Molecular pathogenesis of OA pain: past, present, and future

Abstract

Objective:

To provide a historical perspective and narrative review on research into the molecular pathogenesis of osteoarthritis pain.

Design:

PubMed databases were searched for combinations of “osteoarthritis”, “pain” and “animal models” for papers that represented key phases in the history of osteoarthritis pain discovery research including epidemiology, pathology, imaging, preclinical modelling and clinical trials.

Results:

The possible anatomical sources of osteoarthritis pain were identified over 50 years ago, but relatively slow progress has been made in understanding the apparent disconnect between structural changes captured by radiography and symptom severity. Translationally relevant animal models of osteoarthritis have aided in our understanding of the structural and molecular drivers of osteoarthritis pain, including molecules such as nerve growth factor (NGF) and C-C motif chemokine ligand 2 (CCL2). Events leading to persistent osteoarthritis pain appear to involve a two-step process involving changes in joint innervation, including neo-innervation of the articular cartilage, as well as sensitization at the level of the joint, dorsal root ganglion and central nervous system.

Conclusions:

There remains a great need for the development of treatments to reduce osteoarthritis pain in patients. Harnessing all that we have learned over the past several decades is helping us to appreciate the important interaction between structural disease and pain, and this is likely to facilitate development of new disease modifying therapies in the future.

Introduction.

Significant progress has been made in unravelling molecular pathogenic mechanisms in osteoarthritis (OA) over the past 30 years and beyond. These have been aided in recent decades by large scale genomic studies, clinically relevant preclinical models, refinement of clinical outcome measures and novel molecular targeting. Whilst we are still some way from having full disease modifying drugs that will modulate structural disease as well as symptoms, we have seen the emergence of a number of pharmacological approaches that modulate aspects of disease such as structure modification by sprifermin, a truncated FGF18 analogue, and strategies that neutralize nerve growth factor (NGF) in the treatment of OA pain. Accompanying these studies is a greater understanding of their mechanism of action, their site of action and other molecules that contribute to the pathway and which could themselves also be potential targets. These studies build on several earlier decades of careful observation, pathological investigation and early molecular discovery to end up where we are today. Harnessing this collective knowledge and working across research silos is bringing us closer to effect real change to benefit patients.

Pathological correlates with Clinical OA pain.

The study of OA pain has a long history. It was well established over 50 years ago that OA pain originated in the joint; indeed, both neurectomy and intra-articular injection of local anaesthetic were able to relieve pain ¹. It was also appreciated that pain was mechanically induced; osteotomy, by altering the loaded regions of the joint, was able to alleviate symptoms ². Pain thresholds varied between and within individuals and this was attributed to the influence of social and other patient factors that affected the pain experience. In 1989, Brandt speculated on the pathologies within the OA joint that could give rise to pain ³. These included synovitis (possibly induced by crystals or by ‘damage particles’ from abrided cartilage), stretching of the periosteum over osteophytes, microfractures within the subchondral bone, venous hypertension due to distortion of medullary blood flow in bone (synonymous with bone marrow oedema on magnetic resonance imaging), distension of the joint capsule and muscle spasm. Altman later added ligament insertions and damaged menisci to this list ⁴. These interpretations were partly based on the sensitivity of these tissues in a healthy joint to sharp pressure. Indeed, in their seminal paper of 1950, Kellgren and Samuel used healthy human volunteers and of patients undergoing arthrotomy under local anesthesia to document the degree to which capsular tissues and ligaments within the knee were sensitive to sharp pressure ⁵. They concluded that fibrous ligaments appeared to be a major source of joint pain. A later study performed by orthopedic surgeon brothers without intra-articular anesthesia confirmed the sensitivity of the cruciate ligament insertions and also noted the insensitivity of articular cartilage, consistent with it being aneural and avascular ⁶. As a result, cartilage was largely excluded as being relevant to the origin of pain in OA.

The role of inflammation, specifically synovitis, in OA pain has remained controversial. Whilst Kellgren and Samuel documented the relative insensitivity of most regions of the healthy synovium, the Dye study reported that palpation of the suprapatellar pouch, capsule, and the medial and lateral retinacula produced moderate to severe localized pain ⁶. Likewise, innervation of the synovium was demonstrated in 1990 ⁷ and in OA tissue these fibres were strongly stained for substance P (also known as tachykinin), a neuropeptide that, along with CGRP, increased the sensitivity of pain fibres and also contributed to neurogenic inflammation through inciting cytokine release from mast cells and stromal cells ^8–10. Such neurotransmitters and other inflammatory molecules such as prostaglandins were proposed to drive pain in arthritis ¹¹. Part of this might also have been due to the oedematous response in the tissue, leading to increased mechano-sensitivity. Importantly, in the 1980s the concept of ‘silent nociceptors’ took shape - nociceptors that would not normally respond to mechanical stimuli became sensitized to do so after an inflammatory insult ¹². Electrophysiology techniques have also been used to show that injection of a variety of inflammatory agents into the joint produced peripheral sensitization and increased activity of the nerve afferents innervating the joint (reviewed in ¹³). Whether these findings were relevant to osteoarthritis with lower levels of inflammation was unclear at the time.

Anti-inflammatory agents in OA pain:

The late 1970s and 1980s was the age of the non-steroidal anti-inflammatory drug (NSAID) revolution with many head-to-head studies comparing analgesic efficacy of different proprietary treatments. NSAIDs can promote analgesia through inhibition of the production of prostaglandin E2, reducing sensitization of nerves, as well as by reducing local inflammation. However, as a class, for the treatment of OA pain these drugs were analgesic but without evidence of a true anti-inflammatory effect (i.e. they didn’t reduce inflammatory cell infiltration). Indeed, it was noted that the synovium was little changed after treatment raising some authors to suggest that the mechanism of action was not synovium based. A similar phenomenon was also described following treatment with intraarticular and oral glucocorticoid use in OA; a number of studies have demonstrated a temporary analgesic response with i.a. use of long-acting glucocorticoids, but a direct correlation with synovial volume or vascularity, a surrogate marker for inflammation, was missing. Low dose oral prednisolone is not analgesic ¹⁴, but 10mg oral prednisolone daily was able to reduce symptoms ¹⁵. In the latter study there was a reduction in synovial volume but little effect on power doppler signal on ultrasound or vascularity of the synovium. The pain target for NSAIDs and corticosteroids therefore remains uncertain.

Of note, for both of these classes of anti-inflammatories, a concern over accelerating structural joint disease has been raised. In the case of NSAIDs, these were shown to accelerate structural disease in canine OA models and could suppress the normal anabolic response of damaged cartilage in vitro ^{16, 17 18, 19}. This might be one of the mechanisms behind “indomethacin knee” or “indomethacin hip” a phenomenon where disease was seemingly accelerated by prolonged use of indomethacin ^{20, 21}. Similarly, i.a. corticosteroid may accelerate cartilage loss in OA after prolonged use. In one study where individuals were given 3 monthly i.a. steroid injections over a two-year period, no analgesic effect over placebo was noted (time points earlier than 3 months were not measured) but a worsening of cartilage loss was observed by MRI in the steroid group compared with placebo ²².

Discordance between pain and pathology:

In recent decades much investment has been made in developing quantitative measures of OA damage within the joint using magnetic resonance imaging (MRI), and to a lesser extent computerised tomography (CT). MRI has the advantage of being able to reveal the soft tissues, e.g. synovium and bone marrow, and to correlate these pathologies with patient reported pain over time. Whilst these clinical studies have significantly increased our knowledge of pathology within the joint ²³, how this changes with disease course ²⁴, how age-related OA resembles post traumatic OA ²⁵, and how individual pathologies relate to one another geographically within the joint ^26–29, the studies have not helped to elucidate the primary pathology that drives pain in the OA joint. This is largely because all pathologies correlate moderately well with pain and they also correlate with one another ^30–32. Building on previous earlier descriptions, it has become clear that much of the discordance of reported pain and pathological change in the joint is due to extra-articular factors that have a large and unpredictable impact on patient perceived pain. Indeed, controlling for these person level effects improves the correlation of pain score with radiographic score ^{33, 34}.

Another area where clinical research has contributed to our understanding of OA pain is through the use of quantitative sensory testing to assess sensitization. From the body of work carried out in the past 15 years, a few important concepts have emerged: measures of peripheral and central sensitization are associated with OA-related pain ³⁵, presence of both peripheral and central sensitization predicts development of persistent knee pain ³⁶ and abnormalities of somatosensory perception normalize following successful surgical intervention ^{37 38, 39}, as do functional MRI images of the brain ^{40, 41} indicating that ongoing pathology within the joint is driving these changes. Further understanding of the relationship between sensitization and persistent OA pain may provide opportunities for prevention, delay or reversal of chronic pain ⁴².

Pre-clinical models in OA pain discovery research.

In the early 2000s, a push to develop models of osteoarthritis suitable for pharmaceutical testing was made (with criteria listed at the time including: reproducibility, ability to predict efficacy in humans, similar pathology as human disease). The monoiodoacetate (MIA) rat model emerged which induced pathology by targeting chondrocyte metabolism and exhibited development of weight-bearing asymmetry ^{43, 44}. Surgical rat ⁴⁵ and mouse models of OA ^{46, 47} also began to be tested for behavioral changes associated with pain. Reversal of these behaviors were often tested using NSAIDs and behaviors were largely found to be sensitive in a dose-dependent manner. However, even at this time discrepancies in reports across laboratories in the responses of models to these standard analgesics were noted ^{43, 48}. This highlights the challenges experienced when performing these types of tests, the dependence on the expertise within specific laboratories, and almost certainly, differences in bias mitigation strategies used.

In a 2018 review ⁴⁹, we had noted that the number of studies examining pain in animal models of osteoarthritis was limited, and an updated search since that time indicates that the landscape in terms of numbers of studies has not increased much (Table 1). A PubMed search from 2018-August 2023 for “osteoarthritis mouse” yielded 2,306 papers, however adding the word ‘pain’ decreased this number to 381. Of these, only 145 papers used a model of OA (rather than an acute inflammatory joint pain model) and performed at least one pain behavior assessment. However, the types of studies being performed has changed over this time. Since 2018, of the studies that have measured any pain-related behavioral assessment, the inclusion of both evoked and non-evoked types of pain-related behavior has become more routine. Currently, the majority of mouse testing involves surgical induction of OA. For historical reasons this is generally performed in male mice where structural disease is significantly greater. Where comparisons have been made, female mice appear to develop pain at an earlier stage of joint (cartilage) damage and this can be seen in slower as well as more aggressive surgical models, such as that induced by DMM, partial medial meniscectomy (PMX) ⁵⁰ and intact ACL rupture models ⁵¹. Continued use of models that facilitate development of joint damage in both sexes should help to improve this area of research moving forward. When it comes to reporting, most papers include specification regarding blinding when describing histological scoring, but less than 50% of papers performing behavioral testing stated whether these tests were performed blinded or not. In addition, the details included in these sections could be improved in order to increase replicability. Finally, in contrast to how the study of joint inflammatory pain started with a major focus on electrophysiological measurements, assessment of functional changes in nervous system tissues remains relatively uncommon in the setting of long-term osteoarthritis models (32%) – as the majority of studies include behavioral testing as a secondary outcome to structural joint changes. Through the use of both electrophysiology and calcium imaging, we have learned that osteoarthritis is associated with increased responsiveness of nociceptors to mechanical stimuli as well as with a lower action potential firing threshold ^52–54. Moving forward, continued adoption of neuroscience technologies combined with the use of these well-established pre-clinical models of osteoarthritis is expected to yield novel analgesic targets.

Table 1.

PubMed search ‘osteoarthritis pain mouse’: 2018-Aug 13, 2023

	% Papers since 2018 (out of 145 papers total)
Monosodium iodoacetate (MIA) model	24
Surgical joint destabilization model	61
Collagenase model	6
Systemic OA models (aging or obesity)	6
Non-surgical joint injury model	6
Two model comparison included	7
Included analysis (molecular or functional) of neural tissue	32
Assessed measure of sensitization (allodynia/hyperalgesia)	71
Assessed non-evoked behaviors	61
Pain assessed at more than one time point	77
Assessed pain and joint damage	88
Blinding of pain assessments included	45
Testing performed in both sexes	12

The discovery of NGF as a principal driver of OA pain:

An important recent breakthrough in molecular pain pathogenesis, that has been facilitated by the use of murine models of disease, was the discovery that the neurotrophin, nerve growth factor (NGF) was an important driver of OA pain. NGF had a strong established role in neuronal development; causing sprouting and elongation along gradients of NGF ^{55, 56}. The notion that NGF also had sensitizing activities at the level of nociceptive fibres was important and paved the way for considering it as an analgesic target ^{57, 58}. In murine OA, NGF mRNA was transiently regulated in whole joint extracts immediately in the post operative phase and then again at late stages after destabilization of the joint (from 8-10 weeks post operation according to type of surgery), correlating with the time of spontaneous pain behaviour in mice ^{59, 60}. Delivery of the soluble high affinity receptor (TrkA) to mice with established pain led to rapid analgesia ⁵⁹. Analgesia through neutralization of NGF was also demonstrated using anti-NGF antibodies in rat surgically induced OA ⁶¹, and by NGF vaccination in mice ⁶⁰.

These studies have been substantially validated by the success of clinical trials using neutralizing antibodies to NGF. First published in 2010, anti-NGF delivered subcutaneously (two doses over 12 months) demonstrated strong, dose-dependent analgesia, quite unlike anything seen previously ⁶². Anecdotal reports from patients recruited in this study from the lead author indicated that this treatment appeared to lead to a staggering improvement in symptoms such that patients were able to “go out dancing for the first time in years!”. However, the clinical trials that followed experienced a number of setbacks. The Federal Drug Association (FDA), noted early on that some trial participants (around 6%) were developing rapidly progressive joint damage. This was deemed to be due to rapidly progressive OA (RPOA), rather than a Charcot joint (in which the joint has lost its proprioception), or avascular necrosis ⁶³. It was also observed that this risk appeared to increase at the higher doses of anti-NGF and especially when combined with NSAIDs. The risk was not only in index joints (the joint with the primary diagnosis of OA), but was also appearing in joints which were not known to have disease. Although phase II and phase III clinical studies were restarted (with mitigation strategies in place), RPOA was still seen in around 8% of participants ⁶⁴, and in 2021 the FDA failed to grant a licence for anti-NGF treatment for OA pain. However, feline and canine versions have gone on to receive approval for treating joint pain in these species, which could help inform the longer-term effects of such agents ⁶⁵.

The source of NGF in the OA joint has been subject to some debate over the years. The protein has been immunolocalized to human OA synovium ^{66, 67}, while mRNA analyses of synovium from normal, painful and painless human OA joints failed to show NGF mRNA regulation by this tissue ⁶⁸. Immunostaining was also evident in osteochondral channels, and associated with pain, in late human OA ^{69, 70} and more recently in surgically induced rat OA ⁷¹. This appears to accord with the discovery that NGF is strongly induced by injury to articular cartilage, both in vitro and in vivo after joint destabilization and correlates with the timing of pain behaviour ⁷². There was little evidence of an increase in inflammatory genes at the time NGF became elevated in late murine OA further implicating a non-inflammatory drive of NGF regulation when mice develop pain behaviour. While OA subchondral bone marrow lesions also failed to show regulation of NGF mRNA ⁷³, production of NGF by other cells in the subchondral bone such as macrophages and osteoclasts ⁷⁰ has been demonstrated. Together this supports the hypothesis whereby the injured basal cartilage and/or subchondral bone makes NGF and this is both necessary for the neoinnervation seen in advanced disease, as well as changing the firing threshold for neurons in the vicinity (reviewed in ^{74 75}) (Figure 1). Evidence to support this is suggested by the appearance of CGRP-positive and Na_V1.8-positive neuronal fibres that develop in the subchondral bone below damaged regions of the articular cartilage ^{67, 76, 77}. This process may be facilitated by secretion of netrin-1 by subchondral bone osteoclasts. Netrin-1 promotes neuronal growth and its deletion protects mice from development of mechanical allodynia after surgically induced OA ⁷⁸.

Figure 1. Molecular model of OA pain.

Model requires a two-step process: neoinnervation of joint tissues, including the osteochondral junction, and sensitization of the nociceptor response at the level of the joint and dorsal root ganglion (DRG) which may lead to central sensitization. Key molecular players include nerve growth factor (NGF), C-C chemokine ligand 2 (CCL2) and Piezo2. Netrin-1 has been implicated in guiding axons through the subchondral bone. A number of other molecules have been implicated in nociceptor sensitization in OA. Little is known about molecules that influence central sensitization. GDNF – glial derived neurotrophic factor; BDNF – brain derived neurotrophic factor; GMCSF – granulocyte-macrophage colony stimulating factor.

Other candidate molecular pathways:

Whilst the current evidence suggests that NGF is a primary driver of pain in OA, other robust pathways have also emerged that influence pain sensitivity in experimental OA. One of these is the C-C chemokine, CCL2 (also known as monocyte chemotactic protein-1, MCP-1) through its receptor, CCR2. CCL2 was transiently increased in the dorsal root ganglia (DRG) of mice at 8 weeks post joint destabilization (by DMM) and deletion of CCR2 was associated with loss of macrophage infiltration in the DRG and a failure to develop late spontaneous pain behaviour ⁷⁹. In a separate study, both CCR2 and CCL2 null mice had delayed onset of late spontaneous OA pain behaviour despite structural damage in the joint being similar to wild type mice over the same period ⁸⁰. Long-term systemic therapeutic targeting of CCR2 via a receptor antagonist also displayed an analgesic benefit ⁸¹. Further work has demonstrated that intra-articular targeting of CCR2 may also be sufficient to reduce symptoms in a mouse model of OA ⁸².

Strong evidence for Piezo2 has also emerged. This mechanosensitive ion channel expressed by nociceptors modulates sensitivity of murine OA pain and has a direct effect on NGF-mediated responses. Indeed, deletion of Piezo2 in Na_V1.8-specific pain fibres reduced spontaneous pain behaviour in mice 12 weeks after surgical joint destabilization (without affecting structural joint disease) and abrogated pain sensitization (and joint swelling) induced by injecting NGF directly into the joint. As Piezo2 is a mechanosensor, it potentially provides another link to explain the highly mechanosensitive nature of OA pain ⁸³.

Finally, multiple other candidate pathways have been identified that can influence sensitization potentially through neuroimmune interactions such as Toll-like receptor signaling, GM-CSF, and other inflammatory cytokines (reviewed in ⁸⁴).

Influence of biological sex on pain.

It has been known for a long time from epidemiological studies that females experience increased symptomatic OA associated with higher pain scores and pain persistence. This is evidence largely obtained from post-menopausal women in whom most OA cases are described, raising the possibility that sex hormones may influence the experience of pain. Sexual pain dimorphism has also been described in murine OA (also noted above). It is striking and consistent that female mice develop pain at a seemingly earlier stage of structural (cartilage damage) disease ^{50, 85, 86}. The molecular profiles within the whole joint as well as the synovium also differ between sexes ^{50, 51} and joint shape may also contribute to risk ⁸⁷. Spontaneous OA associated with aging also demonstrates differences between the molecular and cellular profiles within the DRGs between sexes ⁸⁸. Interestingly, while antibodies against CGRP failed to provide analgesic relief for human OA knee pain⁸⁹, data in non-arthritis rodent models of pain have demonstrated more robust effects in female compared with male animals⁹⁰. Functional MRI imaging has been used in rat OA, albeit in the monosodium iodoacetate model, where differences in the periaqueductal grey matter connectivity was seen in females. Such findings would be consistent with increasing the predisposition to developing chronic pain ⁹¹.

Future directions

It is fair to say that good progress has been made in the past few decades at unravelling molecular mechanisms that underly pain in OA, aided by careful mechanistic molecular studies and clinically relevant in vivo models. However, we still do not have treatments that improve the lives of patients and there is still a nudging concern that targeting pain without targeting structural disease may have detrimental outcomes regarding structural disease, at least in a subset of patients. It is intuitive that powerful analgesia per se is going to encourage increased use of a joint and thereby drive the very mechano-inflammatory pathways that drove disease in the first place ⁹². But there are other possible ways in which neutralizing targets, such as NGF, could worsen joint damage. In the case of NSAIDs, these were shown to suppress the normal anabolic response of damaged cartilage in vitro. In the case of NGF, its receptors (TrkA and p75) are expressed on a variety of non-neuronal cell types and may therefore have additional effects on mesenchymal tissues of the joint. For instance, p75 is known to be expressed on mesenchymal stem cells within the synovial fluid and osteoblasts. Chondrocytes also express both TrkA and p75 but the role of these has not been elucidated ⁹³. On the other hand, while CCR2 is also expressed by multiple cell types within the joint, including nerves, its deletion does not appear to lead to worsening structural disease. Indeed in some studies mild structural protection has been demonstrated. Finally, there is some emerging evidence that local ablation of knee innervating nerve afferents, for example through the use of resiniferatoxin or cryoneurolysis, may provide analgesic relief, but whether this strategy will have adverse structural effects is not yet clear⁹⁴. Therefore, considering how targeting may affect not only the target cell type but all cells within the joint would likely help to refine how we can harness the analgesic powers of a given drug whilst minimising effects on other joint tissues.