Spinal Cord Synaptic Plasticity and Chronic Pain

This is the authors' abstract. We don't add key points for this paper.

The word "pain" encompasses several phenomena, which probably have several different mechanisms. Immediately after a noxious stimulus there comes an initial, short-lasting component of pain, which is described well by the gate-control theory. A little later comes a slow-onset, long-lasting component mediated by C-fibers from deep tissue, if that has been injured. This may be combined with sensitization of peripheral receptors and their continued activation by products of peripheral inflammatory processes. Relatively rarely, a pathologic state of chronic pain may develop, which can involve more or less permanent changes in the central nervous system (CNS), persisting long after the original injury has healed [1]. Various forms of synaptic plasticity that may underlie these pathologic CNS changes are the subject of this review. There has been much interest recently in the involvement of spinal cord synaptic plasticity in the hyperalgesia and allodynia of chronic pain syndromes [2,3]. Several forms of activity-dependent plasticity have been described in the spinal cord. Well established cases that have been widely associated with chronic pain are "windup" [4], which lasts seconds to minutes, and increases [5,6] or decreases [7] in the excitability of spinal reflexes lasting 1 to 3 h. More recently, long-term potentiation and depression of synaptic transmission, of the same types that occur in the hippocampus and neocortex (where they have been shown to last for days), have also been described in several regions of the spinal cord [8-10]. Until now, there has been little cross-fertilization between the literature on spinal cord changes concerned with chronic pain and that on long-term potentiation and depression. It is, therefore, useful to summarize very briefly what is known of the mechanisms of these various forms of physiologic plasticity in such a way as to allow comparison between them and to clarify similarities and differences. Windup Windup is a progressive increase in the number of action potentials elicited per stimulus that occurs in dorsal horn neurons [4] and in the large motor neurons of the ventral horn [11] when the stimulus frequency to C-fibers exceeds 0.5 Hz. The mechanism of windup was revealed by Thompson et al. [12]. These workers recorded intracellularly from ventral horn cells and stimulated dorsal roots at increasing frequencies. When the stimulation frequency reached about 0.5 Hz, the postsynaptic depolarizing responses to individual dorsal root stimuli (i.e., the excitatory postsynaptic potentials [EPSPs]) in the ventral horn cells summated to produce a cumulative depolarization, which resulted in a burst of action potential discharge instead of a single action potential in the ventral horn cell in response to each dorsal root stimulus. After the high-frequency stimulation was stopped, action potentials continued to fire for as long as the depolarization of the ventral horn cell lasted (approximately 60 s) and then ceased. The duration of individual EPSPs produced by low-frequency stimulation was shortened, and the windup effect was completely blocked by antagonists of the N-methyl-D-aspartic acid (NMDA) type of glutamate receptor [12]. These results suggest that windup can be regarded as temporal summation of NMDA receptor-mediated EPSPs. Repeated episodes of windup are quite likely to trigger synaptic long-term potentiation (LTP) and/or long-term depression (LTD), as defined and described in the next sections, but this has not so far been studied. Long-Term Potentiation LTP was first described in the hippocampus as a long-lasting increase in the efficacy of synaptic transmission (i.e., an increase in the amplitude of the EPSP recorded in response to a test presynaptic stimulus and also in the excitability of the postsynaptic cell) caused by brief high-frequency (100 Hz) stimulation of synaptic pathways [13]. The hippocampus is involved in declarative memory and LTP fulfilled an important criterion laid down in Hebb's [14] model of how memory might work, so the process of LTP seized the imagination of those concerned with the mechanism of the laying down of memory traces. There has been an exponentially increasing amount of work on the mechanism of LTP over the last 20 yr and it has turned out to be a many-staged, highly complicated process (for reviews see [15-19]). A very brief and simplified summary of the mechanism of LTP follows, with only those findings which have appeared since the most recent major review documented: for documentation of most of the following, the reader is referred to Bliss and Collingridge's extensive exposition [19]. The summary is itself summarized in Figure 1.Figure 1: Mechanism of long-term potentiation. A, Normal synaptic transmission. Invasion of the presynaptic nerve terminal by an action potential causes exocytosis of vesicles of neurotrasmitter (glutamate). Glutamate binds to three kinds of postsynaptic receptor: alpha-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid (AMPA) receptors, metabotropic receptors, and N-methyl-D-aspartic acid (NMDA) receptors. Glutamate binding to AMPA receptors causes opening of channels permeable to Na+ and K (+) ions. Na+ flows into and K+ out of the cell in such a way that the net effect is an influx of positive charge, resulting in a transient depolarization of the postsynaptic cell (excitatory postsynaptic potential [EPSP]). Glutamate binding to metabotropic receptors causes binding of guanosine triphosphate (GTP) to G-proteins in the cell membrane and the activation of various so-called second messenger enzymes such as protein kinase C and adenyl cyclase. Binding of glutamate to NMDA receptors potentially opens channels permable to Na+, K+, and Ca++, but no ions flow because the channels are blocked by Mg++ ions in the extracellular fluid. B, Induction of LTP. A burst of action potentials in the presynaptic nerve terminal causes release of more glutamate than in Part A. More AMPA receptors are activated and the larger resultant postsynaptic depolarization ejects the Mg++ from NMDA receptor-associated channels. Ca++ flows in through these channels. This Ca (++) interacts with the second messenger enzymes and causes more AMPA receptors to become available (so that a given amount of glutamate will cause a larger EPSP in future), various ion channels in the membrane to be phosphorylated (making the cell more excitable, i.e., more likely to fire in response to a given depolarization) and a retrograde messenger to be released. The retrograde messenger acts by an unknown mechanism on presynaptic nerve terminals which have been active recently to cause release of more glutamate per action potential. The postsynaptic calcium surge can also activate genes in the nucleus of the postsynaptic cell and cause growth of dendritic spines.The most common form of LTP is initiated when bursts of high-frequency presynaptic activity release bursts of the neurotransmitter glutamate from presynaptic terminals. This glutamate activates the alpha-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid (AMPA) type of glutamate receptor on the postsynaptic cell. AMPA receptors open their associated cation channels, allowing sodium and potassium ions to flow into and out of the cell in such a way that the postsynaptic cell is depolarized in a fast EPSP. The depolarization, if large enough, relieves a magnesium-dependent block which, at normal resting membrane potentials, inactivates the NMDA type of glutamate receptor. NMDA receptors to which glutamate has bound open their associated ion channels, which are permeable to sodium, potassium, and calcium ions. Calcium flows into the postsynaptic cell. Influx of calcium triggers a further release of calcium from intracellular stores. Usually, the critical factor which determines whether calcium enters the postsynaptic cell to start the whole process leading to LTP is the release of enough glutamate by the initial burst of presynaptic activity to depolarize the postsynaptic cell enough for NMDA receptors to be activated. In some cases, a form of LTP called NMDA-independent LTP, which is not prevented by NMDA antagonists and occurs in the apparent absence of NMDA receptors, can be induced. The mechanism of this form of LTP is not clear, but it could be due to the activation of voltage-sensitive calcium channels by a large postsynaptic depolarization [20]. The forms of LTP so far found in the spinal cord are NMDA-dependent. Both forms of LTP (in the hippocampus at least) require activation of metabotropic glutamate receptors in addition to NMDA receptors [21]. Metabotropic receptors (see Figure 1) are not coupled to ion channels but instead activate G proteins in the cell membrane and in conjunction with the initial calcium transient activate various second messengers further down the pathway below. The transient increase in calcium concentration in the postsynaptic cell initiates a whole chain of events. Not necessarily in this order: (a) The postsynaptic cell releases a retrograde factor or factors, the identity of which is currently a hot topic, with arachidonic acid, nitric oxide [19], and most recently, carbon monoxide [22,23] being candidates. The retrograde factor(s) diffuse(s) back to the presynaptic cell and causes (b) more transmitter release in response to each presynaptic action potential: this only occurs in terminals which have been primed by recent activity. (c) Calcium-dependent enzymes are activated, such as protein kinase C, calcium/calmodulin kinase, and the cyclic AMP-dependent protein kinase A [19,24], resulting in (d) membrane proteins including receptors and ion channels being phosphorylated, so that, (e) the postsynaptic cell becomes more excitable, in the sense that a given size of EPSP is more likely to initiate an action potential: this is called E-S potentiation (EPSP-Spike) and (f) more AMPA receptors become available on the postsynaptic cell. (g) Gene transcription is affected, probably [but see Madison et al. [18]] to make these and/or other changes very long-lasting (weeks to months). (h) Growth of spines on the postsynaptic dendrites and/or presynaptic terminals occurs. Although most of the earlier work was on LTP in the hippocampus, LTP has also been found to occur in visual cortex [25], sensorimotor cortex [26], and prefrontal cortex [27], where it gives every indication of having essentially the same mechanism as it does in the hippocampus [17]. This is also the case in the spinal cord [8,10]. Differences Between LTP and Windup 1. Probably the most obvious difference between LTP and windup is in the time-course of the two processes. Windup lasts a few minutes at most. LTP comprises several different processes with different time-courses that may occur in different combinations in response to any particular conditioning episode(s): the term has been used to describe variations lasting from 20 min to several months. The current consensus seems to be that a potentiation has to last at least 1 h to be called LTP. Shorter lasting forms (20-60 min) are now generally called short-term potentiation. 2. It should be noted that after the initial depolarization involved in the induction of LTP, the resting membrane potential, input resistance, and passive membrane properties of postsynaptic cells are identical with those at nonpotentiated sites [28]. This, of course, is in contrast to the situation with windup, which is due to continuing depolarization. 3. It is also the case that LTP can (usually) be prevented from occurring but is not reversed by blockers of NMDA receptors. This again contrasts with windup, which is prevented and would also be reversed (if one could do the experiment before it decayed) by NMDA blockers. In fact once LTP has been induced, very little is known that will reverse it. Cooling a hippocampal slice for 5 min within the first 20 min after conditioning will reverse LTP--or perhaps simply prevent induction of some of its slow-starting components [29]. The protein kinase C inhibitor staurosporine [30] and the haem oxygenase (carbon monoxide production) inhibitor zinc protoporphyrin IX [22] have been reported to reverse LTP if applied in the first hour or so after induction. However, once the potentiation has been established for some time there is nothing known currently which will reverse it except the induction of LTD, at lease some forms of which may represent depotentiation of LTP [31]. Long-Term Depression LTD is, as its name implies, a long-lasting decrease in the efficacy of synaptic transmission, which can also be caused by high-frequency conditioning stimulation. In fact, in contrast to LTP, which requires high-frequency conditioning, a wide variety of both high- and low-frequency conditioning protocols have been shown to cause LTD [17,31-33]. Although it was first described in the cerebellum 12 yr ago [34], the mechanism of LTD has only recently begun to be characterized. This is possibly because in the hippocampus even the existence of LTD has been controversial, with some influential laboratories unable to induce it all, although many others can. However, there is no controversy about the fact that LTD occurs in other regions of the CNS, including the visual cortex [25], sensorimotor cortex [26] and prefrontal cortex [27], the striatum [35], and the spinal cord [8,9]. Interest in LTD is increasing as both its existence and its importance become more widely recognized: this is reflected in a recent burst of reviews [17,31-33]. LTD in the hippocampus and the neocortex seems to be induced by a surge of calcium concentration in the postsynaptic cell similar to, but smaller than, that which induces LTP. As with LTP, the source of the calcium surge seems not to be important--it can come from a large surge at a neighboring synapse, from internal calcium stores, from influx through voltage sensitive calcium channels, or even through NMDA receptor-associated channels--what is important is the amplitude of the increase [32]. Probably, a large increase triggers LTP and a moderate increase triggers LTD. This fits with a model proposed by Lisman [36], which suggests that a large increase of calcium concentration activates protein kinases, which phosphorylate phosphoproteins such as transmitter receptors, while a more moderate increase in calcium concentraion activates phosphatases, which dephosphorylate such phosphoproteins. Possibly, LTD represents a dephosphorylation of the critical molecules which were phosphorylated to produce LTP. For example, phosphorylation of masked AMPA receptors in the postsynaptic membrane may make them available, and dephosphorylation may make them unavailable again. In the cerebellum [37], and possibly in one region of the spinal cord [10], the two enzymes may well have the opposite calcium dependence, as the evidence suggests that a moderate postsynaptic calcium surge induces LTP and a larger surge induces LTD. Therapeutic use of a large calcium surge to induce spinal LTD is probably not a viable proposition at present, as a calcium ionophore specific for the postsynaptic cell is not available, and a large calcium influx in presynaptic terminals would cause catastrophic levels of transmitter release. However, future pharmacologic developments may allow reassessment of this possibility. The "calcium hypothesis" of the mechanisms of LTP and LTD explains the observation that in many regions of the CNS the same conditioning stimulus parameters give rise sometimes to LTP and sometimes to LTD, depending on factors which have not been elucidated. This is reminiscent of the induction of either long-lasting increases [5] or long-lasting decreases [7], in spinal reflexes by essentially the same conditioning stimulation (see below). Factors that might influence the probability of induction of LTP or LTD include modulating influences from other synaptic projections, which modify postsynaptic excitability and influence the probability with which a given input activity pattern induces postsynaptic calcium entry reaching thresholds for induction of LTP or LTD. For example, combined application of cholinergic and adrenergic agonists substantially lowers the threshold for LTP induction in the neocortex [38], which suggests that activity of adrenergic and cholinergic projections onto a postsynaptic cell would increase the likelihood of LTP induction at glutamate synapses. Endogenous opioids and mu opioid receptors appear to be involved in NMDA-independent LTP in the hippocampus, as their antagonist, naloxone, blocks induction of NMDA-independent LTP [39,40-43]. kappa Opioid receptors may have the opposite effect, with their agonist, dynorphin, reported either to inhibit [44,45] or have no effect on [42] NMDA-independent LTP. Naloxone is reported to partially reverse LTD in the spinal cord [10] and to block heterosynaptic depression in the CA3 region of the hippocampus [44]. Prolonged Activity-Induced Changes in Reflex Excitability If C fibers in the decerebrate, spinalized rat are stimulated at 2 Hz for 10 min the ipsilateral flexion reflex is increased for 3 h [6]. If C fibers in the decerebrate spinal cat are stimulated at 2 Hz for 15-30 min the flexion reflex is inhibited for 1 h, in a naloxone-reversible fashion [7]. The mechanism of these long-lasting changes is not clear, but it seems quite likely that the long-lasting increases and decreases in reflexes described here could be due to induction of LTP and LTD, in the spinal cord. findings could be as with this and the fact that the long-lasting reflex increases found in their are which they that a conditioning stimulus in one of increases the response to other of to that an excitability increase of postsynaptic cells is forms of LTP are i.e., to the that were activated by the conditioning However, since the reflex by and is it could be that the long-lasting increase is due to LTP of from the on which the different A second that could be as that the long-lasting increase in reflex excitability is mediated by a mechanism other than LTP is that the application of antagonists or the long-lasting increase in excitability of the flexion reflex caused by application of or to the and long-lasting produced by chronic nerve injury LTP is but not by blockers. However, while the activation of by causes a burst of activity which could be regarded as that produced by other forms of nerve the situation with is complicated by which causes continuing input and peripheral sensitization of In one of hyperalgesia produced by in the central changes on stimulation of A were found to on a continuing of input from the chronic nerve injury would in continued of noxious input from the individual EPSPs in the spinal cord are by it might be that the peripheral component of an increased reflex excitability would be by these blockers There is as no evidence that increases in reflex excitability produced by brief stimulation of peripheral are reversed by NMDA blockers. the other there are several similarities between increases of reflex excitability and LTP. forms of LTP are for their induction on the activation of NMDA receptors (see and the long-lasting of the reflex produced by either brief stimulation of the nerve or application of to the is also prevented by NMDA antagonists hyperalgesia caused by nerve injury is also on the activation of NMDA receptors for induction hyperalgesia caused by is not on NMDA receptors but on metabotropic glutamate receptors metabotropic agonists will also produce LTP in the hippocampus hyperalgesia caused in by nerve injury is with an increase in membrane binding of protein kinase C in spinal and which an increase in the membrane binding of protein kinase C in hippocampal LTP oxide which also prevent of cyclic guanosine of reflexes and there is some evidence that nitric oxide is the retrograde messenger in LTP although this is by no [19]. the long-lasting reflex decrease described by et al. [7] is identical with LTD is not However, one between LTD in the spinal cord [10] and the long-lasting reflex decrease is that both are by More evidence is on this The that central nervous system changes produced by bursts of activity can induce long-lasting pain syndromes has to the of was normal pain can be by before and as well as after This represents a major in and When chronic pain are the situation is not so If chronic pain syndromes are the of windup in the spinal they should be reversed by NMDA blockers. least one case has been in which NMDA blockers were It was found that one component of the chronic pain short-lasting of pain after a was by the NMDA It seems likely that this component of the pain However, the deep pain and the allodynia also by the were not by the NMDA LTP may well underlie these forms of chronic If LTP in the spinal cord does underlie chronic pain, induction of LTD by stimulation of the peripheral nerve should such In fact, current in cases of chronic pain sometimes does include the of nerve stimulation which may well work by LTD. evidence in of this is given below. of peripheral to chronic pain was because the of pain that stimulation of large nerve fibers should The was in several cases in a not by the gate-control was that a few minutes of stimulation sometimes produced of pain This of chronic pain is by a than a mechanism The of at frequency Hz) are not by in although it is that the of used in these while to block the action of on mu receptors, were to the action of on kappa or receptors However, with low-frequency Hz or burst is reversed by that both the depression of reflex excitability [7] and LTD in the spinal cord [10] are by naloxone, as described in sections, this the that at low-frequency by LTD in the spinal cord. In LTP and LTD of synaptic transmission in the spinal cord may be the processes chronic pain of the mechanisms of LTP and LTD in many regions of the CNS is and may well suggest useful for the of chronic