadro's number

NHS N-hydroxysulfosuccinimide

NMP N-methylpyrrolidone

NMR nuclear magnetic resonance

OEGMA oligo(ethylene glycol) methyl ether methacrylate

xxix

P(x) fraction of microgels expected to contain x cells and λ is the average number of cells per droplet

PCL polycaprolactone

PDMS poly(dimethylsiloxane)

PE pass energy

PEO poly(ethylene oxide)

PFPE perfluoropolyether

PMMA poly(methyl methacrylate)

PNIPAM cast poly(N-isopropylacrylamide)

POEGMA poly(oligoethylene glycol methacrylate)

POM polarized optical microscopy

PPO polypropylene oxide

PTHF polytetrahydrofuran

PU polyurethane

Qag volume flow rate of agarose solution

Qcont volume flow rate of continous phase

Qcross volume flow rate of crosslinker solution

Qgel volume flow rate of gelatin solution

QXY degree of swelling in the X- or Y-direction

QXY,eq degree of swelling in the X- or Y-direction at equilibrium

xxx

QZ degree of swelling in the Z-direction

QZ,eq degree of swelling in the Z-direction at equilibrium

 density

Rf radius of CNC fibrils forming the mesh

RGD arginylglycylaspartic acid

Rh hydrodynamic radius of the probe

Rf radius of the opening between the fibrils (half the mesh size)

Rp radius of micropipette

S slope of the plot ΔP versus (x − xo )/Rp

SAXS small-angle X-ray scattering

SEM scanning electron microscopy t time

T temperature

TEM transmission electron microscopy

TGA thioglycolic acid

THF tetrahydrofuran

V volume of stressed microgel

V0 volume of unstressed microgel x number of cells x length of the intrusion of a microgel into the mircopipette

xxxi

X-direction a direction perpendicular to the direction of ice growth

XFA xenon flash analysis

xo original length of the intrusion of a microgel into the mircopipette

XRD X-ray diffraction

XY-plane the plane orthogonal to the direction of ice growth

Y-direction a direction perpendicular to the direction of ice growth

Z-direction a direction parallel to the direction of ice growth

zi charge number of species, i

γ gyromagnetic ratio

δ gradient pulse length

Δ diffusion delay

δ phase angle

ΔP negative pressure differential applied to the microgel

ε dielectric constant of the electrolyte solution

ε0 permittivity of free space

κpara thermal conductivity in the parallel direction

κperp thermal conductivity in the perpendicular direction

λ average number of cells per droplet

xxxii

Chapter 1 Polymer Scaffolds as Artificial Extracellular Matrices

Introduction

The work presented in this thesis explores the fabrication and characterization of scaffolds that mimic the extracellular matrix (ECM). In this chapter, we discuss the synthesis and fabrication of microgels, nanofibrillar hydrogels, as well as anisotropic aerogels, hydrogels, and foams for development of potential artificial ECMs. This section is partially adapted with permission from “Supramolecular nanofibrillar polymer hydrogels,” a chapter in Supramolecular Polymer Networks and Gels. S. Seiffert, Ed., Adv. Polym. Sci., Springer International Publishing, Switzerland, 2015, 167-208.

Hierarchically assembled complex systems created by nature have inspired material chemists and engineers to create many new materials such as Velcro, adhesives, and photonic crystals from burr hooks, gecko feet, and butterfly wings, respectively. The mammalian extracellular matrix (ECM), which supports cells and imparts tissue function, are also an interesting material to mimic because of its diversity and ability to derive different properties and functions from structure. This thesis describes our efforts into creating materials that imitate the natural ECMs.

1.1 The natural ECMs

Before we begin to reproduce these natural ECMs, we must first understand their structure and composition. The ECM is the material that fills the extracellular spaces between the cells in tissue.1 Major components of the ECM are fibrous proteins and proteoglycans (Figure 1.1). Fibrous proteins consist of collagen, elastin, and laminin, which provide structural support and mechanical resilience to imposed stresses and strains. Proteoglycans are proteins that are covalently attached to long, linear polysaccharides called glycosaminoglycans.2 The glycosaminoglycans, negatively charged at physiological pH 7.4, are highly hydrophilic and allow proteoglycans to form a gel, which surrounds the fibrous protein components in the ECM. The gel-like property of proteoglycans allow ECMs to hydrate and withstand compressive

1 2 forces.3 Depending on the type of tissue, the ECMs are composed of different types and concentrations of these components.

The structure and organization of the ECM components greatly impact tissue function. Most tissues in vivo are anisotropic.5 Cardiac tissues consists of discrete, aligned muscle layers.6 Figure 1.2a show the structure of the ventricular myocardium containing lamellae of organized myocytes in a collagen matrix. This lamellar architecture in the myocardium (muscular wall of the heart) allow muscle fiber extension and wall stress to be distributed through the ventrical wall during systole and diastole. Furthermore, as a result of the structural anisotropy, the electrical conduction velocity was observed to be 2 to 3 times greater in plane than transverse to it in ventricular muscle and ~10 time greater in the crista terminalis.7 The cornea stroma is composed of 200 lamellae each containing thin nanofibrils of collagen (~36 nm) aligned parallel

3 to each other.8 The destructive interference of the scattered light by the evenly spaced nanofibrils results in the cornea transparency.9 In tendon, collagen fibers are arranged in rope-like bundles providing very high tensile strength.10

In addition to the lamellar structure mentioned above, the natural ECMs adopt other structures including columnar, and fibrillar morphologies. Figure 1.2b shows the SEM image of the columnar structure of decellularized endomysium in the intramuscular connective tissue of semitendinosus muscles.11 In vivo, the endomysium ensheaths individual skeletal muscle fiber, which are aligned to the long axis of the sheaths. This columnar arrangement guides and supports the myocytes within. Also, decellularized adipose tissue, shown in Figure 1.2c, has a network architecture composed of fibrils of collagen.12 The fibrillar network acts as an exoskeleton, protecting the adipocyte within from mechanical disruption.

In addition to imparting tissue-specific mechanical properties, ECM components contain biochemical cues. In particular, arginylglycylaspartic acid (RGD) peptide sequences, found on both collagen and fibronectin, are bound by the transmembrane integrin receptors on a cell’s surface.13 Cell anchorage induces various signal transduction cascades and rearrangement of cell shape, both of which are important for many cellular processes including migration,

4 proliferation, and apoptosis.13 For anchorage-dependent cells this binding is required for the survival.

1.2 Mimicking natural ECMs

Scaffolds that recapitulate the natural ECM have received much attention for the following purposes:

1. To create scaffolds in which cells grow for tissue engineering applications;

2. To create in vitro cell and tissue culture models; and

3. To create biomimetics which have a structures or functions analogous to the native ECM.

Note that the first two purposes listed above require a more rigorous imitation of the natural ECM, than the third purpose, since tissue engineering scaffolds and cell cultures both have cell- related biological applications. The third listed purpose, to mimic the ECM to derive materials with similar structure or chemistry, may or may not have an intended biological application.

The aim of tissue engineering is to replace, repair, or regenerate organ or tissue functions by creating artificial tissues and organs for transplantation.14 Artificial ECMs can also be used to create controlled, reproducible in vitro cell culture models for the biological studies including the studies of biochemical mechanisms, cell-ECM interactions, effects of disease and drugs on cells. Both tissue engineering and cell culture development face similar challenges, namely, that of mimicking the dimensionality, structure, and mechanical property of the natural ECM.

In vivo, cells are surrounded by a three dimensional (3D) ECM, and thus receiving biophysical and biochemical cues from all directions. On a two-dimensional (2D) substrate, however, cells are exposed to only a fraction of these cues and encounter little resistance in terms of migration. This difference is striking for 2D versus 3D in vitro cultures. For example, in studies of the response of tumors to anti-cancer drugs, cancer cells cultured under 3D biomimetic conditions were more resistant to anti-cancer drugs, and expressed a higher level of multidrug resistance-associated protein than cells cultured on 2D substrates.15,16 For cell culture development, 3D models bridge the gap between the less rigorous 2D cell culture and animal models. Results from 3D models would be a more accurate prediction of in vivo cellular

5 responses, saving time and perhaps even animal subjects that would have been sacrificed for in vivo testing. When designing an artificial ECM, one should keep in mind the surrounding complete surrounding of the cell is accounted for.

In addition to dimensionality, the structural and mechanical properties of the scaffold should match that of the native ECM. The native ECM can assume various structure depending on the type of tissue. It is important to mimic the structure to impart the desired function to the specific tissue. The function of the scaffold also depends on its mechanical properties such as stiffness, elasticity, and strength of the ECM itself. For examples, in bone ECM, the mineral deposits provides compressive strength, while the collagen fibrils give flexibility.17 From the perspective of the resident cells, the rigidity of the ECM to which it binds affects cell motility, proliferation, spreading, and differentiation.18 For example, mesenchymal stem cells that grow on soft substrates with a similar stiffness to brain, commit to a neurogenic fate, while the same cells grown on hard substrates, similar in stiffness to bone, express osteogenic phenotypes.19 As a design principle for artificial ECMs, the structure mechanical properties of the artificial ECM should match as closely as possible to those of the native ECM.

1.3 Nanofibrils as building blocks of the ECM

To mimic the dimensionality, structure, and mechanical properties of native ECM, it is crucial to also imitate the properties of the building blocks that make up this matrix in vivo. The proteinaceous component of the natural ECM is nanofibrillar. Generally, the diameters of the fibrils are on the order of nanometers, while their lengths can vary from hundreds of nanometers to micrometers. Collagen, the most abundant protein in the mammalian ECM, exists primarily Type 1 collagen, which has a hierarchically assembled fibrillar structure. This hierarchical assembly is shown in Figure 1.3. The primary building blocks are collagen molecules containing repeating units of XaaYaaGly, where Xaa and Yaa are commonly proline and hydroxyproline.20 Under appropriate conditions, collagen molecules organize into a right-handed triple-helix (tropocollagens) that is composed of three left-handed polyproline II helices. Tropocollagens are stabilized by inter-strand hydrogen bonding between the amides along the backbone of the three protein strands, where N–H(Gly). . .O=C(Xaa). Tropocollagens in type 1 collagen are <2 nm in diameter and ~300 nm in length. They further assemble laterally in a staggered fashion into nanofibrils of up to 500 nm in diameter and up to 1 cm in length, with a 64 nm D-periodicity

(Figure 1.3).21 Tropocollagens are held to each other by inter-helical water bridges with no direct contact between the triple helices.22

Figure 1.3 The biosynthetic route to collagen fibers. Reproduced with permission from Annual Reviews.20

In vitro, nanofibrillar collagen gels can be formed by adjusting the pH, temperature, and ionic strength of the solution.23 Generally, an acid-soluble collagen solution is neutralized and warmed to 30–34 ºC to induce fibrillogenesis and gelation.23,24 The gel network is formed by entanglement of nanofibrils, as well as by hydrophobic and electrostatic attraction forces.25 Nanofibrillar collagen gels show a complex response to deformation. Figure 1.4 shows a representative stress–strain curve for Type 1 collagen hydrogel matrices under tensile

7 deformation in physiological conditions.26 The toe region (defined as the region between zero strain and the intersection of linear fit and strain axis) corresponds to the straightening of crimps in the fibrils due to the flexibility of the fibers and the presence of non-helical telomeric regions on the molecular level.27 In the linear region, an increase in the Young’s modulus of the gel is associated with stretching the collagen triple helices and with sliding of the collagen molecules past each other.28,29 At even higher strains, disruption of the fibrillar structure results in failure of the gel. Therefore, the collagen gels owe their non-linear mechanical response to deformation to their hierarchical structure.

1.4 Artificial nanofibrils as building blocks of artificial ECMs

Artificial nanofibrils can be designed emulating the structure, composition, and properties of the natural collagenous fibers. For cell and tissue culture applications, these artificial

8 nanofibrils are often assembled into networks that swell in water and have pores sizes on the order of magnitude similar to the fibrillar building blocks themselves. Due to the large building blocks (compared to molecular gels), the mesh size of the nanofibrillar hydrogel can be significantly larger than those of molecular gels, allowing a higher degree of mass and cell transfer. For example, the pores of nanofibrillar agarose hydrogels could be tuned from <100 to ~1,200 nm by decreasing the concentration of agarose from 3 % to 0.5 % (w/w), respectively.30

Generally, nanofibrillar hydrogels form in a hierarchical multistep process (Figure 1.5) that begins with the association of individual molecules into discrete high-aspect-ratio supramolecular structures, associating to form larger nanofibrils, which subsequently develop into a 3D network.31,32 The driving forces behind the assembly of molecule to nanofibrils and, subsequently, to hydrogel networks are discussed in the following section.

1.4.1 Driving forces for formation of nanofibril building blocks from polymer molecules

It requires some molecular engineering (by nature or artificially), to assemble molecules into shape-anisotropic, well-defined fibrils (Figure 1.5). For nanofibrils to form from polymers, a

9 well-defined way of assembling is required. Non-covalent interactions, such as H-bonding, electrostatic, and van der Waals forces, are favorable because the weak nature of these bonds allows molecules to rearrange and assemble into an optimum configuration.

The driving forces for the assembly of polymeric molecules into one-dimensional (1D) fibers arise when the enthalpic gain from intermolecular interactions outweighs the entropic loss from reduced polymer flexibility.34 Inter- and intra-chain hydrogen bonding is by far the most important driving force for the association of biopolymer chains and the stabilization of fibrils in aqueous media.

1.4.2 Driving forces for the formation of a nanofibrillar network

To achieve gelation instead of uncontrolled aggregation and precipitation, a balance needs to be maintained between attractive forces that lead to network formation and solvophilic forces that keeps parts of the nanofibrils solvated. The coexistence of ‘soluble’ and ‘insoluble’/’constrained’ regions within the nanofibrils makes the nanofibrillar network stable and swollen. To generate a hydrated, non-collapsed nanofibrillar network, the parts of the nanofibrils need to be stabilized by sufficient solvent-fibril interactions and/or electrostatic repulsion. A 3D network can be formed by nanofibril association, branching, or entanglement.

Nanofibrils can associate into a 3D network by covalent or non-covalent bonds between the nanofibrils. These bonds can be triggered by changes such as chemical reactions, temperature, increase in polymer concentration, increase in ionic strength, or by the addition of ions charged oppositely to the nanofibrils. During assembly, the enthalpic gain associated with the formation of a network should exceed the entropic loss due to gelation. Non-covalent bonds responsible for the formation of a network structure are case-specific and can include hydrogen bonding, electrostatic attraction, hydrophobic interactions, or guest–host interactions, depending on the type of functional groups on the surface of individual nanofibrils. For instance, alginate networks can be formed via the interactions between the carboxylic groups present on rod-like alginate complexes and the cations present.35

A network can be formed when the nanofibrils split into branches that subsequently reform nanofibrils with branching counterparts from another nanofibril. For example, irregularities in the agarose molecule structure results in disruptions in the agarose double helix

10 and branching.36 Branching agarose molecules reform double helices intermolecularly with other branching agarose molecules.

If the nanofibrillar building blocks are sufficiently long and their concentration is sufficiently high these nanofibrils can entangle to form a gel. The relaxation time, τ, of gels made by fibrillar objects scales with the length of the fibrils, L, at τ ~ L3.37 With fibrils that are sufficiently long, topographical constrains can be used to create materials with effectively infinitely long τ, such that the properties of the material is practically ‘gel-like.’ An example of such gels is one made from cellulose nanofibirils (CNFs), which have widths of 3-4 nm and lengths up to 1000 nm. Unlike their shorter cellulose nanocrystal (CNC) counterpart, CNFs can gel via entanglement at much lower concentrations.38

1.4.3 Artificial nanofibrillar ECM made from synthetic and natural building blocks

A large number of nanofibrillar gels both from artificially and naturally derived polymers have been used as artificial ECM for cell encapsulation. In the case of synthetic hydrogels formed by long wormlike micelles of block copolymers39 or fiber-like structures of short amphiphilic peptides,40 hydrophobic forces govern the association of hydrophobic blocks in an aqueous environment, in order to minimize the surface energy of the system. The segregated hydrophobic blocks form the core of the wormlike micelle, resulting in 1D supramolecular aggregates that have enhanced stability in water in comparison with the original molecules.41 The self-assembly of fiber-like structures can be fine-tuned by varying, for example, the copolymer composition and concentration.41 During the assembly of peptide amphiphiles, the formation of β-sheets in the peptide region of the molecule is crucial to the creation of fiber-like supramolecular structures.42

In natural systems, biopolymers are the building blocks of nanofibrillar materials. Advantage of using natural polymers as building blocks is their inherent biocompatibility and bioavailability. Biological fiber-forming polymers such as collagen, fibrin, agarose, alginate, and chitosan have also been used to regenerate or repair damaged tissues and organs.43 The unique properties and broad range of applications of naturally derived supramolecular nanofibrillar hydrogels have motivated the design of their synthetic analogues, such as hydrogels formed by wormlike block copolymer micelles44 and amphiphilic peptides.45 Proteins and polysaccharides

11 often form double and triple helices that are the origin of the fibrillar structure. The inclusion of rigid repeating units in the polymer backbone dictates the helical structure, because these rigid units control the torsion angles within the helical structures. For instance, in collagen, these repeat units are proline and hydroxyproline residues,20 whereas anhydrogalactose cages are the monomeric units in agarose and κ- and ι-carrageenans.36 In the following sections, we will focus on two example: agarose and cellulose nanocrystals.

1.4.4 Nanofibrillar hydrogels of agarose

Agarose is a linear polysaccharide extracted from red algae composed of alternating 1,3- linked β D-galactopyranosyl and 1,4-linked 3,6-anhydro-α-L-galactopyranosyl residues; a fraction of these residues (~2 %) contains sulfate groups.46,47 Agarose solutions can form thermoresponsive gels. Temperatures as high as 95 °C are required for agarose molecules in solution to completely adopt a random coil conformation.48,49 Upon cooling the solution below the temperature of complete melting (95 ºC), typically to 10–40 °C, agarose coils assemble into left-handed double helices. The presence of a 3,6-anhydro bridge in the covalent structure and numerous hydrogen bonds contribute to the formation of left-handed helices.48 The formation of agarose networks occurs on both the supramolecular and molecular levels. On the supramolecular level, agarose double helices aggregate to form bundles.50 At temperatures of ~20–40 ºC, depending on the agarose concentration in solution, 7–11 double helices assemble into these bundles, or nanofibrils (Figure 1.6a).48,51 On the molecular level, some of the agarose repeat units lack the regular 3,6-anhydro residues, which leads to disruption of the double helix and formation of “soluble kinks.”36 The kinks result in branching, which contributes to formation of the 3D gel network. Agarose sol–gel transitions exhibit strong hysteresis in response to changes in temperature. Depending on the agarose concentration and type, the gelling temperature of agarose solutions is between 10 and 40 "C, which is significantly lower than the melting temperature of the gel (approximately 90 ºC).49 The hysteresis and thermal stability of agarose gels originate from cooperative hydrogen bonding in both the double helices and their bundles.51 The “memory” of intermolecular associations can be erased upon heating agarose solutions to ~90 ºC, at which agarose molecules adopt a random coil conformation.

The mechanical and structural properties of agarose gels are sensitive to their thermal history.49 For example, when the cooling temperature was maintained below 35 ºC, fast agarose gelation resulted in strong, homogenous and elastic gels with a pore size of ~100 nm. When the gelation temperature was held above 35 ºC, phase separation competed with the gelation process to form heterogeneous, turbid gels with poor mechanical properties. The elastic modulus of 2 % (w/w) agarose gels increased from 12 to 78 kPa as the curing temperature ranged from 43 to 5 ºC, respectively (Figure 1.6b).

High melting temperature and resistance to degradation make agarose gels suitable for autoclaving. In addition, agarose gels are being explored as artificial 3D extracellular matrices because of their biocompatibility, lack of cell-adhesion, and tunable mechanical properties (Young’s and shear moduli), which are achieved by varying the concentration of agarose.52–54

1.4.5 Cellulose nanocrystals as a building blocks for nanofibrillar hydrogels

Cellulose, the most abundant renewable organic material produced in the biosphere, is extracted from plants, bacteria, algae, fungi, and tunicates.55 The structure of nature-derived cellulosic materials is hierarchical. The primary building unit of cellulose is β-1,4,-linked- anhydro-D-glucose. A discrete number of cellulose molecules pack in a parallel fashion to form

13 elementary fibrils, also known as microfibrils.55 Depending on the source, elementary fibrils have widths of 3–50 nm55 and lengths that can exceed 10 μm.56 Elementary fibrils are composed of alternating regions of crystalline and amorphous cellulose. The crystalline regions of elementary fibrils can be isolated by selectively degrading the amorphous regions by acid hydrolysis. These isolated highly crystalline regions are called cellulose nanocrystals, also known as nanocrystalline cellulose or cellulose nanowhiskers. Wood fibers and other plant fibers can be used to produce cellulose nanofibrils, which have dimensions similar to that of elementary fibrils.55

Acid hydrolysis of cellulose fibers degrades the amorphous regions of the microfibers and yields cellulose nanocrystals (CNCs) with an average diameter of 5–70 nm and length of 100–250 nm (Figure 1.7a).55,58 Different cellulose sources such as wood or algae yield CNCs with different dimensions, even under similar preparation conditions.59 For example, cotton and wood yield highly crystalline CNCs with narrow size distribution, whereas tunicin and algae generate CNCs with larger dispersities and lengths that range from 100 nm to several micrometers.60 Various TEM images of CNC from different sources are shown in Figure 1.7b. During hydrolysis, sulfuric acid reacts with surface hydroxyl groups on CNCs leading to the functionalization of the surface of the CNC with sulfate ester groups. Stability of aqueous CNC suspensions results from an electrostatic repulsion between individual CNCs that counteracts their attraction due to van der Waals forces and hydrogen bonding.61,62

Low-concentration CNC suspensions are clear isotropic fluids, whereas beyond a critical concentration the solution phase separates into a birefringent chiral nematic liquid crystalline phase and an isotropic phase.63 As the CNC content is further increased, a critical concentration is reached where the entire suspension forms a chiral nematic liquid crystalline phase with a characteristic fingerprint pattern.64 The origin of this chirality is not entirely clear and was proposed to be a result of the helicoidal structure of CNCs.55 At a higher CNC content, a gel is formed. The aspect ratio of the CNCs is a key variable in determining CNC gelation and phase separation during the formation of a liquid crystalline phase.55,65,66 Suspensions of long CNCs tend to gel before attaining the liquid crystalline structure.63 The degree of sulfation of CNCs determines the surface charge density and also significantly affects the critical concentration at which the transition isotropic to liquid crystal to gel takes place. It was shown that at lower degrees of sulfation the electrostatic repulsion between CNCs decreases, which leads to gel formation at lower CNC concentrations than for CNCs with a higher degree of sulfation.67,68 An early work in the field revealed the formation of a birefringent gel, when a suspension of CNCs was heated on a steam bath.59 This gelation was likely due to desulfation of CNC surfaces under heating.

Gelation of CNC suspensions is also induced by suppressing electrostatic repulsion between CNCs, either by decreasing the surface density of charged sulfate ester groups or by increasing the ionic strength of the aqueous medium. For example, shear thinning (thixotropic) CNC hydrogels were obtained by desulfation of CNCs with glycerol.60 Alternatively, the

15 addition of NaCl can be used to control the rheological behavior of CNC suspensions in the isotropic, chiral nematic and gel states over a range of CNC and NaCl concentrations.69 For biphasic samples (above the threshold of isotropic-to-chiral nematic transition), the addition of NaCl up to 5 mM concentration decreased the size of chiral nematic domains and increased the sample viscosity at low shear rates, while for gel samples, the addition of NaCl decreased CNC gel viscosity.69

Surface modification of CNCs results in a decrease in surface negative charge and enables the formation of gels by changes in temperature, pH, or ionic strength.70,71 For example, cationic surface functionalization of CNCs resulted in the formation of thixotropic gels at CNC concentrations of 3.5 % (w/w) or greater.72 In addition, functionalization of the CNC surface with carboxylic acid or amine groups rendered the CNCs responsive to pH.73 Sol–gel transitions occurred at pH values corresponding to neutral or weakly charged CNCs, thereby allowing hydrogen bonding to dominate.

Hydrogels of CNCs have many desirable properties, including low cost, nontoxicity, hydrophilicity, biocompatibility, and biodegradability, all of which contribute to their potential applications in bioengineering and biomedicine. For example, CNC dispersed in a solution of cellulose, sodium hydroxide, and urea formed a gel that steadily released bovine serum albumin into a simulated body fluid.74 Nanocomposites of CNC and poly(vinyl alcohol), a hydrophilic non-cytotoxic polymer, exhibited a broad range of mechanical properties that could be tuned to mimic those of cardiovascular tissues implants.75 In Chapter 4 of this thesis, we will explore how cellulose nanocrystals (CNCs) can be used to make nanofibrillar hydrogels, in which the mechanical properties and pore sizes can be tuned by the addition of salts consisting of various cations.

1.5 Microgels for tissue engineering

Hydrogels in the form of microgels are advantageous for cell encapsulation for several reasons.76 Microgels with dimensions not exceeding ~200 μm allow the diffusion of oxygen, nutrients, and metabolic products to the cells, particularly when the 3D cell culture lacks vasculature. If the microgels are monodispersed, the narrow size distribution enables control over the average number of cells in the microgels, thus the effect of confinement and intercellular distances.

Microgels with well-defined, monodispersed dimensions ca be generated using microfluidics (MFs). In one MF approach, streams of aqueous solutions of gel precursors were introduced into the device and mixed. The mixed stream was broken up into droplets that subsequently gel to form hydrogel particles (microgels). The composition of the microgel can be adjusted by varying the relative flow rates of each stream. In this way, the microgel composition and properties (e.g. stiffness) as well as the number of encapsulated cells can be varied.77,78 An example is the generation of agarose microgels shown in Figure 1.8a. Agarose solutions of two different concentrations were combined, mixed, and the mixed stream was sheared in a T- junction device by an immiscible oil phase at a T-junction.78 The droplets were collect in buffer and cooled to gel the agarose component. This method was used to encapsulate murine embryonic stem cells, as shown in Figure 1.8b. Combinatorial libraries of microgels can be generated to screen the desired hydrogel properties.52 In Chapter 3 of this thesis, we will explore a MF platform for generating biopolymer composites, in which the mechanical properties and structure can be tuned in high-throughput manner.

1.6 Freeze-casting method to make anisotropic monolithic scaffolds

Most ECMs are inherently anisotropic on the microscale to impart geometric asymmetry and directionality to the residing cells and tissues on the macroscale. Muscle fibers, for example,

17 exert forces parallel to the direction of contraction allowing us to move and our hearts to beat. This anisotropy is translated from long, tubular myocytes that are aligned by the collagen network that surround them.8

The topographical anisotropy of the ECM affects cells on a biochemical level.79 When cells bind to an aligned substrate through their integrin receptors, filaments within the cytoskeleton connected to the integrin also align, resulting in cell anisotropy and changes in up- stream signalling. The filaments in the cytoskeleton are mechanically integrated with the cell nucleus, such that reorganization of the cytoskeletal filaments can affect the nucleus shape, signaling transduction, and gene regulation.80–84 As a result, cell alignment, induced by aligned or anisotropic ECMs, can affect cell attachment, proliferation, metabolic activity and differentiation.85 Mimicking the anisotropic structures of the native ECMs, tissue engineering uses a “top-down” approach to generating artificial ECMs predicted to satisfy the desired organization of cells seeded within.

Three-dimensional anisotropic scaffolds for cell culture can be fabricated from polymeric 86 87 materials using techniques such as 3D printing, supercritical CO2 foaming, and freeze- casting88–90. Directional freeze-casting, in particular, is a technique in which dispersions are frozen in a uni-directional temperature gradient. The procedure for freeze-casting ceramic slurries is shown in Figure 1.9. Unidirectional ice crystal growth excludes the materials in the dispersed phase (e.g. particles, polymers, monomers) into the space between the ice. After freeze-casting, the solvent (typically water for tissue engineering scaffolds), could either be removed through sublimation or the sample could be placed directly back into water after crosslinking the components. If the former procedure is the case, the aerogel intermediate could be rehydrated forming a hydrogel, though the resulting hydrogels must be crosslinked either covalently or physically in order for the hydrogel not to disintegrate upon rehydration. Aligned features in the lamellar structure made by freeze-casting have been used as nerve guidance channels across spinal cord injury lesions91 and scaffolds for tendon tissue regentation.85

An advantage of freeze-casting is that various structures of polymers and particles can be realized by varying the concentration of the constituents and the ice front velocity (Figure 1.10).93 At fast cooling rates (region a in Figure 1.10), the particles are entrapped in the ice, resulting in dense structures. When the velocity is decreased below the critical value, the materials in the dispersed phase are expelled from the growing ice to form columnar structures (region b in Figure 1.10) and lamellar structures (region c and d in Figure 1.10). If the structure is lamellae, the distance between the lamellae increases with decreasing freeze-cast velocity.

Figure 1.10 Parameters that affect freeze-cast structures. Reproduced with permission from The Royal Society of Chemsitry.93

The freeze-casting method also provides the interconnectivity of the pores, which is particularly important for tissue engineering scaffolds without vasculature. Hydrogels made by freeze-casting have been used to culture cells using biopolymers such as collagen,89 gelatin,94 chitosan,95 and composites.96 Gelatin/chitosan freeze-cast hydrogels could support and accelerate fibroblast in filtration with good cytocompatibility.90 On open cells of freeze-cast scaffolds were helpful for mass transport of nutrients, oxygen and cells.

Anisotropic hydrogels with fibrillar, columnar, lamellae structures similar to those in vivo (Figure 1.2) have yet to be report. Chapter 5 of this thesis describes the freeze-casting of CNCs and poly(oligoethylene glycol methacrylate) (POEGMA) dispersions into anisotropic composite aerogels and their corresponding hydrogels with fibrillar, columnar, and lamellae structures.

1.7 Polyurethane scaffolds as mimics of the ECM

Polyurethane (PU) have long been used in in vivo biomedical devices such as catheters,97 stents98, cardiac pacing leads,99 neurological leads100, and heart valve implants,101 due to PU biocompatibility, hemocompatibility, abrasion resistance, high tensile strength, and elasticity.102

Because of these reasons, polyurethanes have also demonstrated much potential as building blocks for scaffolds used in tissue engineering for several reasons. The type of polyol, isocyanate, and chain extender, used for PU synthesis, can be changed modularly to yield the desired mechanical property.

The elasticity and high stiffness of PUs have also made them promising materials for culturing cells native to relatively stiff tissues such as cardiovascular and musculoskeletal tissues. For example, PUs made from lysine-derived polyisocyanate were used to fabricate artificial ECMs with Young’s modulus between 1.20 to 1.43 GPa to mimic the in vivo cancellous bone ECM which have a Young’s modulus of 0.1 to 2 GPa.103,104 These scaffolds, supporting the growth and attachment of osteoblasts, can be used for bone grafting. In addition, an advantage of using PUs is the ability to create foams with open pores by solvent casting, salt leaching, thermally induced phase separation, melt moulding, gas foaming, and emulsion freeze-drying.105 The porous structure of PU can improve mass transport and allow cells to propagate through the scaffold. For instance, PU scaffolds used for vascular prosthesis require pores for anchoring neointimal and perigraft cells, and encouraging angiogenesis.106

Directional freeze-casting can be used to create anisotropic scaffolds. Scaffolds with tubular pores have been freeze-casted from dimethyl siloxane solutions of PUs made from polycaprolactone, polycaprolactone-b-polyethylene glycol-b-polycaprolactone, 1,4- diisocyanatobutane and putrescine.107–109 Freeze-casting of PU solutions in dioxane solutions have also been reported.110 However, to our knowledge, anisotropic, elastomeric PU foams made by freeze-casting polyurethane dispersions have yet to be reported in the literature. Chapter 6 describes the freeze-casting of PU dispersions into anisotropic open-cell foams, which have lamellar structures very similar to the ones shown in Figure 1.2. We exploit this anisotropic lamellar structure to guide heat in anisotropic and isotropic modes: a function which may prove useful in thermal management materials.

1.8 Overall goals and summary of projects

The overall goal of this work is to develop artificial ECMs for potential cell encapsulation applications. The objectives of each project in this thesis are as follows:

 Chapter 3: To develop a microfluidic platform for generating biopolymer composite microgels for which the composition, rigidity, and structure could be tuned in high- throughput manner

 Chapter 4: To form nanofibrillar hydrogels by adding inorganic salts to cellulose nanocrystal (CNC) suspensions and to tune the structural and mechanical properties of the resulting hydrogels

 Chapter 5: To fabricate of anisotropic composite aerogels and their resulting hydrogels by freeze-casting aldehyde-functionalized CNCs with hydrazide-functionalized POEGMA

 Chapter 6: To fabricate polyurethane foams with anisotropic structural, mechanical, and thermal properties

Chapter 2 Materials and Methods

Experimental 2.1 Materials

2.1.1 Materials for the microfluidic generation of agarose and gelatin composites microgels

Gelatin (type A from porcine skin, 300 Bloom), morpholinoethanesulfonic acid, tyramine hydrochloride, 1-ethyl-3-[3-(dimethylamino)propyl]carbodiimide hydrochloride (EDC), N-hydroxysulfosuccinimide (NHS), horseradish peroxidase (HRP, 52 units/mg), 1,1,1,3,3,3- hexamethyldisilazane (99.9% pure), Basic Blue 41, fluorescein isothiocyanate isomer I (FITC), 30% (w/w) hydrogen peroxide, and tetrahydrofuran (THF) were purchased from Sigma-Aldrich. NHS-fluorescein was received from Thermo Fischer Scientific. Hank’s Balanced Salt Solution (HBSS) was purchased from Gibco-Invitrogen. Ultralow gelling temperature agarose (SeaPrep) was purchased from Lonza Group Ltd. SU-8 photoresist was purchased from MicroChem. Poly- (dimethylsiloxane) (PDMS) (Sylgard 184) was supplied by Dow Corning. Fluorinated oil HFE- 7500 3M Novec was purchased from 3M. All chemicals were used as received. The triblock copolymer surfactant of perfluoropolyethers and a poly(ethylene oxide)-polypropylene copolymer [PFPE-P(EO-PO)-PFPE] was synthesized as described previously.111 Krytox 157 FSH was purchased from DuPont. HFE 7100 and HFE 7500 was purchased from 3M. Jeffamine ED-900 was generously donated by Huntsman Corporation.

Perfluoroalkoxyalkane tubing was received from Upchurch Scientifics. The dialysis membrane (molecular weight cutoff of 6000 Da) was purchased from Spectra Pro. Dimethyl sulfoxide (DMSO) was supplied by Chimiques ACP Chemicals Inc. Borosilicate glass capillaries (outer and inner diameters of 1.0 and 0.75 mm, respectively) were purchased from Sutter Instruments Co. The ceramic scoring wafer was purchased from Restek. Flow rates of the liquids were controlled using PhD 200 Harvard Apparatus PHD 2000 syringe pumps.

22 23

2.1.2 Materials used for the investigation of the ionic gelation of cellulose nanocrystals

An aqueous CNC suspension of 6.43% w/w was provided by FPInnovations (Quebec,

Canada). NaCl, AlCl3 (anhydrous), and SrCl2 (anhydrous) were purchased from Acros Organics.

MgCl2 (anhydrous) was supplied by Alfa Aesar. CaCl2 (anhydrous) was purchased from Fisher Scientific. Dialysis membrane (molecular weight cut-off = 6000 Da) was purchased from

Spectrum Laboratories, Inc. Dextran analytical standard (Mw = 147, 600 Da, Mw/Mn = 1.47) was purchased from Sigma-Aldrich. Deuterium oxide (D2O) was supplied by Cambridge Isotope Laboratories, Inc. (USA). Deionized water was obtained from a Millipore Milli-Q water purification system.

2.1.3 Materials used for the formation of anisotropic hydrogels

Oligo(ethylene glycol) methyl ether methacrylate (OEGMA500, Sigma Aldrich, 95%, Mn

~ 500) and di(ethylene glycol) methyl ether methacrylate (M(EO)2MA, Sigma Aldrich, 95%) were purified using a column of basic aluminum oxide (Sigma Aldrich, type CG-20). Acrylic acid (AA, Sigma Aldrich, 99%), 2,2-azobisisobutyric acid dimethyl ester (AIBMe, Wako Chemicals, 98.5%), adipic acid dihydrazide (ADH, Alfa Aesar, 98%), N’-ethyl-N-(3- dimethylaminopropyl)-carbodiimide (EDC, Carbosynth, Compton CA, commercial grade), thioglycolic acid (TGA, Sigma Aldrich, 98%), sodium periodate (NaIO4, Sigma Aldrich, >99.8%), sodium hydroxide (EMD Millipore Germany), sodium chloride (Sigma Aldrich, ≥99.5%), hydrochloric acid (LabChem Inc., 1M), silver(I) oxide (Sigma Aldrich, ≥99.99% trace metals basis), dioxane (Caledon Laboratory Chemicals, reagent grade) and sulfuric acid (Sigma Aldrich, 95-98%) were used as received. Whatman cotton ashless filter aid was purchased from GE Healthcare Canada. In all the syntheses and fabrication protocols, Milli-Q grade distilled deionized water (DIW, 18.2 MΩ cm resistivity) was used.

2.1.4 Materials used for the formation of anisotropic polyurethane foams

Isophoronediisocyante (IDPI), dimethylolpropionic acid (DMPA), triblock copolymer of polycaprolactoned and polytetrahydrofuran (PCL-b-PTHF-b-PCL) (Mw = 1850 g/mol) with O-H end-groups, carbon nanofibers (CNFs), and N-(3-Dimethylaminopropyl)-N′-ethylcarbodiimide hydrochloride (EDC) were purchased from Sigma Aldrich, Canada. Acetone was purchased from

Caledon Laboratories Ltd. Deionized water was prepared using a Milli-Q purification system. Cellulose nanocrystals (CNCs) were purchased from Forest Products Laboratory (FPL)- University of Maine. Conductive-grade carbon black (CB), VXC 72, was generously donated by Cabot.

2.2 Methods

Experiments, data processing, and data interpretation were performed by Mo Kit Chau, unless specified otherwise. The experiments were performed at the University of Toronto, unless specified otherwise.

2.2.1 Methods for the microfluidic generation of agarose and gelatin composites microgels

2.2.1.1 Functionalization of gelatin

Chemical modification of gelatin with tyramine was conducted as reported by Sakai et al.112 Gelatin (3 g) was dissolved in 300 mL of 50 mM aqueous morpholinoethanesulfonic acid. The solution was maintained at 60 °C for 4 h and subsequently cooled to room temperature. Tyramine hydrochloride (1.5 g), EDC (1.10 g), and NHS (0.332 g) were added to this solution. The mixture was stirred for 30 min, and the resulting polymer solution was dialyzed against deionized water, until the fluid outside the dialysis bag no longer exhibited the absorbance peak at 275 nm characteristic of tyramine. The product, gelatin with increased phenolic hydroxyl content (gelatin-Ph), was lyophilized.

2.2.1.2 Synthesis of the agarose−fluorescein conjugate

The agarose−fluorescein isothiocyanate (agarose−FITC) conjugate was synthesized following the reported protocol.113 Agarose (0.2 g) was dissolved in 10 mL of anhydrous DMSO at 50 °C. A solution of FITC in DMSO (200 μL at 100 mg/mL FITC) was added to the agarose solution. The resulting solution was mixed for 3 h at 50 °C, cooled to room temperature, and dialyzed against water (6 × 3L). The solution was lyophilized.

2.2.1.3 Synthesis of triblock copolymer surfactant

The synthesis of the triblock copolymer surfactant was carried out as reported by Holtze et al.111 Krytox 157 FSH (15.47 g) were degassed under stirring in a Schlenk tube. HFE 7100 (15

25 mL) was added and the mixture was manually shaken. Under nitrogen flow, approximately 2 mL of oxalyl chloride was added. After adding a drop of dimethyl formamide, the solution was stirred overnight at room temperature.

The solvent was removed by rotary evaporation and 15 mL of dry HFE 7500 was added. 1.12 g of Jeffamine ED-900 dissolved in 1 mL of dry pyridine is added to the Krytox solution drop-wise. The turbid reaction mixture was stirred overnight. The reaction mixture was poured into 50 mL of methanol. After vigorous shaking, methanol was decanted and the residue was washed with another 30 mL of methanol. The residue was dissolved in 10 mL of HFE 7100. These steps were repeated with acetone, THF, and methanol. The residue was dried in vacuo overnight using a rotary evaporator. The residue was dissolved in 15 mL of HFE 7100 and filtered through a syringe filter (0.45 μm, cellulose). Then, the solvent was removed in vacuo overnight yielding a viscous, yellow-brown, clear oil.

2.2.1.4 Fabrication of microfluidic devices

Photolithographic masters were prepared from SU-8 50 photoresist in bas-relief on silicon wafers. Microfluidic (MF) devices were fabricated in PDMS following a soft lithography procedure114 and maintained in the oven at 140 °C for 12 h. To hydrophobize the surface of the microchannels, the MF device was placed for 6 h under reduced pressure in a desiccator containing trichloro(1H,1H,2H,2H-perfluorooctyl)silane. To further hydrophobize the surface of the microchannels, 1,1,1,3,3,3- hexamethyldisilazane was introduced into the MF device using

N2 gas at 80°C for 3h.

2.2.1.5 Quantification of the compositions of the droplets

The composition of the microgels was quantified by determining the concentration of the dye Basic Blue 41 (added to the solution of agarose at a concentration of 1 mM) in the precursor droplets. Droplets containing a mixture of agarose, gelatin, and hydrogen peroxide were generated by varying the ratio of flow rates of the agarose and gelatin solutions while keeping constant the flow rate of the solution of hydrogen peroxide (for these experiments, we replaced gelatin−Ph with nonfunctionalized gelatin). The images of the composite droplets were captured using a Lumenera Infinity 2-1M camera. The pixel intensity of the image of droplets was determined using ImageJ and corrected by subtracting the pixel intensity value of the

26 background. A calibration curve was constructed to relate the pixel intensity value to the concentration of the dye in the solution.

2.2.1.6 Construction of the calibration curve for the determination of dye concentration in the droplets

The concentration of the Basic Blue 41 dye in the precursor droplets was determined by constructing a calibration curve. Droplets were generated by MF emulsification of aqueous solutions with varying concentrations of the dye. The droplets were imaged using an optical microscope. From the images captured using a Lumenera Infinity 2-1M camera, the pixel intensities were analyzed for the solution in the droplets and the background using by ImageJ software. The corrected pixel intensity values were calculated by subtracting the pixel intensity values for dye-free droplets from the pixel intensity values for the droplets with a varying concentration of the dye, and correcting for the background outside of the droplet.

2.2.1.7 Characterization of the mixing of agarose and gelatin−Ph solutions in droplets

The extent of mixing of polymer solutions in the droplets was examined by emulsifying a mixture of agarose− FITC, gelatin−Ph, and hydrogen peroxide solutions and analyzing the distribution of agarose−FITC within the droplets. A Leitz filter cube with a blue excitation filter and a green barrier filter was used in conjunction with a Leitz Aristoplan microscope equipped with a Lumenera Infinity 2-1M camera to image the fluorescent droplets and the resulting microgels.

2.2.1.8 Viscosity of polymer solutions

The viscosities of the solutions with varying concentrations of agarose and gelatin were determined using a Brookfield rheometer, which was equipped with the Brookfield UL adapter at 37 °C. The total polymer concentration was maintained constant at 4% w/w, while the ratio of gelatin concentration, Cgel, to agarose concentration, Cag was varied.

2.2.1.9 Characterization of microgel morphology

Laser confocal fluorescence microscopy experiments were performed on a Leica TCS SP2 microscope with a HeNe laser operating at 1.2 mW. Imaging of cross sections of the

27 composite microgels in the equatorial plane was conducted using excitation and emission wavelengths of 488 and 490− 540 nm, respectively.

2.2.1.10 Determination of the microstructure of the composite gels

The microstructure of the composite gels was examined using scanning electron microscopy (SEM). Solutions of agarose and gelatin−Ph in HBSS with different polymer weight concentrations were prepared at 60 °C. The solutions were cooled to 37 °C, and HRP was added to achieve a concentration of 10 units/mL. Following the addition of 5 mM hydrogen peroxide, the resulting solution was cooled to 4 °C and maintained at this temperature for 20 min to form a composite gel.

A sample of the gel was placed in a microporous specimen capsule (30 μm pore size, Canemco-Marivac), and water in the gel was gradually replaced with methanol by consecutively submerging the capsules in 20, 40, 60, 80, and 100% (v/v) methanol/water mixtures. The capsule was placed in an Autosamdri-810 Tousimis critical point drier, in which the methanol was exchanged with liquid CO2. The liquid CO2 was brought to a supercritical state. Slow venting of the chambers produced the dried samples, which were imaged using a Quanta FEI 250 scanning electron microscope (10.0 kV).

2.2.1.11 Determination of the stiffness of the composite microgels

The stiffness of the microgels with various compositions was determined using a micropipet aspiration technique.115 A borosilicate glass capillary with an inner diameter of 0.75 mm was pulled using a Vertical Pipette Puller 700C (David Kopf Instrument) to make a “zero diameter” micropipet. The tip of the pipet was cut using the ceramic scoring wafer (Restek) to give an inner diameter of 105 μm. The pipet was connected to a manometer using perfluoroalkoxyalkane tubing (Upchurch Scientifics). The pipet and tubing were filled with deionized water. The pressure at the pipet tip was controlled with a manometer. A droplet of the microgel suspension was deposited on a glass slide mounted on the stage of an Olympus MC3529 inverted microscope, and a microgel particle was trapped at the opening of the micropipet without excess pressure. The pressure drop was increased in a stepwise manner at a rate of 98.1 Pa/ min, and the change in the microgel shape was monitored at each step. The microgel stiffness was quantified using a method reported previously.116

2.2.2 Methods for the investigation of the ionic gelation of cellulose nanocrystals

2.2.2.1 Preparation of CNC suspensions

An 6.43% w/w aqueous suspension of CNC, generously supplied by FP Innovations, was dialyzed against deionized water. The water-to-suspension volume ratio was 100:1, and over the course of dialysis, the water was changed 6 times (the first and the second water changes were 2 and 4 h and subsequent changes were 12 h after the beginning of dialysis). After dialysis, the CNC suspension was filtered using No. 41 and 42 Whatman filter papers, and reconcentrated by evaporating water under ambient conditions until the CNC concentration, CCNC = 5.53% w/w was reached. This suspension was diluted with deionized water to prepare CNC suspensions with the required concentrations.

2.2.2.2 Preparation of CNC gels and sols

Hydrogels of CNCs were prepared by adding a metal chloride, that is, NaCl, MgCl2,

AlCl3, CaCl2, or SrCl2 solution to a CNC suspension to reach a salt molality in the CNC hydrogel in the range of 1−50 mm (we expressed concentrations in molality (m), that is, the number of moles of solute dissolved in 1 kg of solvent). Suspensions with CNC concentration, CCNC, in the range of 0.5−4% w/w were used.

2.2.2.3 State diagram of CNC dispersions and gels

These experiments were performed by Shivanthi Sriskandha.

The inversion test was used to determine the difference between the gel and sol states. A sample was identified as a “gel”, if it did not flow upon inversion of the vial, and as a “sol”, if it did. The states of CNC samples were determined at 25 and 37 °C to foreshadow their behavior during cell culture at physiological temperatures.

2.2.2.4 Scanning electron microscopy

The structure of CNC gels was imaged using scanning electron microscopy (SEM). Supercritical point drying was used to prepare hydrogel samples. First, a CNC gel was placed in

29 a microporous specimen capsule (30 μm pore size, Canemco-Marivac). Next, water was gradually replaced with methanol by consecutively submerging the capsules into 20, 40, 60, and 80% (v/ v) methanol/water mixtures and, finally, in pure methanol. Afterwards, the capsule with the CNC gel was placed in an Autosamdri-810 Tousimis critical point drier. The methanol in the sample was exchanged with liquid CO2, which was subsequently brought to a supercritical state and removed by slow venting. The dried gels were sputter-coated with gold and imaged using a Quanta FEI 250 scanning electron microscope (5 kV).

2.2.2.5 Characterization of rheological properties of CNC gels

The rheological properties of CNC gels were studied using an ARES (TA Instruments) rheometer with a parallel plate geometry. The diameter of the plates and the gap between them were 25 mm and 1 mm, respectively. Amplitude strain sweeps were performed to determine the range of the linear viscoelastic region. Following this experiment, dynamic frequency sweeps were performed with oscillatory frequencies between 0.1 and 100 rad/s at a constant strain of 0.5%.

The range of the linear viscoelastic response of the CNC gels was determined by performing strain amplitude sweeps on gel samples. Strain was applied from 0.01 to 400 % at an oscillatory frequency of 1 Hz.

The hysteresis of shear-induced deformation of CNC gels was determined by performing two consecutive dynamic frequency strain sweeps at 0.5 % strain.

2.2.2.6 Imaging of CNCs

Transmission electron microscopy (TEM) experiments were performed on a Hitachi H- 7000 transmission electron microscope. The CNCs were deposited from an aqueous 0.001 % w/w suspension on copper grids (Electron Microscopy Sciences). The average diameter and length of CNCs were 31 ± 3 and 194 ± 56 nm, respectively, as measured using ImageJ software.

2.2.2.7 Polarized optical microscopy

Gels and suspensions of CNCs were imaged using polarized optical microscopy. The CNC suspensions at CCNC = 4% w/w were prepared as described in section 2.2.2.1. The CNC suspensions at CCNC = 2% w/w were prepared by diluting a CNC suspension of 2.67% w/w with

30 deionized water. Gelation was induced by adding a metal chloride solution to a suspension of CNCs. A 0.2 mm-thick sample of CNC gel was confined between two glass slides to prevent water evaporation, and then placed between two cross-polarizers (U-AN360 and UP110 U-POT, Olympus, Japan). The sample was imaged using a Canon EX-F1 digital camera.

2.2.2.8 Determination of the size of dextran probes for NMR experiments

Dextran with a molecular weight of 150 kDa was dissolved in deionized water at concentration of 5 mg/mL. The size of molecules was measured using dynamic light scattering (DLS). The average size of the dextran molecules was found to be 20 ± 4 nm.

2.2.2.9 Determination of mesh size of CNC gels

These experiments were performed jointly by Mo Kit Chau and Dmitry Pichugan.

The mesh sizes in CNC hydrogels were determined using pulsed field gradient NMR (PFG NMR) by measuring the diffusion coefficients of dextran molecules in solution and in 117 CNC gels (D0 and D, respectively). Dextran was used as a probe because it does not interact with CNCs.118

A suspension of CNC (0.75 g) in H2O at CCNC of 2.67 or 5.33% w/w was added to an NMR tube. Then, 0.25 g of a 5 or 50 mm solution of salt in H O was added, and the contents of the tube were mixed. After gel formation, 1 g of a 20 mg/mL dextran solution in D2O was introduced in the NMR tube above the gel. The sample was allowed to equilibrate overnight, and the supernatant solution was removed. The NMR spectra were acquired using an Agilent DD2 500 MHz instrument equipped with a OneNMR direct detect probe with a 1H 90° pulse width of 8.4 μs. The bipolar pulse pair stimulated echo sequence was used as supplied by Agilent. For the CNC gels, a 2 s saturation pulse was used to suppress the water signal. The recycle delay was set to be 5-fold the longitudinal relaxation time of the dextran in the gel. The gradient pulse length, δ, was 7 ms, and the diffusion delay, Δ, was set to 300 ms. The gradient strength, g, was varied in 15 increments from 1.9 G/cm to a gradient strength that yielded 85−90% attenuation of the dextran signals around 45 G/cm. All spectra were collected using 32 scans per gradient strength with 8 steady states, 4.5 s acquisition time, 6 s recycle delay, and 8 kHz spectral window

centered on the water signal. Attenuations were also measured for dextran in D2O at 10 mg/mL. The signal attenuation was fitted using Origin to the equation

Equation 2.1

where I0 and I are the intensities of the 3.79 ppm dextran signal at the lowest gradient and at different gradient amplitudes, respectively. The γ, δ, g, and Δ are the gyromagnetic ratio of the observed nucleus (1H), gradient pulse magnitude, gradient pulse length, and diffusion time, respectively. D′ is the fitting parameter equal to the diffusion coefficient of the probe. The ratio of D/D0 is related to the hydrodynamic radius of the probe, Rh, the radius of CNC fibrils forming the mesh, Rf, and the radius of the opening between the fibrils, Rp, (half the mesh size) as follows119

Equation 2.2

The hydrodynamic radius of dextran was determined using dynamic light scattering. The average mesh size of each CNC gel sample was calculated as 2Rp and was determined from a set of triplicate experiments.

2.2.2.10 Determination of the size of dextran probes for NMR experiments

2.2.2.11 Small-angle X-ray scattering (SAXS)

These experiments were performed by Dmitro Nykypanchuk and Oleg Gang in Brookhaven National Laboratories.

High-resolution synchrotron-based SAXS measurements were performed at the X9 beamline at the National Synchrotron Light Source, Brookhaven National Laboratory (USA).

For measurements, a CNC gel was placed inside a 1 mm-diameter quartz capillary (Charles Supper, MA), and data were collected using a PILATUS detector. Camera length was calibrated against silver behenate. The data were corrected for sample absorption using a semitransparent beam stop, and the background was corrected for water and capillary scattering.

2.2.3 Methods for the formation of anisotropic freeze-cast aerogels and hydrogels

2.2.3.1 Synthesis of hydrazide-functionalized poly(oligoethylene glycol methacrylate) (H-POEGMA)

These polymers were synthesize by Kevin De France at McMaster University.

The synthesis procedure for the H-POEGMA used in this work was reported by Smeet et 120,121 al. Briefly, AIBMe (74 mg), M(EO)2MA (6.2 g), OEGMA500 (1.8 g), AA (1046 µL) and TGA (150 µL, 10 w/w% in dioxane) were introduced into a round-bottom flask. Following the addition of dioxane (20 mL) to this mixture, it was purged with nitrogen for 30 min. The reaction proceeded for 4 h in an oil bath at 75 °C, after which the solvent was removed via rotary evaporation. Following this step, 200 mL of deionized water was added to the resulting POEGMA solution along with ADH (8.66 g). To adjust the mixture pH to 4.75 ± 0.1, 1M HCl was used, after which EDC (3.87 g) was added to mediate conversion of carboxylic acid groups to hydrazide groups. The value of pH was maintained at 4.75 ± 0.1 via a dropwise addition of 1M HCl over 4 h, until no further pH change was noted, after which the reaction was allowed to proceed overnight. The product was dialyzed (molecular weight cutoff of 3,500 g mol-1) against deionized water for a minimum of six 6 h cycles and then lyophilized. Polymers were stored as a 10 wt% suspension in deionized water at 4 °C.

Hydrazide content (1.61 mmol/g hydrazide groups) was determined via conductometric titration (ManTech, 0.1 M NaOH titrant). The molecular weight of H-POEGMA of Mn = 17.7 kDa and polydispersity of 3.2 were determined by aqueous size exclusion chromatography on a Waters 515 HPLC pump with three Ultrahydrogel columns (30 cm x 7.8 mm i.d. with exclusion limits of 0–3 kDa, 0–50 kDa and 2–300 kDa) and a Waters 2414 refractive index detector. A mobile phase containing 25 mM N-cyclohexyl-2-aminoethanesulfonic acid (CHES) buffer, 500 -1 mM NaNO3 and 10 mM NaN3 was used at a flow rate of 0.8 mL min . The molar ratio of hydrazide-functionalized AA: M(EO)2MA:OEGMA500 was 0.28:0.64:0.061.

2.2.3.2 Preparation of cellulose nanocrystals

These CNCs were prepared by Kevin De France at McMaster University.

Cellulose nanocrystals were generated using the sulfuric acid-mediated hydrolysis of cotton.122 Briefly, blended cotton filter aid was exposed to 64 wt% sulfuric acid solution at 45 oC for 45 min with mechanical stirring, diluted 10-fold in deionized water, and centrifuged for 10 min at ~5000 g. The acidic supernatant was decanted from the centrifuge tubes, leaving a cellulose pellet, since the CNCs were insoluble at acidic pH. The cycles of water adding, centrifuging, and decanting were repeated until a pellet no longer formed upon the addition of water. The resulting suspension was subsequently dialyzed (molecular weight cutoff 12-14,000 g mol-1) against deionized water for a minimum of ten cycles of at least, 12 h each. The CNC suspension was then sonicated using a probe sonicator (Sonifier 450, Branson Ultrasonics, Danbury, CT) for three 15 min-long cycles and stored at 4 °C as a 1 wt% suspension in acid form (pH = 3.2). Sulfate half-ester content was determined by conductometric titration,123 yielding a sulfur content of 0.42 wt % (~ 0.30 charges/nm2). The CNC apparent diameter of 71 nm and electrophoretic mobility of -1.86 × 10-8 m2 V-1 s-1 were determined by dynamic light scattering and electrokinetic potential measurements using a 0.25 wt % CNC suspension in 10 mM NaCl solution (Zetasizer Nano, Malvern, UK).

2.2.3.3 Preparation of aldehyde-functionalized cellulose nanocrystals (A- CNCs)

Functionalized was performed by Kevin De France at McMaster University.

The selective oxidation of CNC surface hydroxyl groups to aldehyde groups was optimized as described Sun et al.124 Sodium periodate and a 1.0 wt% suspension of CNCs were added to a round-bottom flask at the NaIO4:CNC weight ratio of 4:1. The flask was covered with aluminum foil to prevent NaIO4 photo-decomposition. The pH value of the suspension was then adjusted to 3.5 and the suspension was placed under magnetic stirring in an oil bath at 45 oC for 4 h. The reaction was subsequently quenched by cooling the reaction mixture and exposing it to light to de-activate residual NaIO4. The resulting product was dialyzed (molecular weight cutoff 12-14,000 g mol-1) against deionized water for a minimum of ten cycles of at least, 12 h each. The suspension was concentrated to 8 wt% by ambient evaporation and stored at 4 oC. Aldehyde content of 3.66 mmol/g was determined by conductometric titration (ManTech, 0.1 M NaOH

34 titrant) after selectively oxidizing aldehyde groups to carboxylic acids using silver(I) oxide, as described previously.48

2.2.3.4 Atomic force microscopy imaging of aldehyde-modified CNCs

Functionalization was performed by Kevin De France at McMaster University.

Aldehyde-modified CNCs (A-CNCs) were imaged in contact mode in the alternating current (AC) mode with an Asylum MFP-3D atomic force microscope (Asylum Research, Santa Barbara, CA). Rectangular FMR cantilevers (NanoWorld) with normal spring constants of 1.2– 5.5 N/m and resonance frequencies of 60–90 kHz were used to obtain the AFM image.

2.2.3.5 Preparation of anisotropic aerogels and hydrogels by freeze- casting

The experimental setup was developed by Mo Kit Chau and Bernd Kopera. The foams were prepared jointly by Vanessa Machado and Mo Kit Chau.

Two different assemblies were used for freeze-casting. Assembly 1 consisted of an aluminum block equilibrated in a Dewar filled with liquid nitrogen, enabling freeze-casting at a - 196 oC. The setup for freeze-casting at the temperature in the range from -80 to -20 °C (Assembly 2) consisted of an aluminum rod, 2” in diameter and 12” in height (McMaster-Carr 8974K552), topped with six circular polyimide round heaters (Omega Product Number KHR- 2/10) that were spaced by copper plates. Above the heaters was a copper cylinder, 2” in diameter and 1” in height (McMaster-Carr 9103K2). The assembly was placed in a stainless steel Dewar filled with liquid nitrogen. The temperature of the heating elements was controlled using a proportional-integral-derivative controller connected to a thermocouple inserted in the copper cylinder.

An aqueous mixture of A-CNCs and H-POEGMA was charged into a Teflon or a polycarbonate tube closed on one end with a 0.9 mm-thick copper lid. Aerogels for SEM imaging and photographing were fabricated in a 1.3 cm inner diameter Teflon tube. Aerogels for compression tests and swelling experiments were fabricated in a polycarbonate cuboidal tube with the inner length and width of 0.95 cm. The filled mold was placed atop the cooled aluminum block (Assembly 1) or atop the cooled copper block (Assembly 2). Freezing of the

35 sample was assessed visually. After that, the samples were maintained on the cooling block for an additional minute to ensure complete freezing and subsequently, freeze-dried in lyophilizer. To form hydrogels, freeze-cast aerogels were swollen in deionized water.

2.2.3.6 Scanning electron microscopy imaging of aerogels

The resulting A-CNC-H-POEGMA aerogels were freeze-fractured either parallel, or perpendicular to the direction of freeze-casting, sputter-coated with gold, and imaged using a scanning electron microscope (SEM, Quanta FEI 250 scanning electron microscope, 10 kV).

2.2.3.7 Determination of surface area of aerogels

These experiments were performed by Laura Reyes at the University of Toronto.

The surface area of aerogels was determined by volumetric nitrogen adsorption at 77 K using a Quantachrome Autosorb-1-C and calculated using a Brunauer-Emmet-Teller (BET) equation.125 Aerogels were outgassed overnight at 80 oC (at 100°C the samples discoloured, possibly due to the polymer degradation).

2.2.3.8 Hydrogel swelling

These experiments were performed jointly by Vanessa Machado and Mo Kit Chau.

To characterize the anisotropic swelling of the hydrogels in deionized water, we defined a coordinate system for the cuboidal aerogel/hydrogel samples, in which the X- and Y-axes were orthogonal to the direction of ice growth, and the Z-axis was parallel to the direction of ice growth. The original cuboid aerogels had an equal length, l0, and height, h0. Upon swelling, at time t, the hydrogels acquired the corresponding dimensions, l(t) and h(t). The dimensions of aerogels and hydrogels were measured using a caliper at different time intervals. The degree of swelling in the XY-plane was determined as

푙(푡)−푙o 푄XY(푡) = Equation 2.3 푙o

The degree of hydrogel swelling in the Z-direction was characterized as

ℎ(푡)−ℎo 푄푧(푡) = Equation 2.4 ℎo

2.2.3.9 Mechanical testing

These experiments were performed by Kevin De France at McMaster University.

Cyclic compression tests for hydrogel samples were conducted in an aqueous environment at 22 oC using a Mach-1 Mechanical Tester (Biomomentum Inc, QC) operating under parallel-plate geometry. Prior to testing, aerogels with a width and length of 9.5 mm and a height of 11 to 14 mm were immersed in water for 10 min. The resulting hydrogels were compressed in either the XY-direction (perpendicular to the direction of freeze-casting), or the Z-direction (parallel to the direction of freeze-casting) over 50 cycles, by applying XY-strains of 50% or Z-strains of 10% (the samples buckled at Z-strains exceeding 10%). The tests were performed in triplicate, with the Young’s modulus values calculated from the first compression cycle (after an initial pre- compression/pre-loading step to condition the hydrogels).

2.2.3.10 Small-angle X-ray scattering

These experiments were prepared by Sabine Rosenfeldt and Bernd Kopera at the University of Bayreuth.

The aerogels were characterized at room temperature using a small-angle X-ray system (Double Ganesha AIR, SAXSLAB, Denmark). The X-ray source was a rotating anode (copper, MicoMax 007HF, Rigaku Corporation, Japan) that provided a micro-focused beam. The data were recorded by a position-sensitive detector (PILATUS 300K, Dectris). To cover the range of scattering vectors between 0.004-2.0 Å-1 different detector positions were used. The circularly averaged data were normalized to incident beam intensity, sample thickness, and measurement time before subtracting the background (air). The data analysis was performed with the software Scatter (version 2.5), which was also used to perform calculations based on simple geometric models.126 The irradiation volume was approximately 200 μm x 200 μm x 1 cm, where 1 cm is the sample thickness.

2.2.4 Preparation and characterization of anisotropic polyurethane foams

2.2.4.1 Synthesis and dispersion of polyurethane in water

Triblock copolymer of polycaprolactoned and polytetrahydrofuran (PCL-b-PTHF-b-PCL, 45.5 g) and dimethylolpropionic acid (DMPA, 1.8 g) were added to a 1 L jacketed three-necked

37 round-bottom flask, which was then heated to 60 C overnight under vacuum. The round-bottom flask was fitted with a condenser, nitrogen inlet, and nitrogen outlet. The content of the flask was stirred with an overhead stirrer at 400 rpm. Then, the flask was heated to 85 C. To form the prepolymer, 11.9 g of isophoronediisocyante (IPDI) was added to the heated flask. The reaction was monitored using an attenuate total reflectance-Fourier transform infrared (ATR-FTIR) spectrometer (Vertex 70, Bruker Corp.) with a single reflection diamond ATR crystal (MIRacle, Pike Technologies). When the intensity of the isocyanate peak at 2260 cm-1 leveled off, the temperature was lowered to 60 C, 200 mL of acetone was added, and the prepolymer was chain- extended using 0.8 g of ethylene diamine.

An infrared (IR) spectrum of the PCL-b-PTHF-b-PCL oligomer and PU polymer product was acquired using the ATR-FTIR spectrometer. A drop of the aqueous PU dispersion was placed on the ATR crystal and left to dry before the measurement was taken.

Gel permeation chromatography (GPC) characterization of the PU polymer was conducted at 85 °C (Perkin Elmer Oven Column Selector) using a 1.0 g/L solution of lithium chloride in N-methylpyrrolidone (NMP) as eluent, at a flow rate of 1.0 mL/min through two Agilent PLgel 5μm MIXED-C columns equipped with a Perkin Elmer Refractive Index Detector. Poly(methyl methacrylate) (PMMA) standards were used for calibration. The polyurethane (PU) was dissolved at the concentration of 2 mg/mL in the GPC eluent.

To disperse the PU in water, first, the PU was dispersed by stirring the polymer solution in acetone at 700 rpm, while adding 140 mL of water drop-wise. The acetone was subsequently removed by rotatory evaporation under reduced pressure. The concentration of the dispersion was found to be 22.4 wt% by gravimetric analysis.

2.2.4.2 Cryogenic transmission electron microscopy

Cryo-TEM imaging was performed by Dr. Markus Drechsler at the University of Bayreuth.

A drop of the dilute PU dispersion in water was placed on a copper grid coated with a lacey carbon film. The water was then removed with a filter paper. Immediately after, the grid was instantly shock-frozen by plunging it rapidly into liquid ethane. The grid was cooled to approximately 90 K by liquid nitrogen in a temperature-controlled freezing unit (Zeiss Cryobox).

The frozen specimen was inserted into a cryogenic transfer holder (CT3500, Gatan) and transferred to a Zeiss EM922 Omega energy-filtered TEM instrument. The sample was imaged with an acceleration voltage of 200 kV and temperatures of ~90 K. The particle size was determined using the ImageJ software.

2.2.4.3 Electrokinetic potential characterization of PU particles

The electrokinetic potential of the PU particles in the aqueous dispersion, diluted to 0.1 wt% using deionized water, was determined using a Malvern Zetasizer Nano ZS instrument.

2.2.4.4 Preparation and characterization of CNCs

The concentration of the CNC suspension, received from FPL University of Maine, was 11.8 wt%. The CNC suspension was diluted to 2.5 wt% with deonized water and dialyzed against deionized water using regenerated cellulose membrane with a 12 kDa molecular weight cut-off (Sigma Aldrich). The volume ratio of the suspension to dialysate was 1:100 per volume change. The water was changed a total of 10 times at 1 h, 2 h, 6 h, and subsequently, at 12 h intervals. After dialysis, the CNC suspension was filtered using No. 41 and 42 Whatman filter papers and then, using 0.45 m syringe filters (Starstedt Filtropur, PES-membrane). The CNC suspensions were re-concentrated by centrifugation for 10 h at 14 000 x g (Thermo Scientific, Heraeus Multifuge X1R Centrifuge) and the supernatant was discarded. The final concentration of the suspension was 8.3 wt%, as determined by gravimetric analysis.

The size of the CNCs was characterized by TEM. A drop of CNC suspension, diluted to 0.1 wt% with deionized water, was dried on a carbon-coated grid (Ted Pella Inc.). A 2 wt% aqueous uranyl acetate solution at pH = 4.2 (adjusted by HCl) was used as a negative stain. The CNCs were imaged using a Hitachi H-7000 Transmission Electron Microscope. The dimensions of the CNCs were determined using ImageJ software.

For the dialyzed CNC suspensions, inductively-coupled plasma atomic emission spectroscopy (ICP-AES) was performed on an Optima 7300 ICP-OES spectrometer, Perkin Elmer. The CNCs contained 1.86  0.02 wt% of Na and 2.67  0.06 wt% of S per dry weight of CNC. These values infer that there are 8.0 wt% sulfate groups per dry weight of CNC present and 50 % of these sulfate groups are associated with Na+.

Electrokinetic potential measurements of the CNC suspensions (diluted to 0.5 wt% with deionized water) were performed using a Malvern Zetasizer Nano ZS instrument.

2.2.4.5 Characterization of carbon black and carbon nanofibers

SEM imaging was performed by Vanessa Machado. XPS experiments were performed by Rana Sodhi. The fitting and data analysis were performed by Mo Kit Chau.

Carbon black (CB) powder were imaged using a high-resolution SEM (HRSEM) on a Hitachi S-5200 (5.0 kV). Carbon nanofibers (CNFs) were imaged using a Quanta FEI 250 in TEM mode.

X-ray photoelectron spectroscopy (XPS) spectra of CB and CNFs were obtained on a ThermoFisher Scientific K-Alpha XPS system (ThermoFisher Scientific, E. Grinstead, UK). The samples were attached to the sample plate using conductive, double-sided tape. A monochromatic Al Kα X-ray was used with a nominal spot size of 400 µm. Following the collection of survey spectra (pass energy (PE) - 200 eV), high energy resolution spectra (PE – 25 eV) were obtained for C 1s, O 1s, and C KLL Auger peaks. Relative atomic percentages were obtained from the C 1s and O 1s peak areas using the supplied sensitivity factors (modified Scofield – 1.000 and 2.881 respectively). Charge compensation was applied using the combined e-/Ar+ floodgun. The energy scale was not adjusted. All data acquisition and peak fitting were performed with the supplied software (Advantage v5.949 and v5.926 respectively).

2.2.4.6 Stabilization of graphitic components in water using CNCs

These experiments were performed by Vanessa Machado.

To assess the performance of CNCs as dispersants for CB and CNFs, the dispersions of graphitic materials in water with and without CNCs were prepared. Carbon black or CNFs (a sufficient amount to result in a final 5wt% loading in the foam) were weighed in a vial to which a 4 wt% CNC suspension or deionized water was added. For example, if the intended final weight of the freeze-cast dispersion was 1 g, 0.05 g of the graphitic material should be added. The suspension was vortexed for 30 s and then sonicated for 30 min. Next, the suspension was vortexed again for 30 s and then sonicated for 30 min. Then, water was added such that the final concentration of the graphitic material was 5 wt% and the final concentration of CNC, if present,

40 was 2.5 wt%. The dispersions were vortexed again for 30 s and photographs of the vial were immediately taken. These suspension were diluted by a factor of 1000 with deionized water. The samples were subsequently deposited on carbon-coated copper TEM grids (Ted Pella Inc.) and imaged using TEM (Quanta FEI 250 Electron Microscope).

2.2.4.7 Freeze-casting of PU foams

Freeze-casting of PU foams was performed by Mo Kit Chau and Vanessa Machado.

The setup for freeze-casting dispersions at -20 °C consisted of an aluminum rod (2” in diameter and 12” tall, McMaster-Carr 8974K552), topped with six circular polyimide round heaters (Omega Product Number KHR-2/10) with each heater spaced by copper plates. Above the heaters was a copper cylinder (2” in diameter and” tall, McMaster-Carr 9103K2). The assembly was contained in a stainless steel container, which was filled with liquid nitrogen. The temperature of the heaters was controlled using a proportional-integral-derivative controller connected to a thermocouple embedded in the copper cylinder.

Foams of pure PU were denoted as “PUpure,” polyurethane foams loaded with 2.5 wt%

CNC were denoted as “PUCNC,” polyurethane foams loaded with 2.5 wt% CNC, and 5 wt% CB or 5 wt% CNF were denoted as “PUCNC-CB” or “PUCNC-CNF”, respectively.

For PUpure, the PU dispersion was diluted to 20 wt% with deionized water and transferred to the mold. Foams of PUCNC were prepared by the CNCs and PU dispersions such that the final loading of CNC in the foams is 2.5 wt%. For PUCNC-CB and PUCNC-CNF, the CB or CNF were first, weighed into a vial and a desired amount of the CNC suspension was added. The graphitic materials were dispersed in water by the following steps: 30 s vortexing, 30 min sonicating, 30 s vortexing, 30 min sonicating, and 30 s vortexing. The PU dispersion was added to this mixture and the mixture was vortexed, again, for 30 s before the mixture was transferred into a mold.

Two types of molds were used for freeze-casting. When freeze-casting samples for imaging and mechanical testing, smaller, polycarbonate cuboidal tubes (outer width = 1/2”, length = 1/2”, and wall thickness = 1/16”, McMaster-Carr) were used. When freeze-casting samples for thermal measurements, larger, chemical-resistant PVC cuboidal tubes (outer width = 3/4”, length = 3/4” and wall thickness = 0.060”, McMaster-Carr) were used. In either case, the molds were closed at one end with a 0.9 mm-thick copper bottom lid.

Immediately prior to freeze-casting, a solution containing a stoichiometric amount of EDC to react with one third of the carboxylic groups present on the PU particles was added to the dispersion mixture. For example, if 1 mole of carboxylic groups was present in the dispersion, 0.3 moles of EDC should be added. For all samples, the final concentration of PU in the dispersion mixture prior to freeze-casting was 20 wt%. The mold, filled with the dispersion mixture, was placed atop the cooled copper block. Once the suspension was frozen, as assessed visually, it remained on the cooling block for an addition 3 min for smaller molds and an additional 5 min for larger molds before it was removed from the cooling block and subsequently, freeze-dried. After freeze-drying, the foams were annealed at 90 ºC for 8 h. The loading concentrations of the graphitic material (CB or CNFs) and CNCs, if present, in the final foams were 5 and 2.5 wt%, respectively.

2.2.4.8 Scanning electron microscopy imaging of PU foams

Experiments were conducted by Mo Kit Chau and Bernd Kopera.

The foam samples were cut parallel or perpendicular to the ice-growth direction. The samples were then sputter-coated with gold and imaged using a Quanta FEI 250 scanning electron microscope (15 kV).

2.2.4.9 X-ray microtomography imaging of PU foams

Experiments were conducted by Jacqueline Uhm.

A polyurethane foam containing 2.5 wt% CNC and 97.5 wt% PU (PUCNC) was scanned with a Skyscan 1072 Micro-CT (Bruker, Artselaar, Belgium) with a linear resolution of 2.34 µm at a magnification of 120 with an accelerating voltage of 61 kV and a tube current of 170 μA. Projection images were acquired over 180° at angular increments of 0.23° with an exposure time of 1.3 seconds per frame, averaged over four frames. Three-dimensional images were reconstructed using the reconstruction software provided by the manufacturer (NRecon Version 1.6.4.1), where the ring artefact reduction was applied as needed.

2.2.4.10 Differential scanning calorimetry of PU foams

These experiments were performed by Fabian Nutz at the University of Bayreuth.

Heat capacity of the PU foams was determined by differential scanning calorimetry (DSC) on a Mettler Toledo DSC 2 according to ASTM E1269 standard testing method. The foam mass varied between 10 and 15 mg. The measurements were performed under a nitrogen flow of 50 mL min-1 at a heating rate of 20 K min‑1. Two heating-cooling cycles between 0 and

150 °C were performed for each sample. The second heating cycle was used for the calculation of specific heat capacity (cp) at constant pressure.

2.2.4.11 X-ray diffraction measurements for PU foams, CNCs, and CNFs

These experiments were performed by Dr. Wolfgang Milius at the University of Bayreuth.

X-ray diffraction (XRD) measurements were performed on a X'Pert MPD Pro Powder Diffractometer from PANalytical in reflection geometry with Cu-Kα radiation filtered with a nickel filter. The scattering beam stop was on the primary and secondary side 1/8 º. The diffraction angle, 2θ, range was 2 to 60 º.

2.2.4.12 Compression testing of PU foams

Compression tests were performed using an Instron 5848 Microtester with a 500.0 N load cell. Any menisci present on the top of the foams were removed by a razor blade. Samples were either compressed perpendicular, or parallel to the ice-growth direction. Foams were first, pre- compressed to 20 % of the original dimension. The samples were then compressed to 50 % strain at 3 mm/min. The recoverability of the foams was characterized by comparing the initial dimensions of the foam and the dimensions after 50 % compressive strain.

2.2.4.13 Compression-decompression cycles for PU foams

The foams were pre-compressed to 20 % of the original dimension. Samples were then subjected to 20 compression-decompression cycles up to 20 % compressive strain. The compression rate was 3 mm/min with 2 min time intervals between compressions for sample relaxation.

2.2.4.14 Thermal diffusivity measurements and thermal conductivity calculation

Preliminary thermal diffusivity experiments were performed by Mo Kit Chau and Bernd Kopera jointly. Thermal conductivity measurements were performed by Bernd Kopera at the University of Bayreuth.

The PU foams were sliced parallel and perpendicular to the ice-growth direction using a razor blade (Canemco-Marivac, Double Edge Razor Blade, 314-2). Each slice was 1-2 mm thick. The thickness near the center of the samples was determined with a high-resolution digital height measuring unit (Mitutoyo Litematic VL-50). The thermal diffusivity of the PU foams was measured using xenon flash analysis (XFA) on a Linseis XFA 500 Xenon Flash apparatus with an InSb infrared detector. White PUpure and PUCNC foams were either coated with a thin layer of CB powder on each side or wrapped with one layer of gold leaf. Data evaluation was conducted using the software Aprosoft Laser Flash Evaluation v1.06 using the radiation or combined model, which considers finite pulse, heat loss, and radiative transport effects.

The thermal diffusivities,  of the foams in the parallel and perpendicular directions were determined by xenon flash analysis (XFA). The thermal diffusivities were then used to calculate the thermal conductivity, , as

휅 = 훼 ∙ 푐푝 ∙ 휌 Equation 2.5,

where the specific heat capacity, cp, of the PU matrix was obtained by DSC, and  is the density of the foam. The dimensions of the foam were determined using a caliper. The mass of the foam was determined by weighing. The density of the foam was calculated by dividing the mass by the volume.

2.2.4.15 Infrared thermogram images

Infrared thermogram images were taken by a Variocam HD research from Infratec (Dresden,

Germany). The sliced PUCNC-CNF samples were illuminated by a 488 nm laser spot with < 10 mW for local thermal excitation. The images were recorded before thermal equilibrium was reached.

Chapter 3 Microfluidic generation of composite biopolymer microgels with tunable compositions and mechanical properties

Contribution: Mo Kit Chau designed and carried out experiments, conducted data analysis, interpreted the results and wrote the manuscript.

Introduction

Cell fate is, to a large extent, governed by complex, spatiotemporally regulated interactions between cells and the local extracellular matrix (ECM).128 The complexity of natural ECMs poses a challenge to understanding the mechanisms of external chemical and biophysical regulation of cell fate and reproducing ECM properties in vitro. With the goal of recapitulating the nature of cell−ECM interactions, biological and synthetic hydrogels acting as model three- dimensional (3D) cellular environments are being developed at a rapid pace.129 Important

130 applications of such hydrogels include screening of stem cell niches for tissue engineering, clinical diagnosis of mechanoregulation-related diseases,131 producing scaffolds for tissue engineering,132 and 3D cell cultures for elucidating biochemical mechanisms ex vivo.133

Synthetic hydrogels offer a promising approach to artificial, instructive ECMs with controllable cell-specific properties, yet generation of synthetic ECMs requires an interconnected porosity on the length scale of cellular processes, appropriate mechanical properties, a structure that mimics protein fibrils interwoven within a hydrated glycosaminoglycan network, and, most importantly, low cytotoxicity. These features are challenging to achieve in covalently crosslinked synthetic hydrogels.

Hydrogels of biopolymers such as natural polysaccharides (e.g., alginate, agarose, pectin, or chitosan) and proteins (e.g., gelatin, collagen, or elastin) are non-cytotoxic and can be generally formed under mild conditions, thereby preserving cell viability,134 yet a particular spectrum of hydrogel properties, including their chemical composition and adhesion site density,

45 porosity, permeability, and elasticity, may not be realized using a single biopolymer. The combination of desired properties can be achieved by generating artificial ECMs from composite hydrogels formed by more than one biopolymer and optimization of their properties for cell encapsulation.

Artificial ECM hydrogel particles (microgels) make superior candidates for cell encapsulation over bulk gels for several reasons.76,135 The small (<200 μm) dimensions of microgels allow diffusion of oxygen, nutrients, and metabolic products to and from encapsulated cells.136 The smaller size of the polymer matrix also decreases light scattering that may interfere with characterization. The monodispersity of microgels allows accurate elucidation of the structure−cell activity relationship. Microfluidics (MFs) has been used to generate monodispersed cell-laden microgels.78,137

Microfluidics involves the manipulation of small amounts of fluids in channels with dimensions of tens to hundreds of micrometers.138 Droplet MF, in which discrete volumes of fluid are formed in an immiscible phase inside the MF channels, is an avenue to creating monodispersed polymer particles.139 The advantages of using droplet MFs to make polymer particles is the frugal use of reagents, fast mixing times between multiple components, precise control over dynamics and high-throughput generation of combinatorial libraries. Droplet MFs can be used to generate monodisperse microgels allowing control over the cell encapsulation rate, the average cell−cell distance, the delivery of functional molecules to a particular number of the encapsulated cells, and the ability to characterize cell feedback to their respective microenvironment.76 Importantly, MF generation of cell-laden microgels offers the ability to produce vast combinatorial libraries of instructive ECMs with varying chemical and biophysical properties.140

Microgel compositions can be tuned in a throughput manner by varying the relative volumetric flow rates of multiple streams of precursor polymer solutions supplied to the MF device. A mixed solution is then emulsified, and composite precursor droplets are ultimately gelled to yield microgels. The applicability of this strategy was demonstrated for the generation of cell-laden microgels with varying elasticities53 and for controllable cell co-culture.137

The ability to vary microgel compositions is especially beneficial for generating multicomponent hydrogels. The ECMs in vivo are complex mixtures of biopolymers, including

46 collagen, proteoglycans, and glycoproteins.128 To explore a broad range of available hydrogel components, MFs offers the ability to prepare composite microgels from a mixture of polymers, with each component playing a particular role, e.g., controlling the structural stability or adhesive properties of the hydrogel.

In this chapter, we report MF generation of composite agarose−gelatin microgels with variable compositions and mechanical properties. Agarose is a thermoresponsive, non-adhesive polysaccharide that gels upon cooling at 17−28 °C, depending on the concentration and source of agarose used.141 Agarose gels withstand a cell culture temperature of 37 °C without liquification, which makes them particularly useful in cell encapsulation experiments, yet because of the lack of adhesive and signaling domains that are present on natural ECM proteins,142 agarose hydrogels pose a limitation for their use as mimetics of the native ECM. Gelatin is the hydrolysis product of collagen, the most abundant protein present in the mammalian ECM.143 Gelatin contains adhesive Arg-Gly-Asp (RGD) sequences that bind to integrin receptors on the cell surface. Below 8−18 °C, gelatin molecules associate into triple helices that form a gel;144,145 however, these gels dissociate upon being heated to 37 °C, the cell culture temperature. A combination of agarose and gelatin in composite microgels can be used to control the composition of the hydrogels and associated rigidity, structure, and cell adhesion properties.

We used a MF strategy to generate composite agarose/gelatin microgels with varying compositions, structures, and mechanical properties. The agarose component of the microgels gelled at the reduced temperature, while gelatin modified with phenolic hydroxyl groups underwent peroxidase-catalyzed gelation. The composition of the microgels was changed in a high-throughput manner by mixing aqueous solutions of agarose, chemically modified gelatin, and hydrogen peroxide in different volume ratios and generating composite precursor droplets. The tuning of the composition of the microgels was used to control their morphology, structure, and stiffness. The developed MF strategy is an efficient approach to the generation of combinatorial libraries of composite microgels with tunable compositions, structures, and mechanical properties, which can be used as instructive artificial microenvironments for cell encapsulation and culture.

3.1 Results and Discussion

3.1.1 Design of the Microfluidic Device

Soft lithography114 was used to generate the MF devices used in this work (see Section 2.2.1.4). The design of the MF device is shown in

Figure 3.1a and enlarged in Figure 3.1b which show the dimensions of the microchannels. Gelatin with increased phenolic hydroxyl content (gelatin-Ph) was synthesized using a reported literature procedure by coupling the carboxylic groups on the gelatin with tyramine using EDC coupling (Scheme 1).112 Gelatin-Ph was required since the enzyme horseradish peroxidase (HRP) crosslinks the gelatin via these functionalities. Inlets 1, 2, 3, and 4 supply the continuous (oil) phase, gelatin-Ph solution containing enzyme, agarose solution and a crosslinker solution, respectively.

3.1.2 Gelation Time of Gelatin-Ph

To determine gelation time of gelatin-Ph, HRP was first added to a gelatin-Ph solution. Final concentrations of gelatin-Ph between 2 and 3.5 % (w/w) and a final concentration of 1 or 5 unit/mL of HRP were used. The gelation time of was determined as the time required for the solution to immobilize a stir bar at 80 rpm after the addition of 1 mM of hydrogen peroxide. The final total volume of solution per sample containing the gelatin-Ph, HRP, and hydrogen peroxide was 1.5 mL. Figure 3.2 shows that the gelation times of the solutions increased with decreasing concentrations of gelatin-Ph and hydrogen peroxidase.

3.1.3 Microfluidic Generation of Composite Microgels

Figure 3.3 shows an optical microscopy image of the MF flow-focusing droplet generator used in this work. Aqueous solutions of agarose, gelatin−Ph mixed with HRP, and hydrogen peroxide were supplied to the MF device as the components of the droplet phase via three separate inlets at flow rates of Qag, Qgel, and Qcross, respectively. Three aqueous solutions, namely, an agarose solution, a gelatin−Ph solution containing 10 units/mL HRP, and a 5 mM

50 hydrogel peroxide solution, were supplied to the MF flow-focusing droplet generator as the constituents of the droplet phase.146 The concentration of hydrogen peroxide in the droplets and in the corresponding microgels was comparable to the 1 mM of hydrogen peroxide proven to be non-toxic for the encapsulated cells (approximately 95% cell viability in gelatin−Ph gels).112 Fluorinated oil mixed with 1.0 % (w/w) a triblock copolymer surfactant was supplied to the MF device as a continuous phase at a flow rate Qcont. The surfactant is composed of perfluoroalkylether outer blocks, and a polyethylene glycol-polypropylene glycol copolymer inner block. Immediately prior to the orifice, three aqueous streams formed a single stream, which was subsequently broken into droplets in the orifice by the shear force imposed by the continuous fluorinated oil phase. Figure 3.3 shows the generation of precursor droplets composed of the mixed solution of agarose, gelatin−Ph, hydrogen peroxidase, and hydrogen peroxide. In the MF experiments, the diameter of droplets was tuned from 175 to 105 μm by varying the flow rate of the continuous phase from 0.6 to 24 mL/h, respectively.

The droplets traveled downstream of the orifice toward a serpentine channel, where they were partly gelled due of the enzymatic cross-linking of gelatin−Ph by the reaction of hydrogen peroxide in the presence of HRP. The partially gelled droplets exited the MF device and were collected in a solution of HBSS buffer. Following a 20 min cooling at 4 °C, the agarose component of the droplets gelled. The fluorinated oil was removed by evaporation. In our earlier

51 work, we verified that keeping cells at 4 °C for 45 min does not significantly affect their viability.53,137 Figure 3.4a shows a representative optical microscopy image of the composite microgels. The distribution of microgel dimensions is shown in Figure 3.4b. The particles had an average diameter of 175 μm and a polydispersity or coefficient of variation (defined as the standard deviation in the diameter of the droplets divided by the mean diameter) of 3%.

For the microgels generated in this work, we estimated a minimal cell concentration in a feeding suspension that would lead to a 100% encapsulation rate. Using the Poisson distribution equation147

Equation 3.1,

52 where P(x) is the fraction of microgels expected to contain x cells and λ is the average number of cells per droplet. For microgels with diameters of 105 and 175 μm, a 100% fraction of microgels containing at least one cell (that is, leading to a 100% encapsulation rate) could be achieved for minimal cell concentrations in the droplet phase of 7.6 × 106 and 1.6 × 106 cells mL−1, respectively.

3.1.4 Composition of the Droplets

To develop a MF approach to the generation of composite biopolymer microgels, we determined the relationship between the flow rates of the individual solutions supplied to the MF device and the composition of the composite droplets (and microgel particles). While for liquids with low viscosities, the fraction of each component in the droplet is determined directly from its relative volumetric flow rate,148 for liquids with a large difference in viscosity such as agarose and gelatin−Ph solutions, an adjustment in relative flow rates may be needed to generate droplets with a particular composition.149 The viscosity of agarose, gelatin and their mixtures are shown in Figure 3.5. The total polymer concentration was maintained constant at 4% w/w, while the ratio of gelatin concentration, Cgel, to agarose concentration, Cag was varied. Viscosity decreases with increasing Cgel. Since the viscosity of the agarose and gelatin solutions are significantly different, the concentration of agarose and gelatin in the resulting droplets must be determined empirically as a function of their respective flow rates.

The concentration of agarose in the droplets was determined by the addition of a dye to the agarose precursor solution. The concentration of the dye is proportional to the concentration biopolymer in the respective precursor solution. The concentration of the dye can be determined using an optical imaging technique in which the pixel intensity could be determined for varying concentration of dye. First, calibration curve must be constructed to for relating the pixel intensities to known concentrations of dye in formed droplets. Calibration curve for droplets of water containing varying concentrations of Basic Blue 41 is shown in Figure 3.6.

Figure 3.6 Calibration curve used for the determination of dye concentrations in droplets.

The variation in the weight concentration of agarose, Cag, in the composite droplets was determined by adding 1 mM Basic Blue 41 dye to the agarose solution and measuring the concentration of the dye in the composite droplets using a calibration curve (Figure 3.6). Aqueous solutions of agarose, gelatin−Ph mixed with horseradish peroxidase, and hydrogen peroxide were supplied to the MF device, as shown in Figure 3.3, at volumetric flow rates of Qag,

Qgel, and Qcross, respectively. In the emulsification process, we varied the relative flow rate of the agarose solution, while maintaining Qag + Qgel + Qcross = 0.6 mL/h, Qcross = 0.2 mL/h, and Qcont =

0.6 mL/h, such that the Qag/(Qag + Qgel + Qcross) was varied from 0.1 to 0.4.

Figure 3.7 shows the experimentally measured variation in Cag in the droplets, plotted as a function of the relative flow rate of the agarose solution, as well as the estimated concentration of agarose in the droplets, Cag,est, calculated as

° 푪풂품,풆풔풕 = (푪풂품푸풂품흆풂품)/(푸풂품흆풂품 + 푸품풆풍흆품풆풍 + 푸풄풓풐풔풔흆풄풓풐풔풔) Equation 3.2, where Cag° is the weight concentration of agarose in the solution supplied to the MF device [Cag°

= 3% (w/w)] and ρag, ρgel, and ρcross are the densities of the agarose, gelatin−Ph, and hydrogen

55 peroxide solutions, respectively. For the dilute aqueous polymer solutions used in this work, we assumed that ρag = ρgel = ρcross ≈ 1 g/mL. Figure 3.7 shows that the experimental value of Cag was consistently lower than Cag,est. This is likely a result of the higher resistance to flow of viscous liquids through confined microfluidic channels, because of viscose dissipation.150

On the basis of Figure 3.7, we conclude that the concentration of agarose in the composite droplets (and the corresponding microgels) was lower than expected from its relative flow rate. In the rest of the paper, for the characterization of the properties of microgels, we relied on the agarose−gelatin−Ph concentration ratio based on the experimentally determined microgel composition using Figure 3.7.

3.1.5 Microgel Morphology

Microgel morphology was examined using confocal fluorescence microscopy. For these experiments, we generated microgels from gelatin−Ph mixed with horseradish peroxidase,

56 hydrogen peroxide, and agarose covalently labeled with a fluorescent dye, FITC (agarose− FITC). The FITC-labeled agarose was synthesized by conjugating FITC with agarose as described previously.113 Figure 4 shows images of individual composite microgels containing

Cag‐FITC and Cgel‐Ph at Cag‐FITC/Cgel‐Ph concentration ratios varying from 1.43/0 to 0.63/1.15, where Cag‐FITC is the weight concentration of agarose−FITC and Cgel‐Ph is the weight concentration of gelatin−Ph.

The images of microgels formed by agarose−FITC had a uniform distribution of fluorescence intensity throughout the entire microgel particle. The images of the composite gels were darker, because the concentration of agarose—FITC was lower in the composite microgels. The distribution of fluorescence intensity throughout the microgel uniform; darker domains indicated partial segregation of gelatin−Ph. The average size of these domains increased with gelatin−Ph fraction from ∼10 μm at a Cag‐FITC/Cgel‐Ph ratio of 0.95/0.72 to ∼25 μm at a Cag‐

FITC/Cgel‐Ph ratio of 0.63/1.15. Gelation of the composite droplets occurred in two steps: by the rapid chemical gelation of gelatin−Ph forming a loose network and then thermal gelation of agarose. Because agarose is more hydrophilic than gelatin,151 to minimize the surface energy of

57 the system, agarose molecules were expected to migrate to the surface of the microgel. Another possible explanation for the observed agarose distribution is that since the crosslinking of the gelatin component is faster than the gelation of the agarose, the agarose component could be rejected towards the outer surface of the microgel.

To assess whether the phase separation had occurred as a result of poor mixing on-chip, optical microscopy was used to image droplet formation. Optical fluoresnce microscopy was used to image droplet formation with agarose-FITC, gelatin-Ph, and crosslinker (Figure 3.9). The fluorescence intensity in the droplets appear to be uniform downstream though the resolution of the fluorescent images was insufficient to resolve for certain whether or not the droplets were well-mixed after emulsification. In another experiment, Basic Blue 41 dye at the concentration of 1 mM is added either to agarose, gelatin, or H2O2 precursor solution for visualization. Basic Blue 41 is added either to agarose, gelatin, or H2O2 precursor solution for visualization. Optical image of the droplets during droplet formation and 15.5 mm downstream are shown in Figure 3.10 and Figure 3.11for agarose and gelatin, respectively. The uniform distribution of Basic Blue 41 in the droplets shortly after droplet formation implied good mixing of the the dye in the droplet, though the mixing of the agarose and gelatin components could not be inferred. Because MF emulsification of the mixture of agarose, gelatin−Ph, and hydrogen peroxide led to the complete mixing of multiple components in the composite droplets, we conclude that partial phase separation occurred in the stage of gelation and was consistent with earlier reports of phase separation in an agarose−gelatin mixture.152,153

a) b)

Figure 3.9 Optical fluorescence microscopy images of the composite droplets containing agarose-FITC, gelatin-Ph, and crosslinker, taken immediately after the orifice (a) and 15.5 mm downstream of the orifice (b). Qag,= Qgel = Qcross = 0.1 mL/hr. Qcont = 2 mL/hr. The scale bars are 500 μm.

Figure 3.10 Optical image of a) droplet formation and b) 15.5 mm downstream on chip. Basic blue is added to the agarose component for visualization. Scale bars are 500 μm.

Figure 3.11 Optical image of a) droplet formation and b) 15.5mm downstream on chip. Basic blue is added to the gelatin component for visualization. Scale bars are 500 μm.

Control over the extent of phase separation in composite agarose−gelatin microgels microgels can be advantageous for cell encapsulation purposes: the coexistence of domains with dimensions on the order of, or larger than, the cell size can provide encapsulated cells with heterogeneous microenvironments similar to those in vivo. The degree of phase separation within the microgels can be controlled in a throughput manner by tuning the composition of the microgels and the relative rates of gelation of agarose and gelatin−Ph. Furthermore, phase separation can also be tuned by controlling kinetic trapping (rate and temperature of cooling in this case).154

3.1.6 Microstructure of Composite Gels

We examined the microstructure of the composite hydrogels with varying compositions using SEM. Figure 3.12 shows representative SEM images of agarose−gelatin−Ph gels prepared under conditions similar to those used for the MF preparation of the microgels. The structure of agarose gels appeared to be fiberlike and porous. The average width of the fibers of ∼40 nm was consistent with the reported agarose structure.155 With an increasing content of gelatin−Ph in the composite gel, the structure of the gel became less porous. At a relatively high gelatin−Ph

59 content, the nanofibrils of agarose appeared to be embedded in the gelatin−Ph host (Figure 3.12d). A gel formed from the chemically cross-linked gelatin−Ph exhibited a globular structure with an average globule size of ∼100 nm. We estimate the average pore size to be in the range from 79 ± 77 nm in pure agarose to 58 ± 51 nm in chemically modified gelatin. The average pore size in nonmodified gelatin hydrogels was reported to be 320−640 nm.153 In our work, smaller pores could be a result of syneresis, contraction of a gel exuding liquid, due to the chemical cross-linking of gelatin−Ph.

Figure 3.12 SEM images of agarose−gelatin−Ph gels. The Cag/Cgel‐Ph ratios in the gels were (a) 2/0, (b) 1.5/0.5, (c) 1/1, (d) 0.5/1.5, and (e) 0/2. The scale bar is 250 nm.

3.1.7 Mechanical Properties of Composite Microgels

The mechanical properties of soft particles are typically measured using atomic force microscopy,52,156 micropipette aspiration,157 and microfluidic confinement.158 In the present work, the mechanical properties of the composite microgels were characterized using a micropipette aspiration method.157 The microgels with different compositions were aspirated into a glass micropipette with a diameter of 150 μm, and the deformation of the microgels was monitored as a function of the decrease in pressure applied to the microgel. The negative pressure differential

applied to the microgel, ΔP, was related to the aspirated length of the microgel, (x − xo), by the relationship157

Micropipette aspiration experiments Equation 3.3, A representative image of a microgel aspirated into a micropipette is shown in Figure S8. where xo and x are the lengths of the intrusion of the microgel into the micropipette at ΔP = 0 and after the finite suction pressure, ΔP, is applied, respectively, and Rp is the radius of the micropipette. The quantity S, which is the slope of the plot of ΔP versus (x − xo)/Rp, is the microgel stiffness or a measure of the degree of deformation of the microgel. While for incompressible, linear elastic materials, the stiffness S can be related to the Young’s modulus of the material,159–161 the microgels studied in this work were compressible and could also lose water, when subjected to a stress.158 Figure S8. Aspiration of 150 µm-diameter agarose microgel into a glass micropipette with the inner diameter of 53 µm. Scale bar is 50 µm. A representative image of a microgel aspirated into a micropipette is shown in Figure 3.14. WeW havee ha alsove a studiedlso stud theied susceptibilitythe susceptib ioflit ythe o fmicrogel the mic rtoog volumeel to v olosslum eby l oexaminingss by the variation of the change in volume (푉 − 푉)/푉 with (x − x )/Rp where V and V are the examining the variation of relative ch0ange in v0o lume ()V0 -Vo /V0 with x /0R p , where V0 unstressed and stressed volumes of the microgel. The volume change and the extent of the and V are the unstressed and stressed volumes of the microgel. microgel intrusion into the micropipette, can be related by

()V -V x - x 0 = D o (3) V0 Rp Equation 3.4, The slope of this graph, D, is expected to be proportional to E / K , where K is the bulk in which D, is the proportionality constant. D, is expected to be proportional to E / K, where K is modulus of the microgel. D can, therefore, be regarded as the dewatering propensity of the bulk modulus of the microgel. D can, therefore be regarded as the dewatering propensity of the microgel. In Figure 9, we have shown D for microgel particles with different ratios of the microgel. Parameters, V0, V, x, x0, and Rp can be calculated from analyzing the images, and D canCa gbe/C gdetermined.el and Cag + CInge lFigure mainta 3in.14ed ,c weons thaveant. W shownithin a Dn eforxp emicrogelrimental eparticlesrror, the vwithalue differentof C ratios of Cag/Cgel and Cag + Cgel maintained constant. Within an experimental error, the value of C was was approximately constant over the range of Cag/Cgel ratios explored in the experiments, approximately constant over the range of Cag/Cgel ratios explored in the experiments, which which suggested that the Poisson’s ratio, which is directly related to D, was relatively suggested that the Poisson’s ratio, which is directly related to D, was relatively insensitive to the compositioninsensitive tofo tthehe cmicrogel.ompositio nThis of tresulthe mi cwasroge consistentl. This res uwithlt wa thes c ofactnsis tthatent wtheith bulk the ftoac sheart modulusthat the ratiobulk tofo smicrogelshear modu showedlus ratio ao flittle mic rdeviationogels show fromed a lanitt lordere devi ofati ounity.n from We an haveorder found that the propensity to dewater is relatively insensitive to the composition of the microgels (Figure of unity.

3.14). Therefore, we adhered to the description of the slope of ΔP versus (x − x0)/Rp as an apparent “stiffness”, which was done previously.157

Figure 3.14 Variation of the dewatering propensity, D, as a function of the fraction of the agarose concentration, Cag/ (Cag +Cgel).

Figure 3.15a shows the variation of ΔP with (x − xo)/Rp for microgels generated at varying Cag/Cgel‐Ph weight ratios, while Cag + Cgel‐Ph was maintained at 4% (w/w). The variation was linear, justifying the use of a linear elastic theory in interpreting the data. The variation in stiffnesss S extracted as the slope of the ΔP versus (x − xo)/Rp lines is shown in Figure 3.15b for

62 composite microgels with different compositions. The microgel stiffness increased with an increasing weight ratio of agarose to gelatin from ∼200 to 850 Pa, which agreed with the fact that agarose gels are significantly stiffer than gelatin gels of the same concentration at room temperature.162

3.2 Conclusions

Composite biopolymer microgels of agarose and chemically modified gelatin have been generated by the microfluidic emulsification of mixtures of the corresponding solutions and subsequent gelation of the precursor droplets. The compositions of the droplets could be varied in a high-throughput manner by changing the volume flow rate of the components. Agarose underwent gelation at a reduced temperature. Chemically modified gelatin underwent enzymatically catalyzed gelation in the presence of a subtoxic concentration of hydrogen

63 peroxide. The microstructure, morphology, and stiffness of the composite microgels were controlled in a throughput manner by changing the microgel composition. The microfluidic approach can be extended to other combinations of biopolymers composites. This work demonstrates a possible platform for the on-demand generation of microgels that can be used as instructive artificial extracellular matrices with desired properties.

Chapter 4 Ion-Mediated Gelation of Aqueous Suspensions of Cellulose Nanocrystals

Contribution: Mo Kit Chau designed and carried out experiments, conducted data analysis, interpreted the results and wrote the manuscript.

Introduction

Many biological polymers, including proteins and polysaccharides, form hydrogels by the reversible physical association or entanglement of high aspect-ratio structural units called nanofibrils, strands, or filaments.33,164 Nanofibrils are built by the hierarchical assembly of many molecules and have diameters in the range of tens to hundreds of nanometers and lengths up to micrometers. The association of nanofibrils in networks is generally driven by hydrogen bonding, hydrophobic forces, electrostatic forces, and van der Waals interactions. If the concentration of nanofibrils in suspension is sufficiently high and their persistence length is not too large, networks can be formed via entanglement of nanofibrils.

The hierarchical nature of nanofibrillar hydrogels often imparts improved mechanical properties, nonlinear viscoelastic behavior,26 larger pore sizes,30 and enhanced thermal stability165 compared to the hydrogels made from the molecular solutions of the same polymer. Nanofibrillar gels formed by biopolymers, such as collagen and agarose, are generally non-cytotoxic, biodegradable, and noncytotoxic.33 These properties make them promising candidates for applications in catalytic scaffolding,166 templating polymer composites,167 drug delivery,25 and tissue engineering,168 to name a few.

Shape-anisotropic, high-aspect ratio cellulose nanocrystals (CNCs) have recently gained great interest in the materials science and nanoscience fields. These nanoparticles are composed of cellulose molecules packed in a parallel fashion with a helical twist.169 The molecules are held

65 together in the CNCs by hydrogen bonds. The diameter and length of CNCs are typically in the range of 10−30 and 50−500 nm, respectively, depending on their source. They have a high degree of crystallinity (54−88%) and excellent mechanical properties with an estimated tensile strength of 110−220 GPa.169 Furthermore, CNCs bear surface hydroxyl groups, which can be used for their chemical functionalization with low-molecular weight molecules or polymers.170 Importantly, CNCs are environmentally friendly, abundant in nature, and inexpensive.170

Gelation of aqueous CNC suspensions has been achieved in several ways. For example, an increase in CNC concentration in the suspension (up to 14.5 wt%) has led to the formation of lyotropic liquid crystalline gels.66 Alternatively, composite hydrogels have been formed from a mixture of CNCs with a gelling component (e.g., methylcellulose,171 hydroxyethyl cellulose, or hydroxypropyl guar).118 Other methods for the preparation of CNC gels utilize a reduction in electrostatic repulsion between negatively charged CNCs. More specifically, preparation of CNCs by acid hydrolysis of wood pulp results in CNCs with surface anionic sulfate groups. Electrostatic repulsion between these anionic groups renders CNCs colloidally stable. Desulphation of the CNC surface by heating their suspension in the presence of glycerol leads to a decrease in colloidal stability and favors attraction between CNCs, thereby yielding thixotropic CNC hydrogels.60 In this process, however, the replacement of water with glycerol may limit the range of biorelated applications of the CNC hydrogels. An alternative method relies on increasing the ionic strength of CNC suspensions by adding salts. The addition of salts reduces the Debye length of CNCs and suppresses electrostatic repulsion between them, thereby leading to dominant attractive interactions, such as van der Waals forces and hydrogen bonding. For example, CNC hydrogels have been formed upon the addition of NaCl to an aqueous CNC suspension.172

Gelation using multivalent cations has been studied for 500− 2000 nm long carboxylate- decorated cellulose nanofibrils,173 which are significantly longer than CNCs and contain alternating amorphous and crystalline domains. The storage moduli of such hydrogels increased with increasing cation charge number, which enabled tuning of the gel mechanical properties. To the best of our knowledge, a systematic study of the properties and structure of CNC hydrogels formed in the presence of cations with different charge numbers and dimensions has not been reported even though CNC gels can offer improved colloidal stability, enhanced mechanical

66 properties, and a higher propensity for alignment under shear in comparison to cellulose nanofibrils.

This chapter describes the results of a comprehensive experimental study of the formation and properties of CNC hydrogels formed by the addition of cations with varying charge numbers and ionic radii to aqueous CNC suspensions. More specifically, we examined the rheological properties by oscillatory rheometry and structure of CNC gels using scanning electron microscopy, NMR, polarization optical microscopy, and small-angle X-ray scattering.

For a particular CNC concentration, we found that hydrogel stiffness increased with increasing charge number and ionic radius of the added cations. The increase in gel stiffness was accompanied by an increase in mesh size, in contrast to the prediction of conventional poroelastic theory of molecular gels. Both features were attributed to the stronger side-by-side CNC association in the presence of added cations, which led to the formation of a stiffer network. As a result, the mechanical properties of the CNC gels could be accurately tuned by varying the type and concentration of the cations. The established structure−property relationships have important implications for the use of CNC gels as drug delivery vehicles and as scaffolds for tissue engineering.

4.1 Results

4.1.1 Ionically Mediated Gelation of CNC Suspensions

Prior to experiments, CNCs were dialyzed against deionized water. The water-to- suspension volume ratio was 100:1, and over the course of dialysis, the water was changed 6 times (the first and the second water changes were 2 and 4 h and subsequent changes were 12 h after the beginning of dialysis). The average diameter and length of CNCs used in the present work were 31 ± 3 and 194 ± 56 nm, respectively (Figure 4.1). Gelation of dialyzed CNC suspensions was induced by adding metal chloride solutions of NaCl, CaCl2, MgCl2, SrCl2, or

AlCl3. The final salt concentration in the suspension varied from 1 to 50 mm (millimolal), and the weight concentration of cellulose, CCNC was varied from 0.5 to 4% w/w. Table 4.1 (left column) summarizes the notations used in the present work. For example, the Ca50−4 gel was prepared at a 50 mm concentration of CaCl2 and CCNC = 4% w/w. Table 4.1 also shows the charge numbers and ionic radii of the added cations in columns 3 and 4, respectively.

Figure 4.1 Transmission electron microscopy image of CNCs after dialysis against water. The scale bar is 500 nm.

Table 4.1 Nomenclature used for the CNC samples and characteristics of the cations.

CNC Solubility of Salt concentration metal sulphates concentration in gelling Cation Cation radius* in water** in the gel suspension Salt charge Sample added number (Å) (g/100g) (mm) (% w/w)

CNC0-4 ------0 4.0

CNC0-2 ------0 2.0

Na50-4 NaCl 1+ 1.02 28.1 50 4.0

Mg50-4 MgCl2 2+ 0.72 35.7 50 4.0

Ca50-4 CaCl2 2+ 1.00 0.205 50 4.0

Sr50-4 SrCl2 2+ 1.18 0.0135 50 4.0

Al50-4 AlCl3 3+ 0.54 38.5 50 4.0

Ca5-4 CaCl2 2+ 1.00 0.205 5.0 4.0

Ca50-2 CaCl2 2+ 1.00 0.205 50 2.0 * Ionic radii were taken from ref 174. ** The solubility of metal sulphates in water are taken from reference 175.

In a qualitative study of gelation of CNC suspensions, we focused on sol−gel transitions induced by the addition of metal chlorides containing cations with different ionic radii and charge numbers. Figure 4.2 a−c shows state diagrams characterizing the effect of CCNC and salt concentration on gel formation. The critical concentration of added metal chloride required for gelation reduced with increasing charge number of the added cation. For example, a higher concentration of NaCl was required to trigger gelation in comparison with MgCl2 or AlCl3. The

69 diagrams replotted for the corresponding Debye lengths of CNCs are given in Figure 4.2 a′−c′. The Debye length of the CNCs used in this work was calculated as176

Equation 4.1,

where ε is the dielectric constant of the electrolyte solution, ε0 permittivity of free space

(vacuum), kB is the Boltzmann constant, T is the temperature, e is the electron charge, NA

Avagadro’s constant, zi is the charge number of species i, and Mi is the molar concentration of that species. The similarity between the state diagrams shown in Figure 4.2 a′−c′ for different added salts implied that screening of electrostatic repulsion between the CNCs played an important role in the formation of the CNC network structure.

70 represent the boundaries between the sol and gel states. The lines between the sol and gel states in each diagram were drawn by visual estimation.

The difference in ionic radii between Mg2+, Ca2+, and Sr2+ cations did not significantly affect sol−gel boundaries in the state diagrams (Figure 4.3). The threshold value of CCNC required for the sol−gel transition remained similar for Na+, Mg2+, Al3+, Mg2+, Ca2+, and Sr2+, which suggested that a minimum value of CCNC of ∼1.5 wt % was required to form a network in the presence of cations.

Figure 4.3 State diagrams of aqueous CNC suspensions in the prescence of cations. The sol () and gel () states are observed following the addition of metal salts solutions of a) NaCl, b)

MgCl2, c) AlCl3, d) CaCl2, and e) SrCl2 at 25 °C (a-e) at 25 °C; and f) NaCl, g) MgCl2, h) AlCl3, i) CaCl2, and j) SrCl2 at 37 °C. The lines between the sol and gel states in each diagram were manually estimated.

Qualitatively similar sol−gel transitions were observed at 25 and 37 °C (the physiological temperature), as shown in Figure 4.3, suggesting that CNC gels were stable at physiological temperature and, thus, may be suitable for cell culture and in vivo applications if their cytotoxicity is appropriate.

4.1.2 Rheological Properties of CNC Gels

The rheological properties of the CNC gels were characterized by their storage modulus (G′), loss modulus (G”), complex shear modulus (G*), and tan δ (loss tangent). The rheological properties were studied for the CNC gels formed by adding 50 mm solution of a particular metal chloride to a CNC suspension such that the final value of CCNC was 4% w/w. Prior to the performing dynamic frequency sweeps, amplitude strain sweeps were performed on the CNC gels to determine the region of their linear viscoelastic response. A representative dependence of shear moduli on strain for the gel Ca50-4 is shown in Figure 4.4.

Dynamic frequency sweeps were performed at 0.5% strain to obtain G*, G′, and G′′. The complex shear modulus, G*, characterizes the rigidity of a gel subjected to deformation below the yield stress.177 The complex shear modulus is defined as G* = G′ + G′′. Figure 4.5 shows the effect of the cation charge number and ionic radii on the G′ and G′′ of the hydrogels. Over the entire range of oscillatory frequencies from 0.1 to 100 rad/s and for all the gel samples examined, the value of G′ was greater than G′′, which signified gel-like properties of the sample. Furthermore, for all of the CNC gels, the values of loss factor (defined as, tan δ = G′′/ G′) were significantly smaller than unity, which suggested that elastic behavior dominated.177 Table 4.2 summarizes the values of |G*|, G′, G”, and tan δ for the CNC gels, all at the oscillatory frequency of 1 rad/s.

Figure 4.4 Strain amplitude sweep for the Ca50-4 gel at 25 oC.

Table 4.2 Rheological properties and mesh size of CNC gels* Sample G' G'' tan δ |G*| Average mesh size** (kPa) (kPa) (kPa) (nm)

Na50-4 1.5 0.1 0.08 1.5 71 ± 6 Mg50-4 7.7 1.0 0.13 7.8 80 ± 2 Ca50-4 10.0 1.4 0.14 10.1 85 ± 1 Sr50-4 11.8 1.7 0.14 12.0 92 ± 7 Al50-4 13.9 2.2 0.16 14.1 83 ± 1 Ca5-4 1.6 0.4 0.12 3.0 79 ± 0.4

Ca50-2 0.6 0.1 0.09 0.6 156 ± 5

* Rheological experiments were performed at 0.5 % strain at a frequency of 1 rad/s. ** The uncertainty in the mesh size represents a standard deviation, based a set of triplicate experiments.

The results presented in Figure 4.5a indicate that both elastic and viscous contributions to gel rigidity increased with an increasing cation charge number in the order Na+ < Mg2+ < Al3+. This trend was consistent with an earlier study of gels formed by cellulose nanofibers functionalized with carboxylate surface groups.173 The increase in strength of CNC gels with an increasing cation charge number was caused by the reduction in Debye length and stronger screening of electrostatic repulsion between the CNCs, thereby making attractive van der Waals and hydrogen bonding interactions dominant forces favoring CNC association, in agreement with the Derjaguin−Landau−Verwey−Overbeek (DLVO) theory.178

Figure 4.5b shows the variation in G′ and G′′ values for the hydrogels formed in the presence of divalent cations with various ionic radii. The values of G′ and G′′ at the oscillatory frequency of 1 rad/s are also shown in Table 4.2. The increase in ionic radii of the cations led to an increase in G′ and G′′ values over the entire frequency range. The increase in gel rigidity with increasing ionic radii (Mg50−4 < Ca50−4 < Sr50−4) correlated with the decrease in solubility of the corresponding metal sulfates in water: the solubility of MgSO4, CaSO4, and SrSO4 are 35.7,

0.205, and 0.0135 g per 100 g of water, respectively.179 This trend suggested that the divalent cations, whose metal sulfates have low solubility in water, could bridge two adjacent sulfate half- ester groups of CNCs.

A stronger association between sulfate groups and metal ions of varying ionic radii was rationalized using the hard-soft acid-base (HSAB) theory. The HSAB theory states that hard (highly electrophilic, non-polarizable) acids prefer to coordinate with hard (highly electronegative, non-polarizable) bases, while soft (weakly electrophilic, polarizable) acids prefers to coordinate with soft (weakly electronegative, polarizable) bases.180 The softness of the cations increases with ionic radii, Mg2+

Gels formed in the presence of Ca2+ cations were used to study the effect of salt and CNC concentrations on the rheological properties of the system (Figure 4.6). Upon the increase of Ca2+ cation concentration from 5 to 50 mm (corresponding to the Debye length of CNCs from

2.35 to 0.78 nm, respectively) at CCNC = 4% w/w, the magnitude of the complex modulus, |G*|, increased from 1.6 to 10.1 kPa (Table 4.2). Similarly, increasing CCNC from 2 to 4% w/w at a Ca2+ cation concentration of 50 mm increased the value of |G*| from 0.6 to 10.1 kPa, respectively. Thus, a 10-fold increase in divalent salt concentration resulted in a 6-fold increase in gel rigidity, whereas a 2-fold increase in CCNC resulted in a 17-fold increase in |G*|. Thus, we conclude that (i) the mechanical properties of the CNC hydrogels were more sensitive to changes in CCNC than to the concentration of added cations and (ii) the addition of cations can be used to fine-tune the mechanical properties of CNC hydrogels.

The values of G′ and G′′ shown in Figure 4.5 and Figure 4.6 were dependent on the oscillatory frequency. At low frequencies (from 0.1 to 1 rad/s), the value of G′ increased with frequency, whereas G′′ decreased. At higher oscillatory frequencies (from 1 to 100 rad/s), both G′ and G′′ increased with frequency. These trends were rationalized as follows. At low frequency, the dissipation of energy occurred due to the motion of large sections of the CNC network,181 whereas at high frequencies, the motion of small segments of the network (associated with a higher rigidity of the CNC network) led to a smaller change in G′ and G′′ with frequency.

The hysteresis in shear of the gels was examined by repeating a dynamic frequency sweep immediately after completion of the first one. For Mg50−4, Al50−4, Ca50−4, Sr50−4, and Ca50−2 gels (Figure 4.7 and Figure 4.8), similar values of G′ and G′′ were for two consecutive sweeps, which implied that the gels either did not change or rapidly recovered their structure after deformation. Significant hysteresis was observed for Na50−4 and Ca5−4; over the entire

77 frequency range, the value of G′ increased in the second experiment (Figure 4.7 and Figure 4.8), which was ascribed to the shear-induced reorganization of CNCs in the gel.

4.1.3 Characterization of Hydrogel Structure

4.1.3.1 Electron microscopy characterization of hydrogel structure

Figure 4.9 shows scanning electron microscopy (SEM) images of the CNC hydrogels. The samples were prepared by the supercritical point drying method extensively used for imaging of hydrogel structures.173,182,183 All the hydrogels prepared in the presence of different cations exhibited a similar nanofibrillar network structure with a random orientation of nanofibrils on a length scale of several micrometers.

Figure 4.9 Scanning electron microscopy images of CNC hydrogels: (a) Na50−4, (b) Mg50−4, (c) Al50- 4, (d) Ca50−4, and (e) Sr50−4. The scale bars are 1 μm.

4.1.3.2 NMR characterization of hydrogel structure

The structure of CNC hydrogels was further characterized by their mesh size by measuring the diffusion coefficients of the dextran molecular probe in solution and in the CNC hydrogels using pulsed field gradient NMR. The 1H NMR spectra of 150 kDa dextran probe at various values of Z, in water and in the Ca50-4 are shown in Figure 4.10 and Figure 4.11, respectively, in which

Equation 4.2, where γ, δ, g, and Δ are the gyromagnetic ratio of the observed nucleus (1H), gradient pulse magnitude, gradient pulse length, and diffusion time, respectively. As can be inferred from Figure 4.10, the Z value is proportional to the gradient pulse magnitude. The attenuation of

79 dextran resonance with increasing Z was related to the apparent diffusion coefficient of the dextran molecules, D’, by

Equation 4.3,

where I0 and I are the intensities at a particular dextran resonance at the lowest gradient amplitude and at different gradient amplitudes, respectively. The normalized echo attenuations at

3.79 ppm for 150 kDa dextran in D2O and in Ca50-4 are both plotted against Z in Figure 4.12. The attenuation curve for Ca50-4 is representative of and similar to that of all other CNC gels. The data points were fitted (using the least squares method) to Equation 4.3 (the fit is also shown in Figure 4.12). The resulting diffusion coefficients and their corresponding variance are summarized in Table 4.3.

Figure 4.10 1H NMR spectra of free dextran molecules at varying gradient strengths.

Figure 4.11 1H NMR spectra for dextran molecules embedded in the Ca50-4 gel at varying gradient strengths.

Figure 4.12 Experimental and fitted values of the normalized echo attenuation for 150 kDa dextran in the solution D2O and in a Ca50-4 gel, plotted against Z.

Table 4.3 Diffusion coefficients of dextran in D2O and in CNC hydrogels

Diffusion coefficient Standard

Sample of dextran deviation*

( 10-7 cm2/s) ( 10-7 cm2/s)

In Solution (D2O) 1.95 0.21

Na50-4 1.32 0.07

Mg50-4 1.31 0.06

Al50-4 1.38 0.00

Ca50-4 1.40 0.04

Sr50-4 1.42 0.05

Ca50-2 1.70 0.02

Ca5-4 1.34 0.02

*Standard deviation was calculated from the results of three measurements.

The mesh size of the CNC hydrogels were calculated using the equation for the obstruction-model derived by Amsden119

Equation 4.4,

where D0 and D are the apparent diffusion coefficients of dextran molecule in solution and in the

CNC gels. The hydrodynamic radius of the probe is Rh, the radius of CNC fibrils forming the mesh is Rf, and the radius of the opening between the fibrils (corresponding to half the mesh size) is Rp. The hydrodynamic radius of dextran was determined using dynamic light scattering.

The average mesh size of each CNC gel was calculated as 2Rp and was determined from a set of triplicate experiments.

Table 4.2 shows the average mesh sizes of CNC hydrogels formed in the presence of different cations. At 50 mm salt concentration and CCNC = 4% w/w, the mesh size increased in the gels formed by adding cations with higher charge numbers and larger ion sizes. We attribute the increase in mesh size to the same two factors that affected gel rigidity. First, reduction in the Debye length of the CNCs (which for Na50−4, Ca50−4, and Al50−4 was 1.34, 0.78, and 0.55 nm, respectively) led to the screening of electrostatic repulsion between the CNCs. Second, an increase in the cation radius enhanced CNC attraction by increasing metal−ligand affinity between the metal acid and the sulfate groups. Both effects favored association of CNCs and led to the formation of denser and/or thicker fibrils, which for a particular CCNC, formed a gel with a larger mesh size.

Even though there was a logarithmic dependence of the mesh size on D0 / D, the standard deviations for each set of triplicates were fairly low attesting to the high precision of the measurements. The dextran used had a Mw = 147, 600 Da with Mw/Mn = 1.47. The large dispersity of the polymer may affect the absolute value of the determined mesh sizes, though the trend of mesh size with varying cation charge and size could still be observed. In the future, it is possible to use monodispersed polymers with varying molecular weights to determine if there are observable differences in the mesh size values. However, it is important to note that the spin- echo technique has been used to measuring the dimensions of restricting geometries by measuring the self-diffusion of the solvent molecule itself,184,185 therefore it is not necessary to use probes that have sizes similar to the pore dimensions.The variation in CCNC at a constant salt content (e.g., at 50 mm CaCl2) also influenced the gel mesh size. A decrease in CCNC from 4 to 2% w/w led to an increase in mesh size from 85 to 156 nm, respectively (Table 4.2). In this case, a larger mesh size was the result of a smaller number of associating CNCs and a lower number of contact points between CNCs per gel volume.

2+ At CCNC = 4% w/w, with a 10-fold reduction in the concentration of Ca cations from Ca50−4 to Ca5−4 (an increase in Debye length from 0.78 to 2.35 nm), the mesh size decreased from 85 to 79 nm for Ca50−4 and Ca5−4, respectively. This effect occurred due to a weaker association between the CNCs at a lower cation concentration.

4.1.3.3 Polarized optical microscopy

Figure 4.13 shows polarized optical microscopy (POM) images of the CNC suspension

(sample CNC0−4) and CNC gels with CCNC = 4% w/w. The POM image of the CNC0−4 sample exhibited a streaked texture referred to as “pre-cholesteric order” (Figure 4.14a).66,186 Upon the introduction of 50 mm salt (Figure 4.13b−f), the streaked texture of all of the gels was replaced by a marble-like texture, which suggested that the addition of salt destroys any order present before. At a low concentration of CaCl2 at 5 mm, the streaked structure was partly preserved, although the density of the streaks increased (Figure 4.14a).

Figure 4.14 Polarization optical microscopy images of (a) Ca5−4 gel, (b) CNC0−2 suspension, and (c) Ca50−2 gel. The scale bars are 100 μm.

The variation in CCNC also affected the gel texture examined by POM. At CCNC = 2% w/w, the CNC suspension phase-separated into anisotropic and isotropic phases,187 and the POM image of the gel featured small chiral nematic domains (tactoids) dispersed in an isotropic continuous phase (Figure 4.14b). In the presence of 50 mm of CaCl2, both Ca50−4 and Ca50−2 featured a similar marble-like texture (Figure 4.13e and Figure 4.14c, respectively). Again, the addition of salt destroys the presence of order in the gel.

4.1.4 Characterization of the Gel Structure by Small-Angle X-ray Scattering

Small angle X-ray scattering (SAXS) profile of the CNC0−4 sample revealed well- defined interference peaks with a correlation length of ∼40 nm (Figure 4.15). We used the rod model to describe the scattering profile of this system.188 The fit of the experimental profile to the theoretical one yielded an inter-rod distance of ∼40 nm and an average CNC diameter of ∼6 nm (the small value of the CNC diameter in comparison with those found from TEM imaging was ascribed to their whisker-like geometry). Thus, long-range repulsive interactions between the CNCs led to well-defined distances between the nanofibrils.

In contrast, for CNC gels prepared by adding cations, no features could be attributed to a well-defined correlation length in the scattering vector varying from 0.005 to 0.12 Å−1 (corresponding to length scales from approximately 5 to 125 nm). Such behavior can be explained either by a highly irregular distance between the CNCs associating in fibrils or the dense packing of CNCs into fibrils. In the latter case, the scattering contrast diminished due to the negligible gap between the CNCs, which resulted in the featureless SAXS profiles.

Because the average mesh size in the CNC gels was <120 nm (Table 4.2), we conclude that the gels obtained in the presence of cations did not exhibit an ordered structure with the characteristic length scale up to 125 nm (the upper limit of SAXS measurements). The CNC gels formed without the addition of cations (Figure 4.13a and Figure 4.14b) exhibited a precholesteric order with the length scale exceeding 125 nm, in addition to the periodic distance between CNCs in the fibrils. Instead, gels formed in the presence of cations showed a “marble” POM structure, which most likely originated only from the CNC association side-by-side.

4.2 Discussion

Ionically mediated gelation of CNC suspensions yielded isotropic nanofibrillar gels. The rheological properties and structure (mesh size) of the gels depended on the type of cation introduced to the CNC suspension. To emphasize this effect, we plotted the variations in |G*| and mesh size as a function of the Debye length of the CNCs for gels formed by the addition of different types of salts at constant CCNC and salt concentration (Figure 4.16). The change in |G*| and mesh size followed the same trend: with an increasing charge number and ionic radii of the cation added, the value of |G*| and mesh size increased. The CNC suspensions without the addition of cations (Figure 4.13a and Figure 4.14b) exhibited a precholesteric order with the length scale exceeding 125 nm, in addition to the periodic distance between CNCs in the fibrils. Gels formed in the presence of added salts showed a “marble” POM structure, which suggest the loss of precholesteric order.

The value of |G*| and mesh size increased as Na50−4 < Mg50−4 < Al50−4, due to the association of CNCs into stronger and thicker fibrils. We stress that for hydrogels Mg50−4, Ca50−4, and Sr50−4 at the same Debye length, the values of |G*| and mesh size increased with

87 increasing ionic radii, which implied that an effect other than electrostatic screening could be in play. This effect is most likely bridging between CNCs.

Increasing rigidity with increasing mesh size was an unexpected result and was contrary to the conventional poroelastic theory,189,190 which rationalizes that gel rigidity increases with reducing mesh size. In the case of ionically mediated CNC gelation, the gels became stronger with increasing mesh size, due to the association of CNCs into fibrils to form a hierarchical structure of nanofibrillar hydrogels.

4.3 Conclusions

Chapter 5 Anisotropic Hydrogels Derived from Cellulose Nanocryals

Contribution: M. Chau contributed to the manuscript by designing and carrying out experiments, data analysis and interpretation, and article writing. B. Kopera helped with the freeze-casting setup design. K. J. De France and K. J. W. Chan prepared POEGMA solutions and CNC suspensions. K. J. De France performed the compression tests. S. Rosenfeldt and B. Kopera performed the SAXS experiments. V. Machado prepared some of the freeze-casted samples and performed the swelling experiments. E. Cranston, T. Hoare, and S. Förster provided guidance and suggestions on experimental design, data interpretation, and article writing.

Introduction

Hydrogels have a broad range of applications in drug delivery,191 tissue engineering,14,192 separation of biological molecules,193 and water purification.194 The utilization of hydrogels as biomaterials has a particular appeal, since they can be engineered to exhibit biophysical and chemical properties that are similar to native extracellular matrix. Since many tissues, e.g., striated muscle,195 cartilage,196 or cornea197, to name just a few examples, have anisotropic hierarchical morphologies, there is a growing interest in developing approaches for the fabrication of anisotropic hydrogels that exhibit direction-dependent pore shape, microstructure, stiffness, and conductivity.195–210 In tissue engineering, aside from biomimicry, anisotropic pore shape and hydrogel structure, in general, are important for cell guidance211 and differentiation,19 as well as mass transport of biofactors and nutrients throughout the scaffold.95,208,212 In bioseparation, control over the shape anisotropy of hydrogel pores may enhance the selectivity of the filtration of biological species and/or minimize the pressure drop across the matrix.213

Anisotropic hydrogels have been fabricated by applying tensile or compressive forces to shape-anisotropic gel components, e. g., carbon nanotubes or cellulose nanocrystals, within an isotropic hydrogel matrix.202,204,214,215 Self-assembled fibrils of peptide amphiphiles26 or lamellar bilayers of polymerizable surfactants25, 27 have been oriented within a hydrogel matrix using shear forces. Alternately, dielectrophoresis has been utilized to align carbon nanotubes in an isotropic hydrogel matrix.206 Micropatterning approaches 7 and 3D printing218 have been applied to create anisotropic hydrogels with pores of well-defined sizes and geometries.

Directional freeze-casting (also known as ice-templating) is another promising method for fabricating anisotropic gels with a well-defined porous structure.95,208,209,212,219–223 In this method, precursor solutions or suspensions of monomers or polymers are frozen under a unidirectional temperature gradient, thereby excluding the solute from the ice lattice into the space between the growing ice crystals.92,224,225 To form a hydrogel, the resulting free-standing microporous scaffold is swollen with water. The elongated pore shapes in the hydrogel replicate the shapes of the unidirectionally grown ice crystals.

Generally, in order to prevent re-dissolution of the freeze-cast scaffold upon its immersion in water, post-processing material photocrosslinking or photopolymerization are used.207,209,213,219,226–228 A more efficient strategy would be simultaneous freeze-casting and cross-linking, as demonstrated for anisotropic hydrogels of agarose221 and glutaraldehyde- crosslinked gelatin94. This alternative approach can be readily implemented in composite hydrogels, in addition, to greater morphological control achieved by varying the ratio of hydrogel components.

Recently, cellulose nanocrystals (CNCs) have gained interest as a component of composite hydrogels, due to the mechanical strength, commercial availability, and biocompatibility of CNCs.55 Cellulose nanocrystals have been used as reinforcing agents for isotropic hydrogels.229–237 Hydrogels have been prepared from freeze-cast suspensions of CNCs and xylan, in which the xylan component modified with aldehyde groups formed hemiacetal bonds with the xylan.222 In these hydrogels, however the fast hydrolysis of hemiacetals may limit the stability of the resulting material.238

Here, we report the fabrication of anisotropic CNC-containing hydrogels that have two novel and advantageous features: (i) they are derived from precursor aerogels prepared via a single-step freeze-casting and crosslinking process, and (ii) they are cross-linked via more stable, slowly hydrolyzable bonds, which is critical in the context of tissue engineering. Importantly, we show the ability to generate precursor aerogels with fibrillar, columnar and lamellar morphologies, leading to direction-dependent swelling and mechanical properties of the resulting hydrogels.

We used aldehyde-functionalized CNCs and hydrazide-functionalized poly(oligo ethylene glycol methacrylate) (POEGMA) as aerogel components. Hydrazone bonds formed

90 between aldehyde and hydrazide groups are hydrolytically degradable, relatively slowly (over months) at neutral pH but significantly faster at acidic pH values.120 The microstructure of the aerogels and the resulting hydrogels was controlled by varying the weight ratio of CNC-to- POEGMA, the total concentration of these components in the precursor suspension, and the freeze-casting temperature. The well-established cytocompatibility of CNCs55,169,239,240 and POEGMA,121,241 the degradability of the hydrazone cross-linked networks, and the anisotropy of these hydrogels suggests their potential utility in biomedical applications.

5.1 Fabrication and microstructure of anisotropic aerogels

Atomic force microscopy was used to determine A-CNC dimensions to be 60-220 nm in length and 2-10 nm in diameter (Figure 5.1).

Figure 5.1 Atomic force microscopy height image of aldehyde-functionalized CNCs.

H-POEGMA and A-CNCs were mixed and immediately freeze-cast in either cylindrical or cuboidal molds (Scheme 1.1). The final hydrogels were held together by physical crosslinks via non-covalent, hydrophobic interactions between H-POEGMA and A-CNCs, in addition to the covalent hydrazone cross-linking.237 The aerogel structure and hydrogel mechanical properties were examined as a function of the weight ratio of A-CNC-to-H-POEGMA and the

91 total concentration of A-CNCs and H-POEGMA in the mixture used for freeze-casting (this concentration was denoted as CA-CNC+H-POEGMA).

Table 5.1 shows the recipes used for the preparation of the aerogels. Sample names are denoted as a:b-W@T, where a:b is the weight ratio of A-CNC-to-H-POEGMA, W is the CA-

CNC+H-POEGMA, and T is the temperature of freeze-casting. When no temperature T is specified in the sample notation, the freeze-casting process was conducted at -196 °C.

Scheme 1.1 Cross-linking reaction of aldehyde-modified CNCs (A-CNCs) with hydrazide- functionalized POEGMA (H-POEGMA) and illustration of morphologies of hydrogels achieved via the freeze-casting process.

Table 5.1 Recipes of freeze-cast aerogels and hydrogels and Young’s moduli of the rehydrated hydrogels

Weight ratio of Sample name* CA-CNC+H-POEGMA Freeze-cast A-CNC:H- (a:b-W@T) (wt%) temperature (°C) POEGMA

1:1-4 1:1 4.0 -196

1:3-4 1:3 4.0 -196

1:5-4 1:5 4.0 -196

1:5-2.5 1:5 2.5 -196

1:5-7 1:5 7.0 -196

1:5-4 @-80 1:5 4.0 -80

1:5-4 @-20 1:5 4.0 -20

* When no temperature T is specified in sample name, the freeze-casting process was conducted at -196 °C.

** Calculated from the linear portion of the first stress-strain cycle applied.

Figure 5.2a shows SEM images of the cross-section of the aerogels in the XY-plane perpendicular to the ice-growth direction. Three general trends are evident from this figure. First, at the low CA-CNC+H-POEGMA and high weight fraction of A-CNCs in the precursor suspension, the aerogels had a fibrillar structure with fiber width of ~0.5 μm (top left image in Figure 5.2a).

Second, the fibrillar structure transformed into a columnar structure when the CA-CNC+H-POEGMA was increased 4-7-fold at high A-CNC content (Figure 5.2a, top row, two right images). The pores had the width of ~4 μm and a wall thickness of ~100 nm. Third, at a high CA-CNC+H-POEGMA and a low weight fraction of A-CNCs, the aerogels had a lamellar morphology (Figure 5.2a, right

93 column, two bottom images). The lamellae were ~200 nm-thick and an inter-lamellar distance was ~7 μm. The fibrillar, columnar, and lamellar structures are labeled as "F", "C" and "L", respectively, in the SEM images in Figure 5.2a. In the intermediate range of sample compositions, the aerogel morphologies were less defined and showed a transition from fibrillar- like to lamellar-like structures.

The ability of the free-standing aerogels to maintain the shape of the mold, that is, their structural integrity upon lyophilization, was examined with respect to both CA-CNC+H-POEGMA and the A-CNC-to-H-POEGMA weight ratio (Figure 5.2b). Although all samples did not disintegrate upon lyophilization, the greatest structural stability was observed for the aerogels with the CA-

CNC+H-POEGMA of 2.5 and 7.0 wt%. Aerogels formed at lower CA-CNC+H-POEGMA did not contain a sufficient amount of material to preserve the aerogel shape. At high A-CNC fraction, the dimensional stability of aerogels after lyophilization improved, which was attributed to the mechanical strength of A-CNCs.

In addition to varying aerogel composition, the freeze-cast temperature and thus the rate of ice crystal growth in the samples was used to control aerogel morphology (Figure 5.3). The aerogel freeze-cast at -20 °C formed a more well-defined lamellar-like structure, in comparison with the same sample prepared at -80 °C. At lower freezing velocities achieved at smaller

94 temperature gradients, A-CNCs and H-POEGMA had sufficient time to be excluded from the growing ice crystals, thus yielding more organized structures.

5.2 Examination of the surface area of aerogels

The diverse morphologies of the aerogels were reflected by changes in their surface area (Figure

5.4a). At CA-CNC+H-POEGMA = 4.0 wt%, with increasing the A-CNC:H-POEGMA ratio the aerogel surface area increased from 18 (sample 1:1-4) to 33 m2/g (sample 1:5-4), concurrent with the observed change in microstructure from columnar to sheet-like, respectively (Figure 5.2a). In contrast, at a constant A-CNC:H-POEGMA ratio of 1:5, increasing the CA-CNC+H-POEGMA from 4.0 to 7.0 wt% resulted in the decrease of the surface area from 33 to 5 m2/g (Figure 5.4), which

95 correlated with both the higher density of the aerogels and the structural transition from sheet- like to lamellar morphology (Figure 5.2a). Increasing the CA-CNC+H-POEGMA from 2.5 to 4.0 wt% at 1:5 A-CNC:H-POEGMA ratio did not significantly change aerogels (Figure 5.4b). Here, the effect of higher density on the surface area of aerogels was compensated for by the change in their morphology. Overall, the surface area of the aerogels was in the rage from 5 to 33 m2/g, which is lower than that of aerogels formed CNCs only.242,243

Figure 5.4 Surface area of (a) aerogels at a constant CA-CNC+H-POEGMA and varying A-CNC:H-

POEGMA ratio and (b) aerogels at constant A-CNC-to-H-POEMGA ratio and varying CA-CNC+H-

POEGMA.

5.3 Small-angle X-ray scattering in aerogels

Small-angle X-ray scattering (SAXS) was used to supplement the characterization of the aerogels by SEM by providing insight into the volume-averaged aerogel structure. Figure 5.5 shows the SAXS data for the 1:5-4 aerogel (used as an example), as well as theoretical simulations of 2D SAXS patterns. Other 2D SAXS patterns for fibrillar, columnar and lamellar aerogels are given in Figure 5.6. The SAXS measurements were performed with the X-ray beam irradiating the aerogel in the Z- or XY-directions (corresponding to the right panels in Figure 5.5a and b, respectively). In the case of irradiation in the Z-direction (Figure 5.5a, right panel), the experimental 2D pattern was isotropic (circular), while irradiation in the XY-direction

96 yielded the anisotropic (ellipsoidal) 2D pattern (Figure 5.5b, right panel). These results suggested that scattering objects within the aerogel were preferentially aligned in the Z-direction but not within the XY-plane.

97 cylinders (red triangles) with a radius of 65 nm (± 10 %) and a length of 50 μm; and discs (black dots) with a radius of 5 μm and a diameter of 260 nm (± 10 %). The dotted line marks the lower limit of the experimentally reachable q range.

Figure 5.6 SEM images and corresponding 2D SAXS patterns of fibrillar (F), columnar (C), and lamellae (L) aerogels. The aerogels were irradiated in the Z- or XY-direction in SAXS experiments.

Theoretical 2D SAXS patterns were simulated to determine the length scale of the aligned scattering objects. Figure 5.5a and b (both left panels) show the simulated patterns for model discs preferentially aligned in the Z-direction and discs with random spatial orientation, respectively. The dimensions of these discs - a radius of 5 μm and thickness of 260 nm (± 10 % - were chosen to be comparable to those of mesostructured fibrils, columnar walls, and lamellae

98 shown in the SEM images (Figure 5.2a). Figure 5.5a (left panel) shows the simulated 2D SAXS pattern for the discs distributed isotopically. The pattern was similar to the experimental pattern (Figure 5.5a, right), which suggested that the mesostructures are the source of scattering and that they are not aligned in the XY-plane. Figure 5.5b (left panel) shows the simulated ellipsoidal 2D SAXS pattern for the discs preferentially aligned in Z-direction with a separation distance between the discs of 10 μm. The pattern was similar to the experimental pattern (Figure 5.5b, right), implying that the experimental scattering pattern likely originated from the scattering of mesostructures aligned in the Z-direction. In addition, the 1D radial-averaged plots for the 2D scattering patterns of 1:5-4 (Figure 5.5c) showed a q-4 scaling for q < 0.2 Å-1, which implied that the scattering data were in the Porod region244 and the scattering was dominated by objects with a size >100 nm. Thus, we conclude that the mesostructures were responsible for the main scattering contribution.

While the A-CNC components of the aerogel could be aligned by shear forces imposed by the growing ice front during uni-directional freezing, as previously suggested for freeze-cast CNC aerogels;245 it is difficult to prove such an alignment by low-angle SAXS measurements. This is because the scattering of small cylinder-like objects such as CNCs is concealed by the scattering from larger mesoscopic fibrils, columnar walls, and lamellae. To demonstrate this effect, we showed in Figure 5.5d the theoretical 1D radial-averaged scattering plots for small cylinders with diameters < 10 nm (similar to A-CNCs) and for large cylinders and discs with diameters > 100 nm (similar to mesostructures in SEM images in Figure 5.2a). These plots show that in the low q range, scattering contributions from the small cylinders are concealed by the contribution of large structures, if both are to be present in the same sample.

Although the SAXS data at low experimental q ranges could not provide insight into the alignment of the A-CNCs in the aerogel, the scattering patterns at high q values were sensitive to the crystalline cores of A-CNCs and could reveal their alignment. The 1D radial-averaged SAXS plots for 1:5-4 aerogel irradiated in the Z- and XY-directions (Figure 5.5c) both show a peak at 1.6 Å-1 and a broad shoulder at ~0.5 Å-1, attributable to the (200) and (110) reflections, respectively, from the cellulose crystal planes.242 If the A-CNCs were aligned in the Z-direction, well-defined spot- or arc-like Bragg reflections would be expected at these q values. The 2D SAXS patterns, for the 1:5-4 aerogel irradiated in the XY- and Z-directions, show Debye- Scherrer rings (Figure 5.7), as opposed to spot-like Bragg reflections or defined arcs, which

99 suggested that there is no preferential orientation of the A-CNCs in neither the Z-direction, nor in the XY-plane.

Figure 5.7 2D SAXS patterns of a 1:5-4 aerogel irradiated in the (a) Z- and (b) XY-directions.

To demonstrate that Debye-Scherrer rings exist at all azimuths, Figure 5.8a and b show the 2D SAXS patterns for the 1:1-4 aerogel irradiated in Z- and XY- directions, while Figure 5.8c shows the 2D SAXS pattern for the same aerogel as in Figure 5.8b, irradiated in the same direction, after the sample was rotated by 90° about the axis parallel to the direction of irradiation. Debye-Scherrer rings were observed in all cases, which suggested that there is no alignment of the A-CNCs within the aerogels.

100

5.4 Swelling behavior of anisotropic hydrogels

The key advantage of the material studied is that upon aerogel rehydration, chemical cross-linking between A-CNCs and H-POEGMA constituents prevents the disintegration of the aerogel, unlike with hemiacetal crosslinked CNC hydrogels which would readily dissolve in water.222 In addition, chemical cross-linking of the aerogel can preserve its anisotropy upon swelling.

Indeed, the aerogels examined in the present work did not disintegrate when exposed to water for the time interval up to 93 h, while Equilibrium swelling was achieved within 24 h (Figure 5.9). The swelling kinetics of the anisotropic hydrogel samples are shown in Figure 5.9. The equilibrium degree of swelling was reached for all the samples in <24 h, with a final measurement taken after 93 h.

101

Figure 5.9 Swelling kinetics for A-CNC-H-POEGMA samples prepared at (a) CA-CNC+H-POEGMA = 4 wt% and varying A-CNC:H-POEGMA ratios and (b) at the weight ratio A-CNC:H-POEGMA of 1:5 ratio and varying CA-CNC+H-POEGMA.

A coordinate system was used to characterize anisotropy in aerogel swelling (and mechanical properties of the hydrogels described in the following section). In Figure 5.9a, the Z- and XY-directions refer to the direction parallel to and perpendicular to the direction of ice growth, respectively. The anisotropic degrees of swelling of the aerogels are shown in Figure

5.10b and c. At CA-CNC+H-POEGMA= 4 wt%, a higher equilibrium swelling was achieved at a higher fraction of the hydrophilic H-POEGMA (Figure 5.10b), in agreement with earlier reports for isotropic A-CNC-H-POEGMA hydrogels.237

In addition, increasing CA-CNC+H-POEGMA led to stronger swelling, which was attributed to both the increased osmotic pressure governing swelling in more concentrated hydrogels and a larger amount of H-POEGMA in the sample. The structural anisotropy of the aerogel led to a higher degree of swelling in the XY-direction than in the Z-direction (Student’s t-test, p < 0.05 except for samples 1:3-4 and 1:5-7). Swelling in the Z-direction was mostly driven by the H- POEGMA component, whereas swelling in the XY-direction occurred also due to filling of the anisotropic pores with water and the resulting increase in pore size.

102

Figure 5.10 (a) Coordinate system used for swelling and compression tests. (b, c) Degree of swelling in the XY (QXY,eq) and Z (QZ,eq) directions of aerogels cast from suspensions at varying

CNC:POEGMA weight ratio and CA-CNC+H-POEGMA=4 wt% (b) and varying CA-CNC+H-POEGMA and CNC:POEGMA weight ratio of 1:5 (c). (d, e) Young’s moduli of hydrogels prepared at varying

5.5 Mechanical properties of anisotropic hydrogels

Mechanical properties of anisotropic hydrogels. To characterize the mechanical integrity and direction-dependent Young's moduli of the anisotropic hydrogels (important for potential tissue engineering applications246), compression tests were performed on swollen hydrogels in

103 the XY-direction (up to 50 % strain) and the Z-direction (up to 10 % strain). All hydrogels exhibited 100% shape recovery in both XY- and Z-directions after 50 compression- decompression cycles at a strain rate of 3 % of the original hydrogel dimension per second. Since the stain rate was low, hydrodynamic stresses caused by the flow of water out of the hydrogel during compression were assumed to be negligible. Representative stress-strain curves for sample 1:5-4 are shown in Figure 5.12a (Figure 5.11 shows the stress-strain curves for other hydrogels). The direction-specific Young’s moduli of the hydrogels are shown in Figure 5.10d and e. The Young’s modulus in the Z-direction (Ez) was higher than that in the XY-direction

(Exy) in each sample tested, consistent with the anisotropic hydrogel structures. Compression in the Z-direction led to buckling of the hydrogel fibrillar, columnar and lamellar mesostructures. In the XY-direction, at low strain, compression led to the collapse of the hydrogel pores, while at higher strain, the composite material within the hydrogel walls/fibrils was compressed. The anisotropy in hydrogel compression was less pronounced in fibrillar hydrogels. The CNC-to- POEGMA weight ratio did not have a significant effect on hydrogel weakening upon cyclic loading (Figure 5.11), however as the CA-CNC+H-POEGMA was increased from 2.5 to 7 wt%, hydrogels exhibited fatigue over 50 compression cycles, which was attributed to the irreversible deformation of the lamellar structures (delamination).

104

Figure 5.11 Stress strain curves for A-CNC-H-POEGMA hydrogels compressed parallel (Z) and perpendicular (XY) to the direction of ice growth (50 compression cycles).

The direction-dependent Young’s moduli of anisotropic hydrogels correlated with their compositions and underlying morphologies. As shown in Figure 5.12b and Figure 5.10d, at CA-

CNC+H-POEGMA = 4 wt%, at lower A-CNC:H-to-POEGMA weight ratios (a higher fraction of

POEGMA in the hydrogel), Exy changed weakly, whele Ez strongly increased. The latter trend was counter-intuitive, since the fraction of rigid CNCs in the system reduced. We attribute this effect to the need in a higher fraction of POEGMA in the hydrogel to achieve the highest degree of chemical cross-linking between A-CNCs and H-POEGMA.

In contrast, at the A-CNC-to-H-POEGMA weight ratio of 1:5, increasing the CA-CNC+H-

POEGMA from 2.5 to 7% led to an increase in both Exy and Ez (Figure 5.12d and Figure 5.10e). We ascribe the increase in Young’s moduli in both directions to the increase in the total density of

105 the hydrogel at a higher precursor concentration. In addition, a transition from a sheet-like structure (for 1:5-2.5 hydrogels) to a lamellar morphology (for 1:5-7 hydrogels) resulted in the increase in the Young’s moduli.

Similarly, decreasing the freeze-cast temperature yielded hydrogels with higher Exy and

Ez values (Figure 5.12d). As the freeze-cast temperature decreased, the ice front velocity increased,93 which resulted in a tighter packing of the the CNC/POEGMA mixture excluded from the ice growing front. These tightly packed structures resulted in stiffer hydrogels.

106

Thus we conclude that structurally and mechanically anisotropic precursor aerogels and resulting composite hydrogels with a high structural integrity can be formed by freeze-casting a dispersion of A-CNCs and H-POEGMA. By varying the precursor composition, and/or the freeze-cast temperature, the morphology and the direction-dependent moduli of the hydrogels can be tuned and controlled. Such control may be of particular interest for mimicking biological tissues that exhibit analogous directional mechanics due to their internal orientation.

5.6 Conclusions

Aerogels and hydrogels of A-CNCs and H-POEGMA have been produced in a single freeze-casting/cross-linking procedure. Chemical cross-linking between A-CNCs and H- POEGMA resulted in mechanically stable hydrogels, with A-CNC component contributing to improved dimensional stability, although without structural alignment. The lamellar, columnar, and fibrillar morphologies of the material were realized by varying the total amount of A-CNCs and H-POEGMA in the precursor dispersion and the weight ratio between these constituents. The composition and morphology of the material determined anisotropic swelling and mechanical properties of the composite hydrogels. The structural integrity of the hydrogels and the capability to control their direction-dependent swelling and mechanical properties suggest that these materials may function as effective biomimetic scaffolds for tissue engineering of oriented tissues.

Chapter 6 Reversible transition between anisotropic and isotropic thermal conductivity in elastic polyurethane foams

Contribution: Bernd Kopera conducted the thermal diffusivity experimental and performed some SEM imaging. Vanessa Machado contributed to some of the sample preparation, HRSEM and TEM imaging. XRD experiments were performed by Dr. Wolfgang Milius. DSC experiments were performed by Fabian Nutz. Graphitic dispersion stability experimental and characterization were performed by Vanessa Machado. XPS was performed by Rana Sodhi. Cyro-TEM imaging was performed by Dr. Markus Drechsler. The remaining experiments, data analysis, and data processing was performed by Mo Kit Chau.

Introduction

The ability to control the thermal conductivity, , of a foam is imperative for its specific thermal application, and has been the focus of intensive research.247 Much effort has been invested to extend the boundaries of what is currently possible for the highest  (such as in heat exchangers248) and lowest  (such as in foam insulators249). However, these materials are mostly isotropic, leaving the directional thermal properties unoptimized. The intended purpose of most thermal materials is to guide or prevent heat flow in a particular direction. For example, the intended purpose of building insulation is to prevent heat transfer in the direction perpendicular to the wall, even though traditional insulation materials are isotropic on the microscale. More advanced insulation systems, such as in dynamic insulation materials, could prevent heat loss and promote heat recovery by utilizing anisotropic open-cell foams that are thermally insulating in the direction perpendicular to the walls, yet thermally conductive in the parallel direction. To address this lack of directional thermal control, anisotropically structured foams with corresponding anisotropic thermal conductivities can be used to guide the flow of heat in a desired direction.

Most conventional polymeric foams prepared by the gas foaming process are inherently anisotropic due to the rise of gas bubbles;250 however, due to the high viscosity of the polymer precursor, the anisotropy generated in this manner is weak with low pore aspect ratios of ~1.3.251 Other methods to generate anisotropic foams include applying large pressure drops to gas-

107 108 saturated polymers252 and applying uniaxial strain while thermally annealing isotropic foams.253 However, these methods lead to limited success in creating highly anisotropic pores that percolate rectilinearly through a structure. Recently, highly anisotropic foams with pore aspect ratios of ~100 have been fabricated using directional freeze-casting.223 This technique involved freezing a dispersion in a uni-directional temperature gradient, where dispersants are excluded to the spaces between the growing solvent crystals.92,254,255 The frozen solvent was then sublimed leaving a free-standing, anisotropic structure, which was templated by the solvent crystals. Foams made by freeze-casting graphene oxide and cellulose nanocrystals exhibited a large ratio of thermal conductivities parallel and perpendicular to the direction of ice growth (~ 11) with a perpendicular thermal conductivity of 15 mWm-1K-1, which is significantly lower than the value of thermal conductivity of air (26 mWm-1K-1).256 Although these foams had excellent insulation properties, they lacked the elasticity, which is important in applications that demand resilience, impact absorbance, and durability.

In this chapter, we report the freeze-casting of polyurethane (PU) dispersions mixed with carbon black (CB) or carbon nanofibers, both stabilized by cellulose nanocrystals (CNCs), into elastomeric foams with highly anisotropic, lamellar structures, which led to concomitant direction-dependent thermal conductivity and Young’s modulus.

6.1 Synthesis and characterization of polyurethane

Fabricating anisotropic foams involved five steps: (i) synthesizing a PU polymer of isophorone diisocyanate (IPDI), dimethylolpropionic acid (DMPA), and polycaprolactone-block- polytetrahydrofuran-block-polycaproclatone (PCL-b-PTHF-b-PCL), chain-extended with ethylene diamine dispersed in water; (ii) mixing in the additives; (iii) freeze-casting the mixture; (iv) freeze-drying the resulting frozen mixture; and (v) thermal annealing the composite foam. The polyol of PCL-b-PTHF-b-PCL (1850 g/mol), DMPA, and IPDI were used to synthesize a PU prepolymer (Scheme 1). A block copolymer, rather than a homopolymer, was chosen as the polylol component to minimize the crystallization of the PU.257 The prepolymer was then chain extended with ethylene diamine.

109

Scheme 1. Synthesis of water-dispersible PU, where x = 5, y = 12, m = 6, n = 58, p = 62. The latter three values were determined by feed ratios.

Infrared (IR) spectra of the resulting polymer and PCL-b-PTHF-b-PCL oligomer are shown in Figure 6.1. The peak at 1731 cm-1, observed in both the PCL-b-PTHF-b-PCL oligomer and PU spectra, was attributed to the esters C=O stretching vibrations from the polycaprolactone component. In the IR spectrum of the PU polymer, a shoulder at 1650 cm-1 was related to the H- bonded C=O stretching vibrations of the polycaprolactone esters. The IR spectrum of the PCL-b- PTHF-b-PCL oligomer showed a broad peak at 3350-3600 cm-1, attributable to the O-H stretching vibration of the alcohol end groups. This peak was missing in the PU spectrum, which implied that the O-H end groups had reacted during the reaction of the polyol and DMPA with IDPI. A peak in the PU spectrum at 3200-2400 cm-1, not present in spectrum of the oligomer diol, was attributed to the N-H stretching vibration from urethane and urea groups. In addition, a peak at 1526 cm-1 in the PU spectrum was attributed to the C=O stretching vibration from the urethane and urea groups. Thus, the IR spectra suggested the successful formation of the desired PU product.

110

Figure 6.1 IR spectra of the PCL-b-PTHF-b-PCL oligomer and the PU polymer.

The gel-permeation chromatography (GPC) trace of the PU polymer is shown in Figure

6.2. Nominal number average molecular weight, Mn, was found to be 58, 600 g/mol with a polydispersity index of 1.8, based on the PMMA calibration.

Figure 6.2 GPC trace for the PU polymer in NMP eluent.

111

The PU polymer was dissolved acetone to form a 30 wt% solution and trimethylamine was added to the solution, such that the moles ratio of isocyanate to hydroxyl groups was 1.1. Water was then added under vigorous stirring to form a dispersion of PU in water. The acetone was subsequently removed by rotary evaporation. The resulting PU particles dispersed were imaged using cryogenic transmission electron microscopy (cryo-TEM) (Figure 6.3a). The particle size was determined to be 13.5  4 nm. The PU particles were negatively charged with an electrokinetic potential of -52  2 mV.

112

Figure 6.3 a) Cyro-TEM image of PU particles. The scale bar is 100 nm. b) TEM image of CNCs. The scale bar is 250 nm. c) HRSEM image of CB particles. The scale bar is 250 nm. d)

TEM image of CNFs. Scale bar is 1 m. e—h) Photographs of PUpure, PUCNC, PUCNC-CB, and

113

The length of the cube is 930 µm. The lamellae are oriented along the ice-growth direction, without preferential orientation within the perpendicular planes.

6.2 Characterization of CB and CNF additives

Additives, CB and CNFs, were used to vary the thermal and mechanical properties of PU foams. These additives were stabilized by CNCs as described below. Figure 6.3c shows the high- resolution scanning electron microscopy (HRSEM) image of CB particles, which were spheroidal with a diameter of 32  18 nm. Figure 6.3d shows the TEM image of CNFs, which had an average diameter of 150  51 nm, and the lengths of the CNFs were ~20—200 μm, the latter as specified by the manufacturer.258

Figure 6.4a shows the X-ray photoemission and Auger spectra for CB and CNFs. To determine the ratio of sp2-to-sp3 carbons in CB and CNFs, the D parameter for each material was calculated using a previously reported method,259 where the D parameter isx the difference in binding energies between the positive maximum and negative minimum of the first derivatives of the C KLL spectrum. The C KLL spectra, in general, are obtained by exciting and removing a core electron in the K shell of the carbon atom. A higher electron in L shell relaxes to fill the vacancy in the K shell, thereby resulting in the emission of another electron in the L shell. This latter electron is known as an Auger electron. The C KLL spectrum can be obtained by measuring the kinetic energies of these Auger electrons. The first derivatives of the C KLL spectrum for CB and CNFs are shown in Figure 6.4b. The D parameters of CB and CNFs were compared with the D parameters of diamond (14.2 eV, assumed to have only sp3 carbons) and graphite (22.5 eV, assumed to have only sp2 carbons). A linear approximation was used to interpolate sp2/sp3 carbon ratio for CB and CNFs, yielding approximately 71 % and 76 %, respectively.

114

To determine the amount and type of carbon and oxygen on the surface of each graphitic material, the C1s and O1s spectra for CB and CNFs are shown in Figure 1.1c—f. For the O1s spectra of CB, the high binding energy of the O1s peaks suggested that the O1s spectra were affected by differential charging. However, the peaks in the C1s spectra did not appear to be affected by differential charging, which implied that O1s contained contributions from oxygens

115 which were not bound to carbon, potentially from adsorbed moisture. Therefore, we did not further interpret the peaks from the O1s spectra. The C1s spectra for CB and CNFs were fitted to Lorentzian-Gaussian product functions using an Powell optimisation fitting algorithm and the resulting peaks and assignments are shown in Table 6.1 and

Peak label Peak BE Atomic % Assignment

C1s A 284.43 58 Carbon sp2

C1s B 285.31 24 Carbon sp3/C-O

C1s C 287.12 5 Possibly satellite structure

C1s D 289.26 3 π to π* transition

C1s E 290.19 10 π to π* transition

Table 6.2

, respectively.

Table 6.1 Peak fits and assignments from the C1s spectra of CB.

Peak label Peak BE Atomic % Assignment

C1s A 284.43 58 Carbon sp2

C1s B 285.31 24 Carbon sp3/C-O

C1s C 287.12 5 Possibly satellite structure

C1s D 289.26 3 π to π* transition

C1s E 290.19 10 π to π* transition

Table 6.2 Peak fits and assignments from the C1s spectra of CNF.

Peak label Peak BE Atomic % Assignment

116

C1s A 284.55 83 Carbon sp2

C1s B 285.35 3 Carbon sp3

C1s C 286.43 3 C-O

C1s D 291.28 11 π to π* transition

The CB had O1s peaks with binding energies (BEs) between 522.6 to 537.9 eV. By the comparing the area under the peaks in this region with the total area in the XPS spectrum, the contribution of oxygen to the total atomic profile was 6 %. The remaining 94 % was attributed to carbon from the C1s peaks that appeared between BE 284.5 to 291.27 eV. Carbon nanofibers had fewer oxygen containing functionalities with an atomic profile of 99 % from C1s and 1 % from O1s. It is important to note that these atomic percentages are approximations and the O1s peak may also be affected by the presence of moisture.

A high carbon content in CB and CNFs suggested that these materials were hydrophobic and that a stabilizer, such as CNCs, was required for their dispersion in water. Cellulose nanocrystals are amphiphilic, with hydrophobic (200) facets of CNCs that can interact with the graphitic surfaces.260 The hydrophobic facets of CNCs can interact with aromatic groups through van der Waals interactions as reported previously for the binding of cellulose to CNCs through their tryptophan residues.260 The hydrophobic facets of CNCs have been demonstrated to interact with hydrophobic carbon nanotubes.260,261 Cellulose nanocrystals also have hydrophilic (110) and (11̅0) facets, rich in hydroxyl groups, which facilitate the dispersal hydrophobic materials in water.261,262 The amphiphilic nature of CNCs have been utilized to stabilize Pickering emulsions of styrene in water.262

To assess the ability of CNCs to disperse CB and CNFs into an aqueous dispersion, graphitic materials in water with and without CNCs were prepared. The CNCs used in this work had a width of 8  2 nm and length of 110  28 nm (by TEM, Figure 6.3b) with a negative ζ- potential of -49  1 mV. The photographs in Figure 6.5a—d in show that the CB and CNFs in water without CNCs settled to the bottom of the container immediately. The TEM image (Figure

117

6.5e) of the corresponding samples shows that CB aggregated into clusters (~200 nm in size) without the addition of CNCs. As shown in Figure 6.5f The CNCs facilitated the dispersal of CB into a homogenous and stable dispersion. The TEM images of CB dispersions containing CNCs in Figure 6.5b showed that clusters of CB particles, which were likely stabilized by the CNCs. The appearance of these clusters is consistent with the literature reports which describe CB existing as 10—500 nm sized clusters of the indivisible units of quasi-spherical particles that are fused.263 Similarly, the TEM images of the CNFs without CNCs (Figure 6.5g) showed that the CNFs were highly entangled. The TEM image of the CNFs with CNCs (Figure 6.5h) showed that the CNFs with CNCs were more evenly distributed.

Figure 6.5 a—d) Photographs and e—h) TEM images of CB and CNF with and without CNCs. The final concentration of CB or CNF was 5 wt%. The final concentration of CNCs, if present, was 2.5 wt%. Scale bars for the TEM images are 500nm.

6.3 Fabrication and morphology of freeze-cast PU foams

The PU dispersion and their additives were freeze-cast into anisotropic PU and PU composite foams. The graphitic additives were first, dispersed in an aqueous suspension of CNCs. The composition of the dispersion before freeze-casting was 20 wt% PU, 1 wt% graphitic additives, and 0.5 wt % CNC. Then, the PU dispersion was added, followed by the addition of an

118

N-(3-dimethylaminopropyl)-N′-ethylcarbodiimide hydrochloride (EDC) solution. The concentration of EDC in the dispersion before freeze-casting was 0.26 wt%. The EDC reacted with the carboxylic groups on the PU and reduced the repulsive energy between the PU particles, thereby promoting coalescence of these particles during freeze-casting.264,265 The dispersion containing PU dispersion, EDC, and the relevant additives was introduced into a mold, which was placed on a cold finger set to -20 C. This temperature was chosen to minimize crystallization of the polyester and polyether components of the PU. After freeze-casting, the ice was sublimed to obtain free-standing PU foams, which were subsequently annealed at 90 C for 8 h in vacuum to promote crosslinking via hydrogen bonding between the hard segments of PU, and to reduce the tackiness of the foam surface.

Later in the text, foams of pure PU are denoted as “PUpure.” Polyurethane foams loaded with 2.5 wt% CNC are denoted as “PUCNC.” Polyurethane foams loaded with 2.5 wt% CNC, as well as 5 wt% CB or 5 wt% CNF are denoted as “PUCNC-CB” or “PUCNC-CNF”, respectively.

Photographs of PU foams are shown in Figure 6.3e—h. These foams were cut perpendicular and parallel to the ice-growth direction and imaged using SEM (Figure 6.3i—p). As shown in SEM images, all samples were composed of continuous, oriented lamellae with the long axis of the lamellae oriented parallel to the ice-growth direction. We term this axis as the “parallel” axis, throughout this work. The axis orthogonal to the “parallel” axis is termed “perpendicular.” As shown in the three-dimensional micro-computed tomography (microCT) image for PUpure (Figure 6.1q), these lamellae were curvilinear, without any preferential alignment in the plane perpendicular to the ice-growth direction. The average lamellar thickness was ~5 – 10 µm and was statistically invariant for all the samples (Table 6.3). Their inter- lamellar distances were on the order of ~ 17 µm.

119

Table 6.3 Thicknesses of the lamellae, inter-lamellar distances, and cp (at 25 ºC) for various PU foams.

Sample Lamellae Inter-lamellar 휿∥⁄휿⊥ 휿∥⁄휿⊥ 휿∥⁄휿⊥ cp

name thickness* distance* (m) air vacuum helium (J g-1 K-1)

(m)

PUpure 7  4 12  9 2.0 5.1 1.2 1.678

PUCNC 8  3 21 18 1.8 4.3 1.1 1.854

PUCNC-CB 8  3 18  12 1.4 2.2 1.0 1.938

PUCNC-CNF 6  2 17  11 2.1 2.7 1.6 1.865 *Determined from SEM using 100 measurements

Three types of inter-lamellar features were observed in the SEM images of the foams in Figure 6.3 and the corresponding low-magnification SEM images in Figure 6.6. First, short, rail- like protrusions on the side of the lamellae along the lamellae parallel to the direction of ice growth. As observed in Figure 6.3m and Figure 6.6b (marked with red lines), these protrusions were most pronounced for PUpure with a regular spacing of ~ 13 μm, which resulted from the formation of dendritic ice.266 These protrusions could form inter-lamellar bridges upon annealing. For PUCNC, PUCNC-CB, and PUCNC-CNF (Figure 6.3n—p and Figure 6.6 perpendicular views) the introduction of additives reduced the regularity of the spacing. Second, as shown in Figure 6.6 f and g (marked by red arrows), inter-lamellar strut-like bridges were observed for

PUCNC-CB, and PUCNC-CNF. This bridging may have resulted from the entrapment of particles within the growing ice crystals. Increasing the amount of additives increased the viscosity of the dispersion, which favored the entrapment of PU particles, CNCs, and graphitic materials during freeze-casting. The third type of inter-lamellar features is shown for PUCNC-CNF in Figure 6.7a (marked with arrows), and stemmed from bundles or individual CNFs that bridged the lamellae.

120

Figure 6.6 Low-magnification SEM images of various PU foams cut (a—d) parallel and (e—h) perpendicular direction of ice growth. Scale bars are 400 μm. The red lines emphasize regular spacing in the projections of PUpure. The red arrows mark the strut-like bridging features in

PUCNC-CB and PUCNC-CNF.

121

The PUCNC-CNF foams exhibited more structural variability compared to the other PU foams. Figure 6.7 shows SEM images of PUCNC-CNF with various microstructural features. Most of the CNFs were embedded in the PU lamellae, with the occasional inter-lamellar CNF bridges and strut-like PU bridges (Figure 6.7a, marked red arrows). In several regions, a high concentration of CNFs resulted in a closure of the pores along the ice-growth direction (Figure 6.7b, marked with a red circle). Regions of high CNF content (Figure 6.7c, marked with a red circle) and bundles of pure CNFs (Figure 6.7d, marked with a red circle) also bridged between the lamellae. These CNFs may have phase-separated during the freeze-cast process into the pores of the foams.

Figure 6.7 SEM images of the same PUCNC-CNF foam show various structural features including a) CNFs embedded within the lamellae, inter-lamellar CNFs bridges, strut-like PU bridges, b) blocked pores, c) pores bridged by a high concentration of CNFs, and d) isolated bundle of CNFs.

122

To assess the potential crystallization of the PU foams, differential scanning calorimetry (DSC, Figure 6.8) and X-ray diffraction (XRD, Figure 6.10) measurements of the foams were performed.

Differential scanning calorimetry measurements for the PU foams were performed before and after annealing at 90°C for 8 h (Figure 6.8). In order to assess the thermal transitions of the PU foams, the first heating cycles of each sample are shown. All DSC traces featured a monotonic increase in heat flow with increasing temperature, which was attributed to the increase in specific heat capacity, cp, of the foam with increasing temperature. The differences in absolute heat flow between annealed and non-annealed samples were attributed to a lack of contact between the polymer sample and the DSC aluminum pan during the first heating cycle.

To test this effect, Figure 6.9 shows the first and the second heating cycles for PUpure. The absolute heat flow was drastically different between the two cycles. For all annealed samples, a small and very broad endothermic contribution between 60 to 90 °C could caused by the breakage of hydrogen bonds between the hard segments of PU (urethane and urea functionalities). These hydrogen bonds may have formed during the thermal annealing step.267 Nevertheless, such hydrogen-bonded regions were very broadly distributed and constituted only a small fraction of the overall sample. Overall, these DSC measurements indicated an amorphous polymer structure, since no prominent peaks for crystallization was observed.268,269

123

-1 Figure 6.8 DSC curves measured at a heating rate of 5 K min for a) PUpure, b) PUCNC, c) PUCNC-

CB, and d) PUCNC-CNF before (red curves) and after (blue cruves) annealing at 90 °C for 8 h.

124

Figure 6.9 First (red curve) and second (blue curve) heating cycle for PUpure at a heating rate of 5 K min-1.

X-ray diffraction experiments were performed on the PU foams to characterize their microstructure. The X-ray diffractograms for PU foams, CNCs, CNFs are shown in Figure 6.10. For all the foams, X-ray diffractograms showed only a broad peak with a maximum at 2θ ≈ 19 º, which could be attributed to the amorphous nature of the hard segment domains (containing the hydrogen bonded urea and urethane functionalities).268 No sharp peaks associated with the soft- segment crystallization were observed.17,270

No diffraction peaks from the CNCs were observed in the X-ray diffractograms of the composite foams, likely because the concentration of CNCs (2.5 wt%) was too low. In addition, the crystalline CNC components did not induce the formation of ordered domains in the PU 271 matrix. The addition of 5 wt% CB did not affect the XRD pattern of PUCNC-CB when compared to that of PUCNC. Though CB may contain varying amounts of graphitic quasi-crystalline 272 domains depending on the CB source, XRD peaks were not observed for PUCNC-CB possibly because the concentration of CB in the foam was no sufficient. Carbon nanofibers are cylindrical nanostructures composed of graphene layers arranged in stacked cones.273 The X-ray diffractogram of CNFs showed a sharp Bragg peak at 27 º, which corresponded to the d002 ~ 3.28 274 Å of graphene layers in the CNFs. The X-ray diffractogram of PUCNC-CNF also showed a sharp

125

Bragg diffraction peak at 27 º, which likely arose from the CNFs.275 The broad diffraction peak at ~45 º was attributed to the presence of amorphous hard segment domains containing hydrogen bonded polyurea segments.271 Overall, the XRD measurements, agreed with the conclusions from DSC measurements, in that the PU in the compressible foams had an amorphous structure.

Figure 6.10 X-ray diffractograms for PUpure, PUCNC, PUCNC-CB, PUCNC-CNF, CNC, and CNF.

6.4 Mechanical properties of PU foams

The mechanical properties of the PU foams were determined by compression testing.

Figure 6.11a—d shows the stress-strain curves of PUpure, PUCNC, PUCNC-CB, and PUCNC-CNF foams compressed up to 50 % strain in the parallel and perpendicular directions. When compressed in the parallel direction, the samples displayed a classical deformation behavior, characteristic of elastomeric cellular foams.251,276 Below ~10 % strain, increasing the strain resulted in a linear increase in stress, attributable to the bending of the lamellae oriented parallel to the loading direction. The Young’s modulus for each foam compressed in the parallel direction, Epara, (summarized in Figure 6.11e) was determined from the slope of this linear elastic region, between 0 and 5 % strain. Increasing the strain beyond the linear elastic region led to a plateau, where stress was largely independent of strain. This plateau was associated with the collapse of the pores by elastic buckling of the lamellae. Further increase in strain beyond the plateau region

126 yielded a steep increase in stress as a result of densification, in which the lamellae were forced into contact and further buckling was not possible.

Figure 6.11 Representative stress-strain curves for samples a) PUpure, b) PUCNC, c) PUCNC-CB, and d) PUCNC-CNF compressed up to 50 % strain in the parallel (blue) and perpendicular (red) directions, after pre-compression to 20 % strain. e) The value of E of each sample was calculated using the linear elastic region of each compression curve. f) The recovery after 50 % strain for each sample compressed in the parallel and perpendicular directions.

Compression of the foams in the perpendicular direction likely involved bending of the lamellae without their buckling. The values of Young’s modulus for the foams compressed in the

127

perpendicular direction, Eperp, are shown in Figure 6.11e, and were also determined from the slope of the stress-strain curve in the linear regime up to 5 % strain.

For all the samples, the values of Epara were greater than the values of Eperp because buckling upon compressing the edges of PU lamellae required more force than compressing the lamellae towards each other. For PUpure, Epara = 1.9  0.4 MPa, while Eperp was considerably lower (0.302  0.002 MPa). As illustrated in the Ashby plot (Figure 6.13i), the Young’s moduli of these PU foams were comparable to those reported for flexible foams, elastomeric foams, and

277 rubber. The addition of CNCs decreased values of Epara for PUCNC foams to 1.3  0.1 MPa. This effect was unexpected since CNCs have a Young’s modulus of 105 GPa278 and have been used as reinforcement agents in PU films.279 We ascribed this effect to the CNCs potentially disrupting the formation of hydrogen bonds in the PU, thus weakening the material. The addition of CB and CNFs increased values of Epara to 2.9  0.4 MPa for PUCNC-CB and to 3.3  0.5 MPa for PUCNC-CNF. The values of Eperp also increased from 0.41  0.13 MPa in PUCNC to 0.81  0.18 and 0.65  0.13 MPa for PUCNC-CB and PUCNC-CNF, respectively. This increase may be attributed to the reinforcement of the PU foams with the stiff graphitic materials.280

An advantageous feature of these anisotropic foams was their elasticity. To demonstrate recoverability of the PU foams after compression, the foams were subjected to 20 compression- decompression cycles up to 20% strain, which was well beyond the onset of buckling in the lamellae for compression in the parallel direction (Figure 6.12). The mechanical response was highly reproducible: the compression set of the foams was ~8 %, with irreversible losses mostly over the first several cycles. (Compression set is defined as the percentage of original specimen thicken after the load has been removed). In another experiment, foams subjected to 50% compressive strain demonstrated high recoverability with up to 95 % in the parallel direction and 71—90 % in the perpendicular direction (Figure 6.11f). We attribute the lower recoverability in the perpendicular direction to adhesion between the lamellae via hydrogen bonding. To the best of our knowledge, this is the first report of anisotropic, elastomeric foams made by freeze-casting PU dispersions.

128

Figure 6.12 Stress-strain curves from the compression-decompression cycles on PUpure, PUCNC,

PUCNC-CB, and PUCNC-CNF. The scales of the compressive stress for the perpendicular cases are significantly lower than those in the parallel plots.

129

6.5 Thermal properties of PU foams

The thermal diffusivity,  of the foams in the parallel and perpendicular directions was determined by xenon flash analysis (XFA). The value of was then used to calculate the thermal conductivity, , as

휅 = 훼 ∙ 푐푝 ∙ 휌 Equation 6.1,

where the specific heat capacity, cp, of the PU matrix was obtained by DSC, and the density of the foam,  , was determined by dividing the weight of the foam by the volume of the foam.

The values of κ of PU foams in vacuum, air, and helium are shown in Figure 6.13a, b, and c, respectively. In vacuum, thermal conductivity in the parallel direction, κpara, for PUpure was -1 -1 - 59 ± 3 mW m K , while value of κ in the perpendicular direction, κperp, was only 11 ± 1 mW m 1 -1 K . The value of κ of PUpure and PUCNC were similar, in their respective directions, within experimental error, which suggested that a small amount of CNCs (2.5 wt% loading) was not sufficient to affect the value of κ of PU foams. The incorporation of 5 wt% of CNCs also did not significantly the affect value of κ for PUCNC-CB in either direction likely because the CB was present at sub-percolating concentrations (below the percolation threshold). In contrast, PUCNC-

CNF samples had a significantly higher thermally conductivity than PUpure, PUCNC, PUCNC-CB with -1 -1 -1 -1 κpara = 158 ± 11 mW m K , and κperp = 59 ± 5 mW m K . The high aspect ratio of CNFs enabled the formation of a percolating network of CNFs in the aligned lamellae. Higher values of κ’s were achieved via thermal transport along the graphitic CNF backbone.

130

131

Figure 6.13 (a—c) Thermal conductivity of PU foams in the parallel (blue) and perpendicular (red) directions measured in a) vacuum (pressure < 1 mbar), b) air (ambient pressure, 990 mbar), and c) helium (pressure = 1000 mbar). (d, f) Thermal conductivity of PUpure in the parallel (blue) and perpendicular (red) directions upon d) cycling between vacuum (pressure < 1 mbar) and helium (pressure = 1000 mbar) and f) cycling between air and helium at a constant pressure of

The structural anisotropy of the lamellae in the foams resulted in an orientation- dependent thermal conductivity. The thermal anisotropy was most pronounced for PUpure and

PUCNC in vacuum with κpara/κperp of 5.1 and 4.3, respectively. This anisotropy was attributed to the presence of very long pores, as inferred from the SEM images and the microCT reconstruction (Figure 6.3). For PUCNC-CB and PUCNC-CNF, the thermal anisotropy was less pronounced with κpara/κperp ratios of 2.2 and 2.7, respectively. The introduction of graphitic materials increased inter-lamellar bridging, which reduced the aspect ratio of the pores, thus allowing more inter-lamellar thermal transport. The thermal anisotropy was also evident from the

IR thermograms (taken under ambient conditions) of PUCNC-CNF sliced in the parallel and perpendicular directions (Figure 6.13d). Irradiating foam samples cut in the perpendicular and parallel directions gave circular (isotropic) and ellipsoidal (anisotropic) temperature distributions, which suggested that thermal transport was more efficient along the lamellae. (The ellipsoidal heat dissipation was a result of anisotropic thermal diffusivity in a material, in which heat is diffused faster in one direction than another).

The presence of air (at 1000 mbar) in the PU foams allowed additional thermal transport through the gas phase, resulting in a higher value of  in air than in vacuum (Figure 6.13c). The thermal anisotropy in air was less pronounced than in vacuum. For PUpure and PUCNC, the ratio of

κpara/κperp decreased from ~ 5 to 2 from vacuum to air, respectively (see Table 6.3).

132

The Ashby plot in Figure 6.13 shows that the κpara values were comparable to the values of κ of bulk thermoplastic elastomers (e.g. PU elastomers have values of κ from 190 to 250 mW m-1 K-1 281), which typically have a significantly higher density (~1000 kg m-3), compared to the -3 -3 277 density of the PU-based foams under consideration here (210 kg m to 300 kgm ). The κperp values were comparable to values of κ of isotropic flexible foams (e.g. open cell, isotropic PU -1 -1 282 foams have values of κ ~30 mW m K ). Overall, the Ashby plot showed that these PU foams have Young’s moduli similar to those of elastomeric, isotropic PU foams have a Young’s modulus of ~0.08—0.93 MPa283, and rubbers have a Young’s modulus of ~ 1—2 MPa,284 while their κpara and κperp were comparable to both bulk and foam materials, respectively.

A unique and outstanding feature of the PU foam described in this work is their ability to camouflage their thermal anisotropy derived from their structural anisotropy, when an appropriate atmosphere is applied. We demonstrate two different methods for changing the thermal anisotropy of the foam by changing the value of value of  of the gas component: (i) by varying pressure and (ii) varying gas composition.

In the first method, the pressure was reduced to decrease the value of  of the gas phase via the Knudsen effect. We note that for this effect to occur, the pore sizes must be small enough to reduce the mean free path of the confined gas molecules at a particular pressure range. The pressure of the gas can be lowered to reduce the value of  of the gas, until a desirable difference in  between the solid and the gas phase is achieved, thereby controlling the degree of anisotropic thermal conductivity. For instance, Figure 6.13e shows the decrease in the values of

κpara and κperp for PUpure with decreasing helium pressure. The thermal conductivity progressed from a thermally isotropic state to a thermally anisotropic state.

Figure 6.14a—d shows values of κpara and κperp for various PU foams, when the surrounding atmospheric conditions were cycled between helium (1000 mbar) and vacuum (0.6 mbar). For PUpure, PUCNC, and PUCNC-CB, the difference between the values of κpara and κperp vanished in the presence of a helium atmosphere at a pressure of 1000 mbar, resulting in a thermally isotropic foam (Figure 6.14a—c). The loss of thermal anisotropy was attributed to the small difference between the values of κ of helium (156 mW m-1 K-1 at 300 K)256 and PU (~200 mW m-1 K-1).285 The relatively close matching between the value of κ of the gas phase and the solid phase allowed structurally anisotropic foams to behave as single-phase materials with

133 isotropic values of κ (assuming a low thermal interface resistance between the solid and the gas phases).

The reproducibility of the switching was excellent, due to the open-cell structure and elastomeric properties of the foams, as discussed above. The transition between the vacuum and helium atmospheres was very fast, though the time resolution of the xenon flash analysis did not allow for a detailed assessment of any changes faster than one minute.

For PUCNC-CNF shown in Figure 6.14d, the mismatch in κ values between the solid and helium (at 1000 mbar) led to the retention of the thermal anisotropy regardless of pressure. The presence of highly conductive CNFs created a mismatch between the values of κ of the lamellae and the values of κ of the helium, such that thermal anisotropy of the foam was maintained at all the pressures.

Figure 6.14 Thermal switching for various PU foams between helium (pressure = 1000 mbar) and vacuum (pressure < 1 mbar) atmospheres.

The second method for altering the gas  is by changing the composition of the gas phase. For example, by decreasing the fraction of helium in a helium/air mixture from 100 % to 0 %, value of  of the gas phase was lowered monotonically (helium has a higher value of  than air). For PUpure in an air/helium mixture (Figure 6.13h), the degree of anisotropy increased when the fraction of helium in the mixture was decreased. Figure 6.13g shows PUpure switching between isotropic and anisotropic modes of thermal conductivity when the composition of the

134 gas phase cycled between helium and air, while the total pressure maintained at 980 mbar. This effect is possible because helium at this pressure has a similar thermal conductivity to PU, while air has a much lower thermal conductivity. To the best of our knowledge, this report is the first to demonstrate materials that can transition between isotropic and anisotropic modes of thermal transport by triggering a change in the surrounding atmosphere.

Based on these findings, we derive two design principles for achieving a unique thermal switching behavior. First, the structural skeleton (solid phase) of the porous material must provide an anisotropic pathway distinct from that of the gas phase for thermal energy transport. In the present work, this pathway was the percolating network of PU created by the directional growth of the ice crystals during the freeze-cast process. The second principle is that the values of κ of the solid and the gas phases must match as closely as possible under a particular condition (e.g. pressure, gas composition, and temperature) but be mismatched at another. Here, value of κ of the solid phase in PUpure, PUCNC, and PUCNC-CB matched the values of κ of helium at 1000 mbar, such that the thermal anisotropy of the solid phase was camouflaged, resulting in isotropic thermal transport. To obtain the anisotropic mode, a thermal mismatch between the values of  of the gas and the solid was introduced either by changing the values of  of the gas, or the solid component; the former being the easiest to achieve reversibly.

It is important to note that even though we demonstrate thermal switching abilities for specific PU foams, which are anisotropic porous materials, the underlying design principles are generally applicable to any continuous biphasic material, in which the values of  of one phase can conveniently and reversibly altered.

6.6 Conclusion

In summary, we synthesized PU polymer with by reacting IPDI, PTHF-b-PCL-b-PTHF, DMPA, and ethylene diamine. The carboxylic groups on the PU polymer was deprotonated with triethylamine and the resulting polymer was dispersed into water. The aqueous PU dispersion was combined with CNC, CB, and CNF additives and freeze-cast to fabricate PU foams with long, oriented, interconnected lamellae and highly anisotropic pores, which were templated by the directional ice growth. This anisotropic structure manifested itself in anisotropic mechanical and thermal properties. Incorporating graphitic additives, CB and CNFs, increased the value of

135

Young’s modulus in both direction. The inclusion of CB did not significantly affect the value of κ, whereas the incorporation of CNFs doubled the values of thermal conductivity in all directions. The PU foams described in this work can be reversibly switch between anisotropic and isotropic modes of heat conduction by either changing the pressure via the Knudsen effect, or by varying the composition of the gas. To our knowledge, this is the first system in which the values of κ of a porous material was reversibly switched between anisotropic and isotropic states. Such materials may find application as dynamic insulation materials,286 in which both the amplitude and direction of heat conduction need to be simultaneously controlled. We hope that this work initiates a new field of research into the generation of cellular materials, in which functionality is derived from the structural anisotropy.

136

Chapter 7 Conclusion, Summary, and Future Works

Conclusion

Three-dimensional (3D) scaffolds that recapitulate the mammalian extracellular matrix (ECM) have potential applications as tissue engineering scaffolds and as in vitro cell culture. This thesis describes the fabrication and characterization of microgels, aerogels, hydrogels, and foams biomimic the mechanical properties and structures of the natural ECM.

7.1 Microfluidically generated biocomposite microgels. Summary and future works

In Chapter 3, we explored a microfluidic platform for generating biopolymer composite microgels, for which the composition, rigidity, and structure could be tuned in high-throughput manner. The microgels were generated by combining streams of agarose and gelatin solutions and mixing these solutions in the resulting droplets. The gelatin was chemically modified to incorporate additional phenolic groups, which were crosslinked enzymatically in the presence hydrogen peroxide. The thermoresponsive agarose component, underwent gelation at reduced temperature. By increasing the ratio of the gelatin-to-agarose solution flow rates, the morphology underwent a transition from globular to fibrillar, while the apparent Young’s modulus of the microgel increased.

In the future, this type microgel modules could be combined with 3D printing to generate tissues with controllable heterogeneity on-demand. To entertain this idea, the mechanical property and composition of in vivo tissue could first be mapped in 3D using ultrasound imaging and then modelled using computer-aided design software. The flow rates of precursor streams containing gelling solutions could be introduced into a microfluidic device (as the one described in Chapter 3). The hydrogels modules exiting the microfluidic device would be precisely positioned into a collector via a nozzle whose location is controlled by a computer system. To mimic the heterogeneity of tissue, the composition of the individual modules can be tuned by varying the flow rate, and the spatial and mechanical properties of the resulting tissue could be defined from the 3D mapping of in vivo tissues.

137

7.2 Nanofibrillar hydrogels. Summary and future works

Chapter 4 described the formation of nanofibrillar hydrogels by the addition of inorganic salts to cellulose nanocrystal (CNC) suspensions. The effect of cation size, charge, and concentration on the rheological and structural properties were investigated. The increase in the cations charge number or ionic radii both suppressed the Debye length of the CNCs, thus resulting in network structures with a thicker filaments that make up the meshes, higher stiffness, and larger pores. The trend in cation charge number can be explained by Derjaguin−Landau−Verwey−Overbeek (DLVO) theory, while the effect of the cation ionic radii is rationalized by the hard-soft acid-base (HSAB) theory. Furthermore, cations with larger ionic radii can bridge between CNCs by interacting with the sulfate groups present on the surface of CNCs, thereby increasing gel stiffness. Interestingly, the mesh size increased simultaneously with the stiffness. This phenomenon is not typically observed for molecular hydrogels and ascribed to the hierarchical assembly of the anisotropic building blocks.

For further exploitation of ionic crosslinks for forming nanofibrillar hydrogels of CNCs, polycationic copolymers could be used as the crosslinking agent. For example, an A-B-A triblock copolymer could be used, where A is a water-soluble, polycationic block and B is a neutral, hydrophilic block. Polycations, such as poly(2-(trimethylamino)ethyl methacrylate), could bind and bridge CNCs though electrostatic interactions. The presence of multiple associative interactions per block should result in stronger crosslinking. The purpose of the hydrophilic B block, such as poly[oligo(ethylene glycol) methyl ether methacrylate] (POEGMA), would be to interact with the solvent, resulting a swollen network rather than a floc. Such ionic crosslinking can be also be explored for dendrimers and hyperbranched polymers with a hydrophilic core and peripheries decorated with cationic functional groups.

7.3 Anisotropic hydrogels. Summary and future works

Chapter 5 described the fabrication of anisotropic composite aerogels and their resulting hydrogels made by freeze-casting aldehyde-functionalized CNCs together with hydrazide- functionalized POEGMA. The hydrogels were held together by covalent hydrazone crosslinks, as well as by physical adsorption between the CNCs and POEGMA. The morphology of the

138 aerogels could be tuned from fibrillar to columnar and lamellar by varying the total concentration of precursor in the freeze-cast dispersion and the weight ratio between the CNC and POEMGA. The hydrogels exhibited shape-anisotropic Young’s moduli and swelling behavior.

A future research direction for these anisotropic hydrogels could be to freeze-cast poly(N- isopropylacrylamide) (PNIPAM) together with gold nanoparticles to make anisotropic hydrogel actuators. PNIPAM is a thermoresponsive polymer with a lower-critical solution temperature (LCST) at 32 ºC.287 Gold nanoparticles are able to convert light into heat up to 70—80 ºC depending on the power of the laser used and the absorbance of the hydrogels containing the nanoparticles.288,289 Freeze-casting would be an excellent method to create hydrogels of PNIPAM, in which the features, likely lamellar, are thin enough, such that the PNIPAM would have a fast-response time to the heat released by the gold nanoparticles upon the irradiation of light. These hydrogels could be used potentially as actuators that mimic moving tissues, particularly upon the exposure to strobe lights. In addition, the gold-thiol chemistry could be exploited to tether biologically active molecules relevant for 3D cell cultures applications.

7.4 Anisotropic polyurethane foams. Summary and future works

Chapter 6 described the structural, mechanical, and thermal properties of anisotropic polyurethane foams made by freeze-casting. The foams had a lamellar structure with the lamellae oriented parallel to the direction of ice growth. Carbon black and carbon nanofibers were incorporated to tune the Young’s modulus and thermal conductivity of the resulting foams. The carbon black and carbon nanofibers increased the Young’s modulus of the foams in both direction. Carbon black did not have a significant effect on thermal conductivity of the foams while carbon nanofibers doubled the thermal conductivity in all directions. By exploiting the Knudsen effect, the thermal conductivity of the foams could reversibly toggle between anisotropic and isotropic modes by either changing the gas pressure, or by changing the composition of the gas at a constant pressure.

Though the method of switching conductivity by changing gaseous environment may be interesting, it may be difficult to implement it in real applications, since the system will need to be sealed and gases would have to be supplied and the gas pressure would have to be regulated. Instead, a future direction could be to fill the anisotropic pores of the anisotropic foams with a phase change material, which would change from a solid to liquid state upon heating or cooling

139 above or below a critical temperature. Phase change materials themselves have been used as temperature regulating material in buildings.290 The critical phase change temperature is usually adjusted to 25 ºC, such that if the temperature is above 25 ºC, the PCM decreases the temperature of the room by absorbing the heat and changing states from a solid to a liquid. When the temperature is below 25 ºC, the PCM increases the room’s temperature by releasing the energy and freezing into a solid. The thermal conductivity of the PCM either in the solid or liquid state could be chosen to match the thermal conductivity of the solid skeleton of the foam. In these cases, temperature, rather than gas composition, could be used to switch between anisotropic and isotropic modes of thermal conductivity.

In addition, anisotropic PU foams containing a percolating network of electrically conductive fillers, such as carbon nanofibers, carbon nanotubes, and conjugated polymers, could be used as cardiac patches. The lamellar structure, elastic properties, and electric conductivity would make these foams ideal scaffolds for heart tissue engineering.

140

References

(1) Alberts, B. Molecular biology of the cell; Garland Science: New york, 2008.

(2) Frantz, C.; Stewart, K. M.; Weaver, V. M. J. Cell Sci. 2010, 123 (Pt 24), 4195–4200.

(3) Yanagishita, M. Pathol. Int. 1993, 43 (6), 283–293.

(4) Wade, R. J.; Burdick, J. A. Mater. Today 2012, 15 (10), 454–459.

(5) Atala, A.; Lanza, R.; Thomson, J. A.; Nerem, R. Principles of Regenerative Medicine; Elsevier Science, 2010.

(6) LeGrice, I. J.; Smaill, B. H.; Chai, L. Z.; Edgar, S. G.; Gavin, J. B.; Hunter, P. J. Am J Physiol Hear. Circ Physiol 1995, 269 (2), H571–H582.

(7) Saffitz, J. E.; Kanter, H. L.; Green, K. G.; Tolley, T. K.; Beyer, E. C. Circ. Res. 1994, 74 (6 ), 1065–1070.

(8) Hayes, S.; Lewis, P.; Islam, M. M.; Doutch, J.; Sorensen, T.; White, T.; Griffith, M.; Meek, K. M. Acta Biomater. 2015, 25, 121–130.

(9) Freegard, T. J. Eye (Lond). 1997, 11 ( Pt 4) (4), 465–471.

(10) James, R.; Kesturu, G.; Balian, G.; Chhabra, A. B. J. Hand Surg. Am. 2008, 33 (1), 102– 112.

(11) Nishimura, T.; Hattori, A.; Takahashi, K. Cells Tissues Organs 1994, 151 (4), 250–257.

(12) Flynn, L. E. Decellularized Adipose Tissue. US20110151011 A1, 2011.

(13) Ruoslahti, E. Annu. Rev. Cell Dev. Biol. 1996, 12, 697–715.

(14) Khademhosseini, A.; Vacanti, J. P.; Langer, R. Sci. Am. 2009, 300, 64–71.

(15) Xu, X.; Farach-Carson, M. C.; Jia, X. Biotechnol. Adv. 2014, 32 (7), 1256–1268.

(16) Xu, X.; Sabanayagam, C. R.; Harrington, D. A.; Farach-Carson, M. C.; Jia, X. Biomaterials 2014, 35 (10), 3319–3330.

(17) Gupta, K. K.; Kundan, A.; Mishra, P. K.; Srivastava, P.; Mohanty, S.; Singh, N. K.; Mishra, A.; Maiti, P. Phys. Chem. Chem. Phys. 2012, 14 (37), 12844–12853.

(18) García, J. R.; García, A. J. Nat. Mater. 2014, 13 (6), 539–540.

(19) Engler, A. J.; Sen, S.; Sweeney, H. L.; Discher, D. E. Cell 2006, 126 (4), 677–689.

(20) Shoulders, M. D.; Raines, R. T. Annu. Rev. Biochem. 2009, 78, 929–958.

(21) Kadler, K. E.; Holmes, D. F.; Trotter, J. A.; Chapman, J. A. Biochem. J. 1996, 316, 1–11.

141

(22) Kuznetsova, N.; Rau, D. C.; Parsegian, V. A.; Leikin, S. Biophys. J. 1997, 72 (1), 353– 362.

(23) Vader, D.; Kabla, A.; Weitz, D.; Mahadevan, L. PLoS One 2009, 4 (6), e5902.

(24) Shayegan, M.; Forde, N. R. PLoS One 2013, 8 (8), e70590.

(25) Wallace, D. Adv. Drug Deliv. Rev. 2003, 55 (12), 1631–1649.

(26) Roeder, B. a.; Kokini, K.; Sturgis, J. E.; Robinson, J. P.; Voytik-Harbin, S. L. J. Biomech. Eng. 2002, 124 (2), 214.

(27) Diamant, J.; Keller, A.; Baer, E.; Litt, M.; Arridge, R. G. C. Proc. R. Soc. London. Ser. B, Biol. Sci. 1972, 180 (1060), 293–315.

(28) Fratzl, P.; Misof, K.; Zizak, I.; Rapp, G.; Amenitsch, H.; Bernstorff, S. J. Struct. Biol. 1998, 122 (1-2), 119–122.

(29) Folkhard, W.; Mosler, E.; Geercken, W.; Knörzer, E.; Nemetschek-Gansler, H.; Nemetschek, T.; Koch, M. H. J. Int. J. Biol. Macromol. 1987, 9 (3), 169–175.

(30) Narayanan, J.; Xiong, J.-Y.; Liu, X.-Y. J. Phys. Conf. Ser. 2006, 28 (1), 83–86.

(31) Jung, J. P.; Gasiorowski, J. Z.; Collier, J. H. Pept. Sci. 2010, 94 (1), 49–59.

(32) Palmer, L. C.; Leung, C.-Y.; Kewalramani, S.; Kumthekar, R.; Newcomb, C. J.; Olvera de la Cruz, M.; Bedzyk, M. J.; Stupp, S. I. J. Am. Chem. Soc. 2014, 136 (41), 14377–14380.

(33) Chau, M.; Sriskandha, S. E.; Thérien-Aubin, H.; Kumacheva, E. In Supramolecular Polymer Networks and Gels; Springer International Publishing, 2015; pp 167–208.

(34) Thomas, S.; Durand, D.; Chassenieux, C.; Jyotishkumar, P. Handbook of Biopolymer- Based Materials: From Blends and Composites to Gels and Complex Networks, Wiley.; New York, 2013.

(35) Smidsrod, O. Faraday Discuss. Chem. Soc. 1974, 57, 263–274.

(36) Morris, E. R.; Rees, D. A.; Thom, D.; Welsh, E. J. J. Supramol. Struct. 1977, 6 (2), 259– 274.

(37) Raghavan, S. R.; Douglas, J. F. Soft Matter 2012, 8 (33), 8539.

(38) Pääkkö, M.; Ankerfors, M.; Kosonen, H.; Nykänen, a; Ahola, S.; Osterberg, M.; Ruokolainen, J.; Laine, J.; Larsson, P. T.; Ikkala, O.; Lindström, T. Biomacromolecules 2007, 8 (6), 1934–1941.

(39) Warren, N. J.; Armes, S. P. J. Am. Chem. Soc. 2014, 136 (29), 10174–10185.

(40) Stendahl, J. C.; Rao, M. S.; Guler, M. O.; Stupp, S. I. Adv. Funct. Mater. 2006, 16 (4), 499–508.

142

(41) Mai, Y.; Eisenberg, A. Chem. Soc. Rev. 2012, 41 (18), 5969–5985.

(42) Paramonov, S. E.; Jun, H.; Hartgerink, J. D. J. Am. Chem. Soc. 2006, 128, 7291–7298.

(43) Lee, K. Y.; Mooney, D. J. Chem. Rev. 2001, 101 (7), 1869–1880.

(44) Blanazs, A.; Verber, R.; Mykhaylyk, O. O.; Ryan, A. J.; Heath, J. Z.; Douglas, C. W. I.; Armes, S. P. J. Am. Chem. Soc. 2012, 134 (23), 9741–9748.

(45) Zhao, X.; Pan, F.; Xu, H.; Yaseen, M.; Shan, H.; Hauser, C. a E.; Zhang, S.; Lu, J. R. Chem. Soc. Rev. 2010, 39 (9), 3480–3498.

(46) Stanley, F. N. In Food Polysaccharides and Their Applications; Stephen, A. M., Phillips, G. O., Williams, P. A., Eds.; CRC Press: Boca Raton, 2006; pp 217–238.

(47) Domozych, D. S. In eLS; John Wiley & Sons: Chichester, 2011.

(48) Arnott, S.; Fulmer, A.; Scott, W. E.; Dea, I. C. M.; Moorhouse, R.; Rees, D. A. J. Mol. Biol. 1974, 90 (2), 269–284.

(49) Aymard, P.; Martin, D. R.; Plucknett, K.; Foster, T. J.; Clark, A. H.; Norton, I. T. Biopolymers 2001, 59 (3), 131–144.

(50) Indovina, P. L.; Tettamanti, E.; Micciancio-Giammarinaro, M. S.; Palma, M. U. J. Chem. Phys. 1979, 70 (6), 2841.

(51) Djabourov, M.; Clark, A. H.; Rowlands, D. W.; Ross-Murphy, S. B. Macromolecules 1989, 22 (1), 180–188.

(52) Kumachev, A.; Tumarkin, E.; Walker, G. C.; Kumacheva, E. Soft Matter 2013, 9 (10), 2959–2965.

(53) Kumachev, A.; Greener, J.; Tumarkin, E.; Eiser, E.; Zandstra, P. W.; Kumacheva, E. Biomaterials 2011, 32 (6), 1477–1483.

(54) Chau, M.; Abolhasani, M.; Thérien-Aubin, H.; Li, Y.; Wang, Y.; Velasco, D.; Tumarkin, E.; Ramachandran, A.; Kumacheva, E. Biomacromolecules 2014, 15 (7), 2419–2425.

(55) Habibi, Y.; Lucia, L. a; Rojas, O. J. Chem. Rev. 2010, 110 (6), 3479–3500.

(56) Abdul Khalil, H. P. S.; Bhat, A. H.; Ireana Yusra, A. F. Carbohydr. Polym. 2012, 87 (2), 963–979.

(57) Domingues, R. M. A.; Gomes, M. E.; Reis, R. L. Biomacromolecules 2014, 15 (7), 2327– 2346.

(58) Ranby, B. G. Discuss. Faraday Soc. 1951, 11, 158–164.

(59) Marchessault, R. H.; Morehead, F. F.; Koch, M. J. J. Colloid Sci. 1961, 344, 327–344.

143

(60) Dorris, A.; Gray, D. G. Cellulose 2012, 19 (3), 687–694.

(61) Dufresne, A. Nanocellulose: From Nature to High Performance Tailored Materials.; Walter de Gruyter GmbH: Berlin, 2012.

(62) Lin, N.; Dufresne, A. Nanoscale 2014, 6 (10), 5384–5393.

(63) Klemm, D.; Kramer, F.; Moritz, S.; Lindström, T.; Ankerfors, M.; Gray, D.; Dorris, A. Angew. Chem. Int. Ed. Engl. 2011, 50 (24), 5438–5466.

(64) Dong, X. M.; Revol, J.-F.; Gray, D. G. Cellulose 1998, 5, 19–32.

(65) Liu, D.; Chen, X.; Yue, Y.; Chen, M.; Wu, Q. Carbohydr. Polym. 2011, 84 (1), 316–322.

(66) Urena-Benavides, E. E.; Ao, G.; Davis, V. A.; Kitchens, C. L. Macromolecules 2011, 44, 8990–8998.

(67) Shafeiei-Sabet, S.; Hamad, W. Y.; Hatzikiriakos, S. G. Rheol. Acta 2013, 52 (8-9), 741– 751.

(68) Derakhshandeh, B.; Petekidis, G.; Shafiei Sabet, S.; Hamad, W. Y.; Hatzikiriakos, S. G. J. Rheol. (N. Y. N. Y). 2013, 57 (1), 131.

(69) Shafiei-Sabet, S.; Hamad, W. Y.; Hatzikiriakos, S. G. Cellulose 2014, 21 (5), 3347–3359.

(70) Habibi, Y. Chem. Soc. Rev. 2014, 43 (5), 1519–1542.

(71) Eyley, S.; Thielemans, W. Nanoscale 2014, 6 (14), 7764–7779.

(72) Hasani, M.; Cranston, E. D.; Westman, G.; Gray, D. G. Soft Matter 2008, 4 (11), 2238.

(73) Way, A. E.; Hsu, L.; Shanmuganathan, K.; Weder, C.; Rowan, S. J. ACS Macro Lett. 2012, 1 (8), 1001–1006.

(74) Wang, Y.; Chen, L. Carbohydr. Polym. 2011, 83 (4), 1937–1946.

(75) Kalia, S.; Dufresne, A.; Cherian, B. M.; Kaith, B. S.; Avérous, L.; Njuguna, J.; Nassiopoulos, E. Int. J. Polym. Sci. 2011, 2011, 1–35.

(76) Velasco, D.; Tumarkin, E.; Kumacheva, E. Small 2012, 8 (11), 1633–1642.

(77) Tumarkin, E.; Tzadu, L.; Csaszar, E.; Seo, M.; Zhang, H.; Lee, A.; Peerani, R.; Purpura, K.; Zandstra, P. W.; Kumacheva, E. Integr. Biol. (Camb). 2011, 3 (6), 653–662.

(78) Kumachev, A.; Greener, J.; Tumarkin, E.; Eiser, E.; Zandstra, P. W.; Kumacheva, E. Biomaterials 2011, 32 (6), 1477–1483.

(79) Dalby, M. J.; Riehle, M. O.; Yarwood, S. J.; Wilkinson, C. D. .; Curtis, A. S. . Exp. Cell Res. 2003, 284 (2), 272–280.

144

(80) Maniotis, A. J.; Chen, C. S.; Ingber, D. E. Proc. Natl. Acad. Sci. 1997, 94 (3), 849–854.

(81) Curtis, A. S. G.; Seehar, G. M. Nature 1978, 274 (5666), 52–53.

(82) Curtis, A. S. G. Biomechanics and Cells; Lyall, F., Haj, A. J. El, Eds.; Cambridge University Press: Cambridge, 1994.

(83) Bruinink, A.; Wintermantel, E. Biomaterials 2001, 22 (18), 2465–2473.

(84) Goldspink, G.; Scutt, A.; Loughna, P. T.; Wells, D. J.; Jaenicke, T.; Gerlach, G. F. Am J Physiol Regul. Integr. Comp Physiol 1992, 262 (3), R356–R363.

(85) Caliari, S. R.; Harley, B. A. C. Biomaterials 2011, 32 (23), 5330–5340.

(86) Murphy, S. V; Atala, A. Nat. Biotechnol. 2014, 32 (8), 773–785.

(87) Mathieu, L. M.; Mueller, T. L.; Bourban, P.-E.; Pioletti, D. P.; Müller, R.; Månson, J.-A. E. Biomaterials 2006, 27 (6), 905–916.

(88) Ho, M.-H.; Kuo, P.-Y.; Hsieh, H.-J.; Hsien, T.-Y.; Hou, L.-T.; Lai, J.-Y.; Wang, D.-M. Biomaterials 2004, 25 (1), 129–138.

(89) Murphy, C. M.; Haugh, M. G.; O’Brien, F. J. Biomaterials 2010, 31 (3), 461–466.

(90) Lie Ma, Changyou Gao, Zhengwei Mao, Jie Zhou, Jiacong Shen, Xueqing Hu, C. H. Biomaterials 2003, 24 (26), 4833–4841.

(91) Francis, N. L.; Hunger, P. M.; Donius, A. E.; Riblett, B. W.; Zavaliangos, A.; Wegst, U. G. K.; Wheatley, M. A. J. Biomed. Mater. Res. A 2013, 101 (12), 3493–3503.

(92) Deville, S. Adv. Eng. Mater. 2008, 10 (3), 155–169.

(93) Deville, S.; Saiz, E.; Tomsia, A. P. Acta Mater. 2007, 55 (6), 1965–1974.

(94) Wu, X.; Liu, Y.; Li, X.; Wen, P.; Zhang, Y.; Long, Y.; Wang, X.; Guo, Y.; Xing, F.; Gao, J. Acta Biomater. 2010, 6 (3), 1167–1177.

(95) He, J.; Mao, M.; Liu, Y.; Zhu, L.; Li, D. Mater. Lett. 2013, 90, 93–96.

(96) Aranaz, I.; Gutiérrez, M. C.; Ferrer, M. L.; del Monte, F. Mar. Drugs 2014, 12 (11), 5619–5642.

(97) Raumedic. Catheter Range and Sets https://www.raumedic.com/fileadmin/user_upload/PDF/catheter-range-and-sets.pdf (accessed Jan 26, 2016).

(98) AERO® (OTW) Tracheobronchial Stent http://endotek.merit.com/products/pulmonary/aero-tracheobronchial-stent.html (accessed Jan 30, 2016).

145

(99) Medtronic. Medtronic Pacemaker Leads http://www.medtronic.com/for-healthcare- professionals/products-therapies/cardiac-rhythm/pacemakers/pacing-leads/ (accessed Jan 26, 2016).

(100) Tarsy, D.; Vitek, J. L.; Starr, P.; Okun, M. Deep Brain Stimulation in Neurological and Psychiatric Disorders; Current Clinical Neurology; Humana Press, 2008.

(101) Kütting, M.; Roggenkamp, J.; Urban, U.; Schmitz-Rode, T.; Steinseifer, U. Expert Rev. Med. Devices 2014.

(102) Thomson, T. Polyurethanes as Specialty Chemicals: Principles and Applications; CRC Press, 2004.

(103) Guelcher, S. A.; Srinivasan, A.; Dumas, J. E.; Didier, J. E.; McBride, S.; Hollinger, J. O. Biomaterials 2008, 29 (12), 1762–1775.

(104) Porter, J. R.; Ruckh, T. T.; Popat, K. C. Biotechnol. Prog. 25 (6), 1539–1560.

(105) Janik, H.; Marzec, M. Mater. Sci. Eng. C. Mater. Biol. Appl. 2015, 48, 586–591.

(106) Zhang, Z.; Wang, Z.; Liu, S.; Kodama, M. Biomaterials 2004, 25 (1), 177–187.

(107) Guan, J.; Fujimoto, K. L.; Sacks, M. S.; Wagner, W. R. Biomaterials 2005, 26 (18), 3961– 3971.

(108) Guan, J.; Stankus, J. J.; Wagner, W. R. J. Control. Release 2007, 120 (1-2), 70–78.

(109) Zhisen Shen, Jingjing Chen, Cheng Kang, Changfeng Gong, Y. Z. Biomed Res. Int. 2013, No. Article ID 138504, 11 pages.

(110) de Mulder, E. L. W.; Hannink, G.; Verdonschot, N.; Buma, P. Biomed. Mater. 2013, 8 (2), 025004.

(111) Christian Holtze, Jeremy Jon Agresti, David A. Weitz, Keunho Ahn, John Brian Hutchison, Andrew Gifiths, Abdesiam El Harrak, Olive John Miller, Jean-Christophe Barat, Valorie Taly, Michael Ryckelynck, C. M. Fluorocarbon emulsion stabilizing surfactants. 20100105112 A1, 2010.

(112) Sakai, S.; Hirose, K.; Taguchi, K.; Ogushi, Y.; Kawakami, K. Biomaterials 2009, 30 (20), 3371–3377.

(113) Horger, K. S.; Estes, D. J.; Capone, R.; Mayer, M. J. Am. Chem. Soc. 2009, 131 (5), 1810–1819.

(114) Xia, Y.; Whitesides, G. M. Angew. Chemie Int. Ed. 1998, 37 (5), 550–575.

(115) Kleinberger, R. M.; Burke, N. A. D.; Dalnoki-Veress, K.; Stöver, H. D. H. Mater. Sci. Eng. C 2013, 33 (7), 4295–4304.

(116) Theret, D. P.; Levesque, M. J.; Sato, M.; Nerem, R. M.; Wheeler, L. T. J. Biomech. Eng.

146

1988, 110 (3), 190–199.

(117) Wallace, M.; Adams, D. J.; Iggo, J. A. Soft Matter 2013, 9 (22), 5483–5491.

(118) Hu, Z.; Cranston, E. D.; Ng, R.; Pelton, R. Langmuir 2014, 30 (10), 2684–2692.

(119) Amsden, B. Macromolecules 1999, 32 (3), 874–879.

(120) Smeets, N. M. B.; Bakaic, E.; Patenaude, M.; Hoare, T. Acta Biomater. 2014, 10 (10), 4143–4155.

(121) Smeets, N. M. B.; Bakaic, E.; Patenaude, M.; Hoare, T. Chem. Commun. 2014, 50 (25), 3306–3309.

(122) Beck-candanedo, S.; Roman, M.; Gray, D. G. Biomacromolecules 2005, 6, 1048–1054.

(123) Beck, S.; Méthot, M.; Bouchard, J. Cellulose 2014, 22 (1), 101–116.

(124) Sun, B.; Hou, Q.; Liu, Z.; Ni, Y. Cellulose 2015, 22 (2), 1135–1146.

(125) Brunauer, S.; Emmett, P. H.; Teller, E. J. Am. Chem. Soc. 1938, 60 (2), 309–319.

(126) Förster, S.; Fischer, S.; Zielske, K.; Schellbach, C.; Sztucki, M.; Lindner, P.; Perlich, J. Adv. Colloid Interface Sci. 2011, 163 (1), 53–83.

(127) Chau, M.; Abolhasani, M.; Li, Y.; Wang, Y.; Velasco, D.; Tumarkin, E.; Ramachandran, A.; Kumacheva, E. Biomacromolecules 2014, 15, 2419–2425.

(128) Bosman, F. T.; Stamenkovic, I. J. Pathol. 2003, 200 (4), 423–428.

(129) Tibbitt, M. W.; Anseth, K. S. Biotechnol. Bioeng. 2009, 103 (4), 655–663.

(130) Kobel, S.; Lutolf, M. Biotechniques 2010, 48 (4), ix – xxii.

(131) Chen, J.-H.; Liu, C.; You, L.; Simmons, C. A. J. Biomech. 2010, 43 (1), 108–118.

(132) Holy, C. E.; Shoichet, M. S.; Davies, J. E. J. Biomed. Mater. Res. 2000, 51 (3), 376–382.

(133) Yamada, K. M.; Cukierman, E. Cell 2007, 130 (4), 601–610.

(134) Teixeira, L. S. M.; Feijen, J.; van Blitterswijk, C. A.; Dijkstra, P. J.; Karperien, M. Biomaterials 2012, 33 (5), 1281–1290.

(135) Wan, J. Polymers (Basel). 2012, 4 (4), 1084–1108.

(136) McGuigan, A. P.; Sefton, M. V. Tissue Eng. 2007, 13 (5), 1069–1078.

(137) Tumarkin, E.; Tzadu, L.; Csaszar, E.; Seo, M.; Zhang, H.; Lee, A.; Peerani, R.; Purpura, K.; Zandstra, P. W.; Kumacheva, E. Integr. Biol. 2011, 3 (6), 653–662.

147

(138) Whitesides, G. M. Nature 2006, 442 (7101), 368–373.

(139) Teh, S.-Y.; Lin, R.; Hung, L.-H.; Lee, A. P. Lab Chip 2008, 8 (2), 198–220.

(140) Velasco, D.; Chau, M.; Thérien-Aubin, H.; Kumachev, A.; Tumarkin, E.; Jia, Z.; Walker, G. C.; Monteiro, M. J.; Kumacheva, E. Soft Matter 2013, 9 (8), 2380.

(141) Feke, G. T., Prins, W. Macromolecules 1974, 7 (4), 527–530.

(142) Hynes, R. O. Science (80-. ). 2009, 326 (5957), 1216–1219.

(143) Eastoe, J. E. Biochem. J. 1955, 61 (4), 589–600.

(144) Ross-Murphy, S. B. Polymer (Guildf). 1992, 33 (12), 2622–2627.

(145) Gornall, J. L.; Terentjev, E. M. Soft Matter 2008, 4 (3), 544.

(146) Anna, S. L.; Bontoux, N.; Stone, H. a. Appl. Phys. Lett. 2003, 82 (3), 364–366.

(147) Haight, F. A. Handbook of the Poisson distribution; Wiley, 1967.

(148) L. D. Landau, E. M. L. Fluid mechanics; Pergamon Press: Oxford, England, 1987.

(149) Chan, K. L. A.; Niu, X.; de Mello, A. J.; Kazarian, S. G. Lab Chip 2010, 10 (16), 2170– 2174.

(150) Koo, J.; Kleinstreuer, C. Int. J. Heat Mass Transf. 2004, 47 (14-16), 3159–3169.

(151) Yasuda, T.; Okuno, T.; Yasuda, H. Langmuir 1994, 10 (7), 2435–2439.

(152) Clark, A. H.; Richardson, R. K.; Ross-Murphy, S. B.; Stubbs, J. M. Macromolecules 1983, 16 (8), 1367–1374.

(153) Rahbani, J.; Behzad, A. R.; Khashab, N. M.; Al-Ghoul, M. Electrophoresis 2013, 34 (3), 405–408.

(154) Langton, M.; Hermansson, A.; Lore, N. Food Hydrocoll. 1999, 13, 185–198.

(155) Tuvikene, R.; Truus, K.; Kollist, A.; Volobujeva, O.; Mellikov, E.; Pehk, T. J. Appl. Phycol. 2007, 20 (5), 527–535.

(156) Ma, S.; Natoli, M.; Liu, X.; Neubauer, M. P.; Watt, F. M.; Fery, A.; Huck, W. T. S. J. Mater. Chem. B 2013, 1 (38), 5128.

(157) Kleinberger, R. M.; Burke, N. A. D.; Dalnoki-Veress, K.; Stöver, H. D. H. Mater. Sci. Eng. C. Mater. Biol. Appl. 2013, 33 (7), 4295–4304.

(158) Li, Y.; Kumacheva, E.; Ramachandran, A. Soft Matter 2013, 9 (43), 10391–10403.

(159) Mansoor, A. H.; Guilak, F. J. Biomech. Eng. 2002, 124 (5), 586–595.

148

(160) Jones, W. R.; Ping Ting-Beall, H.; Lee, G. M.; Kelley, S. S.; Hochmuth, R. M.; Guilak, F. J. Biomech. 1999, 32 (2), 119–127.

(161) Boudou, T.; Ohayon, J.; Arntz, Y.; Finet, G.; Picart, C.; Tracqui, P. J. Biomech. 2006, 39 (9), 1677–1685.

(162) Tolstoguzov, V. B. Food Hydrocoll. 1995, 9 (4), 317–332.

(163) Chau, M.; Sriskandha, S. E.; Pichugin, D.; Thérien-Aubin, H.; Nykypanchuk, D.; Chauve, G.; Méthot, M.; Bouchard, J.; Gang, O.; Kumacheva, E. Biomacromolecules 2015, 16 (8), 2455–2462.

(164) Xu, B. Langmuir 2009, 25 (15), 8375–8377.

(165) Kara, S.; Tamerler, C.; Bermek, H.; Pekcan, Ö. J. Bioact. Compat. Polym. 2003, 18 (1), 33–44.

(166) Hirst, L. S.; Pynn, R.; Bruinsma, R. F.; Safinya, C. R. J. Chem. Phys. 2005, 123 (10), 104902.

(167) Fall, A. B.; Lindström, S. B.; Sprakel, J.; Wågberg, L. Soft Matter 2013, 9 (6), 1852–1863.

(168) Dugan, J. M.; Gough, J. E.; Eichhorn, S. J. Nanomedicine 2013, 8 (2), 287–298.

(169) Moon, R. J.; Martini, A.; Nairn, J.; Simonsen, J.; Youngblood, J. Chem. Soc. Rev. 2011, 40 (7), 3941–3994.

(170) Lin, N.; Huang, J.; Dufresne, A. Nanoscale 2012, 4 (11), 3274–3294.

(171) McKee, J. R.; Hietala, S.; Seitsonen, J.; Laine, J.; Kontturi, E.; Ikkala, O. ACS Macro Lett. 2014, 3 (3), 266–270.

(172) Shafiei-Sabet, S.; Hamad, W. Y.; Hatzikiriakos, S. G. Cellulose 2014, 21 (5), 3347–3359.

(173) Dong, H.; Snyder, J. F.; Williams, K. S.; Andzelm, J. W. Biomacromolecules 2013, 14, 3338–3345.

(174) Shannon, R. D. Acta Crystallogr. Sect. A 1976, 32 (5), 751–767.

(175) Haynes, W. M. CRC Handbook of Chemistry and Physics, 95th Edition; CRC Handbook of Chemistry and Physics; Taylor & Francis, 2014.

(176) Israelachvili, J. N. Intermolecular and Surface Forces; Intermolecular and Surface Forces; Elsevier Science, 2010.

(177) Mezger, T. G. The rheology handbook: For users of rotational and oscillatory rheometers; Vincentz Network: Hannover, 2006.

(178) Likos, C. N. Phys. Rep. 2001, 348 (4-5), 267–439.

149

(179) Haynes, W. M. CRC Handbook of Chemistry and Physics, 95th Edition; 2014.

(180) Pearson, R. G. J. Chem. Educ. 1968, 45 (10), 643.

(181) Goodwin, J. W.; Hughes, R. W. Rheology for Chemists; Royal Society of Chemistry: Cambridge, 2008.

(182) O’Leary, L. E. R.; Fallas, J. A.; Bakota, E. L.; Kang, M. K.; Hartgerink, J. D. Nat. Chem. 2011, 3 (10), 821–828.

(183) Ulrich, T. A.; Jain, A.; Tanner, K.; MacKay, J. L.; Kumar, S. Biomaterials 2010, 31 (7), 1875–1884.

(184) Ohkubo, T.; Kidena, K.; Ohira, A. Macromolecules 2008, 41 (22), 8688–8693.

(185) Stepišnik, J.; Fritzinger, B.; Scheler, U.; Mohorič, A.; Mohorič, A. EPL (Europhysics Lett. 2012, 98 (5), 57009.

(186) Lagerwall, J. P. F.; Schütz, C.; Salajkova, M.; Noh, J.; Hyun Park, J.; Scalia, G.; Bergström, L. NPG Asia Mater. 2014, 6 (1), e80.

(187) Dong, X. M.; Kimura, T.; Gray, D. G. Langmuir 1996, 12, 2076–2082.

(188) Ilavsky, J.; Jemian, P. R. J. Appl. Crystallogr. 2009, 42 (2), 347–353.

(189) Hui, C.-Y.; Muralidharan, V. J. Chem. Phys. 2005, 123 (15), 154905.

(190) Biot, M. A. J. Appl. Phys. 1962, 33 (4), 1482–1498.

(191) Hoare, T. R.; Kohane, D. S. Polymer (Guildf). 2008, 49 (8), 1993–2007.

(192) Hoffman, A. S. Adv. Drug Deliv. Rev. 2002, 54 (1), 3–12.

(193) Kim, J. J.; Park, K. Bioseparation 1999, 7 (4-5), 177–184.

(194) Jing, G.; Wang, L.; Yu, H.; Amer, W. a.; Zhang, L. Colloids Surf., A 2013, 416, 86–94.

(195) Haque, M. A.; Kurokawa, T.; Gong, J. P. Soft Matter 2012, 8 (31), 8008–8016.

(196) Walker, K. J.; Madihally, S. V. J. Biomed. Mater. Res. Part B Appl. Biomater. 2014, No. August, 1146–1160.

(197) Rudisill, S. G.; DiVito, M. D.; Hubel, A.; Stein, A. Acta Biomater. 2014, 12, 122–128.

(198) Aubin, H.; Nichol, J. W.; Hutson, C. B.; Bae, H.; Sieminski, A. L.; Cropek, D. M.; Akhyari, P.; Khademhosseini, A. Biomaterials 2010, 31 (27), 6941–6951.

(199) Yamada, M.; Sugaya, S.; Naganuma, Y.; Seki, M. Soft Matter 2012, 8 (11), 3122–3130.

(200) Li, Y.; Huang, G.; Zhang, X.; Wang, L.; Du, Y.; Lu, T. J.; Xu, F. Biotechnol. Adv. 2014,

150

32 (2), 347–365.

(201) Camarero-Espinosa, S.; Rothen-Rutishauser, B.; Weder, C.; Foster, E. J. Biomaterials 2015, 74, 42–52.

(202) Swan, M. C.; Bucknall, D. G.; Goodacre, T. E. E.; Czernuszka, J. T. Acta Biomater. 2011, 7 (3), 1126–1132.

(203) Tseng, H.; Puperi, D. S.; Kim, E. J.; Ayoub, S.; Shah, J. V; Cuchiara, M. L.; West, J. L.; Grande-Allen, K. J. Tissue Eng. Part A 2014, 20 (19-20), 2634–2645.

(204) Voge, C. M.; Kariolis, M.; MacDonald, R. A.; Stegemann, J. P. J. Biomed. Mater. Res. A 2008, 86 (1), 269–277.

(205) Dvir, T.; Timko, B. P.; Brigham, M. D.; Naik, S. R.; Karajanagi, S. S.; Levy, O.; Jin, H.; Parker, K. K.; Langer, R.; Kohane, D. S. Nat. Nanotechnol. 2011, 6 (11), 720–725.

(206) Ahadian, S.; Ramón-Azcón, J.; Estili, M.; Liang, X.; Ostrovidov, S.; Shiku, H.; Ramalingam, M.; Nakajima, K.; Sakka, Y.; Bae, H.; Matsue, T.; Khademhosseini, A. Sci. Rep. 2014, 4, 4271.

(207) Barrow, M.; Zhang, H. Soft Matter 2013, 9 (9), 2723–2729.

(208) Mao, M.; He, J.; Liu, Y.; Li, X.; Li, D. Acta Biomater. 2012, 8 (6), 2175–2184.

(209) Chen, M.; Zhu, J.; Qi, G.; He, C.; Wang, H. Mater. Lett. 2012, 89, 104–107.

(210) Zhu, J.; Wang, J.; Liu, Q.; Liu, Y.; Wang, L.; He, C.; Wang, H. J. Mater. Chem. B 2013, 1 (7), 978–986.

(211) Paz, A. C.; Soleas, J.; Poon, J. C. H.; Trieu, D.; Waddell, T. K.; McGuigan, A. P. Tissue Eng. Part B. Rev. 2014, 20 (1), 56–72.

(212) He, J.; Wang, Y.; Liu, Y.; Li, D.; Jin, Z. Biofabrication 2013, 5 (2), 025002.

(213) Barrow, M.; Eltmimi, A.; Ahmed, A.; Myers, P.; Zhang, H. J. Mater. Chem. 2012, 22 (23), 11615–11620.

(214) Chatelin, S.; Bernal, M.; Deffieux, T.; Papadacci, C.; Flaud, P.; Nahas, A.; Boccara, C.; Gennisson, J.-L.; Tanter, M.; Pernot, M. Phys. Med. Biol. 2014, 59 (22), 6923–6940.

(215) Osorio-Madrazo, A.; Eder, M.; Rueggeberg, M.; Pandey, J. K.; Harrington, M. J.; Nishiyama, Y.; Putaux, J.-L.; Rochas, C.; Burgert, I. Biomacromolecules 2012, 13 (3), 850–856.

(216) Yang, W.; Furukawa, H.; Gong, J. P. Adv. Mater. 2008, 20 (23), 4499–4503.

(217) Zhang, S.; Greenfield, M. a; Mata, A.; Palmer, L. C.; Bitton, R.; Mantei, J. R.; Aparicio, C.; de la Cruz, M. O.; Stupp, S. I. Nat. Mater. 2010, 9 (7), 594–601.

151

(218) Hockaday, L. A.; Kang, K. H.; Colangelo, N. W.; Cheung, P. Y. C.; Duan, B.; Malone, E.; Wu, J.; Girardi, L. N.; Bonassar, L. J.; Lipson, H.; Chu, C. C.; Butcher, J. T. Biofabrication 2012, 4 (3), 035005.

(219) Barrow, M.; Zhang, H. Soft Matter 2013, 9 (9), 2723–2729.

(220) Liu, H.; Nakagawa, K.; Chaudhary, D.; Asakuma, Y.; Tadé, M. O. Chem. Eng. Res. Des. 2011, 89 (11), 2356–2364.

(221) Yokoyama, F.; Achife, E. C.; Momoda, J.; Shimamura, K.; Monobe, K. Colloid Polym. Sci. 1990, 268 (6), 552–558.

(222) Köhnke, T.; Elder, T.; Theliander, H.; Ragauskas, A. J. Carbohydr. Polym. 2014, 100, 24– 30.

(223) Wicklein, B.; Kocjan, A.; Salazar-Alvarez, G.; Carosio, F.; Camino, G.; Antonietti, M.; Bergström, L. Nat. Nanotechnol. 2015, 10 (3), 277–283.

(224) Darder, M.; Aranda, P.; Ferrer, M. L.; Gutiérrez, M. C.; del Monte, F.; Ruiz-Hitzky, E. Adv. Mater. 2011, 23 (44), 5262–5267.

(225) Gutiérrez, M. C.; Ferrer, M. L.; del Monte, F. Chem. Mater. 2008, 20 (3), 634–648.

(226) Wu, J.; Zhao, Q.; Sun, J.; Zhou, Q. Soft Matter 2012, 8 (13), 3620–3626.

(227) Bai, H.; Polini, A.; Delattre, B.; Tomsia, A. P. Chem. Mater. 2013, 25 (22), 4551–4556.

(228) Wu, J.; Lin, Y.; Sun, J. J. Mater. Chem. 2012, 22 (34), 17449–17451.

(229) Yang, X.; Bakaic, E.; Hoare, T.; Cranston, E. D. Biomacromolecules 2013, 14 (12), 4447– 4455.

(230) Yang, J.; Zhao, J.-J.; Xu, F.; Sun, R.-C. ACS Appl. Mater. Interfaces 2013, 5 (24), 12960– 12967.

(231) Zhou, C.; Wu, Q.; Yue, Y.; Zhang, Q. J. Colloid Interface Sci. 2011, 353 (1), 116–123.

(232) Yang, J.; Zhao, J.-J.; Han, C.-R.; Duan, J.-F.; Xu, F.; Sun, R.-C. Cellulose 2013, 21 (1), 541–551.

(233) Yang, J.; Han, C.-R.; Zhang, X.-M.; Xu, F.; Sun, R.-C. Macromolecules 2014, 47 (12), 4077–4086.

(234) Siham Atifi, Shunxing Su, W. Y. H. Nord. Pulp Pap. Res. J. 2014, 29 (1), 95–104.

(235) Hebeish, A.; Farag, S.; Sharaf, S.; Shaheen, T. I. Carbohydr. Polym. 2014, 102, 159–166.

(236) Yang, J.; Han, C.; Xu, F.; Sun, R. Nanoscale 2014, 6 (11), 5934–5943.

(237) De France, K. J.; Chan, K. J. W.; Cranston, E. D.; Hoare, T. Biomacromolecules 2016, 17

152

(2), 649–660.

(238) Chiang, Y.; Kresge, A. J. J. Org. Chem. 1985, 50 (25), 5038–5040.

(239) Eichhorn, S. J. Soft Matter 2011, 7 (2), 303–315.

(240) Azizi Samir, M. A. S.; Alloin, F.; Dufresne, A. Biomacromolecules 2005, 6 (2), 612–626.

(241) Lutz, J.-F. J. Polym. Sci. Part A Polym. Chem. 2008, 46 (11), 3459–3470.

(242) Kumar, A.; Negi, Y. S.; Choudhary, V.; Bhardwaj, N. K. J. Mater. Phys. Chem. 2014, 2 (1), 1–8.

(243) Heath, L.; Thielemans, W. Green Chem. 2010, 12 (8), 1448–1453.

(244) Seeck, O. H.; Murphy, B. X-Ray Diffraction: Modern Experimental Techniques; Pan Stanford Publishing, 2015.

(245) Han, J.; Zhou, C.; Wu, Y.; Liu, F.; Wu, Q. Biomacromolecules 2013, 14 (5), 1529–1540.

(246) Tibbitt, M. W.; Rodell, C. B.; Burdick, J. A.; Anseth, K. S. Proc. Natl. Acad. Sci. 2015, 112 (47), 201516247.

(247) Öchsner, A.; Murch, G. E.; de Lemos, M. J. S. Cellular and Porous Materials: Thermal Properties Simulation and Prediction; Wiley, 2008.

(248) Zhao, C. Y. Int. J. Heat Mass Transf. 2012, 55 (13-14), 3618–3632.

(249) BASF Neopor ® Thermal Conductivity http://www.plasticsportal.net/wa/plasticsEU~en_GB/portal/show/content/products/foams/ neopor_thermal_conductivity (accessed Dec 31, 2015).

(250) Huber, A. T.; Gibson, L. J. J. Mater. Sci. 1988, 23 (8), 3031–3040.

(251) Gibson, L. J.; Ashby, M. F. Cellular Solids: Structure and Properties; Cambridge Solid State Science Series; Cambridge University Press, 1999.

(252) Antunes, M.; Realinho, V.; Velasco, J. I.; Solórzano, E.; Rodríguez-Pérez, M.-Á.; de Saja, J. A. Mater. Chem. Phys. 2012, 136 (1), 268–276.

(253) Lee, T.; Lakes, R. S. J. Mater. Sci. 1997, 32 (9), 2397–2401.

(254) Deville, S.; Maire, E.; Bernard-Granger, G.; Lasalle, A.; Bogner, A.; Gauthier, C.; Leloup, J.; Guizard, C. Nat. Mater. 2009, 8 (12), 966–972.

(255) Deville, S.; Saiz, E.; Nalla, R. K.; Tomsia, A. P. Science 2006, 311 (5760), 515–518.

(256) Haynes, W. M. CRC Handbook of Chemistry and Physics, 96th Edition; CRC Press, 2015.

(257) Rueda-Larraz, L.; d’Arlas, B. F.; Tercjak, A.; Ribes, A.; Mondragon, I.; Eceiza, A. Eur.

153

Polym. J. 2009, 45 (7), 2096–2109.

(258) Aldrich, S. Carbon Nanofibers-Sigma Aldrich http://www.sigmaaldrich.com/catalog/product/aldrich/719781?lang=en®ion=CA (accessed Aug 2, 2016).

(259) Lascovich, J. C.; Giorgi, R.; Scaglione, S. Appl. Surf. Sci. 1991, 47 (1), 17–21.

(260) Olivier, C.; Moreau, C.; Bertoncini, P.; Bizot, H.; Chauvet, O.; Cathala, B. Langmuir 2012, 28 (34), 12463–12471.

(261) Moreau, C.; Villares, A.; Capron, I.; Cathala, B. Ind. Crops Prod. 2016.

(262) Kalashnikova, I.; Bizot, H.; Cathala, B.; Capron, I. Biomacromolecules 2012, 13 (1), 267– 275.

(263) Harris, P. J. F. Crit. Rev. Solid State Mater. Sci. 2005, 30 (4), 235–253.

(264) Coogan, R. G. Prog. Org. Coatings 1997, 32 (1-4), 51–63.

(265) Yuan, J.; Lin, S.; Shen, J. Colloids Surf. B. Biointerfaces 2008, 66 (1), 90–95.

(266) Deville, S.; Saiz, E.; Tomsia, A. P. Biomaterials 2006, 27 (32), 5480–5489.

(267) Chen, T. K.; Shieh, T. S.; Chui, J. Y. Macromolecules 1998, 31 (4), 1312–1320.

(268) Yanagihara, Y.; Osaka, N.; Murayama, S.; Saito, H. Polymer (Guildf). 2013, 54 (8), 2183– 2189.

(269) Seymour, R. W.; Cooper, S. L. J. Polym. Sci. Part B Polym. Lett. 1971, 9 (9), 689–694.

(270) Bogdanov, B.; Toncheva, V.; Schacht, E.; Finelli, L.; Sarti, B.; Scandola, M. Polymer (Guildf). 1999, 40 (11), 3171–3182.

(271) Xia, H.; Song, M. J. Mater. Chem. 2006, 16 (19), 1843.

(272) Ungár, T.; Gubicza, J.; Ribárik, G.; Pantea, C.; Zerda, T. W. Carbon N. Y. 2002, 40 (6), 929–937.

(273) Poveda, R. L.; Gupta, N. Carbon Nanofiber Reinforced Polymer Composites; SpringerBriefs in Materials; Springer International Publishing, 2015.

(274) Ponnamma, D.; Thomas, S. Non-Linear Viscoelasticity of Rubber Composites and Nanocomposites: Influence of Filler Geometry and Size in Different Length Scales; Advances in Polymer Science; Springer International Publishing, 2014.

(275) Sonnenschein, M. F. Polyurethanes: Science, Technology, Markets, and Trends; Wiley Series on Polymer Engineering and Technology; Wiley, 2014.

(276) Ashby, M. F. Philos. Trans. R. Soc. London A Math. Phys. Eng. Sci. 2006, 364 (1838),

154

15–30.

(277) Granta Design Limited, 2015 2015.

(278) Rusli, R.; Eichhorn, S. J. Appl. Phys. Lett. 2008, 93 (3), 033111.

(279) Cao, X.; Dong, H.; Li, C. M. Biomacromolecules 2007, 8 (3), 899–904.

(280) Antunes, M.; Velasco, J. I. Prog. Polym. Sci. 2014, 39 (3), 486–509.

(281) Thermoplastic Polyurethane Elastomers (TPU) http://www.polyurethanes.basf.de/pu/solutions/us/function/conversions:/publish/content/g roup/Arbeitsgebiete_und_Produkte/Thermoplastische_Spezialelastomere/Infomaterial/elas tollan_material_uk.pdf (accessed Mar 18, 2016).

(282) Wu, J.-W.; Chu, H.-S. Heat Mass Transf. 1998, 34 (4), 247–254.

(283) Patel, P. S. D.; Shepherd, D. E. T.; Hukins, D. W. L. BMC Musculoskelet. Disord. 2008, 9, 137.

(284) Ellis, B.; Smith, R. Polymers: A Property Database, Second Edition; CRC Press, 2008.

(285) Benli, S.; Yilmazer, U.; Pekel, F.; Ozkar, S. J. Appl. Polym. Sci. 1998, 68 (7), 1057–1065.

(286) Jelle, B. P. Energy Build. 2011, 43 (10), 2549–2563.

(287) Schild, H. G. Prog. Polym. Sci. 1992, 17 (2), 163–249.

(288) Huang, X.; Jain, P. K.; El-Sayed, I. H.; El-Sayed, M. A. Photochem. Photobiol. 82 (2), 412–417.

(289) Jiang, K.; Smith, D. A.; Pinchuk, A. J. Phys. Chem. C 2013, 117 (51), 27073–27080.

(290) Zalba, B.; Marín, J. M.; Cabeza, L. F.; Mehling, H. Appl. Therm. Eng. 2003, 23 (3), 251– 283.

155