Background
Osteoarthritis (OA) is the most common musculoskeletal disorder, characterized by chronic joint pain and substantial functional impairment [
1]. Although OA has historically not been considered an inflammatory condition, a growing body of evidence supports the involvement of synovial inflammation in the observed cartilage degradation and bone erosion [
2‐
6]. Proinflammatory cytokines are likely to be critical in driving such inflammation.
The collagenase-induced OA (CiOA) model involves the induction of joint instability by intra-articular injection of collagenase [
7,
8] leading to joint damage, including cartilage matrix breakdown [
9,
10], macrophage-mediated osteophyte formation [
11,
12], as well as pain [
8,
13], thereby mimicking features of the human disease [
14]. We have previously demonstrated that granulocyte macrophage-colony stimulating factor (GM-CSF) is a key mediator in the CiOA model [
8]. Both prophylactic, and notably therapeutic, blockade of GM-CSF using a neutralizing monoclonal antibody (mAb) have been shown to be effective at ameliorating CiOA-induced pain and disease [
8]. As a result there is a current phase II trial in hand OA using this approach [
15].
In addition to OA, GM-CSF has been implicated in the development of inflammatory pain and arthritic pain and disease [
16‐
19], and blockade of GM-CSF and its receptor are currently showing promise in rheumatoid arthritis (RA) trials [
15]. Regarding the mode of action of GM-CSF, we have recently reported that GM-CSF induces the chemokine, chemokine (c-c motif) ligand 17 (CCL17), via Jmjd3-regulated interferon regulatory factor 4 (IRF4), to mediate inflammation, and that blockade of CCL17 can ameliorate GM-CSF-dependent inflammatory pain and arthritic pain and disease [
18]. Furthermore, models in which tumour necrosis factor (TNF) is necessary can utilize this pathway [
20].
In the current study we provide evidence that the GM-CSF→Jmjd3→IRF4→CCL17 pathway, originally identified in human and murine monocytes/macrophages [
18], is required for CiOA pain and optimal arthritis development; however TNF is not involved.
Discussion
We have previously reported the importance of GM-CSF in the progression of pain and disease in the CiOA model [
8]. Here we demonstrate that IRF4, CCL17 and CCR4 are also required. Therapeutic inhibition of CCL17 or Jmjd3 was successful in ameliorating the already established arthritic pain and disease. Thus our previously proposed GM-CSF→Jmjd3→IRF4→CCL17 pathway, first identified in human and murine monocytes/macrophages, appears to be important not only in the context of inflammatory arthritis and pain (e.g. ZIA) [
18], but also in a model of OA, including in the development of the associated pain. However, TNF, which we have found before to be important for the initiation of ZIA pain and disease and mechanistically can use the same pathway leading to CCL17 formation via GM-CSF and JMJD3-regulated IRF4 formation [
20], was not required for the development of CiOA pain or disease. CiOA pain, first detected at 3 weeks, was rapidly reversed by treatment with a specific COX-2 inhibitor but there was no effect on histologic scoring or osteophyte size.
CiOA shares some features with human OA, such as the development of synovitis, cartilage erosion and osteophytes [
8,
11], which we have shown to be GM-CSF dependent [
8]; there are a number of studies using this macrophage-dependent model in rodents (see, for example [
8‐
13,
37,
38]). Interestingly, the proportion of synovial macrophages was not altered in the absence of GM-CSF, although there was a reduction in the proportion of neutrophils. A lack of IRF4 resulted in a slight reduction in early synovitis, with once again a reduction in the proportion of synovial neutrophils, while a lack of CCL17 had no effect; however, in the absence of either GM-CSF [
8], IRF4 or CCL17 there was no pain development at week 3 and significantly reduced histologic scores and osteophyte size at week 6 (Fig.
1a-c). Thus, the degree of early synovitis observed upon deletion of IRF4 and CCL17, in contrast to that observed upon deletion of GM-CSF, does not correlate with the subsequent pain levels and histologic changes. In line with these observations, in OA the degrees of joint pain and structural change do not always overlap [
5]. In the absence of IRF4 or CCL17, the activation states of the cells and the levels of associated inflammatory mediators present during synovitis may differ, which are likely important for the subsequent pathologic changes. In support of this mechanism, our gene expression analysis indicated that
Ccl17 mRNA is expressed in the CiOA synovial macrophages from WT mice, but not in those from
GM-CSF
−/−
and
IRF4
−/−
mice, data consistent with the involvement of the GM-CSF→IRF4→CCL17 pathway in the CiOA synovial macrophages, a cell type considered to be important in OA pathogenesis [
2‐
6,
33,
39].
Importantly, therapeutic blockade of CCL17 ameliorated both CiOA pain and disease. CCL17 was originally considered to be a M2 cytokine due to its preferential attraction of T
H2 lymphocytes [
25,
40]. It can be produced by certain macrophage/dendritic cell populations [
18,
23,
41,
42] and is elevated in many inflammatory conditions [
18,
43‐
45] and in synovial fluid in OA [
46]. CiOA synovitis in the absence of CCL17 suggests that CCL17 has other functions, apart from a chemotactic role [
18]. In this connection, we have also observed in models of inflammatory arthritis, including lymphocyte-independent models, that a lack of CCL17 has more profound effects on cartilage damage and bone erosion than on cellular infiltration [
18]. We have reported that systemic administration of CCL17 can drive arthritic pain in an inflamed joint in a COX-dependent manner [
18]. Our data showing the therapeutic efficacy of a COX-2 inhibitor on CiOA pain (Fig.
5) is consistent with CCL17 being able to regulate joint eicosanoid levels in some manner. There are conflicting data as to whether the CCL17 receptor, CCR4, is expressed in neurons [
47‐
50] as such expression would indicate the possibility of their direct activation by CCL17. Human microglial cells have been reported to express CCR4 [
51], suggesting CCL17 could also act at this level in pain development. However, the blockade of CiOA pain by systemic anti-CCL17 mAb administration suggests a peripheral algesic action for CCL17, at least in this model.
The clinical syndrome of “OA” affects not only the composition, structure and function of articular cartilage but also the integrity of multiple joint tissues such as synovium, bone, etc., i.e. an appreciation has emerged that OA is a “whole joint” disease (see, for example, previous work [
2‐
6,
33]). Also, as adult articular cartilage is avascular and aneural, pathogenic changes to non-cartilaginous joint tissues are of particular interest in understanding the source of pain generation in OA [
33]. During OA progression, the synovial membrane is one source of proinflammatory and catabolic products, including matrix metalloproteinases (MMPs) and aggrecanases, which potentially contribute to articular matrix breakdown [
33]. The synovial cell gene expression analysis described here indicated that
Ccr4 mRNA was expressed in a number of cell types, including fibroblasts, in the CiOA model while only fibroblasts appeared to express MMP3 and MMP13 which have been shown to be important for macrophage-mediated cartilage breakdown in experimental OA, including in CiOA [
8,
52‐
54]. We found that, compared to WT mice, synovial fibroblasts from
GM-CSF
−/−
,
Irf4−/− and
Ccl17
E/E
mice all had reduced mRNA expression of these MMPs and these gene-deficient mice also had reduced joint destruction and osteophyte size at 6 weeks post CiOA induction. As one possible mechanism, it could be that macrophage-derived CCL17 activates directly CCR4-expressing fibroblasts, which in turn augment MMP3 and MMP13 expression, leading to joint damage - both MMPs are expressed in synovial tissue from patients with early symptomatic OA [
55]. Notably, CCL17-mediated CCR4 activation in other models is reported to up-regulate MMP13 [
56,
57]. Importantly, cells in both cartilage and bone are also likely to express MMPs [
58] and whether CCL17 acts directly on cartilage and bone requires further investigation, as does the relative contribution of the synovium, cartilage and bone to pain and joint destruction in this model.
Blockade of TNF has revolutionized RA treatment. We have recently shown that TNF can be linked to the GM-CSF→IRF4→CCL17 pathway, with the actions of TNF and GM-CSF being interdependent [
20]. Interestingly, TNF is not required for CiOA pain and disease. It is not the only proinflammatory cytokine capable of associating with GM-CSF biology nor does GM-CSF-initiated inflammation necessarily involve TNF. For example, IL-1 can induce GM-CSF in a number of cell populations [
59‐
61] and an IL-1-driven monoarticular arthritis model is GM-CSF dependent [
62]. However, IL-1 is also not required for CiOA development [
37]. How collagenase initiates the GM-CSF-dependent pathway in the CiOA model is unknown, as is the relationship of the pathway to other mediators important in this model, for example, the alarmins, S100A8 and S100A9 [
38]. Consistent with our findings, anti-TNF treatment in the destabilization of the medial meniscus model did not ameliorate late-stage OA pain [
63]. Some studies have assessed the benefit of anti-TNF therapies in patients with hand or knee OA with variable results [
4,
6,
64,
65]. Better patient stratification and large, randomized, placebo-controlled trials would appear to be needed [
4].
Conclusions
From the findings described, we have evidence consistent with the GM-CSF→Jmjd3→IRF4→CCL17 pathway being important in CiOA, regulating both pain and disease. Whether the pathway is relevant to other experimental OA models is being explored. As for GM-CSF [
8], therapeutic blockade of CCL17 was effective at ameliorating both pain and disease. As mentioned, low-grade inflammation is now being recognized as important in OA pathogenesis and progression [
2‐
6,
33,
66] and pain has the highest impact on its burden [
65]. Targeting GM-CSF or its receptor in RA is yielding promising results [
15] and, as a result of prior findings in the CiOA model [
8], a phase II trial in hand OA is currently underway [
15]; however, given the possible adverse side effects associated with GM-CSF/GM-CSFR blockade, such as pulmonary alveolar proteinosis and infections [
16,
67], targeting CCL17, which is elevated in the synovial fluid of patients with OA [
46] and which we have found to be downstream of GM-CSF [
18], may have some advantages for the treatment of not only inflammatory arthritis but also at least for some patients with OA.