Osteoarthritis (OA), the most common of all arthropathies in our aging population, is a leading cause of disability and represents a large (and growing) worldwide socio-economic cost [1]. It affects approximately 30 million adults in the USA [2], and this number is expected to double by 2020 [3], with longer life expectancy and the increasing incidence of obesity, two major risk factors for the disease. Despite critical importance in drug development, translation of OA therapies focusing either on structure (disease-modifying OA drugs) or pain (symptom-modifying OA drugs) from the bench to bedside has slowed [1, 4, 5]. Differences between preclinical OA models and the disease evaluated in clinical trials contribute to this failure. Rising criticism is noted over the classic translational research, which has failed to predict the efficacy of chronic pain treatments [6–9]. Most critics have targeted the poor validity and clinical relevance of experimental pain models using laboratory animals [10, 11]. It has also been hypothesized that current animal models are too reliant on evoked (reflexive) withdrawal responses and that development of meaningful assessment tools allowing, for instance, the measurement of continuous spontaneous pain, might help to translate experimental data to clinical practice [6, 12].

Naturally occurring OA models are recognized to present pathophysiological changes closest to clinical OA, particularly in large animals [13], but also entail experimental disadvantages (long period to onset, and variability of disease development). In contrast, chemical models cause the most rapidly progressing OA, requiring less invasive procedures and enabling standardization (with increased sample homogeneity). The monosodium iodoacetate (MIA) chemical OA model as described 25 years ago induces cartilage degeneration by disruption of chondrocyte metabolism (i.e., breaking down the cellular aerobic glycolysis). In rats, the MIA model is well-established and resembles the histological and pain-related characteristics of human degenerative OA [14–29]. Owing to the extensive description of the pain response in rats, the MIA OA model was proposed as a standard OA model for pain assessment [23, 30].

The quality of pain assessment methodologies is a cornerstone of preclinical studies targeting new analgesics [7, 10–12, 16, 31, 32]. Three different categories of pain expression can be evaluated in rats: reflexive measures, spontaneous measures, and operant responses [8, 31]. First, reflexive measures using stimulus-evoked responses are commonly used in rats to assess potential hyperalgesia and allodynia, recognized as a clinical expression of the neuropathic pain (nociceptive sensitization) component [8, 33–35]. These measures are generated by exposure to thermal, mechanical, or electrical stimulus, involving mainly spinal-level pain processing, and are also increasingly present in human quantitative sensory testing characterization of pain [36–38]. Second, spontaneous measures can be useful to quantify pain and/or wellbeing [8]. For example, kinetic (static or dynamic weight distribution) [14] or kinematic [39] (ambulation evaluation or characterization) measurement, and spontaneous activity [24] can indirectly assess quality of life in OA models. Pain-induced behaviors (scratching/licking/biting, hypophagia, vocalization, etc.) should also be considered in this category [37]. Finally, operant responses have been more recently introduced to characterize pain in animal models [8, 34, 40–44]. Operant testing is opposite to reflexive response testing as it allows the quantification of behavioral responses at higher levels of the brain, reproducing multiple dimensions of pain, including affective and cognitive changes and not only sensory-discriminative perception [42, 43, 45, 46]. This type of measure allows the observer to evaluate the aversive component of pain as operant tests give the animal an opportunity to avoid the painful condition [33, 40–43, 47–49].

In patients OA of the knee joint, pain is a combination of inflammatory, immune and neurogenic components participating in the hypersensitivity syndrome. Central sensitization mechanisms [50] include various biochemical processes such as increased spinal release of neurotransmitters and neuromodulators, and increased excitability of postsynaptic neurons. In an OA [51] and arthritis [52] rat model, higher levels of neuropeptides, such as substance P (SP) and calcitonin gene-related peptide (CGRP) have been found in the spinal cord. Thus, nervous system modulation seems to play a critical role in the development of the disease [53]. The contribution of these spinal neuromediators to neurogenic inflammation-mediated chronic pain in OA, and concomitant changes in functional pain assessment methods has not been fully established.

With such a variety of methods for pain and analgesic response assessment, it is difficult to opt for the method(s) most adapted to specific conditions. The current study undertook to establish, the reliability of a panel of pain assessment methods (including reflexive, spontaneous, and operant testing) in normal rats, and the influence of environmental conditions, including acclimatization and experimental conditions of manipulation (observer, inverted circadian cycle, and exercise). The most reliable methods were then used to characterize OA pain in the well-established chemical MIA model in rats, while conducting concurrent validation of pain assessment methods in relation to the expression of different spinal neuropeptides and their responsiveness to treatment with intra-articular lidocaine.

Rat models are common in OA research as they are easy to customize and are cost-effective [58]. The MIA model in particular can be standardized and is associated with rapidly developing well-characterized lesions [23, 30]. In an effort to improve the translation of preclinical OA research to the clinical field, we conducted a two-phase study, first, to determine the most reliable pain assessment method protocol, and second, to validate this protocol in the most common chemical model of OA in rats with concomitant changes in spinal neuropeptide concentrations.

Initially, the effect of the environment (inverted circadian cycle, activity level (treadmill exercise), and observer) was tested using well-known pain assessment tools. Prior studies have demonstrated an effect of the inverted circadian cycle on pain research protocols with rodents [59–61]. Our results did not suggest any impact of conducting the evaluation during daytime (more convenient for the investigator). Our group reported a significant reduction in variability of kinetics measures after exercise in cats [62] and dogs [63] with OA. The current study confirmed beneficial effects of exercise to reduce SWB (or other outcome) variability in the MIA rat model of OA. This study also qualified TNTC as a quantitative pain measure using spontaneous behavior. Importantly, the study tested reliability and validity of TNTC, and the potential impact on results obtained with other pain assessment methods, which may be used concurrently. Our results confirmed that a broad range of methods can be combined for pain assessment in the same animals while maintaining reliability and scientific validity.

Finally, as different observers can introduce some degree of bias in pain assessment outcomes, inter-rater reliability was tested by observers with different levels of expertise (one intermediate and one advanced). The methodology included in the current study was accessible to an observer with intermediate experience, as no significant difference was identified during analysis based on the level of experience. As a limitation, the number of observers was minimal, as both observers were women, and only objective assessment methods were selected for this study (limiting any bias related to subjective observation). Therefore, such a hypothesis (the potential influence of experience, and/or gender) would need to be tested further before making inferences from the results. This is particularly important, as recent work has established the influence of the observer’s gender in inducing stress-related analgesia in rodents [64]. Similarly, for limiting the influence of interferential factors in studying the effect of environment, only female rats were used. A possible gender effect would need to be tested in future experiments. Indeed, male rodents are recognized as more sensitive to olfactory exposure to males, including men, causing stress and related analgesia [64]. Moreover, sexual dimorphism [65] and hormonal influence [66, 67] have been observed in endogenous pain modulation mechanisms. Finally, women are more represented in the field of chronic pain [68]; however, many reasons have been explored to investigate this finding [69]. Also, for decades, males have been overrepresented in preclinical research. This situation can definitely lead to a certain bias [70]. All these previous studies justify our decision to use female rats.

As a recognized indicator of test-retest reliability [57], the ICC demonstrated that assessment of mechanical sensitivity using the Randall Selitto test presents poor repeatability, and this outcome cannot be recommended for a valid reflexive measure of pain. The phase 1 experiments demonstrated that SWB and tactile sensitivity can produce more repeatable data when animals (and the observer) are allowed to acclimate to the test device for at least one week. Similarly, there was a slight reduction in the variability of SWB when measured after treadmill exercise. However, the beneficial effects of exercise were not as significant in the chemically induced OA rat model as those observed in cats with naturally occurring OA [62, 71]. The PEAP, rotarod, and treadmill activity measured as TNTC appeared to be highly repeatable without requiring prolonged acclimatization. It must be noted that in both treadmill and PEAP, the intermediate periods 2 and 3 (i.e., 5–10 minutes, and 10–15 minutes, respectively) demonstrated the highest repeatability. These results also suggest that the treadmill and PEAP sessions are a little too long, so for future experimentation the session could be reduced in both cases to 15 minutes instead of 20 minutes. To our knowledge, this study is the first to evaluate the test-retest (repeatability) and inter-rater (reproducibility) reliability of a complete set of pain assessment methods in normal rodents.

As a measure of distribution dispersion that does not require similar units and therefore allows comparison of different variables, the CV of each pain assessment method was verified. At the first evaluation, we again observed the poor metrological property of mechanical sensitivity, presenting the highest inter-individual variability (around 70 % CV). Moreover, the different acclimatization protocols did not help to decrease this variability as the variation in CV from the last to the first evaluation was between 0.53 and 1.07. On the other side, our results support the importance of choosing the optimal acclimatization protocol for pain assessment in the rat model. To our knowledge, this study represents the first systematic evaluation of the effect on data variability of different acclimatization protocols, including four to eight assessments over a 2-week period, with a series of tests near or far from the others. The most reliable acclimatization protocol included five assessments with exposure to the testing methods every other day for the last week (days -14, -7, -5, -3, and -1), with baseline values acquired at day -1.

The second exploratory phase of our project evaluated the validity of the most promising pain assessment methods as determined during phase 1, when applied to the MIA model of OA in rats. Briefly, SWB, PTAE, PEAP, and TNTC detected pain-related changes following OA induction with an intra-articular MIA injection and were validated by increased release of spinal neuropeptides such as SP and CGRP. However, the rotarod assessment, as used in our experimental conditions, was not sensitive to induction of OA pain. Moreover, the PTAE and PEAP methods demonstrated that the sham injection of 0.9 % NaCl was not totally neutral, which was confirmed by the augmentation of spinal liberation of SP (26 times higher) and CGRP (22 times higher) in the sham group. Interestingly, PTAE and PEAP also confirmed the analgesic effect of intra-articular lidocaine injection by the downregulation of spinal neuropeptides, whereas TNTC and SWB did not detect the expected analgesic effect. These results suggest that SWB detects more biomechanical alterations of the joint than ongoing pain and consequently, could be a sensitive method to detect knee joint dysfunction. Of the four pain assessment methods evaluated for concurrent validity, only PEAP detected a treatment effect of lidocaine with a significant difference between the MIA-L group and both the sham and MIA groups. Interestingly, body weight was not affected, either by possible manipulation-related stress in phase 1, or by MIA induction of pain in phase 2. This confirms the possible lack of sensitivity of different endpoints used in research, such as feeding, drinking, etc., for determining quality of life.

The enhanced escape/avoidance behavior and lower PWT were previously demonstrated in both neuropathic and inflammatory pain models [33, 41, 47, 72, 73]. The higher sensitivity of PEAP compared to PTAE in detecting the efficacy of pain relief was also demonstrated in other studies [33].

It would be logical to consider that animals would allocate roughly a similar amount of time in both environments if they do not show natural preference or aversion to one of the two environments in the PEAP testing. This study clearly establishes a strong preference of the rat, a nocturnal animal, to the dark side of the test apparatus. The acclimatization of rats to the test apparatus is fast, the establishment of a baseline is preferable, and both intermediate periods 2 and 3 are highly repeatable. Moreover, PEAP assessment includes both classical (Pavlovian) and operant conditioning in the process of training [42]. When compared with PEAP, PTAE required a longer acclimatization period for both the animal and the observer (at least one week as demonstrated in this study). However, assessing the escape/avoidance behavior using the PEAP required a much longer evaluation time (with 20 to 30 minutes required for each animal), being too labor-intensive to the experimenters, while no automated apparatus is commercially available for this test.

Interestingly, the intra-articular lidocaine injection affected both PTAE and PEAP, and to a lesser degree TNTC, but did not alter changes in weight bearing on the RHP. The lack of effect of lidocaine on SWB could be related to reduced pain in the absence of movement in this model. In a recent study [25], intra-articular lidocaine (200 μL) was efficient to reduce the shift in weight bearing at day 14 post-MIA injection, but only for the highest dose of MIA (4.8 mg). The lower MIA dose and volume in our study (2 mg and 50 μL) combined with the lower sensitivity of SWB may be responsible for the lack of effect with this method, while PEAP and PTAE accurately captured the expected pharmacological effects of lidocaine. Intra-articular lidocaine was chosen for the analgesic test in this study, because of the apparently controversial results obtained in conditioning procedures with non-steroidal anti-inflammatory and opioid drugs (for review, see [42]).

These findings may be useful when designing studies of the efficacy of analgesia using the MIA-induced OA rat model. Moreover, there is also some evidence that the combination of the quick-acting effect of lidocaine (reaching a peak effect at 10 minutes after the intra-articular injection on a CatWalk) [74] and the necessary time to induce a change in the distribution of gait in supraspinal locomotor areas in patients with OA [75], seems to explain the lack of detection of lidocaine analgesia by SWB. Previous studies combined with our results, provide evidence for future use of a continuous infusion of lidocaine to obtain a more sustained analgesic effect, attaining higher lidocaine synovial levels for a prolonged time period.

The intra-articular injection of saline (sham group) generated some hyperalgesia or allodynia, as assessed by PTAE and reflected by the observed change in the operant testing. This is supported by the recent finding of some increased NF-kB activity on days 3 and 7 measured by in vivo luminescent imaging in a transgenic mouse model receiving an intra-articular injection of saline [76]. Moreover, in the MIA mouse model tested in the same study, temporal kinetics of NF-kB activity were strongly correlated with mechanical allodynia (PTAE) and serum interleukin (IL)-6 levels in the inflammatory phase (day 3) of this model, while serum IL-1β was strongly correlated with pain sensitivity in the chronic pain phase (up to day 28) [76]. An increase in the intra-articular pressure and possible injection-related inflammation are proposed to explain this finding. Based on these results, a neutral control group (without intra-articular injection) may be valuable in future experiments.

The MIA model is recognized as valuable in OA research for its ability to detect analgesic effects of different drugs and compounds. The initial inflammatory phase of this model allows the evaluation of various non-steroidal anti-inflammatory drugs and cyclo-oxygenase inhibitors [14, 24, 27, 77]. Moreover, the efficacy of morphine, gabapentin, pregabalin, and the transient receptor potential vanilloid receptor antagonist was successfully demonstrated in this model [15–17, 24, 77]. Moreover, some studies showed that MIA-induced OA leads to an increase in the neuron firing rate and a reduced activation threshold of the afferent nerve fibers [78], which consequently leads to sensitization of spinal neurons in the dorsal horn [79]. In this study, spinal cord neuropeptide quantification suggests and supports development of central sensitization in this model. Indeed, our study confirms an increase in spinal biomarkers of SP and CGRP, as previously observed in the MIA model [51].

The upregulation of spinal neuropeptides observed in this study suggests activation of the peptidergic afferent C-fibers, resulting in central sensitization. It has been well-demonstrated [80], by relative increasing expression of target gene mRNA like pro-inflammatory cytokines (IL-1 and tumor necrosis factor) and pain mediators (CGRP, SP, neuropeptide Y, and galanin), that MIA-induced joint degeneration in rats generates an animal model suitable for mechanistic and pharmacologic studies on nociceptive pain pathways caused by OA. Altogether, this provides further key in vivo evidence that OA pain could be caused by central sensitization through communication between peripheral OA nociceptors and the central sensory system [81, 82]. Despite the fact that SWB did not detect lidocaine treatment on day 21, or asymmetry of weight distribution, our study clearly demonstrated that lidocaine analgesic effects noted by PEAP were translated by concomitant significant downregulation of spinal SP, which was 34 times lower, and of CGRP, which was 20 times lower on day 21.

These results mimic similar therapeutic effects on behavior and SP and CGRP spinal cord expression of intra-articular resiniferatoxin [83] and proteasome inhibitor MG132 [84] in the MIA OA pain model in rats. Unfortunately, we observed that the MIA model caused temporary changes of short duration (return to baseline values at day 21 post injection) and relies on a disease mechanism (chemical inhibition of glyceraldehyde-3-phosphate dehydrogenase activity in chondrocytes, resulting in cell death following the disruption of its cellular glycolysis process [15, 16, 18]) that differs from human natural OA, which could limit the predictability of the therapeutic effect of analgesic and disease-modifying agents. Finally, higher levels of spinal neuropeptides at sacrifice clearly confirms that our model caused some long-term pain or OA damage.

Pain assessment methods used with the MIA model should be selected and scheduled appropriately. In this study only mechanical sensitivity had poor metrological properties, but SWB, the operant PEAP testing, tactile sensitivity, rotarod, and treadmill (TNTC) were repeatable under different environmental conditions. The rotarod test did not achieve sufficient sensitivity to detect OA pain induced by MIA injection in rats and may not be included in future studies. For detecting the analgesic effect of local administration of lidocaine, the pain assessment method that demonstrated the best results was the operant testing, which had the greatest sensitivity, followed by PTAE, whereas SWB had some limitation in sensitivity. Spinal neuropeptide quantification at the end of the experiment has allowed us to validate the effect of positive lidocaine treatment in a more objective manner, as MIA can induce pain. However, the main limitation of this study was the small sample size. Furthermore, it was possible to increase the validity and reliability of pain assessment methods with an optimal acclimatization protocol (five assessments over 2 weeks). In addition, the sham intra-articular saline injection was not totally neutral, particularly with more sensitive methods such as PEAP, PTAE, and this was confirmed by the release of spinal neuropeptides. We therefore recommend the addition of a naive control group (without intra-articular injection). Moreover, increased neuropeptide levels obviously support the central sensitization observed in the MIA rat model. The present results highlight potential for these neuro-mediators as pharmacological biomarkers for analgesic testing in association with sensitive functional assessment methods.

Adj-P, adjusted p value; BW, body weight; CGRP, calcitonin gene-related peptide; CV, coefficient of variation; D, day; HPLC-MS/MS, tandem mass spectrometry coupled to high-performance liquid chromatography; ICC, intraclass correlation coefficient; IL, interleukin; L, lidocaine; LHP, left hind paw; LS-Mean, least square means; MIA, monosodium iodoacetate; MRM, multiple reactions monitoring; MS, mechanical sensitivity; n, number of animals; OA, osteoarthritis; P, probability; PEAP, place escape/avoidance paradigm; PTAE, punctate tactile allodynia evaluation; PWT, paw withdrawal threshold; RHP, right hind paw; RPM, revolutions per minute; RR, relative ratio; SD, standard deviation; SEM, standard error of the mean; SP, substance P; SWB, static weight bearing; TNTC, treadmill number of total crossings; TS, tactile sensitivity; V/V, volume/volume