We demonstrated above for the first time that GM-CSF is required for both pain and optimal disease development in a joint instability OA model. Furthermore, GM-CSF neutralization by a therapeutic mAb-based protocol rapidly and completely abolished existing arthritic pain. Experimental OA, induced by intra-articular collagenase injection as above, has been shown to share some important features with human OA, such as the development of osteophytes and cartilage erosion [
20,
21]. There is also a low grade inflammatory response (synovitis), which can also be seen in the human disease [
37]. The mild synovitis observed early in disease development was virtually absent when GM-CSF was not present. This is in agreement with previous studies using systemic inflammatory models of arthritis in which blockade or absence of GM-CSF led to a reduced inflammatory response and fewer macrophages in the joint [
8,
10]. Synovial macrophages have been shown to be important in this OA-like model, with both synovial inflammation and cartilage damage being ameliorated when they are selectively depleted from the synovial lining layer by clodronate liposomes [
21]. As GM-CSF can directly promote macrophage survival [
38], its neutralization may also lead to direct depletion of joint macrophages, in this case by apoptosis [
38]. In the current study, the degree of MMP-mediated cartilage damage was also lower in the absence of GM-CSF, suggesting that GM-CSF may be important for sustained and ongoing joint damage, possibly by limiting synovial macrophage activation in addition to their numbers [
2,
4,
38].
Osteophyte formation, which is a common feature of OA and which may be a repair mechanism to help stabilize joints [
39], is prevented by synovial macrophage depletion in the collagenase-induced model [
21]. An association between osteophytes and pain has been reported for hand OA [
40]. In the absence of GM-CSF, the size of the osteophytes formed was smaller than in the presence of GM-CSF. This may be related to the lack of pain seen in the absence of GM-CSF. It could also indicate some requirement of (GM-CSF dependent) synovitis for osteophyte development, as has been previously suggested [
41]; however, given there was still some osteophyte formation present in GM-CSF-/- mice, other factors may also be important, such as transforming growth factor β [
42]. The degree of cartilage damage seen at 6 weeks post collagenase injection in the absence of GM-CSF was significantly decreased compared with that in wild-type mice. Interestingly, antibody-mediated GM-CSF blockade both prevented (prophylactic treatment) and reversed completely (therapeutic treatment) the arthritic pain and modulated the degree of osteoarthritic changes in the joint, indicating that GM-CSF is important both in the early and late stages in this model, even though the features of the model change over time. Of note, the neutralization of GM-CSF abolished existing pain within 3 days, the earliest time point measured. The effectiveness of the therapeutic protocol was similar to the findings for disease progression in systemic inflammatory and autoimmune models [
7,
10]. In addition to its direct effects on macrophages, GM-CSF may also have direct effects on neuronal sensitization in the OA lesion, similar to that suggested by Schweizerhof
et al. [
43] for bone cancer pain. Knee pain was evident 3 weeks post collagenase injection. The synovitis seen early in disease development in wild-type mice thus did not lead to any detectable pain, that is, a difference in weight distribution, unlike in more severe inflammatory models of arthritis [
24,
28,
29]. This observation fits with the mild nature of the inflammatory response in this model. It would be of interest to determine the degree of synovitis at 3 weeks post collagenase injection, at the time of detectable pain, and to determine the effect of GM-CSF blockade at this time point on the synovitis. Recently, van Lent
et al. [
31] showed that synovitis peaks at week 2 in this model, before we were able to detect pain, further suggesting that the synovitis alone is unlikely to account for the pain. Exactly how GM-CSF is acting at the different stages is currently being investigated.
For the collagenase-induced model, cartilage damage is also less in IL-1β-/- mice [
44], suggesting a link between GM-CSF and IL-1. Such a link has been found before with the GM-CSF dependence of an IL-1β-induced inflammatory monoarthritis model [
8,
24], consistent with the concept of a 'colony-stimulating factor network' linking GM-CSF, IL-1, and other cytokines [
2,
4,
45]. That such a link between IL-1 and GM-CSF may also be operating in the collagenase-induced arthritis model is supported by the fact that IL-1 can enhance MMP-mediated DIPEN neoepitope expression in cartilage [
46], similar to our data above for GM-CSF and neoepitope expression.