GM-CSF gene-deficient (GM-CSF-/-) mice were backcrossed onto the C57BL/6 background for 12 generations 82526. Mice of both sexes, 8 to 12 weeks of age, were used in all experiments. Mice were housed five per cage and the male:female distribution was comparable for all experimental groups. All experiments were approved by The University of Melbourne Animal Ethics Committee.

Arthritis was induced as published 21. Briefly, mice received an intra-articular injection of one unit of collagenase type VII (Sigma-Aldrich, St. Louis, Missouri, USA) on days 0 and 2 to induce joint instability. Because of the focal nature of the damage seen in this model 27, the region of the joint most affected (lateral or medial side) is dependent on the placement of the initial injection (that is, from which side). This placement varied between experiments but for individual experiments was kept constant. Mice were sacrificed at weeks 1, 2 and 6 post collagenase injection and knee joints were collected for histology.

As an indicator of knee pain, the differential distribution of weight on the hind limbs was measured using an incapacitance meter (IITC Life Science Inc., Woodland Hills, CA, USA). This validated technique for arthritic knee pain 242829 measures changes in weight distribution between the arthritic hind limb relative to the non-arthritic hind limb. Mice were allowed to acclimatize to the equipment on three occasions prior to experiment. Weight placed on each limb was measured over a 5-second period. Three separate measurements were taken per mouse for each time point by an independent observer without knowledge of the experimental groups. The readings were then averaged. Results are expressed as a percentage of the weight placed on the arthritic limb verses the contralateral control limb (arthritic limb/control limb × 100). A decrease in weight applied to the arthritic limb (that is, a reading <100) indicates pain 24.

Mice with collagenase-induced arthritis were treated by intraperitoneal injection, either prophylactically or therapeutically, with anti-mouse GM-CSF monoclonal antibody (mAb; MP1-22E9, AbD Serotec, Oxford, UK) or IgG2a isotype control mAb (AbD Serotec) 710. For prophylactic treatment, mice were given 250 µg per mouse per treatment of anti-GM-CSF or control mAb beginning 4 days prior to the first collagenase injection (day 0), with subsequent treatment on days -2 and 0, and thereafter three times per week for 6 weeks. Serum samples were collected at 2, 4 and 6 weeks post collagenase injection and investigated for their anti-GM-CSF mAb concentration. High serum levels of MP1-22E9 were detected at each time point, which, in the light of published potency data on MP1-22E9 30, prompted us to reduce the dose used and extend the dosing intervals for the therapeutic treatment. For therapeutic treatment, mice were given 150 µg per mouse per treatment of anti-GM-CSF or control mAb twice per week following the onset of pain, until the end of the experiment at 6 weeks. The onset of pain was defined as a significant difference in the average pain compared with t = 0.

At termination after arthritis induction, the knee joints were removed, fixed, decalcified, and paraffin embedded, as previously described 710. Frontal sections (5 µm) were stained with either H & E to examine joint architecture or with safranin O, fast green and hematoxylin for proteoglycan loss, and evaluated by two independent observers without knowledge of the experimental groups, using the histologic assessment as follows.

Week 1 and 2 sections were scored for cellular infiltration and synovial hyperplasia from 0 (normal) to 3 (severe) as previously 21. Week 6 sections were scored for cartilage damage in terms of the OA depth into cartilage (grade) and amount of cartilage affected (stage), as described by van Lent et al. 31. This scoring system is modified from that of Pritzker et al. 32 to make it more suitable for measuring pathologic changes in murine cartilage. OA grade was scored from 0 (normal) to 6 (bone loss, remodeling, deformation) and OA stage from 0 (<10% involvement) to 4 (>50% involvement) for the lateral tibia, lateral femur, medial tibia, and medial femur. The grade and stage were then multiplied to give an overall OA score, to represent a combined assessment of OA severity and extent 31. Up to six sections were scored per mouse, and an average taken. Safranin O-stained sections were used to assess the presence of osteophytes at the lateral and medial femur and tibia. Their size was analyzed using the image software Image J (National Institutes of Health, Bethesda, MD, USA ). The mean area per knee joint (three sections) was calculated and expressed in arbitrary units.

The matrix metalloproteinase (MMP)-induced neoepitope, DIPEN, was detected as published 33. Briefly, paraffin-embedded sections were deparaffinized and rehydrated, and endogenous peroxidise blocked with 3% (vol/vol) H₂O₂ (Sigma-Aldrich). Sections were digested with chondroitinase ABC (0.1 units/mL; Sigma-Aldrich) for 2 hours at 37°C to remove chondroitin sulfate from the proteoglycans, prior to blocking with 5% goat serum. Sections were then incubated overnight at 4°C with anti-DIPEN 34 (a gift from A/Professor Amanda Fosang, University of Melbourne, Victoria, Australia) and detected with a biotinylated anti-rabbit IgG (Dako, Glostrup, Denmark), followed by a streptavidin-peroxidase conjugate (BD Biosciences, San Jose, CA, USA). Peroxidase activity was demonstrated by incubation with 3,3'-diaminobenzidine/tetrahydrochloride (Sigma-Aldrich)-H₂O₂ solution. Slides were counterstained with hematoxylin.

DIPEN staining was scored from 0 to 3 based on amount of staining, where 0 = no staining and 3 = maximal staining, as previously described 35.

For pain readings, Student's t-test or two-way analysis of variance was used. For histologic scores, the Mann-Whitney two-sample rank test or two-way analysis of variance was used; values are expressed as the mean ± standard error of the mean. P * 0.05 was considered statistically significant.

Several RA-like inflammatory arthritis models have been shown to be GM-CSF dependent 78910. To determine whether an OA-like model was also dependent on GM-CSF, histology was performed on the joints of C57BL/6 (wild-type) and GM-CSF-/- mice at 6 weeks post collagenase intra-articular injection and disease scored according to the published protocol 31. At this time point, synovitis was not evident (data not shown). GM-CSF-/- mice showed less knee joint tissue damage in the lateral tibia (P = 0.003), lateral femur (P = 0.02) and medial tibia (P = 0.001) compared with wild-type mice (Figure 1A,B). Overall the mean arthritis score for all regions was significantly lower in GM-CSF-/- mice compared with that in wild-type mice (P = 0.002) (Figure 1B).

Synovitis and osteophyte formation have previously been reported to be present early in disease development in this model 21. Therefore, a role for GM-CSF in the development of these disease features was examined at 1 and 2 weeks post collagenase injection. Mild synovitis was observed in wild-type mice at both time points (Figure 2A,B); however, this was virtually absent in GM-CSF-/- mice with only one mouse at week 1 having a very low score. Expression of the MMP-generated proteoglycan neoepitope, DIPEN, was assessed by immunohistochemistry as a measure of MMP-mediated cartilage damage 36. At week 2, there was significantly greater expression of the neoepitope in the cartilage of wild-type mice compared with that in GM-CSF-/- mice (P = 0.02; Figure 2C,D). At both 1 and 2 weeks, there was a trend towards there being fewer osteophytes in GM-CSF-/- mice compared with wild-type mice, and a trend towards the mean osteophyte size being smaller in the former strain (data not shown). Therefore, GM-CSF plays a role throughout the arthritis progression in this model.

We have recently shown that GM-CSF is key to the development of arthritic pain in a number of inflammatory arthritis models 24. We next explored whether GM-CSF was critical also for joint pain development in this OA-like model by administering a neutralizing mAb to GM-CSF prophylactically. Pain has not been recorded in this model. Pain was measured by the changes in weight distribution using an incapacitance meter and has been validated previously to measure arthritic knee pain 242829. To assess whether there is pain induction, C57BL/6 mice received control mAb beginning 4 days prior to collagenase injection. Knee pain was evident from around 3 weeks and remained significant until 6 weeks when the mice were sacrificed (Figure 3A). Mice pretreated with anti-GM-CSF mAb, however, did not develop any detectable pain. From around day 22, control mAb-treated mice showed significantly more pain compared with anti-GM-CSF mAb-treated mice (P < 0.05).

In agreement with the GM-CSF-/- mouse arthritis data above, at 6 weeks the anti-GM-CSF mAb-treated mice also had significantly less arthritis in the lateral femur (P = 0.03) and medial femur (P = 0.004), compared with the control mAb-treated group (Figure 3B). To confirm that the isotype control mAb had no effect on disease development, a PBS control group was also included. There was no significant difference in the histologic scores for the PBS-treated group and the control mAb-treated group for any of the regions scored (data not shown).The mean level of arthritis development was significantly lower in the anti-GM-CSF mAb-treated group (P = 0.003). Once again, synovitis could not be detected at this time point in any of the mice, and in mice where osteophytes were detected, they were significantly smaller in the anti-GM-CSF mAb-treated group compared with the control mAb-treated group (mean size: anti-GM-CSF mAb-treated group, 238 ± 114 versus control mAb-treated group, 577 ± 48, P < 0.05).

Thus GM-CSF is an essential mediator of knee pain progression in this OA-like model.

The mAb-based approach allowed us to next determine whether therapeutic treatment with anti-GM-CSF mAb could also suppress the pain and disease, and how rapidly it might reverse the pain. Once knee pain was evident in collagenase-injected C57BL/6 mice, those with similar pain readings were treated with either control mAb or anti-GM-CSF mAb twice weekly until week 6. Treatment with the control mAb had no effect on the pain, which continued to increase (Figure 4A). Treatment with anti-GM-CSF mAb completely suppressed the pain within 3 days, this being the earliest time at which it was measured following mAb treatment. This abolition of knee pain was maintained until sacrifice (day 42). By histology, at 6 weeks the anti-GM-CSF mAb-treated mice also had significantly milder disease in the medial tibia (P < 0.01) and medial femur (P < 0.05) regions of the joints (Figure 4B,C) with the mean level of arthritis development also being significantly lower in the anti-GM-CSF mAb-treated group for this therapeutic treatment protocol. Once again, in mice where osteophytes were detected, they were significantly smaller following anti-GM-CSF mAb treatment as compared with control mAb treatment (mean size: 267 ± 17 versus 406 ± 14, P < 0.05).