Major findings for Objective

The first objective of this chapter was to measure candidate molecules of bone and cartilage turnover, mechanical loading of bone and inflammation in synovial fluid (FCD compared to uOA) and serum (FCD compared to healthy controls) to identify suitable biomarkers and mechanisms of FCD pathogenesis.

Synovial fluid biomarkers are poorly reproduced in serum

Notably, out of the 14 biomarkers measured, only CTX-I, sclerostin and IL-13 significantly (p≤0.05) positively correlated across fluid types, indicating that several mechanisms of activity in the joint are poorly reflected systemically. Most non-significant biomarkers were either moderately (RANKL/OPG, glutamate and IL-10) or weakly (RANKL, OPG, ALP, IL-12p70) correlated, however in contrast pro-inflammatory cytokines (TNF-α, IL-6, IL-8) and IFN-γ had no positive association. This is likely due to the short half-lives of most cytokines and chemokines, which are required to act only on cells in close proximity to their release (Zhou et al., 2010). Furthermore, RANKL and OPG only have purpose in bone signalling, thus may exhibit similar properties for this reason. The lack of commonality for many candidates is a considerable factor when identifying appropriate biomarkers of discriminatory potential between FCD subjects and controls and must be taken into account to reduced false interpretation.

Loading and inflammation pathways in osteoblast-mediated osteoclastogenic activity

Synovial fluid OPG levels were significantly elevated in uOA group to FCD subjects, however this was concurrent with a trend of increased RANKL expression, leading to an overall higher RANKL:OPG ratio in uOA subjects. However, the overall group differences were not significant at the p≤0.05 level due to the considerable number of overlapping

182

values. The RANK:OPG ratio is a surrogate indicator of net osteoclast activation through the canonical osteoblast-mediated pathway, indicating that a larger proportion of uOA subjects may be exhibiting higher subchondral bone activated osteoclast numbers and increased bone resorption activity (Tat et al., 2008b). The increased RANKL:OPG ratio is likely resultant of the increased osteoblastic RANKL expression induced by higher levels of pro-inflammatory activity (Kim et al., 2017, Lam et al., 2000), which is evidenced by the significant positive association of IL-6 (and moderate trends of IL-8) with soluble RANKL found in the correlation analysis. On the other hand, OPG is regulated by mechanical loading of subchondral bone (Kim et al., 2006a, Kusumi et al., 2005), evidenced by the significantly elevated levels in uOA subjects who exhibit higher knee loading relative to FCD subjects (evidenced in chapter 4). Together, the findings suggest that mechanical loading and inflammation of subchondral bone have contradictory effects on canonical NF-κB-induced osteoclastogenesis, however subjects with increased knee loads accompanied by up-regulated pro-inflammatory activity in the joint may be exposed to higher RANKL/OPG ratios, and thus, increased recruitment of mature osteoclasts in subchondral bone.

Increased bone resorption is indicative of FCD presence, but not of disease state

The trend of increased RANKL:OPG signalling in the uOA group was not reflected in osteoclastic bone resorption activity represented by CTX-I levels (Garnero et al., 2003), which showed no clear trends in disease group analysis, but larger individual differences in the FCD group. Furthermore, CTX-I showed no clear trend with canonical osteoclastogenic signalling markers (RANKL, OPG and RANKL/OPG) in synovial fluid, and only weak trends in serum. In contrast, the moderate relationships between CTX-I and inflammatory mediators TNF-α and IL-10 in synovial fluid imply a larger influence of bone resorption activity through non-canonical pathways (Knowles and Athanasou, 2009, Sabokbar et al., 2016). Notably, synovial fluid CTX-I significantly (≤0.05) correlated with BMI, which is evident of a positive involvement of mechanical signalling pathways on osteoclast activity, despite the lack of associations to glutamate or sclerostin levels. Although bone resorption activity did not favour a disease group, there was a trend of declining bone formation marker ALP in uOA synovial fluids relative to FCD, which may signify a net shift towards a bone resorption-formation imbalance of resorption consistent with disease advancement. In contrast of the disease group analysis, there was increased bone resorption activity in FCD subjects relative to controls evidenced the significantly (p≤0.05) increased serum CTX-I levels, which justifies serum CTX-I as a suitable biomarker of FCD presence. However, in similar respects to the

183

disease group analysis, there were no clear favoured mechanisms of this increase when considering associations to other serum biomarkers.

Involvement of glutamatergic signalling in FCD pathogenesis

Group comparisons found significantly (p≤0.05) higher levels of glutamate in FCD serum relative to controls, implying an involvement of glutamatergic signalling in FCD pathogenesis. Glutamatergic signalling has been largely implicated in OA as it is functional in bone, cartilage, meniscus and synovium tissues (Wen et al., 2015) and is involved in mechanically regulated bone remodelling (Brakspear and Mason, 2012). Higher FCD serum levels may be resultant of aberrant exocytotic release and transporter activity that regulate extracellular glutamate in response to altered loading of subchondral bone in the FCD joint. The positive correlation of synovial fluid glutamate and ALP may be indicative of osteoblast ionotropic glutamate receptor (iGluR) activation responsible for increased Runx2 activity, a positive regulator of osteogenic gene expression (Hinoi et al., 2002, Ho et al., 2005). Activation of iGluRs on the surface of progenitor/mature osteoclasts up-regulates NF-κB-induced osteoclastogenesis (Merle et al., 2003) and activity (Mentaverri et al., 2003), though these mechanisms were not evident in correlations of synovial fluid glutamate and CTX-I levels or the RANKL:OPG ratio. Furthermore, experimental models of inflammatory arthritis have also revealed glutamatergic modulation of inflammatory and nociceptive pathways (Bonnet et al., 2015, Flood et al., 2007), however this was not evident in correlations with pro- inflammatory cytokines or KOOS scores. These incongruities may be related to the distinctive regulation of extracellular glutamate levels in the joint by exocytotic release mechanisms and elimination by glutamate transporters (Wen et al., 2015).

Elevated pro-/anti-inflammatory imbalance concurrent with disease progression

A substantial finding is the clear pattern of pro-inflammatory cytokine (IL-6, IL-8, TNF- α) activity in uOA joints compared to FCD joints, with significant increases in IL-6 and IL- 8 levels, coupled with the decline in anti-inflammatory mediator IL-10 and IL-13 activity. Collectively, these findings represent an increasingly pathogenic inflammatory regulation of the anabolic/catabolic imbalance in subjects representing increased disease severity and joint loading (uOA relative to FCD). This is a likely a critical mechanism for disease progression, since pro-inflammatory cytokines including TNF-α, IL-1β, IL-6 and IL-8 are of the most consistent and notable soluble biomarkers mediators associated with alterations in bone and cartilage degeneration in animal models of OA

184

(Legrand et al., 2017, Pelletier et al., 2001) and human OA studies (Findlay and Kuliwaba, 2016). This is linked their self-propagation, promotion of tissue turnover (osteoclastogenic/osteoblast-inhibiting) pathways, upregulation of proteolytic matrix degrading enzymes (MMPs, ADAMTS), and production of destructive and nociception- inducing molecules such as prostaglandins E2 (PGE2) and nitric oxide (Fernandes et al.,

2002, Lotz et al., 2013). Correlation analysis corroborated the self-stimulatory mechanism in which occurs, evidenced by the significant relationships of IL-8 with IL-6 and TNF-α, as well as IL-6 with TNF-α. However, only the association between IL-8 and TNF-α was retained in serum which is reflective of the low associations between joint inflammation and systemic levels.

Conversely, anti-inflammatory mediators such as IL-10 and IL-13 on the other hand play an antagonistic role to disease progression, through the potent inhibition of pro- inflammatory cytokine expression and osteo-/chondro-protective down-regulation of tissue turnover pathways (e.g. inhibition of canonical osteoclastogenesis pathway), and suppression of PGE2 production through the suppression of cyclooxygenase-2 (COX-2)

expression (Wojdasiewicz et al., 2014, Onoe et al., 1996, Scanzello, 2017). Additionally, IL-10 levels were significantly lower in FCD relative to control serum and found a moderate inter-fluid correlation, therefore may be a promising serum biomarker representing inflammatory dysregulation with its declining levels with tibiofemoral FCD progression towards an established OA state.

Nociceptive pathways in FCD pathogenesis

Notably, synovial fluid IL-13 and IFN-γ inversely correlated with KOOS pain, symptoms and function scores, implying they may be responsible for worsening clinical factors. Conversely, serum IFN-γ and IL-10 which positively associated with KOOS scores may be protective of the burden. The paradoxical opposing findings for anti-inflammatory cytokines IL-10 and IL-13, as well as for IFN-γ in serum compared to synovial fluid, indicates that they may act on differing pathways in the joint relative to systemically. This may be related to selective activation of peripheral or central sensitization pathways, however to date, these paradoxical effects are undocumented. As previously discussed, declining serum IL-10 levels are consistent with FCD presence (and progression in synovial fluid), therefore may also be useful as an indicator of increased pain induction pathway activation alongside patient-reported outcome measures, which are prone to response shift (lack of an internal perceptual reference of pain).

185

4.4.3 Major findings for Objective 2

The second objective of this chapter was to use multivariate analysis tools to identify inter- and intra-group variances that are indicative of disease subgroups, or ‘phenotypes’ of disease. PCA revealed three main clusters and three individuals that are representative of varying disease processes, which will be discussed with regards to their individual distinctive features.