Introduction

Neuropathic pain is defined by the International Association for the Study of Pain (IASP) as “pain caused by a lesion or disease of the somatosensory nervous system” [1•]. This definition is broad, covering over 100 conditions [2], and it involves injuries which span the entire pain neuro-axis. These injuries are often initially painful, in which case the pain serves to protect the damaged region until it can heal. However, in chronic neuropathic pain, the nervous system responds inappropriately to the damage through multiple mechanisms involving both the nervous system and its modulators. The unfortunate result is an unbalanced sensory system that misreads sensory inputs and can spontaneously generate painful sensations. Approximately 20 million people in the USA suffer from chronic neuropathic pain, with sometimes devastating losses of quality of life [2]. Treatments for neuropathic pain are non-specific and often insufficiently effective [3]. These treatments are not innocuous, and, for patients treated with opioids, can generate life-threatening side effects, highlighting the critical societal need for improved and customized strategies.

Therapeutic strategies for treatment of chronic neuropathic pain are limited by an incomplete understanding of how the nervous system maintains spontaneous pain following resolution of the initial injury. Before clinicians can provide precise treatment strategies for neuropathic pain patients, essential targets in the pathway must be identified. To achieve this goal, it is necessary to determine if maladaptive signaling in the central parts of the somatosensory system are sufficient to generate spontaneous pain. In this review, we focus on this key issue, by first presenting a brief review of both peripheral and central mechanisms in neuropathic pain and then presenting the preclinical and clinical evidence for each potential framework.

Common Neuropathic Pain Syndromes and Overview of Mechanisms

Neuropathic pain syndromes can be divided into two general categories: those that are consequences of a peripheral lesion or disease and those that are consequences of a central lesion or disease. This review focuses on conditions that are considered consequences of a peripheral insult. Central neuropathic pain conditions, such as central post-stroke pain (CPSP), are likely to possess different underlying mechanisms and warrant separate consideration.

Table 1 summarizes by general etiology some of the more common (and typically irreversible) neuropathic pain syndromes that originate from damage to the peripheral nervous system (PNS). As these conditions demonstrate, there are multiple routes to peripheral nerve damage, including mechanical, chemical, and infectious. These conditions share some general features, including spontaneous pain that is shooting, lancinating, or burning [4, 5]. Allodynia—i.e., a painful response to non-painful stimuli—as well as hyperalgesia, are also common features. The overlapping features of these syndromes can lend themselves to common treatment strategies and underscore the likelihood of shared pathophysiologic mechanisms.

Table 1 Some common neuropathic pain syndromes originating from damage to the peripheral nervous system (PNS)
Full size table

Peripheral Mechanisms in Neuropathic Pain

Peripheral nerve damage can result in chronic neuropathic pain through multiple routes [6••]. While the insult may be localized, the responses that lead to chronic pain are not. Peripheral terminals of pain-processing unmyelinated C fibers and thinly-myelinated Aδ fibers can spur the development of neuropathic pain after being affected by metabolic damage, toxins, medications, cytokines, and other inflammatory mediators [7], resulting in fiber density changes and neuronal hyperexcitability [8, 9, 10, 11, 12••]. Along the axon, injuries such as trauma, compression, hypoxia, inflammation, overstimulation, and chemical damage can induce fiber degeneration and alterations in channel expression and composition [13], in turn resulting in ectopic firing and faulty signal transmission [14]. In response to axonal damage and its sequelae, satellite glia and autonomic neurons can incur pain-promoting states though alterations in their overall numbers, distribution, sprouting patterns, and channel expression [15,16,17].

In the DRG and trigeminal ganglia, primary afferent cell bodies can be exposed to chemical, mechanical, and excitotoxic damage, and in neuropathic pain states demonstrate maladaptive changes in their membrane composition, synapse properties, and synapse location(s) [18,19,20]. The probability of peripheral nerve damage or its progression to neuropathic pain can also be increased by genetic predispositions and/or hereditary conditions [21, 22]. The ultimate result of the maladaptive mechanisms following peripheral nerve damage is a state of inappropriate signaling from the peripheral neuron to its second-order targets, with multi-factorial errors in both transduction and transmission [4, 23, 24] (Fig. 1).

Fig. 1
figure 1

Overview of peripheral and central changes contributing to neuropathic pain

Full size image

Central Mechanisms in Neuropathic Pain

With repeated or sufficiently intense stimulation, spinal and supraspinal nociceptive pathways can become sensitized to subsequent stimuli. With persistent nociceptive input [25•], like that seen in peripheral neuropathy, this central sensitization [26] becomes maladaptive. IASP defines central sensitization as “increased responsiveness of nociceptive neurons in the central nervous system to their normal or subthreshold afferent input” [27]. At the synapse of second-order neurons, this increased responsiveness can involve changes in calcium permeability, receptor overexpression, and synapse location [18, 28]. Also promoting a chronic pain state are microglia, whose hyperactivation triggers the release of pain-promoting mediators [29]. In supraspinal regions, the resulting misbalance between descending facilitation and inhibition is another major contributor to ongoing pain [30,31,32]. Maladaptive subcortical and cortical plasticity also contributes to painful interpretation of incoming signals [31, 32], with the ultimate result promoting a chronic pain state (Fig. 1).

Evidence for Peripheral Mechanisms: Preclinical

Injury and/or damage to the nociceptive afferents predominantly accounts for the onset of neuropathic pain. Peripheral mechanisms that initiate and maintain sustained excitation of afferent nerve fibers in neuropathic pain have been extensively studied utilizing multiple rodent models, such as spared nerve injury (SNI), chronic constriction injury (CCI), and spinal nerve ligation (SNL) [33]. In addition, specific disease-related neuropathies and the associated peripheral sensitization mechanisms have also been studied in rodent models of diabetes, chemotherapy, herpes zoster, and HIV-induced peripheral neuropathy [33]. In rodent spinal/sciatic nerve injury or constriction models, increased ectopic electrical discharge in myelinated axons (A fibers) begins generally within several hours of the induction of injury, and subsequently appears in unmyelinated axons (C fibers) within several days to weeks [12••, 34]. A wide variation in the fiber specificity, frequency, type, timeline of increased and/or sustained ectopic discharge, and cross-sensitization among A and C fibers at both peripheral and DRG cell body levels have been reported, which could be linked to the type of target nerve, injury, and the species/strain of animals studied. Multiple sources have subsequently shown that these changes in nerve fiber discharge lead to the development of various reflexive alterations in rodents that are referred to as neuropathic pain behaviors [12••]. Looking from a cellular/molecular aspect, distinct classes of receptors and ion channels in specific sensory neuron subtypes have been implicated for increased/sustained ectopic discharge. Due to the hyperexcitable nature of these neuronal injuries, voltage-gated Na+ (NaV) channels account for the primary molecular entity implicated in peripheral neuropathic pain conditions. Increased expression, trafficking, and peripheral targeting of several NaV channel isoforms, such as NaV1.3 and NaV1.6 (on myelinated axons) and NaV 1.7 and NaV 1.8 (on unmyelinated axons), have been shown in multiple rodent neuropathic models [35,36,37]. In addition, modifications in channel function, which lead to fast channel activation and increased current density, account for hyperexcitation of peripheral nerve fibers in response to neuropathy [32]. Several studies utilizing mouse genetics and pharmacological interventions targeting NaV channels have confirmed their involvement in peripheral nerve fiber excitation and neuropathic pain-related behaviors in rodent models [35,36,37].

Transient receptor potential (TRP) channels account for the major class of sensory detection/transduction channels, which upon activation by multiple pain-producing physico-chemical stimuli, provide the generator potential that is often needed to activate the NaV channels to elicit action potential firing (or electrical discharge) on nerve fibers (reviewed in [38]). Under patho-/physiological conditions, TRPA1 and TRPV4 could be activated in part by mechanical stimuli, TRPA1 and TRPM8 are activated by cold temperatures, and TRPV1 is activated by hot temperatures, as well as by acidic pH. Upon nerve injury/neuropathic conditions, TRPA1 has been shown to be directly activated by cell damage-related mediators, such as reactive oxygen/nitrogen species (ROS/RNS), leading to increased nerve fiber excitation and manifestation of mechanical and cold hypersensitivity behaviors in rodents (reviewed in [38]). Similarly, administration of paclitaxel-based chemotherapeutic drugs that cause peripheral neuropathy in rodents has been suggested to induce mechanical activation/transduction through TRPV4 [39]. Nerve injury, including neuroma formation, involves an inflammatory component, both at the site of injury and at the level of cell body in DRG, with local enrichment of (pro-)inflammatory mediators that provide the spices for nerve fiber sensitization. Modulation of TRPV1 channel function accounts for a major proportion of such sensitization via inflammatory mediators. Specifically, modulated TRPV1 gets activated by minimally acidic pH and at body temperatures, leading to sustained generator potentials and electrical discharge (reviewed in [38]). Both nerve damage/injury and the increased inflammatory microenvironment have been shown to upregulate the expression of these predominant sensory TRP channels, which in addition to functional changes lead to increases in the magnitude and duration of hyperexcitability of nerve fibers [reviewed in [38]. A large number of studies utilizing genetically modified mice lacking specific functional TRP channels and with the use pharmacological blockers of individual TRP channels have shown their critical involvement in peripheral nerve fiber excitation and neuropathic pain-related behaviors in rodent models (reviewed in [38, 40]).

Contrary to NaV and TRP channels, voltage-gated K+ (Kv), leak/two-pore domain K+ (K2P), and Ca2+/voltage-activated K+ (KCa) account for the vast majority of repolarizing or regulatory channels on sensory neurons/afferents (reviewed in [41]). Activation of these channels lead to membrane repolarization, thereby resulting in the suppression of electrical discharge/firing. Decreases in the protein expression of Kv1.1, Kv1.2, K1.4, Kv2.1, Kv2.2, Kv4.3, Kv7.2, Kv7.3, and Kv9.1, as well as of a number of K2P, KCa, and Kir/KATP have been shown in multiple rodent neuropathic pain models, which lead to a decrease in K+ currents and a resultant hyperexcitation of sensory nerves (reviewed in [41]). Except for Kv7 channels, extensive validation of the role of altered expression and/or function of most K+ channels utilizing pharmacological and mouse genetic approaches remains to be explored in nerve injury/neuropathic conditions.

In addition to neuronal channels and receptors, accumulation of infiltrating immune cells such as neutrophils, macrophages, and mast cells at the site of nerve injury constitute yet another peripheral cellular mechanism for nerve fiber hyperexcitation and sustained electrical discharge in majority of neuropathic conditions [42]. Continued supply of (pro-)inflammatory mediators by these immune cells account for both nerve fiber sensitization and neuronal damage, thereby exacerbating the neuropathy. In summary, numerous preclinical studies collectively suggest that (1) multiple mechanisms of peripheral nerve fiber excitation and sensitization operate in nerve injury/neuropathy conditions; (2) these mechanisms lead to sustained electrical discharge that feeds to the CNS and (3) which presumably accounts for continued excitatory ascending pain signal propagation to the brain. Pharmacological interventions aimed at reduction and/or blockage of peripheral nerve fiber excitation in rodent neuropathic pain models by targeting several abovementioned nociceptive ion channels/receptors have shown significant blockade of neuropathic pain-related behaviors [43]. Therefore, it is reasonable to argue that hyperexcitation and sustained electric discharge of peripheral nerve fibers constitute a predominant mechanism for peripheral neuropathic pain conditions.

Evidence for Peripheral Mechanisms: Clinical

In patients with phantom limb pain, single-fiber recordings of sensory fibers projecting into the neuroma demonstrate direct evidence of spontaneous ectopic activity and excessive action potential firing in [44]. Altered firing patterns in afferent neurons are also present in patients with primary erythromelalgia, for whom a mutation in the Nav1.7 channel can cause shifts in nociceptor activation thresholds [45]. As summarized in Table 2, in multiple types of chronic neuropathic pain, studies that block peripheral activity with a local anesthetic have resulted in significant alleviation or complete reduction of pain. Peripheral nerve stimulation, which disrupts incoming sensory signaling, has also been shown to provide significant pain relief in patients with neuropathic pain from post-herpetic neuralgia (PHN), complex regional pain syndrome (CRPS) type II, and traumatic and surgical nerve damage [53,54,55,56,57]. Collectively, these results suggest that peripheral input is an essential and necessary component for spontaneous neuropathic pain.

Table 2 Peripheral nerve blockade: effects on spontaneous neuropathic pain [25•, 46, 47, 48•, 49,50,51,52]
Full size table

Studies have also utilized DRG blockade techniques to demonstrate its key role in spontaneous pain generation. In amputees with phantom limb pain, Vaso et al. demonstrated that dilute lidocaine applied directly to the DRG in concentrations sufficient to suppress DRG ectopic firing, but not transmission of other sensory information, was capable of abolishing phantom limb pain in topographically appropriate regions [58•]. There is also growing evidence for the effectiveness of targeted DRG stimulation in the effective alleviation of chronic neuropathic pain [59, 60], and this evidence may expand as novel interfacing technologies continue to advance [61].

Evidence for Central Mechanisms: Preclinical

Changes in the Spinal Cord

Neuropathy-induced increases in spinal neuronal activity can be partly attributed to increased synaptic efficacy in the spinal cord dorsal horn. Activation of several protein kinases, including PKA, PKC, p38 MAPK, Src, ERK, and CaMKII, is observed in animal models of nerve injury. In painful neuropathy, ionotropic and metabotropic glutamate receptors exhibit phosphorylation and changes in trafficking that increase excitatory post-synaptic potential (EPSP) frequency and amplitude [62,63,64]. Increased post-synaptic activity is also achieved by alterations in glutamate homeostasis, resulting from increased expression of the vesicular glutamate transporters Vglut2 and Vglut3 in the superficial and deep dorsal horn, respectively [65]. This glutamate accumulation in synaptic vesicles is thought to increase EPSP amplitudes [66].

Spinal cord neurons also alter ion channel expression levels to acutely modify their properties following neuropathy. Examples include the voltage-gated calcium channel subunit α2δ-1 in the dorsal horn following induction of CIPN [67]. The ionotropic serotonin receptor 5-HT3 in the dorsal horn is the target of descending serotonergic facilitation of pain from the rostral ventromedial medulla (RVM). Activation of spinal 5-HT3 receptors is also associated with pro-inflammatory cytokine release and glial cell activation, changes that appear to be crucial for the maintenance of central sensitization [68]. Enhanced excitability is also brought about by a reduction in inhibitory tone. BDNF, in addition to its effects on microglia [69] and GluN2B phosphorylation, also inhibits presynaptic GABAA receptors, reducing presynaptic inhibition and causing spontaneous activity in lamina I output neurons, along with increased responsiveness to nociceptive input and the relaying of innocuous mechanical input [70, 71, 72•, 73]. Similar disinhibitory effects have been noted with radial neurons (morphologically distinct excitatory interneurons located in lamina II of the dorsal horn that show diminished inhibitory post-synaptic currents following injury [74]) and presynaptic reductions in GIRK potassium channel expression [75].

The production of inflammatory mediators by injured neurons and activated glial cells drives many of the physiological CNS changes associated with neuropathic pain. For example, dorsal horn neurons exhibit elevated expression of the chemokine SDF-1α/CXCL12 in a CIPN model [76,77,78], CXCL13 in a rat SNL model [79], and CCL3 and its receptor CCR5 in CCI in rats [80•, 81, 82]. Proinflammatory cytokines such as interferon-γ activate spinal microglia, a process that underlies many of the neuropathy-induced changes in spinal neuron behavior, most notably the hyperresponsiveness of wide dynamic range (WDR) neurons following CCI [83], and the activation of convergent nociceptive inputs following injury [84, 85]. Astrocyte activation is also crucial to the manifestation of neuropathic pain [86]. Resident astrocytes, as well as CD4+ T cells infiltrating the dorsal horn, secrete IL-17 following SNL. The resultant expression of IL-1β and IL-6 is, along with TNF-α, important in the maintenance of neuropathic pain [87,88,89]. ATP is released by injured dorsal horn neurons [90], whereupon microglial purinergic receptors are activated, leading to microglial proliferation and neuropathic pain [91,92,93,94, 95•]. The apparent reduction in the importance of microglial activity in the later stages of neuropathic pain models has led to the suggestion that microgliosis and inflammatory mediator production may be most important in the initiation of hypersensitivity and promoting the transition to chronic pain [96].

Changes in Brain Regions

In the ventral posterior thalamus (the major site of projection from the spinothalamic tract), wide dynamic range and nociceptor-specific neurons have shown hyperexcitability in neuropathy models [97]. As in the spinal cord, the vesicular glutamate transporter Vglut2 is increased in the thalamus, periaqueductal gray (PAG), and amygdala following SNI [98•]. The anterior cingulate cortex (ACC) shows increased expression of the astrocyte marker glial fibrillary acidic protein (GFAP) following CIPN—whether this is related to neuropathy-induced changes in glutamate and voltage-gated sodium channel expression in the same region remains to be investigated [99, 100]. Expression of the voltage-gated calcium channel Cav3.2 is upregulated in the ACC of rats after chronic constriction injury (CCI)—a finding that corresponds with enhanced T-type calcium currents in ACC neurons. In addition, the microinjection of a T-type calcium channel inhibitor partially relieves mechanical and thermal hypersensitivity post-CCI [101].

Microglial activation occurs in the mouse brain following CCI in regions associated with pain transmission and affect: the thalamus, sensory cortex, and amygdala [102]. Descending facilitation of neuropathic pain from the PAG is promoted by such glial cell activation in a CCI model [103]. The hippocampus has been reported to exhibit impaired long-term potentiation in SNI mice, an effect that was recently suggested to originate from the effects of tumor necrosis factor-alpha and microglial activation in this brain region [104]. This glia-driven change in synaptic plasticity and resultant mechanical hypersensitivity has also been reported in the primary somatosensory cortex in a mouse SNL model [105•].

Electrical stimulation of the thalamus causes spinal serotonin (5-HT) release that relieves neuropathic pain [106], consistent with the observation that intrathecal injection of serotonin can reverse allodynia [107]. The somatosensory cortex is also involved in descending anti-nociception through reducing “on” cell discharge in the rostral ventromedial medulla (RVM) in a 5-HT1A receptor-dependent fashion [108]. Interestingly, the administration of lidocaine to the RVM of SNL rats relieved allodynia in animals exhibiting pain, but precipitated allodynia in rats that had also undergone surgery, but did not exhibit pain-related behaviors [109], consistent with the bi-directional influence of the RVM in descending modulation. Projections from noradrenergic brainstem nuclei such as the locus coeruleus (LC), and other brain regions which project to the LC, are also regarded as mediators of descending pain inhibition [110, 111]. SNL in rats is associated with increased glutamate concentration in the LC and spinal norepinephrine release. These changes are proposed to underlie the impairment of endogenous analgesia following nerve injury [112] and can provide the rationale for the use of serotonin-norepinephrine reuptake inhibitors (SNRIs) in neuropathic pain. These combined data demonstrate a wide range of structural and functional changes occurring within the CNS following peripheral nerve injury. Spontaneous neuronal activity following neuronal disinhibition has been demonstrated in spinal cord and brainstem neurons, although whether this activity may occur in the absence of afferent (even trivial) input still requires further investigation.

Evidence for Central Mechanisms: Clinical

In human studies, features of central sensitization have been evaluated through multiple approaches [113]. Two testable parameters related to dorsal horn-level central sensitization are wind-up (exaggerated response to a train of stimuli) and secondary hyperalgesia (an increase in pain sensitivity to regions surrounding, but not including, the area of injury). In studies of humans with painful neuropathy, including CRPS type II, phantom limb, CIPN, and PHN, both wind-up and secondary hyperalgesia responses are significantly increased. Altered descending inhibition can be interrogated via conditioned pain modulation (CPM) studies, which test the endogenous ability of the CNS to inhibit painful stimuli. Studies comparing healthy volunteers with patients with peripheral polyneuropathy have demonstrated significantly impaired CPM values in painful neuropathy [114].

In patients with peripheral neuropathies, neuroimaging studies have shown multiple changes in activity and functional connectivity in CNS regions involved in pain processing and pain modulation [115, 116]. Neuroimaging of the cortical and subcortical regions in patients with painful neuropathies have identified alterations in activity and functional connectivity that correlate with the subjects’ neuropathic pain characteristics and treatment, including in patients with low back pain [32], PHN [117], diabetic polyneuropathy [118], neuroma pain [119], phantom limb [120], and CRPS [121, 122]. Additionally, structures in the mesencephalic reticular formation (including possibly the PAG and nucleus cuneiformis) that, in preclinical studies, have been shown to be essential to mechanical allodynia after peripheral nerve injury, demonstrate increased neuronal activity on functional neuroimaging in a human capsaicin-evoked secondary hyperalgesia model [123]. Cerebrospinal fluid cytokine levels in neuropathic pain patients have demonstrated increased levels of pain-promoting mediators including TNF-α, IL-6, IL-8, and IL-1β, as well as low levels of pain-decreasing IL-10 [124,125,126], providing further evidence that multiple central processes are responsible for creating a neuropathic pain state.

Recently, Alshelh et al. used resting-state fMRI in orofacial neuropathic pain patients to identify increased infra-slow oscillatory activity in the ascending pain pathway, including the spinal trigeminal nucleus, somatosensory thalamus, thalamic reticular nucleus, and primary somatosensory cortex; this increased oscillatory activity was not seen in control patients without orofacial pain [127]. This rhythm showed increased regional homogeneity in the spinal trigeminal nucleus region, consistent with a local spread of neural activity by astrocytes, and was suggestive of a self-sustaining thalamocortical dysrhythmia. While a variety of imaging studies provide evidence that critical pain pathway CNS components can generate autonomous signals, they provide neither evidence of causality between these oscillations and pain nor evidence that this activity is sustainable without afferent input.

Discussion

In contrast to the growing clinical evidence of peripheral contributions to neuropathic pain maintenance, studies demonstrating the ability of central sensitization mechanisms to independently generate neuropathic pain remain elusive. One key challenge in generating such potential evidence is the absence of agreed terms or criteria for diagnosing the presence of central sensitization in humans. Despite the existing definition (increased responsiveness of nociceptive neurons in the CNS to their normal or subthreshold afferent input), IASP taxonomy also notes that conclusions about the presence of central sensitization can only be made from indirect findings such as hyperalgesia and allodynia. Additional aspects of central sensitization, such as wind-up, long-term potentiation, and increased receptive fields—as well as potential testable criteria such as nociceptive flexion reflex or central sensitization inventory [128]—are not accounted for in the current IASP taxonomy.

Due to the above challenges, the presence of autonomic CNS pain-generating mechanisms could be tested by confirmation of the following hypothesis: “There are cases in which pain that was initiated by a peripheral nerve damage is independently maintained by central mechanisms.” To confirm this hypothesis, the following supporting data would be needed: (1) Evidence of spontaneous activity/firing of CNS neurons which does not occur under normal (non-injured) conditions, (2) causative relationship between this spontaneous/ectopic CNS firing and human pain, and (3) evidence that this spontaneous firing and pain persist despite the removal of afferent input. As of now, we are not aware of evidence confirming these three criteria. Indeed, there is evidence of spontaneous firing in the CNS neurons. The caveat is that some spontaneous activity can occur under non-painful conditions as well. Therefore, the relationship between the spontaneous activity and pain remains associative, and criterion (2) has not been met. Criterion (3) has been refuted in studies blocking peripheral input for a growing number of peripheral neuropathic pain states. Interestingly, this last criterion may also be unmet for central neuropathic pain states; it is yet to be shown whether blocking the peripheral input from areas of perceived spontaneous pain in central neuropathic pain states affect the experience of spontaneous pain.

Conclusions

Peripheral nerve damage provides opportunity for maladaptation at every point along the pain pathway. It is clear that profound CNS changes occur following peripheral nerve injury, and these changes contribute to the central sensitization. There is also evidence of spontaneous activity in CNS neurons after peripheral nerve damage, although this activity does not necessarily persist without afferent input. In peripheral neuropathic pain, effective blockade of afferent input seems to abolish spontaneous pain, even in the presence of signs suggesting central sensitization.

The nature of clinical studies—and the potential need for more definitive, agreed-upon criteria for confirming the clinical presence of central sensitization—has made it challenging to demonstrate the presence of an independent generator of neuropathic pain in the CNS. As a result, the relationship between spontaneous burst activity in the CNS and pain experience still remains associative rather than causative. In comparison, evidence continues to accumulate for the essential role that peripheral signaling plays in generation of neuropathic pain.

All these points together suggest that although many in the scientific community support the autonomous central pain-generating hypothesis, direct clinical evidence supporting this notion is yet to be generated. Therefore, our conclusion at this point in time is that central sensitization acts rather as an amplifier of peripheral signals, and not an independent pain generator in peripheral neuropathic pain conditions.