Long-term potentiation in spinal nociceptive pathways as a novel target for pain therapy

LTP is defined as a long-lasting increase of synaptic strength [5] that can be mediated by either pre- or postsynaptic mechanisms, or both. Synaptic strength is the magnitude of the postsynaptic response (i.e. postsynaptic current or potential) in response to a single presynaptic action potential at a monosynaptic connection. Recording of LTP therefore has two prerequisites (1) investigation of a monosynaptic connection and (2) recording of postsynaptic currents or potentials. In the spinal cord, there are currently two methods to record synaptic strength in nociceptive pathways that fulfil the above requirements [2, 27]. Both investigate the synaptic connection between primary afferent C-fibres (many of which are nociceptive) and superficial dorsal horn neurons, which is therefore the focus of the present review. In vivo, synaptic strength between primary afferent C-fibres and superficial dorsal horn neurons can be measured in adult rodents by stimulating the sciatic nerve and recording C-fibre-evoked field potentials in superficial dorsal horn that are known to reflect summation of postsynaptic, mainly monosynaptically evoked currents [3, 28]. In vitro, spinal cord slice preparations from young rodents with long dorsal roots are most often used to selectively investigate the synapse between C-fibres and neurons with a known role in nociceptive processing, e.g. lamina I projection neurons that express the neurokinin 1 (NK1) receptor [4, 17].

Several alternative methods have been used to investigate spinal LTP, but may not fulfil all of the above requirements. C-fibre evoked field potentials recorded in deep dorsal horn [14, 29] are very similar to those recorded in superficial dorsal horn, but it is not clear if they reflect monosynaptic transmission from C-fibres. Action-potential firing recorded extracellularly from deep dorsal horn wide dynamic range (WDR) neurons [30, 31] may in part reflect synaptic strength at the first nociceptive synapse but may also be affected by modifications of membrane excitability and synaptic inhibition. Optical imaging after bulk-loading of spinal cord slices with voltage-sensitive dyes does not allow distinction between neuronal and non-neuronal structures and between pre- and postsynaptic structures [4, 32]. Where data from these studies is used in the text or tables, it is specifically indicated.

Voltage-sensitive dye can also be loaded into the presynaptic terminals of primary afferents over the dorsal root. This approach allows to selectively monitor presynaptic electrical activity, but the exact relationship to transmitter release is not known [32].

LTP at the synapse between primary afferent C-fibres and superficial dorsal horn neurons can be induced by various protocols, including strong noxious stimulation of the input pathway and application of certain drugs (Table 1). Most studies use noxious electrical stimulation of the dorsal root or sciatic nerve that can be exactly controlled regarding stimulus intensity and duration and is therefore highly reproducible. Both high frequency stimulation (HFS, several bursts at 100 Hz) and low frequency stimulation (LFS, 2 Hz for several min) of primary afferent C-fibres induce LTP at the first nociceptive synapse in vivo [3, 4] and in vitro [4, 17]. While HFS may reflect the discharge of a subtype of C-fibres at the beginning of noxious mechanical stimuli [33], LFS is similar to discharge rates of C-fibres during peripheral inflammation [34]. Indeed, LTP can also be induced by peripheral inflammation (injection of formalin into the hindpaw, [4]) and, after removal of descending inhibition, by noxious heat or mechanical stimulation of the skin [13]. Mechanical nerve injury is a frequently used animal model of neuropathic pain and also induces LTP [11, 13]. A subset of primary afferent C-fibres express the transient receptor potential channel subfamily V member 1 (TRPV1) that is activated by both noxious heat and capsaicin and plays a major role in the induction of heat hyperalgesia [35]. Selective activation of these fibres by injection of capsaicin into the hindpaw has been shown to be sufficient for LTP induction [4], making TRPV1 antagonists or other methods that target the function of TRPV1-expressing C-fibres a potentially attractive target for prevention or modification of LTP at nociceptive spinal synapses. However, this has not been tested directly.

LTP at the synapse between primary afferent C-fibres and superficial dorsal horn neurons can also be induced by manipulations not directly activating the input pathway. In spinalized animals, prolonged burst stimulation of primary afferent Aδ-fibres induces LTP of C-fibre-evoked field potentials, possibly reflecting heterosynaptic potentiation [36]. LTP can also be induced in the absence of presynaptic activity by application of certain drugs (Table 1). Of special interest may be the induction of LTP by abrupt opioid withdrawal that may represent a cellular mechanism of opioid-induced hyperalgesia [21].

Intracellular Ca²⁺ rise in the postsynaptic neuron is a central step in the induction of many forms of LTP [5, 37], including LTP in spinal dorsal horn [4, 17, 21, 38]. When spinal LTP is induced by HFS or LFS, the massive release of glutamate from nociceptive primary afferents is thought to induce a postsynaptic depolarisation (primarily via α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid [AMPA] receptors) strong enough to remove the Mg²⁺ block from the N-methyl-D-aspartate (NMDA) receptor. Ca²⁺ influx through the NMDA receptor is one of the key signals that activates the intracellular machinery involved in LTP induction [2, 27, 39]. However, the postsynaptic Ca²⁺ rise achieved by NMDA receptor activation alone seems to be insufficient to induce LTP, as several parallel pathways that increase intracellular Ca²⁺ have been shown to be necessary for LTP induction (e.g., Ca²⁺ influx through T-type voltage-gated Ca²⁺ channels [VGCCs] and Ca²⁺ release from intracellular stores, triggered by activation of NK1 receptors and metabotropic glutamate receptors of group I [mGluRIs], see [3, 4, 7, 17, 38, 40]).

Therefore, LTP induction by conditioning stimulation can be interfered with at different stages: (1) Manipulations that reduce basal synaptic transmission at the first nociceptive synapse have the potential to prevent induction of LTP by indirectly preventing NMDA receptor activation. This is likely the case for μ-opioid-receptor antagonists (reduction of transmitter release and reduction of postsynaptic depolarization), AMPA receptor antagonists and γ-aminobutyric acid receptors of type A (GABA_AR) agonists/current enhancers (prevention of postsynaptic depolarization) (2) Drugs that directly interfere with NMDA receptor activation (e.g. NMDA receptor antagonists, Xenon, possibly EphB receptor antagonists) (3) Drugs that interfere with additional sources of activity-dependent intracellular Ca²⁺ rise (e.g. antagonists of T-type VGCCs, NK1 receptors or mGluRIs) (4) Drugs that interfere with intracellular pathways downstream from Ca²⁺ influx (see section on signal transduction pathways). Targets for prevention of LTP induction are summarized in Table 2, illustrated in Figure 1 and are discussed below. Table 2 also shows that the pharmacology of prevention of LTP induction is equivalent to the pharmacology of the prevention of hyperalgesia induction in animal models of inflammation and neuropathic pain.

Synaptic strength between primary afferent C-fibres and superficial dorsal horn neurons can be modified bidirectionally, with LTP or long-term depression (LTD) being induced depending on modalities of stimulation and on the stimulated pathway [36]. For cortical synapses, it has been proposed that the quantitative level of the activity-dependent rise in postsynaptic Ca²⁺ determines whether synaptic strength will increase or decrease. LTP is believed to occur with higher Ca²⁺ elevations that activate protein kinases while LTD would occur at lower Ca²⁺ elevations that activate protein phosphatases, possibly with a large "neutral" Ca²⁺ range between both states, where neither LTP nor LTD is induced [37, 41]. In spinal cord, this has not been tested directly. However, drugs that interfere with intracellular Ca²⁺ levels, like mGluRI receptor antagonists, can convert spinal LTP into LTD when applied during conditioning stimulation [38], suggesting that Ca²⁺ dependence of LTP vs. LTD may be similar in spinal cord and cortex.

In addition to conditioning stimulation, LTP between primary afferent C-fibres and superficial dorsal horn neurons can also be induced by abrupt opioid withdrawal [21]. It has been proposed that this novel form of LTP is induced postsynaptically, sharing mechanisms with stimulation-induced LTP, as it is abolished by preventing postsynaptic Ca²⁺ rise and by blocking postsynaptic G-protein coupled receptors or postsynaptic NMDA receptors [21]. The pre- vs. postsynaptic expression of opioid withdrawal LTP is currently a matter of debate, see [42] and our eLetter commenting on this paper available on the journal's web site.

In the clinical context, patients often present with already established hyperalgesia, e.g. in the form of chronic pain. If LTP indeed contributes to certain forms of chronic pain, then the question arises how established LTP can be therapeutically modified. Reduction of synaptic strength during established LTP may be differentiated into transient ("symptomatic") and permanent ("causal") approaches. Symptomatic approaches will temporarily suppress synaptic transmission at the potentiated synapse but not affect the causal processes that maintain LTP, so that synaptic strength will return to elevated levels after wash-out of the drug. In contrast, causal approaches will reverse the intracellular modifications that maintain LTP and thus permanently revert (depotentiate) synaptic strength towards normal values.

In hippocampus, the maintenance of LTP induced by electrical stimulation can be divided into two distinct phases [107, 130]. The early phase of LTP (E-LTP) sets in immediately after LTP induction but gradually fades away over the first few hours. It involves modification of pre-existing proteins like phosphorylation of synaptic AMPA receptors [131]. Consolidation of LTP requires expression of the late phase of LTP (L-LTP), which slowly develops during the hours after LTP induction and relies on de novo protein synthesis and gene transcription, e.g. resulting in the insertion of new AMPA receptors in the subsynaptic membrane [132]. According to the different mechanisms underlying the two phases of LTP, they may be affected by different drugs. In the rat spinal cord, the late, protein-synthesis-dependent consolidation phase of LTP slowly develops during the first few hours after stimulation, reaching its full expression between 3 and 6 hours after LTP induction [133]. Some drugs do not affect LTP induction but selectively interfere with spinal LTP consolidation by inhibiting the development of L-LTP when given before spinal LTP induction (antagonists at D1/D5 dopamine receptors, TrkB receptors, poly-ADRPTs, see Table 3). Other drugs induce a slow decay of LTP when given very early (15 min) but not later (30 min) after LTP induction (inhibitors of PKA, PKC, ERK, see Table 3). Kinetics and time course suggest that these drugs act by interfering with L-LTP development while leaving established E-LTP unaffected.

Although the time course of the different phases of LTP in humans is currently unknown, modification of fully established L-LTP is presumably most important for possible clinical applications. Thus, animal experiments identifying drugs or interventions of possible clinical interest for the causal treatment of established LTP-associated hyperalgesia should be designed as follows: (1) induction of LTP by HFS, LFS, natural noxious stimulation or opioid withdrawal, (2) application of the drug during fully established L-LTP (i.e. at least 3 h, better 6 h after LTP induction [133]) and (3) if LTP is depressed, true reversal should be differentiated from prolonged drug action by application of an antagonistic drug to ensure that the effect persists after the drug action has been terminated. Alternatively, recording should be continued for a time period ensuring complete washout of the drug. Few studies have tested the effect of drugs or interventions during established L-LTP (≥ 3 h after LTP induction, see Table 3). Currently, only two drugs have been identified that depress established L-LTP (diazepam and clonidine), and only for diazepam, true reversal of L-LTP has been corroborated by use of an antagonistic drug.

Targets for modification of LTP during the maintenance phase are summarized in Table 3, illustrated in Figure 2 and are discussed below.

In rodents, LTP in nociceptive spinal pathways can be induced by noxious stimulation. This has led to the notion that human pain following intense noxious stimulation, e.g. acute postoperative pain or chronic pain developing after an initial strongly painful event, may in part be due to LTP in spinal nociceptive pathways.

Clinical pain manifests as a variable combination of spontaneous pain, hyperalgesia and allodynia (Footnote: according to the new definition of allodynia proposed by the IASP task force in 2008, only pain induced by stimuli not capable of activating nociceptors is classified as allodynia. At present, brush-induced allodynia, that has been shown to rely on transmission via primary afferent Aβ-fibres [145], is the only established example of allodynia according to this new definition). In humans, intense noxious stimulation or tissue injury typically evoke thermal and mechanical hyperalgesia within the stimulated/injured region (primary hyperalgesia) and mechanical hyperalgesia and brush-induced allodynia within a larger surrounding region of non-injured skin (secondary hyperalgesia). While primary hyperalgesia reflects sensitization of nociceptive primary afferents and also includes central mechanisms, secondary hyperalgesia is thought to selectively rely on central (spinal and/or supraspinal) mechanisms [2, 146]. In chronic pain, spread of hyperalgesia to sites distant from the initial site of injury or even affecting the whole body, manifesting as a general elevation of pain sensitivity, may occur [26, 147‐150].

Before discussing the possible implications of injury-induced LTP for human experimental and clinical pain, it is important to determine which of the above manifestations of pain may be due to or enhanced by spinal LTP. LTP at synapses between nociceptive primary afferent C-fibres and superficial spinal dorsal horn neurons amplifies nociceptive signals. Therefore, LTP can account for hyperalgesia and possibly for small reductions in nociceptive thresholds and increases in size of hyperalgesic area. Hyperalgesia to pinprick stimuli is a frequent finding in human experimental or clinical hyperalgesia. Under normal conditions, pinprick stimuli are thought to be conducted by Aδ-fibres [151]. It is presently not known if spinal LTP also affects Aδ-fibre mediated synaptic transmission. However, recent work shows that pinprick hyperalgesia after inflammatory or nerve injury can be mediated by a subclass of C-fibres [152], suggesting that pinprick hyperalgesia might also relay on spinal LTP at C-fibre synapses. Brush-induced allodynia is thought to rely on input via primary afferent non-nociceptive Aβ-fibres [145]. Whether maintenance or modulation of allodynia outside the stimulated or damaged area is dependent on C-fibre sensitization remains controversial [153‐155]. Therefore, the LTP at spinal C-fibre synapses described in the present review is unlikely to solely account for the origin of brush allodynia, although it might contribute to its modulation or maintenance. Although LTP at C-fibre synapses cannot induce spontaneous pain, it may exacerbate spontaneous pain in the region of an injury. Spontaneous pain appears as the result of spontaneous activity in primary nociceptive afferents or central nociceptive neurons. Spontaneous activity in primary afferents, e.g. resulting from peripheral sensitization or from ectopic activity [1], may be amplified in the spinal cord if LTP is present, leading to enhanced pain intensity.

LTP has a homosynaptic component, expressed at the same synapse that was activated by the conditioning stimulation. Homosynaptic spinal LTP may contribute to primary, but not to secondary hyperalgesia. However, synaptic plasticity may in addition be heterosynaptic, i.e. spread to neighboring synapses that have not been directly affected by the conditioning stimulation [156‐158]. Studies investigating spinal LTP in rodents typically use supramaximal stimulation of the whole nerve trunk (sciatic nerve in vivo, dorsal root in vitro), presumably activating all intact fibres and consequently reaching all functional synapses between these fibres and second order neurons. Therefore, it is currently not possible to conclude whether this type of LTP is purely homosynaptic or also includes heterosynaptic components.

However, there is some direct evidence that heterosynaptic LTP occurs in spinal cord. When descending inhibition is removed, conditioning stimulation of Aδ-fibres induces LTP of C-fibre-evoked field potentials [36]. In addition, HFS of the tibial nerve or injury of the gastrocnemius/soleus motor nerve induces LTP of spinal field potentials evoked by stimulation of C-fibres in the sural nerve [12].

Heterosynaptic LTP may rely on various mechanisms. One possibility is that increased intracellular Ca²⁺ and second messengers spread intracellularly to neighboring nociceptive synapses within the same neuron and induce LTP at these synapses. In addition, several neuromodulators, e.g. ATP and BDNF, have been shown to induce LTP in the absence of conditioning stimulation of the input pathway [12, 122]. Intense noxious stimulation is known to release BDNF and ATP into the spinal cord [94, 159]. Diffusion of these substances through the extracellular space may induce heterosynaptic LTP at synapses and neurons not directly activated by the injury or conditioning stimulation and thus contribute to secondary hyperalgesia. In fact, heterotopic LTP has been shown to rely on release of BDNF in spinal cord [12]. It is not known how far these substances can diffuse through the spinal cord. At least, diffusion within the same segment to affect synapses in the termination territory of a neighbouring nerve is possible in rodents [12]. In contrast, diffusion within the spinal cord tissue to distant segments or affecting synaptic transmission in the entire spinal cord seems improbable. On the other hand, more widespread effects could result if sufficient concentrations of these substances reached the cerebrospinal fluid. Whether LTP induced by an initial painful event can account for the spread of hyperalgesia to distant sites of the body or for the generalized hyperalgesia typical for chronic pain [147, 149, 160‐163] is presently not known. Therefore, this manifestation of clinical pain will not be discussed in the present paper.

It has recently been discovered that in rodents, LTP in nociceptive spinal pathways can also be induced by abrupt withdrawal from opioids [21]. It has therefore been hypothesized that LTP may also contribute to the clinically important phenomenon of hyperalgesia following opioid withdrawal [21, 22, 42]. Although this has not been demonstrated directly, opioid-withdrawal LTP would be expected to affect nociceptive synapses throughout all spinal segments. Although it seems likely that opioid-withdrawal LTP can also lead to exacerbation of preexisting hyperalgesia or spontaneous pain, this has not been directly studied so far.

In conclusion, spinal LTP induced by an initial injury or noxious input may contribute to both primary and secondary hyperalgesia. LTP may also contribute to exacerbation of spontaneous pain. However, LTP induced by an initial painful event cannot explain brush allodynia. LTP induced by abrupt opioid withdrawal is proposed to lead to generalized hyperalgesia, possibly also including exacerbation of preexisting hyperalgesia.

It must be emphasized that although the above described sensory phenomena are compatible with spinal LTP, they may also be explained by other mechanisms. This is especially the case in primary hyperalgesia, where a substantial part of the hyperalgesia has been demonstrated to rely on sensitization of primary afferents [146]. The presence of secondary hyperalgesia is not in itself proof of the existence of LTP (i.e. altered synaptic strength), as secondary hyperalgesia can - and has - also been explained by changes in neuronal excitability (e.g. changes in neuronal membrane excitability) as well as changes in segmental or descendng inhibitory control [1, 2, 27, 88, 112, 164‐167]. Definitive proof of the existence of LTP depends on the direct measurement of synaptic strength, which is currently not feasible in humans. Therefore, we will, for the time being, have to accept that evidence for the existence of LTP in human pain pathways will remain indirect and circumstantial.

The following sections contain a more detailed description of those manifestations of human clinical and experimental pain that may principally be due to or exacerbated by spinal LTP, and compares their pharmacology to the known pharmacology of LTP in rodents. As primary hyperalgesia is in most cases accompanied by sensitization of nociceptive nerve endings, we will focus on secondary hyperalgesia (i.e. mechanical hyperalgesia in unstimulated or undamaged tissues) because this, at least, can safely be assumed to be due to central mechanisms [146, 168]. In order to provide relevance to the clinical situation, we will also mention the impact of secondary hyperalgesia induction - or its modulation - on clinical pain measures. Typical measures of clinical pain outcome are pain scores, particularly on movement, and analgesia consumption, particularly in the acute or postoperative context. However, it must be emphasized that such clinical measures reflecting subjective pain experience are regularly found to be only weakly correlated to alterations in pain processing as quantified by various forms of formal sensory testing [161, 162, 169, 170].

Perioperative sensory testing of the secondary hyperalgesia surrounding surgical incision is an attractive way of studying the time course of central pain amplification - and hence potentially LTP - in the clinical context. However, as already mentioned, it should be realised that due to the length of surgery, the effects of perioperative therapeutic intervention will not only influence LTP induction, but also its consolidation.

If LTP is involved in the maintenance of some forms of chronic pain, then therapeutic manoeuvres modifying established LTP in animal models could be expected to impact the hyperalgesia associated with established chronic pain in patients. As already discussed above, we will restrict this discussion to secondary hyperalgesia (i.e. surrounding the initial lesion). Not many human clinical studies using formal sensory testing have been performed in this context; most are small and have been carried out in the context of patients suffering from chronic pain associated with peripheral nerve injury. This is a relevant model as LTP has also been shown to play a role in nerve-injury related pain in rodent models [13].

Comparison of the human and animal literature presented above shows that established rodent LTP and established human hyperalgesia share a similar pharmacology with one major exception.

μ-opioid agonists decrease established secondary hyperalgesia in human volunteer and patient models. Gabapentinoids, too, have been shown to be effective against established hyperalgesia in human volunteer models. This is consistent with the results from animal models where μ-opioids and gabapentinoids suppress LTP during LTP maintenance phase.

Antidepressants have been shown to be effective against established hyperalgesia in pain patients. As antidepressants and central α₂-adrenergic-agonists such as clonidine share central monoaminergic mechanisms, the antihyperalgesic effectiveness of antidepressants in humans might find its animal equivalent in the effectiveness of clonidine in inhibiting established LTP.

However, in animal models, NMDA receptor blockade has no effects on established LTP, which contrasts with the evidence presented that NMDA receptor blockade by ketamine interferes with established secondary hyperalgesia in both human volunteer and patient models. One possible hypothesis explaining this difference would be that in the context of the human models presented, ongoing nociceptive input - albeit at a low level - leads to continuing induction of LTP, contributing to the maintenance of LTP, and thus explaining the sensitivity of apparently established secondary hyperalgesia to NMDA receptor blockade. Alternatively, LTP may be only one of various central mechanisms contributing to established human hyperalgesia and chronic pain, with alternative, NMDA receptor sensitive mechanisms participating in the maintenance phase.

COX inhibition, general anaesthetics, intravenous lidocaine or adenosine, have all been shown to be effective against established hyperalgesia in human volunteer or patient models but have not been tested in animal models of LTP.

Both in human and animal studies, it has often not been tested whether inhibition of established LTP/hyperalgesia outlasts drug effects, precluding differentiation between symptomatic (acute antinociceptive or antihyperalgesic) and causal (reversal of LTP/hyperalgesia) effects. In humans, there is some evidence that for ketamine, lidocaine, paracetamol/parecoxib and the atypical opioid buprenorphine, antihyperalgesia may outlast drug effects, suggesting that causal actions might be operating. However, all of these studies have been performed only in the ongoing transdermal electrical stimulation model, and their applicability to other models and the clinical context remains to be proved.

Our conclusions regarding LTP in rodents vs. humans and its pharmacological modulation are contrasted and summarised in Table 8.