Transcriptomic signatures in cartilage ageing

Ageing presents huge challenges for society because whilst the lifespan increases, the quality of life faced by individuals in old age is often poor [1]. The musculoskeletal system in particular is severely affected by the ageing process, with many tissues undergoing changes that lead to loss of function and frailty. Articular cartilage is susceptible to age-related diseases, such as osteoarthritis (OA), although it is not an inevitable result of ageing but rather a consequence of a complex inter-relationship between age and further predisposing factors such as obesity [2], injury [3], genetics [4] and anatomical configuration [5].

A number of studies have interrogated ageing cartilage in order to elucidate the underlying mechanisms that contribute to OA. An age-related reduction in response to insulin-like growth factor in rats resulted in a decline in synthetic activity [6]. Furthermore, using whole mouse joints, Loeser and colleagues demonstrated that there was a reduction in extracellular matrix (ECM) gene expression in older sham-operated mice following surgical destabilisation of the medial meniscus [7]. A characteristic of ageing articular cartilage is the reduction in the number of chondrocytes within the tissue [8, 9] and there is evidence of chondrocyte senescence [10]. Chondrocyte senescence is believed to be one cause of a decline in the ability of chondrocytes to respond to growth factors; resulting in the anabolic/catabolic imbalance evident in OA [11]. One of the consequences of cell senescence is an alteration in cell phenotype [12] characterised by increased production of cytokines and growth factors. The increase in ageing chondrocytes expressing this phenotype has been proposed to contribute to cartilage ageing and, given the rise in cytokine production in OA, could directly connect ageing to OA development [13]. Furthermore, there is evidence for the role of oxidative damage in cartilage ageing from reactive oxygen species [14, 15], which can result in damage to cartilage DNA [16], whilst a link between reactive oxygen species and development of OA has also been established [17]. Hence, the outcome of ageing on chondrocyte function is an inability to maintain homeostasis when stressed.

There is a need to examine and understand the processes and mechanisms involved specifically in cartilage ageing. Whilst some insights into cartilage ageing have been learnt from transcriptome profiling studies in ageing joints using microarrays [7], these data did not identify a specific chondrocyte phenotype associated with ageing alone. Limitations in coverage and sensitivity mean that a significant part of the chondrocyte ageing transcriptomic phenotype is as yet poorly defined. Advances in high-throughput sequencing methodologies are allowing a new approach to studying transcriptomes: massively parallel sequencing of short reads derived from mRNAs known as RNA-Seq [18]. Compared with microarray technologies, RNA-Seq is demonstrated to enable more accurate quantification of gene expression levels [19]. Furthermore, RNA-Seq is an effective approach for gene expression profiling in ageing tissues with a greater dynamic range and the ability to detect noncoding RNAs [20].

Here we examine the effect of ageing on gene expression in cartilage. Using RNA-Seq analysis of RNA extracted from whole cartilage of young and old equine donors, we elucidate the differential transcriptional signatures associated with ageing and identify some of the molecular mechanisms associated with these changes.

Ageing has an important role in the development of OA by making the joint more susceptible to OA risk factors. To provide interventions to prevent age-related changes and reduce the risk of developing OA, the underlying mechanisms involved in age-related changes of cartilage require elucidation. Characterisation of both young and old cartilage at the molecular level is essential for identifying the important signalling pathways in OA development. In the present study, we used the RNA-Seq technique to undertake deep transcriptome profiling of young and old cartilage. This is the first time that, to our knowledge, this technique has been used to interrogate transcriptional changes in cartilage ageing and, importantly, validation studies using RT-PCR demonstrated high correlation between methodologies and demonstrated reproducibility using a different donor set.

This study built on previous findings that identified a reduction in matrix gene expression with joint ageing [7]. We took a single tissue, articular cartilage, and undertook RNA-Seq in order to interrogate a greater range of genes for differential expression. Not surprisingly, our experiments identified that the age of the donor accounted for the principal variability in the data. The major findings of this study were as follows: the age-related gene expression changes identified were most notably involving reduced differential gene expression in older cartilage; there was an over-representation of genes with reduced expression relating to the ECM, degradative proteases, matrix synthetic enzymes, cytokines and growth factors in cartilage derived from older donors compared with young donors; cartilage ageing caused a decrease in many important Wnt signalling genes; IPA revealed that the top-scoring network for differentially expressed genes was from connective tissue disorders and connective tissue development; IPA also demonstrated significant canonical pathways for atherosclerosis signalling, prothrombin activation and rheumatoid arthritis; and there was differential expression of pseudogenes and small noncoding RNAs in cartilage ageing with increased expression of 12 pseudogenes and six noncoding RNAs in older cartilage, and of one pseudogene and nine small noncoding RNAs in younger cartilage.

Equine tissue was readily obtained, enabling collection of cartilage samples from macroscopically normal, skeletally mature young and aged horses. Importantly, the horse suffers clinical joint diseases similar to man (reviewed [36]), and as such has been used as a model for naturally occurring OA [37] due to extensive knowledge of its pathogenesis and clinical experience of the disease [38]. Indeed, the incidence of equine metacarpophalangeal OA in young racehorses [39] in training is similar to the incidence of post-traumatic OA in man [40]. Additionally, the articular cartilage thickness is also comparable between species [41].

For young horses one year is equivalent to about 3.5 years of a human [42, 43]. The rate of equine ageing relative to equivalent human age is greatest within the first two years of life and decreases after the horse reaches maturity at 4 years of age [44]. Hence, horses >15 years old, as used in this study, are likely to equate to humans older than 52 years. The average lifespan of a horse is 25 to 30 years and so it is possible that the obvious differences in lifespan may yield significant differences in the effect of ageing amongst animal species due to cumulative lifetime load. However, whilst the work in this study may not be directly applied to humans, it does enable an insight into human cartilage ageing by studying a population at skeletal maturity to one beyond the middle age equivalent in man.

This study utilised the entire articular surface of distal metacarpal III bone. High and low load-bearing cartilage was thus used. An assessment of macroscopic changes revealed no abnormalities in our samples. Previous studies indicated a high correlation between gross scoring and Mankin's grading in equine cartilage from the distal metacarpal III bone [28, 45]. To validate that the RNA extracted from the harvested tissue was articular cartilage, the expression level of several genes typically expressed and those of bone were measured. There was a high expression of articular cartilage genes only (data not shown).

Previous studies have identified a number of age-related changes in chondrocyte metabolism (reviewed in [46]). Most of these studies demonstrate changes at the protein level, such as an age-related decline in matrix production when equine chondrocytes were stimulated with TGFβ1 [47]. Others have provided evidence for a chondrocyte senescence secretory phenotype in ageing, demonstrated by an increase in cytokines [48, 49] along with matrix metalloproteinase (MMP) production and a reduction in growth factors [50, 51]. These studies did not interrogate transcript changes and of course simple deduction of protein from mRNA expression is insufficient because post-translational regulation, small noncoding RNAs, decay differences in mRNA and proteins, and locations or molecular associations of proteins affect overall protein levels [52]. However, a recent whole mouse-joint study demonstrated a reduction in matrix genes with age [7] in agreement with our findings. Furthermore, a study of equine articular cartilage concluded that although there was no change in the age-related expression of MMP-13 there was a reduction in MMP-3 and interleukin (IL)-1β gene expression in cartilage from older donors [53]. Annotations of genes at reduced levels in older samples included many relating to the ECM, degradative proteases, matrix synthetic enzymes, cytokines and growth factors. In contrast, within these annotations those at higher levels in older cartilage were very small: COLX, COLXXV, lubricin and fibroblast growth factor 9.

There appears to be an age-related failure of matrix, anabolic and catabolic cartilage factors. This is interesting because a recent study on postnatal and skeletally mature equine cartilage identified a reduction in collagens, matrix modelling and noncollagenous matrix transcripts with age [54]. ADAMTS-4 expression was reduced in the older cartilage in this study, which is in agreement with findings in ageing rat cartilage [55]. In contrast, previous studies have demonstrated an increase in IL-7 in ageing chondrocytes and in response to fibronectin fragments or IL-1 [49]. Although our experiment did not identify IL-7, interestingly one of the most downregulated genes identified in this study was the IL-7 receptor. A reduction in IL-7 receptor signalling in ageing β-progenitor cells has been demonstrated previously to result in ageing-like gene expression profiles [56]. Also, whereas other studies have demonstrated an increase in IL-1 [48] (where an increase in IL-1 protein was evident in older cultured human chondrocytes) and MMP-13 [48, 57] in ageing human cartilage, this study identified an age-related decline in their transcript abundance. However, one MMP-13 study looked at catabolic responsiveness with age whilst another used immunolocalisation of MMP-13 to identify protein. These two factors are not always related [58]. Whilst differences could also be attributed to our age classification of young and old and species distinctions, increased matrix enzymes (MMP-1, MMP-13) and cytokines such as IL-1, IL-8 and IL-11 identified in younger cartilage could be due to increased turnover. Interestingly a recent study identified that low innate capacity to produce IL-1β and IL-6 was associated with the absence of OA in old age [59]. The reduction in IL-1β evident in older cartilage may represent a protective mechanism against OA.

We noted in cartilage derived from old donors that there was primarily a reduction in the expression of some key Wnt signalling genes plus an increase in the Wnt antagonist DKK1 and a reduction in RUNX2, a downstream target of Wnt. Wnt signalling is active in adult cartilage, with deregulation being detrimental, resulting in age-associated joint pathologies due to excessive remodelling and degradation [60]. This signalling pathway has also been found to both regulate matrix synthesis in chondrocyte cell lines [61] and stimulate catabolic genes such as MMP-13 and ADAMTS-4 in chondrocytes [62]. A recent study demonstrated a potential protective function of Wnt in ageing. The activation of the Wnt pathway inhibited IL-1-mediated MMP-13 expression in human chondrocytes through the direct interaction between nuclear factor-κB and β-catenin [63]. One study has linked Wnt signalling with chondrocyte hypertrophy through RUNX2 activation [64], whilst elsewhere it was shown that DKK1 is a major player in the cessation of hypertrophic differentiation that can contribute to OA [65]. Interestingly, COL10A1, a marker of chondrocyte hypertrophy, was increased in old cartilage. However, COL10A1 has also been identified in the transitional zone of cartilage and may have a role in the modification of collagen fibril arrangement [66]. A recent study in mesenchymal stems cells derived from OA patients found that COL10A1 downregulation played a role in the establishment of a defective cartilage matrix in OA [67]. It would seem that this increased expression with ageing is not through the Wnt signalling interaction with subsequent RunX2 activation as described previously [64]. Further credence is given to this hypothesis by our findings that alkaline phosphatase expression, also regulated through RunX2, was downregulated in old cartilage. Overall Wnt signalling is involved in maintenance of cartilage, and the dysregulation event here in ageing may be an important episode. Interfering with the pathway may contribute to improvements in cartilage regeneration.

Using IPA, this study identified age-related changes in pathways and processes including connective tissue disorders and development in which a significant number of genes, regulated both strongly and subtly, were enriched. This is not remarkable given the number of matrix genes differentially identified in the study. Care should also be taken in overinterpretation of this finding because some of the genes in this network are minor components of cartilage, such as COL12A, COL16A, COL25A, LINGO and COCH. Canonical pathways identified as significantly affected by ageing, such as the role of osteoblasts and osteoclasts in rheumatoid arthritis, were not surprising. Interestingly, age-affected atherosclerosis signalling pathways follow the differential expression of a mixture of proteases and lipoproteins. In ageing cartilage, further studies to investigate the significance of this are clearly required.

One advantage for the use of RNA-Seq to undertake differential gene expression studies is that other sets of RNA molecules from the transcriptome can be identified, such as nonprotein coding RNAs (for example, miRNA and small nucleolar RNA (snoRNA)) that constitute a significant part of the transcriptome [68] as well as pseudogenes.

Pseudogenes provide a novel tier of gene regulation through the generation of endogenous silencing RNA or miRNA binding sites, which act as decoys for miRNAs [69]. Indeed some miRNAs have been demonstrated to target the genes [70]. It is hypothesised that pseudogenes act as post-transcriptional regulators of the corresponding parental gene [71]. Whilst possessing very similar sequences to their counterpart coding genes, they are unable to be transcribed due to mutation/deletion or insertion of nucleotides. Transcription of pseudogenes has tissue specificity and can be activated or reduced in disease, indicating a possible functional role in cells [72]. Interestingly, pseudogenes have been identified as increasing with age, such as pseudogene cyclin D₂ in the ovary [73]. Whilst this study identified the differential expression of pseudogenes in cartilage of different ages, it is not known whether these are functional or have relevance to cartilage ageing. Recent work by the Encyclopaedia of DNA Elements (ENCODE) Consortium identified that 8% of the pseudogenes in the human genome are functional [74], and so with the publication of GENCODE, a reference human genome annotation for The ENCODE Project [75], more light may be shed relating to the role of pseudogenes in cartilage ageing in the near future. Pseudogenes thus present an interesting area for future research in cartilage ageing and disease.

The methodology used here does not enrich for miRNAs. To increase the identifications of small miRNAs using RNA-Seq, specific techniques are used for their enrichment in conjunction with additional miRNA abundance quantification algorithms. A single miRNA, miR-21, was however identified as increased in ageing cartilage. miRNAs are short noncoding RNAs that regulate the translation [76] and/or degradation of target message [77]. miR-21 has been implicated in inflammation [78], cancers including osteosarcomas [79], and hypomethylation [80]. The role of miR-21 in cartilage is not fully elucidated, although a study in rats found that it promoted increased proliferation and matrix synthesis in chondrocytes embedded in atelocollagen gel [81]. However, our finding is interesting because epigenetic changes such as hypomethylation occur with ageing, a risk factor contributing to several age-related pathologies [82].

A further set of small noncoding RNAs, snoRNAs - a class of small guide RNAs found in the nucleolus - were also identified in the study. The snoRNAs direct chemical modification of other RNAs, and like miRNAs are emerging as important regulators of cellular function and disease development. There are two principle classes: the C/D box snoRNAs (SNORDs) and H/ACA box snoRNAs (SNORAs), which are associated with methylation and pseudouridylation of ribosomal and other RNAs. In addition, RNase MRP and RNaseP are the only members of a further special class of snoRNAs [83]. Both were significantly reduced in older cartilage in this study. Interestingly, mutations in RNase MRP cause cartilage hair hypoplasia in which patients display dwarfism [84]. In recent work, RNase MRP was identified as a regulator of chondrocyte hypertrophy, demonstrating functional cross talk with chondrogenic pathways [85]. snoRNAs fine-tune the ribosome to accommodate changing requirements for protein production during development, normal function and disease [86]. Indeed, control of snoRNA expression may play a pivotal role in the regulation of high protein-producing cells such as chondrocytes, as demonstrated by the phenotypes of ribosomopathies [87]. Whilst there are very few studies into the significance of snoRNAS in cartilage ageing or disease, a recent study proposed the use of serum snoRNA U38 and U48 as biomarkers of early cartilage damage. These snoRNAs was detected in serum following anterior cruciate ligament injury, but were not associated with normal ageing [88]. The snoRNA transcriptome signatures in ageing cartilage provide an interesting set of genes for further studies to determine their role in ageing.

A major strength of this study is that it represents the first application of RNA-Seq technology for transcriptomic studies in cartilage ageing. The study has increased our knowledge of transcriptional networks by providing a global view of the transcriptome. The molecular signatures described in this paper reflect a combination of degenerative processes and transcriptional responses to the process of ageing. This analysis further supports the use of next-generation sequencing as an ideal quantitative framework to study pathways and networks as an integrated system in order to understand the complex processes of cartilage ageing.