Structural Characterisation of Aggrecan in Cartilaginous Tissues and Tissue Engineered Constructs a Thesis Submitted to the Univ

Structural Characterisation of Aggrecan in Cartilaginous Tissues and Tissue Engineered Constructs

A thesis submitted to the University of Manchester for the degree of Doctor of Philosophy in the Faculty of Biology, Medicine and Health

2017

Russell J Craddock

The School of Biological Sciences Division of Cell Matrix Biology and Regenerative Medicine

List of Contents List of Contents ...... 2 List of Figures ...... 6 List of Tables ...... 8 List of Equations ...... 8 List of Conference Presentations and Publications ...... 8 List of Abbreviations ...... 9 Abstract ...... 10 Declaration ...... 11 Copyright Statement ...... 11 Acknowledgements ...... 12 1 Chapter 1: General Introduction ...... 13 1.1 Overview ...... 14 1.1.2 Intervertebral Disc Biology ...... 15 1.1.3 Intervertebral Disc Macrostructure...... 16 1.1.3.1 Nucleus Pulposus ...... 16 1.1.3.2 Annulus Fibrosus ...... 16 1.1.3.3 Cartilage End-plates ...... 17 1.1.4 Articular Cartilage ...... 17 1.1.5 Structural differences between nucleus pulposus and articular cartliage ...... 18 1.1.6 Extracellular Matrix Molecules and Assemblies in Intervertebral Discs ...... 19 1.1.6.1 Collagen ...... 19 1.1.6.2 Proteogylcans ...... 20 1.1.6.3 Aggrecan ...... 21 1.1.6.3.1 Transcription of Aggrecan ...... 21 1.1.6.3.2 Aggrecan Core Protein Structure...... 22 1.1.6.3.3 Biosynthesis of Aggrecan ...... 24 1.1.6.3.4 Aggrecan Functionality and Protein Interactions ...... 27 1.1.6.3.5 Tissue-specific Variability in Aggrecan Structure ...... 27 1.1.6.3.6 Fragmentation/degradation of aggrecan ...... 29 1.2 Hypothesis and aims of the project ...... 33 2 Chapter 2: General Methods ...... 34 2.1 Cartilaginous tissue collection ...... 35 2.1.1 Native tissue sample collection ...... 35 2.1.2 Isolation and culture of nucleus pulposus cells ...... 35 2.1.3 Encapsulation of nucleus pulposus cells in a type I collagen hydrogel ...... 36 2.1.4 Chondrogenic stimulation of encapsulated nucleus pulposus cells with GDF6/TGFβ3 ...... 36 2.2 Isolation and characterisation of aggrecan ...... 36 2.2.1 Aggrecan isolation by dissociative caesium chloride centrifugation ...... 36 2.2.2 Preparation of aggrecan for structural characterisation ...... 38 2.2.3 Structural Characterisation of aggrecan by atomic force microscopy and multi- angle light scattering ...... 41 2.3 Histological assessment of tissue composition and collagen fibril alignment ...... 45 2.3.1 Histological staining ...... 45 2.3.2 Characterisation of collagen fibril ultrastructure by atomic force microscopy ...... 46 2.4 Measurement of micro-mechanical compressive stiffness by atomic force microscopy indentation ...... 48 2.5 Relative gelatinase activity determined by in situ gelatinase zymography ...... 50 2.6 Computational modelling of aggrecan packing ...... 51

2.7 Gene expression analysis using real-time quantitative-PCR ...... 52 2.7.1 Isolation of RNA from NP cell constructs ...... 52 2.7.2 Reverse transcription of RNA ...... 53 2.7.3 Assessment of NP marker gene expression with real-time quantitative-PCR ...... 53 2.8 Statistical Analysis ...... 54 3 Chapter 3: Characterisation of aggrecan and collagen structure in young, healthy cartilaginous tissues ...... 55 3.1 Introduction ...... 56 3.1.2 Hypothesis and aims ...... 57 3.2 Materials and methods ...... 58 3.2.1 Experimental design ...... 58 3.3 Results ...... 59 3.3.1 The ultrastructure of intact aggrecan monomers was tissue-specific ...... 59 3.3.2 Most isolated aggrecan molecules were fragmented and structurally heterogeneous ...... 59 3.3.3 Aggrecan fragmentation was not associated with aberrant structural or mechanical tissue remodelling or excessive protease activity ...... 62 3.3.4 Aggrecan fragmentation was predicted to modulate molecular packing density .... 68 3.4 Discussion ...... 70 3.5 Conclusions ...... 75 4 Chapter 4: The effect of culture conditions on the nanostructure of aggrecan and mechanical properties in intervertebral disc tissue engineered constructs ...... 76 4.1 Introduction ...... 77 4.1.2 Hypothesis and aims ...... 79 4.2 Materials and methods ...... 80 4.2.2 Experimental design ...... 80 4.3 Results ...... 82 4.3.1 The effect of oxygen on aggrecan synthesis and structure ...... 82 4.3.1.1 Oxygen tension did not affect the proportion of intact aggrecan, but did affect molecular size ...... 82 4.3.1.2 Gene expression aggrecan was reduced in hypoxic conditions compared to normoxia ...... 84 4.3.1.3 Extracellular matrix deposition was more homogeneous and fibrillar collagen content was substantially lower in hypoxic compared to normoxic conditions ...... 86 4.3.1.4 Hypoxia treated tissue engineered intervertebral disc constructs exhibited mechanical properties similar to native NP tissue ...... 87 4.3.1.5 Oxygen tension did not influence protease expression ...... 89 4.3.1.6 In situ gelatin analysis showed gelatinase activity was greater in hypoxia ...... 91 4.3.2 The effect of growth factor on aggrecan synthesis and structure ...... 93 4.3.2.1 Newly synthesised aggrecan molecules were less fragmented in the presence of GDF6 ...... 93 4.3.2.2 The ultrastructure of intact aggrecan monomers was similar between TGFβ3 and GDF6 treated tissue engineered intervertebral disc constructs ...... 96 4.3.2.3 collagen II and aggrecan gene expression was reduced in the presence of growth factors ...... 98 4.3.2.4 Extracellular matrix deposition was similar between growth factor treated TE IVD constructs, but exogenous growth factors suppressed fibrillar collagen organisation ...... 100 4.3.2.5 Both GDF6 and TGFβ3 reduced the construct stiffness ...... 100

4.3.2.6 Gene expression analysis showed greater ADAMTS5 and MMP13 expression in +GDF6 tissue engineered intervertebral disc constructs ...... 102 4.3.2.7 In situ gelatin analysis showed similar gelatinase activity in both growth factor treated and untreated tissue engineered intervertebral disc constructs ...... 104 4.3.3 Manipulation of cell culture conditions replicated the extracellular matrix and mechanical properties of native nucleus pulposus tissue ...... 105 4.3.3.1 In the absence of growth factor, nucleus pulposus cell constructs were structurally and mechanically dissimilar to native tissue ...... 105 4.3.3.2 Hypoxia treated nucleus pulposus cells did not synthesise tissue engineered intervertebral disc tissue similar to native nucleus pulposus tissue ...... 107 4.3.3.3 Growth factor treated tissue engineered intervertebral disc construct derived aggrecan was structurally similar native nucleus pulposus ...... 108 4.4 Discussion ...... 110 4.4.1 The effect of oxygen tension and growth factors on stimulating NP cells to synthesise aggrecan similar to native nucleus pulposus tissue ...... 110 4.4.1.1 Normoxia and hypoxia ...... 110 4.4.1.2 TGFβ3 and GDF6 ...... 111 4.4.2 Aggrecan fragmentation may be due to the activity of constituent aggrecanase activity ...... 113 4.5 Conclusions ...... 113 5 Chapter 5: Characterisation of aggrecan structure in young, healthy porcine tissues ...... 115 5.1 Introduction ...... 116 5.1.1 The origin and phenotype of notochordal cells ...... 116 5.1.2 Hypothesis and aims ...... 117 5.2 Materials & methods ...... 118 5.2.1 Experimental design ...... 118 5.3 Results ...... 119 5.3.1 Porcine nucleus pulposus was smaller than porcine articular cartilage derived intact aggrecan ...... 119 5.3.2 NP porcine aggrecan monomers were similar to bovine NP derived intact aggrecan ...... 119 5.3.3 Porcine nucleus pulposus aggrecan fragments were smaller than porcine articular cartilage ...... 121 5.3.4 Porcine nucleus pulposus aggrecan fragments were similar in size compared to bovine nucleus pulposus ...... 122 5.3.5 Aggrecan fragmentation was not associated with aberrant structural remodelling or excessive protease activity ...... 124 5.4 Discussion ...... 127 5.5 Conclusions ...... 129 6 Chapter 6: General Discussion and Future Work ...... 130 6.1 Study rationale ...... 131 6.2 Key outcomes: ...... 132 6.3 Conclusions ...... 133 6.3 Future work ...... 134 6.3.1 Completion of immediate studies ...... 134 6.3.1.1 Chapter 3 ...... 134 6.3.1.2 Chapter 4 ...... 136 6.3.1.3 Chapter 5 ...... 137

6.3.2 Beyond the remit of this study: creating a viable tissue engineered intervertebral disc replacement ...... 137 7 Chapter 7: References ...... 139

Final Word Count: 40,463

List of Figures Figure 1.1. Structure of an adult human intervertebral disc...... 15 Figure 1.2. Collagen fibril orientation and cell morphology of articular cartilage...... 18 Figure 1.3. Fibrillar collagen structure...... 20 Figure 1.4. Alternative splice variants of the aggrecan globular-3 domain...... 22 Figure 1.5. Aggrecan structure and molecular associations...... 23 Figure 1.6. Biosynthesis of sulfated glycosaminoglycans...... 26 Figure 1.7. Major aggrecanase cleavage sites along the aggrecan core protein...... 32 Figure 2.1. Aggrecan was only present in fractions with a density of greater than 1.54 g/mL...... 38 Figure 2.2. Optimisation of aggrecan storage buffer: water vs tri-buffered saline vs phosphate buffered saline...... 39 Figure 2.3. Protease inhibitors were only required in the extraction buffer...... 40 Figure 2.4. Effect of isolation procedure on intact aggrecan structure...... 40 Figure 2.5. Ultrastructural characterisation of aggrecan...... 42 Figure 2.7. Definition of minimum and maximum force boundaries...... 49 Figure 2.8. Force plot with a linearised fit...... 50 Figure 3.1. Surface-adsorbed intact aggrecan monomers were structurally dissimilar in nucleus pulposus and articular cartialge tissues...... 60 Figure 3.2. Both articular cartilage and nucleus pulposus were composed primarily of fragmented aggrecan...... 61 Figure 3.3. Polydispersity of solution aggrecan indicated that it was highly fragmented and structurally heterogeneous...... 62 Figure 3.4. Young, mature tissues appeared histologically healthy...... 63 Figure 3.5. Collagen fibril ultrastructure and architecture in mature articular cartilage and nucleus pulposus...... 65 Figure 3.6. Articular cartilage was significantly stiffer than nucleus pulposus...... 66 Figure 3.7. Low levels of matrix metalloproteinase activity in young healthy bovine articular cartilage and nucleus pulposus tissue...... 67 Figure 3.8. Simulation of aggrecan packing...... 69 Figure 3.9. Proteoglycan structure in cartilaginous extracellular matrix...... 74 Figure 4.1. Newly synthesised aggrecan was mostly fragmented in nomoxia and hypoxia...... 83 Figure 4.2. Normoxia derived aggrecan had greater glycosaminoglycan content per unit length of core protein...... 84 Figure 4.3. Chondrogenic gene expression in Day 28 constructs...... 85 Figure 4.4. Histological staining of native NP tissue and tissue engineered intervertebral disc tissue...... 87 Figure 4.5. Hypoxia treated constructs were more mechanically heterogeneous and stiffer compared to normoxia...... 89 Figure 4.6. Aggrecanase gene expression in hypoxia treated constructs after 28 days culture...... 90

Figure 4.7. High levels of matrix metalloproteinase activity in the absence of a growth factor in hypoxia...... 92 Figure 4.8. Newly synthesised aggrecan was mostly fragmented...... 94 Figure 4.9. GDF6 treated tissue engineered intervertebral disc construct derived aggrecan had greater glycosaminoglycan content per unit length of core protein...... 95 Figure 4.10. Surface-adsorbed intact aggrecan monomers were structurally similar in +TGFβ3 and +GDF6 tissue engineered intervertebral disc constructs...... 97 Figure 4.11. Chondrogenic gene expression in growth factor treated and untreated constructs cultured for 28 days...... 99 Figure 4.12. Constructs were more mechanically homogenous in the presence of growth factor...... 101 Figure 4.13. +GDF6 and +TGFβ3 tissue engineered intervertebral disc constructs were mechanically similar, but are more compliant than no growth factor treated constructs. 102 Figure 4.14. Aggrecanase gene expression in growth factor treated and untreated constructs cultured for 28 days...... 103 Figure 4.15. Similar levels of matrix metalloproteinase activity in both the presence and absence of a growth factor...... 104 Figure 4.16. Native nucleus pulposus and newly synthesised tissue engineered intervertebral disc aggrecan was mostly fragmented...... 106 Figure 4.17. +GDF6 and +TGFβ3 tissue engineered intervertebral disc constructs were mechanically similar to nucleus pulposus tissue...... 107 Figure 4.20. Surface-adsorbed intact aggrecan monomers were structurally similar between nucleus pulposus, +TGFβ3, and +GDF6 tissue engineered intervertebral disc constructs...... 109 Figure 5.1. Surface-adsorbed intact aggrecan monomers were structurally dissimilar in porcine nucleus pulposus and articular cartilage tissues...... 120 Figure 5.2. Porcine nucleus pulposus and articular cartilage derived intact aggrecan was structurally similar to bovine nucleus pulposus and articular cartilage...... 121 Figure 5.3. Porcine nucleus pulposus derived aggrecan had greater glycosaminoglycan content per unit length of core protein...... 122 Figure 5.4. Both porcine articular cartilage and nucleus pulposus were composed primarily of fragmented aggrecan...... 123 Figure 5.5. Aggrecan fragmentation was similar between porcine and bovine nucleus pulposus...... 124 Figure 5.6. Skeletally mature porcine tissues appeared histologically healthy...... 125 Figure 5.7. Low levels of matrix metalloproteinase activity in healthy porcine articular cartilage...... 126

List of Tables Table 1.1. Summary of intact aggrecan core protein contour length described in the literature...... 28 Table 1.2. Aggrecan cleavage sites and proteases responsible for cleavage...... 31 Table 2.1. Primer sequences and working concentrations...... 54 Table 3.1. Predicted protease specific amino acid sequence cleavage sites of aggrecan. .. 72 Table 3.2. Predicted protease specific amino acid sequence cleavage sites of collagen II. 73 Table 4.1. Experimental design for the manipulation of cell culture conditions...... 80

List of Equations Equation 1: CsCl calculation…………………………………………………..…………………37 Equation 2: Sneddon Equation…………………………………………………………………....49 Equation 3: Linearised Sneddon Equation……………………………………………………...49

List of Conference Presentations and Publications Poster Presentation, Orthopaedic Research Symposium, Las Vegas, USA, March 2015. Title: Aggrecan Ultrastructure is Tissue-Specific

Oral and Poster Presentation, Philadelphia Spine Research Symposium, Philadelphia, USA, November 2015. Title: Aggrecan Ultrastructure is Tissue-Specific

Oral and Poster Presentation, British Society for Matrix Biology, Chester University, April 2016. Title: Aggrecan Ultrastructure is Tissue-Specific

Poster presentation, Postgraduate Research Summer Showcase Symposium, The University of Manchester, August 2016. Title: Molecular structure and mechanical properties of intervertebral disc and articular cartilage

Oral Presentation, Tissue and Cell Engineering Society Annual Meeting, Manchester Metropolitan University, July 2017 Title: Nanostructure of aggrecan in native tissue and engineered intervertebral discs

List of Abbreviations AC Articular Cartilage MMP Matrix Metalloproteinase iAC Immature Articular Cartilage NC Notochordal Cell ACAN Aggrecan (gene) Neu5Ac N-acetylneuraminic Acid ADAMTS A Distintergrin and Matrix NGF No Growth Factor Metalloproteinase AF Annulus Fibrosis NP Nucleus Pulposus AFM Atomic Force Microscopy iNP Immature Nucleus Pulposus BCP 1-Bromo-3-chloropropane PBS Phosphate Buffered Saline BMP Bone Morphogenic Protein PCR Polmerase Chain Reaction C Carbon PSR Picrosirius Red CCP Complement Regulatory SEC- Size Exclusion Protein MALS Chromatography Multi Angle Light Scattering CLD C-type Lectin Domain SD Standard Deviation COL Collagen (gene) SOX9 SRY Box Protein 9 (gene) CS Chondroitin Sulfate TBS Tris-buffered Saline CsCl Caesium Chloride TE Tissue Engineered ECM Extracellular Matrix TEM Transmission Electron Microscopy EGF Epidermal Growth Factor TGFβ Transforming Growth Factor β GAG Glycosaminoglycan TNFα Tumour Necrosis Factor α sGAG Sulfated Glycosaminoglycan XYL Xylose Gal Galactose GalNAc N-acetyl-galactosamine GAPDH Glyceraldehyde 3-phosphate dehydrogenase (gene) GDF Growth Differentiation Factor GlcA Glucuronic Acid GlcNAc N-acetyl-glucosamine HA Hyaluronic Acid IVD Intervertebral Disc IVDD Intervertebral Disc Degeneration iGAGL Individual glycosaminoglycan chain length KS Keratin Sulfate LBP Lower Back Pain LCP Core protein Length LGAG Glycosaminoglycan binding region length LTBP Latent Transforming Growth Factor β Binding Protein MAGAG molecular area of the aggrecan monomer covered by GAG

Abstract Collagen II and the proteoglycan aggrecan are key extracellular matrix (ECM) proteins in cartilaginous tissues such as the intervertebral disc (IVD). Given the functional role that these structural and functional proteins have in the IVD, ECM in tissue engineered intervertebral disc (TE IVD) constructs needs to recapitulate native tissue. As such, there is a need to understand the structure and mechanical function of these molecules in native tissue to inform TE strategies. The aims here were to characterise aggrecan and collagen II using atomic force microscopy (AFM), size-exclusion chromatography multi angle light scattering (SEC-MALS), histology, quantitative PCR, nanomechanical and computational modelling in: (i) skeletally immature and mature bovine articular cartilage (AC) and nucleus pulposus (NP), (ii) TE IVD constructs cultured in hypoxia or treated with transforming growth factor beta [TGFβ3] or growth differentiation factor [GDF6]), and (iii) porcine AC and NP tissue.

No variation in collagen II structure was observed although the proportion of organised fibrillar collagen varied between tissues. Both intact (containing all three globular domains) and non-intact (fragmented) aggrecan monomers were isolated from both AC and IVD and TE IVD constructs. Mature intact native NP aggrecan was ~60 nm shorter (core protein length) compared to AC. In skeletally mature bovine NP and AC tissue, most aggrecan monomers were fragmented (99% and 95%, respectively) with fragments smaller and more structurally heterogeneous in NP. Similar fragmentation was observed in skeletally immature bovine AC (99.5%), indicating fragmentation occurs developmentally at an early age. Fragmentation was not a result of enhanced gelatinase activity. Aggrecan monomers isolated from notochordal cell rich porcine NP were also highly fragmented, similar to bovine NP. Application of a computational packing model suggested fragmentation may affect porosity and nutrient transfer. The reduced modulus was greater in AC than NP (497 kPa and 76.7 kPa, respectively) with the difference likely due to the organisation and abundance of ECM molecules, rather than individual structure. Growth factors (GDF6 and TGFβ3), and not oxygen tension treated TE IVD constructs were structurally (with >95% fragmented monomers), histologically and mechanically (GDF6: 60.2 kPa; TGFβ3; 69.9 kPa) similar to native NP tissue (76.7 kPa) and there was evidence of gelatinase activity. To conclude, these results show that the ultrastructure of intact aggrecan was tissue and cell dependent, and could be modified by manipulation of cell culture conditions, specifically GDF6 which may play a role in aggrecan glycosylation.

Declaration No portion of the work referred to in the thesis has been submitted in support of an application for another degree or qualification of this or any other university or other institute of learning.

Copyright Statement i. The author of this thesis (including any appendices and/or schedules to this thesis) owns certain copyright or related rights in it (the “Copyright”) and s/he has given The University of Manchester certain rights to use such Copyright, including for administrative purposes.

ii. Copies of this thesis, either in full or in extracts and whether in hard or electronic copy, may be made only in accordance with the Copyright, Designs and Patents Act 1988 (as amended) and regulations issued under it or, where appropriate, in accordance Presentation of Theses Policy You are required to submit your thesis electronically Page 11 of 25 with licensing agreements which the University has from time to time. This page must form part of any such copies made.

iii. The ownership of certain Copyright, patents, designs, trademarks and other intellectual property (the “Intellectual Property”) and any reproductions of copyright works in the thesis, for example graphs and tables (“Reproductions”), which may be described in this thesis, may not be owned by the author and may be owned by third parties. Such Intellectual Property and Reproductions cannot and must not be made available for use without the prior written permission of the owner(s) of the relevant Intellectual Property and/or Reproductions.

iv. Further information on the conditions under which disclosure, publication and commercialisation of this thesis, the Copyright and any Intellectual Property and/or Reproductions described in it may take place is available in the University IP Policy (see http://documents.manchester.ac.uk/DocuInfo.aspx?DocID=2442 0), in any relevant Thesis restriction declarations deposited in the University Library, The University Library’s regulations (see http://www.library.manchester.ac.uk/about/regulations/) and in The University’s policy on Presentation of Theses

Acknowledgements Firstly, I would like to thank my family, whose emotional, educational and financial support over the years has helped me immeasurably. Without your care, I would never have developed the critical, scientific mind I have today, always questioning the world around me, wanting to understand how everything works, and trying to make the word a better place. No words can express how grateful I am.

I would also like to take this opportunity to thank my late grandparents, in particular my grandfather; whose long standing suffering from osteoporosis and osteoarthritis was the reason I chose to specialise in musculoskeletal disease. Without your support, both financially and otherwise, I may never have succeeded as I have.

Without the support of my close network of friends/colleagues, I probably wouldn’t have lasted a year, so I’d like to say thank you to the following friends and colleagues for their support: Christopher Storer, James Amphlett, Tim Hughes, Kim Hughes, Joanna Gould, Melody Obeng, Zuzanna Bogdanowicz, Pauline Baird, Sonal Patel, Matthew Humpreys, Julen Mancebo, Abbie Binch, Lizzie Ward, Nigel Hodson, James McConnell, Sarah Hibbert, Patrick Costello, and Amaar Razaq.

Last, but certainly not least, I’d like to say thank you the best supervisors I could have ever asked for; Professor Judith Hoyland, Dr Michael Sherratt and Professor Sarah Cartmell. Without your tutorage, encouragement, dedication, expertise and support I would never have had as much opportunity and of progressed as successfully as I have.

1 Chapter 1: General Introduction

1.1 Overview Lower back pain (LBP) is an age-related musculoskeletal disorder that most commonly affects working-age (35—60-year-old) adults (NICE, 2009). Although it is not a life threatening condition, LBP is a painful and debilitating disorder affecting an estimated 80 per cent of adults at some point in their life (Walker, 2000). Treating this disorder is a massive economic burden in the UK; when considering the cost of NHS care, community health, social services, employment and informal care, the total expenditure exceeds £12 billion pounds per year (Maniadakis and Gray, 2000). This is expected to rise with our ever increasingly ageing population unless a corrective therapy is found (Gorman et al., 2013). Lower back pain has a variety underlying causes including mechanical (80—90%; e.g. intervertebral disc degeneration (IVDD)), neurogenic (5—15%; e.g. herniated disc), non- mechanical spinal conditions (1—2%; e.g. inflammatory arthritis) and referred visceral pain (1—2%; e.g. gastrointestinal disease) (Cohen et al., 2008). IVDD is the most common source of LBP (Malik et al., 2013), which accounts for approximately 40 per cent of cases (Aladin et al., 2010, Cheung et al., 2009, Colombier et al., 2014).

IVDD currently has no cure, and treatments primarily aim to manage the disorder through the relief of symptomatic pain using various painkillers (Cohen et al., 2008). Alternate treatments are available, for example low level laser therapy, therapeutic ultrasound and surgical intervention (NICE, 2009, Cohen et al., 2008). IVDD is known to originate in the central region of the IVD, known as the nucleus pulposus (NP) and end-stage IVDD requires discectomy (removal of the IVD) and/or spinal fusion, or replacement of the IVD with a prosthesis (NICE, 2009, Bertagnoli et al., 2006). However, discectomy and spinal fusion do not treat the underlying pathogenesis, therefore there is a clear need for novel therapies that target the underlying degenerative process to induce endogenous repair by restoring the cell population and function of the tissue. Tissue engineered (TE) IVD constructs are currently being developed for the treatment of IVDD (Clarke et al., 2014), which aims to replicate both the composition and function of native tissue by promoting synthesis of key extracellular matrix (ECM) components (aggrecan and collagen type II) in order to restore mechanical functionality to the damaged disc (Clarke et al., 2014). This is achieved by manipulating factors such as cell type and cell culture conditions when creating constructs. However, to date, the effect of cell culture conditions has been studied predominantly in the context of gene expression and gross characterisation of ECM components through histological, immunohistochemical and biochemical analysis. It is not clear to what extent

14 the molecular structure of key ECM components (e.g. the proteoglycan aggrecan and collagen II) dictate variability in tissue morphology, composition and mechanical functionality, which exists between cartilaginous tissues (e.g. hyaline/articular cartilage [AC] and NP). An increasing body of evidence is highlighting that aggrecan structure is variable and can be altered by manipulating cell culture conditions which ultimately would influence the mechanical properties of any engineered construct. However, the key factors that may influence structure are not known. Additionally, in order to understand what we want to recapitulate in IVD tissue engineered constructs, we must first fully characterise native tissue-derived aggrecan, to ensure we are synthesising aggrecan with the appropriate molecular structure, to create a mechanically viable NP replacement.

1.1.2 Intervertebral Disc Biology IVDs join the spinal vertebrae together and hence enable extension, mobility and flexibility of the spine. They absorb and uniformly redistribute loads created by muscle movement and body mass that are passed through the spinal column (Raj, 2008). In order to carry out these functions the IVD has a unique structure comprised of three morphologically distinct regions ((Smith et al., 2011)); Fig 1.1).

Figure 1.1. Structure of an adult human intervertebral disc. (A) Cross-section of an intervertebral disc highlighting the central gelatinous nucleus pulposus NP tissue contained within the outer rings of the annulus fibrosus (AF) and the thin horizontal hyaline cartilage endplates (approximately 1 mm thick) that lie superior and inferior to the NP and AF. (B) Three-dimensional representation of the AF lamellar structure which contains the NP. IVD’s are typically 4 cm in diameter and 7-10 mm thick across the anterior-posterior plane (Raj, 2008). Image adapted from Smith et al., 2011.

1.1.3 Intervertebral Disc Macrostructure 1.1.3.1 Nucleus Pulposus In the mature IVD the central nucleus pulposus NP is an ECM-rich tissue which is comprised primarily of collagen II, the proteoglycan aggrecan (substituted with sulfated glycosaminoglycans [GAGs] and non-sulfated oligosaccharides), and traces of elastin and fibrillin-1 (Raj, 2008, Smith et al., 2011, Yu et al., 2007, Yu et al., 2005). Collagen II forms a disorganised fibrillar network, providing the NP molecular structure with tensile strength to withstand the osmotic pressure of the NP which is generated by the hydrophilic potential of aggrecan. Long, straight elastic fibres are arranged in a radial pattern, which in combination with aggrecan enables the NP to regain its shape after deformation under compressive and shear loads (Raj, 2008, Smith et al., 2011, Yu et al., 2002, Yu et al., 2007, Yu et al., 2005). Various proteoglycans can be found in the ECM, the most predominant of which is aggrecan. Other proteoglycans present within the IVD include versican, biglycan and decorin whose functions include regulation of collagen fibril diameter/fusion (Smith et al., 2011, Raspanti et al., 2007, Watanabe et al., 2005, Zhang et al., 2005).

The human NP is unique compared to other cartilaginous tissues in the body due to the high proportion of aggrecan to collagen as determined by biochemical assay (27:1) (Mwale et al., 2004), as it is required to resist dynamic compressive and shear loads. On the other hand, the cartilage-like end plates (i.e. with a 2:1 ratio of aggrecan to collagen (Mwale et al., 2004)) function to provide frictionless movement between the IVD and adjacent vertebrae and provide greater tensile strength. Human collagen and aggrecan have a half-lives of approximately 95—215 years (Sivan et al., 2008) and 12 years respectively in the human IVD (Maroudas et al., 1998, Sivan et al., 2006). Whilst the rate of turnover of collagen and aggrecan differs between tissues (e.g collagen II has a half-life of 95—215 years in IVD and 15—95 years in human skin) in general ECM proteins are long lived and therefore prone to damage compared to intracellular proteins (Naylor et al., 2011, Sivan et al., 2008). The human NP is comprised of resident chondrocyte-like cells which are sparsely distributed at approximately 5000 cells/mm throughout the ECM (Raj, 2008).

1.1.3.2 Annulus Fibrosus The NP is surrounded by a highly organised fibrous structure known as the annulus fibrosus (AF), which functions to provide tensile strength and to hold the NP in position during compression and connect the IVD to adjacent vertebrae (Raj, 2008, Colombier et al., 2014,

Smith et al., 2011). The highly organised outer AF structure consists of bundles of collagen I fibrils that form 15—25 rings of distinct concentric lamellae “orientated at 60 degrees to the vertical axis that alternate left and right to adjacent lamellae” resulting in an angle-ply structure (Raj, 2008, Smith et al., 2011). However, in the less organised and fibrous inner AF structure, there is a transition from collagen I to collagen II with an increasing proteoglycan content (Smith et al., 2011). A dense network of elastic fibres exists between the concentric rings of lamellae and individual elastic fibres orientate parallel to collagen fibres that comprise each lamella (Raj, 2008, Smith et al., 2011, Yu et al., 2002, Yu et al., 2007, Yu et al., 2005). Unlike the NP, cells distributed within the AF display a fibroblast- like phenotype, and synthesise mainly collagen I and are positioned parallel along the collagen fibres at a greater density of approximately 9000 cells/mm3 (Colombier et al., 2014). Unlike other cartilaginous tissues, cells in both regions of the IVD extend thin cytoplasmic projections approximately 30 µm in length from the cell body, which may function as mechanosensors and cell-to-cell communication (Raj, 2008, Errington et al., 1998, Bruehlmann et al., 2002).

1.1.3.3 Cartilage End-plates Superior and inferior to the NP and AF are the hyaline cartilage endplates whose function is to act as an interface between the IVD’s and adjacent vertebral bodies regulating nutrient diffusion between both tissues (Figure 1.1; Raj, 2008, Smith et al., 2011). Endplate collagen II fibres run parallel to the vertebral body and anchor into the IVD, connecting both together (Raj, 2008, Smith et al., 2011). Although the endplates act as an interface between the IVD’s and vertebral bodies, collagen fibres from the outer AF penetrate directly into the bone (Raj, 2008, Smith et al., 2011).

1.1.4 Articular Cartilage AC covers knee joints (and most other joint positions) within the body where its role is to resist compressive forces and provide frictionless movements of joints. Similar to NP, AC is predominantly comprised of a collagen II fibrillar network providing tensile strength interspersed with aggrecan-hyaluronic acid (HA) macromolecular aggregates that generate an osmotic potential providing resistance against compressive forces (Goldring, 2000, Fischer et al., 2010). However, unlike NP, in AC the orientation of collagen fibres is semi- ordered and depth dependent (Fig. 1.2); fibres run parallel to the surface in the cartilage superficial zone, followed by disorganised, randomly orientated fibrils in the transitional

17 zone and fibres that arrange perpendicular to the surface in the deep zone (Inamdar et al., 2017). In comparison to NP tissue, where NP cells are randomly, sparsely distributed, AC chondrocyte distribution is depth-dependent. In the AC superficial zone, chondrocytes are flattened disc shapes due to low proteoglycan content and dense fibrillar collagen network, whereas in the transitional zone where proteoglycan content increases and the collagen network is more disorganised, chondrocytes are more rounded (Poole, 1997). In the deep zone, chondrocytes are arranged in columns due to the high proteoglycan content and radially organised collagen fibrils (Poole, 1997).

Figure 1.2. Collagen fibril orientation and cell morphology of articular cartilage. Schematic diagram illustrating the depth-dependent change in type II collagen structure and cell morphology in articular cartilage (Mansfield et al., 2015).

1.1.5 Structural differences between nucleus pulposus and articular cartliage There are three main structural differences between NP and AC ECM, namely: (i) the proportion of aggrecan to collagen (Chapter 1.1.3.1) (Mwale et al., 2004), (ii) the variability in the structure of aggrecan between (NP and AC) tissues (Buckwalter et al., 1989, Kopesky et al., 2010, Lee et al., 2010, Lee et al., 2013), and (iii) the variation in the orientation of collagen fibres between (NP and AC) tissues (Chapter 1.1.3.1 and 1.1.4). NP, AF and AC function is determined by the nanostructure and overall macrostructure of key ECM components, including but not limited to collagen II and aggrecan. The nanostructure includes molecules and proteins that work together to form larger more complex bodies and

18 other tissue components. These interactions will be discussed in greater detail in the following sections.

1.1.6 Extracellular Matrix Molecules and Assemblies in Intervertebral Discs 1.1.6.1 Collagen Assembled collagen I and II monomers are characterised by a helical structure (van der Rest and Garrone, 1990, Shoulders and Raines, 2009). Every third amino acid is glycine, which resides within the centre of the helix (van der Rest and Garrone, 1990, Shoulders and Raines, 2009). The first and second amino acids are typically proline and hydroxyproline, which maintain the integrity of the helical structure (van der Rest and Garrone, 1990, Shoulders and Raines, 2009); this results in a rigid rod-like structure with regions of flexibility (van der Rest and Garrone, 1990, Shoulders and Raines, 2009). Three of these collagen I or collagen II helical monomers interact to form a triple helix (Figure 1.3) (van der Rest and Garrone, 1990, Shoulders and Raines, 2009). These triple helices can form multiple interactions with surrounding molecules as 70% of the amino acid residues project outwards from the structure (van der Rest and Garrone, 1990, Shoulders and Raines, 2009). The presence of charged and hydrophilic amino acid residues between these triple helical domains results in packing into fibrils; triple helices overlap with a staggered structure by integral multiples of 67 nm (van der Rest and Garrone, 1990, Shoulders and Raines, 2009, Sherman et al., 2015). Collagen fibrils are bundled to form fibres (Figure 1.3d). This complex nanostructure provides collagen I and II with its characteristic properties: thermo- stability, mechanical strength, and formation of interactions with other molecules (Shoulders and Raines, 2009). It is also enables the collagen I and II fibrillar networks to form a scaffold that supports and prevents loss of the macromolecular aggrecan complexes and is vital in maintaining the microenvironment of the IVD, as a major mechanical component of IVD ECM (Shoulders and Raines, 2009).

Figure 1.3. Fibrillar collagen structure. (a) Collagen triple-helix. (b) Schematic representation of the triple-helix of collagen comprised of three procollagen chains. (c) The arrangement and size of overlapping collagen triple-helices into collagen fibrils. (d) The hierarchical arrangement of collagen fibrils into collagen fibres, fascicles and tendon. Adapted from Sherman et al., (2015) and Mienaltowski and Birk (2014).

1.1.6.2 Proteogylcans Proteoglycans act as both signalling and structural molecules in both cartilaginous and non- cartilaginous tissues. Proteoglycans consist of a core protein backbone varying in size from approximately 10 kDa to 500 kDa (depending on the proteogylcan) substituted by unbranched polysaccharide moieties. There can be as few as one or two of these GAGs, or hundreds. GAGs also differ in length and charge due to varying chemical modification of the sugars or alternative splicing variants of the proteogylcan core protein. One thing all proteoglycans have in common is an overall net negative charge generated by GAG

20 substitution, enabling them to attract large quantities of positively charged molecules such as growth factors, cytokines, and water. Highly negatively charged sulphate and carbohydrate groups incorporated into GAG structure attracts counter ions which in turn create an ion deficiency in surrounding areas, attracting water molecules to the proteoglycan. By attracting and trapping large quantities of water molecules, particularly larger proteoglycans (i.e. aggrecan) maintain hydration which is important for transport of solutes and chemical reactions and can generate an osmotic potential/swelling pressure, so that when compressed, ‘forests’ of aggrecan are able to resist considerable force.

Proteoglycans are not merely simple scaffolds with GAG attachment sites; they include biologically active functional domains which confer specific functions depending on the proteoglycan (Nastase et al., 2012). Biglycan and decorin regulate ECM structure by stabilising and organising collagen fibres in addition to interacting with various growth factors, for example decorin sequesters TGFβ3 and biglycan interacts with bone morphogenic protein (BMP)-2, -4 and -6, TGFβ3, tumour necrosis factor (TNF)α and Wnt induced signalling protein-1 (Neill et al., 2012, Nastase et al., 2012, Hildebrand et al., 1994, Tufvesson and Westergren-Thorsson, 2002, Mochida et al., 2006, Desnoyers et al., 2001, Chen et al., 2004). These multifunctional hybrid molecules are clearly vital to the homeostasis and functionality of tissues, not only providing structural support, but biological queues controlling cell activity. The most abundant of these and therefore the most structurally relevant in cartilaginous tissues is aggrecan, which is the focus of this study.

1.1.6.3 Aggrecan 1.1.6.3.1 Transcription of Aggrecan Unprocessed human aggrecan core protein is a 2415 amino acid protein with a 16 amino acid signal peptide (Ilic et al., 1995). Alternative splicing can result in 3 isoforms; P16112- 1, P16112-2, p16112-3 (Fig. 1.4). Isoform-1 is the most common sequence, whereas isoforms-2 and -3 are missing the amino acid sequence 2163—2200 (epidermal growth factor [EGF]-like domain) and 2330-2390 (an evolutionarily conserved short consensus repeat) (Doege et al., 1991, Dudhia et al., 1996, UniProt, 2014b); the absence of these sequences have no known functional consequences. Unprocessed bovine aggrecan is a 2364 amino acid protein with a 16 amino acid signal peptide (Hering et al., 1997, UniProt, 2014a). The structure of bovine articular cartilage aggrecan is also variable; alternative splicing results in 2 isoforms; P13608-1 and P13608-2 (Hering et al., 1997, UniProt, 2014a). Similar

21 to human aggrecan, isoform -1 is the most common sequence, whereas isoform-2 is missing the amino acid sequence 2114—2150 (Hering et al., 1997, UniProt, 2014a). Full-length human aggrecan is 55 amino acids longer than bovine aggrecan; consequently, human aggrecan has a longer LCP than bovine aggrecan. To our knowledge, alternative splicing of aggrecan has only been assessed in AC-derived chondrocytes. The complete amino acid structure of porcine aggrecan core protein has yet to be assessed.

Figure 1.4. Alternative splice variants of the aggrecan globular-3 domain. Alternative splicing of aggrecan occurs within the G3 domain. Isoform-1 is the most prevalent isoform in both bovine and human; the G3 domain consists of two epidermal growth factor (EGF)- like domains, a c-type lectin domain (CLD) and a complement regulatory protein-like module (CCP) (Doege et al., 1991, Dudhia et al., 1996, Hering et al., 1997).

1.1.6.3.2 Aggrecan Core Protein Structure Aggrecan core protein consists of a G1 domain, G1—G2 interglobular (IGD) domain, G2 domain, G2—G3 IGD and a G3 domain (Fig 1.5). The G1 domain consists of link domains (that facilitate interactions with link proteins and other link domain containing proteins) and three disulphide-bonded loop structures termed A, B, and B’. Despite the similarity between the G1 and G2 domains (the G2 domain lacks A, but contains B and B’), G2 cannot interact with HA and its functional role is unknown. The reason for this is unclear; however it may be due to glycosylation (Sivan et al., 2014). Arguably the most important part of the aggrecan

22 core protein is the G2—G3 IGD; immediately following the G2 domain is a Keratin Sulfate (KS)-rich region. This region is abundant in Serine amino acid residues that can be only be substituted by O-linked KS or unsulfated O-linked oligosaccharides (Doege et al., 1991). Adjacent are two consecutive Chondroitin Sulfate (CS) substitution regions designated CS1 and CS2. CS1 contains a repeating pattern of 19 amino acids including two CS glycine- serine residue attachment sites; however these can vary depending on the species (Doege et al., 1991). Typically there is anywhere between 13—33 repeats, with a mode of 26—28 repeats (Doege et al., 1997). Unlike the CS1 region, the CS2 region has variability in CS substitution due to much greater variability amino acid structure and is much more susceptible to proteolytic cleavage by aggrecanases, whereas the CS1 region is not (Sivan et al., 2014).

Figure 1.5. Aggrecan structure and molecular associations. HA binds to aggrecan at the N- terminal globular (G)1 domain via a type-C HA-binding domain comprised of two contiguous link modules forming a ternary complex with link protein (LP) (Day and Prestwich, 2002). The G1—G2 and G2—G3 are separated by interglobular domains with the latter functioning as a GAG binding region for KS and CS. The CS binding region is comprised of two regions termed CS1 and CS2 separated based on the variability in the amino acid repeating sequence (Hardingham and Fosang, 1992, Kiani et al., 2002, Day and Prestwich, 2002, Aspberg, 2012).

One of the most important domains in aggrecan is the multifunctional G3 domain; it functions to enable normal trafficking during translation and post-translational modification

23 of the core protein, as well as secretion into the surrounding ECM (Zheng et al., 1998). G3 interacts non-covalently with the molecular chaperone Hsp25 to facilitate G1 domain secretion, correct folding and secretion of aggrecan (Zheng et al., 1998). All G3 domains contain a c-type lectin domain that enables aggrecan to interact with fibrillin-2 through a calcium-dependent mechanism, tenascin R, tenacin-C and sulphated glycolipids (Olin et al., 2001). Fibrillin-2 has been shown to facilitate cross-linking between aggrecan-HA complexes (Olin et al., 2001) and can interact with the aggrecanase (a disintergrin and matrix metalloproteinase (MMP) with thrombospondin repeats (ADAMTS5)) (Bader et al., 2012). Fibrillin-2 and latent transforming growth factor beta (TGFβ) binding protein (LTBP)-2 are present in the IVD co-localised with fibrillin-1 resulting in microfibrils that run parallel to collagen fibrils in the lamellae of the AF. These three proteins co-localise to form dense networks around cells in the NP where it co-localises with elastin (Li et al., 2012). LTBP2 binds latent TGFβ proteins which can be activated to regulate cell growth and immune responses (Saharinen et al., 1999). The G3 domain has been reported to be absent in approximately 70% of synthesised aggrecan monomers as a result of enzymatic activity (Morgelin et al., 1988), the loss of which increases with age (Watanabe et al., 1997). Importantly, this indicates that the majority of aggrecan is fragmented and not in an intact form.

1.1.6.3.3 Biosynthesis of Aggrecan After synthesis of the core protein aggrecan is transported to the endoplasmic reticulum where glycosylation is initiated. The tetrasaccharide linker of CS is constructed in the pre-- golgi compartment or cis-golgi region and GAG polymerisation takes place as aggrecan moves through the lumen of the golgi apparatus (Garcia-Suarez et al., 2014, Kearns et al., 1991, Kearns et al., 1993, Vertel et al., 1993, Nuwayhid et al., 1986, Lohmander et al., 1989). KS and CS are covalently ‘built’ directly onto the aggrecan core protein by the action of various enzymes (Fig. 1.6). Approximately 8—10 KS chains and ~100 CS chains are present on every aggrecan monomer, each GAG is spaced ~1—1.5 nm apart and range from 14— 21 nm in length (Chandran and Horkay, 2012). CS synthesis is initiated through the formation of a tetrasaccharide linker at Serine residues (O-linked glycosylation) and KS can either be N-linked (KSI) or O-linked (KSII) (Garcia-Suarez et al., 2014). Figure 1.6 depicts the synthesis of CS and KSI and KSII chains. Sulfation of CS occurs following chain elongation at both the carbon (C)2 site of glucuronic acid (GlcA) and the C4 and/or C6 sites of N-acetyl-galactosamine (GalNAc) (Huckerby et al., 2001, Garcia-Suarez et al., 2014).

Sulfation is semi-random modulating each GAGs specificity for a substrate (Garcia-Suarez et al., 2014).The ratio of sulfation of C6 to C4 increases with age; foetal CS is predominantly 4C sulfated, whereas in mature CS C4 sulfation is replaced by C6 sulfation and becomes highly C6 sulfated (Plaas et al., 2001a, Vynios, 2014). This change is particularly evident between the ages of 10—20 years old in AC (Lauder et al., 2001), and the change in sulfation is associated with a decrease in the molecular weight of CS to decrease from 25 kD to 16 kD reflecting a decrease in overall GAG chain length (Plaas et al., 2001a). With increasing age the opposite happens in KS where GAG chain length, sulfation and fucosylation increases (Nieduszynski et al., 1990, Montplaisir, 1979, Ruter and Kresse, 1984, Plaas et al., 2001b, Lauder et al., 1998). As cartilage becomes more hypertrophic, changes to the structure of CS and KS covalently attached to the aggrecan core protein have been shown such as “chain length, chain termination, sulfate ester substitution, and substitution of CS chains with KS chains” (Nieduszynski et al., 1990, Bayliss et al., 1999, Lauder et al., 1998, Plaas et al., 1998, Plaas et al., 1997, Brown et al., 1998, Holmes et al., 1988, Huckerby et al., 2001). Glycosylation of the G1 domain with N-linked oligosaccharides and/or KS have been previously documented; however this has no known functionality (Barry et al., 1995). Glycosylation of aggrecan varies depending up the tissues, for example brain-derived aggrecan has no KS chains and fewer CS chains, which likely reflects its function (Matthews et al., 2002, Domowicz et al., 1995, Li et al., 1996, Schwartz et al., 1996). Aggrecan is secreted into the extracellular space via secretory granules (Garcia-Suarez et al., 2014).

Figure 1.6. Biosynthesis of sulfated glycosaminoglycans. (A) KSI and KSII. (B) CS. Sulfation of the disaccharide repeats occurs randomly. Adapted (Esko et al., 2009, Garcia- Suarez et al., 2014).

1.1.6.3.4 Aggrecan Functionality and Protein Interactions Aggrecan is essentially a space filling molecule with various functionalities. It carries out its’ primary function using covalently bound KS and CS. Steric hindrance between adjacent GAG chains aligned along the proteoglycan core enables the proteoglycan to become extended in its ‘bottlebrush’ conformation providing rigidity (Nia et al., 2015). Negatively charged and aligned along aggrecan core protein KS and CS attract positively charged ions within the ECM creating an ion gradient that draws and traps water molecules onto aggrecan. This results in an osmotic hydrostatic pressure that provides resistance against compressive forces allowing ‘forests’ of aggrecan monomers to uniformly redistribute loads. These ‘forests’ or macromolecular aggregates of aggrecan are formed through the binding of 50— 100 aggrecan monomers spaced ~12 nm apart along chains of non-sulfated glycosaminoglycan hyaluronic acid (HA; Fig 1.5) (Hardingham and Fosang, 1992, Kiani et al., 2002, Day and Prestwich, 2002, Blundell et al., 2005, Rosenberg et al., 1975). Aggrecan binds HA through the G1 domain via interaction with link protein and through the G1–G2 interglobular domain (Neame and Barry, 1993, Barry et al., 1995). The aggregates range from 0.5–4µm in length (Rosenberg et al., 1975).

1.1.6.3.5 Tissue-specific Variability in Aggrecan Structure The structure of intact aggrecan varies depending on the tissue it is isolated from and the age of the specimen (summarised in Table 1.1). Aggrecan core protein length (LCP) and individual GAG chain length (iGAGL) are age-dependent in human AC (Lee et al., 2013).

Mature human aggrecan has a shorter LCP and iGAGL than newborn aggrecan and isolated aggrecan has been shown to have lower compressive stiffness under chemically controlled conditions (Lee et al., 2013, Lee et al., 2010, Dean et al., 2006). Mature equine AC aggrecan has been shown to have a significantly lower LCP and iGAGL compared to aggrecan isolated from bone marrow (BM-) MSCs cultured in a peptide hydrogel stimulated with transforming growth factor beta-1 (TGFβ-1) (Lee et al., 2010). Decreased GAG chain length has been associated with a reduced core protein length in bovine nasal cartilage (Mörgelin et al., 1989). These studies highlight the presence of age-, species-, and culture condition- dependent differences in aggrecan structure; however, this has been investigated primarily in the context of AC.

Table 1.1. Summary of intact aggrecan core protein contour length described in the literature.

All Observable Aggrecan Intact Aggrecan Sample L iGAGL L iGAGL CP n n CP n n (nm) (nm) (nm) (nm) Newborn

Articular 477 ± 554 ± 139 55 ± 3 102 62 57 ± 5 34 Cartilage 16 10

(human) Mature Articular 246 ± 510 ± 450 33 ± 2 117 26 35 ± 5 25 Cartilage 13 24 (human)

Articular 220 ± 20 ± 475 ± Cartilage 140 87 20 21 ± 1.3 20 23.5 0.8 19.3 (equine) BM-MSC 440 ± 44 ± 487 ± aggrecan 299 299 119 46 ± 1.8 119 15.6 1.1 9.7 (equine) Compiling data from the literature suggests cell-derived aggrecan may be similar to immature-derived aggrecan and mature aggrecan is similar across species. Data is taken from Lee et al., (2010) and Lee et al., (2013). Note: mean ± 95% confidence interval of means. LCP = core protein contour length; iGAGL = individual glycosaminoglycan chain length; BM-MSC = bone marrow mesenchymal stem cell.

In contrast, our understanding of aggrecan structure within the NP is poorly developed. Several studies by Buckwalter and colleagues utilised electron microscopy to compare mean core protein length of aggrecan between AC and IVD which was found to decrease with age and was highly variable within each tissue due to fragmentation (Buckwalter et al., 1985, Buckwalter et al., 1994, Buckwalter et al., 1989). These observations of aggrecan fragmentation in young healthy tissues have been confirmed by more recent studies (Kopesky et al., 2010, Lee et al., 2010, Lee et al., 2013); however, the extent of aggrecan fragmentation and the functional consequences remain poorly characterised. Instead,

28 attention has focussed on the intact aggrecan population (defined as containing all three globular domains) (Roughley et al., 2014, Kopesky et al., 2010, Lee et al., 2010, Lee et al., 2013).

These studies highlight how the structure of aggrecan impacts on the mechanical behaviour of ECM, but primarily within the context of AC; aggrecan structure is poorly defined in NP and the effect of cell culture conditions (i.e. cell origin, oxygen concentration, growth factor exposure and nutrient availability) on aggrecan structure is poorly understood. From these studies one may infer that providing aggrecan core protein structure is invariant, variation in aggrecan structure is a consequence of differential glycosylation/post-translational modification of aggrecan. Multiple factors could impact on this including cell type- dependent expression of saccharide- and sulfo-tranferases, and the ability of these enzymes to synthesise GAGs, which is modulated by tissue-dependent availability of substrate sugars, oxygen and pH (Matthews et al., 2002, Urban, 2002). Understanding what drives this is vital to successfully generating AC and NP tissue engineered constructs, the mechanical behaviour of which is predominantly dictated by aggrecan structure.

1.1.6.3.6 Fragmentation/degradation of aggrecan Fragmentation or degradation of aggrecan is controlled through proteolysis by a group of enzymes designated as aggrecanases. Aggrecan fragmentation occurs in three main stages: (i) proteolysis by various members of the MMP and ADAMTS family of enzymes (ii) endocytosis and (iii) recycling of remaining CS-containing peptite fragments by glycosidases and sulfatases (Vynios, 2014). Both human and bovine aggrecan have six major cleavage sites (five of which are situated in the CS2 region) targeted by various proteases (Table 1.2). The position of these major cleavage sites along aggrecan core protein is depicted in Figure 1.7. Cleavage at Glu373-Ala374 (NITEGE373-ARGSV374) is the major G1 cleavage site, which is cleaved by the proteolytic enzymes aggrecanse-1 (ADAMTS4) and aggrecanase-2 (ADAMTS5). Cleavage at this site is responsible for the formation of the two most abundant aggrecan fragments in AC (Huang and Wu, 2008, Miwa et al., 2009). The following enzymes are also capable of cleaving aggrecan: the major tissue matrix metalloproteinases (i.e. MMP-1, 2, 3, 7, 8, 9, 10, 13) and cathepsins K, L and m-calpain are found in both the IVD and AC (Chakraborti et al., 2003, Gruber et al., 2011, Konttinen et al., 1999, Ariga et al., 2001, Vo et al., 2013, Fukuta et al., 2011b, Struglics and Hansson, 2010). Cathepsins B, D and H are only found in the AC and cathepsin G is only found in the

IVD (Hembry et al., 1995, Okada et al., 1992, Hamamura et al., 2013, Hembry et al., 2007, Lipari and Gerbino, 2013, Ruettger et al., 2008, Salminen-Mankonen et al., 2007, Cawston and Wilson, 2006, Ohta et al., 1998, Vittorio et al., 1986, Nakase et al., 2000). Glycosylation plays an important role in aggrecan fragmentation; KS substitution of aggrecan increases with age, potentiates aggrecanase attachment and therefore cleavage Glu373-Ala374 (Pratta et al., 2000, Barry et al., 1995, Poon et al., 2005). GAGs also mediate activation of latent Cathepsins (Fonovic and Turk, 2014). In particular, CS and KS fragments generated by cleavage of aggrecan by Cathepsin K and other proteolytic enzymes potentiates the ability of Cathepsin K to degrade collagen, resulting in a positive feedback mechanism that further promotes the destruction of ECM (Li et al., 2000, Li et al., 2004, Li et al., 2002).

Table 1.2. Aggrecan cleavage sites and proteases responsible for cleavage.

All Proteases Responsible for Sequence Reference Cleavage (Porter et al., 2005, Westling et 341 342 MMP-1, 3, 7, 9, 13, 14, 19, and al., 2002, Hou et al., 2003, Mort Asn –Phe 20 at neutral pH, capthesin B et al., 1998, Fosang et al., (human and bovine) and K at low pH 1992a) 373 (Huang and Wu, 2008, Vo et al., NITEGE - 374 ADAMTS-1, 4, 5, 8, 9, 11 and 2013, Porter et al., 2005, Pockert ARGSV 15 et al., 2009, Abbaszade et al., (human and bovine) 1999) PGVA709-AVPV710 954 955 (Struglics and Larsson, 2010, GDLS -GLPS Oshita et al., 2004, Yamamoto et GDLS1353-GLPS1354 Calpain-2 (m) al., 1992, Szomor et al., 1999, 1411 1412 EDLS -RLPS Maehara et al., 2007) GDLS1431-GVPS1432 Leu1462-Val1463 Capthesin D (Handley et al., 2001) 1480 ASELE - 1481 GRGTI (bovine) 1545 ADAMTS-4, 5 (Huang and Wu, 2008, Pockert ASELE - et al., 2009) 1546 GRGTI (human) Leu1654-Val1655 Capthesin D (Handley et al., 2001) 1667 FKEEE - 1668 GLGSV (bovine) 1714 ADAMTS-4, 5 (Huang and Wu, 2008, Pockert FKEEE - et al., 2009) 1715 GLGSV (human) 1771 PTAQE - 1772 (Huang and Wu, 2008, Pockert AGEGP (bovine) 1714 ADAMTS-4, 5, 9 et al., 2009, Somerville et al., FKEEE - 2003) 1715 GLGSV (human) Phe1754-Val1755 Capthesin D (Handley et al., 2001) Leu1854-Ile1855 Capthesin D (Handley et al., 2001) 1871 TISQE - 1872 LGQRP (bovine) (Huang and Wu, 2008, Kuno et 1919 ADAMTS-1, 4, 5 TISQE - al., 2000, Pockert et al., 2009) 1920 LGQRP (human) Summary of all proteolytic cleavage sites along the core protein of aggrecan and the proteolytic enzymes that act on those sites. Major cleavage sites are highlighted in white.

Figure 1.7. Major aggrecanase cleavage sites along the aggrecan core protein. Aggrecanase cleavage sites are shown for both human and bovine aggrecan. Reproduced from Huang and Wu, (2008).

Chapter 1.4.1.3 highlights that aggrecan is an important structural protein in cartilaginous tissues, with multiple functional roles. The presence of variable glycosylation and cleavage sites indicate that aggrecan is structurally modifiable, highlighting the potential for manipulation of its mechanical performance to meet the specific needs of the host niche microenvironment. The purpose of this study is to characterise the structure of aggrecan in cartilaginous tissues — specifically NP and AC — and determine if aggrecan structure can be modified following manipulation of cell culture conditions.

1.2 Hypothesis and aims of the project

Aggrecan plays a major structural role in cartilaginous tissues. Variation in aggrecan structure is known to affect the mechanical functionality of individual monomers, and therefore likely impacts on overall tissue function. Crucially, aggrecan fragmentation and loss has been associated with tissue ageing and degeneration (e.g. IVDD). There is currently a lack of therapies for IVDD; one potential therapy is replacing lost tissue with TE IVD constructs. However, there is a lack of understanding on the structure and function of aggrecan within the IVD. If we wish to synthesise a viable IVD replacement, we must first characterise the nanostructure of aggrecan in native tissue. There are also various considerations for engineering IVD constructs which must be investigated, for example cell type, availability of bioactive growth factors and oxygen concentration.

Therefore we hypothesise that:

Aggrecan ultrastructure may be variable in response to cell type, tissue of origin, species, and microenvironmental factors, and needs to be understood in more detail in order for tissue engineered constructs to recapitulate native NP tissue.

The hypothesis and aims for each experimental chapter (3—5) of this project have been provided at the end of each individual chapter’s introduction. Provided below is a list of the individual chapters:

 Chapter 2: General methods  Chapter 3: Characterisation of aggrecan and collagen structure in young, healthy native cartilaginous tissues  Chapter 4: The effect of culture conditions on the nanostructure of aggrecan and mechanical properties in IVD tissue engineered constructs  Chapter 5: Characterisation of aggrecan structure in young, healthy porcine tissues  Chapter 6: Conclusions and future work

2 Chapter 2: General Methods

In order to avoid undue repetition, the methods for the following chapters are located in this comprehensive general methods chapter. This chapter also details the steps taken to optimise the experimental methodology throughout. As such the subsequent experimental chapters contain a detailed overview of the experimental rationale and design.

2.1 Cartilaginous tissue collection 2.1.1 Native tissue sample collection All reagents were sourced from Sigma-Aldrich (Gillingham, UK) unless otherwise stated. Bovine tissues were sourced from Kurpas Meats PLC (Manchester, UK) abattoir within 24 h of death. NP and AC tissue samples were dissected from bovine tail IVDs and knee joints, respectively (mature tissue from animal’s ≤30 month’s old, immature tissue from neonates). Porcine tissue was sourced from G and Gb Hewitt Ltd (Chester, UK) NP and AC tissue samples were dissected from spine IVDs and knee joints, respectively from female piglets (≤6 weeks old, 10—12kg). Dissected tissue was frozen at -80oC prior to aggrecan extraction. Additional tissue was prepared for histological sectioning by snap freezing in optimal cutting temperature compound (OCT).

2.1.2 Isolation and culture of nucleus pulposus cells Bovine NP tissue was dissected into 1—2mm3 pieces and treated with type II collagenase (0.25% w/v in AQmedia™ with 4500 mg/L glucose, L-alanyl-glutamine, and sodium bicarbonate, without sodium pyruvate, supplemented with 200 units/mL penicillin, 200 µg/mL streptomycin, 0.50 µg/mL amphotericin; approximately 10 mL/g tissue) at 37oC for 2—4 h. Following digestion, the cell suspension was passed through a 40 μm cell strainer and centrifuged at 500 x g for 5 mins. The supernatant was discarded and the cells were resuspended in 10 mL standard disc cell media (AQmedia™ with 4500 mg/L glucose, L- alanyl-glutamine, and sodium bicarbonate, without sodium pyruvate, supplemented with 100 units/mL penicillin, 100 µg/mL streptomycin, 0.25 µg/mL amphotericin, 10% foetal calf serum [FCS], 10 mM ascorbic-2-phosphate [A2P] and 1 mM sodium pyruvate) and cultured o at 37 C 5% CO2. Once cells adhered to the cell culture plate surface, the medium was aspirated, the cells were washed once with PBS (without calcium) and fresh disc cell media was added. Upon 70—80% confluency, cells were passaged using trypsin/EDTA solution (1x) and expanded for encapsulation in type I collagen hydrogels.

2.1.3 Encapsulation of nucleus pulposus cells in a type I collagen hydrogel Cells were expanded to passage 1—3, then trypsinised, and centrifuged at 500 x g for 5 mins. Following aspiration of the supernatant, cells were resuspended in 10 mL disc cell medium and counted using a haemocytometer. NP cells were seeded at a density of 4x106 cells into type I collagen hydrogels: the appropriate number of cells were pelleted and resuspended in disc cell media, type I collagen and 1 M NaH2CO3 at a ratio of 0.05:9:0.05. Each hydrogel was comprised of 100 µL of this stock solution which was subsequently transferred to a transwell insert (high density, translucent PET membrane, 24 well, 0.4 µm pore size) housed o within a 24 well plate, and incubated at 37 C, 5% CO2 for 1—2 h until the hydrogel set.

2.1.4 Chondrogenic stimulation of encapsulated nucleus pulposus cells with GDF6/TGFβ3 NP induction media (AQmedia™ with 4500 mg/L glucose, L-alanyl-glutamine, and sodium bicarbonate, without sodium pyruvate, supplemented with 100 units/ml penicillin, 100 µg/mL streptomycin, 0.25 µg/mL amphotericin, 1.25 mg/mL bovine serum albumin [BSA] 1% foetal calf serum [FCS], 100 µM ascorbic-2-phosphate [A2P] and 1 mM sodium pyruvate, ITS-X [10 µg/ml Insulin, 5.5 µg/mL Transferrin, 6.7 ng/mL Selenium], 40 µg/mL L-proline, 10-7 M dexamethasone, 5.4 µg/mL linoleic acid) was either added to each transwell insert to stimulate condrogenic differentiation; depending on the experiment, NP induction media was either supplemented with 10 ng/mL TGFβ3, 100 ng/mL GDF6 or not supplemented (in no growth factor conditions). In order to replicate the nutrient exchange between the cartilage end plate barriers in the intervertebral disc 750 mL NP induction media was added into the well and 500 mL was added inside the interior of the transwell insert. NP cell constructs were cultured for 28 days prior to processing for aggrecan isolation, histological assessment, mechanical testing, real-time quantitative polymerase chain reaction (PCR) analysis, and in situ gelatin zymography. In preparation for histological assessment, mechanical testing, and in situ gelatin zymography, 28 days NP cell constructs were was in PBS and snap frozen in OCT using liquid nitrogen.

2.2 Isolation and characterisation of aggrecan 2.2.1 Aggrecan isolation by dissociative caesium chloride centrifugation Aggrecan was isolated from native tissue and NP cell constructs as reported previously by Lee et al., (2010) using the well-established dissociative caesium chloride (CsCl) gradient centrifugation technique. Specifically, 1—3 g of native tissue was frozen using liquid

36 nitrogen and homogenised into a fine powder with a Retsch MM301 Ball Mill at a frequency of 26 Hz for 3 mins. This powder was then dissolved in ice-cold aggrecan extraction buffer (4 M guanidine-hydrochloride, 100 mM Na acetate, pH 7) in the presence of a protease inhibitor cocktail (Protease Complete Tablets [Roche, Germany; 1 tablet per 50 mL extraction solution], ratio of 12 mL of buffer per gram of tissue) for 48 h at 4oC on a roller and was then centrifuged at 29000 × g for 30 min at 4°C. In order to isolate aggrecan from cultured cells, 20 NP cell constructs were washed in PBS, drained and then mechanically dissolved on a roller in 11 mL aggrecan extraction buffer, prior to centrifugation. In both cases, after centrifugation the supernatant was collected and CsCl added to a final density of 1.58 g/mL. The following equation was used to calculate the required amount of CsCl in order to achieve a starting density of 1.58 g/mL:

X = a[{1.347p} – {0.0318M} – 1.347] (Equation 1: CsCl calculation)

Where: a = final volume of solution to be centrifuged in mL: 13.5 mL p = final density of solution to be centrifuged in g/mL: 1.58 g/mL M = molarity of GuHCl of the extraction buffer: 4 M X = the amount (g) of CsCl to be added: 8.83 g

Subsequently, supernatant was added to a final weight of 21.33 g (i.e. final weight = p x a; 1.58 g/mL x 13.5 mL = 21.33 g). If the supernatant extracted was not sufficient to achieve a final weight of 21.33 g then extraction buffer was added. Samples were mixed on a roller at 4oC for approximately 1 h until the CsCl was completely dissolved. The final solution of proteoglycan supernatant with CsCl was transferred into a 13.5 mL Quick Seal centrifugation tube (Beckman Coulter, USA), heat sealed and centrifuged at 56,000 rpm for 72 h at 4oC (Ti 70.1 rotor, L-90K ultracentrifuge (Beckman Coulter)). From each tube eleven 1 mL fractions were collected using a 1 mL syringe and 19 G needle. Fractions with density of 1.54 g/mL or greater (D1 fractions) were collected, and fractions with a density of less than 1.54 g/mL were discarded in accordance with established protocols as aggrecan was not identified in these fractions, as assessed by AFM (see Fig. 2.1) (Lee et al., 2010).

Figure 2.1. Aggrecan was only present in fractions with a density of greater than 1.54 g/mL. AFM height images of fractions taken from CsCl-treated AC (2x2 µm; scale bar is 0.2 µm; z-scale 2 nm). (a) 1.38-1.40 g/mL, (b) 1.41-1.45 g/mL, (c) 1.46-1.50 g/mL (d) ≥1.54 g/mL (blue arrows highlight aggrecan monomers).

2.2.2 Preparation of aggrecan for structural characterisation In order to prepare aggrecan for structural characterisation by AFM and size exclusion chromatography multi-angle light scattering (SEC-MALS) it was necessary to determine an appropriate storage buffer that would minimise aggregation of isolated aggrecan in solution, but still enable maximal extension of GAG chains on a surface for structural characterisation. Initially D1 fractions were transferred to high retention cellulose dialysis tubing (12.4 kDa molecular weight cut-off, 99.99% retention) which was boiled in MilliQ water for 2 mins prior to use. Subsequently, the D1 fractions were dialysed exhaustively against 1 M NaCl overnight followed by 48 h against either tris-buffered saline (TBS), MilliQ water (<17.3 MΩ cm resistivity) or phosphate buffered saline (PBS) Dulbecco’s A (0.137 M NaCl; OXOID, Thermo Scientific, UK). Aggrecan was generally found to be more prone to

38 aggregation when stored in TBS or MilliQ water and therefore dialysis was ultimately performed against PBS (Fig. 2.2).

Figure 2.2. Optimisation of aggrecan storage buffer: water vs tri-buffered saline vs phosphate buffered saline. Scanasyt AFM height images of aggrecan isolated from nucleus pulposus tissue prepared in (a) water; (b) tris-buffered saline; (c) phosphate buffered saline. Scale bar represents 0.2 µm.

There is a clear consensus in the literature that protease inhibitors are only required in the extraction buffer (Ng et al., 2003, Lee et al., 2010, Lee et al., 2013); however as aspects of this study are focused on aggrecan fragmentation, the effects of protease inhibitors on aggrecan fragmentation were also tested in the dialysis buffer to ensure elimination of protease activity. No significant difference was found in aggrecan fragmentation between the methodologies (Fig. 2.4). Therefore, protease inhibitors were only used in the extraction buffer. We also wanted to determine if the isolation procedure could affect aggrecan morphology. To investigate this, D1 samples underwent a second dissociative CsCl density centrifugation (D1D1 fractions) which was dialysed exhaustively against PBS. D1D1 samples were found to be affected by the isolation procedure; however, it was unclear as to whether prolonged exposure to CsCl affected aggrecan structure or that the larger molecules were extracted in the first round of dissociative CsCl density centrifugation (Fig. 2.5). Therefore, only one round of CsCl density centrifugation was used across all samples in line with previous studies (Ng et al., 2003, Kopesky et al., 2010, Lee et al., 2010, Lee et al., 2013, Roughley et al., 2014).

Figure 2.3. Protease inhibitors were only required in the extraction buffer. Molecular area size distribution of fragments of all observable (i.e. both intact and non-intact) aggrecan regardless of fragment size and orientation (n=1). There is no significant difference in the molecular area of aggrecan when using protease inhibitors through the entire aggrecan isolation procedure versus only the extraction buffer. Values shown within the histogram are the mean (± SD).

Figure 2.4. Effect of isolation procedure on intact aggrecan structure. (a) Core protein contour length distributions for AC-derived intact aggrecan (D1 vs D1D1). Values shown within the histogram are the mean LCP (± SD). (b) Mean core protein length for AC-derived intact aggrecan (D1 vs D1D1). Box shows mean values and whiskers show minimum and maximum values.

2.2.3 Structural Characterisation of aggrecan by atomic force microscopy and multi- angle light scattering Aggrecan structure was characterised both following adsorption to a surface (by AFM) and in solution (by SEC-MALS). All AFM imaging was carried out at the University of Manchester BioAFM Facility. Isolated aggrecan was prepared for visualisation as previously reported (Kopesky et al., 2010, Lee et al., 2010, Lee et al., 2013). Briefly, isolated and purified aggrecan was adhered to freshly cleaved mica sheets (Agar Scientific, Stanstead, UK) pre-coated with (3-aminopropyl)triethoxysilane (APTES; 0.01% v/v). Mica specimen discs were then attached to metal specimen discs (15 mm diameter; Agar Scientific), washed with double-filtered MilliQ water (<17.3 MΩ cm resistivity; Purite, Thame, UK) and allowed to air-dry for at least 2 h at room temperature. AFM scans of prepared aggrecan samples from native tissue were captured using Peak Force Tapping ModeTM (ScanAsyst) in air at environmental humidities of less than 45%. Imaging was performed using a Multimode 8 AFM (Bruker, MA, USA) fitted with a “J” scanner and controlled by a Nanoscope V controller and using pyramidal ScanAyst-Air (Bruker) probes (with a nominal spring constant k of 0.4 N/m and tip radius of 2 nm). Scan parameters were adjusted automatically by the ScanAsyst software to optimise image quality (Su et al., 2012). AFM scans for prepared aggrecan samples from cell culture-derived aggrecan were captured using AC modes scanning in air at environmental humidity’s of less than 45%. Imaging was performed using a Nanowizard 4 AFM (JPK, Germany) operated on a JPK motorised precision stage mounted on a Zeiss AxioObserver A1 inverted optical microscope encased within a JPK acoustic enclosure (Halcyonics i4 benchtop active vibration isolation system [Accurion, US]) to remove background noise. The system was controlled by a JPK Vortis SPM controller and operated using SPM control software (V4) and JPK DP data processing software. Scanning was performed using FASTSCAN-A (Bruker ASX SAS, France) probes (with a nominal spring constant k of 18 N/m and tip radius of 600 nm) in air at environmental humidity’s of less than 45%.

Quantification of aggrecan ultrastructure was undertaken as follows. AFM height topography images (2 x 2 µm at a sampling frequency of 3.9 nm) were flattened and converted to ASCII matrix files using WSxM 5.0 Develop 7.0 (Horcas et al., 2007). The mean background was subtracted from these images using routines written in Microsoft Visual Basic 6.0. (Sherratt et al., 2004), and converted to RAW files. Curved aggrecan monomers were straightened using the straighten plugin (Kocsis et al., 1991) for ImageJ by

41 manually tracing along the core protein; the molecular ultrastructure of intact aggrecan monomers (comprising all three globular domains) was then characterised with regards to key morphological parameters (core protein contour length [LCP], glycosaminoglycan

[GAG] binding region length [LGAG] and molecular area of the aggrecan monomer covered by GAG [MAGAG]). Previously, GAG chain length has been estimated from measurements of individual GAG chains (Lee et al., 2013); however, individual GAG chains are rarely discernible by either transition electron microscopy (TEM) or AFM. By measuring MAGAG and LGAG we can calculate the mean length of all GAG chains for each monomer. Approximately 60—100 AFM scans were taken to acquire and measure sufficient intact aggrecan monomers (n=~30—40 per sample).

Figure 2.5. Ultrastructural characterisation of aggrecan. Aggrecan was characterised according to key morphological parameters (core protein contour length [LCP], glycosaminoglycan [GAG] binding region length [LGAG] and molecular area of the aggrecan monomer covered by GAG [MAGAG]).

Aggrecan fragmentation was assessed in terms of the ratio of intact to non-intact aggrecan monomers and the average molecular area/size. The first n=300—400 aggrecan monomers were identified across the original AFM scans (including both intact and non-intact) for each sample. Molecular area was measured using ImageJ (http://rsbweb.nih.gov/ij/ [version 1.48/Java 1.7_51 32-bit]). Following subtraction of the mean average background (using routines written in Microsoft Visual Basic 6.0) (Sherratt et al., 2004), each AFM image underwent the following steps: (i) thresholding (16.71%, 58/255, Default, no dark background); (ii) analysis of particles with an area of at least 100 pixels. AFM images were analysed using a Matlab (v. R2014a) code (https://github.com/maxozo/Agracan_Length_Volume_AFM) that was written using the Miji platform (Fiji 1.3.6-fiji and Matlab interface). This code removes noise, smoothing

42 areas (using the despeckle filter) before converting to a binary image using the default binarize command. Instrumental/sample noise was reduced by excluding elements with a radius <3 pixels and infilling elements with a radius >5 pixels. The images were further processed using the regionprops function in Matlab to select each aggrecan molecule, and the number of pixels in each selection was counted to determine the aggrecan area measurement. These areas were subsequently skeletonised using the bwmorph function (Fig. 2.6), and the lengths of the resulting skeletons were used as measurements of aggrecan length. Each molecule within the image was characterised in terms of molecular length and area. Aggregated molecules and sample contamination were manually removed. The average length of the substituted sulphated GAGs was calculated by dividing the molecular area by the aggrecan length, providing the area per unit of length, which was then divided by two.

Figure 2.6. AFM analysis of aggrecan skeletonised using the bwmorph function. (a) Aggrecan was converted to a RAW image then the lengths of the (b) resulting skeletons were used as measurements of aggrecan length.

Adsorption of three dimensional molecules to two dimensional surfaces inevitably affects their structure (Sherratt et al., 2004). As such we also characterised the structure of AC- and NP-derived aggrecan by SEC-MALS; (Jowitt et al., 2010). This method allowed characterisation of the behaviour of the entire aggrecan population (both intact and non- intact) in solution. Purified suspensions were chromatographed on a Shodex 806M SEC gel filtration column in PBS driven by a Bio-Rad GPC system (Bio-Rad, Hercules, California)

44 at 0.75 mL/mina DAWN HELEOS II MALS system with a Wyatt QELS detector (WYATT- 288-H2; Wyatt Technologies, Santa Barbara, US). The absolute molecular mass, polydispersity index and root mean square (RMS) conformation of the isolated aggrecan was calculated using the supplied Astra 6.1 software (Wyatt Technologies, Santa Barbara, US; (Jowitt et al., 2010)). SEC-MALS was also used to determine the shape (e.g. cylindrical/rod- like or globular/spherical) of aggrecan monomers across a range of salt concentrations (0.001 M NaCl, PBS and 1 M NaCl) using RMS conformation.

2.3 Histological assessment of tissue composition and collagen fibril alignment 2.3.1 Histological staining Safranin-O/Fast Green (proteoglycan), Alcian Blue (pH 2.5) (sulphated GAGs), Masson’s Trichrome (non-fibrillar collagen), and Picrosirius Red staining (fibrillar collagen) were used to characterise gross ECM composition. For all stains, tissue cryosections (7 µm thickness) were allowed to dry on to SuperFrost Plus glass slides (Thermoscientific, USA). Following staining all sections were dehydrated in industrial methylated spirits, cleared in xylene and mounted using Pertex (HistoLab, Sweden). Bright-field (and polarised for picrosirius red) images were acquired using a Leica Leitz DMRB microscope (Leica Microsystems, UK) with an Infinity X camera (Lumenera Corporation, Canada) and DeltaPix camera software (DeltaPix, Denmark).

Alcian Blue (pH 2.5) stains weakly sulphated GAGs blue and nuclei may also be stained red by calcium salts present in tissue. Staining was carried out as follows: sections were rinsed for 3 min in 3% acetic acid (v/v), and then stained in Alcian Blue solution (1% Alcian blue, Cat ref: DYE011-B [Solmedia, UK]) w/v in 3% acetic acid (v/v) at pH 2.5) for 30 min. Sections were subsequently washed in 3% acetic acid and rinsed in cold tap water for 10 min prior to dehydration and mounting.

Safranin-O/Fast Green stains nuclei black, cytoplasm green and proteoglycans are orange/red. Staining was carried out as follows: sections were stained in Weigerts haematoxylin (mix Weigerts A and Weigerts B 50:50, Cat ref: AS054-C [A], Cat ref: AS055-C [B] [TCS Biosciences Limited, UK]) for 3 min, washed in tap water for 10 min, stained in filtered 0.02% aqueous Fast Green (Raymond A Lamb, Cat No: S142-2 (CI 42053)) for 5 min, then washed briefly in 1% acetic acid. Each slide was drained and then

45 stained in filtered 0.1% aqueous Safranin O (Cat No: 34312 [VWR International, USA]) for 5 min, then rinsed in IMS prior to dehydration and mounting.

Masson’s Trichrome with Aniline Blue stains nuclei and gametes black, cytoplasm is red, and non-fibrillar collagen is stained blue. Sections were stained using a kit from Surgipath Europe Ltd (Bio-Optica; Cat No: 04010802) as follows: Weigerts haematoxylin for 10 min. Without washing, the slides were drained and 10 drops of panchofusion mason were added for 4 min, and then washed quickly (3—4 seconds) in deionised water. Next, 10 drops of phospho-ktungtastic acid were added for 4 min, sections were washed in deionised water and then 10 drops of Aniline Blue were added for 10 min. Without washing the slides were drained and 10 drops of 1% acetic acid were added for 5 min and subsequently washed in deionised water.

The relative abundance (area fraction) of collagen fibril bundles within a tissue can be assessed using picrosirius red (PSR) staining to enhance collagen birefringence (McConnell et al., 2016). PSR stains nuclei blue. Under polarised light fibrillar collagen exhibit regions with green, red, and orange; the thickest filaments appear red and the thinnest filaments are green. Tissue cryosections (7 µm thickness) were allowed to dry onto SuperFrost Plus glass slides (Thermoscientific, USA) prior to fixation with 4% v/v paraformaldehyde in PBS. PSR staining was achieved as previously described (McConnell et al., 2016). Briefly, sections were stained in Mayor’s haematoxylin (Solmedia Laboratory Supplies, Shrewbury, UK) for 5 min, rinsed in deionised water, stained for 1 h in 0.1% Sirius Red (F3BA, Direct Red 80) in saturated aqueous Picric, washed in 0.1% acetic acid for two washes (1 min each) and without washing in water, sections were drained, dehydrated, cleared and mounted. Bright-field and polarised light images were acquired. When visualised under cross-polarised light, the resulting collagen birefringence may be semi-quantitatively assessed against total tissue area (Graham et al., 2011). This was repeated at three locations per sample, either across variable gradients in tissue composition [e.g. AC surface through to deep zones] or three equidistant points in homogenous tissue (e.g. NP).

2.3.2 Characterisation of collagen fibril ultrastructure by atomic force microscopy Collagen fibril periodicity and orientation was assessed in situ in AC and NP tissue cryosections using established AFM protocols (Graham et al., 2004, Graham and Trafford, 2007, Graham et al., 2010, Fang et al., 2012). Briefly, cryosections (10 µm) were adsorbed

46 onto ethanol-cleaned glass slides and allowed to air-dry for 24 h at room temperature (Sherratt et al., 2004, Graham et al., 2010). In order to reveal the underlying collagen network, NP tissue sections were additionally incubated with 100 µL of Sorensen’s phosphate buffer (95.1 mM Na2HPO4 and 37.9 mM KH2PO4 at pH 7.2) containing hyaluronidase (400-1000 U/mL; from bovine testes) for 24 h at 37oC (Plodinec et al., 2010a). Digested sections were then washed three times with PBS and allowed to air-dry for several hours at room temperature prior to AFM scanning. Collagen fibrils were visualised within AC and NP tissue sections by Tapping Mode AFM in air using a Catalyst AFM, Nanoscope V controller and pyramidal OTESPA probe (Bruker: nominal spring constant k = 12—103 N/m, nominal tip radius = 7 nm). The amplitude set-point was adjusted to optimise image quality. Collagen fibril periodicity and orientation was determined from amplitude images (4 x 4 µm at a sampling frequency of 3.9 nm). These images were processed using the 2D Fast Fourier Transform (2D FFT [for periodicity]) and power spectrum density (for orientation) tools found within WSxM (Wallace, 2015). Orientation coherency was assessed using the OrientationJ plugin in ImageJ. OrientationJ determines a vector for each pixel and calculates the average vector across an image and expresses the orientation coherency (i.e. the percentage of pixels that are aligned in the same direction) (Rezakhaniha et al., 2012). This analysis was repeated across 4 sites per AC and NP sample (biological replicates n=3).

Collagen fibril diameter was quantified by characterising the height of isolated fibrils (Plodinec et al., 2010b). AC and NP tissue samples were dissected into small chunks, then digested with 100 µL of Sorensen’s phosphate buffer containing hyaluronidase, as previously described (Plodinec et al., 2010a). The remaining collagen matrix was resuspended in PBS and homogenised using a T10 basic Ultra-Turrax homogeniser (IKA, UK) leaving isolated collagen fibrils. Approximately 10 µL of the collagen fibril isolate was pipetted onto an ethanol cleaned uncoated glass slide and allowed to air-dry at room temperature for several hours, after which the samples were washed three times with MilliQ water and allowed to air-dry at room temperature prior to visualisation by Tapping modeTM AFM. Collagen fibril diameter was then quantified from 1st order flattened AFM height images. Ten transverse profile measurements were taken across 50 individual isolated collagen fibrils per sample and the average diameter was calculated (biological replicates n=2).

2.4 Measurement of micro-mechanical compressive stiffness by atomic force microscopy indentation The compressive (reduced) modulus of hydrated AC and NP cryosections was assessed by AFM nanomechanical assessment on a Bruker Catalyst AFM with a Nanoscope V controller (Su et al., 2012). AC and NP cryosections (nominal 20 µm thickness) were adsorbed to SuperFrost Plus slides (ThermoScientific, UK) and allowed to air-dry for 24 h at room temperature. Dried sections were washed with MilliQ water to remove OCT and allowed to rehydrate for 1 h prior to mechanical assessment. AC tissue sections were characterised using a pyramidal ScanAsyst-Fluid probe (Bruker; nominal spring constant k = 0.7 N/m, nominal tip radius = 20 nm). The greater compliance of NP tissue required the use of a softer cantilever - the pyramidal MLCT-E probe (Bruker; nominal spring constant k = 0.1 N/m, tip radius 20 nm). Each sample was mechanically assessed at four 100 µm2 sites per cross section (~400 points per site). These sites were assessed as follows: in AC (n=3 biological replicates) from the cartilage superficial zone to the deep zone and throughout the middle zone of NP tissue (as the tissue was homogenous [n=3 biological replicates]; Figure 6). AC tissue was probed at a constant indentation depth of 20 nm, with a 1 Hz sampling time between indents. The more compliant NP tissue was probed at an indentation depth of 20 nm, a retracted delay of 1 s, and with a 0.250 Hz sampling time between indents. The reduced modulus was derived from force curves generated by indentation of tissue samples (Figure 2.7).

Figure 2.7. Definition of minimum and maximum force boundaries. Figure shows a standard force curve generated by indentation of a tissue. The dark blue curve is the extend Z direction and the light blue is the retard Z direction. (http://becs196a.caltech.edu/NanoScopeAnalysisV150Manual.pdf)

The mean compressive (reduced) modulus was derived using Nanoscope Analysis software (Bruker) using a Sneddon (conical) contact point based indentation model with the extension curve using equation 2 (McConnell et al., 2016). The Young’s modulus was computer from the slope of the linearised Sneddon equation (equation 3).

2 퐸 퐹 = tan (훼)훿2 (Equation 2: Sneddon Equation) 휋 (1−휐2) Where: F = force (from force curve) E = Young’s modulus (fit parameter) υ = Poisson’s ratio α = half-angle of the indenter δ = indentation

2 퐸 (퐹)2 = ( tan(훼)) 1/2 (Equation 3: Linearised Sneddon Equation) 휋 (1−휈2) 훿

Before the force curves were linearised the minimum force (Fig. 2.7) was manually subtracted from the maximum force. Adhesion force was included in the modulus calculation. A baseline correction was applied to each force curve to account for baseline tilt generated by the detector configuration and offset with a linear correction that was applied to each whole force curve. The linear Sneddon model is fit to the initial slope of the force curve (Fig. 2.8). Reduced modulus calculated from force curves with a goodness of fit (R2) lower than 0.99 and calculated reduced modulus greater than two standard deviations from the mean were discarded.

Figure 2.8. Force plot with a linearised fit. Figure shows a standard force fit against a linearised Sneddon model. (http://becs196a.caltech.edu/NanoScopeAnalysisV150Manual.pdf)

2.5 Relative gelatinase activity determined by in situ gelatinase zymography In order to determine if aggrecan fragmentation might be driven by chronic protease activity we used in situ gelatin zymography to compare the relative gelatinase activity in bovine AC and NP with a tissue (young healthy rat skin) which displays minimal activity (Tewari et al., 2014). Matrix metalloproteinases (MMPs) are synthesised and sequestered in tissues in an inactive form, therefore MMP expression and immunohistochemical localisation alone are poor predictors of protease activity (Chakraborti et al., 2003, Mook et al., 2003, Frederiks and Mook, 2004). The major tissue matrix metalloproteinases (i.e. MMP-1, 2, 3, 7, 8, 9, 10, 13) and cathepsins K and L are found in both the IVD and AC (Konttinen et al., 1999, Ariga et al., 2001, Chakraborti et al., 2003, Gruber et al., 2011, Vo et al., 2013). Cathepsins B, D and H are only found in the AC and cathepsin G is only found in the IVD (Vittorio et al.,

1986, Okada et al., 1992, Hembry et al., 1995, Nakase et al., 2000, Cawston and Wilson, 2006, Hembry et al., 2007, Salminen-Mankonen et al., 2007, Ruettger et al., 2008, Hamamura et al., 2013, Lipari and Gerbino, 2013, Ohta et al., 1998). All of these enzymes exhibit gelatinase activity and are capable of proteolytically cleaving both aggrecan and gelatin (collagen); therefore the active forms of these enzymes should be detected by in situ gelatin zymography (Nguyen et al., 1990, Nguyen et al., 1991, Fosang et al., 1992b, Konttinen et al., 2002, Chakraborti et al., 2003, Hou et al., 2003, Pelletier et al., 2005). In situ gelatin zymography can be used to localise and quantify relative ECM protease activity in tissue cryosections (Frederiks and Mook, 2004). Gelatinase activity was localised and quantified in tissue cryosections (nominal 15 µm thickness) cut from young healthy bovine AC and NP tissue (n=2) and from healthy rat skin (n=1) as previously described (Akhtar et al., 2014). Briefly, low gelling temperature agarose (1% w/v PBS) was dissolved at 80oC then cooled to 37oC. DAPI and DQ Gelatin stock solution were added to agarose at a final concentration of 1 µg/mL and 10 µg/mL respectively. Approximately 40 µL of agarose/DAPI/DQ gelatin solution was added to each section and cover-slipped immediately. DQ-gelatin coated cryosections were incubated at 4oC for 1 h then at room temperature in the dark for 18 h prior to fluorescent imaging using an Olympus BX51 microscope using an exposure time of 200 ms. Gelatinase activity was quantified using ImageJ: mean fluorescence intensity was quantified from the surface to ~150 µm into the tissue (i.e. from the articular cartilage surface to the middle/deep zones [n=3] and epidermal surface to endodermis in rat skin [n=1]). Fluorescence activity measurements were taken over a 150 µm x 150 µm area in NP tissue as it is a homogeneous tissue (n=3). Background fluorescence was removed across each tissue section to determine mean gelatinase activity/fluorescent intensity in ImageJ. Tissue auto-fluorescence was accounted for using an unstained control.

2.6 Computational modelling of aggrecan packing In order to model the potential effects of aggrecan fragmentation on molecular packing within tissues Thomas Shearer developed a simulation using Mathematica 11.0 (Wolfram Research, Inc., Champaign, Illinois, 2008) which randomly places aggrecan monomers and/or fragments within a virtual cubic volume of side length 40 (all lengths are non- dimensionalised on the aggrecan monomer radius, which was assumed to be constant). Based on the SEC-MALS data, the aggrecan monomers were modelled as rigid cylinders with hemispherical ends. One end of each cylinder was randomly located within the volume

51 by selecting a random coordinate within the range (0,0,0) to (40,40,40), then the orientation of the cylinder relative to that point was defined via a unit vector in a spherical coordinate system described by azimuthal angle θ and polar angle φ, which were randomly selected from the ranges 휃 ∈ {0,2휋} and 휑 ∈ {0, 휋} (molecular extremities were allowed to project beyond the volume boundaries). Each time a new cylinder is placed, the algorithm tests whether it overlaps with an existing cylinder and rejects it if it does (we call this a failed placement) or keeps it if it does not. This process is repeated until there have been a critical number, c, of failed placements, at which point the algorithm terminates. To determine an appropriate value for c, we monitored the final aggrecan volume fraction, f, as c was increased in increments of 100 until f no longer increased. The final value used for all simulations was c=700. Three simulated molecular populations were investigated, which were categorised according to their length (relative to their radius). The first had a distribution of (relative) lengths defined by a Gamma distribution which was fitted to the experimental data for NP aggrecan and found to have shape parameter α=2.97 and scale parameter β=3.02 (we call this population the experimental population). The second had a constant (relative) length of 17.3, which was the average (relative) length calculated from the experimental data on intact NP aggrecan (called the intact population), and the third had constant (relative) length a quarter of that for the intact population in order to simulate entirely fragmented aggrecan (called the fragmented population). The influence of molecular length on total aggrecan volume fraction was then determined for each of the three populations (n=12 simulations/population) by determining what fraction of the central region of the cube (another cube of side length 20 with opposite corners at the points (10,10,10) and (30,30,30)) was occupied by cylinders.

2.7 Gene expression analysis using real-time quantitative-PCR 2.7.1 Isolation of RNA from NP cell constructs After 28 days, each NP cell construct (n=3 per experimental parameter) was washed in PBS, drained and treated with 1 mL of Tri-reagent (TRIzol), homogenised and incubated at room temperature for 10 mins. Samples were then vortexed for 20 s and either stored at -80oC for up to 2 months or processed immediately. Following cell lysis, each sample was centrifuged at 12,000 x g for 15 mins at 4oC, the supernatant was retained, 100 µL 1-Bromo-3- chloropropane was added, vortexed for 20 s, incubated at room temperature for 3 mins and centrifuged at 12,000 x g for 15 mins at 4oC. The resulting upper aqueous phase was separated and 500 µL isopropanol and 2 µL GlycoBlue™ (15 mg/mL; Invitrogen, Ambion,

UK) was added. Tubes were inverted 10 times, incubated at room temperature for 10 mins, centrifuged at 12,000 x g for 20 mins at 4oC, the supernatant was removed leaving a small blue pellet which was washed in ice cold 75% ethanol, vortexed briefly and centrifuged at 12,000 x g for 5 mins at 4oC. After removing the ethanol, the pellet was air dried and resuspended in 21.2 µL Tris-EDTA buffer, 1.2 µL of which was used to perform a RNA quantification using a Nanodrop ND-1000 spectrophotometer (Thermofisher, UK).

2.7.2 Reverse transcription of RNA RNA samples were converted to complementary DNA. All reagents were sourced from ThermoFischer Scientific, Applied Biosciences (UK) unless otherwise stated. For each 20 µL reaction 10 µL isolated RNA (max 200 ng/µL) was added to 10 µL of the following mastermix: reverse transcription buffer (2x), deoxynucleotide mix (20 mM), reverse transcriptase random primers (2x), MultiScribe™ Reverse Transcriptase, RNAse Inhibitor and topped up to 10 µL using molecular biology grade H2O. After briefly vortexing to mix, samples were incubated under the following conditions using a PTC-2000, Peltier thermal cycler (MJ Research, BIO RAD, UK): 25°C for 10 mins, 37°C for 120 mins, 85°C for 5 s,

4°C for ever. The cDNA was diluted to 5 ng/µL using molecular grade H2O and was stored at -20°C, prior to real-time quantitative-polymerase chain reaction.

2.7.3 Assessment of NP marker gene expression with real-time quantitative-PCR All samples were performed in triplicate, with three experimental repeats per experiment and three biological replicates. For each gene analysis, the following mastermix was made per well: ABI SYBR Green mastermix (1x), forward primer (optimised concentration: see primer table), reverse primer (optimised concentration), complementary DNA (0.5 ng/µL) and topped up to 20 µL with molecular grade H2O. In negative controls, molecular grade water was used in place of cDNA. Next, the ABI 96-well plate was sealed, centrifuged at 500 x g for 30 s and placed in a StepOnePlus Real-Time PCR machine (Applied Biosystems, Thermo Fisher Scientific, UK) and analysed for gene expression using StepOnePlus software (V2.2 and 2.3). The primer sequences and working concentrations used can be found in table 2.1.

Table 2.1. Primer sequences and working concentrations. Primer Working Forward sequence Reverse sequence Name Concentration GGGAGGAGACGACT CCCATTCCGTCTTGTTT ACAN 300 nM GCAATC TCTG GGGAAGCCTCACAT GGACATTACCTCATGG SOX9 CGACTTC 300 nM CTGATCT

CATGACCGAGACGT TTGCCGTTGTCGCAGAC COL1 GTGGAA A 300 nM

CGGGCTGAGGGCAA CGTGCAGCCATCCTTCA COL2 CA GA 300 nM

TGCCGCCTGGAGAA CGCCTGCTTCACCACCT GAPDH ACC T 300 nM

GATACGCAAGCCCC TGGGCGACCCAGGAAA MMP-3 300 nM GATGT G CTTCTACTGGCGCGT GAAGGTCACGTAGCCC MMP-9 300 nM GAGTTC ACATAGT TCCGCGGAGAAACA TTCAACCTGCTGAGGAT MMP-13 900 nM CTGATC GCA TCAGGAAATTCAGGT CGTGTATTCACCATTGA ADAMTS4 ACG G 900 nM

CGCTTAATGTCTTCC GGATCTGCTTTCGTGGT ADAMTS5 ATCCTTA AG 900 nM

2.8 Statistical Analysis Statistical significance for aggrecan structure, collagen fibril diameter characterisation and mechanical data was calculated using the non-parametric Mann-Whitney U test (p<0.05) as the data was not normally distributed. Statistical analysis of parametric data was performed using the unpaired two-sample t-test on the following: collagen periodicity, birefringence and in situ gelatinase zymography. Real-Time PCR data was assessed in InStat software (GraphPad, US) using the rank based Kruskal-Wallace H non-parametric test.

3 Chapter 3: Characterisation of aggrecan and collagen structure in young, healthy cartilaginous tissues

3.1 Introduction In mammals, avascular and ECM-rich cartilaginous tissues resist cyclical compressive loads in articular joints and the IVD. In these tissues, age-related degeneration has major clinical consequences causing osteoarthritis and IVD degeneration contributing to chronic low back pain (Roberts et al., 2006, Zhao et al., 2007, Colombier et al., 2014, Greene and Loeser, 2015, Malfait, 2016, Wang et al., 2016). Although, collectively, these disorders cost in excess of $100 billion in the US alone, the pathological mechanisms remain poorly understood and it has not yet proven possible to reverse accumulated damage or engineer mechanically competent replacement tissues (Vedicherla and Buckley, 2016, Zeckser et al., 2016).

In order to identify the pathological mechanisms which initiate and drive degeneration and hence develop effective treatment strategies, it is necessary to characterise the composition and structure of healthy cartilage. Both AC and the NP of the IVD are composed primarily of collagen II fibrils and the large proteoglycan aggrecan, which resist tensile and compressive forces, respectively (Yu et al., 2007, Smith et al., 2011). However, there are also differences between the tissues; mechanically, the NP is significantly more compliant than AC and contains a substantially higher proportion of aggrecan (Mwale et al., 2004). Despite these differences, and in common with other ECM-rich tissues, these proteins are not only highly stable (half-lives are measured in decades) but also function as large, supramolecular assemblies (Sivan et al., 2008, Sivan et al., 2014). As a consequence, transcriptomic and proteomic approaches to characterising cartilage are of limited value and must be complemented with methods which can analyse the macro- and supra-molecular structures of collagen II and aggrecan.

Our understanding of the role played by aggrecan structure in mediating tissue mechanical properties is derived primarily from studies of AC (Ng et al., 2003, Kopesky et al., 2010, Lee et al., 2010, Lee et al., 2013). From this work it is clear that the structures of both the aggrecan core protein and GAG side chains are dependent on age, species and pathology and influence the mechanical behaviour of isolated molecules. Our knowledge of IVD aggrecan structure is more limited. Thirty years ago, Buckwalter and colleagues used electron microscopy to show that the mean core protein length of aggrecan monomers differed between AC and IVD, decreased with age and varied within each tissue (Buckwalter et al., 1985, Buckwalter et al., 1989, Buckwalter et al., 1994). Although these observations of

56 aggrecan fragmentation in young healthy tissues have been confirmed by more recent studies (Lee et al., 2010, Lee et al., 2013) the extent of aggrecan fragmentation remains poorly characterised. Instead, attention has focussed on the intact aggrecan population (defined as containing all three globular domains) (Kopesky et al., 2010, Lee et al., 2010, Lee et al., 2013, Roughley et al., 2014).

The perception of aggrecan as a large, uniformly structured bottlebrush is clear from schematics of cartilage structure (Sophia Fox et al., 2009, Sivan et al., 2014). However, there is evidence that aggrecan is highly fragmented in cartilaginous tissues (Morgelin et al., 1988). In order to recognise age- and disease-induced molecular damage in cartilaginous tissues and design functionally competent engineered tissue replacements, it is imperative to understand the structural differences between AC and NP-derived aggrecan and quantify the relative proportion of intact and non-intact molecules.

3.1.2 Hypothesis and aims The hypothesis of this Chapter was that aggrecan and collagen II ultrastructure will be variable between NP and AC tissue which will affect the mechanical function of these tissues.

The aims of this Chapter were to:

(i) Use AFM combined with solution biophysical methods to characterise and compare the structure of all aggrecan molecules isolated from the AC and NP of young, healthy skeletally immature and mature bovine tissue.

(ii) Use nanomechanical, histological and zymography approaches to confirm that these tissues were non-degenerate and not subject to excessive protease activity.

(iii) Use computational modelling to determine the effect of aggrecan fragmentation on molecular packing.

3.2 Materials and methods 3.2.1 Experimental design This chapter characterises and compares the structure of aggrecan isolated from healthy native skeletally immature and mature bovine AC and NP. Aggrecan structure was characterised adsorbed to a surface using AFM and in solution using SEC-MALS (detailed in Chapter 2.1.6). As the second major ECM protein in cartilaginous tissues, collagen structure and abundance was also assessed by AFM (detailed in Chapter 2.3.2), and picrosirius red staining (detailed in Chapter 2.3.1), respectively. AFM and SEC-MALS were used to characterise and compare the structure of all aggrecan molecules isolated from the AC and NP of young, healthy, skeletally mature cows and new-born, skeletally immature cows. Nanomechanical (detailed in Chapter 2.4), histological (detailed in Chapter 2.3.1) and in situ gelatinase zymography (detailed in Chapter 2.5) approaches were used to confirm that these tissues were non-degenerate and not subject to excessive protease activity. Finally, the effect of aggrecan fragmentation on molecular packing was modelled computationally (detailed in Chapter 2.6).

3.3 Results 3.3.1 The ultrastructure of intact aggrecan monomers was tissue-specific Intact aggrecan molecules adsorbed to surfaces can be readily identified and their ultrastructure characterised by AFM or TEM (Buckwalter et al., 1985, Kopesky et al., 2010, Lee et al., 2010, Lee et al., 2013). Here, we used AFM to compare three structural parameters of intact AC- and NP-derived aggrecan (LCP, LGAG and MAGAG) (Figure 3.1a). Mean aggrecan LCP was significantly lower for NP- (395 ± 36 nm) compared to AC- (460 ± 30 nm) derived aggrecan (p<0.05: Figure 3.1b). Similarly, the mean lengths (LGAG: NP = 310 ± 39 2 nm and AC = 393 ± 31 nm) and areas (MAGAG: NP = 9091 ± 3301 nm , AC = 12652 ± 3573 nm2) of the glycosylated regions were significantly shorter and smaller respectively in NP compared with AC intact aggrecan (p<0.05; Figure 3.1a). Average GAG chain length was significantly different between AC and NP (p<0.01, 18.1 nm and 16.3 nm, respectively). It was clear therefore that the ultrastructure of intact aggrecan differs between tissues.

3.3.2 Most isolated aggrecan molecules were fragmented and structurally heterogeneous In a typical AFM height image of AC and NP-derived aggrecan (Fig. 2.2a and b) only two out of ~100 (NP) and two out of 16 molecules (AC) were intact. The remainder were either missing the G3 domain or had no visible globular domains and varied considerably in length (Fig. 3.2c). In the total molecular population (AC n=933; NP n=1061), fragmented aggrecan accounted for 95% of the observed molecules in AC and 99.5% in NP. Aggrecan fragmentation was higher in NP compared with AC in all three biological replicates and consequently the mean molecular area was significantly lower for NP compared with AC 2 2 (MAGAG: 4625.47 nm and 8543.07 nm respectively: p<0.0001) (Fig. 2D). Aggrecan fragmentation has previously been identified as a consequence of ageing (Sivan et al., 2014). Therefore, to determine if fragmentation in mature, yet young (<30 month), animals was indicative of early onset age-related degeneration we next isolated aggrecan from the AC of neonatal calves. This immature bovine AC contained an even lower proportion of intact aggrecan (0.5%, n = 792) than mature AC (5%) and the mean molecular area of the total 2 2 aggrecan population was also significantly lower ((MAGAG: 5137 nm and 8667 nm respectively: p<0.0001) (Fig. 2.2f).

Figure 3.1. Surface-adsorbed intact aggrecan monomers were structurally dissimilar in nucleus pulposus and articular cartialge tissues. (a) AFM height image (scale bar 0.2 µm; z-scale 2 nm) of an intact aggrecan monomer comprising three globular (G1—G3) domains. Three structural parameters were measured for each intact monomer: core protein length

(LCP), GAG length (LGAG) and molecular area (MAGAG). (b) Combined frequency distributions and mean (+/-SD) of LCP, LGAG and MAGAG for n=3 NP and AC replicates. (c) Relative mean molecular dimensions of intact AC- and NP-derived aggrecan monomers.

Figure 3.2. Both articular cartilage and nucleus pulposus were composed primarily of fragmented aggrecan. (a and b) AFM height images of AC (a) and NP (b) isolated aggrecan molecules (2 x 2 µm; scale bar is 0.2 µm; z-scale 2 nm). Arrows indicate intact monomers. (c) Only a small proportion of aggrecan molecules were in the intact form, most molecules lacked either the G3 domain alone or all globular domains. (d) Molecular area size distribution of all observable (i.e. both intact and non-intact) aggrecan molecules, derived from skeletally mature bovine AC and NP tissue (n=3 biological replicates). (e) Representative AFM image of aggrecan isolated from immature AC (image is 2 x 2 µm; scale bar is 0.2 µm); (f) Molecular area size distribution of fragments of all observable aggrecan regardless of fragment size or orientation, derived from skeletally immature bovine AC tissue (n=2 biological replicates). Values shown within the histogram are the mean (± SD).

In order to exclude the possibility that the apparent dominance of fragmented aggrecan was due to preferential adsorption of non-intact species and also to avoid the influence of surface chemistry on molecular structure we used a complimentary biophysical technique (SEC- MALS) to characterise aggrecan structure in solution (Sherratt et al., 2004, Chandran et al., 2010, Kienle et al., 2014). SEC-MALS was used to assess the mean absolute molecular weight average (Mw), polydispersity (a measure of structural heterogeneity) and elongation (RMS conformation) of aggrecan isolated from mature NP and AC tissue (n=3 biological replicates) at physiological salt concentrations. Although there was no significant difference in the Mw of AC- and NP-derived aggrecan (1762 ± 692 kDa and 867 ± 139 kDa, respectively; Figure 3.3a) the AC-derived aggrecan was more elongated compared to NP- derived (RMS values of 0.54 and 0.45 respectively). Additionally, NP-derived aggrecan was approximately 2—3 times more polydisperse than AC-derived (5.95 and 2.28 respectively, p<0.001; Figure 3.3b). Thus, whether characterised in solution or adsorbed to a surface (by AFM) there were structural differences between AC- and NP-derived aggrecan and both tissues are composed of highly fragmented and heterogeneous molecules.

Figure 3.3. Polydispersity of solution aggrecan indicated that it was highly fragmented and structurally heterogeneous. The structure of AC- and NP-derived aggrecan was characterized by SEC-MALS in PBS: (a) absolute molecular weight and (b) polydispersity index.

3.3.3 Aggrecan fragmentation was not associated with aberrant structural or mechanical tissue remodelling or excessive protease activity As molecular fragmentation is usually considered to be indicative of ageing and/or disease (Sztrolovics et al., 1997, El Bakali et al., 2014), we next characterised the histological

62 composition, collagen ultrastructure and micro-mechanical properties of young, mature AC and NP (Fig. 3.4). Whilst there are well-established GAG-specific stains such as Alcian blue and Safranin O, there are no reliable techniques for the histological quantification of aggrecan. In contrast, fibrillar collagen content can be semi-quantitatively characterised histologically by measuring PSR-enhanced collagen birefringence (McConnell et al., 2016). In accordance with previous studies of healthy tissue (Issy et al., 2013, Iijima et al., 2014), AC was enriched in fibrillar collagen (75.5% ± 13.6%) compared with NP 33.0% (± 7.60%; p<0.01). Conversely, NP was qualitatively also enriched in GAGs compared with AC (Fig. 3.4). The distribution of these ECM components was consistent with healthy tissues (Fig. 3.4).

Figure 3.4. Young, mature tissues appeared histologically healthy. Images are representative of three biological replicates for mature and immature AC and NP tissue. Scale bars are 100 µm. (a—b) bright-field images of picrosirius red staining (mature); (c— d) polarized light images of picrosirius red staining revealing fibrillar collagen (mature); (e—f) Safranin O with Fast Green staining (mature); (g—h) Alcian Blue pH 2.5 staining (mature); (i—j) Safranin O with Fast Green staining (immature); (k—l) Alcian Blue pH 2.5 staining (immature).

AFM and PSR staining was used to characterise collagen fibril morphology and architecture (periodicity and orientation), both in situ and as isolated fibrils (diameter). Collagen fibril periodicity was invariant between AC and NP (mean of 63.48 nm) (Fig. 3.5a) and both tissues were composed of two fibrillar populations which differed in diameter (thick: AC, 39.3 nm; NP, 41.4 nm and thin: AC, 20.03 nm; NP, 19.93 nm) (Fig. 3.5d). The presence of thick and thin collagen II fibrils in AC is well established (Holmes and Kadler, 2006) but to our knowledge this is the first report of thick collagen fibrils in NP. In these samples, no evidence of fibril alignment was observed either by AFM (Figure 3.5c) or histologically (AC alignment: 0.076%, NP alignment (0.096%) (Fig. 3.4c and d)

As ECM structure mediates the mechanical properties of tissues (Akhtar et al., 2011), pathological fragmentation of aggrecan would be expected to alter the compressive modulus (Sivan et al., 2014). Using AFM indentation the mean reduced modulus of AC was 497 kPa (± 232 kPa) which is comparable with previously reported values (Fig. 3.6) (Treppo et al., 2000). NP tissue was found to be significantly more compliant than AC (reduced compressive modulus of 76.7 kPa (± 48.1 kPa) (Fig. 3.6).

Finally, degeneration of cartilaginous tissues is commonly associated with the upregulation of protease expression and activity (El Bakali et al., 2014). However, in situ gelatin zymography (Fig. 3.7) demonstrated that regardless of age gelatinase activity was predominantly localised at the surface of AC and homogenously distributed throughout the NP, with no significant difference between immature and mature AC and NP (immature AC: 50.25 ± 29.66 pixels, immature NP 51.81 ± 32.55 pixels, mature AC: 58.37 ± 56.51 pixels, NP 35.53 ± 10.45 pixels). Importantly, both rat skin epidermis and hypodermis exhibit significantly higher levels of fluorescence and therefore gelatinase activity (epidermis 128.9 ± 49.73 pixels, hypodermis 175.0 ± 15.13 pixels, p<0.0001) than either bovine AC or NP (Fig. 3.7B). It appears therefore that these tissues exhibit no signs of pathological remodelling with regards to collagen fibril and GAG composition, collagen fibril ultrastructure and alignment and micro-mechanical stiffness or ECM protease activity.

Figure 3.5. Collagen fibril ultrastructure and architecture in mature articular cartilage and nucleus pulposus. Data is representative of two biological replicates for AC and NP. (a) Topographical amplitude images representative of native AC and NP (proteogylcan was digested from NP tissue in order to reveal collagen fibrils). (b) Topographical height image of in situ collagen fibrils in AC and NP. (c) Orientation and periodicity of collagen fibrils in situ in AC and NP measured using 2D fast fourier transform. The first ring corresponds to the fundamental frequency (due to collagen periodicity) in the image. The presence of rings rather than spots indicates that the fibrils have no preferred alignment. (d) Topographical height image of isolated collagen fibrils highlighting two distinct populations present in AC and NP. Histogram fitted with Gaussian curves showing the distribution of the diameters of the two populations of collagen fibrils isolated from AC and NP.

Figure 3.6. Articular cartilage was significantly stiffer than nucleus pulposus. Data is representative of three biological replicates for AC and NP tissue. (a) Optical images representative of native AC and NP tissue identifying the location of mechanical measurements (1—4). Scale bar is 1 mm. (b) AFM nanomechnical data displayed as a histogram fitted with Gaussian curves. (c) Bar chart displays the mean reduced modulus (± SE).

Figure 3.7. Low levels of matrix metalloproteinase activity in young healthy bovine articular cartilage and nucleus pulposus tissue. (a) Haematoxylin and Eosin staining of healthy mature and immature (iAC, iNP) bovine AC and NP tissue and healthy rat skin. (b) Respective in situ gelatinase zymography. White lines and boxes indicate regions measured for mean fluorescent activity. (c) Mean fluorescent gelatinase activity between iAC (n=2), iNP tissue (n=2), AC (n=2), NP tissue (n=2), rat skin (n=2). All scale bars are 100 µm.

3.3.4 Aggrecan fragmentation was predicted to modulate molecular packing density In order to characterise the potential influence of aggrecan fragmentation on cartilage and NP structure, we initially used computational approaches to model the effects of molecular dimension on predicted packing density. Molecular dimensions in solution were characterised experimentally using MALS of isolated aggrecan from low (10 mM NaCl), physiological (136 mM) and high (1 M) salt conditions. Salt concentration is thought to mediate interactions between GAG chains and therefore aggrecan morphology (Ng et al., 2003, Chandran et al., 2010, Chandran and Horkay, 2012). Aggrecan elongation was inversely correlated with salt concentration (Ng et al., 2003, Chandran et al., 2010) and RMS conformation values for both NP and AC derived aggrecan were between 0.5—0.8, suggesting that the molecule has a rod-like structure in solution (Kulicke and Clasen, 2004).

Individual aggrecan molecules were therefore modelled as rigid cylinders with hemispherical ends. Population dimensions were modelled as experimental (drawn from the experimentally determined length distributions), intact and fragmented (one quarter length). Using the experimental population, we established that a cut-off of 700 failures indicated a virtually saturated volume (Fig. 8a). Using this cut-off and 12 computational simulations for each length distribution it was clear that the experimental length distribution not only resulted in a higher packing density (10.01% +/- 0.72%) than either intact (9.85% +/- 1.40%) or fragmented (9.27% +/- 0.75%) distributions (significant for experimental vs fragmented, p<0.05 Students t-test) but also reduced heterogeneity (Fig. 8b). These simulations suggest that partial fragmentation of aggrecan may confer an advantage by promoting uniformly dense packing and therefore higher charge density, osmotic potential and mechanical stiffness.

Figure 3.8. Simulation of aggrecan packing. (a) To determine an appropriate number of failures before the algorithm terminated we investigated the effect of this parameter on the predicted aggrecan volume fraction for intact aggrecan (bars represent standard deviation). The predicted volume fraction stops increasing after c=700 (arrow). (b) Predicted aggrecan volume fraction for each length distribution. The volume fraction predicted by the simulation of the experimental distribution has a high mean and low standard deviation compared with the intact and fragmented simulations. (c) Visual representation of the computational models of aggrecan packing within a set volume.

3.4 Discussion This study demonstrates that the molecular structure of aggrecan is highly heterogeneous in developing and mature (yet young) tissues and suggests, using modelling approaches, that programmed fragmentation may be an adaptive process to optimise tissue structure and function

Studies have previously focused on the structure of intact aggrecan (Roughley et al., 2014, Kopesky et al., 2010, Lee et al., 2010, Lee et al., 2013). In this Chapter, a small population of intact aggrecan monomers was observed that were structurally similar to those reported for equine AC tissues and equine BM-MSCs (Kopesky et al., 2010, Lee et al., 2010); however these intact monomers had a smaller LCP compared to those reported in young and adult human AC (Lee et al., 2013). Importantly, LCP was significantly smaller (by ~60 nm) in NP compared to AC-derived aggrecan due to a reduction in LGAG and MAGAG. This difference in LCP, combined with the reduced protein elongation in solution points to variability in the biosynthesis/glycosylation of aggrecan rather than alternative splicing which is restricted to the C-terminal G3 domain and does not appear to affect LCP (Sivan et al., 2014). Differential glycosylation could alter length as reduced GAG load may cause NP- derived aggrecan to adopt a lower energy state and hence a relaxed random coil as opposed to an extended, rigid structure (Ng et al., 2003). Variability in aggrecan structure may also be a consequence of differential sulfation as repulsion of adjacent negatively charged GAG chains along the aggrecan core protein can be modified by changes in solute concentration (Chandran and Horkay, 2012). Other groups have shown an age-related increase in KS chain length and abundance (Santer et al., 1982) and an increase in CS 4S/6S sulfation (Plaas et al., 1997, Bayliss et al., 1999, Lauder et al., 2001, Plaas et al., 2001a) which could potentially affect the overall charge, repulsion and spacing of these GAGs and therefore affect aggrecan core protein extension. In mature, intact AC and NP, mean GAG chain length was significantly different between tissues (p<0.01; Fig. 3.1) suggesting GAG-spacing mediated variation in core protein length is either due to variation in GAG elongation, sulfation patterning or occupation of GAG substitution sites.

The influence of structural difference in intact aggrecan on tissue function may however be marginal. We show that intact monomers only account for a small population (~0.5—5%) of the total molecular population. Although aggrecan fragmentation was observed more than twenty years ago in baboon AC and IVD by Buckwalter and colleagues (Buckwalter et al.,

1989), the ratio of intact to fragmented molecules was not well characterised and molecular fragmentation was assumed to be a consequence of ageing (Buckwalter et al., 1994). More recent studies have focused mainly on characterising the population of intact aggrecan (Kopesky et al., 2010, Lee et al., 2010, Lee et al., 2013). As this study has shown that non- intact aggrecan accounts for at least 95% of the molecules in both mature and developing tissues aggrecan fragmentation is unlikely to be maladaptive.

Enzymatic proteases are the most likely mediators of aggrecan fragmentation. Previously, gelatin zymography has been used to characterise relative aggrecanase activity in developing, diseased and degenerate tissues (Freemont et al., 1999, Gepstein et al., 2002, Weiler, 2013) and crucially, MMP activity in the IVD has been shown to be maximal during foetal development (Rutges et al., 2010). In our study, however, we detected only minimal gelatinase activity in neonatal and mature (yet young) tissues. Therefore, as both AC and the IVD continue to grow postpartum (as a consequence of new ECM synthesis), aggrecan fragmentation may peak during foetal development, but must also continue at low levels during development of the young animal. Identifying which proteases drive this fragmentation is complicated by the structural heterogeneity of the fragmented aggrecan and the low substrate specificity of many ECM proteases (Chakraborti et al., 2003). The following aggrecanases are expressed in either AC and/or IVD but have not been reported experimentally to cleave collagen II: MMP-2, 3, 7, 9, 10, 19, cathepsins B, D, G (IVD only) and H (AC only) and Calpain-2(m) (Vittorio et al., 1986, Okada et al., 1992, Hembry et al., 1995, Turk et al., 1997, Ohta et al., 1998, Nakase et al., 2000, Ariga et al., 2001, Konttinen et al., 2002, Chakraborti et al., 2003, Cawston and Wilson, 2006, Hembry et al., 2007, Minarowska et al., 2008, Ruettger et al., 2008, Salminen-Mankonen et al., 2007, Struglics and Hansson, 2010, Fukuta et al., 2011a, Gruber et al., 2011, Hamamura et al., 2013, Lipari and Gerbino, 2013, Vo et al., 2013). Putative cleavage sites for a sub-set of these enzymes (cathepsin D, calpain-2(m) and the MMPs 2, 3, 7 and 9) can be predicted using the machine learning approaches developed by the PROSPER project (Song et al., 2012) (Tables 3.1 and 3.2). Each of these enzymes is predicted to cleave both ECM components, but cathepsin D is the only enzyme with predicted cleavage sites in processed aggrecan, but not processed collagen. Calpain-2 is present in both the AC and IVD and is the most likely candidate for c-terminal truncation of aggrecan, but there are three predicted cleavage sites in collagen II (Oshita et al., 2004, Fukuta et al., 2011b).

Table 3.1. Predicted protease specific amino acid sequence cleavage sites of aggrecan. C-frag. Cleavage N-frag. Enzyme Cleavage sequence Size Score position size (kDa) (kDa) 29 DNSL SVSI 3.21 280.68 0.96 83 KEKE VVLL 9.22 274.67 1.00 Cathepsin-D 461 SEDL VVQV 54.45 229.43 0.99 699 VEEW IVTQ 83.27 200.62 0.99 1537 FSGL PSGF 180.24 103.65 0.99

1681 PDLS GQPS 197.24 86.65 0.95 Calpain-2 (m) 2070 PGLP SATP 242.21 41.68 1.00

MMP-2 1278 TAGD ISGA 149.63 134.26 1.00

952* GVGD LSGL 112.17 171.72 0.99 971* GVGD LSGL 114.39 169.50 0.96 MMP-3 1351 AVGD LSGL 158.03 125.85 0.95 1421 EVLE ISAS 166.22 117.67 0.96 1788 GIAE VSGE 209.91 73.97 1.00

263 SPEK FTFQ 30.73 253.16 0.97 512 SPEQ LQAA 60.48 223.41 0.95 952* GVGD LSGL 112.17 171.72 0.96 971* GVGD LSGL 114.39 169.50 0.94 MMP-7 1351* AVGD LSGL 158.03 125.85 0.98 1450 VGTD LSGL 169.69 114.20 0.94 1488 ASGD LDLG 174.20 109.68 0.99 1963* TAGD ISGA 230.31 53.58 1.00

868 TAGD ISGA 230.31 53.58 1.00 MMP-9 1963* TAGD ISGA 230.31 53.58 1.00 *=non-unique cleavage sites Grey text indicates predicted cleavage sites (from the PROSPER server) present in immature, pre-processed aggrecan (Song et al., 2012). Black text indicates cleavage sites present in mature, processed aggrecan.

Table 3.2. Predicted protease specific amino acid sequence cleavage sites of collagen II. C-frag. Cleavage N-frag. Enzyme Cleavage sequence Size Score position size (kDa) (kDa) 114 DIKD IVGP 13.42 175.35 0.96 Cathepsin D 143 KGEK GAPG 17.29 171.48 0.93 1477 EQEF GVDI 187.54 1.23 1.00

509 PGER GAPG 64.61 124.16 0.99 Calpain-2 (m) 716 QGLQ GPRG 91.64 97.13 1.00 1142 QGLP GPPG 146.54 42.23 0.98

499 EPGG VGPI 63.30 125.47 0.95 MMP-2 805 EKGE VGPP 103.22 85.55 1.00

37 CVQD GQRY 4.23 184.54 0.98 103 PGPK GQKG 11.95 176.82 0.96 111 EPGD IKDI 13.07 175.70 0.94 MMP-3 134 QGPR GDRG 16.00 172.77 1.00 344 AGAR GNDG 43.44 145.33 0.99 1451 TVIE YRSQ 184.37 4.40 0.95

358 PPGP VGPA 45.22 143.55 0.97 MMP-7 663 GPSG FQGL 84.59 104.18 1.00 1204 PPGN PGPP 154.38 34.39 0.94

222 GPQG FQGN 27.46 161.31 1.00 373 APGA KGEA 47.11 141.67 0.96 798 GPAG ANGE 102.31 86.47 0.99 MMP-9 834 GPAG FAGP 106.84 81.93 1.00 879 GPTG VTGP 112.53 76.24 1.00 1006 EPGK QGAP 128.95 59.82 0.94 Grey text indicates predicted cleavage sites (from the PROSPER server) present in immature, pre-processed collagen (Song et al., 2012). Black text indicates cleavage sites present in mature, processed collagen.

Protease driven fragmentation of aggrecan is likely to influence tissue function Previously aggrecan structure has been shown to mediate single molecule mechanics and GAG-GAG interactions are thought to account for 30—50% of the compressive modulus in AC (Eisenberg and Grodzinsky, 1985, Seog et al., 2005). However, these models assume that aggrecan structure is homogeneous (with regards to core-protein length and GAG sulfation) rather than, as demonstrated in this study, heterogeneous (both within and between tissues) (Figures 3.1—3.3). Our measurements of bovine AC stiffness are similar to those previously reported in humans (Treppo et al., 2000, Seog et al., 2005) and so the fragmentation which we observe did not adversely affect compressive modulus. Therefore, contrary to previous models (Eisenberg and Grodzinsky, 1985, Seog et al., 2005), mechanically competent tissues could be comprised of fragmented, non-HA associated aggrecan (Fig. 3.9). Here we also report for the first time the nanomechanical stiffness of NP which is approximately one fifth that of AC. As over 95% of aggrecan molecules are fragmented in both tissues the difference in compressive modulus between AC and NP are likely to be mediated by the aggrecan to collagen II ratio (Mwale et al., 2004).

Figure 3.9. Proteoglycan structure in cartilaginous extracellular matrix. Traditionally, aggrecan is depicted as an intact bottlebrush structure, forming aggregates through interaction with hyaluronan via its G1 domain. Data from this study challenges this conventional model; aggrecan is mostly fragmented within a cartilaginous matrix, whilst still being capable of generating an osmotic potential.

Aggrecan fragmentation may indirectly influence tissue stiffness by mediating molecular packing density. Aggrecan concentration for example has been shown to be positively correlated with osmotic potential in AC (Chandran and Horkay, 2012) whilst the density of aggrecan-bound GAGs can influence mobility of water, solutes and proteins by modulating osmolarity and porosity (Gribbon and Hardingham, 1998). At high aggrecan concentrations, molecular diffusion through the ECM is significantly reduced which may in turn influence the availability of oxygen and metabolites (Gribbon and Hardingham, 1998). Porosity and permeability also affect resistance to compressive forces; using an AC-mimetic hydrogel scaffold with a similar modulus to AC, tissue modulus was inversely correlated with micro- porosity and permeability (Woodfield et al., 2004, Vikingsson et al., 2015). As our computational simulations predict that a heterogeneous aggrecan population is desirable in order to optimise molecular packing, we suggest that enzymatic processing of aggrecan may be an important adaptive mechanism to mediate the structure and function of cartilaginous tissues.

3.5 Conclusions The observation of extensive aggrecan fragmentation challenges our understanding of the pathology and tissue engineering of cartilaginous tissues. The increase in aggrecanase activity which characterises degenerating tissues is likely to result in further degradation of aggrecan into non-functional fragments which are small enough to diffuse from the matrix. When considering the development of viable tissue engineered constructs for the repair of cartilaginous tissues, it may be necessary to replicate the structurally heterogeneous population which characterises native AC and NP.

4 Chapter 4: The effect of culture conditions on the nanostructure of aggrecan and mechanical properties in intervertebral disc tissue engineered constructs

4.1 Introduction Cartilaginous tissues are comprised of numerous ECM proteins, each having unique structural and bioactive roles. The most ubiquitous and arguably most structurally relevant of these are collagen II and the proteoglycan aggrecan. Steric hindrance and electrostatic interactions between adjacent GAG chains enable aggrecan to generate an osmotic potential enabling it to carry out its role as the major load bearing molecule. Chapter 3 demonstrated that extensive aggrecan fragmentation is a feature of healthy skeletally immature and mature native NP and AC tissues. Accumulation of aggrecan fragments or degradation products (as they are more commonly termed) have traditionally been associated with progression of cartilaginous tissue degeneration (Janusz et al., 2004, Sivan et al., 2014); however, increasing evidence (Kopesky et al., 2010, Lee et al., 2010), shows that aggrecan fragmentation is ubiquitous in perfectly healthy and newly synthesised tissues (Chapter 3.3.2). Fragmentation it seems, is actually a structural adaptation, and not necessarily indicative of pathology, which previous studies (Kopesky et al., 2010, Lee et al., 2010) have poorly discussed in the context of tissue function. Fragmentation could indirectly influence tissue stiffness by mediating molecular packing density, potentially affecting porosity, permeability and consequently the availability of oxygen and other metabolites (Grodzinsky, 1983, Gribbon and Hardingham, 1998, Jackson and Gu, 2009). Therefore, when considering the development of viable tissue engineered constructs for the repair of cartilaginous tissues, it may be necessary to recapitulate the structurally heterogeneous aggrecan population which characterises native AC and NP.

Tissue engineering of viable cartilaginous tissues that are mechanically robust and capable of performing their intended function requires the consideration of several factors; cell type (e.g. notochordal cells, NP cells, BM-MSCs, adipose-derived MSCs and chondrocytes), scaffold biocompatibility, delivery of therapy (e.g. injectable or implantable), appropriate growth factors to stimulate and maintain cellular differentiation, homeostasis and ECM synthesis, and local “niche”/microenvironmental factors (i.e. oxygen concentration and nutrient availability) (Chung and Burdick, 2008, Lee et al., 2010). Although it is possible to create cartilage-like ECM (i.e. with evidence of synthesis of proteoglycans, sGAG and collagen) in tissue engineered constructs the resultant material fails to recapitulate structurally and mechanically viable cartilaginous constructs capable of withstanding the environment in vivo (Hunziker, 2002, Gorensek et al., 2004, Iu et al., 2017).

Intervertebral discs have limited repair potential due to poor vascularisation (limiting oxygen and nutrient availability) which is particularly evident within the centre NP. Any tissue engineering or regenerative/repair strategy incorporating the implantation of cells into the NP will need to promote appropriate and substantial ECM synthesis in an extreme environment that limits a cells metabolism and its capacity to synthesise protein. Therefore, it is necessary to understand the factors that may influence extracellular matrix structure and composition, particularly those that may influence aggrecan structure and hence functional properties of a construct.

It is already known that a number of factors affect ECM synthesis, for example growth factors, oxygen and nutrient availability and cell type; however, this has primarily been studied using gene expression, histological assessment and gross mechanical characterisation of cartilaginous tissues. There is a lack of knowledge regarding the nanostructure of key ECM components (i.e. aggrecan), and how this influences the nanomechanical properties of IVD tissue engineered constructs. Previous work has shown that growth factors (e.g. insulin growth factor [IGF], TGFβ-1, 3, GDF-5 and 6) drive differentiation of MSCs into an NP-like cell phenotype and promote anabolic ECM synthesis in IVD cells (Lee et al., 2010, Stoyanov et al., 2011, Clarke et al., 2014, Colombier et al., 2016). However, some of these factors, in particular TGFβ1 and TGFβ3 have also been used to promote chondrogenesis of MSCs and adult chondrocytes, resulting in a tissue which is more akin to AC. Tissue synthesised in this manner exhibited greater levels of collagen II and lower levels of proteoglycan compared to NP resulting in a hyaline AC that was less hydrated and lacking elastic fibres and as such would not likely function appropriately in the NP (Gorensek et al., 2004). To date, only two previous studies have investigated the nanostructure of aggrecan synthesised in vitro. BM-MSCs cultured with TGFβ1 synthesised aggrecan monomers with greater LCP and GAG chain length compared to animal-matched adult chondrocytes (Kopesky et al., 2010) and native AC tissue (Lee et al., 2010). Longer GAG chains and core protein length were found to positively correlate with greater tissue stiffness (Lee et al., 2010). Although the mechanical stiffness of individual aggrecan monomers was shown to be modulated in response to cell culture manipulation, the overall mechanical function of the engineered tissue was not assessed. Emphasis was also placed on cell origin rather than cell culture conditions (i.e. growth factor and nutrient/oxygen availability). Other studies have shown that oxygen concentration affects the amount of proteoglycan synthesised by NP cells (Yang et al., 2017). However, this has only been

78 investigated in the context of gene expression and gross ECM synthesis using histology and biochemical analyses. Low oxygen concentration is also essential for maintaining IVD cell phenotype, metabolism and ECM synthesis by regulating expression of hypoxia inducible factor in IVD cells (Urban et al., 2004, Chen et al., 2014). Thus, whilst we have some understanding of the gross biochemical and macromechanical effect of bioactive factors and the niche NP environment on ECM synthesis, our knowledge of these factors on the molecular structure of key ECM proteins (e.g. aggrecan) is limited.

In recent years, emphasis has shifted from gross mechanical assessment of whole tissues to dissecting the complex interplay between individual protein structure and the micromechanical function of tissues which governs their overall gross mechanical function. When considering IVD tissue engineering, it is important that the tissue formed recapitulates native tissue, and as such aggrecan nanostructure should be similar so that the engineered tissue is mechanically robust and capable of performing its intended function. Importantly, it is vital to understand how factors in the IVD niche (e.g. cell type, oxygen and nutrient availability, and bioactive factors) affect synthesis and importantly aggrecan structure.

4.1.2 Hypothesis and aims The hypothesis of this chapter was that NP cells synthesise aggrecan with a molecular structure similar to aggrecan derived from native NP tissue, and manipulation of cell culture conditions would alter aggrecan structure.

Bovine cells were utilised rather than human, due to the limited availability of non- degenerate human NP samples. To test this hypothesis, we used AFM combined with SEC- MALS, histology, qPCR, micromechanical assessment and in situ gelatinase zymography to address the following aims:

(i) the effects of varying oxygen concentration in culture on aggrecan synthesis and structure

(ii) the effects of an exogenous growth factor on aggrecan synthesis and structure

4.2 Materials and methods 4.2.2 Experimental design This chapter investigates the effect of two independent culture conditions on the structure of aggrecan synthesised by bovine NP cells encapsulated in a type I collagen gel (i.e in an IVD tissue engineered construct): (i) the effect of oxygen (normoxia vs hypoxia), and (ii) the effect of exogenous addition of two growth factors (TGFβ3 and GDF6) in normoxia (Table 4.1). To investigate the first of these, bovine NP cells were cultured in the presence or absence of either TGFβ3 or GDF6 in normoxia (n=3; henceforth referred to as +TGFβ3 and +GDF6, respectively). The effect of hypoxia was assessed following culture of bovine NP cells (encapsulated in a type I collagen gel) in the absence of growth factor, in 2% O2 (n=3) and normoxia (n=3; NGF), but in the presence of NP induction media (components detailed in Chapter 2.1.6) to maintain NP cell phenotype.

Table 4.1. Experimental design for the manipulation of cell culture conditions. No Growth Factor TGFβ3 GDF6 Normoxia (n=3)   

Hypoxia (n=3) 

NP cells were cultured in a 3D culture system utilising an exemplar collagen I hydrogel. As hydrogels are hydrophilic jelly-like materials that exhibit biochemical properties similar to native ECM structures found within mammalian tissues (Gauvin et al., 2012). This makes them particularly useful as biomaterials for cell culture and tissue engineering strategies. Collagen I hydrogels are an exemplar system to encapsulate the NP cell and has previously been used in our lab to promote differentiation of MSCs into a NP cell-like phenotype (Clarke et al., 2014). In combination with growth factor stimulation, MSCs and NP cells cultured in collagen I demonstrate increased expression and deposition of key IVD ECM proteins (i.e. aggrecan and collagen II) (Bertolo et al., 2012, Lee et al., 2012, Clarke et al., 2014).

Aggrecan was isolated from the tissue engineered constructs following 28 days of culture as described in Chapter 2.1.4—2.1.6 and the molecular structure was assessed as described in Chapter 2.1.9. AFM was used to characterise the structure of intact and non-intact aggrecan adhered to a charged surface. ECM composition (proteoglycan, sGAG, non-fibrillar and fibrillar collagen content) was assessed as described in Chapter 2.1.10, and in turn this was

80 linked to micromechanical assessment (Chapter 2.1.12) of the NP-like constructs by AFM. Real-time PCR was used (Chapter 2.7) to assess the expression of anabolic genes for key ECM proteins (aggrecan [ACAN], collagen II [COL2] and SRY box-9 [SOX9]) and the major aggrecanases (ADAMTS-4, 5, and MMP-3, 9, and 13), whilst in situ gelatin zymography was performed (Chapter 2.5) to determine if aggrecanases (most of which exhibit gelatinase activity) were active in the newly synthesised tissues.

4.3 Results 4.3.1 The effect of oxygen on aggrecan synthesis and structure 4.3.1.1 Oxygen tension did not affect the proportion of intact aggrecan, but did affect molecular size In both normoxia and hypoxia, most of the aggrecan monomers were fragmented (Fig. 4.3a and b) however aggrecan fragments were significantly smaller in hypoxia (2936 nm2; p<0.0001) compared to normoxia (3924 nm2). This was supported by assessment of skeleton

LCP compared against molecular area (Fig 4.2): calculation of the linear gradient showed that a greater amount of GAG was present per unit length of core protein in hypoxia (0.0394x) compared to normoxia (0.0429x), i.e. the average length of GAG chains bound to aggrecan was greater under hypoxic conditions compared to normoxic. The reduced molecular area in hypoxia was due to decreased GAG length.

Figure 4.1. Newly synthesised aggrecan was mostly fragmented in nomoxia and hypoxia. (a and b) AFM height images of (a) normoxia and (b) hypoxia (2 x 2 µm; scale bar is 0.2 µm; z-scale 2 nm). Blue arrows indicate non-intact aggrecan. (c) Molecular area size distribution of all observable aggrecan molecules (i.e. both intact and non-intact) regardless of fragment size or orientation (n=3 biological replicates). Values shown within the histogram are the mean (± SD).

Figure 4.2. Normoxia derived aggrecan had greater glycosaminoglycan content per unit length of core protein. (a and b) Scatter graphs comparing LCP to molecular area of all observable aggrecan molecules (i.e. both intact and non-intact) regardless of fragment size or orientation (n=2 biological replicates). (a) Normoxia and (b) hypoxia.

4.3.1.2 Gene expression aggrecan was reduced in hypoxic conditions compared to normoxia There are no reliable techniques for the histological quantification of aggrecan, and whilst ACAN gene expression doesn’t necessarily translate to protein synthesis, quantification does provide insight into the effect of cell culture conditions. Real-time PCR assessment at day

28 of cell culture (Fig. 4.3) showed significantly lower ACAN gene expression in hypoxic conditions (p<0.05) compared to normoxic.

Figure 4.3. Chondrogenic gene expression in Day 28 constructs. Normoxia (NGF; n=3) and hypoxia (n=3). Bar charts display the mean relative gene expression (∆ct) normalised to the house keeping gene GAPDH (± SE). *p<0.05.

4.3.1.3 Extracellular matrix deposition was more homogeneous and fibrillar collagen content was substantially lower in hypoxic compared to normoxic conditions Qualitatively, proteoglycan staining intensity in normoxia treated TE IVD constructs was minimal and heterogeneously distributed (Safranin O/Fast Green); however, sGAG staining (Alcian Blue pH 2.5) intensity was higher and distribution localised to the centre of the pellet under normoxic conditions (Fig. 4.4). TE IVD constructs cultured in hypoxia showed evidence of homogenous deposition of proteoglycan and sGAG deposition. Fibrillar collagen content was semi-quantitatively characterised histologically by measuring PSR- enhanced collagen birefringence (Fig. 4.4) (McConnell et al., 2016). TE IVD treated constructs cultured in hypoxia showed evidence of collagen birefringence; however, it appeared unorganised. Under normoxic conditions, collagen appeared highly localised and organised (Fig. 4.4); however, the enhanced PSR/birefringence could be due to cellular synthesis of collagen I or re-organisation of existing collagen I within the gel. There was significantly less percentage area of organised fibrillar collagen staining in hypoxia compared to normoxia treated constructs (1.15% ± 1.18% vs 12.7% ± 5.3%, respectively, p<0.001). The distribution of fibrillar collagen was consistent between biological replicates (hypoxia, n=3; normoxia, n=2). In addition, high abundance of total fibrillar and non-fibrillar (Masson’s Trichrome) collagen was observed in hypoxia.

Figure 4.4. Histological staining of native NP tissue and tissue engineered intervertebral disc tissue. Picrosirius Alcian Blue Safranin Masson’s Picrosirius Red Red pH 2.5 O/Fast Green Trichrome (Polarised) (Brightfield)

+TGFβ3

+GDF6

Normoxia

Hypoxia

Images are representative of three biological replicates. Bright-field images show Alcian Blue pH 2.5 staining revealing proteoglycan, Safranin O/Fast Green staining revealing sGAG, Masson’s Trichrome revealing non-fibrillar collagen, and picrosirius red staining revealing fibrillar collagen under polarized light. Scale bars are 100 µm.

4.3.1.4 Hypoxia treated tissue engineered intervertebral disc constructs exhibited mechanical properties similar to native NP tissue In order to account for variability in the ECM composition and cell density through cross sections of the synthesized constructs (as determined by histology; Fig. 4.4), AFM

87 mechanical measurements were acquired at four areas from the construct surface to the middle of the interior (A1—4; Fig 4.5a, b, and c). The mean reduced modulus of normoxia TE IVD constructs was significantly stiffer compared to hypoxia treated TE IVD constructs (350.8 kPa ± 364.5 kPa vs 30.9 kPa ± 13.3 kPa, respectively; p<0.0001) (Fig 4.5e and f). One area (A4) in one of two biological replicates of hypoxia treated TE IVD constructs appeared to be an outlier in that it was significantly different from the other areas in the same sample and different from the data set obtained in the other biologically samples. As such analysis was undertaken following removal of this data set, giving a mean reduced modulus of 30.9 kPa (± 13.3 kPa). This data suggests that hypoxic conditions result in a more compliant, homogenous ECM.

Figure 4.5. Hypoxia treated constructs were more mechanically heterogeneous and stiffer compared to normoxia. Normoxia (n=3) and hypoxia (n=2). AFM micromechanical data across four tissue sites, displayed as a (a) histogram and (b) bar chart. Bar chart displays the mean reduced modulus (± SE). (c) Optical images of Safranin O/Fast Green stained constructs identifying the location of mechanical measurements (A1—4) in relation to proteoglycan content. (e and f) AFM micromechanical data of the average reduced modulus across all four sites, displayed as a (a) histogram and (b) bar chart. Scale bar is 100 µm.

4.3.1.5 Oxygen tension did not influence protease expression Quantitative PCR analysis was used to assess the gene expression of major aggrecanases ADAMTS4, 5, MMP3, 9 and 13 (Fig. 4.6). There was no significant difference in aggrecanase expression between normoxia and hypoxia treated TE IVD constructs.

Figure 4.6. Aggrecanase gene expression in hypoxia treated constructs after 28 days culture. Normoxia (NGF) (n=3) and hypoxia (n=3). Bar charts display the mean relative gene expression (∆ct) normalised to the house keeping gene GAPDH (± SE).

4.3.1.6 In situ gelatin analysis showed gelatinase activity was greater in hypoxia In situ gelatin zymography (Fig. 4.7) demonstrated that gelatinase activity was significantly greater in hypoxia (104.30 ± 10.61 pixels; p<0.0001) compared to normoxia (55.78 ± 26.32 pixels). Gelatinase activity was predominantly localised at the edge of the TE IVD construct in in normoxia, and homogeneous in hypoxia treated TE IVD constructs, which importantly showed higher levels of gelatinase activity throughout the construct. Therefore, increased activity was due to either differential activation of sequestered proteases or the expression of different proteases from those in Figure 4.7.

The next step was to determine the effect of growth factors TGFβ3 and GDF6 on aggrecan structure, in normoxic conditions, and relate this to the mechanical responsiveness of TE IVD constructs.

Figure 4.7. High levels of matrix metalloproteinase activity in the absence of a growth factor in hypoxia. (a) Haematoxylin and Eosin staining of engineered constructs. (b) In situ gelatinase zymography. White lines indicate regions measured for mean fluorescent activity. (c) Quantification of mean fluorescent gelatinase activity (n=3 biological replicates). Bar chart indicates the mean (± SE). All scale bars are 100 µm. ****p<0.0001.

4.3.2 The effect of growth factor on aggrecan synthesis and structure 4.3.2.1 Newly synthesised aggrecan molecules were less fragmented in the presence of GDF6 Regardless of the culture conditions, most of the aggrecan monomers were fragmented. In GDF6 and +TGFβ3 and +GDF6 TE IVDs (Fig. 4.8a and b) only 1.0% and 4.95% of monomers were intact. Aggrecan fragmentation was higher in +TGFβ3 compared with +GDF6 in two out of three biological replicates and consequently the mean molecular area was significantly lower in +TGFβ3 than +GDF6 (molecular area: 4038 ± 2536 nm2 and 4561 ± 3318 nm2 respectively; p<0.01) (Fig. 4.8). In normoxia, the degree of fragmentation resulted in similar sized fragments to +TGFβ3 TE IVD constructs (3924 ± 2181 nm2). Despite the presence of smaller fragments in +TGFβ3, a greater amount of GAG was present per unit length of core protein in +TGFβ3 (0.041x) compared to +GDF6 (0.0438x) and normoxia (0.0429x), as calculated by skeletal analysis (Fig 4.9). I.e. the average length of GAG chains bound to aggrecan was greater in +TGFβ3 compared to +GDF6 and normoxia.

Figure 4.8. Newly synthesised aggrecan was mostly fragmented. (a, b, and c) AFM height images of (a) +TGFβ3, (b) +GDF6, and (c) normoxia (2 x 2 µm; scale bar is 0.2 µm; z- scale 2 nm). (d) Average molecular area of all observable aggrecan molecules (i.e. both intact and non-intact) regardless of fragment size or orientation (n=3 biological replicates). Values shown within the bar chart are the mean (± SE). **p<0.01.

Figure 4.9. GDF6 treated tissue engineered intervertebral disc construct derived aggrecan had greater glycosaminoglycan content per unit length of core protein. (a, b, and c) Scatter graphs comparing LCP to molecular area of all observable aggrecan molecules (i.e. both intact and non-intact) regardless of fragment size or orientation (n=2 biological replicates). (a) +TGFβ3, (b) +GDF6, and (c) normoxia.

4.3.2.2 The ultrastructure of intact aggrecan monomers was similar between TGFβ3 and GDF6 treated tissue engineered intervertebral disc constructs In contrast to the oxygen tension experiments, where no intact aggrecan monomers were found; intact aggrecan monomers were observed in the presence of growth factors. In normoxia, no intact molecules were present. Structurally, there was no difference in three of the morphological parameters assessed for intact aggrecan monomers (LCP, LGAG and

MAGAG); however, the GAG chains (iGAGL) were noticeably longer and was found to be significantly greater in +GDF6 (15.3 nm; p<0.01) monomers compared to +TGFβ3 (12.8 nm; Fig 4.10). This was in contrast to assessment of skeleton LCP compared against molecular area (Fig 4.9): calculation of the linear gradient showed that a greater amount of GAG was present per unit length of core protein in +TGFβ3 (0.041x) compared to +GDF6 (0.0438x). In other words, according to skeletal analysis the average length of GAG chains bound to aggrecan was greater in +TGFβ3 compared to +GDF6.

Figure 4.10. Surface-adsorbed intact aggrecan monomers were structurally similar in +TGFβ3 and +GDF6 tissue engineered intervertebral disc constructs. (a) Four structural parameters were measured for each intact monomer: core protein length (LCP), GAG length

(LGAG), molecular area (MAGAG) and average GAG chain length (iGAGL). Combined frequency distributions and mean (+/-SD) of LCP, LGAG and MAGAG for n=3 TGFβ3+ and GDF6+ replicates. (b) AFM height image (scale bar 0.2 µm; z-scale 2 nm) of intact aggrecan monomers comprising three globular (G1-G3) domains. (c) Relative mean molecular dimensions of intact TGFβ3+ and GDF6+ derived aggrecan monomers.

4.3.2.3 collagen II and aggrecan gene expression was reduced in the presence of growth factors Real-time PCR assessment at day 28 of cell culture (Fig. 4.11) showed significantly lower COL2 and ACAN gene expression in +TGFβ3 TE IVD constructs (p<0.05) compared to no growth factor treated constructs. The addition of growth factor has no effect on SOX9 and can suppress ACAN and COL2 synthesis.

Figure 4.11. Chondrogenic gene expression in growth factor treated and untreated constructs cultured for 28 days. +TGFβ3 (n=2), +GDF6 (n=2), normoxia (n=3. Bar charts display the mean relative gene expression (∆ct) normalised to the house keeping gene GAPDH (± SE). *p<0.05.

4.3.2.4 Extracellular matrix deposition was similar between growth factor treated TE IVD constructs, but exogenous growth factors suppressed fibrillar collagen organisation GAG deposition appeared similarly abundant in TGFβ3 and GDF6 treated constructs and GAGs were homogeneously distributed (Fig. 4.4). In the absence of growth factors however, GAG staining was highly heterogeneous. In contrast, GDF6 appeared to suppress the deposition of total collagen compared with normoxia, whilst TGFβ3 had no effect. Again, collagen was heterogeneously distributed in the absence of growth factors. Organised collagen was largely absent from all constructs; however, growth factor addition was associated with suppressed deposition of organised collagen. Under normoxic conditions, both in the presence and absence of growth factor, there was no significant difference in percent area of fibrillar collagen staining (+GDF6: 19.1% ± 17.9%; +TGFβ3: 17.3% ± 15.8%; normoxia: 12.7% ± 5.3%). The distribution of fibrillar collagen was consistent between biological replicates (+GDF6: n=3; +TGFβ3: n=3; normoxia: n=2).

4.3.2.5 Both GDF6 and TGFβ3 reduced the construct stiffness AFM mechanical measurements were acquired at four areas from the construct surface to the middle of the interior (A1–4; Fig 4.12). The mean reduced modulus of normoxia TE IVD (350.8 kPa ± 364.5 kPa) constructs was mechanically more heterogeneous (Fig 4.12) and significantly stiffer overall compared to +TGFβ3 (69.9 kPa ± 77.5 kPa; p<0.0001) and +GDF6 TE IVD constructs (60.2 kPa ± 69.3 kPa; p<0.0001) (Fig 4.13). Mechanical assessment showed +TGFβ3 TE IVD constructs were significantly stiffer compared to +GDF6 (p<0.0001); however, in comparison to normoxia, they are mechanically similar.

100

Figure 4.12. Constructs were more mechanically homogenous in the presence of growth factor. TGFβ3+ N (n=3), GDF6+ N (n=3), and normoxia (n=3). AFM micromechanical data displayed as a (a) histogram and (b) bar chart. Bar chart displays the mean reduced modulus (± SE). (c) Optical images of Safranin O/Fast Green stained constructs identifying the location of mechanical measurements (A1—4) in relation to proteoglycan content. Scale bar is 100 µm.

101

Figure 4.13. +GDF6 and +TGFβ3 tissue engineered intervertebral disc constructs were mechanically similar, but are more compliant than no growth factor treated constructs. +TGFβ3 (n=3), +GDF6 (n=3), and normoxia (n=3). AFM micromechanical data displayed as a (a) histogram and (b) bar chart. Bar chart displays the mean reduced modulus (± SE). ****p<0.0001.

4.3.2.6 Gene expression analysis showed greater ADAMTS5 and MMP13 expression in +GDF6 tissue engineered intervertebral disc constructs Quantitative PCR analysis was used to assess the gene expression of major aggrecanases ADAMTS4, 5, MMP3, 9 and 13 in growth factor treated and untreated TE IVD constructs at 28 days of culture (Fig. 4.14). There was significantly greater ADAMTS5 expression in +GDF6 compared to +TGFβ3 TE IVD constructs (p<0.05) and significantly greater MMP13 expression in +GDF6 compared to TE constructs in normoxia (p<0.05) constructs.

102

Figure 4.14. Aggrecanase gene expression in growth factor treated and untreated constructs cultured for 28 days. TGFb3+ N (n=2), GDF6+ N (n=2), and normoxia (n=3). Bar charts display the mean relative gene expression (∆ct) normalised to the house keeping gene GAPDH (± SE). MMP13 data for TGFb3+ N has been omitted as the PCR was unsuccessful. *p<0.05.

103

4.3.2.7 In situ gelatin analysis showed similar gelatinase activity in both growth factor treated and untreated tissue engineered intervertebral disc constructs In situ gelatin zymography (Fig. 4.15) demonstrated that gelatinase activity was not significantly different between +GDF6 (47.98 ± 12.59 pixels), +TGFβ3 (64.98 ± 45.34 pixels), and normoxia (55.78 ± 26.32 pixels) treated TE IVD constructs. Gelatinase activity was predominantly localised at the edge of the construct in both growth factor treated and untreated TE IVD constructs.

Figure 4.15. Similar levels of matrix metalloproteinase activity in both the presence and absence of a growth factor. (a) Haematoxylin and Eosin staining of engineered constructs. (b) In situ gelatinase zymography. White lines indicate regions measured for mean fluorescent activity. (c) Quantification of mean fluorescent gelatinase activity (n=3 biological replicates). Bar chart indicates the mean (± SE). All scale bars are 100 µm.

104

4.3.3 Manipulation of cell culture conditions replicated the extracellular matrix and mechanical properties of native nucleus pulposus tissue NP cells were treated according to two regimes; (i) oxygen tension and (ii) the presence of growth factors. Exposure of NP cells to hypoxia resembled some aspects of NP-like ECM; however, oxygen concentration alone was insufficient to stimulate the synthesis of an NP- like ECM. Treatment of NP cells with growth factors in normoxic conditions promoted the deposition of an ECM, more mechanically similar to NP-like matrix. The following sections compare the synthesised ECM of TE IVD constructs treated under these two regimes, to native NP tissue.

4.3.3.1 In the absence of growth factor, nucleus pulposus cell constructs were structurally and mechanically dissimilar to native tissue Prior to manipulation of cell culture conditions, aggrecan structure was assessed in normoxia, to determine if NP cells alone synthesised a matrix similar to native NP. No intact monomers were identified under NGF conditions, compared to 1.1% intact monomers in NP; however, assessment of all observable aggrecan (Fig. 4.16) showed that in NGF, fragmentation resulted in significantly smaller fragments overall in NP (3560 ± 2179 nm2) compared to newly synthesised aggrecan derived from normoxia TE IVD (3924 ± 3924 nm2; p<0.01). In addition, unlike native NP tissue, NGF TE IVD constructs were highly heterogeneous both histologically (Fig. 4.4) and mechanically (Fig. 4.17). There was significantly less fibrillar collagen in normoxia (12.7% ± 5.3%; p<0.01) compared to native NP tissue (33.0% ± 7.60%). Normoxia TE IVD constructs were significantly less compliant (350.8 ± 364.5 kPa) compared to native NP tissue (76.7 ± 48.1 kPa; p<0.0001) (Fig. 4.17).

105

4 0 0 0 0 * * * * * * * * * *

3 5 0 0 0 * * * * )

2 3 0 0 0 0

(

a 2 5 0 0 0

r 2 0 0 0 0

u c

e 1 5 0 0 0

l o

M 1 0 0 0 0

5 0 0 0

0 N P + T G F b 3 + G D F 6 N o rm o x ia H y p o x ia

C o n str u ct

Figure 4.16. Native nucleus pulposus and newly synthesised tissue engineered intervertebral disc aggrecan was mostly fragmented. Box and whisker plot of molecular area size of all observable aggrecan molecules (i.e. both intact and non-intact) regardless of fragment size or orientation (n=3 biological replicates). Box shows the mean and whiskers indicate the minimum and maximum values. **p<0.01; ****p<0.0001.

106

2 0 0 0 1 9 0 0 * * * * 1 8 0 0 1 7 0 0 1 6 0 0

1 5 0 0 )

a 1 4 0 0 P

k 1 3 0 0

(

s 1 2 0 0 u

l 1 1 0 0 u

d 1 0 0 0 o

M 9 0 0

e 8 0 0 c

u 7 0 0 d

e 6 0 0 R 5 0 0 4 0 0 3 0 0 2 0 0 1 0 0 0 -1 0 0 N P + T G F b 3 + G D F 6 N o rm o x ia H y p o x ia

T issu e

Figure 4.17. +GDF6 and +TGFβ3 tissue engineered intervertebral disc constructs were mechanically similar to nucleus pulposus tissue. NP (n=3) +TGFβ3 (n=3), +GDF6 (n=3), normoxia (n=3), and hypoxia (n=2). AFM micromechanical data displayed as a box and whisker plot. Boxes display the mean reduced modulus and whiskers show the minimum and maximum values. ****p<0.0001.

Under normoxic conditions — in the absence of growth factor — nucleus pulposus cells did not synthesise TE IVD tissue that resembled native nucleus pulposus tissue. Therefore, NP cells were exposed to hypoxic conditions; however, this was also found to be insufficient to stimulate the synthesis of an NP-like ECM.

4.3.3.2 Hypoxia treated nucleus pulposus cells did not synthesise tissue engineered intervertebral disc tissue similar to native nucleus pulposus tissue Following normoxic conditions, the capacity for hypoxia to stimulate NP cells to synthesise native NP tissue was assessed. Aggrecan structural anaylsis found no intact monomers and assessment of all observable aggrecan (Fig. 4.17) showed that in hypoxia, fragmentation resulted in substantially smaller fragments overall in hypoxia TE IVD constructs (2963 ± 1617 nm2) compared to NP (3560 ± 2179 nm2; p<0.01). Exposure to hypoxia stimulated NP

107 cells to synthesise a histologically (Fig. 4.18) and mechanically (Fig. 4.19) homogeneous extracellular matrix. However, hypoxia TE IVD constructs were significantly more compliant (30.9 ± 13.2 kPa) compared to native NP tissue (76.7 ± 48.1 kPa; p<0.0001) (Fig. 4.19). In addition, there was significantly less fibrillar collagen both in hypoxia TE IVD constructs (1.15% ± 1.18%; p<0.001) compared to native NP tissue (33.0% ± 7.60%).

Under normoxic conditions — in the presence of a growth factor (i.e. TGFβ3 and GDF6) — nucleus pulposus cells were found to synthesise TE IVD tissue that resembled some aspects of native nucleus pulposus tissue.

4.3.3.3 Growth factor treated tissue engineered intervertebral disc construct derived aggrecan was structurally similar native nucleus pulposus Addition of both GDF6 and TGFβ3 in normoxia resulted in an ECM similar to native NP tissue. Aggrecan structural analysis found a similar proportion of intact monomers in +TGFβ (1.0%) and a greater number in +GDF6 TE IVD constructs (4.95%) compared to NP (1.1%).

Intact aggrecan monomers (Fig. 4.20) were similar in LCP (+TGFβ: 378 ± 24 nm; +GDF6:

372 ± 27 nm; NP: 389 ± 37 nm) and LGAG (+TGFβ: 299 ± 24 nm; +GDF6: 298 ± 25 nm;

NP: 310 ± 39 nm); however, the MAGAG was significantly smaller in +TGFβ (7368 ± 1767 nm2; p<0.05) and +GDF6 (7894 ± 2325 nm2; p<0.05) TE IVD constructs compared to NP 2 (9091 ± 3301 nm ). In addition, iGAGL was found to be significantly smaller in +TGFβ3 TE IVD construct aggrecan (12.8 ± 2.8 nm; p<0.001) compared to NP (16.3 ± 4.1 nm) (Fig. 4.20). Assessment of all observable aggrecan (Fig. 4.17) showed that in both +TGFβ3 and +GDF6 TE IVD constructs aggrecan fragments were significantly larger (+TGFβ3: 4038 ± 2535 nm2; +GDF6: 4561 ± 3318 nm2; p<0.0001) compared to NP (3560 ± 2179 nm2; p<0.01). Exposure to both growth factors stimulated NP cells to synthesise a histologically (Fig. 4.18) and mechanically (Fig. 4.19) homogeneous extracellular matrix similar to native NP tissue. However, both +TGFβ3 and +GDF6 TE IVD constructs were significantly more compliant (+TGFβ3: 69.9 ± 77.5 kPa; +GDF6: 60.2 ± 69.3 kPa; p<0.0001) compared to native NP tissue (76.7 ± 48.1 kPa; p<0.0001) (Fig. 4.19). In addition, there was significantly less fibrillar collagen both in +TGFβ3 (17.3% ± 15.8%; p<0.01) and +GDF6 TE IVD constructs (19.1% ± 17.9%; p<0.01) compared to native NP tissue (33.0% ± 7.60%).

108

Figure 4.20. Surface-adsorbed intact aggrecan monomers were structurally similar between nucleus pulposus, +TGFβ3, and +GDF6 tissue engineered intervertebral disc constructs. (a) Four structural parameters were measured for each intact monomer (n=3 biological replicates): core protein length (LCP), GAG length (LGAG), molecular area

(MAGAG) and average GAG chain length (iGAGL). (b) Box and whisker plot of mean LCP,

LGAG, MAGAG, and iGAGL. Box shows the mean and whiskers indicate the minimum and maximum values. (c) AFM height image (scale bar 0.2 µm; z-scale 2 nm) of intact aggrecan monomers. *p<0.05; ***p<0.001.

109

4.4 Discussion This study aims to provide insight into the optimisation of culture conditions for the fabrication of TE IVD constructs with a matrix akin to native tissue in order to function appropriately. Here, we demonstrate that the molecular structure of aggrecan is highly heterogeneous in newly synthesised cartilaginous tissue similar to native NP tissue, but can be modified by manipulation of cell culture conditions. This study suggests that both protein molecular structure and deposition need to be optimised to engineer viable NP tissue replacements.

4.4.1 The effect of oxygen tension and growth factors on stimulating NP cells to synthesise aggrecan similar to native nucleus pulposus tissue 4.4.1.1 Normoxia and hypoxia Assessment of aggrecan structure, protein expression, deposition (i.e. fibrillar collagen, proteoglycan, sGAG content) and tissue micromechanics showed that culturing NP cells in normoxia, in the absence of a growth factor resulted in TE IVD constructs that were histologically and mechanically heterogeneous and much less compliant than native NP tissue. Therefore, normoxia alone was unsuitable for stimulating ECM synthesis similar to native NP.

Compared to cells grown in ambient oxygen (i.e. normoxia), under reduced oxygen (i.e. hypoxia), aggrecan structure was significantly more fragmented, which correlated with greater gelatinase activity. However, aggrecanase gene expression data showed it was unlikely that ADAMTS4, 5, MMP-3, 9, or 13 were the aggrecanases responsible for the observed proteolysis of aggrecan. According to the literature, MMP13 exhibited higher (although not significantly) expression in normoxia compared to hypoxia, which supports this data (Yang et al., 2013).

Similar to ambient oxygen, reduced oxygen was also found to be insufficient for stimulating ECM synthesis of NP-like tissue; however, hypoxia (histologically/qualitatively) stimulated greater proteoglycan deposition, and the synthesised tissue was histologically homogeneous similar to native NP tissue. In contrast to histological data, ECM gene expression analysis showed that hypoxia reduced aggrecan expression compared to normoxia, with a similar trend in collagen II. This may have been due to the late time point at which samples were collected for analysis; hypoxia-induced aggrecan and collagen II synthesis may have

110 plateaued. Although high total non-fibrillar collagen was observed under reduced oxygen, very little organised fibrillar collagen was synthesised, which contributed to the more compliant ECM. As a consequence, hypoxia TE IVD constructs were much less compliant than native NP tissue. In comparison, under normoxic conditions, collagen was highly localised and organised, resulting in a much stiffer ECM.

Taken together, neither ambient nor reduced oxygen conditions were sufficient to stimulate NP cells to synthesise an NP-like ECM and synthesised TE IVD constructs under these conditions are therefore likely be inappropriate as an NP replacement.

4.4.1.2 TGFβ3 and GDF6 Culturing NP cells in normoxia, in the presence of a growth factor resulted in TE IVD constructs that were histologically and mechanically homogeneous, exhibiting similar properties to native NP tissue. Assessment of aggrecan structure, protein expression, deposition (i.e. fibrillar collagen, proteoglycan, sGAG content) and tissue micromechanics showed that growth factors are required to stimulate ECM synthesis in order to recapitulate and ECM similar to native NP.

Intact aggrecan monomers have historically been used in a small number of studies to determine variability in aggrecan biosynthesis in response to manipulation of cell culture conditions (Lee et al., 2010). In this study, a small population (<10%) of intact aggrecan monomers derived from cultured NP cells was observed, that were structurally similar to those reported in Chapter 3 for bovine NP tissue. Whilst, manipulation of NP cell culture through exogenous addition of growth factor had no significant effect on LCP, and LGAG we did identify variation in MAGAG and iGAGL pointing to variability in the biosynthesis/glycosylation of aggrecan in response to exogenous addition of a growth factor. Previous work (Lee et al., 2010) has shown that variability in the biosynthesis/glycosylation of aggrecan exists. Adult BM-MSCs stimulated with TGFβ1 produce aggrecan similar in length to skeletally immature/growth plate-derived aggrecan, which are synthesised by chondrocytes derived from the mesenchymal cell lineage (Pittenger et al., 1999, Williams and Hare, 2011, Sheng, 2015). Adult articular chondrocytes stimulated with TGFβ1 produce aggrecan that has shorter LCP, individual GAG length and is more fragmented compared to BM-MSCS (Kopesky et al., 2010).

111

As previously mentioned, although our data has demonstrated that intact aggrecan monomers derived from cultured NP cells are similar to native NP aggrecan, intact monomers are in the minority and therefore, likely have little functional relevance. By characterising the degree of fragmentation of the entire aggrecan population, this study has shown that growth factors (i.e. TGFβ3 and GDF6) can modulate the length of GAG chains. Previous studies have shown that a variation in GAG length can affect steric hindrance and repulsive electrostatic forces between adjacent GAG chains, resulting in extension or contraction of the core protein (Kopesky et al., 2010, Chandran and Horkay, 2012); however, this was not observed in growth factor treated constructs.

Both TGFβ3 and GDF6 stimulated similar amounts of ECM synthesis resulting in an engineered tissue that was histologically similar to native NP tissue. In agreement with the literature, histological data suggested that both TGFβ3 and GDF6 stimulated proteoglycan and collagen II deposition (Clarke et al., 2014). In contrast to histological data, ECM gene expression analysis showed that both growth factors reduced aggrecan and collagen II expression compared to normoxia. This may have been due to the late time point at which samples were collected for analysis; growth factor induced aggrecan and collagen II synthesis may have plateaued by Day 28.

Bovine native NP tissue stiffness was similar to that reported for +TGFβ3 and +GDF6 TE IVD constructs, showing potential for utilising either of these growth factors in IVD tissue engineering. However, there was some variation between the compressive modulus of +TGFβ3 and +GDF6 TE IVD constructs. Both +TGFβ3 and +GDF6 TE IVD constructs exhibited similar amounts of unorganised fibrillar collagen and was therefore an unlikely contributor to the variation in tissue compliance seen between the two conditions. As over 95% of aggrecan molecules were fragmented in all newly synthesised construct tissues, the difference in compressive modulus in response to construct conditions was likely to be mediated by the aggrecan to collagen II ratio and not the degree of fragmentation (Mwale et al., 2004). Aggrecan fragmentation may have indirectly influenced tissue stiffness by mediating molecular packing density, as previously discussed (Chapter 3.4).

112

4.4.2 Aggrecan fragmentation may be due to the activity of constituent aggrecanase activity As shown in Chapter 3.3.2, non-intact aggrecan accounts for at least 95% of the molecules in cartilaginous tissues suggesting aggrecan fragmentation is unlikely to be maladaptive. Enzymatic proteases are the most likely mediators of aggrecan fragmentation. In situ gelatin zymography (Freemont et al., 1999, Gepstein et al., 2002, Weiler, 2013) was used to characterise relative aggrecanase activity in NP cell constructs; despite high levels of fragmentation we detected only minimal gelatinase activity in the presence of GDF6, TGFβ3 and in normoxia with no growth factor. Interestingly, gelatinase activity was greatest in hypoxia, where the greatest degree of fragmentation was observed.

Identifying which proteases drive this fragmentation is complicated by the structural heterogeneity of the fragmented aggrecan and the low substrate specificity of many ECM proteases (Chakraborti et al., 2003). A number of aggrecanases were chosen which have been reported in NP disc tissue (Sztrolovics et al., 1997, Sivan et al., 2014). Some of the most active aggrecanases in the disc are ADAMTS4, ADAMTS5, and MMP3, and the least active are MMP9, and 13 (Gendron et al., 2007, Durigova et al., 2011). Interestingly, +GDF6 TE IVD constructs, with the lowest levels of fragmentation, show the highest expression of ADAMTS5 and MMP13. This suggests that although GDF6 may promote MMP expression, this does not necessarily mean the protease is active. There is a large number of potential candidate proteases that may be responsible for cleaving aggrecan, which requires further investigation (Vittorio et al., 1986, Okada et al., 1992, Hembry et al., 1995, Turk et al., 1997, Ohta et al., 1998, Nakase et al., 2000, Ariga et al., 2001, Konttinen et al., 2002, Chakraborti et al., 2003, Cawston and Wilson, 2006, Hembry et al., 2007, Minarowska et al., 2008, Ruettger et al., 2008, Salminen-Mankonen et al., 2007, Struglics and Hansson, 2010, Gruber et al., 2011, Fukuta et al., 2011, Vo et al., 2013, Hamamura et al., 2013, Lipari and Gerbino, 2013).

4.5 Conclusions Aggrecan structure and NP cell ECM composition can be manipulated primarily by modulating growth factors rather than oxygen tension, and requires further investigation to understand the appropriate stimuli for NP tissue engineering. Structural and mechanical characterisation of TE IVD constructs suggested that hypoxia is necessary to stimulate ECM synthesis by NP cells and both GDF6 and TGFβ3 have the potential to promote synthesis of

113

NP like tissue; however, more work is required to determine which combination of oxygen concentration and growth factor is more appropriate to recapitulate NP tissue. In particular, this study lacks biochemical data (e,g, quantification of proteoglycan, sGAG and collagen) which would enable correlation analysis between the functional properties and matrix constituents of TE IVD constructs. Initially, assessment of sGAG content was attempted using the dimethylmethylene blue assay (Clarke et al., 2014); however, this was not achieved due to low aggrecan yield and insensitivity of the assay. Collagen content could have been determined by measuring hydroxyproline content using high precision liquid chromatography (Hardinham and Fosang, 1992); however, the necessary facilities and expertise were not available. Aggrecanase activity does appear to play a role in aggrecan fragmentation in newly synthesised cartilaginous tissue, but there is little direct evidence that aggrecan fragmentation affects mechanical function, as fragmentation is similar between cell culture treatment conditions. Aggrecan fragmentation may modulate porosity and osmolarity; however, aggrecan and collagen II deposition appeared to play a more prominent role in modulating mechanical function. When considering the development of viable tissue engineered constructs for the repair of NP tissue, it may be necessary to replicate the structurally heterogeneous aggrecan population which characterises native NP, as well as sufficient ECM deposition.

114

5 Chapter 5: Characterisation of aggrecan structure in young, healthy porcine tissues

115

5.1 Introduction Chapters 3 and 4 showed that aggrecan was highly fragmented in healthy skeletally immature and mature native bovine NP and AC tissue, as well as in newly synthesised TE IVD constructs using bovine NP cells. ECM of the porcine NP is synthesised by a population of notochordal cells (NCs), from which the smaller NP cells are thought to be derived from (Gottschalk et al., 2001, Yang et al., 2009, Rodrigues-Pinto et al., 2014). Highly gelatinous in comparison to bovine NP tissue and human adult IVD; the functional behaviour of porcine NP must be a consequence of its constituent components (Mwale et al., 2004). Aggrecan is the main water retaining molecule in cartilaginous ECM, and is therefore a likely candidate. This chapter aims to determine if the highly gelatinous nature of porcine NP tissue can be attributed to variation in aggrecan structure.

5.1.1 The origin and phenotype of notochordal cells The notochord develops from the mesoderm; enclosed within the peri-notochordal sheath which forms the outer and inner AF and vertebrae, the notchord condenses into the NP, which contains a population of large, highly vacuolated NCs. At birth, in mammals, such as humans, bovine and sheep, NCs are depleted (Hunter et al., 2003, Choi et al., 2008, McCann et al., 2012, Rodrigues-Pinto et al., 2014). There are two theories regarding the fate of NCs: (i) NCs differentiate into chondrocyte-like NP cells (Gottschalk et al., 2001, Yang et al., 2009, Rodrigues-Pinto et al., 2014), and (ii) NCs die and are replaced by migrating chondrocytes from the cartilage end plates (Kim et al., 2003). Other species, for example porcine and rodents, retain a high proportion of NCs cells throughout their lifetime (Hunter et al., 2004).

ECM structure plays a major role in maintaining NC differentiation and phenotype (Omlor et al., 2014, Navaro et al., 2015); in particular proteoglycan and GAG structure, which control osmolarity and oxygen concentration (by inhibiting neovascularisation and diffusion of solutes) (Pettway et al., 1996, Spillekom et al., 2014, Cornejo et al., 2015, Mongiat et al., 2016). ECM structure changes during development of the notochord (Pettway et al., 1996, Gotz and Quondamatteo, 2001, Hannesson et al., 2015), and consequently, so does the NC population, stimulating differentiation into chondrocyte-like NP cells (Oegema, 2002). However, this shift from NC to NP cell depends upon the species, suggesting that mammals that retain an NC rich population (e.g. porcine) comprise distinct variation in ECM composition (i.e. proteoglycan and GAG) compared to mammals that lose NCs (i.e. bovine

116 and human). NCs have a demarcated golgi and large endoplasmic reticulum, pointing to some variation in protein glycosylation compared to NP cells (Risbud et al., 2010). There is increasing evidence that NCs act as progenitor-like signal control cells that have the capacity to stimulate GAG and proteoglycan synthesis by NP cells (Aguiar et al., 1999, Arkesteijn et al., 2017), but synthesise little proteoglycan themselves (de Vries et al., 2015). Additionally, treatment of BM-MSCs with NC conditioned media has shown to stimulate proteoglycan and collagen synthesis to the same extent as exogenous addition of TGFβ (Potier et al., 2014).

Choosing the appropriate cell type for effective regenerative therapies is essential. It is necessary to understand the phenotypes of candidate cells, and how those cells respond to the ECM and bioactive factors within the local niche microenvironment. There are clear differences between articular chondrocytes, NP cells and NCs with regards to their phenotype, morphology and the structure/quantity of ECM components produced. Cells from different lineages (i.e. mesoderm vs notochordal) have been shown to synthesise aggrecan of different structures. Mesodermal in origin, adult articular chondrocytes and BM-

MSCS synthesise aggrecan (see Chapter 3.3.1) (Lee et al., 2010) that is longer in LCP compared to NP derived aggrecan (see Chapter 3.3.1). Previous studies have shown that an increase in LCP is potentially due to variation in glycosylation resulting in in greater steric hindrance and repulsive electrostatic forces between adjacent GAG chains, resulting in extension of the core protein (Lee et al., 2010). This data suggests that cells of different lineages and differentiation have different metabolic phenotypes resulting in differential pre- programmed biosynthesis/glycosylation machinery that can modulate the mechanical performance of aggrecan to meet the specific needs of the host niche microenvironment.

5.1.2 Hypothesis and aims The hypothesis of this Chapter was that porcine NP tissue, rich in notochordal derived NC cells, is composed of aggrecan with a similar molecular structure to NP tissue (i.e. bovine NP), which is predominantly populated by NP cells. NP tissue that is notochordal in origin synthesises aggrecan which is structurally different compared to AC tissue, which is mesodermal in origin.

117

As a preliminary investigation, the aims of this Chapter were to use AFM, histology, and in situ gelatinase zymography to show that:

(i) NC cells synthesise aggrecan smaller in size compared to AC derived aggrecan

(ii) NC cells synthesise aggrecan similar in structure to NP cell derived aggrecan

5.2 Materials & methods 5.2.1 Experimental design This chapter investigates the ultrastructure of aggrecan isolated from the notochordal cell- rich skeletally immature porcine NP ECM, in comparison to skeletally immature porcine AC and mature bovine NP and AC. AFM was used to characterise the structure of intact and non-intact aggrecan adhered to a charged surface. Histological staining (Safranin-O/Fast Green, Alcian Blue pH 2.5, Masson’s Trichrome and Picrosirius Red) was used to visualise the overall ECM composition (proteoglycan, sGAG, non-fibrillar and fibrillar collagen content, respectively). In situ gelatin zymography was performed to determine if aggrecanases (most of which exhibit gelatinase activity) were active in the skeletally immature porcine NP and AC tissues. This was then used to provide insight into a potential mechanism for the fragmentation of aggrecan.

118

5.3 Results 5.3.1 Porcine nucleus pulposus was smaller than porcine articular cartilage derived intact aggrecan AFM was used to compare three structural parameters of intact AC- and NP-derived aggrecan (LCP, LGAG and MAGAG) (Figure 5.1a). Mean aggrecan LCP was significantly lower for NP- (394 ± 35 nm) compared to AC- (455 ± 30 nm) derived aggrecan (p<0.05: Figure

5.1a). However, the mean lengths (LGAG: NP = 324 ± 39 nm and AC = 393 ± 31 nm) and 2 2 areas (MAGAG: NP = 9808 ± 3301 nm , AC = 11169 ± 2978 nm ) of the glycosylated regions were not significantly shorter and smaller respectively in NP compared with AC intact aggrecan (Fig. 5.1a). There was no significant difference in average GAG chain length was between AC and NP (15 ± 3.7 nm and 16 ± 2.8 nm, respectively). Due to the low number of intact monomers (AC: n=2; NP: n=11) it was difficult to draw any conclusions on aggrecan structural variability between porcine NP and AC.

5.3.2 NP porcine aggrecan monomers were similar to bovine NP derived intact aggrecan Again, due to the low number of intact monomers in porcine NP and AC it was difficult to draw any conclusions on aggrecan structural variability between porcine and bovine tissue. When compared against native bovine NP tissue (Fig. 5.2), porcine NP derived aggrecan

(Fig. 5.2) was not significantly different in LCP (bovine NP: 389 nm; porcine NP: 394 nm), 2 LGAG (bovine NP: 310 nm; porcine NP: 324 nm), MAGAG (bovine NP: 9091 nm ; porcine 2 NP: 9808 nm ) or iGAGL (bovine NP: 16.3 nm; porcine NP: 16.1 nm). In addition, native bovine AC tissue porcine AC derived aggrecan (Fig. 5.2) was not significantly different in

LCP (bovine AC: 457 nm; porcine AC: 455 nm), LGAG (bovine AC: 388 nm; porcine AC: 371 2 2 nm), MAGAG (bovine AC: 12733 nm ; porcine AC: 11169 nm ) or iGAGL (bovine AC: 18.1 nm; porcine AC: 15 nm).

119

Figure 5.1. Surface-adsorbed intact aggrecan monomers were structurally dissimilar in porcine nucleus pulposus and articular cartilage tissues. (a) Four structural parameters were measured for each intact monomer: core protein length (LCP), GAG length (LGAG), molecular area (MAGAG) and average GAG chain length (iGAGL). Combined frequency distributions and mean (+/-SD) of LCP, LGAG, MAGAG and iGAGL for n=2 NP and AC replicates. (b) AFM height image (scale bar 0.2 µm; z-scale 2 nm) of an intact porcine aggrecan monomer comprising three globular (G1—G3) domains. (c) Relative mean molecular dimensions of intact AC- and NP-derived aggrecan monomers.

120

Figure 5.2. Porcine nucleus pulposus and articular cartilage derived intact aggrecan was structurally similar to bovine nucleus pulposus and articular cartilage. Four structural parameters were measured for each intact monomer: core protein length (LCP), GAG length

(LGAG), molecular area (MAGAG) and average GAG chain length (iGAGL). Box shows the mean and whiskers indicate the minimum and maximum values.

5.3.3 Porcine nucleus pulposus aggrecan fragments were smaller than porcine articular cartilage As molecular fragmentation was prevalent in bovine NP and AC tissue, we wanted to determine if fragmentation was also present in porcine tissues. In the total molecular population (AC n=737; NP n=685; Fig 5.4), fragmented aggrecan accounted for 99.67% of the observed molecules in AC and 98.52% in NP. Fewer molecules were intact in AC compared to NP; however, NP derived aggrecan fragments were smaller compared with AC and consequently the mean molecular area was significantly lower for NP compared with AC (Molecular area: 4778 ± 4008 nm2 and 3382 ± 1949 nm2 respectively: p<0.05) (Fig. 5.4c).

121

In contention with assessment of iGAGL in intact monomers (Fig. 5.1a), analysis of all observable aggrecan (both intact and non-intact; n=337 aggrecan monomers), using skeletal analysis (Fig 5.3) showed that through calculation of the linear gradient; a greater amount of GAG was present per unit length of core protein in porcine AC aggrecan (0.0301x) compared to porcine NP aggrecan (0.0425x).

Figure 5.3. Porcine nucleus pulposus derived aggrecan had greater glycosaminoglycan content per unit length of core protein. (a and b) Scatter graphs comparing LCP to molecular area of all observable aggrecan molecules (i.e. both intact and non-intact) regardless of fragment size or orientation (n=2 biological replicates). (a) NP and (b) AC.

5.3.4 Porcine nucleus pulposus aggrecan fragments were similar in size compared to bovine nucleus pulposus Importantly, when compared against native bovine NP tissue, fragment size from porcine NP derived aggrecan (Fig. 5.5) was not significantly different (bovine NP: 3560 nm2; porcine NP: 3382 nm2; p<0.0001). However, fragment size from porcine AC derived aggrecan (Fig. 5.3) was significantly smaller in comparison to bovine AC derived aggrecan (bovine AC: 8667 nm2; porcine AC: 4778 nm2; p<0.0001).

122

Figure 5.4. Both porcine articular cartilage and nucleus pulposus were composed primarily of fragmented aggrecan. (a and b) AFM height images of AC (a) and NP (b) isolated aggrecan molecules (2 x 2 µm; scale bar is 0.2 µm; z-scale 2 nm). Arrows indicate intact monomers. Only a small proportion of aggrecan molecules were in the intact form, most molecules lacked either the G3 domain alone or all globular domains. (c) Molecular area size distribution of all observable aggrecan molecules (n=2 biological replicates). Values shown within the histogram are the mean (± SD).

123

Figure 5.5. Aggrecan fragmentation was similar between porcine and bovine nucleus pulposus. Box and whisker plot of molecular area size of all observable aggrecan molecules (i.e. both intact and non-intact) regardless of fragment size or orientation (n=2 biological replicates). Box shows the mean and whiskers indicate the minimum and maximum values. ****p<0.0001.

5.3.5 Aggrecan fragmentation was not associated with aberrant structural remodelling or excessive protease activity As in Chapter 3, we wanted to ensure that isolated porcine tissue was histologically healthy and non-degenerate, as fragmentation is usually considered to be indicative of ageing and/or disease (Sztrolovics et al., 1997, El Bakali et al., 2014). Therefore, we next characterised the histological composition and collagen ultrastructure skeletally immature mature AC and NP (Fig. 5.6). GAG-specific stains such as Alcian blue and Safranin O indicated porcine NP and AC tissue was enriched in proteoglycan and sGAG. In contrast, there was little non-fibrillar and fibrillar collagen in NP, the latter being semi-quantitatively characterised histologically by measuring PSR-enhanced collagen birefringence (McConnell et al., 2016). In accordance with a previous study of healthy porcine AC tissue (McLeod et al., 2013), AC exhibited fibrillar collagen (9.73% ± 5.38%); however, NP did not (0.0% ± 0.0%; p<0.01). Conversely, NP was qualitatively enriched in GAGs compared with AC (Fig. 5.6). The distribution of these ECM components was consistent with healthy tissues (Fig. 5.6) (Lindburg et al., 2013).

124

Alcian Blue Safranin Masson’s Picrosirius Picrosirius Red pH 2.5 O/Fast Green Trichrome Red (Polarised) (Brightfield)

Figure 5.6. Skeletally mature porcine tissues appeared histologically healthy. Images are representative of two biological replicates for mature and immature AC and NP tissue. Bright-field images show Alcian Blue pH 2.5 staining revealing sGAG, Safranin O/Fast Green staining revealing proteogylcan, Masson’s Trichrome revealing non-fibrillar collagen, and picrosirius red staining revealing fibrillar collagen under polarized light. Scale bars are 100 µm.

In situ gelatin zymography was used to determine if fragmentation of aggrecan was associated with the upregulation of protease activity (El Bakali et al., 2014). Embedding porcine NP for cryosectioning was not possible as the sections fell apart due to the highly gelatinous nature of the NP. Therefore, in situ gelatin zymography was performed on porcine AC tissues and a healthy rat skin control (Fig. 5.7). Protease activity was predominantly localised at the surface of AC (7.79 ± 6.19 pixels) and was significantly lower compared to healthy rat epidermis (128.9 ± 49.73 pixels; p<0.0001) and endodermis (175.0 ± 15.13 pixels; p<0.0001). It appears therefore that these porcine AC exhibited no signs of pathological remodelling with regards to collagen fibril and GAG composition or ECM protease activity.

125

Figure 5.7. Low levels of matrix metalloproteinase activity in healthy porcine articular cartilage. (a) Haematoxylin and Eosin staining of healthy mature and immature (iAC, iNP) bovine AC and NP tissue and healthy rat skin. (b) Respective in situ gelatinase zymography. White lines and boxes indicate regions measured for mean fluorescent activity. (c) Mean fluorescent gelatinase activity between AC (n=2) and rat skin (n=2). All scale bars are 100 µm. ****p<0.0001.

126

5.4 Discussion This preliminary study demonstrated three key points: (i) that all observable porcine NP aggrecan was structurally variable (in terms of molecular area/fragment size) compared to porcine AC (as seen in bovine tissue), (ii) that in a highly gelatinous, young notochordal cell rich porcine tissue, the molecular structure of aggrecan is highly heterogeneous, similar to the less gelatinous, bovine NP tissue, and (iii) porcine AC aggrecan was more fragmented compared to bovine AC.

Very few intact monomers could be identified in porcine AC (n=2) and NP (n=11), making conclusions difficult. This low number of intact monomers in porcine AC was likely due to poor total yield of isolated monomers. Porcine AC and NP monomers appeared structurally similar to bovine (Lee et al., 2010). Intact porcine aggrecan monomers appeared to follow the trend between bovine AC and NP and as previously reported in young and adult human

AC (Lee et al., 2013); porcine NP tissue aggrecan had a smaller LCP, LGAG, and MAGAG compared to AC. Skeletal analysis of all observable aggrecan suggested that porcine NP derived aggrecan monomers had a greater amount of GAG per unit of LCP, which was not evident from analysis of iGAGL of intact monomers. As previously stated, differential glycosylation could alter LCP as reduced GAG load may cause NP-derived aggrecan to adopt a lower energy state and hence a relaxed random coil as opposed to an extended, rigid structure (Ng et al., 2003, Sattelle et al., 2015). Variability in aggrecan structure may also be a consequence of differential sulfation and/or abundance of GAG chains (Chandran and Horkay, 2012). Porcine NP derived aggrecan is structurally similar to that derived from bovine NP tissue, both of which originate from the notochord and structurally dissimilar to AC derived aggrecan which originates from the mesoderm. This data supports the theory (Lee et al., 2010) that biosynthesis/glycosylation of aggrecan varies between cells of different lineages (i.e. notochord vs mesoderm) which may modulate the mechanical performance of aggrecan, tissue osmolarity, porosity (and therefore metabolite transfer) to meet the specific needs of the host niche microenvironment. Interestingly, assessment of all observable fragmentation indicated that the overall size of aggrecan fragments in porcine AC was smaller than in bovine AC. This may be a result of variable glycosylation between species.

As in bovine (see Chapter 3.3.1) and TE IVD (see Chapter 4.3.1), the influence of the structural difference in intact aggrecan on tissue function is likely to be marginal. This study

127 shows that intact monomers only account for a small population (0.33–1.47%) of the total molecular population. The high proportion of non-intact aggrecan accounts in healthy young porcine tissue suggests aggrecan fragmentation is unlikely to be maladaptive and may play a role in mediating solute diffusion within the ECM (see Chapter 3). Interestingly, in contrast to bovine NP tissue (see Chapter 3.3.3), no fibrillar collagen was identified in porcine NP tissue. The absence of a fibrillar collagen network may contribute to the highly fluidic nature of porcine NP. Further investigation is required to understand what effect this has (if any) on maintaining NC cell viability, metabolism and phenotype.

As in Chapter 3, gelatin zymography was used to characterise relative aggrecanase activity in porcine tissues (Freemont et al., 1999, Gepstein et al., 2002, Schulze Willbrenning et al., 2010, Weiler, 2013). In line with bovine and TE IVD construct work, we detected only minimal gelatinase activity in young, healthy porcine tissue. Similar to previously reported data, our work suggests that gelatinase activity, and therefore aggrecan fragmentation may continue at low levels throughout the life of an animal (Schulze Willbrenning et al., 2010). Although aggrecanases have metalloproteinase characteristics, aggrecanase activity is independent to MMP activity (Hughes et al., 1998), and aggrecanases, not MMPs, are speculated to be primarily responsible for aggrecan proteolysis (Little et al., 1999); however, as previously mentioned, identifying which proteases drive fragmentation is complicated by the structural heterogeneity of the fragmented aggrecan and the low substrate specificity of many ECM proteases (Chakraborti et al., 2003).

Protease driven fragmentation and concentration of aggrecan is likely to influence tissue function/stiffness by mediating molecular packing density. Porcine NP is a highly gelatinous tissue which may be attributed to aggrecan abundance; far more aggrecan appeared to be isolated from NP tissue compared to AC tissue when visualized under AFM; however, this was not assessed quantitatively due to time constraints. Loaded with large, vacuolated NCs, histologically porcine NP is a proteoglycan enriched tissue compared to bovine NP (Chapter 3.3.3). Aggrecan concentration has been definitively been shown to be positively correlated with osmotic potential in AC (Chandran and Horkay, 2012), whilst the density of aggrecan- bound GAGs can influence mobility of water, solutes, proteins and matrix synthesis by modulating osmolarity and porosity (Gribbon and Hardingham, 1998, Li et al., 2016). At high aggrecan concentrations, molecular diffusion through the ECM is significantly reduced which may in turn influence the availability of oxygen and metabolites (Gribbon and

128

Hardingham, 1998). Enzymatic processing of aggrecan may be an important adaptive mechanism to mediate the structure and function of cartilaginous tissues; aggrecan fragmentation may act to modulate molecular diffusion. In addition, low protease activity suggests that porcine NP may be a highly preserved, relatively unchanging microenvironment; a niche environment enabling NCs to remain viable and retain their phenotype.

5.5 Conclusions Cells originating from the notochord (i.e. NC and NP) synthesised aggrecan that was structurally similar to each other, and chondrocytes originating from the mesoderm follow a similar trend (i.e. porcine AC versus bovine AC). NC and NP derived aggrecan is smaller in size compared to aggrecan derived from AC. This suggests that cells of different lineages variably glycosylate aggrecan modulating the mechanical performance of aggrecan to meet the specific needs of the host niche microenvironment. Further investigation is required to understand if the highly gelatinous porcine NP is a consequence of variability in the nanostructure of key ECM molecules, or solely the proportion of aggrecan to collagen.

129

6 Chapter 6: General Discussion and Future Work

130

6.1 Study rationale LBP is a debilitating, age-related musculoskeletal disorder that affects millions of people worldwide (Maniadakis and Gray, 2000, Walker, 2000, NICE, 2009). Approximately 40% of LBP cases are caused by IVDD, which are initiated within the central gelatinous NP and are characterised by the pathological proteolytic degradation and subsequent loss of the key ECM proteins aggrecan and collagen type II. These two major ECM proteins are vital to the structure and function of the NP. Collagen forms an unorganised fibrillar network providing tensile strength and acting as a net, ensuring aggrecan does not escape. Aggrecan provides resistance against compressive forces by generating an osmotic swelling pressure, which is attributed to substitution of the core protein with highly sulphated GAG chains (Yu et al., 2007, Smith et al., 2011). Current therapies only aim to provide short-term symptomatic pain relief and do not treat the underlying pathology. In order to replace/repair the degenerate NP with a tissue engineered construct, it is imperative to understand the composition and structure of the major extracellular matrix components within the NP (i.e. aggrecan).

Aggrecan structure has been previously assessed predominantly in the context of AC and was found to be species, age, and cell-dependent (Buckwalter et al., 1985, Lee et al., 2010, Lee et al., 2013). Whilst this provides valuable insight into aggrecan structure, NP cannot be replaced with AC-like tissue; therefore, it is essential to understand the aggrecan structure in NP tissue. There is also evidence that exogenous addition of growth factors can play a role in modulating aggrecan biosynthesis/glycosylation (Kopesky et al., 2010, Lee et al., 2010). GDF6 is a candidate growth factor of choice for tissue engineering NP tissue, as it has been shown to have advantages over TGFβ, which has traditionally been utilised in cell- based cartilage therapies (Lee et al., 2010, Clarke et al., 2014). Any viable therapy must function within the IVD niche, hence understanding the effect of hypoxia on ECM synthesis is essential. Therefore, the studies in this thesis have focused on a number of experiments to characterise the nanostructure of aggrecan in native bovine and porcine NP and AC tissue and in in vitro tissue engineered NP following manipulation of cell culture conditions (i.e. growth factor and hypoxia).

131

6.2 Key outcomes:

1. In Chapter 3, extensive aggrecan fragmentation was observed in both skeletally immature and mature NP and AC tissue, challenging our understanding of the pathology and tissue engineering of cartilaginous tissues. Computational modelling suggested that the degree of aggrecan fragmentation may affect porosity and nutrient transfer within an ECM. Therefore, when considering the development of viable TE IVD constructs for the repair of cartilaginous tissues, it may be necessary to replicate the structurally heterogeneous aggrecan population which characterises native AC and NP.

2. Chapter 4 demonstrated that NP cells synthesise aggrecan with a similar molecular structure and promote synthesis of a histologically and mechanically similar ECM to NP tissue, following treatment with a growth factor (i.e. TGFβ3 and GDF6). Although oxygen tension did promote the synthesis of an ECM by NP cells which mimicked ECM deposition, but not the mechanical function of NP-like ECM; it was insufficient to produce a mechanically viable TE IVD construct. As in Chapter 3, no direct link was identified between aggrecan structure, and the mechanical function of tissues. Exogenous addition of growth factors promoted the synthesis of an NP- like ECM with a similar mechanical function to NP tissue. When considering the development of viable TE IVD constructs for the repair of NP tissue, growth factors, rather than oxygen tension, are necessary to stimulate sufficient and appropriate deposition of ECM which characterises native NP.

3. Chapter 5 demonstrated that cells of different lineages (i.e. notochord and mesoderm) variably glycosylate aggrecan which may modulate the mechanical performance of aggrecan and solute transfer to meet the specific needs of the host niche microenvironment. NC cells synthesised aggrecan similar in size to NP cells, but smaller in size compared to AC. When considering the development of viable TE IVD constructs for the repair of NP tissue, an appropriate cell type should be used; many studies are considering the use of BM-MSCs for TE of IVDs which are mesodermal in origin (Kopesky et al., 2010, Lee et al., 2010). This could result in the synthesis of key ECM components (i.e. aggrecan) that are inappropriately

132

glycosylated and may result in poor mechanical performance or affect porosity and solute transfer.

6.3 Conclusions Taken together, aggrecan was found to be structurally heterogeneous in native bovine tissue; however, it must be acknowledged that increased accumulation of fragmented aggrecan in NP tissue may be the result of a different physical barrier to fragment loss compared to articular cartilage (surrounded by the synovial fluid). The NP is surrounded by a highly ordered and compact AF, which functions to constrain movement of the NP and its constituent components, whereas the AC is only surrounded by the synovial fluid (i.e. there is no physical barrier to the loss of fragments). The findings of this study are contrary to the textbook depiction of aggrecan, which challenges our current understanding of how aggrecan nanostructure contributes to ECM porosity, solute transfer, osmolarity and the mechanical function of tissues it comprises. Variation in aggrecan structure was identified between tissues (i.e. NP and AC) and species (i.e. porcine AC and bovine AC), contributing to previously reported data on age, species and cell dependent variation (Buckwalter et al., 1989, Lee et al., 2010, Lee et al., 2013). This provided the rationale that aggrecan structure can be modified and in order to recapitulate native NP tissue it is necessary to determine what factors affect aggrecan structure (e.g. cell culture conditions).

Following manipulation of cell culture conditions (i.e. oxygen tension and growth factor), this study showed that growth factors (i.e. GDF6 and/or TGFβ3) were required to promote tissue akin to native NP. However, due to lack of biochemical data, the importance of aggrecan structure versus the abundance of aggrecan on the mechanical function of TE IVD constructs could not be resolved. Whilst having the appropriate aggrecan structure is likely important in water accumulation, it is more likely that the overall abundance (i.e. having enough) of aggrecan is the main contributor to generating an osmotic potential and maximising tissue hydration. Another limitation is the interpretation of the mechanical data; the mechanical function of cartilaginous (i.e. AC and NP) tissue is primarily dictated between the interplay of aggrecan and type II collagen. By calculating the reduced modulus from the initial curve between the maximum and minimum force required to indent each tissue, the combined contribution of both aggrecan and collagen was measured. Further dissection of force curve data would have enabled identification of the individual

133 contributions of aggrecan and collagen, which could have been correlated with biochemical quantification of both proteins.

Whilst NP cells were used in this study, they are unlikely to be considered for any regenerative therapy, due to difficulty obtaining healthy NP cells without causing harm to the donor. Therefore, it is more likely that stem cells will be used to synthesise TE IVD tissue; however, MSCs may be inappropriate for recapitulating native NP tissue as they originate from the mesoderm: taken together, Chapter 3 and Chapter 5 demonstrated that aggrecan structure varies between notochord (NP) and mesoderm tissues (AC), and notochord originating cells are required to synthesise aggrecan with a structure similar to native NP tissue. Embryonic or induced puripotent stem cells may be a suitable alternative as both have the capacity to be stimulated to differentiate into notochord progenitor cells (Chen et al., 2013, Liu et al., 2014, Liu et al., 2015).

To conclude, this study shows that intact monomers only account for a small population (0.33—4.95%) of the total molecular population in native tissue. The high proportion of non- intact aggrecan in healthy cartilaginous tissues suggests aggrecan fragmentation is unlikely to be maladaptive and may play a role in mediating solute diffusion within the ECM (Chapter 3). The observation of extensive aggrecan fragmentation across cartilaginous tissues (i.e. tissue engineered, immature and mature) and between species (i.e. human, bovine and porcine) continues to challenge our understanding of the pathology and tissue engineering of cartilaginous tissues. Aggrecanase activity likely plays a functional role, as fragmentation may act to modulate molecular diffusion (porosity) and osmolarity and consequently the ability to resist compressive forces. In order to successfully create a TE IVD construct, it is necessary to choose appropriate biological stimuli to recapitulate native NP tissue.

6.3 Future work 6.3.1 Completion of immediate studies These studies have raised a number of questions that can be addressed through the following future studies.

6.3.1.1 Chapter 3 Variable sulfation or occupation of GAG sites along aggrecan core protein is the most likely contributor to the size difference observed between intact AC and NP tissue derived

134 aggrecan (Lee et al., 2010). Therefore, a comparison of the sulfation patterning of sGAG in relation to LCP, LGAG, MAGAG and iGAGL between skeletally immature and mature bovine AC and NP would be useful. This could be achieved using fluorophore-assisted carbohydrate electrophoresis (Plaas et al., 2001a). Biochemical analysis of aggrecan and type II collagen would also provide additional insight into the importance of protein structure versus abundance in relation to the mechanical function of cartilaginous (i.e. AC and NP) tissue and TE IVD constructs. Further investigations into the glycosylation machinery that regulate glycolysis of aggrecan would enhance understanding of the mechanism behind aggrecan structural variability and function. It would also be beneficial to determine which individual or combination of aggrecanases is responsible for fragmentation of aggrecan in skeletally immature and mature AC and NP tissue. This would provide insight into the difference between normal processing and pathological degradation of aggrecan. To achieve this, it would require assessing both the gene expression and protein activity of the full panel of aggrecanases known to be capable of cleaving aggrecan, as previously detailed in Chapter 3.

There are also several ways in which the packing model in Chapter 3 could be improved. Chapter 3—5 demonstrated that structural heterogeneity of aggrecan exists between NP and AC and aggrecan structure can be manipulated by cell culture conditions; however, none of these studies investigated a direct mechanism that links aggrecan structural variation to mechanical responsiveness and tissue microenvironment. This study has provided the basic observations required to generate a model to understand the implications of aggrecan structural variability; however, in order to derive a mechanism, further observations are required, such as assessing the (i) GAG content, (ii) sulfation patterning of GAGs bound aggrecan, and (iii) tissue ionic strength which affect the electrostatic interactions between adjacent and opposing GAGs (Eisenberg and Grodzinsky, 1985, Buschmann and Grodzinsky, 1995, Dean et al., 2006, Nia et al., 2015), and consequently (iii) porosity/hydraulic permeability (as GAG spacing is reduced at high ionic strengths [e.g 0.1—1.0 M NaCl] which results in decreased hydraulic permeability and reduced tissue stiffness) (iv) osmolarty, and (iv) water content (Nia et al., 2015).

Modelling variation in the compressive modulus of tissues in response to aggrecan structure has previously been met with limited success; previously Dean and colleagues (Han et al., 2005), modelled the compressive modulus of opposing monolayers of chemically end

135 grafted aggrecan monomers; however these monolayers do not include all of the fragments that are lacking their N-terminal G1 domains. Other models investigating the opposing GAG chains provide insight into the contribution of GAGs and therefore proteogylcans to the compressive modulus of tissues, but are limited by their misconception of aggrecan monomers as predominantly intact bottlebrush structures (Nia et al., 2015). It is important to consider how external force is resisted by aggrecan. Upon compression, water (an incompressible material) — accumulated by the osmotic potential generated by aggrecan — resists compression. This osmotic potential depends upon the charge density of aggrecan which is dictated by its structure (sGAG substitution). Therefore, it is important to characterise the relationship between the nanostructure of aggrecan and net molecular charge density. Biomimetic aggrecan may be useful for modelling the function of native aggrecan as it can be synthesised to mimic fragment sizes and variable GAG chain length; when combined with osmolarity and mechanical assessment, it may be possible to derive the mechanical responsiveness of synthetic cartilaginous tissues in response to fragment size (Sharma et al., 2013, Prudnikova et al., 2017). The packing model reported in Chapter 3 of this study is also limited, but finite element modelling has the potential to accurately model the contribution of aggrecan structure on tissue mechanics. Currently this basic model only incorporates the observed structural dimensions of aggrecan (see Chapter 3) in relation to porosity; however, additional data such as the overall net negative charge of each aggrecan monomer and the overall ECM osmolarity generated by different populations of aggrecan structural variants correlated with compressive modulus would be conceptually possible (Glover et al., 1991, Sivan et al., 2013). In addition, an experimental model tissue could be created using varying sizes of biomimetic aggrecan and/or collagen, and potentially incorporate 3D printing to create the architecture.

6.3.1.2 Chapter 4 As in Chapter 3, characterisation of the sulfation patterning and quantification of occupied GAG substitution sites on each aggrecan monomer would provide further insight into variability in the biosynthesis of aggrecan as a consequence of the manipulation of cell culture conditions. TE IVD constructs were initially assessed using SEC-MALS; however, this was not successful due to either a technical or sample error. Therefore, it would be useful to characterise the behaviour of TE IVD derived aggrecan in solution, similar to the conditions in vivo. Further investigation of aggrecanase expression and activity at additional time points (e.g. 1, 3, 4, 7, 14, and 21 days of culture) would identify the individual or group

136 of aggrecanases responsible for programmed fragmentation of newly synthesised aggrecan and provide insight into the functional relevance of aggrecan fragmentation. Furthermore, aggrecanase and anabolic gene expression of GDF6 and TGFβ3 TE IVD constructs was only assessed in two biological replicates and requires at least on additional replicate.

Mechanical characterisation of hypoxia TE IVD requires a third biological replicate. In addition, Chapter 4 would have benefited from the characterisation of a TE IVD cultured under hypoxia with GDF6 or TGFβ3, as data suggests that a combination of growth factor and hypoxia has the potential to stimulate ECM more akin to native NP tissue than either isolated condition (Chen et al., 2014).

6.3.1.3 Chapter 5 Chapter 5 is a preliminary study and requires an additional biological replicate for both porcine NP and AC. Data in this Chapter showed that aggrecan structure was invariable between porcine and bovine NP suggesting the highly gelatinous nature of porcine NP is due to variation in the overall negative charge of individual aggrecan monomers or the quantity of aggrecan and collagen. This could be further investigated through characterisation of the sulfation patterning and quantification of occupied GAG substitution sites on each aggrecan monomer. Unfortunately, there are no reliable direct methods for the quantification of aggrecan. Due to methodological limitations and the highly compliant nature of porcine NP it was not possible to assess its compressive modulus using AFM; however, mechanical characterisation correlated with aggrecan and collagen content would provide insight into the effect of aggrecan structure on the mechanical function of cartilaginous tissues that are either notochord or mesoderm in origin. This would provide further insight into the most suitable cell type for TE IVD constructs. In addition, it would be useful to characterise the behaviour of porcine NP and AC derived aggrecan in solution (with SEC-MALS), similar to the conditions in vivo, as this may provide insights into the highly gelatinous nature of porcine NP tissue.

6.3.2 Beyond the remit of this study: creating a viable tissue engineered intervertebral disc replacement In conclusion, the ultimate goal of creating a viable TE IVD replacement that mimics native NP tissue requires the consideration of multiple factors. This study only considers the contributions of aggrecan and collagen II structure, which is a gross simplification of the

137 many ECM proteins which have known roles in cartilaginous tissue assembly, stabilisation and function. To truly understand the cumulative effect of all these ECM proteins on tissue function, we must understand how they interact. For example, the HA mediates molecular packing density of aggrecan aggregates, which interact with collagenous networks present in the tissue (Hardingham and Fosang, 1992). In addition, we need to understand the interaction of resident cells with the surrounding ECM. IVD cells have recently been shown to stabilise ECM through pericellular colocalisation of perlecan and collagen IV (Hayes et al., 2016).

138

7 Chapter 7: References

139

ABBASZADE, I., LIU, R. Q., YANG, F., ROSENFELD, S. A., ROSS, O. H., LINK, J. R., ELLIS, D. M., TORTORELLA, M. D., PRATTA, M. A., HOLLIS, J. M., WYNN, R., DUKE, J. L., GEORGE, H. J., HILLMAN, M. C., MURPHY, K., WISWALL, B. H., COPELAND, R. A., DECICCO, C. P., BRUCKNER, R., NAGASE, H., ITOH, Y., NEWTON, R. C., MAGOLDA, R. L., TRZASKOS, J. M. & BURN, T. C. 1999. Cloning and characterization of ADAMTS11, an aggrecanase from the ADAMTS family. J Biol Chem, 274, 23443-50. AGUIAR, D. J., JOHNSON, S. L. & OEGEMA, T. R. 1999. Notochordal cells interact with nucleus pulposus cells: regulation of proteoglycan synthesis. Exp Cell Res, 246, 129- 37. AKHTAR, R., CRUICKSHANK, J. K., ZHAO, X., WALTON, L. A., GARDINER, N. J., BARRETT, S. D., GRAHAM, H. K., DERBY, B. & SHERRATT, M. J. 2014. Localized micro- and nano-scale remodelling in the diabetic aorta. Acta Biomater, 10, 4843-4851. AKHTAR, R., SHERRATT, M. J., CRUICKSHANK, J. K. & DERBY, B. 2011. Characterizing the elastic properties of tissues. Mater Today (Kidlington), 14, 96- 105. ALADIN, D. M., CHEUNG, K. M., NGAN, A. H., CHAN, D., LEUNG, V. Y., LIM, C. T., LUK, K. D. & LU, W. W. 2010. Nanostructure of collagen fibrils in human nucleus pulposus and its correlation with macroscale tissue mechanics. J Orthop Res, 28, 497-502. ARIGA, K., YONENOBU, K., NAKASE, T., KANEKO, M., OKUDA, S., UCHIYAMA, Y. & YOSHIKAWA, H. 2001. Localization of cathepsins D, K, and L in degenerated human intervertebral discs. Spine (Phila Pa 1976), 26, 2666-2672. ARKESTEIJN, I. T. M., POTIER, E. & ITO, K. 2017. The Regenerative Potential of Notochordal Cells in a Nucleus Pulposus Explant. Global Spine J, 7, 14-20. ASPBERG, A. 2012. The different roles of aggrecan interaction domains. J Histochem Cytochem, 60, 987-96. BADER, H. L., WANG, L. W., HO, J. C., TRAN, T., HOLDEN, P., FITZGERALD, J., ATIT, R. P., REINHARDT, D. P. & APTE, S. S. 2012. A disintegrin-like and metalloprotease domain containing thrombospondin type 1 motif-like 5 (ADAMTSL5) is a novel fibrillin-1-, fibrillin-2-, and heparin-binding member of the ADAMTS superfamily containing a netrin-like module. Matrix Biol, 31, 398-411. BARRY, F. P., ROSENBERG, L. C., GAW, J. U., GAW, J. U., KOOB, T. J. & NEAME, P. J. 1995. N- and O-linked keratan sulfate on the hyaluronan binding region of aggrecan from mature and immature bovine cartilage. J Biol Chem, 270, 20516-24. BAYLISS, M. T., OSBORNE, D., WOODHOUSE, S. & DAVIDSON, C. 1999. Sulfation of chondroitin sulfate in human articular cartilage. The effect of age, topographical position, and zone of cartilage on tissue composition. J Biol Chem, 274, 15892- 15900. BERTAGNOLI, R., YUE, J. J., NANIEVA, R., FENK-MAYER, A., HUSTED, D. S., SHAH, R. V. & EMERSON, J. W. 2006. Lumbar total disc arthroplasty in patients older than 60 years of age: a prospective study of the ProDisc prosthesis with 2-year minimum follow-up period. J Neurosurg Spine, 4, 85-90. BERTOLO, A., MEHR, M., AEBLI, N., BAUR, M., FERGUSON, S. J. & STOYANOV, J. V. 2012. Influence of different commercial scaffolds on the in vitro differentiation of human mesenchymal stem cells to nucleus pulposus-like cells. Eur Spine J, 21 Suppl 6, S826-38. BLUNDELL, C. D., ALMOND, A., MAHONEY, D. J., DEANGELIS, P. L., CAMPBELL, I. D. & DAY, A. J. 2005. Towards a structure for a TSG-6.hyaluronan complex by

140

modeling and NMR spectroscopy: insights into other members of the link module superfamily. J Biol Chem, 280, 18189-201. BROWN, G. M., HUCKERBY, T. N., BAYLISS, M. T. & NIEDUSZYNSKI, I. A. 1998. Human aggrecan keratan sulfate undergoes structural changes during adolescent development. J Biol Chem, 273, 26408-14. BRUEHLMANN, S. B., RATTNER, J. B., MATYAS, J. R. & DUNCAN, N. A. 2002. Regional variations in the cellular matrix of the annulus fibrosus of the intervertebral disc. J Anat, 201, 159-71. BUCKWALTER, J. A., PEDRINI-MILLE, A., PEDRINI, V. & TUDISCO, C. 1985. Proteoglycans of human infant intervertebral disc. Electron microscopic and biochemical studies. J Bone Joint Surg Am, 67, 284-294. BUCKWALTER, J. A., ROUGHLEY, P. J. & ROSENBERG, L. C. 1994. Age-related changes in cartilage proteoglycans: quantitative electron microscopic studies. Microsc Res Tech, 28, 398-408. BUCKWALTER, J. A., SMITH, K. C., KAZARIEN, L. E., ROSENBERG, L. C. & UNGAR, R. 1989. Articular cartilage and intervertebral disc proteoglycans differ in structure: an electron microscopic study. J Orthop Res, 7, 146-151. BUSCHMANN, M. D. & GRODZINSKY, A. J. 1995. A molecular model of proteoglycan- associated electrostatic forces in cartilage mechanics. J Biomech Eng, 117, 179-92. CAWSTON, T. E. & WILSON, A. J. 2006. Understanding the role of tissue degrading enzymes and their inhibitors in development and disease. Best Pract Res Clin Rheumatol, 20, 983-1002. CHAKRABORTI, S., MANDAL, M., DAS, S., MANDAL, A. & CHAKRABORTI, T. 2003. Regulation of matrix metalloproteinases: an overview. Mol Cell Biochem, 253, 269-85. CHANDRAN, P. L., DIMITRIADIS, E. K., BASSER, P. J. & HORKAY, F. 2010. Probing Interactions between Aggrecan and Mica Surface by the Atomic Force Microscopy. J Polym Sci B Polym Phys, 48, 2575-2581. CHANDRAN, P. L. & HORKAY, F. 2012. Aggrecan, an unusual polyelectrolyte: review of solution behavior and physiological implications. Acta Biomater, 8, 3-12. CHEN, J., LEE, E. J., JING, L., CHRISTOFOROU, N., LEONG, K. W. & SETTON, L. A. 2013. Differentiation of mouse induced pluripotent stem cells (iPSCs) into nucleus pulposus-like cells in vitro. PLoS One, 8, e75548. CHEN, J. W., LI, B., YANG, Y. H., JIANG, S. D. & JIANG, L. S. 2014. Significance of hypoxia in the physiological function of intervertebral disc cells. Crit Rev Eukaryot Gene Expr, 24, 193-204. CHEN, X. D., FISHER, L. W., ROBEY, P. G. & YOUNG, M. F. 2004. The small leucine- rich proteoglycan biglycan modulates BMP-4-induced osteoblast differentiation. FASEB J, 18, 948-58. CHEUNG, K. M., KARPPINEN, J., CHAN, D., HO, D. W., SONG, Y. Q., SHAM, P., CHEAH, K. S., LEONG, J. C. & LUK, K. D. 2009. Prevalence and pattern of lumbar magnetic resonance imaging changes in a population study of one thousand forty- three individuals. Spine (Phila Pa 1976), 34, 934-40. CHOI, K. S., COHN, M. J. & HARFE, B. D. 2008. Identification of nucleus pulposus precursor cells and notochordal remnants in the mouse: implications for disk degeneration and chordoma formation. Dev Dyn, 237, 3953-8. CHUNG, C. & BURDICK, J. A. 2008. Engineering cartilage tissue. Adv Drug Deliv Rev, 60, 243-62. CLARKE, L. E., MCCONNELL, J. C., SHERRATT, M. J., DERBY, B., RICHARDSON, S. M. & HOYLAND, J. A. 2014. Growth differentiation factor 6 and transforming

141

growth factor-beta differentially mediate mesenchymal stem cell differentiation, composition and micromechanical properties of nucleus pulposus constructs. Arthritis Res Ther, 16, R67. COHEN, S. P., ARGOFF, C. E. & CARRAGEE, E. J. 2008. Management of low back pain. BMJ, 337, a2718. COLOMBIER, P., CLOUET, J., BOYER, C., RUEL, M., BONIN, G., LESOEUR, J., MOREAU, A., FELLAH, B. H., WEISS, P., LESCAUDRON, L., CAMUS, A. & GUICHEUX, J. 2016. TGF-beta1 and GDF5 Act Synergistically to Drive the Differentiation of Human Adipose Stromal Cells toward Nucleus Pulposus-like Cells. Stem Cells, 34, 653-67. COLOMBIER, P., CLOUET, J., HAMEL, O., LESCAUDRON, L. & GUICHEUX, J. 2014. The lumbar intervertebral disc: from embryonic development to degeneration. Joint Bone Spine, 81, 125-129. CORNEJO, M. C., CHO, S. K., GIANNARELLI, C., IATRIDIS, J. C. & PURMESSUR, D. 2015. Soluble factors from the notochordal-rich intervertebral disc inhibit endothelial cell invasion and vessel formation in the presence and absence of pro-inflammatory cytokines. Osteoarthritis Cartilage, 23, 487-96. DAY, A. J. & PRESTWICH, G. D. 2002. Hyaluronan-binding proteins: tying up the giant. J Biol Chem, 277, 4585-8. DE VRIES, S. A., POTIER, E., VAN DOESELAAR, M., MEIJ, B. P., TRYFONIDOU, M. A. & ITO, K. 2015. Conditioned medium derived from notochordal cell-rich nucleus pulposus tissue stimulates matrix production by canine nucleus pulposus cells and bone marrow-derived stromal cells. Tissue Eng Part A, 21, 1077-84. DEAN, D., HAN, L., GRODZINSKY, A. J. & ORTIZ, C. 2006. Compressive nanomechanics of opposing aggrecan macromolecules. J Biomech, 39, 2555-65. DESNOYERS, L., ARNOTT, D. & PENNICA, D. 2001. WISP-1 binds to decorin and biglycan. J Biol Chem, 276, 47599-607. DOEGE, K. J., COULTER, S. N., MEEK, L. M., MASLEN, K. & WOOD, J. G. 1997. A human-specific polymorphism in the coding region of the aggrecan gene. Variable number of tandem repeats produce a range of core protein sizes in the general population. J Biol Chem, 272, 13974-9. DOEGE, K. J., SASAKI, M., KIMURA, T. & YAMADA, Y. 1991. Complete coding sequence and deduced primary structure of the human cartilage large aggregating proteoglycan, aggrecan. Human-specific repeats, and additional alternatively spliced forms. J Biol Chem, 266, 894-902. DOMOWICZ, M., LI, H., HENNIG, A., HENRY, J., VERTEL, B. M. & SCHWARTZ, N. B. 1995. The biochemically and immunologically distinct CSPG of notochord is a product of the aggrecan gene. Dev Biol, 171, 655-64. DUDHIA, J., DAVIDSON, C. M., WELLS, T. M., VYNIOS, D. H., HARDINGHAM, T. E. & BAYLISS, M. T. 1996. Age-related changes in the content of the C-terminal region of aggrecan in human articular cartilage. Biochem J, 313 ( Pt 3), 933-40. DURIGOVA, M., NAGASE, H., MORT, J. S. & ROUGHLEY, P. J. 2011. MMPs are less efficient than ADAMTS5 in cleaving aggrecan core protein. Matrix Biol, 30, 145- 53. EISENBERG, S. R. & GRODZINSKY, A. J. 1985. Swelling of articular cartilage and other connective tissues: electromechanochemical forces. J Orthop Res, 3, 148-159. EL BAKALI, J., GRAS-MASSE, H., MAINGOT, L., DEPREZ, B., DUMONT, J., LEROUX, F. & DEPREZ-POULAIN, R. 2014. Inhibition of aggrecanases as a therapeutic strategy in osteoarthritis. Future Med Chem, 6, 1399-1412.

142

ERRINGTON, R. J., PUUSTJARVI, K., WHITE, I. R., ROBERTS, S. & URBAN, J. P. 1998. Characterisation of cytoplasm-filled processes in cells of the intervertebral disc. J Anat, 192 ( Pt 3), 369-78. ESKO, J., KIMATA, K. & LINDAHL, U. 2009. Essentials of Glycobiology, Cold Spring Harbour (NY), Cold Spring Harbor Laboratory Press. FANG, M., GOLDSTEIN, E. L., TURNER, A. S., LES, C. M., ORR, B. G., FISHER, G. J., WELCH, K. B., ROTHMAN, E. D. & BANASZAK HOLL, M. M. 2012. Type I collagen D-spacing in fibril bundles of dermis, tendon, and bone: bridging between nano- and micro-level tissue hierarchy. ACS Nano, 6, 9503-9514. FISCHER, J., DICKHUT, A., RICKERT, M. & RICHTER, W. 2010. Human articular chondrocytes secrete parathyroid hormone-related protein and inhibit hypertrophy of mesenchymal stem cells in coculture during chondrogenesis. Arthritis Rheum, 62, 2696-706. FONOVIC, M. & TURK, B. 2014. Cysteine cathepsins and extracellular matrix degradation. Biochim Biophys Acta, 1840, 2560-70. FOSANG, A. J., NEAME, P. J., LAST, K., HARDINGHAM, T. E., MURPHY, G. & HAMILTON, J. A. 1992a. The interglobular domain of cartilage aggrecan is cleaved by PUMP, gelatinases, and cathepsin B. J Biol Chem, 267, 19470-4. FOSANG, A. J., NEAME, P. J., LAST, K., HARDINGHAM, T. E., MURPHY, G. & HAMILTON, J. A. 1992b. The interglobular domain of cartilage aggrecan is cleaved by PUMP, gelatinases, and cathepsin B. J Biol Chem, 267, 19470-19474. FREDERIKS, W. M. & MOOK, O. R. 2004. Metabolic mapping of proteinase activity with emphasis on in situ zymography of gelatinases: review and protocols. J Histochem Cytochem, 52, 711-22. FREEMONT, A. J., BYERS, R. J., TAIWO, Y. O. & HOYLAND, J. A. 1999. In situ zymographic localisation of type II collagen degrading activity in osteoarthritic human articular cartilage. Ann Rheum Dis, 58, 357-365. FUKUTA, S., MIYAMOTO, K., SUZUKI, K., MAEHARA, H., INOUE, T., HARA, A., KIKUIKE, K., TAGUCHI, A. & SHIMIZU, K. 2011. Abundance of calpain and aggrecan-cleavage products of calpain in degenerated human intervertebral discs. Osteoarthritis Cartilage, 19, 1254-1262. GANTENBEIN, B., CALANDRIELLO, E., WUERTZ-KOZAK, K., BENNEKER, L. M., KEEL, M. J. & CHAN, S. C. 2014. Activation of intervertebral disc cells by coculture with notochordal cells, conditioned medium and hypoxia. BMC Musculoskelet Disord, 15, 422. GARCIA-SUAREZ, O., GARCIA, B., FERNANDEZ-VEGA, I., ASTUDILLO, A. & QUIROS, L. M. 2014. Neuroendocrine tumors show altered expression of chondroitin sulfate, glypican 1, glypican 5, and syndecan 2 depending on their differentiation grade. Front Oncol, 4, 15. GAUVIN, R., PARENTEAU-BAREIL, R., DOKMECI, M. R., MERRYMAN, W. D. & KHADEMHOSSEINI, A. 2012. Hydrogels and microtechnologies for engineering the cellular microenvironment. Wiley Interdiscip Rev Nanomed Nanobiotechnol, 4, 235-46. GENDRON, C., KASHIWAGI, M., LIM, N. H., ENGHILD, J. J., THOGERSEN, I. B., HUGHES, C., CATERSON, B. & NAGASE, H. 2007. Proteolytic activities of human ADAMTS-5: comparative studies with ADAMTS-4. J Biol Chem, 282, 18294-306. GEPSTEIN, A., SHAPIRO, S., ARBEL, G., LAHAT, N. & LIVNE, E. 2002. Expression of matrix metalloproteinases in articular cartilage of temporomandibular and knee joints of mice during growth, maturation, and aging. Arthritis Rheum, 46, 3240-3250.

143

GLOVER, M. G., HARGENS, A. R., MAHMOOD, M. M., GOTT, S., BROWN, M. D. & GARFIN, S. R. 1991. A new technique for the in vitro measurement of nucleus pulposus swelling pressure. J Orthop Res, 9, 61-7. GOLDRING, M. B. 2000. The role of the chondrocyte in osteoarthritis. Arthritis Rheum, 43, 1916-26. GORENSEK, M., JAKSIMOVIC, C., KREGAR-VELIKONJA, N., GORENSEK, M., KNEZEVIC, M., JERAS, M., PAVLOVCIC, V. & COR, A. 2004. Nucleus pulposus repair with cultured autologous elastic cartilage derived chondrocytes. Cell Mol Biol Lett, 9, 363-73. GORMAN, M., ZAIDI, A., AZAD, B., BEALES, S., KULSCAR, A., KNOX- VYDMANOV, C., MIHNOVITS, A., MIKKONEN-JEANNERET, E., ONG, P., SCOBIE, J. & SKINNER, M. 2013. Global AgeWatch Index 2013: Insight report [Online]. HelpAge International. Available: http://www.helpage.org/global- agewatch/data/global-rankings-table/gclid=CO37iKaOoLoCFfHItAodviIAOA [Accessed 18/10/13]. GOTTSCHALK, D., FEHN, M., PATT, S., SAEGER, W., KIRCHNER, T. & AIGNER, T. 2001. Matrix gene expression analysis and cellular phenotyping in chordoma reveals focal differentiation pattern of neoplastic cells mimicking nucleus pulposus development. Am J Pathol, 158, 1571-8. GOTZ, W. & QUONDAMATTEO, F. 2001. Glycoconjugate distribution in early human notochord and axial mesenchyme. Acta Histochem, 103, 21-35. GRAHAM, H. K., AKHTAR, R., KRIDIOTIS, C., DERBY, B., KUNDU, T., TRAFFORD, A. W. & SHERRATT, M. J. 2011. Localised micro-mechanical stiffening in the ageing aorta. Mech Ageing Dev, 132, 459-467. GRAHAM, H. K., HODSON, N. W., HOYLAND, J. A., MILLWARD-SADLER, S. J., GARROD, D., SCOTHERN, A., GRIFFITHS, C. E., WATSON, R. E., COX, T. R., ERLER, J. T., TRAFFORD, A. W. & SHERRATT, M. J. 2010. Tissue section AFM: In situ ultrastructural imaging of native biomolecules. Matrix Biol, 29, 254-60. GRAHAM, H. K. & TRAFFORD, A. W. 2007. Spatial disruption and enhanced degradation of collagen with the transition from compensated ventricular hypertrophy to symptomatic congestive heart failure. Am J Physiol Heart Circ Physiol, 292, H1364- 1372. GRAHAM, J. S., VOMUND, A. N., PHILLIPS, C. L. & GRANDBOIS, M. 2004. Structural changes in human type I collagen fibrils investigated by force spectroscopy. Exp Cell Res, 299, 335-42. GREENE, M. A. & LOESER, R. F. 2015. Aging-related inflammation in osteoarthritis. Osteoarthritis Cartilage, 23, 1966-1971. GRIBBON, P. & HARDINGHAM, T. E. 1998. Macromolecular diffusion of biological polymers measured by confocal fluorescence recovery after photobleaching. Biophys J, 75, 1032-1039. GRODZINSKY, A. J. 1983. Electromechanical and physicochemical properties of connective tissue. Crit Rev Biomed Eng, 9, 133-99. GRUBER, H. E., INGRAM, J. A., HOELSCHER, G. L., ZINCHENKO, N., NORTON, H. J. & HANLEY, E. N., JR. 2011. Constitutive expression of cathepsin K in the human intervertebral disc: new insight into disc extracellular matrix remodeling via cathepsin K and receptor activator of nuclear factor-kappaB ligand. Arthritis Res Ther, 13, R140. HAMAMURA, K., ZHANG, P., ZHAO, L., SHIM, J. W., CHEN, A., DODGE, T. R., WAN, Q., SHIH, H., NA, S., LIN, C. C., SUN, H. B. & YOKOTA, H. 2013. Knee loading reduces MMP13 activity in the mouse cartilage. BMC Musculoskelet Disord, 14, 312.

144

HAN, L., DEAN, D., MAO, P., ORTIZ, C. & GRODZINSKY, A. J. 2005. Nanoscale Conformation and Compressibility of Cartilage Aggrecan Using Microcontact Printing and Atomic Force Microscopy. Macromolecules, 38, 4047-4049. HANDLEY, C. J., MOK, M. T., ILIC, M. Z., ADCOCKS, C., BUTTLE, D. J. & ROBINSON, H. C. 2001. Cathepsin D cleaves aggrecan at unique sites within the interglobular domain and chondroitin sulfate attachment regions that are also cleaved when cartilage is maintained at acid pH. Matrix Biol, 20, 543-53. HANNESSON, K. O., YTTEBORG, E., TAKLE, H., ENERSEN, G., BAEVERFJORD, G. & PEDERSEN, M. E. 2015. Sulphated glycosaminoglycans and proteoglycans in the developing vertebral column of juvenile Atlantic salmon (Salmo salar). Fish Physiol Biochem, 41, 1029-51. HARDINGHAM, T. E. & FOSANG, A. J. 1992. Proteoglycans: many forms and many functions. FASEB J, 6, 861-70. HAYES, A. J., SHU, C. C., LORD, M. S., LITTLE, C. B., WHITELOCK, J. M. & MELROSE, J. 2016. Pericellular colocalisation and interactive properties of type VI collagen and perlecan in the intervertebral disc. Eur Cell Mater, 32, 40-57. HEMBRY, R. M., ATKINSON, S. J. & MURPHY, G. 2007. Assessment of gelatinase expression and activity in articular cartilage. Methods Mol Med, 135, 227-238. HEMBRY, R. M., BAGGA, M. R., REYNOLDS, J. J. & HAMBLEN, D. L. 1995. Immunolocalisation studies on six matrix metalloproteinases and their inhibitors, TIMP-1 and TIMP-2, in synovia from patients with osteo- and rheumatoid arthritis. Ann Rheum Dis, 54, 25-32. HERING, T. M., KOLLAR, J. & HUYNH, T. D. 1997. Complete coding sequence of bovine aggrecan: comparative structural analysis. Arch Biochem Biophys, 345, 259-70. HILDEBRAND, A., ROMARIS, M., RASMUSSEN, L. M., HEINEGARD, D., TWARDZIK, D. R., BORDER, W. A. & RUOSLAHTI, E. 1994. Interaction of the small interstitial proteoglycans biglycan, decorin and fibromodulin with transforming growth factor beta. Biochem J, 302 ( Pt 2), 527-34. HOLMES, D. F. & KADLER, K. E. 2006. The 10+4 microfibril structure of thin cartilage fibrils. Proc Natl Acad Sci U S A, 103, 17249-17254. HOLMES, M. W., BAYLISS, M. T. & MUIR, H. 1988. Hyaluronic acid in human articular cartilage. Age-related changes in content and size. Biochem J, 250, 435-41. HORCAS, I., FERNANDEZ, R., GOMEZ-RODRIGUEZ, J. M., COLCHERO, J., GOMEZ- HERRERO, J. & BARO, A. M. 2007. WSXM: a software for scanning probe microscopy and a tool for nanotechnology. Rev Sci Instrum, 78, 013705. HOU, W. S., LI, Z., BUTTNER, F. H., BARTNIK, E. & BROMME, D. 2003. Cleavage site specificity of cathepsin K toward cartilage proteoglycans and protease complex formation. Biol Chem, 384, 891-897. HUANG, K. & WU, L. D. 2008. Aggrecanase and aggrecan degradation in osteoarthritis: a review. J Int Med Res, 36, 1149-60. HUCKERBY, T. N., LAUDER, R. M., BROWN, G. M., NIEDUSZYNSKI, I. A., ANDERSON, K., BOOCOCK, J., SANDALL, P. L. & WEEKS, S. D. 2001. Characterization of oligosaccharides from the chondroitin sulfates. (1)H-NMR and (13)C-NMR studies of reduced disaccharides and tetrasaccharides. Eur J Biochem, 268, 1181-9. HUGHES, C. E., LITTLE, C. B., BUTTNER, F. H., BARTNIK, E. & CATERSON, B. 1998. Differential expression of aggrecanase and matrix metalloproteinase activity in chondrocytes isolated from bovine and porcine articular cartilage. J Biol Chem, 273, 30576-82.

145

HUNTER, C. J., MATYAS, J. R. & DUNCAN, N. A. 2003. The notochordal cell in the nucleus pulposus: a review in the context of tissue engineering. Tissue Eng, 9, 667- 77. HUNTER, C. J., MATYAS, J. R. & DUNCAN, N. A. 2004. Cytomorphology of notochordal and chondrocytic cells from the nucleus pulposus: a species comparison. J Anat, 205, 357-62. HUNZIKER, E. B. 2002. Articular cartilage repair: basic science and clinical progress. A review of the current status and prospects. Osteoarthritis Cartilage, 10, 432-63. IIJIMA, H., AOYAMA, T., ITO, A., TAJINO, J., NAGAI, M., ZHANG, X., YAMAGUCHI, S., AKIYAMA, H. & KUROKI, H. 2014. Immature articular cartilage and subchondral bone covered by menisci are potentially susceptive to mechanical load. BMC Musculoskelet Disord, 15, 101. ILIC, M. Z., MOK, M. T., WILLIAMSON, O. D., CAMPBELL, M. A., HUGHES, C. E. & HANDLEY, C. J. 1995. Catabolism of aggrecan by explant cultures of human articular cartilage in the presence of retinoic acid. Arch Biochem Biophys, 322, 22- 30. INAMDAR, S. R., KNIGHT, D. P., TERRILL, N. J., KARUNARATNE, A., CACHO- NERIN, F., KNIGHT, M. M. & GUPTA, H. S. 2017. The Secret Life of Collagen: Temporal Changes in Nanoscale Fibrillar Pre-Strain and Molecular Organization During Physiological Loading of Cartilage. ACS Nano. ISSY, A. C., CASTANIA, V., CASTANIA, M., SALMON, C. E., NOGUEIRA- BARBOSA, M. H., BEL, E. D. & DEFINO, H. L. 2013. Experimental model of intervertebral disc degeneration by needle puncture in Wistar rats. Braz J Med Biol Res, 46, 235-244. IU, J., MASSICOTTE, E., LI, S. Q., HURTIG, M. B., TOYSERKANI, E., SANTERRE, J. P. & KANDEL, R. A. 2017. In Vitro Generated Intervertebral Discs: Toward Engineering Tissue Integration. Tissue Eng Part A. JACKSON, A. & GU, W. 2009. Transport Properties of Cartilaginous Tissues. Curr Rheumatol Rev, 5, 40. JANUSZ, M. J., LITTLE, C. B., KING, L. E., HOOKFIN, E. B., BROWN, K. K., HEITMEYER, S. A., CATERSON, B., POOLE, A. R. & TAIWO, Y. O. 2004. Detection of aggrecanase- and MMP-generated catabolic neoepitopes in the rat iodoacetate model of cartilage degeneration. Osteoarthritis Cartilage, 12, 720-8. JOWITT, T. A., MURDOCH, A. D., BALDOCK, C., BERRY, R., DAY, J. M. & HARDINGHAM, T. E. 2010. Order within disorder: aggrecan chondroitin sulphate- attachment region provides new structural insights into protein sequences classified as disordered. Proteins, 78, 3317-3327. KEARNS, A. E., CAMPBELL, S. C., WESTLEY, J. & SCHWARTZ, N. B. 1991. Initiation of chondroitin sulfate biosynthesis: a kinetic analysis of UDP-D-xylose: core protein beta-D-xylosyltransferase. Biochemistry, 30, 7477-83. KEARNS, A. E., VERTEL, B. M. & SCHWARTZ, N. B. 1993. Topography of glycosylation and UDP-xylose production. J Biol Chem, 268, 11097-104. KIANI, C., CHEN, L., WU, Y. J., YEE, A. J. & YANG, B. B. 2002. Structure and function of aggrecan. Cell Res, 12, 19-32. KIENLE, S., GALLEI, M., YU, H., ZHANG, B., KRYSIAK, S., BALZER, B. N., REHAHN, M., SCHLUTER, A. D. & HUGEL, T. 2014. Effect of molecular architecture on single polymer adhesion. Langmuir, 30, 4351-4357. KIM, K. W., LIM, T. H., KIM, J. G., JEONG, S. T., MASUDA, K. & AN, H. S. 2003. The origin of chondrocytes in the nucleus pulposus and histologic findings associated

146

with the transition of a notochordal nucleus pulposus to a fibrocartilaginous nucleus pulposus in intact rabbit intervertebral discs. Spine (Phila Pa 1976), 28, 982-90. KOCSIS, E., TRUS, B. L., STEER, C. J., BISHER, M. E. & STEVEN, A. C. 1991. Image averaging of flexible fibrous macromolecules: the clathrin triskelion has an elastic proximal segment. J Struct Biol, 107, 6-14. KONTTINEN, Y. T., KAAPA, E., HUKKANEN, M., GU, X. H., TAKAGI, M., SANTAVIRTA, S., ALARANTA, H., LI, T. F. & SUDA, A. 1999. Cathepsin G in degenerating and healthy discal tissue. Clin Exp Rheumatol, 17, 197-204. KONTTINEN, Y. T., MANDELIN, J., LI, T. F., SALO, J., LASSUS, J., LILJESTROM, M., HUKKANEN, M., TAKAGI, M., VIRTANEN, I. & SANTAVIRTA, S. 2002. Acidic cysteine endoproteinase cathepsin K in the degeneration of the superficial articular hyaline cartilage in osteoarthritis. Arthritis Rheum, 46, 953-960. KOPESKY, P. W., LEE, H. Y., VANDERPLOEG, E. J., KISIDAY, J. D., FRISBIE, D. D., PLAAS, A. H., ORTIZ, C. & GRODZINSKY, A. J. 2010. Adult equine bone marrow stromal cells produce a cartilage-like ECM mechanically superior to animal-matched adult chondrocytes. Matrix Biol, 29, 427-38. KULICKE, W.-M. & CLASEN, C. 2004. Viscosimetry of polymers and polyelectrolytes, Berlin ; New York, Springer. KUNO, K., OKADA, Y., KAWASHIMA, H., NAKAMURA, H., MIYASAKA, M., OHNO, H. & MATSUSHIMA, K. 2000. ADAMTS-1 cleaves a cartilage proteoglycan, aggrecan. FEBS Lett, 478, 241-5. LAUDER, R. M., HUCKERBY, T. N., BROWN, G. M., BAYLISS, M. T. & NIEDUSZYNSKI, I. A. 2001. Age-related changes in the sulphation of the chondroitin sulphate linkage region from human articular cartilage aggrecan. Biochem J, 358, 523-8. LAUDER, R. M., HUCKERBY, T. N., NIEDUSZYNSKI, I. A. & PLAAS, A. H. 1998. Age-related changes in the structure of the keratan sulphate chains attached to fibromodulin isolated from articular cartilage. Biochem J, 330 ( Pt 2), 753-7. LEE, H. Y., HAN, L., ROUGHLEY, P. J., GRODZINSKY, A. J. & ORTIZ, C. 2013. Age- related nanostructural and nanomechanical changes of individual human cartilage aggrecan monomers and their glycosaminoglycan side chains. J Struct Biol, 181, 264-73. LEE, H. Y., KOPESKY, P. W., PLAAS, A., SANDY, J., KISIDAY, J., FRISBIE, D., GRODZINSKY, A. J. & ORTIZ, C. 2010. Adult bone marrow stromal cell-based tissue-engineered aggrecan exhibits ultrastructure and nanomechanical properties superior to native cartilage. Osteoarthritis Cartilage, 18, 1477-86. LEE, K. I., MOON, S. H., KIM, H., KWON, U. H., KIM, H. J., PARK, S. N., SUH, H., LEE, H. M., KIM, H. S., CHUN, H. J., KWON, I. K. & JANG, J. W. 2012. Tissue engineering of the intervertebral disc with cultured nucleus pulposus cells using atelocollagen scaffold and growth factors. Spine (Phila Pa 1976), 37, 452-8. LI, B., URBAN, J. P. & YU, J. 2012. The distribution of fibrillin-2 and LTBP-2, and their co-localisation with fibrillin-1 in adult bovine tail disc. J Anat, 220, 164-72. LI, H., DOMOWICZ, M., HENNIG, A. & SCHWARTZ, N. B. 1996. S103L reactive chondroitin sulfate proteoglycan (aggrecan) mRNA expressed in developing chick brain and cartilage is encoded by a single gene. Brain Res Mol Brain Res, 36, 309- 21. LI, P., GAN, Y., XU, Y., LI, S., SONG, L., LI, S., LI, H. & ZHOU, Q. 2016. Osmolarity affects matrix synthesis in the nucleus pulposus associated with the involvement of MAPK pathways: A study of ex vivo disc organ culture system. J Orthop Res, 34, 1092-100.

147

LI, Z., HOU, W. S. & BROMME, D. 2000. Collagenolytic activity of cathepsin K is specifically modulated by cartilage-resident chondroitin sulfates. Biochemistry, 39, 529-36. LI, Z., HOU, W. S., ESCALANTE-TORRES, C. R., GELB, B. D. & BROMME, D. 2002. Collagenase activity of cathepsin K depends on complex formation with chondroitin sulfate. J Biol Chem, 277, 28669-76. LI, Z., YASUDA, Y., LI, W., BOGYO, M., KATZ, N., GORDON, R. E., FIELDS, G. B. & BROMME, D. 2004. Regulation of collagenase activities of human cathepsins by glycosaminoglycans. J Biol Chem, 279, 5470-9. LINDBURG, C. A., WILLEY, J. S. & DEAN, D. 2013. Effects of low dose X-ray irradiation on porcine articular cartilage explants. J Orthop Res, 31, 1780-5. LIPARI, L. & GERBINO, A. 2013. Expression of gelatinases (MMP-2, MMP-9) in human articular cartilage. Int J Immunopathol Pharmacol, 26, 817-23. LITTLE, C. B., FLANNERY, C. R., HUGHES, C. E., MORT, J. S., ROUGHLEY, P. J., DENT, C. & CATERSON, B. 1999. Aggrecanase versus matrix metalloproteinases in the catabolism of the interglobular domain of aggrecan in vitro. Biochem J, 344 Pt 1, 61-8. LIU, Y., FU, S., RAHAMAN, M. N., MAO, J. J. & BAL, B. S. 2015. Native nucleus pulposus tissue matrix promotes notochordal differentiation of human induced pluripotent stem cells with potential for treating intervertebral disc degeneration. J Biomed Mater Res A, 103, 1053-9. LIU, Y., RAHAMAN, M. N. & BAL, B. S. 2014. Modulating notochordal differentiation of human induced pluripotent stem cells using natural nucleus pulposus tissue matrix. PLoS One, 9, e100885. LOHMANDER, L. S., SHINOMURA, T., HASCALL, V. C. & KIMURA, J. H. 1989. Xylosyl transfer to the core protein precursor of the rat chondrosarcoma proteoglycan. J Biol Chem, 264, 18775-80. MAEHARA, H., SUZUKI, K., SASAKI, T., OSHITA, H., WADA, E., INOUE, T. & SHIMIZU, K. 2007. G1-G2 aggrecan product that can be generated by M-calpain on truncation at Ala709-Ala710 is present abundantly in human articular cartilage. J Biochem, 141, 469-77. MALFAIT, A. M. 2016. Osteoarthritis year in review 2015: biology. Osteoarthritis Cartilage, 24, 21-6. MALIK, K. M., COHEN, S. P., WALEGA, D. R. & BENZON, H. T. 2013. Diagnostic criteria and treatment of discogenic pain: a systematic review of recent clinical literature. Spine J. MANIADAKIS, N. & GRAY, A. 2000. The economic burden of back pain in the UK. Pain, 84, 95-103. MANSFIELD, J. C., BELL, J. S., WINLOVE, C. P. 2015. The micromechanics of the superficial zone of articular cartilage. Osteoarthritis Cartilage, 23, 1806-16. MAROUDAS, A., BAYLISS, M. T., UCHITEL-KAUSHANSKY, N., SCHNEIDERMAN, R. & GILAV, E. 1998. Aggrecan turnover in human articular cartilage: use of aspartic acid racemization as a marker of molecular age. Arch Biochem Biophys, 350, 61-71. MATTHEWS, R. T., KELLY, G. M., ZERILLO, C. A., GRAY, G., TIEMEYER, M. & HOCKFIELD, S. 2002. Aggrecan glycoforms contribute to the molecular heterogeneity of perineuronal nets. J Neurosci, 22, 7536-47. MCCANN, M. R., TAMPLIN, O. J., ROSSANT, J. & SÉGUIN, C. A. 2012. Tracing notochord-derived cells using a Noto-cre mouse: implications for intervertebral disc development. Dis Model Mech, 5, 73-82.

148

MCCONNELL, J. C., O'CONNEL, O. V., BRENNAN, K., WEIPING, L., HOWE, M., JOSEPH, L., KNIGHT, D., O'CUALAIN, R., LIM, Y., LEEK, A., WADDINGTON, R., ROGAN, J., M., A. S., GANDHI, A., KIRWAN, C. C., SHERRATT, M. J. & STREULI, C. H. 2016. Increased peri-ductal collagen micro-organization may contribute to raised mammographic density. Breast Cancer Research, 18, 5. MCLEOD, M. A., WILUSZ, R. E. & GUILAK, F. 2013. Depth-dependent anisotropy of the micromechanical properties of the extracellular and pericellular matrices of articular cartilage evaluated via atomic force microscopy. J Biomech, 46, 586-92. MIENALTOWSKI, M. J. & BIRK, D. E. 2014. Structure, physiology, and biochemistry of collagens. Adv Exp Med Biol, 802, 5-29. MINAROWSKA, A., GACKO, M., KARWOWSKA, A. & MINAROWSKI, L. 2008. Human cathepsin D. Folia Histochem Cytobiol, 46, 23-38. MIWA, H. E., GERKEN, T. A., HUYNH, T. D., DUESLER, L. R., COTTER, M. & HERING, T. M. 2009. Conserved sequence in the aggrecan interglobular domain modulates cleavage by ADAMTS-4 and ADAMTS-5. Biochim Biophys Acta, 1790, 161-72. MOCHIDA, Y., PARISUTHIMAN, D. & YAMAUCHI, M. 2006. Biglycan is a positive modulator of BMP-2 induced osteoblast differentiation. Adv Exp Med Biol, 585, 101- 13. MONGIAT, M., ANDREUZZI, E., TARTICCHIO, G. & PAULITTI, A. 2016. Extracellular Matrix, a Hard Player in Angiogenesis. Int J Mol Sci, 17. MONTPLAISIR, S. 1979. Latency and activation of cytomegalovirus in man and in mice. Can J Microbiol, 25, 261-6. MOOK, O. R., VAN OVERBEEK, C., ACKEMA, E. G., VAN MALDEGEM, F. & FREDERIKS, W. M. 2003. In situ localization of gelatinolytic activity in the extracellular matrix of metastases of colon cancer in rat liver using quenched fluorogenic DQ-gelatin. J Histochem Cytochem, 51, 821-9. MORGELIN, M., PAULSSON, M., HARDINGHAM, T. E., HEINEGARD, D. & ENGEL, J. 1988. Cartilage proteoglycans. Assembly with hyaluronate and link protein as studied by electron microscopy. Biochem J, 253, 175-85. MÖRGELIN, M., PAULSSON, M., MALMSTRÖM, A. & HEINEGÅRD, D. 1989. Shared and distinct structural features of interstitial proteoglycans from different bovine tissues revealed by electron microscopy. J Biol Chem, 264, 12080-90. MORT, J. S., MAGNY, M. C. & LEE, E. R. 1998. Cathepsin B: an alternative protease for the generation of an aggrecan 'metalloproteinase' cleavage neoepitope. Biochem J, 335 ( Pt 3), 491-4. MWALE, F., ROUGHLEY, P. & ANTONIOU, J. 2004. Distinction between the extracellular matrix of the nucleus pulposus and hyaline cartilage: a requisite for tissue engineering of intervertebral disc. Eur Cell Mater, 8, 58-64. NAKASE, T., KANEKO, M., TOMITA, T., MYOUI, A., ARIGA, K., SUGAMOTO, K., UCHIYAMA, Y., OCHI, T. & YOSHIKAWA, H. 2000. Immunohistochemical detection of cathepsin D, K, and L in the process of endochondral ossification in the human. Histochem Cell Biol, 114, 21-7. NASTASE, M. V., YOUNG, M. F. & SCHAEFER, L. 2012. Biglycan: a multivalent proteoglycan providing structure and signals. J Histochem Cytochem, 60, 963-75. NAVARO, Y., BLEICH-KIMELMAN, N., HAZANOV, L., MIRONI-HARPAZ, I., SHACHAF, Y., GARTY, S., SMITH, Y., PELLED, G., GAZIT, D., SELIKTAR, D. & GAZIT, Z. 2015. Matrix stiffness determines the fate of nucleus pulposus-derived stem cells. Biomaterials, 49, 68-76.

149

NAYLOR, E. C., WATSON, R. E. & SHERRATT, M. J. 2011. Molecular aspects of skin ageing. Maturitas, 69, 249-56. NEAME, P. J. & BARRY, F. P. 1993. The link proteins. Experientia, 49, 393-402. NEILL, T., SCHAEFER, L. & IOZZO, R. V. 2012. Decorin: a guardian from the matrix. Am J Pathol, 181, 380-7. NG, L., GRODZINSKY, A. J., PATWARI, P., SANDY, J., PLAAS, A. & ORTIZ, C. 2003. Individual cartilage aggrecan macromolecules and their constituent glycosaminoglycans visualized via atomic force microscopy. J Struct Biol, 143, 242- 57. NGUYEN, Q., LIU, J., ROUGHLEY, P. J. & MORT, J. S. 1991. Link protein as a monitor in situ of endogenous proteolysis in adult human articular cartilage. Biochem J, 278 ( Pt 1), 143-7. NGUYEN, Q., MORT, J. S. & ROUGHLEY, P. J. 1990. Cartilage proteoglycan aggregate is degraded more extensively by cathepsin L than by cathepsin B. Biochem J, 266, 569-73. NIA, H. T., HAN, L., BOZCHALOOI, I. S., ROUGHLEY, P., YOUCEF-TOUMI, K., GRODZINSKY, A. J. & ORTIZ, C. 2015. Aggrecan nanoscale solid-fluid interactions are a primary determinant of cartilage dynamic mechanical properties. ACS Nano, 9, 2614-25. NICE. 2009. National costing report: low back pain (May 2009) [Online]. NHS. Available: http://www.nice.org.uk/nicemedia/pdf/CG88CostReport.pdf [Accessed 15/10/2013. NIEDUSZYNSKI, I. A., HUCKERBY, T. N., DICKENSON, J. M., BROWN, G. M., TAI, G. H. & BAYLISS, M. T. 1990. Structural aspects of skeletal keratan sulphates. Biochem Soc Trans, 18, 792-3. NUWAYHID, N., GLASER, J. H., JOHNSON, J. C., CONRAD, H. E., HAUSER, S. C. & HIRSCHBERG, C. B. 1986. Xylosylation and glucuronosylation reactions in rat liver Golgi apparatus and endoplasmic reticulum. J Biol Chem, 261, 12936-41. OEGEMA, T. R., JR. 2002. The role of disc cell heterogeneity in determining disc biochemistry: a speculation. Biochem Soc Trans, 30, 839-44. OHTA, S., IMAI, K., YAMASHITA, K., MATSUMOTO, T., AZUMANO, I. & OKADA, Y. 1998. Expression of matrix metalloproteinase 7 (matrilysin) in human osteoarthritic cartilage. Lab Invest, 78, 79-87. OKADA, Y., SHINMEI, M., TANAKA, O., NAKA, K., KIMURA, A., NAKANISHI, I., BAYLISS, M. T., IWATA, K. & NAGASE, H. 1992. Localization of matrix metalloproteinase 3 (stromelysin) in osteoarthritic cartilage and synovium. Lab Invest, 66, 680-90. OLIN, A. I., MORGELIN, M., SASAKI, T., TIMPL, R., HEINEGARD, D. & ASPBERG, A. 2001. The proteoglycans aggrecan and Versican form networks with fibulin-2 through their lectin domain binding. J Biol Chem, 276, 1253-61. OMLOR, G. W., NERLICH, A. G., TIRLAPUR, U. K., URBAN, J. P. & GUEHRING, T. 2014. Loss of notochordal cell phenotype in 3D-cell cultures: implications for disc physiology and disc repair. Arch Orthop Trauma Surg, 134, 1673-81. OSHITA, H., SANDY, J. D., SUZUKI, K., AKAIKE, A., BAI, Y., SASAKI, T. & SHIMIZU, K. 2004. Mature bovine articular cartilage contains abundant aggrecan that is C-terminally truncated at Ala719-Ala720, a site which is readily cleaved by m-calpain. Biochem J, 382, 253-9. PELLETIER, J. P., BOILEAU, C., BOILY, M., BRUNET, J., MINEAU, F., GENG, C., REBOUL, P., LAUFER, S., LAJEUNESSE, D. & MARTEL-PELLETIER, J. 2005. The protective effect of licofelone on experimental osteoarthritis is correlated with the downregulation of gene expression and protein synthesis of several major

150

cartilage catabolic factors: MMP-13, cathepsin K and aggrecanases. Arthritis Res Ther, 7, R1091-102. PETTWAY, Z., DOMOWICZ, M., SCHWARTZ, N. B. & BRONNER-FRASER, M. 1996. Age-dependent inhibition of neural crest migration by the notochord correlates with alterations in the S103L chondroitin sulfate proteoglycan. Exp Cell Res, 225, 195- 206. PITTENGER, M., MACKAY, A., BECK, S., JAISWAL, R., DOUGLAS, R., MOSCA, J., MOORMAN, M., SIMONETTI, D., CRAIG, S. & MARSHAK, D. 1999. Multilineage potential of adult human mesenchymal stem cells. Science, 284, 143-7. PLAAS, A. H., WEST, L., MIDURA, R. J. & HASCALL, V. C. 2001a. Disaccharide composition of hyaluronan and chondroitin/dermatan sulfate. Analysis with fluorophore-assisted carbohydrate electrophoresis. Methods Mol Biol, 171, 117-28. PLAAS, A. H., WEST, L. A. & MIDURA, R. J. 2001b. Keratan sulfate disaccharide composition determined by FACE analysis of keratanase II and endo-beta- galactosidase digestion products. Glycobiology, 11, 779-90. PLAAS, A. H., WEST, L. A., WONG-PALMS, S. & NELSON, F. R. 1998. Glycosaminoglycan sulfation in human osteoarthritis. Disease-related alterations at the non-reducing termini of chondroitin and dermatan sulfate. J Biol Chem, 273, 12642-9. PLAAS, A. H., WONG-PALMS, S., ROUGHLEY, P. J., MIDURA, R. J. & HASCALL, V. C. 1997. Chemical and immunological assay of the nonreducing terminal residues of chondroitin sulfate from human aggrecan. J Biol Chem, 272, 20603-10. PLODINEC, M., LOPARIC, M. & AEBI, U. 2010a. Atomic force microscopy for biological imaging and mechanical testing across length scales. Cold Spring Harb Protoc, 10, 86. PLODINEC, M., LOPARIC, M. & AEBI, U. 2010b. Imaging collagen II using atomic force microscopy (AFM). Cold Spring Harb Protoc, 2010, pdb prot5501. POCKERT, A. J., RICHARDSON, S. M., LE MAITRE, C. L., LYON, M., DEAKIN, J. A., BUTTLE, D. J., FREEMONT, A. J. & HOYLAND, J. A. 2009. Modified expression of the ADAMTS enzymes and tissue inhibitor of metalloproteinases 3 during human intervertebral disc degeneration. Arthritis Rheum, 60, 482-91. POOLE, C. A. 1997. Articular cartilage chondrons: form, function and failure. J Anat, 191 ( Pt 1), 1-13. POON, C. J., PLAAS, A. H., KEENE, D. R., MCQUILLAN, D. J., LAST, K. & FOSANG, A. J. 2005. N-linked keratan sulfate in the aggrecan interglobular domain potentiates aggrecanase activity. J Biol Chem, 280, 23615-21. PORTER, S., CLARK, I. M., KEVORKIAN, L. & EDWARDS, D. R. 2005. The ADAMTS metalloproteinases. Biochem J, 386, 15-27. POTIER, E., DE VRIES, S., VAN DOESELAAR, M. & ITO, K. 2014. Potential application of notochordal cells for intervertebral disc regeneration: an in vitro assessment. Eur Cell Mater, 28, 68-80; discussion 80-1. PRATTA, M. A., TORTORELLA, M. D. & ARNER, E. C. 2000. Age-related changes in aggrecan glycosylation affect cleavage by aggrecanase. J Biol Chem, 275, 39096- 102. PRUDNIKOVA, K., YUCHA, R. W., PATEL, P., KRIETE, A. S., HAN, L., PENN, L. S. & MARCOLONGO, M. S. 2017. Biomimetic Proteoglycans Mimic Macromolecular Architecture and Water Uptake of Natural Proteoglycans. Biomacromolecules, 18, 1713-1723. RAJ, P. P. 2008. Intervertebral disc: anatomy-physiology-pathophysiology-treatment. Pain Pract, 8, 18-44.

151

RASPANTI, M., VIOLA, M., SONAGGERE, M., TIRA, M. E. & TENNI, R. 2007. Collagen fibril structure is affected by collagen concentration and decorin. Biomacromolecules, 8, 2087-91. REZAKHANIHA, R., AGIANNIOTIS, A., SCHRAUWEN, J. T., GRIFFA, A., SAGE, D., BOUTEN, C. V., VAN DE VOSSE, F. N., UNSER, M. & STERGIOPULOS, N. 2012. Experimental investigation of collagen waviness and orientation in the arterial adventitia using confocal laser scanning microscopy. Biomech Model Mechanobiol, 11, 461-73. RISBUD, M. V., SCHAER, T. P. & SHAPIRO, I. M. 2010. Toward an understanding of the role of notochordal cells in the adult intervertebral disc: from discord to accord. Dev Dyn, 239, 2141-8. ROBERTS, S., EVANS, E. H., KLETSAS, D., JAFFRAY, D. C. & EISENSTEIN, S. M. 2006. Senescence in human intervertebral discs. Eur Spine J, 15 Suppl 3, S312-6. RODRIGUES-PINTO, R., RICHARDSON, S. M. & HOYLAND, J. A. 2014. An understanding of intervertebral disc development, maturation and cell phenotype provides clues to direct cell-based tissue regeneration therapies for disc degeneration. Eur Spine J, 23, 1803-14. ROSENBERG, L., HELLMANN, W. & KLEINSCHMIDT, A. K. 1975. Electron microscopic studies of proteoglycan aggregates from bovine articular cartilage. J Biol Chem, 250, 1877-83. ROUGHLEY, P. J., GENG, Y. & MORT, J. S. 2014. The non-aggregated aggrecan in the human intervertebral disc can arise by a non-proteolytic mechanism. Eur Cell Mater, 28, 129-136. RUETTGER, A., SCHUELER, S., MOLLENHAUER, J. A. & WIEDERANDERS, B. 2008. Cathepsins B, K, and L are regulated by a defined collagen type II peptide via activation of classical protein kinase C and p38 MAP kinase in articular chondrocytes. J Biol Chem, 283, 1043-51. RUTER, E. R. & KRESSE, H. 1984. Partial purification and characterization of 3'- phosphoadenylylsulfate:keratan sulfate sulfotransferases. J Biol Chem, 259, 11771- 6. RUTGES, J. P., NIKKELS, P. G., ONER, F. C., OTTINK, K. D., VERBOUT, A. J., CASTELEIN, R. J., CREEMERS, L. B. & DHERT, W. J. 2010. The presence of extracellular matrix degrading metalloproteinases during fetal development of the intervertebral disc. Eur Spine J, 19, 1340-6. SAHARINEN, J., HYYTIAINEN, M., TAIPALE, J. & KESKI-OJA, J. 1999. Latent transforming growth factor-beta binding proteins (LTBPs)--structural extracellular matrix proteins for targeting TGF-beta action. Cytokine Growth Factor Rev, 10, 99- 117. SALMINEN-MANKONEN, H. J., MORKO, J. & VUORIO, E. 2007. Role of cathepsin K in normal joints and in the development of arthritis. Curr Drug Targets, 8, 315-23. SANTER, V., WHITE, R. J. & ROUGHLEY, P. J. 1982. O-Linked oligosaccharides of human articular cartilage proteoglycan. Biochim Biophys Acta, 716, 277-82. SATTELLE, B. M., SHAKERI, J., CLIFF, M. J. & ALMOND, A. 2015. Proteoglycans and their heterogeneous glycosaminoglycans at the atomic scale. Biomacromolecules, 16, 951-61. SCHULZE WILLBRENNING, G., HISS, S., THEUNE, C., MIELENZ, M., SCHELLANDER, K. & SAUERWEIN, H. 2010. Gelatinase activities and haptoglobin concentrations in healthy and in degenerative articular cartilage of pigs. J Anim Physiol Anim Nutr (Berl), 94, 757-66.

152

SCHWARTZ, N. B., DOMOWICZ, M., KRUEGER, R. C., JR., LI, H. & MANGOURA, D. 1996. Brain aggrecan. Perspect Dev Neurobiol, 3, 291-306. SEOG, J., DEAN, D., ROLAUFFS, B., WU, T., GENZER, J., PLAAS, A. H., GRODZINSKY, A. J. & ORTIZ, C. 2005. Nanomechanics of opposing glycosaminoglycan macromolecules. J Biomech, 38, 1789-97. SHARMA, S., LEE, A., CHOI, K., KIM, K., YOUN, I., TRIPPEL, S. B. & PANITCH, A. 2013. Biomimetic aggrecan reduces cartilage extracellular matrix from degradation and lowers catabolic activity in ex vivo and in vivo models. Macromol Biosci, 13, 1228-37. SHENG, G. 2015. The developmental basis of mesenchymal stem/stromal cells (MSCs). BMC Dev Biol, 15, 44. SHERMAN, V. R., YANG, W. & MEYERS, M. A. 2015. The materials science of collagen. J Mech Behav Biomed Mater, 52, 22-50. SHERRATT, M. J., HOLMES, D. F., SHUTTLEWORTH, C. A. & KIELTY, C. M. 2004. Substrate-dependent morphology of supramolecular assemblies: fibrillin and type- VI collagen microfibrils. Biophys J, 86, 3211-22. SHOULDERS, M. D. & RAINES, R. T. 2009. Collagen structure and stability. Annu Rev Biochem, 78, 929-58. SIVAN, S. S., MERKHER, Y., WACHTEL, E., URBAN, J. P., LAZARY, A. & MAROUDAS, A. 2013. A needle micro-osmometer for determination of glycosaminoglycan concentration in excised nucleus pulposus tissue. Eur Spine J, 22, 1765-73. SIVAN, S. S., TSITRON, E., WACHTEL, E., ROUGHLEY, P. J., SAKKEE, N., VAN DER HAM, F., DEGROOT, J., ROBERTS, S. & MAROUDAS, A. 2006. Aggrecan turnover in human intervertebral disc as determined by the racemization of aspartic acid. J Biol Chem, 281, 13009-14. SIVAN, S. S., WACHTEL, E. & ROUGHLEY, P. 2014. Structure, function, aging and turnover of aggrecan in the intervertebral disc. Biochim Biophys Acta, 1840, 3181-9. SIVAN, S. S., WACHTEL, E., TSITRON, E., SAKKEE, N., VAN DER HAM, F., DEGROOT, J., ROBERTS, S. & MAROUDAS, A. 2008. Collagen turnover in normal and degenerate human intervertebral discs as determined by the racemization of aspartic acid. J Biol Chem, 283, 8796-801. SMITH, L. J., NERURKAR, N. L., CHOI, K. S., HARFE, B. D. & ELLIOTT, D. M. 2011. Degeneration and regeneration of the intervertebral disc: lessons from development. Dis Model Mech, 4, 31-41. SOMERVILLE, R. P., LONGPRE, J. M., JUNGERS, K. A., ENGLE, J. M., ROSS, M., EVANKO, S., WIGHT, T. N., LEDUC, R. & APTE, S. S. 2003. Characterization of ADAMTS-9 and ADAMTS-20 as a distinct ADAMTS subfamily related to Caenorhabditis elegans GON-1. J Biol Chem, 278, 9503-13. SONG, J., TAN, H., PERRY, A. J., AKUTSU, T., WEBB, G. I., WHISSTOCK, J. C. & PIKE, R. N. 2012. PROSPER: an integrated feature-based tool for predicting protease substrate cleavage sites. PLoS One, 7, e50300. SOPHIA FOX, A. J., BEDI, A. & RODEO, S. A. 2009. The basic science of articular cartilage: structure, composition, and function. Sports Health, 1, 461-8. SPILLEKOM, S., SMOLDERS, L. A., GRINWIS, G. C., ARKESTEIJN, I. T., ITO, K., MEIJ, B. P. & TRYFONIDOU, M. A. 2014. Increased osmolarity and cell clustering preserve canine notochordal cell phenotype in culture. Tissue Eng Part C Methods, 20, 652-62. STOYANOV, J. V., GANTENBEIN-RITTER, B., BERTOLO, A., AEBLI, N., BAUR, M., ALINI, M. & GRAD, S. 2011. Role of hypoxia and growth and differentiation factor-

153

5 on differentiation of human mesenchymal stem cells towards intervertebral nucleus pulposus-like cells. Eur Cell Mater, 21, 533-47. STRUGLICS, A. & HANSSON, M. 2010. Calpain is involved in C-terminal truncation of human aggrecan. Biochem J, 430, 531-8. STRUGLICS, A. & LARSSON, S. 2010. A comparison of different purification methods of aggrecan fragments from human articular cartilage and synovial fluid. Matrix Biol, 29, 74-83. SU, H. N., CHEN, Z. H., LIU, S. B., QIAO, L. P., CHEN, X. L., HE, H. L., ZHAO, X., ZHOU, B. C. & ZHANG, Y. Z. 2012. Characterization of bacterial polysaccharide capsules and detection in the presence of deliquescent water by atomic force microscopy. Appl Environ Microbiol, 78, 3476-9. SZOMOR, Z., SHIMIZU, K., YAMAMOTO, S., YASUDA, T., ISHIKAWA, H. & NAKAMURA, T. 1999. Externalization of calpain (calcium-dependent neutral cysteine proteinase) in human arthritic cartilage. Clin Exp Rheumatol, 17, 569-74. SZTROLOVICS, R., ALINI, M., ROUGHLEY, P. J. & MORT, J. S. 1997. Aggrecan degradation in human intervertebral disc and articular cartilage. Biochem J, 326 ( Pt 1), 235-41. TEWARI, A., GRYS, K., KOLLET, J., SARKANY, R. & YOUNG, A. R. 2014. Upregulation of MMP12 and its activity by UVA1 in human skin: potential implications for photoaging. J Invest Dermatol, 134, 2598-609. TREPPO, S., KOEPP, H., QUAN, E. C., COLE, A. A., KUETTNER, K. E. & GRODZINSKY, A. J. 2000. Comparison of biomechanical and biochemical properties of cartilage from human knee and ankle pairs. J Orthop Res, 18, 739-48. TUFVESSON, E. & WESTERGREN-THORSSON, G. 2002. Tumour necrosis factor-alpha interacts with biglycan and decorin. FEBS Lett, 530, 124-8. TURK, B., TURK, V. & TURK, D. 1997. Structural and functional aspects of papain-like cysteine proteinases and their protein inhibitors. Biol Chem, 378, 141-50. UNIPROT. 2014a. Aggrecan Core Protein P16112 (PGCA_BOVIN) [Online]. Available: http://www.uniprot.org/uniprot/P13608 [Accessed 26/07/2014. UNIPROT. 2014b. Aggrecan Core Protein P16112 (PGCA_HUMAN) [Online]. Available: http://www.uniprot.org/uniprot/P16112 [Accessed 26/07/2014. URBAN, J. P. 2002. The role of the physicochemical environment in determining disc cell behaviour. Biochem Soc Trans, 30, 858-64. URBAN, J. P., SMITH, S. & FAIRBANK, J. C. 2004. Nutrition of the intervertebral disc. Spine (Phila Pa 1976), 29, 2700-9. VAN DER REST, M. & GARRONE, R. 1990. Collagens as multidomain proteins. Biochimie, 72, 473-84. VEDICHERLA, S. & BUCKLEY, C. T. 2016. Cell-based therapies for intervertebral disc and cartilage regeneration - current concepts, parallels and perspectives. J Orthop Res, 35, 8-22. VERTEL, B. M., WALTERS, L. M., FLAY, N., KEARNS, A. E. & SCHWARTZ, N. B. 1993. Xylosylation is an endoplasmic reticulum to Golgi event. J Biol Chem, 268, 11105-12. VIKINGSSON, L., CLAESSENS, B., GOMEZ-TEJEDOR, J. A., GALLEGO FERRER, G. & GOMEZ RIBELLES, J. L. 2015. Relationship between micro-porosity, water permeability and mechanical behavior in scaffolds for cartilage engineering. J Mech Behav Biomed Mater, 48, 60-9. VITTORIO, N., CRISSMAN, J. D., HOPSON, C. N. & HERMAN, J. H. 1986. Histologic assessment of cathepsin D in osteoarthritic cartilage. Clin Exp Rheumatol, 4, 221-30.

154

VO, N. V., HARTMAN, R. A., YURUBE, T., JACOBS, L. J., SOWA, G. A. & KANG, J. D. 2013. Expression and regulation of metalloproteinases and their inhibitors in intervertebral disc aging and degeneration. Spine J, 13, 331-41. VYNIOS, D. H. 2014. Metabolism of cartilage proteoglycans in health and disease. Biomed Res Int, 2014, 452315. WALKER, B. F. 2000. The prevalence of low back pain: a systematic review of the literature from 1966 to 1998. J Spinal Disord, 13, 205-17. WALLACE, J. M. 2015. Effects of fixation and demineralization on bone collagen D- spacing as analyzed by atomic force microscopy. Connect Tissue Res, 56, 68-75. WANG, F., CAI, F., SHI, R., WANG, X. H. & WU, X. T. 2016. Aging and age related stresses: a senescence mechanism of intervertebral disc degeneration. Osteoarthritis Cartilage, 24, 398-408. WATANABE, H., CHEUNG, S. C., ITANO, N., KIMATA, K. & YAMADA, Y. 1997. Identification of hyaluronan-binding domains of aggrecan. J Biol Chem, 272, 28057- 65. WATANABE, T., HOSAKA, Y., YAMAMOTO, E., UEDA, H., SUGAWARA, K., TAKAHASHI, H. & TAKEHANA, K. 2005. Control of the collagen fibril diameter in the equine superficial digital flexor tendon in horses by decorin. J Vet Med Sci, 67, 855-60. WEILER, C. 2013. [In situ analysis of pathomechanisms of human intervertebral disc degeneration]. Pathologe, 34 Suppl 2, 251-9. WESTLING, J., FOSANG, A. J., LAST, K., THOMPSON, V. P., TOMKINSON, K. N., HEBERT, T., MCDONAGH, T., COLLINS-RACIE, L. A., LAVALLIE, E. R., MORRIS, E. A. & SANDY, J. D. 2002. ADAMTS4 cleaves at the aggrecanase site (Glu373-Ala374) and secondarily at the matrix metalloproteinase site (Asn341- Phe342) in the aggrecan interglobular domain. J Biol Chem, 277, 16059-66. WILLIAMS, A. R. & HARE, J. M. 2011. Mesenchymal stem cells: biology, pathophysiology, translational findings, and therapeutic implications for cardiac disease. Circ Res, 109, 923-40. WOODFIELD, T. B., MALDA, J., DE WIJN, J., PETERS, F., RIESLE, J. & VAN BLITTERSWIJK, C. A. 2004. Design of porous scaffolds for cartilage tissue engineering using a three-dimensional fiber-deposition technique. Biomaterials, 25, 4149-61. YAMAMOTO, S., SHIMIZU, K., SHIMIZU, K., SUZUKI, K., NAKAGAWA, Y. & YAMAMURO, T. 1992. Calcium-dependent cysteine proteinase (calpain) in human arthritic synovial joints. Arthritis Rheum, 35, 1309-17. YANG, F., LEUNG, V. Y., LUK, K. D., CHAN, D. & CHEUNG, K. M. 2009. Injury- induced sequential transformation of notochordal nucleus pulposus to chondrogenic and fibrocartilaginous phenotype in the mouse. J Pathol, 218, 113-21. YANG, S. H., HU, M. H., LO, W. Y., SUN, Y. H., WU, C. C. & YANG, K. C. 2017. The influence of oxygen concentration on the extracellular matrix production of human nucleus pulposus cells during isolation-expansion process. J Biomed Mater Res A, 105, 1575-1582. YANG, S. H., HU, M. H., SUN, Y. H. & LIN, F. H. 2013. Differential phenotypic behaviors of human degenerative nucleus pulposus cells under normoxic and hypoxic conditions: influence of oxygen concentration during isolation, expansion, and cultivation. Spine J, 13, 1590-6. YU, J., FAIRBANK, J. C., ROBERTS, S. & URBAN, J. P. 2005. The elastic fiber network of the anulus fibrosus of the normal and scoliotic human intervertebral disc. Spine (Phila Pa 1976), 30, 1815-20.

155

YU, J., TIRLAPUR, U., FAIRBANK, J., HANDFORD, P., ROBERTS, S., WINLOVE, C. P., CUI, Z. & URBAN, J. 2007. Microfibrils, elastin fibres and collagen fibres in the human intervertebral disc and bovine tail disc. J Anat, 210, 460-71. YU, J., WINLOVE, P. C., ROBERTS, S. & URBAN, J. P. 2002. Elastic fibre organization in the intervertebral discs of the bovine tail. J Anat, 201, 465-75. ZECKSER, J., WOLFF, M., TUCKER, J. & GOODWIN, J. 2016. Multipotent Mesenchymal Stem Cell Treatment for Discogenic Low Back Pain and Disc Degeneration. Stem Cells Int, 2, 1-13. ZHANG, G., YOUNG, B. B., EZURA, Y., FAVATA, M., SOSLOWSKY, L. J., CHAKRAVARTI, S. & BIRK, D. E. 2005. Development of tendon structure and function: regulation of collagen fibrillogenesis. J Musculoskelet Neuronal Interact, 5, 5-21. ZHAO, C. Q., WANG, L. M., JIANG, L. S. & DAI, L. Y. 2007. The cell biology of intervertebral disc aging and degeneration. Ageing Res Rev, 6, 247-61. ZHENG, J., LUO, W. & TANZER, M. L. 1998. Aggrecan synthesis and secretion. A paradigm for molecular and cellular coordination of multiglobular protein folding and intracellular trafficking. J Biol Chem, 273, 12999-3006.

156