lncRNAs: novel players in intervertebral disc degeneration and osteoarthritis

Abstract The term long non‐coding RNA (lncRNA) refers to a group of RNAs with length more than 200 nucleotides, limited protein‐coding potential, and having widespread biological functions, including regulation of transcriptional patterns and protein activity, formation of endogenous small interfering RNAs (siRNAs) and natural microRNA (miRNA) sponges. Intervertebral disc degeneration (IDD) and osteoarthritis (OA) are the most common chronic, prevalent and age‐related degenerative musculoskeletal disorders. Numbers of lncRNAs are differentially expressed in human degenerative nucleus pulposus tissue and OA cartilage. Moreover, some lncRNAs have been shown to be involved in multiple pathological processes during OA, including extracellular matrix (ECM) degradation, inflammatory responses, apoptosis and angiogenesis. In this review, we summarize current knowledge concerning lncRNAs, from their biogenesis, classification and biological functions to molecular mechanisms and therapeutic potential in IDD and OA.

ncRNAs, with less than 200 nucleotides, comprise several distinct types of ncRNAs, such as microRNAs (miRNAs), small-interfering RNAs (siRNAs) and Piwi-interacting RNAs (piRNAs). 7 Long non-coding RNAs (lncRNAs), earliest identified from cDNA in 1991, are defined as RNA>200 nucleotides in length without an open reading frame (ORF). 8,9 Current GENCODE annotation for humans (v25) lists over 15767 lncRNAs. Like miRNAs, lncRNAs are also thought to be major contributors to normal cellular physiological processes and their expression patterns are tissue-and cell-specific. 10 Although studies on lcnRNAs are still in their infancy, they have emerged as critical players in the onset and development of OA and IDD. 11,12 In this review, we overview the biogenesis, classification and functions of lncRNAs, and focus on their emerging pathological implications and therapeutic potential in IDD and OA.

| BIOGENESIS AND CLASSIFICATION OF LNCRNAS
The vast majority of lncRNAs are produced by the same transcriptional machinery as are other mRNAs, as evidenced by RNA polymerase II (Pol II) occupancy and histone modifications related to transcription initiation and elongation. 13 These lncRNAs possess a 5′ terminal methylguanosine cap and are usually spliced and polyadenylated. 14 In addition, alternate pathways have been found to promote the generation of some lncRNAs. 15 These pathways include a poorly characterized contingent of non-polyadenylated lncRNAs likely expressed from RNA polymerase III promoter, 10,16 and lncRNAs that are excised during splicing and snoRNA biogenesis. 17 Based on their locations and characteristics, lncRNAs can be classified into five subgroups ( Figure 1): (1) sense (when it overlaps with one or more exons of another transcript on the same strand), (2) antisense (when it overlaps with one or more exons of another transcript on the opposite strand), (3) bidirectional (when its transcription and a neighbouring coding transcript on the opposite strand is initiated in close genomic proximity), (4) intergenic (when it lies within the genomic interval between two genes) and (5) intronic (when it transcribes from an intron of a second transcript). 18 However, given the fact that one type of lncRNAs shares with the characteristics of other types of ones, this categorization still remains controversial. 19 A more accurate categorization of lncRNAs is required to investigate their functions.

| BIOLOGICAL FUNCTIONS OF LNCRNAS
As may be expected from their structural diversity, a wide range of biological functions have been described for lncRNAs. These functions primarily include gene transcription, epigenomic regulation, translation of protein-coding genes, RNA turnover, chromatin organization and genome defence. 20,21 It is worth noting that most of the functions described to date are implicated in regulation of gene expression, both of protein-coding genes and other non-coding RNAs, through a multitude of mechanisms of action as presented in Figure 2. 22 (1) Transcript of a lncRNA across the promoter region of a downstream protein-coding gene is able to directly interfere with RNA polymerase II recruitment, thereby inhibiting the expression of protein-coding gene. 23 (2) LncRNAs recruit chromatin remodelers to facilitate histone modifications at specific gene loci, which can activate the transcription of target genes. 24 (3) An antisense lncRNA can hybridize to overlap with a sense transcript, which blocks recognition of the splice sites by the spliceosome, thus leading to an alternatively spliced transcript. 25 (4) Alternatively, lncRNAs hybridize with the sense or antisense transcripts to form double-stranded RNAs (dsRNAs) that are subsequently cleaved by Dicer to produce endogenous siRNAs F I G U R E 1 A schematic presentation of ncRNA categories. NcRNAs contain a much larger portion of the human genome than protein-coding RNA, which comprise <3% of the genome. NcRNAs are classified into housekeeping ncRNAs, and regulatory ncRNAs that are subdivided into small ncRNAs and lncRNAs based on transcript size. LncRNAs can be designated as sense, antisense, bidirectional, intergenic or intronic according to their genomic location relative to that of nearby protein-coding genes (endo-siRNAs). 26 (5) LncRNAs interact with specific protein partners and then regulate their activity. 27 (6) Some lncRNAs can bind to the proteins, which forms the RNA-protein complexes for maintaining protein stability. 28 (7) In addition to modulating protein activity and serving as structural components, binding of lncRNAs to the proteins can affect their subcellular localization. 29 (8) A number of lncRNAs can be post-transcriptionally processed to generate small ncRNAs, such as miRNAs and piRNAs. 30 (9) LncRNAs act as competing endogenous RNAs (ceRNAs) or natural miRNA sponges. These ceRNAs communicate with, and co-regulate, each other by competing to bind to shared miRNAs, thereby titrating miRNA availability. [31][32][33][34] Understanding of the crosstalk between lncRNAs and miRNAs may provide novel insight into gene regulatory networks and have significant implications in human development and disease. Collectively, lncRNAs play key roles in a majority of cell molecular functions, from various aspects of transcription to translation. Although only a very small portion of known lncRNAs have been thoroughly characterized to date, future work will likely identify many more transcripts that fit into these and other functional paradigms.

| ROLE OF LNCRNAS IN IDD
The intervertebral disc (IVD) is a viscoelastic weight-bearing "cushion" and plays a crucial role in the maintaining flexibility and stability of the spine. It consists of three morphologically distinct regions: nucleus pulposus (NP), annulus fibrosus (AF) and cartilaginous end plates (CEPs). Its central region is the NP, which is surrounded by the AF laterally and CEPs inferiorly and superiorly. NP is a gelatinous matrix that is rich in type II collagen (Col II) and proteoglycans, especially aggrecan. 35 AF is a thick, dense structure, and it is divided into the outer and inner annulus. The extracellular matrix (ECM) in the outer annulus is dominated by type I collagen (Col I) and also contains relatively low amounts of proteoglycans; however, the inner annulus is made up of both Col I and Col II, with a higher proteoglycan content.
It is worth noting that the IVD is the largest avascular structure in the body, with nerve endings only reaching the inner annulus. 36 Because of these structural features, degeneration is apt to occur within the

IVD.
IDD is the most common cause of chronic low back pain, a public healthy question severely affecting quality of life. It constitutes the pathological foundation of most musculoskeletal disorders of the spine, including spinal stenosis, structural instability, disc herniation, radiculopathy and myelopathy. 1 The aetiology of IDD is currently attributed to the interaction between environmental and genetic factors ( Figure 3). Notably, environmental factors such as vibration, mechanical loading, physical activities and tobacco smoking are responsible for only a small portion of IDD. [37][38][39] In contrast, genetic heredity is the predominant risk factor for degenerative disc disease and is estimated to cause over 70% of cases. 40,41 Although the pathogenesis of IDD is not completely understood, it is well established that progressive loss of ECM, cell proliferation, inflammatory response, angiogenesis, apoptosis, autophagy, and oxidative stress play critical roles in the occurrence and development of disc degeneration. [42][43][44][45][46][47] Like the well-known short non-coding RNAs such as miRNAs, 1 49 Overexpression of FAF1 is known to potentiate Fas-mediated apoptosis and suppress the degradation of ubiquitinated proteins, leading to a significant increase of cell death. 50,51 These authors also found that in addition to upregulation of the nearby enhancer-like lncRNA, RP11-296A18.3, FAF1 is significantly increased in degenerative discs compared with the normal discs. 48 It has been reported that enhancer-like lncRNAs are able to activate the proximal promoter and then stimulate transcription of their nearby coding genes. 52 Conversely, knockout of enhancer-like lncRNAs dramatically inhibit the expression of their neighbouring protein-coding genes. 53 Thus, it is likely that upregulated RP11-296A18.3 induces overexpression of its nearby gene FAF1, which eventually facilitates excessive apoptosis of IVD cells. Further studies are required to test this possibility.
In a later study, a same lncRNA-mRNA microarray technique was used to analyse the levels of lncRNAs in human degenerative and normal NP tissues. 54  However, this speculation needs to be further validated by experimental evidence, and whether other differentially expressed lncRNAs are implicated in IDD is yet to be investigated.

| ROLE OF LNCRNAS IN OA
OA, also called osteoarthrosis deformans, is the most common form of arthritis worldwide. As a highly heterogeneous disease, it affects all synovial joints, including the hand, knee, hip and spine. The major pathological features of OA are characterized by the progressive degradation of the articular cartilage along with secondary bone remodelling and episodic synovitis. 56 There has been a large body of evidence supporting that ageing is the most important aetiological risk factor. 57,58 Other factors such as obesity, genetics and acute destabilising joint injuries also contribute to the development and progression of OA. [58][59][60][61][62] Although the underlying molecular mechanisms are not fully understood, it is generally known that ECM disruption, 63 In another study, in comparison with normal cartilage, 82 lncRNAs are significantly upregulated, but 70 are significantly downregulated in OA cartilage. 71 Collectively, there are a number of differently expressed lncRNAs in human OA cartilage, and these lncRNAs may be involved in the pathology of OA. Further functional investigation of single lncRNA is essential to confirm the association with OA and to explore novel potential targets for therapy.  78 It has been reported that HOTAIR expression is significantly increased in hepatocellular cancer, 79 81 and acute myeloid leukaemia. 82 Overexpression of HOTAIR markedly promotes MMP-9 synthesis in SiHa cells, a human cervical cancer cell line. 83 Similarly, HOTAIR is significantly upregulated in the synovial fluid of temporomandibular joint OA patients compared with that of normal controls. 84 Increased HOTAIR levels are also observed in the synovial fluid of temporomandibular joint OA rabbits as compared to control rabbits. 84 Furthermore, in chondrocytes isolated from the temporomandibular joint condylar cartilage of New Zealand white rabbits, interleukin (IL)-1β treatment dramatically enhances the expression of MMP-1, MMP-3 and MMP-9, whereas these effects are reversed by HOTAIR knockdown. 84 Taken together, these observations support the idea that HOTAIR is involved in upregulation of MMPs induced by IL-1β and also functions as a contributor to OA.

| lncRNAs regulate ECM degradation
H19 was the first lncRNA gene discovered, 85 and is paternally imprinted. In humans, H19 maps to chromosome 11p15.5, in close proximity to the reciprocally imprinted insulin-like growth factor 2 (IGF2) gene. 86 It transcribes a 2.3 kb non-coding RNA transcript by RNA polymerase II and splices to five exons. H19 is highly expressed throughout development of the embryo and foetus, and also to the large extent in the placenta, but it is shut down in the vast majority of tissues shortly after birth. In cultured human chondrocytes under chemically induced hypoxic condition, the expression of H19 and Col IIA1 (a subtype of Col II) is significantly increased, and a positive correlation exists between H19 and Col IIA1 levels. 87 H19 is known to serve as a precursor for miR-675. 88 Transfection with the miR-675 mimic also dramatically enhances Col IIA1 levels in human normal chondrocytes lacking H19. 89 These findings suggest a protective role for H19 in regulating the balance between matrix anabolism and catabolism within the articular cartilage.
HOTTIP resides in 5′ end of the HoxA cluster and encodes a ln-cRNA that is known to downregulate the expression of HoxA-13, a murine abdominal-B-type homeobox gene, one of the members in the Hox gene family, by directly inhibiting histone modifications. 90,91 Knockout of HoxA-13 has been shown to cause digital reduction. 92 Also, mis-expression of HoxA-13 in chick limb leads to homeotic transformation of the cartilage condensations and reduces cell death by altering cell adhesiveness in chick limb buds. 93 In human OA chondrocytes, HOTTIP expression is significantly increased, with concurrent downregulation of HoxA-13. 94 Moreover, introduction of HoxA-13 siRNA to human OA chondrocytes markedly decreases integrin-α1 levels. 94 Of important, overexpression of integrin-α1 subunit contributes to chondrogenesis, and mice lacking integrin-α1 develop cartilage degradation at a younger age and show increased MMP-2 synthesis. 95 Thus, HOTTIP may promote the degradation of cartilage ECM by suppressing the HoxA-13/integrin-α1/MMP-2 signalling pathway, which plays an important role in the pathogenesis of OA. Nevertheless, this speculation needs to be further proved by experimental evidence, and whether HOTTIP can affect other MMP production for ECM loss in OA is yet to be investigated.
In addition to lncRNA-CIR, HOTAIR, H19 and HOTTIP, GAS5 is associated with modulation of ECM anabolism and catabolism in chondrocytes. GAS5 was originally identified from a subtraction cDNA library and named after the finding that its expression levels increased upon growth arrest in mammalian cells. 96 It is located at 1q25 and contains 12 exons that are alternatively spliced to produce two possible mature lncRNAs (GAS5a and GAS5b) and 11 introns. These introns can encode 10 box C/D snoRNA. In spite of its short ORF, GAS5 does not possess protein-coding capacity and serves as a so-called host gene for snoRNAs. 97 In SK-Mel-110 cells, a human melanoma cell line, overexpression of GAS5 markedly downregulates the expression of MMP-2, leading to reduced migration and invasion ability. 98 Conversely, GAS5 levels are higher in human OA chondrocytes than those in normal chondrocytes, and its exogenous induction in human OA chondrocytes diminishes miR-21 levels and then increases the expression of MMP-2, MMP-3, MMP-9, MMP-13 and ADAMTS-4, with concomitant decrease in the Col II and aggrecan contents. 99 This discrepancy may be attributed to the differences in the types of cells, and cultural system.

| lncRNAs regulate the inflammatory response
Inflammation is increasingly being recognized as an important driver of OA cartilage pathology. 100 They found that their expression is significantly decreased in both knee and hip OA cartilage compared with non-OA cartilage. 106 In human chondrocyte TC28 cell line, knockdown of CILinc01 or CILinc02 dramatically enhances IL-1β-stimulated production of IL-6, IL-8, TNF-α, macrophage inflammatory protein (MIP-1β) and granulocyte-colony stimulating factor (G-CSF), suggesting a protective role of these two lncRNAs in controlling inflammation-driven cartilage degeneration in OA.

| LncRNAs regulate chondrocyte and synoviocyte apoptosis
Apoptosis, also called type I programmed cell death, is a highly regulated pathway that involves specific sets of intracellular signals and genes. Dysregulation of apoptosis results in pathological states, such as cancer, developmental anomalies and degenerative diseases. 107,108 During apoptosis, cells show morphological characteristics including cell shrinkage, plasma membrane blebbing, chromatin condensation, DNA fragmentation and apoptotic body formation. 109 Chondrocytes play a crucial role in maintaining articular integrity and physiology via the synthesis of ECM components to resist mechanical loads. It is well established that reduced chondrocyte number due to apoptosis is a critical cause leading to cartilage degeneration in the process of OA. 67,110 Interestingly, some studies have focused on the association of lncRNAs with chondrocyte apoptosis. Several lines of evidence have revealed GAS5 as an inducer of apoptosis in numerous human cancers, such as breast cancer, 111 hepatocellular carcinoma, 112 ovarian cancer 113 and prostate cancer. 114 Similarly, the exogenous induction of GAS5 in normal human chondrocytes decreases miR-21 levels and then stimulate apoptosis, leading to a severe degenerative morphology. 99 In primary rabbit condylar chondrocytes, knockdown of HOTAIR is also found to attenuate the apoptosis rate induced by IL-1β. 84 Therefore, inhibition of GAS5 or HOTAIR may have a potential therapeutic benefit for OA patients through the blockage of chondrocyte apoptosis.
Synovitis is one of the most important characteristics of OA, presenting with pathological features including hyperplasia of the synovial lining and inflammatory cell infiltration. 115 Activated synoviocytes secrete a number of proinflammatory cytokines such as IL-1β, TNF-α and IL-6, which block cartilage turnover and promote the produce of MMPs and cathepsins, resulting in the destruction of bone and cartilage. 116 In contradiction with chondrocytes, increased number of synoviocytes thus contributes to the development of OA. A recent experimental study by Kang et al. has demonstrated that prostate cancer gene expression marker 1 (PCGEM1), a lncRNA, is highly expressed in human OA synoviocytes. 117 Moreover, exogenous overexpression of PCGEM1 markedly inhibits apoptosis and stimulates the proliferation of human OA synoviocytes by directly binding to miR-770. 117 Thus, reduced synoviocyte apoptosis may be one of the mechanisms by which activation of PCGEM1 promotes OA progression.

| LncRNAs regulate the angiogenesis
Angiogenesis, defined as blood vessel outgrowth from pre-existing vasculature, is indispensable for growth and development, the reproductive cycle and tissue repair. 118 As articular cartilage and the inner two-thirds of the meniscus are normally avascular, they depend on oxygen and nutrients from adjacent synoviocytes and synovial blood vessels via the synovial fluid. 119,120 There has been a large body of evidence supporting the involvement of angiogenesis in OA development. [121][122][123] Angiogenesis is modulated by the equilibrium of proangiogenic and anti-angiogenic factors, which is regulated by the presence of either a facilitating or inhibitory ECM environment. Pathological neovascularization in OA represents a breakdown of these normal homeostatic mechanisms. Although the precise molecular pathways that regulate angiogenesis in the osteoarthritic joint are not still fully understood, proangiogenic factors produced within the osteoarthritic joint include prostaglandins, nitric oxide, regulatory peptides, cytokines, chemokines and growth factors, particularly vascular endothelial growth factor (VEGF). Anti-angiogenic factors contain protease inhibitors, matrix fragments and factors involved in the regression of inflammation. 123,124 Maternally expressed gene 3 (MEG3) is an imprinted gene that belongs to the imprinted delta-like 1 homologue (DLK1)-MEG3 locus located at chromosome 14q32.3 in humans. Its mouse orthologue, Meg3, also known as gene trap locus 2 (Gtl2), is localized to distal chromosome 12. The MEG3 gene encodes a lncRNA, which is expressed in many normal tissues. It has already been reported that VEGF expression and cortical microvessel formation are significantly increased in the brains of mouse Meg3-null embryos. 125,126 In agreement, downregulated MEG3 and upregulated VEGF are observed in articular cartilage samples from OA patients compared with normal controls. 127 Moreover, MEG3 levels are inversely correlated with VEGF levels. 127

| POSSIBLE THERAPEUTIC APPROACHES FOR TARGETING LNCRNAS IN IDD AND OA
The onset and development of IDD and OA has been regarded as a multifactorial and complex process. Currently, there is still a lack of effectively biological treatment in IDD and OA patients, highlighting the need to search for novel therapeutic approaches. As mentioned above, many lncRNAs are differently expressed in human degenerative IVD tissue and OA cartilage tissue and some lncRNAs have been shown to involve multiple pathological processes of OA development, suggesting lncRNAs as novel contributors to both diseases. Thus, lncRNA targeting therapy is expected to open a new hope for the management of IDD and OA.

| Silencing of lncRNAs
RNA interference (RNAi) is a process through which doublestranded RNA induces the activation of cellular pathways, leading to potent and selective silencing of genes with homology to the double strand. 129 Induction of RNAi through administration of siRNAs has been successfully used in treatment of hepatitis, viral infections and cancers. 130 Similar to other genes, lncRNAs are also able to be silenced by using specific siRNAs. For example, siRNA against HOTAIR appears to protect against OA development. 84 However, several obstacles need to be overcome in the application of RNAi to knockdown a lncRNA of interest for therapeutic purposes: (1) siRNAs often do not only silence their specific target genes but also interfere with the expression of other genes (off-target effects). 131 (2) The efficiency of a siRNA is not predictable, so that finding a good -efficient and specific -siRNA can be time and cost-intensive. (3) Some transcripts can be hard to target because of their strong secondary structure, incorporation into large protein complexes or their intracellular localization. (4) The siRNA-mediated knockdown is not permanent, making it unsuitable for long-term treatment.

| Antisense oligonucleotides
Antisense oligonucleotides (ASOs) are single-stranded nucleotides or nucleotide analogues, which predominantly act in the nucleus by selectively cleaving pre-mRNAs having complementary sites via an RNase H-dependent mechanism. 132 Although ASOs can also act by translation arrest, they are currently primarily used as 'GapmeRs', having a central region that supports RNase H activity flanked by chemically modified ends that enhance affinity and decrease susceptibility to nucleases. 133 Like siRNAs, ASOs are a viable approach to target lncRNAs for therapeutic purposes. In recent years, targeting lncR-NAs with ASO technology has been applied in cancer therapy. 134 For example, blockade of lncRNA MALAT1 (metastasis-associated lung adenocarcinoma transcript 1) using ASO has been shown to inhibit metastasis formation after lung cancer implantation. 135 It is anticipated that lncRNA ASOs could be potential candidates for treating IDD and OA. However, effective delivery of ASOs to their intracellular sites of action remains a major challenge. by GapmeRs markedly attenuates blood flow recovery and capillary density after hindlimb ischaemia. 137 Hence, this approach to target lncRNAs may also have potential therapeutic application in the treatment of IDD and OA.

| Small molecule inhibitors
Small molecule inhibitors are another therapeutic strategy. These inhibitors are designed to suppress lncRNA expression or to hide the binding sites for lncRNAs and thereby antagonize the interaction with specific partners. It has been reported that bromodomain-containing 4 (BRD4), a small molecule inhibitor, can downregulate HOTAIR expression and then inhibit glioblastoma tumour growth. 138 Another small molecule inhibitor that blocks the binding of HOTAIR to its partners is also found to protect against tumour cell proliferation, invasion and metastasis. 139 Given the fact that HOTAIR plays an important role in promoting OA progression, 84 in vivo delivery of small molecule inhibitors for HOTAIR may have a potential therapeutic benefit for OA patients.

| Zinc finger nucleases
Zinc finger nucleases (ZFNs) are genetically engineered proteins that contain a DNA-binding domain composed of at least three Cys 2 His 2 zinc fingers and a non-specific DNA cleavage domain derived from the restriction endonuclease FokI. 140 The zinc finger domains can be engineered to target a specific nucleotide sequence. The fused nuclease domain creates a DNA double-strand break at this specific site after dimerization, thus conferring a knockout-like effect on gene functions. Recently, ZFNs have been successfully employed to downregulate MALAT1 expression, with a 1000-fold reduction. 141 Thus, this approach is far superior to RNAi technique so as to knockdown a lncRNA.

| CONCLUSIONS AND FUTURE DIRECTIONS
Despite the investigations of lcnRNAs remaining in their infancy, they have been suggested as new contributors to IDD and OA. This provides novel insight into the pathogenesis of IDD and OA. With continued efforts, some dysregulated lncRNAs may be used as valuable diagnostic biomarkers and therapeutic targets. However, many challenges are still ahead. Thousands of lncRNAs have been shown to be differentially expressed in IDD and OA, whereas there is still a lack of the effects of single lncRNA on IDD and only nine lncR-NAs are proved to involve OA to date. Thus, it is important to elucidate the role of more other lncRNAs in IDD and OA development.
LncRNAs are known to act as natural miRNA sponges, and there has been a large body of evidence supporting the involvement of miR-NAs in IDD 1 and OA. [142][143][144] Future studies should also focus on the novel crosstalk between lncRNAs and miRNAs during IDD and OA.
Although in vitro studies may provide promising results with regard to lncRNAs involving both diseases, in vivo validation will be necessary. As discussed above, several approaches to target lncRNAs for therapeutic purposes can be considered once key disease-relevant contributions of these genes have been identified. A major challenge of all of these approaches is to accomplish target-specific delivery.
Recently, several novel delivery strategies have been developed to reduce off-target effects, especially nanoparticles that are characteristic by improved stability, extremely small size, biocompatibility and selfassembly. 145,146 Administration of HOTAIR siRNA by using magnetic nanoparticles has been shown to effectively inhibit the proliferation, invasion and in vivo tumourigenicity of human glioma stem cells. 147 The use of nanoparticles as effective delivery vehicles for lncRNAs is thus highly attractive and deserves further studies. As the pace of research in lncRNAs progresses, addressing these issues will provide opportunities for the development of novel therapeutic strategies based on targeting lncRNAs for IDD and OA.