Computational functions of neurons and circuits signaling injury: Relationship to pain behavior

Abstract

The basic circuitry of the “pain pathway” mediating transmission of information from the periphery to the brain is well known, consisting of specialized sensory fibers known as nociceptors projecting to specific spinal cord neurons, which in turn project on to the thalamus and cerebral cortex. Here we survey some of the unique properties of these circuits, such as peripheral and central sensitization, and the segmental and descending modulatory control of synaptic transmission. We also review evidence indicating dissociation between nociceptor activity and behavioral indications of pain. Together, these considerations point to the need for a more quantitative approach to the nociceptive system, specifically the interactions at peripheral, spinal, and supraspinal levels as well as between them, to more fully understand how the activity in nociceptive neurons individually and collectively is related to the pain response.

This brief review surveys our current understanding of how activity in pathways activated by nociceptors contributes to the experience of pain. In attempting this task we must recognize certain difficulties at the outset, particularly the fact that the same stimulus might or might not be considered painful, depending on factors such as sex (1) and the genetic profile of the individual (2). More significantly, the pain produced by a nociceptive stimulus in the same individual is influenced by situational variables, for example the level of stress, as part of the adaptive response to such challenges (3). To help explain this property of nociception, we will focus on the modifiability of these neurons and circuits because this is an important determinant of the failure of activated nociceptive afferents to elicit a stereotyped pain response.

Periphery

The concept of a neural mechanism for pain was advanced explicitly by the 17th century philosopher René Descartes, who posited a neural “channel” connecting the site of peripheral damage to the brain. The concept of the nociceptor was introduced more than 100 y ago by Sherrington (4), who defined it as a sensory receptor responsive to stimuli that are potentially damaging to the organism; more precisely that in the skin there had evolved “a special sense of its own injuries.” His focus on reflex action led him further to postulate that the action of nociceptors would lead to withdrawal of the affected body part from the source of damage (5). Electrical stimulation of peripheral nerves revealed a characteristically high threshold for the withdrawal reflex compared with proprioceptive reflexes; this was interpreted as indicating that the peripheral axons responsible were relatively inexcitable (i.e., were small myelinated or unmyelinated rather than large myelinated) (6). Early attempts to record the adequate stimulus for such fibers were only partially successful, in large part because of the difficulty in recording from individual small fibers. This left open the possibility that pain results from patterned activity in sensory fibers also sensitive to nonnoxious stimuli rather than activity in a special population of nociceptors. This uncertainty was resolved in the late 1960s and early 1970s when Perl and collaborators (7, 8) succeeded in recording from a population of small-diameter axons with properties of nociceptors. It is now well established that nociceptors are a distinct population of sensory fibers with unique physiological, anatomical, and chemical properties. However, as we shall see below, activation of nociceptors can be prevented from causing pain, and conversely, pain is reported in situations in which nociceptors are not activated.

The defining characteristic of nociceptors is their high threshold to natural stimulation of their receptive field. This has been investigated mostly in skin, but nociceptors exist also in muscle and viscera. The initial work from Perl's laboratory in the cat demonstrated that nociceptors could be subdivided into two general groups: high-threshold mechanoreceptors whose axons conduct in the small myelinated Aδ range, and polymodal nociceptors sensitive to a range of modalities, including mechanical, thermal, and/or chemical, with unmyelinated C-fiber axons (6). Studies using these defining characteristics as starting points have revealed many additional unique characteristics of nociceptors, including their Na channel composition (9–11), their action potential configuration (12, 13), their receptors for numerous inflammatory molecules (14), their transmitters (15, 16), and their projections into the spinal cord (17, 18).

Relationship of Peripheral Nociceptors to Pain

The recent advances in genetic methods have offered the opportunity to test the effects of removing individual molecular receptor classes or specific classes of nociceptors on the ability to respond to nociceptive stimuli. Because of evidence that the TRPV1 receptor responds to noxious heat with a threshold (43 °C) that is very close to the threshold for heat pain (19), it was expected that knockout mice lacking the TRPV1 receptor would be unresponsive to noxious thermal stimulation. Dissociated dorsal root ganglion cells from these preparations exhibited the expected loss of sensitivity to noxious heat, but the mice themselves displayed largely normal responses to noxious heat (20, 21). However, they exhibited a deficit in sensitization to noxious agents such as carrageenan. The full explanation for these findings is not yet available. One possibility is that receptor(s) other than TRPV1 respond to thermal stimulation but are not subject to sensitization. DRG cells expressing the unknown receptor might be either very fragile or rare and thus unlikely to be observed in dissociated cell culture.

A more recent approach has involved eliminating entire populations of nociceptors and evaluating the resulting behavioral sensory loss. Cavanaugh et al. (22) have used genetic and pharmacological approaches to selectively eliminate entire classes of nociceptors. In mice peptidergic nociceptors cells terminating in lamina I were selectively killed by treatment with capsaicin. In companion experiments using transgenic mice, a different population of polymodal nociceptors terminating in outer lamina II (lamina II_o) and expressing a unique G protein-coupled receptor (MGPCR) could be selectively killed after treatment with diphtheria toxin. Because both of these populations are polymodal nociceptors, it was expected that elimination of either class would result in a similar behavioral deficit involving noxious heat and high-threshold mechanical stimulation. An unexpected result was obtained, namely that eliminating virtually all peptidergic nociceptors terminating in lamina I reduced sensitivity to thermal nociception but not to mechanical nociception. The opposite result was obtained when the MGPCR-expressing polymodal nociceptors projecting to lamina II_o were eliminated (i.e., mechanical nociceptive threshold was elevated but thermal nociception threshold was unaffected).

It is difficult at present to give a definitive interpretation of these unexpected results. They suggest that the information coded in the discharge of the nociceptors does not affect behavior directly; only certain components seem to affect behavior: noxious heat in the case of the TRPV1-expressing nociceptors projecting to lamina I and mechanical nociception in the case of the MGPCR-expressing nociceptors projecting to lamina II_o. One possible interpretation is that subtle variations in the discharges associated with the different modalities are filtered differently by synapses in the pathways projecting to lamina I and lamina II_o owing to differences in discharge properties, synaptic properties, or both. Synaptic filtering is known to vary at synapses, for example those made between group Ia stretch-sensitive spindle afferents and different target motoneurons (23, 24). An additional factor that might contribute to filtering the transmitted nociceptive discharge is the contribution of NMDA receptors in the postsynaptic cells (25, 26), which cause substantial nonlinearity in the transmission process due to the voltage dependence of the Mg²⁺ block associated with these receptors (27). This might provide an additional basis for variable filtering based on the spike intervals associated with the responses of the different afferent fiber classes to specific stimulus modalities.

These data suggest that certain pathways projecting from cells in different laminae of the dorsal horn are specialized to abstract certain aspects of the information coded in the discharge of polymodal nociceptors. This may be important in helping the organism distinguish between the different modalities to which the polymodal nociceptors can respond. The details of this computational achievement remain to be fully understood.

Sensitization

One of the hallmarks of nociceptive pathways is sensitization in response to a damaging stimulus whereby the behavioral response to subsequent stimuli is enhanced. In some cases normally nonnociceptive stimuli elicit pain because of a decrease in threshold (allodynia); in others, a normally painful stimulus becomes even more painful (hyperalgesia). Two general mechanisms have been identified (Fig. 1): one acting at the periphery (peripheral sensitization) and the other acting centrally (central sensitization). Peripheral sensitization results from injury-induced release of a number of sensitizing agents from damaged cells and also from immune-competent cells recruited into the vicinity of the nociceptive nerve terminals (28). These substances act to either reduce the threshold or enhance the magnitude of the response to a given stimulus. These sensitizing agents have been referred to collectively as the “inflammatory soup.” Central sensitization of nociceptive pathways has been studied in most detail in the spinal cord (29), although there is evidence that similar mechanisms occur at synapses in other regions of the nociceptive system (30). As with peripheral sensitization, a number of molecules and signaling pathways are involved in eliciting central sensitization (29).

Fig. 1.

Schematic diagram of major mechanisms involved in peripheral and central sensitization. Injury results in release of inflammatory mediators (inflammatory soup) that enhance the nociceptor discharge produced directly by the injury. The increased activity in sensory neurons results within hours in increased levels of peptide (calcitonin gene-related peptide, substance P, BDNF) in the dorsal root ganglion and subsequent release of these into the spinal cord and into the injured tissue. Two frequency-dependent effects occur in the spinal cord: short-term temporal summation, known as windup (supported by peptide release), and LTP. Both of these require NMDA receptor activity in superficial dorsal horn neurons. Many other receptors in dorsal horn neurons contribute to the enhanced activity (e.g., trkB and NK1). Activated microglia also release BDNF, which can decrease E_Cl⁻ via effects on the KCC2 transporter, thereby converting inhibitory actions into excitatory ones, thus resulting in increased discharge in nociceptive pathways. Influx of Ca²⁺ participates in second messenger activation of gene transcription mechanisms, which prolongs the effect of nociceptor activity.

A fundamental question for both peripheral and central sensitization is how the different molecular mechanisms interact. In the case of peripheral sensitization a number of molecules, including bradykinin, prostaglandin, histamine, and NGF, are released, generally in response to an inflammatory injury. They are released from immune-competent cells, such as mast cells, macrophages, and neutrophils, recruited into the injured region. Each agent acts on the nociceptive nerve ending via its receptor: BK-receptors for bradykinin, EP-receptors for prostaglandin, H-receptors for histamine, and trkA receptors for NGF, etc. It is not clear whether these molecules play a similar role and are redundant to ensure pain and subsequent guarding of inflamed tissues, or whether they have independent roles to play in eliciting inflammatory pain. The potential for interaction among these agents exists at the level of intracellular signaling in the nociceptive nerve terminal (e.g., PKA and/or PKC), but it is also possible that interaction takes place at the level of the inflammatory cells whose output can influence the release from other inflammatory cells (e.g., release from mast cells is influenced by NGF). There is some evidence in behavioral experiments for interaction between the effects of these different agents (31), but few studies of this sort are available. Another way of framing this problem is to ask whether these sensitizing mechanisms act in series or in parallel. It seems likely that both types of interaction operate to determine the level of peripheral sensitization. Because many pharmacological interventions in pain involve interference with single mechanisms (e.g., aspirin inhibiting prostaglandin synthesis), it is important to resolve the mechanisms linking injury and discharge properties of nociceptors at a computational level.

As pointed out above, nociceptors are unique in releasing peptidergic transmitters such as substance P and calcitonin gene-related peptide into the spinal cord in addition to the common excitatory transmitter glutamate. This prolongs the depolarization of postsynaptic cells, resulting in temporal summation in response to successive stimuli, originally described as windup (32). The result is a discharge of increasing length in response to successive C-fiber volleys, provided they are elicited no more than a few seconds apart. Responses to A-fiber stimulation do not exhibit this temporal summation. In accordance with this pattern, it has been found that “first pain” elicited by activity in Aδ-fibers does not sensitize with successive stimulation, whereas “second pain” elicited by activity in C-fibers does sensitize, with parameters consistent with a role for windup (33).

The cumulative depolarization of second-order cells has consequences beyond the increased impulse activity because the postsynaptic cells in the superficial dorsal horn activated by nociceptive inputs have NMDA receptors in addition to AMPA receptors. The NMDA receptors become permeable to Ca²⁺ when subjected to depolarization, and this acts as a second messenger initiating a series of events that can lead to long-lasting transcriptional changes in gene activity (29). NMDA receptors also contribute to another use-dependent long-term potentiation (LTP) of activity in postsynaptic cells in the superficial dorsal horn. Unlike the windup, which does not require high-frequency activity, LTP lasts for hours. Application of this stimulus paradigm to the human forearm also results in long-lasting behavioral sensitization (34).

As with peripheral sensitization, there are numerous mechanisms contributing to central sensitization. The neurotrophin BDNF plays an important role. BDNF is up-regulated in the soma of nociceptors several hours after the onset of peripheral sensitization (35, 36), produced for example by NGF. This BDNF is released into the superficial dorsal horn (37), and it has been shown to sensitize the synaptic response of AMPA receptors in cells of lamina II via a PKC-dependent mechanism (38). BDNF plays another role in central sensitization: it is released from microglia in the superficial dorsal horn activated by ATP release from spinal terminals of nociceptive afferents and inhibits the KCC2 chloride transporter in relay cells, thereby converting GABA-driven hyperpolarization to depolarization (39). The interaction of these effects of BDNF in the spinal cord remains to be evaluated. Although trkB agonists have been implicated mainly in central sensitization, BDNF and NT-4/5 also sensitize the discharge of nociceptors (40–42).

Nociceptive Circuits in the Spinal Cord and Brain

Cells in the dorsal horn responding to nociceptive inputs are of two general types: nociceptive specific (NS) and wide dynamic range (WDR). NS neurons are located primarily in lamina I and the outer portion of lamina II (lamina II_o). These cells receive their primary afferent input exclusively from nociceptors, either Aδ-fibers (lamina I) or C-fibers (lamina II_o), and respond exclusively to nociceptive inputs. In deeper laminae of the dorsal horn are cells that receive input from Aβ- and Aδ-peripheral fibers as well as input from nociceptive cells in the superficial laminae. This convergent input makes them responsive to both gentle and noxious stimulation, hence the descriptor WDR.

Both WDR and NS cells project from the spinal cord to the brain. The major ascending pathway is via the crossed spinothalamic tract. Many such cells terminate in the somatosensory nuclei in the lateral thalamus, particularly the ventrobasal nucleus; the midline intralaminar nuclei of the thalamus are another major termination site of spinothalamic axons (43). Cells in the ventrobasal nucleus project largely to the SI somatosensory cortex, whereas the midline intralaminar nuclei project to the anterior cingulate cortex (ACC). Other destinations of thalamic cells are the insular cortex and SII. There is considerable species variation in the pattern of thalamo-cortical projections. We will see below that these different cortical regions are concerned with different components of the nociceptive stimulus and its transformation into the pain response.

Some ascending fibers carrying nociceptive information from the spinal cord terminate caudal to the thalamus in the nucleus of the rostral ventromedial medulla (RVM), and in the periaqueductal gray (PAG) in the midbrain. These projections activate systems that feed back to the spinal cord to modulate the input from nociceptors (see below).

Processing the Nociceptive Message

We have already discussed evidence that processing the nociceptive message may not be straightforward because deleting a population of nociceptors does not necessarily predict the behavioral deficit. A parallel conclusion can be drawn from studies of central pathways responding to activation of the nociceptors. For example, behavioral experiments in the trigeminal system with recordings from NS and WDR neurons have revealed that the magnitude of nociceptive stimuli is coded by WDR cells, not by NS cells (44). The role of NS cells is not well established, although there have been suggestions that lamina I NS neurons project via a “private” pathway to the thalamus and that from there this information is relayed to the anterior cingulate and insular cortex (45). The role of this system according to its proponents is a homeostatic one, similar to temperature processing, whereby the body maintains an equilibrium state in response to the nociceptive stimulus.

Neurons in the somatosensory cortex seem highly specialized for discriminative aspects of pain because they have small receptive fields, unlike cells in the ACC, which have whole-body receptive fields (46). The function of ACC has been best elucidated using PET imaging to measure activity in conscious human experimental subjects. The findings from these experiments suggest that anterior cingulate cortical activity is related to the unpleasantness of a painful stimulus rather than its magnitude (47). ACC projections to the amygdala indicate a close connection with the “emotional brain,” which adds an important affective component to the response to the painful stimulus.

Circuits That Control Nociceptive Input to the Brain

Numerous studies in the past 50 y have revealed that nociceptive pathways are subject to control at many different levels. The gate theory of pain proposed by Melzack and Wall (48) was a landmark in this area because it suggested that activation of large-diameter afferents exerted a mixed excitatory and inhibitory effect on spinal neurons transmitting nociceptive information to the brain, whereas small-diameter afferents elicited a purely excitatory influence on the activity of these cells. They suggested that it was the balance in activity between small- and large-diameter sensory fibers that determined the level of nociceptive activity forwarded to the brain. Although some details of the circuitry proposed in that article (48) were not confirmed, particularly the idea that small-diameter afferents resulted in presynaptic disinhibition of cutaneous inputs to the spinal neurons (49, 50), the ideas expressed by these authors opened a significant new chapter in studies of nociceptive pathways. One important outcome was the suggestion that pain might be modified by selective electrical stimulation of large-diameter fibers in peripheral nerve (transcutaneous electrical nerve stimulation) (51) or via dorsal column stimulation (52). This has proven to be very useful clinically in certain painful conditions (53), although the mechanism is probably more complex than the original proposal based on the gate theory because the pain relief can persist for up to hours after the large fiber stimulation is stopped, possibly because of the involvement of opiate mechanisms (54, 55).

Another system with strong influence on transmission from afferent activity in nociceptors to spinal transmission neurons originates from the RVM (Fig. 2) (3). Two major groups of neurons projecting to the spinal cord have been identified from recordings in vivo. OFF cells are tonically active and have an inhibitory action on nociceptive transmission in the spinal cord. ON cells are normally silent and facilitate nociceptive transmission (56). Ascending activity elicited by a painful stimulus inhibits OFF cells and increases activity in ON cells, which act together to amplify the spinal effects of the nociceptive activity, for example in facilitating the withdrawal response (57). Opiates influence activity in OFF and ON cells oppositely, with OFF cells being facilitated and ON cells being inhibited. This combination acts to reduce the gain of nociceptive pathways at the level of the dorsal horn when opiate receptors are activated. The RVM is influenced by higher centers such as the PAG in the midbrain, which in turn receives input from cortical centers including the ACC and thus plays a role in conditioning nociceptor input as a function of behavioral context. This system has also been proposed to play a role in the analgesia reported in response to placebo administration; the reduced pain is associated with activation of descending pathways emanating from cingulate cortex (58) and is mediated by opiate receptor activity (59).

Fig. 2.

Simplified diagram displaying some of the potential major interactions between different levels of the nociceptive system; the sign of the effect is given only in the dorsal horn of the spinal cord and the RVM. Segmental inputs to the superficial dorsal horn from Aδ- and C-fibers are excitatory (+). Inputs from Aβ-fibers also have an inhibitory component (denoted +/−) as described in the gate theory (see text). The specific pathways mediating the inhibition are not specified in this figure. Descending ON and OFF cells from the RVM can excite and inhibit transmission, respectively, from peripheral nociceptive afferent fibers to cells in the spinal dorsal horn. OFF cells are excited and ON cells are inhibited by projections from PAG. In turn, the PAG is influenced by projections from hypothalamus and amygdala and indirectly through these nuclei from the ACC. RVM, RVM, PAG, and hypothalamus are also influenced by ascending connections from the spinal cord, providing the substrate for multiple feedback loops influencing transmission from primary afferent fibers to spinal neurons.

Administration of peripheral sensitizing agents affects the balance of activity in the ON and OFF cells in the RVM (57). This suggests that this system can set the gain of the spinal nociceptive pathway on a moment-to-moment basis.

A more diffuse inhibitory mechanism has been demonstrated in experimental animals and human subjects. Known as diffuse noxious inhibitory control (DNIC), this mechanism acts over the entire body, such that inputs from one region (e.g., the head) can inhibit inputs from any other part of the body (e.g., the hind limb/leg) (60). This mechanism is independent of the PAG/RVM system described above and has been suggested to suppress weak nociceptive inputs in favor of the strongest ones.

The inhibitory mechanisms outlined above provide a potential explanation for the finding that activation of nociceptors does not necessarily result in the experience of pain. There are also examples of the reverse (i.e., when pain is experienced in response to stimulation that does not include nociceptors). For example, human subjects report pain from placing their hand on a series of bars arranged such that cool and warm bars alternate (#1, #3, etc. cool; #2, #4, etc. warm; referred to as a thermal grill); touching the individual bars of either temperature is not painful. The report of pain in the absence of nociceptors activation is called the thermal grill illusion (61). Imaging the brain of subjects touching the thermal grill reveals paradoxical activation of areas associated with pain (i.e., the insula and ACC), but these areas are not activated by the cool and warm temperatures touched individually (62). The authors have advanced a computational model based on recordings from lamina I cells in the spinal cord; they found that cells responding exclusively to the cool temperature had a significantly decreased response to the grill stimulus (using the same cool temperature), whereas a different population of lamina I cells responsive to noxious heat, pinch, and cool (HPC cells) displayed no change in response when tested with the grill. They postulate that cells responsive to cool stimuli inhibit the responsiveness of thalamic or cortical cells to inputs from HPC cells. Thus, the cool stimuli in the thermal grill cause disinhibition of neurons coding the “painfulness” of the stimulus (HPC cell discharge).

Perspective

It is clear from this brief summary that input from nociceptors can be affected by changes in excitability of the sensory neurons themselves as well as their synapses in the spinal cord. The nociceptor discharge is influenced at the periphery by sensitization mechanisms driven by substances released in the vicinity of the sensory terminals during a diverse group of events associated with inflammation. Beyond that, the transmission of the nociceptive message is subject to modulation by segmental, descending, and ascending inhibitory mechanisms. The latter seem to consist of at least two different types of pathways, some local and another (DNIC) more global.

Not surprisingly, much of the interest in these mechanisms has been driven by attempts to attenuate persistent or chronic pain in disease states. In the case of peripheral sensitization these efforts are driven toward conditions such as arthritis, burns, or headache and involve the use of inhibitors of cyclooxygenase enzymes responsible for prostaglandin synthesis (63). Other agents, such as molecules with anti-NGF activity, are currently being tested for a possible therapeutic role in certain types of pain (64). Segmental control mechanisms have been activated using electrical stimulation of peripheral nerves or their rostral projection in the spinal cord (53). Descending systems have been manipulated in most cases by delivery of opiates because the cingulate cortex→PAG→RVM→spinal cord descending control system expresses opiate receptors at every site, and on balance, activation of this system acts to inhibit nociceptive input to the spinal cord (65). This list is not exhaustive; other pharmacological approaches to pain at the spinal level include the tricyclic antidepressants, NMDA receptor blockade, and Ca channel blockers (66). Although there has been modest success in controlling pain pharmacologically, there remain serious deficits. Significantly, neuropathic pain caused by direct injury to the peripheral nerves or the spinal cord is relatively refractory to control by these methods (67). Furthermore, these drugs all have side effects, some quite serious, such as addiction, gastrointestinal bleeding, etc.

One of the hallmarks of chronic pain is its association with plasticity of the nociceptive system. In the case of inflammatory pain this plasticity is largely functional, involving increases in the number and sensitivity of high-threshold receptors at the periphery, changes in synaptic efficacy in the spinal cord, and alterations in the balance of excitatory and inhibitory inputs. In the case of neuropathic pain caused by damage to peripheral nerves or central pathways, structural plasticity (e.g., collateral sprouting) might be expected to occur because of partial denervation of neurons resulting from the injury.

The multiplicity of changes makes evaluation of any specific strategy very difficult, particularly in view of the potential for interaction among these different mechanisms. For example, there is evidence that descending control mechanisms from the RVM are altered in a time-dependent manner after inflammatory injury (57). Thus, any treatment of the pain at the periphery might have paradoxical effects due to changes in function of the descending control mechanism. These considerations call for a more integrated and quantitative approach to the behavior of the nociceptive pathway. A problem that requires attention in this regard is how any manipulations should be evaluated. The most realistic measure is behavior, because this is the desired endpoint and because it is carried out in the absence of anesthesia, which interferes with evaluation of pain mechanisms. However, this is complicated in nonhuman species such as rodents owing to the inability to obtain direct measures (reports) of pain intensity. Brain activity might be a useful surrogate, but because multiple brain regions acting together are probably responsible for the pain behavior, this would be difficult to interpret unless the combinatorial rules were understood. Thus, despite tremendous progress in establishing neuronal pathways, cellular and physiological mechanisms, and molecular entities involved in conversion of nociceptors activity to pain behavior, it is clear that we require more information on the behavior of this system as a whole as well as the interactions between the different components to enhance our ability to better treat painful conditions.