Background
Chronic pain, caused by diseases such as arthritis or by nerve damage, affects millions of people worldwide. Osteoarthritis (OA) is a widespread condition affecting the elderly where ~70–90% of over 75 year olds are affected [
1], and represents one of the largest cost burdens to healthcare in the western world accounting for 1–2.5% of their gross national product [
2]. The majority of complaints concerning the disease are of chronic pain as well as a loss of joint function. Although the prevalence of neuropathic pain is difficult to ascertain due to its multifarious pathologies (e.g. postherpetic neuralgia, painful diabetic peripheral neuropathy, HIV polyneuropathy, etc.) it also presents a significant burden both to individuals and healthcare systems alike [
3].
The majority of OA research has concentrated on the biochemical and anatomical pathology observed during the progression of the disease rather than studies involving the ensuing pain. Since there are no current therapies to slow the disease progression [
4], analgesia is the first line treatment for OA. OA joint pain is described as a chronic and deep aching poorly localised pain, which is aggravated by physical activity and changes in the weather. Classical analgesics include paracetamol (acetaminophen), nonsteroidal anti-inflammatory drugs (NSAIDs), opioids and steroids. In the majority of patients, these treatments do not provide full pain relief and also display significant side effect profiles Although more is known with regards to the mechanisms underlying neuropathic pain [
5,
6], the knowledge is far from complete. Moreover, opioids, antidepressants and anticonvulsants (e.g. gabapentin) are used in the treatment of neuropathic pain but these therapies often display mixed efficacy in different patient groups. It is therefore important to try to elucidate the mechanisms responsible for the induction and maintenance of these pain states to help in the development of more effective analgesics for the treatment of OA and neuropathic pain. One such approach is to use transgenic mice to identify possible novel molecular targets. Success in these approaches requires robust and translational models.
The monosodium iodoacetate (MIA) OA model has been well described in the rat especially in terms of the pathological progression of the disease [
7] and more recently pain behaviour [
8,
9]. Iodoacetate disrupts glycolysis by inhibition of glyceraldehyde-3-phosphate dehydrogenase, and subsequently causes chondrocyte death
in vitro and
in vivo [
10]. This, therefore, represents an effective model of OA since it is thought that an imbalance occurs between the synthetic and degradative pathways within the articular cartilage resulting in abnormal cellular metabolism and that this represents a major causative factor of OA [
11,
12]. Since the structural integrity of cartilage relies on the normal functioning of chondrocytes, intra-articular injection with MIA produces cartilage degeneration and perturbations of the subchondral bone consistent with the clinical histopathology of OA [
12‐
14]. As this degenerative model progresses, the subchondral bone becomes exposed generating joint impairment and associated pain [
12] and mechanical hypersensitivity [
8,
9]. The pain-related behaviour in this model is thought to be characterised by an early acute inflammatory phase resulting from a fluid expansion of the synovial membrane followed by a persistent phase where the inflammation is largely resolved and is not thought to contribute to the pain pathogenesis [
15].
The emphasis of this study concerns the pain associated with OA rather than the degenerative process
per se since numerous papers have investigated the biochemical and pathological processes associated with this model [
7,
16]. Nevertheless, there are no studies that have compared both the behavioural and neuronal phenotypes of OA and neuropathy, a key to understanding the relations between pain mechanisms and symptoms in animals, especially the mouse where transgenic approaches can be used to reveal potential target mechanisms.
In this study we characterise, for the first time: MIA induced OA in C57Bl/6 mice using both behavioural and in vivo electrophysiological measurements; and compare the neuronal attributes of dorsal horn neuronal recordings in, partial sciatic nerve ligation (PSNL), a model of neuropathic pain. Whereas behaviour informs on threshold responses in awake animals, in vivo electrophysiology can quantify suprathreshold responses which can equate intensities that produce high levels of pain scores in patients.
Discussion
At present, the clinical treatment of OA is at best only partially effective and with the elderly population increasing, the prevalence of OA pain will rise requiring a greater need for clinically effective drugs to treat the pain associated with this disease. This highlights the need for effective models of OA pain in order to study OA pain pathogenesis and to identify novel therapeutic targets. The same holds for neuropathic pain. Here we report clear changes in behavioural and neuronal responses in the MIA model.
This study provides the first characterisation of MIA-induced OA in mice in terms of both pain behaviour and subsequent neuronal characterisation. The behavioural results, in this study, clearly demonstrate that a single infrapatellar injection of MIA into mice results in an early incidence of mechanical hypersensitivity on the hindpaw that is maintained throughout the testing period (28 days). The prevalence of behavioural mechanical hypersensitivity and absence of thermal hypersensitivity is consistent with the findings of other studies conducted in rats [
9,
16,
18‐
20] and correlates well with clinical observations where, in some cases, patients report improvements in pain scores after application of heat or cooling packs. In rheumatoid arthritis in mice, induced using CFA, behavioural hypersensitivities are observed both for mechanical and thermal stimuli [
21], thus highlighting the differences between these two disorders. Generally, OA has been regarded primarily as a noninflammatory arthropathy, however, local inflammation and synovitis are clinically reported and have been observed in animal models of OA. In this study, OA behaviour developed with a slight temporal biphasic profile. Although this early phase may represent the early synovial inflammation (synovitis), which contributes to the early development of OA [
22]; it might be that the MIA itself is having a direct pro-inflammatory effect. Similar, but more marked, temporal patterns of OA behaviour have been observed following MIA injection in rats, where the first phase peaked at day 4 and was largely resolved by day 7, the second phase was initiated at day 14 and remained unresolved for the study duration [
9,
15,
18,
19]. In this study, therefore, the neuronal characterisation was performed from day 14, when the inflammation is thought to have been largely resolved and a more persistent pain state exists, since NSAIDs have previously only been found to be effective in the 3 days following MIA injection [
8].
Although von Frey filaments were applied to the plantar aspect of the hind paw rather than the skin overlying the knee joint due to experimental hindrances; referred pain in the thigh, leg and foot of OA patients has been reported [
23,
24]. Cell bodies of afferents from the knee are thought to co-localise in DRGs with those of the hindpaw since retrograde labelling studies have shown that L3-L5 predominantly receive primary afferent input from the knee [
25] and L3, L4 and L5 DRGs receive hind paw afferents [
26]. This juxtaposition permits crosstalk such that pain transmitted by the afferents supplying the knee would engender pain in the hind paw. The behavioural studies of evoked pain in the hind paw support secondary hyperalgesia as this is characterised by mechanical hypersensitivity and often lacks heat hypersensitivity [
27]. The spinal cord neuronal recordings bear out this premise in that measures of central hypersensitivity were observed.
Although cartilage itself is considered aneural, bone is densely innervated with Aδ and C fibres especially in regions of maximum load and high turnover such as those of the proximal and distal head of the femur [
28]. During arthritis, pro-inflammatory agents such as bradykinin, substance P, and prostaglandins are released into the joint [
29] which can lead to afferent fibre sensitisation and reductions in fibre thresholds (peripheral sensitisation). The establishment of OA in this study was accompanied by increases in electrically evoked A- and C-fibre responses with no changes in their activation thresholds. This might also be explained by the activation of silent nociceptors, which do not normally respond to stimuli but become active in response to tissue damage/inflammation [
30]. These increases in A- and C-fibre firing may underlie OA patients' accounts of aching and throbbing pain interspersed with activity related sharp/stabbing pain [
31]. The very clear increase in the non-potentiated 'input' response, indicative of increased peripheral drives, would result in an increased nociceptive drive onto dorsal horn neurons thus facilitating central mechanisms of hypersensitivity such as wind-up. The application of suprathreshold stimuli, during the electrophysiological characterisation, evoked enhanced responses to noxious mechanical stimuli. This electrophysiology is similar to a study on CFA-induced arthritis in the rat, where facilitated mechanical responses in the noxious range and increases in the proportion of neurones responding to mechanical stimulation were observed in neurones receiving input for the joints [
32]. Thus the increased input, wind-up, A- and C-fibre firing, response to pinch and mechanical hypersensitivity observed both in the behaviour and electrophysiology imply a role for central sensitisation in the MIA model of OA pain.
Although, peripheral neuropathic trauma may induce the release of inflammatory mediators this is only likely in the early stage of neuropathic pain. Likewise, osteoarthritic degeneration could lead to nerve compression, a common cause of neuropathic pain but again, this is not likely to be a major mechanism. It is likely that OA initiates an inflammatory state as cartilage degrades and then moves to a nociceptive pain in the later stages with direct physical stress on bones providing the peripheral drive. Neuropathic pain, by contrast, involves disordered and altered ion channel and other functions in the damaged and spared peripheral nerves and is so restricted to the nerve territory and involves both loss and gain of function. Both the behavioural and neuronal measures in these models show overlapping alterations produced by the very different pathologies. This suggests some commonalities in the central processing of different peripheral pain states but with marked upregulations in the neuropathy model, possibly compensating for the afferent fibre loss [
33]. Mechanical hypersensitivity was marked in both models but greater in PSNL, perhaps relating to the segmental changes in the nerve injured zone, where testing was done. In the MIA model the enhanced mechanical responses are likely to represent referred pain since the hindpaws were tested after knee MIA. Both groups showed larger mechanical evoked responses than their controls, significant in the MIA group. Input was enhanced in the MIA group but unchanged after nerve injury, perhaps reflecting the loss of input caused by the neuropathy yet enhanced primary drives onto deep dorsal horn neurones after MIA. The trend towards an increased thermal responses of neurones was not reflected in the behaviour and this is likely a result of the suprathreshold nature of the neuronal responses – the responses of the neurones to lower temperatures, akin to the behaviour were not altered.
Clinical observations have also suggested central changes in nociceptive information processing in patients with OA. Support for this comes from a study conducted in patients with bilateral symptomatic OA of the knee, where a single intra-articular injection of bupivacaine (local anaesthetic) into the most painful knee caused a reduction in pain scores for both the ipsilateral and contralateral knee [
34]. In this study, it seems likely that there is an important central involvement in OA pain as typical indicators for peripheral sensitisation, namely decreased activation thresholds for A- and C-fibres (where stimuli would bypass sensitized peripheral terminals), were not observed in the neuronal characterisation, although an increase in input was observed that could reflect augmented excitability. A role for peripheral sensitisation or altered peripheral drives cannot, therefore, be ruled out and may be more important during OA pain induction rather than its maintenance.
At early stages of this model in rats, NSAIDs display good efficacy but become ineffective at later time points [
8,
15,
18], and pain behaviour can only be attenuated using morphine [
18] and Gabapentin [
16]. Furthermore, repeat dosing studies, a regimen more consistent with the clinical setting, found that both gabapentin, used for neuropathic pain in patients, and morphine displayed efficacy in this model [
9]. This pharmacology suggests that the early stages of the model/disease reflect acute inflammatory pain, but as the disease progresses the inflammation is largely resolved and starts to share similarities with those of neuropathic pain. Furthermore, in one study, a significant increase in the levels of ATF-3 (a marker for neuropathy) in L5 dorsal root ganglia was found at days 8 and 14 following MIA injection in rats, however, no differences were found at later time points up to days 35 [
16].
We demonstrate increases in C-fibre evoked responses and wind-up of WDR neurones in PSNL mice. Along with the slight increase in the proportion of neurones displaying spontaneous activity (and at a slightly higher rate), these changes are consistent with those expected for central hyperexcitability. Similarly, a trend was observed for thermally and mechanically evoked responses, although this did not achieve significance. This is consistent with electrophysiological studies performed in spinal nerve ligated rats, where L5 and L6 dorsal roots are tightly ligated reducing spinal afferent input to the hindpaw to 30–40% [
35], but comparable responses are observed in the sham groups to peripherally applied stimuli [
36‐
38]. This suggests, in concordance with the rat, that these remaining hyperexcitable neurones can compensate for the sensory loss occurring in these pain states. Following nerve injury, α
2δ subunits of voltage gated calcium channels are slowly upregulated in the central terminals of peripheral nerves leading to spinal hyperexcitability [
39], spinal neurones receive greater descending facilitatory influences [
36] and receptive fields expand [
40]. The former changes are permissive for the actions of gabapentin and pregabalin [
41] and may therefore also reflect both the spinal hyperexcitability observed in OA in this study and the efficacy of gabapentin observed in rat studies of OA [
9]. Studies of this nature in the MIA model will allow further comparative mechanistic studies across different pain aetiologies.
Methods
All procedures were performed according to current UK home office legislature (Animals Scientific procedures Act, 1986). Adult (8–12 weeks; 20–30 g) C57Bl/6 mice of mixed gender were used for these studies (Harlan, UK). Animals were housed in groups of 3–4, with 12 h light/dark cycle and allowed food and water ad libitum except during procedural testing. Surgical procedures were performed under general anaesthesia with halothane (1–2%) delivered in O2.
Monosodium iodoacetate (MIA) injection
OA was induced, in briefly anaesthetised mice, by a single intra-articular injection of MIA (Sigma, UK) into the knee. Knee joints were shaved and flexed at a 90° angle; 5 μl of 5 mg/ml MIA in sterile saline (0.9%) was injected through the infra-patellar ligament into the joint space of the left (ipsilateral) knee using a 30-gauge 0.5" needle. This concentration of MIA has been found previously, in mice, to precipitate histological changes in the cartilage, consistent with those of osteoarthritis [
43]. Control mice received an intra-articular injection of vehicle (5 μl sterile saline, 0.9%).
Partial Sciatic Nerve Ligation (PSNL)
Nerve injury was induced in deeply anaesthetised mice using a method based on that previously described by [
44,
45]. The left (ipsilateral) sciatic nerve was exposed above its trifurcation and one third to one half of the nerve was tightly ligated using 7-0 non-absorbable silk suture (Mersilk, Ethicon). The wound was closed in layers using 4-0 suture (Mersilk, Ethicon), and animals were allowed to recover. In sham mice, the sciatic nerve was exposed, but not ligated and was closed as before.
Nociceptive behaviours
All behavioural testing was preceded by two baseline measurements taken 2–4 days prior to injection/surgery. A single observer was used for the study, and was blind to the treatment given to each animal.
Behavioural tactile hypersensitivity (von Frey hairs)
Animals were placed in a Perspex chamber with a wire mesh floor, and allowed to acclimatise for at least 2 hours prior to testing. Tactile hypersensitivity was tested by touching the plantar surface of the hindpaw with von Frey filaments using the "up-down method" [
46], starting with 0.6 g then ranging from 0.07 g to 1.4 g. Positive withdrawals were counted as biting, licking and withdrawal during or immediately following the 3 s stimulus. The strength of the von Frey filament was increased or decreased following a negative or positive response respectively. This up-down procedure was applied 4 times following the first change in response and stimuli were not re-applied within a 5 minute period. Data are presented as 50% paw withdrawal threshold (PWT) for each group ± SEM.
Heat hypersensitivity (Hargreaves'test)
Mice were assessed for thermal hypersensitivity as described by [
47]. Mice were placed in translucent chambers and were allowed to acclimatise for at least 1 hour. Noxious heat sensitivity was measured using a radiant heat device directed at the plantar aspect of the paw. Paw withdrawal latencies, accurate to the 0.1 s were noted, where a maximal cut off time of 20 s was used to minimise paw damage. The stimulus intensity was adjusted to ~8 s withdrawal latency for naive animals. Measurements were taken 3–5 times with at least 5 minutes between tests. Data are presented as mean withdrawal latency for each group ± SEM.
Rotarod
Animals (n = 10 per group) were placed on the rotarod at a speed of 4–40 rpm for a maximum of 300 s. During baseline tests two training trials were given followed by two recorded trials. Two sets of baseline values were recorded and their mean was taken. Animals were excluded from further testing if they had not achieved a trial of > 250 s in any of their 8 baseline trials. After surgery, three trials were performed and the mean latency was recorded. Animals were allowed to rest for at least 15 min between testing. Data are represented as mean latency to fall for each group ± SEM, where the latency to fall for mice freely rotating on the drum was recorded following two successive rotations.
Spinal cord electrophysiology
In vivo electrophysiology was performed at days 14–21 following MIA/sham injection or surgery using parylene coated tungsten electrodes (A-M Systems, USA). Animals were anaesthetised using urethane (240 mg/Kg), and a laminectomy was performed exposing L3-L5 of the spinal cord. Once a single wide dynamic range neurone had been isolated in the ipsilateral dorsal horn receiving inputs from the hindpaw, A- and C-fibre thresholds were measured and noted. A train of 16 electrical stimuli (2 ms wide pulses, 0.5 Hz), delivered transcutaneously by means of pins inserted in to the hindpaw, was applied at 3 times the activation threshold for C-fibre responses. A post stimulus time histogram was constructed and fibre responses were separated according to latency: A (0–50 ms) C (50–250 ms). Responses occurring after the C fibre latency band were characterised as post discharge (250–800 ms). Input (non-potentiated response) was calculated as the number of action potentials in response to the first stimulus × total number of stimuli (16). Wind up, a measure of the neuronal activity in response to a succession of stimuli and hence synaptic strength, was calculated as the total number of action potentials evoked at the end of the train – the input. A wide range of natural stimuli including brush, von Frey filaments and heat were applied to the hindpaw for a period of 10 seconds. Data were captured and analysed using a CED 1401 interface coupled to a Pentium computer running Spike 2 software (Cambridge Electronic Design).
Data analyses
Data are expressed as mean ± SEM, where n represents the number of individual experiments performed. Raw data were analysed using two-way repeated measure ANOVAs, Kruskal-Wallis, Mann-Whitney and unpaired student T tests as appropriate where p < 0.05 was considered significant.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
VLH and AHD conceived, designed and performed the experiments, analysed and wrote the manuscript.