Hyperalgesia by synaptic long-term potentiation (LTP): an update.

Long-term potentiation of synaptic strength (LTP) in nociceptive pathways shares principle features with hyperalgesia including induction protocols, pharmacological profile, neuronal and glial cell types involved and means for prevention. LTP at synapses of nociceptive nerve fibres constitutes a contemporary cellular model for pain amplification following trauma, inflammation, nerve injury or withdrawal from opioids. It provides a novel target for pain therapy. This review summarizes recent progress which has been made in unravelling the properties and functions of LTP in the nociceptive system and in identifying means for its prevention and reversal. Copyright Â

Introduction

The vulnerability of tissues rises significantly in case of an injury or an inflammation. The nociceptive systems adapts to this by lowering response thresholds and by increasing response magnitude in order to maintain its protective function [1]. Behaviourally these adaptations manifest as hyperalgesia in experimental animals [2,3], in volunteers [4] and in patients [5]. Pro-nociceptive adaptations may occur at all levels of the neuraxis from nociceptive nerve endings, to spinal dorsal horn and all the way up to cortical neuronal networks. In contrast to sensitization of nociceptive nerve endings, some of the central mechanisms may persist long after the initial cause for pain and the need for special tissue protection has disappeared. Hyperalgesia then becomes maladaptive. The underlying central mechanisms can be grouped into two major categories: Impaired inhibition and enhanced excitation in nociceptive pathways. Multiple mechanisms have been identified so far which relate to the synthesis and/or the release of neurotransmitters, the density, the distribution and the activation of neurotransmitter receptors, the single channel conductance or the open time probability of ion channels and the number and morphology of synapses and dendritic spines. All of which can ultimately modulate the neurons’ intrinsic properties or synaptic strength.

From early on important insights into central components of hyperalgesia have been obtained in humans [6], from reflex measurements in experimental animals, for example [7], as well as from single neuron recordings in ventral- [8] and more importantly in the dorsal horn [9] of the spinal cord. It was only in later studies that synaptic plasticity has been assessed in the nociceptive system [10,11].

Nociceptive neurons are defined by their input (i.e. the excitatory mono- or polysynaptic input from nociceptive nerve fibres), but not by their function. Consequently nociceptive neurons comprise a very heterogeneous group of neurons including excitatory and inhibitory interneurons, projection neurons and motoneurons. Changes in the responsiveness of nociceptive neurons may thus have different and even opposing effects on pain depending upon the neurons’ function. For better understanding, here, the term ‘Principle Pain Neurons’ is used for neurons which, when discharging action potentials, trigger the perception of pain (see discussion in [12]). Here we review recent progress in understanding synaptic plasticity in spinal nociceptive pathways which, when expressed in principle pain neurons, amplify pain. The focus is on most recently published data. Comprehensive reviews on the synaptic mechanisms of hyperalgesia have been published [3,13].

Activity-dependent LTP at the first synapse in nociceptive pathways

Hyperalgesia and LTP are induced in an activity-dependent manner by strong or lasting discharges in C-fibres generating a central amplification of nociceptive responses. Typically LTP is induced by conditioning high frequency electrical stimulation (∼100 Hz; HFS, Figure 1A) at most synapses in the central nervous system and also at C-fibre synapses in the superficial spinal dorsal horn [14]. At C-fibre synapses LTP can further be induced by conditioning low frequency stimulation (∼2–10 Hz, LFS, Figure 1B) [15], but also by natural noxious stimulation (subcutaneous capsaicin, Figure 1C, formalin, noxious heat or pinching) and by acute nerve injury (sciatic nerve transection or crush) [15-17]. LTP has been demonstrated in vivo and in vitro, mainly in rats (e.g. [18] for recent studies) but also in mice [29]. As a general rule, conditioning stimuli which induce LTP at C-fibre synapses also cause hyperalgesia in behaving animals or human subjects (see below and [3,13] for reviews).

Figure 1

Induction of LTP at C-fibre synapses.

The figure illustrates different activity-dependent and -independent forms of LTP at C-fibre synapses. The graphs display mean time courses of amplitudes of C-fibre-evoked field potentials measured in the superficial spinal dorsal horn of adult, deeply anaesthetized rats. Field potentials were evoked by stimulation of sciatic nerve fibres at C-fibre intensity. Conditioning stimulation consisted of electrical stimulation of sciatic nerve fibre afferents at a high frequency (A, HFS, 100 Hz given four times for 1 s at 10 s intervals), at a low frequency (B, LFS, 2 Hz for 2 min), or subcutaneous injection of transient receptor potential vanilloid 1 channel agonist capscaicin (C, 1%, 100 μl). In D LTP was induced upon withdrawal from a brief (1 h) intravenous application of a high dose of remifentanil (450 mg kg−1 h−1 for 1 h, black horizontal bar). Modified from [15,18].

While LTP can be induced at most, if not all synapses in the central nervous system, the susceptibility for LTP induction and suitable parameters for LTP induction vary, however, considerably. A good example are synapses of nociceptive skin afferents which are apparently less prone to express LTP as compared to synapses from muscle afferents [26]. Conditioning stimulation of C-fibre afferents which innervate the skin may fail to induce LTP while identical conditioning stimulation of afferents in a mixed nerve or in a muscle nerve induces robust LTP [26]. This difference disappears when brain-derived neurotrophic factor (BDNF) is applied directly onto the spinal cord at a low concentration suggesting that lack of this neurotrophic factor in cutaneous afferents renders them less prone to express LTP. The differential susceptibility of skin versus muscle afferents to express LTP correlates well with their respective ability to trigger prolonged facilitation of nociceptive reflexes [30].

Activity-independent forms of LTP

Opioid withdrawal LTP Hyperalgesia and spinal LTP can also be induced in the absence of any activity in nociceptive nerve fibres. A clinically relevant example is hyperalgesia which develops after abrupt withdrawal from opioids. This form of hyperalgesia may also involve expression of LTP at C-fibre synapses [18]. A brief application of the ultra-short acting μ-opioid receptor (MOR) agonist remifentanil in vivo or d-Ala2, N-MePhe4, Gly-ol]-enkephalin (DAMGO) in vitro leads to an acute depression of synaptic strength in C-fibres (Figure 1D). Upon withdrawal synaptic strength not only quickly returns to normal but becomes potentiated for prolonged periods of time (Figure 1D). The induction of withdrawal LTP at C-fibre synapses in vitro [18] requires activation of postsynaptic G-proteins, postsynaptic NMDA-receptors and a rise in postsynaptic Ca2+ levels [18]. Some MOR agonists activate additional pro-nociceptive mechanisms. For example, withdrawal from fentanyl or morphine not only causes opioid withdrawal LTP but in addition activates descending, facilitatory, serotonergic pathways acting on spinal 5-HT3 receptors [31]. Hyperalgesia results when descending facilitation and/or opioid withdrawal LTP are expressed at synapses between nociceptive C-fibres and principle pain neurons.

Other activity-independent forms of LTP are induced at C-fibre synapses by spinal application of BDNF [20], adenosine triphosphate (ATP) [27] or reactive oxygen species donors [24] and in nerve injured rats also by tumour necrosis factor-α (TNFα) [32].

Distinct signalling pathways for LTP induction versus LTP maintenance

The signalling pathways which are involved in the induction of LTP are different from those which are required for its maintenance. They further differ between different induction protocols for spinal LTP which are expressed at C-fibre synapses, see Figures 2 and 3.

Postsynaptic signalling for LTP induction

Virtually all known forms of LTP induction at spinal C-fibre synapses require a rise in postsynaptic Ca2+ concentration [14,15,18]. Postsynaptic Ca2+ rises by opening of postsynaptic NMDA receptors [14], T-type voltage-gated calcium channels [15], Ca2+-permeable AMPA receptors [33] and by Ca2+ release from intracellular Ca2+ stores triggered by activation of metabotropic glutamate receptors or neurokinin 1 receptors [19]. Metabotropic receptors mobilise intracellular Ca2+ by activation of ryanodine and inositol-1,4,5 trisphosphate (IP3) receptors via phospholipase C [19], see Figure 2A. The rise in postsynaptic Ca2+ then activates Ca2+-dependent signalling pathways involving protein kinase C (PKC), calcium-calmodulin-dependent protein kinase II (CaMKII) and nitric oxide synthase (NOS) [15]. Other enzymes involved are extracellular signal-regulated kinase (ERK) which induces a lasting phosphorylation and activation of the transcription factor cAMP responsive element-binding protein (CREB), see Figure 2A [34].

Figure 2

Signalling pathways of LTP induction at C-fibre synapses.

The schemes summarize elements of the signalling pathways which are required (A) or sufficient (B) for the induction of LTP at spinal C-fibre synapses. The elements involved in LTP induction are typically identified by the respective blockers (required elements) or activators (sufficient elements) which were applied topically to the spinal cord. Many of the involved signalling elements are expressed at more than one cellular site as shown in the figure. The cellular site(s) of action is/are thus in most cases not known, except when substances were applied directly into the postsynaptic neuron as shown for Ca2+, NMDAR, and GPCR (which are in boxes here) in A. Suggested signalling pathways are indicated by arrows. Diffusion of elements is illustrated by dotted lines. * indicates that activation of this element induces LTP in spinalised animals only.

Abbreviations and literature:

AMPAR: α-amino-3-3hydroxy-5-5methyl-4-4isoxazoleproprionic acid receptor [72]; ATP: Adenosine triphosphate [22]; BDNF: Brain derived neurotrophic factor [20,34]; CaMKII: calcium/calmodulin-dependent protein kinase II [15,18]; D1,5R: Dopamine receptor D1,5 [74]; EphB: Ephrin B receptor [29,75]; ERK: Extracellular signal-regulated kinase [76]; GLT-1: Glutamate transporter 1 [77]; GPCR: G-protein coupled receptor [18]; IP3R: Inositol triphosphate receptor [14,15]; mGluR1: Metabotropic glutamate receptor group 1 [78,79]; NK1R: Neurokinin 1 receptor [14,15,19,80,81]; NK2R: Neurokinin 2 receptor [80,81]; NMDAR: N-methyl d-aspartate receptor [11,14–16,18]; NO: Nitric oxide [15,65]; NOS: Nitric oxide synthase [15,65]; PKA: Protein kinase A [73]; PKC: Protein kinase C [15,18]; PLC: Phospholipase C [14,15,19]; P2X7, P2X4: Ionotropic purinergic receptor [22]; p38MAPK: p38 mitogen-activated protein kinases [20,27]; ROS: Reactive oxygen species [24]; RyR: Ryanodine receptor [18]; SFK: Src family kinases [25]; TNFα: Tumour necrosis factor α [25,32]; TNFαR: Tumour necrosis factor α receptor [25,32]; TrkBR: Neurotrophic tyrosine kinase receptor type 2 [20]; T-type VGCC: T-type voltage gated calcium channel [14,15,19]; 5HT3R: Serotonin type 3 receptor [82].

Not all forms of spinal LTP require the activation of all of these signalling elements for induction. For example, LTP induced by spinal application of ATP or BDNF [27,34], but not high frequency stimulation-induced LTP [32] depends upon p38 mitogen-activated protein kinase (p38 MAPK). Likewise, high- and low frequency stimulation-induced LTP [15,19] but not opioid-withdrawal LTP [18] requires activation of CaMKII.

Pre- and postsynaptic signalling for LTP maintenance

While the induction of all known forms of LTP at spinal C-fibre synapses requires postsynaptic signalling, recent studies suggest that the maintenance of LTP may involve both, post- as well as presynaptic signalling, see Figure 3A. The early phase of LTP consists of the first few hours of LTP expression. Early phase LTP involves posttranslational modifications of synaptic proteins, such as phosphorylation of synaptic AMPA receptors [35]. Conditioning LFS of primary afferent C-fibres induces phosphorylation of the GluR1 subunit of spinal AMPA receptor channels at Ser831 [36] which increases their unitary single channel conductance. AMPA receptor-mediated currents in spinal nociceptive neurons are further elevated by enhanced AMPA receptor expression and by modified trafficking [35]. AMPA receptors are largely located postsynaptically where they mediate neuronal excitation. AMPA receptors may also be expressed at or near the terminals of a subset of dorsal root ganglion cells where they, by contrast, mediate presynaptic inhibition but not facilitation [37]. Taken together these findings suggest a postsynaptic component to LTP maintenance via enhanced AMPA receptor function.

Figure 3

Signalling pathways of LTP maintenance and LTP reversal at C-fibre synapses.

The schemes summarize elements of signalling pathways which are required for the maintenance of LTP at spinal C-fibre synapses. Thus, when any of these elements is blocked established LTP diminishes or disappears (required elements for LTP maintenance, A). The diagram in B summarizes elements which, when activated reverse established LTP. These sufficient elements for the reversal of LTP are underlined. Elements which are not underlined are required for the reversal of LTP. When blocked these elements prevent the reversal of LTP by at least one of the sufficient elements. Blockers and activators of the respective elements were usually applied topically to the spinal cord. Many of the known signalling elements are expressed at more than one cellular site as shown in the figure. The cellular site(s) of action is/are thus not known in most cases. Suggested signalling pathways are indicated by arrows. Diffusion of elements is illustrated by dotted lines.

Abbreviations and literature:

AMPAR: α-amino-3-3hydroxy-5-5methyl-4-4isoxazoleproprionic acid receptor (unpubl.); A1R: Adenosine 1 receptor [83]; α2-AR: α2-adrenergic receptor [84]; α2δ VGCC: Voltage gated calcium channel [85]; CaMKII: Calcium/calmodulin-dependent protein kinase II [73]; cGMP: Cyclic guanosine monophosphate [84]; D1,5R: Dopamine receptor D1,5 [74]; ERK: Extracellular signal-regulated kinase [76,86]; GABAAR: γ-aminobutyric acid A receptor [57]; mAChR: Muscarinic acetylcholine receptor [84]; mGluR1: Metabotropic glutamate receptor group 1 (unpubl.); MOR: μ-opioid-receptor [85]; NMDAR: N-methyl d-aspartate receptor (unpubl.); NO: Nitric oxide [84]; NOS: Nitric oxide synthase [84] PKA: Protein kinase A [73]; PKC: Protein kinase [73]; PP1: Protein phosphatase 1 (unpubl.); RyR: Ryanodine receptor (unpubl.); TrkBR: Neurotrophic tyrosine kinase receptor type 2 [34].

By contrast, H.-L. Pan and his colleagues concluded from their data [38] that both, the induction and the maintenance of opioid withdrawal LTP at unidentified synapses in spinal dorsal horn is presynaptic. We provided, however, evidence that the induction of withdrawal LTP is postsynaptic at C-fibre synapses [18], see paragraph above and our eLetter to their report. After any postsynaptic induction the expression of LTP may, nonetheless, involve presynaptic mechanisms. And indeed upon withdrawal from fentanyl or morphine but not from remifentanil the paired-pulse ratio of C-fibre-evoked field potentials decreases suggesting an increased neurotransmitter release [31]. At synapses in the brain protein kinase M ζ (PKMζ) is one of the key factors responsible for the maintenance of LTP [39]. Recent studies suggest that this kinase is also required for plasticity in nociceptive pathways in the spinal cord [40] and in the anterior cingulate cortex [41]. It is presently unknown whether PKMζ is also involved in the maintenance of LTP at C-fibre synapses.

Late phase LTP develops slowly over the first hours after LTP induction and persists for days, weeks or even longer. Expression of late phase LTP requires synapse-to-nucleus signalling via signalling molecules such as ERK1/2 and cAMP all of them may trigger the activation of CREB. The transcription factor CREB controls the expression of a myriad of proteins, many of which are relevant for synaptic transmission. Late phase LTP can consequently be blocked by protein synthesis inhibitors [34] and may involve incorporation of new AMPA receptors into the postsynaptic membrane [42], see Figure 2B and [13,43,44] for recent reviews.

Role of glial cells for LTP induction

In the central nervous system neurons and glial cell heavily interact and mutually influence their functions [45]. This also applies to the nociceptive system where excitation of nociceptive nerve fibres not only activates spinal neurons but also spinal microglia and astrocytes which, in turn, release neuroactive substances [46]. The release of these gliotransmitters contributes to the induction and perhaps also to the maintenance of LTP at C-fibres, see Figures 2A, B, 3A, and B. For example, HFS-induced LTP can be prevented by blocking or silencing spinal P2X7 receptors which are largely expressed on microglia [22]. Activated glial cells, Src-family kinases and p38 MAPK all contribute to the induction of LTP at C-fibre synapses via release of TNFα and activation of TNF receptor-1 [25]. In the absence of any C-fibre activation, spinal application of BDNF [20] induces late-phase LTP which requires activation of spinal microglia, Src-family kinases and p38 MAPK, see Figure 2A, B. This will consequently induce the release of TNFα, interleukin-1 and interleukin-6, among others [47,48].

Is spinal LTP homo- or heterosynaptic in nature?

Activity-dependent LTP may not only affect synapses which were activated by the conditioning stimulus. LTP may also ‘spread’ to inactive synapses converging onto the same postsynaptic neuron. It is still unknown if LTP in nociceptive pathways is homosynaptic in nature. Homosynaptic LTP at nociceptive synapses with principle pain neurons leads to primary hyperalgesia. Heterosynaptic LTP at synapses between nociceptive afferents and principle pain neurons would cause pain amplification outside but close to the area of injury or inflammation, that is, secondary hyperalgesia. A recent study by Carole Torsney [49] suggests that hindpaw inflammation by complete Freund's adjuvant leads to a heterosynaptic facilitation of monosynaptic Aδ-fibre input to spinal lamina I neurons expressing the neurokinin-1 receptor. This finding could well explain heterosynaptic mechanisms underlying mechanical hyperalgesia. Ongoing studies in our laboratory further suggest that in superficial spinal dorsal horn homo- and heterosynaptic forms of LTP are expressed at C-fibre- and GABAergic synapses, respectively [50].

Prevention of LTP induction in nociceptive pathways

Previous studies have identified a growing number of targets for preventing LTP induction [13]. Clinically useful tools for preventing LTP induction include NMDA receptor antagonists, for example, [11,15,51], opioids [52,53] and the noble gas xenon [54]. NMDA receptor antagonists proved effective also in volunteers [55]. The inducibility of spinal LTP is further modulated by descending systems originating from various brain sites. When descending pathways are interrupted conditioning stimuli which are normally ineffective may now induce LTP [16] indicating a pre-emptive function of endogenous antihyperalgesic systems. These include descending oxytocinergic hypothalamic pathways from the paraventricular nucleus [56]. The induction of LTP thus not only depends upon the parameters of conditioning stimulation and the type of afferent fibres involved, but also upon the modulation by endogenous pro- and antinociceptive systems.

Reversal of established LTP in nociceptive pathways

LTP at synapses between nociceptive nerve fibres and principle pain neurons causes hyperalgesia. Reversal of LTP, that is, ‘depotentiation’ thus constitutes a potential means to erase a memory trace of pain. When benzodiazepines are applied directly onto the spinal cord during early phase LTP its consolidation is impaired [57], see Figure 3B. A brief (1 h), systemic application of a high dose of the ultra-short acting MOR agonist remifentanil reverses LTP induced by low- or high-frequency conditioning stimulation of C-fibre afferents or by subcutaneous capsaicin [36]. The opioid-induced depotentiation involves activation of NMDA receptors, metabotropic glutamate receptors, Ca2+ release from ryanodine-sensitive intracellular stores and activation of protein phosphatase 1, see Figure 3B. AMPA receptor channels are phosphorylated at Ser831 by LTP-inducing stimuli. This leads to enhanced single channel conductance and thus synaptic strength. AMPA receptors are dephosphorylated at Ser831 by protein phosphatase 1 after high dose opioid administration. This probably constitutes a key mechanism for opioid-induced depotentiation [36]. Thus, in contrast to current believes opioids may not only temporarily dampen pain, they may also eliminate an important cause for hyperalgesia.

Hyperalgesia resulting from a biological cascade amplifier in a nociceptive daisy chain

LTP is a form of synaptic plasticity which can be induced at many different, it not all types of excitatory and inhibitory synapses in the central nervous system. It is thus not surprising that LTP is not only expressed at the first synaptic relays in nociceptive pathways [14,15,18]. LTP has also been observed at synapses of glutamatergic [28] and GABAergic [50] interneurons in superficial spinal dorsal horn and at excitatory synapses between neurons in the spinal trigeminal subnucleus caudalis and -oralis [58]. Furthermore, LTP can be elicited at synapses in putatively nociceptive relays in the anterior cingulate cortex [59-61]. When LTP is simultaneously expressed at multiple sites connected serially along excitatory nociceptive pathways it will boost nociception exponentially. In the first instance trauma or inflammation trigger sensitization of nociceptors [1]. The resulting enhanced, ongoing discharges in nociceptive nerve fibres induce LTP at the first synaptic relays [3,13]. An obvious outcome of LTP at excitatory synapses is the increased firing of action potentials of the postsynaptic neuron in response to presynaptic activity. LTP-inducing stimuli indeed lead to elevated C-fibre-evoked discharges in dorsal horn neurons, see [56,62,63] for recent studies. This in turn will probably facilitate LTP induction at synapses further downstream in nociceptive pathways [61]. And indeed, conditioning, LTP-inducing stimulation of sciatic nerve fibres causes enhanced positron-emission tomography signals in the primary somatosensory cortex and delayed responses in the amygdala, the periaqueductal grey, the rostral ventromedial medulla, and the dorsolateral pontomesencephalic tegmentum [64]. Such a sequence of events constitutes a biological cascade amplifier in a nociceptive daisy chain.

Behavioural correlates of LTP in nociceptive pathways

Conditioning stimuli which induce LTP at spinal C-fibre synapses lead to hyperalgesia in behaving animals. Under local muscular paralysis with lidocaine, brief conditioning HFS of sciatic nerve fibre afferents at C-fibre strength leads to thermal [65] and to mechanical [66] hyperalgesia for 6–9 days at the ipsilateral but not at the contralateral hindpaw.

The final proof for any pain mechanism is a perceptual correlate in the human subject. In volunteers, transcutaneous conditioning HFS of cutaneous nerve fibres induces a long-lasting increase in pain sensitivity at the stimulation site (homotopic facilitation) as well as in the immediately surrounding skin area (heterotopic facilitation) [67]. A number of recent studies have confirmed and extended the initial reports of perceptual [55,68-70] and electrophysiological [71] correlates of spinal LTP in volunteers, see [3,13] for reviews. At present a direct comparison between the results obtained from experimental animals with those from human subjects is, however, hampered by the fact that in humans conditioning stimulation was always applied to a small set of cutaneous afferents, while in most previous animal experiments conditioning electrical nerve stimulation recruited virtually all fibres in large mixed nerves, thus including muscle afferents. This probably makes a major difference as muscle and skin afferents differ substantially in their ability to express synaptic LTP [26]. Nonetheless the volunteer studies importantly demonstrate that perceptual correlates of LTP can be demonstrated in humans and they are probably also relevant for pain in patients.

Concluding remarks

LTP is a feature of most, if not all synapses in the central nervous system but its properties much depend on the type of synapse involved, the induction protocols used and the context of its induction. LTP at C-fibre synapses constitutes a powerful model system for the prolonged amplification of nociception. LTP probably contributes to enhanced pain-related behaviour in experimental animals and to the amplification of pain perception in human subjects. Understanding LTP in nociceptive pathways appears to be promising for developing better strategies for the prevention and the treatment of some types of chronic pain.

References and recommended reading

Papers of particular interest, published within the period of review, have been highlighted as:

• of special interest

•• of outstanding interest