Both the histologic and gene expression analysis support a phasic process for the development of OA in this model. The early phases at 2 and 4 weeks after DMM surgery were the most active in terms of gene expression at a time when the earliest cartilage lesions in the medial tibial plateau were very mild but abaxial chondrophytes had already formed. The chondrophytes matured to osteophytes as the articular cartilage lesions became more severe but, interestingly, this was accompanied by a significant decline in overall gene expression at the 8 week time point. In addition to the differences in overall gene expression, the findings from cluster analysis showed that each time point studied had a unique gene signature that also supports a phasic process.
Perhaps most striking was the decline in gene expression at 8 weeks after surgery. There was a significant down-regulation of genes regulating cell proliferation while matrix remodeling genes that were up-regulated at 2 and 4 weeks were mostly off, suggesting that a quiescent phase had been reached. However, at 16 weeks this pattern had changed and was accompanied by significant cartilage loss in the medial tibial plateau while cartilage damage and chondrocyte death was beginning to be evident in the lateral tibial plateau along with the appearance of lateral axial osteophytes. Because RNA for the gene arrays was only isolated from the medial side of the joint, the changes in gene expression observed at 16 weeks were not due to the disease progressing laterally but rather reflect more advanced disease medially. The lack of time points between 4 and 8 weeks and 8 and 16 weeks prevents any conclusions about the length of the quiescent phase.
The significant expression of Prelp, Col3a1, and fibromodulin at 16 weeks suggests that a matrix repair or a “wound healing response” was active. At 16 weeks, the articular cartilage from the DMM group had severe lesions in the medial tibial plateau with some loss of cartilage. This suggested that the signal at 16 weeks for these genes may have come from either early changes on the femoral side, since some femoral tissue was included for RNA extraction, or more likely, from other joint structures such as the meniscus, synovium, ligaments and subchondral bone. Immunostaining revealed that ligaments and the fibrocartilage found over osteophytes were common locations for all three proteins. These results are consistent with previous studies that examined the location and function of these proteins. Prelp is thought to play a role in matrix organization through interactions with collagen, as does the small leucine-rich proteoglycan fibromodulin 
. Deletion of fibromodulin in mice leads to OA, possibly via weakened ligaments and tendons 
. However, a recent study that revealed a role for complement activation in OA demonstrated increased levels of fibromodulin in OA synovial fluid and found that fibromodulin could promote the activation of complement 
. Overexpression of Prelp has been shown to disrupt normal collagen fiber formation in skin 
and, in contrast to fibromodulin, it is an inhibitor of complement activation 
. Thus, too little or too much fibromodulin and Prelp could contribute to OA.
Increased expression of type III collagen has been previously noted in human OA cartilage 
and it has been found in normal cartilage in fibrils containing type II collagen 
. We noted a pericellular location for type III collagen in articular cartilage and the meniscus. It is not known what role type III collagen could have in the OA joint. In the set of 371 genes that were differentially expressed in DMM vs sham joints, we observed increased expression, at one or more time points, in a large number of collagen genes that are likely part of an attempted repair response. These included the alpha 1 chain of types III, V, XII, XIV, XVI, and XXII collagen and the alpha 2 chain of type IV collagen, along with the collagen processing genes, procollagen c-endopeptidase enhancer protein and procollagen-lysine, 2-oxoglutarate 5-dioxygenase 1.
The phasic changes in gene expression noted in the DMM knees were not the result of aging of the animals over the 16 week time course. Changes in gene expression were noted when we compared the 12 week old animals at baseline (un-operated) and the sham controls over the 16 week time course but these differed significantly from the changes in gene expression noted in the DMM relative to sham joints. This was seen in the overlap analysis where the changes over time in expression in the DMM joints were most often in the opposite direction to changes over time in the sham joints. This was most evident for Prelp and fibromodulin which were significantly down-regulated at week 16 in the sham controls but increased in the DMM joints. Interestingly, immunopositivity for both of these proteins was notably stronger and more extensive in fibrocartilage over osteophytes than in articular or meniscal cartilage. In addition, immunopositivity for Prelp was increased in degenerative cartilage matrix compared with normal cartilage. In the sham time course analysis, most genes exhibited an up-regulation of expression at 2 weeks followed by a progressive decline out to 16 weeks, compared to time zero. Many of these genes were involved in morphogenesis and development, suggesting the animals were more actively growing at the start of the study and then growth likely slowed. Importantly, the progressive increase in subchondral bone area and thickness were the only histologic measures that changed with time in all groups (control, sham and DMM).
We have recently reported a comparison of the 8 week time point results included in the present study with results from the same time point in animals that were 12 months-old at the time of DMM surgery 
. The histologic OA severity in the older mice was about twice that of the younger mice and significantly more genes were up-regulated in the older mice. The 8 week comparisons suggest that the older mice develop OA more rapidly after DMM surgery than the younger mice, a result which could have contributed to the differences in gene expression at 8 weeks. A major difference between the young and older mice was in the expression of muscle related genes which were up-regulated in the older mice and either down-regulated or unchanged in the younger mice. Consistent with greater importance of the muscle-related genes in the older mice, we did not find significant annotations for muscle in the time course study using young mice. It would be of interest to do a time course study in the 12 month-old mice to determine if a phasic progression is also found. We did compare the DMM vs sham gene expression from the 8 week time point in the 12 month old mice to the other time points in the younger mice and found the most genes in common at 16 weeks (158 genes) suggesting more rapid progression in the older mice (data not shown).
A phasic progression of OA was suggested by a previous study in humans where radiographic progression was related to changes in serial measures of serum COMP, as an OA biomarker 
. Interestingly, we found that COMP, along with MMP-13, clustered with genes that were up-regulated during times of increased activity at 2, 4, and 16 weeks and were down-regulated at 8 weeks when activity was lowest. One potential explanation for a phasic process would be the maturation of osteophytes which have been proposed to stabilize the joint mechanics 
. Our observation that chondrophytes had already formed by 2 weeks after surgery to destabilize the meniscus is consistent with the hypothesis that joint destabilization stimulates their formation. Increased expression of Bmp-1, TGFβ2 and TGFβ3 at the early time points is also consistent with their presumed role in stimulating osteophyte formation 
. The stabilization, however, appeared to be transient because cartilage loss was marked at 16 weeks and gene expression was increased relative to 8 weeks.
There are important limitations to the present study. Because RNA was not isolated from individual tissues it is not possible to determine which specific tissue (cartilage, bone, meniscus, ligament, synovium) contributed to the changes in gene expression. Also, if a gene is expressed in a single tissue, pooling of tissues will reduce the signal to a lower level than what would have been observed in the individual tissue. That is the likely explanation for the overall lower levels of gene expression observed in the present study when compared to work using a single tissue. Based on the immunostaining for type III collagen, fibromodulin, and Prelp, as well as immunostaining for IL-33, periostin, and CCL21 reported in our previous study 
, it appears that multiple joint tissues can be responsible for changes in gene expression consistent with the joint operating as an organ rather than as individual tissues. Therapeutic interventions for OA which target key processes occurring in multiple joint tissues rather than a single tissue should be more effective than those targeting a single tissue. Another limitation is that the DMM model may or may not reflect human OA disease pathogenesis. Certainly the histologic features in cartilage and bone are similar, but in human OA there may be more disease activity in the synovium than in the DMM model. Finally, we studied gene expression, and not protein production, and these do not always correlate. Immunohistochemical studies to assess changes at the protein level are not sensitive enough to detect differences noted at the level of gene expression, unless either the differences in expression are marked, or the protein being examined is only detected in diseased tissue and not in normal tissue.
Further studies will be needed to determine the importance of these findings for human OA. An important implication of the results is that the degree of response to an intervention given during a phasic process will depend on the timing of the intervention. Pre-clinical studies using animal models most often start the intervention at the same time or just after the start of the OA process while in human trials participants are likely to be at various stages of the disease process when the intervention is initiated. Finding markers of various disease stages in OA could be used to direct targeted therapy to the proper phase of the disease when the target is most active.