Bioinformatics and Molecular Approaches for the Development of Biological Artificial Cartilage
By
Nguyen P.T. Huynh
Department of Cell Biology Duke University
Date: ______Approved:
______Farshid Guilak, Supervisor
______Blanche Capel, Chair
______Michel Bagnat
______Charles Gersbach
______Eda Yildirim
Dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the Department of Cell Biology in the Graduate School of Duke University
2018 ABSTRACT
Bioinformatics and Molecular Approaches for the Development of Biological Artificial Cartilage
By
Nguyen P.T. Huynh
Department of Cell Biology Duke University
Date: ______Approved:
______Farshid Guilak, Supervisor
______Blanche Capel, Chair
______Michel Bagnat
______Charles Gersbach
______Eda Yildirim
An abstract of a dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the Department of Cell Biology in the Graduate School of Duke University
2018
Copyright by Nguyen P.T. Huynh 2018
Abstract Osteoarthritis (OA) is one of the leading causes of disability in the United
States, afflicting over 27 million Americans and imposing an economic burden of more than $128 billion each year (1, 2). OA is characterized by progressive degeneration of articular cartilage together with sub-chondral bone remodeling and synovial joint inflammation. Currently, OA treatments are limited and inadequate to restore the joint to its full functionality.
Over the years, progress has been made to create biologic cartilage substitutes. However, the repair of degenerated cartilage remains challenging due to its complex architecture and limited capability to integrate with surrounding tissues. Hence, there exists a need to create not only functional chondral constructs, but functional osteochondral constructs, which could potentially enhance affixing properties of cartilage implants utilizing the underlying bone. Furthermore, the molecular mechanisms driving chondrogenesis are still not fully understood. Therefore, detailed transcriptomic profiling would bring forth the progression of not only genes, but gene entities and networks that orchestrate this process.
Bone-marrow derived mesenchymal stem cells (MSCs) are routinely utilized to create cartilage constructs in vitro for the study of chondrogenesis. In this work, we set out to examine the underlying mechanisms of these cells, as well as the intricate gene correlation networks over the time course of lineage development. We first asked the question of how transforming growth factors are
iv determining MSC differentiation, and subsequently utilized genetic engineering to manipulate this pathway to create an osteochondral construct. Next, we performed high-throughput next-generation sequencing to profile the dynamics of
MSC transcriptomes over the time course of chondrogenesis. Bioinformatics analyses of these big data have yielded a multitude of information: the chondrogenic functional module, the associated gene ontologies, and finally the elucidation of GRASLND and its crucial function in chondrogenesis. We extended our results with a detailed molecular characterization of GRASLND and its underlying mechanisms. We showed that GRASLND could enhance chondrogenesis, and thus proposed its therapeutic use in cartilage tissue engineering as well as in the treatment of OA.
v
Dedication
“We are all in the gutter, but some of us are looking at the stars.” – Oscar Wilde
To the good fights.
To the perseverance.
To the American Dream.
To my 19-year-old self.
To Ely, and to Ely’s Appa.
vi
Table of Contents
Abstract ...... iv
List of Tables ...... xii
List of Figures ...... xiii
List of Abbreviations ...... xv
Acknowledgements ...... xviii
1. Introduction ...... 1
1.1. Cartilage and Bone Development ...... 1
1.2. Osteoarthritis and Cartilage Regeneration ...... 3
1.3. The human genome project and the discovery of long non-coding RNAs . 5
1.4. The study and identification of functional lncRNAs ...... 6
1.4.1. Identification of novel lncRNAs ...... 7
1.4.2. Identifying clues to lncRNA function ...... 8
1.4.3. Validation of lncRNA function through reverse genetics ...... 11
1.5. Long non-coding RNAs in the regulation of stem cell maintenance or differentiation ...... 12
1.6. Long non-coding RNAs in the regulation of chondrocytic and osteoblastic differentiation ...... 15
1.6.1. LncRNAs regulating chondrocyte differentiation ...... 15
1.6.2. LncRNAs regulating osteoblast differentiation ...... 21
1.7. Long non-coding RNAs in cartilage- and bone-related diseases ...... 25
1.7.1. LncRNAs in cartilage homeostasis and osteoarthritis ...... 25
1.7.2. LncRNAs in osteosarcoma ...... 29
vii
2. Genetic Engineering of Mesenchymal Stem Cells for Differential Matrix Deposition on 3D Woven Scaffolds ...... 38
2.1. Introduction ...... 38
2.2. Materials and Methods ...... 40
2.2.1. MSC culture and differentiation ...... 40
2.2.2. PCL scaffold processing ...... 42
2.2.3. SMAD3 knockdown with shRNA ...... 42
2.2.4. RUNX2 overexpression ...... 43
2.2.5. Lentivirus production ...... 43
2.2.6. Lentivirus transduction of MSC ...... 44
2.2.7. qRT-PCR ...... 44
2.2.8. Alcian blue and alizarin red staining and quantification ...... 45
2.2.9. Biochemical analyses for DNA and GAG content of engineered scaffolds ...... 46
2.2.10. Histological processing of scaffold samples ...... 46
2.2.11. Alkaline phosphatase assay ...... 47
2.2.12. Statistical analysis ...... 47
2.3. Results ...... 48
2.3.1. SMAD3 knockdown or RUNX2 overexpression inhibits TGFβ3 induced cartilaginous matrix deposition ...... 48
2.3.2. SMAD3 knockdown and RUNX2 overexpression act to enhance mineral deposition in a chondrogenic environment ...... 49
2.3.3. Differential GAG deposition was achieved in the dual-scaffold culture system ...... 50
2.3.4. Differential type II collagen production was achieved in the dual- scaffold culture system ...... 51
viii
2.3.5. Differential mineral deposition was achieved in the dual-scaffold culture system ...... 52
2.4. Discussion ...... 53
2.5. Conclusions ...... 57
3. High-Depth Transcriptomic Profiling reveals the Temporal Gene Signature of Mesenchymal Stem Cells During Chondrogenesis ...... 68
3.1. Introduction ...... 68
3.2. Materials and methods ...... 70
3.2.1. MSC isolation and expansion ...... 70
3.2.2. Chondrogenic differentiation ...... 70
3.2.3. Flow cytometry ...... 71
3.2.4. Biochemical analyses for DNA and GAG content ...... 71
3.2.5. Histological analyses for GAG and collagen content ...... 72
3.2.6. RNA isolation, library preparation and sequencing ...... 73
3.2.7. Data processing for computational analysis ...... 73
3.2.8. Computational analysis ...... 74
3.2.9. Determination of k-mean value ...... 75
3.2.10. Webtool development ...... 75
3.2.11. Statistical analysis ...... 76
3.3. Results ...... 76
3.3.1. Surface marker distribution of mesenchymal stem cells ...... 76
3.3.2. MSCs synthesized a cartilaginous matrix under chondrogenic conditions ...... 77
3.3.3. Interactions of biological pathways during chondrogenesis ...... 78
ix
3.3.4. Modules of gene co-expression networks highly correlated with GAG accumulation ...... 81
3.3.5. Characteristics of highly correlated modules ...... 83
3.3.6. Identification of miRNAs and lncRNAs from chondrogenic functional module...... 85
3.3.7. Web-based encyclopedia of MSC chondrogenesis ...... 86
3.4. Discussion ...... 86
3.5. Conclusions ...... 93
4. The long non-coding RNA GRASLND enhances chondrogenesis via suppression of interferon type II ...... 105
4.1. Introduction ...... 105
4.2. Materials and Methods ...... 107
4.2.1. Cell culture ...... 107
4.2.2. Plasmid construction ...... 108
4.2.3. Lentivirus production ...... 109
4.2.4. Lentivirus transduction ...... 111
4.2.5. Chondrogenesis assay ...... 111
4.2.6. Osteogenesis and adipogenesis assays ...... 112
4.2.7. Biochemical assays ...... 112
4.2.8. Immunohistochemistry and histology...... 113
4.2.9. RNA fluorescence in situ hybridization (RNA FISH) ...... 113
4.2.10. RNA isolation and quantitative RT-PCR ...... 114
4.2.11. Luminescence assay ...... 114
4.2.12. Western blot ...... 115
4.2.13. Cytotoxicity assay ...... 115 x
4.2.14. RNA-Seq and data analysis ...... 115
4.2.15. Statistical analyses ...... 116
4.3. Results ...... 116
4.3.1. RNF144A-AS1 is a crucial and specific regulator of chondrogenesis ...... 116
4.3.2. Characterization of RNF144A-AS1 ...... 118
4.3.3. Enhanced chondrogenesis for cartilage tissue engineering with GRASLND ...... 120
4.3.4. GRASLND inhibits type II interferon signaling ...... 122
4.3.5. GRASLND enhanced the chondrogenesis of adipose-derived stem cells ...... 125
4.4. Discussion ...... 126
4.5. Conclusions ...... 130
5. Conclusions ...... 141
Appendices ...... 145
Supplementary Data for “Genetic Engineering of Mesenchymal Stem Cells for Differential Matrix Deposition on 3D Woven Scaffolds” ...... 145
Supplementary Data for “High-Depth Transcriptomic Profiling reveals the Temporal Gene Signature of Mesenchymal Stem Cells During Chondrogenesis” ...... 150
Supplementary Data for “The Long Non-Coding RNA GRASLND Enhances Chondrogenesis Via Suppression of Interferon Type II” ...... 155
References ...... 163
Biography ...... 193
xi
List of Tables
Table 1: Percentage of cell populations that were positive for specific markers. n=3 for each donor. Mean ± SD...... 102
Table 2: Differentially expressed genes between MSC-derived pellets and human cartilage samples ...... 103
Table 3: Differentially expressed genes through MSC chondrogenesis ...... 104
Table 4: Long non-coding RNA candidates shortlist ...... 140
xii
List of Figures Figure 1: HOTAIR function in regulating HOXD gene expression ...... 33
Figure 2: lncRNA-HIT and Hottip function in regulating Hoxa gene expression . 35
Figure 3: Different mechanisms of H19 in osteogenesis ...... 36
Figure 4: MALAT-1 function in osteosarcoma ...... 37
Figure 5: Overview of experimental approach ...... 58
Figure 6: Reduction in GAG deposition with SMAD3 knockdown or RUNX2 overexpression ...... 60
Figure 7: Mineral deposition was achieved in the presence of TGFβ3 ...... 61
Figure 8: SMAD3 knockdown resulted in reduced GAG deposition ...... 62
Figure 9: SMAD3 knockdown resulted in reduced GAG deposition, quantified by biochemical assays ...... 63
Figure 10: Differential COLII expression...... 64
Figure 11: Synergistic effect of SMAD3 knockdown and RUNX2 overexpression for osteogenesis ...... 65
Figure 12: Modulation of TGF signaling for simultaneous cartilaginous and mineralized matrices in one culture well ...... 67
Figure 13: Distribution of MSC surface markers ...... 94
Figure 14: MSC chondrogenesis in pellet culture ...... 95
Figure 15: Differential expression (DE) analysis ...... 97
Figure 16: Modules and their relationships to biochemical changes ...... 99
Figure 17: Characteristics of module steel blue ...... 100
Figure 18: Characteristics of module magenta ...... 101
Figure 19: RNF144A-AS1 is important and specific to chondrogenesis in MSCs...... 131
Figure 20: RNF144A-AS1 expression is increased with chondrogenesis...... 133
xiii
Figure 21: RNF144-AS1 (GRASLND) relationship to RNF144A and SOX9. .... 134
Figure 22: GRASLND enhances chondrogenesis...... 136
Figure 23: GRASLND acts to suppress interferon type II signaling...... 138
Figure 24: GRASLND enhances chondrogenesis in adipose-derived stem cells...... 139
xiv
List of Abbreviations
ACAN or Acan Aggrecan
ALPL or Alpl Alkaline phosphatase
ASC Adipose derived stem cell
BMP or Bmp Bone morphogenic protein
CDH2 or Cdh2 N-cadherin
CASP3 or Casp3 Caspase 3
COL10A1 or Col10a1 Collage type X alpha chain 1
COL2A1 or Col2a1 Collagen type II alpha chain 1
CRISPR/Cas9 Clustered regularly interspaced short palindromic repeat/ CRISPR-associated protein 9
CRISPR/dCas9 Clustered regularly interspaced short palindromic repeat/ catalytically dead CRISPR-associated protein 9
DMMB assay 1,9-dimethylmethylene blue assay
ECM Extracellular matrix
ESC Embryonic stem cell
FGF or Fgf Fibroblast growth factor
GRASLND Glycosaminoglycan regulatory associated long noncoding RNA
HOTAIR or Hotair Hox transcript antisense RNA
IHH or Ihh Indian hedgehog
KO Knock-out lincRNA Large intergenic non-coding RNA lncRNA Long non-coding RNA
xv miRNA Micro RNA
MMP13 or Mmp13 Matrix metallopeptidase 13
MS Mass spectrometry
MSC Mesenchymal stem cell
NCAM or Ncam Neural cell adhesion molecule
OA Osteoarthritis
ORF Open reading frame piRNA Piwi-interacting RNA
PRC Polycomb complex protein
PTHrP Parathyroid hormone-related protein
RNA-Seq RNA sequencing
RNF144A Ring finger protein 144A
RNF144A-AS1 RNF144A antisense RNA 1
RUNX2 or Runx2 Runt related transcription factor 2
SHH or Shh Sonic hedgehog shRNA Short hairpin RNA
SMAD3 or Smad3 Mothers against DPP homolog 3
SOX5 or Sox5 SRY-Box 5
SOX6 or Sox6 SRY-Box 6
SOX9 or Sox9 SRY-Box 9
TF Transcription factor
TP53 Tumor protein 53
TGF-β3 Transforming growth factor beta 3
xvi
VEGF or Vegf Vascular endothelial growth factor
WGCNA Weighted gene correlation network analysis
xvii
Acknowledgements My first pipetting act happened in the laboratory of Dr. Steven A. Williams at
Smith College. To me, he is not only a professor, but a mentor with big heart. In the lab, I learned for the first time how to use Qiagen kits, how to run gels, and how to make my LB plates. When I met Steve, I was a junior in college with no laboratory experience. If he had not believed in me and offered me a position, I would not have probably gained the skills and blossomed. In life, he had guided me through difficult and important decisions. I remember how he would host
Thanksgiving dinners at his place and invite over the students who did not go home. My undergraduate years could not have been more fulfilled. He kept my mind hungry, and my heart warm. And for this, I owe my many thanks and gratitude to him.
My journey at Smith ended as my journey at Duke began. I would like to thank Dr. Brigid Hogan for the opportunity to work in her lab, and for all the inspiring conversations. She had taught me many wonders in life. She passed on this poem to me, which I have kept near and dear to my heart.
“I keep six honest serving-men
(They taught me all I knew);
Their names are What and Why and When
And How and Where and Who.” - Rudyard Kipling
I then applied to the Developmental and Stem Cell Biology training program at Duke, which brought me to my most memorable on-campus visit. Without the
xviii admission committee and their decision, I would not have been where I am today. Thank you Dr. David McClay, Dr. David Sherwood, Dr. Blanche Capel, Dr.
Deborah Silver, and Dr. Donald Fox. You gave me the ticket to make a difference.
Through different rotations, I joined the lab of Dr. Farshid Guilak, and unknowingly began the most exciting and eventful six years of my life. My dissertation would not have taken shapes without Dr. Jonathan Brunger – a mentor and a brother I am forever in debt with. I would also like to thank my undergraduate students, whose aids helped me tremendously in achieving my goals. Thank you Catherine Gloss and Jeremiah Lorentz for all the missed lunches and the late nights. Thank you for always having my back.
The Guilak lab is one of the most nurturing environments I have ever been immersed in. We bonded. We laughed. We cried. We did it all together. Thank you for making my youth an eventful one. To the “outpost” back at Duke: Dr.
Vincent Willards, Dr. Johannah Sanchez-Adams, Dr. Shaunak Adkar, Dr. Chris
Gilchrist, Dr. Brian Diekman, Dr. Holly Leddy. To the Washington University crew: Dr. Natalie Kelly, Dr. Chris Rowland, Dr. Chia-Lung Wu, Dr. Robert Nims,
Dr. Natalia Harasymowicz, Dr. Ruhang Tang, Nancy Steward, Nick Thompson,
Jim Maus, and all lab members. I could not have done it without you.
I would also like to thank my dissertation committee members – Dr. Blanche
Capel, Dr. Charles Gersbach, Dr. Michel Bagnat, and Dr. Eda Yildirim. Thank
xix you for spending your precious time giving me the insights and the support much needed.
I am grateful for the financial support through the National Institutes of Health, the Nancy Taylor Foundation for Chronic Diseases, and the Arthritis Foundation.
The work presented in this dissertation was made possible through them.
Through the years of hardship, the people who matter come into focus. Mom,
Dad. You are far away but I love you always. Ba, mẹ. Dù có xa cách nhưng con lúc nào cũng thương ba mẹ. I also owe my happiness in life to Minh and Ely.
Thank you for being the pillars, and the shoulders to lean on. Your loves transpire through space and time. And forever they will.
I saved the best for last. I owe all of my success to Farsh. I could not have imagined a better relationship forged throughout this journey. When I started, the questions I asked were not part of any research projects. Instead of shooting me down, Farsh had given me all his support. He provided me with expertise, constructive criticism, aids, and most importantly, the perseverance to come through. He saw it till the end. GRASLND was a story of RE-search. I had major failures before I found GRASLND. If it had not been for Farsh, I would have stopped and missed out on this awesome gene. To me, Farsh is both an advisor and a father figure, who I will always look up to and seek advice from. He is one heck of a boss: he is the “World’s Best Boss”.
xx
1. Introduction This chapter is an adaptation of: Huynh, N.P.T., et al. (2016). “Emerging roles for long noncoding RNAs in skeletal biology and disease.” Connective Tissue
Research. 58(1): 116-141.
1.1. Cartilage and Bone Development
Cartilage is a connective tissue that can be classified into three groups: hyaline (e.g., joint articular cartilage), elastic (e.g., ear) and fibrocartilage (e.g., meniscus and intervertebral discs in the spine). These cartilage types differ in extracellular matrix (ECM) composition and cell (chondrocyte) phenotype, thereby providing each tissue type with specific mechanical properties to carry out their function in the body.
Hyaline cartilage development is a complex process that begins with the condensation of mesodermal cells that are derived from the mesoderm germ layer (68). In response to various signaling pathways, chondro-progenitor cells are formed. These cells proliferate and differentiate into chondrocytes that are responsible for generating the extracellular matrix (ECM) which, in hyaline cartilage, consists primarily of collagen type II and the large aggregating proteoglycan, aggrecan. Two types of hyaline cartilage exist in the limb: articular cartilage and growth plate (epiphyseal) cartilage. Some controversy remains as to whether articular and growth plate cartilages are derived from common precursor cells or if the outer region of articular cartilage (i.e., the superficial zone) is derived from distinct precursor cells during synovial joint development
1
(3, 4). In any case, we know that chondrocytes of articular cartilage remain as permanent, differentiated cells that synthesize an ECM with appropriate mechanical properties to permit transmission of loads within the synovial joint during development and aging. Chondrocytes of the growth plate, however, are distinctly localized in columns where they continue to proliferate, thereby lengthening the limb during development. These cells then terminally differentiate to form large hypertrophic chondrocytes. The production of hypertrophic chondrocytes is essential to permit long bone formation via the process of endochondral ossification, whereby the cartilage template is essentially replaced by bone (5). It was thought that the majority of hypertrophic chondrocytes become apoptotic during endochondral ossification, but recent studies have shown that these cells also contribute toward the osteoblast pool or that there are osteoprogenitor cells in hypertrophic cartilage that can differentiate toward bone- producing osteoblasts (6-9).
Unlike cartilage, bone is a rigid tissue made up mainly of collagen fibers, calcium carbonate, and hydroxyapatite. Three cell types reside within the bone tissue: osteoblasts that synthesize bone extracellular matrix, osteoclasts that resorb bone, and osteocytes that are believed to be the mechano-sensing cells embedded deep within the bone (10-12). During embryonic development, bone is formed from cells of three sources: the somites, the lateral plate mesoderm, and the cranial neural crest. Formation of bone can occur via two processes: endochondral ossification or intramembraneous ossification.
2
The vertebral, limb, and rib skeletons are formed by endochondral ossification that begins with formation of a cartilage intermediate template (5). The chondrocyte-like cells later enlarge and alter their gene expression, leading to generation of an extracellular matrix that can be mineralized and vascularized
(13-15). Following vascularization of hypertrophic cartilage, pre-osteoblasts and pre-osteoclasts infiltrate this primary center of ossification. Mature osteoblasts then function to generate new bone matrix and, at the same time, a bone marrow cavity is created. Mature osteoclasts function to resorb the trabecular bone within the ossification center. Overall, there is a critical balance between osteoblasts and osteoclasts to maintain bone homeostasis.
In contrast, intramembraneous ossification occurs when the stem cells in mesenchymal tissue differentiate directly toward the osteoblast lineage without the requirement of a cartilage template. This can be seen perhaps most clearly by the construction of our facial skeleton. These cells subsequently mature into osteoblasts that can produce an osteoid matrix rich in collagen, proteoglycan and calcium (16).
1.2. Osteoarthritis and Cartilage Regeneration
Osteoarthritis (OA) is a painful and debilitating disease of the joints that is characterized by progressive degenerative changes in the articular cartilage and other joint tissues such as the synovium, and subchondral bone. Under normal circumstances, mature articular cartilage is maintained in a healthy balance of anabolic and catabolic processes by the activity of the chondrocytes. While many 3 factors can contribute to the onset and progression of OA, disruption of cartilage homeostasis can occur if a joint is subjected to altered loading caused by mal- alignment or trauma (e.g. meniscal or ligament injury, articular cartilage fracture)
(17-22). Obesity can also affect joint homeostasis due to increased loads as well as chronic inflammation (19, 23-25). As a result, catabolism often predominates, causing joint abnormalities including articular cartilage degradation, subchondral bone sclerosis, osteophyte formation and synovitis which, collectively, can be defined as OA (26). OA is a challenging disease to treat given that articular cartilage has little or no intrinsic regenerative capacity. Many of the cellular and molecular changes that occur in OA are known, such as alterations in growth factor and cytokine signaling, inflammation, oxidative stress, chemokine signaling and metabolism (26-28). However, no effective therapies or disease-modifying drugs have yet been discovered to ameliorate or stop OA progression.
Articular cartilage is the connective tissue that lines the ends of long bones in diarthrodial joints, providing a low-friction, load-bearing surface to help distribute loads between opposing bones. While cartilage provides protection against load- bearing and impact upon motion between the articular surfaces, it is an aneural, avascular tissue that has little intrinsic capacity for repair (29). Thus, focal chondral or osteochondral injuries result in significant pain and disability, and may lead to OA (28). The definitive treatment for end-stage OA is total joint arthroplasty that seeks to replace the diarthrodial joint with a metal and plastic prosthesis (30-34). However, artificial joints tend to wear out in 10 to 20 years,
4 with a steady rise in the total volume of revision procedures in the United States
(35, 36). The most common reason for revision arthroplasty is mechanical loosening: a failure of the bond between the implant and the bone (35). Despite significant progress toward developing cartilage substitutes in recent years, there remains a need for regenerative medicine approaches that can enhance the repair of large subchondral cartilage defects using biomimetic implants with improved fixation properties and similar mechanical and biochemical properties to those of native articular cartilage. The development of an osteochondral construct is attractive in this aspect since the osseous and mineralized phase could provide an anchoring layer to affix the cartilaginous phase within the joint
(37).
1.3. The human genome project and the discovery of long non- coding RNAs
The human genome project commenced in 1990, and its first complete assembly was announced in 2003. From this endeavor, it was found that while
70-90% of the human genome is actively transcribed, there is only 1-2% of the genome that contains protein-coding information (38, 39). The remaining non- coding RNA (ncRNA) transcripts were found to lack conserved open reading frames (ORF) and were predicted to have no biological functions in the cell. As such, these non-coding RNAs were usually initially referred to as “junk RNA”.
5
However, the past two decades of research on ncRNAs has overwhelmingly shown that these RNAs are in fact not “junk”, but rather play crucial and multifunctional roles in the regulation of cell phenotype. For example, classes of short ncRNAs [e.g.microRNAs (miRNAs) and Piwi-interacting RNAs (piRNAs)] have been shown to function as gene silencing agents (40-50). More recently, much interest has focused on another class of ncRNAs, known as long non- coding RNAs (lncRNAs). LncRNAs are defined as any ncRNA containing 200 or more nucleotides. The majority of lncRNAs are transcribed by RNA polymerase II
(RNA Pol II), and they are often spliced (and even alternatively spliced) during
RNA processing. From the most recent release of GENCODE, it is predicted that there are currently 15,779 human lncRNAs and 12,726 mouse lncRNAs
(Gencode Human version 28 and Gencode Mouse version M18). In this work, we will utilize bioinformatics and molecular tools to discover novel lncRNAs and characterize their roles in skeletal development and maintenance.
1.4. The study and identification of functional lncRNAs
The field of lncRNA biology is relatively young by scientific standards, and significant investments are still being made to identify new lncRNAs across different tissues and among species. Many techniques, which have been utilized to study well-characterized lncRNAs for some time, will continue to be used to interrogate the functions of these newly identified lncRNAs. However, advances in other areas such as next-generation sequencing have improved our ability to characterize lncRNAs and have drastically improved assay throughput. Various 6 approaches being utilized to determine lncRNA localization and function will be discussed in this section.
1.4.1. Identification of novel lncRNAs
While protein-coding genes can be predicted computationally from genomic sequences by virtue of their coding capacity and high levels of conservation, the identification of new lncRNAs cannot be performed in this manner, as non-coding transcripts lack these features. Notably, although comparisons of known lncRNAs demonstrated some degree of positive selection and conservation of secondary structure (51-53), predictive algorithms have a high false-positive rate for identification of new lncRNAs (54, 55). Furthermore, some well-characterized lncRNAs such as HOTAIR (HOX Transcript Antisense RNA) show highly divergent sequences and gene structures across species and are missed by these predictions, yet the function appears to be conserved (56).
As a result of this and significant advances in sequencing-based technologies and analysis pipelines, RNA sequencing (RNA-Seq) experiments remain the gold standard for identification of lncRNAs and have been implemented to identify novel lncRNAs across a variety of tissues and species. While sufficient depth of sequencing (100-200 million reads) is a critical parameter for high-quality studies seeking to exhaustively identify novel low-abundance lncRNAs in a given tissue
(57), lower-coverage experiments such as those utilized by Iyer and colleagues
(58) can be pooled together to facilitate preliminary identification of novel transcripts. Identification of novel lncRNAs from RNA-Seq datasets typically rely
7 on analytical pipelines such as TopHat, which aligns reads that may contain splice junctions to the genome, and Cufflinks, which assembles transcripts de novo (59, 60).
In addition to these first-pass strategies for identification of novel lncRNAs, additional methods are typically used to define the ends of the transcript and to validate the presence of additional splice isoforms. In many cases, these methods have been updated such that high-throughput sequencing-based methods may be used. While isoforms for a single gene can be identified from cDNA by PCR, detailed interrogation of many loci of interest can be accomplished utilizing protocols such as Capture-Seq, which enrich for sequences of interest using hybridization-based methods and significantly reduce the required depth of sequencing for identification of novel transcripts (61). In addition, identification of the 5’ end of the transcript, which is often underrepresented in sequencing libraries, using rapid amplification of cDNA ends
(RACE) can be parallelized using methods like Deep-RACE (62).
1.4.2. Identifying clues to lncRNA function
Perhaps the most telling indicator of lncRNA function is its subcellular localization, as this provides insight into its potential binding partners and signals its capacity for re-wiring gene regulation at the transcriptional or post- transcriptional levels. A prominent method for lncRNA localization is single molecule RNA in situ hybridization. When combined with fluorescence microscopy, these techniques can provide information relating to not only
8 whether the lncRNA is nuclear or cytoplasmic, but whether it is present near putative mRNA or genomic targets within the cell (63). Furthermore, this methodology can also allow one to estimate the range of expression among single cells, and as a result, to more rigorously quantify the lncRNA of interest
(64). Other methods relying on cellular fractionation can yield additional complementary information regarding lncRNA function. While the functions of ribosome-associated lncRNAs are not yet clear, ribosome profiling and ribosome fractionation experiments identified a number of ribosome-associated lncRNAs
(65), and nuclear fractionation studies revealed a large number of lncRNAs associated to actively transcribed chromatin (66).
Identification of the DNA, RNA, and/or proteins that interact with a given lncRNA generates a second tier of information regarding its function (67).
Unbiased approaches to identify these partners often rely on hybridization-based methods to isolate the lncRNA and its binding partners from the cell.
Subsequently, next-generation sequencing can be used to identify interactions between other RNA molecules using RNA antisense purification (RAP-Seq (68)), chromatin domains using chromatin isolation by RNA purification (ChIRP (69)), capture hybridization analysis of RNA targets (CHART (70)), or proteins using
RNA pull down technique coupled to mass spectrometry (RAP-MS (71)). Notably, more direct interrogation of the functions of nuclear lncRNAs in regulation of individual loci is now possible with the development of the CRISPR-Display technique (CRISPR: Clustered Regularly Interspaced Short Palindromic Repeat)
9
(72). This technique uses locus-specific guide RNAs to target a Cas9-lncRNA
(Cas9: CRISPR associated protein 9) complex to the desired genomic location, which not only facilitates the detailed characterization of the lncRNA on the locus of interest, but also allows the identification of specific features within the lncRNA that are required for its function.
A third powerful predictor of lncRNA function is the secondary and tertiary structure of the transcript. For example, stem-loop structures are frequently present in noncoding RNAs that interact with the Polycomb complex protein
PRC2, a key epigenetic regulator (73). As mentioned previously, a variety of methods are available to predict RNA secondary structure (52, 53, 74). However, more rigorous experimental analysis of RNA secondary and tertiary structure is frequently performed using chemical and enzymatic probes. Most notable is the
Selective 2’-hydroxyl acylation analyzed by primer extension (SHAPE) method
(75).
Finally, while RNA-Seq is a powerful technique for identifying new lncRNAs, it also proves useful in predicting potential functions of lncRNAs in vivo through the identification of co-regulated gene networks. By identifying mRNAs that are regulated similarly to the lncRNAs of interest in response to developmental or environmental stimuli, a network of genes affecting specific pathways can be developed and tested (76).
10
1.4.3. Validation of lncRNA function through reverse genetics
Probing the functions of lncRNAs using reverse genetics in vivo is complex, but a number of techniques can be used easily in vitro. Antisense knockdown of lncRNAs has become commonplace, and techniques exist for robust knockdown of both nuclear and cytoplasmic lncRNAs (77, 78). However, strategies relying on
RNA interference may be limited by delivery of the siRNA or shRNA, low knockdown efficiency, and off-target effects, and the importance of including well- designed experimental controls can complicate the design of in vivo studies.
The utility of knockout (KO) animal models for protein-coding genes has been essential for thorough investigations of their functions, but the design of similar models for lncRNAs is much more challenging. Exon-deletion strategies and the use of Clustered regularly interspaced short palindromic repeats (CRISPR) –
CRISPR associated protein (Cas) technologies to introduce frameshifts in protein-coding genes have been very successful (79), but these are likely much less effective for lncRNAs, as there is no open reading frame to be altered and a lncRNA may retain its function even if significant regions are deleted (80-83).
Furthermore, as the entire genomic region encoding the lncRNA needs to be removed in order to generate a true null mutant, it is especially important to ensure that regulatory regions that may be encoded within the locus are not perturbed.
Despite these challenges, a number of KO animal models have been developed for lncRNAs and have been reviewed elsewhere (84). In addition, a
11 number of lncRNAs have been modeled successfully by a variety of genetic manipulations including gene disruptions (Xist; (85)), targeted promoter deletions
(Kcnq1ot1 (86)), and premature termination strategies (Kcnq1ot1 (87)). Thus, as with other genetic models in cell culture and in vivo, recent advances in CRISPR-
Cas technologies, especially those improving efficiency of targeted knock-in strategies, will likely improve the speed at which lncRNA targeting studies can be accomplished (88).
1.5. Long non-coding RNAs in the regulation of stem cell maintenance or differentiation
LncRNAs have rigorously been shown to play a crucial role in maintaining pluripotency (89, 90). One study (89) performed a systematic loss of function experiment in which 147 lincRNAs were knocked down in mouse embryonic stem cells (ESCs). While the differentiation capacity of these cells after lincRNA (large intervening non-coding RNA) knockdown was not assessed, they observed that
93% of the lincRNAs targeted likely played an important role in maintenance of pluripotency, as their depletion resulted in gene expression profiles that were highly similar to those obtained when key pluripotency markers such as Nanog or
Oct4 were depleted.
A subsequent study demonstrated that lncRNAs are not only important in maintaining pluripotency, but may also act as key regulators in promoting and preserving cell identity after differentiation (90). Using a microarray chip to
12 quantify expression of 6,671 lncRNA transcripts, Ng et al. identified 934 differentially expressed ncRNAs as human embryonic stem cells were differentiated into neuron progenitors in vitro. Their functional studies focused on two classes of lncRNAs: those that were much more abundant in ESCs than in neural progenitors (NPCs) and likely played a role in maintenance of pluripotency, and those that were much more abundant in NPCs than in ESCs and likely played a role in neuronal differentiation. As expected, knockdown of several lncRNAs more abundant in ESCs led to spontaneous differentiation into a variety of cell types while knockdown of several neurogenic lncRNAs resulted in inhibition of neuronal differentiation. Functional characterization of some of these neuronal lncRNAs (e.g. RMST, LINC01109, and CACNA2D1) showed that they were localized to the nucleus and interacted with either the Polycomb repressive complex PRC2 or the neuronal transcription factor REST. In contrast, the neurogenic lncRNA MIR100HG was localized to the cytoplasm. This lncRNA also serves as a precursor to the miRNAs miR-125b and let-7 within its introns, which are known to function in neural development (91, 92). Knockdown of
MIR100HG substantially reduced expression of these miRNAs, suggesting this lncRNA may also function as a reservoir for miRNAs important for differentiation.
Although the majority of lncRNA studies in stem cell differentiation have been carried out in vitro, comprehensive libraries of lncRNA expression during different stages of murine neurogenesis in vivo have also been described (93). In the study from Ramos et al., this library was developed by integrating multiple RNA-
13
Sequencing approaches (RNA-Seq and RNA Capture-Seq (61)) with assays of chromatin state using chromatin immunoprecipitation and sequencing (ChIP-
Seq), which provides not only a comprehensive list of expressed transcripts and splice isoforms, but also some indication of their regulation in vivo. Interestingly, comparisons of murine ESCs and micro-dissected neural stem cells (NSCs) demonstrated that many lncRNAs exhibit substantial changes in chromatin state during differentiation that correspond with transitioning from a “poised” state to an active state. For example, lnc-Pou3f2, which is encoded upstream of the well- established neurogenic transcription factor Pou3f2, is bivalently marked with both activating (H3K4me3) and repressive (H3K27me3) histone modifications in
ESCs, but transitions to a fully active state marked by only H3K4me3 in NSCs.
However, a substantial number of lncRNAs remain bivalently marked in NSCs, suggesting that these lncRNAs may function later during neural lineage specification. Functional validation of several lncRNAs confirmed their important roles in NSC self-renewal or differentiation, demonstrating that well-constructed lncRNA expression libraries are extremely valuable resources for investigating the roles of lncRNAs in development and disease in vivo across various cell types and tissues.
Other studies have demonstrated the function of lncRNAs in the definition of other cell/tissue fates, including cardiac development. The lncRNA Fendrr was identified by differential expression in posterior mesoderm compared to other early somite-stage mouse tissues, and in vivo studies in which the transcript was
14 prematurely terminated clearly demonstrated that loss of Fendrr was embryonic lethal (94). Defects in both cardiac development and in development of the ventral body wall suggested defects in differentiation of the lateral mesoderm, which was verified using gene expression analysis and chromatin profiling.
Similarly, Klattenhoff and colleagues again utilized RNA-Seq to identify lncRNAs that were most abundant in the murine heart compared to other tissues and identified the candidate Braveheart (Bvht), a lncRNA which has no obvious homologs in human or rat tissues (95). In vitro studies of Bvht in ESC differentiation showed that this lncRNA is not required for ESC maintenance or global differentiation but plays an important role in cardiac differentiation via interactions with the Polycomb repressive complex, PRC2. Notably, these studies were possible due to the natural propensity for ESC-derived embryoid bodies to generate cardiac tissue and well-established cardiomyocyte differentiation assays, highlighting the importance of developing well-defined in vitro differentiation assays for other cell types of interest.
1.6. Long non-coding RNAs in the regulation of chondrocytic and osteoblastic differentiation
1.6.1. LncRNAs regulating chondrocyte differentiation
A common method to study chondrogenesis in vitro involves three dimensional culture of human mesenchymal stem/stromal cells isolated from bone marrow (96, 97) or other sources such as adipose, synovium and umbilical cord (98, 99). A recent study utilized human bone marrow-derived MSCs (MSCs) 15 to determine lncRNA expression signatures in cells two weeks following in vitro chondrogenesis compared to day 0 (control) of the differentiation assay (100).
Following microarray analysis (lncRNA + mRNA Human Gene Expression
Microarray V4.0; CapitalBio Corp, Beijing, China), over three thousand lncRNAs were found to be significantly differentially expressed (|Fold change| > 2) between control and chondro-induced cells. Among these, three lncRNAs were identified as upregulated following chondrogenesis (ZBED3-AS1, CTA-941F9.9,
ENST00000433576.1) and one was identified as downregulated (LINC00707).
Moreover, it is not known if these lncRNAs showed the most dramatic fold changes in this work since the entire array data was not reported. Another caveat with this study involves the relatively small increase in type II collagen expression
(~ 2 fold increase) after 2 weeks of chondrogenesis. Normally, COL2A1 levels are increased by several orders of magnitude during such MSC chondrogenesis assays (101, 102). Moreover, no COL2A1 expression data was shown from their array study. It will be interesting to determine lncRNA expression by RNA-Seq analysis in MSC-induced cells that produce higher levels of COL2A1 during chondrogenesis and investigate if the same lncRNAs found in this study by Wang et al., are also found to be differentially-expressed.
Another recent study identified that the transcription factor, SOX4, can regulate the proliferation and chondrogenic differentiation of human synovium- derived MSCs (SMSCs) and that activation of the lncRNA DANCR is involved in this process (103). Here, SMSCs were extracted from the synovium of OA
16 patients and aggregate cultures were utilized for in vitro chondrogenesis.
Apparently, the differentiation medium in these assays did not include transforming growth factors TGF-β1/3, and so it is surprising that efficient chondrogenic differentiation was achieved. In any case, SOX4 over-expression was found to increase expression of chondrocyte markers over two weeks in culture. To attempt to decipher SOX4 function in this scenario, promoter analysis of the lncRNA DANCR revealed a SOX4 binding site. Luciferase and ChIP assays showed that SOX4 can interact with its binding site on the DANCR promoter. Inhibition of DANCR by RNAi apparently suppressed the effects of
SOX4 in promoting chondrogenesis, thereby suggesting that SOX4 functions upstream of DANCR. DANCR (Differentiation Antagonizing Non-protein Coding
RNA) was first described as a lncRNA (previously named ANCR), that functions in suppression of progenitor cell differentiation (104). Two recent studies have shown that DANCR can increase stem cell-like features in hepatocellular carcinoma cells (105) and that it can suppress odontoblast-like differentiation of human dental pulp cells by inhibiting the Wnt/β-catenin pathway (106). Given this
“anti-differentiation” function, it is interesting that DANCR activation via SOX4 in
SMSCs resulted in increased proliferation and differentiation toward the chondrocyte lineage in this study. Further research is required to thoroughly decipher the role of DANCR in chondrogenesis.
With respect to human skeletal pathology, a lncRNA was recently found to be associated with brachydactyly type E (BDE), a condition characterized by
17 shortening of metacarpals and metatarsals. In this study (107), two families with autosomal dominant BDE were identified with translocations in chromosome 12 that resulted in genomic disruption of an important cis-regulatory element named
CISTR-ACT, which also produces the lncRNA DA125942. Notably, the CISTR-
ACT locus interacted in cis with the PTHLH locus, which encodes the chondrogenic regulator parathyroid hormone-like hormone (108), and in trans with the SOX9 locus, which encodes the master chondrogenic transcription factor
SRY-Box 9 (109). Importantly, knockdown of DA125942 in a chondrogenic cell line resulted in downregulation of PTHLH and SOX9, as well as a plethora of other genes important in skeletal patterning (HOXB cluster, ETV4) or chondrogenesis (HIF1A), suggesting a potentially important role for this lncRNA in regulating chondrogenic differentiation.
Studies on lncRNAs associated with HOX genes have provided insights into their functions in regulating skeletal development. HOX genes are a conserved family of developmental transcription factors that elicit specific developmental programs along the head-to-tail axis of animals (110). For example, lncRNA genes in the 5’ HOXA and HOXD clusters are known to regulate limb and spine growth and patterning (111-114). HOTAIR (HOX Transcript Antisense RNA) was the first vertebrate lncRNA described to regulate HOX function (115).
Specifically, HOTAIR is expressed from the HOXC locus and functions to recruit
EZH2 and SUZ12, part of the Polycomb repressive complex PRC2, to the HOXD cluster, thereby establishing a silent chromatin state using H3K27 methylation
18 and resulting in repression of several 5’ HOXD genes (Figure 1). Targeted deletion of Hotair in mice resulted in patterning malformations in the lumbosacral junction and in the metacarpal and carpal bones in the limbs (56). Interestingly, cyclic stretch has been shown to decrease HOTAIR levels in human aortic interstitial cells (AVICs), and reducing HOTAIR levels via siRNA in AVICs results in increased expression of calcification genes (116). These findings suggest that
HOTAIR is mechanoresponsive and may thus may play a potential role in mechanically-regulated calcification, although additional work is needed to test this hypothesis in other cells types.
In contrast, HOTTIP (HOXA Transcript at the Distal Tip) is an enhancer lncRNA that regulates expression of 5’ HOXA genes to control growth and elongation of skeletal elements of the limb (117). HOTTIP functions by regulating chromosome looping to recruit the WDR5/MLL histone methyltransferase complex to the 5’ HOXA genes, conferring an active chromatin state via H3K4 methylation (117, 118) (Figure 2B). While HOTAIR has been shown to act in trans (located on the HOXC locus and regulates HOXD genes) (Figure 1),
HOTTIP was noted to act in cis (Figure 2B) and its proximity was crucial for
HOXA gene expression, as expression of HOXA genes was not changed when
HOTTIP was ectopically expressed via lentivirus.
An elegant study published recently by Stadler’s group has identified another lncRNA located within the HOXA locus, named lncRNA-HIT, that appears to function as a critical epigenetic regulator of chondrogenesis (119). LncRNA-HIT
19
(HOXA Transcript Induced by TGF-β) was initially characterized as a TGF-β- responsive lncRNA during epithelial-to-mesenchymal transition in mammary epithelia (120). LncRNA-HIT was mapped as a single exon in the mouse genome between Hoxa11 and Hoxa13 (121). Carlson et al. found that lncRNA-HIT was expressed in the developing mouse limb and hypothesized that it may play a role in chondrogenesis. RNA FISH analysis showed that lncRNA-HIT is localized to the nucleus of limb mesenchymal cells, while mass spectroscopy and immunoprecipitation experiments revealed that it associates with p100/CBP complexes (Figure 2A). Knock-down of lncRNA-HIT in micromass cultures of murine limb mesenchymal cells was found to inhibit chondrogenesis (i.e. cartilage nodule formation) in these cultures. Further experiments in this study strongly suggests that, mechanistically, lncRNA-HIT functions as an enhancer lncRNA via association with p100/CBP complexes to maintain H3K27ac at specific chromatin sites thereby resulting in the activation of 5’ HoxA genes. In addition to affecting expression of HoxA genes, lncRNA-HIT siRNA experiments also revealed decreased expression of a number of other genes, including
Bmpr1b. This finding may also explain why knock-down of lncRNA-HIT reduces cartilage formation of murine limb mesenchyme in vitro, since loss of Bmpr1b is known to negatively affect chondrogenesis in mice and humans (122-125).
Overall, this study is the first to provide detailed mechanistic insights into the role of a specific lncRNA in chondrogenesis and reveals another level of complexity
20 toward our understanding of how chondro-progenitor commitment and cartilage tissue development is controlled.
1.6.2. LncRNAs regulating osteoblast differentiation
To date, studies on the role of lncRNAs in bone biology have been limited.
However, one report showed that homozygous mice lacking the lncRNA, Hotair, resulted in lumbosacral transformation and fusion of the metacarpal and carpal bones (56). This skeletal phenotype was similar to transgenic mice with ectopic expression of HoxD, resulting in elevation and anteriorization of Hoxd10 and
Hoxd11 expression. It has been suggested earlier that Hotair acts in trans and regulates the expression of HoxD genes by recruiting the Polycomb repressive complex PRC2 to the HoxD locus (115). RNA-Seq data of WT, Hotair +/-, and
Hotair -/- confirms the earlier hypothesis: Hotair deletion leads to demethylation of H3K27 and methylation of H3K4 at targeted genes, resulting in derepression of HoxD genes. Interestingly, several imprinted loci were also identified as Hotair targets, including Dlk1-Meg3 and Igf2-H19. Since Hotair and its well-known related gene Hottip regulate the Hox loci, and Hox genes are involved in body patterning and limb development, it was not surprising that mutation of this lncRNA led to such striking skeletal patterning defects. It will be interesting to investigate whether Hottip is also involved in the regulation of osteogenesis.
Recent studies have demonstrated that H19 is upregulated during osteogenesis (126, 127). Different mechanisms, which may not be exclusive of one another, have been proposed to describe H19 function in this context. Huang
21 and colleagues demonstrate that H19 can function to promote osteogenesis.
They showed that H19 is processed into miR-675, which targets TGFB1 for degradation (127) (Figure 3A). Since TGFβ signaling has been previously shown to impede osteogenesis by inducing heterochromatin formation at several promoters such as RUNX2 and OCN (128), a drop in TGF-β1 would lead to derepression of these osteogenic markers and subsequently promote osteogenesis. Additionally, this study also showed that H19/miR-675 downregulated levels of HDAC4/5, which also resulted in increased expression of osteoblast genes, although it was not clear whether this effect is direct or indirect.
Interestingly, a specific function for H19 outside of miR-675 biogenesis was not described by Huang et al. In contrast, the proposed mechanism described by
Liang and colleagues demonstrated that H19 itself functions as a competing endogenous RNA (ceRNA), binding to miR-141 and miR-22 and sequestering them away from their mRNA targets (Figure 3B). Notably, these miRNAs were previously established to target and degrade CTNNB1 (β-catenin) (129), and H19 may relieve their repressive effect on the Wnt-signaling pathway by competitively binding to these miRNAs, thereby inducing osteoblast differentiation. As H19 had previously been shown to be enriched during chondrogenesis (130), it may be the case that H19 suppresses the stemness of MSCs in general instead of promoting specific differentiation programs in chondrogenesis and/or osteogenesis, although further evidence on the molecular mechanism of H19 in
MSCs is required to prove this hypothesis.
22
There have been two reports of the lncRNA DANCR in osteogenesis. In one study, it was proposed that DANCR regulates expression of the osteogenic regulator RUNX2 by recruiting the Polycomb repressive complex component
EZH2 to the RUNX2 promoter, thereby resulting in suppression of osteogenesis
(131). While RNA pulldowns showed a direct association of DANCR and EZH2, there was no evidence that DANCR was directly binding or recruiting EZH2 to directly bind to RUNX2 promoter, which will be of interest for future investigation.
A similar study manipulated DANCR expression in human periodontal ligament stem cells, and also showed that downregulation of DANCR was critical for osteogenesis (132). Here, DANCR knockdown upregulated β-catenin/Wnt signaling pathway in these cells, which has been well-documented to promote osteoblastogenesis. However, this study only suggests a correlation between
DANCR and the osteogenic markers, and no direct interactions between the lncRNA and the genomic elements regulating osteogenic genes were investigated in the study. In addition, since DANCR was first proposed to function as an anti-differentiation player (104, 133, 134), it might be the case that reducing DANCR expression evokes a global effect on the epigenetic state of the cells. Instead of directly affecting their osteogenic capability, DANCR may be acting to make the cells more competent for differentiation into specific cell types given the appropriate medium conditions.
Additionally, a set of lncRNA microarray expression data was also generated from a mouse MSC cell line that was stimulated with BMP2 (135). The study
23 identified 116 lncRNAs that were differentially expressed: 59 upregulated and 57 downregulated. However, several issues were noted in this study: i) the cut-off values used for differential expression were arbitrary and changed between time points, ii) the experimental methods were not well-described, and iii) data showing appropriate osteogenic induction were relatively weak. Furthermore, no functional characterization was performed to test the requirements for these lncRNAs in osteogenesis. Therefore, additional validation of differentially- expressed lncRNAs in this system is still needed.
The expression of the lncRNA HIF1A-AS1 in human bone marrow-derived
MSCs was also characterized, and the authors imply a potential role for this lncRNA in osteogenesis which is dependent on the histone deacetylase SIRT1, an important positive regulator of osteoblastogenesis and bone mass (136, 137).
Here, overexpression of SIRT1 downregulated HIF1A-AS1 expression, while knockdown and pharmacological inhibition of SIRT1 upregulated HIF1A-AS1 expression. Interestingly, HIF1A-AS1 knockdown decreased expression of
HOXD10, suggesting a potential role for this lncRNA in HOX gene regulation.
However, the authors describe a pro-osteogenic role for TGF-β, which has been well documented to function in an anti-osteogenic capacity. In addition, the regulation of HIF1A-AS1 during in vitro osteogenic assays was not described, and it was not clear whether these experiments were performed under osteogenic conditions, so further work is needed to determine the function of this lncRNA during osteogenesis.
24
Lastly, the function of the lncRNA MEG3 was shown to play an important role in osteogenesis, especially in the context of multiple myeloma, a hematological cancer that also affects bone mass by inhibition of osteoblastogenesis of MSCs
(138). Here, it was demonstrated that MEG3 expression increases during osteogenic differentiation of MSCs from normal patients, but its expression is reduced overall and is unchanged during osteogenesis. In normal cells, MEG3 knockdown reduced osteogenic differentiation and functioned by reducing transcription of BMP4, and ectopic MEG3 overexpression or treatment with exogenous BMP4 rescued the osteogenic defect of MSCs from multiple myeloma patients. Mechanistically, it was shown that MEG3 regulates BMP4 transcription by disrupting the interaction between the SOX2 transcription factor and the
BMP4 promoter. Interestingly, overexpression of MEG3 in MSCs from multiple myeloma patients also improved chondrogenic differentiation, suggesting a potential role for this lncRNA in chondrocyte differentiation as well.
1.7. Long non-coding RNAs in cartilage- and bone-related diseases
1.7.1. LncRNAs in cartilage homeostasis and osteoarthritis
To date, a small number of studies have been published on lncRNA expression and potential function in mature chondrocytes within the context of
OA. The lncRNA H19 was first reported by Dudek et al. to be highly expressed in mature primary articular chondrocytes and to be regulated by SOX9 (130). This study also showed that H19 knock-down in primary chondrocytes reduced 25
COL2A1 levels; however, H19 also serves as the precursor for the microRNA miR-675, and they demonstrated that the effects of H19 knockdown could be rescued by over-expression of miR-675. Thus, they concluded that miR-675 is an important functional component of H19 to regulate the chondrocyte phenotype.
In addition, Steck et al. reported that H19 expression, as well as IGF2, was elevated in human OA cartilage compared to healthy, control tissue; however
H19 and IGF2 levels did not correlate significantly in cartilage (139). This study also showed that H19 levels are increased in chondrocytes under hypoxic conditions, while levels decreased in response to pro-inflammatory cytokines.
From these findings, it was speculated that H19 may play a role in regulating chondrocyte metabolism in response to stress as well as induce chondrocyte anabolism. As discussed earlier, H19 was recently shown to induce osteoblast differentiation via TGF-β1/Smad3/HDAC or β-catenin/Wnt signaling pathways
(126, 127). It remains to be discovered if H19 functions in a similar manner to induce chondrogenesis. Finally, a recent report revealed a new function for H19 in regulating DNMT3B-mediated DNA methylation (140). This is interesting from the standpoint that changes in DNA methylation patterns have been detected in human OA cartilage and that H19 levels have been reported to be higher in OA
(139, 141).
A study by Kim et al. showed that HOTTIP expression was increased
(assessed by microarray analysis) in chondrocytes from OA patients compared to non-OA or normal control cells (142). They also detected a decrease in
26
HOXA13 expression as a result of increased HOTTIP levels. However, it has been reported (and mentioned earlier in this review) that HOTTIP functions to facilitate H3K4 methylation and, therefore, activation of distal HOX genes. It is therefore curious that an increase in HOTTIP would lead to a reduced HOXA13 expression in this study. Further mechanistic studies are needed to clarify the functional role of HOTTIP in regulating cartilage homeostasis.
Other microarray-based studies have been published reporting lncRNA expression changes in OA chondrocytes compared to healthy control cells (141,
143, 144). One of these studies reported 121 differentially-expressed lncRNAs,
73 of which were significantly upregulated including HOTAIR, GAS5, PMS2L2,
RP11-445H22.4, H19 and CTD-2574D22.4 (141). Of these lncRNAs, HOTAIR and GAS5 were also found to be upregulated in OA chondrocytes in the Kim et al. study mentioned above (142). The study by Liu et al. (144) identified 152 differentially expressed lncRNAs and focused specifically on one lncRNA
(lncRNA-CIR; Cartilage Injury Related) which was upregulated over 10 fold in OA chondrocytes, as confirmed by qRT-PCR, and was also shown to increase in chondrocytes following treatment with IL-1β and TNF-α. Attempts to decipher the function of lncRNA-CIR were done by knock-down (siRNA) and over-expression approaches; it was concluded from these experiments that lncRNA-CIR can induce chondrocyte catabolic events via increased expression of MMP13 and
ADAMTS-5. The study by Fu et al. (143) provided expression data of differentially expressed lncRNAs followed by target predictions and pathway
27 analyses of those lncRNAs found to be most highly up- or down-regulated in OA.
They utilized the same microarray (Human lncRNA Array v2.0; 8x60K; Arraystar) that was used by Liu et al. (144) and reported many more differentially expressed lncRNAs (4714 lncRNAs) based on the fact that they included lncRNAs whose levels changed by 2 fold or more (as opposed to the Liu et al. study that reported lncRNAs with an 8-fold or more change in expression).
Another recent report also showed increased expression of GAS5 (Growth
Arrest-Specific 5) in OA cartilage compared to healthy control tissue (145). RNA
FISH analysis showed nuclear localization of GAS5 in healthy chondrocytes while localization in OA chondrocytes also revealed cytoplasmic distribution.
Over-expression of GAS5 in chondrocytes in vitro resulted in increased metalloproteinase expression, decreased expression of markers associated with autophagy and decreased expression of miR-21. Manipulation of miR-21 levels in murine articular cartilage in vivo via lentivirus showed some degree of chondro-protection following OA-induced surgery by destabilization of the medial meniscus (DMM) (146). The overall conclusion from these findings is that GAS5 over-expression appears to contribute to a catabolic phenotype via regulation of miR-21 in chondrocytes. Another study was published recently reporting decreased expression of the lncRNA MEG3 (Maternally Expressed Gene 3) in
OA and that its expression levels were inversely associated with VEGF levels
(147). However, no functional data was included in this study to be able to make any conclusions on MEG3 regulation of VEGF in chondrocytes.
28
Finally, a study by Pearson et al. aimed to identify lncRNAs associated with the inflammatory response in human primary hip OA chondrocytes (148). These cells were treated for 4h with IL-1β and differentially-expressed lncRNAs between treated and untreated cells were identified following RNA-Seq analysis
(GSE74220). Of the 983 lncRNAs identified, 125 were found to be differentially expressed following IL-1β treatment, including PACER (p5-Associated COX2-
Extragenic RNA) and two novel chondrocyte inflammation-associated lincRNAs
(CILinc01 and CILinc02). The increase in expression of these three lincRNAs was confirmed by additional in vitro assays using IL-1β-induced primary knee or hip chondrocytes followed by qPCR analysis. Interestingly, expression of these lncRNAs was found to be lower in OA cartilage compared to healthy control tissue. This finding may be due to the fact that the inflammatory status (and hence lncRNA expression) of chondrocytes embedded within late stage OA cartilage tissue may not correlate with that in cultured chondrocytes isolated from cartilage following treatment with IL-1β. In any case, this report has provided in- depth information on a number of inflammatory-regulated lncRNAs that may be worth pursuing in greater detail with respect to determining their biological functions in regulating chondrocyte homeostasis.
1.7.2. LncRNAs in osteosarcoma
Osteosarcoma is a type of bone cancer that occurs at a rate of 5% in pediatric malignancies (149, 150). In humans, osteoblastic osteosarcoma is characterized by highly mineralized tissues that are abundant in osteoids with
29 multinucleated cells (151, 152). From recent improvements in surgical, therapeutic, and radiation care, the survival rate of osteosarcoma patients has increased from less than 20% to 70% over the past forty years (153-157).
However, elucidation of the molecular basis of disease development and progression is required to establish better treatment options. Recently, several lncRNAs have been identified that show differential expression in osteosarcoma tissue.
One lncRNA that is dysregulated in osteosarcoma is MALAT-1 (Metastasis-
Associated Lung Adenocarcinoma Transcript 1), a 6.8kb transcript that is conserved among 33 mammalian species (158). Unlike most long noncoding
RNA, MALAT-1 is not poly-adenylated at its 3’ end but instead folds into a triple helix (Figure 4), a unique structure that protects the RNA from degradation by nucleases (159). MALAT-1 is expressed in many types of cancers, where it often promotes proliferation (reviewed in (160)). Therefore, it is not surprising that
MALAT-1 was also observed to be upregulated in osteosarcoma (161, 162).
These independent studies established that knockdown of MALAT-1 in osteosarcoma cells not only decreased proliferation but also induced apoptosis, and knockdown reduced osteosarcoma growth and metastasis in vivo. In addition, a third study utilizing the MG63 osteosarcoma cell line demonstrated that MALAT-1 was downregulated by high-dose estrogen, and this downregulation reduced proliferation, migration, and invasion (163). MALAT-1 appeared to sequester the splicing complex factor SFPQ from its splicing co-
30 regulator PTBP2 (Figure 4), and downregulation of MALAT-1 resulted in a higher accumulation of the SFPQ/PTBP2 complex. Interestingly, similar results have been observed in colon cancer cells, in which sequestration of SFPQ by MALAT-
1 results in PTBP2-mediated increases in cellular proliferation and migration
(164).
Many lncRNAs have been identified using similar methods, and in vitro studies show that they may play important roles in osteosarcoma. Much like
MALAT-1, the lncRNA SNHG12 (Small Nucleolar RNA Host Gene 12) is upregulated in osteosarcoma patients (165, 166). Both studies demonstrated that expression of SNHG12 correlated with expression of AMOT (angiomotin), a gene with unknown functions in osteosarcoma. While modulation of SNHG12 substantially altered expression of AMOT, the mechanism of regulation is not clear, and the function of this lncRNA and AMOT have not been clearly established (165). Similarly, the lncRNA HULC (Highly Upregulated in Liver
Cancer) is upregulated in osteosarcoma tissues (167), and knockdown of HULC in osteosarcoma cell lines in vitro suppressed proliferation, migration, and invasion. In contrast, the lncRNA TUSC7 (Tumor Suppressor Candidate 7), also known as LOC285194, is substantially downregulated in osteosarcoma, and low expression of this lncRNA correlates with poor survival in osteosarcoma patients
(168, 169). Interestingly, expression changes in this lncRNA seemed to correlate with copy number alterations rather than an epigenetic silencing of the genes. In contrast to MALAT-1, knockdown of TUSC7 or LOC285194 led to increased
31 proliferation and/or decreased apoptosis, and in vivo xenograft studies showed that TUSC7 silencing promoted tumor growth. Lastly, the promoter-upstream transcript HIF2PUT has been proposed to regulate expression of HIF2A in osteosarcoma (170). While the expression of HIF2PUT and HIF2A in osteosarcoma appeared to be relatively weakly correlated, knockdown of
HIF2PUT increased cell proliferation and invasion in vitro in an osteosarcoma cell line, suggesting that this class of lncRNAs may also function in osteosarcoma.
In addition, lncRNAs that function in chondrocyte or osteoblast differentiation and homeostasis have also been found to be dysregulated in osteosarcoma, including DANCR, PACER, HOTAIR, HOTTIP, and H19. Much like MALAT-1, knockdown of DANCR in osteosarcoma cell lines reduced cell proliferation and resulted in cell cycle arrest (171). Similarly, the lncRNAs PACER, HOTAIR, and
HOTTIP were shown to be upregulated in osteosarcoma, and knockdown of these lncRNAs inhibited proliferation and/or invasion of osteosarcoma cells (172-
174). Mechanistic investigation of PACER function demonstrated that it upregulates the expression of the oncogene COX2 in an NF-κB-dependent manner. A role for H19 in osteosarcoma has been proposed by Chan and colleagues, who developed a model for spontaneous development of osteoblastic osteosarcoma in which p53-heterozygous mice containing an osteoblast-specific inactivation of Patch1 results in partial upregulation of
Hedgehog signaling (151). In this model, the Hippo pathway transcriptional coactivator yes-associated protein 1 (Yap1) was upregulated by Hedgehog
32 signaling, and in turn upregulates H19. Interestingly, loss of imprinting at the
H19-IGF2 has previously been observed in osteosarcoma (175), suggesting a potential function for H19 and genomic imprinting in this setting.
Finally, other clinical investigations have revealed correlations between lncRNA expression levels and osteosarcoma patient survival. Decreased expression of MEG3 (176) and upregulation of TUG1 (Taurine-Upregulated Gene
1) (177) correlate strongly with poor prognosis in osteosarcoma, while the lncRNA ODRUL (Osteosarcoma Doxorubicin-Resistance Upregulated LncRNA) correlates strongly with poor response to doxorubicin chemotherapy (178). These preliminary data suggested that we may eventually be able to utilize lncRNAs as predictive biomarkers.
Figure 1: HOTAIR function in regulating HOXD gene expression Expressed from the HOXC locus, HOTAIR binds to SUZ12 and EZH2, part of the PRC2 complex. HOTAIR acts in trans by recruiting the PRC2 complex to the HOXD locus, and silences the locus by establishing H3K27me3 marks. 33
34
Figure 2: lncRNA-HIT and Hottip function in regulating Hoxa gene expression (A) lncRNA-HIT binds to the p100/CBP complex and maintains H3K27ac marks at HoxA genes, resulting in gene activation. (B). Hottip acts in cis by recruiting the WDR5/MLL histone methyltransferase complex to the Hoxa genes, activating gene by establishing H3K4me3 marks.
35
Figure 3: Different mechanisms of H19 in osteogenesis (A) H19 is processed into miR-675, which targets TGFB1 for degradation. (B) H19 acts as a sponge to sequester miR-141 and miR-22 away from their targets, thus leading to an upregulation of targeted mRNAs that may be required for osteoblast differentiation.
36
Figure 4: MALAT-1 function in osteosarcoma (A) MALAT-1 lncRNA is not polyadenylated, but instead possesses a triple helix formation at its 3’ end, and this protects the RNA from degradation. (B) MALAT-1 binds to SFPQ in the nuclear speckle, as a result releasing PTBP2 from the SFPQ/PTBP2 complex. Subsequently, unbound PTBP2 can induce tumor growth by increasing cellular proliferation and migration.
37
2. Genetic Engineering of Mesenchymal Stem Cells for Differential Matrix Deposition on 3D Woven Scaffolds This chapter is an adaptation of: Huynh, N.P.T., et al. (2018). “Genetic
Engineering of Mesenchymal Stem Cells for Differential Matrix Deposition on 3D
Woven Scaffolds.” Tissue Engineering Part A. Epub ahead of print.
2.1. Introduction
In the field of tissue engineering, there has been extensive interest in creating cartilage, bone, or combined osteochondral constructs that can provide enhanced fixation of engineered tissues into defect sites. Several previous approaches have demonstrated the ability to develop osteochondral constructs by combining multiple cells types (179-183), multilayered scaffolds (184-187), or multistep differentiation protocols (180, 188-194). However, it is still a major challenge to differentially direct cell fate determination into distinct lineages (i.e., cartilage and bone) from a single cell source, in a single culture system, while utilizing only one scaffold material. If proven efficacious, a single stage approach could streamline the engineering of multiphase tissues by circumventing the need for multiple cell types or multiple differentiation culture conditions.
MSCs are an attractive cell source for cartilage and bone tissue engineering mainly because of their easy access, high capability of in vitro expansion, and multipotent ability to express various cellular phenotypes, including osteogenic and chondrogenic lineages (195, 196). For example, previous work from our group has demonstrated the capability of growing engineered cartilage or bone
38 on three-dimensionally (3D) woven poly(ε-caprolactone) (PCL) scaffolds seeded with human MSCs under separate, prescribed chondrogenic or osteogenic conditions (197-201). However, MSCs require high concentrations of exogenous growth factors to enter distinct lineage programs in vitro (202-204). For example, transforming growth factor beta 3 (TGFβ3) is commonly used to induce MSC chondrogenesis (96, 97, 205, 206), while bone morphogenic protein 2 (BMP2) is often used to induce osteogenesis (207-209). These growth factor and environmental cues that define chondrogenic differentiation may interfere with osteogenic differentiation (128, 210-212), and vice versa. Thus, it remains challenging to direct chondrogenesis and osteogenesis simultaneously and in a site-specific manner within one in vitro culture environment due to inhibitory effects of chondrogenic-inducing TGFβ3 and osteogenic-inducing BMP2 on one another. For example, upon TGFβ3 signaling, Mothers against DPP homolog 3
(Smad3) is phosphorylated and translocates into the nucleus to repress Runt related transcription factor (Runx2) and Runx2-induced transcriptional activation of osteoblast differentiation genes (212). The inhibition is achieved by recruiting
Class IIa histone deacetylases (HDAC) 4 and 5 to the Smad3/Runx2 complex at
Runx2-binding DNA sequences, repressing the expression of Runx2 and its downstream targets (128, 213).
It had been suggested that medium supplemented with TGFβ3 can induce chondrogenesis in MSCs through the canonical SMAD2/3 pathway (97, 206,
214). In addition, RUNX2 is a transcription factor with multiple binding sites in the
39 promoter regions of bone matrix markers (215), and its osteogenic potential has been suggested in several studies in vivo or in vitro with defined osteogenic conditions (216-221). However, it remains unclear whether RUNX2 would exert its osteogenic effect in a chondrogenic environment with TGFβ3 stimulation.
Here, we developed a method to concomitantly induce osteogenic and chondrogenic differentiation from human bone-marrow derived MSCs on two separate 3D woven PCL scaffolds in one single culture system. We employed the TGFβ3/SMAD3 axis to produce a cartilaginous matrix on one scaffold.
Meanwhile, we engineered MSCs to potentiate mineralized matrix by overexpressing RUNX2 with SMAD3 modulation on the other scaffold. We hypothesized that a combination of SMAD3 knockdown and RUNX2 overexpression would significantly enhance mineral deposition even under the influence of TGFβ3 stimulation. The novelty of this research lies within our capability to, under the same biochemical cues, produce two scaffolds with distinct extracellular matrix (ECM) compositions by modulating intracellular signaling pathway of TGFβ3.
2.2. Materials and Methods
2.2.1. MSC culture and differentiation
Bone marrow was obtained from discarded and de-identified waste tissue from adult bone marrow transplant donors in accordance with the Institutional
Review Board of Duke University Medical Center. Adherent cells were expanded and maintained in expansion medium: DMEM-low glucose (Gibco), 1% 40
Penicillin/streptomycin (Gibco), 10% FBS (ThermoFisher), and 1 ng/mL basic fibroblast growth factor (Roche) (222). MSCs from three donors were expanded until end of passage 2, then an equal number of MSCs from each donor was combined to make a superlot. This approach has been shown to increase experimental throughput and utilize assay resources more efficiently, while effectively representing the average differentiation behavior of each of their contributing cell populations (223). Experiments with individual donors showed similar responses of MSCs to SMAD3 knockdown and RUNX2 overexpression regarding GAG and mineral depositions (Figure S1). MSCs were used at the end of passage 5 for all experiments, unless otherwise noted.
Cells were then induced to differentiate in a defined medium consisting of
DMEM-high glucose (Gibco), 1% ITS+ (Corning), 1% penicillin/streptomycin
(Gibco), 100 nM dexamethasone (Sigma), 50 µg/mL ascorbic acid (Sigma), 40
µg/mL L-proline (Sigma), and 100 nM β-glycerophosphate (Chem-Impex
International) (97, 224, 225). rhTGF-β3 (R&D Systems) was supplemented at 5 ng/mL.
We examined the effect of TGFβ3 signaling manipulation in both monolayer and scaffold culturing system, as outlined in Figure 5A, B. All media were exchanged every 3 days. Monolayer cells were harvested at 1 week post- induction for qRT-PCR analysis, and 3 weeks post-induction for alcian blue staining/ quantification, and alizarin red staining/ quantification. Scaffolds were harvested 5 weeks post-induction for histological and biochemical analysis.
41
2.2.2. PCL scaffold processing
PCL scaffold fabric was produced as previously described (199, 226), and cylindrical samples (4 mm Ø) were obtained, aseptically processed, and coated with 0.002% poly-L lysine (Sigma) (197).
2.2.3. SMAD3 knockdown with shRNA
Sequences for short hairpin RNA (shRNA) targeting the human SMAD3 transcript were designed using the RNAi consortium (TRC) GPP Web Portal
(Broad Institute) (227). shRNA specific to the SMAD3 transcript (SMAD3 shRNA) was selected for its superior efficiency compared to two other sequences (data not shown), and had the following target sequence: 5’-
TGAGCAGAACAGGTAGTATTA-3’. A vector delivering a scrambled sequence was used as control. The scrambled shRNA had the following sequence: 5’-
TTCTCCGAACGTGTCACGTT-3’. Both the target and the scrambled sequences have been confirmed to have no homology that is longer than 15 base pairs with any other human transcripts (NCBI Blast, Homo sapiens annotation release 108). shRNA sequences were cloned into a modified lentiviral vector (Figure 5C)
(Addgene #12247) (228) as described elsewhere (229) using MluI and ClaI restriction sites. This modified lentiviral transfer vector also contains a dsRedExpress2 expression cassette, which enabled fluorescence microscopy imaging. Although there was a puromycin resistant cassette in this vector, we did not perform antibiotic selection of transduced MSCs, as we have found that puromycin supplementation in the culture can inhibit the ability of MSCs to
42 proliferate and differentiate post-selection (data not shown). An overview of this vector is outlined in Figure 5C.
2.2.4. RUNX2 overexpression
The coding sequence of RUNX2 (NM_001024630.3) was synthesized and cloned into a modified lentiviral transfer vector (Addgene #12250) (228) using
MluI and XmaI restriction sites. An overview of this vector is outlined in Figure
5C.
2.2.5. Lentivirus production
To produce vesicular stomatitis virus glycoprotein pseudotyped lentivirus,
HEK293T cells were plated at 3.8 x 106 cells per 10 cm dish in DMEM-high glucose (Gibco) supplemented with 10% heat inactivated FBS (Gibco). The following day, cells were co-transfected with the appropriate transfer vector (20
µg), the second-generation packaging plasmid psPAX2 (Addgene #12260, 15
µg), and the envelope plasmid pMD2.G (Addgene #12259, 6 µg) by calcium phosphate precipitation (230). After 14-16 hours of incubation, 12 mL of fresh medium was exchanged. Twenty-four hours later, supernatant was harvested
(Harvest 1) and stored at 4°C, and another 12 mL of fresh medium was again exchanged. After another 24 hours, supernatant was harvested a second time
(Harvest 2). Harvest 1 and harvest 2 were pooled together, filtered through 0.45
µm cellulose acetate filters (Corning) to clear out producer cells, aliquoted, and stored at -80°C until future use.
43
2.2.6. Lentivirus transduction of MSC
One day before transduction, MSCs were plated at 4,500 cells/ cm2. The next day, MSCs were transduced in the presence of 4 µg/mL polybrene (Sigma). Four groups of lentivirus were used: scrambled shRNA, SMAD3 shRNA, scrambled shRNA with RUNX2, SMAD3 shRNA with RUNX2, and these transductions were carried out separately in individual culture vessels. Twenty-four hours post transduction, MSCs were washed once with PBS, and fresh expansion medium was exchanged. Non-transduced and transduced MSCs were expanded separately, each in expansion medium for 1 week. Subsequently, MSCs were either cultured in induction medium in monolayer experiments, or digested with
0.05% Trypsin-EDTA (Gibco) and seeded on 4-mm Ø poly-L-lysine coated scaffolds at 250,000 cells per scaffold. Scaffolds seeded with non-transduced or virally transduced cells were combined at this step and cultured in expansion medium for 1 week, and then in induction medium for 5 weeks. Due to the dsRedExpress2 expression cassette in our shRNA transfer vector (Figure 5C), scaffolds seeded with non-transduced and virally-transduced MSCs could be easily distinguished by fluorescence microscopy (Figure 5D).
2.2.7. qRT-PCR
Monolayer cells were harvested for quantitative real-time polymerase chain reaction (qRT-PCR) 1 week post-induction. RNA isolation was carried out following manufacturer’s protocol (Norgen). Reverse transcription was performed immediately after RNA was obtained, using Superscript VILO cDNA master mix
44
(Invitrogen). qRT-PCR was performed using Fast SyBR Green master mix
(Applied Biosystems) following manufacturer’s protocol. Primer pairs (Table S1) were synthesized by Integrated DNA Technologies, Inc. (IDT). Results are reported as log10 of fold change in expression of the gene of interest, normalized to ribosomal 18S expression by the ΔΔCt method.
2.2.8. Alcian blue and alizarin red staining and quantification
Three weeks post-induction, monolayer cells were washed once with PBS and fixed in 4% paraformaldehyde for 30 minutes at room temperature.
Wells were then washed with DI H2O, and submerged in 1 mL of alcian blue
(0.1% w/v in 0.1N hydrochloric acid) (Acros) for 4 hours or in 1 mL of alizarin red
(40mM, pH 4.1-4.3) (EMD Millipore) for 20 minutes at room temperature with gentle shaking. After staining, non-specific dye was rinsed off by washing with
1mL of DI H2O three times.
For quantification of the amount of alcian blue in stained wells, wells were incubated in 500 µL of 6M guanidine hydrochloride solution (Sigma) overnight at room temperature. The next morning, supernatant was collected into clear bottom 96-well plates. Known concentrations of alcian blue were used as standards, and absorbance was recorded at 595 nm with the Cytation 5 microplate reader (BioTek).
For quantification of the amount of alizarin red in stained wells, dye was released as previously described (231). Known concentrations of alizarin red
45 were used as standards, and absorbance was recorded at 405 nm with the
Cytation 5 reader (BioTek).
2.2.9. Biochemical analyses for DNA and GAG content of engineered scaffolds
On the day of harvest, fluorescence microscopy was used to distinguish between virally-transduced (red fluorescent signal) and non-transduced (no fluorescent signal) scaffolds for harvest. Scaffolds were subsequently transferred to individual wells, washed once with PBS, and stored at -20ºC until processing.
Scaffolds were lyophilized overnight, then digested in 125 µg/mL papain at 58ºC for 20 hours for biochemical analyses. DNA content was measured with the
PicoGreen assay (ThermoFisher), and glycosaminoglycan (GAG) was measured using the 1,9-dimethylmethylene blue assay (DMMB) (232) at 525 nm wavelength.
2.2.10. Histological processing of scaffold samples
Virally-transduced and non-transduced scaffolds were separated as described above. Scaffolds were washed once with PBS, fixed in 4% paraformaldehyde for 48 hours, paraffin embedded, and sectioned at 10 µm thickness. Slides were stained for Safranin O and Fast Green using a standard protocol (233), and for Von Kossa following manufacturer’s protocol (Abcam
#ab150687). For immunohistochemistry, the following primary monoclonal antibodies were used: type I collagen (Abcam #ab90395), type II collagen
(Developmental Studies Hybridoma bank #II-II6B3), type X collagen (Sigma
46
#C7974). All immunohistochemistry staining procedures were carried out with biotinylated goat anti-mouse secondary antibodies (Abcam #ab97021) as described previously (197). Human osteochondral sections were used as positive controls, and sample sections were incubated without primary antibody for negative controls.
2.2.11. Alkaline phosphatase assay
Virally-transduced and non-transduced scaffolds were separated as described above. Scaffolds were washed once with PBS, snap frozen in liquid nitrogen, and stored at -80ºC until processing. Scaffolds were homogenized with a biopulverizer (Biospec Products) chilled in liquid nitrogen. Subsequent steps were carried out following manufacturer’s protocol (Abcam #ab83371).
2.2.12. Statistical analysis
Statistical analysis was performed using Graphpad Prism version 7.03
(GraphPad Software). Alizarin red (n=4 per group) and alcian blue (n=5 per group) samples were analyzed using two-way ANOVA with Tukey post-hoc test
(α=0.05) using samples from two independent experiments. For biochemical and alkaline phosphatase assays, two-tailed paired t-tests were performed on scaffold pairs of non-transduced and virally transduced samples of each experimental group. These two experiments were independent from each other.
Three scaffold pairs were collected for biochemical assays per experimental group. Additional scaffold pairs (n=4-5) were collected for alkaline phosphatase assay per experimental group. Results were reported with mean values ±
47 standard error of the means (SEM). For qRT-PCR analysis, calculated fold change values were log-transformed before statistical analysis with Student’s t- test, and standard errors were calculated using the standard propagation of error method (234).
2.3. Results
2.3.1. SMAD3 knockdown or RUNX2 overexpression inhibits TGFβ3 induced cartilaginous matrix deposition
We examined the effect of SMAD3 knockdown and RUNX2 overexpression on MSCs in monolayer (Figure 5A). We utilized this as a rapid and simple screening system before moving to 3D scaffold culture (Figure 5B).
Human MSCs transduced with SMAD3 shRNA exhibited a significant reduction in SMAD3 level compared to those transduced with scrambled shRNA
(p=0.002) (Figure S2A). Bright field and fluorescence microscopy suggested that transduction efficiency was high (Figure S2C-D). We also investigated the efficiency of the RUNX2 overexpression, with the result showing a ~6-fold increase in the level of RUNX2 compared to non-transduced cells (p < 0.0001)
(Figure S2B).
In the absence of TGFβ3, there was minimal GAG deposition, as indicated by low level of alcian blue staining and quantification (Figure 6). When TGFβ3 was supplemented in the medium, GAG-rich matrix was observed in scrambled shRNA transduced wells as expected (5680 ± 412 μM) (Figure 6A, C). On the other hand, GAG production was inhibited either by knocking down SMAD3
48
(SMAD3 shRNA) (2820 ± 90 μM), by overexpressing RUNX2 (scrambled shRNA with RUNX2) (770 ± 202 μM), or by both (SMAD3 shRNA with RUNX2) (400 ± 50
μM) (p < 0.0001). Our results suggested that TGFβ3-induced GAG production could be inhibited. This raised the question whether these MSCs, inhibited from propagating TGFβ3 signaling, would start producing mineralized matrix instead.
2.3.2. SMAD3 knockdown and RUNX2 overexpression act to enhance mineral deposition in a chondrogenic environment
In monolayer, when MSCs were induced in the absence of TGFβ3, mineral deposition was observed in all samples by alizarin red staining (Figure 7A).
Quantitative analysis confirmed that there was no statistical difference among the groups in the amount of mineral produced when TGFβ3 was absent (Figure 7B).
However, in the presence of 5 ng/mL of TGFβ3, no staining was observed for the scrambled shRNA transduced wells. This was expected since TGFβ3 would induce MSCs towards a chondrogenic fate, and thus abolish mineral production.
When SMAD3 level was reduced in SMAD3 shRNA transduced wells, TGFβ3 signaling was dampened, and thus mineral deposition was detected to some extent, as observed with both alizarin red staining and quantification (87.83 ±
27.10 μM) (Figure 7A, C). When RUNX2 level was elevated in the scrambled shRNA with RUNX2 transduced group, mineral deposition was also detected
(228.25 ± 13.80 μM). When SMAD3 shRNA was used in conjunction with RUNX2 overexpression (SMAD3 shRNA with RUNX2), mineral deposition was achieved at the highest level (342.00 ± 32.17 μM) within the groups under 5 ng/mL of
49
TGFβ3, although the detected difference was not significant. Altogether, the data suggests that engineered MSCs could undergo matrix mineralization in a chondrogenic environment.
2.3.3. Differential GAG deposition was achieved in the dual-scaffold culture system
To test whether these engineered MSCs could be applied towards a tissue engineering context, we explored the effect of SMAD3 knockdown and RUNX2 overexpression in a system where two 3D PCL scaffolds of non-transduced (NT) and transduced cells were cultured adjacently in the same well (Figure 5B).
After 5 weeks in culture, NT scaffolds from all groups stained positive for
Safranin-O (Figure 8, left panel). There was a slight reduction in GAG deposition in scaffolds transduced with scrambled shRNA (Figure 8A, right panel). However,
GAG production was substantially reduced in scaffolds transduced with either
SMAD3 shRNA, or RUNX2, or both (Figure 8B-D, right panel). The finding was confirmed quantitatively by DMMB assay (Figure 9A) (p < 0.001). In addition, there was a difference in the amount of total DNA between NT and virally transduced samples, although there was no difference among the virally transduced groups (Figure 9B) (p < 0.05). When total GAG is normalized to the total amount of DNA produced, there was no statistical difference between the
NT and the scrambled shRNA transduced scaffolds. On the contrary, a decrease in GAG production was apparent in the virally-transduced scaffolds, when the
SMAD3 shRNA with RUNX2 samples exhibited statistically significant difference
50 from their NT counterparts (6.55 ± 1.64 mg/mg vs 18.86 ± 0.87 mg/mg, p =
0.0355). (Figure 9C). This suggested that non-transduced MSCs could be induced to make GAG-rich matrix in the dual-scaffold culture condition.
Moreover, SMAD3 knockdown and RUNX2 overexpression would inhibit the chondrogenic effect of TGFβ3 when used separately or together.
2.3.4. Differential type II collagen production was achieved in the dual-scaffold culture system
Type II collagen (COLII), another major component of articular cartilage, was observed in both the non-transduced and the scrambled shRNA transduced scaffolds. This indicated that scrambled shRNA did not interfere with collagen production (Figure 10A). On the contrary, COLII production was significantly reduced in scaffolds transduced with either SMAD3 shRNA, or RUNX2, or both
(Figure 10B-D). This recapitulated the differential GAG productions in non- transduced versus virally-transduced scaffolds, suggesting that TGFβ3 in the medium can induce nascent MSCs to make COLII-rich matrix in the dual-scaffold culture system. Moreover, either SMAD3 shRNA or RUNX2 may inhibit the chondrogenic effect of TGFβ3. It is important to note that this negative effect of
SMAD3 knockdown and RUNX2 overexpression on the ability of MSCs to produce GAG and type II collagen rich matrix was expected, since these MSCs were specifically engineered for the purpose of producing mineralization instead.
We also investigated the expression of type I collagen (characteristic of bone)
(Figure S3), and type X collagen (characteristic of hypertrophic cartilage) (Figure
51
S4). Scaffolds cultured in the defined induction medium condition exhibited low levels of both collagens, and there was no observable difference in type I collagen staining between the NT and the virally-transduced scaffolds in all four groups. However, there was a slight increase in type X collagen staining in the group of SMAD3 shRNA with RUNX2 compared to their NT counterparts (Figure
S4).
2.3.5. Differential mineral deposition was achieved in the dual- scaffold culture system
Mineral deposition in tissue engineered constructs was first assessed by Von
Kossa staining. NT scaffolds from all groups did not exhibit positive staining
(Figure 11A-D, left panel), indicating that there was no mineral deposition when
MSCs were cultured in the presence of TGFβ3. The strongest staining was observed in scaffolds containing cells transduced with SMAD3 shRNA and
RUNX2 overexpression (Figure 11D, right panel), suggesting that there was a synergistic effect of these two treatments on mineralized matrix in the culture condition. This phenomenon was confirmed quantitatively with alkaline phosphatase assay (p=0.0123) (Figure 11E). The amount of alkaline phosphatase was highest and most distinctive in the virally transduced scaffolds with both SMAD3 shRNA and RUNX2. Together, these data indicate that we could achieve matrix mineralization by knocking down SMAD3 in conjunction with overexpressing RUNX2 in our defined biochemical culture condition.
52
2.4. Discussion
A persistent challenge in the field of regenerative medicine has been the ability to engineer complex tissues comprised of multiple cell types and organized into regions of distinct ECM constituents, such as the osteochondral interface between cartilage and bone. Our study demonstrated that cartilage-like and bone-like matrix deposition can be achieved simultaneously by manipulation of the intracellular TGFβ3 signaling pathway and facilitation of RUNX2 expression. Here, we created a dual-scaffold single culture system by (1) supplementing exogenous TGFβ3 to induce chondrogenesis in one subset of non-genetically engineered MSCs (NT scaffolds) and (2) introducing shRNA to suppress SMAD3 level in conjunction with overexpressing RUNX2 to drive matrix mineralization in the other subset of MSCs (transduced scaffolds). Our histological and biochemical data showed distinct matrix phenotypes between the two scaffolds that were cultured in the same well. While the production of GAG and type II collagen was enriched in the NT scaffolds, transduced scaffolds with suppressed SMAD3 and enhanced RUNX2 expression did not display evidence of chondrogenesis. On the other hand, these genetically engineered scaffolds showed signs of mineralization, as demonstrated by Von Kossa staining and alkaline phosphatase activity quantification. Our study presents novel findings that blocking a specific node in the TGFβ3 signaling pathway in conjunction with enhancing RUNX2 can lead to mineralization even under TGFβ3 stimulation.
This method eliminates the need for prolonged culture in various differentiation 53 conditions and multi-stage processes to generate two distinct tissue matrices
(bone and cartilage). More importantly, it opens new avenues for functional interfacial tissue engineering for the treatment of subchondral defects.
While our base medium condition was selected to allow MSCs to differentiate towards both chondrogenic and osteogenic lineages, the addition of TGFβ3 abolished the matrix mineralization potential. However, this inhibitory effect of
TGFβ3 on mineral deposition was overcome by knocking down SMAD3. The effect was observed in monolayer and subsequently recapitulated in our dual- scaffold, single culture system. Our findings extend previous work on the intricate network of the TGFβ superfamily and their control of MSC differentiation. In addition to its stimulatory effects in chondrogenesis, TGFβ has also been suggested to play an inhibitory role in osteoblastic maturation through Smad3
[47,49]. Indeed, blocking of TGFβ type I receptor (TGFβRI) kinase through pharmacology has been demonstrated to enhance osteoblastic differentiation of mouse C2C12 cells (235). Pharmacologic inhibition of TGFβRI also increased bone mass and bone formation in postnatal mice (236). When we inhibited TGFβ signaling, albeit through the intracellular SMAD3 axis, we were able to detect comparable trends in MSC matrix mineralization and alkaline phosphatase activity.
A key finding of this study was the successful production of mineralized matrix under TGFβ3 stimulation, while preserving the potency to induce GAG and type
II collagen rich matrix in another subset of cells. We showed that medium
54 supplemented with TGFβ3 can induce robust chondrogenesis in non-transduced scaffolds, consistent with previous reports on cartilage tissue engineering using
MSCs (97, 206, 214). In this same environment, we demonstrated the ability to produce mineralization by inducing RUNX2 overexpression with simultaneous
SMAD3 modulation. To our knowledge, this is the first experiment implementing
RUNX2 overexpression in human MSCs with successful matrix mineralization and high alkaline phosphatase activity.
At the cellular level, TGFβ activates chondrogenic genes via the SMAD2/3- dependent pathway (Figure 12, left). When TGFβ3 binds to the receptor,
SMAD2/3 is phosphorylated, binds to SMAD4, and translocates into the nucleus, leading to transcriptional activation of chondrogenic genes. As a result, a cartilaginous matrix is produced. Simultaneously, TGFβ3 inhibits osteoblast differentiation genes via phosphorylated SMAD3 (Figure 12, top right). After translocation into the nucleus, SMAD3 recruits HDAC4/5 to the promoters of
RUNX2 and RUNX2 downstream targets, inhibiting their transcriptions (128,
213). In cells genetically engineered to express SMAD3 shRNA (Figure 12, bottom right), SMAD3 mRNA is either degraded or translationally inhibited, leading to de-repression of RUNX2 and its downstream targets. Removal of
HDAC4/5 at these genomic regions makes the promoters accessible to exogenous RUNX2 and thus synergistically enhances transcriptions of bone matrix markers, resulting in a mineralized matrix.
55
While scaffolds seeded cells transduced with SMAD3 shRNA and RUNX2 displayed matrix mineralization, they exhibited low levels of type I collagen. On the contrary, there was a slight increase in type X collagen staining in these scaffolds compared to the NT counterparts. While type I collagen is the most abundant protein of bone matrix (237), type X collagen is more characteristic of hypertrophic cartilage in the deep zone (238). This suggests that engineered
MSCs may not have undergone full osteoblastic differentiation, but instead matured into hypertrophic chondrocytes that eventually progressed towards matrix mineralization. Future work focusing on dissecting the intracellular network of endochondral and intramembranous ossification may shed light on the pathway underlying mineral deposition in these engineered MSCs.
Previous work from our group and others have demonstrated the feasibility of biomaterial-mediated gene delivery using viral vectors. This technique allows for spatially defined and directed cellular differentiation and tissue development by utilizing poly-L-lysine to immobilize gene delivery vectors (197, 219, 239). An important next step in our dual-scaffold single-culture system would be to apply this scaffold-mediated lentiviral transduction method for derivation of an osteochondral construct from one cell source in a single culture system. In more details, a bilayered scaffold with differential coating of lentivirus would induce a layer of cartilage (non-coated) on top of a layer of bone (coated with SMAD3 shRNA and RUNX2). Towards future clinical application, deliveries of these bilayered scaffolds into patients’ joint defects as cell-free systems can induce
56 differential matrix depositions of homing MSCs in situ. Thus, our findings provide proof-of-concept for the ability to simultaneously induce cartilaginous and mineralized matrices in spatially defined, bilayered scaffolds under such environment.
2.5. Conclusions
This study demonstrates the ability to genetically engineer MSCs for differential cartilaginous and mineralized matrix depositions on one type of scaffold in one culture system. A defined chondrogenic environment with medium-supplemented TGFβ3 facilitated the production of GAG and type II collagen on one scaffold. Simultaneously, engineered MSCs with enhanced
RUNX2 levels in modulation of SMAD3 generated mineralized matrix on scaffolds cultured in the same conditions. Controlling differential cell fate determination by this approach would allow for construction of complex multiphasic tissues without the need for prolonged culture in various medium conditions or elaborative scaffold structures. In the future, the efficacy of this method could be applied to fabricate an osteochondral construct from one single cell source for enhanced graft osseo-integration.
57
Figure 5: Overview of experimental approach (A) MSCs were cultured in monolayer. The use of SMAD3 shRNA and RUNX2 overexpression was hypothesized to induce MSCs towards osteogenesis, despite the presence of TGFβ3. (B) MSCs were pretransduced with SMAD3 shRNA or SMAD3 shRNA with RUNX2 before being seeded on the woven
58 scaffolds. Scrambled shRNA and scrambled shRNA with RUNX2 were used as control groups. Square box indicates two scaffolds cultured in the same well. Yellow and blue bars represent warp and weft fibers. (C) Schema of viral backbones used. A detailed description of the backbone components is provided in Table S2. We expect transduced cells to emit a strong signal under the RFP channel (due to dsRedExpress2 in our transfer vector construct). (D) Distinct differences between non-transduced and virally-transduced scaffolds cultured in the same well. Representative microscopic images of scaffolds in culture (week 5). Only the scaffold with virally transduced cells (scrambled control in this image) emitted RFP signal, while the NT scaffold did not. Upper panel: stitched image from multiple views to capture the entire scaffold pair. Lower panel: single images of scaffold pairs. Scale bar = 500 µm.
59
Figure 6: Reduction in GAG deposition with SMAD3 knockdown or RUNX2 overexpression (A) Alcian blue staining of monolayer cells cultured without (left) and with (right) TGFβ3. Distinct GAG rich matrix was observed in the scrambled shRNA transduced wells when stimulated with 5 ng/mL TGFβ3. All other wells exhibited very low levels of alcian blue staining. Scale bar = 500 µm. (B) Quantification of alcian blue dye in wells with 0 ng/mL (purple dots) or 5 ng/mL (red squares) of TGFβ3. n=5. Points represent independent specimen. Bars represent geometric means for each group. Two-way ANOVA with Tukey post- hoc test. Effect of TGFβ3 dose: p < 0.0001. Effect of virus type: p < 0.0001. Interaction between TGFβ3 dose and virus type: p < 0.0001. Groups of different letters are statistically different from one another.
60
Figure 7: Mineral deposition was achieved in the presence of TGFβ3 (A) Alizarin red staining of monolayer cells cultured without (left) and with (right) TGFβ3. The presence of TGFβ3 abolished mineral deposition. This effect was reversed in SMAD3 shRNA, scrambled shRNA with RUNX2, or SMAD3 shRNA with RUNX2. Scale bar = 500 µm. (B) Quantification of alizarin red dye in wells with 0 ng/mL (purple dots) or 5 ng/mL (red squares) of TGFβ3. n=5. Points represent independent specimen. Bars represent geometric means for each group. Two-way ANOVA with Tukey post-hoc test. Effect of TGFβ3 dose: p < 0.0001. Effect of virus type: N.S. Interaction between TGFβ3 dose and virus type: N.S. Groups of different letters are statistically different from one another. N.S. (not significant).
61
Figure 8: SMAD3 knockdown resulted in reduced GAG deposition Safranin O-Fast Green staining of scaffold pairs cultured in 5ng/mL TGFβ3. (A) NT and scrambled shRNA; (B) NT and SMAD3 shRNA; (C) NT and scrambled shRNA with RUNX2; (D) NT and SMAD3 shRNA with RUNX2. Square brackets indicate scaffold pairs cultured in the same well. GAG-rich matrix was observed in NT samples across all groups, while little GAG was observed in shRNA SMAD3 or RUNX2 transduced scaffolds. Scale bar = 100 µm.
62
Figure 9: SMAD3 knockdown resulted in reduced GAG deposition, quantified by biochemical assays Dot plots representing: (A) Total GAG production as quantified by DMMB assay, (B) Double-stranded DNA production as quantified by PicoGreen assay, (C) GAG normalized to DNA amount. Points represent independent specimen; bars represent geometric means for each group. Differential GAG deposition was observed in culture of scaffold pairs. SMAD3 knockdown with RUNX2 overexpression resulted in reduced GAG accumulation in the ECM matrix of virally transduced scaffolds (p < 0.001). n=3. Points represent independent specimen. Bars represent geometric means for each group. Red dots represent scaffolds seeded with non-transduced cells; orange squares represent scaffolds seeded with virally transduced cells. Two-tailed paired t tests for scaffold pairs in the same group (α = 0.05). N.S. (not significant)
63
Figure 10: Differential COLII expression. Immunohistochemistry staining of scaffold pairs cultured in 5 ng/mL TGFβ3. (A) NT and scrambled shRNA; (B) NT and SMAD3 shRNA; (C) NT and scrambled shRNA with RUNX2; (D) NT and SMAD3 shRNA with RUNX2. (E) Human osteochondral control. (F) Representative negative control with no primary antibody. Square brackets indicate scaffold pairs cultured in the same well. Type II collagen rich matrix was observed in both NT and scrambled shRNA transduced scaffolds, while differential type II collagen staining was observed for all the other groups. Scale bar = 100 µm.
64
Figure 11: Synergistic effect of SMAD3 knockdown and RUNX2 overexpression for osteogenesis (A-D) Von Kossa staining with nuclear red counterstain of scaffold pairs cultured in 5 ng/mL TGFβ3. (A) NT and scrambled shRNA; (B) NT and SMAD3 shRNA; (C) NT and scrambled shRNA with RUNX2; (D) NT and SMAD3 shRNA with RUNX2. Square brackets indicate scaffold pairs cultured in the same well. No mineral staining was observed except for the SMAD3 shRNA with RUNX2 transduced groups (brown). Scale bar = 100 µm.
65
Figure 11 (cont.): (E) Alkaline phosphatase activity quantification. n=4-5. Points represent independent specimen. Bars represent geometric means for each group. Red dots represent scaffolds seeded with non-transduced cells; orange squares represent scaffolds seeded with virally transduced cells. Two-tailed paired t tests for scaffold pairs in the same group (α = 0.05). N.S. (not significant)
66
Figure 12: Modulation of TGF signaling for simultaneous cartilaginous and mineralized matrices in one culture well Left: On scaffold with non-transduced cells, TGFβ3 binds to the receptor complex and phosphorylates SMAD2 and SMAD3. Phosphorylated SMAD2/3 binds to SMAD4, translocates into the nucleus and forms a complex with respective transcription factors to turn on chondrogenic genes. Top right: On scaffold with scrambled shRNA with RUNX2 transduced cells, TGF signaling phosphorylates SMAD3, which translocates into the nucleus and recruit HDAC4/5 to the promoters of RUNX2 and RUNX2 downstream targets, repressing their transcriptions. Bottom right: On scaffold with SMAD3 shRNA with RUNX2 transduced cells, SMAD3 shRNA is expressed, and thus degrades SMAD3 mRNA or inhibits its translation. This leads to de-repression at the promoters of RUNX2 and RUNX2 downstream targets, activating their expression. [*] Adapted from (213).
67
3. High-Depth Transcriptomic Profiling reveals the Temporal Gene Signature of Mesenchymal Stem Cells During Chondrogenesis This chapter is an adaptation of: Huynh, N.P.T., et al. (2018). “High-Depth
Transcriptomic Profiling Reveals the Temporal Gene Signature of Mesenchymal
Stem Cells During Chondrogenesis.” FASEB J. Epub ahead of print.
3.1. Introduction
Articular cartilage serves as a low-friction and load-bearing tissue, uniquely providing for joint motion for daily activities in life. However, articular cartilage has little or no capacity to heal or regenerate itself with injury or disease. Therefore, diseases associated with cartilage loss or degeneration often lead to progressive long-term pain and disability (240). Thus, there has been significant effort to develop new tissue-engineering approaches to enhance cartilage repair, often based on combinations of biomaterials and stem cells, to regenerate new cartilage to repair focal defects or resurface entire joints (241).
Mesenchymal stem/stromal cells (MSCs) were originally identified as a multipotent cell type found in the bone marrow (242), and more recently, similar cells have been identified throughout the body as populations of pericytes (243,
244). MSCs are easily accessible, expandable, and capable of differentiating into multiple skeletal lineages (196). For these reasons, they have been intensively investigated as a cell source for tissue engineering of cartilage, as well as a variety of other tissues (197, 198, 200, 201, 245-253). A number of in vitro studies of MSC-derived engineered cartilage have been performed in “pellet” 68 cultures, which are 3D aggregates of cells that produce cartilage-like structure after approximately three weeks (96, 97). Several molecular pathways have been proposed to play an important role during this process (reviewed in (254)).
However, the time course of transcriptomic events involved in the interplay and dynamic regulation of these pathways during MSC chondrogenesis has not been fully elucidated. While a number of individual genes (e.g., transcription factors, growth factors, signaling molecules, etc.) have been characterized at specific time points, it is likely that the coordinated activity of a number of factors make up the gene regulatory network (GRN) governing MSC chondrogenesis (255). For example, a recent study has examined this underlying process by differential gene expression analysis on microarray platforms (256).
To examine quantitatively the transcriptomic landscape of MSC chondrogenesis, we performed RNA-Seq at multiple time points during human
MSC chondrogenesis in pellet culture. Compared to microarray analysis, RNA-
Seq presents several additional advantages. For example, RNA-Seq possesses greater precision due to higher signal-to-noise ratio, independence from manufacturer’s limitation in the number of designed probes, a broader dynamic range to identify low abundance transcripts and highly differentially expressed genes, and the ability to probe for novel genes (257). Bioinformatic analysis of this RNA-Seq data has uncovered several dominant gene signatures, including intertwining changes of specific molecular pathways, subsets of highly-correlated genes of the chondrogenic functional module, and identification of long non-
69 coding RNA and micro-RNA candidates as regulators of chondrogenesis.
Furthermore, we have assembled this information into an online searchable encyclopedia of MSC chondrogenesis, providing a comprehensive knowledge database of both coding and non-coding RNAs that describes the transcriptome of MSCs through progression towards a chondrocytic fate.
3.2. Materials and methods
3.2.1. MSC isolation and expansion
The process of MSC isolation and expansion was carried out as described in section 2.B.1. For this chapter, three individual donors were used as biological replicates in subsequent experiments.
3.2.2. Chondrogenic differentiation
At the end of passage 4, MSCs were dissociated from the culture vessels with
0.05% Trypsin-EDTA (Sigma-Aldrich, St. Louis, MO, USA). Trypsin was subsequently inactivated with equal volume of expansion medium. Detached cells were collected by centrifugation (200 x g for 5 min), followed by 3 washes in pre-warmed DMEM-high glucose (Gibco, Thermo Fisher Scientific, Waltham,
MA, USA). The resulting cell mixtures were resuspended at 5.0 x 105 cells/mL in chondrogenic medium: DMEM-high glucose (Gibco, Thermo Fisher Scientific,
Waltham, MA, USA), 1% Penicillin/streptomycin (Gibco, Thermo Fisher Scientific,
Waltham, MA, USA), 1% ITS+ (Corning, Corning, NY, USA), 100 nM dexamethasone (Sigma-Aldrich, St. Louis, MO, USA), 50 µg/mL ascorbic acid
(Sigma-Aldrich, St. Louis, MO, USA), 40 µg/mL L-proline (Sigma-Aldrich, St. 70
Louis, MO, USA), and 10 ng/mL rhTGF-β3 (R&D Systems, Minneapolis, MN,
USA). To form pellets, 2.5 x 105 cells (500 µL) was dispended into each 15 mL conical tube, then centrifuged at 200 x g for 5 min. Pellets were cultured for 21 days at 37ºC and 5% CO2 with medium exchange every three days. Pellets were harvested on day 0, day 1, day 3, day 7, day 14, and day 21 for RNA isolation, and on day 3, day 14, and day 21 for histological and biochemical analyses.
3.2.3. Flow cytometry
Passage 4 MSCs were digested with 0.05% Trypsin-EDTA (Sigma-Aldrich,
St. Louis, MO, USA) and washed three times in PBS/ 1% FBS. Cells were incubated with a cocktail of positive surface markers (CD44, CD73, CD105,
CD90) (BD Biosciences, Franklin Lakes, NJ, USA) or with a negative surface marker (CD45) (BioLegend, San Diego, CA, USA) for 45 minutes. Unstained samples and samples with appropriate isotypes were used as controls. Samples with staining on separate channels were used for compensation. After incubation, cells were washed three times with PBS/ 1% FBS, and flow cytometry was performed. Data were analyzed with FlowJo, and reported as mean values ± standard deviation (n=3 independent specimens for each donor).
3.2.4. Biochemical analyses for DNA and GAG content
Harvested pellets were washed once in PBS, and stored at -20ºC until processing. Pellets were digested in 125 µL/mL papain at 60ºC overnight for biochemical analyses. DNA content was measured with the PicoGreen assay
(Thermo Fisher Scientific, Waltham, MA, USA), and glycosaminoglycan (GAG)
71 was measured as previously described (232) with the DMMB assay at 525 nm wavelength.
3.2.5. Histological analyses for GAG and collagen content
Harvested pellets were washed once in PBS, fixed in 4% paraformaldehyde for 48 hours, paraffin embedded, and sectioned at 10 µm thickness. A standard protocol for Safranin O and Fast Green staining was performed (233). For immunohistochemistry, the following primary monoclonal antibodies were used: type I collagen (#ab90395; Abcam, Cambridge, MA, USA), type II collagen (#II-
II6B3; Developmental Studies Hybridoma Bank, University of Iowa, Iowa City, IA,
USA), type X collagen (#C7974; Sigma-Aldrich, St. Louis, MO, USA). After clearing and rehydration, slides were digested in pepsin (Thermo Fisher
Scientific, Waltham, MA, USA) for epitope retrieval, and treated with methanol/ hydrogen peroxide for endogenous peroxidase quenching. The following steps were performed at room temperature: slides were incubated in blocking goat serum (Invitrogen, Carlsbad, CA, USA) for 30 min, followed by the appropriate primary antibodies for 1 hour, then with biotinylated goat anti-mouse secondary antibodies (#ab97021; Abcam, Cambridge, MA, USA) for 30 min. Streptavidin- horse radish peroxidase conjugate (Invitrogen, Carlsbad, CA, USA) with aminoethyl carbazole chromogen (AEC) (Invitrogen, Carlsbad, CA, USA) were used to develop signals. Staining was performed on human osteochondral sections as positive controls. Slides were also incubated without primary antibodies as negative controls.
72
3.2.6. RNA isolation, library preparation and sequencing
On day of harvest, pellets were washed once with PBS, snap frozen in liquid nitrogen, and stored at -80ºC until processing. Isolation of RNA was performed following manufacturer’s protocol (Norgen Biotek, Thorold, ON, Canada #48500) with the following modifications: for each time point, three pellets were homogenized with a pestle prior to addition of lysis buffer RL; RNA was eluted in
20 µL of diethyl pyrocarbonate (DEPC)-treated water. Samples were stored at -
80ºC and submitted to the Duke Center for Genomic and Computational Biology for library preparation and sequencing. Information on sample QC was provided in Table S1. Libraries were prepared using TruSeq Stranded Total RNA with
Ribo-Zero Gold kit (Illumina, San Diego, CA, USA). Sequencing was performed on a HiSeq2500 (paired end and 100 base-pair reads) with sequencing depth of
100 million reads per samples.
3.2.7. Data processing for computational analysis
Raw fastq files were processed with Trim Galore! (Babraham Bioinformatics,
Cambridge, UK) to eliminate low quality reads and remnant adapter sequences.
After trimming, processed reads were aligned to the human reference genome
(version GRCh38) by STAR (258), and the number of aligned reads to each annotated genes or transcripts (GENCODE v21) were counted using HTseq
(259). Normalized reads for each samples were performed with DESeq (259).
73
3.2.8. Computational analysis
At each pair of time points, we collected two lists of differentially expressed genes (upregulated and downregulated) using DESeq (260). Gene lists were uploaded to DAVID Bioinformatics Resources (https://david.ncifcrf.gov/) to perform gene ontology enrichment analysis (261, 262). The following parameters were employed for enrichment annotation: Homo sapiens was chosen as background, and results were obtained for biological pathways
(GOTERM_BP_DIRECT) and functional categories (UP_KEYWORD). After functional annotation clustering, an enrichment score cut-off of 2.5 was imposed to extract meaningful clusters out of resulting lists. For each cluster, pathways or terms with the highest Benjamini-Hochberg adjusted p values were chosen. We also record pathways or terms that appear multiple times across gene lists
(enriched at various time point transitions), provided their adjusted p values were the second highest in each cluster. Bar graphs were generated based on log transforms of adjusted p values. A similar pipeline was also used to compare the transcriptomic landscapes between MSC-derived pellets and human cartilage samples (downloaded from GSE106292) using DESeq and DAVID
Bioinformatics Resources with the same parameters.
Weighted gene correlation network analysis (WGCNA) was performed on normalized counts (as calculated by DESeq) of RNA-Seq data. Day 0 samples were excluded from this analysis for two reasons: their long distance from the other samples, and the lack of biochemical data at earlier time points for accurate
74 correlation of gene expression and trait. Adjacency matrix was built with a soft thresholding value of 7, based on recommendation from the WGCNA tutorial.
Gene cluster dendrogram was performed with a height cut-off of 0.25. Resulting gene lists of interesting modules were extracted and uploaded to DAVID for gene ontology analysis with the aforementioned parameters. Transcription factors from each module were extracted with information curated by JASPAR (263). Gene co-expression networks were prepared with Cytoscape (264) representing the top 20 hub genes of each module or transcription factors in each module with their top 3 connected genes.
3.2.9. Determination of k-mean value
A list of unique genes identified after DESeq analysis was clustered using the k-means clustering algorithm. K-mean value was determined by the elbow method.
3.2.10. Webtool development
The webpage was developed using the Shiny R package (265). The graph depicts normalized gene counts, with error bars representing the standard deviation of each gene at each time point. Data were fitted for a polynomial function of power 3. The heatmap depicts relative expression of transcript variants across time points. The color panel was scaled for each row, with dark purple being the highest expression and white being the lowest.
75
3.2.11. Statistical analysis
Statistical analysis was performed using R Statistical Software (Foundation for Statistical Computing, Vienna, Austria). Biochemical data were analyzed using two-way ANOVA with Tukey post-hoc test (α=0.05). Results were reported with mean values ± standard deviation.
3.3. Results
3.3.1. Surface marker distribution of mesenchymal stem cells
An overview of our experimental strategy is outlined in Figure 13A. MSCs from 3 separate donors were cultured and expanded in monolayer for pellet formation, with the pellets subsequently collected at six different time points for
RNA isolation and sequencing. At three of the collection time points (days 3, 14 and 21), biochemical and histological analyses were also completed. In addition, we collected MSCs cultured in monolayer for flow cytometry to characterize their surface marker distribution. MSCs from all three donors exhibited similar patterns of surface markers (Figure 13B): high in CD44, CD73, CD105, CD90, and low or non-expressive in CD45. Our MSC populations were mainly positive for CD44
(>97%), CD73 (>97%), and CD105 (>93%), with a majority (>75%) expressing
CD90 (Table 1). More importantly, the pattern was conserved across all three donors, suggesting a certain level of similarity in surface marker distribution of our biological replicates (Table 1).
76
3.3.2. MSCs synthesized a cartilaginous matrix under chondrogenic conditions
Pellets exhibited an accumulation of glycosaminoglycan (GAG) over our time course of analysis (Figure 14A), with a significant increase from day 3 (8.4 ± 3.1
µg; average from three donors) to day 14 (65.6 ± 15.5 µg; average from three donors). At later time points (day 14 and day 21), there was a slight increase in the amount of GAG produced within individual donors. In addition, on day 21, donor 3 exhibited the highest level of GAG (123.6 ± 9.0 µg), followed by donor 2
(109.7 ± 6.8 µg), then donor 1 (65.5 ± 5.3 µg). In contrast, the level of DNA was similar between days. When GAG was normalized to DNA, we observed the same phenomenon: a significant boost from day 3 to day 14 (4.6 ± 2.1 mg/mg compared to 29.9 ± 3.9 mg/mg; average from three donors), and a slight increase from day 14 to day 21 (29.9 ± 3.9 mg/mg compared to 39.7 ± 4.2 mg/mg; average from three donors). Interestingly, normalized data showed little donor to donor variability, with donor 3 still exhibiting the highest GAG per DNA amount on day 21 (42.2 ± 1.2 mg/mg).
Increased GAG accumulation over time was also reflected in SafraninO/ Fast
Green staining (Figure 14B). While pellets from day 3 exhibited little or no GAG staining, day 21 pellets from all three donors exhibited high level of GAG.
Immunohistochemistry indicated that in addition to GAG accumulation, these pellets were also synthesizing type II collagen – another marker of chondrogenesis – with defined positive staining on day 21 compared to day 3
77
(Figure 14C). In addition, we also performed immunohistochemistry on type I collagen (marker for fibrosis) (Figure 14D) and type X collagen (marker for cartilage hypertrophy) (Figure 14E). The protein levels of type I and type X collagen observed in our pellets were low during our time course of analysis.
3.3.3. Interactions of biological pathways during chondrogenesis
We performed principal component analysis of RNA-Seq data on normalized samples to explore their interrelationships (Figure 15A). All of the biological replicates were clustered together by time points, showing that significant changes in the transcriptome were observed over time in culture as the samples underwent chondrogenesis. In this analysis, day 0 samples were separate from the other days. Day 1 and day 3 samples appeared similar to one another as compared to those of later time points. Day 7 samples signified the transition moving from the multipotent to a more defined, differentiated chondrocyte-like state. There was no apparent difference among day 14 and day 21 samples.
Principal component analysis between our acquired dataset and publicly available RNA-Seq data from Evseenko and colleagues (GSE106292) indicated that engineered cartilage is transcriptomically distinct from human articular cartilages of all stages reported (embryonic, adolescent, and adult) (Figure 15B), although MSC pellets at all time points appeared more similar to human adult cartilage (compared to human embryonic cartilage or human adolescent cartilage). In-depth analysis of differentially expressed genes revealed similar patterns in gene ontologies. Compared to human cartilage samples, day 21
78
MSC-derived pellets contained highly expressed genes in collagen organization, extracellular matrix organization, skeletal system development (Table 2, Table
S2). In addition, genes that were lowly expressed in pellets were pertaining to transcriptional regulation (Table S2). MSC-derived pellets were different from human embryonic samples at the limb bud stage (samples harvested at 6 weeks), indicating the in vitro system utilized to engineer cartilage followed a distinct path and thus may resemble other pathways in skeletal system development.
To further analyze the transcriptomic profiles of MSC chondrogenesis, we used RNA-Seq to investigate the expression patterns of markers of MSC and their potential lineages (Figure 15C). As expected, in our defined induction medium, the expression of chondrogenic markers was significantly upregulated
(e.g., COL2A1, ACAN, COL9A1, COL11A1, COMP). This process occurred concurrently with a decrease in the expression of MSC markers (e.g., CD44,
ENG, NT5E, THY1). In addition, we examined how markers of other potential mesenchymal lineages responded, especially osteogenesis and adipogenesis.
Almost all non-chondrogenic cell fate markers either remained constant or were downregulated in our time course differentiation, with the exception of SP7. While
SP7 was upregulated, its expression as measured by RPKM value was relatively lower compared to those of chondrogenic markers (Table S1). Overall, transcriptomes recapitulated the trend detected on a protein level (Figure 14A),
79 suggesting robust MSC chondrogenesis, with RNA-Seq data similarly reflecting the regulation of gene expression during chondrogenesis.
To examine the biological pathways involved in this process, we performed differential gene expression analysis between pairs of consecutive time points.
Table 3 summarizes the number of genes that are significantly up- or down- regulated for at least 2 folds (|log2(foldchange)| > 1 and p ≤ 0.05). MSCs exhibited the greatest change from day 0 to day 1, with 2,087 genes being upregulated and 1,860 genes being downregulated. On the contrary, the least noticeable difference was between day 14 and day 21, with only 11 genes being upregulated and 5 genes being downregulated. A complete list of differentially expressed genes is shown in Table S3. Differentially expressed genes emerged into four main expression patterns as identified by the k-means clustering algorithm (Figure 15D). The first cluster was composed of genes that were continuously upregulated, while the second cluster contained genes that were continuously downregulated during chondrogenesis. The third and the forth clusters were constituted by genes that were more dynamically regulated during earlier time points (day 1 and day 3), but subsequently returned to day 0 expression levels past the day 7 timepoint. Subsequently, we performed gene ontology analysis with DAVID (261, 262) to pinpoint the orchestration of biological pathways involved during chondrogenesis. Figure 15E depicts the most enriched pathways as MSCs were differentiating towards a chondrocyte- like state. Because the number of differentially expressed genes from day 14 to
80 day 21 was small, we did not find any significantly enriched pathways for this pair of time points (adjusted p values > 0.05). Therefore, we examined whether there would be any significant pathways from day 7 to day 21. Importantly, pathways related to chondrogenic development were enriched from lists of upregulated genes (glycoprotein, growth factor, extracellular matrix organization, collagen catabolic process). In addition, cellular proliferation was highly upregulated during day 1 to day 3, and subsequently downregulated at later time points (DNA replication, nucleosome assembly, chromosome, cell cycle), suggesting a switch from a proliferative to a biosynthetic state during chondrogenic differentiation.
3.3.4. Modules of gene co-expression networks highly correlated with GAG accumulation
Hierarchical clustering of RNA-Seq data of all samples recapitulated that samples showed greater similarity by day of chondrogenic induction than by donor of origin, particularly at earlier time points. Day 21 samples separate themselves from the rest, signifying a distinct state in transcriptomic composition.
The content of GAG, DNA, and GAG per DNA increased over time, and correlated with sample groups going from day 3 to day 14 to day 21 (Figure 16A heatmap). Since donor 1 had the least chondrogenic potential (lowest GAG level)
(Figure 14A), samples derived from donor 1 at mid time points (day 7, day 14) were associated with samples of earlier time points from the other two donors: donor 1-day 14 transcriptome was more similar to those of donor 2-day 7 and donor 3-day 7.
81
Unsigned weighted gene correlation network analysis (WGCNA) (266, 267) was utilized to identify subsets of genes that were highly correlated with changes in GAG, DNA, and GAG per DNA, hereby referred to as traits. Genes were next clustered into 94 modules of correlated expression patterns (Figure S1A and
Table S4). Subsequently, the eigengene of each module was investigated for correlation with changes in GAG per DNA. WGCNA defines eigengene as the first principal component of a gene cluster, thus effectively represents the gene expression profile of that module. For ease of tracking and visualization, modules are designated with arbitrary colors. The adjacency heatmap (Figure 16B) grouped modules whose eigengenes are highly correlated with the trait of interest into blocks of meta-module. WGCNA defines the adjacency value as follow: Aj = (1 + correlation)/2. As a module is highly correlated with the investigated trait (GAG per DNA), its adjacency value will either be close to 0, or close to 1. With an absolute correlation value cut-off of 0.8 (Ai < 0.1 or Aj > 0.9),
21 interesting modules emerged and required further investigation (Figure 16C).
Module steel blue was the most positively correlated with the traits and module magenta4 (referred to as magenta from this point) was the most negatively correlated. We also examined the relationships between these modules to changes in GAG (Figure S1B), and changes in DNA (Figure S1C). As expected, correlations between modules and GAG as trait was very similar to those between modules and GAG per DNA as trait. Interestingly, with DNA as trait, we observed a different order in correlated modules, possibly due to a relatively
82 moderate increase in DNA compared to the magnitude of increase in GAG over time.
A limitation in our above analysis of gene-trait correlation was the missing values from biochemical data, which could potentially lead to identification of insignificant gene clusters. To validate our finding, we extracted the expression values of chondrogenic markers (COL2A1 and ACAN) and performed WGCNA with these vectors as traits (Figure S2A, B). Our rationale was that modules highly correlated to chondrogenic expression patterns would also be composed of genes heavily involved in the process. Consistent with our initial findings, module steel blue and module magenta were again established as the most interesting gene clusters that were highly correlated to the expression patterns of both COL2A1 and ACAN (Figure S2C, D).
3.3.5. Characteristics of highly correlated modules
From WGCNA, we identified the modules steel blue and magenta as most relevant for further investigation. Co-expression networks were built for the 20 top hubs (Figure 17A and Figure 18A), where genes with the most connections were centered inside the networks. It has been proposed that centrally located hubs were essential as shared links among interacting pathways (268), and that these hubs were key control points in module formation (269). We were also interested in the relationship between transcription factors and their potential downstream effectors. Thus, we constructed a gene co-expression network of all transcription factors (TFs) with defined DNA binding motifs identified in each
83 module with the top three genes that were most correlated to each TF (Figure
17B and Figure 18B). For ease of visualization, connections among non-TF genes were hidden in represented figures. Highlights within the network from module steel blue involved TFs established for their roles in skeletal system development (e.g., FOS, JUNB, FOXA2, SOXs, RARs) (270-275). In contrast, transcription factors involved in module magenta mostly have unknown functions in the skeletal tissue, except for JUN and HOXD8. While JUN had been suggested to form complexes with FOS to regulate bone development (274),
HOXD8 had been shown to affect mesodermal axial patterning and positioning of the hindlimbs (276).
Defined expression patterns for each module (Figure 17 C-D, Figure 18 C-D) are visualized by a heatmap of modular genes and the pattern of eigengene vector. It is important to note that by constructing unsigned networks, negatively and positively correlated genes would be clustered into one module provided they passed the thresholding cut-off. Therefore, although heatmaps of modular genes showed patterns that closely resemble the eigengene, there were subsets of genes whose expressions were reversed compared to the eigengene pattern.
The distinction between module steel blue and module magenta lay within modular gene dynamics between day 1 and day 7. Magenta modular genes exhibited a more rapid, steeper change in their expression pattern (rapid changes) compared to those of module steel blue (delayed changes) during these earlier time points.
84
We also examined the next five most correlated modules (mediumpurple2, cyan, sky blue, dark violet, sienna3). Their top 20 hub genes and TF networks were depicted in corresponding gene co-expression networks (Figure S3 A, D, H,
L, O and Figure S3 E, I) with accompanying characteristic expression patterns for each module (Figure S3 B, F, J, M, P and Figure S3 C, G, K, N, Q).
To elucidate the biological pathways underlying module steel blue and module magenta, we performed enrichment analysis on modular genes with
DAVID (Figure 17E and Figure 18E). There were 1,172 genes for module steel blue and 395 genes for module magenta that were used as independent inputs to
DAVID. Encouragingly, skeletal system development was detected as the most enriched biological pathway involved in module steel blue. Pathways previously suggested to be connected to musculoskeletal diseases such as Dwarfism, polymorphism, and Sticker’s syndrome were also enriched in our analysis. For these reasons, we propose that this cluster of genes represent crucial players in chondrogenesis, and thus hereby refer to it as the “chondrogenic functional module”. The genes comprising module magenta were mainly involved in cytoskeletal organization, albeit possessing a relatively smaller significant value.
3.3.6. Identification of miRNAs and lncRNAs from chondrogenic functional module
We proceeded to further investigate module steel blue. Within this module, we identified 230 annotated lncRNA genes (Table S5). In addition, we examined whether known targets of miRNAs would be enriched in module steel blue since
85
RNA-Seq data do not capture small non-coding information. We selected miRNAs with fewer than 50 predicted targets and high prediction scores utilizing data provided by miR database (277, 278). By this method, we could detect 498 associated upstream miRNAs whose target mRNAs were enriched in our chondrogenic functional module (Table S5).
3.3.7. Web-based encyclopedia of MSC chondrogenesis
Our processed RNA-Seq data were developed into a web-based encyclopedia for easy access and user-friendly interface (http://msc- browser.guilaklab.com). Here, we have arranged data into two categories: expression of genes and expression of transcript variants over the time course of chondrogenesis. Their patterns can be readily retrieved from a drop-down menu of gene names of interest. As an example, we looked up ACAN in our web browser (Figure S4). The resulting graph portrays the expression pattern of
ACAN during chondrogenesis, and the accompanying heatmap reveals that there were 4 splice variants of ACAN, whose patterns conform to that of the gene itself.
3.4. Discussion
Advances in high throughput sequencing have provided critical new methods for rapid and cost-effective genomic and transcriptomic research. Our study provides an extensive RNA-Seq database that followed MSC chondrogenesis in vitro at a deep sequencing level (100 million reads per library). This depth of sequencing is particularly valuable as it provides the possibility to identify novel, unknown transcripts, as well as the capability to distinguish differential levels of 86 lowly expressed transcripts (279). With three biological replicates and little donor- to-donor variation, our data presents an in-depth encyclopedia of the gene expression changes induced during MSC chondrogenesis. Our study demonstrated that this was a complicated process with many concerted pathways, beginning with a transient upregulation in cellular proliferation and transitioning towards the building of collagen, glycoprotein, and extracellular matrix. Differentially expressed genes were divided into four main expression patterns, exhibiting continuous or transient upregulation and downregulation.
Furthermore, gene co-expression network analysis on this dataset led us to identify a chondrogenic functional module composed of 1,172 genes. The eigengene (first principal component) of this module is highly correlated with changes in primary chondrocytic markers, including GAG protein level, COL2A1 expression, and ACAN expression over time, signifying that corresponding modular genes may be heavily involved in chondrogenesis. We speculate that in- depth investigation of this gene cluster will identify many more crucial players and potentiate our ability to understand the signaling axes involved in chondrogenesis. In this respect, a more comprehensive understanding of this process will hopefully provide important insights that improve MSC-based cartilage tissue engineering. More importantly, we identified a list of non-coding genes (small non-coding and long non-coding) that could potentially be both prognostic and therapeutic. For example, understanding of the gene regulatory networks or specific non-coding RNAs that regulate chondrogenesis may allow
87 for development of new protocols for more rapid or complete differentiation of
MSCs in this context.
An accompanying phenotype of chondrogenically induced MSCs is their potential for hypertrophic differentiation with an upregulation of COL10A1,
MMP13, and sometimes fibroblastic markers such as COL1A1, particularly when implanted in vivo. In our study, pellets harvested on day 21 showed increased staining for collagen type I and collagen type X compared to earlier time points, signifying that at the protein level, the cells had started to express a hypertrophic or fibroblastic phenotype in this defined culture system. Similarly, at the gene expression level, we observed a gradual increase in COL1A1, COL10A1, and
MMP13. This suggests that these MSC pellets may be en route to becoming endochondral cartilage as have been reported in the literature (280, 281). Further investigation of the pathways involved in these processes may provide important insights into regulating MSC hypertrophic differentiation, either positively or negatively, for tissue engineering applications.
Pellet culture has been extensively used as an in vitro model for chondrogenesis in cartilage tissue engineering (96, 97). Under appropriate culture conditions, these 3D aggregates usually exhibit extracellular matrix rich in
GAG and collagen type II, concomitant with an increase in gene expression of well-defined chondrogenic markers such as COL2A1, ACAN, and SOX9 (282). In contrast, it has been suggested that there was little proliferation and, to some extent, gradual cell loss within these pellets (282). Conforming to existing
88 literature, our biochemical and histological assays confirmed robust chondrogenesis with progressive accumulation of GAG and collagen type II in the extracellular matrix, with relatively stable DNA content over time.
Nonetheless, we observed via principal component analysis that at these early time points (up to 21 days), cartilage engineered from MSCs is transcriptomically distinct from human articular cartilages (Figure 15B). This observation reiterates current opinions that early MSC derived pellet aggregates form a tissue in vitro that is distinct from native cartilage, indicating that further studies of MSC differentiation or longer maturation times may be necessary for improving the hyaline-like quality of cartilage engineered from MSCs.
While pellet culture was originally thought to mimic embryonic limb development where mesenchymal condensation initiates chondrogenesis (283), interplay between pathways and dynamics of chondrogenic differentiation had not been fully elucidated. Here, transcriptomic profiling of MSC chondrogenesis has identified several key logical pathways, such as glycoprotein, extracellular matrix organization, and collagen catabolic process. At first glance, it was counterintuitive to establish collagen catabolic process – the breakdown of collagen – as one of the upregulated pathways in our analysis. However, further investigation indicated that genes identified within this ontology term were mostly collagen (COL2A1, COL11A1, COL9A1) instead of proteases or collagenases, thus we propose that the ontology term would be more correctly categorized as
“collagen regulation”. It is important to note that there were also a few matrix
89 metalloproteinases within this GO term that were upregulated from day 0 to day
1, and MMP13 was upregulated from day 7 to day 21 – a phenomenon that has been reported previously (280, 281).
Although several cell surface markers are often used to identify MSCs, it is widely accepted that the distribution of surface markers can be influenced by many factors, including the culture conditions (222), isolation techniques (284,
285), and cell source (286). Therefore, it is valuable to characterize the surface marker distribution of specific MSCs populations. Here, we showed that MSCs from three donors are relatively homogeneous for three surface markers (CD44+/
CD73+/ CD105+) (> 95% of cell populations), although CD90 – another characteristic cluster of differentiation – was only positive in about 80% of the populations. Early publications on isolated human MSCs indicated that cells were at least homogeneous at passage 1 and passage 2 (196) for a composition of surface markers, including CD44, CD73, CD105, and CD90 (287, 288). While
CD73 and CD105 appeared to be stable post monolayer expansion, some cases indicated that MSCs in fact gained CD44 expression (reviewed in (289)).
Complementary to the MSC markers, we found that MSCs were negative for expression of hematopoietic stem cell specific CD45. Furthermore, the cell surface marker profiles in the present study are consistent with those reported under similar expansion culture condition, which have subsequently been identified to enhance chondrogenic potential (222).
90
Weighted gene co-expression network analysis (WGCNA) is a powerful tool for extracting meaningful biological cues from large, high throughput datasets.
The use of a gene co-expression network is most advantageous in its ability for global interpretation without reliance on prior knowledge (290). It has been extensively documented that co-expressed genes are functionally related and co- regulated (291, 292). Identification of genes within the same module can offer profound insight into uncharacterized genes with likely similar biological functions, or involvement of new genes in previously curated pathways. With
WGCNA, we obtained a chondrogenic functional module of 1,172 genes. This module contains characteristic markers for cartilage matrix: COL2A1, ACAN,
COMP, COL9A1, COL11A1. Mechanosensitive ion channels - TRPV4 and
PIEZO1 - crucial for cartilage maintenance, homeostasis, and injury response, were also present in this module (293-296).
A subset of transcription factors and their highly co-expressed genes in module steel blue was highlighted. Among the participants were members of the
SOX family, retinoic acid receptors, FOS/ JUN complex, and FOXA2. While it is widely accepted that the SOX trio (SOX5/6/9) controls chondrogenesis (270-
272), SOX6 was surprisingly not a steel blue modular gene. We speculate that
SOX6 may be regulated by a different mechanism, and thus exhibited a different expression behavior. The other interesting transcription factors had been established to function in skeletal system development. For example, mice deficient in retinoic acid receptors (RARB/ RARG double knock-out) displayed
91 stunted skeletal growth (275). FOS and JUN proteins form the transcription factor complex AP1 and play an important role in osteoblast maturation and osteoclastogenesis (274). FOXA2 factor regulates chondrocyte hypertrophy by upregulating collagen type X and MMP13 (273). Transcription factors crucial for chondrogenesis and hypertrophy were highly connected to one another, reiterating the entangled pathways that essentially result in endochondral ossification (281, 297). Hence, the ability to dissect and separate these routes will provide insights into preventing unwanted differentiation pathways such as
MSC hypertrophy.
By further examining this module, we identified 230 lncRNAs and 498 associated miRNAs. Evident physiological connection between correlated genes indicates the significance of these non-coding transcripts and their roles in chondrogenic regulation. Initially referred to as genomic remnants, lncRNAs are gaining remarkable attention as crucial molecules in cellular maintenance and lineage commitment (89, 298). Recently, a growing number of lncRNAs have been identified as important players in musculoskeletal development and diseases (reviewed in (299)). In this study, we identify 230 modular lncRNA candidates. Among these were a few whose functions in skeletal system development had been elucidated, such as H19 (130, 139) and HIF1A-AS1 (137,
300). This list therefore represents a composition of long non-coding genes that requires further investigation to fully understand their underlying mechanisms.
92
Historically, miRNAs represent a major class of non-coding RNAs that are more established and extensively investigated. In the realm of musculoskeletal system development, let-7, miR-20b, miR-363 had been shown to upregulate during limb formation in chick embryos (301). miR-31 and miR-338 were differentially expressed during blastema formation in appendage regeneration and were conserved across evolutionarily distant species (302). miR-145 was upregulated during MSC chondrogenic differentiation and regulated Sox9 expression in mice (303). It was promising that we identified the aforementioned miRNAs in our list of module-associated miRNAs. Thus, further investigation is prompted to harness the full potential of manipulating these miRNA levels for functional cartilage tissue engineering.
3.5. Conclusions
In summary, we have provided an in-depth database of MSC chondrogenesis. Rigorous analysis of this dataset identified several subsets of interesting genes: protein-coding, long non-coding, and small non-coding.
Routinely, genes are studied as individual concepts and while a singular pathway can be examined in detail, their integration into larger genomic/ transcriptomic scheme is often lost. Thus, a more robust gene signature like the chondrogenic functional module could broaden the knowledge of chondrogenesis and cartilage pathology. The data has been developed into an easily accessible web-based encyclopedia that can act as a reference point for additional pathway analysis and hypothesis-driven experiments designed to improve MSC differentiation. 93
Figure 13: Distribution of MSC surface markers (A) Experimental design. MSCs were formed into pellets after passage 4 and cultured under chondrogenic conditions for up to 21 days. Pellets were harvested at days 0, 1, 3, 7, 14, and 21. (B) Representative histograms from three donors. Control samples stained with isotypes were depicted in red; samples stained with surface markers were depicted in blue.
94
Figure 14: MSC chondrogenesis in pellet culture (A) Biochemical assays measuring GAG and DNA over the time course of analysis. There was an increase in GAG accumulation from day 3 to later time points (left and right panel), while DNA levels remained similar across days (middle panel). Mean ± SD. n = 3 independent specimens per donor (3 donors as biological replicates). Two-way ANOVA followed by Tukey post-hoc test (α=0.05). Groups of different letters are statistically different from one another. (B) Safranin-O/ Fast Green staining on pellets from three donors at different
95 time points. Scale bar = 500 µm. (C-E) Immunohistochemistry staining on pellets from three donors at different time points: (C) collagen type II; (D) collagen type I; (E) collagen type X. Scale bar = 500 µm.
96
Figure 15: Differential expression (DE) analysis
97
(A) Principal component analysis of 18 MSC-derived pellet samples. Samples were closely related by days, and there was minor donor to donor variation. (B) Principal component analysis of MSC-derived pellet samples versus human cartilage at different developmental stages. (C) Expression patterns of MSC markers and markers of potential lineages. Gene expression was represented as log2 fold-change compared to day 0 samples. (D) k-means clustering grouped DE genes into four patterns. Gene expression was represented as log2 fold-change compared to day 0 samples. (E) Bar graphs of enriched pathways from gene ontology analysis.
98
Figure 16: Modules and their relationships to biochemical changes (A) Dendrogram of RNA-Seq samples and corresponding changes in traits. Lowest values were depicted in white; highest values were depicted in red; missing values were depicted in grey. (B) Eigengene adjacency heatmap with blocks of meta-modules. (C) Modules whose eigengenes are highly correlated with changes in GAG per DNA and their corresponding Pearson correlation values.
99
Figure 17: Characteristics of module steel blue (A) Co-expression network of 20 top hubs. (B) Co-expression network of involved transcription factors (red) and their top 3 connected genes (steel blue). (C) Heatmap of steel blue modular gene expression. (D) Eigengene vector of module steel blue. (E) Enriched GO terms and pathways from modular genes with gene ontology analysis.
100
Figure 18: Characteristics of module magenta (A) Co-expression network of 20 top hubs. (B) Co-expression network of involved transcription factors (red) and their top 3 connected genes (magenta). (C) Heatmap of magenta modular gene expression. (D) Eigengene vector of module magenta. (E) Enriched GO terms and pathways from modular genes with gene ontology analysis.
101
Table 1: Percentage of cell populations that were positive for specific markers. n=3 for each donor. Mean ± SD. CD44 CD73 CD105 CD90 CD45 Donor 1 99.0 ± 0.5 99.1 ± 0.5 94.3 ± 0.1 78.0 ± 3.3 0.1 ± 0.1 Donor 2 97.2 ± 4.5 97.3 ± 4.5 93.0 ± 5.8 75.9 ± 1.6 0.4 ± 0.1 Donor 3 98.7 ± 1.2 98.7 ± 1.0 96.2 ± 0.6 77.2 ± 2.2 0.1 ± 0.0
102
Table 2: Differentially expressed genes between MSC-derived pellets and human cartilage samples
Number of genes Number of genes upregulated in Day 21 downregulated in Day pellets 21 pellets Day 21 MSC pellets 1332 764 compared to embryonic cartilage Day 21 MSC pellets 2075 798 compared to adolescent cartilage Day 21 MSC pellets 1017 566 compared to adult cartilage
103
Table 3: Differentially expressed genes through MSC chondrogenesis Days Number of genes Number of genes upregulated downregulated Day 0 to Day 1 2087 1860 Day 1 to Day 3 322 170 Day 3 to Day 7 354 518 Day 7 to Day 14 32 135 Day 14 to Day 21 11 5 Day 7 to Day 21 326 452
104
4. The long non-coding RNA GRASLND enhances chondrogenesis via suppression of interferon type II
4.1. Introduction
Despite completion of the human genome project, it remains challenging to understand the function of a large part of our species genome. Since most computational predictions have relied on the presence of open reading frames
(ORFs), previously annotated cDNAs, or sequence conservation with genes from other organisms, non-coding RNAs (ncRNAs) were difficult to identify because they neither possessed ORFs nor were evolutionarily highly conserved (304).
The initial draft reported canonical classes of ncRNAs, such as transfer RNAs, ribosomal RNAs, small nucleolar RNAs, and small nuclear RNAs. There was, however, a special appearance of the X Inactive Specific Transcript (XIST) – then referred to as “ncRNA(s) of enigmatic function”. Years of studies have revealed that XIST in fact belongs to a novel and defined class of ncRNAs now known as long non-coding RNAs (lncRNAs). In 2009, Guttmann and colleagues used chromatin-state mapping and identified transcriptional units of functional large intervening non-coding RNAs (lincRNAs): those actively transcribed in regions flanking protein-coding loci (305). Follow-up loss-of-function studies by the same group indicated that these lincRNAs were indeed crucial for the maintenance of pluripotency in embryonic stem cells. This was perhaps the inception of lncRNAs and their regulatory importance. While there exists a growing literature of lncRNA discoveries and characterizations in a multitude of
105 tissues and cellular processes, research on their roles in the musculoskeletal system remained limited. Pioneering examples include: lncRNA-HIT (HOXA
Transcript Induced by TGFβ) (119), functioning in the epigenetic regulation during early limb development, and ROCR (Regulator of Chondrogenesis RNA)
(306), acting upstream of SOX9 to regulate chondrocyte differentiation, among others (299).
While interferons (IFN) are widely known for their antiviral response, they can also act in the regulation of other cellular processes (307). There are two types of
IFN: type I includes mainly IFN alpha (IFN-α), and IFN beta (IFN-β); type II contains IFN gamma (IFN-). While IFN type I ligands are encoded in genes found on human chromosome 9, IFN-is located on human chromosome 12.
Bearing no structural homology to type I, IFN- also does not compete for the same receptors: IFN-α and IFN-β form complexes with IFNARs (Interferon Alpha and Beta Receptors), whereas IFN- binds to IFNGRs (Interferon Gamma
Receptors) (307-309). However, both types transmit through the Janus kinase/ signal transducers and activators of transcription (JAK/STAT) pathway to activate similar set of interferon stimulated genes (ISGs), despite possessing distinct DNA responsible elements (interferon-stimulated responsible element (ISRE) for type I and gamma interferon activation site (GAS) for type II). In addition, IFN- has been implicated in other non-viral responses, most notably its priming effect in auto-immune diseases such as lupus nephritis, multiple sclerosis, or rheumatoid arthritis (310). In musculoskeletal system development, interferon signaling has 106 been shown to induce osteoblast and osteoclast differentiation (311-314).
However, it remains to be determined whether, and through what mechanisms,
IFNs participate in stem cell induction towards the chondrocyte lineage or if they play a role in the development of OA.
Our recent publication that quantifies the transcriptomic trajectory of MSC differentiation into chondrocytes has offered unprecedented opportunities to investigate crucial lncRNAs in this process (315). Nevertheless, it is often difficult to pick from the enormous number of potential candidate genes for subsequent functional characterization. Here, we used a bioinformatics approach to integrate such data set with other publicly available ones, applying a rational and systematic data mining method to define a manageable list of final candidates for follow-up experiments. As a result, we identified GRASLND (Glycosaminoglycan
Regulatory ASsociated Long Non-coDing RNA) to be a crucial regulator of chondrogenesis. We showed that GRASLND enhances chondrogenesis by acting to suppress the IFN- signaling pathway, and this effect was prevalent across different cell types and conditions. Together, these results highlight novel roles of both GRASLND and IFN in cartilage development and maintenance, as well as their potential therapeutic uses to combat OA.
4.2. Materials and Methods
4.2.1. Cell culture
Culture and expansion of bone marrow MSCs were described in section
2.2.1. 107
Adipose derived mesenchymal stem cells (ASCs) were purchased from
ATCC (SCRC-4000) and cultured in complete growth medium: Mesenchymal stem cells basal medium (ATCC PCS-500-030), mesenchymal stem cell growth kit (ATCC PCS-500-040) (2% FBS, 5 ng/mL basic recombinant human FGF, 5 ng/mL acidic recombinant human FGF, 5 ng/mL recombinant human EGF, 2.4 nM L-alanyl-L-glutamine), 0.2 mg/mL G418. ASCs were stored in liquid nitrogen in freeze medium: 80% FBS (Northwest), 10% complete growth medium, 10%
DMSO.
4.2.2. Plasmid construction shRNA
Short hairpin RNA (shRNA) sequences for specific genes of interest were designed with the Broad Institute GPP Web Portal. For each gene, six different sequences were selected for screening, after which the two most efficient were chosen for downstream experiments in chondrogenic assays. A complete list of efficient shRNA sequences was presented in Table S1. shRNA was cloned into a modified lentiviral vector (Addgene #12247) using MluI and ClaI restriction sites, as described previously (229).
Transgene overexpression of GRASLND
A derivative vector from modified TMPrtTA (198, 316) was created with
NEBuilder® HiFi DNA Assembly Master Mix (New England Biolabs). Backbone was digested with EcoRV-HI (New England Biolabs) and PspXI (New England
Biolabs). The following fragments were amplified and assembled into the cut
108 plasmid: Tetracycline responsive element and minimal CMV promoter
(TRE/CMV), Firefly luciferase, bGH poly(A) termination signal (BGHpA). Primers and plasmids for cloning was provided in Table S2.
The full sequence of GRASLND transcript variant 1 (RefSeq NR_033997.1) was synthesized by Integrated DNA Technologies, Inc. GRASLND or the dsRed fluorescent protein coding sequence were cloned into the above derivative tetracycline inducible plasmid with NEBuilder® HiFi DNA Assembly Master Mix
(New England Biolabs) at NheI and MluI restriction sites (pLVD-GRASLND and pLVD-dsRed). Amplifying primers were provided in Table S2.
CRISPR-dCas9 activation of GRASLND
Guide RNA sequences were designed using the UCSC genome browser
(http://genome.ucsc.edu/) (317), integrated with the MIT specificity score calculated by CRISPOR and the Doench efficiency score (318, 319). Sequences were cloned into the pLV-hUbC-dCas9-VP64 lentiviral transfer vector (Addgene
#53192) using BsmBI restriction sites (320). Eleven potential sequences were picked and screened for their efficiency, and the gRNA with the highest activation potential was chosen (Figure S5). The synthetic gRNA used in all CRISPR- dCas9 activation experiments has the following sequence: 5’-
CCACTGGGGATAGTTCCCTG-3’.
4.2.3. Lentivirus production
HEK 293T producer cells were maintained in 293T medium: DMEM-high glucose (Gibco), 10% heat inactivated FBS (Atlas), 1% Penicillin/streptomycin.
109
To produce lentivirus, HEK 293T cells were plated at 3.8 x 106 cells per 10 cm dish (Corning) or at 8.3 x 106 cells per 15 cm dish (Falcon) in 293T medium. The following day, cells were co-transfected by calcium phosphate precipitation with the appropriate transfer vector (20 µg for 10 cm dish; 60 µg for 15 cm dish), the second-generation packaging plasmid psPAX2 (Addgene #12260) (15 µg for 10 cm dish; 45 µg for 15 cm dish), and the envelope plasmid pMD2.g (Addgene
#12259) (6 µg for 10 cm dish; 18 µg for 15 cm dish). Cells were incubated at
37°C overnight. The following day, fresh medium consisting of DMEM-high glucose, 10% heat-inactivated FBS, 1% Penicillin/streptomycin, 4 mM caffeine was exchanged (12 mL for 10 cm dish; 36 mL for 15 cm dish). Lentivirus was harvested 24 hours post medium change (harvest 1), when fresh medium was exchanged again. 48 hours post medium change, harvest 2 was collected.
Harvest 1 and harvest 2 supernatant were pooled, filtered through 0.45 µm cellulose acetate filters (Corning), concentrated, aliquoted and stored at -80°C for future use.
To produce lentivirus for shRNA and gRNA screening, HEK 293T cells were plated at 1.5-2 x 106 cells per well in a 6-well plate in DMEM-high glucose
(Gibco), 10% heat inactivated FBS (Atlas). The following day, cells were co- transfected with 2 µg of the appropriate transfer vector, 1.5 µg of the packaging plasmid psPAX2 (Addgene #12260), and 0.6 µg of the envelope plasmid
(Addgene #12259) with Lipofectamine 2000 (ThermoFisher) following
110 manufacture’s protocol. Harvest and storage were performed as described above.
For knockdown experiments, lentivirus were titered by determining the number of antibiotic resistant colonies after puromycin treatment and used at
MOI 2. For overexpression experiments, lentivirus were titered by measuring integrated lentiviral copy number with qPCR and used at MOI ~ 0.1-0.2 for transgene overexpression and at MOI ~ 0.4-0.9 for CRISPR-dCas9 mediated overexpression.
4.2.4. Lentivirus transduction
One day before transduction, cells were plated at 4,500 cells/ cm2. Cells were transduced with appropriate lentivirus in expansion medium supplemented with 4
µg/mL polybrene (Sigma-Aldrich). Twenty-four hours post transduction, cells were rinsed once in PBS. Fresh expansion medium was exchanged. Cells were cultured with medium exchange every three days until future use.
4.2.5. Chondrogenesis assay
MSCs or ASCs were digested in 0.05% Trypsin-EDTA (Gibco), and trypsin was inactivated with 1.5X volume of expansion medium. Digested cells were centrifuged at 200 x g for 5 minutes, and supernatant was aspirated.
Subsequently, cells were washed in pre-warmed DMEM-high glucose (Gibco) three times, and resuspended at 5 x 105 cells/mL in complete chondrogenic medium: DMEM-high glucose (Gibco), 1% Penicillin/ streptomycin (Gibco), 1%
ITS+ (Corning), 100 nM Dexamethasone (Sigma-Aldrich), 50 µg/mL ascorbic
111 acid (Signma-Aldrich), 40 µg/mL L-proline (Sigma-Aldrich), 10 ng/mL rhTGF-β3
(R&D Systems). 500 µL of the above cell mixture was dispensed into 15 mL conical tubes, and centrifuged at 200 x g for 5 minutes. Formed pellets were cultured at 37°C, 5% CO2 for 21 days with medium exchange every three days.
4.2.6. Osteogenesis and adipogenesis assays
MSCs were plated at 2 x 104 cells/ well in 6-well plates (Corning) and cultured for 4 days in MSC expansion medium, followed by induction medium for 7 days.
Osteogenic induction medium includes: DMEM-high glucose (Gibco), 10% FBS,
1% Penicillin/ streptomycin (Gibco), 10 nM Dexamethasone (Sigma-Aldrich), 50
µg/mL ascorbic acid (Sigma-Aldrich), 40 µg/mL L-proline (Sigma-Aldrich), 10 mM
β-glycerol phosphate (Chem-Impex International), 100 ng/mL rh-BMP2
(ThermoFisher). Adipogenic induction medium includes: DMEM-high glucose
(Gibco), 10% FBS (ThermoFisher), 1% Penicillin/ streptomycin (Gibco), 1% ITS+
(Corning), 100 nM Dexamethasone (Sigma-Aldrich), 450 µM 3-isobutyl-1- methylxanthine (Sigma-Aldrich), 200 µM indomethacin (Sigma-Aldrich).
4.2.7. Biochemical assays
Harvested pellets were stored at -20°C until further processing. Collected samples were digested in 125 µg/mL papain at 60°C overnight. DMMB assay was performed as previously described to measure GAG production (232).
PicoGreen assay (ThermoFisher) was performed to measure DNA content.
112
4.2.8. Immunohistochemistry and histology
Harvested pellets were fixed in 4% paraformaldehyde for 48 hours, and processed for paraffin embedding. Samples were sectioned at 10 µm thickness, and subjected to either Safranin O – Fast Green standard staining (233) or to immunohistochemistry of collagen type II (Developmental Studies Hybridoma
Bank, University of Iowa; #II-II6B3). Human osteochondral sections were stained simultaneously to serve as positive control. Sections with no primary antibodies were used as negative control for immunohistochemistry.
4.2.9. RNA fluorescence in situ hybridization (RNA FISH)
Harvested pellets or human cartilage samples were snap frozen in Tissue-
Plus O.C.T. Compound (Fisher HealthCare) and stored at -80°C until further processing. Samples were sectioned at 5 µm thickness and slides were stored at
-80°C until staining. Probe sets for RNA FISH were conjugated with Quasar®
670 dye, and were synthesized by LGC Biosearch Technologies and listed in
Table S3 (GRASLND). GAPDH probe set was pre-designed by the manufacturer.
Staining was carried out according to manufacturer’s protocol for frozen tissues.
Slides were mounted with Prolong Gold antifade mountant with DAPI
(ThermoFisher) and imaged with the Virtual Slide Microscope VS120 (Olympus) at lower magnification and with the confocal microscope (Zeiss) at higher magnification.
113
4.2.10. RNA isolation and quantitative RT-PCR
Norgen Total RNA isolation kits were used for all RNA isolation (Norgen
Biotek). For monolayer, cells were lysed in buffer RL and stored at -20°C until further processing. For pellets, harvested samples were snap frozen in liquid nitrogen and stored at -80°C until further processing. On day of RNA isolation, pellets were homogenized in buffer RL using a bead beater (BioSpec Products) at 2,500 oscillations per minute for 20 seconds for a total of three times.
Subsequent steps were performed following manufacturer’s protocol.
4.2.11. Luminescence assay
MSCs were plated at 8.5 x 104 cells per well in 24-well plates (Corning).
Lentivirus carrying the response elements for type I (ISRE - #CLS-008L-1) or type II (GAS - #CLS-009L-1) upstream of firefly luciferase was purchased from
Qiagen. After 24 hours incubation, cells were co-transduced with virus in the following groups: ISRE with scrambled shRNA, ISRE with GRASLND shRNA,
GAS with scrambled shRNA, GAS with GRASLND shRNA. 24 hours post- transduction, cells were rinsed once in PBS and fresh medium was exchanged.
Three days later, medium was switched to expansion medium with 100 ng/mL
IFN-β (PeproTech) for wells with ISRE or with 5 ng/mL IFN- (PeproTech) for wells with GAS. MSCs were cultured for another 22 hours, and then harvested for luminescence assay using Bright-Glo Luciferase Assay System (Promega).
Luminescence signals were measured using the Cytation 5 Plate reader
(BioTek).
114
4.2.12. Western blot
On day of harvest, cells were homogenized with complete lysis buffer in ice cold phosphate buffered saline: 10X RIPA buffer (Cell Signaling Technology),
100X phosphatase inhibitor cocktail A (Santa Cruz Biotechnology), 100X HaltTM protease inhibitor cocktails (ThermoScientific). Lysates were subsequently centrifuged at 14,000 x g for 15 minutes at 4°C, and supernatants were collected and stored at -20°C until further processing. Western blot was serviced by
RayBiotech with the following antibodies: primary anti-β-actin (RayBiotech), primary anti-RNF144A (Abcam), and secondary anti-rabbit-HRP (horse radish peroxidase) (RayBiotech).
4.2.13. Cytotoxicity assay
Seven days post viral transduction, medium was collected and the amount of lactose dehydrogenase (LDH) was measured as indirect output for cellular toxicity. Assays were performed following manufacturer’s protocol (Promega).
Absorbance signal was recorded at 490 nm with the Cytation 5 instrument
(BioTek).
4.2.14. RNA-Seq and data analysis
Isolated RNAs were stored at -80°C and submitted to the Genome
Technology Access Center at Washington University in St Louis for library preparation and sequencing on a HiSeq 2500. Library preparation was carried out as previously described in section 3.B.6. Samples were sequenced with 2 x
101 parameter.
115
4.2.15. Statistical analyses
All statistical analyses were performed using R version 3.4.3 (2017-11-30)
4.3. Results
4.3.1. RNF144A-AS1 is a crucial and specific regulator of chondrogenesis
First, we utilized the published database on MSC chondrogenesis
(https://msc-browser.guilaklab.com) (GSE109503) (315) to identify long non- coding RNA candidates. We investigated the expression patterns of MSC markers (ALCAM, ENG, VCAM1), chondrogenic markers (ACAN, COL2A1,
COMP), and SOX transcription factors (SOX5, SOX6, SOX9) (Figure S1A).
Pearson correlation analysis revealed 245 lncRNAs whose expression was highly correlated to those of MSC markers, 207 lncRNAs to chondrogenic markers, and 34 lncRNAs to SOX transcription factors (Figure S1B, C). Among those, six were downregulated and one was upregulated upon SOX9 overexpression (Table 4) (GSE69110) (321). To validate their functions in chondrogenesis, we systematically designed small hairpin RNAs (shRNAs) targeting each candidate and assessed knockdown effect after 21 days of chondrogenic induction. Five out of six MSC-related lncRNAs did not affect glycosaminoglycan (GAG) production – an important extracellular matrix component in cartilage (Figure S2). While these lncRNAs may be regulating
MSC maintenance or homeostasis, their roles in chondrogenesis appeared to be minimal. Moreover, we found that lower levels of MSC-correlated lncRNAs did
116 not prime the mesenchymal stem cells towards better chondrogenesis. On the contrary, depletion of RNF144A-AS1 resulted in decreased expression of chondrogenic markers (COL2A1, ACAN), and upregulation of apoptotic (CASP3) and cellular senescence (TP53) markers. This effect was not due to nonspecific cytotoxicity of examined shRNAs, as released levels of lactase dehydrogenase
(LDH) were similar among the groups (Figure S3). On a protein level, biochemical assays indicated a reduction in GAG deposition, although DNA level was only slightly affected (Figure 19 C-E). We observed the same phenotypic loss of GAG and collagen type II in the extracellular matrices (ECM) of pellet samples with RNF144A-AS1 targeted shRNAs, while the scrambled controls displayed explicit staining of these proteins (Figure 19F). Taken together, this data indicates that RNF144A-AS1, while not required for cellular proliferation, plays a crucial role in cartilage-like matrix production.
To establish whether RNF144A-AS1 is expressed exclusively in chondrogenesis, we performed a time course experiment. We induced MSCs towards other potential lineages, and measured RNF144A-AS1 levels throughout these processes. Successful differentiation was established with an increase in lineage-specific markers: PPARG and ADIPOQ for adipogenesis, COL1A1 and
COL10A1 for osteogenesis, and ACAN, SOX9 and COL2A1 for chondrogenesis
(Figure 19 G-I). We found that RNF144A-AS1 expression was particularly enriched in chondrogenesis (Figure 19H). In contrast, RNF144A-AS1 peaked at earlier timepoints during adipogenesis but decreased at later time points (Figure
117
19G), and downregulated when MSCs underwent osteogenic induction (Figure
19I). Thus, it is evident that RNF144A-AS1 is specific to chondrogenesis, and its down-regulation suggested an inhibitive role in other processes.
This pattern was recapitulated with our RNA in situ hybridization experiment
(Figure 20). MSC pellets were shown to start accumulating RNF144A-AS1 signals at later time points in chondrogenesis. Interestingly, we observed that
RNF144A-AS1 was localized to the cellular cytoplasm in puncta, suggesting it may be a part of important RNA-protein complexes.
4.3.2. Characterization of RNF144A-AS1
We examined the characteristics of RNF144A-AS1 by first exploring its evolutionary conservation or divergence. Except for exon 1, the genomic region of RNF144A-AS1 is highly conserved in primates whose common ancestor traced back to 25 million years ago (322): Homo sapiens, Pan troglodytes, and
Rhesus macaque (Figure 21A). Sequences become less conserved in other mammalians: Mus musculus, Canis lupus familiaris, and Loxodonta africana, or in other non-mammalian vertebrates: Gallus gallus, Xenopus tropicalis, Danio rerio, and Petromyzon marinus. This suggests that RNF144A-AS1 was gained during the ape divergence, and thus may potentially had been selected under evolutionary pressure.
Per GENCODE categorization, the AS (antisense) suffix indicates a group of lncRNAs that are positioned on the opposite strand, with overlapping sequences to their juxtaposed protein-coding genes. Often, these lncRNAs play a role in
118 regulating the expression of their protein-coding counterparts (reviewed in (299)).
Therefore, we set out to examine whether this is also characteristic of RNF144A-
AS1, in the absence or presence of TGF-β3, a potent growth factor in chondrogenesis (Figure 21B-C). Neither knockdown nor overexpression of
RNF144A-AS1 affected RNF144A transcript levels in both tested doses of TGF-
β3. Moreover, RNF144A protein translation also remained unaffected with variations of RNF144A-AS1 levels, as indicated by western blot (Figure 21C).
These results indicate that RNF144A-AS1 displayed a distinct function and is not involved in the regulation of RNF144A. For these reasons, we proposed an alternative name for RNF144A-AS1: GRASLND - Glycosaminoglycan Regulatory
ASsociated Long Non-coDing RNA.
Next, we explored the signaling axis of GRASLND. Data mining and computational analysis on earlier published data suggested that GRASLND was a downstream effector of SOX9 (GSE69110) (321). When SOX9 was overexpressed in fibroblasts, GRASLND expression was increased. We further confirmed this by utilizing SOX9 transgene overexpression in our MSCs culture
(Figure 21E). Interestingly, while TGF-β3 has been demonstrated to act upstream of SOX9, exogenous addition of this growth factor alone did not result in enhanced GRASLND expression. It is notable that SOX9 levels in GFP controls were indistinguishable between TGF-β3 conditions at the time of investigation (1 week in monolayer culture), consistent with our previous finding that SOX9 was not upregulated until later timepoints in MSC chondrogenesis
119
(msc-browser.guilaklab.com). Therefore, TGF-β3, despite being a potent growth factor, is not sufficient to elevate GRASLND level. It is likely that TGF-β3 acts indirectly through SOX9 in order to turn on GRASLND during chondrogenesis.
4.3.3. Enhanced chondrogenesis for cartilage tissue engineering with GRASLND
As depletion of GRASLND resulted in reduced GAG and collagen deposition, we sought to investigate whether overexpression of GRASLND would enhance chondrogenesis. We assessed this question by both transgene ectopic expression and by CRISPR-dCas9 (Clustered regularly interspaced short palindromic repeats – catalytically dead Cas9) mediated in-locus activation.
We designed our lentiviral transfer vector to carry a BGH-pA (Bovine Growth
Hormone Polyadenylation) termination signal downstream of GRASLND to allow for its correct processing (Figure S4A). Later, we found that the poly-A signal was not necessary in this case since a different backbone design could recapitulate similar trends in our results (data not shown). Additionally, GRASLND was also driven under a doxycycline inducible promoter, enabling the temporal control of its expression. We utilized this feature to induce GRASLND only during chondrogenic culture (Figure 22A). This experimental design focused solely on the role of GRASLND during chondrogenesis, while successfully eliminating its effect in MSC maintenance and expansion from our analysis. Since doxycycline was most potent at 1 µg/mL (Figure S4 B-C), this dose was used for all following experiments.
120
To determine whether GRASLND would improve chondrogenesis at lower doses of growth factor or at earlier time points, we compared DNA and GAG levels from pellets cultured under different TGF-β3 concentrations on day 7, day
14, and day 21 (Figure S4 D-F). In agreement with our knockdown data, DNA contents were mostly unaffected. On the other hand, increases in GAG were observed at higher doses and at later time points, especially at 10 ng/mL of TGF-
β3. It appears that an elevated level of GRASLND alone was not sufficient for
GAG deposition. Quite to the contrary, GRASLND may act in concert with other downstream effectors, which were not present at lower doses of TGF-β3 or at earlier time points in the process.
Elevated levels of GRASLND resulted in higher amounts of GAG protein deposition (Figure 22B), and this phenomenon was recapitulated on a transcriptomic level (Figure 22C). We observed a slight increase in chondrogenic markers (COL2A1, ACAN), and a slight decrease in the apoptotic marker
CAPS3, while cellular senescence was not different between the two groups
(TP53) (Figure 22D). Histologically, pellets made from dsRed-transduced MSCs exhibited normal GAG and collagen type II staining, indicating successful chondrogenesis. The control pellets were non-distinguishable from those made with GRASLND-transduced MSCs (Figure 22D), albeit macroscopically smaller at the time of harvest (data not shown).
These findings were further confirmed using CRISPR-dCas9-VP64 mediated activation of endogenous GRASLND. This system had been previously utilized to
121 up-regulate neuronal transcription factors, successfully inducing embryonic fibroblasts into neurons (323). Among eleven tested synthetic gRNAs, one had the highest effectivity (Figure S5). When GRASLAND was transcriptionally activated with CRISPR-dCas9, chondrogenesis was enhanced as evidenced by elevated amount of GAG deposition, while DNA amount was again mostly unaffected (Figure 22E). Similar trends were detected by qRT-PCR (Figure 22F) and histology (Figure 22G). It is worth noting, that relative to transgene ectopic expression, CRISPR-dCas9 mediated activation only resulted in a moderate up- regulation of GRASLND (2 folds versus 100 folds). However, the functional outcome was more pronounced with CRISPR-dCas9. We observed a 50% increase in the level of GAG produced when normalized to DNA (9.4 ± 2.19 mg/mg vs 16.3 ± 2.08 mg/mg), compared to 30% detected with ectopic expression (10.5 ± 0.84 mg/mg vs 13.9 ± 0.52 mg/mg).
4.3.4. GRASLND inhibits type II interferon signaling
To elucidate the signaling pathways of GRASLND, we conducted an RNA-
Seq experiment with subsequent gene ontology analysis. As expected,
GRASLND depletion resulted in impaired expression of chondrogenic genes, with COL2A1 and COL9A1 among the most significantly downregulated genes
(ranked by p-values) (Figure 23A). Skeletal system development and extracellular matrix organization were among the pathways most affected by the knockdown (Figure 23B). Surprisingly, pathways pertaining to viral infection (viral transcription, translational initiation) were highly enriched in the upregulated gene
122 list. Specifically, interferon (IFN)-related signaling was highly elevated upon silencing of GRASLND. The top 20 upregulated genes involved many IFN downstream targets (IFI6, IFIT1, IFIH1, MX1, OAS1, ISG15, IFIT3, MX2, IFI27,
IFIT2, OASL), with both type I (IFN-α, IFN-β) and type II (IFN-) found to be enriched in our gene ontology analysis (Figure 23B). Taken together, GRASLND may potentially act through the interferon signaling pathway during chondrogenesis.
To further confirm this relationship, we performed luciferase reporter assays for interferon signaling upon GRASLND knockdown. Utilizing specific reporter constructs, we were able to determine whether GRASLND acted in type I or type
II IFN. Our results indicated that decreased level of GRASLND led to heightened type II (IFN-) (Figure 23D) response but not type I (IFN-β) (Figure 23C).
Importantly, luminescence activities between scrambled control and GRASLND knockdown with no exogeneous IFN were indistinguishable from each other. This indicates that at basal level, the two groups responded similarly to lentiviral transduction, and the observed difference in IFN signal was a consequence of
GRASLND down-regulation.
Interestingly, by mining a published microarray database (GSE57218) (324), we found that IFN-related genes were highly elevated in cartilage tissues of osteoarthritis patients (Figure S6A). However, the effect was only potent in older subjects (> 75 years old), with young subjects displaying no difference in their
IFN expression levels. Since the microarray did not contain probes for 123
GRASLND, no information on its expression could be extracted. However, we found an independent study reported changes in the transcriptomes of intact and damaged cartilage tissues by RNA-Seq (E-MTAB-4304) (325). Analysis of this data set suggested a negative correlation between GRASLND and IFN-related genes (Figure S6C). Recapitulating earlier microarray data, we observed an upregulation of IFN signals in patients 70 years or older. Interestingly, GRASLND expression seemed to be downregulated at damaged cartilage sites compared to intact ones (Figure S6B). Therefore, we proposed that GRASLND may possess some therapeutic potential through suppressing IFN signaling in osteoarthritis. To evaluate the possibility, we implemented the use of the GRASLND transgene in engineered cartilage cultured under IFN addition (100 ng/mL of IFN-β or 5 ng/mL of IFN-). Consistent with the luciferase reporter assay, GRASLND protective effect was encouragingly observed upon IFN- assault but not IFN-β (Figure 23E,
F). However, we observed a reduced level of GAG production compared to normal condition, suggesting that although GRASLND can protect the ECM from degradation, its effect was not potent enough to restore engineered cartilage to a state of no IFN assault.
Since GRASLND was found in the cytoplasm, we hypothesized that it is part of an RNA-protein complex. To test this theory, we performed an RNA pull down assay, followed by mass spectrometry. Resulting proteins that emerged as potential binding partners to GRASLND included: Gamma-Interferon Inducible
Protein 16 (IFI16), Interferon-Stimulated 20kDa Exonuclease-Like 2 (ISG20L2), 124
Interferon-Induced Double-Stranded RNA-Activated Protein Kinase (EIF2AK2), and Interferon-Inducible Double-Stranded RNA-Dependent Protein Kinase
Activator A (PRKRA). Subsequent RNA pull down followed by western blot confirmed EIF2AK2 was a binding partner of GRASLND (Figure S7). Upon activation of IFN, EIF2AK2 phosphorylates translation initiation factor EIF2S1, which as a result inhibits protein synthesis. Here, we proposed a working model where GRASLND binds to EIF2AK2, successfully sequestering it and preventing the phosphorylation of EIF2S1.
4.3.5. GRASLND enhanced the chondrogenesis of adipose-derived stem cells
In the field of cartilage tissue engineering, adipose-derived stem cells (ASCs) can be used as an alternative cell source to MSCs. Nevertheless, it remains challenging to induce ASCs into a more chondrogenic fate, possibly due to their inherent potential to advance towards adipogenesis. Here, we addressed whether modulating GRASLND expression could also serve to improve ASC chondrogenesis, consistent with the effects we observed on MSCs, as a means of enhancing the use of ASCs for cartilage engineering. We observed an increase in GAG production when GRASLND was overexpressed in ASCs compared to control (2.8 ± 0.22 mg/mg vs 1.7 ± 0.14 mg/mg) (Figure 24A). More importantly, COL2A1 expression was highly elevated (~ 5 fold) with increased
GRASLND level (Figure 24B). There was also a small increase in ACAN level, although not statistically significant. When examining these cartilage constructs
125 histologically, we observed a more slightly positive staining of collagen type II in pellets with GRASLND overexpression compared to the dsRed controls (Figure
24C). Based on these data, it appears that GRASLND utilized the same mechanism across these two cell types, asserting a pan effect on potentiating their chondrogenic capabilities.
4.4. Discussion
Several previous studies seeking to elucidate lncRNA functional conservation during evolution had posited that they are rapidly evolving, with prominently diverse repertories and expression across species (80). As protein-coding genes are constrained within their highly conserved functions, lncRNAs may have evolved as a means to cope with environmental adaptations and regulatory plasticity (326). In fact, evolutionarily young lncRNAs may partially explain our organismal complexity compared to other distant species (327), given how
Caenorhabditis elegans displays a similar number of protein-coding genes as
Homo sapiens (328). Human lncRNA transcripts were estimated to be about 80% orthologous in chimpanzee, but only 38% in mouse. Species diverged further along the evolutionary tree contained even less (~ 3-19% orthologous transcripts if diverged 90 million to 300 million years ago) (80, 326). Most importantly, while lncRNAs can be diverse in their sequence or expression levels, they are extremely conserved in their tissue specificities (80, 326). Since GRASLND is a primate-specific lncRNA, it might have been placed under positive selection during evolution and expressed in specific tissues (within them the cartilage) 126 similar to other primate-specific transcripts. Unfortunately, lack of a mouse ortholog renders in vivo studies in this model organism unattainable. To tackle its functionality in vivo, we may have to resort to transgenic models of other primates, such as the Rhesus monkey or chimpanzees.
We found that GRASLND acts to enhance chondrogenesis through negative regulation of the interferon pathway. However, it remains unclear whether
GRASLND functions exclusively in MSC differentiation, or whether it also plays a role in host defense against viral infection. Several lncRNAs had been previously identified bearing their links to interferon regulation in this aspect (review in
(329)). For example, NeST suppresses IFN- by recruiting WDR5 to the locus and altering its chromatin state (330). lncRNA-CMPK2 resulted in heightened
IFN-α signal and ultimately a reduction in HCV replication (331). The ability of these lncRNAs to regulate viral-host dynamics poses them as useful clinical targets for intervention upon viral infection. Although not within the scope of our investigation, the probable function of GRASLND in microbial response will be of great interest for future studies.
Since lentivirus was employed to manipulate the expression of GRASLND, it is possible our observations were confounded by the cellular response to viral infection. However, luciferase reporter assays suggested that this was not the case, since basal luminescence levels (no interferon supplementation) between the scrambled controls and the KD were indistinguishable, suggesting that the detected elevation of interferon signaling (Figure 23C-D) was likely attributed to 127 varying levels of GRASLND in examined groups. Additionally, our evidence indicates that GRASLND acts through type II rather than type I IFN. Even when
IFN-β required a higher dose to evoke responses in MSCs, we found that 5 ng/mL of IFN- was still more detrimental to chondrogenic constructs compared to 100 ng/mL of IFN-β. We speculated that a skewed distribution of available surface receptors between type I and type II (IFNAR vs IFNGR) may explain this phenomenon. In fact, MSCs expressed a much lower level of IFNAR2 compared to IFNAR1, IFNGR1, or IFNGR2 (315) (msc-browser.guilaklab.com). As these receptors work in dimers (308, 309), response to type I may be stunted due to
IFNAR2 deficiency.
Originally discovered by its ability to induce cellular resistance upon viral infection, interferon has since then been proven to play many roles in other cellular processes. Probably most prominent is its participation in hematopoiesis, as the presence of IFN-β can induce terminal differentiation of several hematopoietic cell lines (307). Moreover, interferons are also known for their inhibitory effects on cellular proliferation and growth, and thus are sometimes referred to as tumor suppressors (332, 333). Additionally, during the early stage of embryonic development when massive cell proliferation occurs, interferon is suppressed, and the embryonic cells are IFN-resistant (334). In agreement with previous research, we showed that as MSCs progressed towards a more differentiated fate during chondrogenesis, interferon downstream signaling effectors were also repressed by GRASLND. 128
In the context of the musculoskeletal system, IFN is mostly recognized for its connection in bone development and homeostasis (311-314). IFN is also involved in an orchestration with TGF-β in wound healing and myogenesis (335-
337). However, the role of IFN in chondrogenesis remained under-appreciated.
On the other hand, IFN had been implicated in arthritis by several in vivo studies
(338, 339). Our curation of publicly available databases suggested similar patterns of IFN in degenerated cartilage. As GRASLND inhibits IFN, utilization of the lncRNA poses high potentials in both MSC cartilage tissue engineering and in
OA treatment. As a proof of concept, we showed that GRASLND could enhance matrix deposition across cell types of origin, with and without interferon attack in vitro. It would be interesting to next investigate whether GRASLND can protect ex vivo or in vivo cartilages from degradation in a milieu of pro-inflammatory cytokines.
Interestingly in the case of ASCs, we observed not only an increase in GAG and collagen type II matrix deposition, we also showed an increase in COL2A1 expression. Detection of relatively high fold change in the samples with
GRASLND overexpression compared to the controls may be attributable to low level of COL2A1 in ASCs. Alternatively, it has been proposed that SOX9, when expressed at high level, could bind to multiple active enhancer elements to turn on cartilage-related gene targets (321), whereas this transcription factor was involved in other basal cell activities at lower level. We hypothesized that
129
GRASLND may be raising SOX9 expression over the threshold and as a result engaging ASCs in a chondrogenic program.
Finally, we demonstrated novel success in utilizing the modified CRISPR- dCas9 system for endogenous transcriptional activation of GRASLND. This system had been previously used in other cell types to turn on protein-coding genes, mostly transcription factors. Here, we potentiated this technology to study long non-coding RNAs. We showed that CRISPR may be more effective than transgene expression, as indicated by a larger increase in GAG production.
While evidence suggested that GRASLND acts in trans, we speculate the
CRISPR-dCas9 system could be useful to investigate those acting in cis, as well as lncRNAs that are difficult in molecular cloning for gain of function studies.
4.5. Conclusions
Here, we have identified GRASLND as an important regulator of chondrogenesis. GRASLND acts downstream of SOX9 and enhances cartilage- like matrix deposition in stem cells-derived constructs. Moreover, GRASLND functions to suppress IFN, and as a result induces adult stem cells towards a more chondrocytic linage. We propose that GRASLND can potentially be applied as a therapeutic use for both cartilage tissue engineering and for the treatment of
OA.
130
Figure 19: RNF144A-AS1 is important and specific to chondrogenesis in MSCs. (A) Expression pattern of RNF144A-AS1 in chondrogenesis (data retrieved from: GSE109503). (B) Effect of RNF144A-AS1 knockdown on chondrogenic, apoptotic, and cell cycle inhibition markers. n = 5. Statistical analysis was performed on log transformed fold changes. One-way ANOVA followed by Tukey post-hoc test (α =0.05). Groups of different letters are statistically different from one another. (C-E) Effect of RNF144A-AS1 knockdown on pellet matrix synthesis. (C) GAG (D) DNA (E) GAG as normalized to DNA. n = 5.
131
One-way ANOVA followed by Tukey post-hoc test (α =0.05). Groups of different letters are statistically different from one another. (F) Representative histological images of MSC pellets cultured in chondrogenic condition for 21 days. Reduced level of GAG and collagen type II was observed in RNF144A- AS1 KD samples compared to the scrambled controls. Scale bar = 200 µm. SafO-FG: SafraninO-Fast Green staining. COLII IHC: collagen type II immunohistochemistry. hOC: human osteochondral control. (G-I) qRT-PCR analysis of MSC samples cultured in (G) adipogenic condition (n=6), (H) osteogenic condition (n=6), (I) chondrogenic condition (n=3-4). Lineage- specific markers were upregulated throughout the time course of analysis. RNF144A-AS1 level: (G) peaked at earlier time points and decreased afterwards in adipogenesis, (H) continually decreased in osteogenesis, (I) continually increased in chondrogenesis.
132
Figure 20: RNF144A-AS1 expression is increased with chondrogenesis. (A) RNA in situ hybridization of MSC-derived pellets at different time points during chondrogenesis. GAPDH was used as a control. GAPDH and RNF144A-AS1 probes were hybridized on separate slides. RNF144A-AS1 was observed at later time points in chondrogenesis. Scale bar = 20 µm. (B) Confocal microscopy on MSC-derived pellets. RNF144A-AS1 was observed in puncta in the cellular cytoplasm. Scale bar = 5 µm.
133
Figure 21: RNF144-AS1 (GRASLND) relationship to RNF144A and SOX9. (A) RNF144A-AS1 is located on chromosome 2 of the human genome, overlapping RNF144A in the opposite direction. RNF144A-AS1 is conserved in primates but not in mammalians. (B-C) RNF144A-AS1 did not regulate the expression of RNF144A, as evidenced by (B) knockdown and (C) overexpression of the lncRNA. RNF144A level did not change under all conditions. n=4. (D) RNF144A-AS1 did not regulate the translation of RNF144A, as semi-quantified by western blot. (E) RNF144A-AS1 is downstream of SOX9. RNF144A-AS1 level was measured in GFP- or SOX9- transduced MSCs under different doses of TGF-β3. n=6. Statistical analysis was performed on log-transformed fold changes. Welch’s t-test between 134 samples for each qRT-PCR gene target. Groups of different letters are statistically different.
135
Figure 22: GRASLND enhances chondrogenesis. (A) Experimental timeline. (B, E) Biochemical analyses of MSC-derived pellets at day 21. (C, F) qRT-PCR analyses of MSC-derived pellets at day 21. (D, G) Representative histological images of MSC-derived pellets at day 21. Scale bar = 200 µm. SafO-FG: SafraninO-Fast Green staining. COLII IHC: collagen type II immunohistochemistry. hOC: human osteochondral control. (B-D) Transgene
136 ectopic expression of GRASLND. (E-G) CRISPR-dCas9-VP64 induced activated of GRASLND.
137
Figure 23: GRASLND acts to suppress interferon type II signaling. (A) Top 20 up- and down-regulated genes in GRASLND KD pellets compared to scrambled controls. (B) Gene ontology analysis of affect pathways when GRASLND level is reduced. (C) Luciferase reporter assays on MSCs transduced with (C) ISRE promoter element, specific for interferon type I response. MSCs were cultured in either 0 ng/mL or 10 ng/mL of IFN-β. GRASLND KD did not alter IFN type I response compared to scrambled control. n=3. (D) uciferase reporter assays on MSCs transduced with GAS promoter element, specific for interferon type II response. MSCs were cultured in either 0 ng/mL or 5 ng/mL of IFN-. GRASLND KD increased IFN type II response compared to scrambled control. n=3. Two-way ANOVA followed by Tukey post-hoc test (α=0.05). Groups of different letters are statistically different. (E) Biochemical assays on MSC-derived pellets cultured under 100 ng/mL of IFN-β. n=4 (F) Biochemical assays on MSC-derived pellets cultured 138 under 5 ng/mL of IFN-. n=6. Welch’s t-test. Groups of different letters are statistically different.
Figure 24: GRASLND enhances chondrogenesis in adipose-derived stem cells. (A) Biochemical analyses. n= 5 (B) qRT-PCR analyses. n=6 (C) Representative histological images. Scale bar = 200 µm. ASC-derived pellets were cultured under chondrogenic condition for 21 days. Increase in GAG, collagen type II in the matrix were observed in the overexpression compared to the control. COL2A1 expression was also upregulated when GRASLND level was elevated with transgene ectopic expression. Welch’s t-test. For qRT-PCR, statistical analysis was performed on log-transformed fold change. Groups of different letters are statistically different.
139
Table 4: Long non-coding RNA candidates shortlist Gene symbol RefSeq Official name Relationship to MSC Relationship to SOX9 Reference chondrogenesis EMX2OS NR_002791 EMX2 opposite Correlated with MSC Downregulated upon SOX9 strand/antisense RNA markers’ expression overexpression LOXL1-AS1 NR_040066 LOXL1 antisense RNA 1 Correlated with MSC Downregulated upon SOX9 markers’ expression overexpression LINC01279 NR_109779 Long Intergenic Non- Correlated with MSC Downregulated upon SOX9 Protein Coding RNA markers’ expression overexpression 1279 MIR4435-2HG Not MIR4435-2 host gene Correlated with MSC Downregulated upon SOX9 available markers’ expression overexpression RP11-366L20.2 NR_120478 uncharacterized Correlated with MSC Upregulated upon SOX9 LOC100129940 markers’ expression overexpression
140 RNF144A-AS1 NR_033997 RNF144A antisense Correlated with Upregulated upon SOX9
RNA 1 chondrogenic overexpression markers’ expression RP11-79H23.3 Not Not available Correlated with MSC Upregulated upon SOX9 available markers’ expression overexpression
5. Conclusions
In this work, we showed how multidisciplinary approaches can be used to better understand the engineering of artificial cartilage. From a tissue engineering angle, we presented a proof of principle for simultaneous production of mineralized and cartilaginous extracellular matrices (ECMs). This method can potentially present new avenues to produce osteochondral constructs that both bypass prolonged, separated cultured times, and enable in situ site-directed multi-tissue formation. From a bioinformatics angle, the transcriptomics of in vitro chondrogenesis of mesenchymal stem cells were profiled for the first time. With a compilation of multiple donors and multiple time points, we were able to integrate various bioinformatics tools for detailed description of the dynamics and interplays between different pathways as MSCs progressed through this trajectory. From a molecular and cellular biology angle, we functionally validated candidates identified in silico, and as a result discovered novel roles of
GRASLND and IFN in MSC chondrogenesis. Altogether, this work highlighted the power of interdisciplinary studies. More importantly, this approach led us to recognize novel long non-coding RNA player with no previous known functions.
In Chapter 2, we presented evidence showing that manipulation of the transforming growth factor-β3 (TGF-β3) pathway leads to simultaneous formation of both mineralized and cartilaginous matrices. By utilizing lentiviral transduction, we genetically engineered MSCs to (1) become insensitive to TGF-β3 stimulation
141 via SMAD3 knockdown and (2) overexpress osteo-lineage transcription factor
RUNX2. A compilation of both (1) and (2) resulted in synergistic production of mineralized matrix in MSCs that would produce glycosaminoglycan ECM otherwise. Cultured together, genetically engineered MSCs started depositing minerals while unmodified MSCs produced GAG. Thus, manipulation of the TGF-
β3 pathway is a promising strategy for tissue engineering of an osteochondral construct deriving from one cell source and one culture condition.
In Chapter 3, we strived to better understand the molecular pathways governing MSC differentiation towards a more chondrocyte-like fate. By transcriptomic profiling of this process, we showed that in vitro chondrogenesis started with proliferation of MSCs, followed by ECM anabolic production.
Weighted gene correlation network analysis (WGCNA) identified a chondrogenic functional module, composed of 1,172 genes whose expression patterns are correlated with GAG deposition. From these 1,172 genes, we identified 230 long non-coding RNAs (lncRNAs) and 498 associated micro-RNAs (miRNAs) whose functional validations would further the knowledge of MSC biology.
In Chapter 4, we utilized the above dataset together with publicly available databases and pinpointed seven lncRNA candidates for further validation and characterization. Knockdown of MSC-related lncRNAs showed no difference in
MSC chondrogenesis, while knockdown of GRASLND, a chondrogenic-related marker, displayed impaired GAG production. On the other hand, enhancement of
GRASLND level, either by transgene ectopic expression or by Clustered 142 regularly interspaced short palindromic repeat (CRISPR) – catalytically dead
CRISPR Associated Protein 9 (dCas9) mediated in locus activation, resulted in higher levels of GAG deposition in the ECM. Furthermore, we showed that
GRASLND acts downstream of SOX9 and suppresses the IFN type II signaling pathway. The discovery of GRASLND and its interaction with IFN is advantageous in combatting against OA. Previous publications suggested that
IFN is upregulated in OA, and thus GRASLND as an inhibitory agent against IFN can be a potential therapeutic utilization. More importantly, the effect of
GRASLND in chondrogenesis seemed to be potent in different cell types: we showed that both MSCs and ASCs responded similarly to GRASLND upregulation. Taken together, these data suggest that GRASLND is an important regulator of chondrogenesis, adding to the growing literature of long non-coding
RNAs as well as presenting novel roles of both GRASLND and IFN in the chondrogenic process.
Clearly, at the wake of big data and readily-available high-throughput sequencing services, it is important to increase the level of integration between bioinformatics and other fields. Here, while we presented many potential lncRNA candidates (230 from the chondrogenic functional module, or 7 from cross- referencing with public data), implementation of molecular and cellular biology was crucial to follow through with functional validation and characterization. This interdisciplinary approach has come to fruition in multiple fronts in the current work. As we derived a method to simultaneously direct bone-like and cartilage-
143 like ECMs in a single culture system, we realized the limitation of supra- physiological doses of growth factor. To understand this process better and to avoid the use of growth factor at such high concentrations in the future, we realized in-depth detailing of MSC transcriptomics was necessary. In doing so, we described a chondrogenic functional module that composed of 1,172 genes, a small subset to investigate compared to the current annotation of around 60,000 genes. Finally, further bioinformatics analyses narrowed our list of potential candidates to seven, whose functional validation led us to GRASLND. GRASLND is an important lncRNA in chondrogenesis, acting to enhance chondrogenesis by suppressing IFN. Employing tissue engineering, cellular and molecular biology, and bioinformatics in this new age of technology promises to bring forth great advances in the field of regenerative medicine for functional and biologic artificial cartilage replacements.
144
Appendices Supplementary Data for “Genetic Engineering of Mesenchymal Stem Cells for Differential Matrix Deposition on 3D Woven Scaffolds”
Figure S1: (A,B) Quantification of alcian blue dye in wells with 0 ng/mL (purple dots) or 5 ng/mL (red squares) of TGFβ3. (C, D) Quantification of alizarin red dye in wells with 0 ng/mL (purple dots) or 5 ng/mL (red squares) of TGFβ3. Experiments were performed on individual donors with Donor 1 (A, C) and Donor 2 (B, D). n=3. Points represent independent specimen. Bars represent
145 geometric means for each group. Two-way ANOVA with Tukey post-hoc test. Groups of different letters are statistically different from one another.
Figure S2: Manipulation of SMAD3 and RUNX2 level by lentiviral delivery. (A) qRT-PCR shows efficiency of target shRNA. Student’s t test on log transformed values of fold changes. n=6 (p = 0.002) (B) qRT-PCR shows upregulation of RUNX2 with lentiviral construct. Student’s t test on log transformed values of fold changes. n=6 (p < 0.0001). Groups of different letters are statistically different from one another. (C) Representative raw image of transduced MSCs under bright-field and (D) red-fluorescent channel. Scale bar = 400 µm.
146
Figure S3: Non-transduced and virally transduced scaffolds displayed low expression of collagen type I. Immunohistochemistry staining of scaffold pairs cultured in 5 ng/mL TGFβ3. (A) NT and scrambled shRNA; (B) NT and SMAD3 shRNA; (C) NT and scrambled shRNA with RUNX2; (D) NT and SMAD3 shRNA with RUNX2. (E) Human osteochondral control. (F) Representative negative control without primary antibody. Square brackets indicate scaffold pairs cultured in the same well. Type I collagen was not observed in either non- transduced or virally transduced scaffolds. Scale bar = 100 µm.
147
Figure S4: SMAD3 shRNA with RUNX2 transduced scaffolds displayed slightly higher level of type X collagen compared to the non-transduced counterparts. Immunohistochemistry staining of scaffold pairs cultured in 5 ng/mL TGFβ3. (A) NT and scrambled shRNA; (B) NT and SMAD3 shRNA; (C) NT and scrambled shRNA with RUNX2; (D) NT and SMAD3 shRNA with RUNX2. (E) Human osteochondral control. (F) Representative negative control without primary antibody. Square brackets indicate scaffold pairs cultured in the same well. Type X collagen was not observed in non-transduced or virally transduced scaffolds. Scale bar = 100 µm.
148
Table S1: List of qRT-PCR primers Primer name Primer sequence r18S forward primer 5’-CGGCTACCACATCCAAGGAA-3’ r18S reverse primer 5’-GGGCCTCGAAAGAGTCCTGT-3’ RUNX2 forward primer 5’-TCCCTGAACTCTGCACCAAG-3’ RUNX2 reverse primer 5’-ATCTGGCTCAGGTAGGAGGG-3’ SMAD3 forward primer 5’-GCAGAACGTCAACACCAAGTGC-3’ SMAD3 reverse primer 5’-GTGCAGGTCTGGCCATCGC-3’
Table S2: Viral backbone components
5’ LTR 5’ long terminal repeat ψ Psi packaging signal EF1α Elongation factor 1 alpha promoter cPPT Central polypurine tract dsRedExpress2 Red fluorescent protein coding sequence from Discosoma sp. IRES Internal ribosomal entry site PuroR Puromycin resistance gene WPRE Woodchuck hepatitis virus post-transcriptional regulatory element U3PPT U3 polypurine tract tetO Tetracycline operator sequence H1 H1 promoter shRNA shRNA sequence SIN Self inactivating sequence 3’ LTR 3’ long terminal repeat
149
Supplementary Data for “High-Depth Transcriptomic Profiling reveals the Temporal Gene Signature of Mesenchymal Stem Cells During Chondrogenesis”
Figure S1: (A) Cluster dendrogram of genes into modules with arbitrary colors. (B, C) Modules whose eigengenes are highly correlated with changes in (B) GAG content and (C) DNA content and their corresponding Pearson correlation values.
150
Figure S2: Modules and their relationships to COL2A1 and ACAN expression patterns. (A) Dendrogram of RNA-Seq samples and corresponding changes in traits. Lowest values were depicted in white; highest values were depicted in red; missing values were depicted in grey. (B) Eigengene adjacency heatmap with blocks of meta-modules. (C) Modules highly correlated with changes in COL2A1 and ACAN expression and their corresponding Pearson correlation values.
151
Figure S3: Characteristics of four second most highly correlated modules. First row: module mediaumpurple2. Second row: module cyan. Third row: module sky blue. Forth row: module dark violet. Fifth row: module sienna3. (A, D, H, L, O) Co-expression network of the 20 top hubs. (E, I) Co-expression network of involved transcription factors and their top 3 connected genes. Module mediumpurple2, sienna3 did not contain any transcription factors as curated by JASPAR, and module dark violet only contains one. (B, F, J, M, P) Eigengene vector of corresponding module. (C, G, K, N, Q) Heatmap of corresponding modular gene expression.
152
Figure S4: Browser Interface. Left panel is a drop down menu with gene name as options. Selection of gene name returns a graph of gene expression (right upper panel), and a heatmap of transcript variants over the time course of analysis (right lower panel).
153
Supplementary documents: Submitted as electronic excel files
Table S1 – Samples QC values and expression of chondrogenic and osteogenic markers as measured by RPKM values.
Table S2 – List of differentially expressed genes (MSC-derived pellets versus human cartilage samples) by DAVID gene ontology terms.
Table S3 – List of differentially expressed genes by pair of days.
Table S4 – Gene name by modules
Table S5 – lncRNA and miRNA candidates in module steel blue
154
Supplementary Data for “The Long Non-Coding RNA GRASLND Enhances Chondrogenesis Via Suppression of Interferon Type II”
Figure S1: Identification of lncRNA candidates. LncRNAs whose expression patterns are correlated to crucial markers are of interest. (A) Expression patterns of previously identified MSC markers (left), chondrogenic markers (middle), and SOX transcription factors (right). Data retrieved from: GSE109503. (B) Expression patterns of correlated genes or lncRNAs (Pearson correlation > 0.9) to MSC markers (left), chondrogenic markers (middle), and SOX transcription factors (right). Lower boundary: 25th percentile; upper
155 boundary: 75the percentile; blue line: median of correlated gene set. (C) Number of correlated genes and correlated lncRNAs.
Figure S2: Functional validation of identified lncRNA candidates. Left: Expression pattern of candidates during chondrogenesis (data retrieved from GSE109503). Middle: Efficiency of designed target shRNAs. Fold change was
156 calculated in relative to scrambled controls. Individual graphs indicate experiments on target number 1 and target number 2 were performed separately. Right: Quantitative analysis of synthesized GAG matrix normalized to DNA amount. (A-C) EMX2OS, LOXL1-AS1, LINC01279: no change in GAG/DNA level compared to the scrambled control. (D) RP11-366L20.2: reduced level of GAG/DNA compared to the scrambled control. (E) MIR4435- 1HG: reduced level of GAG/DNA in target #1 knockdown but not target #2 knockdown samples.
Figure S3: Cytotoxicity assay of target shRNAs. Cytotoxicity level was measured indirectly via the amount of LDH in the medium. No differences between groups were observed. n = 4. One-way ANOVA followed by Tukey post-hoc test (α =0.05). Groups of different letters are statistically different from one another.
157
Figure S4: Effect of RNF144A-AS1 overexpression across time points and tested doses. (A) Overview of designed lentiviral backbone. RNF144A-AS1 is driven under a Doxycycline inducible promoter, and poly-adenylated with an artificial BGH-pA signal. (B) Doxycycline (Dox) titration. Dox is most potent at 1 μg/mL under both conditions of TGF-β3. (C-E) Biochemical analyses of MSC pellets cultured under chondrogenic condition with different doses of TGF-β3. n
158
= 3-4. Welch’s t-test. Groups of different letters are statistically different from one another.
Figure S5: Synthetic guide RNA screening for efficient activation of endogeneous RNF144A-AS1.
159
Figure S6: IFN is upregulated in OA. yo: years old.
Figure S7: RNA pull down followed by western blot confirmed EIF2AK2 as the binding partner of GRASLND.
160
Table S1: shRNA target sequences. 5’ – sequence – 3’ CTGGCATTTCTTGTCATTATT EMX2OS GAATGGCATTTAGTCTTTAAA CAGAAGAGGTGCTCGATAAAT LOXL1-AS1 TGGGTACTTTATTTGCTATTA GCACCATCCTCTAACAATAAT LINC01279 CAGGGCTCTGAGGTGAATATT ACAAGATGGTTAAACTCATTT MIR4435-2HG CTATTCAGCAGACAATGATAA RP11-366L20.2 GGTGATGTATGGCCCATAAAT TACAAAGGTGGCAAGATAAAT RNF144A-AS1 GGCAAGATAAATGACAATAAA
Table S2: Cloning primer sequences. 5’ – sequence – 3’ BGHp A TATCGATCACGAGACTAGCCTCGAGTCCATAGAGCCCACCGC forward AT primer BGHp A CGCCGTGTAAACGCGTCGACTGTGCCTTCTAGTTG reverse primer Lucifer ase GGCACAGTCGACGCGTTTACACGGCGATCTTGCC forward primer Lucifer ase GCCCCGAATGCTAGCATGGAAGATGCCAAAAACATTAAG reverse primer CMV promot er CATCTTCCATGCTAGCATTCGGGGCCGCGGAGGC forward primer CMV promot CAACCCCGTGCGAATTCGATATCAATTTTATCGATCACGAGAC er TAGCCTCGAGTTTACCAC reverse primer 161
dsRed CAACTAGAAGGCACAGTCGACGCGTTTACAATTCGTCGTGCTT forward G primer dsRed AGCCTCCGCGGCCCCGAATGCTAGCATGGATAGCACTGAGAA reverse C primer RNF14 4A- CAACTAGAAGGCACAGTCGACGCGTCTTTCATTCAACAAATTT AS1 TTATGAAGTGCC forward primer RNF14 4A- AGCCTCCGCGGCCCCGAATGCTAGCACGCCATTCTCCTGCCT AS1 C reverse primer
Table S3: RNF144A-AS1 probe set sequences. 5’ – sequence – 3’ Probe #1 GTTCGTGGTCTTTTTTCTAG Probe #2 TAGGCAGGTCTCAGGATGTC Probe #3 CTTCTGGAGGCCACCTAATG Probe #4 CGGTGTAGGAATCAGGGGAG Probe #5 GGTGAACAGACAGACTTTCC Probe #6 AGATGTCCATTCCACCTTTT Probe #7 TTCATCCTCTAGTGAGAGGC Probe #8 CAGGCCTTTATGCAGAACGA Probe #9 GTACTCAACCAGGAACTTCT Probe #10 CCCTGAATCCTTCATTCATA Probe #11 AACCATAATATCCCCCAGAT Probe #12 AGACGTTACATTCCACATTC Probe #13 TAGCACAGATGGGCTGAAGA Probe #14 TGGCTTCTGGAATGAGATGA Probe #15 TGTCTGCTCCTGAGAGAAGG Probe #16 CTCTGGCAGGAAAGTCTTGT Probe #17 TGTCTTATGTGGATGCTACT Probe #18 GCTTGAAGACATCTCTGCAT Probe #19 GCACTTTCTTTCTTGAAGTT Probe #20 CAGTGCTGTGTTAAGTGACT Probe #21 ATGAGTAGTCACCTTCCATG Probe #22 CCTTGTTCTCTAAGTTGCAG
162
References
1. Bitton R (2009) The economic burden of osteoarthritis. The American journal of managed care 15(8 Suppl):S230-235.
2. Helmick CG, et al. (2008) Estimates of the prevalence of arthritis and other rheumatic conditions in the United States. Part I. Arthritis and rheumatism 58(1):15-25.
3. Hyde G, Dover S, Aszodi A, Wallis GA, & Boot-Handford RP (2007) Lineage tracing using matrilin-1 gene expression reveals that articular chondrocytes exist as the joint interzone forms. Dev Biol 304(2):825-833.
4. Koyama E, et al. (2008) A distinct cohort of progenitor cells participates in synovial joint and articular cartilage formation during mouse limb skeletogenesis. Dev Biol 316(1):62-73.
5. Kronenberg HM (2003) Developmental regulation of the growth plate. Nature 423(6937):332-336.
6. Jing Y, et al. (2015) Chondrocytes Directly Transform into Bone Cells in Mandibular Condyle Growth. J Dent Res 94(12):1668-1675.
7. Park J, et al. (2015) Dual pathways to endochondral osteoblasts: a novel chondrocyte-derived osteoprogenitor cell identified in hypertrophic cartilage. Biol Open 4(5):608-621.
8. Yang L, Tsang KY, Tang HC, Chan D, & Cheah KS (2014) Hypertrophic chondrocytes can become osteoblasts and osteocytes in endochondral bone formation. Proc Natl Acad Sci U S A 111(33):12097-12102.
9. Zhou X, et al. (2014) Chondrocytes transdifferentiate into osteoblasts in endochondral bone during development, postnatal growth and fracture healing in mice. PLoS Genet 10(12):e1004820.
10. Gilbert SF (2010) Developmental biology (Sinauer Associates, Sunderland, Mass.) 9th Ed pp xxi, 711, 780 p.
11. Ash P, Loutit JF, & Townsend KM (1980) Osteoclasts derived from haematopoietic stem cells. Nature 283(5748):669-670.
12. Blair HC, Kahn AJ, Crouch EC, Jeffrey JJ, & Teitelbaum SL (1986) Isolated osteoclasts resorb the organic and inorganic components of bone. J Cell Biol 102(4):1164-1172.
163
13. Gerber HP, et al. (1999) VEGF is required for growth and survival in neonatal mice. Development 126(6):1149-1159.
14. Haigh JJ, Gerber HP, Ferrara N, & Wagner EF (2000) Conditional inactivation of VEGF-A in areas of collagen2a1 expression results in embryonic lethality in the heterozygous state. Development 127(7):1445- 1453.
15. Vega RB, et al. (2004) Histone deacetylase 4 controls chondrocyte hypertrophy during skeletogenesis. Cell 119(4):555-566.
16. Abzhanov A, Rodda SJ, McMahon AP, & Tabin CJ (2007) Regulation of skeletogenic differentiation in cranial dermal bone. Development 134(17):3133-3144.
17. Buckwalter JA (1995) Osteoarthritis and articular cartilage use, disuse, and abuse: experimental studies. J Rheumatol Suppl 43:13-15.
18. Gilbertson EM (1975) Development of periarticular osteophytes in experimentally induced osteoarthritis in the dog. A study using microradiographic, microangiographic, and fluorescent bone-labelling techniques. Ann Rheum Dis 34(1):12-25.
19. Guilak F (2011) Biomechanical factors in osteoarthritis. Best Pract Res Clin Rheumatol 25(6):815-823.
20. Hoch DH, Grodzinsky AJ, Koob TJ, Albert ML, & Eyre DR (1983) Early changes in material properties of rabbit articular cartilage after meniscectomy. Journal of orthopaedic research : official publication of the Orthopaedic Research Society 1(1):4-12.
21. Moskowitz RW, et al. (1973) Experimentally induced degenerative joint lesions following partial meniscectomy in the rabbit. Arthritis and rheumatism 16(3):397-405.
22. Pond MJ & Nuki G (1973) Experimentally-induced osteoarthritis in the dog. Ann Rheum Dis 32(4):387-388.
23. Griffin TM, et al. (2010) Diet-induced obesity differentially regulates behavioral, biomechanical, and molecular risk factors for osteoarthritis in mice. Arthritis research & therapy 12(4):R130.
24. Wu CL, Diekman BO, Jain D, & Guilak F (2013) Diet-induced obesity alters the differentiation potential of stem cells isolated from bone marrow,
164
adipose tissue and infrapatellar fat pad: the effects of free fatty acids. Int J Obes (Lond) 37(8):1079-1087.
25. Hugle T, Geurts J, Nuesch C, Muller-Gerbl M, & Valderrabano V (2012) Aging and osteoarthritis: an inevitable encounter? J Aging Res 2012:950192.
26. Goldring MB (2012) Articular cartilage degradation in osteoarthritis. HSS J 8(1):7-9.
27. Sokolove J & Lepus CM (2013) Role of inflammation in the pathogenesis of osteoarthritis: latest findings and interpretations. Ther Adv Musculoskelet Dis 5(2):77-94.
28. Loeser RF, Goldring SR, Scanzello CR, & Goldring MB (2012) Osteoarthritis: a disease of the joint as an organ. Arthritis and rheumatism 64(6):1697-1707.
29. Sophia Fox AJ, Bedi A, & Rodeo SA (2009) The basic science of articular cartilage: structure, composition, and function. Sports Health 1(6):461- 468.
30. Hootman JM & Helmick CG (2006) Projections of US prevalence of arthritis and associated activity limitations. Arthritis Rheum 54(1):226-229.
31. Lawrence RC, et al. (2008) Estimates of the prevalence of arthritis and other rheumatic conditions in the United States. Part II. Arthritis Rheum 58(1):26-35.
32. Math KR, Zaidi SF, Petchprapa C, & Harwin SF (2006) Imaging of total knee arthroplasty. Semin Musculoskelet Radiol 10(1):47-63.
33. Mulcahy H & Chew FS (2013) Current concepts in knee replacement: features and imaging assessment. AJR Am J Roentgenol 201(6):W828- 842.
34. Mulcahy H & Chew FS (2014) Current concepts in knee replacement: complications. AJR Am J Roentgenol 202(1):W76-86.
35. Bozic KJ, et al. (2010) The epidemiology of revision total knee arthroplasty in the United States. Clin Orthop Relat Res 468(1):45-51.
36. Kurtz S, Ong K, Lau E, Mowat F, & Halpern M (2007) Projections of primary and revision hip and knee arthroplasty in the United States from 2005 to 2030. J Bone Joint Surg Am 89(4):780-785. 165
37. Mikos AG, et al. (2006) Engineering complex tissues. Tissue Eng 12(12):3307-3339.
38. Kapranov P, Willingham AT, & Gingeras TR (2007) Genome-wide transcription and the implications for genomic organization. Nature reviews. Genetics 8(6):413-423.
39. Mercer TR, Dinger ME, & Mattick JS (2009) Long non-coding RNAs: insights into functions. Nature reviews. Genetics 10(3):155-159.
40. Aravin A, et al. (2006) A novel class of small RNAs bind to MILI protein in mouse testes. Nature 442(7099):203-207.
41. Grishok A & Mello CC (2002) RNAi (Nematodes: Caenorhabditis elegans). Advances in genetics 46:339-360.
42. Montgomery MK & Fire A (1998) Double-stranded RNA as a mediator in sequence-specific genetic silencing and co-suppression. Trends in genetics : TIG 14(7):255-258.
43. Ambros V (2001) microRNAs: tiny regulators with great potential. Cell 107(7):823-826.
44. Amaral PP & Mattick JS (2008) Noncoding RNA in development. Mammalian genome : official journal of the International Mammalian Genome Society 19(7-8):454-492.
45. Girard A, Sachidanandam R, Hannon GJ, & Carmell MA (2006) A germline-specific class of small RNAs binds mammalian Piwi proteins. Nature 442(7099):199-202.
46. Grivna ST, Beyret E, Wang Z, & Lin H (2006) A novel class of small RNAs in mouse spermatogenic cells. Genes & development 20(13):1709-1714.
47. Lau NC, et al. (2006) Characterization of the piRNA complex from rat testes. Science 313(5785):363-367.
48. Lee HC, et al. (2009) qiRNA is a new type of small interfering RNA induced by DNA damage. Nature 459(7244):274-277.
49. Vagin VV, et al. (2006) A distinct small RNA pathway silences selfish genetic elements in the germline. Science 313(5785):320-324.
50. Watanabe T, et al. (2006) Identification and characterization of two novel classes of small RNAs in the mouse germline: retrotransposon-derived 166
siRNAs in oocytes and germline small RNAs in testes. Genes & development 20(13):1732-1743.
51. Ponjavic J, Ponting CP, & Lunter G (2007) Functionality or transcriptional noise? Evidence for selection within long noncoding RNAs. Genome research 17(5):556-565.
52. Smith MA, Gesell T, Stadler PF, & Mattick JS (2013) Widespread purifying selection on RNA structure in mammals. Nucleic acids research 41(17):8220-8236.
53. Washietl S, Hofacker IL, Lukasser M, Huttenhofer A, & Stadler PF (2005) Mapping of conserved RNA secondary structures predicts thousands of functional noncoding RNAs in the human genome. Nature biotechnology 23(11):1383-1390.
54. Nielsen MM, et al. (2014) Identification of expressed and conserved human noncoding RNAs. Rna 20(2):236-251.
55. Torarinsson E, et al. (2008) Comparative genomics beyond sequence- based alignments: RNA structures in the ENCODE regions. Genome research 18(2):242-251.
56. Li L, et al. (2013) Targeted disruption of Hotair leads to homeotic transformation and gene derepression. Cell reports 5(1):3-12.
57. Tarazona S, Garcia-Alcalde F, Dopazo J, Ferrer A, & Conesa A (2011) Differential expression in RNA-seq: a matter of depth. Genome research 21(12):2213-2223.
58. Iyer MK, et al. (2015) The landscape of long noncoding RNAs in the human transcriptome. Nature genetics 47(3):199-208.
59. Trapnell C, et al. (2012) Differential gene and transcript expression analysis of RNA-seq experiments with TopHat and Cufflinks. Nature protocols 7(3):562-578.
60. Trapnell C, et al. (2010) Transcript assembly and quantification by RNA- Seq reveals unannotated transcripts and isoform switching during cell differentiation. Nature biotechnology 28(5):511-515.
61. Mercer TR, et al. (2012) Targeted RNA sequencing reveals the deep complexity of the human transcriptome. Nature biotechnology 30(1):99- 104.
167
62. Olivarius S, Plessy C, & Carninci P (2009) High-throughput verification of transcriptional starting sites by Deep-RACE. BioTechniques 46(2):130- 132.
63. Raj A, van den Bogaard P, Rifkin SA, van Oudenaarden A, & Tyagi S (2008) Imaging individual mRNA molecules using multiple singly labeled probes. Nature methods 5(10):877-879.
64. Cabili MN, et al. (2015) Localization and abundance analysis of human lncRNAs at single-cell and single-molecule resolution. Genome biology 16:20.
65. van Heesch S, et al. (2014) Extensive localization of long noncoding RNAs to the cytosol and mono- and polyribosomal complexes. Genome biology 15(1):R6.
66. Werner MS & Ruthenburg AJ (2015) Nuclear Fractionation Reveals Thousands of Chromatin-Tethered Noncoding RNAs Adjacent to Active Genes. Cell reports 12(7):1089-1098.
67. McHugh CA, Russell P, & Guttman M (2014) Methods for comprehensive experimental identification of RNA-protein interactions. Genome Biol 15(1):203.
68. Engreitz JM, et al. (2014) RNA-RNA interactions enable specific targeting of noncoding RNAs to nascent Pre-mRNAs and chromatin sites. Cell 159(1):188-199.
69. Chu C, Quinn J, & Chang HY (2012) Chromatin isolation by RNA purification (ChIRP). Journal of visualized experiments : JoVE (61).
70. Simon MD, et al. (2011) The genomic binding sites of a noncoding RNA. Proceedings of the National Academy of Sciences of the United States of America 108(51):20497-20502.
71. McHugh CA, et al. (2015) The Xist lncRNA interacts directly with SHARP to silence transcription through HDAC3. Nature 521(7551):232-236.
72. Shechner DM, Hacisuleyman E, Younger ST, & Rinn JL (2015) Multiplexable, locus-specific targeting of long RNAs with CRISPR-Display. Nature methods 12(7):664-670.
73. Kanhere A, et al. (2010) Short RNAs are transcribed from repressed polycomb target genes and interact with polycomb repressive complex-2. Molecular cell 38(5):675-688. 168
74. Pedersen JS, et al. (2006) Identification and classification of conserved RNA secondary structures in the human genome. PLoS computational biology 2(4):e33.
75. Merino EJ, Wilkinson KA, Coughlan JL, & Weeks KM (2005) RNA structure analysis at single nucleotide resolution by selective 2'-hydroxyl acylation and primer extension (SHAPE). J Am Chem Soc 127(12):4223- 4231.
76. Liao Q, et al. (2011) Large-scale prediction of long non-coding RNA functions in a coding-non-coding gene co-expression network. Nucleic acids research 39(9):3864-3878.
77. Lennox KA & Behlke MA (2016) Cellular localization of long non-coding RNAs affects silencing by RNAi more than by antisense oligonucleotides. Nucleic acids research 44(2):863-877.
78. Goff LA & Rinn JL (2015) Linking RNA biology to lncRNAs. Genome research 25(10):1456-1465.
79. Shalem O, et al. (2014) Genome-scale CRISPR-Cas9 knockout screening in human cells. Science 343(6166):84-87.
80. Necsulea A, et al. (2014) The evolution of lncRNA repertoires and expression patterns in tetrapods. Nature 505(7485):635-640.
81. Guttman M, et al. (2010) Ab initio reconstruction of cell type-specific transcriptomes in mouse reveals the conserved multi-exonic structure of lincRNAs. Nature biotechnology 28(5):503-510.
82. Derrien T, et al. (2012) The GENCODE v7 catalog of human long noncoding RNAs: analysis of their gene structure, evolution, and expression. Genome research 22(9):1775-1789.
83. Cabili MN, et al. (2011) Integrative annotation of human large intergenic noncoding RNAs reveals global properties and specific subclasses. Genes & development 25(18):1915-1927.
84. Li L & Chang HY (2014) Physiological roles of long noncoding RNAs: insight from knockout mice. Trends Cell Biol 24(10):594-602.
85. Penny GD, Kay GF, Sheardown SA, Rastan S, & Brockdorff N (1996) Requirement for Xist in X chromosome inactivation. Nature 379(6561):131-137.
169
86. Fitzpatrick GV, Soloway PD, & Higgins MJ (2002) Regional loss of imprinting and growth deficiency in mice with a targeted deletion of KvDMR1. Nature genetics 32(3):426-431.
87. Mancini-Dinardo D, Steele SJ, Levorse JM, Ingram RS, & Tilghman SM (2006) Elongation of the Kcnq1ot1 transcript is required for genomic imprinting of neighboring genes. Genes & development 20(10):1268-1282.
88. Ho TT, et al. (2015) Targeting non-coding RNAs with the CRISPR/Cas9 system in human cell lines. Nucleic acids research 43(3):e17.
89. Guttman M, et al. (2011) lincRNAs act in the circuitry controlling pluripotency and differentiation. Nature 477(7364):295-300.
90. Ng SY, Johnson R, & Stanton LW (2012) Human long non-coding RNAs promote pluripotency and neuronal differentiation by association with chromatin modifiers and transcription factors. The EMBO journal 31(3):522-533.
91. Le MT, et al. (2009) MicroRNA-125b promotes neuronal differentiation in human cells by repressing multiple targets. Molecular and cellular biology 29(19):5290-5305.
92. Rybak A, et al. (2008) A feedback loop comprising lin-28 and let-7 controls pre-let-7 maturation during neural stem-cell commitment. Nature cell biology 10(8):987-993.
93. Ramos AD, et al. (2013) Integration of genome-wide approaches identifies lncRNAs of adult neural stem cells and their progeny in vivo. Cell stem cell 12(5):616-628.
94. Grote P, et al. (2013) The tissue-specific lncRNA Fendrr is an essential regulator of heart and body wall development in the mouse. Developmental cell 24(2):206-214.
95. Klattenhoff CA, et al. (2013) Braveheart, a long noncoding RNA required for cardiovascular lineage commitment. Cell 152(3):570-583.
96. Mackay AM, et al. (1998) Chondrogenic differentiation of cultured human mesenchymal stem cells from marrow. Tissue engineering 4(4):415-428.
97. Johnstone B, Hering TM, Caplan AI, Goldberg VM, & Yoo JU (1998) In vitro chondrogenesis of bone marrow-derived mesenchymal progenitor cells. Exp Cell Res 238(1):265-272.
170
98. Boeuf S & Richter W (2010) Chondrogenesis of mesenchymal stem cells: role of tissue source and inducing factors. Stem cell research & therapy 1(4):31.
99. Beane OS & Darling EM (2012) Isolation, characterization, and differentiation of stem cells for cartilage regeneration. Ann Biomed Eng 40(10):2079-2097.
100. Wang L, Li Z, Li Z, Yu B, & Wang Y (2015) Long noncoding RNAs expression signatures in chondrogenic differentiation of human bone marrow mesenchymal stem cells. Biochemical and biophysical research communications 456(1):459-464.
101. Barry F, Boynton RE, Liu B, & Murphy JM (2001) Chondrogenic differentiation of mesenchymal stem cells from bone marrow: differentiation-dependent gene expression of matrix components. Experimental cell research 268(2):189-200.
102. Hardingham TE, Oldershaw RA, & Tew SR (2006) Cartilage, SOX9 and Notch signals in chondrogenesis. J Anat 209(4):469-480.
103. Zhang L, et al. (2015) Sox4 enhances chondrogenic differentiation and proliferation of human synovium-derived stem cell via activation of long noncoding RNA DANCR. J Mol Histol 46(6):467-473.
104. Kretz M, et al. (2012) Suppression of progenitor differentiation requires the long noncoding RNA ANCR. Genes & development 26(4):338-343.
105. Yuan SX, et al. (2016) Long noncoding RNA DANCR increases stemness features of hepatocellular carcinoma by derepression of CTNNB1. Hepatology 63(2):499-511.
106. Chen L, et al. (2015) lncRNA DANCR suppresses odontoblast-like differentiation of human dental pulp cells by inhibiting wnt/beta-catenin pathway. Cell and tissue research.
107. Maass PG, et al. (2012) A misplaced lncRNA causes brachydactyly in humans. J Clin Invest 122(11):3990-4002.
108. Vortkamp A, et al. (1996) Regulation of rate of cartilage differentiation by Indian hedgehog and PTH-related protein. Science 273(5275):613-622.
109. Bi W, Deng JM, Zhang Z, Behringer RR, & de Crombrugghe B (1999) Sox9 is required for cartilage formation. Nature genetics 22(1):85-89.
171
110. Pineault KM & Wellik DM (2014) Hox genes and limb musculoskeletal development. Curr Osteoporos Rep 12(4):420-427.
111. Favier B, Le Meur M, Chambon P, & Dolle P (1995) Axial skeleton homeosis and forelimb malformations in Hoxd-11 mutant mice. Proceedings of the National Academy of Sciences of the United States of America 92(1):310-314.
112. Gerard M, et al. (1996) In vivo targeted mutagenesis of a regulatory element required for positioning the Hoxd-11 and Hoxd-10 expression boundaries. Genes & development 10(18):2326-2334.
113. Hashimoto K, Yokouchi Y, Yamamoto M, & Kuroiwa A (1999) Distinct signaling molecules control Hoxa-11 and Hoxa-13 expression in the muscle precursor and mesenchyme of the chick limb bud. Development 126(12):2771-2783.
114. Salsi V & Zappavigna V (2006) Hoxd13 and Hoxa13 directly control the expression of the EphA7 Ephrin tyrosine kinase receptor in developing limbs. The Journal of biological chemistry 281(4):1992-1999.
115. Rinn JL, et al. (2007) Functional demarcation of active and silent chromatin domains in human HOX loci by noncoding RNAs. Cell 129(7):1311-1323.
116. Carrion K, et al. (2014) The long non-coding HOTAIR is modulated by cyclic stretch and WNT/beta-CATENIN in human aortic valve cells and is a novel repressor of calcification genes. PloS one 9(5):e96577.
117. Wang KC, et al. (2011) A long noncoding RNA maintains active chromatin to coordinate homeotic gene expression. Nature 472(7341):120-124.
118. Schuettengruber B, Chourrout D, Vervoort M, Leblanc B, & Cavalli G (2007) Genome regulation by polycomb and trithorax proteins. Cell 128(4):735-745.
119. Carlson HL, et al. (2015) LncRNA-HIT Functions as an Epigenetic Regulator of Chondrogenesis through Its Recruitment of p100/CBP Complexes. PLoS genetics 11(12):e1005680.
120. Richards EJ, et al. (2015) Long non-coding RNAs (LncRNA) regulated by transforming growth factor (TGF) beta: LncRNA-hit-mediated TGFbeta- induced epithelial to mesenchymal transition in mammary epithelia. The Journal of biological chemistry 290(11):6857-6867.
172
121. Kawai J, et al. (2001) Functional annotation of a full-length mouse cDNA collection. Nature 409(6821):685-690.
122. Suzuki T, Hasso SM, & Fallon JF (2008) Unique SMAD1/5/8 activity at the phalanx-forming region determines digit identity. Proceedings of the National Academy of Sciences of the United States of America 105(11):4185-4190.
123. Chimal-Monroy J, et al. (2003) Analysis of the molecular cascade responsible for mesodermal limb chondrogenesis: Sox genes and BMP signaling. Dev Biol 257(2):292-301.
124. Lehmann K, et al. (2003) Mutations in bone morphogenetic protein receptor 1B cause brachydactyly type A2. Proceedings of the National Academy of Sciences of the United States of America 100(21):12277- 12282.
125. Baur ST, Mai JJ, & Dymecki SM (2000) Combinatorial signaling through BMP receptor IB and GDF5: shaping of the distal mouse limb and the genetics of distal limb diversity. Development 127(3):605-619.
126. Liang WC, et al. (2016) H19 activates Wnt signaling and promotes osteoblast differentiation by functioning as a competing endogenous RNA. Sci Rep 6:20121.
127. Huang Y, Zheng Y, Jia L, & Li W (2015) Long Noncoding RNA H19 Promotes Osteoblast Differentiation Via TGF-beta1/Smad3/HDAC Signaling Pathway by Deriving miR-675. Stem cells 33(12):3481-3492.
128. Kang JS, Alliston T, Delston R, & Derynck R (2005) Repression of Runx2 function by TGF-beta through recruitment of class II histone deacetylases by Smad3. EMBO J 24(14):2543-2555.
129. Saydam O, et al. (2009) Downregulated microRNA-200a in meningiomas promotes tumor growth by reducing E-cadherin and activating the Wnt/beta-catenin signaling pathway. Molecular and cellular biology 29(21):5923-5940.
130. Dudek KA, Lafont JE, Martinez-Sanchez A, & Murphy CL (2010) Type II collagen expression is regulated by tissue-specific miR-675 in human articular chondrocytes. The Journal of biological chemistry 285(32):24381- 24387.
131. Zhu L & Xu PC (2013) Downregulated LncRNA-ANCR promotes osteoblast differentiation by targeting EZH2 and regulating Runx2 173
expression. Biochemical and biophysical research communications 432(4):612-617.
132. Jia Q, Jiang W, & Ni L (2015) Down-regulated non-coding RNA (lncRNA- ANCR) promotes osteogenic differentiation of periodontal ligament stem cells. Arch Oral Biol 60(2):234-241.
133. Nagase T, et al. (1995) Prediction of the coding sequences of unidentified human genes. III. The coding sequences of 40 new genes (KIAA0081- KIAA0120) deduced by analysis of cDNA clones from human cell line KG- 1. DNA Res 2(1):37-43.
134. Kikuchi K, et al. (2009) Transcripts of unknown function in multiple- signaling pathways involved in human stem cell differentiation. Nucleic acids research 37(15):4987-5000.
135. Zuo C, et al. (2013) Expression profiling of lncRNAs in C3H10T1/2 mesenchymal stem cells undergoing early osteoblast differentiation. Mol Med Rep 8(2):463-467.
136. Backesjo CM, Li Y, Lindgren U, & Haldosen LA (2006) Activation of Sirt1 decreases adipocyte formation during osteoblast differentiation of mesenchymal stem cells. J Bone Miner Res 21(7):993-1002.
137. Cohen-Kfir E, et al. (2011) Sirt1 is a regulator of bone mass and a repressor of Sost encoding for sclerostin, a bone formation inhibitor. Endocrinology 152(12):4514-4524.
138. Zhuang W, et al. (2015) Upregulation of lncRNA MEG3 Promotes Osteogenic Differentiation of Mesenchymal Stem Cells From Multiple Myeloma Patients By Targeting BMP4 Transcription. Stem cells 33(6):1985-1997.
139. Steck E, et al. (2012) Regulation of H19 and its encoded microRNA-675 in osteoarthritis and under anabolic and catabolic in vitro conditions. Journal of molecular medicine 90(10):1185-1195.
140. Zhou J, et al. (2015) H19 lncRNA alters DNA methylation genome wide by regulating S-adenosylhomocysteine hydrolase. Nat Commun 6:10221.
141. Xing D, et al. (2014) Identification of long noncoding RNA associated with osteoarthritis in humans. Orthop Surg 6(4):288-293.
174
142. Kim D, et al. (2013) Two non-coding RNAs, MicroRNA-101 and HOTTIP contribute cartilage integrity by epigenetic and homeotic regulation of integrin-alpha1. Cellular signalling 25(12):2878-2887.
143. Fu M, et al. (2015) Expression profile of long noncoding RNAs in cartilage from knee osteoarthritis patients. Osteoarthritis Cartilage 23(3):423-432.
144. Liu Q, et al. (2014) Long noncoding RNA related to cartilage injury promotes chondrocyte extracellular matrix degradation in osteoarthritis. Arthritis & rheumatology 66(4):969-978.
145. Song J, Ahn C, Chun CH, & Jin EJ (2014) A long non-coding RNA, GAS5, plays a critical role in the regulation of miR-21 during osteoarthritis. Journal of orthopaedic research : official publication of the Orthopaedic Research Society 32(12):1628-1635.
146. Glasson SS, Blanchet TJ, & Morris EA (2007) The surgical destabilization of the medial meniscus (DMM) model of osteoarthritis in the 129/SvEv mouse. Osteoarthritis Cartilage 15(9):1061-1069.
147. Su W, Xie W, Shang Q, & Su B (2015) The Long Noncoding RNA MEG3 Is Downregulated and Inversely Associated with VEGF Levels in Osteoarthritis. Biomed Res Int 2015:356893.
148. Pearson MJ, et al. (2016) Long Intergenic Noncoding RNAs Mediate the Human Chondrocyte Inflammatory Response and Are Differentially Expressed in Osteoarthritis Cartilage. Arthritis & rheumatology 68(4):845- 856.
149. Haddox CL, et al. (2014) Osteosarcoma in pediatric patients and young adults: a single institution retrospective review of presentation, therapy, and outcome. Sarcoma 2014:402509.
150. Hagleitner MM, de Bont ES, & Te Loo DM (2012) Survival trends and long-term toxicity in pediatric patients with osteosarcoma. Sarcoma 2012:636405.
151. Chan LH, et al. (2014) Hedgehog signaling induces osteosarcoma development through Yap1 and H19 overexpression. Oncogene 33(40):4857-4866.
152. Sathiyamoorthy S & Ali SZ (2012) Osteoblastic osteosarcoma: cytomorphologic characteristics and differential diagnosis on fine-needle aspiration. Acta Cytol 56(5):481-486.
175
153. Bernthal NM, et al. (2012) Long-term results (>25 years) of a randomized, prospective clinical trial evaluating chemotherapy in patients with high- grade, operable osteosarcoma. Cancer 118(23):5888-5893.
154. Eilber F, et al. (1987) Adjuvant chemotherapy for osteosarcoma: a randomized prospective trial. J Clin Oncol 5(1):21-26.
155. Link MP, et al. (1986) The effect of adjuvant chemotherapy on relapse- free survival in patients with osteosarcoma of the extremity. N Engl J Med 314(25):1600-1606.
156. Meyers PA, et al. (1998) Intensification of preoperative chemotherapy for osteogenic sarcoma: results of the Memorial Sloan-Kettering (T12) protocol. J Clin Oncol 16(7):2452-2458.
157. Rosen G, et al. (1983) Primary osteogenic sarcoma: eight-year experience with adjuvant chemotherapy. J Cancer Res Clin Oncol 106 Suppl:55-67.
158. Ji P, et al. (2003) MALAT-1, a novel noncoding RNA, and thymosin beta4 predict metastasis and survival in early-stage non-small cell lung cancer. Oncogene 22(39):8031-8041.
159. Wilusz JE, et al. (2012) A triple helix stabilizes the 3' ends of long noncoding RNAs that lack poly(A) tails. Genes & development 26(21):2392-2407.
160. Yoshimoto R, Mayeda A, Yoshida M, & Nakagawa S (2016) MALAT1 long non-coding RNA in cancer. Biochim Biophys Acta 1859(1):192-199.
161. Dong Y, et al. (2015) MALAT1 promotes the proliferation and metastasis of osteosarcoma cells by activating the PI3K/Akt pathway. Tumour Biol 36(3):1477-1486.
162. Cai X, et al. (2015) Long noncoding RNA MALAT1 as a potential therapeutic target in osteosarcoma. Journal of orthopaedic research : official publication of the Orthopaedic Research Society.
163. Fang D, et al. (2015) 17beta-estradiol regulates cell proliferation, colony formation, migration, invasion and promotes apoptosis by upregulating miR-9 and thus degrades MALAT-1 in osteosarcoma cell MG-63 in an estrogen receptor-independent manner. Biochemical and biophysical research communications 457(4):500-506.
164. Ji Q, et al. (2014) Long non-coding RNA MALAT1 promotes tumour growth and metastasis in colorectal cancer through binding to SFPQ and 176
releasing oncogene PTBP2 from SFPQ/PTBP2 complex. Br J Cancer 111(4):736-748.
165. Ruan W, Wang P, Feng S, Xue Y, & Li Y (2015) Long non-coding RNA small nucleolar RNA host gene 12 (SNHG12) promotes cell proliferation and migration by upregulating angiomotin gene expression in human osteosarcoma cells. Tumour Biol.
166. Li JP, et al. (2013) Microarray expression profile of long noncoding RNAs in human osteosarcoma. Biochemical and biophysical research communications 433(2):200-206.
167. Sun XH, Yang LB, Geng XL, Wang R, & Zhang ZC (2015) Increased expression of lncRNA HULC indicates a poor prognosis and promotes cell metastasis in osteosarcoma. Int J Clin Exp Pathol 8(3):2994-3000.
168. Pasic I, et al. (2010) Recurrent focal copy-number changes and loss of heterozygosity implicate two noncoding RNAs and one tumor suppressor gene at chromosome 3q13.31 in osteosarcoma. Cancer research 70(1):160-171.
169. Cong M, Li J, Jing R, & Li Z (2016) Long non-coding RNA tumor suppressor candidate 7 functions as a tumor suppressor and inhibits proliferation in osteosarcoma. Tumour Biol.
170. Wang Y, et al. (2015) A novel long non-coding RNA, hypoxia-inducible factor-2alpha promoter upstream transcript, functions as an inhibitor of osteosarcoma stem cells in vitro. Mol Med Rep 11(4):2534-2540.
171. Min L, et al. (2016) Antidifferentiation Noncoding RNA Regulates the Proliferation of Osteosarcoma Cells. Cancer Biother Radiopharm 31(2):52-57.
172. Li F, Cao L, Hang D, Wang F, & Wang Q (2015) Long non-coding RNA HOTTIP is up-regulated and associated with poor prognosis in patients with osteosarcoma. Int J Clin Exp Pathol 8(9):11414-11420.
173. Wang B, et al. (2015) Overexpression of Long Non-Coding RNA HOTAIR Promotes Tumor Growth and Metastasis in Human Osteosarcoma. Mol Cells 38(5):432-440.
174. Qian M, et al. (2015) P50-associated COX-2 extragenic RNA (PACER) overexpression promotes proliferation and metastasis of osteosarcoma cells by activating COX-2 gene. Tumour Biol.
177
175. Ulaner GA, et al. (2003) Loss of imprinting of IGF2 and H19 in osteosarcoma is accompanied by reciprocal methylation changes of a CTCF-binding site. Human molecular genetics 12(5):535-549.
176. Tian ZZ, Guo XJ, Zhao YM, & Fang Y (2015) Decreased expression of long non-coding RNA MEG3 acts as a potential predictor biomarker in progression and poor prognosis of osteosarcoma. Int J Clin Exp Pathol 8(11):15138-15142.
177. Ma B, et al. (2015) Upregulation of long non-coding RNA TUG1 correlates with poor prognosis and disease status in osteosarcoma. Tumour Biol.
178. Zhang CL, Zhu KP, Shen GQ, & Zhu ZS (2015) A long non-coding RNA contributes to doxorubicin resistance of osteosarcoma. Tumour Biol.
179. Lam J, et al. (2014) Generation of osteochondral tissue constructs with chondrogenically and osteogenically predifferentiated mesenchymal stem cells encapsulated in bilayered hydrogels. Acta Biomater 10(3):1112- 1123.
180. Schaefer D, et al. (2000) In vitro generation of osteochondral composites. Biomaterials 21(24):2599-2606.
181. Schaefer D, et al. (2002) Tissue-engineered composites for the repair of large osteochondral defects. Arthritis Rheum 46(9):2524-2534.
182. Gao J, et al. (2001) Tissue-engineered fabrication of an osteochondral composite graft using rat bone marrow-derived mesenchymal stem cells. Tissue Eng 7(4):363-371.
183. Cao T, Ho KH, & Teoh SH (2003) Scaffold design and in vitro study of osteochondral coculture in a three-dimensional porous polycaprolactone scaffold fabricated by fused deposition modeling. Tissue Eng 9 Suppl 1:S103-112.
184. Moutos FT & Guilak F (2008) Composite scaffolds for cartilage tissue engineering. Biorheology 45(3-4):501-512.
185. O'Shea TM & Miao X (2008) Bilayered scaffolds for osteochondral tissue engineering. Tissue Eng Part B Rev 14(4):447-464.
186. Khanarian NT, Haney NM, Burga RA, & Lu HH (2012) A functional agarose-hydroxyapatite scaffold for osteochondral interface regeneration. Biomaterials 33(21):5247-5258.
178
187. Khanarian NT, Jiang J, Wan LQ, Mow VC, & Lu HH (2012) A hydrogel- mineral composite scaffold for osteochondral interface tissue engineering. Tissue Eng Part A 18(5-6):533-545.
188. Guo X, et al. (2009) In vitro generation of an osteochondral construct using injectable hydrogel composites encapsulating rabbit marrow mesenchymal stem cells. Biomaterials 30(14):2741-2752.
189. Guo X, et al. (2010) Effects of TGF-beta3 and preculture period of osteogenic cells on the chondrogenic differentiation of rabbit marrow mesenchymal stem cells encapsulated in a bilayered hydrogel composite. Acta Biomater 6(8):2920-2931.
190. Cheng HW, Luk KD, Cheung KM, & Chan BP (2011) In vitro generation of an osteochondral interface from mesenchymal stem cell-collagen microspheres. Biomaterials 32(6):1526-1535.
191. Grayson WL, Bhumiratana S, Grace Chao PH, Hung CT, & Vunjak- Novakovic G (2010) Spatial regulation of human mesenchymal stem cell differentiation in engineered osteochondral constructs: effects of pre- differentiation, soluble factors and medium perfusion. Osteoarthritis Cartilage 18(5):714-723.
192. Re'em T, Witte F, Willbold E, Ruvinov E, & Cohen S (2012) Simultaneous regeneration of articular cartilage and subchondral bone induced by spatially presented TGF-beta and BMP-4 in a bilayer affinity binding system. Acta Biomater 8(9):3283-3293.
193. Chen J, et al. (2011) Simultaneous regeneration of articular cartilage and subchondral bone in vivo using MSCs induced by a spatially controlled gene delivery system in bilayered integrated scaffolds. Biomaterials 32(21):4793-4805.
194. Tuli R, et al. (2004) Human mesenchymal progenitor cell-based tissue engineering of a single-unit osteochondral construct. Tissue Eng 10(7- 8):1169-1179.
195. Bhatia R & Hare JM (2005) Mesenchymal stem cells: future source for reparative medicine. Congest Heart Fail 11(2):87-91; quiz 92-83.
196. Pittenger MF, et al. (1999) Multilineage potential of adult human mesenchymal stem cells. Science 284(5411):143-147.
179
197. Brunger JM, et al. (2014) Scaffold-mediated lentiviral transduction for functional tissue engineering of cartilage. Proc Natl Acad Sci U S A 111(9):E798-806.
198. Glass KA, et al. (2014) Tissue-engineered cartilage with inducible and tunable immunomodulatory properties. Biomaterials 35(22):5921-5931.
199. Moutos FT, Freed LE, & Guilak F (2007) A biomimetic three-dimensional woven composite scaffold for functional tissue engineering of cartilage. Nat Mater 6(2):162-167.
200. Moutos FT, et al. (2016) Anatomically shaped tissue-engineered cartilage with tunable and inducible anticytokine delivery for biological joint resurfacing. Proc Natl Acad Sci U S A 113(31):E4513-4522.
201. Abrahamsson CK, et al. (2010) Chondrogenesis and mineralization during in vitro culture of human mesenchymal stem cells on three-dimensional woven scaffolds. Tissue Eng Part A 16(12):3709-3718.
202. Somoza RA, Welter JF, Correa D, & Caplan AI (2014) Chondrogenic differentiation of mesenchymal stem cells: challenges and unfulfilled expectations. Tissue Eng Part B Rev 20(6):596-608.
203. Hillel AT, et al. (2010) Characterization of human mesenchymal stem cell- engineered cartilage: analysis of its ultrastructure, cell density and chondrocyte phenotype compared to native adult and fetal cartilage. Cells Tissues Organs 191(1):12-20.
204. Zhai LJ, Zhao KQ, Wang ZQ, Feng Y, & Xing SC (2011) Mesenchymal stem cells display different gene expression profiles compared to hyaline and elastic chondrocytes. Int J Clin Exp Med 4(1):81-90.
205. Kruger JP, Machens I, Lahner M, Endres M, & Kaps C (2014) Initial boost release of transforming growth factor-beta3 and chondrogenesis by freeze-dried bioactive polymer scaffolds. Ann Biomed Eng 42(12):2562- 2576.
206. Madry H, Rey-Rico A, Venkatesan JK, Johnstone B, & Cucchiarini M (2014) Transforming growth factor Beta-releasing scaffolds for cartilage tissue engineering. Tissue Eng Part B Rev 20(2):106-125.
207. Lian JB, et al. (2006) Networks and hubs for the transcriptional control of osteoblastogenesis. Rev Endocr Metab Disord 7(1-2):1-16.
180
208. Kempen DH, et al. (2008) Retention of in vitro and in vivo BMP-2 bioactivities in sustained delivery vehicles for bone tissue engineering. Biomaterials 29(22):3245-3252.
209. Musgrave DS, et al. (1999) Adenovirus-mediated direct gene therapy with bone morphogenetic protein-2 produces bone. Bone 24(6):541-547.
210. Song B, Estrada KD, & Lyons KM (2009) Smad signaling in skeletal development and regeneration. Cytokine Growth Factor Rev 20(5-6):379- 388.
211. Kim DW & Lassar AB (2003) Smad-dependent recruitment of a histone deacetylase/Sin3A complex modulates the bone morphogenetic protein- dependent transcriptional repressor activity of Nkx3.2. Mol Cell Biol 23(23):8704-8717.
212. Alliston T, Choy L, Ducy P, Karsenty G, & Derynck R (2001) TGF-beta- induced repression of CBFA1 by Smad3 decreases cbfa1 and osteocalcin expression and inhibits osteoblast differentiation. EMBO J 20(9):2254- 2272.
213. Tang SY & Alliston T (2013) Regulation of postnatal bone homeostasis by TGFbeta. Bonekey Rep 2:255.
214. Bian L, et al. (2011) Enhanced MSC chondrogenesis following delivery of TGF-beta3 from alginate microspheres within hyaluronic acid hydrogels in vitro and in vivo. Biomaterials 32(27):6425-6434.
215. Stein GS, et al. (2004) Runx2 control of organization, assembly and activity of the regulatory machinery for skeletal gene expression. Oncogene 23(24):4315-4329.
216. Byers BA, Guldberg RE, & Garcia AJ (2004) Synergy between genetic and tissue engineering: Runx2 overexpression and in vitro construct development enhance in vivo mineralization. Tissue Eng 10(11-12):1757- 1766.
217. Byers BA, Pavlath GK, Murphy TJ, Karsenty G, & Garcia AJ (2002) Cell- type-dependent up-regulation of in vitro mineralization after overexpression of the osteoblast-specific transcription factor Runx2/Cbfal. J Bone Miner Res 17(11):1931-1944.
218. Gersbach CA, Byers BA, Pavlath GK, & Garcia AJ (2004) Runx2/Cbfa1 stimulates transdifferentiation of primary skeletal myoblasts into a mineralizing osteoblastic phenotype. Exp Cell Res 300(2):406-417. 181
219. Gersbach CA, Coyer SR, Le Doux JM, & Garcia AJ (2007) Biomaterial- mediated retroviral gene transfer using self-assembled monolayers. Biomaterials 28(34):5121-5127.
220. Gersbach CA, Guldberg RE, & Garcia AJ (2007) In vitro and in vivo osteoblastic differentiation of BMP-2- and Runx2-engineered skeletal myoblasts. J Cell Biochem 100(5):1324-1336.
221. Zheng H, Guo Z, Ma Q, Jia H, & Dang G (2004) Cbfa1/osf2 transduced bone marrow stromal cells facilitate bone formation in vitro and in vivo. Calcif Tissue Int 74(2):194-203.
222. Hagmann S, et al. (2013) FGF-2 addition during expansion of human bone marrow-derived stromal cells alters MSC surface marker distribution and chondrogenic differentiation potential. Cell Prolif 46(4):396-407.
223. Bodle JC, et al. (2014) Age-related effects on the potency of human adipose-derived stem cells: creation and evaluation of superlots and implications for musculoskeletal tissue engineering applications. Tissue Eng Part C Methods 20(12):972-983.
224. Boskey AL, et al. (1996) The mechanism of beta-glycerophosphate action in mineralizing chick limb-bud mesenchymal cell cultures. J Bone Miner Res 11(11):1694-1702.
225. Coe MR, Summers TA, Parsons SJ, Boskey AL, & Balian G (1992) Matrix mineralization in hypertrophic chondrocyte cultures. Beta glycerophosphate increases type X collagen messenger RNA and the specific activity of pp60c-src kinase. Bone Miner 18(2):91-106.
226. Moutos FT & Guilak F (2010) Functional properties of cell-seeded three- dimensionally woven poly(epsilon-caprolactone) scaffolds for cartilage tissue engineering. Tissue Eng Part A 16(4):1291-1301.
227. Moffat J, et al. (2006) A lentiviral RNAi library for human and mouse genes applied to an arrayed viral high-content screen. Cell 124(6):1283-1298.
228. Wiznerowicz M & Trono D (2003) Conditional suppression of cellular genes: lentivirus vector-mediated drug-inducible RNA interference. J Virol 77(16):8957-8961.
229. Diekman BO, et al. (2015) Knockdown of the cell cycle inhibitor p21 enhances cartilage formation by induced pluripotent stem cells. Tissue Eng Part A 21(7-8):1261-1274.
182
230. Salmon P & Trono D (2006) Production and titration of lentiviral vectors. Curr Protoc Neurosci Chapter 4:Unit 4 21.
231. Diekman BO, et al. (2013) Intra-articular delivery of purified mesenchymal stem cells from C57BL/6 or MRL/MpJ superhealer mice prevents posttraumatic arthritis. Cell Transplant 22(8):1395-1408.
232. Farndale RW, Buttle DJ, & Barrett AJ (1986) Improved quantitation and discrimination of sulphated glycosaminoglycans by use of dimethylmethylene blue. Biochim Biophys Acta 883(2):173-177.
233. Estes BT, Diekman BO, Gimble JM, & Guilak F (2010) Isolation of adipose-derived stem cells and their induction to a chondrogenic phenotype. Nat Protoc 5(7):1294-1311.
234. Livak KJ & Schmittgen TD (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.
235. Maeda S, Hayashi M, Komiya S, Imamura T, & Miyazono K (2004) Endogenous TGF-beta signaling suppresses maturation of osteoblastic mesenchymal cells. EMBO J 23(3):552-563.
236. Mohammad KS, et al. (2009) Pharmacologic inhibition of the TGF-beta type I receptor kinase has anabolic and anti-catabolic effects on bone. PLoS One 4(4):e5275.
237. Boskey AL (2013) Bone composition: relationship to bone fragility and antiosteoporotic drug effects. Bonekey Rep 2:447.
238. Schmid TM, Bonen DK, Luchene L, & Linsenmayer TF (1991) Late events in chondrocyte differentiation: hypertrophy, type X collagen synthesis and matrix calcification. In Vivo 5(5):533-540.
239. Phillips JE, Burns KL, Le Doux JM, Guldberg RE, & Garcia AJ (2008) Engineering graded tissue interfaces. Proc Natl Acad Sci U S A 105(34):12170-12175.
240. Nguyen US, et al. (2011) Increasing prevalence of knee pain and symptomatic knee osteoarthritis: survey and cohort data. Ann Intern Med 155(11):725-732.
241. Ahmed TA & Hincke MT (2014) Mesenchymal stem cell-based tissue engineering strategies for repair of articular cartilage. Histol Histopathol 29(6):669-689. 183
242. Caplan AI (1991) Mesenchymal stem cells. J Orthop Res 9(5):641-650.
243. Caplan AI (2008) All MSCs are pericytes? Cell Stem Cell 3(3):229-230.
244. Crisan M, et al. (2008) A perivascular origin for mesenchymal stem cells in multiple human organs. Cell Stem Cell 3(3):301-313.
245. Kwong FN, Richardson SM, & Evans CH (2008) Chordin knockdown enhances the osteogenic differentiation of human mesenchymal stem cells. Arthritis Res Ther 10(3):R65.
246. Meinel L, et al. (2006) Osteogenesis by human mesenchymal stem cells cultured on silk biomaterials: comparison of adenovirus mediated gene transfer and protein delivery of BMP-2. Biomaterials 27(28):4993-5002.
247. Baker BM, Nathan AS, Gee AO, & Mauck RL (2010) The influence of an aligned nanofibrous topography on human mesenchymal stem cell fibrochondrogenesis. Biomaterials 31(24):6190-6200.
248. Bian L, Zhai DY, Mauck RL, & Burdick JA (2011) Coculture of human mesenchymal stem cells and articular chondrocytes reduces hypertrophy and enhances functional properties of engineered cartilage. Tissue Eng Part A 17(7-8):1137-1145.
249. Bian L, Guvendiren M, Mauck RL, & Burdick JA (2013) Hydrogels that mimic developmentally relevant matrix and N-cadherin interactions enhance MSC chondrogenesis. Proc Natl Acad Sci U S A 110(25):10117- 10122.
250. Li WJ, Tuli R, Huang X, Laquerriere P, & Tuan RS (2005) Multilineage differentiation of human mesenchymal stem cells in a three-dimensional nanofibrous scaffold. Biomaterials 26(25):5158-5166.
251. Kuo CK & Tuan RS (2008) Mechanoactive tenogenic differentiation of human mesenchymal stem cells. Tissue Eng Part A 14(10):1615-1627.
252. Awad HA, et al. (2000) In vitro characterization of mesenchymal stem cell- seeded collagen scaffolds for tendon repair: effects of initial seeding density on contraction kinetics. J Biomed Mater Res 51(2):233-240.
253. Young RG, et al. (1998) Use of mesenchymal stem cells in a collagen matrix for Achilles tendon repair. J Orthop Res 16(4):406-413.
184
254. Bhaskar B, Mekala NK, Baadhe RR, & Rao PS (2014) Role of signaling pathways in mesenchymal stem cell differentiation. Curr Stem Cell Res Ther 9(6):508-512.
255. Yang Y, et al. (2014) Gene co-expression network analysis reveals common system-level properties of prognostic genes across cancer types. Nat Commun 5:3231.
256. Somoza RA, et al. (2017) Transcriptome-Wide Analyses of Human Neonatal Articular Cartilage and Human Mesenchymal Stem Cell-Derived Cartilage Provide a New Molecular Target for Evaluating Engineered Cartilage. Tissue Eng Part A.
257. Wang Z, Gerstein M, & Snyder M (2009) RNA-Seq: a revolutionary tool for transcriptomics. Nat Rev Genet 10(1):57-63.
258. Dobin A, et al. (2013) STAR: ultrafast universal RNA-seq aligner. Bioinformatics 29(1):15-21.
259. Anders S, Pyl PT, & Huber W (2015) HTSeq--a Python framework to work with high-throughput sequencing data. Bioinformatics 31(2):166-169.
260. Anders S & Huber W (2010) Differential expression analysis for sequence count data. Genome Biol 11(10):R106.
261. Huang da W, Sherman BT, & Lempicki RA (2009) Bioinformatics enrichment tools: paths toward the comprehensive functional analysis of large gene lists. Nucleic Acids Res 37(1):1-13.
262. Huang da W, Sherman BT, & Lempicki RA (2009) Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nat Protoc 4(1):44-57.
263. Khan A, et al. (2018) JASPAR 2018: update of the open-access database of transcription factor binding profiles and its web framework. Nucleic Acids Res 46(D1):D260-D266.
264. Shannon P, et al. (2003) Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res 13(11):2498- 2504.
265. J. CWCJAJJXYM (2017) shiny: Web Application Framework for R. . R package version 1.0.5.
185
266. Langfelder P & Horvath S (2008) WGCNA: an R package for weighted correlation network analysis. BMC Bioinformatics 9:559.
267. Langfelder P & Horvath S (2012) Fast R Functions for Robust Correlations and Hierarchical Clustering. J Stat Softw 46(11).
268. Jeong H, Mason SP, Barabasi AL, & Oltvai ZN (2001) Lethality and centrality in protein networks. Nature 411(6833):41-42.
269. Doering TA, et al. (2012) Network analysis reveals centrally connected genes and pathways involved in CD8+ T cell exhaustion versus memory. Immunity 37(6):1130-1144.
270. Ikeda T, et al. (2004) The combination of SOX5, SOX6, and SOX9 (the SOX trio) provides signals sufficient for induction of permanent cartilage. Arthritis Rheum 50(11):3561-3573.
271. Lefebvre V, Li P, & de Crombrugghe B (1998) A new long form of Sox5 (L- Sox5), Sox6 and Sox9 are coexpressed in chondrogenesis and cooperatively activate the type II collagen gene. EMBO J 17(19):5718- 5733.
272. Lefebvre V, Huang W, Harley VR, Goodfellow PN, & de Crombrugghe B (1997) SOX9 is a potent activator of the chondrocyte-specific enhancer of the pro alpha1(II) collagen gene. Mol Cell Biol 17(4):2336-2346.
273. Ionescu A, et al. (2012) FoxA family members are crucial regulators of the hypertrophic chondrocyte differentiation program. Dev Cell 22(5):927-939.
274. Wagner EF (2002) Functions of AP1 (Fos/Jun) in bone development. Ann Rheum Dis 61 Suppl 2:ii40-42.
275. Williams JA, et al. (2009) Retinoic acid receptors are required for skeletal growth, matrix homeostasis and growth plate function in postnatal mouse. Dev Biol 328(2):315-327.
276. van den Akker E, et al. (2001) Axial skeletal patterning in mice lacking all paralogous group 8 Hox genes. Development 128(10):1911-1921.
277. Wang X (2016) Improving microRNA target prediction by modeling with unambiguously identified microRNA-target pairs from CLIP-ligation studies. Bioinformatics 32(9):1316-1322.
186
278. Wong N & Wang X (2015) miRDB: an online resource for microRNA target prediction and functional annotations. Nucleic Acids Res 43(Database issue):D146-152.
279. Sims D, Sudbery I, Ilott NE, Heger A, & Ponting CP (2014) Sequencing depth and coverage: key considerations in genomic analyses. Nat Rev Genet 15(2):121-132.
280. Mueller MB & Tuan RS (2008) Functional characterization of hypertrophy in chondrogenesis of human mesenchymal stem cells. Arthritis Rheum 58(5):1377-1388.
281. Mueller MB, et al. (2010) Hypertrophy in mesenchymal stem cell chondrogenesis: effect of TGF-beta isoforms and chondrogenic conditioning. Cells Tissues Organs 192(3):158-166.
282. Sekiya I, Vuoristo JT, Larson BL, & Prockop DJ (2002) In vitro cartilage formation by human adult stem cells from bone marrow stroma defines the sequence of cellular and molecular events during chondrogenesis. Proc Natl Acad Sci U S A 99(7):4397-4402.
283. Hall BK (2015) Bones and Cartilage (Academic Press) 2 Ed p 920.
284. Posel C, et al. (2012) Density gradient centrifugation compromises bone marrow mononuclear cell yield. PLoS One 7(12):e50293.
285. Mareschi K, et al. (2012) Multipotent mesenchymal stromal stem cell expansion by plating whole bone marrow at a low cellular density: a more advantageous method for clinical use. Stem Cells Int 2012:920581.
286. Sacchetti B, et al. (2016) No Identical "Mesenchymal Stem Cells" at Different Times and Sites: Human Committed Progenitors of Distinct Origin and Differentiation Potential Are Incorporated as Adventitial Cells in Microvessels. Stem Cell Reports 6(6):897-913.
287. Barry F, Boynton R, Murphy M, Haynesworth S, & Zaia J (2001) The SH-3 and SH-4 antibodies recognize distinct epitopes on CD73 from human mesenchymal stem cells. Biochem Biophys Res Commun 289(2):519-524.
288. Barry FP, Boynton RE, Haynesworth S, Murphy JM, & Zaia J (1999) The monoclonal antibody SH-2, raised against human mesenchymal stem cells, recognizes an epitope on endoglin (CD105). Biochem Biophys Res Commun 265(1):134-139.
187
289. Bara JJ, Richards RG, Alini M, & Stoddart MJ (2014) Concise review: Bone marrow-derived mesenchymal stem cells change phenotype following in vitro culture: implications for basic research and the clinic. Stem Cells 32(7):1713-1723.
290. Zhao W, et al. (2010) Weighted gene coexpression network analysis: state of the art. J Biopharm Stat 20(2):281-300.
291. Lee HK, Hsu AK, Sajdak J, Qin J, & Pavlidis P (2004) Coexpression analysis of human genes across many microarray data sets. Genome Res 14(6):1085-1094.
292. Mansson R, et al. (2004) Pearson correlation analysis of microarray data allows for the identification of genetic targets for early B-cell factor. J Biol Chem 279(17):17905-17913.
293. Lee W, Guilak F, & Liedtke W (2017) Role of Piezo Channels in Joint Health and Injury. Curr Top Membr 79:263-273.
294. O'Conor CJ, Leddy HA, Benefield HC, Liedtke WB, & Guilak F (2014) TRPV4-mediated mechanotransduction regulates the metabolic response of chondrocytes to dynamic loading. Proc Natl Acad Sci U S A 111(4):1316-1321.
295. Phan MN, et al. (2009) Functional characterization of TRPV4 as an osmotically sensitive ion channel in porcine articular chondrocytes. Arthritis Rheum 60(10):3028-3037.
296. Lee W, et al. (2014) Synergy between Piezo1 and Piezo2 channels confers high-strain mechanosensitivity to articular cartilage. Proc Natl Acad Sci U S A 111(47):E5114-5122.
297. Ng JJ, et al. (2017) Recapitulation of physiological spatiotemporal signals promotes in vitro formation of phenotypically stable human articular cartilage. Proc Natl Acad Sci U S A 114(10):2556-2561.
298. Geisler S & Coller J (2013) RNA in unexpected places: long non-coding RNA functions in diverse cellular contexts. Nat Rev Mol Cell Biol 14(11):699-712.
299. Huynh NP, Anderson BA, Guilak F, & McAlinden A (2017) Emerging roles for long noncoding RNAs in skeletal biology and disease. Connect Tissue Res 58(1):116-141.
188
300. Backesjo CM, Li Y, Lindgren U, & Haldosen LA (2009) Activation of Sirt1 decreases adipocyte formation during osteoblast differentiation of mesenchymal stem cells. Cells Tissues Organs 189(1-4):93-97.
301. Darnell DK, et al. (2006) MicroRNA expression during chick embryo development. Dev Dyn 235(11):3156-3165.
302. King BL & Yin VP (2016) A Conserved MicroRNA Regulatory Circuit Is Differentially Controlled during Limb/Appendage Regeneration. PLoS One 11(6):e0157106.
303. Yang B, et al. (2011) MicroRNA-145 regulates chondrogenic differentiation of mesenchymal stem cells by targeting Sox9. PLoS One 6(7):e21679.
304. Lander ES, et al. (2001) Initial sequencing and analysis of the human genome. Nature 409(6822):860-921.
305. Guttman M, et al. (2009) Chromatin signature reveals over a thousand highly conserved large non-coding RNAs in mammals. Nature 458(7235):223-227.
306. Barter MJ, et al. (2017) The long non-coding RNA ROCR contributes to SOX9 expression and chondrogenic differentiation of human mesenchymal stem cells. Development 144(24):4510-4521.
307. Hertzog PJ, Hwang SY, & Kola I (1994) Role of interferons in the regulation of cell proliferation, differentiation, and development. Mol Reprod Dev 39(2):226-232.
308. Brierley MM & Fish EN (2002) Review: IFN-alpha/beta receptor interactions to biologic outcomes: understanding the circuitry. J Interferon Cytokine Res 22(8):835-845.
309. Hu X & Ivashkiv LB (2009) Cross-regulation of signaling pathways by interferon-gamma: implications for immune responses and autoimmune diseases. Immunity 31(4):539-550.
310. Green DS, Young HA, & Valencia JC (2017) Current prospects of type II interferon gamma signaling and autoimmunity. J Biol Chem 292(34):13925-13933.
311. Dieudonne FX, et al. (2013) Promotion of osteoblast differentiation in mesenchymal cells through Cbl-mediated control of STAT5 activity. Stem Cells 31(7):1340-1349.
189
312. Rostovskaya M, et al. (2018) Clonal Analysis Delineates Transcriptional Programs of Osteogenic and Adipogenic Lineages of Adult Mouse Skeletal Progenitors. Stem Cell Reports 11(1):212-227.
313. Takayanagi H, et al. (2002) RANKL maintains bone homeostasis through c-Fos-dependent induction of interferon-beta. Nature 416(6882):744-749.
314. Takayanagi H, Kim S, & Taniguchi T (2002) Signaling crosstalk between RANKL and interferons in osteoclast differentiation. Arthritis Res 4 Suppl 3:S227-232.
315. Huynh NPT, Zhang B, & Guilak F (2018) High-depth transcriptomic profiling reveals the temporal gene signature of human mesenchymal stem cells during chondrogenesis. FASEB J:fj201800534R.
316. Barde I, et al. (2006) Efficient control of gene expression in the hematopoietic system using a single Tet-on inducible lentiviral vector. Mol Ther 13(2):382-390.
317. Kent WJ, et al. (2002) The human genome browser at UCSC. Genome Res 12(6):996-1006.
318. Doench JG, et al. (2016) Optimized sgRNA design to maximize activity and minimize off-target effects of CRISPR-Cas9. Nat Biotechnol 34(2):184-191.
319. Haeussler M, et al. (2016) Evaluation of off-target and on-target scoring algorithms and integration into the guide RNA selection tool CRISPOR. Genome Biol 17(1):148.
320. Kabadi AM, Ousterout DG, Hilton IB, & Gersbach CA (2014) Multiplex CRISPR/Cas9-based genome engineering from a single lentiviral vector. Nucleic Acids Res 42(19):e147.
321. Ohba S, He X, Hojo H, & McMahon AP (2015) Distinct Transcriptional Programs Underlie Sox9 Regulation of the Mammalian Chondrocyte. Cell Rep 12(2):229-243.
322. Rhesus Macaque Genome S, et al. (2007) Evolutionary and biomedical insights from the rhesus macaque genome. Science 316(5822):222-234.
323. Black JB, et al. (2016) Targeted Epigenetic Remodeling of Endogenous Loci by CRISPR/Cas9-Based Transcriptional Activators Directly Converts Fibroblasts to Neuronal Cells. Cell Stem Cell 19(3):406-414.
190
324. Ramos YF, et al. (2014) Genes involved in the osteoarthritis process identified through genome wide expression analysis in articular cartilage; the RAAK study. PLoS One 9(7):e103056.
325. Dunn SL, et al. (2016) Gene expression changes in damaged osteoarthritic cartilage identify a signature of non-chondrogenic and mechanical responses. Osteoarthritis Cartilage 24(8):1431-1440.
326. Washietl S, Kellis M, & Garber M (2014) Evolutionary dynamics and tissue specificity of human long noncoding RNAs in six mammals. Genome Res 24(4):616-628.
327. Awan HM, Shah A, Rashid F, & Shan G (2017) Primate-specific Long Non-coding RNAs and MicroRNAs. Genomics Proteomics Bioinformatics 15(3):187-195.
328. Sin O, Michels H, & Nollen EA (2014) Genetic screens in Caenorhabditis elegans models for neurodegenerative diseases. Biochim Biophys Acta 1842(10):1951-1959.
329. Valadkhan S & Gunawardane LS (2016) lncRNA-mediated regulation of the interferon response. Virus Res 212:127-136.
330. Gomez JA, et al. (2013) The NeST long ncRNA controls microbial susceptibility and epigenetic activation of the interferon-gamma locus. Cell 152(4):743-754.
331. Kambara H, et al. (2014) Negative regulation of the interferon response by an interferon-induced long non-coding RNA. Nucleic Acids Res 42(16):10668-10680.
332. Lengyel P (1993) Tumor-suppressor genes: news about the interferon connection. Proc Natl Acad Sci U S A 90(13):5893-5895.
333. Pitha PM (1990) Interferons: a new class of tumor suppressor genes? Cancer Cells 2(7):215-216.
334. Drasner K, Epstein CJ, & Epstein LB (1979) The antiproliferative effects of interferon on murine embryonic cells. Proc Soc Exp Biol Med 160(1):46- 49.
335. Cheng M, Nguyen MH, Fantuzzi G, & Koh TJ (2008) Endogenous interferon-gamma is required for efficient skeletal muscle regeneration. Am J Physiol Cell Physiol 294(5):C1183-1191.
191
336. Ishida Y, Kondo T, Takayasu T, Iwakura Y, & Mukaida N (2004) The essential involvement of cross-talk between IFN-gamma and TGF-beta in the skin wound-healing process. J Immunol 172(3):1848-1855.
337. Londhe P & Davie JK (2011) Gamma interferon modulates myogenesis through the major histocompatibility complex class II transactivator, CIITA. Mol Cell Biol 31(14):2854-2866.
338. Boissier MC, et al. (1995) Biphasic effect of interferon-gamma in murine collagen-induced arthritis. Eur J Immunol 25(5):1184-1190.
339. Cooper SM, Sriram S, & Ranges GE (1988) Suppression of murine collagen-induced arthritis with monoclonal anti-Ia antibodies and augmentation with IFN-gamma. J Immunol 141(6):1958-1962.
192
Biography
Nguyen P.T. Huynh was born and raised in Ho Chi Minh City, Vietnam.
Nguyen received a B.A. from Smith College in Biochemistry in May 2012. She then went on to join the Developmental and Stem Cell Biology training program, and subsequently the Cell Biology department at Duke University. Nguyen received a Ph.D. degree in Cell Biology in December 2018. Her publications include:
Huynh, N.P.T., Zhang, B.*, Guilak, F.* (2018) High-Depth Transcriptomic Profiling Reveals the Temporal Gene Signature of Mesenchymal Stem Cells During Chondrogenesis. FASEB J. Epub ahead of print. (*equal contribution)
Huynh, N.P.T., Brunger, J.M., Gloss, C.C., Moutos, F., Gersbach, C., Guilak, F. (2018) Genetic Engineering of Mesenchymal Stem Cells for Differential Matrix Deposition on 3D Woven Scaffolds. Tissue Engineering Part A. Epub ahead of print.
Rowland, C.R.*, Glass, K.A.*, Ettyreddy, A.R., Gloss, C.C., Matthews J.R.L., Huynh, N.P.T., Guilak, F. (2018) Regulation of Decellularized Tissue Remodeling Via Scaffold-Mediated Lentiviral Delivery in Anatomically-Shaped Osteochondral Constructs. Biomaterials. 177, 161-175. (*equal contribution)
Huynh, N.P.T.*, Anderson, B.A.*, Guilak, F., McAlinden, A. (2017) Emerging roles for long noncoding RNAs in skeletal biology and disease. Connective Tissue Research. 58(1), 116-141. (*equal contribution)
Brunger, J.M., Huynh, N.P.T.¸ Guenther, C.M., Perez-Pinera, P., Moutos, F.T., Sanchez-Adams, J., Gersbach, C.A., Guilak, F. (2014) Scaffold-mediated lentiviral transduction for functional tissue engineering of cartilage. Proceedings of the National Academy of Sciences. 111(9), E798-E806.
193