Inhibiting tumor necrosis factor‐alpha at time of induced intervertebral disc injury limits long‐term pain and degeneration in a rat model

Background Painful intervertebral disc (IVD) degeneration has tremendous societal costs and few effective therapies. Intradiscal tumor necrosis factor‐alpha (TNFα) is commonly associated with low back pain, but the direct relationship remains unclear. Purpose Treatment strategies for low back pain require improved understanding of the complex relationships between pain, intradiscal pro‐inflammatory cytokines, and structural IVD degeneration. A rat in vivo lumbar IVD puncture model was used to 1) determine the role of TNFα in initiating painful IVD degeneration, and 2) identify statistical relationships between painful behavior, IVD degeneration, and intradiscal pro‐inflammatory cytokine expression. Methods Lumbar IVDs were punctured anteriorly and injected with TNFα, anti‐TNFα, or saline and compared with sham and naive controls. Hindpaw mechanical hyperalgesia was assayed weekly to determine pain over time. 6‐weeks post‐surgery, animals were sacrificed, and IVD degeneration, IVD height, and intradiscal TNFα and interleukin‐1 beta (IL‐1β) expressions were assayed. Results Intradiscal TNFα injection increased pain and IVD degeneration whereas anti‐TNFα alleviated pain to sham level. Multivariate step‐wise linear regression identified pain threshold was predicted by IVD degeneration and intradiscal TNFα expression. Pain threshold was also linearly associated with IVD height loss and IL‐1β. Discussion The significant associations between IVD degeneration, height loss, inflammation, and painful behavior highlight the multifactorial nature of painful IVD degeneration and the challenges to diagnose and treat a specific underlying factor. We concluded that TNFα is an initiator of painful IVD degeneration and its early inhibition can mitigate pain and degeneration. Intradiscal TNFα inhibition following IVD injury may warrant investigation for its potential to alter downstream painful IVD degeneration processes.

degeneration is highly associated with back pain. [4][5][6][7][8] However, many patients with magnetic resonance imaging evidence for IVD degeneration do not report back pain, 9,10 highlighting challenges in diagnosis and treatment. Consequently, nonsurgical interventions are most commonly recommended, and often of limited efficacy. 11,12 Surgery for disc degeneration has also yielded mixed clinical results. Specific phenotypes of painful IVD degeneration have been introduced to help distinguish aging from painful IVD degeneration conditions, and these phenotypes often involve structural defects and pro-inflammatory conditions. 4,7,[13][14][15][16] IVD injuries are known to precipitate a strong proinflammatory responses with macrophage infiltration that can induce permanent alterations to IVD structure and function. [17][18][19] Together, these findings demonstrate that IVD degeneration and proinflammatory cytokines play important roles in back pain, but the relationships are complex and nonlinear. A better understanding of the relationship between IVD degeneration and back pain is an important societal research priority, and a clearer understanding may provide insights to identifying therapeutic targets and to developing preventative treatment strategies for discogenic pain.
IVD degeneration-related pain is multifactorial and associated with intradiscal inflammation, neurovascular ingrowth into the IVD, sensitization of nervous system (ie, upregulation of pain-related neuropeptides in the dorsal root ganglia), impingement of adjacent nerve roots, biomechanical instability, and increased demand on the surrounding spinal muscles. [20][21][22][23][24] In particular, intradiscal proinflammatory cytokines tumor necrosis factor-alpha (TNFα) and interleukin-1 beta (IL-1β) are highly implicated in the development and progression of IVD degeneration and discogenic pain, and these cytokines can be produced by native nucleus pulposus (NP) and annulus fibrosus (AF) cells as well as infiltrating inflammatory cells. 18,[25][26][27][28][29][30][31][32][33][34] Upon IVD injury, macrophages and mast cells are recruited and release TNFα and IL-1β in the IVD and also induce native IVD cells to further produce pro-inflammatory cytokines, including TNFα, IL-1β, IL-6, and IL-8. [35][36][37] TNFα is highly associated with IVD degeneration because it has been demonstrated to promote extracellular matrix degradation via upregulation of catabolic mediators, including matrix metalloproteinases (MMPs) and a disintegrin and metalloproteinase with thrombospondin motifs (ADAMTS). [38][39][40] TNFα can also upregulate substance P, nerve growth factor (NGF), and vascular endothelial growth factor production, which may induce painful conditions through sensitization of the nervous system and neurovascular ingrowth into the IVD. [41][42][43][44] The importance of TNFα in discogenic back pain led to multiple human clinical trials treating chronic discogenic pain with TNFα inhibitors, yet results were mixed highlighting a need for further research. [45][46][47] There is a high research priority for investigating a direct relationship between TNFα and discogenic back pain that can further assess contribution of specific pain generators, including intradiscal pro-inflammatory cytokines and structural IVD degeneration, and such a controlled investigation requires an animal study.
Animal models are commonly used for studying IVD degeneration, with pain often used as motivation for studying IVD degeneration. In vivo rodent models play an important role in determining the underlying pathophysiology of painful IVD degeneration because known quantitative methods are available that can characterize specific painful responses. 24,48 However, few studies have directly assessed the pain associated with IVD degeneration, and no study to date has investigated the direct associations between pain, structural degeneration, and intradiscal pro-inflammatory cytokines in order to characterize the most important contributors to the painful process.
Furthermore, rats offer advantages over mice because the larger size enables more precise control of induced injuries. Annular injury models (using puncture or stab injury) are commonly adopted to induce IVD degeneration in rat models because the severity of injury can be controlled. Our group previously developed an in vivo painful IVD degeneration model in rat and showed that AF puncture with intradiscal injection of phosphate buffered saline (PBS) produced painful behavior and degeneration in the injured IVDs, and intradiscal injection of TNFα into the injured IVD further increased painful behavior up to 6 weeks post-injury. 48,49 The findings suggested that TNFα plays an important role in the development of IVD degeneration-related pain; however, it is unclear whether inhibiting TNFα at the time of IVD injury can prevent or mitigate IVD degeneration and degeneration-related pain. Furthermore, the relationships between pain, intradiscal pro-inflammatory cytokines, and IVD structural degeneration in this model are unknown.
This study applied an in vivo rat model of painful IVD degeneration to determine (1) a role of TNFα in causing painful IVD degeneration following an IVD puncture injury; and (2) the relationships between pain, IVD structural degeneration, and intradiscal proinflammatory cytokines in this model. We hypothesized that annular injury with intradiscal administration of TNFα would increase painful IVD degeneration, whereas blocking TNFα at the time of injury with intradiscal injection of anti-TNFα would prevent this painful IVD degeneration cascade. We also hypothesized that the painful behavior would be multifactorial but be predicted by intradiscal proinflammatory cytokines expression and IVD structural degeneration.

| Experimental design
A total of 36 skeletally mature 50 male Sprague-Dawley rats (4-5 months old) were used and randomly assigned into one of five experimental groups: naive (n = 8), sham (n = 8), PBS (n = 8), TNFα (n = 6), or anti-TNFα (n = 6). In the PBS, TNFα, and anti-TNFα groups, rat lumbar IVDs were exposed, punctured, and intradiscally injected with PBS, TNFα, or anti-TNFα, respectively. The injection of TNFα aimed to further increase the content of intradiscal TNFα in addition to the expected inflammatory response resulting from IVD injury; while anti-TNFα was injected to downregulate the intradiscal TNFα response induced by IVD injury. The sham group involved surgery to expose lumbar IVDs only without puncture or injection, and naive group involved no surgical procedures at all. Pain behavior and IVD height were measured before surgery and then weekly after surgery using hindpaw mechanical hyperalgesia test and lateral radiographs, respectively. After pain behavior and radiographic measurements were taken at the 6-week time point, the rats were euthanized, and the lumbar spines were isolated for morphological and biochemical analyses. All experimental procedures were approved by the Institutional Animal Care and Use Committee.

| Pain behavior measurement
Severity of pain was measured using hindpaw mechanical hyperalgesia test, shown to be a sensitive measurement technique to quantify the pain associated with IVD degeneration in rats. 48 A decrease in paw withdrawal threshold indicated an increase in pain sensitivity. The test was carried out in wire mesh-floored cages, and the rats were placed in the cages at least 20 min prior to measurement for acclimation. Calibrated von Frey filaments, ranging from 0.4 to 26.0 g (Stoelting, Wood Dale, Illinois), were then applied to the plantar surface of the hindpaw with sufficient force to cause buckling of the filament. The 50% paw withdrawal thresholds were determined for both left and right hindpaws using the up-down method with three trials for each side. 52 The withdrawal thresholds from left and right hindpaws were then averaged for analysis and normalized to preoperative values.

| Radiographic IVD height measurement
IVD heights of noninjured IVD (ie, L2/3 IVD) and injured IVDs (i.e. L4/5 and L5/6 IVDs) were measured via lateral radiographs with the animals under general anesthesia. Animals were anesthetized for approximately 10 min before taking the radiographs to minimize the effect of anesthesia on IVD height resulting from muscle relaxation and IVD swelling. 53 After the radiographic film was digitized using a flatbed scanner with backlighting, the image was magnified and analyzed in ImageJ. Both upper and lower vertebral boundaries of the motion segment were manually identified, and coordinates of the vertebral boundaries were obtained. The coordinates were then transferred to the MATLAB (Mathworks, Inc., Natick, Massachusetts), 49 which measured the averaged distance between the vertebral boundaries. A step wedge was used as a scale reference. Each disc was measured thrice and averaged for analysis. The IVD heights from the injured IVDs were also averaged for analysis and normalized by dividing by the presurgical heights of each IVD. Interrater and intrarater reliability tests were used to determine the consistency and repeatability for measuring IVD height using this technique. A total of 30 IVDs were randomly picked and measured by three researchers on two separate days with 7 days apart. The interrater reliability was excellent with intraclass correlation coefficient, ICC(3,3), of 0.983; and the interrater reliability was also excellent with ICC(3,3) for the three researchers of 0.992, 0.993, and 0.977 for the three researchers, respectively.

| IVD morphology and semi-quantitative degeneration grading
Mid-sagittal sections were stained with safranin-O/light green/hematoxylin to assess IVD morphology and evaluated using bright-field microscopy. The severity of IVD degeneration of each section was assessed using a semi-quantitative IVD degeneration grading scale adapted from the grading scales proposed by Masuda et al 54

and
Rutges et al. 55 The combined degeneration scale consisted of five categories, including (1) AF, (2) border between AF and NP, (3) cellularity of NP, (4) matrix of NP, and (5) endplate (Table 1). Each category received a score of 0-2 with score 0 for normal morphology and score 2 for characteristic severe degeneration; therefore, the overall degenerative score for each section was between 0 and 10. All sections were assessed by two researchers blinded to the experimental groups.
Intra-and inter-rater reliability scores were obtained for error analysis

| Intradiscal pro-inflammatory cytokine expressions
Expression of intradiscal pro-inflammatory cytokines TNFα and IL-1β was determined using immunohistochemistry. Mid-sagittal IVD sections were selected from each rat. Sections were rehydrated, treated with protein block reagent, and incubated overnight with either rabbit polyclonal primary antibody against rat TNFα (NB600-587; Novus, Littleton, Colorado), rabbit polyclonal primary antibody against rat IL-1β (bs-6319R; Bioss, Woburn, Massachusetts), or normal rabbit serum (Biocare Medical, Concord, California) as negative control. Sections were incubated with a horseradish peroxidase-conjugated anti-rabbit secondary antibody (MP-7451, ImmPRESS VR Reagent; Vector Laboratories, Burlingame, California) for 30 min and 3% hydrogen peroxide for 10 min, then treated with diaminobenzidine-based horseradishperoxidase substrate (ImmPACT DAB; Vector Laboratories) to visualize the immunoreactivity. Sections were counterstained with toluidine blue, dehydrated, mounted, and evaluated using bright-field microscopy.
Nine area-of-interest regions were identified from each IVD section, including outer anterior AF (three), inner anterior AF (two), NP (two), and posterior AF (two). The percentage of immunopositive cells from each area-of-interest region was determined via manual cell counting and then averaged for analysis.

| Statistical analysis
The results of hindpaw mechanical hyperalgesia and IVD height

| Surgery did not affect rat general health
Both sham surgery and spinal injury procedures were well-tolerated by the rats. The rat mean AESD body weight of 495 AE48 g, 504 AE52 g, 522 AE51 g, 533 AE52 g, 544 AE53 g, 555 AE53 g, and 565 AE55 g, for presurgery and postsurgery weeks 1, 2, 3, 4, 5, and 6, respectively. There were no significant differences between groups at any time point. No intraoperative complications or obvious stress or discomfort were observed from the general physical examination.

| TNFα inhibition at time of IVD injury prevents development of acute and long-term painful behavior
The continuous changes of normalized hindpaw withdrawal threshold within each group, as well as the comparison of withdrawal threshold  54,55 across groups at acute (ie, 1-week postsurgery) and long-term (ie, The IVD heights obtained at 6-week postsurgery were normalized to presurgery baseline. The normalized heights of all injured IVDs (including PBS, TNFα, and anti-TNFα) were significantly smaller than those of naive and sham IVDs ( Figure 3B). There was no significant difference between naive and sham groups, or among the three injury groups ( Figure 3C).

| Painful behavior predicted by IVD degeneration using stepwise linear regression analysis
Multivariate stepwise linear regression analysis showed that the nor- TNFα and IL-1β immunopositivity with R 2 of 0.453 ( Figure 5F).

| Pain threshold, structural degeneration and intradiscal pro-inflammatory cytokines correlate with each other
Pearson's correlation analysis showed that the normalized paw withdrawal threshold, IVD degeneration grade, normalized IVD height, as well as intradiscal TNFα and IL-1β immunopositivity were significantly correlated with each other (P < .05), except for the association between IVD degeneration and intradiscal IL-1β expression, which showed a trend of correlation with R 2 of 0.351 (P = .053) ( Table 3).

| DISCUSSION
This study identified multifactorial causes of painful IVD degeneration and determined a causal role for TNFα on the initiation of    a Relationship of paw withdrawal threshold to structural IVD degeneration, IVD height, and intradiscal pro-inflammatory cytokines using multivariate stepwise linear regression. In this in vivo rat study, the pain behavior was significantly associated with IVD degeneration and intradiscal TNFα expression as predictive factors using multivariate stepwise linear regression, and univariate correlation analysis also showed that the pain was significantly associated with loss of IVD height and intradiscal IL-1β expression. These results are consistent with another study that used an annular puncture model with NP removal that found a correlation between IVD degeneration and mechanical hyperalgesia, which was assayed directly on the low back. 65 The multiple associations of pain with IVD structural and biochemical degenerative changes in this in vivo rat model have many similarities with painful IVD degeneration in humans, which supports the feasibility of using this rat model to understand the pathophysiology of discogenic pain. 4 76 TNFα has also been shown to induce pain on the behavioral level. TNFα injection directly into rat hindpaws resulted in increased hyperalgesia demonstrating a role of TNFα in the initiation of inflammatory pain, and this painful behavior was inhibited by local TNFα inhibition, demonstrating the potential of local TNFα inhibition to alleviate the inflammatory pain in the periphery. 82 In an investigation of the inflammatory response and effect of anti-TNFα on pain in a surgically induced IVD herniation model with autologous IVD autograft and spinal nerve ligation, local application of anti-TNFα at the time of herniation surgery significantly improved paw withdrawal threshold in a rat model. 37 These finding further support the role of TNFα in peripheral inflammatory pain and the concept of locally inhibiting TNFα at early stages after injury to prevent the development of long-term pain. However, it should be noted that the pain pathophysiology is different between the two models. The pain pathophysiology is much clearer in IVD herniation models with inflamed IVD tissue directly impinging a nerve but more difficult to identify in discogenic pain models, which highlights the importance of the validation in the current model. TNFα upregulates NGF in IVD cells, and nociceptive nerves, usually restricted to the outer AF in normal healthy IVDs, can be stimulated with NGF to grow into the inner AF and NP as was found in the painful human IVDs. 23,41,83,84 In this model, anti-TNFα may have worked in several different ways: mitigating intradiscal inflammation, thereby preventing innervated nerve stimulation, decreasing the inflammatory environment of adjacent nerve roots, or inhibiting neovascular innervation. Likely, the mitigation of pain behavior following annular injury is a combination of the aforementioned factors, and the specific mechanism requires future investigations. However, histological assessments with this model did not identify obvious nerves or vessels, suggesting neovascularization is not likely to be a dominant cause of observed pain.
The current study focused on the changes in painful behavior and the structural and biochemical changes in the IVD in response to IVD injury and was limited because we did not investigate changes in the central nervous system, which plays important roles in both acute and chronic pain and warrants future investigation. Previous work has shown TNFα upregulates NGF in NP and AF cells in vitro, 41  NGF to pain and the broader inflammatory changes after TNFα injection or inhibition is an interesting area of future investigation. Additionally, only male rats were used in this study to minimize variability in behavioral and biochemical measurements, and it should be noted that there is evidence that pain pathways differ between sexes. 86,87 While it is not expected that the conclusions of the current findings would differ across sexes, male and female rats likely have sex differences in the neural pathophysiology of painful IVD degeneration that warrant further investigation. Lastly, rodent IVDs contain different cell types to those during human degeneration, and thus the responses may vary. Nonetheless, rodents are widely used to study IVD pathologies and treatment strategies. 88 In conclusion, the results demonstrate that intradiscal TNFα plays a critical role initiating an inflammatory cascade that results in painful IVD degeneration and suggest that inhibiting TNFα at early stages after IVD injury may offer potential benefit for preventing pain symptoms and structural IVD degeneration. The pain behavior following IVD injury was significantly associated with IVD degeneration and intradiscal pro-inflammatory cytokine expressions, which supports that this in vivo rat model has a phenotype similar to the human condition of painful IVD degeneration and could be a useful tool to study the underlying pathophysiology of painful IVD degeneration and for screening therapeutic strategies for discogenic pain.