**4. Discussion**

Cell-based therapies have shown some degree of success in restoring damaged cartilage [30,31], but no study to date has described the presence of the proteoglycan perlecan in either the natural repair or cell therapy repair of cartilage in humans. Perlecan contributes to processes that are essential to the functioning of chondrocytes such as cell attachment, differentiation and production of extracellular matrix components [12,13], which makes it an ideal candidate molecule to assess in the formation of new cartilage. There is a longstanding biological paradigm that once damaged, articular cartilage cannot heal itself. However, evidence is now mounting to indicate that actually, to a limited extent, articular cartilage does have an innate ability to repair [2,32], although the mechanism and pathways are poorly defined. To our knowledge, this study is the first to assess and compare the differences in perlecan immunolocalisation in matched patient cartilage samples that were repaired either naturally or with autologous cells, while assessing the effects of heparanase on the phenotype of human chondrocytes in vitro.

The variety of tissue morphologies observed in the repair tissues, i.e., fibrous, hyaline, fibrocartilage and a mixture of the two, demonstrates the unpredictable and variable nature of cartilage repair. Some of these differences could be donor dependent, but since the repair tissues have been collected from two different sources (one from the harvest site and the other post-treatment with cell therapy), the repair could have been the result of two different biological mechanisms. Furthermore, the lack of an identifiable pattern of morphology in donor-matched natural and cell therapy repaired tissues could be due to differences in the microenvironment of the location where these defects were found. The high incidence of vascularisation present in the naturally repaired biopsies is of concern, as in its native state, cartilage is avascular. One could hypothesise that there may be a temporary invasion of blood vessels as a means of instigating the repair processes and over time with tissue remodelling and maturation, this vascularisation may disappear. Synovial infiltrates are often vascularised and usually associated with poor cartilage repair [33], but a recent study has provided evidence of the contribution of synovial cells in the repair of cartilage surface injuries in mice [34]. Adhesions identified by MRI (which are likely to be vascularised) have been shown to correlate with better histological features of cartilage repair twelve months after ACI [35].

Perlecan was immunolocalised in the pericellular matrix in healthy cartilage, which was is in line with previous findings [16]. In contrast, the immunolocalisation of perlecan in the repair tissue differed depending upon the type of tissue morphology present, for example, in areas of hyaline cartilage, perlecan appeared to have a more "normal" pericellular appearance whereas in areas of fibrocartilage, it was more associated within the interterritorial extracellular matrix. The latter appears to resemble the disposition of perlecan observed in foetal patella, femoral condyle and tibial plateau tissues [36]. This, combined with the fact that perlecan is a marker of early chondrogenic activity [37], suggests that embryological mechanisms could be contributing to the repair of damaged adult cartilage, either naturally or post-cell therapy. This is further evidenced by the observation that in cell therapy repaired samples, perlecan is associated with better tissue morphology and increased proteoglycan content, more resembling normal, healthy cartilage. (One slight caveat in comparing this immunolocalisation between healthy and surgical samples, however, is that there was some disparity in times between ex-vivo collection and processing; for healthy donors, time to fixation was ~24 h + 2–4 h but for surgical samples it was much quicker (2–4 h).)

Fibrocartilage commonly forms in repair sites following cell therapy, at least in biopsies obtained ~12 months post-treatment [25]. Whilst the aim of cell therapy in the treatment of cartilage defects is the formation of hyaline cartilage, there is evidence that the initial repair tissue which forms is remodelled [38] and does indeed mature towards hyaline cartilage with time post-treatment [39]. The distribution of perlecan seen in our study is perhaps further evidence of this, with the more diffuse and widespread location seen in fibrocartilage resembling that of developing or rudimentary cartilages, some of which

subsequently mature to form hyaline cartilage with its definite pericellular staining pattern. The strong vascular localisation of perlecan in the naturally repaired tissues is expected, confirming reports of its role in angiogenesis [40,41].

In healthy articular cartilage, collagen type III has a diffuse localisation in the territorial regions around the chondrocytes, i.e., beyond the pericellular capsule [28]. We found this pattern only in the repair tissues where some hyaline-like cartilage was present, suggesting more matured repair or regeneration. Collagen type III is often associated with collagen type I and is abundant in damaged tissues that are attempting to repair [28,42].

Collagen type VI is a microfibrillar collagen, accounting for approximately 1% of total collagen in adult articular cartilage [43]. Predominantly located in the pericellular matrix (PCM) in developing and mature cartilage, collagen type VI has been demonstrated to be integral for regulating chondrocyte swelling and contributing to the biomechanical integrity of the PCM; indeed, it also binds to the chondrocyte membrane via the RGD sequences [44–47]. During osteoarthritis, however, the localisation of collagen type VI changes to more interterritorial matrix expression, possibly reflecting increased degradation of the collagen fibrils [48,49]. The diffuse pattern of immunolocalisation of collagen type VI in the majority of the repair tissues tested in our study is similar to that of perlecan and indicates an immature PCM in regenerating cartilage. Despite our observations of collagen types III and VI immunlocalisation in repair tissue being similar to those found in OA, it is also possible that they are indicative of an immature, developing cartilage rather than degeneration. Both perlecan and collagen type VI have shown to be pivotal to the biomechanical function of the PCM [50]. As a result, and due to the ability of cartilage to detect and respond to mechanical loading, perlecan in particular could be an active participant in the loading-related aspects of cartilage repair and remodelling. Perlecan's role of "mechanosensing" in tissue maintenance has been demonstrated in bone [51,52], while its ability to influence the elastic modulus of the PCM has been proven in cartilage [50].

Given the unique glycolytic capability of heparanase, this enzyme has been proposed to be a valuable therapeutic target in repair biology [53,54]. The fragments released from the HS by the action of heparanase are often more bioactive than the native molecule [55,56]. For example, when heparanase cleaves HS from perlecan in the basement membrane it releases bound FGF2, which promotes angiogenesis, wound healing and tumour formation [57,58]. In our study, we tested the effects of heparanase on chondrocytes in terms of cell morphology, proliferation, and the expression of surface and genetic markers. Although no noticeable difference in cell morphology was noted, chondrocytes treated with heparanase showed higher proliferation compared to the untreated controls. This finding corroborates a previous study showing a heparanase-induced increase in proliferation and migration of the ATDC5 chondrocyte cell line [59], and supports the theory that the removal of HS encourages an increase in cell proliferation. Further investigations are needed to determine whether this stimulation of chondrocytes by heparanase is reproduced in vivo, and what the pathophysiological implications are, notably in the modulation of tissue repair. One should also consider the source/s of the HS that has been depleted, as perlecan is not the only HS-containing proteoglycan found in cartilage.

The flow cytometry analysis conducted in our experiment produced the first data looking at the effects of heparanase treatment on the expression of a comprehensive panel of surface markers in human chondrocytes. Exposing human chondrocytes to exogenous heparanase did not influence the expression of either surface stem cells markers (CD70, CD90 and CD105), or chondrogenic markers (CD44, CD151 and CD166). Interestingly, another study in mice MSCs has shown similarly that the inhibition of endogenous heparanase has no effect on these stem cells markers [60].

Of the five chondrocyte populations tested for their response to heparanase, three of them showed a marginal increase in FGFR3 as assessed by flow cytometry, while the gene expression of FGFR3 was stable. This is of particular interest in the context of cartilage repair, since signalling through the FGFR3 pathway is essential to chondrocyte function

during chondrogenesis. During the embryological development of cartilage rudiments, FGFR1c, FGFR2c, FGFR3c and perlecan are employed by mesenchymal cells to promote the production of extracellular matrix production [17,61,62]. FGF-18 has also been shown to signal through FGFR3 in the cartilaginous development of the human foetal spine [63]. Furthermore, a mouse knockout model revealed that the deletion of domain I in HS improved the symptoms of OA and preserved the expression of FGFR3 with disease progression [64]. We hypothesise that the positivity of FGFR3 on reparative cells in the de novo formation of cartilage could be an essential mediator of natural and CT repaired tissues. Additional work would, however, be needed to investigate this further.

Integrins are a family of cell adhesion receptors that are vital to the interactions between chondrocytes and the cartilage extracellular matrix, that is mediated through the binding of matrix components such as collagen types II and VI, vitronectin and fibronectin [65]. The heparanase treatment of chondrocytes in our study did not affect CD29, which is the β1 integrin subunit. CD29 couples with the α1 integrin subunit (CD49a) to form the α1β1 complex, and facilitates the binding of collagen types II and VI [66,67]. The reduced positivity of CD49a in four of the five cell populations treated with heparanase suggests a possible interaction between HS and the integrins that warrants further characterisation in cartilage repair. The heterogeneity of the expression of integrin subunits CD49b(α2), CD49c(α3) and the complex CD51/61(αV/β3) in response to heparanase, may be indicative of the versatility of chondrocytes when interacting with their pericellular environs and extracellular matrix. The marked lower levels in CD49a, CD49b and CD49c detected in chondrocytes from the oldest donor (donor 17) after heparanase treatment could reflect the age-induced decrease in integrins in cartilage that has been previously shown [68].

We tested the effects of heparanase on the expression of key chondrogenic genes and found no significant change in the expression of SOX9, collagen type II and aggrecan. This differed from a previous study that showed an increased gene expression for collagen type II and aggrecan after heparanase treatment, but this was using a more appropriate 3D culture system [59]; even in our monolayer system, heparanase had no inhibitory effect on chondrocytes. The marginal increase in expression of the hypertrophic genes for collagen type X and ALK-1 could be an indirect effect of the increased cell proliferation and is not conducive to the repair of hyaline articular cartilage. This observation should act as a reminder that the mechanisms triggered by the removal of HS would need to be controlled to avoid undesirable matrix formation [69].

The strong immunolocalisation of perlecan in chondrocyte cell clusters suggests that the pericellular matrix of these cells may still be intact, or at least being maintained, in some monolayer cultures with close cell contact. This finding confirms previous studies showing pronounced perlecan staining in clusters found in OA cartilage [70,71]. It was found that domain IV-3 of perlecan was responsible for chondrocyte clustering, by mediating a decrease in ERK1/2 signalling [72]. The presence of perlecan persisted despite after the assumed removal of HS in our study, which could indicate that the heparanase-induced response from chondrocytes is due to the loss of HS from perlecan, and not perlecan itself. Such observations were made in a mouse study where a *Hspg2* exon 3 null strain continued to produce perlecan without the native HS [13].

The present study is not without its limitations. For example, we acknowledge that a bigger sample size would make this study more robust; however, we are confident, based on our experiences, that the tissue morphologies presented here are in line with our previous observations. The naturally and cell therapy repaired tissues that we studied formed at different locations in the joint. This may have limited the direct comparison of the two cartilages, for instance due to differences in biomechanical forces betweeen different regions of the knee joint [73,74]. Regarding the in vitro cell experiments, chondrogenic differentiation may have provided additional insight into the effects of heparanase on chondrocyte function. It is also important to note that the enzymatic activity of heparanase is not specifically targeted to the HS on perlecan and that other HS proteins such as agrin, syndecan 1 and syndecan 4 may also be affected by heparanase [75,76]. This study does not identify an exact pathway or mechanism per se whereby perlecan influences cartilage repair, but it does indicate that it appears to be an integral player and so worthy of further investigation.
