IDENTIFICATION OF GENES REGULATING
OSTEOARTHRITIS DEVELOPMENT USING A
MOUSE PHENOTYPE LIBRARY
JACOB JOHN KENNY
This thesis is presented for the degree of Doctor of Philosophy of The University of Western Australia School of Biomedical Sciences Pathology 2018
i
THESIS DECLARATION
I, Jacob John Kenny, certify that: This thesis has been substantially accomplished during enrolment in the degree. This thesis does not contain material which has been accepted for the award of any other degree or diploma in my name, in any university or other tertiary institution. No part of this work will, in the future, be used in a submission in my name, for any other degree or diploma in any university or other tertiary institution without the prior approval of The University of Western Australia and where applicable, any partner institution responsible for the joint-award of this degree. This thesis does not contain any material previously published or written by another person, except where due reference has been made in the text. The work(s) are not in any way a violation or infringement of any copyright, trademark, patent, or other rights whatsoever of any person. The research involving animal data reported in this thesis was assessed and approved by The University of Western Australia Animal Ethics Committee. Approval #: RA/3/500/87. The research involving animals reported in this thesis followed The University of Western Australia and national standards for the care and use of laboratory animals. The work described in this thesis was funded by the National Health and Medical Research Council. APP1127156 – “Gene mining for novel molecular determinants of the skeleton”. This research was supported by an Australian Government Research Training Program (RTP) Scholarship. Technical assistance was kindly provided by Professor Grant Morahan, Sylvia Young and Dr Ramesh Ram for their support with gene mapping and the management of Geneminer software that is described throughout the thesis. Mr Kai Chen and Ms Dian Teguh for support in primary cell culture used for RNA analysis described in Chapter 3, section 3.2.7. Mr Jinbo Yuan for his fantastic help managing the CC mice collection as we received them. This thesis does not contain work that I have published, nor work under review for publication.
Signature:
Date: 13/11/2018
ii ABSTRACT
Osteoarthritis (OA) is a degenerative joint disease with a slow progression that affects the articular cartilage of synovial joints. It is characterized by the destruction of articular cartilage resulting in pain and limited mobility in the affected joint of sufferers, with treatment being limited to pain management and joint replacement. It is well understood that the end stage of OA is the loss of cartilage due to wearing of the joint, however the contribution of many factors in the synovial joint such as the bone, connective tissue and muscle, synovium and immune system all contribute to the pathogenesis of OA.
Considering that not all cases of OA are a result of a post-traumatic event, investigation of spontaneous forms of OA is of great importance to observe the disease progression without a mechanical initiation. Investigation of spontaneous OA in human genome wide association (GWAS), twin and family studies has suggested a heritability component that requires the investigation of genetic factors affecting OA development. Current mouse models rely on the initiation of knee OA by mechanical or chemical intervention, with limited consistent spontaneous models available. Without a better understanding of the genetics and how each of the factors contribute to the end stage of OA, alternative interventions and diagnostic tools will remain limited to very few aspects of the disease.
Using recombinant inbred mouse strains is an effective way of investigating the genetic component of OA. The Collaborative Cross is a large scale funnel breeding scheme derived from eight founder mouse strains for the purpose of capturing approximately 90% of variation in the mouse genome. The resulting recombinant inbred strains are vastly phenotypically diverse and can be screened for a range of genetically heritable traits such as knee OA. This study utilized knee joints of over 275 aged CC mice across 50 strains representing middle and old aged populations that were each screened using histological
iii techniques and scored for severity using the Osteoarthritis Research Society International
(OARSI) guidelines for knee OA assessment in the mouse. Analysis of parameters such as severity in the total cohort, 8-12 month and >12 month cohorts and the incidence of any OARSI scored OA and severe scores >3 were performed. Quantitative Trait Locus
(QTL) mapping was conducted on 222 mice across 43 strains to identify candidate genes responsible for the severity and risk of OA. Confirmation of candidate genes was conducted by investigating gene expression, protein interaction, correlation with human
GWAS datasets and publications and qPCR. A total of 7 loci were identified with 2 reaching genome wide significance with more than 90 genes being identified in the study.
Analysis of these genes for potential candidates revealed 8 genome wide significant genes: Myo1f, Daxx, Col11a2, Tap2, Glis3, Cd274 Pdcd1lg2 and Il33. Of these genes,
Col11a2, Il33 and Glis3 have been confirmed in human studies of OA, while Daxx, Tap2,
Cd274 and Pdcd1lg2 all have been implicated in studies of other forms of arthritis. Myo1f was identified as a novel candidate as it showed expression in osteoclasts and bone and interacted as part of a network with other known OA related candidates. A further 8 genes that reached a suggestive threshold which were also investigated: Cdh4, Col9a3, Cdh13,
Obscn, Tab1, Epas1, Jak2 and Dkk1. Of these Col9a3, Epas1, Jak2 and Dkk1 were confirmed in human studies of OA with Tab1 and Cdh13 being implicated in other forms of arthritis. Of the potentially novel candidates, Obscn was found to interact with RhoA, which has a role in a pathway responsible for the induction of chondrocyte differentiation.
Future studies of candidate gene knockout animal models and investigation of protein interactions will provide a clearer insight into the involvement of these candidate genes and how they contribute to the development of OA.
iv In summary, the results presented in this study validates the use of the collaborative cross for the characterization of spontaneous OA and the identification of known and novel OA susceptibility genes across all of the involved cell and tissue types. The identification of genes found in human OA studies also validates the use of mouse models for future studies of heritable spontaneous OA.
v TABLE OF CONTENTS
Thesis Declaration……………………………………………………….……………...…….ii Abstract..……………………………………………………….……………………..iii Table Of Contents……………………………………………………………………vi List Of Figures and Tables………………………………….………………………..x Acknowledgements…………………………….…………………………………….xv Published Abstracts and Articles...……………………………………..……….....xix Abbreviations………………………………………….………………………….....xxi
Chapter 1 ...... 1
Literature Review...... 1 1.1 Arthritis ...... 2 1.2 Osteoarthritis ...... 2 1.3 Characterization ...... 3 1.4 Clinical Presentation ...... 4 1.5 Diagnosis ...... 4 1.6 Synovial Joint Anatomy ...... 5 1.6.1 Bone ...... 5 1.6.2 Articular Cartilage ...... 8 1.6.3 Synovium ...... 10 1.6.4 Ligaments and Meniscus ...... 11 1.6.5 Muscle ...... 12 1.7 Treatment ...... 12 1.8 Risk Factors ...... 14 1.9 Biological Processes ...... 15 1.9.1 Endochondral Ossification ...... 15 1.9.2 Pro-inflammatory Cytokines ...... 18 1.9.3 Anti-inflammatory Cytokines ...... 20 1.9.4 Matrix Degrading Enzymes ...... 21 1.9.4.1 Matrix Metalloproteinases ...... 21 1.9.4.2 Aggrecanases...... 23 1.9.5 Growth Factors ...... 25 1.9.5.1 VEGF ...... 25 1.9.6 Biological Pathways Involved With OA ...... 25 1.9.6.1 TGF-β/BMP Pathway ...... 25 1.9.6.2 Wnt/β-catenin pathway ...... 28 1.9.6.3 RANKL/RANK/OPG pathway ...... 30 1.10 Genetic Research Models of OA ...... 34 1.10.1 Heritability ...... 34 1.10.2 Genome Wide Association Studies ...... 34
vi 1.10.3 Animal Models ...... 38 1.10.4 The Collaborative Cross ...... 39 1.11 Histological OA Grading Systems ...... 42 1.12 Summary...... 43
Chapter 2 ...... 45
Hypothesis and Aims ...... 45
Chapter 3 ...... 47
Materials and Methods ...... 47 3.1 Materials ...... 48 3.1.1 Equipment ...... 48 3.1.2 Chemical Reagents ...... 49 3.1.3 Buffers and Solutions ...... 51 3.1.4 Tissue Culture Materials and Reagents ...... 52 3.1.5 Oligonucleotide Primers and Housekeeping Genes Primers ...... 53 3.1.6 Software ...... 55 3.1.7 Gene Mapping Links ...... 55 3.1.8 Bioinformatics Tools ...... 56 3.2 Methods...... 56 3.2.1 Workflow ...... 56 3.2.2 Animal model ...... 56 3.2.3 Histological analysis of mice ...... 59 3.2.3.1 Staining ...... 60 3.2.3.2 OARSI Scoring ...... 61 3.2.3.3 Defining the OA phenotype ...... 64 3.2.3.4 Classification and ranking of phenotypes ...... 65 3.2.4 Haplotype reconstruction and QTL mapping ...... 67 3.2.5 Candidate gene verification ...... 70 3.2.6 Genotyping ...... 72 3.2.7 Primary Cell Isolation and Culture ...... 72 3.2.7.1 Tissue Isolation ...... 73 3.2.7.2 Osteoclast Culture ...... 73 3.2.7.3 Osteoblast Culture ...... 74 3.2.8 Molecular analysis ...... 75 3.2.8.1 RNA extraction and purification (using Trizol) ...... 75 3.2.8.2 Reverse transcription of extracted RNA ...... 76 3.2.8.3 QPCR analysis ...... 77 3.2.8.4 Real time PCR analysis ...... 78 3.2.9 Statistical analysis ...... 79 Chapter 4 ...... 80
Characterization of Osteoarthritic phenotype using OARSI histologic grading scale...... 80 4.1 Introduction ...... 81
vii 4.2 Screening and grading of OA phenotypes using the OARSI method ...... 82 4.3 Analysis of total joint OA in CC mice >8 months of age ...... 83 4.4 Analysis of most severe single surface OA phenotype in CC mice, >8 months of age ...... 86 4.5 Analysis of OA incidence in CC mice >8 months of age ...... 89 4.6 Analysis of total joint and single surface OA in CC mice >12 months of age.92 4.7 Discussion ...... 95
Chapter 5 ...... 99
Quantitative Trait Locus Haplotype Mapping of Osteoarthritis Phenotypes in the Collaborative Cross ...... 99 5.1 Introduction ...... 100 5.2 Results ...... 102 5.2.1 Median total joint OA in the >8 month cohort ...... 102 5.2.2 Median highest single surface measured OA in the >8 month cohort ...... 105 5.2.3 Median total joint OA in the >12 month cohort ...... 107 5.2.4 Median highest single surface measured OA in the >12 month cohort ...... 111 5.2.5 Median total joint score of middle aged 8-12 month cohort ...... 117 5.2.6 Incidence of any OA grade in the >8 month cohort ...... 121 5.3 Discussion ...... 126
Chapter 6 ...... 132
Characterization of Candidate Genes Associated With Osteoarthritis Development ...... 132 6.1 Introduction ...... 133 6.2 Results ...... 135 6.2.1 Identification of candidate genes implicated above the 95th percentile threshold ...... 135 6.2.2 Characterization of genome wide significant candidate genes...... 137 6.2.2.1 Potential novel candidates ...... 137 6.2.2.2 Candidate genes implicated in OA studies...... 140 6.2.2.3 Candidate genes implicated in studies of other forms of arthritis ...... 151 6.2.3 Identification of candidate genes implicated above the 37th percentile threshold ...... 163 6.2.4 Characterization of suggestive candidate genes ...... 165 6.2.4.1 Candidate genes implicated in OA studies...... 165 6.2.4.2 Candidate genes implicated in studies of other forms of arthritis ...... 178 6.3 Discussion ...... 192
Chapter 7 ...... 196
viii General Overview and Future Directions ...... 196 7.1 Overview of thesis and general discussion ...... 197 7.2 Future directions ...... 205
Chapter 8 ...... 209
References ...... 209
Appendix ...... 231
ix
LIST OF FIGURES AND TABLES
Figure 1.1 Diagram of healthy and osteoarthritic synovial joint 6
Figure 1.2 Diagram of endochondral ossification 17
Figure 1.3 TGF-β/BMP signalling pathways 27
Figure 1.4 Wnt signalling pathway 29
Figure 1.5 RANKL/RANK/OPG signalling pathway 33
Figure 1.6 The Collaborative Cross funnel breeding scheme 40
Figure 3.1 Workflow of the Collaborative Cross OA screening project 57
Table 3.1 Range of OARSI scores observed and represented in the
CC mice 62
Figure 3.2 Knee joint OA phenotype classifications 66
Figure 3.3 Example of Geneminer data normalization 69
Figure 4.1 Assessment of total knee joint and score calculation
example 84
Figure 4.2 Spread of OA phenotypes as measured via OARSI grading
of the total joint 85
Figure 4.3 Assessment of single surface knee joint and score
calculation example 87
Figure 4.4 Spread of OA phenotypes as measured via OARSI grading
of the highest scoring single joint surface 88
Figure 4.5 Percentage of any OA score measured in CC populations 90
Figure 4.6 Percentage of OA single joint score >3 measured in CC
populations 91
x Figure 4.7 Spread of OA phenotypes in >12 month population of CC
mice, as measured via OARSI grading of the total joint
score 93
Figure 4.8 Spread of OA phenotypes in >12 month population of CC
mice, as measured via OARSI grading of the highest single
surface score 94
Figure 5.1 QTL mapping of median total joint score >8 months. 103
Figure 5.2 QTL mapping of median total joint score >8 months
normalized. 104
Figure 5.3 QTL mapping of median single surface joint score >8
months. 106
Figure 5.4 QTL mapping of median total joint score >12 months for
chromosome 2. 108
Figure 5.5 QTL mapping of median total joint score >12 months for
chromosome 17. 109
Figure 5.6 QTL mapping of median total joint score >12 months
normalized. 110
Figure 5.7 QTL mapping of median single surface joint score >12
months for chromosome 17. 112
Figure 5.8 QTL mapping of median single surface joint score >12
months for chromosome 4. 113
Figure 5.9 QTL mapping of median single surface joint score >12
months for chromosome 4, normalized. 115
Figure 5.10 QTL mapping of median single surface joint score >12
months for chromosome 15, normalized. 116
xi Figure 5.11 QTL mapping of median total joint score 8-12 months. 118
Figure 5.12 QTL mapping of median total joint score 8-12 months
normalized. 119
Figure 5.13 QTL mapping of OA incidence in the >8 month cohort for
chromosome 11. 122
Figure 5.14 QTL mapping of OA incidence in the >8 month cohort for
chromosome 19. 123
Figure 5.15 QTL mapping of OA incidence with a score >3, in the >8
month cohort. 125
Table 5.1 Compiled list of genes implicated in all QTL maps 129
Table 6.1 Candidate genes implicated above the 95th percentile. 136
Figure 6.1 Gene expression profile of mouse Myo1f 138
Figure 6.2 Visual protein-protein interaction networks for MYO1F in
humans 139
Figure 6.3 Gene expression profile of mouse Col11a2 141
Figure 6.4 Visual protein-protein interaction networks for COL11A2
in humans 143
Figure 6.5 Gene expression profile of mouse Glis3. 144
Figure 6.6 Expression of Glis3 during chondrogenesis from rabbit
derived MSCs 145
Figure 6.7 Visual protein-protein interaction networks for GLIS3 in
humans 147
Figure 6.8 Gene expression profile of mouse Il-33. 148
Figure 6.9 Visual protein-protein interaction networks for IL-33 in
humans 150
xii Figure 6.10 Gene expression profile of mouse Daxx 152
Figure 6.11 Visual protein-protein interaction networks for DAXX in
humans 153
Figure 6.12 Gene expression profile of mouse Pdcd1lg2 154
Figure 6.13 Visual protein-protein interaction networks for
PDCD1LG2 in humans 156
Figure 6.14 Gene expression profile of mouse Cd274 158
Figure 6.15 Visual protein-protein interaction networks for CD274 in
humans 159
Figure 6.16 Gene expression profile of mouse Tap2 161
Figure 6.17 Visual protein-protein interaction networks for TAP2 in
humans 162
Table 6.2 Candidate genes implicated above the 37th percentile. 164
Figure 6.18 Gene expression profile of mouse Col9a3 166
Figure 6.19 Visual protein-protein interaction networks for COL9A3 in
humans 167
Figure 6.20 Gene expression profile of mouse Epas1 169
Figure 6.21 Visual protein-protein interaction networks for EPAS1 in
humans 170
Figure 6.22 Gene expression profile of mouse Jak2 172
Figure 6.23 Visual protein-protein interaction networks for JAK2 in
humans 173
Figure 6.24 Gene expression profile of mouse Dkk1 176
Figure 6.25 Visual protein-protein interaction networks for DKK1 in
humans 177
xiii Figure 6.26 Gene expression profile of mouse Tab1 179
Figure 6.27 Visual protein-protein interaction networks for TAB1 in
humans 180
Figure 6.28 Gene expression profile of mouse Obscn 182
Figure 6.29 Visual protein-protein interaction networks for OBSCN in
humans 183
Figure 6.30 Gene expression profile of mouse Cdh13 185
Figure 6.31 Visual protein-protein interaction networks for CDH13 in
humans 186
Figure 6.32 Gene expression profile of mouse Cdh4 188
Figure 6.33 Visual protein-protein interaction networks for CDH4 in
humans 189
Figure 6.34 Gene expression of murine Cdh4 in osteoclasts and
osteoblasts 191
Figure 7.1 Candidate gene annotation of an osteoarthritic joint 204
Supplementary Strain sample sizes
Table 9.1 232
Supplementary QTL mapping of single surface score >4, preliminary
Figure 9.1 cohort 234
Supplementary QTL mapping of single surface score, preliminary cohort
Figure 9.2 235
Supplementary Expression of Zfhx4 during chondrogenesis from rabbit
Figure 9.3 derived MSCs 236
xiv ACKNOWLEDGEMENTS
I would like to acknowledge and thank my two supervisors W/Prof Jiake Xu and Dr
Jennifer Tickner, for their encouragement and ongoing support on my journey from research assistant to PhD student. This adventure of self and scientific discovery would not have been possible without the both of them. Jiake’s optimistic and energetic attitude towards our field of study was the driving force to constantly aim higher and make the best of every opportunity. His infectious passion and faith in my abilities was a constant motivator during my time spent as his staff member and his student. Without Jiake, my academic goals and PhD candidature would not have been possible.
A special thanks to Dr Jennifer Tickner, for treating me as an equal and as a friend during my candidature. She is truly the backbone of the laboratory group and there are no words to truly describe the amazing effort, encouragement and support she provides to each and every student and visiting scholar that passes through our laboratory. Her expertise in and out of the laboratory and her ability to put results into context were without a doubt the reason I am able to complete the writing of this thesis and all of the work within it.
I would like to acknowledge that this research was supported by an Australian
Government Research Training Program (RTP) Scholarship.
I would like to acknowledge and thank Prof Grant Morahan and his support team Ramesh
Ram and Sylvia Young at Geniad, for providing us with the CC mice and the supporting
Geneminer software. Managing this enormous cohort and the even larger data collection is no simple exercise, I am very grateful for Grant’s contributions to my mapping approaches and Sylvia and Ramesh’s assistance with the maintenance of the software.
xv
I would like to acknowledge and thank Prof Richard Allcock, an inspiring mentor, brilliant scientist and good friend. His vast knowledge of genetic analyses and experience in academia were invaluable to me throughout my time at UWA as both staff and student.
I would also like to thank Nina Kresoje and Dr Kyle Yau for their ongoing support in the laboratory and their friendship.
I would like to thank Prof Kimberley Strong for being a fantastic mentor since my arrival at the university and for her ongoing support and friendship. She is an inspiration and the reason I set out to challenge myself with this candidature. I would also like to thank Prof
Clayton Fragall for his encouragement and sharing his experience with me over the years.
I would also like to thank W/Prof Wendy Erber for her encouragement over the years and her support before and during my candidature.
I would like to acknowledge William Tomlinson for his efforts in validating the OARSI scores and joining me in the lab for the later parts of this work as part of his medical scholarly activity.
I would like to acknowledge Dr Peter Robbins of PathWest for his generous assistance in validating my identification of OA phenotypes in my first year of study.
I would like to thank my fellow lab group students, past and present: Benjamin Ng, Baysie
Lim, Chao Wang, Colin Wang, Dian Teguh, Gaurav Jadhav, Heng Qiu, Jinbo Yuan, Kai
Chen and Vincent Kuek. Jinbo’s assistance with preparing the mice was extremely valuable along with his advice on the CC project that we both used to conduct our
xvi research. Dian’s assistance with cell culture was greatly appreciated and her friendship and mentorship over the years will not be forgotten. Kai and Colin’s assistance with extractions and qPCR protocols was most helpful.
I would like to thank my fellow post graduate students and office mates: Han Ng Leng,
Olivia Ruhen, Britt Clynick, Jarrad Hall, Nicole Bzdyl, Josh Clayton, Bob Mirzai, Henry
Hui and Thaddaeus Teo. I am so fortunate to have such a fantastic network of people to spend my days with and to lean on when times are tough. A special thanks to Han for allowing me to use his computer on occasion and his expertise in qPCR.
A special thanks to Prof Kristen Nowak for her advice and feedback on my interpretation of the QTL mapping, Prof Nathan Pavlos for his mentorship and Prof Fiona Pixley for offering me a quiet space to complete my writing.
Thanks to the staff at the Animal Resource Centre for their support with the maintenance of the CC mouse colonies.
A special thanks to my co-workers, the UWA biomedical sciences teaching team for their ongoing support. Your understanding and patience during my candidature was greatly appreciated and will not be forgotten.
A special thanks to all of my students, who offered their help over the duration of my candidature no matter how big or small, your contribution is appreciated.
xvii To my partner Gemma, it because of this experience that I was able to have met you and it is because of you that I am able to see it through to the end. Your strength is inspiring and your love and support is what drives me to succeed.
To my family, my Dad Ian, my Mum Margaret and my sister Jessica, for your ongoing love, support, and patience, and for accepting me back into your home during my candidature. This accomplishment is a testament to your faith in me, and the opportunities you presented me throughout my life. For that I thank you.
xviii PUBLISHED ABSTRACTS AND ARTICLES
ABSTRACTS
1. Kenny J, Ram R, Morahan G, Tickner J, Xu J (2016) “Identification of genes regulating osteoarthritis development using a mouse phenotype library” Australian Scientific Medical Research Symposium, Perth, Western Australia (Oral presentation).
2. Kenny J, Ram R, Morahan G, Tickner J, Xu J (2016) “Identification of genes regulating osteoarthritis development using a mouse phenotype library” Australian and New Zealand Bone and Mineral Society, Annual scientific meeting, Gold Coast, Queensland, Australia (Poster presentation).
3. Kenny J, Ram R, Morahan G, Tickner J, Xu J (2017) “Identification of genes regulating osteoarthritis development using a mouse phenotype library” 1st Australia-China Conference of Science, Technology and Innovation (Oral presentation).
4. Kenny J, Ram R, Morahan G, Tickner J, Xu J (2017) “Identification of genes regulating osteoarthritis development using a mouse phenotype library” The Australian and New Zealand Orthopaedic Research Society 23rd annual conference, Adelaide, South Australia (Oral presentation).
5. Kenny J, Ram R, Morahan G, Tickner J, Xu J (2017) “Identification of genes regulating osteoarthritis development using a mouse phenotype library” Combined Biomedical Sciences Meeting 27th annual meeting, Perth, Western Australia (Oral presentation and poster presentation).
xix ARTICLES
1. Jadhav, G., Teguh, D., Kenny, J., Tickner, J. and Xu, J., 2016. Morc3 mutant mice exhibit reduced cortical area and thickness, accompanied by altered haematopoietic stem cells niche and bone cell differentiation. Scientific reports, 6, p.25964.
2. Boutilier, J.K., Taylor, R.L., Mann, T., McNamara, E., Hoffman, G.J., Kenny, J., Dilley, R.J., Henry, P., Morahan, G., Laing, N.G. and Nowak, K.J., 2017. Gene Expression Networks in the Murine Pulmonary Myocardium Provide Insight into the Pathobiology of Atrial Fibrillation. G3: Genes, Genomes, Genetics, 7(9), pp.2999-3017.
3. Tan, Z., Cheng, J., Liu, Q., Zhou, L., Kenny, J., Wang, T., Lin, X., Yuan, J., Quinn, J.M., Tickner, J. and Hong, G., 2017. Neohesperidin suppresses osteoclast differentiation, bone resorption and ovariectomised-induced osteoporosis in mice. Molecular and cellular endocrinology, 439, pp.369-378.
xx ABREVIATIONS
ACL Anterior Cruciate Ligament
ACR American College of Rheumatology
A disintegrin and metalloproteinase with thrombospondin motif 4
ADAMTS4 (aggrecanase 1)
A disintegrin and met alloproteinase with thrombospondin motif 5
ADAMTS5 (aggrecanase 2)
ALK Activin receptor-like kinase
ALP Alkaline phosphotase
AP-1 Activator protein 1
AS Ankylosing spondylitis
ATP Adenosine triphosphate
BGLAP Bone gamma-carboxyglutamic acid-containing protein (osteocalcin)
BMI Body mass index
BMP Bone morphogenic protein
CALM2 Calmodulin 2
CC Collaborative cross
CDH13 Cadherin 13
CDH4 Cadherin 4
cDNA Complementary Deoxyribonucleic acid
COL11A2 Collagen XI alpha 2 chain
Col2A1 Collagen II alpha chain 1
COL9A3 Collagen IX alpha 3 chain
CREB cAMP-response element-binding protein
CTSK Cathepsin K
xxi DAXX Death domain associated protein
DKK Dickkopf protein
DMEM Dulbecco's modified eagle medium
DMM Destabilisation of the medial meniscus
DPX Dibutylphthalate Polystyrene Xylene
ECM Extracellular matrix
EDTA Ethylenediaminetetraacetic acid
ELISA Enzyme linked immunoabsorbant assay
EPAS1 Endothelial PAS domain protein 1
ERK Extracellular signal-regulated kinase
FBS Fetal bovine serum
GDFs Growth differentiation factors
GLIS3 GLIS family zinc finger 3
GWAS Genome wide association studies
GWS Genome wide significance
H&E Haymatoxylin and eosin
HBSS Hanks' balanced salt solution
HPRT Hypoxanthine-guanine phosphoribosyltransferase
IFNγ Inteferon gamma
IHH Indian hedgehog
IL-1 Interleukin 1
IL-10 Interleukin 10
IL-10R Interleukin 10 receptor
IL-13 Interleukin 13
IL-15 Interleukin 15
xxii IL-17 Interleukin 17
IL-18 Interleukin 18
IL-18Ra Interleukin 18 receptor antagonist
IL-1Ra Interleukin 1 receptor antagonist
IL-1α Interleukin 1 alpha
IL-1β Interleukin 1 beta
IL-21 Interleukin 21
IL-33 Interleukin 33
IL-36Ra Interleukin 36 receptor antagonist
IL-36α Interleukin 36 aplha
IL-36γ Interleukin 36 gamma
IL-37 Interleukin 37
IL-38 Interleukin 38
IL-4 Interleukin 4
IL-4Ra Interleukin 4 receptor antagonist
IL-6 Interleukin 6
IL-8 Interleukin 8 iNOS Inducible Nitric Oxide Synthase
JAK2 Janus kinase 2
JIA Juvenile idopathic arthritis
JNK c-Jun N-terminal kinase
KOA Knee Osteoarthritis
LFC Lateral femoral condyle
LOD Log of the odds ratio
LPS Lipopolysaccharide
xxiii LRP5 Low density lipoprotein receptor related protein 5
LRP6 Low density lipoprotein receptor related protein 6
LTP Lateral tibial plateau
MAPK Mitogen activated protein kinase
MFC Medial femoral condyle
MITF Microphthalmia-associated transcription factor
MKK Map kinase kinase
MMP Matrix Metalloproteinase
MMP-1 Matrix Metalloproteinase 1 (Collagenase)
MMP-13 Matrix Metalloproteinase 13 (Collagenase 3)
MMP-3 Matrix Metalloproteinase 3 (Stromelysin 1)
MRI Magnetic Resonance Imaging mRNA Messenger RNA
MSC Mesenchymal Stem Cell
MTP Medial tibial plateau
MUGA Mouse universal genotyping array
MYO1F Myosin IF
NCBI National center for biotechnology information
NF-κB Nuclear factor kappa B
NFATc1 Nuclear factor of activated T cells cytoplasmic 1
NSAIDs Non-steroidal anti-inflammatory drugs
OA Osteoarthritis
OARSI Osteoarthritis research society international
OBSCN Obscurin
OPG Osteoprotegrin
xxiv PBS Phosphate buffered saline
PCL Posterior cruciate ligament
PCR Polymerase chain reaction
PDCD1LG2 Programmed cell death ligand 2
PKC protein kinase C
PPI Protein-protein interaction
QPCR quantitative polymerase chain reaction
QTL Quantitative trait locus
RA Rheumatoid Arthritis
RANK Receptor activator of neuclear factor kappa-B
RANKL Receptor activator of neuclear factor kappa-B ligand
RI Recombinant inbred
RNA Ribonucleic acid
SD Standard deviation
SMAD1 Mothers against decapentaplegic homolog 1
SMAD5 Mothers against decapentaplegic homolog 5
SMAD8 Mothers against decapentaplegic homolog 9
SMURF SMAD ubiquitination regulatory factor
SNP Single Nucleotide Polymorphism
STAT Signal transducer and activator of transcription
TAB1 TGF-beta activated kinase 1 (MAP3K7) binding protein 1
TAE Tris base, acetic acid, EDTA buffer
TAK1 Transforming growth factor beta-activated kinase 1
TAP2 Transporter 2, ATP binding cassette subfamily B member
Tg Transgenic
xxv TGF-β Transforming growth factor beta
TIMP Tissue inhibitor of metalloproteinase
TKR Total Knee Replacement
TNF-α Tumor Necrosis Factor Alpha
TRAF TNF receptor-associated factor
TRAP Tartrate resistant acid phosphatase
TYK2 Tyrosine kinase 2
VEGF Vascular Endothelial Growth Factor
VNTR Variable Number Tandem Repeat
ZFHX4 Zinc finger homeobox 4
α-MEM Minimal essential medium alpha
xxvi Chapter 1 Literature Review
1
1.1 Arthritis
Arthritis diseases affect 3.85 million people in Australia, with an estimated economic burden of $23.9 billion each year in the form of medical care and loss of production and earnings. With the increase in life expectancy and greater numbers of aging individuals in the Australian population, current trends suggest the number of affected Australians will increase to 7 million by 2050 (Economics Access 2007).
Nearly 2.4 million sufferers of arthritis in Australia are considered to be of working age
(Australian Bureau of Statistics [ABS] 2015). While some symptoms of arthritis can be managed, there is no disease-modifying cure for the disease. Due to the progressive nature of arthritis, early detection and intervention can benefit sufferers and delay the early onset of the disease, maximising productivity and quality of life (Roos and Arden 2015).
There are over 100 diseases of joints classified as forms of arthritis. In Australia, osteoarthritis, rheumatoid arthritis and gout account for 95% of arthritis cases (ABS
2015).
1.2 Osteoarthritis
Osteoarthritis (OA) is the most common degenerative joint disorder that affects mainly load bearing joints such as hips and knees (Litwic, Edwards et al. 2013). OA is associated with aging but is not considered to be a natural component of aging (Felson, Lawrence et al. 2000). The Australian bureau of statistics states that in 2014-15, 3.5 million people, equivalent to 15.3% of the Australian population had arthritis with a higher prevalence in women. Of this population, 58.9% were classified specifically as OA sufferers (ABS
2 2015). While OA affects all age groups, cases become more prevalent after 50 years of age in men and 40 in women (Breedveld 2004; Cross, Smith et al. 2014).
1.3 Characterization
OA affects the entire synovial joint. It is characterized by pain, inflammation and stiffness of the joint resulting in limited mobility (Hutton 1989). The disease affects joints that are frequently exposed to mechanical stress, most commonly hands, hips and knees (Loeser,
Goldring et al. 2012). OA is the result of cartilage degradation, leading to fibrillation, erosion and cracking in the superficial layer. As severity of degradation increases over time, the pathological process of OA penetrates the deeper layers of cartilage (Litwic,
Edwards et al. 2013; Goldring and Goldring 2010; Buckwalter, Mankin et al. 2005).
A pathological indicator of OA progression is re-establishment of the endochondral ossification process. This is a process of cartilage becoming mineralised to form solid bone, which is observed during embryonic skeletal development (Mackie, Ahmed et al.
2008; Burr 2004). Endochondral ossification can be detected histologically by the presence of many tidemarks. Tidemarks are the result of progressive thickening of calcified cartilage, which is more highly mineralised than bone. As the thickening progresses, more tidemarks are observed, which signifies tidemark advancement and the advancement of mineralised cartilage. This thickening results in a reduction of overlaying articular cartilage. This layer is not very dynamic and is not capable of making enough new cartilage to maintain its volume (Burr and Gallant 2012).
3 1.4 Clinical Presentation
Patients suffering from suspected OA present clinically with pain in the affected joint associated with limited swelling and a limited range of motion and or stiffness. In cases of knee osteoarthritis (KOA), patients will often present with an adjusted gait favouring a less affected limb or avoiding load bearing of the joint in question. Patients will express a higher pain response when the knee is under load and will dictate a grinding sensation when attempting flexion/extension. In severe cases, there will be a visible valgus or varus change in the alignment of the tibia in relation to the femur (Gok, Ergin et al. 2002).
1.5 Diagnosis
Diagnosis of KOA can be defined as clinical, radiological or subjective. Clinical KOA is diagnosed by patient history and presentation of pain, joint stiffness and bony enlargement. Classification of KOA by this method is made in accordance with the
American College of Rheumatology (ACR) criteria (Altman, Asch et al. 1986).
Radiological diagnosis is confirmed via X-ray and patients are graded using the Kelgren-
Lawrence grading scale to track their progression on a 0-4 scale. KOA phenotypes are categorised by visible joint space loss, osteophyte formation, bone sclerosis and cyst formation (Kellgren J 1963). Patients who are diagnosed with KOA with a Kelgren-
Lawrence grade of 2-3 will be monitored over a period of 8 years for signs of progression.
Patients who reach a grade of 4 within 8 years will be candidates for total knee arthroplasty. MRI is often used for a more refined evaluation of changes in bone and cartilage (Litwic, Edwards et al. 2013). Subjective KOA is defined by the patient’s personal assessment of condition, ultimately by presence of pain and joint stiffness. There is a weak correlation between subjectively defined and radiological KOA there is a clear heterogeneity in presentation. For example, some patients presenting with pain may have
4 no radiographic changes, whilst patients shown to have radiographic KOA may not experience no pain, or mobility symptoms (Hannan, Felson et al. 2000). This indirect correlation further supports the concept of KOA being a disease exhibiting multiple symptoms manifested in multiple tissues that make up complex synovial joints.
Therefore, this disease cannot be categorised simply by bony changes and pain alone.
1.6 Synovial Joint Anatomy
The synovial joint is comprised of multiple tissues. Each tissue contributes structurally to the joint as well as interacting with each other to maintain the function and maintain proper joint homeostasis. The tissues that make up the internal synovial joint are cartilage, calcified cartilage, bone, synovium and ligament. External tissues such as muscles and tendons also contribute to joint homeostasis by maintaining normal joint biomechanics such as alignment, proper articulation of surfaces and range of motion (Herzog, Longino et al. 2003). Maintenance of joint homeostasis is dependent on the integrity of all of the involved tissues. Degradation of any one of the involved tissues will have a degenerative effect on the others, ultimately resulting in the failure of the joint as a whole (Figure 1.1)
(Burr 2004).
1.6.1 Bone
The skeleton is an organ that is both rigid and dynamic. Bone is specifically structured to be rigid and capable of supporting load while providing leverage for locomotion. Bones are constantly undergoing remodelling, a process where they are shaped, reshaped and repaired (Boyle, Simonet et al. 2003). As well as being rigid, bone is required to be lightweight and flexible to enable it to absorb energy under load without fracturing. It is composed of type-I collagen, enabling elasticity with the addition of calcium hydroxyapatite mineral deposits within the collagen matrix to maintain the rigidity of the
5 Figure 1.1 Diagram of healthy and osteoarthritic synovial joint
The functional synovial joint is comprised of many different tissues and cell types interfacing with each other to perform both independent and collaborative functions to maintain joint homeostasis. Degradation or altered function of each of the tissues and cell types involved can contribute to the osteoarthritic phenotype. Figure information sourced from (Herzog, Longino et al. 2003; Goldring and Goldring 2010;
Buckwalter, Mankin et al. 2005).
6 tissue. This mixture of organic and inorganic compounds ensure that the structure is capable of sustaining impact whilst maintaining its structure (Seeman and Delmas 2006).
The formation of long bones occurs via a process called endochondral ossification. This process involves the formation of a cartilage scaffold, which is then replaced by mineralised bone. Comparatively, the formation of flat bones occurs without the cartilage scaffold, instead forming by the condensing of mesenchymal stem cells. This process is called intramembranous ossification. Maintenance of the structural integrity and mineral homeostasis of bone is managed via bone remodelling. The skeleton is constantly being remodelled in response to bone damage via mechanical loading, or for the removal of unwanted bone. Due to this process occurring asynchronously throughout the skeleton, locally generated and regulated factors are considered to play an important role in facilitating the communication between the participating cells (Henriksen, Karsdal et al.
2014). The removal or resorption of mineralised bone is conducted and regulated by osteoclasts, while the formation of new bone is conducted and regulated by osteoblasts.
Osteoclasts take approximately three weeks to complete a phase of resorption. The osteoblast facilitated bone formation phase takes approximately three to four months.
Approximately 10% of the human skeleton is regenerated yearly, with a complete mature skeleton renewed in 10 years (Sims and Martin 2014). The ability to remodel the rigid skeletal structure enables effective healing of the damaged mineralised tissue. However, the cartilage surfaces of adjoining long bones, which remain non-mineralised at the end of the endochondral ossification process, are not remodelled to re-establish normal tissue architecture.
7
1.6.2 Articular Cartilage
Articular cartilage is a smooth shock-absorbing layer lining subchondral bone in synovial joints. This layer is capable of withstanding repetitive load whilst enabling frictionless movement to facilitate normal joint operation (Buckwalter, Mankin et al. 2005). Articular cartilage can be divided into four zones, the superficial, transitional, deep and calcified or mineralised zone. These zones are classified by chondrocyte morphology and the distribution of collagen, in particular type II collagen (Schiller 1995; Cole and Gomoll
2009). The superficial zone is the surface layer that comes in contact with the adjoining bone’s cartilage surface and is designed to offer a frictionless surface and withstand gliding forces. It’s characterised by the elongated morphology of the chondrocytes and the collagen fibres running parallel to the surface. This zone has the highest density of collagen and fibronectin while containing the lowest amount of proteoglycans. The superficial zone also has the greatest tensile stiffness of all the zones and due to it’s density, acts as a barrier for larger molecules attempting to access the deeper layers
(Aydelotte and Kuettner 1988). These properties suggest the superficial layer plays a role in isolating articular cartilage from the immune system; therefore irreparable damage to this layer may expose and release molecules that may facilitate an immune or inflammatory response (Buckwalter, Mankin et al. 2005).
The transitional zone is the intermediate space between the superficial and deep cartilage zones. Chondrocytes within the transitional zone are spheroidal and contain higher concentrations of synthetic organelles, golgi membranes and endoplasmic reticulum when compared with superficial zone chondrocytes (Buckwalter, Mankin et al. 2005).
Though the transitional zone contains a lower collagen density when compared with the
8 superficial zone, the collagen fibrils have a larger diameter and a higher concentration of proteoglycan is present in the matrix (Buckwalter, Mankin et al. 2005).
The middle or deep zone contains spheroid chondrocytes arranged in columns perpendicular to the cartilage surface. This zone contains the largest diameter collagen fibrils and the highest proteoglycan concentration. The collagen fibrils extend perpendicular to the cartilage surface and pass into the tidemark or cartilage calcification front (Buckwalter, Mankin et al. 2005). The orientation of the fibrils and arrangement of chondrocytes, along with the proteoglycan content in the deeper zones suggest that these layers are structured to cope with compressive and tensile forces (Buckwalter, Mankin et al. 2005).
The calcified zone is the deepest zone of articular cartilage, which separates the middle zone and the subchondral bone. Chondrocytes in this zone have less volume and contain small amounts of endoplasmic reticulum and golgi membranes, suggesting that the cells within the calcified layer have a low metabolic activity. It is this zone where the end of endochondral ossification is defined and the tidemark of mineralisation is observed
(Buckwalter, Mankin et al. 2005).
The majority of articular cartilage is comprised of extracellular matrix with embedded chondrocytes accounting for less than 10% of the total cartilage tissue (Walker 1996).
The structure of the extracellular component of articular cartilage is a fibrillar collagen network that contains both small and large aggregating proteoglycans, aggrecan and other molecules in conjunction with embedded chondrocytes. These collagen fibres are made up of type II collagen fibrils that internally contain type XI collagen. Type IX collagen is
9 integrated into the surface of the fibril to facilitate proteoglycan retention and association with other articular cartilage matrix components. The pericellular matrix of the embedded chondrocyte contains type VI collagen microfibrils (Goldring and Goldring 2010).
Superficial zone chondrocytes express proteoglycan-4 or otherwise known as lubricin, an essential component of lubrication of the articular boundaries. Deep zone chondrocytes primary function is the production and maintenance of extracellular components such as aggrecan and collagen II. These components are essential to enable the articular cartilage to cope with biomechanical stresses. Chondrocytes situated at the bottom of the deep zone are larger, though they do not differ in function from other deep zone chondrocytes
(Broom and Poole 1982).
1.6.3 Synovium
Synovium is a soft tissue lining that occupies the space of diarthrodial joints. It is comprised of a surface layer (intima) containing macrophages and fibroblasts and an underlying layer (subintima) containing blood vessels, lymphatic vessels and fibroblasts in a collagen based extracellular matrix. Synovial fluid, which is rich in hyaluronan, is present between these two layers. These structures operate in conjunction to provide a non-adherent surface between articular cartilage surfaces. The synovial tissue contains a rich nerve supply around the vascular networks and extends into intimal layers. Non- pathogenic synovium has been shown to produce Interleukin-1 (IL-1), Interleukin-6 (IL-
6) and Tumor Necrosis Factor alpha (TNF-α) in minimal quantities. However, larger volumes of these cytokines are detected in the synovium of joints affected by inflammatory arthritis (Smith, Barg et al. 2003). Larger quantities of IL-1 receptor antagonist are present in normal synovium for the suppression of inflammatory processes.
10 There are also low levels of Osteoprotegrin (OPG) and considerably lower levels of
Receptor Activator of Nuclear factor-Kappa B (RANKL) in synovium, to prevent the formation of osteoclasts and to maintain normal joint homeostasis (Smith 2011). The avascular nature of articular cartilage requires embedded chondrocytes to obtain nutrients predominantly via diffusion from the synovium which is facilitated by joint loading
(Levick and McDonald 1995; Buckwalter, Mankin et al. 2005).
1.6.4 Ligaments and Meniscus
The ligaments are comprised of fibrous connective tissue inserted into the non-articular regions of the synovial joint. Ligaments are responsible for maintaining the attachment of the bones while providing support to resist migration of the adjoining bones during extension, flexion and/or rotation (Flandry and Hommel 2011). Ligament failure or laxity due to injury or disease results in a failure to maintain normal joint biomechanics both passively and under load. This failure can cause an increase in load and progressive wear of articular cartilage, initiation of inflammation within the affected joint and ultimately initiate osteoarthritic changes (Halewood and Amis 2015). The menisci are responsible for increasing congruency of the tibiofemoral joint along with limiting tibiofemoral rotation and translation (Halewood and Amis 2015). Meniscal damage is a significant risk factor for OA development (Englund, Guermazi et al. 2009), and the rate of OA progression in animal models of post partial or total menisectomy is higher than those with anterior cruciate ligament (ACL) injury (Karahan, Kincaid et al. 2001). This suggests that the meniscus serves a role not only as an articular surface locator, but also in load distribution (McDermott and Amis 2006).
11 1.6.5 Muscle
Quadriceps muscles are important for maintaining proper gait and joint kinematics during locomotion. Quadriceps strength was found to be 20% reduced in radiographically diagnosed OA patients. This weakness may be present in OA patients who have no knee pain or muscle atrophy, suggesting muscle dysfunction as a cause for the measured weakness (Slemenda, Brandt et al. 1997). Instability of knee joints due to ligament injury will also contribute to an avoidance gait, which further exacerbates muscle weakness.
This weakness can lead to gradual joint degeneration and OA, as eccentric quadriceps contractions are responsible for controlling knee flexion under load. This can interfere with the ability of the joint to attenuate shock during the transfer of weight to the affected limb (Lewek, Rudolph et al. 2002).
1.7 Treatment
Treatment of OA is confined to the management of symptoms and ultimately the replacement of articular cartilage with a prosthesis. There is currently no standard treatment available to reverse the progression of the disease. Patients who progress slowly can manage their KOA symptoms with a combination of pain management via Non- steroidal anti-inflammatory drugs (NSAIDs) and physical therapy (Adams, Band-Entrup et al. 2013; Pendleton, Arden et al. 2000). As the symptoms progress, patients can increase their dosage or seek anti-inflammatory (Pelletier, Martel-Pelletier et al. 2001) or prescription opioid pain medication (Gore, Tai et al. 2012). The negative implications of long-term opioid pain medication may outweigh the benefits as the disease progresses
(Rolita, Spegman et al. 2013). In these cases, along with cases of severely impacted mobility, surgical intervention is required.
12 Total knee arthroplasty or replacement (TKR) surgery is the current standard for treatment of severe, end stage KOA. The surgery involves excision of the tibial plateau, femoral condyles and patellarfemoral cartilage surfaces and both the anterior and posterior cruciate ligaments. The surfaces are replaced with titanium implants and polyethylene bearings, which are affixed to the bones using bone cement and pins or osteo-conductive implant surfaces including hydroxyapatite coatings (Voigt and Mosier
2011). The pros and cons of these techniques have been studied at length as stated in this review by (Aprato, Risitano et al. 2016), ultimately concluding that both of these techniques can significantly reduce pain levels and restore mobility in patients. However, both of these techniques suffer from the same shortcomings being that the implants that approach 10 years in situ increasingly require revision surgery at rates between 3-13% depending on implant due to pain, infection, polyethylene wear and aseptic loosening
(Australian Orthopaedic Association National Joint Replacement Registry annual report
2017; Roberts, Esler et al. 2007).
When considering the groups of OA sufferers that are potential TKR patients of working age (<55 years), the long-term satisfaction of TKR surgery is a significant concern. This cohort of patients also experience greater complications post TKR revision surgery
(Stambough, Clohisy et al. 2014). This is of great significance considering the increase in life expectancy and patients “outliving” their primary TKR and requiring revision surgery. This highlights the importance of early identification and intervention of potential OA sufferers, with the interest of prolonging healthy joint function prior to primary TKR.
13 1.8 Risk Factors
When assessing human TKR incidence, the risk of KOA increases with age, suggesting that increasing age is the most significant contributor to the development of KOA
(Blanco, Möller et al. 2015; Felson, Zhang et al. 1995). Although severe cases are mainly associated with older age, any age group is at risk of developing KOA. Post-traumatic cases of KOA are induced via high impact, compression and destabilisation of the knee joint, resulting in articular cartilage lesions. These lesions heal over time but are replaced with fibrocartilage, which has poor mechanical loading properties and is prone to breaking down and reproducing the injury. Patients with history of ACL rupture are of increased risk due to the impact lesions that occur on articular cartilage and other comorbidities at the time of the injury (Lohmander, Englund et al. 2007). Similar lesions are prone to development after meniscal tears that progressively damage articular cartilage over time (Englund, Guermazi et al. 2009). While these tears can sometimes be repaired, a partial or total menisectomy is often the only treatment. The consequence of this treatment also leaves patients with a considerably higher risk of developing OA
(McDermott and Amis 2006; Roos, Lauren et al. 1998).
Arthrofibrosis is a formation of scar tissue within a synovial joint, which can form after knee surgeries. Formation of this scar tissue also puts patients at higher risk of KOA development due to the limited range of motion and associated limited activity (Mayr,
Weig et al. 2004). As a result of cartilage surface changes or pain adjusting gait, malalignment of the joint is itself a risk factor for KOA development as discussed in the review by (Tanamas, Hanna et al. 2009). This malalignment may be caused by impaired muscle function, particularly in the quadriceps. It has been demonstrated that there is a significant association between muscle strength and proprioception and an increase in
14 KOA risk (Slemenda, Brandt et al. 1997). This weakness may be pre-existing or developed after ACL rupture (Lewek, Rudolph et al. 2002).
A high body mass index and obesity has been demonstrated as a high risk factor for the development of KOA as a result of additional mechanical stress on knee joints (Breedveld
2004; Grotle, Hagen et al. 2008). The suggestion of mechanical stress alone as a contributor to the development of KOA was demonstrated by (Muraki, Akune et al. 2009) where occupations involving heavy lifting can contribute to an increased risk of KOA development. Spontaneous forms of KOA however, may not be associated with any immediate event and may be the result of abnormalities that progressively influences the damage of articular cartilage, such as inherited genetic variance.
1.9 Biological Processes
1.9.1 Endochondral Ossification
Endochondral ossification is a process by which cartilage is formed and is gradually replaced by mineralised bone to form the skeleton. This process begins with mesenchymal stem cells (MSCs), which undergo condensation and differentiate into chondrocytes. The chondrocytes secrete an extracellular matrix that begins to form a skeletal template of hyaline cartilage. This template grows interstitially and the cells located in the centre of the perichondrium differentiate into osteoblasts, resulting in mineralisation of the tissue to form a bone collar that ultimately becomes cortical bone
(Shapiro 2008). During this process, the chondrocytes within the cartilage template continue to proliferate and undergo hypertrophy as they mature. The hypertrophic chondrocytes secrete a collagen X matrix, Vascular Endothelial Growth Factor (VEGF) to induce angiogenesis, and alkaline phosphatase, the result of this process is the
15 mineralisation of the tissue (Mackie, Tatarczuch et al. 2011). To facilitate the invasion of blood vessels into the cartilage template, matrix metalloproteinases (MMPs) are produced to degrade the cartilage matrix. The vessels allow access for osteoclasts and osteoprogenitor cells to begin matrix remodelling and ossification; this area is called the primary ossification centre (Mackie, Ahmed et al. 2008). Within the primary ossification centre, osteoclasts begin to resorb the cartilage template while osteoblasts lay down a woven bone matrix. A secondary ossification centre is also formed in the epiphyses, resulting in a stable layer of hyaline cartilage at the joint surface and a section of cartilage between the primary and secondary ossification centres, referred to as a growth plate.
This secondary ossification centre as it matures is then referred to as the subchondral bone region. This growth plate is responsible for the continual elongation of the bone as it continues to grow until the skeleton is fully matured, in which case the growth place closes and is replaced entirely with mineralised bone (Figure 1.2) (Mackie, Tatarczuch et al. 2011).
16
Figure 1.2 Diagram of endochondral ossification
Endochondral ossification is initiated by the condensation and proliferation of MSCs, which undergo differentiation into chondrocytes. The chondrocytes secrete extracellular matrix and form a cartilage template. VEGF expression induces angiogenesis, which initiates the mineralisation of the template and begins forming bone in the mid shaft primary ossification centre. A secondary ossification centre forms in the epiphyses and forms the subchondral bone and hyaline cartilage regions Figure adapted from (Shapiro
2008).
17 1.9.2 Pro-inflammatory Cytokines
The interleukin family of cytokines consists of 11 members, 7 of which are pro- inflammatory agonists (IL-1α, IL-1β, IL-18, IL-33, IL-36α, IL-36β and IL-36γ) and 4 antagonists (IL-1Ra, IL-36Ra, IL-37 and IL-38). IL-1 cytokines induce inflammatory responses with their activity tightly regulated at the production, processing and maturation, receptor binding and post receptor signalling stages (Palomo, Dietrich et al.
2015). It has become increasingly evident that the inflammatory cytokine network is substantially involved in the pathogenesis of OA. The most significant group of inflammatory cytokines involved in this process has been identified as IL-1β, TNFα, IL-
6, IL-15, IL-17 and IL-18 and anti-inflammatory cytokines IL-4, IL-10 and IL-13
(Wojdasiewicz, Poniatowski et al. 2014). Of this group, IL-1β and TNFα, are considered to be the major contributors to OA changes in the synovial joint withIL-1β affecting cartilage destruction and TNFα influencing the inflammatory response (Kapoor, Martel-
Pelletier et al. 2011). The same cells that synthesise IL-1β in the synovial joint secrete
TNFα. The increase in concentration of TNFα is also observed in the same regions as IL-
1β such as the synovial fluid and membrane, cartilage and subchondral bone (Melchiorri,
Meliconi et al. 1998; Farahat, Yanni et al. 1993). In addition, TNFα is also responsible for the synthesis of IL-6, IL-8 and VEGF (Guerne, Carson et al. 1990; Honorati, Cattini et al. 2007).
IL-6 and IL-8 levels have been shown to increase in the synovial fluid of RA and OA affected joints, with significantly higher IL-6 levels in lower grade OA cases (Kaneko,
Satoh et al. 2000). An increase in the circulating levels of IL-6 has also been found to be a significant predictor of OA progression (Livshits, Zhai et al. 2009). A potential role of
IL-6 and its effect on articular cartilage may be its ability to inhibit the production of type
18 II collagen (Poree, Kypriotou et al. 2008). Although it has been shown that an increase in
IL-6 production was associated with ageing, when knocked out in mice, a more severe
OA phenotype was observed rather than a reduction (Greene and Loeser 2015).
Investigation of cytokine levels in the synovial membrane of early stage knee OA patients showed an increase in levels of IL-15 along with IL-6, IL-21 and TNFα, along with the matrix degrading enzymes MMP-1 and MMP-3. The levels of IL-15 were also correlated with the levels of IL-6 (Scanzello, Umoh et al. 2009).
IL-17 has been found to have a role in the inhibition of chondrocyte proteoglycan synthesis. It is secreted by CD4+ activated memory T cells and was shown to significantly suppress proteoglycan synthesis in murine articular cartilage. Pre-treating the cartilage with IL-4 reduces the inhibitory effects of IL-17 on proteoglycan synthesis; this is likely via a reduction of inducible nitric oxide synthase (iNOS) (Lubberts, Joosten et al. 2000).
Chondrocytes produce a precursor to IL-18 and secrete a mature form in response to IL-
1 stimulation. The effect of IL-18 on chondrocytes is by the upregulation of IL-18Rα and inducing the production of MMP-1, MMP-3 and MMP-13 (Olee, Hashimoto et al. 1999;
Dai, Shan et al. 2005). It has been shown that SNPs in the IL-18 gene may be a marker of susceptibility to knee OA (Hulin-Curtis, Bidwell et al. 2012).
While the suggestion of elevated pro-inflammatory markers being a significant contributor to continuous osteoarthritic change is valid, the findings of Ahlen et al, show that there is no difference in pro-inflammatory cytokine levels in the synovial fluid of
ACL reconstructed joints, at the 8 years post operative mark when compared with the contralateral knee. However, osteoarthritic changes of cartilage and meniscus damage were still detected in operated knees (Ahlen, Roshani et al. 2015). This data supports the
19 notion of multiple factors including inflammatory cytokines have a role in the pathogenesis of OA.
1.9.3 Anti-inflammatory Cytokines
While the causation of OA may be facilitated in part by the presence of inflammatory cytokines, a group of anti-inflammatory cytokines, namely IL-4, IL-10 and IL-13 have been shown to be involved in the pathogenesis of OA. Both IL-4 and IL-13 activity is mediated through the same receptor system IL-4Rα (Mueller, Zhang et al. 2002). It has been shown that IL-4 has a significant chondroprotective effect by inhibiting the degradation of proteoglycans in articular cartilage, by the inhibition of MMP secretions
(Doi, Nishida et al. 2008), although chondrocytes in OA affected cartilage show a decrease in the protective effects of IL-4 (Millward-Sadler, Wright et al. 2000). Studies have also identified variable number tandem repeat (VNTR) polymorphisms in intron 3 of the IL-4 gene that may be a marker of susceptibility to the development of knee OA
(Yigit, Inanir et al. 2014). IL-10 also demonstrates a chondroprotective effect during the development of OA. Both IL-10 and the receptor IL-10R have been shown to be expressed in chondrocytes and to inhibit the expression of IL-1 and TNFα, as well as the production of MMPs (Iannone, De Bari et al. 2001; Wang and Lou 2001). It has been demonstrated that IL-10 activates STAT3 and the kinase pathway
SMAD1/SMAD5/SMAD8 and ERK1/2 MAPK and induces expression of the bone morphogenic proteins BMP-2 and BMP-6, which belong to the TGF-β family and have an important role in chondrogenesis (Jung, Kim et al. 2013; Umulis, O'Connor et al.
2009). It is clear that IL-10 has a chondroprotective and anti-inflammatory effect on OA affected joints. Production of IL-10 has shown to be induced by mechanical loading and exercise. Studies by Helmark et al and Lems et al demonstrate this with a group of knee
OA sufferers performing exercise for 3 hours alongside a resting control group of OA
20 sufferers of the same severity. All participants had the affected knee joint connected to catheters periarticularly, to examine the levels of IL-10 and other cytokines before, during and after exercise (Helmark, Mikkelsen et al. 2010; Lems and den Uyl 2010). The IL-13 anti-inflammatory cytokine has been shown to have anti-inflammatory and chondroprotective effects on immune cells and articular cartilage (Hart, Ahern et al.
1995), and has the ability to transfer intracellular signals via IL-4Rα/JAK2/STAT3 and
IL-13Rα1/TYK2/STAT1/STAT6 signalling cascades (Bhattacharjee, Shukla et al. 2013).
The role of IL-13 in fibroblasts in the synovium is of particular importance; this is outlined in a study of synovium samples taken from patients undergoing TKR. The synovium samples were administered with IL-13 alongside proinflammatory LPS samples and incubated for 72 hours, then tested via ELISA, northern blotting and binding tests. It was shown that IL-13 inhibited the synthesis of IL-1β, TNFα and MMP-3 while simultaneously increasing the levels of IL-1Ra. In the synovium cells, the mRNA levels of IL-1β were reduced while IL-1Ra mRNA levels were increased, relative to the control samples. The increase in IL-1Ra production in the synovial fibroblasts resulted in a reduction in binding between IL-1β and it’s receptor (Jovanovic, Pelletier et al. 1998).
The ability of these anti-inflammatory cytokines to have a protective effect against OA via the suppression of MMPs and pro-inflammatory cytokines, and their role in mediating inflammation in the early stages of OA highlight the importance of these cytokines and the inflammatory process in the pathogenesis of OA.
1.9.4 Matrix Degrading Enzymes
1.9.4.1 Matrix Metalloproteinases
Chondrocytes in resting state of normal adult articular cartilage are considered quiescent, with very little cartilage matrix turnover. Disruption of this resting state due to cartilage
21 injury causes chondrocytes to proliferate and form clustered cell lacunae. Chondrocytes also have receptors that respond to mechanical stimulation via the extracellular matrix
(ECM) which are activated via injury to the cartilage. This activation results in an increase of matrix degrading enzymes and matrix proteins (Figure 1.1) (Goldring and Marcu
2009). Members of the matrix metalloproteinase (MMP) family, which include aggrecanases and collagenases, can be found in the articular cartilage of OA affected joints. Of the group of MMPs considered most important in the pathogenesis of OA, interstitial collagenase (MMP-1), stromelysin-1 (MMP-3) and collagenase-3 (MMP-13) have been found to have a destructive effect on cartilage and stimulation from cytokines such as IL-1β, IL-17 and TNFα results in the increase in the levels of these MMPs
(Mengshol, Vincenti et al. 2000). Early OA cartilage matrix degradation may be initiated by the degradation of aggrecan via MMP-3 and A Disintegrin and Metalloproteinase with
Thrombospondin Motif 5 (ADAMTS-5). The degradation of type II collagen is also initiated by an increase in activity of MMP-13, resulting in an irreversible destruction of the collagen network. Proteoglycans offer a protective coating for the type II collagen network that inhibits proteinases from degrading them; this highlights the significance of proteoglycan loss as a precursor to cartilage erosion (Loeser, Goldring et al. 2012). Recent studies have demonstrated that cartilage erosion in Mmp13 knockout mice is reduced, however the depletion of aggrecan showed no reduction (Little, Barai et al. 2009). This suggests that while the loss of proteoglycans may accelerate cartilage erosion, it is not an indicator of progressive OA.
Mechanical stimuli such as compression and loading as a result of frequent heavy use and injury can also cause an increase in mRNA levels of MMP-3 along with ADAMTS-5 and the tissue inhibitor of metalloproteinases (TIMP-1) (Lee, Fitzgerald et al. 2005). The
22 levels of MMPs compared to endogenous inhibitors is tipped in favour of the inhibitors in normal human cartilage. If these levels were to shift in favour of MMPs the result would be a shift towards cartilage degeneration (Dean, Martel-Pelletier et al. 1989). The secretion of MMPs in the pathogenesis of OA is not completely localised to the chondrocyte in response to mechanical stimuli and inflammatory cytokines. Osteoblasts residing in sclerotic subchondral bone have been shown to reduce production of collagen
II markers and increase production of MMP-3 and MMP-13. This suggests that the osteoblasts in the subchondral bone may contribute to the initiation of chondrocyte hypertrophy and the mineralisation of the cartilage matrix (Sanchez, Deberg et al. 2005;
Sanchez, Deberg et al. 2005).
1.9.4.2 Aggrecanases
The cartilage matrix is made up of many components. Aggrecan is one of the first of these components that when lost leads to loss of cartilage integrity. Loss of aggrecan is considered to be important in the initiation of OA, which is followed by the irreversible collagen degradation (Hollander, Pidoux et al. 1995; Mankin and Lippiello 1970). The depletion of aggrecan in OA affected cartilage is considered to be due to an increase in proteolytic cleavage of the core protein due to the activity of aggrecanases.
Aggrecanases belong to the family of ADAMTS and are the main proteinases responsible for the cleavage of aggrecan in early stage s of cartilage remodelling. There are 19
ADAMTS family members that are known to influence OA development and progression along with influencing angiogenesis. More specifically ADAMTS-1, ADAMTS-4,
ADAMTS-5, ADAMTS-8 and ADAMTS-9 have been shown to degrade proteoglycan
(Porter, Clark et al. 2005). While ADAMTS-8 and ADAMTS-9 cleavage is highly
23 inefficient, a study of ADAMTS-1 knockout mice showed that there were no morphological differences in cartilage or degree of inflammation between the wild type and the knockout mice, even in response to antigen induced arthritis (Little, Mittaz et al.
2005). Of these ADAMTS family members, the remaining and the most widely studied are aggrecanase-1 (ADAMTS-4) and aggrecanase-2 (ADAMTS-5). These two aggrecanases have been shown to be the most efficient aggrecanases in vitro and are therefore considered to be the major aggrecanases in the pathogenesis of OA (Kashiwagi,
Enghild et al. 2004; Gendron, Kashiwagi et al. 2007). Regulation of the activity of aggrecanases is important in maintaining the balance between the anabolism and catabolism of aggrecan. Inhibitors such as tissue inhibitor of matrix metalloproteinase
(TIMP-3) are control mechanisms of aggrecan catabolism and disturbing the balance between TIMP-3 and ADAMTS-4 in the favour of catabolism can contribute to the degeneration of cartilage in OA (Sahebjam, Khokha et al. 2007; Yamanishi, Boyle et al.
2002). It has been shown that ADAMTS-4 mRNA can be induced by IL-1β or TNFα in human cartilage, however ADAMTS-5 mRNA was not regulated by these cytokines
(Koshy, Lundy et al. 2002). Investigation of ADAMTS-5 has revealed that it is constitutively expressed in human chondrocytes and synovial fibroblasts (Fosang,
Rogerson et al. 2008). Studies involving the inhibition of either of these aggrecanases resulted in an observed reduction of OA degradation in vivo (Malfait, Liu et al. 2002).
Deletion of ADAMTS-4 and ADAMTS-5 in mouse models showed that ADAMTS-5 knockout mice exhibited a reduction in cartilage degradation after surgically inducing OA and that deletion of ADAMTS-4 has no effect on skeletal development, (Glasson, Askew et al. 2004; Glasson, Askew et al. 2005), thus suggesting that ADAMTS-5 may be the primary aggrecanase involved in the development of OA.
24 1.9.5 Growth Factors
1.9.5.1 VEGF
Vascular Endothelial Growth Factor (VEGF) is an angiogenic factor, which is a critical component of angiogenesis during bone remodelling and skeletal development and is an important component of the pathogenesis of OA (Neve, Cantatore et al. 2013). The secretion of VEGF is also important for the stimulation of endothelial cells to undergo differentiation, proliferation and migration for tube formation (Wise, Inder et al. 2012).
Induction of angiogenesis is a key component of endochondral ossification due to the promotion of osteoblast differentiation, thus increasing mineralisation. It is by this process, that the role of VEGF in OA is considered to be of importance. The increase in expression of VEGF in OA affected cartilage in turn stimulates the growth of blood vessels from the subchondral bone into the cartilage and advances the mineralisation tidemark (Lingaraj, Poh et al. 2010). Further studies have shown the presence of higher levels of VEGF in plasma and synovial fluid (Saetan, Honsawek et al. 2014). A significant association of SNPs in VEGF and a human OA affected cohort further demonstrates the role of VEGF in the pathogenesis of OA (Rodriguez-Fontenla, Calaza et al. 2014).
1.9.6 Biological Pathways Involved With OA
1.9.6.1 TGF-β/BMP Pathway
The TGF-β superfamily is a group of secreted polypeptides that include TGF-βs (TGF-
β1, TGF-β2, TGF-β3), activins, inhibins nodal, Growth Differentiation factors (GDFs) and Bone Morphogenic Proteins (BMPs). The TGF-β/BMP signalling pathway is involved in critical cellular processes such as the embryonic stages of skeletal development and post natal bone homeostasis (Guo and Wang 2009; Chen, Deng et al.
2012). TGF-β and BMP signal transduction occurs via heteromeric complexes of
25 transmembrane serine/threonine type I and type II receptors at the cell surface. TGF-β binds to a type II receptor which then recruits and phosphorylates a type I receptor. The type I receptors are also known as Activin receptor-like kinases (ALKs) which are active downstream of the type II receptor and ultimately determine the receptor specificity.
Intracellular signalling begins upon the activation of type I receptors by the phosphorylation of R-Smad. The type I receptor ALK5 stimulates Smad2/3 phosphorylation while ALK1,2,3 and 6 mediate Smad1/5/8 activation (Davidson, Remst et al. 2009). TGF-β signalling generally depends on Smad2/3 while BMP signalling generally depends on Smad1/5/8, both of these signalling cascades have been reported to be involved in chondrocyte differentiation (van der Kraan, Goumans et al. 2012; Plaas,
Velasco et al. 2011). Smad activity is modulated by inhibitors Smad6/7 and Smad ubiquitination regulatory factors (Smurfs). Smurfs target Smad pathways specifically, with Smurf1 targeting Smad1/5 and Smurf2 targeting Smad2/3, and the inhibitor Smad6 and Smurf1/2 are known to inhibit Smad1/5/8 signalling (Figure 1.3) (van der Kraan,
Goumans et al. 2012; Wu, Wang et al. 2008; Horiki, Imamura et al. 2004). The non- canonical TGF-β/BMP signalling requires activation of p38 MAPK instead of Smads, by phosphorylation of TAK1 that recruits TAB1 to initiate MKK-p38 MAPK or MKK-
ERK1/2 signalling cascades.
TGF-β is implicated in mesenchymal condensation, proliferation of chondroblasts and the deposition of extracellular matrix molecules specific to cartilage. At younger ages,
TGF-β is considered to be protective for cartilage. However, TGF-β has a significant role on OA development including the degradation of cartilage and osteophyte formation in older individuals. ALK5 expression is found to decrease more than ALK1 in aged and
OA articular chondrocytes. It has been shown that there is a high correlation with ALK1
26
Figure 1.3 TGF-β/BMP signalling pathways
BMP binds to ALK 1,2,3 and 6 via type I receptors on the cell surface and results in
Smad1/5/8 phosphorylation resulting in downstream expression of MMPs. TGF-β binds to ALK 5 via type I receptors on the cell surface resulting in Smad2/3 phosphorylation resulting in downstream expression of type II collagen and aggrecan.
Figure information sourced and adapted from (Chen, Deng et al. 2012; van der Kraan,
Goumans et al. 2012).
27 and the expression of MMP-13, which ultimately contributes to the formation of the OA phenotype. Conversely, the expression of collagen II and aggrecan is correlated with
ALK5 thus demonstrating the shift from Smad2/3 to Smad1/5/8 signalling results in the reduction of chondroprotective markers and the dominant production of degenerative markers (Davidson, Remst et al. 2009).
1.9.6.2 Wnt/β-catenin pathway
Wnt proteins have an integral role in developmental processes including organogenesis, cell differentiation, morphogenesis and tissue remodelling. There are two distinct pathways that Wnt signalling can be divided into. The canonical (Wnt/β-catenin pathway) and the non-canonical, which can be further divided into the planar cell polarity pathway
(PCP) and the Wnt/Ca2+ pathway (Clevers 2006; B Blom, L van Lent et al. 2010). In the canonical signalling pathway, Wnt binds to Frizzled on the cell surface and the co- receptors low density lipoprotein receptor-related protein 5 and 6 (LRP-5/6), which leads to the accumulation of β-catenin in the cytoplasm. β-catenin then translocates into the nucleus where it affects transcription factors for activation of target genes (Clevers 2006).
The non-canonical pathways do not involve β-catenin and instead involve the activation of protein kinase C (PKC), calmodulin dependent kinase II and c-Jun N-terminal kinase
(JNK) (Figure 1.4). β-catenin is known to induce moderate chondrocyte maturation.
Activation of this pathway during skeletal development stimulates chondrocyte hypertrophy and calcification, and the expression of MMPs and VEGF. Under physiological conditions, the canonical Wnt signalling has also been shown to inhibit endochondral ossification, chondrocyte maturation and proliferation, and cartilage matrix production (Corr 2008). Therefore, both insufficient and excessive levels of β-catenin may impact on the homeostasis of articular chondrocytes. β-catenin regulates OA
28
Figure 1.4 Wnt signalling pathway
The canonical Wnt/β-catenin pathway in resting state without Wnt binding results in the degradation of β-catenin in the cytoplasm. Binding of Wnt to Frizzled and co- receptors LRP5/6 results in the accumulation of β-catenin, which translocates into the nucleus where it affects transcription factors for target gene activation. The non- canonical pathway does not involve β-catenin and instead utilises PKC, Camk2 and c-
Jun N-terminal kinase (JNK). Figure information sourced and adapted from (Clevers
2006; B Blom, L van Lent et al. 2010)
29 development by inducing apoptosis of chondrocytes in the middle section of articular cartilage, resulting in degradation. Simultaneously, the maturation of chondrocytes at the periphery combined with the availability of blood vessels from the synovium and tendons could complete the endochondral ossification process to induce the formation of osteophytes (Kawaguchi 2009).
Antagonists of the canonical Wnt/β-catenin signalling pathway such as Dickkopf (DKK) proteins have important roles in bone formation and embryonic cell fate. The antagonism of Wnt by DKK occurs via the binding of the C-terminal cysteine rich domain of DKK to the Wnt co-receptor LRP5/6. DKK1 is a key mediator in the remodelling of joints and a negative correlation exists between DKK1 expression and OA development.
Blockading DKK1 leads to the inhibition of bone resorption in joints and results in the rapid formation of osteophytes, which therefore indicates the involvement of Wnt signalling in osteophyte formation in OA development. The mechanism of how blockading DKK1 induces osteophyte formation is linked with the ability of Wnt proteins to induce OPG expression, which blocks RANKL mediated bone resorption (Diarra,
Stolina et al. 2007).
1.9.6.3 RANKL/RANK/OPG pathway
Subchondral bone changes and the formation of osteophytes are considered to be hallmarks of OA development. The mechanism of bone remodelling is tightly managed by the molecular interactions of OPG, RANK and RANKL. RANKL binds to RANK to activate an intracellular signalling cascade. Studies have shown that human chondrocytes express and produce OPG, RANK and RANKL and that OPG may increase catabolic factors that are involved in the development of OA (Kwan Tat, Amiable et al. 2009).
30 RANKL is produced by cell of the osteoblastic lineage and is a key component in the mediation of bone resorption by the regulation of osteoclastogenesis and mature osteoclast activation, by binding to RANK, which is a receptor located on the cell surface of osteoclasts and osteoclast precursors (Kearns, Khosla et al. 2008). OPG is secreted by osteoblasts and other cell types and through the binding of OPG with RANKL; OPG is able to inhibit the interaction between RANKL and RANK thus preventing the activation of RANK and ultimately osteoclastogenesis (Khosla 2001). The role of
RANKL/RANK/OPG has also been observed in the maintenance of the immune response. Activated T cells express RANKL, which then binds to RANK, which is expressed on the surfaces of dendritic cells and regulates their function and survival
(Cremer, Dieu-Nosjean et al. 2002; Yu, Gu et al. 2003). Investigation of osteoblastic cells derived from OA affected patients showed significantly lower levels of RANKL expression which was attributed to the higher levels of OPG inhibiting the binding to
RANK thus resulting in bone formation phenotypes such as osteophytes (Corrado, Neve et al. 2013).
The binding of RANKL with RANK initiates an intracellular signalling cascade within osteoclast precursors, which results in the activation of NF-κB, NFATc1, Akt/PKB, JNK,
ERK and p38 MAPK pathways. RANK lacks intrinsic kinase activity and therefore uses interactions with adaptors and docking proteins such as tumour necrosis factor receptor- associated factors (TRAFs) upon the activation of the receptor by RANKL (Wong, Besser et al. 1999; Darnay, Haridas et al. 1998). RANK utilises TRAFs 1, 2, 3, 5 and 6 for the activation of downstream signalling (Darnay, Haridas et al. 1998). TRAFs activate the
TAB1/TAB2/TAK1 complex, which results in the activation of IKKβ and MAPKs (JNK,
ERK and p38). Following the activation of these pathways, activation and translocation
31 of transcription factors such as NF-κB, NFATc1, CREB, AP-1 and MITF, which are in turn responsible for the differentiation of osteoclasts and the initiation of bone resorption
(Figure 1.5) (Takayanagi, Kim et al. 2002). NF-κB signalling mediates the inflammatory response by chondrocytes, which leads to the destruction of the extracellular matrix. NF-
κB and MAP kinases can mediate the expression of MMPs induced by TNFα or IL-1β, suggesting that NF-κB may be a potential therapeutic target for the treatment of OA
(Roman-Blas and Jimenez 2006).
32
Figure 1.5 RANKL/RANK/OPG signalling pathway
Osteoblasts express RANKL on the cell surface, which bind to RANK, which is found on the cell surface of precursor and mature osteoclasts. OPG is secreted by osteoblasts, chondrocytes and stromal cells and is a decoy receptor to inhibit the binding of
RANKL to RANK. The RANK receptor lacks intrinsic kinase activity and uses docking proteins and adaptors such as TRAFs 2,3,5 and 6 to activate the
TAB1/TAB2/TAK1 complex, resulting in IKK and MAPK activation. The activation of this pathway results in the translocation and activation of transcription factors involved in osteoclast differentiation such as NF-κB, NFATc1, CREB, AP-1 and
MITF. Figure information sourced and adapted from (Khosla 2001; Wong, Besser et al. 1999; Takayanagi, Kim et al. 2002)
33 1.10 Genetic Research Models of OA
1.10.1 Heritability
A number of lines of evidence suggest that there is a genetic component to the development of OA. A recent study has shown that screening for a collection of 23 SNPs that have been associated with OA phenotypes in patient cohorts can be an accurate prognostic tool (Blanco, Möller et al. 2015). Studies of 130 identical and 120 non- identical twins also showed that 35% to 65% of hand and knee osteoarthritis could be influenced by genetic factors (Spector, Cicuttini et al. 1996). Additional studies performed with inbred mouse strains such as STR/ort, have observed the development of spontaneous OA phenotypes (Mason, Chambers et al. 2001), indicating a strong genetic basis for the disease.
1.10.2 Genome Wide Association Studies
Investigation of OA in human patient cohorts is the standard of clinical research into the disease. Categorising patients by history of knee injury, self reported injury, heavy mechanical stress and surgery can classify post-traumatic OA cases within patient cohorts
(Grotle, Hagen et al. 2008; Hannan, Felson et al. 2000; Muraki, Akune et al. 2009;
Roos, Lauren et al. 1998). Investigating genetic components of OA is conducted using the common disease-common variant hypothesis, which lead to the creation of Genome
Wide Association Studies (GWAS) (Reich and Lander 2001). These studies have been responsible for the identification of many loci associated with a range of diseases
(Hindorff LA et al, NHGRI GWAS catalog).
34 Identification of genetic loci that are uniquely associated with OA patients using this method has resulted in the identification of many genes worthy of further investigation into their role in the manifestation of OA. The most powerful human GWAS to date was conducted by Zengini et al (Zengini, Hatzikotoulas et al. 2018), using the UK Biobank data of 30,727 OA cases and 297,191 controls to uncover potentially novel OA risk loci.
This study identified nine novel risk loci, six of which were genome wide significant.
Further investigation of the genes containing or located near the implicated variants were
MAP2K6, ZC3H11B, ZNF345, ANXA3, TGFA and JPH3. The variant rs116882138 was located downstream of MOB3B and upstream EQTN and the variant rs6516886 was located upstream of RWDD2B and LTN1 was found to be located 28kb from rs6516886.
The variant rs11780978 was found in the intronic region of PLEC, which interestingly has been shown to have a functional effect on skeletal muscle in mice (Castanon, Walko et al. 2013). The functional analysis looking at differential expression of genes located near the novel variants implicated PCYOX1, FAM136A, BACH1, MAP3K7CL, BOP1,
PLAA and ZNF382. The study also showed a strong positive genetic correlation between body mass index and related anthropometric traits with both hip and knee OA from self reported and hospital diagnosed cases, along with a strong correlation between hip and knee OA. These results suggest that genetic factors that influence OA development may not be completely site specific when only hips and knees are considered. Including the new loci identified in this study, the collective OA loci identified to date accounts for
26.3% of the disease trait variance.
A previous study also highlighted the association of diabetes with OA by identifying a genome wide significant variant in the GLIS3 gene using datasets from individuals who
35 underwent total joint replacement in the knee or hip as a result of OA (Casalone,
Tachmazidou et al. 2018).
An earlier study known as the arcOGEN study was previously the largest GWAS undertaken for OA. This GWAS utilised 7,410 unrelated patients, 80% of which were candidates for total joint replacement and a cohort of 11,009 unrelated controls. The results of this study revealed eight novel loci that were associated with OA development risk in a European population, with five crossing the genome wide significant threshold.
The most significant signal contained two candidates GLT8D1 and GNL3 being highly associated with total joint replacement patients. Of the remaining candidates ASTN2,
CHST11, rs10492367 located between KLHDC5 and PTHLH, and rs9350591 located between FILIP1 and SENP6 were all associated with hip arthritis and total hip replacement. TP63 was the only gene associated with total knee replacement in females only and the affected gender analysis showed that FTO was associated with OA affected females and rs10948172 located between SUPT3H and CDC5L was associated with OA affected males (Zeggini, Panoutsopoulou et al. 2012).
Prior to the arcOGEN and UK biobank studies, only three loci were found to be associated with OA in European populations at the genome wide significant level. A meta-analysis of European and Asian cohorts by Chapman et al (Chapman, Takahashi et al. 2008), consisting of 11,000 individuals implicated a functional SNP in GDF5, which was later validated with genome wide statistical significance by (Valdes, Evangelou et al. 2011).
A genome wide significant locus on chromosome 7q22 was identified implicating the
GPR22 gene as a potential candidate (Kerkhof, Lories et al. 2010). Soon after this study, another analysis consisting of 3,177 OA affected individuals and 4,894 controls revealed
36 SNPs in MCF2L to be highly associated with the European OA affected cohort (Day-
Williams, Southam et al. 2011). The arcOGEN study also included the regions implicated from these earlier studies and was able to detect a signal for MCF2L in their cohort but showed no evidence of association with GDF5. It was also suggested that previously improper phenotype definition might have contributed to the lack of ability to detect strong and reproducible signals (Zeggini, Panoutsopoulou et al. 2012).
A later study that included the arcOGEN consortium conducted an assessment of OA candidate genes in a meta-analysis across nine different GWAS. This study showed that out of 199 potential candidate genes only two, COL11A1 and VEGF were found to be significantly associated with OA (Rodriguez-Fontenla, Calaza et al. 2014).
While these human genome wide association studies are adequately powered and have revealed a great deal about the genetic component of OA, the disadvantages of these studies have been described within them from a poorly defined phenotype to the inability account for the external factors contributing to the disease such as environment, body mass index and lifestyle of the large patient cohort. While human studies are appropriate from a true phenotype and translatable genetic perspective, the need for non-invasive procedures to identify and track disease progression can be a limiting factor when considering the multifactorial nature of OA due to the variability of symptomatic versus radiographic diagnosis (Hannan, Felson et al. 2000). A true macroscopic and/or microscopic assessment of the OA phenotype in a research setting requires a more invasive approach that can’t be considered in humans.
37 1.10.3 Animal Models
Large animal models such as sheep and goats are useful for investigating OA due to the size of the affected joints and the anatomical similarities shared with the human ACL and meniscus (Proffen, McElfresh et al. 2012). Articular cartilage thickness is also sufficient for the creation of osteochondral defects for inducing post-traumatic OA phenotypes and assessing cartilage healing (Brehm, Aklin et al. 2006). The disadvantages of these models are the cost to maintain the animals required and the inability to genetically modify and the lack of spontaneous OA observed. The slow progression of OA also makes large animal models very time consuming to generate phenotypes to study, as they take a minimum of two years to reach skeletal maturity (Little, Smith et al. 2010).
Murine models are inexpensive and have a much shorter time to complete spontaneous disease progression, with spontaneous OA developing within the ages of 9-12 months
(Poole, Blake et al. 2010). Murine models of genetically influenced spontaneous OA can be utilised via knockout mice such as the STR/ort mouse model (Mason, Chambers et al.
2001, Glasson 2007). A quantitative trait locus mapping study for spontaneous OA in the
STR/ort mice identified a locus on chromosome 4 that contained several genes such as
Mmp16 (MT3-MMP), Gdf6 (a member of the TGF-β superfamily), Map3k7 (TAK1) and
Tgfbr1 (ALK5) (Watanabe, Oue et al. 2012). There are several observed limitations of this study, the first being the small number of replicates (n=100), the limited mapping resolution due to the use of only 122 autosomal microsatellite markers for the QTL mapping and the histological analysis of only tibias which had been separated from the femurs prior to the analysis.
38 Recombinant Inbred (RI) strains of mice that best mimic the characteristics of a human
GWAS, while overcoming it’s limitations by controlling the external factors may be used to investigate spontaneous forms of OA with better mapping resolution using SNPs to genotype the RI strains (Aylor, Valdar et al. 2011). The ability to cull mice at desirable time points and collect fresh tissue from these cohorts will also benefit researchers by allowing a thorough histological evaluation of the intact murine joint.
1.10.4 The Collaborative Cross
The Collaborative Cross (CC) is a large scale Recombinant Inbred (RI) murine strain panel, assembled for the purpose of generating a mouse phenotype library for genetic linkage analyses (Churchill, Airey et al. 2004). The CC project began in 2004 at the
Jackson Laboratory in the United States and later expanded to collaborators in Kenya,
Israel and Australia. The Australian version of the project is also known as “The Gene
Mine”, and was established by Geniad Ltd (Morahan, Balmer et al. 2008).
The CC is in essence a murine based GWAS. The project generates RI strains that are derived from eight founder strains, five of which are classical laboratory inbred strains
(A/J, C57BL/6J, 129S1/SvImJ, NOD/LtJ and NZO/H1LtJ), and three of which are wild derived subspecies (CAST/EiJ, PWK/PhJ and WSB/EiJ) (Figure 1.6) (Churchill, Airey et al. 2004; C.C. Consortium 2012). The selection of these strains as founders ensured the maximum coverage of diversity in the mouse genome, with the variance randomly distributed amongst the CC lines. The CC lines are generated via a funnel-breeding scheme of the eight founders to generate a strain carrying contributions from all eight.
This strain is then inbred for a further 20 generations to be considered as fully inbred. The
39 Figure 1.6 The Collaborative Cross funnel breeding scheme
Eight founder mouse strains have been selected to encompass approximately 90% of the genetic variance in the mouse genome. RI strains are generated via sibling mating from G3 onwards until G20 when they are considered to be fully inbred. Figure adapted from (Morahan, Balmer et al. 2008).
40 aim of the CC project is to generate over 1,000 RI lines that present with a wide variation of phenotypes for the purpose of screening for heritable disease traits.
The application of CC mice in disease studies have also included strains that are not considered to be fully inbred, however a wide range of phenotypes have been identified in these studies thus far (Aylor, Valdar et al. 2011; Bottomly, Ferris et al. 2012;
Phillippi, Xie et al. 2014). Simulation studies investigating the power and the precision of the CC’s genetic mapping ability demonstrated significant performance improvements when compared to two-way models and was able to provide sufficient resolution even at half of the desired number of RI lines (Broman 2005; Valdar, Flint et al. 2006). Studies using low strain numbers to map known major effect genes affecting coat colour in the
CC, also showed that using less than 100 strains was still sufficient to map significant quantitative trait loci with confidence (Ram, Mehta et al. 2014).
The inbred platform of the CC enables the users to utilize public databases for genotyping information of the founder strains and haplotype reconstruction of the RI strains, avoiding the need to genotype each strain used in the study by researchers undertaking the screening. The major advantages of the CC phenotype library over traditional human
GWAS is predominantly the ability to control the environmental effect that may contribute to phenotypes that are being screened for heritable effect genes. The fast breeding and maturity of the animals is also an asset to researchers undertaking studies of age related diseases such as OA, along with the ability to readily source the relevant tissue for a detailed phenotypic analysis. Finally, the ease of genetic manipulation of the mouse is a valuable resource for confirmation of traits observed in the CC cohort. The ability to back-cross the affected strains or knock out potential targets identified in the
41 mapping process will aid in further understanding of the relationship between the loci and the disease phenotype.
While the CC mice breeding and strain generation is still progressing, there is still scope to utilize the CC mice in genetic research. Currently, there has been no published work in the field of OA using CC mice, with the closest comparable study being the QTL mapping using STR/ort mice (Watanabe, Oue et al. 2012). There has also been one study investigating genes for bone microarchitecture using the CC cohort (Levy, Mott et al.
2015). This study utilized 31 RI strains from the CC cohort (n=160 mice) and conducted genome wide mapping across 77,808 SNP markers. This study was successful in identifying six loci associated with a range of skeletal traits including previously validated genes such as Avp and Oxt, along with potentially novel candidates. While this study is an encouraging proof of principle for the use of the CC mouse cohort for investigation of genes that are regulating complex diseases, there are limitations of limited strain numbers and a young and narrow age range (10-13 weeks). Therefore, further and better powered studies into a wider range of diseases such as spontaneous OA are needed.
1.11 Histological OA Grading Systems
A significant advantage of the small animal model is the assessment of the entire joint histologically including subchondral bone, synovium and meniscus. This enables the implementation of scoring systems such as the Mankin system (Mankin, Dorfman et al.
1971), which has been the most widely used amongst the literature. Despite the prolific use of the Mankin system, it’s appropriateness has been questioned due to the differences between human and mouse cartilage structure (Glasson, Chambers et al. 2010). A revision
42 of this scoring system by McNulty et al was recommended to overcome the shortcomings of the Mankin system (McNulty, Loeser et al. 2012). While the McNulty system is comprehensive in assessing many of the OA phenotypes observed in the mouse, its technical complexity has restricted its use.
The Osteoarthritis Research Society International (OARSI) recommended a system for the histological grading of OA in the murine knee joint that is both comprehensive and easy to implement. It involves a semi quantitative score ranging from 0-6 of structural cartilage damage in all four observable quadrants in frontally sectioned knee joints, medial/lateral femoral condyles and medial/lateral tibial plateaus. An additional osteophyte scoring range of 0-3 can also be implemented. This method of scoring requires multiple sections across the entire joint depth with a summed total score or a maximum observed score per quadrant reported. The OARSI system also has good intra-observer reproducibility in both experiences and novice assessors making it a versatile tool for high volume screening (Glasson, Chambers et al. 2010).
1.12 Summary
OA is a significant and complex disease involving many systems, cells and tissue types, which affects a significant portion of the population. It is clear that there is a strong genetic component to the disease although current human research fails to control for external factors affecting the experimental cohorts. While animal studies can overcome this, there is a lack of spontaneous OA phenotype studies that can focus purely on the genetic influence on the phenotype. RI models such as the CC mice can offer a large cohort of genetically diverse animals for screening for OA phenotypes. The current
43 treatment of OA is symptomatic and in order to develop new disease modifying treatments, more research into the genetic basis of the disease is required.
44 Chapter 2 Hypothesis and Aims
45 The pathogenesis of OA is extremely complex with many factors from multiple cells, tissues and systems contributing to the development of the cartilage degradation. Many genes have been characterised in human GWAS although these studies suffer from poor phenotype definition and the inability to control for external factors, which have been shown to initiate OA. Current animal studies are capable of overcoming such factors however the majority of studies induce the OA phenotype and have a poor representation of spontaneous OA. Linkage analyses of spontaneous OA models suffer from low mapping resolution and are underpowered. The CC is capable of generating strains of significant genetic diversity to observe disease phenotypes with a large cohort and high mapping resolution. I hypothesise that screening the Collaborative Cross mouse library strains will result in identifying a range of OA phenotypes. Mapping genetic loci will identify novel gene loci associated with the development of spontaneous OA.
In order to test this hypothesis, the following aims are proposed:
a) To screen CC mice for osteoarthritic phenotypes using histology.
b) Categorize osteoarthritic phenotypes using the OARSI grading method.
c) Use quantitative trait locus mapping to identify common gene loci consistent with
prominent osteoarthritic phenotypes using genotyping data of the CC mice.
d) Prioritize genes implicated at the resulting loci and identify potential candidates.
e) Confirm roles of candidate genes using in house and publically available
expression analysis and bioinformatics approaches.
46 Chapter 3 Materials and Methods
47 3.1 Materials
3.1.1 Equipment
Equipment Name / Model Manufacturer
A&D® HT500 Compact Digital Scale A&D Engineering Inc., USA
Shimadzu Scientific ATX124 Analytical Balance Instruments Inc., USA
B-28 Incubator BINDER Inc., USA
Biological Safety Cabinet Class II Gelaire Pty. Ltd., Australia
Centrifuge 5430R Eppendorf AG, Germany
Centrifuge 5810R Eppendorf AG, Germany
Easypet® Electronic Pipette Eppendorf AG, Germany
Eclipse Ti Fluorescent Microscope with Intenslight C- Nikon Instruments Inc., USA HGFIE Precentered Fiber Illuminator
ESCO® PCR Cabinet Esco Technologies Inc. USA
Grant-Bio PV-1 benchtop vortex Thermo Fisher Scientific, USA
Johndec Engineering Plastics Fume Control System FE2000 Pty. Ltd, Australia
Leica RM2155 automated microtome Leica Microsystems, Germany
Mastercycler® Pro Thermal Cycler Eppendorf AG, Germany
Micro One TOMY centrifuge Quantum Scientific, Australia
Milli-Q Integral Water Purification System Millipore Corporation, USA
Minispin® plus centrifuge Eppendorf AG, Germany
Model 680 Microplate Reader Bio-Rad, USA
48 PowerPac™ HC Power Supply Bio-Rad, USA
Sterile-Cycle CO2 Incubator Thermo Fisher Scientific, USA
Tissue-Tek TEC 5 embedding station Sakura Finetek, Europe
Thermomixer® Comfort Eppendorf AG, Germany
Tissue Processor TP1020 Leica Microsystems, Germany
Veriti® 96-Well Thermal Cycler Applied Biosystems, USA
ViiA™ 7 Real-Time PCR System Applied Biosystems, USA
Xplorer® Electronic Pipette (5-100 μl) Eppendorf AG, Germany
3.1.2 Chemical Reagents
Reagent Manufacturer
2-Ethoxyethanol Sigma-Aldrich, USA
Agarose Powder Promega Corporation, USA
Aluminium Sulfate Sigma-Aldrich, USA
Bromophenol Blue Sigma-Aldrich, USA
Chloroform Sigma-Aldrich, USA
Dibutylphthalate Polystyrene Xylene Sigma-Aldrich, USA (DPX)
Di-Sodium Phosphate (Na2HPO4) Sigma-Aldrich, USA
Eosin-Y Disodium Salt Sigma-Aldrich, USA
Ethanol BDH Laboratory Supplies, England
Ethylene Glycol Sigma-Aldrich, USA
Ethylenediaminetetraacetic Acid (EDTA) BDH Laboratory Supplies, England
Fast Green FCF BDH Laboratory Supplies, England
49 Formaldehyde Solution Sigma-Aldrich, USA
Glacial Acetic Acid Sigma-Aldrich, USA
Glycine Sigma-Aldrich, USA
Glycerol BDH Laboratory Supplies, England
Haematoxylin Sigma-Aldrich, USA
Hydrochloric Acid BDH Laboratory Supplies, England
Methanol BDH Laboratory Supplies, England
Mono-Sodium Phosphate (NaH2PO4) Sigma-Aldrich, USA
Nuclease-Free Water Promega Corporation, USA
Paraformaldehyde Sigma-Aldrich, USA
Phloxine B Sigma-Aldrich, USA
Propan-2-ol BDH Laboratory Supplies, England
Protease Inhibitor Cocktail Roche Diagnostics, Germany
Safranin-O BDH Laboratory Supplies, England
Sodium Acetate Trihydrate Sigma-Aldrich, USA
Sodium Chloride BDH Laboratory Supplies, England
Sodium Deoxycholate BDH Laboratory Supplies, England
Sodium Hydroxide BDH Laboratory Supplies, England
Sodium Iodate Sigma-Aldrich, USA
Sodium Tartrate Dehydrate AJAX chemicals, Australia
Triton X-100 Sigma-Aldrich, USA
Trizma Base Sigma-Aldrich, USA
Trizma Hydrochloride Sigma-Aldrich, USA
TRIzol® Reagent Invitrogen, Australia
50 Tween-20 MP Biochemicals, France
Xylene Hurst Scientific, Australia
3.1.3 Buffers and Solutions
Solution Composition, Preparation and Storage
100 mM stock solution of each dNTP was
prepared. 25 % (v/v) of each dNTP was
combined to give a 25 mM stock of dNTPs (dATP, dCTP, dGTP, dTTP) combined dNTPs.
5 nM dNTP was prepared by dilution of
the 25 mM stock in ddH20.
Aliquots stored at -20oC.
14 % (w/v) EDTA diluted in Milli-Q
EDTA water. Adjusted to pH 8 to dissolve EDTA
and then adjusted to pH 7.4.
250 ml ethylene glycol dissolved in 750
ml in ddH20, 6 g haematoxylin, 0.6 g
Haematoxylin (Gill’s) sodium iodate, 80 g aluminium sulfate and
20 ml glacial acetic acid. Diluted 1:5 in
ddH20 and filtered before use.
10 % (v/v) formaldehyde solution, 45 mM
Neutral Buffered Formalin (NBF) Na2HPO4 and 33 mM NaH2PO4 in Milli-Q
water.
Paraformaldehyde 4 % (w/v) paraformaldehyde in 1x PBS.
51 Stored at -20oC.
10x PBS stock solution: 70 mM Ha2HP04,
30 mM HaH2PO4, 1.3 M NaCl dissolved
Phosphate Buffered Saline (PBS) in Milli-Q water.
1x PBS: 10x PBS stock solution diluted
1:10 in ddH20 and adjusted to pH 7.4.
Sodium Chloride 5 M stock solution in ddH20.
Sodium Hydroxide 5 M stock solution in ddH20.
50x TAE stock solution: 2 M Trizma base,
5.71 % (v/v) glacial acetic acid and 50
Tris base, acetic acid, EDTA buffer (TAE) mM EDTA.
1x TAE solution: dilution of 50x stock
solution.
3.1.4 Tissue Culture Materials and Reagents
Material / Reagent Manufacturer
1α25-dihydroxyvitamin D3 Sigma-Aldrich, USA
β-glycerophosphate Sigma-Aldrich, USA
Cell Dissociation Solution, Non-enzymatic Sigma-Aldrich, USA
Cell Scraper Sarstedt, Germany
Cryopure Cryogenic Vials (1.6ml) Sarstedt, Germany
Dexamethasone Sigma-Aldrich, USA
Dimethyl Sulfoxide (DMSO) Sigma-Aldrich, USA
Falcon Cell Strainer (100 µm) BD Biosciences, USA
52 Fetal Bovine Serum Invitrogen, New Zealand
Gibco® Minimum Essential Medium Invitrogen, Australia alpha (α-MEM)
Gibco® Dulbecco’s modified eagle Invitrogen, Australia medium (DMEM)
Glass Bottom Culture dish (10 mm with 3 MatTek Corporation, USA mm No. 1.5 glass cover slip)
GlutaMAX™ Invitrogen, Australia
Hanks’ Balanced Salt Solution (HBSS) Invitrogen, New Zealand
L-ascorbic Acid Sigma-Aldrich, USA
Penicillin/Streptomycin Invitrogen, Australia
Serological Pipettes (10 ml, 25 ml) Sarstedt, Germany
Tissue Culture Flask (25 cm2) Nunc, Denmark
Tissue Culture Flask (75 cm2) Nunc, Denmark
Tissue Culture Plate (96-Wells) Nunc, Denmark
TrypLE™ Express Invitrogen, Australia
Trypsin-EDTA (0.25%) Invitrogen, Australia
3.1.5 Oligonucleotide Primers and Housekeeping Genes Primers
Oligonucleotide primer sequences for Real-time PCR amplifications were designed using
Primer-BLAST (https://www.ncbi.nlm.nih.gov/tools/primer-blast/), and screened through NCBI-BLAST to ensure specificity (Ye, Coulouris et al. 2012). The primers were purchased from Geneworks, Australia, and shipped in freeze-dried form. They were then reconstructed in sterile and nuclease-free ddH2O at a final stock concentration of 100 µM.
53 Aliquots of resuspended oligonucleotides were stored at -20 ºC and placed on ice during usage.
Primer Sequence (5’ – 3’)
Acp5 (TRACP) Forward CAGCAGCCAAGGAGGACTAC
Acp5 (TRACP) Reverse ACATAGCCCACACCGTTCTC
Bglap (Osteocalcin) Forward GCGCTCTGTCTCTCTGACCT
Bglap (Osteocalcin) Reverse ACCTTATTGCCCTCCTGCTT
Hprt Forward CAGTCCCAGCGTCGTGATTA
Hprt Reverse TGGCCTCCCATCTCCTTCAT
Cdh4 Forward CCTCTTACCATCTCCGAGCCC
Cdh4 Reverse GACCCATGTCCCGTGGAGTG
Col2A1 Forward ATGAGGGAGCGGTAGAGACC
Col2A1 Reverse GCCCTAATTTTCGGGCATCC
Actb (β-actin) Forward CACCCGCGAGCACAGCTTCTT
Actb (β-actin) Reverse CCACCATCACACCCTGGTGCCT
Ctsk (Cathepsin K) Forward CCAGTGGGAGCTATGGAAGA
Ctsk (Cathepsin K) Reverse AAGTGGTTCATGGCCAGTTC
Alpl (Alkaline phosphatase) Forward AACCCAGACACAAGCATTCC
Alpl (Alkaline phosphatase) Reverse GCCTTTGAGGTTTTTGGTCA
54 3.1.6 Software
Software Description Manufacturer
Endnote (Version X8) Clarivate Analytics, NY, USA
Microsoft Office 2013 Microsoft, USA
Graphpad Prism (version 7) GraphPad Software, Inc. USA
NIS-Elements Basic Research Nikon Instruments Inc., USA (version: 3.22.14)
NIS-Elements C Nikon Instruments Inc., USA
NRecon (version: 1.6.9.8) Bruker® Skyscan 1176, Belgium
ViiA™ 7 software (version 1.2.1) Applied Biosystems, USA
3.1.7 Gene Mapping Links
Gene Mapping Links Website
GeneMiner (Geniad) http://www.sysgen.org/GeneMiner/
http://www.sanger.ac.uk/science/data/mou Mouse Genomes Project (Sanger Institute) se-genomes-project dbSNP Database (National Institutes of https://www.ncbi.nlm.nih.gov/SNP/ Health)
The Collaborative Cross Database (UNC http://csbio.unc.edu/CCstatus/index.py Systems Genetics)
Entrez Gene (National Institutes of Health) https://www.ncbi.nlm.nih.gov/gene
Ensembl Genome Browser 87 www.ensembl.org/
55 3.1.8 Bioinformatics Tools
Bioinformatics Links Website
BioGPS www.biogps.org
Uniprot http://www.uniprot.org/
EMBOSS Needle – Pairwise https://www.ebi.ac.uk/Tools/psa/emboss_needle/ sequence alignment
GeneCards http://www.genecards.org/
PathCards http://pathcards.genecards.org/
STRING https://string-db.org/
3.2 Methods
3.2.1 Workflow
This study will begin by screening the available CC mice for OA phenotypes, mapping the genes associated and validating the potential candidate genes for each phenotype.
Validation of genes will be conducted using bioinformatics approaches by literature searches, expression studies both internally and by use of external expression information sources and correlations with human genome wide association studies (Figure 3.1).
3.2.2 Animal model
The Collaborative Cross Gene Mine has been generated by Grant Morahan at Geniad.
Eight founder strains, A/J, C57BL/6J, 129S1/SvImJ, NOD/ShiLtJ, NZO/H1LtJ,
CAST/EiJ, PWK/PhJ and WSB/EiJ, (these will be referred to throughout the results as
A/J, C57BL/6J, 129, NOD, NZO, CAST, PWK and WSB). (G0) are crossed to generate the first generation (G1), which are then bred to generate the parent strain (G2). The offspring of the parent strain
56
Figure 3.1 Workflow of the Collaborative Cross OA screening project
The study begins with the dissection and sample preparation of hind limbs of the CC mice. Each mouse is individually screened via serial histology sectioning and given an OARSI score. Scores are then used to map genetic loci using QTL mapping software. Genes in resulting loci are investigated further using bioinformatics approaches and in house qPCR expression experiments. Convincing candidate genes are then considered for downstream knock out animal studies.
57 (G3) are considered to have a contribution of all eight founder strains (G0). Mice generated from G3 are then inbred for a minimum of 20 generations at which point they are considered fully inbred. Strains generated at G23 and over are each given identifiers used to classify mice under recombination headings. Mice under each heading will then be culled and have their hind limbs removed for analysis. The use of all animals in this study was conducted in accordance with the appropriate ethics approvals and the
Australian Code for the Care and Use of Animals for Scientific Purposes.
Geniad have over 100 CC strains breeding and available beyond G15 (Morahan, Balmer et al. 2008). Genotyping of the founder strains and CC strains have been conducted via the MegaMUGA array by typing 77,808 SNPs throughout the founder genomes, to provide genome-wide coverage. The initial power calculations of the CC demonstrated that the use of 500 strains provided a detection power of 67% to detect a QTL with an additive effect of 5%. Detection power reached approximately 100% when the effect of the QTL exceeded 10% (Valdar, Flint et al. 2006). The ability to generate and test 500 strains is however unrealistic, therefore the majority of studies thus far have utilised less than 100 strains for screening and mapping loci (Levy, Mott et al. 2015, Boutilier, Taylor et al. 2017). Ram et al. has demonstrated identification of genes using small strain numbers as a proof of principle (Ram, Mehta et al. 2014). This study utilised a total number of 291 Geniad mice and 50 strains for phenotype screening. After the exclusion of poor tissue condition and strains without replicates the total numbers used for QTL mapping are 222 Geniad mice and 43 strains.
58 3.2.3 Histological analysis of mice
Fully intact femurs and tibias were processed for histological examination of the knee joints. Surrounding flesh will be removed to expose the long bones without disturbing the flesh around the knee joints or over flexing the joints. The limbs will be fixed in 10% neutral buffered formalin for 24 hours, following a rinse in PBS and being transferred to
14% EDTA for 3-5 days at 37ᵒ Celsius for decalcification. The EDTA solution was changed daily until the bones appear soft. The limbs were then rinsed in PBS and then processed using a vacuum equipped tissue processor according to the following conditions.
Reagent Time (Hours)
70% Ethanol 0.5
95% Ethanol 1
95% Ethanol 1.5
100% Ethanol 1.5
100% Ethanol 1.5
100% Ethanol 1.5
50% Ethanol/ 50% Xylene 2
Xylene 2
Xylene 2
Molten Wax 2.5
Molten Wax 2.5
Molten Wax 2.5
59 Specimens will be transferred to an embedding station where they will be placed in metal moulds and positioned for the appropriate cutting plain. Moulds will be filled with molten wax and set to cool on a cryo plate until they are completely solidified. The specimen blocks will then be sectioned on a Leica RM2155 microtome.
Each limb will be sectioned at 5µm at a frontal orientation to view calcified and articular cartilage, joint space, osteophyte formation and subchondral bone volume. Bones will be trimmed to a depth that will display the initial tibial attachment of the ACL. Sections will be collected every 40µm for observation and for retention for further staining. Sections will be collected until such depth that the Posterior Cruciate Ligament (PCL) is the dominant feature in the centre of the joint and the lateral and medial meniscus have lengthened to identify the posterior horn for both menisci.
3.2.3.1 Staining
Sections will be stained using H&E and Safranin O for microscopic analysis using the following solutions and protocol.
Reagent Time (minutes)
Xylene 2
Xylene 2
Xylene 2
100% Ethanol 1
100% Ethanol 1
100% Ethanol 1
95% Ethanol 1
60 70% Ethanol 1
Running tap water Wash
1% Acetic acid Wash
0.005% Fast Green 5
1% Acetic Acid Wash
0.5% Safranin O 5
1% Acetic Acid Wash
100% Ethanol Wash
100% Ethanol Wash
100% Ethanol 1
50% Ethanol/ 50% Xylene 2
Xylene 2
Xylene 2
Xylene 2
Coverslip and mount using DPX synthetic mounting media.
3.2.3.2 OARSI Scoring
Scoring of osteoarthritic severity was conducted using the using the Osteoarthritis
Research Society International (OARSI) method (Glasson, Chambers et al. 2010). A subset of samples was co-scored by pathologist Dr Peter Robbins (PathWest) to validate the scoring system. The classification of OA phenotypes observed in the CC according to the OARSI grading scale is shown in Table 3.1.
61 Table 3.1 Range of OARSI scores observed and represented in the CC mice Score Histology Observation 0 Normal
1 Small fibrillations without loss of
cartilage (Loss of Safranin-O
without structural changes
considered a score of 0.5)
2 Vertical clefts down to the layer
immediately below the superficial
layer and some loss of surface
lamina
62 3 Vertical clefts/erosion to the
calcified cartilage extending to
<25% of the articular surface
4 Vertical clefts/erosion to the
calcified cartilage extending to
25-50% of the articular surface
5 Vertical clefts/erosion to the
calcified cartilage extending to
50-75% of the articular surface
63 6 Vertical clefts/erosion to the
calcified cartilage extending
>75% of the articular surface
3.2.3.3 Defining the OA phenotype
The OARSI scoring allows for ease of classification of OA severity, replicable amongst multiple observers. This assessment of phenotypes is simple and effective in grades 3 – 6 as it comprises an easily measurable range of damage. The 1 – 2 ranges are more difficult to classify, as the mouse articular cartilage does not have well defined superficial and middle zones (Glasson, Chambers et al. 2010). The majority of aged individuals exhibit a phenotype in the 1 – 2 range resulting in a non normal distribution of affected against non affected strains. For the purpose of classifying the incidence of OA phenotypes per strain, a threshold of a single surface cartilage surface defect minimum score of 3 was used. This is based on grade 3 showing a definitive vertical cleft reaching the calcified layer of cartilage thus exposing the tidemark to the joint space (Table 3.1).
64 3.2.3.4 Classification and ranking of phenotypes
Each knee joint will be scored based upon the most severe phenotype observed throughout the screening region of each individual. These scores will be given based on the following phenotype observations.
Total joint score
A complete score of the assessed knee joint consisting of the combined scores of the medial tibial plateau (MTP), lateral tibial plateau (LTP), medial femoral condyle (MFC) and lateral femoral condyle (LFC) (Figure 3.2). The scored cohort will be assessed as a middle-aged cohort (ages 8-12 months), an elderly cohort (> 12 months) and a combined cohort (> 8 months).
Single surface score
An assessment of the single most severe cartilage defect present on any given cartilage load-bearing surface (Figure 3.2). The scored cohort will also be assessed as a middle- aged cohort (ages 8-12 months), an elderly cohort (> 12 months) and a combined cohort
(> 8 months).
Incidence of phenotype
An assessment of the frequency of which OA phenotypes are observed within replicates of CC strains. These will be separated into two groups, measuring incidence of any score of OA and scores greater than 3.
65 A
B
Figure 3.2 Knee joint OA phenotype classifications
A total joint score of OA is a combined score of the medial femoral condyle
(MFC), medial tibial plateau (MTP), lateral femoral condyle (LFC) and lateral
tibial plateau (LTP) (A). The highest observed score within the joint defines a
highest joint score or single surface score (B).
66 3.2.4 Haplotype reconstruction and QTL mapping
The genotypes for the eight founders were obtained via the University of North Carolina
CC website (http://csbio.unc.edu/CCstatus/index.py?run=GeneseekMM). The genotypes of the founders were separated into two sets per strain, homozygous genotypes of allele
1 and homozygous genotypes of allele 2 for their genomes to be considered haploid or fully inbred. Researchers at Geniad, led by Prof Grant Morahan inputted this data into
HAPPY software using a method called “hdesign” (Ram, Mehta et al. 2014; Mott, Talbot et al. 2000) which is designed to estimate the founder haplotype having the maximum likelihood probability for genotype sets of allele 1 and 2 independently. After verification of the haplotype data, users are able to interface with the dataset via online software called
Geneminer (http://www.sysgen.org/GeneMiner/). This software has been developed for the purpose of correlating phenotypes observed in the CC mice with their genotypes for the purpose of gene mapping.
The process of gene mapping is carried out on the Geneminer software. In this study, strains are ranked via numerical values in accordance with the observed phenotype. In this case, each strain is given the median values of OARSI scores observed in their respective phenotype classifications or percentage values given for incidence of phenotypes per CC strain. This data is inputted into Geneminer (version: 19/05/2016) to compare haplotypes of the ranked strains and identify loci that are represented by founder contributions that are unique to the highly ranked CC strains. The QTL analysis is carried out in raw data format (or “as is”) and normalized data format. Each run generates a QTL plot for each respective parameter or phenotype. The purpose of normalisation when running the QTL analysis is for the optimisation of asymmetrical data distribution. Strains scores are ranked from highest to lowest with the majority of the cohort falling within a
67 central bell curve style distribution. In cases where the data distribution is weighted towards higher or lower scores, a normalization function can be selected to clearly define the differences between high scoring strains and the remainder of the cohort (Figure 3.3).
The data output from Geneminer is represented in the form of a plot showing peaks across the genome measured in LOD values. These plots demonstrate the correlation between the founder SNP markers and the phenotype observed within the strains. This correlation produces peak(s) within chromosome(s), which are interpreted as significant based on the
LOD score and passing pre-set thresholds within the software. The thresholds for significance are calculated from 1,000 additive model permutations and set by Geniad and Prof Grant Morahan. This method was constructed based on the approach recommended by (Durrant, Tayem et al. 2011) and has been applied by groups such as
(Boutilier, Taylor et al. 2017). These thresholds are represented by two lines: a red line indicating the 95th percentile of confidence which is equivalent to p <0.05, and a yellow line representing the 37th percentile of confidence which is equivalent to p <0.63. Peaks that cross the 95th percentile are considered as genome wide significant, while peaks crossing only the 37th percentile are considered to be suggestive only.
Gene investigation is carried out by selecting QTL peaks of interest, and observing the
Geneminer output from the genomic interval selection. The output will present SNPs present from founders contributing to the selected interval. In addition to the Geneminer
SNPs, the ENCODE database will provide a more detailed selection of SNPs including the missense/regulatory effects present at the selected locus by inputting the interval and founders that are contributing to the site.
68
A B
Figure 3.3 Example of Geneminer data normalization
(A) An asymmetrical bell curve distribution representative of strain scores weighted towards the high range. (B) A normalized data set clearly defining the high and low scoring strains from the mean.
69 The Geneminer software calculates the contribution of each founder to each observed trait. This is determined by coefficients (log of odds ratio) of the fit from the logistic/multinominal regression model and using the plotting tools in the DOQTL R package (Gatti, Svenson et al. 2014). The founder effects calculated within the system are represented by p-values to represent the dominant founder contributions to each QTL.
The SNPs derived from the prominent founder/s at the QTL in question will be considered as the significant contributor to the phenotype. This founder coefficient is of importance when considering genes implicated at a QTL of interest that are contributed by many evenly distributed founders. When many founders are contributing to a single locus, there is a decreased likelihood of the genes implicated at that locus to be considered strong candidates.
3.2.5 Candidate gene verification
The investigation of candidate genes was undertaken initially using bioinformatics approaches using readily available public resources. Broad tissue and cell expression data was extracted from BioGPS and the expression profile figures were generated using
Microsoft Excel 2011 (Version 14.7.7). Expression profiles of interest showed moderate or high expression of candidate genes in bone, bone marrow, BMMs, osteoclasts, osteoblasts, macrophages and skeletal muscle. Expression in related cell lines was also considered. The National Institutes of Health online database was used to search for gene annotation information such as sequence data, position in the genome, gene structures and variance. Sequence data was also sourced through the National Institutes of Health online database for the purpose of aligning the transcribed proteins with the human equivalent proteins. EMBOSS Needle – Pairwise sequence alignment was used to
70 conduct the sequence alignment to investigate the conservation of the proteins between mice and humans to determine the likelihood of similar functionality.
The literature for each gene was reviewed using online search tools such as Pubmed and
Google Scholar. Searches were conducted for each gene or it’s known aliases, accompanied by key words such as osteoarthritis, arthritis, cartilage, chondrocyte and bone, all management of referencing was conducted using Endnote X8. A combination of both the expression and literature evidence resulted in the classification of the given gene as a potentially strong candidate. Validation of the gene within human studies identified the gene as a primary candidate, while genes fitting these criteria but had not yet been published in OA studies specifically were deemed novel candidates. All remaining genes that showed either expression in related tissues and cells or literature evidence were classified as secondary candidates for future validation studies. Genes that did not fill either of these criteria were considered as unlikely candidates.
Investigation of protein-protein interaction (PPI) networks was carried out on all candidates to observe any links between the candidate genes identified in this study with other known candidates associated with or contributing to the pathogenesis of OA. The
PPI networks were constructed using the online software package STRING (version
10.5). The network nodes represent the proteins within the network. Large nodes represent proteins with a known or predicted 3D structure and small nodes represent proteins with an unknown 3D structure. Nodes that are coloured represent first shell interactions while colourless nodes represent second shell interactions. The coloured lines linking the nodes are referred to as “edges”. These edges define meaningful interactions between proteins and can be divided into three categories: known interactions, predicted functional interactions and others. Known interactions include evidence from databases
71 and experimentally determined interactions. Predicted functional interactions include gene neighbourhood, fusion and co-occurrence. Other interactions include co-expression, homologue and association via text-mining. Further information is available at www.string-db.org (Szklarczyk, Franceschini et al. 2015).
3.2.6 Genotyping
Prof Grant Morahan’s group will maintain genotyping in house and will provide genotyping data upon request. The Mouse Universal Genotyping Array (MUGA) kit will be used to conduct the genotyping. SNP markers will be distributed with an average spacing of 325 kb (SD 191 kb) throughout the mouse genome. The markers were chosen to be maximally informative and maximally independent for the eight founder strains of the CC mice .
3.2.7 Primary Cell Isolation and Culture
Cells were cultured from a known positive OA CC mouse strain of the appropriate age cut off. A negative control will be chosen based on the haplotype data provided. The negative control was required to be of a minimum cut off age of 8 months and have a different dominant founder contributing to the QTL of interest. Bone marrow cells were isolated from the long bones (femur, tibia and humerus) of 50 week old HIP and 36 week old LAX mice. The bones were dissected from the mouse bodies and placed in sterile
PBS. The flesh was removed from the bones in a sterile culture hood and the bone marrow was flushed out of the bones with a complete α-MEM culture media (α-MEM, 10% heat inactivated FBS, 2mM L-glutamine, 100µg/mL Streptomycin and 100 U/mL Penicillin) using a 23 gauge needle. The bone marrow was then filtered through a 100µL cell strainer and centrifuged at 448x g for 15 minutes and the supernatant discarded. The cell pellet
72 was then resuspended in complete α-MEM and cultured in sterile 75cm2 tissue culture flasks at 37°C with 5%CO2 in a tissue culture incubator.
Bone marrow macrophages (BMM) were cultured via the method above with the inclusion of 10 ng/mL of M-CSF to the complete α-MEM media. Once the cells were considered 80% confluent, they were trypsinised to detach them from the culture flask and plated for osteoclast differentiation. Surplus cells were processed for long term cryopreservation and/or repeat experiments.
3.2.7.1 Tissue Isolation
Cartilage surfaces from femurs and tibias were scraped from long bones and quadriceps were removed from the dissected limbs. The tissues were placed in a sterile mortar with an aliquot of liquid nitrogen and then ground with a pestle until the tissue became a fine powder. The powder was resuspended in 1mL of Trizol, homogenised and transferred to a sterile tube and stored at -80°C until RNA extraction was performed.
3.2.7.2 Osteoclast Culture
BMM cells sourced as described in section 2.2.6 were detached from culture flasks by rinsing the cell layer in 1xPBS and incubating with 2mL of Tryple at 37°C for 15 minutes.
The Tryple was then inactivated by adding complete α-MEM and the cell layer detached using a cell scraper. The cells were then pelleted by centrifuging at 448x g for 5 minutes and resuspended in fresh complete α-MEM. The cells were counted using a Neubauer counting chamber and seeded into a 6 well plate at a volume of 8x104 per well. The cells were stimulated with 50ng/mL of RANKL over a time course for 1,3 and 5 days with a day 0 control. At day 5, the cell media was removed and the cell layer rinsed in 1xPBS.
73 The cell layers were treated with Trizol and incubated at room temperature until they detached from the plate. The Trizol cell suspension was then collected in sterile tubes and stored at -80°C until RNA extraction was performed.
3.2.7.3 Osteoblast Culture
Primary osteoblasts were isolated from the long bones in accordance with the protocol recommended by (Bakker and Klein-Nulend 2012). The bone marrow was flushed out and cut into 2-4mm pieces. The pieces were then placed in collagenase II digestion media and incubated at 37°C with vigorous shaking at 200rpm for 2 hours. After digestion, the media was inactivated with an equal volume of Dulbecco’s Modified Eagle’s Medium
(DMEM) supplemented with 10% Foetal bovine serum (heat inactivated), 2 mM L- glutamine, 100U/ml Penicillin and 100μl/ml Streptomycin (otherwise referred to as complete DMEM). The pieces were rinsed 3 times in complete DMEM and transferred to a 25cm2 tissue culture flask containing fresh complete DMEM for culturing. After 3-5 days of culture, mature osteoblastic cells were observed to be migrating from the bone pieces. The culture was maintained by changing complete DMEM media every 3 days until the cells reached 85% confluence.
The cells were then harvested by adding Tryple and incubating at 37°C to detach them from the flask surface. The Tryple was then inactivated by adding complete DMEM, the cell suspension was then pelleted via centrifugation at 1500rpm/448 x g for 5 minutes.
The cell pellet was then resuspended in fresh media and cell counts were performed on a
Neubauer counting chamber. The cells were then seeded into a 24 well tissue culture plate at a cell density of 4x104 cells per well and cultured until 90% confluence prior to differentiation. Cells were differentiated using osteogenic medium (complete DMEM supplemented with 10nM dexamethasone, 2mM β-glycerophosphate and 50μg/ml of
74 ascorbate) and were differentiated for 21 days, replacing the differentiation media every
3-4 days. Cells were harvested at day 0, 7, 14, and 21 time points by rinsing gently in 1x
PBS and treating with Trizol until the cells detached from the culture plate. Trizol cell suspensions were then collected in sterile tubes and stored at -80°C until RNA extraction was performed.
3.2.8 Molecular analysis
3.2.8.1 RNA extraction and purification (using Trizol)
The samples suspended in Trizol were thawed whilst still maintaining a low temperature by storing on ice. 200µL of chloroform was added to the tubes and were mixed vigorously for 15 seconds. The tubes were then kept on ice for 10 minutes and then centrifuged at maximum speed for 15 minutes at 4°C. 450µL of the aqueous phase was carefully transferred to a new sterile tube, to which we then added 450µL of molecular grade 70% ethanol and mixed thoroughly. We then transferred 700µL of this solution to an RNeasy
MinElute spin column. The columns were centrifuged for 15 seconds at 8000x g at room temperature. The flow through was discarded and the rest of the solution was added to the columns. The columns were then centrifuged again for 15 seconds at 8000x g at room temperature and the flow through discarded. We then added 500µL of RPE buffer to the columns and centrifuged for 15 seconds at 8000x g at room temperature, and discarded the flow through. A volume of 500µL of 80% molecular grade ethanol was then added to the columns and centrifuged for 2 minutes at 8000x g at room temperature. After discarding the flow through, the columns were transferred to new tubes and centrifuged for a further 5 minutes at maximum speed, at room temperature to dry. The flow through and tubes were discarded and the columns were transferred to sterile 1.5mL collection tubes. We then added 25µL of RNase free water to the columns and centrifuged at
75 maximum speed at room temperature to elute the RNA. The RNA was then stored at -
80°C until they were required for use.
3.2.8.2 Reverse transcription of extracted RNA
Reverse transcription of single stranded RNA extracted from cultured cells to double stranded cDNA for the purpose of conducting QPCR analysis was conducted according to the following protocol:
Reagent Volume Per Reaction (µl)
OligodT (100 µM) 0.25
Nuclease-Free water 5.25
1 µg RNA diluted in Nuclease-Free water 10
Total volume 15.5
Samples were heated for 3 minutes at 75°C to denature the RNA and allow the binding of OligodT primers to the single stranded RNA. After this process, the following reagents were added to the reaction:
Reagent Volume Per Reaction (µl)
dNTP (5 mM) 2.5
5x M-MLV RT buffer 5
M-MLV RT 1
RNasin 1
Total volume 9.5
76 To amplify the cDNA, the samples were heated to 42°C and held for 1 hour followed by further heating to 92°C for 10 minutes. The resulting cDNA was then stored at -20°C until they were required for use.
3.2.8.3 QPCR analysis
QPCR was performed to investigate the expression profile of genes implicated via the
QTL mapping process within osteoclasts (day 0, 1, 3 and 5), osteoblasts (day 0, 7, 14 and
21) and articular cartilage tissue of affected and non-affected CC mice. The primers used for each reaction are listed in section 2.1.5. The reverse transcribed cDNA was diluted
1:5 in nuclease free water to obtain sufficient quantity for QPCR analysis. Reactions were performed in triplicate according to the following protocol:
Reagent Volume Per Reaction (µl)
Forward Primer (10 µM) 0.3
Reverse Primer (10 µM) 0.3
SYBR Green Master mix 6
cDNA 5.4
Total volume 12
A master mix of SYBR green and primers was prepared, mixed thoroughly by vortexing and then loaded into an optical 384 well reaction plate. The diluted cDNA was loaded into the appropriate wells containing the SYBR green-primer master mix. The plate was then thoroughly sealed and centrifuged briefly before being placed in the ViiA™ 7 Real-
Time PCR System (Applied Biosystems). The conditions were carried out for 40 cycles of denaturation at 95°C for 15 seconds and annealing at 60°C for 60 seconds. The continuous heating steps of 95°C for 15 seconds, 60°C for 60 seconds and 95°C for 15
77 seconds were conducted to analyse the melt curve of the PCR products. The real time
PCR system was programmed as follows:
PCR Stage Melt Curve Stage Holding Stage (x40 cycles) (Continuous)
50oC for 120 sec, 95oC for 15sec, 95oC for 15 sec, 95oC for 600 sec 60oC for 60 sec 60oC for 60 sec, 95oC for 15 sec
3.2.8.4 Real time PCR analysis
The real time-PCR data (CT values) was collected using the ViiA™ 7 software (Applied
Biosystems). Melting curve for the PCR products were analysed to verify the specificity of the PCR primers. The normalized fold expression (Fold change) of genes were obtained relative to the house keeping genes (Hprt and β-actin). The fold change was calculated using the Livak’s ΔΔCT method (Livak and Schmittgen 2001). The Livak’s equation is described as follows:
Fold change = 2-ΔΔCT
CT =
Mean of CT values
ΔCT =
CT of gene of interest of Sample A (Test) - CT housekeeping gene
ΔΔCT =
ΔCT of gene of interest of Sample A (Test) - ΔCT of gene of interest of Sample B
(Control)
78
The Livak’s method can be used to compare the gene expression of two different samples
(e.g. A and B) relative to the housekeeping gene(s). In the case of this study: Sample A represents the OA affected strain, while Sample B represents the non-affected control strain. The gene expression of Sample A is presented as fold change relative to Sample
B. All analyses were conducted in technical triplicates. Due to the very limited access to numbers of appropriately aged strains and poor viability of the cells harvested from aged
CC mice, reliable biological replicates were not able to be tested.
3.2.9 Statistical analysis
Numerical data was entered into Microsoft Excel and statistical analyses were conducted using Excel. Figures were generated using Graphpad® Prism (version 7).
For in vitro QPCR experiments, all are presented as the mean ± SD of the values obtained from the technical replicates. Statistical significance was determined via ANOVA. A P- value of less than 0.05 was considered to be significant.
79 Chapter 4 Characterization of Osteoarthritic phenotype using OARSI histologic grading scale
80 4.1 Introduction
The Collaborative Cross (CC) mice are a collection of recombinant inbred strains derived from eight founder strains (A/J, C57BL/6J, 129S1/ScImJ, NOD/LtJ, NZO/HiLtJ,
CAST/EiJ, PWK/PhJ and WSB/EiJ), for the purpose of capturing the majority of variance observed in the mouse genome. Inbred strains generated from random genetic contributions from these eight founders, via a funnel-breeding scheme are given unique identifiers. Screening these strains for phenotypes of interest has proven to be effective in identifying genetic traits responsible for diseases (C.C. Consortium 2012).
Strains generated within the CC exhibit a range of disease phenotypes, including cartilage degradation and Osteoarthritis (OA). Whilst large cohort studies such as human GWAS have proven to be useful at identifying genes that may affect OA development and disease progression, the significant contribution of external factors such as diet, activity and lifestyle reduce the reliability of identifying purely genetically influenced, spontaneous forms of OA. The CC mouse strains are generated under uniform conditions including diet and housing, therefore increasing the reliability of mapped genes to observed OA phenotypes (Churchill, Airey et al. 2004).
Characterization of OA phenotypes in mouse models according to (OARSI) can be interpreted as total joint phenotype score and individual joint surface scores.
Implementing the OARSI scoring system to characterize the total joint can easily differentiate between mild OA sufferers, who may have a single lesion present and individuals with multiple lesions, indicating a more severe disease process. Total joint scores can be problematic when comparing multiple low-grade lesions with a higher-
81 grade lesion on a single surface. From a phenotype categorization perspective, these two
OA scores represent different disease processes and thus should be investigated independently, as different phenotypes.
4.2 Screening and grading of OA phenotypes using the OARSI method
Characterization of OA phenotypes was conducted via histological grading. The left hind limb of each animal was sectioned frontally with sections collected every 40μm from the beginning of the tibial insertion of the ACL. Sections were collected until both the medial and lateral posterior horn of the meniscus began to extend to the center of the knee, indicating the primary load bearing region had been passed. Sections were observed via
Safranin O and Fast Green staining and graded according to the OARSI grading scale for the mouse knee joint (Glasson, Chambers et al. 2010). Sections presenting with the most severe observed OA phenotype were then used to represent the grade for that individual.
The aim of this study was to characterize an OA phenotype within the CC cohort. The phenotype was broken up into traits that best represent severity and incidence of OA. The traits that were analyzed were a total joint score, single cartilage surface score and percentage incidence of OA. A threshold of individuals aged greater than 8 months was applied to segregate the middle to late aged population of the CC mice, therefore providing the most significant cohort to observe age related OA. This cohort consisted of
43 strains and a total of 222 individuals. Each strain was required to contribute a minimum of 3 individuals to be included in the study.
82
4.3 Analysis of total joint OA in CC mice >8 months of age
OA affected strains were observed variably across the CC population
Total joint OA phenotypes within the CC cohort showed a spread of OA severity scores that varied from strain to strain, as well as some variability of individuals within strains.
Within the strains JUD, PEF, POH, HAX2 and MAK every individual analysed showed some degree of OA. Of the strains that met the inclusion criteria, 11.62% showed OA in all individuals. Strains JEUNE and HIP showed the highest mean values while WOB2 and DET3 showed no evidence of OA (ie. An OARSI score of 0) in any of the individuals.
PEF showed the highest total joint OARSI score with a mean value in the top 10% of scores (Figure 4.2).
83
condyle
femoral
lateral
(MTP),
plateau
tibial
medial
(MFC),
condyle
example
femoral
calculation
medial
the
score
of
and
score
joint
knee
(LTP)
combined
a
total
is
of
plateau
OA
of
tibial
score
Assessment
lateral
4.1 joint
and
total
Figure A (LFC)
84
and
each
scores
of
box
joint
scores
using
total
ed
n=3.
The
of
ys
than
anal
less
method.
then
Distribution
were
grading
population
strains.
a
strain
OARSI
per
with
the
between
ng
strains
usi
joint
individuals
of
populations
total
identifies
assessed
the
was
of
compare
asterisk
Populations
and
LTP)
Red
grading
and
strain
range).
LFC
ARSI
individual.
each
O
P,
that
via
MT
within
for
(MFC,
score
measured
as
distribution
joint
minimum/maximum
surface
and
total
a
cartilage
population
give
phenotypes
the
to
percentile
each
OA
on
of
25/75
observe
OA
n,
to
combined
of
Spread
(mea
plots
were
4.2
severity
strain
Figure The surface whisker per
85 4.4 Analysis of most severe single surface OA phenotype in CC mice, >8 months of age
To properly differentiate between a joint that was mildly affected on all surfaces and a single surface that was considered intermediate or severely affected, individuals were scored within strains by the highest observed OARSI score on any of the cartilage surfaces (Figure 4.3). Strains JEUNE and HIP once again showed the highest mean values, PEF and MAK showed the highest observed scores with mean values in the top
25% of scores (Figure 4.4).
86
surface.
cartilage
any
on
recorded
score
example
highest
calculation
single
the
score
by
and
defined
joint
is
knee
score
surface
surface
single
single
of
or
score
joint
Assessment
4.3
highest
Figure The
87
per
scoring
whisker
scores
and
highest
joint
box
The
total
using
n=3.
of
n
ed
tha
method.
analys
less
surface
then
Distribution
grading
joint
were
population
strains.
a
OARSI
single
strain
the
with
per
between
scoring
using
st
strains
highe
individuals
assessed
of
the
populations
identifies
was
of
LTP)
asterisk
compare
grading
Populations
and
and
Red
LFC
OARSI
strain
range).
individual.
via
MTP,
each
that
for
(MFC,
within
measured
score
as
surface
joint
minimum/maximum
a
lage
distribution
and
give
carti
phenotypes
to
each
OA
used
population
on
percentile
of
the
was
OA
25/75
of
Spread
surface
observe
4.4
(mean,
to
severity
Figure The cartilage plots strain
88 4.5 Analysis of OA incidence in CC mice >8 months of age
We observed that there were strains that showed consistent OA incidence. Using the same inclusion criteria of a minimum of n=3 individuals per strain contribution, we assessed the percentage incidence of OA phenotypes within strains of the CC population. Strains
JUD, MAK, POH, HAX2 and PEF all showed a 100% incidence of OA regardless of severity, accounting for 11.63% of the CC population observed. We observed that 62.79% of the CC cohort registered an incidence of OA above 50% regardless of severity (Figure
4.5). When we apply thresholding to observe only intermediate to severe scores (OARSI scores >3), we observed HIP and JEUNE as having 66.66% incidence of OA with a single surface OARSI score greater than 3. We also observed 4.65% of the CC cohort registered incidence of OA above 50% with a single surface OARSI score greater than 3. This analysis also revealed 34.88% of the CC cohort demonstrated some incidence of OA with an OARSI score greater than 3 within strains (Figure 4.6).
89
populations
CC
in
measured
core
s
OA
any
of
Percentage
4.5
Figure
90
populations
CC
in
measured
>3
e
scor
joint
single
OA
of
Percentage
4.6
Figure
91 4.6 Analysis of total joint and single surface OA in CC mice >12 months of age.
Our analysis showed at times, a wide variation between individuals within any given strain. We hypothesized that this may be due to a wider range of ages displaying OA phenotypes at different rates of progression. Therefore, we segregated the population of individuals that were greater than 12 months of age to observe the strains that were displaying OA phenotypes at older ages. This population contained 32 strains with a total of 128 individuals that met the modified inclusion criteria of n= >2 individuals per strain, to account for the reduced numbers of replicates.
We applied the same principal of observing total joint OA scores and highest single surface scores as found in figure 4.1 and 4.3. We observed HIP to have the highest mean score, with SAT, MEE and BOON all not registering an OA score. The highest observed individual score in the elderly cohort was of the MAK strain, at a total joint score of 11.
Due to the large separation from the mean value, this individual is likely to be an outlier.
Excluding this data point, the highest scoring individual was of the HIP with a score of 9.
Interestingly, 25% of the elderly CC strains had every individual register an OARSI score above 0 (Figure 4.7).
Investigation into the highest single surface scores revealed that both HIP and MOP were the only strains to present with an OARSI score greater than 3, accounting for 6.25% of the CC strains (Figure 4.8).
92
a
with
score
strains
joint
total
identifies
the
of
asterisk
grading
Red
OARSI
range).
via
measured
as
minimum/maximum
mice,
and
CC
of
percentile
population
25/75
month
(mean,
>12
in
strain
per
scores
phenotypes
OA
joint
n=2.
of
total
than
of
Spread
less
4.7
Figure Distribution population
93
less
score
population
surface
a
single with
strains
highest
he
t
of
identifies
grading
asterisk
Red
OARSI
via
range).
measured
as
mice,
minimum/maximum
CC
and
of
percentile
population
25/75
month
(mean,
>12
in
strain
per
phenotypes
scores
OA
joint
of
total
of
Spread
4.8
n=2.
94 Figure Distribution than 4.7 Discussion
We have successfully characterized OA phenotypes within the CC population using the
OARSI grading method. We observed a range of OA severity within and between strains, with several strains that had the majority of their population showing OA phenotypes that we described as severe. This study further solidifies the concept that spontaneous age related OA has a genetic component as the phenotype has been observed to be associated with individuals that share the same genetic lineage. Our study is of comparable statistical power in regards to sample size (220 vs 222) to the GWAS style study performed by
(Blanco, Möller et al. 2015). The advantages of the CC cohort enable us to make these observations without the influence of individual lifestyle variables that have been reported to have an effect on the development of OA (Litwic, Edwards et al. 2013).
We have concluded that when measuring OA in a population, both frequency and severity must be considered as separate entities and therefore separate disease processes. The severity of OA as measured by the OARSI grading scale is of particular interest, as we observed when assessing the incidence of any OA across the CC population, all strains except for DET3 and WOB2 showed a measurable OA phenotype. This raises questions regarding individuals exhibiting mild grade phenotypes being characterized as OA sufferers in phenotype characterization studies.
We observed that 95.35% of the CC cohort achieved an OARSI grade, though 63.41% of those failed to exceed a score of 2. According to Blanco et al (Blanco, Möller et al. 2015), achieving a grade of 2 or 3 on the Kellgren-Lawrence scale of 0 to 4 deemed the participants eligible for a study to investigate SNP association with spontaneous OA.
Using a similar concept in the OARSI grading scale, any individual that was observed to
95 have a vertical cleft extending to the calcified layer (grade 3 or over) was considered a significant OA sufferer due to the observable loss of cartilage. The CC model of spontaneous OA exhibits comparable cartilage degradation as described by Glasson et al and is representative of cartilage degradation of end stage OA in humans (Glasson,
Chambers et al. 2010).
This assessment of OA severity also signifies the importance of the comparison between total joint scores and single highest surface scores. A single surface score of 3 or greater signifies a more severe disease state when compared to a total joint score of the same value that may be due to a combination of scores not exceeding 1 or 2. The difference between these is observed predominantly with the strains JUD and POH, which both showed score reductions that reduced their ranking against other high scoring strains.
However, the most significantly affected strains JEUNE and HIP were not affected by the alternative scoring method. As described by Glasson et al, the mouse OA phenotype is rarely characterized in a mild form. This suggests that the single surface scoring method is more applicable to differentiating multiple mild OA phenotypes from a single but significant lesion (Glasson, Chambers et al. 2010).
When assessing the difference in aged mice, strains that scored in the top 10% of severity that had candidates in both the 8-12 month and the >12 month cohorts, showed a narrower distribution of scores when observing the >12 month cohort alone. These results suggest that with an increase in age, mice from the same genetic lineage present more consistently with cases of severe OA. However, the >12 month cohort study was limited by the available candidates (n=128) and strains JEUNE and PEF that presented with severe OA in the 8-12 month cohort did not have >12 month individuals available for screening.
96 Continuing the breeding of strains and a focus on culling animals beyond 12 months would increase the potential statistical power of this study.
The reported female gender bias for OA development (Srikanth, Fryer et al. 2005; ABS
2015), is also noteworthy, as the ability to source particular gender and strains of certain age groups was limited in our study. With continued breeding of the CC strains that showed interesting phenotypes, a future gender comparison study would be achievable.
A full list of strain numbers assessed in this study is available in the appendix
(Supplementary Figure 9.1).
The utilization of the CC mice for spontaneous OA genetic research, in favor of human
GWAS style approaches has proven to be advantageous due to the greater flexibility in genetic and environmental manipulation and the ability to readily source tissue for phenotype investigation. A wide range of mammals develop OA like symptoms, including the mouse (Glasson, Chambers et al. 2010). While some of the larger animal models better simulate human like OA progression, the colony management, genetic manipulation and time to disease progression can be undesirable for many researchers.
The laboratory mouse has a large reproductive capacity along with a short gestation period of only 3 weeks, and an easily manipulated housing and environment (Jilka 2013).
The laboratory mouse is also capable of developing spontaneous OA within the ages of
9-12 months (Poole, Blake et al. 2010), making it suitable for a genetic based spontaneous
OA study.
In summary, the CC mouse platform has proven to be beneficial in generating strains with genetically influenced spontaneous OA phenotypes. As we have observed OA
97 development in our cohort of CC mice in the form of cartilage degradation, we may also investigate other parameters that are involved in the OA disease process. Investigation of subchondral bone changes, osteophyte formation and quadriceps muscle mass may all be observed in future studies of this cohort. Finally, the addition of the ability to investigate the genetic loci that are uniquely associated with each affected strain will further our understanding of the OA disease process.
98
Chapter 5 Quantitative Trait Locus Haplotype Mapping of Osteoarthritis Phenotypes in the Collaborative Cross
99 5.1 Introduction
Genome Wide Association Studies (GWAS) in humans have been successful in identifying multiple genetic loci associated with heritable diseases such as OA (Zengini,
Hatzikotoulas et al. 2018; Valdes and Spector 2008; Blanco, Möller et al. 2015; arcOGEN Consortium and arcOGEN collaborators 2012; Hackinger, Trajanoska et al.
2017). These studies are limited by their ability to control external factors that may have an effect on the disease state and progression.
The Collaborative Cross founders and the strains generated from the funnel-breeding scheme have a wide variation of physical attributes and phenotypes (C.C. Consortium
2012). Identification of over 38 million SNPs and indels within the founder strains have enabled researchers to efficiently track their heritability within the CC cohort (Munger,
Raghupathy et al. 2014). This has proven to be useful in identifying genetic loci that have been uniquely inherited by strains affected by disease while adequately controlling external factors that may contribute to the development of complex phenotypes such as
OA.
As shown in chapter 4, the CC mice were categorized by severity of total joint and most severe single surface OA in a >8 month cohort and a >12 month cohort. Strains were also categorized by the reported incidence of OA per strain and incidence of OA scores >3 per strain. Using the median values of the OARSI scores and the percentage values of the incidence studies, the strain scores were ranked and input into mapping software to identify genetic loci that are associated with the highly ranked strains.
100 Using quantitative trait locus (QTL) mapping software (www.sysgen.org/GeneMiner), haplotypes of ranked strains were compared to identify loci that showed similar founder contributions across the highly ranked strains whilst being significantly different from low ranking strains. These loci are represented as peaks and can be interpreted by how prominent the peaks are compared to the rest of the genome. GWAS studies utilize a fixed p-value threshold to identify the significance of variants that are identified. Such thresholds are used commonly in QTL mapping, though they vary based on the sample size used. These thresholds are presented as percentile ranges which are representative of p-values, which can be desirable when variable sample sizes are being used to map loci on the same platform. For this study, we used two thresholds set by the Geneminer mapping software.
We applied the methods for QTL peak significance demonstrated by Boutilier, Taylor et al and Durrant et al describing the use of the thresholds presented by the Geneminer software (Boutilier, Taylor et al. 2017; Durrant, Tayem et al. 2011). Yellow represents a suggested significance above the 37th percentile (p < 0.63) and red represents genome wide significance above the 95th percentile (p < 0.05). The most reliable QTL peaks were identified by the lowest p-values, represented by the thresholds the peaks pass, a high
LOD value, few dominant founder strains contributing to the locus and ideally narrow genomic intervals. Reliability of peaks that cross the 37th percentile threshold were considered to be suggestive. However, genes implicated at these sites are still considered interesting and may be worthy of further study. A normalization function was also used to allow Geneminer to rank the highest scoring strains against the remaining cohort more precisely in circumstances where there were greater numbers of strains scoring in the middle to high ranges. Genes were identified at each locus using SNPs categorized by the
101 MegaMUGA Illumina array, which is used to identify 77,808 SNP markers throughout the mouse genome. These SNPs were used to identify which founders contribute to the locus as well as categorize the genes that are present. Additionally, other databases such as ENCODE will be used to provide an in depth assessment of the SNPs present in regions of interest.
5.2 Results
5.2.1 Median total joint OA in the >8 month cohort
We conducted QTL analysis of median total joint OA in the >8 month cohort to investigate the genes associated with a broad spectrum of OA sufferers of the complete
CC cohort, reflecting a standard population of middle to late aged individuals. We observed a locus on chromosome 19 that was derived from the median values of total joint OA scores per strain. The locus, identified by the peak in figure 5.1, was above the
37th percentile threshold between 27.82 and 31.51 Mbp on chromosome 19. SNPs identified via the Geneminer database in this region implicated the following genes: Rfx3,
Jak2, Cd274, Pdcd1lg2, Mlana, Il33, Uhrf2, Gldc, Mbl2 and Dkk1. The CAST founder was the significant contributor to this locus.
Using this data set, we utilized the normalize function on the Geneminer database. We observed the same locus on chromosome 19 appear though the peak passed the 95th percentile threshold (Figure 5.2). While there were similarities in the genes identified when using the normalization function, we observed the SNPs identified via the
Geneminer database were different when compared to the non-normalized mapped loci.
The genes implicated via the normalized dataset were Rln1, Cd274, Pdcd1lg2, Ermp1,
Ranbp6, Il33, Uhrf2 and Gldc. The CAST and PWK founders were the significant
102
Figure 5.1 QTL mapping of median total joint score >8 months.
(A) QTL plot showing a peak crossing the 37th percentile on chr 19. (B) Expansion of chr
19 showing genes between 27.82 and 31.51 Mbp.
103
Figure 5.2 QTL mapping of median total joint score >8 months normalized.
(A) QTL plot showing a peak crossing the 95th percentile on chr 19. (B) Expansion of chr
19 showing genes between 27.82 and 31.51 Mbp.
104 contributors to this locus. Investigation of other SNPs between 27.82 and 31.51 Mbp on chromosome 19 using ENCODE revealed Glis3 as another gene located in this locus.
However, the fact that this gene was not implicated via Geneminer means its significance should be interpreted with caution.
5.2.2 Median highest single surface measured OA in the >8 month cohort
We conducted QTL analysis of median highest single surface OA in the >8 month cohort to investigate the genes associated with the most severe OA defect of the complete CC cohort, reflecting a standard population of middle to late aged individuals. This analysis reflects the rapidly progressing disease state which differs from low scoring OA across many surfaces of the synovial joint. Using the median values of the highest single surface
OA scores of CC strains, we observed that no peaks reached the 37th percentile threshold.
However, there was a prominent peak on chromosome 19 that reached an LOD of 6.9
(Figure 5.3).
Using the Geneminer normalize function on this dataset, we observed a peak on chromosome 19 representing a locus located between 27.82 and 31.51 Mbp. The locus crosses the 95th percentile threshold only between 28.33 and 29.26 Mbp. SNPs identified via the Geneminer database implicated the following genes at this site: Glis3, Rln1,
Cd274, Pdcd1lg2, Ermp1, Ranbp6, Il33, Uhrf2 and Gldc. The CAST and PWK founders were the significant contributors to this locus, with the exception of Glis3, which was represented by B6, NOD and NZO (Figure 5.3).
105 Figure 5.3 QTL mapping of median single surface joint score >8 months.
(A) QTL plot showing no peaks above the 37th percentile. (B) Normalised QTL plot showing a peak crossing the 95th percentile on chr 19. (C) Expansion of chr 19 showing genes between 27.82 and 31.51 Mbp.
106 5.2.3 Median total joint OA in the >12 month cohort
We conducted QTL analysis of median total joint OA in the >12 month CC cohort to investigate the genes associated with a late stage development of OA. We observed two peaks crossing the 37th percentile threshold on chromosome 2 between 174.44 and 181.54
Mbp and on chromosome 17 between 86.62 and 91.75 Mbp. The Geneminer database implicated the following genes at the locus on chromosome 2: Cdh26, Cdh4, Psma7,
Ss18l1, Gtpbp5, Adrm1, Lama5, Rbbp8nl, Gata5, Ogfr, Col9a3, Tcfl5, Dido1, Ythdf1,
Nkain4, Arfgap1, Col20a1, Ppdpf, Ptk6, Helz2, Rtel1, Zgpat, Abhd16b, Tpd52l2, Uckl1 and Myt1. The PWK and WSB founders were the significant contributors to this locus
(Figure 5.4). The Geneminer database implicated the gene Epas1 at the locus on chromosome 17, with the WSB founder as the significant contributor (Figure 5.5).
When using the normalize function, we observed that the peaks on chromosome 2 and 17 disappeared and a peak on chromosome 4 crossed the 37th percentile between 51.74 and
62.95 Mbp. The Geneminer database implicated Smc2 at this site. The founders AJ and
129s were the significant contributors to this locus (Figure 5.6). Using the ENCODE database to investigate other SNPs that have been categorized between 174.44 and 181.54
Mbp on chromosome 2 revealed the following additional genes of potential interest:
Zfp831, Edn3, Etohi1, Zfp931, Phactr3, Sycp2, Fam217b, Taf4a, Hrh3, Osbpl2, Rps21,
Cables2, Gata5os, Slco4a1, Ntsr1, Mrgbp, Gid8, Slc17a9, Bhlhe23, Birc7, Chrna4,
Kcnq2, Eef1a2, Srms, Gmeb2, Stmn3, Arfrp1, Lime1, Slc2a4rg-ps, Zbtb46, Dnajc5,
Samd10, Prpf6, Tcea2, Rgs19 and Oprl1. The ENCODE database revealed the following additional genes on chromosome 17 between 86.62 and 91.75 Mbp: Tmem247, Rhoq,
Socs5, Kcnk12 and Ppp1r21. From the normalized analysis, ENCODE SNPs that have
107
Figure 5.4 QTL mapping of median total joint score >12 months for chromosome
2.
(A) QTL plot showing peaks crossing the 37th percentile on chr 2 and chr 17. (B)
Expansion of chr 2 showing genes between 174.44 and 181.54 Mbp.
108
Figure 5.5 QTL mapping of median total joint score >12 months for chromosome
17.
(A) QTL plot showing peaks crossing the 37th percentile on chr 2 and chr 17. (B)
Expansion of chr 17 showing genes between 86.62 and 91.75 Mbp.
109
Figure 5.6 QTL mapping of median total joint score >12 months normalized.
(A) QTL plot showing a peak crossing the 37th percentile on chr 4. (B) Expansion of chr
4 showing genes between 51.74 and 62.95 Mbp.
110 been categorized between 51.74 and 62.95 Mbp on chromosome 4 revealed no additional genes of potential interest.
5.2.4 Median highest single surface measured OA in the >12 month cohort
We conducted QTL analysis of the highest single surface OA measured in the >12 month
CC cohort to investigate the genes associated with a late stage development of single severe OA lesions. Using the median values of the highest single surface scores of the
>12 month old CC mice, we observed the same peak on chromosome 17 between 86.62 and 91.75 Mbp that we observed in the >12 month total joint analysis. This locus crosses the 37th percentile threshold, the Geneminer database implicated Epas1 at this site. The
WSB founder was the significant contributor to this locus (Figure 5.7). In addition, we also observed a peak on chromosome 4 that was observed in the normalized total joint mapping in section 5.2.3, this time between 51.74 and 58.61 Mbp. In addition to Smc2,
Geneminer implicated Lpar1 at this site with AJ, and 129s as the significant founders.
We also observed an additional peak between 63.07 and 63.92 Mbp on chromosome 4, where the Geneminer software implicated no genes (Figure 5.8).
Using the ENCODE database, we observed the following additional genes of interest between 86.62 and 91.75 Mbp on chromosome 17: Prkce, Epas1, Rhoq, Pigf, Mcfd2,
Ttc7, Calm2, Epcam, Msh2, Kcnk12, Msh6, Fbxo11, Foxn2, Ston1, Fshr and Nrxn1. We also observed the following genes between 51.74 and 58.61 Mbp on Chromosome 4:
Actl7b, Ctnnal1 and Ptpn3. Mup20 and Rgs3 were also observed in the region between
63.07 and 63.92 Mbp.
When using the normalize function, we observed a similar peak as reported in section
5.2.3, on chromosome 4 between 51.74 and 100.54 Mbp. The Geneminer database
111
Figure 5.7 QTL mapping of median single surface joint score >12 months for chromosome 17.
(A) QTL plot showing peaks crossing the 37th percentile on chr 4 and chr 17. (B)
Expansion of chr 17 showing genes between 86.62 and 91.75 Mbp.
112
Figure 5.8 QTL mapping of median single surface joint score >12 months for chromosome 4.
(A) QTL plot showing peaks crossing the 37th percentile on chr 4 and chr 17. (B)
Expansion of chr 4 showing genes between 51.74 and 58.61 Mbp.
113 implicated the following genes at this site: Smc2, Mpdz, Frem1, Haus6, Plin2 and
Dennd4c. The founders AJ, 129s and NZO were the significant contributors to this locus
(Figure 5.9). In addition, we observed a peak on chromosome 15 between 74.65 and 86.78
Mbp. The Geneminer database implicated the following genes at this site: Tmprss6,
Sh3bp1, Fam227a, Sun2, Cbx6 and Tab1. The founders NZO, PWK and WSB were the significant contributors to this locus (Figure 5.10).
Using the ENCODE database, we observed the following additional genes of interest on chromosome 4 between 51.74 and 100.54 Mbp: Actl7b, Actl7a, Ikbkap, Ctnnal1,
Tmem245, Frrs1l, Epb4.1l4b, Susd1, Mup8, Mup20, Rgs3, Ptprd, Mpdz, Fam154a,
Saxo1os, Rraga, Scarna8, Haus6, Plin2, Dennd4c, Rps6, Acer2 and Slc24a2. The following genes were also observed on chromosome 15 between 74.65 and 86.78 Mbp:
Col14a1, Zhx2, Zfat, Il2rb, Pla2g6, Tmem184b, Cby1, Dnalc4, Nptxr, Nptcd, Cbx7,
Syngr1, Slc25a17, Xpnpep3, Ttll1, Ttll12, Scube1, Mpped1, Atxn10, Celsr1 and Gramd4.
114
Figure 5.9 QTL mapping of median single surface joint score >12 months for chromosome 4, normalized.
(A) QTL plot showing peaks crossing the 37th percentile on chr 4 and chr 15. (B)
Expansion of chr 4 showing genes between 51.74 and 100.54 Mbp.
115
Figure 5.10 QTL mapping of median single surface joint score >12 months for chromosome 15, normalized.
(A) QTL plot showing peaks crossing the 37th percentile on chr 4 and chr 15. (B)
Expansion of chr 15 showing genes between 74.65 and 86.78 Mbp.
116 5.2.5 Median total joint score of middle aged 8-12 month cohort
As we have observed in the present study, there are clear differences in loci that are mapped when segregating the >12 month cohort for mapping individually. We applied the same segregation principle and mapping techniques to identify traits that may be unique to an early onset of OA. The segregation allowed the use of 24 strains (n=94) and median values of total joint scores were used to map the loci.
We observed two prominent peaks, one on chromosome 17 between 33.64 and 45.38 Mbp and chromosome 19 between 27.82 and 28.33 Mbp. Both of these peaks crossed the 37th percentile threshold (Figure 5.11). The Geneminer database implicated Rfx3 and Glis3 at the locus on chromosome 19 with the CAST founder being the significant contributor. At the chromosome 17 locus, the following genes were implicated: Esp8, Cdc5l, Ptchd4,
Myo1f, Daxx, Wdr46, Col11a2, H2-DMb2, H2-DMb1, Tap2, H2-Ab1, H2-Aa, H2-Eb1,
Btnl6, Nelfe and Trim15. The founders B6, 129s and CAST were the significant contributors to this locus.
When we utilized the normalize function, we observed that the locus on chromosome 17 crossed the 95th percentile threshold. The same sets of genes were implicated with the exception of Esp8, Cdc5l and Ptchd4 and the same founders contributed to the loci. The locus on chromosome 19 remained between the 37th and 95th percentile (Figure 5.12).
However the number of genes implicated by Geneminer at this site increased with the addition of Jak2, Cd274, Pdcd1lg2, Mlana, Il33, Uhrf2, Gldc, Mbl2 and Dkk1. The CAST founder remained the significant contributor to the locus. We investigated the locus on chromosome 17 using the ENCODE database and revealed the following additional genes of potential interest:
117
Figure 5.11 QTL mapping of median total joint score 8-12 months.
(A) QTL plot showing two peaks crossing the 37th percentile on chr 17 and chr 19. (B)
Expansion of chr 19 showing genes between 27.82 and 28.33 Mbp.
118
Figure 5.12 QTL mapping of median total joint score 8-12 months normalized.
(A) QTL plot showing two peaks. On chr 17 crossing the 95th percentile and on chr 19 crossing the 37th percentile. (B) Expansion of chr 17 showing genes between 33.64 and 45.38 Mbp.
119 Adamts10, H2-K2, H2-Ke6, Rxrb, H2-Eb2, H2-Ea-ps, Bntl5-ps, Bntl7-ps, H2-Q5,
Trim39, Trim26, Trim10, Rcan2, Clic5 and Supt3.
120 5.2.6 Incidence of any OA grade in the >8 month cohort
As we observed in the previous study, the frequency at which individuals develop OA varies from strain to strain in the CC cohort. Strains were given scores based on the percentage of any grade of OA observed to investigate the disease penetrance per strain.
These values were used to map the loci associated with higher incidence of disease. We then applied a threshold of OARSI score >3 to investigate the loci associated with incidence of a severe OA disease state.
We observed three peaks, representing three loci that crossed the 37th percentile threshold.
The peaks appear on chromosome 2 between 3.34 and 8.06 Mbp, chromosome 11 between 55.73 and 61.79 Mbp and chromosome 19 between 27.82 and 30.72 Mbp (Figure
5.13 and 5.14). The locus on chromosome 2 revealed no genes via the Geneminer database. When we utilized the ENCODE database, Dclre1c, Suv39h2 and Hspa14 were identified as genes of potential interest.
The locus on chromosome 11 revealed the following genes via the Geneminer database:
Larp1, Cnot8, Gemin5, Igtp, Irgm2, Zfp672, Lypd8, Trim58, Zfp39, Btnl10, Trim17,
Obscn, and 8 Olfr family genes (Figure 5.13). We used the ENCODE database to investigate genes of potential interest at the site on chromosome 11 and identified the following additional genes: Nmur2, Gria1, Mfap3, Galnt10, Mrpl22, Zfp692, Sh3bp5l,
Fam183b, Rnf187, Hist3h2ba, Trim11, Guk1, Tom1I2 and Lrrc48. The locus on chromosome 19 revealed only Glis3 via the Geneminer database (Figure 5.14).
We then mapped the percentage values of OA incidence with scores >3 and observed two peaks crossing the 37th percentile threshold. These peaks appear on chromosome 8
121
Figure 5.13 QTL mapping of OA incidence in the >8 month cohort for chromosome 11.
(A) QTL plot showing three peaks crossing the 37th percentile, on chr 2, chr 11 and chr19. (B) Expansion of chr 11 showing genes between 55.73 and 61.79 Mbp.
122
Figure 5.14 QTL mapping of OA incidence in the >8 month cohort for chromosome 19.
(A) QTL plot showing three peaks crossing the 37th percentile, on chr 2, chr 11 and chr19. (B) Expansion of chr 19 showing genes between 27.82 and 30.72 Mbp.
123 between 120.39 and 123.90 Mbp and chromosome 19 between 33.45 and 33.80 Mbp
(Figure 5.15). The locus on chromosome 8 implicated Cdh13 via the Geneminer database.
The locus on chromosome 19 revealed no genes via the Geneminer database.
124
Figure 5.15 QTL mapping of OA incidence with a score >3, in the >8 month cohort.
(A) QTL plot showing two peaks crossing the 37th percentile, on chr 8, and chr19. (B)
Expansion of chr 8 showing genes between 120.39 and 123.90 Mbp.
125 5.3 Discussion
The focus of this study has been to further investigate the genetic regulators of OA within the CC mouse cohort. Using the categorically defined phenotypes that we identified in the previous chapter, we successfully identified seven genetic loci across six chromosomes that are uniquely associated with these disease states, two of which passed the 95th percentile threshold, implicating genome wide significance. More than 90 genes have been discovered in this study, with 22 included in the genome wide significance category (Table 5.1).
Throughout this study, the locus on chromosome 19 was recurrently identified with the exception of any mapping conducted on the >12 month threshold data set. This suggests that the genes found in this locus are all considered to be worthy of further study in OA research. The genes identified by the Geneminer database at this locus varied slightly, as the locus varied in size between each mapping data set. The variation in the genes discovered at this locus suggests the genetic differences that may exist between each of the phenotype assessment parameters, particularly which combination of genes may be associated with the regulation of the phenotype in question. For example, Glis3 is present in the chromosome 19 locus only when mapping is conducted from the incidence and the
8-12 month cohort perspective. This suggests that Glis3 may be associated with the initiation of early onset OA.
The additional loci observed when mapping the percentage incidence dataset are interesting as they accompany the commonly identified locus on chromosome 19,
126 suggesting that the genes that they contain may have an effect on an individual’s risk of developing OA when accompanied with the effects of the chromosome 19 locus.
The genes that were identified by Geneminer were assigned p-values that reflected their significance within the locus they are found. These p-values occasionally differed between genes but in many cases they remained the same across all genes found at the locus. Genes with the highest p-values may be considered more viable candidates for further studies, however all genes implicated above any given threshold are considered to be of importance and require further investigation and validation.
The use of normalization of data sets via Geneminer in some cases increased the significance of already visible loci that may be affected by the prominence of phenotypic outliers present in strains not contributing to the locus crossing the 37th percentile. While the locus did not change after normalization, the genes that were present varied slightly.
However, the significance of genes that were still present post normalization had increased. We concluded that the normalization function was beneficial in increasing the significance of some loci. Comparing strains that scored the highest against the remainder of the cohort resulted in a more significant locus that often contained fewer genes.
Although the standard mapping resulted in loci that may not have reached the GWS threshold, the inclusion of additional genes that may have lower effects on the phenotype should be considered in conjunction with the normalized data sets.
In cases of loci that were observed to be significant but contained no genes according to
Geneminer, the use of the ENCODE database was used to observe the genes present at the site. The lack of genes presented by Geneminer may be due to the number of SNPs
127 available in Geneminer’s database that may not provide sufficient coverage at the site of interest or a potential false positive mapping. The use of ENCODE enabled us to collect data in a more classical GWAS style by observing all genes present at the loci. While this yielded some promising results by including genes worthy of further study, they must be treated with caution, as they are not defined by cut off values as those from Geneminer are.
The collection of genes implicated in this study can be investigated further initially by observing correlations with human GWAS data sets, observing expression profiles via open access sources such as BioGPS and searching the literature for relevant publications.
These sources will also be valuable in narrowing down genuine candidate genes for further studies, by segregating genes that are highly expressed in OA affected tissues.
Further investigation of candidate genes can be conducted by performing gene expression experiments in specific tissues from affected strains to compare profiles with unaffected strains.
128
Table 5.1 Compiled list of genes implicated in all QTL maps
95th Percentile P- Chromosome Gene dbSNP Pos(Bp) Type Founders Value Association 8-12 month 17 Myo1f rs6329696 33712873 splice_site - 5.85E-07 total joint 8-12 month Daxx rs241036860 34049604 inframe_del - 5.85E-07 total joint 8-12 month Wdr46 . 34078132 splice_site - 5.85E-07 total joint 8-12 month Col11a2 . 34180179 splice_site B6,129s,CAST 5.85E-07 total joint 8-12 month H2-DMb2 rs47444840 34285498 missense - 5.85E-07 total joint 8-12 month H2-DMb1 rs108371532 34296588 missense - 5.85E-07 total joint 8-12 month Tap2 rs8258631 34351036 missense B6,129s,CAST 5.85E-07 total joint 8-12 month H2-Ab1 rs107595512 34401811 missense B6,129s,CAST 5.85E-07 total joint 8-12 month H2-Aa rs245673796 34421289 missense B6,129s,CAST 5.85E-07 total joint 8-12 month H2-Eb1 rs50132005 34446779 missense - 5.85E-07 total joint 8-12 month Btnl6 rs108319426 34644893 missense B6,129s,CAST 5.85E-07 total joint 8-12 month Nelfe . 34991017 inframe_ins B6,129s,CAST 5.85E-07 total joint 8-12 month Trim15 . 37001808 splice_site B6,129s,CAST 5.85E-07 total joint >8 month single 19 Glis3 rs30370239 28337109 missense B6,NOD,NZO 1.46E-04 surface >8 month single Rln1 rs48035875 29406575 missense CAST,PWK 4.53E-04 surface >8 month single Cd274 rs37044441 29446974 missense CAST,PWK 4.53E-04 surface >8 month single Pdcd1lg2 rs260164394 29520444 missense CAST,PWK 4.53E-04 surface >8 month single Ermp1 rs246267590 29701330 splice_site CAST,PWK 4.53E-04 surface >8 month single Ranbp6 rs233264122 29884144 missense CAST,PWK 4.53E-04 surface >8 month single Il33 rs262710057 30012389 splice_site CAST,PWK 4.53E-04 surface >8 month single Uhrf2 rs260406045 30162964 splice_site CAST,PWK 4.53E-04 surface >8 month single Gldc rs49341516 30207970 splice_site CAST,PWK 4.53E-04 surface
37th Percentile P- Gene dbSNP Pos(Bp) Type Founders Value Association > 12 month 2 Cdh26 rs27670331 178195508 splice_site WSB 1.76E-06 total joint > 12 month Cdh4 rs246485545 179582093 missense PWK 1.76E-06 total joint > 12 month Psma7 . 179771669 missense PWK 1.76E-06 total joint > 12 month Ss18l1 rs27664670 179787243 splice_site PWK 1.76E-06 total joint > 12 month Gtpbp5 rs247315857 179813487 missense PWK 1.76E-06 total joint > 12 month Adrm1 rs27679314 179910556 splice_site PWK 1.76E-06 total joint > 12 month Lama5 rs228260341 179911286 missense PWK 1.76E-06 total joint
129 > 12 month Rbbp8nl rs265937556 180012664 missense PWK 1.76E-06 total joint > 12 month Gata5 rs243672187 180062162 inframe_del PWK 1.76E-06 total joint > 12 month Ogfr rs13475605 180324242 missense PWK 1.76E-06 total joint > 12 month Col9a3 rs238294986 180346021 splice_site PWK 1.76E-06 total joint > 12 month Tcfl5 rs255129129 180377020 missense PWK 1.76E-06 total joint > 12 month Dido1 rs27664097 180418291 missense PWK 1.76E-06 total joint > 12 month Ythdf1 rs27667810 180645385 splice_site PWK 1.76E-06 total joint > 12 month Nkain4 rs27648163 180678898 missense PWK 1.76E-06 total joint > 12 month Arfgap1 rs27647920 180715737 splice_site PWK 1.76E-06 total joint > 12 month Col20a1 rs27694675 180733390 missense PWK 1.76E-06 total joint > 12 month Ppdpf rs27642184 180922702 splice_site PWK 1.76E-06 total joint > 12 month Ptk6 rs230301753 180933086 splice_site PWK 1.76E-06 total joint > 12 month Helz2 rs223230761 180966121 missense PWK 1.76E-06 total joint > 12 month Rtel1 rs243018327 181070591 splice_site PWK 1.76E-06 total joint > 12 month Zgpat rs253861124 181100723 inframe_ins PWK 1.76E-06 total joint > 12 month Abhd16b rs27671657 181228511 missense PWK 1.76E-06 total joint > 12 month Tpd52l2 rs212864047 181242930 missense PWK 1.76E-06 total joint > 12 month Uckl1 rs27654424 181305300 missense PWK 1.76E-06 total joint > 12 month Myt1 rs27685393 181532136 missense PWK 1.76E-06 total joint >12 month 4 Frem1 rs28072410 82559865 splice_site AJ,129s 3.32E-05 single surface >12 month Mpdz rs50311969 80994623 missense AJ,129s,NZO 1.21E-04 single surface splice_dono >12 month Smc2 . 52452284 r AJ,129s 1.47E-04 single surface >12 month Haus6 rs28061696 86228989 missense AJ,129s,NZO 1.49E-04 single surface splice_acce >12 month Plin2 rs3691335 86304077 pt AJ,129s,NZO 1.49E-04 single surface >12 month Dennd4c rs28079697 86471355 missense AJ,129s,NZO 1.49E-04 single surface >12 month Lpar1 rs27850547 58499786 missense 129s 9.75E-05 single surface >12 month Cdk5rap2 rs27839067 69884618 splice_site 129s 9.75E-05 single surface >12 month Tyrp1 . 80496716 missense 129s 6.71E-04 single surface >12 month Tle1 rs27840093 71806421 missense 129s 7.77E-04 single surface Incidence >3 8 Cdh13 rs37221381 121198927 missense - 6.46E-04 OARSI Incidence of 11 Larp1 rs50250532 57855842 missense B6,CAST 2.18E-05 OA splice_dono Incidence of Cnot8 rs245744133 57918296 r B6,CAST 2.18E-05 OA Incidence of Gemin5 rs29427483 57938720 missense B6,CAST 2.18E-05 OA Incidence of Igtp rs13463263 58020620 missense B6,CAST 2.18E-05 OA Incidence of Irgm2 rs6192315 58033414 missense B6,CAST 2.18E-05 OA Incidence of Zfp672 rs29390207 58143128 missense B6,CAST 2.18E-05 OA
130 Incidence of Lypd8 rs227304660 58192604 splice_site B6,CAST 2.18E-05 OA Incidence of Olfr332 rs29484442 58303250 stop_gain B6,CAST 2.18E-05 OA Incidence of Olfr331 rs49521872 58315210 missense B6,CAST 2.18E-05 OA Incidence of Trim58 rs48413015 58465162 missense B6,CAST 2.18E-05 OA Incidence of Olfr179 rs237411242 58480018 missense B6,CAST 2.18E-05 OA Incidence of Olfr320 rs26944296 58497434 missense B6,CAST 2.18E-05 OA Incidence of Olfr319 rs46063371 58515460 missense B6,CAST 2.18E-05 OA Incidence of Olfr317 rs49237152 58545717 missense B6,CAST 2.18E-05 OA Incidence of Olfr316 . 58571375 missense B6,CAST 2.18E-05 OA Incidence of Olfr313 rs47139371 58630798 missense B6,CAST 2.18E-05 OA Incidence of Zfp39 rs26971903 58704281 missense B6,CAST 2.18E-05 OA Incidence of Btnl10 rs243419500 58737429 inframe_ins B6,CAST 2.18E-05 OA Incidence of Trim17 rs26971749 58781984 splice_site B6,CAST 2.18E-05 OA Incidence of Obscn rs26955762 58821321 splice_site B6,CAST 2.18E-05 OA NZO,PWK,WS >12 month 15 Sun2 rs219132203 79567592 splice_site B 4.29E-05 single surface NZO,PWK,WS >12 month Cbx6 rs227221465 79658674 missense B 4.29E-05 single surface >12 month Tmprss6 rs225618358 78272988 splice_site PWK,WSB 6.76E-04 single surface >12 month Sh3bp1 rs251457439 78736941 splice_site - 8.52E-04 single surface >12 month Fam227a rs37552463 79467213 missense - 8.52E-04 single surface >12 month Tab1 rs46904772 79980972 missense - 8.52E-04 single surface
17 Epas1 rs50822893 87227219 missense WSB 1.76E-06 >12 month 8-12 month Esp8 rs108670479 40666992 missense 129s,CAST 4.43E-06 total joint 8-12 month Cdc5l . 45529220 stop_gain 129s 4.43E-06 total joint 8-12 month Ptchd4 . 42453654 missense B6,129s,CAST 7.44E-05 total joint >8 months total 19 Rfx3 rs257182977 28085160 splice_site CAST 1.96E-04 joint >8 months total Jak2 . 29373165 missense CAST 1.96E-04 joint >8 months total Dkk1 rs36488595 30621914 missense CAST 1.96E-04 joint
131
Chapter 6 Characterization of Candidate Genes Associated With Osteoarthritis Development
132 6.1 Introduction
Investigation into the genetic involvement of Osteoarthritis (OA) has offered new insights into a degenerative disease that is historically associated with mechanically induced wear.
Large human Genome Wide Association Studies (GWAS) and investigation of OA induced animal models have revealed several loci and subsequent genes that may play a role in the development of the disease. Further investigation of the effects, function and interactions of these genes and their transcribed proteins are critical to the understanding of the disease process. This understanding will allow researchers to identify and implement novel or existing therapeutic approaches in a targeted fashion to provide treatment for OA.
Investigation into the roles of potential candidate genes can be conducted by many means.
Reviewing the literature related to the genes of interest to correlate a previously identified role on OA or cells/tissues of interest is a useful technique to segregate a likely candidate from potentially false positives. The complexity of OA and the multiple factors that may contribute to the overall phenotype means a broad approach is required for the initial candidate review process. For example, it is well known that inflammatory cytokines play a role in the development of human OA (Wojdasiewicz, Poniatowski et al. 2014;
Daghestani et al. 2015). Correlation of such genes implicated in mouse studies with human published datasets builds confidence in the replication of the human OA phenotype in small animal models. These models equip researchers with viable tools for the investigation and development of treatments for OA in humans. Investigation into gene expression is also an important tool in identifying potential candidates. Observing expression of potential candidates within biological systems involved with OA
133 development builds confidence in the function of the genes, and/or their resulting protein and their potential involvement in the pathogenesis of OA.
Proteins act as part of a dynamic network to perform biological functions. The mapping and understanding of protein-protein interactions (PPIs) is of critical importance to the greater understanding of the complex molecular relationships and their functions within living systems. The term “interactome” defines the complete network of protein interactions that occurs within a living organism (Rual, Venkatesan et al. 2005; De Las
Rivas and Fontanillo 2010). Interactome mapping and the creation and utilization of databases capable of mapping PPIs have become an increasingly important component of biological research (Mehta and Trinkle-Mulcahy 2016). As described in chapter 1, in the context of OA research, it is known that multiple factors from individual systems and structures contribute to the development of the phenotype. The effects of one system often contribute to the pathological change of another, for example; damage to articular surfaces induces increased VEGF expression in chondrocytes, ultimately leading to vascularization of the cartilage matrix, tide mark advancement, mineralization and the instigation of bone formation (Patil, Sable et al. 2012; Honorati et al. 2007). The use of
PPI resources allows biological researchers to better understand the complex functional roles of proteins and to identify pathways that are involved in the regulation of OA, for a more targeted intervention of pathological processes.
In this chapter we have compiled data from a range of resources such as; published human genome studies of OA, gene expression profiles of involved biological systems, and protein interaction networks to classify and prioritize genes identified in chapter 5.
Information gathered from these resources will be used to validate the model used to
134 investigate OA in chapter 4 and to identify candidate genes that are the most likely to be involved in the pathogenesis of OA.
6.2 Results
6.2.1 Identification of candidate genes implicated above the 95th percentile threshold
To segregate potential candidate genes from the genes implicated at the two loci identified in chapter 5, each gene was investigated using online resources Genecards, Uniprot and
BioGPS. The literature for each gene was reviewed using Pubmed to identify any association with arthritis and osteoarthritis specifically. For a gene to be included as a potential candidate, evidence was needed of significant expression in involved tissues or cells and/or recent publications implicating the association and/or role of the gene in the disease process. Out of 22 genes identified as genome wide significant in chapter 5, 8 met the criteria of potential candidate genes (Table 6.1). The genes Daxx and Tap2 showed expression in bone marrow macrophages and Pdcd1lg2 showed expression in mast cells but all showed minimal to no expression in bone or cartilage related cells. These genes were all implicated in publications unique to Rheumatoid Arthritis (RA) with Tap2 also being implicated in studies investigating Juvenile Idiopathic Arthritis (JIA) and
Ankylosing Spondylitis (AS). Myo1f showed significant expression in both osteoclasts
(OC) and bone however no publications relating to arthritis have been identified. The remaining genes Col11a2, Glis3, Cd274 and Il33 were all implicated in studies investigating OA specifically.
135 Table 6.1 Candidate genes implicated above the 95th percentile.
Gene dbSNP P- Association Expression Related
Value published work
Myo1f rs6329696 5.85E 8-12 month OC +++ None
-07 total joint Bone ++
Daxx rs2410368 5.85E 8-12 month BMM +++ RA
60 -07 total joint
Col11a2 . 5.85E 8-12 month OB +++ Early onset OA
-07 total joint Bone +++
Tap2 rs8258631 5.85E 8-12 month OC + RA, JIA, AS
-07 total joint BMM +++
Glis3 rs3037023 1.46E >8 month OB ++ Human OA
9 -04 single GWAS
surface
Cd274 rs3704444 7.02E >8 month BMM +++ Treg enrichment
1 -05 total joint in RA and OA
Pdcd1lg2 rs2601643 7.02E >8 month Mast cells RA Synovial
94 -05 total joint +++ fluid
Il33 rs2627100 7.02E >8 month OB +++ Mouse
57 -05 total joint C2C12 cells microarray,
++ Human OA
serum levels
136 6.2.2 Characterization of genome wide significant candidate genes
6.2.2.1 Potential novel candidates
MYO1F
Myosin IF (Myo1f) is a protein-coding gene located on chromosome 17, which was implicated in the QTL mapping of total joint scores of 8-12 month old or “middle aged”
CC mice. MYO1F is highly conserved between humans and mice sharing 93.7% amino acid identity and 97.7% similarity. Myosins are actin based motor molecules that use ATP from hydrolysis. The protein encoded by this gene is considered to be an unconventional myosin that may have a role in the intracellular movement of membrane-enclosed compartments. The role of MYO1F has yet to be investigated in OA, diseases that have been associated with MYO1F include deafness and acute monocytic leukemia and its known pathways are the PAK pathway and the ERK signaling pathway.
Investigation of expression of Myo1f in mice revealed a high level of expression in osteoclasts and bone (Figure 6.1). This expression suggests that Myo1f may play a role in the subchondral bone-remodeling component of OA. Further investigation of PPI networks involving MYO1F revealed a strong association with CALM2 (Figure 6.2), which has been previously implicated in a study of hip OA in a Japanese population
(Mototani, Iida et al. 2010). This study states that CALM2 is abundantly expressed in articular chondrocytes and OA cartilage and two SNPs in intron 1 were found to be associated with a sub population of hip OA patients without acetabular dysplasia. Further investigation of MYO1F and its role in bone remodeling and it’s interaction with CALM2 would be advantageous in understanding its role in OA.
137
Figure 6.1 Gene expression profile of mouse Myo1f
Myo1f is highly expressed in osteoclasts, BMMs and dendritic cells. It is also found to be expressed in bone and bone marrow.
138
Figure 6.2 Visual protein-protein interaction networks for MYO1F in humans
MYO1F is linked with CALM2, a gene suggested to be involved with human hip OA.
139 6.2.2.2 Candidate genes implicated in OA studies
COL11A2
Collagen XI Alpha 2 chain (Col11a2) is a protein coding gene found on chromosome 17, which was implicated in the QTL mapping of total joint scores of 8-12 month old or
“middle aged” CC mice. The transcribed protein of COL11A2 shares 93.7% amino acid identity and 95.7% similarity between mice and humans. The gene codes for a minor fibrillar collagen and may play a significant role in fibrillogenesis via the control of lateral growth of collagen II fibrils. Its related pathways are the integrin and ERK signaling pathway. Expression of Col11a2 in mice revealed a high expression in differentiating osteoblasts and in bone (Figure 6.3).
Known mutations in COL11A2 have been associated with Otospondylomegaepiphyseal dysplasia (OSMED syndrome), Weissenbacher-Zweymuller syndrome and deafness.
Studies by Jeffries et al have shown differential methylation as well as differential expression of COL11A2 in affected cartilage from 24 patients undergoing hip athroplasty
(Jeffries, Donica et al. 2014). Studies investigating genes associated with heat pain sensitivity in knee OA have identified SNPs within COL11A2 that are associated with pain sensitization in patients with developed knee OA (Ho, Wallace et al. 2017).
Investigation of patients with early onset OA in the hip and/or knee have been shown to carry two sequence variations in COL11A2 when compared with non-affected controls
(Jakkula, Melkoniemi et al. 2005). Stickler syndrome is a disorder that also presents with premature arthritis that has been associated with mutations observed in the COL11A2 gene (Li and Thorne 2010). When observing PPI networks, COL11A2 was co-expressed
140
Figure 6.3 Gene expression profile of mouse Col11a2
Col11a2 is highly expressed in bone and during osteoblast differentiation
141 with COMP, ACAN and HAPLN1 (Figure 6.4). The expression of COMP and ACAN were negatively correlated with the production of ADAMTSs in cartilage tissue from an
OA affected cohort (Zhang, Ji et al. 2015). These studies reflect the association with mice developing OA at 8-12 months results presented in chapter 5, therefore supporting the association between Col11a2 and the early onset of OA.
GLIS3
GLIS Family Zinc Finger 3 (Glis3) is a protein-coding gene found on chromosome 19, which was implicated in the QTL mapping of highest single surface scores of >8 month old complete CC mouse cohort. The transcribed protein of GLIS3 shares 83.8% identity and 88.7% similarity between mice and humans. GLIS3 encodes a nuclear protein with five C2H2-type zinc finger domains. It functions as both a transcription activator and repressor and is involved in the development of structures such as the eye, thyroid, liver, kidney and pancreatic beta cells. Investigation of Glis3 expression in mice showed high expression in differentiating osteoblasts, with maximum expression observed at day 21
(Figure 6.5). We analysed data collected from an in-house rabbit derived MSC microarray to investigate the expression of Glis3 during chondrogenesis. We observed that Glis3 is expressed during chondrogenesis with expression peaking at day 14 (Figure 6.6).
Mutations in GLIS3 have been linked to susceptibility to type 1 and 2 diabetes by modulating pancreatic beta cell apoptosis (Nogueira, Paula et al. 2013). It has also been implicated in the development of congenital hypothyroidism in humans, with knockout models in mice exhibiting similar phenotypes (Lichti-Kaiser, ZeRuth et al. 2014). Glis3 has also been reported to promote osteoblast differentiation via the regulation of FGF18 expression, while acting synergistically with BMP2 and Shh in mice (Beak, Kang et al.
2007). Most recently, a novel variant in GLIS3 has been implicated in a human GWAS
142
Figure 6.4 Visual protein-protein interaction networks for COL11A2 in humans
COL11A2 is linked with COMP and ACAN via co-expression. Expression of COMP and ACAN are negatively associated with the production of aggrecanases in OA affected cartilage
143
Figure 6.5 Gene expression profile of mouse Glis3
Glis3 is expressed during osteoblast differentiation.
144 A
B
Figure 6.6 Expression of Glis3 during chondrogenesis from rabbit derived MSCs
(A) Glis3 is expressed during chondrocyte differentiation with maximum expression at day 14. (B) Expression of Col2a1 during chondrogenesis as a positive control for successful chondrocyte differentiation.
145 of 17,894 OA affected individuals who have undergone hip and/or knee replacement.
This study also suggested GLIS3 was found to be expressed in articular cartilage
(Casalone, Tachmazidou et al. 2018). Investigation of PPI networks revealed that GLIS3 is co-expressed with ERN1 (Figure 6.7), which has been shown to be downregulated in chondrocytes from OA affected individuals (Li, Tardif et al. 2016). The implication of
GLIS3 in a recent GWAS with a large cohort of knee and/or hip OA sufferers is an encouraging proof of principle for the investigation of spontaneous OA in the CC cohort.
IL-33
Interleukin 33 (Il-33) is a protein-coding gene found on chromosome 19, which was implicated in the QTL mapping of total joint and highest single surface scores of >8 month old complete CC mouse cohort. The transcribed protein of IL33 shares 51.6% identity and 69.1% amino acid similarity between humans and mice. The transcribed protein is a cytokine that binds to the IL1RL1/ST2 receptor, which activates NF-kappa-
B and MAPK pathways in targeted cells. It is involved with the maturation of Th2 cells and induces the secretion of T-helper type 2-associated cytokines. IL-33 is also involved in the activation of mast cells, basophils, eosinophils and natural killer cells. Investigation of Il-33 expression showed high levels of expression in differentiating osteoblasts, reaching maximum expression at day 21, and the C2C12 myoblastic cell line (Figure 6.8)
A microarray analysis conducted on a DMM-induced OA mouse model by Loeser et al shows that Il-33 was expressed in multiple joint tissue cells along with extracellular matrix and cell adhesion related genes (Loeser, Olex et al. 2012). A study of IL-33 levels in human serum
146
Figure 6.7 Visual protein-protein interaction networks for GLIS3 in humans
GLIS3 is linked via co-expression with ERN1, which has been found to be downregulated in OA affected chondrocytes.
147
Figure 6.8 Gene expression profile of mouse Il-33
IL-33 is highly expressed during osteoblast differentiation with peak expression at day 21.
148 and synovial fluid by Talabot-Ayer showed high levels of IL-33 mRNA expression in the synovium (Talabot-Ayer, McKee et al. 2012). IL-33 levels were detected in serum by
ELISA, although in lower levels when compared with RA patients. IL-33 functions as an alarmin and cells within an OA affected synovial joint secrete IL-33 along with HMGB1,
S100 proteins and other endogenous molecules (Nefla, Holzinger et al. 2016). Secretion of these alarmins can recruit immune cells to the inflamed synovium and perpetuate the disease by initiating the adaptive immune response. This suggests that IL-33 has a role in the continued progression of cartilage destruction and inflammation that occurs in OA.
Investigation of PPI networks revealed that IL-33 is co-expressed with TLR2 and an experimentally determined interaction with IL1RL1. The PPI network also implicated a network including GATA3, IL-13, IL-1B, IL-4, IL-6 and VEGFA, which have all been implicated in OA studies (Figure 6.9).
149
Figure 6.9 Visual protein-protein interaction networks for IL-33 in humans
IL-33 is co-expressed with TLR2 and an experimentally determined interaction with
IL1RL1. It is also linked with GATA3, IL-13, IL-4, IL1B and VEGFA.
150 6.2.2.3 Candidate genes implicated in studies of other forms of arthritis
DAXX
Death Domain Associated Protein (Daxx) is a protein coding gene found on chromosome
17, which was implicated in the QTL mapping of total joint scores of 8-12 month old or
“middle aged” CC mice. The transcribed protein shares 70.7% identity and 80.1% similarity between mice and humans and is a multifunctional protein that resides in multiple locations in the nucleus and cytoplasm. Investigation of Daxx expression showed high levels of expression in Bone Marrow-derived Macrophages (BMMs) in mice (Figure
6.10). Further investigation of PPI networks showed strong interactions and co- expression of DAXX with SUMO1, FAS, UBC, MDM2 and MAP3K5 in humans (Figure
6.11). Diseases associated with DAXX include alpha-thalassemia and gastric neuroendocrine tumours. A study by Meinecke et al showed that the accumulation of
DAXX in promyelocytic leukaemia protein nuclear bodies by the small ubiquitin-like modifier (SUMO-1) might contribute to the pathogenesis of RA (Meinecke, Cinski et al.
2007). However, a role for DAXX in OA is yet to be determined.
PDCD1LG2
Programmed Cell Death Ligand 2 (PDCD1LG2) is a protein coding gene found on chromosome 19 which was implicated in the QTL mapping of total joint and highest single surface scores of >8 month old complete CC mouse cohort. The transcribed protein shares 63% identity and 73.3% similarity between mice and humans and is involved in the costimulatory signal, which is essential for T-cell proliferation and IFNγ production in a PDCD1-independent manner. Investigation of Pdcd1lg2 expression showed high levels of expression in mast cells in mice stimulated with IgE antigen (Figure 6.12).
151
Figure 6.10 Gene expression profile of mouse Daxx
Daxx is highly expressed in bone marrow macrophages.
152
Figure 6.11 Visual protein-protein interaction networks for DAXX in humans
DAXX is found to have strong interactions and co-expression with SUMO1, FAS,
UBC, MDM2 and MAP3K5. It has been suggested that SUMO1 may have a role in
RA.
153
Figure 6.12 Gene expression profile of mouse Pdcd1lg2
Pdcd1lg2 is highly expressed in mast cells and IgE primed mast cells.
154 Human PPI networks showed that PDCD1LG2 interacts with PDCD1, PTPN11, CD4,
LCK, CD3E, CD3D, CD3G, HLA-DPB1, HLA-DRB1 and CD274 (Figure 6.13).
Interestingly, Cd274 was also implicated within the same locus as Pdcd1lg2. A study by
Bhattacharjee et al, an investigation of the synovial fluid proteome in RA identified the presence of 956 proteins that included PDCD1LG2 among M-CSF, TNFRSF1B, hyaluronan-binding protein 2, semaphorin 4A and osteoclast stimulating factor 2
(Bhattacharjee, Balakrishnan et al. 2016). Although PDCD1LG2 has not been implicated in OA disease studies prior to this present study, its association with RA suggests its potential as a candidate for further OA studies.
155
Figure 6.13 Visual protein-protein interaction networks for PDCD1LG2 in humans
PDCD1LG2 is tightly linked with PDCD1, PTPN11, CD4, LCK, CD3E, CD3D,
CD3G, HLA-DPB1, HLA-DRB1 and CD274.
156 CD274
CD274 molecule or otherwise known as Programmed Cell Death 1 Ligand 1
(PDCD1LG1) is a protein coding gene found on chromosome 19 which was implicated in the QTL mapping of total joint and highest single surface scores of >8 month old complete CC mouse cohort. The transcribed protein is 69.4% identical and 81.1% similar between mice and humans and encodes an immune inhibitory receptor ligand that is expressed by hematopoietic and non-hematopoietic cells. The interaction of this ligand with its receptor inhibits T-cell activation and cytokine production, which is important during infection or inflammation of normal tissue for preventing autoimmunity by maintaining homeostasis of the immune response. An important paralog of CD274 is
PDCD1LG2, which was implicated within the same locus as Cd274. Investigation of
CD274 expression showed high levels of expression in LPS primed BMMs in mice
(Figure 6.14). Investigation of PPI networks showed a tight network of interactions between CD274 and PDCD1, PTPN11, PDCD1LG2, LCK, CD3E, CD4, HLA-DRB1,
HLA-DPB1, HLA-DRA and CD3G in humans (Figure 6.15).
Investigation of Treg enrichment in the joint compartment of RA and OA affected joints shows that Tregs are in fact enriched in both forms of arthritis with only minor differences in expression of markers, including CD274 (Sugita, Hosaka et al. 2015). Further investigation of CD274 and its role in Treg function in OA affected joints is required to gain a better understanding of the underlying role of CD274 in the pathogenesis of OA, therefore suggesting CD274 may be a viable candidate for further studies of OA.
157
Figure 6.14 Gene expression profile of mouse Cd274
Cd274 is highly expressed in LPS primed bone marrow macrophage cells.
158 Figure 6.15 Visual protein-protein interaction networks for CD274 in humans
CD274 is part of a tight network of interactions between PDCD1, PTPN11,
PDCD1LG2, LCK, CD3E, CD4, HLA-DRB1, HLA-DPB1, HLA-DRA and CD3G.
159 TAP2
Transporter 2, ATP Binding Cassette Subfamily B Member (TAP2) is a protein coding gene located on chromosome 17, which was implicated in the QTL mapping of total joint scores of 8-12 month old or “middle aged” CC mice. The transcribed protein shares
75.4% identity and 84.5% similarity between mice and humans and is a member of the
MDR/TAP subfamily of the ATP-binding cassette (ABC) transporter superfamily. This protein forms a heterodimer with ABCB2 in order to transport antigens from the cytoplasm to the endoplasmic reticulum for association with MHC class 1 molecules.
Investigation of Tap2 expression showed high levels of expression in mouse mast cells and LPS primed BMMs, and lower expression in mouse osteoclasts (Figure 6.16). Further investigation of PPI networks involving TAP2 revealed a network of interactions that included HLA-A, HLA-B, HLA-C, HLA-E, HLA-G, TAP1, PDIA3, CALR and B2M
(Figure 6.17). The interaction of TAP2 and beta 2 microglobulin (B2M) may be of particular interest as B2M mRNA is highly expressed in OA cartilage and synovial fluid when compared with unaffected controls (Zhang, Liew et al. 2002).
The involvement of TAP2 in many forms of arthritis has been investigated at length with associations of polymorphisms in TAP2 being implicated in RA and Juvenile Idiopathic
Arthritis (JIA). A recent study showed that TAP2-379Ile allele was significantly associated with a 38% increase in risk of RA in Asians and TAP-2565Thr allele increased the risk of RA in Europeans by 38% (Dai, Chen et al. 2014). A study of JIA showed that multiple SNPs from TAP2/TAP1 haplotypes were associated with JIA (Runstadler, Saila et al. 2004). Studies by Fraile et al and Westman et al both showed that TAP2 had no involvement in ankylosing spondylitis (Fraile, Collado et al. 2000, Westman, 1995 #236).
160
Figure 6.16 Gene expression profile of mouse Tap2
Tap2 is highly expressed mast cells, LPS primed bone marrow macrophages and in lower levels in osteoclasts.
161 Figure 6.17 Visual protein-protein interaction networks for TAP2 in humans
TAP2 is linked in a tight network with HLA-A, HLA-B, HLA-C, HLA-E, HLA-G,
TAP1, PDIA3, CALR and B2M. High B2M expression in cartilage has been associated with OA.
162 While the role of TAP2 has yet to be investigated specifically in OA, it’s identification within this study, its role in other forms of arthritis and it’s interaction with B2M supports the notion that it may be considered as a viable candidate for further OA research.
6.2.3 Identification of candidate genes implicated above the 37th percentile threshold
Segregation of candidate genes from all remaining loci that reached the suggestive threshold of the 37th percentile was conducted using the same online resources that were used in section 6.2.1. A more stringent gene selection criteria was used in the selection of the suggestive genes to separate false positives from true suggested candidates. Both significant expression in involved tissues and recent publications implicating the association and/or role of the gene in the disease process were required to be included as potential candidates. Out of 70 genes identified as suggestive in chapter 5, 39 showed either associated expression or publications supporting a role in some form of arthritis, with 7 meeting the inclusion criteria of both expression and published data available
(Table 6.2). The genes Cdh13 and Obscn showed expression in skeletal muscle with
Cdh13 being implicated in age related disorders and JIA. Tab1 showed expression in osteoclasts but has only been implicated in RA studies thus far. The remaining genes
Col9a3, Epas1, Jak2, and Dkk1 were all expressed in bone or associated cells and have been implicated in previous OA research.
An additional potential candidate Cdh4 was also included in the analysis due to being implicated during the preliminary stages of the study and remaining a suggestive gene of interest after completing the study. Cdh4 also demonstrated expression in skeletal muscle as well as being differentially expressed in osteoclasts between affected and non-affected
OA mice.
163 Table 6.2 Candidate genes implicated above the 37th percentile.
Gene dbSNP P- Association Expression Published work
Value
Cdh4 rs246485 1.76E > 12 month Skeletal None
545 -06 total joint muscle +++
Col9a3 rs238294 1.76E > 12 month OB + Knee and hip OA,
986 -06 total joint Lumbar disk disease,
Stickler syndrome
Cdh13 rs372213 6.46E Incidence Skeletal JIA, age related
81 -04 >3 OARSI muscle +++ disorders
Obscn rs269557 2.18E Incidence of Skeletal Muscle disorders
62 -05 OA muscle +++
Tab1 rs469047 8.52E >12 month OC + Involved in pathways
72 -04 single regulating cartilage and
surface joint development
Epas1 rs508228 1.76E >12 month OB + Hif2a, chondrocytes,
93 -06 Bone + OA
Muscle ++
Jak2 . 1.96E >8 months BMM +++ RA, OA
-04 total joint Bone +
OB +
Dkk1 rs364885 1.96E >8 months Bone + Hip OA, RA, inhibition
95 -04 total joint of OA in mice
164 6.2.4 Characterization of suggestive candidate genes
6.2.4.1 Candidate genes implicated in OA studies
COL9A3
Collagen Type IX alpha 3 Chain (COL9A3) is a protein coding gene found on chromosome 2, which was implicated in the QTL mapping of total joint scores of <12 month old or “aged” CC mice. The transcribed protein shares 90.9% identity and 93.3% similarity between mice and humans. The gene encodes one of the three alpha chains of type IX collagen of which is the major component of hyaline cartilage found in synovial joints. It is responsible for providing tensile strength to the extracellular matrix of hyaline cartilage and while it is known to be expressed by chondrocytes, it has also been found to be expressed in osteoblasts in lower levels (Figure 6.18). COL9A3 interacts with the integrin and ERK pathways. PPI networks show that COL9A3 is part of a tight network of interaction with 5 other collagen family members, COL16A1, COL2A1, COL9A1,
COL9A2 and COL11A1, along with 2 matrix metallopeptidases MMP3 and MMP13, and
PLOD1, PLOD2 and PLOD3 (Figure 6.19).
Many joint diseases have implicated associations and/or mutations in COL9A3. A family study of multiple epiphyseal dysplasia showed that among the three mutations found at the splice donor or acceptor site of COL9A3 that caused skipping of exon 3, resulting in
12 amino acids being deleted from the COL3 domain, there was also an novel missense mutation identified that affected the Gly residue in the COL3 domain (Jeong, Lee et al.
2014). A study of knee and hip OA in a Japanese population also showed a SNP in
COL9A3 to be significantly associated with knee OA specifically (Ikeda, Mabuchi et al.
2002). Lumbar disk disease is another phenotype affected by COL9A3. A study by
Paassilta et al showed that amongst mutations found in the other collagen IX genes,
165
Figure 6.18 Gene expression profile of mouse Col9a3
Col9a3 is expressed during osteoblast differentiation.
166
Figure 6.19 Visual protein-protein interaction networks for COL9A3 in humans
COL9A3 is linked with collagen family members COL16A1, COL2A1, COL9A1,
COL9A2 and COL11A1, along with 2 matrix metallopeptidases MMP3 and MMP13.
167 a mutation consisting of Arg103Trp (argininetryptophan) substitution in the alpha3 chain allele (Trp3 allele) was present in 12.2% of lumbar disk disease cases. Presence of at least one Trp3 allele was found to increase the risk of lumbar disk disease 3-fold
(Paassilta, Lohiniva et al. 2001). The significant association between COL9A3 and OA among other joint diseases suggests the importance of COL9A3 as a potential candidate gene for further studies.
EPAS1
Endothelial PAS Domain Protein 1 (EPAS1) is a protein coding gene found on chromosome 17, which was implicated in the QTL mapping of total joint scores of <12 month old or “aged” CC mice. The transcribed protein shares 87.9% identity and 92.1% similarity between mice and humans and encodes a transcription factor HIF-2-Alpha
(HIF-2α) which is involved in the induction of oxygen related genes. EPAS1 was found to be expressed in low levels in osteoblasts and in skeletal muscle (Figure 6.20).
Investigation of PPI networks revealed that EPAS1 interacts with Vascular Endothelial
Growth Factor A (VEGFA), a pro-angiogenic growth factor involved in endochondral ossification (Patil, Sable et al. 2012) (Figure 6.21).
The role of EPAS1, HIF-2α and VEGFA has been well categorized in OA studies. HIF-
2α expression is increased in human and mouse OA affected cartilage and induces the expression of genes within chondrocytes that encode catabolic factors such as matrix metalloproteinases MMP1, MMP3, MMP9, MMP12 and MMP13, aggrecanase-1
(ADAMTS4), nitric oxide synthase-2 (NOS2) and prostaglandin-endoperoxide synthase-
2 (PTGS2). A transgenic mouse model for Epas1 only in chondrocytes showed
168
Figure 6.20 Gene expression profile of mouse Epas1
Epas1 is expressed in low levels in osteoblasts and skeletal muscle.
169 Figure 6.21 Visual protein-protein interaction networks for EPAS1 in humans
EPAS1 interacts with VEGFA, a pro-angiogenic growth factor involved in endochondral ossification.
170 spontaneous cartilage degradation while heterozygous deletion of Epas1 in mice showed a reduction in cartilage destruction when inducing OA via DMM or collagenase injection
(Yang, Kim et al. 2010). The involvement of HIF-2α in endochondral ossification, matrix degradation and vascular invasion via the enhancement of promoter activities of
COL10A1, MMP13 and VEGFA was reported by Saito et al including the identification of a single SNP in the human EPAS1 gene that was associated with knee OA in a Japanese cohort (Saito, Fukai et al. 2010). While the evidence supporting the role of Epas1 and
HIF-2α in mouse OA is convincing, a more recent investigation could not replicate the association of the SNP in human OA (Nakajima, Shi et al. 2011). Further study of EPAS1 and its role in OA in humans is required, however the evidence suggesting its role in OA mice is encouraging as a potential candidate gene for OA research.
JAK2
Janus Kinase 2 (JAK2) is a protein coding gene found on chromosome 19, which was implicated in the QTL mapping of total joint scores of the >8 month old complete CC mouse cohort. The transcribed protein of JAK2 shares 93.7% identity and 97.7% similarity between mice and humans. JAK2 encodes a protein tyrosine kinase, which has a role in specific cytokine receptor pathways. It is involved in cell growth, development and differentiation and is involved in signaling in both innate and adaptive immunity.
Upon activation of JAK receptors, phosphorylation and activation of STATs occur, which initiates the JAK-STAT signaling pathway. Jak2 is highly expressed in mouse BMMs during LPS stimulation and has a lower expression in bone and osteoblasts (Figure 6.22) while PPI networks show JAK2 interacts in a tight network with STAT1, STAT3,
STAT5A, STAT5B and STAT6 (Figure 6.23).
171
Figure 6.22 Gene expression profile of mouse Jak2
Jak2 is highly expressed in mouse BMMs during LPS stimulation and has a lower expression in bone and osteoblasts.
172 Figure 6.23 Visual protein-protein interaction networks for JAK2 in humans
JAK2 interacts in a tight network with STAT1, STAT3, STAT5A, STAT5B and STAT6.
173 Studies investigating the correlation between obesity and the initiation of OA have shown that inhibition of JAK2-STAT3 signaling pathway alters the expression of proteins associated with leptin induced chondrocyte apoptosis in rats (Zhang, Shen et al. 2016). A study investigating chondrocyte mineralization via calcium phosphate crystals showed basic calcium phosphate crystals stimulated IL-6 secretion via signaling pathways involving Syk and P13 kinases as well as Jak2 and Stat3 molecules (Nasi, So et al. 2016).
Investigation of JAK2 inhibitor Tyrphostin AG490 showed a decrease in expression of signal transducers and activators of STAT3, phosphorylated JAK2, DKK1 and receptor activator of nuclear factor KB ligand (RANKL) in the joints of OA affected mice and inhibited RANKL-induced osteoclast differentiation (Gyurkovska, Stefanova et al. 2014).
Identification of Jak2 in the OA affected CC mice combined with the findings of these studies support the involvement of the JAK2-STAT3 pathway in the pathogenesis of OA and the potential for JAK2 to be targeted in future OA treatment studies.
DKK1
Dickkopf WNT Signaling Pathway Inhibitor 1 (DKK1) is a protein coding gene located on chromosome 19, which was implicated in the QTL mapping of total joint scores of the
>8 month old complete CC mouse cohort. The transcribed protein of DKK1 shares 81.2% identity and 87.1% similarity between mice and humans. It is a member of the dickkopf protein family, which are characterized by two cysteine rich domains that mediate protein-to-protein interactions. The DKK1 protein binds to the LRP6 co-receptor, which then inhibits beta catenin dependent Wnt signaling. DKK1 has been implicated in bone formation and related diseases such as osteoporosis and embryonic development.
Investigation of expression showed that Dkk1 was expressed in bone at low levels
174 (Figure 6.24) and interacted in a PPI network with WNT1, WNT3A, WNT8A, LRP5 and
LRP6 (Figure 6.25). Investigation into the role of DKK1 in OA has revealed a negative correlation between DKK1 expression and OA development. A study of induced experimental OA in mice has shown that expression of Dkk1 was greatly reduced in chondrocytes from individuals that underwent menisectomy to induce OA symptoms when compared with unaffected controls. Conversely, Col1a1-Dkk1-Tg mice, which exhibit an overexpression of Dkk1, showed a lower OA score after menisectomy. The
Col1a1-Dkk1-Tg mice also showed a decrease in VEGF expression in osteoblasts thus decreasing expression of mRNA for MMPs in chondrocytes (Funck-Brentano, Bouaziz et al. 2014). Further evidence of the negative correlation between OA and DKK1 expression has been shown in a human study, demonstrating a reduction in expression of
WNT antagonists DKK1 and FRZB while observing an increase in hypertrophic markers
RUNX2, COL10A1 and IHH via qPCR of cartilage specimens taken after knee replacement (Zhong, Huang et al. 2016). An earlier study by Leijten et al supports the evidence of a reduction of WNT antagonists in OA affected tissue and showed that the reduction is more significant in degrading cartilage in comparison to undamaged cartilage of the same joint (Leijten, Bos et al. 2013). These studies suggest that the expression of
DKK1 has a protective effect against OA development and is therefore a valuable candidate for the investigation of impaired DKK1 function in OA sufferers and as a possible target for the treatment of OA.
175 Figure 6.24 Gene expression profile of mouse Dkk1
Dkk1 is expressed in bone at low levels.
176 Figure 6.25 Visual protein-protein interaction networks for DKK1 in humans
DKK1 interacts in a network with WNT1, WNT3A, WNT8A, LRP5 and LRP6.
177 6.2.4.2 Candidate genes implicated in studies of other forms of arthritis
TAB1
TGF-Beta Activated Kinase 1 (MAP3K7) Binding Protein 1 (TAB1) is a protein coding gene found on chromosome 15, which was implicated in the QTL mapping of the highest single surface scores of >12 month old or “aged” CC mice. The transcribed protein shares
97.2% identity and 99% similarity between mice and humans. It is involved in RANKL signaling in osteoclasts and may have a role as a signaling intermediate between TGFB and MAP3K7/TAK1. Tab1 is ubiquitously expressed and shows a small increase in expression in osteoclasts and skeletal muscle (Figure 6.26). PPI network investigation revealed that TAB1 was part of a network including MAP3K7, MAPK14, RIPK2 and
IRAK1 (Figure 6.27).
The role of TAB1 in OA has not yet been categorized directly. It has been shown that
TAB1 is responsible for the activation of TAK1 (Scholz, Sidler et al. 2010). The role of activated TAK1 in MAPK and BMP signaling pathways is of importance because of its role in cartilage and joint development. The deletion of Tak1 in mouse chondrocytes resulted in developmental cartilage defects and a reduction in chondrocyte proliferation and survival. This suggests that the loss of Tak1 results in the impaired activation of a downstream MAPK target p38, along with a reduction in the activation of the
BMP/SMAD signaling pathway, which are necessary for proper cartilage development
(Gunnell, Jonason et al. 2010). Due to the importance of TAK1 in this signaling, further investigation of TAB1 and its role in cartilage formation is required.
178
Figure 6.26 Gene expression profile of mouse Tab1
Tab1 is ubiquitously expressed and expression is slightly increased in osteoclasts and skeletal muscle.
179 Figure 6.27 Visual protein-protein interaction networks for TAB1 in humans
TAB1 is part of a network including MAP3K7, MAPK14, RIPK2 and IRAK1.
180 OBSCN
Obscurin, Cytoskeletal Calmodulin and Titin-Interacting RhoGEF (OBSCN) is a protein coding gene found on chromosome 11, which was implicated in the QTL mapping of incidence of OA in the >8 month old complete CC mouse cohort. The transcribed protein of OBSCN shares 74.9% identity and 81.7% similarity between mice and humans. The gene is greater than 150kb in size, containing over 80 exons and the transcribed protein is approximately 720kDa. The protein contains 68 lg domains, 2 fibronectin domains, 1 calcium/calmodulin binding domain, 1 RhoGEF domain and 2 serine-threonine kinase domains. It is a structural component of striated muscle and has a role in myofibrillogenesis. Investigation into expression showed that Obscn is highly expressed in skeletal muscle (Figure 6.28) and investigation of PPI networks revealed interaction and co-expression of OBSCN with RHOA, CDC42 and RAC1 (Figure 6.29).
OBSCN has not yet been implicated in OA studies, although a variant has been identified in OBSCN in a family affected by distal muscular dystrophy, which affects the function of the quadriceps, which may have an effect on OA progression (Rossi, Palmio et al.
2017). Investigation of other proteins interacting in a network with OBSCN showed a correlation between RHOA, CDC42 and RAC1 and OA development. The role of the
GTPase RhoA and its effector kinases ROCK1/2 have been investigated in the context of chondrocyte differentiation and proliferation and were shown to increase proliferation and delay hypertrophism, thus showing a decrease in alkaline phosphatase activity and hypertrophic markers collagen X and MMP13 (Wang, Woods et al. 2004). CDC42 has an important role in chondrogenesis and the lifespan of chondrocytes. This has been demonstrated via conditional deletion of Cdc42 in mice showing a range of bone
181
Figure 6.28 Gene expression profile of mouse Obscn
Obscn is highly expressed in skeletal muscle.
182 Figure 6.29 Visual protein-protein interaction networks for OBSCN in humans
OBSCN is part of a network including RHOA, CDC42 and RAC1.
183 phenotypes and a loss of columnar organization of chondrocytes during proliferation
(Suzuki, Yamada et al. 2015). A study by Wang et al showed that overexpression of
Cdc42 or Rac1 in chondrogenic cells will result in a reduction of cell numbers and accelerated hypertrophic differentiation therefore reversing the effect of GTPase RhoA
(Wang and Beier 2005). The role of OBSCN and its effect on RhoA signaling therefore suggest it is a viable candidate for further OA related studies.
CDH13
Cadherin 13 (CDH13) is a protein coding gene found on chromosome 8, which was implicated in the QTL mapping of incidence of OARSI scores >3 in the >8 month old complete CC mouse cohort. The transcribed protein shares 93.4% identity and 96.6% similarity between mice and humans. It acts as a negative regulator of neural cell growth and also protects vascular cells from apoptosis from oxidative stress. Unlike other cadherin family members, CDH13 lacks a cytoplasmic domain and thus is thought not to have a role in cell-cell adhesion. It is highly expressed in skeletal muscle (Figure 6.30), and is part of a PPI network including JUP, CTNND1 and other cadherin family members
(Figure 6.31).
CDH13 has been implicated in a GWAS study of human age related disorders in a group of genes considered to have a role in the causation of age related disorders (Srivastava,
Thukral et al. 2015). It has also been implicated in a GWAS pilot study of familial
Juvenile Idiopathic Arthritis (JIA) (Aydin-Son, Batu et al. 2015), and a study of novel epistatic risk factors with PTPN22 for RA, where SNPs in CDH13 showed evidence for interaction with PTPN22 and may contribute to the susceptibility of RA (Briggs, Ramsay et al. 2010).
184
Figure 6.30 Gene expression profile of mouse Cdh13
Cdh13 is highly expressed in skeletal muscle.
185 Figure 6.31 Visual protein-protein interaction networks for CDH13 in humans
CDH13 is part of a network including JUP, CTNND1 and other cadherin family members.
186 CDH4
Cadherin 4 or R-cadherin (CDH4) is a protein coding gene located on chromosome 2, which was implicated in the QTL mapping of the total joint scores of >12 month old or
“aged” CC mice. The transcribed protein of CDH4 shares 92.2% identity and 96.2% similarity between mice and humans and is a calcium dependent cell-cell adhesion glycoprotein. It is thought to have a role in neuronal outgrowth and kidney and muscle development. It is highly expressed in skeletal muscle (Figure 6.32) and is part of a PPI network that includes other cadherin family members, JUP, CTNND1 and CDC42
(Figure 6.33). The role of CDH4 is yet to be investigated in OA or other forms of arthritis, although it has been implicated in studies of various cancers including chondrosarcoma
(Niini, Scheinin et al. 2012). A study investigating the role of R-cadherin on cell motility in tumour cells demonstrated that expression of R-cadherin downregulates the expression of other cadherins. The expression of R-cadherin showed an increase in cell motility due to an increase in Rho GTPase activity, which resulted in an increase in Rac1 and Cdc42 activation (Johnson, Theisen et al. 2004). This effect on Rho GTPase activity suggests that R-cadherin may have an upstream effect on a pathway that may contribute to cartilage phenotypes.
Cdh4 was one of the initial genes identified during this study, and hence it was used to validate the association between OA development and changes in gene expression in the
CC mice. We performed a preliminary in vitro experiment to characterize the expression of Cdh4 in bone cells derived from the CC mice. Real-time quantitative PCR was used to observe expression levels in osteoclasts, osteoblasts and whole cartilage in OA affected versus non-affected CC mice.
187
Figure 6.32 Gene expression profile of mouse Cdh4
Cdh4 is highly expressed in skeletal muscle.
188 Figure 6.33 Visual protein-protein interaction networks for CDH4 in humans
CDH4 is part of a network including JUP, CTNND1 and CDC42.
189 Hprt and β-actin were used as nominal expression or “housekeeping” controls for all assays. Cdh4 was found to be expressed during osteoclast formation of cells derived from a non-affected CC strain, with expression peaking at day 3. Osteoclasts derived from the
OA affected strain showed a trend of reduced expression of Cdh4 at day 3 and day 5 when compared with the non-affected control, however this was not found to be statistically significant (>p-0.05). The positive expression marker TRAP showed a significant upward trend of expression in both the OA affected and non-affected samples. Analysis of expression of Cdh4 during osteogenesis showed lower levels of expression in OA derived osteoblasts with a trend of gradual increase which was not found to be statistically significant (>p-0.05). Osteoblasts derived from an unaffected strain showed peak expression at day 7 with a gradual tapering off of expression until day 21. Control expression of Alp showed significantly higher levels of expression in osteoblasts derived from the OA affected strain peaking at day 7, when compared with osteoblasts derived from an unaffected strain (Figure 6.34).
190
A
B
Figure 6.34 Gene expression of murine Cdh4 in osteoclasts and osteoblasts
BMMs were isolated from OA affected (HIP strain) mice and unaffected (LAX strain)
mice. RNA was extracted on day 0, 1, 3 and 5 during osteoclastogenesis and 0, 7, 14
and 21 during osteogenesis. QPCR analysis was carried out to observe the expression
of (A) Cdh4, Trap and (B) Cdh4, Alp relative to day 0 of the unaffected control. All
expression was normalized to housekeeping genes Hprt and β-actin. P<0.05 *,
P<0.01**, P<0.001***
191 6.3 Discussion
This chapter focused on the interrogation of genes implicated in the mapping of early onset, old age and the incidence of spontaneous OA as observed within the CC cohort.
With more than 90 genes implicated in the study, investigation of expression via BioGPS and function or association via Pubmed literature searches and investigation of PPI networks were carried out on all genes implicated in the mapping to narrow down a list of potential gene candidates. The genes that met the criteria for investigation were analysed further using datasets from an in house generated, rabbit chondrocyte differentiation microarray, and qPCR analysis.
Genes identified in this study that have also been associated with human knee OA are considered to be the strongest candidates for the regulation of OA development. Genes showing expression and/or interaction with pathways that are considered to be important to the maintenance of joint homeostasis, but without published data implicating them in
OA directly are considered to be novel candidates. Of the genome wide significant candidates, Col11a2, Glis3 and Il33 were all implicated in human knee OA studies
(Jakkula, Melkoniemi et al. 2005; Casalone, Tachmazidou et al. 2018; Nefla, Holzinger et al. 2016). It is of particular interest that Col11a2 was implicated only in the 8-12 month cohort of CC mice and has been shown to be associated with the early onset of human
OA. This result is an encouraging proof of principle for the use of QTL analysis with the
CC mice to detect genes regulating diseases such as OA. Of the suggestive candidates,
Jak2, Dkk1 and Col9a3 were also implicated in studies of human OA. The study conducted by Jakkula et al that implicated COL11A2 in early onset of OA also showed variance in COL9A3 in the early onset OA cohort, however the variance did not alter known splicing consensus sequences or result in amino acid substitutions in the resulting
192 protein (Jakkula, Melkoniemi et al. 2005). This is also an encouraging proof of principle as Col9a3 was only implicated in the >12 month CC cohort in the study presented in chapter 5. In addition, Cd274, Pdcd1lg2, Tap2 and Daxx, which were also found in the loci reaching GWS, and Epas1 and Cdh13, which were found in the suggestive loci, all demonstrated involvement in other forms of arthritis. This suggests that these genes require further study to better understand their potential role in OA specifically.
The genes classified as novel candidates in this study, Cdh4, Obscn, Tab1 and Myo1f, are encouraging new frontiers in understanding the genetic component of OA. Both Tab1 and
Myo1f showed expression in osteoclasts with Tab1 also showing an involvement in joint development pathways. The involvement of Myo1f, Obscn and Cdh4 in skeletal muscle, adds to the potential role of muscles in the development of OA. The decrease in Cdh4 expression during osteoclastogenesis from OA affected CC mice also raises new questions regarding the role of Cdh4 in the bone-remodeling component of OA. In addition, the effects of Obscn and Cdh4 on Rho GTPases, Rac1 and Cdc42 activation present new questions of the downstream implications of this activation on joint homeostasis.
The genes investigated in this study show function in many tissues and systems. The role of some of these genes have been verified in individual tissues/systems and at times have a suggested role in multiple components of the synovial joint. Glis3, which is a GWS candidate in this study and an equally significant candidate in a recent human study by
Casalone et al (Casalone, Tachmazidou et al. 2018), has been shown to have a role in osteoblast differentiation and is expressed in chondrocytes, whilst having a significant involvement in diseases such as diabetes and hypothyroidism. These diseases exhibit a
193 high body mass index phenotype, which is a known risk factor for OA development
(Grotle, Hagen et al. 2008). A greater focus on genes affecting multiple tissues/systems, particularly off target systems such as the muscle and thyroid will offer a new insight into the genetic risk of OA development.
There were over 90 genes implicated in this study. While all should be considered as
“interesting” and may be worthy of further study, a selection criteria was implemented to prioritize the most significant candidates first. While the majority of genes that did not meet the selection criteria were found in loci crossing only the suggestive threshold, there were 14 genes that reached GWS that were not considered as priority candidates due to the lack of supporting information in the literature. Very little is known about these genes; however, they should be considered as secondary candidates for future studies as they may be completely novel regulators of OA development through unknown mechanisms.
The PPI networks generated by the STRING software were valuable in the characterization process of genes that have not yet been implicated in arthritis studies.
While this resource can offer insight into protein interactions based on online databases, textmining, co-expression and more, confirmation of these interactions and their significance in further studies is required. Broad expression profile information sourced from BioGPS was also an effective way to prioritize primary candidate genes. However, this expression data should not be the only method of defining the significance of potential candidates. While this data is a valuable resource for an efficient review of many genes, investigation into expression of primary candidates in affected cells/tissues via additional microarray targets and qPCR will be required for full profiling of these candidates in future studies.
194
In summary, the number of genes identified using QTL mapping of OA phenotypes in the CC mice has demonstrated the potential of the CC platform for the identification of genes involved in complex diseases such as OA. The use of bioinformatics tools to prioritize potential candidate genes has proven to be effective in verifying known candidates and broadening the knowledge of potentially novel candidates by investigation of expression and interactions with pathways that are relevant to the OA disease process.
195
Chapter 7 General Overview and Future Directions
196 7.1 Overview of thesis and general discussion
Osteoarthritis (OA) is a common and debilitating degenerative joint disease affecting synovial joints. It is categorized by the destruction of articular cartilage, which in turn causes pain, and limited mobility in the affected joint of sufferers. The current and most effective treatments available for the disease remain as pain management and total joint replacement. It is well accepted that the end stage of OA is categorized as the destruction of articular cartilage, however the involvement of subchondral bone, connective tissue and muscle, synovium and the immune system are all involved in the pathogenesis of OA.
Post-traumatic OA is a result of significant synovial joint injury; inducing these kinds of injuries in animal models have proven effective in investigating the disease process
(Poole, Blake et al. 2010). While it is widely accepted that injury is a precursor to many of the pathogenic changes in the associated tissues, not all cases of OA development contain a joint injury component. It has been suggested that spontaneous forms of OA may be a result of diet, BMI and diabetic status. However, the most reliable predictor of
OA risk remains the age of the individual (Blanco, Möller et al. 2015). It has been suggested that there is a heritable component of OA; twin and genome wide association studies have demonstrated a heritability component of OA along with implicating a range of genes found to be associated with hip and knee OA (Spector and MacGregor 2004).
The complete collection of risk loci discovered to date is estimated to account for 26.3% of the total disease trait variance, however these studies lack the ability to control for environmental factors that may contribute to the development of OA and also suffer due to a poorly defined phenotype (Zengini, Hatzikotoulas et al. 2018). Therefore, more studies investigating the genetic component of OA that account for these shortcomings are required.
197
The CC is a large scale funnel breeding project consisting of eight founder strains of mice that represent approximately 90% of the mouse genome. The founders are crossed to generate a large panel of RI strains that exhibit a wide range of phenotypes, which can be assessed for the purpose of investigating the heritable genetic links with the observed phenotypes. The resulting RI strains are generated quickly due to fast breeding of mice, and their diet and environments are easily controlled thus overcoming the limitations of the human GWAS (Churchill, Airey et al. 2004). The closest equivalent study conducted by Watanabe et, al using STR/ort mice suffered from small sample size (n=100) and inferior mapping resolution when compared with the CC (Watanabe, Oue et al. 2012).
There have been no studies undertaken using CC mice to investigate spontaneous OA thus far.
In the present study, over 275 CC mice across 50 strains ages 8 months and older were dissected, decalcified and processed into paraffin blocks, and screened using histology for OA phenotypes in accordance with OARSI mouse grading scale. OA phenotype parameters were analyzed as total joint OA and highest single surface OA in complete, middle aged and old age cohorts, as well as incidence of any OA and incidence of OARSI scores >3 in the complete cohort. Strains with less than two replicates were excluded from the age related cohort and strains with less than three replicates excluded from the complete cohort and incidence studies. This technique proved to be an effective means of assessing OA in mice albeit much more time consuming when compared with higher throughput radiographic techniques. Nevertheless, the use of histology yielded more advantages due to a significant improvement in the definition of the OA phenotype.
198 A maximum of 222 mice across 43 strains were analysed for the purpose of the downstream QTL mapping. A limitation placed upon this study was that the mice that were available were a random selection of individuals across the strains. The complete cohort available by the end of this study consisted of over 900 mice spanning 70 strains.
However, there was not an even age distribution across all strains as the culling of mice and the early retiring of entire strains due to lack of breeding affected the ability to source a representative aged cohort. The initial age threshold of >12 months required to investigate OA reduced the available cohort to 174 mice across 40 strains, with 7 strains not meeting the inclusion criteria of a minimum two mice per strain. It was at this point that the inclusion of an 8-12 month cohort was recommended to assist in powering the study with more individual replicates for strains that had already been screened, as well as the addition of 10 extra strains to increase the mapping precision. The excessive handling of the tissues due to their use in other studies resulted in the degradation of the tissue in some cases. While this was sustainable use of animal tissue, it resulted in a reduction of numbers for downstream applications where fresh and immediate fixation of tissue is required. As the breeding of the CC mice is ongoing, the progressive inclusion of more strains and the ability to power strains that did not have significant replicates is possible. While the increase in the numbers will provide a more representative picture of
OA per strain and per gender, it may in turn affect the mapping by improving the resolution. Regardless of these shortcomings, this study is significantly more powered than the closest equivalent QTL mapping study conducted by Watanabe et, al. and the current CC mouse study investigating BMD by Levy et, al. (Levy, Mott et al. 2015;
Watanabe, Oue et al. 2012).
199 The CC mouse cohort was effective in presenting a spontaneous, measurable OA phenotype that varied in severity. The phenotype was characteristic of those represented in existing mouse models for which the OARSI scoring system was created (Glasson,
Chambers et al. 2010). The CC phenotype was characterized on cartilage destruction alone, which presented in much the same way as human OA is observed, with fibrillation and cracking of the superficial layer and progressive erosion into the deeper layers of cartilage and subchondral bone (Litwic, Edwards et al. 2013; Buckwalter, Mankin et al.
2005). There was a wide range of strains that presented with low range OARSI scores.
As stated by Glasson et, al, due to the thin mouse articular cartilage, OA in the mouse is considered to be an all or nothing phenomenon (Glasson, Chambers et al. 2010).
Therefore, the use of a threshold of OARSI scores >3 when measuring the incidence of
OA in the CC cohort was used to segregate the genuine OA phenotypes from the more arbitrary low scores. After thresholding, 34.88% of the CC cohort presented with OARSI scores >3. This incidence is more than double the recorded incidence of arthritis and four times the number of those presenting with OA specifically in 2014-15 by the Australian
Bureau of Statistics (ABS 2015). Observation of strains presenting with more than 50% of individuals with a score >3 resulted in 4.65% of the CC population presenting with a severe and prevalent OA phenotype, measuring much closer to the national average. This highlights the importance of normalizing the CC dataset. The mapping of severity was conducted with raw scores or “as is” and “normalized” to account for the distribution of scores amongst the cohort. The normalization function allows the mapping software to compare the most significant OA scoring strains against the remainder of the CC cohort.
Preliminary OA screening of the >12 month CC cohort implicated a locus on chromosome 3 containing Zfhx4, Pex2 and Il-7 and a locus on chromosome 2 containing
200 Cdh4 (see supplementary figure 9.1 and 9.2). Early investigation of the potentially novel
Zfhx4 revealed significant expression in differentiating chondrocytes as observed in a rabbit MSC microarray conducted by our laboratory group (see supplementary figure
9.3). The later inclusion of additional replicates resulted in a change of mapping which reduced the significance of the locus containing Zfhx4. Due to the lack of association between Zfhx4 and OA in the later powered version of the study, any expression data relating to Zfhx4 should be approached with caution. However, the Cdh4 locus remained significant and as a result, more potential genes including Col9a3 were revealed at this locus, thus confirming the importance of the number of strains used in QTL mapping of the CC mice.
Candidate genes were identified using QTL mapping of values generated from total joint and single surface grades of middle aged and old aged CC mice as well as OA incidence and OA incidence of >3 OARSI grade. More than 90 genes were implicated from a total of seven loci, two of which were considered to be genome wide significant due to the loci crossing the 95th percentile threshold. The GWS locus on chromosome 19 derived from single surface OA severity in the complete cohort contained Glis3, Rln1, Cd274,
Pdcd1lg2, Ermp1, Ranbp6, Il-33, Uhrf2 and Gldc. After investigation of expression, PPI networks and literature searches, Glis3, Cd274, Pdcd1lg2 and Il-33 were chosen as first priority candidates. The GWS locus on chromosome 17 derived from total joint OA severity in the middle-aged cohort contained Myo1f, Daxx, Wdr46, Col11a2, H2-DMb2,
H2-DMb1, Tap2, H2Ab1, H2-Aa, H2-Eb1, Btnl6, Nelfe and Trim15, a section of this locus is considered to be part of the mouse MHC. After investigation of these genes, Myo1f,
Daxx, Col11a2 and Tap2 were chosen as first priority candidates. The mapping of total joint and single surface yielded only marginally different results, mainly affecting the
201 number of genes included in the locus. The mapping of single surface values resulted in the inclusion of Glis3, suggesting that single severe defects in articular cartilage may be associated with diabetes and hypothyroidism phenotypes, which have been associated with Glis3 (Nogueira, Paula et al. 2013; Lichti-Kaiser, ZeRuth et al. 2014).
Of the genes that were implicated above the suggestive threshold, we identified Cdh4,
Col9a3, Cdh13, Obscn, Tab1, Epas1, Jak2 and Dkk1 as the first priority candidates. Cdh4 and Col9a3 were implicated in a locus on chromosome 2 associated with total joint OA scores in the elderly cohort with Tab1 on chromosome 15 implicated in single surface
OA severity. Epas1 on chromosome 17 was implicated in the mapping of both of these parameters. Obscn on chromosome 11 was implicated in the mapping of incidence of any
OA score with Cdh13 on chromosome 8 being implicated in the mapping of OA incidence with an OARSI score >3. Jak2 and Dkk1 were found on the same locus on chromosome
19 that was associated with total joint OA in the complete cohort using the un-normalized dataset. This has shown that the use of normalization of the dataset within Geneminer is beneficial for identifying more significant genetic traits by redistributing the strain scores and focusing on the difference between those with the most severe phenotypes against the remainder of the population. While this is useful to identify the most significant genetic traits, the inclusion of genes such as Jak2 and Dkk1 in the un-normalized mapping is of interest and need to be considered as worthy of investigation, despite the locus not reaching GWS in this format.
Further investigation of genes using bioinformatics approaches showed that a range of genes that have been previously categorized in human and animal studies. These primary candidate genes encompass all the involved tissues and systems that have been implicated
202 in studies of OA and are a convincing representation of the notion of spontaneous OA being a total joint disease (Figure 7.1). The identification of Col11a2 in the middle-aged cohort supports the findings of Jakkula et, al. identifying SNPs in the COL11A2 gene in a human early onset of OA cohort that may result in changes to the transcribed protein
(Jakkula, Melkoniemi et al. 2005). While the same study also investigated SNPs in
COL9A3, the results showed that while there were mutations present, they were not in protein coding regions and were unlikely to have an effect on protein function.
Interestingly, the present study also implicated Col9a3 in an older cohort. This supports the findings of Jakkula et, al. suggesting that COL9A3 may not contribute to the early onset of OA, although the SNPs that were identified may contribute in a different way, not resulting in an immediate effect. Further studies of the SNPs identified in COL9A3 and their involvement in later onset OA are required. The identification of Il-33 in the present study along with Tap2, Cd274 and Pdcd1lg2 draws more attention to the involvement of inflammatory cytokines along with other genes expressed in the synovial fluid of RA sufferers with OA sufferers. The implication of the mouse MHC in the middle-aged cohort provides further evidence for the involvement of the immune system in the pathogenesis of OA. The identification of Epas1 is a convincing candidate for the re-establishment of endochondral ossification and the tide mark advancement component of OA This study also identified the collection of novel genes Myo1f, Obscn and Cdh4.
These genes were all significantly expressed in skeletal muscle or involved with muscle phenotypes with Myo1f also showing expression in osteoclasts and bone. The role of muscle weakness in OA has been investigated though the genetic link between muscle and OA requires further investigation. Expression studies conducted on osteoclasts derived from OA affected and non-affected CC mice showed a lack of expression of Cdh4
203 Figure 7.1 Candidate gene annotation of an osteoarthritic joint
A list of the candidate genes identified in the present study representing their expression and/or involvement in osteoarthritis.
204 during osteoclastogenesis when compared with the unaffected control, however this was not found to be significant due to the large error. The cause of this error is due to the small sample size and the overall low expression of Cdh4 in bone cells. This expression study requires more biological repeats for statistically powered confirmation of these results.
Unfortunately, a significant limitation of this study was the ability to access particular strains of the right age to source bone marrow to conduct downstream experiments, due to the lack of control over the breeding and culling of mice. Another limitation was the poor viability of the bone marrow of the aged mice. Because of this, we were unsuccessful in growing enough cells to conduct chondrogenesis assays to observe Cdh4 expression during chondrocyte differentiation. Due to time restraints and the inability to readily source the appropriate live CC mice, we were unable to investigate the expression of
Myo1f and Obscn in the same way. Future studies will require the use of younger mice of the same CC strains (and potentially other affected strains) to undertake the cell-based assays. Freshly sourced tissue from older mice will also enable us to optimize and use immunohistochemical and in-situ hybridization techniques to observe the expression of these novel candidates in the affected tissue.
7.2 Future directions
This study has unveiled a range of new avenues for future OA studies. The primary task will be to add more replicates to the established CC strains and to expand the study to include more strains. The secondary task will be to investigate the novel genes identified in this study further in two different aspects. Firstly, the investigation of significance in a large human knee OA dataset, secondly, the functional analysis of these genes via in vitro experiments and in vivo transgenic or knock out mouse models.
205
The increase in replicates in the CC mouse screening will offer a better characterization of phenotypes within strains, especially when considering incidence of the phenotype and the ability to segregate males and females into independent studies. The increase in CC strains included in the study will increase the QTL mapping resolution, therefore reducing the number of potential false-positive genes implicated as a result of larger regions being identified (Valdar, Flint et al. 2006). Additional screening of CC mice will enable the categorization of other OA related phenotypes such as osteophyte formation and subchondral bone changes, which in turn can be used to map loci in the same way. The database of microCT images of the CC mice hind limbs created by our lab group can be used to measure the subchondral bone plate thickness to better understand the subchondral bone-remodelling component of OA. Correlation of these results may offer new insights into genes independently associated with each type of OA phenotype or the potential overlap of genes associated with more than one of these phenotypes. Further investigation of aged CC strains showing no OA phenotypes may be of interest to categorize the genetic basis of OA resistance with age. Similarly, access to younger mice of the OA affected CC strains will enable a clearer understanding of the development of
OA, particularly the strains affected by the early onset of the disease.
The continued breeding of the CC cohort will enable the access to appropriately aged biological replicates for further cell-based approaches to gene verification. The use of younger animals will improve cell viability for the generation of chondrogenesis assays along with osteoclastogenesis and osteogenesis assays in strains of interest. These experiments will reveal in greater detail, the effect of the loci on the phenotypes presented within the strains. The potential to assess the effect on one or many cell types
206 simultaneously is of interest to better understand if functional failures in one tissue give rise to pathology found in others or whether they are isolated events contributing synergistically to exacerbate the severity of the disease. The use of immunohistochemistry and western blotting will also offer insights into the protein expression of the genes identified in this study. The combination of SNPs from particular founders contributing to the candidate gene may result in an altered structure and impaired function of the transcribed protein. This study has shown that these proteins have also been shown in some cases to interact with each other, therefore further investigation into protein structure and function will be required to better understand the roles of the genes identified in this study.
The present study did not have access to live CC mice for the purpose of undertaking behavioural studies. A future study investigating gait or running wheel behaviour at age points will offer a better understanding of the effect of activity on the development of OA in the CC mice, and how the OA phenotype that has been observed affects the mobility of the mice during their lifetime. This will also allow an assessment of body mass per strain to correlate with known OA candidate strains to assess if there is a significant association of BMI with OA. Investigation of diabetes incidence in the CC mice and correlation of OA datasets from the present study will aid in the validation of the link between OA development and diabetes, the involvement of Glis3 with either disease, and if that link has any association with the BMI of the affected individuals.
The most recent and powerful human GWAS study investigating OA conducted by
Zengini et, al. utilized data from the UK Biobank of 30,727 OA cases and 297,191 controls (Zengini, Hatzikotoulas et al. 2018). Published data by Casalone et, al. early this
207 year has provided validation of Glis3 as a candidate gene in the present study (Casalone,
Tachmazidou et al. 2018). While the use of mouse models is of great value for OA research due to their versatility, correlation with human datasets are required to validate the potential candidate genes. Accessing the UK Biobank data for the purpose of correlating the full list of genes identified in the present study with variants identified in the human dataset will provide a more thorough validation of potentially translatable target genes for further investigation.
To gain a deeper understanding of the functional effects of the candidate genes identified in the present study, in vivo experiments using animal models (particularly the mouse) are required. These functional effects can be observed by comparing parameters such as cartilage degradation, subchondral bone changes, inflammation, gait and behaviour, and
BMI. The ease of manipulation of the mouse genome allows for such in vivo studies as global gene knockout, or in cases where global knockouts may be embryonically lethal, conditional knockouts, as well as transgenic overexpression of candidate genes. While each of these techniques has its advantages and limitations, the creation of a viable genetically modified animal, capable of presenting with a consistent spontaneous OA phenotype will not only offer a greater understanding of the functional role of the candidate genes identified in this study, but present researchers with a reliable platform to undertake pharmacological studies. These future studies in conjunction with the candidate genes identified in the present study will aid in the development of treatments and early risk screening targets for OA.
208 Chapter 8 References
209 2017, A. O. A. N. J. R. R. a. r. (2017). Australian Orthopaedic Association National Joint Replacement Registry annual report 2017.
Adams, T., D. Band-Entrup, S. Kuhn, L. Legere, K. Mace, A. Paggi and M. Penney (2013). "Physical therapy management of knee osteoarthritis in the middle-aged athlete." Sports Med Arthrosc 21(1): 2-10.
Ahlen, M., L. Roshani, M. Liden, A. Struglics, L. Rostgard-Christensen and J. Kartus (2015). "Inflammatory cytokines and biomarkers of cartilage metabolism 8 years after anterior cruciate ligament reconstruction: results from operated and contralateral knees." Am J Sports Med 43(6): 1460-1466.
Altman, R., E. Asch, D. Bloch, G. Bole, D. Borenstein, K. Brandt, W. Christy, T. D. Cooke, R. Greenwald, M. Hochberg and et al. (1986). "Development of criteria for the classification and reporting of osteoarthritis. Classification of osteoarthritis of the knee. Diagnostic and Therapeutic Criteria Committee of the American Rheumatism Association." Arthritis Rheum 29(8): 1039-1049.
Aprato, A., S. Risitano, L. Sabatini, M. Giachino, G. Agati and A. Massè (2016). "Cementless total knee arthroplasty." Annals of Translational Medicine 4(7): 129. arc, O. C. and O. C. arc (2012). "Identification of new susceptibility loci for osteoarthritis (arcOGEN): a genome-wide association study." Lancet 380(9844): 815-823.
Aydelotte, M. B. and K. E. Kuettner (1988). "Differences between sub-populations of cultured bovine articular chondrocytes. I. Morphology and cartilage matrix production." Connective tissue research 18(3): 205-222.
Aydin-Son, Y., E. D. Batu, E. Demirkaya, Y. Bilginer, O. Kasapcopur, E. Unsal, M. Alikasifoglu and S. Ozen (2015). "Systems-level analysis of genome wide association study results for a pilot juvenile idiopathic arthritis family study." Turk J Pediatr 57(4): 324-333.
Aylor, D. L., W. Valdar, W. Foulds-Mathes, R. J. Buus, R. A. Verdugo, R. S. Baric, M. T. Ferris, J. A. Frelinger, M. Heise and M. B. Frieman (2011). "Genetic analysis of complex traits in the emerging Collaborative Cross." Genome research 21(8): 1213-1222.
B Blom, A., P. L van Lent, P. M van der Kraan and W. B van den Berg (2010). "To seek shelter from the WNT in osteoarthritis? WNT-signaling as a target for osteoarthritis therapy." Current drug targets 11(5): 620-629.
Bakker, A. D. and J. Klein-Nulend (2012). Osteoblast isolation from murine calvaria and long bones. Bone Research Protocols, Springer: 19-29.
Beak, J. Y., H. S. Kang, Y. S. Kim and A. M. Jetten (2007). "Kruppel-like zinc finger protein Glis3 promotes osteoblast differentiation by regulating FGF18 expression." J Bone Miner Res 22(8): 1234-1244.
210 Bhattacharjee, A., M. Shukla, V. P. Yakubenko, A. Mulya, S. Kundu and M. K. Cathcart (2013). "IL-4 and IL-13 employ discrete signaling pathways for target gene expression in alternatively activated monocytes/macrophages." Free Radic Biol Med 54: 1-16.
Bhattacharjee, M., L. Balakrishnan, S. Renuse, J. Advani, R. Goel, G. Sathe, T. S. Keshava Prasad, B. Nair, R. Jois, S. Shankar and A. Pandey (2016). "Synovial fluid proteome in rheumatoid arthritis." Clin Proteomics 13: 12.
Blanco, F. J., I. Möller, M. Romera, A. Rozadilla, J. A. Sánchez-Lázaro, A. Rodríguez, J. Gálvez, J. Forés, J. Monfort and S. Ojeda (2015). "Improved prediction of knee osteoarthritis progression by genetic polymorphisms: the Arthrotest Study." Rheumatology 54(7): 1236-1243.
Bottomly, D., M. T. Ferris, L. D. Aicher, E. Rosenzweig, A. Whitmore, D. L. Aylor, B. L. Haagmans, L. E. Gralinski, B. G. Bradel-Tretheway and J. T. Bryan (2012). "Expression quantitative trait Loci for extreme host response to influenza a in pre- collaborative cross mice." G3: Genes, Genomes, Genetics 2(2): 213-221.
Boutilier, J. K., R. L. Taylor, R. Ram, E. McNamara, Q. Nguyen, H. Goullée, D. Chandler, M. Mehta, L. Balmer, N. G. Laing, G. Morahan and K. J. Nowak (2017). "Variable cardiac α-actin (Actc1) expression in early adult skeletal muscle correlates with promoter methylation." Biochimica et Biophysica Acta (BBA) - Gene Regulatory Mechanisms 1860(10): 1025-1036.
Boyle, W. J., W. S. Simonet and D. L. Lacey (2003). "Osteoclast differentiation and activation." Nature 423(6937): 337-342.
Breedveld, F. C. (2004). "Osteoarthritis--the impact of a serious disease." Rheumatology (Oxford) 43 Suppl 1: i4-8.
Brehm, W., B. Aklin, T. Yamashita, F. Rieser, T. Trüb, R. Jakob and P. Mainil-Varlet (2006). "Repair of superficial osteochondral defects with an autologous scaffold-free cartilage construct in a caprine model: implantation method and short-term results." Osteoarthritis and cartilage 14(12): 1214-1226.
Briggs, F. B., P. P. Ramsay, E. Madden, J. M. Norris, V. M. Holers, T. R. Mikuls, T. Sokka, M. F. Seldin, P. K. Gregersen, L. A. Criswell and L. F. Barcellos (2010). "Supervised machine learning and logistic regression identifies novel epistatic risk factors with PTPN22 for rheumatoid arthritis." Genes Immun 11(3): 199-208.
Broman, K. W. (2005). "The genomes of recombinant inbred lines." Genetics 169(2): 1133-1146.
Broom, N. D. and C. A. Poole (1982). "A functional-morphological study of the tidemark region of articular cartilage maintained in a non-viable physiological condition." J Anat 135(Pt 1): 65-82.
Buckwalter, J. A., H. J. Mankin and A. J. Grodzinsky (2005). "Articular cartilage and osteoarthritis." Instructional Course Lectures-American Academy of Orthopaedic Surgeons 54: 465.
211
Burr, D. B. (2004). "Anatomy and physiology of the mineralized tissues: Role in the pathogenesis of osteoarthrosis." Osteoarthritis and Cartilage 12(Supplement): 20-30.
Burr, D. B. (2004). "Anatomy and physiology of the mineralized tissues: role in the pathogenesis of osteoarthrosis." Osteoarthritis Cartilage 12 Suppl A: S20-30.
Burr, D. B. and M. A. Gallant (2012). "Bone remodelling in osteoarthritis." Nat Rev Rheumatol 8(11): 665-673.
Casalone, E., I. Tachmazidou, E. Zengini, K. Hatzikotoulas, S. Hackinger, D. Suveges, J. Steinberg, N. W. Rayner, J. M. Wilkinson, K. Panoutsopoulou and E. Zeggini (2018). "A novel variant in GLIS3 is associated with osteoarthritis." Ann Rheum Dis 77(4): 620-623.
Castanon, M. J., G. Walko, L. Winter and G. Wiche (2013). "Plectin-intermediate filament partnership in skin, skeletal muscle, and peripheral nerve." Histochem Cell Biol 140(1): 33-53.
Chapman, K., A. Takahashi, I. Meulenbelt, C. Watson, J. Rodriguez-Lopez, R. Egli, A. Tsezou, K. N. Malizos, M. Kloppenburg, D. Shi, L. Southam, R. van der Breggen, R. Donn, J. Qin, M. Doherty, P. E. Slagboom, G. Wallis, N. Kamatani, Q. Jiang, A. Gonzalez, J. Loughlin and S. Ikegawa (2008). "A meta-analysis of European and Asian cohorts reveals a global role of a functional SNP in the 5' UTR of GDF5 with osteoarthritis susceptibility." Hum Mol Genet 17(10): 1497-1504.
Chen, G., C. Deng and Y.-P. Li (2012). "TGF-β and BMP signaling in osteoblast differentiation and bone formation." International journal of biological sciences 8(2): 272.
Churchill, G. A., D. C. Airey, H. Allayee, J. M. Angel, A. D. Attie, J. Beatty, W. D. Beavis, J. K. Belknap, B. Bennett and W. Berrettini (2004). "The Collaborative Cross, a community resource for the genetic analysis of complex traits." Nature genetics 36(11): 1133.
Clevers, H. (2006). "Wnt/β-catenin signaling in development and disease." Cell 127(3): 469-480.
Cole, B. J. and A. Gomoll (2009). Biologic joint reconstruction: alternatives to arthroplasty, SLACK Incorporated.
Consortium, C. C. (2012). "The genome architecture of the Collaborative Cross mouse genetic reference population." Genetics 190(2): 389-401.
Corr, M. (2008). "Wnt–β-catenin signaling in the pathogenesis of osteoarthritis." Nature Reviews Rheumatology 4(10): 550.
Corrado, A., A. Neve, A. Macchiarola, A. Gaudio, A. Marucci and F. P. Cantatore (2013). "RANKL/OPG ratio and DKK-1 expression in primary osteoblastic cultures from osteoarthritic and osteoporotic subjects." The Journal of rheumatology 40(5): 684-694.
212 Cremer, I., M. C. Dieu-Nosjean, S. Marechal, C. Dezutter-Dambuyant, S. Goddard, D. Adams, N. Winter, C. Menetrier-Caux, C. Sautes-Fridman, W. H. Fridman and C. G. Mueller (2002). "Long-lived immature dendritic cells mediated by TRANCE-RANK interaction." Blood 100(10): 3646-3655.
Cross, M., E. Smith, D. Hoy, S. Nolte, I. Ackerman, M. Fransen, L. Bridgett, S. Williams, F. Guillemin, C. L. Hill, L. L. Laslett, G. Jones, F. Cicuttini, R. Osborne, T. Vos, R. Buchbinder, A. Woolf and L. March (2014). "The global burden of hip and knee osteoarthritis: estimates from the Global Burden of Disease 2010 study." Annals of the Rheumatic Diseases.
Daghestani, H. N., & Kraus, V. B. (2015). Inflammatory biomarkers in osteoarthritis. Osteoarthritis and cartilage, 23(11), 1890-1896.
Dai, D., Y. Chen, P. Ru, X. Zhou, J. Tao, H. Ye, Q. Hong, L. Tang, G. Pan, D. Lin, Q. Gong, Y. Lv, L. Xu and S. Duan (2014). "Significant association between TAP2 polymorphisms and rheumatoid arthritis: a meta-analysis." Diagn Pathol 9: 129.
Dai, S. M., Z. Z. Shan, K. Nishioka and K. Yudoh (2005). "Implication of interleukin 18 in production of matrix metalloproteinases in articular chondrocytes in arthritis: direct effect on chondrocytes may not be pivotal." Ann Rheum Dis 64(5): 735-742.
Darnay, B. G., V. Haridas, J. Ni, P. A. Moore and B. B. Aggarwal (1998). "Characterization of the intracellular domain of receptor activator of NF-kappaB (RANK). Interaction with tumor necrosis factor receptor-associated factors and activation of NF-kappab and c-Jun N-terminal kinase." J Biol Chem 273(32): 20551-20555.
Davidson, E. N. B., D. F. Remst, E. L. Vitters, H. M. van Beuningen, A. B. Blom, M.-J. Goumans, W. B. van den Berg and P. M. van der Kraan (2009). "Increase in ALK1/ALK5 ratio as a cause for elevated MMP-13 expression in osteoarthritis in humans and mice." The Journal of Immunology 182(12): 7937-7945.
Day-Williams, A. G., L. Southam, K. Panoutsopoulou, N. W. Rayner, T. Esko, K. Estrada, H. T. Helgadottir, A. Hofman, T. Ingvarsson, H. Jonsson, A. Keis, H. J. Kerkhof, G. Thorleifsson, N. K. Arden, A. Carr, K. Chapman, P. Deloukas, J. Loughlin, A. McCaskie, W. E. Ollier, S. H. Ralston, T. D. Spector, G. A. Wallis, J. M. Wilkinson, N. Aslam, F. Birell, I. Carluke, J. Joseph, A. Rai, M. Reed, K. Walker, S. A. Doherty, I. Jonsdottir, R. A. Maciewicz, K. R. Muir, A. Metspalu, F. Rivadeneira, K. Stefansson, U. Styrkarsdottir, A. G. Uitterlinden, J. B. van Meurs, W. Zhang, A. M. Valdes, M. Doherty and E. Zeggini (2011). "A variant in MCF2L is associated with osteoarthritis." Am J Hum Genet 89(3): 446-450.
De Las Rivas, J. and C. Fontanillo (2010). "Protein-protein interactions essentials: key concepts to building and analyzing interactome networks." PLoS Comput Biol 6(6): e1000807.
213 Dean, D. D., J. Martel-Pelletier, J. P. Pelletier, D. S. Howell and J. F. Woessner, Jr. (1989). "Evidence for metalloproteinase and metalloproteinase inhibitor imbalance in human osteoarthritic cartilage." J Clin Invest 84(2): 678-685.
Diarra, D., M. Stolina, K. Polzer, J. Zwerina, M. S. Ominsky, D. Dwyer, A. Korb, J. Smolen, M. Hoffmann and C. Scheinecker (2007). "Dickkopf-1 is a master regulator of joint remodeling." Nature medicine 13(2): 156.
Doi, H., K. Nishida, M. Yorimitsu, T. Komiyama, Y. Kadota, T. Tetsunaga, A. Yoshida, S. Kubota, M. Takigawa and T. Ozaki (2008). "Interleukin-4 downregulates the cyclic tensile stress-induced matrix metalloproteinases-13 and cathepsin B expression by rat normal chondrocytes." Acta Med Okayama 62(2): 119-126.
Durrant, C., H. Tayem, B. Yalcin, J. Cleak, L. Goodstadt, F. Pardo-Manuel de Villena, R. Mott and F. A. Iraqi (2011). "Collaborative Cross mice and their power to map host susceptibility to Aspergillus fumigatus infection." Genome Research 21(8): 1239-1248.
Economics, A. (2007). "Painful realities: the economic impact of arthritis in Australia in 2007." Sydney: Arthritis Australia.
Englund, M., A. Guermazi, F. W. Roemer, P. Aliabadi, M. Yang, C. E. Lewis, J. Torner, M. C. Nevitt, B. Sack and D. T. Felson (2009). "Meniscal tear in knees without surgery and the development of radiographic osteoarthritis among middle-aged and elderly persons: The Multicenter Osteoarthritis Study." Arthritis Rheum 60(3): 831-839.
Farahat, M. N., G. Yanni, R. Poston and G. S. Panayi (1993). "Cytokine expression in synovial membranes of patients with rheumatoid arthritis and osteoarthritis." Ann Rheum Dis 52(12): 870-875.
Felson, D. T., R. C. Lawrence, P. A. Dieppe, R. Hirsch, C. G. Helmick, J. M. Jordan, R. S. Kington, N. E. Lane, M. C. Nevitt, Y. Zhang, M. Sowers, T. McAlindon, T. D. Spector, A. R. Poole, S. Z. Yanovski, G. Ateshian, L. Sharma, J. A. Buckwalter, K. D. Brandt and J. F. Fries (2000). "Osteoarthritis: new insights. Part 1: the disease and its risk factors." Ann Intern Med 133(8): 635-646.
Felson, D. T., Y. Zhang, M. T. Hannan, A. Naimark, B. N. Weissman, P. Aliabadi and D. Levy (1995). "The incidence and natural history of knee osteoarthritis in the elderly. The Framingham Osteoarthritis Study." Arthritis Rheum 38(10): 1500-1505.
Flandry, F. and G. Hommel (2011). "Normal Anatomy and Biomechanics of the Knee." Sports Medicine and Arthroscopy Review 19(2): 82-92.
Fosang, A. J., F. M. Rogerson, C. J. East and H. Stanton (2008). "ADAMTS-5: the story so far." Eur Cell Mater 15: 11-26.
Fraile, A., M. D. Collado, L. Mataran, J. Martin and A. Nieto (2000). "TAP1 and TAP2 polymorphism in Spanish patients with ankylosing spondylitis." Exp Clin Immunogenet 17(4): 199-204.
214 Funck-Brentano, T., W. Bouaziz, C. Marty, V. Geoffroy, E. Hay and M. Cohen-Solal (2014). "Dkk-1-mediated inhibition of Wnt signaling in bone ameliorates osteoarthritis in mice." Arthritis Rheumatol 66(11): 3028-3039.
Gatti, D. M., K. L. Svenson, A. Shabalin, L.-Y. Wu, W. Valdar, P. Simecek, N. Goodwin, R. Cheng, D. Pomp and A. Palmer (2014). "Quantitative trait locus mapping methods for diversity outbred mice." G3: Genes, Genomes, Genetics 4(9): 1623-1633.
Gendron, C., M. Kashiwagi, N. H. Lim, J. J. Enghild, I. B. Thogersen, C. Hughes, B. Caterson and H. Nagase (2007). "Proteolytic activities of human ADAMTS-5: comparative studies with ADAMTS-4." J Biol Chem 282(25): 18294-18306.
Glasson, S., M. Chambers, W. Van Den Berg and C. Little (2010). "The OARSI histopathology initiative–recommendations for histological assessments of osteoarthritis in the mouse." Osteoarthritis and cartilage 18: S17-S23.
Glasson, S. S. (2007). "In vivo osteoarthritis target validation utilizing genetically- modified mice." Current drug targets 8(2): 367-376.
Glasson, S. S., R. Askew, B. Sheppard, B. Carito, T. Blanchet, H. L. Ma, C. R. Flannery, D. Peluso, K. Kanki, Z. Yang, M. K. Majumdar and E. A. Morris (2005). "Deletion of active ADAMTS5 prevents cartilage degradation in a murine model of osteoarthritis." Nature 434(7033): 644-648.
Glasson, S. S., R. Askew, B. Sheppard, B. A. Carito, T. Blanchet, H. L. Ma, C. R. Flannery, K. Kanki, E. Wang, D. Peluso, Z. Yang, M. K. Majumdar and E. A. Morris (2004). "Characterization of and osteoarthritis susceptibility in ADAMTS-4-knockout mice." Arthritis Rheum 50(8): 2547-2558.
Gok, H., S. Ergin and G. Yavuzer (2002). "Kinetic and kinematic characteristics of gait in patients with medial knee arthrosis." Acta Orthop Scand 73(6): 647-652.
Goldring, M. B. and S. R. Goldring (2010). "Articular cartilage and subchondral bone in the pathogenesis of osteoarthritis." Annals of the New York Academy of Sciences 1192(1): 230-237.
Goldring, M. B. and K. B. Marcu (2009). "Cartilage homeostasis in health and rheumatic diseases." Arthritis Res Ther 11(3): 224.
Gore, M., K. S. Tai, A. Sadosky, D. Leslie and B. R. Stacey (2012). "Use and costs of prescription medications and alternative treatments in patients with osteoarthritis and chronic low back pain in community-based settings." Pain Pract 12(7): 550-560.
Greene, M. A. and R. F. Loeser (2015). "Aging-related inflammation in osteoarthritis." Osteoarthritis Cartilage 23(11): 1966-1971.
Grotle, M., K. B. Hagen, B. Natvig, F. A. Dahl and T. K. Kvien (2008). "Obesity and osteoarthritis in knee, hip and/or hand: an epidemiological study in the general population with 10 years follow-up." BMC Musculoskelet Disord 9: 132.
215 Guerne, P. A., D. A. Carson and M. Lotz (1990). "IL-6 production by human articular chondrocytes. Modulation of its synthesis by cytokines, growth factors, and hormones in vitro." J Immunol 144(2): 499-505.
Gunnell, L. M., J. H. Jonason, A. E. Loiselle, A. Kohn, E. M. Schwarz, M. J. Hilton and R. J. O'Keefe (2010). "TAK1 regulates cartilage and joint development via the MAPK and BMP signaling pathways." J Bone Miner Res 25(8): 1784-1797.
Guo, X. and X.-F. Wang (2009). "Signaling cross-talk between TGF-β/BMP and other pathways." Cell research 19(1): 71.
Gyurkovska, V., T. Stefanova, P. Dimitrova, S. Danova, R. Tropcheva and N. Ivanovska (2014). "Tyrosine kinase inhibitor tyrphostin AG490 retards chronic joint inflammation in mice." Inflammation 37(4): 995-1005.
Hackinger, S., K. Trajanoska, U. Styrkarsdottir, E. Zengini, J. Steinberg, G. R. S. Ritchie, K. Hatzikotoulas, A. Gilly, E. Evangelou, J. P. Kemp, D. Evans, T. Ingvarsson, H. Jonsson, U. Thorsteinsdottir, K. Stefansson, A. W. McCaskie, R. A. Brooks, J. M. Wilkinson, F. Rivadeneira and E. Zeggini (2017). "Evaluation of shared genetic aetiology between osteoarthritis and bone mineral density identifies SMAD3 as a novel osteoarthritis risk locus." Hum Mol Genet 26(19): 3850-3858.
Halewood, C. and A. A. Amis (2015). "Clinically relevant biomechanics of the knee capsule and ligaments." Knee Surgery, Sports Traumatology, Arthroscopy 23(10): 2789- 2796.
Hannan, M. T., D. T. Felson and T. Pincus (2000). "Analysis of the discordance between radiographic changes and knee pain in osteoarthritis of the knee." J Rheumatol 27(6): 1513-1517.
Hart, P. H., M. J. Ahern, M. D. Smith and J. J. Finlay-Jones (1995). "Regulatory effects of IL-13 on synovial fluid macrophages and blood monocytes from patients with inflammatory arthritis." Clin Exp Immunol 99(3): 331-337.
Helmark, I. C., U. R. Mikkelsen, J. Borglum, A. Rothe, M. C. Petersen, O. Andersen, H. Langberg and M. Kjaer (2010). "Exercise increases interleukin-10 levels both intraarticularly and peri-synovially in patients with knee osteoarthritis: a randomized controlled trial." Arthritis Res Ther 12(4): R126.
Henriksen, K., M. A. Karsdal and T. John Martin (2014). "Osteoclast-Derived Coupling Factors in Bone Remodeling." Calcified Tissue International 94(1): 88-97.
Herzog, W., D. Longino and A. Clark (2003). "The role of muscles in joint adaptation and degeneration." Langenbeck's Archives of Surgery 388(5): 305-315.
Hindorff LA, M. J. E. B. I., Morales J (European Bioinformatics Institute), Junkins HA, Hall PN, Klemm AK, and Manolio TA. "A Catalog of Published Genome-Wide Association Studies. http://www.genome.gov/gwastudies."
216 Ho, K. W. D., M. R. Wallace, K. T. Sibille, E. J. Bartley, Y. Cruz-Almeida, T. L. Glover, C. D. King, B. R. Goodin, A. Addison, J. C. Edberg, R. Staud, L. A. Bradley and R. B. Fillingim (2017). "Single Nucleotide Polymorphism in the COL11A2 Gene Associated with Heat Pain Sensitivity in Knee Osteoarthritis." Mol Pain 13: 1744806917724259.
Hollander, A. P., I. Pidoux, A. Reiner, C. Rorabeck, R. Bourne and A. R. Poole (1995). "Damage to type II collagen in aging and osteoarthritis starts at the articular surface, originates around chondrocytes, and extends into the cartilage with progressive degeneration." J Clin Invest 96(6): 2859-2869.
Honorati, M. C., L. Cattini and A. Facchini (2007). "VEGF production by osteoarthritic chondrocytes cultured in micromass and stimulated by IL-17 and TNF-alpha." Connect Tissue Res 48(5): 239-245.
Horiki, M., T. Imamura, M. Okamoto, M. Hayashi, J. Murai, A. Myoui, T. Ochi, K. Miyazono, H. Yoshikawa and N. Tsumaki (2004). "Smad6/Smurf1 overexpression in cartilage delays chondrocyte hypertrophy and causes dwarfism with osteopenia." The Journal of cell biology 165(3): 433-445.
Hulin-Curtis, S. L., J. L. Bidwell and M. J. Perry (2012). "Evaluation of IL18 and IL18R1 polymorphisms: genetic susceptibility to knee osteoarthritis." Int J Immunogenet 39(2): 106-109.
Hutton, C. W. (1989). "Osteoarthritis: the cause not result of joint failure?" Ann Rheum Dis 48(11): 958-961.
Iannone, F., C. De Bari, F. Dell'Accio, M. Covelli, F. P. Cantatore, V. Patella, G. Lo Bianco and G. Lapadula (2001). "Interleukin-10 and interleukin-10 receptor in human osteoarthritic and healthy chondrocytes." Clin Exp Rheumatol 19(2): 139-145.
Ikeda, T., A. Mabuchi, A. Fukuda, A. Kawakami, Y. Ryo, S. Yamamoto, K. Miyoshi, N. Haga, H. Hiraoka, Y. Takatori, H. Kawaguchi, K. Nakamura and S. Ikegawa (2002). "Association analysis of single nucleotide polymorphisms in cartilage-specific collagen genes with knee and hip osteoarthritis in the Japanese population." J Bone Miner Res 17(7): 1290-1296.
Jakkula, E., M. Melkoniemi, I. Kiviranta, J. Lohiniva, S. S. Raina, M. Perala, M. L. Warman, K. Ahonen, H. Kroger, H. H. Goring and L. Ala-Kokko (2005). "The role of sequence variations within the genes encoding collagen II, IX and XI in non-syndromic, early-onset osteoarthritis." Osteoarthritis Cartilage 13(6): 497-507.
Jeffries, M. A., M. Donica, L. W. Baker, M. E. Stevenson, A. C. Annan, M. B. Humphrey, J. A. James and A. H. Sawalha (2014). "Genome‐Wide DNA Methylation Study Identifies Significant Epigenomic Changes in Osteoarthritic Cartilage." Arthritis & Rheumatology 66(10): 2804-2815.
Jeong, C., J. Y. Lee, J. Kim, H. Chae, H. I. Park, M. Kim, O. H. Kim, P. Kim, Y. K. Lee and J. Jung (2014). "Novel COL9A3 mutation in a family diagnosed with multiple epiphyseal dysplasia: a case report." BMC Musculoskelet Disord 15: 371.
217 Jilka, R. L. (2013). "The relevance of mouse models for investigating age-related bone loss in humans." Journals of Gerontology Series A: Biomedical Sciences and Medical Sciences 68(10): 1209-1217.
Johnson, E., C. S. Theisen, K. R. Johnson and M. J. Wheelock (2004). "R-cadherin influences cell motility via Rho family GTPases." J Biol Chem 279(30): 31041-31049.
Jovanovic, D., J. P. Pelletier, N. Alaaeddine, F. Mineau, C. Geng, P. Ranger and J. Martel- Pelletier (1998). "Effect of IL-13 on cytokines, cytokine receptors and inhibitors on human osteoarthritis synovium and synovial fibroblasts." Osteoarthritis Cartilage 6(1): 40-49.
Jung, Y. K., G. W. Kim, H. R. Park, E. J. Lee, J. Y. Choi, F. Beier and S. W. Han (2013). "Role of interleukin-10 in endochondral bone formation in mice: anabolic effect via the bone morphogenetic protein/Smad pathway." Arthritis Rheum 65(12): 3153-3164.
Kaneko, S., T. Satoh, J. Chiba, C. Ju, K. Inoue and J. Kagawa (2000). "Interleukin-6 and interleukin-8 levels in serum and synovial fluid of patients with osteoarthritis." Cytokines Cell Mol Ther 6(2): 71-79.
Kapoor, M., J. Martel-Pelletier, D. Lajeunesse, J.-P. Pelletier and H. Fahmi (2011). "Role of proinflammatory cytokines in the pathophysiology of osteoarthritis." Nature Reviews Rheumatology 7(1): 33.
Karahan, S., S. A. Kincaid, J. R. Kammermann and J. C. Wright (2001). "Evaluation of the rat stifle joint after transection of the cranial cruciate ligament and partial medial meniscectomy." Comp Med 51(6): 504-512.
Kashiwagi, M., J. J. Enghild, C. Gendron, C. Hughes, B. Caterson, Y. Itoh and H. Nagase (2004). "Altered proteolytic activities of ADAMTS-4 expressed by C-terminal processing." J Biol Chem 279(11): 10109-10119.
Kawaguchi, H. (2009). "Regulation of Osteoarthritis Development by Wnt–β‐catenin Signaling Through the Endochondral Ossification Process." Journal of Bone and Mineral Research 24(1): 8-11.
Kearns, A. E., S. Khosla and P. J. Kostenuik (2008). "Receptor activator of nuclear factor kappaB ligand and osteoprotegerin regulation of bone remodeling in health and disease." Endocr Rev 29(2): 155-192.
Kellgren J, L. J. (1963). Atlas of standard radiographs. The epidemiology of chronic rheumatism. Vol 2. Oxford, Blackwell Scientific Publications.
Kerkhof, H. J., R. J. Lories, I. Meulenbelt, I. Jonsdottir, A. M. Valdes, P. Arp, T. Ingvarsson, M. Jhamai, H. Jonsson, L. Stolk, G. Thorleifsson, G. Zhai, F. Zhang, Y. Zhu, R. van der Breggen, A. Carr, M. Doherty, S. Doherty, D. T. Felson, A. Gonzalez, B. V. Halldorsson, D. J. Hart, V. B. Hauksson, A. Hofman, J. P. Ioannidis, M. Kloppenburg, N. E. Lane, J. Loughlin, F. P. Luyten, M. C. Nevitt, N. Parimi, H. A. Pols, F. Rivadeneira, E. P. Slagboom, U. Styrkarsdottir, A. Tsezou, T. van de Putte, J. Zmuda, T. D. Spector, K. Stefansson, A. G. Uitterlinden and J. B. van Meurs (2010). "A genome-wide
218 association study identifies an osteoarthritis susceptibility locus on chromosome 7q22." Arthritis Rheum 62(2): 499-510.
Khosla, S. (2001). "Minireview: the OPG/RANKL/RANK system." Endocrinology 142(12): 5050-5055.
Koshy, P. J., C. J. Lundy, A. D. Rowan, S. Porter, D. R. Edwards, A. Hogan, I. M. Clark and T. E. Cawston (2002). "The modulation of matrix metalloproteinase and ADAM gene expression in human chondrocytes by interleukin-1 and oncostatin M: a time-course study using real-time quantitative reverse transcription-polymerase chain reaction." Arthritis Rheum 46(4): 961-967.
Kwan Tat, S., N. Amiable, J. P. Pelletier, C. Boileau, D. Lajeunesse, N. Duval and J. Martel-Pelletier (2009). "Modulation of OPG, RANK and RANKL by human chondrocytes and their implication during osteoarthritis." Rheumatology (Oxford) 48(12): 1482-1490.
Lee, J. H., J. B. Fitzgerald, M. A. Dimicco and A. J. Grodzinsky (2005). "Mechanical injury of cartilage explants causes specific time-dependent changes in chondrocyte gene expression." Arthritis Rheum 52(8): 2386-2395.
Leijten, J. C., S. D. Bos, E. B. Landman, N. Georgi, H. Jahr, I. Meulenbelt, J. N. Post, C. A. van Blitterswijk and M. Karperien (2013). "GREM1, FRZB and DKK1 mRNA levels correlate with osteoarthritis and are regulated by osteoarthritis-associated factors." Arthritis Res Ther 15(5): R126.
Lems, W. F. and D. den Uyl (2010). "Exercise-induced changes in interleukin-10 in patients with knee osteoarthritis: new perspectives?" Arthritis Res Ther 12(4): 131.
Levick, J. R. and J. N. McDonald (1995). "Fluid movement across synovium in healthy joints: role of synovial fluid macromolecules." Annals of the Rheumatic Diseases 54(5): 417.
Levy, R., R. F. Mott, F. A. Iraqi and Y. Gabet (2015). "Collaborative cross mice in a genetic association study reveal new candidate genes for bone microarchitecture." BMC Genomics 16: 1013.
Lewek, M., K. Rudolph, M. Axe and L. Snyder-Mackler (2002). "The effect of insufficient quadriceps strength on gait after anterior cruciate ligament reconstruction." Clinical Biomechanics 17(1): 56-63.
Li, K. and C. Thorne (2010). "Adult presentation of Stickler syndrome type III." Clin Rheumatol 29(7): 795-797.
Li, Y. H., G. Tardif, D. Hum, M. Kapoor, H. Fahmi, J. P. Pelletier and J. Martel-Pelletier (2016). "The unfolded protein response genes in human osteoarthritic chondrocytes: PERK emerges as a potential therapeutic target." Arthritis Res Ther 18: 172.
219 Lichti-Kaiser, K., G. ZeRuth and A. M. Jetten (2014). "TRANSCRIPTION FACTOR GLI-SIMILAR 3 (GLIS3): IMPLICATIONS FOR THE DEVELOPMENT OF CONGENITAL HYPOTHYROIDISM." J Endocrinol Diabetes Obes 2(2): 1024.
Lingaraj, K., C. K. Poh and W. Wang (2010). "Vascular endothelial growth factor (VEGF) is expressed during articular cartilage growth and re-expressed in osteoarthritis." Annals Academy of Medicine Singapore 39(5): 399.
Little, C., M. Smith, M. Cake, R. Read, M. Murphy and F. Barry (2010). "The OARSI histopathology initiative–recommendations for histological assessments of osteoarthritis in sheep and goats." Osteoarthritis and Cartilage 18: S80-S92.
Little, C. B., A. Barai, D. Burkhardt, S. M. Smith, A. J. Fosang, Z. Werb, M. Shah and E. W. Thompson (2009). "Matrix metalloproteinase 13-deficient mice are resistant to osteoarthritic cartilage erosion but not chondrocyte hypertrophy or osteophyte development." Arthritis Rheum 60(12): 3723-3733.
Little, C. B., L. Mittaz, D. Belluoccio, F. M. Rogerson, I. K. Campbell, C. T. Meeker, J. F. Bateman, M. A. Pritchard and A. J. Fosang (2005). "ADAMTS-1-knockout mice do not exhibit abnormalities in aggrecan turnover in vitro or in vivo." Arthritis Rheum 52(5): 1461-1472.
Litwic, A., M. Edwards, E. Dennison and C. Cooper (2013). "Epidemiology and Burden of Osteoarthritis." British medical bulletin 105: 185-199.
Livak, K. J. and T. D. Schmittgen (2001). "Analysis of relative gene expression data using real-time quantitative PCR and the 2(-Delta Delta C(T)) Method." Methods 25(4): 402- 408.
Livshits, G., G. Zhai, D. J. Hart, B. S. Kato, H. Wang, F. M. Williams and T. D. Spector (2009). "Interleukin-6 is a significant predictor of radiographic knee osteoarthritis: The Chingford Study." Arthritis Rheum 60(7): 2037-2045.
Loeser, R. F., S. R. Goldring, C. R. Scanzello and M. B. Goldring (2012). "Osteoarthritis: A Disease of the Joint as an Organ." Arthritis and Rheumatism 64(6): 1697-1707.
Loeser, R. F., A. L. Olex, M. A. McNulty, C. S. Carlson, M. F. Callahan, C. M. Ferguson, J. Chou, X. Leng and J. S. Fetrow (2012). "Microarray analysis reveals age-related differences in gene expression during the development of osteoarthritis in mice." Arthritis Rheum 64(3): 705-717.
Lohmander, L. S., P. M. Englund, L. L. Dahl and E. M. Roos (2007). "The long-term consequence of anterior cruciate ligament and meniscus injuries: osteoarthritis." Am J Sports Med 35(10): 1756-1769.
Lubberts, E., L. A. Joosten, F. A. van de Loo, L. A. van den Gersselaar and W. B. van den Berg (2000). "Reduction of interleukin-17-induced inhibition of chondrocyte proteoglycan synthesis in intact murine articular cartilage by interleukin-4." Arthritis Rheum 43(6): 1300-1306.
220 Mackie, E., Y. Ahmed, L. Tatarczuch, K.-S. Chen and M. Mirams (2008). "Endochondral ossification: how cartilage is converted into bone in the developing skeleton." The international journal of biochemistry & cell biology 40(1): 46-62.
Mackie, E. J., L. Tatarczuch and M. Mirams (2011). "The skeleton: a multi-functional complex organ. The growth plate chondrocyte and endochondral ossification." Journal of Endocrinology 211(2): 109-121.
Malfait, A. M., R. Q. Liu, K. Ijiri, S. Komiya and M. D. Tortorella (2002). "Inhibition of ADAM-TS4 and ADAM-TS5 prevents aggrecan degradation in osteoarthritic cartilage." J Biol Chem 277(25): 22201-22208.
Mankin, H. J., H. Dorfman, L. Lippiello and A. Zarins (1971). "Biochemical and metabolic abnormalities in articular cartilage from osteo-arthritic human hips. II. Correlation of morphology with biochemical and metabolic data." J Bone Joint Surg Am 53(3): 523-537.
Mankin, H. J. and L. Lippiello (1970). "Biochemical and metabolic abnormalities in articular cartilage from osteo-arthritic human hips." J Bone Joint Surg Am 52(3): 424- 434.
Mason, R. M., M. G. Chambers, J. Flannelly, J. D. Gaffen, J. Dudhia and M. T. Bayliss (2001). "The STR/ort mouse and its use as a model of osteoarthritis." Osteoarthritis Cartilage 9(2): 85-91.
Mayr, H. O., T. G. Weig and W. Plitz (2004). "Arthrofibrosis following ACL reconstruction--reasons and outcome." Arch Orthop Trauma Surg 124(8): 518-522.
McDermott, I. D. and A. A. Amis (2006). "The consequences of meniscectomy." J Bone Joint Surg Br 88(12): 1549-1556.
McNulty, M. A., R. F. Loeser, C. Davey, M. F. Callahan, C. M. Ferguson and C. S. Carlson (2012). "Histopathology of naturally occurring and surgically induced osteoarthritis in mice." Osteoarthritis and cartilage 20(8): 949-956.
Mehta, V. and L. Trinkle-Mulcahy (2016). "Recent advances in large-scale protein interactome mapping." F1000Res 5.
Meinecke, I., A. Cinski, A. Baier, M. A. Peters, B. Dankbar, A. Wille, A. Drynda, H. Mendoza, R. E. Gay, R. T. Hay, B. Ink, S. Gay and T. Pap (2007). "Modification of nuclear PML protein by SUMO-1 regulates Fas-induced apoptosis in rheumatoid arthritis synovial fibroblasts." Proc Natl Acad Sci U S A 104(12): 5073-5078.
Melchiorri, C., R. Meliconi, L. Frizziero, T. Silvestri, L. Pulsatelli, I. Mazzetti, R. M. Borzi, M. Uguccioni and A. Facchini (1998). "Enhanced and coordinated in vivo expression of inflammatory cytokines and nitric oxide synthase by chondrocytes from patients with osteoarthritis." Arthritis Rheum 41(12): 2165-2174.
Mengshol, J. A., M. P. Vincenti, C. I. Coon, A. Barchowsky and C. E. Brinckerhoff (2000). "Interleukin-1 induction of collagenase 3 (matrix metalloproteinase 13) gene
221 expression in chondrocytes requires p38, c-Jun N-terminal kinase, and nuclear factor kappaB: differential regulation of collagenase 1 and collagenase 3." Arthritis Rheum 43(4): 801-811.
Millward-Sadler, S. J., M. O. Wright, L. W. Davies, G. Nuki and D. M. Salter (2000). "Mechanotransduction via integrins and interleukin-4 results in altered aggrecan and matrix metalloproteinase 3 gene expression in normal, but not osteoarthritic, human articular chondrocytes." Arthritis Rheum 43(9): 2091-2099.
Morahan, G., L. Balmer and D. Monley (2008). "Establishment of "The Gene Mine": a resource for rapid identification of complex trait genes." Mamm Genome 19(6): 390-393.
Mototani, H., A. Iida, Y. Nakamura and S. Ikegawa (2010). "Identification of sequence polymorphisms in CALM2 and analysis of association with hip osteoarthritis in a Japanese population." J Bone Miner Metab 28(5): 547-553.
Mott, R., C. J. Talbot, M. G. Turri, A. C. Collins and J. Flint (2000). "A method for fine mapping quantitative trait loci in outbred animal stocks." Proceedings of the National Academy of Sciences 97(23): 12649-12654.
Mueller, T. D., J. L. Zhang, W. Sebald and A. Duschl (2002). "Structure, binding, and antagonists in the IL-4/IL-13 receptor system." Biochim Biophys Acta 1592(3): 237-250.
Munger, S. C., N. Raghupathy, K. Choi, A. K. Simons, D. M. Gatti, D. A. Hinerfeld, K. L. Svenson, M. P. Keller, A. D. Attie and M. A. Hibbs (2014). "RNA-Seq alignment to individualized genomes improves transcript abundance estimates in multiparent populations." Genetics 198(1): 59-73.
Muraki, S., T. Akune, H. Oka, A. Mabuchi, Y. En-Yo, M. Yoshida, A. Saika, K. Nakamura, H. Kawaguchi and N. Yoshimura (2009). "Association of occupational activity with radiographic knee osteoarthritis and lumbar spondylosis in elderly patients of population-based cohorts: a large-scale population-based study." Arthritis Rheum 61(6): 779-786.
Nakajima, M., D. Shi, J. Dai, A. Tsezou, M. Zheng, P. E. Norman, A. Takahashi, S. Ikegawa and Q. Jiang (2011). "Replication studies in various ethnic populations do not support the association of the HIF-2alpha SNP rs17039192 with knee osteoarthritis." Nat Med 17(1): 26-27; author reply 27-29.
Nasi, S., A. So, C. Combes, M. Daudon and N. Busso (2016). "Interleukin-6 and chondrocyte mineralisation act in tandem to promote experimental osteoarthritis." Ann Rheum Dis 75(7): 1372-1379.
Nefla, M., D. Holzinger, F. Berenbaum and C. Jacques (2016). "The danger from within: alarmins in arthritis." Nat Rev Rheumatol 12(11): 669-683.
Neumann, J., F. C. Hofmann, U. Heilmeier, W. Ashmeik, K. Tang, A. S. Gersing, B. J. Schwaiger, M. C. Nevitt, G. B. Joseph, N. E. Lane, C. E. McCulloch and T. M. Link (2018). "Type 2 diabetes patients have accelerated cartilage matrix degeneration
222 compared to diabetes free controls: data from the Osteoarthritis Initiative." Osteoarthritis Cartilage 26(6): 751-761.
Neve, A., F. P. Cantatore, A. Corrado, A. Gaudio, S. Ruggieri and D. Ribatti (2013). "In vitro and in vivo angiogenic activity of osteoarthritic and osteoporotic osteoblasts is modulated by VEGF and vitamin D3 treatment." Regulatory peptides 184: 81-84.
Niini, T., I. Scheinin, L. Lahti, S. Savola, F. Mertens, J. Hollmen, T. Bohling, A. Kivioja, K. H. Nord and S. Knuutila (2012). "Homozygous deletions of cadherin genes in chondrosarcoma-an array comparative genomic hybridization study." Cancer Genet 205(11): 588-593.
Nogueira, T. C., F. M. Paula, O. Villate, M. L. Colli, R. F. Moura, D. A. Cunha, L. Marselli, P. Marchetti, M. Cnop, C. Julier and D. L. Eizirik (2013). "GLIS3, a susceptibility gene for type 1 and type 2 diabetes, modulates pancreatic beta cell apoptosis via regulation of a splice variant of the BH3-only protein Bim." PLoS Genet 9(5): e1003532.
Olee, T., S. Hashimoto, J. Quach and M. Lotz (1999). "IL-18 is produced by articular chondrocytes and induces proinflammatory and catabolic responses." J Immunol 162(2): 1096-1100.
Paassilta, P., J. Lohiniva, H. H. Goring, M. Perala, S. S. Raina, J. Karppinen, M. Hakala, T. Palm, H. Kroger, I. Kaitila, H. Vanharanta, J. Ott and L. Ala-Kokko (2001). "Identification of a novel common genetic risk factor for lumbar disk disease." Jama 285(14): 1843-1849.
Palomo, J., D. Dietrich, P. Martin, G. Palmer and C. Gabay (2015). "The interleukin (IL)- 1 cytokine family--Balance between agonists and antagonists in inflammatory diseases." Cytokine 76(1): 25-37.
Patil, A. S., R. B. Sable and R. M. Kothari (2012). "Occurrence, biochemical profile of vascular endothelial growth factor (VEGF) isoforms and their functions in endochondral ossification." J Cell Physiol 227(4): 1298-1308.
Pelletier, J. P., J. Martel-Pelletier and S. B. Abramson (2001). "Osteoarthritis, an inflammatory disease: potential implication for the selection of new therapeutic targets." Arthritis Rheum 44(6): 1237-1247.
Pendleton, A., N. Arden, M. Dougados, M. Doherty, B. Bannwarth, J. W. Bijlsma, F. Cluzeau, C. Cooper, P. A. Dieppe, K. P. Gunther, H. J. Hauselmann, G. Herrero- Beaumont, P. M. Kaklamanis, B. Leeb, M. Lequesne, S. Lohmander, B. Mazieres, E. M. Mola, K. Pavelka, U. Serni, B. Swoboda, A. A. Verbruggen, G. Weseloh and I. Zimmermann-Gorska (2000). "EULAR recommendations for the management of knee osteoarthritis: report of a task force of the Standing Committee for International Clinical Studies Including Therapeutic Trials (ESCISIT)." Ann Rheum Dis 59(12): 936-944.
Phillippi, J., Y. Xie, D. R. Miller, T. A. Bell, Z. Zhang, A. B. Lenarcic, D. L. Aylor, S. H. Krovi, D. W. Threadgill and F. P.-M. De Villena (2014). "Using the emerging Collaborative Cross to probe the immune system." Genes and immunity 15(1): 38.
223
Piva, S. R., A. M. Susko, S. S. Khoja, D. A. Josbeno, G. K. Fitzgerald and F. G. Toledo (2015). "Links between osteoarthritis and diabetes: implications for management from a physical activity perspective." Clin Geriatr Med 31(1): 67-87, viii.
Plaas, A., J. Velasco, D. Gorski, J. Li, A. Cole, K. Christopherson and J. Sandy (2011). "The relationship between fibrogenic TGFβ1 signaling in the joint and cartilage degradation in post-injury osteoarthritis." Osteoarthritis and cartilage 19(9): 1081-1090.
Poole, R., S. Blake, M. Buschmann, S. Goldring, S. Laverty, S. Lockwood, J. Matyas, J. McDougall, K. Pritzker and K. Rudolphi (2010). "Recommendations for the use of preclinical models in the study and treatment of osteoarthritis." Osteoarthritis and Cartilage 18: S10-S16.
Poree, B., M. Kypriotou, C. Chadjichristos, G. Beauchef, E. Renard, F. Legendre, M. Melin, S. Gueret, D. J. Hartmann, F. Mallein-Gerin, J. P. Pujol, K. Boumediene and P. Galera (2008). "Interleukin-6 (IL-6) and/or soluble IL-6 receptor down-regulation of human type II collagen gene expression in articular chondrocytes requires a decrease of Sp1.Sp3 ratio and of the binding activity of both factors to the COL2A1 promoter." J Biol Chem 283(8): 4850-4865.
Porter, S., I. M. Clark, L. Kevorkian and D. R. Edwards (2005). "The ADAMTS metalloproteinases." Biochem J 386(Pt 1): 15-27.
Proffen, B. L., M. McElfresh, B. C. Fleming and M. M. Murray (2012). "A comparative anatomical study of the human knee and six animal species." The Knee 19(4): 493-499.
Ram, R., M. Mehta, L. Balmer, D. M. Gatti and G. Morahan (2014). "Rapid identification of major-effect genes using the collaborative cross." Genetics 198(1): 75-86.
Reich, D. E. and E. S. Lander (2001). "On the allelic spectrum of human disease." Trends in Genetics 17(9): 502-510.
Roberts, V. I., C. N. A. Esler and W. M. Harper (2007). "A 15-year follow-up study of 4606 primary total knee replacements." Journal of Bone & Joint Surgery, British Volume 89-B(11): 1452.
Rodriguez-Fontenla, C., M. Calaza, E. Evangelou, A. M. Valdes, N. Arden, F. J. Blanco, A. Carr, K. Chapman, P. Deloukas, M. Doherty, T. Esko, C. M. Garces Aleta, J. J. Gomez- Reino Carnota, H. Helgadottir, A. Hofman, I. Jonsdottir, H. J. Kerkhof, M. Kloppenburg, A. McCaskie, E. E. Ntzani, W. E. Ollier, N. Oreiro, K. Panoutsopoulou, S. H. Ralston, Y. F. Ramos, J. A. Riancho, F. Rivadeneira, P. E. Slagboom, U. Styrkarsdottir, U. Thorsteinsdottir, G. Thorleifsson, A. Tsezou, A. G. Uitterlinden, G. A. Wallis, J. M. Wilkinson, G. Zhai, Y. Zhu, D. T. Felson, J. P. Ioannidis, J. Loughlin, A. Metspalu, I. Meulenbelt, K. Stefansson, J. B. van Meurs, E. Zeggini, T. D. Spector and A. Gonzalez (2014). "Assessment of osteoarthritis candidate genes in a meta-analysis of nine genome- wide association studies." Arthritis Rheumatol 66(4): 940-949.
224 Rolita, L., A. Spegman, X. Tang and B. N. Cronstein (2013). "Greater number of narcotic analgesic prescriptions for osteoarthritis is associated with falls and fractures in elderly adults." J Am Geriatr Soc 61(3): 335-340.
Roman-Blas, J. A. and S. A. Jimenez (2006). "NF-κB as a potential therapeutic target in osteoarthritis and rheumatoid arthritis." Osteoarthritis and Cartilage 14(9): 839-848.
Roos, E. M. and N. K. Arden (2015). "Strategies for the prevention of knee osteoarthritis." 12: 92.
Roos, H., M. Lauren, T. Adalberth, E. M. Roos, K. Jonsson and L. S. Lohmander (1998). "Knee osteoarthritis after meniscectomy: prevalence of radiographic changes after twenty-one years, compared with matched controls." Arthritis Rheum 41(4): 687-693.
Rossi, D., J. Palmio, A. Evilä, L. Galli, V. Barone, T. A. Caldwell, R. A. Policke, E. Aldkheil, C. E. Berndsen, N. T. Wright, E. Malfatti, G. Brochier, E. Pierantozzi, A. Jordanova, V. Guergueltcheva, N. B. Romero, P. Hackman, B. Eymard, B. Udd and V. Sorrentino (2017). "A novel FLNC frameshift and an OBSCN variant in a family with distal muscular dystrophy." PLoS ONE 12(10): e0186642.
Rual, J. F., K. Venkatesan, T. Hao, T. Hirozane-Kishikawa, A. Dricot, N. Li, G. F. Berriz, F. D. Gibbons, M. Dreze, N. Ayivi-Guedehoussou, N. Klitgord, C. Simon, M. Boxem, S. Milstein, J. Rosenberg, D. S. Goldberg, L. V. Zhang, S. L. Wong, G. Franklin, S. Li, J. S. Albala, J. Lim, C. Fraughton, E. Llamosas, S. Cevik, C. Bex, P. Lamesch, R. S. Sikorski, J. Vandenhaute, H. Y. Zoghbi, A. Smolyar, S. Bosak, R. Sequerra, L. Doucette- Stamm, M. E. Cusick, D. E. Hill, F. P. Roth and M. Vidal (2005). "Towards a proteome- scale map of the human protein-protein interaction network." Nature 437(7062): 1173- 1178.
Runstadler, J. A., H. Saila, A. Savolainen, M. Leirisalo-Repo, K. Aho, E. Tuomilehto- Wolf, J. Tuomilehto and M. F. Seldin (2004). "HLA-DRB1, TAP2/TAP1, and HLA- DPB1 haplotypes in Finnish juvenile idiopathic arthritis: more complexity within the MHC." Genes Immun 5(7): 562-571.
Saetan, N., S. Honsawek, A. Tanavalee, P. Yuktanandana, S. Meknavin, S. Ngarmukos, T. Tanpowpong and V. Parkpian (2014). "Relationship of plasma and synovial fluid vascular endothelial growth factor with radiographic severity in primary knee osteoarthritis." International orthopaedics 38(5): 1099-1104.
Sahebjam, S., R. Khokha and J. S. Mort (2007). "Increased collagen and aggrecan degradation with age in the joints of Timp3(-/-) mice." Arthritis Rheum 56(3): 905-909.
Saito, T., A. Fukai, A. Mabuchi, T. Ikeda, F. Yano, S. Ohba, N. Nishida, T. Akune, N. Yoshimura, T. Nakagawa, K. Nakamura, K. Tokunaga, U. I. Chung and H. Kawaguchi (2010). "Transcriptional regulation of endochondral ossification by HIF-2alpha during skeletal growth and osteoarthritis development." Nat Med 16(6): 678-686.
Sanchez, C., M. A. Deberg, N. Piccardi, P. Msika, J. Y. L. Reginster and Y. E. Henrotin (2005). "Osteoblasts from the sclerotic subchondral bone downregulate aggrecan but upregulate metalloproteinases expression by chondrocytes. This effect is mimicked by
225 interleukin-6, -1β and oncostatin M pre-treated non-sclerotic osteoblasts." Osteoarthritis and Cartilage 13(11): 979-987.
Sanchez, C., M. A. Deberg, N. Piccardi, P. Msika, J. Y. L. Reginster and Y. E. Henrotin (2005). "Subchondral bone osteoblasts induce phenotypic changes in human osteoarthritic chondrocytes." Osteoarthritis and Cartilage 13(11): 988-997.
Scanzello, C. R., E. Umoh, F. Pessler, C. Diaz-Torne, T. Miles, E. Dicarlo, H. G. Potter, L. Mandl, R. Marx, S. Rodeo, S. R. Goldring and M. K. Crow (2009). "Local cytokine profiles in knee osteoarthritis: elevated synovial fluid interleukin-15 differentiates early from end-stage disease." Osteoarthritis Cartilage 17(8): 1040-1048.
Schiller, A. (1995). "Pathology of osteoarthritis." Osteoarthritic disorders: 95-101.
Scholz, R., C. L. Sidler, R. F. Thali, N. Winssinger, P. C. Cheung and D. Neumann (2010). "Autoactivation of transforming growth factor beta-activated kinase 1 is a sequential bimolecular process." J Biol Chem 285(33): 25753-25766.
Seeman, E. and P. D. Delmas (2006). "Bone quality--the material and structural basis of bone strength and fragility." N Engl J Med 354(21): 2250-2261.
Shapiro, F. (2008). "Bone development and its relation to fracture repair. The role of mesenchymal osteoblasts and surface osteoblasts." Eur Cell Mater 15(53): e76.
Sims, N. A. and T. J. Martin (2014). "Coupling the activities of bone formation and resorption: a multitude of signals within the basic multicellular unit." BoneKEy Reports 3: 481.
Slemenda, C., K. D. Brandt, D. K. Heilman and et al. (1997). "QUadriceps weakness and osteoarthritis of the knee." Annals of Internal Medicine 127(2): 97-104.
Smith, M. D. (2011). "The Normal Synovium." The Open Rheumatology Journal 5: 100- 106.
Smith, M. D., E. Barg, H. Weedon, V. Papengelis, T. Smeets, P. P. Tak, M. Kraan, M. Coleman and M. J. Ahern (2003). "Microarchitecture and protective mechanisms in synovial tissue from clinically and arthroscopically normal knee joints." Ann Rheum Dis 62(4): 303-307.
Spector, T. D., F. Cicuttini, J. Baker, J. Loughlin and D. Hart (1996). "Genetic influences on osteoarthritis in women: a twin study." BMJ 312(7036): 940-943.
Spector, T. D. and A. J. MacGregor (2004). "Risk factors for osteoarthritis: genetics." Osteoarthritis Cartilage 12 Suppl A: S39-44.
Srikanth, V. K., J. L. Fryer, G. Zhai, T. M. Winzenberg, D. Hosmer and G. Jones (2005). "A meta-analysis of sex differences prevalence, incidence and severity of osteoarthritis." Osteoarthritis Cartilage 13(9): 769-781.
226 Srivastava, I., N. Thukral and Y. Hasija (2015). "GENETICS OF HUMAN AGE RELATED DISORDERS." Adv Gerontol 28(2): 228-247.
Stambough, J. B., J. C. Clohisy, R. L. Barrack, R. M. Nunley and J. A. Keeney (2014). "Increased risk of failure following revision total knee replacement in patients aged 55 years and younger." Bone Joint J 96-b(12): 1657-1662.
STATISTICS, A. B. O. (2015). "Australian Bureau of Statistics National Health Survey, 2015."
Sugita, S., Y. Hosaka, K. Okada, D. Mori, F. Yano, H. Kobayashi, Y. Taniguchi, Y. Mori, T. Okuma, S. H. Chang, M. Kawata, S. Taketomi, H. Chikuda, H. Akiyama, R. Kageyama, U. I. Chung, S. Tanaka, H. Kawaguchi, S. Ohba and T. Saito (2015). "Transcription factor Hes1 modulates osteoarthritis development in cooperation with calcium/calmodulin-dependent protein kinase 2." Proc Natl Acad Sci U S A 112(10): 3080-3085.
Suzuki, W., A. Yamada, R. Aizawa, D. Suzuki, H. Kassai, T. Harada, M. Nakayama, R. Nagahama, K. Maki and S. Takeda (2015). "Cdc42 is critical for cartilage development during endochondral ossification." Endocrinology 156(1): 314-322.
Szklarczyk, D., A. Franceschini, S. Wyder, K. Forslund, D. Heller, J. Huerta-Cepas, M. Simonovic, A. Roth, A. Santos, K. P. Tsafou, M. Kuhn, P. Bork, L. J. Jensen and C. von Mering (2015). "STRING v10: protein-protein interaction networks, integrated over the tree of life." Nucleic Acids Res 43(Database issue): D447-452.
Takayanagi, H., S. Kim, T. Koga, H. Nishina, M. Isshiki, H. Yoshida, A. Saiura, M. Isobe, T. Yokochi, J.-i. Inoue, E. F. Wagner, T. W. Mak, T. Kodama and T. Taniguchi (2002). "Induction and Activation of the Transcription Factor NFATc1 (NFAT2) Integrate RANKL Signaling in Terminal Differentiation of Osteoclasts." Developmental Cell 3(6): 889-901.
Talabot-Ayer, D., T. McKee, P. Gindre, S. Bas, D. L. Baeten, C. Gabay and G. Palmer (2012). "Distinct serum and synovial fluid interleukin (IL)-33 levels in rheumatoid arthritis, psoriatic arthritis and osteoarthritis." Joint Bone Spine 79(1): 32-37.
Tanamas, S., F. S. Hanna, F. M. Cicuttini, A. E. Wluka, P. Berry and D. M. Urquhart (2009). "Does knee malalignment increase the risk of development and progression of knee osteoarthritis? A systematic review." Arthritis Care & Research 61(4): 459-467.
Umulis, D., M. B. O'Connor and S. S. Blair (2009). "The extracellular regulation of bone morphogenetic protein signaling." Development 136(22): 3715-3728.
Valdar, W., J. Flint and R. Mott (2006). "Simulating the collaborative cross: power of quantitative trait loci detection and mapping resolution in large sets of recombinant inbred strains of mice." Genetics 172(3): 1783-1797.
Valdes, A. M., E. Evangelou, H. J. Kerkhof, A. Tamm, S. A. Doherty, K. Kisand, A. Tamm, I. Kerna, A. Uitterlinden, A. Hofman, F. Rivadeneira, C. Cooper, E. M. Dennison, W. Zhang, K. R. Muir, J. P. Ioannidis, M. Wheeler, R. A. Maciewicz, J. B. van Meurs,
227 N. K. Arden, T. D. Spector and M. Doherty (2011). "The GDF5 rs143383 polymorphism is associated with osteoarthritis of the knee with genome-wide statistical significance." Ann Rheum Dis 70(5): 873-875.
Valdes, A. M. and T. D. Spector (2008). "The contribution of genes to osteoarthritis." Rheum Dis Clin North Am 34(3): 581-603. van der Kraan, P. M., M.-J. Goumans, E. B. Davidson and P. Ten Dijke (2012). "Age- dependent alteration of TGF-β signalling in osteoarthritis." Cell and tissue research 347(1): 257-265.
Voigt, J. D. and M. Mosier (2011). "Hydroxyapatite (HA) coating appears to be of benefit for implant durability of tibial components in primary total knee arthroplasty." Acta Orthop 82(4): 448-459.
Walker, J. (1996). "Cartilage of human joints and related structures." Athletic injuries and rehabilitation: 120-151.
Wang, G. and F. Beier (2005). "Rac1/Cdc42 and RhoA GTPases antagonistically regulate chondrocyte proliferation, hypertrophy, and apoptosis." Journal of bone and mineral research 20(6): 1022-1031.
Wang, G., A. Woods, S. Sabari, L. Pagnotta, L.-A. Stanton and F. Beier (2004). "RhoA/ROCK signaling suppresses hypertrophic chondrocyte differentiation." Journal of Biological Chemistry 279(13): 13205-13214.
Wang, Y. and S. Lou (2001). "Direct protective effect of interleukin-10 on articular chondrocytes in vitro." Chin Med J (Engl) 114(7): 723-725.
Watanabe, K., Y. Oue, Y. Miyamoto, M. Matsuura, Y. Mizuno and S. Ikegawa (2012). "Identification of a quantitative trait locus for spontaneous osteoarthritis in STR/ort mice." J Orthop Res 30(1): 15-20.
Wise, L. M., M. K. Inder, N. C. Real, G. S. Stuart, S. B. Fleming and A. A. Mercer (2012). "The vascular endothelial growth factor (VEGF)‐E encoded by orf virus regulates keratinocyte proliferation and migration and promotes epidermal regeneration." Cellular microbiology 14(9): 1376-1390.
Wojdasiewicz, P., L. A. Poniatowski and D. Szukiewicz (2014). "The role of inflammatory and anti-inflammatory cytokines in the pathogenesis of osteoarthritis." Mediators Inflamm 2014: 561459.
Wong, B. R., D. Besser, N. Kim, J. R. Arron, M. Vologodskaia, H. Hanafusa and Y. Choi (1999). "TRANCE, a TNF family member, activates Akt/PKB through a signaling complex involving TRAF6 and c-Src." Molecular cell 4(6): 1041-1049.
Wu, Q., M. Wang, M. J. Zuscik, D. Chen, R. J. O'Keefe and R. N. Rosier (2008). "Regulation of embryonic endochondral ossification by Smurf2." Journal of Orthopaedic Research 26(5): 704-712.
228 Yamanishi, Y., D. L. Boyle, M. Clark, R. A. Maki, M. D. Tortorella, E. C. Arner and G. S. Firestein (2002). "Expression and regulation of aggrecanase in arthritis: the role of TGF-beta." J Immunol 168(3): 1405-1412.
Yang, S., J. Kim, J.-H. Ryu, H. Oh, C.-H. Chun, B. J. Kim, B. H. Min and J.-S. Chun (2010). "Hypoxia-inducible factor-2α is a catabolic regulator of osteoarthritic cartilage destruction." Nature medicine 16(6): 687.
Ye, J., G. Coulouris, I. Zaretskaya, I. Cutcutache, S. Rozen and T. L. Madden (2012). "Primer-BLAST: a tool to design target-specific primers for polymerase chain reaction." BMC Bioinformatics 13: 134.
Yigit, S., A. Inanir, A. Tekcan, E. Tural, G. T. Ozturk, G. Kismali and N. Karakus (2014). "Significant association of interleukin-4 gene intron 3 VNTR polymorphism with susceptibility to knee osteoarthritis." Gene 537(1): 6-9.
Yu, Q., J. X. Gu, C. Kovacs, J. Freedman, E. K. Thomas and M. A. Ostrowski (2003). "Cooperation of TNF family members CD40 ligand, receptor activator of NF-kappa B ligand, and TNF-alpha in the activation of dendritic cells and the expansion of viral specific CD8+ T cell memory responses in HIV-1-infected and HIV-1-uninfected individuals." J Immunol 170(4): 1797-1805.
Zeggini, E., K. Panoutsopoulou, L. Southam, N. W. Rayner, A. G. Day-Williams, M. C. Lopes, V. Boraska, T. Esko, E. Evangelou, A. Hoffman, J. J. Houwing-Duistermaat, T. Ingvarsson, I. Jonsdottir, H. Jonnson, H. J. Kerkhof, M. Kloppenburg, S. D. Bos, M. Mangino, S. Metrustry, P. E. Slagboom, G. Thorleifsson, E. V. Raine, M. Ratnayake, M. Ricketts, C. Beazley, H. Blackburn, S. Bumpstead, K. S. Elliott, S. E. Hunt, S. C. Potter, S. Y. Shin, V. K. Yadav, G. Zhai, K. Sherburn, K. Dixon, E. Arden, N. Aslam, P. K. Battley, I. Carluke, S. Doherty, A. Gordon, J. Joseph, R. Keen, N. C. Koller, S. Mitchell, F. O'Neill, E. Paling, M. R. Reed, F. Rivadeneira, D. Swift, K. Walker, B. Watkins, M. Wheeler, F. Birrell, J. P. Ioannidis, I. Meulenbelt, A. Metspalu, A. Rai, D. Salter, K. Stefansson, U. Stykarsdottir, A. G. Uitterlinden, J. B. van Meurs, K. Chapman, P. Deloukas, W. E. Ollier, G. A. Wallis, N. Arden, A. Carr, M. Doherty, A. McCaskie, J. M. Willkinson, S. H. Ralston, A. M. Valdes, T. D. Spector and J. Loughlin (2012). "Identification of new susceptibility loci for osteoarthritis (arcOGEN): a genome-wide association study." Lancet 380(9844): 815-823.
Zengini, E., K. Hatzikotoulas, I. Tachmazidou, J. Steinberg, F. P. Hartwig, L. Southam, S. Hackinger, C. G. Boer, U. Styrkarsdottir, A. Gilly, D. Suveges, B. Killian, T. Ingvarsson, H. Jonsson, G. C. Babis, A. McCaskie, A. G. Uitterlinden, J. B. J. van Meurs, U. Thorsteinsdottir, K. Stefansson, G. Davey Smith, J. M. Wilkinson and E. Zeggini (2018). "Genome-wide analyses using UK Biobank data provide insights into the genetic architecture of osteoarthritis." Nat Genet 50(4): 549-558.
Zhang, H., C. C. Liew and K. W. Marshall (2002). "Microarray analysis reveals the involvement of beta-2 microglobulin (B2M) in human osteoarthritis." Osteoarthritis Cartilage 10(12): 950-960.
229 Zhang, Q., Q. Ji, X. Wang, L. Kang, Y. Fu, Y. Yin, Z. Li, Y. Liu, X. Xu and Y. Wang (2015). "SOX9 is a regulator of ADAMTSs-induced cartilage degeneration at the early stage of human osteoarthritis." Osteoarthritis Cartilage 23(12): 2259-2268.
Zhang, Z. M., C. Shen, H. Li, Q. Fan, J. Ding, F. C. Jin and L. Sha (2016). "Leptin induces the apoptosis of chondrocytes in an in vitro model of osteoarthritis via the JAK2STAT3 signaling pathway." Mol Med Rep 13(4): 3684-3690.
Zhong, L., X. Huang, M. Karperien and J. N. Post (2016). "Correlation between Gene Expression and Osteoarthritis Progression in Human." Int J Mol Sci 17(7).
230 Appendix
231 Supplementary Table 9.1 Strain sample sizes 8-12 Strain months >12 months Total M F M F BEW 0 0 4 2 6 BEM 3 3 0 0 6 BOON 3 1 0 2 6 CAMERON 0 2 0 1 3 CIS 0 0 2 6 8 CIS2 0 0 2 2 4 CIV2 0 0 1 1 2 DAVIS 0 0 3 2 5 DET3 2 2 0 1 5 DONNELL 0 0 3 3 6 FUF 0 1 2 2 5 GALASUPREME 0 1 2 2 5 GAV 3 3 0 0 6 GEK2 3 3 0 0 6 GET 0 0 1 3 4 HAX2 0 3 0 0 3 HAZ 3 3 0 0 6 HIP 0 4 1 1 6 HOE 1 0 0 3 4 JAFFA 0 0 2 2 4 JEUNE 0 3 0 0 3 JUD 1 1 1 3 6 JUNIOR 0 0 2 2 4 MAK 0 0 3 0 3 MEE 0 0 1 1 2 MERCURI 0 0 2 0 2 MOK 0 0 1 1 2 MOP 2 5 1 1 9 PEF 2 3 0 0 5 PEF2 0 0 3 4 7 PER2 1 2 1 1 5 PIPING 3 3 0 0 6 POH 1 2 2 1 6 POT 0 0 1 0 1 RAE2 0 0 2 6 8 REV 0 0 0 4 4 ROGAN 0 0 3 4 7 SAT 1 2 1 1 5 SOLDIER 0 0 1 3 4 STUCKY 2 1 1 0 4
232 TAS 0 1 0 0 1 TOFU 1 2 1 1 5 VIT 3 2 0 0 5 WAB2 0 0 2 3 5 WAD 0 2 0 1 3 WOB2 1 2 1 0 4 WOT2 0 0 1 8 9 YID 3 3 0 0 6 ZIF2 2 0 0 3 5 ZOE 0 0 1 0 1 Total 101 136 237 Samples unfit for use 38 Total samples 275
233
Supplementary Figure 9.1 QTL mapping of single surface score >4, preliminary cohort
QTL plot showing a peak on chromosome 3 crossing the 37th percentile between 0 and 8.5 mbp containing Zfhx4, Pex2 and Il7.
234
Supplementary Figure 9.2 QTL mapping of single surface score, preliminary cohort
QTL plot showing a peak on chromosome 2 crossing the 37th percentile between
178.74 and 180 mbp containing Cdh4.
235
Supplementary Figure 9.3 Expression of Zfhx4 during chondrogenesis from rabbit derived MSCs
Zfhx4 is expressed during chondrocyte differentiation with peak expression at day
14. Col2a1 expression confirms successful chondrocyte differentiation.
236