Pathophysiological Response of Nerves to Injury

”Normal,” “Inflammatory,” and “Neuropathic” Pain

Normally, pain is felt when signals originating in thinly myelinated (Aδ) and/or unmyelinated (C) nociceptive afferents reach a conscious brain. Examples are pinprick or a stubbed toe. The pain is always evoked by stimuli; spontaneous pain is not normal. The sensation felt (pain) corresponds in location, time, and quality to the stimulus (noxious) in the expected manner. This is “normal” (or nociceptive) pain.

In addition to evoking acute nociceptive pain, burns, abrasions, chemical irritations, and infections often cause more prolonged pain, both spontaneous and evoked by stimuli. Pain in response to weak, normally innocuous stimuli is “allodynia”; exaggerated pain in response to stimuli expected to be (moderately) painful is “hyperalgesia” (Merskey and Bogduk 1994). In the case of allodynia, at least, tenderness in the “sensitized” tissue (pain) no longer corresponds to the stimulus (non-noxious). This type of pain is usually called inflammatory pain since it is typically accompanied by an immune response and mediated by pro-inflammatory molecules. It should be noted, however, that pungent substances such as capsaicin and (non-injuring) electrical shocks can provoke pain with these characteristics even though they do not damage tissue or elicit an immune response.

Neuropathic pain, like inflammatory pain, involves pathology; it is not “normal,” and it often features sensitization and allodynia and hence mismatch between stimulus (non-noxious) and response (pain). It differs, however, in that it is caused by a lesion or disease of the somatosensory nervous system itself. This distinction is not without problems. Pain from inflammation in a major nerve trunk (“neuritis”) is generally considered neuropathic. On the other hand, even minor trauma to skin, muscle, or joint always injures the terminal part of some nerve fibers. Despite the neuropathy present, however, such pain is rarely neuropathic. The reason is that except in rare cases, the minor neural injury is not the cause of the pain. Neuropathic pain is also distinguished from inflammatory pain by the frequent presence of unique sensory features such as electric shock–like sensations and hyperpathia.

Normal (nociceptive) pain and inflammatory pain are adaptive design features of the intact pain system—an alarm bell. Neuropathic pain, in contrast, reflects faulty, maladaptive functioning of a pain system that has been damaged. Consider a defective alarm system constantly producing false alarms. All three types of pain can and often do co-exist. For example, a space-occupying tumor may simultaneously apply noxious force to otherwise healthy tissue evoking nociceptive pain, trigger an inflammatory response, and directly injure nerves. The relationships among these different pain terms are illustrated in Figure 61-1.

Sensitization

Sensitization by chemical substances, inflammation, and neuropathy may result from PNS and/or CNS processes. The classic explanation of tissue hypersensibility is the “sensitized nociceptor” hypothesis (Lewis 1942). According to this hypothesis, hypersensibility is due to a reduction in the threshold of nociceptive sensory endings, such as in the skin (“peripheral sensitization”). This is probably the right explanation in the case of heat allodynia. Bradykinin and many other inflammatory mediators are known to cause thermosensitive nociceptors to respond to modest warming at temperatures normally too low to evoke pain. The result is “heat allodynia.” If the threshold falls below the ambient temperature, impulse firing and burning pain will appear to be spontaneous. Likewise, sensitized nociceptors show an exaggerated response to suprathreshold heat and mechanical stimuli, including de novo responses of previously insensitive C fibers (Schmidt et al 1994). This yields “heat and mechanical hyperalgesia.” However, the sensitized nociceptor hypothesis does not explain “tactile allodynia” i.e., pain evoked by light touch. The mechanical response thresholds of Aδ and C nociceptors rarely drop into the range that evokes tactile allodynia no matter what inflammatory mediator is applied (Kocher et al 1987; Reeh et al 1987; Koltzenburg et al 1994a; Schmidt et al 1995; Andrew and Greenspan 1999; Banik and Brennan 2004, 2008; Tsuboi et al 2004, Shim et al 2005). Rather, a considerable body of evidence indicates that tenderness to touch is signaled by low-threshold Aβ touch afferents, not sensitized nociceptors.

The radical idea that pain can be signaled by non-nociceptive low-threshold afferents comes from many observations. First and foremost is response time. If sensitized C-fiber nociceptors were to blame, there ought to be a long delay between the tactile stimulus and the pain, about a second for an inflamed finger (≈1-m conduction distance at ≈1 m/sec) and longer for an inflamed toe. One could argue that the immediate response actually experienced is due to sensitized Aδ nociceptors. However, one would then expect that each touch would evoke two volleys of pain, a rapid Aδ-fiber response and then a later C-fiber response (first and second pain). Tactile allodynia is a common, almost an everyday event, and a 1-second delay between stimulus and response could not be missed. Such delays do not occur (Campbell et al 1988). A second argument is that sensitized nociceptors show only a small reduction in the tactile response threshold. As noted above, few if any come to respond to the light brush, touch, and air puff stimuli that evoke allodynic pain. A variety of additional observations involving afferent-selective nerve block, intraneural electrical stimulation, absence of flare, and others support the conclusion that the signal that evokes tactile allodynia is carried centrally by rapidly conducting, thickly myelinated, Aβ, low-threshold mechanoreceptive touch afferents (Campbell et al 1988, Koltzenburg et al 1994b, Torebjork et al 1992).

The existence of “Aβ pain” constitutes a revolution in our understanding of both inflammatory and neuropathic pain. Indeed, since tactile allodynia is an important cause of suffering and disability in patients with neuropathic pain, pain signaled by Aβ touch afferents may be as important as pain signaled by nociceptors. However, how can Aβ fibers, which normally evoke touch, come to evoke pain? This is due largely to altered central processing of the peripheral signal, or central sensitization (Hardy et al 1952, Woolf 1983, Devor et al 1991, Woolf 2011). Central sensitization, in contrast to peripheral sensitization, is not simply a threshold-lowering process. Indeed, in tactile allodynia there is no noticeable change in the response threshold of Aβ touch-sensitive fibers to touch stimuli. Moreover, the sensation that they evoke is not strong touch. It is pain. Rather than simply amplifying, central sensitization changes the modality of the response from touch to pain. This is accompanied by a corresponding change in the cortical areas activated (Maihofner et al 2006). A large variety of electrophysiological mechanisms have been proposed to explain this transformation. Examples include activation of previously blocked N-methyl-D-aspartate (NMDA)-type glutamate receptors, imbalance in chloride ion equilibrium, loss of inhibitory interneurons, activation of glia, and altered descending control. Mechanisms of central sensitization will not be discussed here, although the concept will be referred to frequently.

Ectopia

Neuromas, sprouts, and patches of dysmyelination are structural entities. Whether they contribute to pain depends on whether their formation is accompanied by the development of electrical hyperexcitability. In classic studies, electrophysiological recordings were made from sensory axons that terminate in an experimental nerve-end neuroma, and massive spontaneous discharge of impulses was observed. The firing was generated in the neuroma and was eliminated (transiently) by resection of the neuroma and by local anesthetic block of the nerve end. Likewise, it was enhanced by mechanically probing the neuroma (Fig. 61-3; also see Fig. 61-2; Wall and Gutnick 1974, Govrin-Lippmann and Devor 1978). Similar ectopic electrogenesis (ectopia), both spontaneous and evoked by stimuli, also occurs at mid-nerve locations such as neuromas-in-continuity, disseminated microneuromas, and sites of demyelination such as experimental entrapment neuropathies (Burchiel 1980, Smith and McDonald 1980, Calvin et al 1982, Baker and Bostock 1992, Kajander and Bennett 1992, Wallace et al 2003). Subsequent research revealed ectopic spontaneous and evoked electrogenesis at additional pacemaker locations in injured nerves: in regenerating sprouts (Scadding 1981, Pinault 1995, Chen and Devor 1998, Han et al 2000, Gorodetskaya et al 2003); at sites of nerve inflammation (neuritis) (Eliav et al 2001, Bove et al 2003); in experimental diabetic polyneuropathy (Burchiel et al 1985, Dobretsov et al 2001, Khan et al 2002); after viral infections (Mayer et al 1986, Kress and Fickenscher 2001); after vincristine, taxol, and methylmercury intoxication (Delio et al 1992, Tanner et al 1998, Xiao and Bennett 2008b); and in hereditary demyelinating polyneuropathies (Martini 1997, Gillespie et al 2000). The specific agent that causes neural injury may have some effect, but not a critical one. Pathophysiological discharge can emerge no matter how axons are injured.

In all these cases the discharge is “ectopic” because it originates away from the normal location of sensory impulse generation, the peripheral sensory ending. As expected, activity generated ectopically drives spinal and higher-order neurons in the CNS in the normal way. This has been confirmed by electrical recording, imaging, and the use of activity markers (Mao et al 1992, Chi et al 1993, Sotgiu et al 1994b, Pitcher and Henry 2004). Note that the terms “ectopic” and “ectopia” do not imply that the abnormal discharge necessarily arises spontaneously. Discharge evoked by the application of stimuli at locations where impulse initiation does not normally occur, at the wrist in carpal tunel syndrome, for example, is also ectopic. With one known exception, heat (Hoffmann et al 2008, 2009; Teliban et al 2010), natural non-traumatic stimuli applied at a mid-nerve location do not evoke impulse discharge or refered sensation. In the presence of neuropathy, however, they frequently do. Nerve injury triggers a fundamental change in the midportion of axons that renders them locally hyperexcitable and capable of ectopic electrogenesis.

Firing Rates and Patterns

Spontaneous ectopic discharge occurs in both myelinated (A) and unmyelinated (C) fibers, although with certain differences. Activity in A fibers tends to commence earlier, is found in a higher proportion of neurons at its peak, and is characterized by higher firing rates and more bursting than is activity in C fibers (Fig. 61-4). In A fibers the ectopia is usually rhythmic; that is, there is a fixed interval between subsequent impulses within a train (usually 65–35 msec, which translates to an instantaneous discharge rate of 15–30 Hz). In many fibers the impulse train is interrupted by silent pauses, which results in a bursting, on–off pattern (“interrupted autorhythmicity”; see Fig. 61-3). The remaining A fibers, as well as most C fibers, fire in a slow, irregular pattern (0.1–10 Hz; Govrin-Lippmann and Devor 1978). Spike patterning, particularly bursting, can have an effect on post-synaptic neurons over and above the average firing rate by virtue of temporal summation of excitatory post-synaptic potentials during bursts (Burke et al 1970, Lever et al 2001).

Different afferent types differ in their tendency to develop spontaneous firing or to become ectopically mechanosensitive. However, it is difficult to identify in which functional types ectopia preferentially develops. The problem is that axotomy disconnects the receptor ending from the rest of the axon. Heroic efforts are required to preserve information about the original receptor type over the many hours or days that elapse before ectopic firing begins. Partial information has been obtained from conduction velocity, from the ability to follow tetanic stimuli (the “marking method”), by comparing recordings from dorsal versus ventral roots, and from examination of cutaneous nerves versus nerves serving muscle, viscera, and other tissues. Injured sensory axons are much more likely than injured motor axons to generate spontaneous ectopic activity. Aβ and Aδ afferents are represented roughly in proportion to their numbers in the nerve. Interestingly, although injured cutaneous afferents frequently show ectopic mechanosensitivity without spontaneous firing, muscle afferents are more likely to fire spontaneously, at least after distal axotomy (Johnson and Munson 1991, Tal et al 1999, Michaelis et al 2000). Perhaps this is because muscle nerves normally contain many tonic and slowly adapting afferent fiber types (muscle spindle afferents, proprioceptors). Neuroma end-bulbs and sprouts tend to develop the same sensitivities that they had before injury, presumably because of fiber type–specific gene expression at the level of the cell soma (Devor et al 1990, Koschorke et al 1991, Teliban et al 2010).

Neurolysis As a Therapeutic Strategy

Since neuromas are an important source of painful ectopia, it seems logical that surgical resection or neurolysis should bring relief. Unfortunately, shortly after neuroma resection the same pathophysiological processes that caused pacemaker activity in the first place are re-engaged at the freshly cut nerve end, perhaps even in intensified form as a result of priming. Surgical mobilization of a neuroma to a site with a reduced likelihood of mechanical compression, however, may provide long-term relief in cases in which mechanosensitivity rather than spontaneous firing is the main problem (Campbell 2007). Different nerves behave differently. An incision for thoracotomy, for example, which frequently damages intercostal nerves, is much more likely than a comparable incision in the abdomen to be followed by neuropathic scar pain. For reasons that are not entirely clear, destruction of the tooth pulp in root canal treatment or tooth extraction does not commonly induce a painful neuroma. The same seems to be true of the intrinsic innervation of long bones. In hip replacement surgery, the femur is cut across and a prosthetic joint and bone cement are introduced into the bone marrow chamber. Yet despite the massive destruction of intrinsic bone afferent axons, the development of chronic neuropathic pain is infrequent. This suggests that pain relief might be achieved in patients with osteoarthritis by intrinsic denervation of the epiphyseal end of the bone, just as in dental root canal treatment (Niv et al 2003).

Microneurography in Humans

The method of percutaneous microneurography has extended observations on ectopia to awake humans, including those with neuropathic pain. It remains a research rather than a diagnostic tool, however, because of its technical difficulty and intrinsic risk. Practitioners have been justifiably reluctant to insert microelectrodes into already problematic nerves. Nonetheless, enough studies have appeared to make it clear that ectopic hyperexcitability occurs in patients as in nerve-injured animals and that it is a fundamental contributor to many clinical neuropathic pain conditions.

Not long after the first observations in animals, Nystrom and Hagbarth (1981) carried out a pioneering study in which they documented ongoing firing in the peroneal nerve in a lower extremity amputee who had ongoing phantom foot pain. Percussion of the neuroma evoked stabbing pain (the Tinel sign) and an intense burst of spike activity. The evoked bursts and the evoked pain were eliminated by local anesthetic block of the neuroma. Interestingly, however, much of the ongoing discharge persisted. The DRG is the most likely source of this persistent activity, but direct evidence of this in humans is still lacking. Subsequent microneurographic studies documented a direct relationship between ectopia and pain in a variety of other neuropathic conditions (e.g., Nordin et al 1984, Campero et al 1998, Burchiel and Baumann 2004, Ochoa et al 2005, Jorum and Schmelz 2006). For example, painful dysesthesia triggered by straight-leg lifting in a patient with radiculopathy (Lasegue’s sign) was accompanied by evoked ectopic bursts recorded in the sural nerve. The ectopic source was the injured spinal root or DRG (Nordin et al 1984).

More recent studies have used the “marking method” to resolve activity in individual C fibers. The results provide evidence that the ongoing, often burning pain characteristic of many peripheral neuropathies is due to spontaneous discharge in C-fiber nociceptors (Serra 2010, Kleggetveit et al 2012). Multiplet and burst firing, afterdischarge, and other interesting peculiarities of ectopia in animal models (below) have also been seen in patients, thus further strengthening the clinical relevance of the experimental models (Weidner et al 2002, Bostock et al 2005). Other lines of evidence support these electrophysiological data. For example, pain is evoked by the application of substances known from animal preparations to excite ectopic pacemaker sites, including K⁺ channel blockers and adrenergic agonists (Chabal et al 1989b, 1992). Likewise, blockers of ectopia such as local, regional, and systemic anesthetics suppress neuropathic pain (Wallace et al 1996). Even the most severe pain, such as occurs in complex regional pain syndrome (CRPS), is reliably stopped by peripheral nerve or brachial plexus block, thereby permitting physiotherapy, albeit only for the duration of the block.

Animal Models of Spontaneous Pain

The foregoing observations leave little doubt that spontaneous ectopic discharge is a primary driver of spontaneous pain in humans. Spontaneous ectopia also occurs in neuropathy models in animals. Although this implies that the animals also experience spontaneous pain, proving it is not trivial. This issue is important because animal models are essential for screening novel analgesic drugs. This topic is discussed in more detail in Chapter 62.

Spontaneously emitted behaviors such as vocalization, abnormal posture and gait, and unprovoked paw lifting occurs in neuropathy models and have been put forward as potential markers of spontaneous pain. However, follow-up study has failed to provide convincing support for this conjecture (Mogil et al 2010, Urban et al 2011). Instinctive behavioral biomarkers of spontaneous pain may not even exist in prey species such as mice and rats since they would signal vulnerability to predators and natural selection is likely to have excluded them. Convincing evidence for spontaneous pain is available from a conditioning paradigm in which animals were trained to show place preference for a chamber in which they were provided with analgesia (King et al 2009).

Autotomy behavior in the neuroma model provides another option (Wall et al 1979). This behavior has been validated as a biomarker of spontaneous neuropathic pain (anesthesia dolorosa) via a number of approaches (Devor 2007). For example, it is reduced by appropriate drugs, it is associated with prominent ectopia, augmenting the ectopia augments the autotomy, and suppressing it suppresses autotomy (Devor et al 1982, Levitt 1985, Coderre et al 1986, Blumenkopf and Lipman 1991, Seltzer 1995, Kauppilla 1998, Liu et al 2001a, Abdulla and Smith 2002, Devor 2007, Nissenbaum et al 2008, Radtke et al 2010). Recently, it has been shown that a gene variant that predisposes to autotomy in rodents is associated with neuropathic pain in humans (Nissenbaum et al 2010). Two other end points proposed recently also offer some hope: the coding of facial expressions in mice (Langford et al 2010) and a new procedure for tracking ultrasonic calls (Kurejova et al 2010).

Mechanosensitivity: Hot Spots, Trigger Points, and the Tinel Sign

Exploring the surface of an injured nerve with a fine probe reveals clusters of tiny mechanosensitive “hot spots.” They do not occur in normal nerves. Such probing excites silent axons and also accelerates discharge in already active ones (Fig. 61-2A and B). Spontaneous and evoked discharges almost certainly arise from the same cellular locus and pacemaker process (Burchiel 1984, Chen and Devor 1998). Mechanosensitivity underlies the Tinel sign, the often stabbing or electric shock–like dysesthesia evoked by percussion of neuromas, along nerves in painful diabetic neuropathy, and in carpal tunnel syndrome, and it is responsible for the shooting leg pain in sciatica (Nystrom and Hagbarth 1981, Kuslich et al 1991, Dorsi et al 2008). Like normal mechanosensitive endings, hot spots typically respond either with a short spike burst at the onset and/or release of a stimulus or with sustained firing for the duration of application of the force (rapid and slow adaptation). Interestingly, however, in some hot spots firing persists beyond the end of the stimulus (“mechanical afterdischarge,” Fig. 61-6A). In effect, the brief stimulus triggers a period of spontaneous firing. This clearly abnormal pattern accounts for the frequent persistence of evoked sensation in neuropathy (Noordenbos 1959, Gottrup et al 2003). Locations where nerves run adjacent to tendons and bone (e.g., the carpal tunnel) or where small nerve branches cross tough fascial planes are particularly at risk for sustaining focal trauma and developing ectopic mechanosensitivity. Pain evoked at such trigger points by local palpation, weight bearing, and untoward movements may represent a neuropathic contribution to chronic musculoskeletal pain, perhaps including fibromyalgia.

Ectopia and Tactile Allodynia

Tactile allodynia, unlike hyperalgesia and tender points, is probably not a result of excessive mechanosensitivity of nociceptive sensory endings for reasons noted above (under Sensitization). Even though tender collateral sprouts and intradermal disseminated microneuromas may well contribute, the major cause of tactile allodynia is thought to be spike activity in Aβ touch afferents processed by sensitized spinal cord circuitry. Here too, however, spontaneous ectopia in the periphery plays a crucial role: it induces and maintains the central sensitization. Thus, tactile allodynia results from pathophysiology in the PNS and in the CNS. Note that the sensory effect of spontaneous ectopia in Aβ fibers is also exacerbated by central sensitization, thereby adding to the spontaneous pain. Finally, central sensitization may amplify the sensory consequences of spontaneous and evoked activity in at least some types of nociceptors (Klein et al 2004).

In Figure 61-7 the sequence of events believed to underlie the tactile allodynia in neuropathy is illustrated by using the spinal nerve ligation model of neuropathic pain (Kim and Chung 1992). In this model the L5 (or L5–6) spinal nerve is cut; about half the afferents in the sciatic nerve are severed and the hindpaw is partially denervated. Cutting the nerve creates two permanent sources of spontaneous ectopic activity, the L5 spinal nerve-end neuroma and the L5 DRG. “Uninjured” nociceptors in the L4 segment may also contribute. This activity drives central pain pathways and also triggers central sensitization, hence augmenting spontaneous pain and inducing allodynia.

Corresponding observations have been made in patients. For example, Gracely and colleagues (1992) observed that when a focal source of ectopia from an old scar was identified and blocked with local anesthetic, not only was the focal pain temporarily eliminated but also the widespread allodynia that the patient suffered. The same principle is also relevant to conditions such as migraine and visceral and musculoskeletal pain (Burstein et al 2000, Giamberardino 2008, Graven-Nielsen and Arendt-Nielsen 2010, Woolf 2011). The focal impulse source dynamically maintains central sensitization. Remove it and hypersensibility fades, usually within minutes or at most a few hours.

Exacerbation of Ectopia by Temperature and by Chemical Mediators

Cold intolerance and cold allodynia are common symptoms in neuropathy; people who live in cold climates suffer in particular during the winter (Lindblom 1994). On the other hand, in some conditions, notably CRPS, patients may obtain relief by wrapping the painful limb in a cold, wet towel. This difference might reflect which primary afferents are involved (Matzner and Devor 1987, Hoffmann et al 2008). Ectopic activity in C fibers tends to be excited by cooling and suppressed by warming. Collateral sprouts (which derive mostly from C fibers) and many intradermal disseminated microneuroma endings are therefore expected to be hypersensitive to cold and to provoke pain when the skin is cooled. In contrast, ectopia in A fibers is enhanced by warming, whereas cooling suppresses it. If the pain in CRPS were primarily due to Aβ fibers, this could explain why cooling provides some relief in this condition. As discussed below, the abnormal sensitivity of injured axons reflects altered regulation and trafficking of ion channels and receptors, including thermoresponsive transient receptor potential (TRP) channels. This varies with fiber type (Schmidt et al 2009).

In addition to temperature, metabolic stress and a wide array of chemical factors can also depolarize sensory neurons and directly excite discharge at ectopic pacemaker sites in neuropathy. All these factors can also sensitize pacemaker sites and make them hyper-responsive to mechanical and other stimuli. Examples include tissue ischemia and hypoxia, changes in blood gases, elevated blood glucose, altered ion concentrations, and numerous endogenous neuroactive substances, including catecholamines, adenosine triphosphate (ATP), nitric oxide, peptides, cytokines, small lipids, neurotrophins, histamine, bradykinin, prostaglandins, TNF-α, and many other pro-inflammatory mediators (Wall and Gutnick 1974, Zimmermann et al 1987, Devor et al 1992b, Noda et al 1997, Levy et al 2000, Rivera et al 2000, Zhou et al 2000, Chen et al 2001, White et al 2005, Binshtok et al 2008, Patwardhan et al 2010, Xu et al 2010). New mediators continue to appear regularly, which suggests that the current inventory is far from complete. The chemical milieu of afferent neurons is a major factor in ectopia and neuropathic pain, but only if ectopic pacemaker capability has developed first.

Systemic and Focal Inflammation (“Neuritis”)

In Guillain-Barré syndrome (GBS) and related inflammatory polyneuropathies, nerves suffer widespread myelin destruction and axonopathy as a result of autoimmune attack. This causes paralysis and sensory loss. However, as in other forms of neuropathy, these negative symptoms are often accompanied by dysesthesias and ongoing pain (Moalem-Taylor et al 2007, Luongo et al 2008, Koike and Sobue 2010). Membrane remodeling and pain behavior have been documented in the animal model of GBS (experimental allergic neuritis), and ectopic hyperexcitability is probably present in this model (Novakovic et al 1998, Sorkin 2007). There is direct electrophysiological evidence that ectopic hyperexcitability occurs at sites of focal nerve inflammation. In the chronic constriction injury model of neuropathic pain, for example, chromic gut ligatures are applied loosely to the sciatic nerve. This causes local inflammation and nerve swelling, which partially strangles the nerve (Bennett and Xie 1988). Variants of this model include focal constriction of trigeminal nerve branches and compression of dorsal root and trigeminal ganglia (Vos et al 1994, Hu and Xing 1998, Zhang et al 2000, Ma and LaMotte 2007, Ahn et al 2009). Such injury causes frank destruction of many axons, focal demyelination of others, and the release of inflammatory mediators.

Peripheral nerve injury also triggers an inflammatory reaction in associated ganglia, including activation of intrinsic glial cells and attraction of invasive immune cells. An inflammatory milieu contributes to the development of ectopic firing at both the nerve injury site and the ganglia, with consequent spontaneous pain and hypersensibility (Kajander et al 1992, Tal and Eliav 1996, Homma et al 2002). The animal constriction and compression models emulate the many clinical conditions that feature focal neuropathy exacerbated by inflammation. Among these are carpal tunnel syndrome, disc herniation with sciatica, and solid tumors that press on nerves. Indeed, a contribution of inflammation in conditions involving neural trauma may be the rule rather than the exception.

There is also tentative evidence that inflammatory mediators might be able to act on mid-nerve fibers directly without concurrent (structural) neuropathy. Specifically, it has been reported that when TNF-α or complete Freund’s adjuvant is applied to the surface of healthy nerves or DRGs, a focus of spontaneous firing and mechanosensitivity may emerge rapidly (Sorkin et al 1997, Eliav et al 2001, Homma et al 2002, Wang et al 2007, Amaya et al 2009). By inference, in systemic and focal inflammatory and toxic neuropathies in which there is minimal concurrent nerve trauma, widespread pain might result from the direct action of inflammatory mediators on midaxon and intraganglionic receptors.

Sympathetic–Sensory Coupling and Sympathetically Maintained Pain

Sympathetic–sensory coupling is a special case of ectopic chemosensitivity. In some patients neuropathic pain appears to be exacerbated by sympathetic efferent activity and relieved by sympathetic block or sympatholysis. This is “sympathetically maintained pain” (SMP). Pain may be evoked by direct electrical stimulation of sympathetic efferents or by maneuvers that increase sympathetic drive, such as the Valsalva maneuver and whole-body cooling (Harden et al 2001, Schattschneider et al 2008). Pain also occurs in animals and in humans when adrenergic agonists are injected into nerve-end neuromas or into the skin in patients with SMP, and it may be rekindled in individuals whose hyperalgesia was previously relieved by treatments that reduce the endogenous sympathetic drive (Chabal et al 1992, Torebjork et al 1995, Ali et al 2000). These clinical observations correspond to the abnormal adrenosensitivity of ectopic pacemaker sites (injured axons and DRG cell somata). As noted, residual “uninjured” afferent endings in the skin also become adrenosensitive, perhaps in association with collateral sprouting (Wall and Gutnick 1974; Devor and Janig 1981; Sato and Perl 1991; Ali et al 1999, 2000).

SMP does not result from excess sympathetic drive as was once assumed. If anything, sympathetic drive is decreased in nerve-injured patients (Harden et al 2001). Rather, sympathetic–sensory coupling reflects the increased adrenergic sensitivity of sensory neurons, presumably caused by either an excess of adrenoreceptors, elevated baseline excitability, or both. Studies using receptor-selective agents indicate the predominance of α₂ pharmacology, but α₁-adrenoreceptors also contribute (Janig 1990, Devor et al 1994a, Chen et al 1996). Two recent studies based on microneurography investigated adrenosensitivity in cutaneous C nociceptors innervating painful skin in patients with CRPS. Although CRPS has long been considered to be the prototypical SMP, this is uncertain. Unfortunately, the results of the two studies were discordant (Campero et al 2010, Orstavik and Jorum 2010).

A unique and potentially important form of sympathetic–sensory coupling occurs within the DRG. In nerve-injured animals and in humans, sympathetic efferents that normally serve local blood vessels proliferate within the ganglion and form unique basket-like skeins around sensory somata. Since the sympathetic axons have not been cut, this is an instance of collateral or “reactive” (not regenerative) sprouting. The sprouting (and associated proliferation of satellite glial cells) may be induced by NGF released by axotomized DRG neurons. Axotomy triggers both elevated NGF synthesis and ectopia, which is expected to enhance NGF release (Devor 1983b, McLachlan et al 1993, Shinder et al 1999).

Role of Ectopia in Tissue Edema and Trophic Changes

Impulses that originate in nerve-end neuromas can travel only centrally, toward the CNS. However, impulses that originate in the mid-nerve region can also propagate distally (antidromically). Mid-nerve pacemaker sites include patches of demyelination and the DRG. Antidromic impulses also arise near sensory endings when one of several preterminal axon branches is abnormally active (“axon reflex”). In each of these cases the pacemaker site drives a sustained barrage of impulses centrally, which causes pain, and an identical barrage antidromically to invade peripheral tissue. The antidromic discharge may play a significant role in neuropathy by releasing vasoactive peptides from the axonal shaft (Sauer et al 2001) and from sensory endings (Ninian 1913, Lembeck and Holzer 1979, Lotz et al 1988, Hsieh et al 1996, Harden et al 2001). Some, notably substance P and calcitonin gene–related peptide (CGRP), cause “neurogenic inflammation,” that is, local vasodilation and hence warming and reddening of the skin, combined with plasma extravasation from postcapillary venules and hence edema. If such activity is sustained over time, it might induce the trophic changes in skin, nails, and bone that are so characteristic of CRPS and some other neuropathies. It might even sensitize nociceptors (but see Reeh et al 1986).

The Blood–Brain Barrier and Neurogenic Inflammation in the Spinal Cord

It has recently been noted that peripheral nerve injury triggers a breakdown of the capillary endothelial barrier of the intrinsic vasculature of the spinal cord. This permits plasma proteins and even circulating immune cells (T lymphocytes) to enter the cord (Gordh et al 2006, Costigan et al 2009, Beggs et al 2010). The process parallels and might even cause “glial activation” in the cord. Glial activation with associated intraspinal release of inflammatory cytokines is thought to be a significant contributor to central sensitization (Watkins and Maier 2002). The mechanism whereby nerve injury brings about breakdown of the spinal blood–brain barrier (BBB) has not been established yet. A likely possibility is the sustained central release of neuropeptides as a result of ectopic pacemaker discharge. This is the central equivalent of cutaneous neurogenic inflammation in the skin.

Induction by Nociceptive C Fibers

Most of the evidence available indicates that the signal that induces and maintains central sensitization in neuropathy is electrical impulse activity originating in the periphery. A less likely possibility is a molecule or molecules delivered centrally by axoplasmic transport independent of spiking. Acute tissue injury and electrical nerve stimulation at C-fiber strength reliably trigger central sensitization within seconds or minutes (Woolf 1983). This is far too fast for axoplasmic transport. Likewise, neuropathic conditions in which there is sustained firing in afferent C fibers often feature tactile allodynia. C-fiber activity originating in inflamed skin also evokes central sensitization, as reflected by tactile allodynia in the region in and surrounding the injury (“primary and secondary hyperalgesia”). In each of these cases silencing the ongoing nociceptor activity (e.g., by cooling or local anesthesia) rapidly reverses central sensitization and the consequent allodynia (Gracely et al 1992, Sheen and Chung 1993, Koltzenburg et al 1994b, Yoon et al 1996, Zhang et al 2000, Sukhotinsky et al 2004, Wen et al 2007, Pitcher and Henry 2008, Thacker et al 2009, Xie et al 2009).

The rapidity of the change, often within minutes, indicates that maintenance of central sensitization by ongoing impulse activity is dynamic. There is little solid evidence that spinal cord sensitization can become “centralized” and independent of peripheral drive. Consider childbirth, the passage of a kidney stone, or hip replacement surgery. Even after weeks or years of sustained nociceptor input, the pain and allodynia resolve quickly when the definitive peripheral driver is identified and removed. There is no “transition to chronicity” except perhaps in the sense of psychosocial deterioration. Likewise, prior afferent block does not prevent the later emergence of pain (“pre-emptive analgesia”) in cases in which peripheral drive continues to be present after the block has worn off (Niv et al 1999, Dahl and Kehlet 2011). Unfortunately, definitive removal of peripheral pain drivers is often beyond our reach.

Gently stroking healthy skin, or nerve stimulation at intensities that activate Aβ fibers only, does not induce central sensitization or pain. For this reason, conventional thought is dominated by the idea that nociceptor input is both sufficient and necessary for triggering and maintaining central sensitization. The presumed mechanism is that neurotransmitters released from nociceptors, but not from Aβ touch afferents, are required to initiate the sensitizing process. However, in animal models of traumatic neuropathy, ectopic spontaneous activity occurs overwhelmingly in Aβ afferents, at least in the early stages of pain when tactile allodynia first appears (see Fig. 61-4; Boucher et al 2000; Han et al 2000; Liu et al 2000a, 2000b; Xie et al 2005; Tal et al 2006). What, then, triggers and maintains central sensitization and tactile allodynia in these models?

A possible answer is C-fiber ectopia in residual “uninjured” neurons or collateral sprouts, but this is uncertain. The exceedingly low firing rates involved may be insufficient to drive central sensitization. Likewise, nerve injury–induced central sensitization and allodynia persist in animals in which most C nociceptors are eliminated with capsaicin and related neurotoxins (Shir and Seltzer 1990, Ossipov et al 1999). In contrast, a variety of procedures that suppress ectopia in axotomized Aβ fibers, without affecting neighboring “uninjured” C fibers, do relieve tactile allodynia (Sheen and Chung 1993, Yoon et al 1996, Na et al 2000, Sukhotinsky et al 2004, Naik et al 2008). Ectopic activity does occur in some Aδ afferents soon after axotomy, presumably including Aδ nociceptors, but the bulk of the activity takes place in Aβ afferents. Can we be certain that afferent nociceptors are the only afferents capable of triggering and maintaining central sensitization?

A Role for Ectopic Activity in Aβ Afferents in the Induction of Central Sensitization

Normally, Aβ touch fibers clearly play no role in the induction of central sensitization and allodynia. However, there is tantalizing evidence that cellular changes triggered by inflammation and axotomy might render Aβ afferents capable of doing so (Devor 2009). For example, following axotomy, Aβ afferents rapidly begin to express substance P and CGRP, the peptides present in nociceptors that are thought to mediate C fiber–induced pain and central sensitization (Noguchi et al 1995, Weissner et al 2006). This is one aspect of the cellular reprogramming (“phenotypic switching”) noted above. Once switching has occurred, impulse activity in injured Aβ fibers releases these peptides in the spinal cord. This, in turn, triggers central sensitization by a direct action on spinal neurons, via glial intermediates, or by opening the BBB (Pitcher and Henry 2004). Interestingly, phenotypic switching in touch afferents appears to account for the genotype-selective neuropathic pain behavior in rats selected for high versus low predisposition to neuropathic pain in the neuroma model (Nitzan-Luques et al 2011). CGRP in Aβ afferents is up-regulated in the pain-prone line but not in the pain-protected line. Phenotypic switching also occurs in animal inflammation models. In these models, activation of cutaneous Aβ fibers by gently brushing the inflamed skin triggers central sensitization (Ma and Woolf 1996).

Additional evidence that activity in injured (but not normal) Aβ fibers may both drive pain and trigger central sensitization comes from studies using c-fos expression as an indicator of central sensitization. In healthy animals, stimulation of C fibers triggers c-fos expression in post-synaptic neurons in the superficial and deep laminae of the dorsal horn, but stimulation of intact Aβ fibers does not. However, if there has been previous nerve injury, Aβ-fiber stimulation does induce c-fos expression (Molander et al 1994, Shortland and Molander 1998, Day et al 2001). Again, the likely explanation is phenotypic switching. Overall, these data suggest that contrary to prevailing beliefs, ectopia in Aβ afferents might well participate in central sensitization in chronic pain conditions, both inflammatory and neuropathic (Devor 2009, Suter et al 2009).

Ephaptic Crosstalk

The most widely cited, though not the most important in terms of the number of neurons affected, is “ephaptic” (i.e., electrical) crosstalk. In the mid-1940s Granit and Skoglund (1945) discovered that acute nerve transection short-circuits the insulation between adjacent axons and thereby permits ionic current from one fiber to directly excite neighboring fibers. This coupling vanishes within minutes and is therefore unlikely to be of much functional significance. However, several weeks later ephaptic crosstalk re-emerges, now in an enduring form (Fig. 61-8). Ephaptic crosstalk occurs when there is a sufficient surface area of close membrane apposition between adjacent neurons in the absence of the normal glial insulation (Fig. 61-9; Fried et al 1993). It has been documented in neuromas, regenerating sprouts, and patches of demyelination (Seltzer and Devor 1979, Rasminsky 1980).

Coupled fibers are frequently of different types. For example, nociceptors can be driven ephaptically by stimulation of low-threshold touch afferents (Amir and Devor 1992). Such coupling is an additional potential basis for tactile allodynia, independent of intracutaneous pacemaker sites and central sensitization. When multiple axons are coupled, discharge in one may activate many neighbors. The resulting signal amplification is instantaneous, and the sensation evoked is expected to be lancinating and paroxysmal. Interestingly, ephapsis does not occur in DRGs, intact or following traumatic nerve injury (Devor and Wall 1990), but it can occur after infection by certain strains of herpes simplex virus (Mayer et al 1986). This implicates ephaptic crosstalk as a factor in the pain of post-herpetic neuralgia. Ephaptic crosstalk does not appear to develop between sensory and sympathetic fibers (Devor and Janig 1981) thus undermining the old theory that such coupling accounts for SMP.

Chemically Mediated (Non-ephaptic) Cross-excitation, Crossed-depolarization, and Crossed-afterdischarge

Lisney and Devor (1987) discovered a very different type of cross-excitation among injured axons and termed it “crossed-afterdischarge.” Unlike electrical coupling, where spikes in the passive fibers follow spikes in the active fibers on a one-to-one basis (see Fig. 61-8), firing in the passive cross-excitated fiber outlasts firing in the active fiber (hence “afterdischarge”). Single-stimulus impulses have little effect on the passive fibers, but repetitive activity does have an effect (Fig. 61-10A). When more fibers are activated for a longer duration and at higher activation frequency, stronger cross-excitation is generated.

Cross-excitation also occurs in sensory and autonomic ganglia (Amir and Devor 1996, Matsuka et al 2001, Oh and Weinreich 2002). In both nerves and ganglia it depends on the passive neighbor having repetitive firing capability. In healthy DRGs, electrical stimulation of the axons of neighboring neurons induces depolarization in a large fraction of the cells sampled, but this “crossed-depolarization” is usually subthreshold for spike discharge (Fig. 61-10). However, in neuropathy models, crossed-depolarization reaches threshold for evoking repetitive discharge in many DRG somata (“cross-excitation”). A particularly salient situation is when pacemaker sites are present at mid-nerve locations (e.g., the DRG) but at least some of the axons in the nerve remain in continuity with the skin. In this situation, electrical nerve stimulation, or simply brushing the skin, can evoke crossed-afterdischarge amplification in the DRG (Fig. 61-10B). As in nerves, cross-excitation in the DRG can involve different fiber types and potentially render weak tactile stimulation of the skin painful (tactile allodynia).

Cross-excitation is thought to be due to the buildup of diffusible chemical mediators released into the extracellular space from the activated neurons. It is a unique form of paracrine communication. Repeated stimulation progressively winds up discharge in the passive neighboring fibers as the concentration of the paracrine mediator (or mediators) in the extracellular compartment increases (Fig. 61-10). The identity of the chemical mediator is not yet known with confidence. There is evidence implicating both a transient, stimuation-evoked rise in extracellular K⁺ concentrations (Utzschneider et al 1992) and released ATP acting through P2X₇ receptors (Zhang et al 2007, Gu et al 2010). Non-synaptic chemically mediated neuron-to-neuron interaction of this sort is also known to occur in the CNS, where it is termed “volume transmission” (Agnati et al 2010). The phenomenon is difficult to study in the brain because of simultaneous synaptic interactions. The DRG is an advantageous preparation for studying volume transmission in isolation.

It has recently been shown that nerve injury (and peripheral inflammation) causes the satellite glial cells that surround each neuron in the DRG to become coupled to one another. The coupling, which in this case is mediated by gap junctions, has the potential to affect the discharge of DRG neurons and hence to signal pain (Hanani 2005). In addition, in neuropathy models satellite glia are known to proliferate and become “activated,” much like spinal astrocytes and microglia. Cytokines, neurotrophins, and other mediators released by activated satellite glia can excite DRG (and trigeminal ganglion) neurons in a bidirectional neuron–glia interaction (Obata et al 2006; Takeda et al 2007, 2009; Ohara et al 2009; Xie et al 2009; Gu et al 2010).

Role of Cross-excitation in the Pain Paroxysms in Trigeminal Neuralgia

Crossed-afterdischarge provides a potential explanation for one of the most distinctive, peculiar and devastating of the chronic neuropathic pain states, TN (tic douloureux). The mechanism may also contribute to pain paroxysms in a wide variety of other neuropathies. Patients with TN suffer from dramatic, brief stabbing or electric shock–like “lightning” pains in the trigeminal distribution, either spontaneously or on gentle tactile stimulation of a trigger point on the face or in the oral cavity. The underlying pathology in most patients is thought to be demyelination in the trigeminal root secondary to microvascular root compression. The discovery 50 years ago that TN responds to the anticonvulsant drug carbamazepine (Tegretol) gave rise to the hypothesis that the pain paroxysms in TN reflect seizure-like activity in the trigeminal brain stem.

More recently, however, Rappaport and Devor (1994) proposed an alternative hypothesis more consistent with the known PNS pathology of TN. According to their “ignition hypothesis” pain paroxysms begin with discharge in a small cluster of trigeminal nerve afferents that are activated by cutaneous trigger point stimulation or spontaneously. Crossed-afterdischarge, which may occur at the trigeminal root compression site or in the trigeminal root ganglion, then “ignites” ectopic activity in previously passive neighboring neurons. This activity ignites additional passive neighbors, and they ignite still more neurons. The resulting positive-feedback chain reaction builds up rapidly to an intense, explosive peak. Since neurons of all types become active simultaneously, an event that otherwise occurs only with electric shocks, the sensation felt is electric shock–like. After a few seconds of massive high-frequency firing, activity-evoked aftersuppression dampens the paroxysm, thereby ending the pain attack and establishing a few minutes of refractoriness. The system is then reset, ready for the next pain attack. The ignition mechanism is expected to be sensitive to at least some of the membrane-stabilizing anticonvulsants that suppress CNS seizure activity, notably carbamazepine. However, the drug’s action is supposed to be on PNS ectopia rather than on CNS seizures and is achieved by membrane stabilization rather than by synaptic suppression.

Stimulus Transduction and Spike Encoding

Signal transduction depends on transducer molecules, large proteins that translate stretch, heat, cold, pH, and other stimuli into a “generator depolarization.” They do this by controlling the flow of ions across the membrane of the sensory ending in response to stimuli. A well-known example is the transient receptor potential vanilloid 1 (TRPV1) channel. In the case of TRPV1 the receptor apparatus, the molecular structure that actually senses the vanilloid (capsaicin) stimulus, is an integral part of the larger transducer channel. Alternatively, the receptor may be physically separate from the ion channel and coupled to it functionally. G protein–coupled receptors (GPCRs), for example, do not have an intrinsic ion channel.

The generator potential is not a nerve impulse. It fades to nearly nothing over the first millimeters of the axon and cannot deliver a sensory signal to the CNS, which is much farther away. Only if the generator potential is encoded into a train of impulses can the pain signal reach the brain. This requires a different set of molecules, voltage-sensitive ion channels, especially Na⁺ and K⁺ channels. With some possible exceptions, transduction and encoding molecules are not the same. Both types are required for electrogenesis.

Single-Spike Threshold

The ion channels responsible for encoding are proteins that respond to changing transmembrane voltage rather than to sensory stimuli. Na⁺ and K⁺ channels each come in different “types” (Jenkinson 2006, Dib-Hajj et al 2010). At rest, most Na⁺ channels are closed. The generator depolarization initiated by the sensory stimulus causes the most sensitive types to open. This results in an inward flow of Na⁺ ions, which augments the generator depolarization. If the inward Na⁺ current is small, it is quenched by an outward counter-flow of K⁺ ions. However, if it is a bit larger, the depolarization resulting from the inward Na⁺ current causes more Na⁺ channels to open, and this in turn causes still more to open. The resulting inward torrent of Na⁺ ions overwhelms the outward flow of K⁺ ions and results in an explosive net inward current, the “action potential” (nerve impulse, spike). The crossover point between quenching and spike generation is the “spike threshold.” Note that threshold is not a fixed value. It is the equilibrium point between inward (mostly Na⁺) and outward (K⁺) currents. Its value (in millivolts) is affected by the density and properties of the channels present, and importantly, it is also affected by the rate of depolarization. Gradual depolarization may not evoke a spike (“accommodation”), whereas rapid depolarization of the same magnitude does.

Having opened, Na⁺ channels rapidly close again and membrane polarity returns toward rest. The rate of repolarization is often accelerated by transient opening of voltage-sensitive K⁺-selective ion channels, although at some locations (e.g., the node of Ranvier in some species), only voltage-insensitive K⁺ “leak channels” are present. When voltage-sensitive K⁺ channels are present, the outward flow of K⁺ ions actually overshoots the resting potential and cause a brief post-spike hyperpolarization. In some cells the after-hyperpolarization is followed by a small rebound depolarization, the “depolarizing afterpotential” (DAP). DAPs bring the membrane potential of the cell back toward spike threshold. If large enough and rapid enough, they may trigger a second spike or even a third and fourth (i.e., a spike burst; Fig. 61-12A).

Pacemaker Capability: Threshold for Repetitive Firing

The individual action potential is the basic unit of neural signaling, but sensory experience and certainly chronic pain depend on sustained firing. Some types of sensory endings, including nociceptors, can generate sustained (pacemaker) activity. Others respond only transiently. However, away from the sensory ending, the axon can convey trains of spikes but cannot generate them (except if injury has occurred). Pacemaker sites (natural and ectopic) use two related electrogenic processes. The first is the classic Hodgkin–Huxley process whereby a persistent generator current repeatedly depolarizes the neuron following each spike, thereby drawing the membrane potential from rest back toward spike threshold (Jack et al 1983). K⁺ channels control the rate of depolarization and hence the probability and rate of repetitive firing. The second pacemaker process is an outcome of “resonance,” or the intrinsic tendency of the membrane potential in some cells to oscillate between minor depolarization and minor hyperpolarization. This property is reflected in damped oscillations at the beginning and end of depolarizing steps and in the DAP that follows single spikes in some cells (Fig. 61-12A). A unique “resurgent” current may contribute to resonance (Cummins et al 2005, Enomoto et al 2006, Kovalsky et al 2009).

Oscillations, in cells that have them, are usually subthreshold. However, when peaks are brought to threshold by gradual depolarization or moderate stimuli, they may initiate action potentials (Fig. 61-12D) (Amir et al 1999, Xing et al 2001, Ma and LaMotte 2007). Oscillations do not promote spiking merely because of the extra few millivolts of depolarization that they contribute. After all, depolarization in a slow ramp usually triggers no spikes at all, even if the depolarization is deep. The key property of oscillation sinusoids is their rapid rate of depolarization. This overcomes membrane accommodation. Thus, even though they are small in amplitude, oscillations riding on a slow depolarization create pacemaker capability (Kovalsky et al 2008).

The “Therapeutic Window”: Threshold for Impulse Initiation versus Threshold for Repetitive Firing

The emergence of pacemaker capability at ectopic locations is arguably the most important process underlying neuropathic hyperexcitability and pain. Figure 61-11B shows the typical relationship between firing frequency and (tonic) stimulus intensity (f-I curve, or “encoding function”) in a neuron that is able to fire repetitively. Below a certain stimulus intensity the neuron is silent, but once threshold is reached (8 μA in Fig. 61-11B), sustained firing begins with a sudden jump to a relatively high “minimum rhythmic firing frequency” (mRFF). If the pacemaker site rests well below its repetitive firing threshold, a small depolarizing stimulus will not evoke any spikes. If it is well above threshold (and already firing repetitively), the small stimulus will accelerate the firing a bit (see Fig. 61-11B, zone B). However, if the pacemaker at rest hovers just below threshold (Fig. 61-11B, zone A), the same small stimulus will cause a shift from silence to just above the mRFF, a major amplification (Matzner and Devor 1992, Kovalsky et al 2009). Correspondingly, if the pacemaker at rest is just above the rhythmic firing threshold (and firing spontaneously), slight suppression will cause silence and thereby “stabilize the membrane.” Straddling the threshold yields burst firing (see Fig. 61-3).

The threshold for initiating a single spike and the threshold for repetitive firing are very different. Both depend on the density of Na⁺ channels, but the rhythmic firing threshold is far more sensitive than the single-spike threshold to changes in channel density. The relationship between the two is plotted in Figure 61-13. When Na⁺ channels are abundant, the two thresholds are similar (see Fig. 61-13A right). As Na⁺ channel density is reduced, the repetitive firing threshold rises sharply until firing can no longer be evoked at all, no matter how strong the stimulus (Fig. 61-13; ≈80 mS/cm²). Single spikes, however, are still evoked normally. Complete conduction block occurs only when very few functioning Na⁺ channels remain and action potentials can no longer be sustained (Fig. 61-13A left). The gap between pacemaker and single-spike firing constitutes a “therapeutic window.” Within this window sustained ectopia (and neuropathic pain) can be suppressed without blocking normal sensory signaling. A variety of “membrane-stabilizing” drugs exploit this therapeutic window to selectively suppress ectopia. This is illustrated for lidocaine in Figure 61-14.

Subthreshold Oscillations and Ectopic Firing after Nerve Injury

Detecting stimuli is the specialized task of the sensory ending. At all other locations nerve fibers are essentially incapable of responding to natural stimuli (with the exception of direct heating and some chemical stimuli; Hoffmann et al 2009). Try applying pressure over the median nerve in your forearm. This does not evoke impulses in the midaxon portion of median nerve axons where the pressure is applied. If it did, you would feel tingling in the medial part of your hand. However, injury and disease change this. If there had been a neuroma or a patch of demyelination where you applied the pressure, it might well have produced a sensation referred to the hand, perhaps a painful sensation (e.g., Tinel’s sign at the carpal tunnel).

Research on the cellular mechanisms underlying this change is based mostly on recordings from DRG somata. Recording transmembrane potentials from axons remains technically challenging. However, the similarity of discharge patterns and responses to stimuli and drugs in DRGs and neuromas and the small number of quasi-intracellular recordings that have been made from axons (Kapoor et al 1997) suggest that the same basic electrogenic processes apply. Nearly all intact DRG neurons respond to prolonged step depolarization with a single spike or a brief burst lasting up to a few hundred milliseconds. They then fall silent even though the depolarization is still present (see Fig. 61-12A). Slow ramp depolarization, which better mimics natural generator potentials, rarely generates even a single spike. A few neurons, however, show sustained firing even with slow ramp depolarization. In these exceptional neurons subthreshold oscillations are present (Amir et al 1999, Wu et al 2001b).

In DRG neurons with subthreshold oscillations, each time an individual oscillation sinusoid crosses threshold, a single spike is generated. Such crossings occur at random intervals, which yields a slow, irregular ectopic firing pattern (Fig. 61-3A [3]). This changes if the oscillating neuron also generates DAPs. In this case the first evoked spike triggers a continuous DAP-sustained impulse train, or “tonic autorhythmicity” (see Fig. 61-3A [1]; Amir et al 2002b). If in addition to oscillations and DAPs factors are present that transiently stop the firing (e.g., an activity-induced hyperpolarizing K⁺ current), the impulse train will eventually stop, only to resume the next time that an oscillation sinusoid happens to reach threshold (see Figs. 61-3A [2] and B and 61-12B and D). The result is firing in bursts (“interrupted autorhythmicity”). Slow irregular firing is characteristic of ectopia in C fibers, whereas all three patterns, slow irregular firing, tonic firing, and burst firing at irregular interburst intervals, occur in A fibers (see Fig. 61-3A).

Axotomy and other forms of injury that induce neuropathic pain trigger a dramatic increase in the proportion of DRG neurons with subthreshold oscillations and hence an increased proportion with sustained discharge at rest and in the presence of tonic or slow ramp depolarization (Amir et al 1999, Xing et al 2001, Liu et al 2002, Ma and LaMotte 2007). Neuropathy also increases the likelihood of DAPs and hence the prevalence of high-frequency tonic and burst firing. Both changes are apparently due to remodeling of the intrinsic electrical properties of the neuronal membrane. The remodeling process appears to be a universal hallmark of ectopia in sensory neurons and is hence a key target for control of neuropathic pain.

Electrogenesis: Excitation, Excitability, and Hyperexcitability

The terms “excitation” and “excitability” sound alike, but they are subtly different. Excitation refers primarily to stimulus transduction, the ability of a stimulus to depolarize the sensory neuron and create a generator potential. It depends on the transduction molecules present in the membrane of the sensory ending and both the properties of the stimulus and changes that occur in the process of its physical conveyance from the point of application to the transduction molecules (e.g., thermal conductivity and viscoelastic properties of the skin). Excitability, the capability of being excited, refers mainly to the process of encoding of the generator potential into a propagated impulse train. Encoding depends primarily on Na⁺ and K⁺ channels and the DAPs, subthreshold oscillations, and spikes that they generate. Spike electrogenesis is an outcome of both processes. Analgesic drugs can target either. However, stimuli that excite are numerous, and while eliminating any one of them might have a significant effect, all the others would remain in play. In contrast, if excitability is suppressed, the neuron loses its ability to signal pain in response to any stimulus. Excitation can be thought of as the wide mouth of a funnel whereas excitability is the narrow outlet. The latter is a uniquely powerful locus for controlling neuropathic pain signals. How is excitability modeled in healthy afferents and how do axotomy and demyelination lead to remodeling?

Retrograde Signaling and Control of Excitability in Normal Afferents

The membrane proteins responsible for transduction and encoding are synthesized on ribosomes in the DRG cell soma. They are then loaded on intracytoplasmic transport vesicles and vectorially transported down the axon. When the proteins arrive at the sensory ending, they are inserted into the membrane by vesicle exocytosis (Fig. 61-15). At first they are free to drift in the membrane plane, but after a short time, many become trapped and anchored to submembrane cytoskeleton-associated proteins such as ankyrin B and G and spectrin. The overall process, membrane “trafficking,” is highly complex and closely regulated in healthy neurons so that the correct molecules arrive at their correct addresses in appropriate numbers (Pfeffer 2007, Bennett and Healy 2009). In the event of nerve injury, regulation can fail, resulting in hyperexcitability and pain. This mechanism is discussed later.

Trafficking is dynamic. Vesicles carrying freshly synthesized transducer and ion channel molecules are delivered continuously to the sensory ending, and this delivery is balanced by endocytotic reuptake and degradation of molecules that were inserted previously (Leterrier et al 2010). The turnover half-life of Na⁺ channels, for example, is about 1–3 days. How trafficking is regulated and kept in balance is understood only sketchily. Control of receptor and channel biosynthesis and delivery is thought to involve neurotrophic molecules supplied, at least in part, by the innervated target tissues. Specific neurotrophins (e.g., NGF, GDNF) are taken up by axon endings that express the corresponding neurotrophin receptors and transported retrogradely to the cell soma. There they regulate gene expression and anterograde trafficking of the gene products to the axon’s sensory ending. This determines afferent sensitivity.

In embryonic development, neurotrophic signaling is critical for the very survival of sensory neurons. Interrupting retrograde flow (e.g., by axotomy) leads to rapid cell death. In adults, DRG neurons are less dependent on neurotrophins for survival, but they continue to play an important role in membrane modeling and neuronal phenotype. This role is illustrated in experiments in which adult DRG neurons were axotomized to block the retrograde flow of endogenous neurotrophins. Exogenous neurotrophins were then provided as a replacement. When the right neurotrophins were chosen, axotomy-induced hyperexcitability was prevented and even reversed (Boucher and McMahon 2001). A different, newly discovered mode of retrograde signaling may also play a role, at least in the event of axonopathy. Specifically, at the site where the axon is breached, importin proteins are locally translated and then transported back to the cell soma where they affect gene expression and neuronal phenotype (Hanz and Fainzilber 2006).

Regulation of Excitability

An important gap in our knowledge concerns how intact neurons of a given type maintain a stable level of excitability over long periods (a lifetime). How do nociceptors remain nociceptors and touch-sensitive fibers remain sensitive to touch? If there were an upward drift in the delivery of Na⁺ channels to nociceptor endings, for example, they might begin responding to weak stimuli and cause pain at every turn. A downward drift would make then unresponsive, thereby eliminating the protective role of pain. It is almost inconceivable that the control is “open loop”—that is, the cell soma delivers an unwavering quantity of channels and receptors to the end of the axon. There must be some sort of homeostatic feedback control.

One possible device is the dependence of gene regulation and trafficking on neuronal spike activity. For example, if afferent activity were too high over a particular integration window (hours, days?), excitability would be reduced; if too low, it would be increased (Devor 1999). In this way the sensitivity of the afferent ending would be stabilized around a functional set-point related to the message that it is intended to transmit. Evidence for homeostatic control of this sort exists for other systems (Dargent and Couraud 1990, Dargent et al 1994, Desai et al 1999, Fields et al 2001, Klein et al 2003, Schmidt et al 2009, Krol et al 2010, Kuba et al 2010, Savtchouk and Liu 2011, Turrigiano 2012). For example, in cardiac myocytes, excess spike activity has been shown to increase intracellular Ca²⁺ levels, which reduces Na⁺ conductance and hence excitability (Brodie et al 1989, Offord and Catterall 1989). Similarly, sound deprivation causes brain stem auditory neurons to increase their Na⁺ conductance and hence response sensitivity (Kuba et al 2010). In DRG neurons, cytoplasmic neurotrophin levels and release are both regulated by spike activity, as is TRPA1 trafficking from the cytoplasm to the surface of the membrane (Balkowiec and Katz 2000, Schmidt et al 2009).

Activity–excitability coupling, however, may not work for nociceptors because they rarely fire action potentials and therefore cannot provide activity-related feedback. Consider, for example, nociceptors in cardiac muscle. Even though such fibers may not fire at all for years, they are nonetheless reliably available to signal cardiac insufficiency by causing chest pain. Perhaps the activity-related feedback signal comes from firing in low-threshold afferents that share the same DRG as the nociceptors. The signal could be delivered by crossed-depolarization. That is, nociceptor phenotype could be regulated by the subthreshold depolarization driven by spike activity in the neighboring low-threshold afferents (Rusnak et al 2007, Sohya et al 2007). Such a feedback control system, if it exists, provides a possible rationale for why DRG somata are electrically excitable in the first place and why subthreshold cross-depolarization exists. There are indicators that control of excitability might also involve non-neuronal cells such as glia and Merkel cells, and extracellular matrix components (Kazarinova-Noyes et al 2001).

Membrane Remodeling after Nerve Injury

Regardless of whether excitability in healthy afferent neurons is regulated over the long run by activity-related signals, axonopathy and demyelination clearly disrupt the regulation. The resulting remodeling involves at least three interrelated cellular processes: membrane trafficking, up- and down-regulation of gene expression, and altered behavior of receptors and ion channels. Although the focus of research has been on receptors and channels, allied molecules such as lipids, anchoring proteins, and structural proteins must also be regulated.

Ion Channel Trafficking and Accumulation

Axotomy or functional block of axoplasmic transport (e.g., by cold or antimitotic drugs) leads to the local accumulation of channel-loaded transport vesicles. Some of these vesicles insert into the membrane and thereby increase local channel density and enhance electrogenicity (Fig. 61-16; also see Fig. 61-15). A variety of voltage-sensitive Na⁺ channel types have been shown to accumulate at neuroma endings and patches of demyelination in both animal neuropathy models and humans with neuropathic pain (Novakovic et al 1998, Devor et al 1993, Kretschmer et al 2002, Rasband et al 2003, Black et al 2008, Coggan et al 2010). This remodeling appears to be a causative factor in the creation of ectopic pacemaker sites following nerve injury.

Functional remodeling by channel accumulation does not require a de novo process, only a quantitative shift in the normal equilibrium between channel insertion and reuptake. Remodeling is a result of both permissive and promotional factors. In myelinated axons the axonal membrane under the myelin normally contains a very low density of Na⁺ channels, perhaps because myelin suppresses their insertion. Demyelination, sprouting, and end-bulb formation remove this suppression and permit excess channel insertion. Axotomy also promotes membrane remodeling. Normally, the large amount of Na⁺ channel protein transported down the axon supplies the rapid turnover of channels at downstream nodes of Ranvier, the internodal membrane, and sensory endings (Leterrier et al 2010, Persson 2010a). Following axotomy these targets no longer exist. As a result, in-transit channels are redistributed into whatever competent membrane sites remain, notably neuroma end-bulbs, sprouts, and plaques of demyelination (see Figs. 61-15 and 61-16). Sites far upstream of the injury might also receive excess channel protein (Gilly and Brismar 1989), a potential mechanism for hyperexcitability in axotomized DRG somata. The same principles hold for other membrane proteins, including K⁺ and Ca²⁺ channels and α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptors (England et al 1998, Garry and Fleetwood-Walker 2004, Nissenbaum et al 2008, Mathie et al 2010, Nuwer et al 2010, Roza et al 2011). In principle, drugs that alter axoplasmic transport could be exploited for pain therapy (Devor and Govrin-Lippmann 1983).

A final mode of altered channel trafficking is in the sensory ending itself. Neurons contain stores of latent cytoplasmic ion channels that do not contribute to the transmembrane ionic current. Impulse traffic and diffusible mediators within innervated tissue can promote Ca²⁺-dependent vesicle exocytosis and local membrane insertion of these latent receptors and channels. The expected result is enhanced ionic currents and electrical hyperexcitability (Huang and Neher 1996, Bao et al 2003).

Regulation of Gene Expression

Axotomy also triggers changes in the expression of various transducer, receptor, and ion channel molecules in DRG neurons. Since more than 99% of the cytoplasmic mass of sensory neurons serving the extremities is in the axon, mid-nerve axotomy represents a major metabolic challenge for the neuron. Interestingly, neuronal gene expression may also change after demyelination (Craner et al 2003, Wallace et al 2003). As noted above, thousands of genes are up- or down-regulated in neuropathic pain models (Persson et al 2009a, Hammer et al 2010, Michaelevski et al 2010), and a much larger number of protein products are altered. For some genes the degree of regulation varies across mouse strains in a manner that correlates with strain-specific pain susceptibility (Persson et al 2009a, 2009b). Such genes are good candidates for being related to pain mechanisms. Likewise, as noted, regulation may differ among afferent types within a single strain (Persson et al 2010b).

It needs to be appreciated, however, that the levels of mRNA expressed and the size of cytoplasmic protein stores do not necessarily translate simply into the density of proteins in the membrane. For example, most Na⁺ channel types accumulate in neuroma endings despite the fact that their synthesis in the soma is down-regulated (Kretschmer et al 2002, Waxman et al 2002, Black et al 2008). Likewise, neurotransmitters may be depleted in the DRG without an obvious loss at the central terminals (Bailey and Ribeiro-da-Silva 2006). The phenotype of a sensory neuron reflects the entire ensemble of molecules present, including modifications as a result of the local environmental influences of nearby non-neural cells and extracellular matrix components. Given that the levels of thousands of interacting molecules may change following nerve injury and disease, it is difficult to intuit or to simulate how all aspects of a neuron’s behavior might be changed by such injury.

Understanding remodeling can be made more tractable by focusing on the facets of the neuronal phenotype that are most closely related to neuropathic pain. I have made the case above that membrane hyperexcitability and ectopic electrogenesis transcend issues of sensory transduction and abnormal synaptic processing in the CNS. If so, a focus on these channels is justified.

Excitability depends mostly on a relatively restricted set of ion channels. Mammalian DRG neurons express, at least in some measure, eight types of voltage-gated Na⁺ channel pore (α) subunits in combination with three modulatory (β) subunits. A similar variety of voltage-sensitive and voltage-insensitive (K2p leak) K⁺ channels are expressed. Some members of each type are selective for afferents and even for nociceptors. To these we need to add channels that are gated by ions and ligands rather than voltage. For example, activity-related entry of Ca²⁺ ions can activate Ca²⁺-gated K⁺ channels and cause hyperpolarization. Changes in the level and disposition of any of these molecules can alter excitability and pain.

Current-Carrying Capacity of Ion Channels Already in Place

Ion channels determine cell excitability by controlling the transmembrane flow of ionic currents. An increase in net flow through a fixed number of channels has much the same effect as an increase in the total number of channels present. Increased net flow can result from a higher probability of channels opening, extended channel open time, incomplete inactivation, or increased unitary channel conductance. Such changes in the intrinsic kinetics of ion channel behavior represent a third process that can contribute to neuronal hyperexcitability.

Evidence is accumulating that modulation of the current-carrying capacity of voltage-gated channels might contribute to neuropathic pain. For example, cyclic adenosine monophosphate–dependent phosphorylation of Na⁺ channels reduces Na⁺ current, wherease dephosphorylation returns it to normal (Li et al 1992, Gold et al 1998, Vijayaragavan et al 2004, Chahine et al 2005, Hudmon et al 2008). Certain hormones, trophic factors, neuromodulatory peptides (e.g., CGRP), and inflammatory mediators (notably prostaglandins) activate dephosphorylating enzymes, including protein kinase A and C. They are therefore positioned to affect afferent excitability. Some of these mediators are known to sensitize afferents by increasing Na⁺ current density, reducing K⁺ current density, and augmenting oscillations and neuronal firing even when the membrane potential remains unchanged (Nicol et al 1997, Gold et al 1998, Hu et al 2001, d’Alcantara et al 2002, Natura et al 2005, Binshtok et al 2008). They alter excitability without necessarily changing excitation. Transduction channels, of course, can also be sensitized.