The role of joint innervation in the pathogenesis of arthritis

Role of neuropeptides

About one-third of the unmyelinated free nerve endings in joints contain substance P or calcitonin gene-related peptide (CGRP), or both,15^,16 peptides that are potent inflammatory mediators. Substance P is a vasodilator, which increases capillary permeability, induces mast cell de-granulation and is a potent leukocyte chemotactic agent.17^,18 It is also mitogenic and chemotactic for endothelial cells and fibroblasts. CGRP is a potent vasodilator, a mitogen for endothelial cells and a potent inhibitor of insulin-mediated glycogen synthesis. 19 Acting together, these neuron-derived peptides can significantly potentiate the inflammatory response, and they are released into the joint and peri-articular tissues during inflammatory arthritis.20

Tissue injury increases the local expression of nerve growth factor by fibroblasts in response to tumour necrosis factor-α or interleukin-1β.21^,22 Levels of nerve growth factor in synovial fluid are significantly increased in patients with inflammatory arthritis. 23 Nerve growth factor is a trophic factor for mast cells, small diameter peptidergic neurons and sympathetic efferent neurons. Substance P and CGRP messenger RNA (mRNA) and peptide content in the dorsal root ganglia and dorsal horn of the spinal cord are increased dramatically after induction of experimental inflammatory arthritis as a result of increased expression of nerve growth factor in the damaged target tissue.17^,24^–29

Two classes of peripheral nociceptor

About one-third of unmyelinated sensory fibres do not express substance P or CGRP. Recent investigations of these neurons have revealed enough characteristic differences to suggest that they should be considered in a separate functional class. They express a receptor for a different neurotrophic factor called glial-derived neurotrophic factor and also selectively express receptors for adenosine triphosphate that are not found on neuropeptide-containing neurons. Furthermore, this subgroup of unmyelinated sensory neurons exhibits a different pattern of axon termination in the dorsal horn of the spinal cord.30^,31 The significance of these differences remains to be determined, but there is a clear indication that these different neuronal subtypes may mediate different types of pain and may therefore be selectively blocked by appropriate ligands.

Sensitization of joint afferent neurons during inflammation

Pain is one of the cardinal signs of arthritis and is usually associated with allodynia (pain to innocuous stimuli) and hyperalgesia (amplified pain to noxious stimuli). These perceptual changes are now known to reflect cellular and molecular changes in the cells of the dorsal root ganglia and the spinal cord. Allodynia and hyperalgesia come about because of changes that originate almost exclusively in the smaller diameter thinly myelinated and unmyelinated populations of sensory fibres. These neurons exhibit enlargement of their receptive fields and decreased sensory thresholds to mechanical and noxious stimuli.4 Many neurons begin to fire spontaneously, and others fire at much higher rates in response to stimulation. Although many of the well-known inflammation-associated mediators (e.g., bradykinin, serotonin, prostaglandins) have been proposed as causes of or contributors to these changes in neuronal response properties, it was recently shown that almost all of the observed changes in the behaviour of these neurons may be induced by locally increased concentrations of nerve growth factor.32^–34

Role of autonomic fibres

Vasomotor reflexes become significantly altered during the inflammatory response to joint inflammation. Normal joint tissues have a relatively low baseline blood flow, which can be modulated by stimulation of sympathetic efferents or application of substance P and CGRP.35^,36 These reflexes are abolished by adjuvant arthritis, which induces significant increases in joint blood flow that can no longer be altered by sympathetic stimulation or the application of neuropeptides.37

Sympathetic fibres make a contribution to plasma extravasation. Increased tissue levels of bradykinin induce the release of serotinin (5-hydroxytryptamine) from mast cells and platelets. Serotonin mediates plasma extravasation from the synovial microcirculation by actions on 5-hydroxytryptamine-2A receptors that are found on sympathetic efferent terminals. It is thought that activation of the 5-hydroxytryptamine-2A receptor induces release of prostaglandins from the sympathetic terminal by a mechanism independent of action potentials or membrane depolarization. These effects are blocked by sympathectomy or selective antagonists at the 5-hydroxytryptamine-2A receptor.38

Sensitization of intraspinal second order sensory neurons

Dorsal horn neurons of the spinal cord receiving input from joints may be classified into 2 types: nociceptive specific and wide dynamic range. Nociceptive-specific neurons respond only to noxious or damaging stimuli. Wide-dynamic-range neurons show a graded response to various stimuli, with very high outputs in response to noxious stimuli.9 Just as inflammation-associated allodynia and hyperalgesia arise from lowered thresholds of primary afferent neurons, these sensory disturbances also reflect significant increases in excitability of the both the wide-dynamic-range and nociceptive-specific types of second order sensory neurons in the spinal cord that show lowered thresholds, higher firing rates and sometimes spontaneous discharges. In recent years, the causes of these changes in excitability have been elucidated. Increased sensory traffic from the inflamed joint leads to localized increases in dorsal horn levels of glutamate, substance P and CGRP. Hyperexcitability of the second order neurons deep in the dorsal horn is principally caused by the actions of glutamate upon metabotropic glutamate receptors43 and of GABA (presumably released from interneurons) upon the GABA_A receptor41 but is further augmented by the actions of intraspinal substance P and CGRP.44^,45 Blockade of the metabotropic glutamate receptor or the GABA_A receptor can prevent the development of hyperexcitability of the second order sensory neurons in the spinal cord without affecting normal responses to innocuous and noxious stimuli.41^,43

Recent experiments have revealed that an experimental inflammatory arthritis can also increase substance P and CGRP production in the dorsal root ganglia and spinal cord contralateral to the inflamed joint and even induce a symmetrical inflammation in the contralateral joint.46^,47 These observations lend further support to the evidence that afferent and spinal mechanisms play a key role in the initiation and promotion of inflammation. These “contralateral effects” can be prevented by blockade of the “dorsal root reflexes” with appropriate antagonists to glutamate and GABA at the appropriate receptors.

Charcot’s arthropathy

Charcot’s arthropathy is a severe form of degenerative arthritis that occurs in patients with peripheral sensory neuropathy.49^–52 Although Charcot believed the disease was caused by a loss of trophic substances derived from the nerves, leading to atrophy of the joint tissues, the views of Volkmann have subsequently dominated modern thinking.51 Volkmann and later Virchow argued that the loss of limb sensation compromised muscular function in the limb, leading to mechanical damage to the joint.53 This became the consensus opinion of subsequent investigators. Although there has been considerable speculation, the potential relationship of the pathogenic mechanisms of Charcot’s arthropathy to the common forms of osteoarthritis remains controversial.12^,13^,54^,55

Osteoarthritis, aging and joint innervation

Osteoarthritis is one of commonest disabling conditions associated with aging. The prevalence of osteoarthritis correlates directly with age and shows an increase with every decade after the age of 25 years. Age is the highest risk factor for osteoarthritis,56 and 27%, 34% and 44% of the population exhibit radiographic osteoarthritis of the knee in the seventh, eighth and ninth decades of life respectively. 57^,58 Yet the normal aging of articular cartilage and subchondral bone does not presage the pathological changes found in osteoarthritis,56 and the underlying cause of the disorder remains unknown. As Brandt59 suggested, the end stage of osteoarthritis represents “joint failure,” but the primary defect may lie in cartilage, synovium, subchondral bone, ligament or the neuromuscular system. Indeed, it has been repeatedly suggested that a failure of protective neuromuscular reflexes could predispose people to osteoarthritis. 60^–62

A number of investigations in humans have hinted that neural mechanisms may play an important protective role in normal joint function. For example, a study using metal-impregnation stains to examine biopsy specimens of human knee joint capsule and ligament obtained from elderly patients with osteoarthritis at the time of joint replacement surgery failed to detect any of the neural elements that could be found in newborn or adolescent tissue.63 Another more recent study identified a group of seemingly normal subjects with very subtle deficiencies in motor coordination during walking gait, manifested by abnormal peak loading during the early stance phase.62 This subgroup of “normal” people might be particularly dependent on joint afferent systems to protect their joints from damaging loads or might even have a deficiency of joint afferent innervation. Taken together, these studies lend support to speculations that an age-related loss of joint sensory innervation could cause or accelerate the development of common human osteoarthritis. 13^,60^,61

Neurosensory function of ligaments

In recent years these speculations have been fuelled by experiments that found that simple knee joint denervation in dogs produced minimal to no cartilage degeneration at up to 16 months. However, if denervation was combined with transection of the anterior cruciate ligament, then an extremely rapid and progressive knee joint degeneration did ensue.64^,65

These observations may be partially explained by electrophysiological studies in cats that have clearly established the existence of the neural circuitry necessary to allow joint afferent neurons to mediate functionally important motor reflexes that are appropriate to protect joint surfaces from damaging loads.12^,13^,54^,66^–72 Briefly, ligament-associated mechanoreceptors have been found to have a major influence on muscle spindle activity by regulating γ-motor neuron activation of the intrafusal muscle fibres. Stimulation of receptors in the cruciate ligaments, for example, increases γ-motor neuron outflow, muscle spindle sensitivity and muscle tension in both the flexor and extensor muscles of the knee joint.69^,70^,73 The net effect is a reflex increase of limb stiffness in response to applied loads. This balancing of muscle activities acting across the joint increases joint stability, which could be expected to protect the joint surfaces from mechanical damage (Fig. 1).

• Download figure
• Open in new tab
• Download powerpoint

FIG. 1

Diagram of a knee joint and associated spinal segment illustrating how joint receptors function as part of a feedback loop within the γ-motor neuron and muscle spindle receptor feedback loop. (A) Under normal circumstances, ligament stretch increases firing in mechanoreceptors, which increases the γ-motor neuron activity. This, in turn increases muscle spindle sensitivity to stretch, which feeds back on to the α-motor neuron, thereby increasing muscle tension, which increases joint stability. (B) Loss of joint afferent input due to ligament disruption or neuronal loss leads to a reduced γ-motor neuron outflow and ultimately a loss of muscle force, which decreases joint stability.

Age-related loss of mouse knee joint innervation

Discussions about protective muscular reflexes mediated by joint receptors and their relationship to “idiopathic” degenerative arthritis are of less interest, however, unless there is a loss of joint innervation with aging. Age-related loss of neurons from the central nervous system is a well-established phenomenon but has not been widely documented in the peripheral nervous system. Experiments in our laboratory revealed that in mice, innervation of the knee joint can be accurately and consistently quantified by retrograde tracing with the fluorescent dye Fluoro-Gold (Fluorochrome, Inc., Inglewood, Colo.). When this technique was used in C57BL/6 mice of different ages, a 60% loss of joint afferent neurons over the life span of the mouse was found.74 Most of the loss occurs in the first 6 months of the mouse’s 2-year mean life span and precedes the onset of degenerative arthritis, which has been reported to occur in this strain of mice.75^,76 Rats are genetically closely related to mice but are rarely afflicted by degenerative arthritis. Interestingly, in experiments using Fluoro-Gold labelling of joint innervation in aging rats I found no losses of joint innervation with aging (unpublished data).

Joint receptor function in humans

Because few investigators studying human specimens have attempted to accurately quantify innervation of the knee joint, it remains unknown whether similar losses can and do occur in humans. However, in a series of studies some have attempted to approach this question indirectly by examining proprioceptive function in the hip and knee joint of awake human subjects. These studies did show some impairment of joint position sense with aging and after ligament injury or joint replacement injury.77^–80 However, the relationship of a diminution of the conscious perception of joint position and movement to the type of protective muscular reflexes I have described is questionable at best. It is now clear that the conscious perception of joint position is mediated almost exclusively by muscle spindle receptors.81 The perception of joint or limb position in space is an important and useful sensory function, but it is not the same as the detection and modulation of joint loads, which is an unconscious, reflex motor function. A more useful (although more technically challenging) study of the function of joint receptors and the consequences of their loss would attempt to quantify muscle activities during load-bearing activity. A loss of joint receptor input would be expected to result in a disturbance of the normal balance of muscle forces across the knee joint. Such a study has been done. In 2 recent papers73^,82 it was reported that in a group of patients with chronic anterior cruciate ligament injury there were measurable changes in motor function. Patients and uninjured control subjects were tested while performing isokinetic resisted knee flexion and extension. Patients generated significantly lower electromyographic activity and muscle force in the hamstring muscles when compared with normal subjects performing the same exercise or when compared with the patient’s contralateral uninjured limb. Of particular note is that surgical reconstruction of the anterior cruciate ligament did not change these observed differences.73^,82 These observations are consistent with our current view of ligament-associated mechanoreceptor function derived from the animal laboratory investigations already described. They lend support to the contention that a major element of ligament function is neurosensory and would appear to indicate that the loss of this function is not adequately addressed by the current standard surgical techniques of ligament reconstruction.

Clinical relevance

A cautious interpretation of the currently available data would suggest that if there is an age-related loss of joint innervation in humans, it could be a contributing factor in the pathogenesis of osteoarthritis. Such a loss would likely be of greater significance in association with some type of injury to the joint. In addition, it is wise to recall that the pathogenesis of osteoarthritis is likely multifactorial. Many extrinsic factors such as body mass, activity levels, limb alignment and soft-tissue constraint are also important determinants of the loading history of articular cartilage surfaces, surfaces that almost certainly vary in their tolerance of load, injury and wear.

Current treatments for anterior cruciate ligament deficiency do not address the potentially important neurosensory properties of the native ligament. This may prove to have a significant impact on the ultimate outcome in these patients.