MSCs are an appealing source for cell-based treatment of musculoskeletal diseases and injury. Ageing is associated with various altered cellular phenotypes. Furthermore, the regeneration potential of MSCs is reduced with increasing age and is correlated with changes in cellular functions [5, 24]. This study of chondrogenic, osteogenic and tenogenic constructs derived from young and old MSCs provides a comprehensive proteomic analysis of tissue constructs whilst concurrently enabling improved understanding of the age-related functional and biological variations, which may affect their applications to regenerative medicine. In addition, our approach enables common and tissue-specific pathways of musculoskeletal ageing in an in-vitro system to be identified.
Characterisation of the tissue constructs was undertaken using standard methods following chondrogenic and osteogenic differentiation. For tenogenic constructs we used histological staining with Masson’s Trichrome and TEM to ascertain the presence and organisation of a collagenous matrix together with collagen type I, THBS4  and SERPINEF1 gene expression. The latter two have recently been identified as being the most DE genes in tendon differentiation but with low expression in chondrogenic differentiation . There was heterogeneity in the response to differentiation of these markers which was not age related, similar to that in aggrecan expression in chondrogenic constructs. The conditions of differentiation may impact on results. Because differentiation involves culture with different factors, which change according to the method used, it is conceivable that there might an age-related change in response to alteration of these factors which could affect results. Further work is required in this area.
The proteomic profiles of the constructs demonstrated using 1D gels showed that there was no gross difference in the profiles within construct type with ageing. When these results were compared with the number of identified proteins there were a higher number of protein identifications within the osteogenic (2233 proteins) and chondrogenic (2226 proteins) constructs, indicating a more complex proteome in these tissues. This was similar to the number of proteins identified in MSCs (2347 proteins) . A Rapigest™-based protein extraction workflow was utilised for tenogenic (615 proteins) constructs whereas a guanidine-based extraction was used for chondrogenic and osteogenic constructs, because Rapigest™ provided superior results compared with guanidine for tenogenic constructs in terms of proteins identified. The difference in number of protein identifications was probably due to the altered cell to ECM ratios within constructs, evident from histological sections. One limitation of the study is that it cannot be ruled out that some of the proteome pathway differences evident could be due to the culture format or duration.
We used label-free quantification to identify age-related DE proteins within each construct type. For our initial analysis we filtered data using p and q values (false-discovery adjusted p values for multiple testing) in order to reduce the number of false positives . This produced 128 DE proteins in chondrogenic constructs, 207 DE proteins in tenogenic constructs but only four proteins for osteogenic constructs. This indicates that at this level of filtering the age of the MSC donor has little effect on the proteome of osteogenic 2D constructs, whilst the protein composition of tendon constructs is most affected by MSC donor age. This disparity with the other two construct types could be due to their differentiation in 3D. We used standard 3D construct differentiation techniques for chondrogenic constructs due to problems with dedifferentiation into monolayers . The pellet characteristics closely mimic cartilage . In tenogenic differentiation, uniaxial tension in 3D is a requirement for differentiation . For osteogenic constructs we used the standard 2D system because it is the best described method and we could foresee problems with the mass spectrometry compatibility of materials included in many of the 3D osteogenic systems . However, 2D techniques inadequately produce the in-vivo environment for stem cells established by extrinsic and intrinsic cell signalling affecting biological function and differentiation capacity over time . Therefore, the effect of MSC donor age on the 3D osteogenic construct proteome should be studied in future.
For the musculoskeletal constructs it is essential to demonstrate the cellular phenotype and tissue composition, especially the ECM molecules that play a structural role and that contribute to the resulting mechanical properties. Therefore we identified the compositional and age-related DE matrisomal proteins in constructs. A number of matrisomal proteins were shared between all constructs such as COMP, TIMP1, decorin and biglycan, whilst some were shared between some types such as TIMP3 between chondrogenic and tenogenic constructs. This demonstrates that, similar to native tissues, the constructs have contrasting ECM profiles [20, 47]. Furthermore, when DE matrisomal proteins were investigated some proteins again shared age-related changes (COL4A2, MXRA5, THBS1 and MMP14) in chondrogenic and tenogenic constructs. Others such as plasminogen in chondrogenic constructs, and lumican in tenogenic constructs, were construct distinct. Interestingly, in agreement with results from 1D gels, the tenogenic constructs contained the most ECM matrisomal proteins as a percentage of all proteins identified within the construct type. Our findings revealed that the age of the donor MSCs had distinct or similar effects on construct ECM, depending on the differentiation lineage. The consequence of these altered matrices on the mechanical competence of the constructs requires further work.
Gene ontology revealed that metabolic processes were overrepresented in tenogenic constructs. This was also evident using IPA, which identified an increase in glucose and protein metabolism (both identified as activated in ageing), the latter related specifically to protein expression, proteolysis, catabolism and anabolism. Protein metabolism was demonstrated, for example, by an increase in abundance in old tendon constructs of TIMP-1, TIMP-3, MMP-2 and MMP-14. Furthermore, this was validated by neopeptide analysis, an indicator of protein turnover [20, 30]. Neopeptides represent ECM fragments produced by tissue remodelling. We have previously shown that neopeptide expression is altered in normal tendon ageing and disease . The changes could be attributed to altered metabolism within the cells present in ageing tendon constructs. However, the increased abundance of proteases could also be due to release of intracellular proteases because of cell death. The DNA content of old tendon constructs was reduced despite all constructs being seeded at an equivalent rate at the start of the experiments. This could be due to cell death in older constructs or reduced proliferation capacity. However, because there was a concomitant increase in DE ECM proteins, these protein metabolism changes seem to be due to a dysregulation of protein metabolism in ageing tendon constructs. These findings may also help understand how the tendon undergoes physiological remodelling that is evident in ageing.
Gene ontology also identified that DE age-related proteins in chondrogenic constructs were higher for the cellular component ECM and extracellular region proteins compared with the other construct types, indicating that donor age affects matrix proteins of chondrogenic constructs the most. These age-related changes in the ECM could have important implications for the quality of engineered tissue.
There were protein changes in lipid metabolism-related proteins in chondrogenic constructs. In cartilage, lipids are a source of energy and are incorporated into structural components and signalling molecules. Chondrocytes express several proteins for fatty acid metabolism and cholesterol biosynthesis and these molecules are increased during chondrogenesis . Chondrocyte lipid peroxidation has been suggested to have a role in cartilage ageing . We identified age-related changes in proteins involved with LXR activation and cholesterol biosynthesis. Given previous findings, it would seem that MSC donor age has an impact on lipid metabolism which could affect their chondrogenic potential further given that our culture conditions were hypoxic.
A further interesting feature derived from pathway analysis was the demonstration of an age-related inflammatory response in chondrogenic constructs, a significant feature in chondrogenic constructs. Ageing MSCs are known to undergo inflammageing and tissue-engineered cartilage may be more predisposed compared with other tissue types, similar to native ageing cartilage . Because MSC-derived chondrocytes are a potential treatment for chondral lesions , the use of allogeneic MSCs from younger donors may be beneficial in treating older patients. These findings demonstrate that our use of young and old donor-derived MSCs to produce musculoskeletal constructs could be a useful model to study musculoskeletal ageing.
A principal age-related feature of osteogenic constructs was mitochondrial dysfunction, when reactive oxygen species (ROS)-mediated oxygen stress overpowers the antioxidant defence system. Oxidative damage affects replication and transcription of mitochondrial DNA, leading to a decline in mitochondrial function and enhanced ROS production with further damage to mitochondrial DNA. In all constructs we demonstrated age-related protein changes involved in cell death and survival, and the alterations in oxidative stress probably contribute to this. Our results imply that MSC-derived tissue engineering from older donors must focus on oxidative stress protection.
Pathway analysis revealed that actin cytoskeleton changes were common to all ageing constructs. Others have identified an age-affected alteration in cytoskeletal organisation in rat MSCs . In all ageing constructs, similar to MSCs  we hypothesise that there is a decline in responsiveness to mechanical and biological signals due to a less dynamic cytoskeleton. The age-related network of proteins involved in cell migration and movement were also affected in all construct types. In total these findings are consistent with altered cytoskeletal dynamics affecting cell movement through coupling to actin organisation and turnover .
A number of upstream regulators of DE proteins were identified in each construct type. Our analysis revealed a number of significant transcriptional regulators potentially responsible for the protein changes. These data provide a starting point for future studies in MSC-derived tissue engineering of musculoskeletal constructs from older patients. One interesting finding was the contrasting roles of TGFβ and HIF1α in the DE proteins from chondrogenic and tenogenic constructs. TGFβ was significantly predicted to affect protein changes relating to tissue development in chondrogenic and tenogenic constructs and relating to differentiation in chondrogenic constructs, but in opposite directions; inhibited in chondrogenic but activated in tenogenic. Whilst there is only indirect evidence for a role of TGFβ in ageing, it is an important growth factor in development and differentiation. In our study, TGFβ signalling was predicted to be inhibited in chondrogenic constructs and activated in tenogenic constructs. This could be due to differing responses to oxidative stress previously identified here or distinctive requirements for culture conditions dependent on construct type. Furthermore, expression of the sets of age and tissue-specific transcriptional regulators may explain these findings, leading to similar molecular scenarios for some pathways (antioxidant, cell survival and cytoskeleton) but contrasting in others (e.g. protein and energy metabolism in tenogenic constructs and lipid metabolism in chondrogenic constructs).
Finally, the functions of the DE proteins identified which relate to cell death and survival, and antioxidant and cytoskeletal changes, are associated and important for chondrogenic, osteogenic and tenogenic differentiation. Chondrogenesis is characterised by changes in cell shape  and actin organisation is essential . In tenogenesis, cytoskeletal organisation is also paramount . Furthermore, osteogenesis is tightly regulated by ROS (reviewed in ). Age-related proteomic changes will thus affect the ability and quality of tissue-engineered constructs.
The study of musculoskeletal ageing in bone, cartilage and tendon is generally undertaken in isolation and it is often difficult to attain aged matched tissue samples in humans. We propose our approach as a model for musculoskeletal ageing that could be probed further to identify factors that may aid in recapitulation of a younger tissue phenotype. This is because as musculoskeletal tissues age they become more prone to age-related musculoskeletal disease such as osteoarthritis, tendinopathy and osteoporosis. We have identified some shared age-related characteristics (inflammageing, oxidative stress, cytoskeletal) which raise the prospects that common therapeutic targets could be developed to prevent these diseases. Understanding what drives these changes in diverse tissues could lead to the development of new therapeutic methods, which are advantageous to the musculoskeletal system in general.