Introduction

Neuropathic pain (NP) is a chronic and extensive condition secondary to dysfunction or nerve damage, usually originated in the periphery. If the injurious stimulus perpetuates, dorsal horn neurons in the spinal cord become sensitized, enduring pain perception; process known as central sensitization (CS) [13]. Because inflammation is essential for NP development and that apoptosis is involved in CS, tumor necrosis factor-alpha (TNF-α) becomes an interesting study target [4, 5]. Several studies have investigated the relation between TNF-α expression in different nervous tissues and pain behavior; hitherto these studies have not been conclusive. Though in general, TNF-α is associated with cellular harm and pain behavior, it has also been shown that these effects are bound by temporal and spatial resolutions.

For decades, animal models have reproduced cardinal aspects of human pain, offering essential features for the preclinical study of NP. The main classification of NP models divides them into central pain, peripheral nerve injury, and secondary disease neuropathy models. In this review, we will focus on the peripheral nerve injury models in order to study the effect of these lesions on the process of central sensitization. The primary anatomical structures involved in these injury procedures are the sciatic nerve and dorsal root ganglion (DRG). These methods comprise partial ligature, complete ligature, root or nerve pinching, crushing, transection, and local inflammatory stimulus. The most representative examples include the chronic constriction injury model (Bennett model), partial sciatic nerve ligation model (Seltzer model), spinal nerve ligation, plexus avulsion, compression of the DRG/root, laminectomy, and disc injury [612].

TNF-α is considered to be one of the principal mediators in pro-inflammatory processes that involve necrosis, apoptosis and proliferation [5, 1315]. It is produced essentially by macrophages and T-lymphocytes in response to stressed or damaged tissue, and it may therefore serve as a systemic marker for tissue injury. For instance, complications found in inflammatory and autoimmune diseases are often accompanied by increased serum levels. In case of increased levels, a systemic inflammatory reaction leading to septic shock might occur. However, if TNF-α is present in low quantities it would generate a process responsible for local inflammatory symptoms, depending on the damaged structure [1618]. In the nervous system, when a local peripheral injury is produced, it generates activation and morphological changes in dorsal horn microglia [1923]. This process includes P2X4 receptors activation, followed by Ca2+ influx and p38 MAPK activation. As a consequence, microglia releases diffusible factors like TNF-α, modifying neuronal excitability and thereby pain perception [2426]. TNF-α has two distinct receptors, TNF receptor-1 (TNFR1 or p55) and TNF receptor-2 (TNFR2 or p75). The main effect of TNFR1 is related to apoptosis and is present in practically every cell; whereas TNFR2 has a limited constitutive expression and is basically associated with cell survival [27, 28]. This is particularly important in pain attributable to different effects depending on specific receptor activation [2931].

In this review, we present the functional and biochemical changes after peripheral nerve injury in animal models, focused on the role of TNF-α in the development of NP.

Search methods

Articles were collected via PubMed from 1999 until July 2010; in order to analyze the understanding progress in the last decade, a period with increased standardized methods since the initial works associating TNF-α and neuropathic pain were performed. The following keywords and their combinations were used: chronic neuropathic pain, tumor necrosis factor, dorsal horn apoptosis, central sensitization, peripheral nerve injury, and neuropathic pain model. The obtained citations were examined by two independent reviewers in order to assure the application of the inclusion and exclusion criteria.

The inclusion criteria were: original full-text articles, preclinical reports, written in English, quantitative data on inflammatory response and pain production, mechanical compression or irritation to neural tissue, involvement of peripheral nervous system injury, and detailed measurement instrument with objective examination protocol. The exclusion criteria comprised: omission of TNF-α levels measurements, absence of standardized behavioral evaluations, presence of co-morbidity, use of TNF-α as inductor of lesion or inflammatory procedure, trauma-related or spinal cord injury models, and involvement of other cytokines.

Results

From the primary search a total of 2,896 publications were identified. Based on titles and abstracts, 2,814 articles were discarded due to irrelevance to the topic or duplication between different combinations. From the 82 prospective studies, 73 were selected according to the inclusion criteria. After applying the exclusion criteria we obtained 33 articles that were finally chosen for this review (Fig. 1) [3264].

Fig. 1
figure 1

Flow diagram representative of search strategy, identification, inclusion and exclusion of articles

From these 33 studies, 25 used rats of different strains and 8 applied genetically modified mice. The sample size of the experiments ranged from 7 to 360 animals, with a mean of 77 rodents per study. The most reproduced model was the ligation of the sciatic nerve denominated chronic constriction injury (n = 14), followed by nerve crushing (n = 8) and nerve transection (n = 4). The most common location to induce the inflammatory process was the sciatic nerve (n = 18) followed by the spinal nerves (n = 5). Tissue samples were obtained from diverse structures, mostly the DRG (n = 16), sciatic nerve (n = 13) and spinal cord (n = 10) (Table 1). Laboratory techniques used for TNF-α determination in these biopsies included primarily immunohistochemistry (n = 22), and secondary ELISA (n = 8), qPCR (n = 7), and immunoblotting (n = 5). Follow-up of pain behavior monitoring ranged from 5 days to 12 weeks, 14 days being the most repeated period (n = 11). Not all of the included studies in this review applied behavioral testing related to allodynia and hyperalgesia (n = 11); the rest assessed mechanical sensitive thresholds primarily with von Frey filaments (n = 18), and thermal responsiveness with the Hargreaves test (n = 10). Some other conduct assessment techniques comprising hot-plate (n = 2) and Randall–Selitto tests (n = 2).

Table 1 Demographic data from all included papers in this review

Mechanical allodynia was assessed in 20 studies through application of pressure in affected and non-affected limbs in order to evaluate painful responses. Each one of these studies showed a significant decrease in all cases in nociceptive thresholds as a consequence of the injury procedure when compared to sham groups. Initiation of pain typically started 24 h after the lesion intervention, reached maximum levels at 1–2 weeks, and returned close to presurgical control levels after 4 weeks. In two reports bilateral allodynia was present, in one case only for 7 days and referred in the second one as significant and long-lasting [42, 64]. Nine articles evaluated pharmacotherapy and found the highest effectiveness when the drugs were applied within 7 days after injury. In a specific case of cannabinoid administration, an antinociceptive effect was achieved after applying high doses from day 7 to 14 after injury (20 mg/kg) [36]. Opioid administration also reverted successfully the acute nociceptive effects, but only at increased doses and tolerance development (10 mg/kg b.i.d. for 5 days) [55].

Thermal hyperalgesia followed a similar pattern as mechanical allodynia, showing lower thresholds at short-term measurements, starting to decrease and even reaching baseline levels after 14 days. In one article TNFR1-deficient rodents did not exhibit significant painful behavior after heating stimulation at any time point after injury [63] (Table 2).

Table 2 Detection methods to asses TNF-alpha expression and behavioral tests to measure pain

All models show increased TNF-α levels after injury. Pharmacological treatments that failed to decrease these levels included monotherapy of paracetamol, tramadol, cannabidiol, morphine, soluble TNF receptor, and IL-1 receptor antagonist. On the other hand, successful interventions that lower TNF- α expression included NMDA receptor antagonist MK-801, thalidomide, combined therapy of soluble TNF receptor with IL-1 receptor antagonist, chronic administration of morphine (dose-dependant with aggravation after interruption), α2-macroglobulin, nimesulide, a combination of paracetamol with tramadol, recombinant human erythropoietin and triamcinolone [32, 3436, 43, 45, 55, 61, 62, 64].

Chronologically, TNF-α expression was characterized by an immediate increase in affected structures starting as soon as 1 h after injury [45, 56, 59]. Peak levels are detected around 24 h, sustain for about 3 days, and in many cases then decreased, occasionally to baseline levels [40, 41, 45, 47]. Some publications describe a lowering of TNF-α expression after this point, but the mainstream pointed out the maintenance of a plateau phase up to 1–2 weeks, followed by a periodical decline. Measurements performed at 3 and 4 weeks reveal weak signaling of the inflammatory process [37, 3941, 45, 47, 54, 56, 59, 60]. Only one study documented prolonged increased TNF-α levels lasting at least 6 and 12 weeks, not only in nervous tissue but also in connective tissue, after applying the high force reaching task model [39].

The degree to which an altered TNFR1 and TNFR2 ratio contributes to pain behavior is still not clear. One study observed an increased TNFR1 expression several days after dorsal root lesion, while TNFR2 levels remained unchanged [44]. In contrast, a similar follow-up after sciatic nerve injury resulted in a twofold increased TNFR1 level and a sevenfold elevated TNFR2 level [41]. Yet another report showed that immediate measurement only after 6 h post-spinal nerve ligation revealed that both receptors showed similar increased expression in the dorsal root ganglia and remained at comparable levels up to the maximum follow-up [59].

Discussion

In this review we have summarized the reports that appeared over the past 10 years and described peripheral injury-induced TNF-α expression in the nervous tissue and its role in the production of CS. To date, there is a considerable amount of evidence to support the hypothesis that inflammatory process following tissue damage can lower pain thresholds. In this respect, many pathways and cytokines have been described to contribute in the development of chronic nociceptive signaling, including the three proinflammatory cytokines that induce activation of p38 MAPK: TNF-α, IL-1β and IL-6 [25]. Focus on TNF-α is because it is regarded as the archetypical proinflammatory agent, attributable to its function as an inductor of diverse processes that initiate and perpetuate the inflammatory response. At the same time, its increased expression in nervous tissues upon a pain stimulus suggests that this cytokine is essential for the development of pain sensation [6569].

Several authors published similar results with regard to TNF-α presence in peripheral and central nervous tissue after neural injury (Fig. 2). However, part of the main divergences remains in the involvement and interactions of TNFR1 and TNFR2. Thus far, it seems that both receptor types play a different role in the development of allodynia and hyperalgesia depending on the site of the lesion and the moment of the cascade activation. For example, TNFR1 has shown constitutive expression in the uninjured spinal cord in modulatory centers like the dorsal horn, whereas these areas hardly express TNFR2 [44, 70]. Nevertheless, this does not imply that after peripheral nerve injury and TNF-α release, TNFR2 could not be involved in nociceptive regulation [47]. This could also be dependant on the proximity of these lesions from the spinal cord, illustrated by earlier manifestation of pain in spinal nerve lesions (h) than in sciatic nerve lesions (days) [45, 59]. In judgment of the literature reported to our days, it is not still clear if the final nociceptive outcome is more dependant on the amount of cytokine or on the activation/inhibition of specific receptors (Fig. 3).

Fig. 2
figure 2

Mechanisms involved in pain. TNF-α increases hyperexcitability and consequently neuropathic pain via different peripheral and central mechanisms. At the site of injury TNF-α leads to activation of p38 and JNK, followed by increased TTX-R and VR1 expression, accompanied by production of other inflammatory agents and substance P. In the dorsal horn of the spinal cord, TNF-α induces increased activity of NMDA and AMPA receptors, leading to depolarization. It is not completely elucidated yet the role of TNF-α on inhibitory neurons, but it is suggested that it increases GABA and glycine release from presynaptic terminals, producing hyperexcitability

Fig. 3
figure 3

Intracellular pathways related to TNFR1 and TNFR2. On the left, TNFR1 is primarily related to the induction of apoptosis; though, it can interact via TRAF2 to promote cell survival under certain circumstances. On the right, TNFR2 function is oriented to follow parallel pathways that will lead to cell survival and proliferation. Probably the balance and interaction between these two receptors determine the final damage produced by lesions, and therefore the amount of pain

The increase of TNF-α immunoreactive cells or mRNA levels showed a similar temporal resolution, regardless of the studied tissue or injury model. As previously described, TNF-α production typically starts to reach maximal levels within 24 h, which are maintained for over 2 weeks and then usually drop [37, 3941, 45, 47, 54, 56, 59, 60]. This paradigm is important for understanding the development of painful symptomatology and for the therapeutic timing approach. Behavioral data show that the clinical presentation follows the presence of increased amounts of TNF-α in neural tissue [7173]. Fluctuation in TNF-α levels tended to match with increased or decreased states of mechanical allodynia and thermal hyperalgesia. Progressive nociception in injured animals was established after the first few days and maintained until its highest point around the second week, followed by a gradual drop up to baseline thresholds [37, 54, 56]. This transient pain behavior may be a consequence of the continuous adaptation of pain perception that familiarizes the animal with the applied stimuli, yet in each study the cytokine levels correlated closely to the pain sensation [41, 53]. In this respect, TNF-α involvement in pain could also be inferred in studies where wild-type mice showed greater levels of pain when compared to genetically TNF knock-out animals [74, 75].

The pharmacological therapies proved to be successful when applied during the first 7 days post-lesion, which is consistent with the temporal resolution of TNF-α expression, probably avoiding the initial inflammatory response [66, 76]. This is also an essential matter in the apparent dual intervention of TNF-α in neurons as toxic or protective agent. Originally, the function of this cytokine has been described as harmful, but previous evidence and data in this review suggest that protection and regeneration could be involved in a receptor-dependant mode. It has been described that TNFR2 has a fundamental participation in remyelination of neurons after injury, probably linked to its late participation in the inflammatory cascade [44, 70, 7779]. This suggestion also comes from delayed expression of TNFR2, which coincides with behavioral recovery in the late phases of pain sensation. Thus, it is of vital importance to understand the temporal expression of TNF-α and its receptors in order to develop a successful therapeutic strategy. In these days, we can observe that the same cytokine at different moments, acting on different receptors, could bring opposite histological and clinical outcomes. These observations bring the query as to which factors are more determinant for the production of pain, the concentration of TNF-α or the TNFR1–TNFR2 expression ratio. In this respect, a key question that remains is why patients with chronic NP do not benefit from this self-limiting protective mechanism. We believe that a balance alteration between the TNF-α receptors could determine the severity of pain, not only in a temporal expression (acute or chronic) but also on the extension of the produced damage. Animal models have proven to develop and reproduce pain in a comparable manner to human observations. In summary, there is evidence from physical injury models, transgenic models, and pharmacological interference studies that TNF-α expression in the spinal cord modulates pain sensation. However, future perspectives for treatment still require better understanding of its receptors and intracellular pathway interactions as pain modulators. Therefore, further translational studies may be performed to address this challenging topic and design more effective therapies by regulating inflammation-induced pain sensation.