Animal models are important and essential tools to unravel the molecular mechanisms of pain in OA and for pre-clinical testing of new therapies. The MIA model has been widely used in pain associated studies and to test potential analgesic agents, since it is a rapid and easily reproducible method, and has been described as the one that best mimics the histopathology of human OA [22–26]. The severity of the MIA-induced OA has been shown to be concentration and time-dependent in a range of concentrations [22, 26, 27]. Upon evaluation of the histopathology profile with different doses across a time period of 31 days, we observed a pattern of lesions clearly dependent of MIA concentration. Behavioural findings also showed a dose and time-dependent pattern on nociception, when evaluated by the Knee-Bend and CatWalk tests. Moreover, the ipsilateral paw-print intensity profile here observed with the three doses of MIA was comparable to the weight bearing evaluation described by Pomonis et al.  using similar doses.
Neuronal hypertrophy and phenotypic alterations in animals injected with 2 mg of MIA , along with data revealing ATF-3 expression in DRG neurons of animals with MIA-induced OA , suggested that neuronal damage might be occurring in this experimental model. In the present study, we further evaluated this possibility by investigating the expression of two widely used neuronal injury markers, ATF3 and NPY, in the DRG neurons, at different time-points of the OA evolution and using different doses of MIA.
The presence of ATF-3 has been reported in L5 DRG neurons of OA animals injected with 1 mg of MIA at day 8 and 14 . In another study ATF-3 expression has been reported in FG labelled L4 DRG neurons of OA animals injected with 2 mg of MIA only from day 14 onwards . However, in the present study, we observed ATF-3 expression in L3, L4 and L5 DRG as early as 3 days post-MIA injection, at doses routinely used to induce OA (1 and 2 mg). Moreover, ATF-3 expression was dose and time-dependent, with a biphasic pattern, and occurred in various sized neurons as previously described in axotomized DRG neurons .
ATF-3 has been described by several authors as a sensitive marker of nerve damage [12, 13]. This transcription factor, barely present in the DRG of naive animals, is dramatically induced in DRG neurons following peripheral axotomy . Therefore, the induction of ATF-3 seems to be a cellular response to some types of stress, and nociceptive stimuli without nerve injury does not seem to induce ATF-3 in sensory neurons [12, 13]. This has been corroborated in studies where sciatic nerve transection induced strong ATF-3 expression, while intra-plantar inflammation induced by Complete Freund’s Adjuvant did not [12, 13]. As Braz and Basbaum  stress, ATF-3 expression is not triggered only by the increased activity of sensory fibres due to the lesion, nerve damage being mandatory for such expression. In MIA-induced OA there is no damage of the nerve per se, but there might be injury of nerve endings located at the joint that trigger a similar response but at a smaller scale. Similarly, intraplantar injection of formalin, causing not only tissue inflammation but also injury of local nerve endings, causes ATF-3 expression in some neurons . Also, Hill et al.  reported that skin incision without real nerve damage induces nerve injury-like responses such as ATF-3 expression.
ATF-3 is also expressed in the DRG neurons in collagen-induced arthritis  and in monoarthritis following complete Freund´s adjuvant injection in the tibiotarsal joint . A possible role of positive regulatory factors, such as tumour necrosis factor α (TNF-α) present in the neuroinflammatory environment, could account for the activation of ATF-3 in those conditions, as previously suggested by others [29, 32]. However, it should be noted that in collagen-induced arthritis ATF-3 expression was not affected by anti-TNF therapy . Furthermore, bone destruction occurs in both models [33, 34] and in CFA induced arthritis there is a highly significant correlation between pain behaviour and joint destruction .
NPY expression in DRG neurons is evoked by injury to sensory neurons [16, 17, 35], therefore, expression of NPY was also evaluated in the present study. NPY has been described to contribute to the excitability of axotomized sensory neurons, which in turn could invoke aberrant spontaneous activity in damaged sensory nerves that contribute to neuropathic pain [35, 36]. Furthermore, it has been suggested that NPY can be released intraganglionically, especially after peripheral nerve injury, and act as a mediator of chemical cell-to-cell signalling . Conversely, painful inflammation of rats’ hind paw does not induce NPY expression in DRG neurons [37, 38]. In our study, NPY expression was induced by 1 or 2 mg of MIA immediately at day 3. From day 3 forward, NPY expression was only significantly different from controls with the 2 mg dose, while the lowest dose used (0.3 mg) never induced sufficient NPY expression to reach statistical significance.
The temporal profile and dose related expression was different for the two markers evaluated in this study, which may be due to different activation mechanisms. In fact, while ATF-3 expression is a transcription factor that may be elicited immediately in DRG neurons, NPY expression might need more time since it could be dependent of sympathetic-sensory coupling within the DRG .
Interestingly, a biphasic pattern was observed both on behavioural data and on the expression of the injury related molecules, though no significant correlation between them was found.
Injury due to mechanical activation of the primary afferent endings present in subcondral bone has been proposed as one of the important mechanisms for the generation of pain in OA [40–42]. Considering that ATF-3 and NPY expression are signalling neuronal injury, that could explain the second wave of increased expression of ATF-3 and NPY observed in the present study at latter time points of the disease, when erosion of the cartilage and exposure of subcondral bone occurs. The absence of articular cartilage leads to bone-to-bone articulation and exposure of nociceptor endings in the bone to biomechanical forces associated with weight bearing . However, both neuronal injury markers showed increased expression at day 3 with the higher doses of MIA. Day 3 corresponds to an initial phase of the disease when an inflammatory component has been described by some authors [22, 23, 26], and no degeneration of the cartilage is present, even with the dose of 2 mg, as observed in the histopathological analysis. A possible explanation for the increased expression of ATF-3 and NPY at this early time point could be an injury to nerve endings in the capsule and synovium. As mentioned above, some authors propose that the neuroinflammatory environment could be the trigger to this increased expression [29, 31, 32], but another possibility is that MIA itself may cause a chemical injury of nerve terminals in the injected knee. In fact, MIA is an inhibitor of glicolysis that ultimately leads to the necrosis of chondrocytes, but is not cell specific, and depending on the concentration used different degrees of cell death can be achieved. Actually, we observe that increasing doses of MIA induce the expression of ATF-3 and NPY in further neurons, which suggests that the chemical stimuli might induce some axonal damage producing a nerve-injury like response. In what concerns the reduction in ATF3- and NPY-expressing neurons in some time-points, the possibility of loss of knee-joint afferent neurons does not seem to explain it, since in our previous study  we did not observed a reduced number of neurons in DRG of MIA injected animals, when total counts were performed at 31 days.
The lack of correlation between neuropathic markers and pain behaviours previously referred, indicate that although a nerve injury may be important for pain derived from OA, other mechanisms might also contribute for nociception. In this context, the inflammatory component is likely to cause a rise in the activation of joint nociceptors, that in turn lead to an increased sensitivity of spinal cord neurons, resulting in enhanced nociception.
After peripheral axonal injury, the perikaria of affected neurons and the surrounding glial cells respond to the insult with morphological, metabolic and biochemical changes , in order to promote survival and regeneration of the lesioned nerves . ATF-3, besides being implicated in cell death, may also have a role in inhibition of apoptosis and induction of neurite elongation, and thus promote neuronal survival, depending on the stress signal, cell type and intracellular pathway activated . In fact, in the oncology field, where ATF-3 has been extensively studied, it seems that the cellular context strongly influences its role in cancer development, acting as an oncogene or as a tumour suppressor . Actually, it is possible that the fluctuation of ATF-3 over time has to do with the trigger signal, and also derives from the fact that ATF-3 expression is transient and regulates the balance between proliferative and apoptotic signals.
On the other hand, GAP-43 expression seems to peak when axons are elongating . Therefore, GAP-43 expression was analysed in order to evaluate whether an enhanced growth state had been activated. An increased expression of GAP-43 in ATF-3 positive neurons was observed immediately at day 3, but that became more pronounced from day 7 onwards. The distance of the axotomy site from the cell body seems to be important in determining GAP-43 expression and the speed of its up-regulation [18, 48, 49]. This could explain the delayed expression in OA animals, since the damage is far from the cell body. It should also be noted that GAP-43 was expressed in all cell size DRG cells, as occurs after peripheral axotomy . The augmented expression of GAP-43 reinforces the hypothesis of nerve damage. ATF-3 along with other factors might be involved in the fate of these neurons after injury , and in some cases it might be having pro-survival, axonal-regeneration effect.