Background

Low back pain brings the patient to the spine surgeon, but the relationship between disc degeneration and pain production in the disc is still poorly understood. Patients with chronic low back pain do not have the leg pain which results when fragments of herniated disc push on nerves; instead, these patients experience pain that is thought to arise from the disc itself. Although some back pain may be related to lumbosacral anatomy/function spinal anatomy [1], age, and other factors (see Fairbank et al. for a recent systematic review of low back pain classification [2]), discogenic low back pain is not well understood [3–8]. As with all pain, the pain-initiating event results from complex cellular, molecular and functional events at the nociceptors (naked nerve endings) [9, 10]. Discogenic pain is believed to result from disc changes (possibly from outer annulus pressure on nerve endings, disc cell dysfunction, products of matrix degradation [11], or unknown events) which influence the nervous system by stimulation of annulus nociceptors.

Recent studies focused over the last decade on molecular events and gene expression patterns related to pain [12], and studies have revealed that genome-wide gene expression studies provide a powerful methodology for identification of clinically relevant genes [13]. The objective of the present study was to perform a genome-wide analysis of annulus tissue from patients with discogenic back pain, compared to disc tissue from control subjects and herniated disc patients, in an analysis of the expression of pain, nerve and neurotrophin-related genes. Ontology searches for these specific topics were utilized in order to avoid searching large gene array data bases gene by gene, and because this technique provides a controlled vocabulary of search terms for gene characteristics [14].

As cell-based therapies for disc degeneration progress, information on pain-, nerve- and neurotrophin-gene expression in disc tissue becomes increasingly important. Findings presented here have potential applications in future treatment modalities, such as perispinal administration of TNF-α inhibitors (such as etanercept/enbrel [15]) or use of specific small molecular antagonists to neurotrophins. New information presented here on the relationship between proinflammatory mediators, nerves and neurotrophins has the potential to contribute to future important antagonist profiles with application to discogenic back pain.

Results

Demographic features of the patient population are summarized in Table 1 which presents spinal site, grade, subject age, and whether surgical specimens were derived from herniated or non-herniated tissue. In the present work, gene expression analyses included three Thompson grade II and five grade III control disc specimens (obtained from Cooperative Human Tissue Network). Surgical specimens were analyzed from three grade II discs (two specimens of which were from herniated discs), four grade III discs (from herniated discs), five grade IV specimens (two of which were from herniated discs), and three grade V specimens (two of which were from herniated discs).

Table 1 Demographic data on disc tissues*
Full size table

The ontology gene expression analysis for pain and nerve categories included a variety of biological processes, molecular functions, and cellular components. To assist the reader in understanding these, Table 2 outlines the categories included in our searches.

Table 2 Outline of ontology search strategies
Full size table

In each of our summary tables on gene expression findings (Tables  3, 4 and 5), genes are listed which have relevance to nerves, pain and neurotrophins (listed in section A of Tables  3, 4 and 5), genes with relevance to proinflammatory cytokines and chemokines (listed in section B of Tables  3, 4 and 5) and genes with specificity not only to nerve, pain and neurotrophins but also to the disc itself (listed in section C of the tables).

Table 3 Significant differences in pain-, neurotrophin-, nerve-, and disc-related genes in more degenerated discs (Grades iv and v) compared to less degenerated discs (Grades i – iii)
Full size table
Table 4 Significant differences in pain-, neurotrophin-, nerve- and disc related gene expression in surgical compared to control (CHTN) specimens
Full size table
Table 5 Significant differences in pain-, neurotrophin-, nerve-related and disc genes in herniated discs compared to control (CHTN) discs
Full size table

Neurotrophin-, nerve-, and pain-related gene, and disc gene expression patterns in the annulus: expression patterns in more degenerate compared to less degenerate discs

In Table 3, findings are presented for selected relevant genes which were significantly elevated in expression in more degenerated Thompson grade IV and V discs compared to findings in grades I, II and III discs.

Of major interest in Table 3 are a number of genes with high relevance to pain, neurotrophins and nerves. These include significantly upregulated expression in the more degenerated discs of the following: bradykinin receptor B1, calcitonin gene-related peptide, catechol-O-methyltransferase, neuron navigator-1 and −2, neuropilin 2, and reticulon 4 (also known as neurite growth inhibitor). These genes showed up regulation fold changes ranging from 1.09 to 4.61.

A large number of genes in the ion transport grouping showed significant changes in this analysis; 69 genes showed significant up regulation, and 59 significant down regulation (most data not shown).

Genes with specific high relationships to disc cell biology included these significantly upregulated genes in the more degenerated discs: apoptosis-inducing factor (mitochondrion associated), BMP receptor, type II, collagens type I, III, IV, V and VI, connective tissue growth factor, fibronectin, hypoxia inducible factor 1 (HIF1), several of the mitogen-activated protein kinases, SPARC (osteonectin), TGF-ß, lumican and TIMP metallopeptidase inhibitors-2 and −3. These genes showed upregulation fold changes ranging from 1.42 to 17.58. Nitric oxide synthase 3 showed a 4.77 fold downregulation.

Neurotrophin-, nerve-, and pain-related gene, and disc gene expression patterns in the annulus: analysis of expression patterns in surgical compared to control (CHTN) specimens

In Table 4, findings are presented for relevant genes which were significantly upregulated in surgically operated disc specimens compared to expression findings in control (CHTN) discs.

Of high interest in Table 4 are a number of genes with high relevance to pain, neurotrophins and nerves. These include the following: calcitonin gene-related peptide, catechol-O-methyltransferase, ciliary neurotrophic factor, nerve growth factor, neuron navigator 1, neuropilin 2, reticulon 4, and roundabout axon guidance receptor. These genes showed up regulation fold changes ranging from 1.2 to 5.96.

A large number of genes in the ion transport grouping also showed significant changes in this analysis; 87 genes showed significant up regulation, and 73 significant down regulation (most of the gene data in the ion transport group are not shown here).

Genes with specific high relationships to disc cell biology included these significantly upregulated genes in surgical vs. control discs: heat shock proteins, fibronectin 1, versican, lumican, several of the mitogen-activated protein kinases, TIMP metallopeptidase inhibitors-3 and −4, TFG-ß, latent TGF binding protein 1, several of the TNF receptors, and collagen type II alpha 1. These genes showed up regulation fold changes ranging from 1.1 to 66.35. Notable down regulated genes included brevican and FGF 5 (with fold changes of 1.3 and 1.4, respectively).

Neurotrophin-, nerve-, and pain-related gene, and disc gene expression patterns in the annulus: expression patterns in herniated compared to Non-herniated discs

Although not related to discogenic low back pain, we were also interested in evaluating our data in terms of expression patterns which were significantly different in herniated discs vs. non-herniated discs (Table 5).

Of high interest in Table 5 are a number of genes with high relevance to pain, neurotrophins and nerves. These include the following: Calcitonin gene related peptide (down regulated 1.2 fold). Upregulated genes included neuropilin 2, nerve growth factor, reticulon 4, roundabout axon guidance receptor; these genes were upregulated 1.5 to 5.58 fold.

A large number of genes in the ion transport grouping also showed significant changes in this analysis; 98 genes showed significant up regulation, and 39 significant down regulation (most data not shown).

Genes with specific high relationships to disc cell biology included these significantly upregulated genes: a number of the collagens, fibronectin 1, hypoxia inducible factors-1 and −2 (alpha subunit), latent TGF-ß binding protein 1, TGF-ß receptor 1, several of the mitogen-activated protein kinases, proteoglycan 4, SOD, and the apoptosis-associated genes caspase 6 and TNFRSF1A-associated via death domain.

Immunohistochemical studies

Paraffin-embedded annulus tissue was available for several of the subjects studied here (subjects # 5, 10, 14, 15, 19, 21 and 22 (Table 1) which enabled us to perform immunolocalization studies for products of three of the genes of special interest here (calcitonin gene-related peptide, catechol-0-methyltransferase and bradykinin receptor B1). For each of these immunolocalizations, cells were present with localization in single cells, clusters of cells, and both rounded and spindle-shaped cells in the outermost region of the annulus. Representative images are shown in Figures 1A-C; Figure 1D presents a negative control with the absence of any localization. Note that in Figures 1A-C adjacent cells were occasionally present which showed no localization.

Figure 1
figure 1

Immunohistochemical localizations of calcitonin gene-related peptide, catechol-0-methyltransferase and bradykinin receptor B1: Representative images showing localization of calcitonin gene-related peptide (Figure 1A), catechol-0-methyltransferase (Figure 1B) and bradykinin receptor B1 (Figure 1C) in annulus regions of the human disc. Figure   1D illustrates a negative control. Arrows mark nearby cells which did not show localization.

Full size image

Discussion

In the present study we performed a genome-wide microarray analysis of human annulus tissue from patients with discogenic back pain compared to disc tissue from control subjects or compared to and herniated disc patients we had a special focus upon analysis of the expression of pain, nerve and neurotrophin-related genes. Ontology searches were an efficient search technique for identification of pain, and nerve-related genes [14]. Although there is a large body of clinical literature on low back pain, molecular studies are few, and those in the literature primarily focus upon population-based genetic studies of polymorphisms (SNPs) (see [13] for a review of pain and spinal disease). To the best of our knowledge, the present work is the first (non-SNP) genome-wide study of pain, neurotrophin and nerve-related genes in disc degeneration.

Pain-related genes

Several well-recognized pain-related genes were found to have significant elevations in our analyses. Bradykinin receptor B1, calcitonin gene-related peptide and catechol-0-methyltransferase were significantly elevated in more degenerated discs (grade IV and V) compared to less degenerated (grades I-III) discs (Table 1). Bradykinin receptor B1 is formed after tissue injury and mediates hyperalgesia in chronic inflammation, but has very low expression in healthy tissue expression [16]. Calcitonin gene-related peptide (CGRP) has been found to be elevated in sensory nerves innervating inflamed tissue [17], in dorsal root ganglia and spinal cord in sciatic nerve injuries in the rat model by Orita et al. [18]. In the latter work, application or antibodies to nerve growth factor or its receptors TrkA or p75NTR blocked CGRP expression. Catechol-0-methyltransferase (COMT) codes for a protein which is important in catabolic pathways of a number of pain-relevant neurotransmitters, including noradrenalin, adrenaline and dopamine [19]. In the present analyses comparing surgically operated discs to control discs (Table 4) and herniated to control discs, calcitonin gene-related peptide was the only one of these genes which was significant, and in these comparisons it was downregulated 1.2-fold. The COMT gene is very interesting since patients with a specific polymorphism identified by Zubieta et al. showed higher sensory and affective pain ratings [20].

Our ability to perform immunolocalization studies on human annulus tissues to determine the presence of bradykinin receptor B1, calcitonin gene-related peptide and catechol-0-methyltransferase (Figure 1) added strength to the present study, and confirmed the presence of products of these gene at the protein level within human disc tissue.

Neurotrophin-related genes

Neurotrophins were also identified in the present analyses, including nerve growth factor in our studies of surgical vs. control discs, and herniated vs. control discs (significantly upregulated, Tables  4 and 5, respectively), and ciliary neurotrophic factor (down regulated, Table 4, surgical vs. control discs).

Studies have recently shown production of several neurotrophins by disc cells. Gigante et al. reported the presence of nerve growth factor (NGF) mRNA and the high affinity tyrosine kinase A receptor (trkA) and the low affinity p75 receptor in the rounded cells in the disc annulus and nucleus pulposus [21]. Recently Abe et al. reported on the expression of nerve growth factor (NGF) by human disc cells in control disc tissue in vivo and in vitro in cells from control discs using immunocytochemistry [22]. Nerve growth factor was found to be high in the outer annulus and herniated disc tissue. That work also demonstrated that the proinflammatory cytokines IL-1 and TNF-alpha had stimulatory effects on NGF. These authors suggested that such actions may play a role in nerve sprouting into the disc and may be associated with discogenic pain. Recent in vitro work from our lab has also confirmed that exposure of disc cells to IL-1ß in three-dimensional culture (which more accurately mimics the in vivo condition) results in elevated production of nerve growth factor by human annulus cells [23].

Ciliary neurotrophic factor whose expression is reported here, has not previously been known to be expressed in the human disc. Work from previous studies has shown that it can act as both a neuroprotective agent [24] and a trophic factor for motor neurons [25].

Nerve-related genes

Several nerve-related genes should be mentioned in our analyses, including neuron navigator-1 and −2, neuropilin 2, reticulon 4 (neurite growth inhibitor), roundabout axon guidance receptor, homolog 3, and neural precursor cell expressed (developmentally down regulated 4-like) (Tables  3, 4 and 5). These genes also have not previously been known to be expressed in the human disc.

Neuron navigator 1 is a microtubule-associated protein involved in neuronal migration [26], and neuron navigator 2 is required for all-trans retinoic acid-mediated neurite outgrowth and axonal elongation [27]. Neuropilin 2 was significantly upregulated in all of our disc comparisons (Tables  3 4 and 5). Neuropilin 2 is another gene which we have found to be significantly upregulated in cultured annulus cells exposed to IL-1ß [23]. Neuropilin-1 and neuropilin-2 are membrane proteins implicated in aspects of neurodevelopment. They are semaphorin III receptors as shown by the work of Kolodkin et al. [28], and are expressed in overlapping populations of neurons in the embryonic nervous system. Semaphorin III is a protein which, when secreted in vitro, results in the collapse of neuronal growth cones and chemorepulsion of neurites. It is also needed for correct sensory afferent innervation [29]. Expression of this gene by annulus cells may provide evidence that annulus cells express this gene to block neurite ingrowth into the disc.

Roundabout (ROBO1) is a gene which encodes a receptor which is a member of the neural cell adhesion molecule family. It functions as an axon guidance receptor [30]. We found upregulation of this gene in surgical vs. control specimens (Table 4) and in herniated vs. control specimens (Table 5).

Reticulon 4 (neurite growth inhibitor) was identified with upregulation in each of our analyses, with 4.6 fold (Table 3, more degenerated discs vs. less degenerated discs), 5.96 fold (Table 4, surgical vs. control discs) and 5.58 fold (Table 5, herniated vs. control discs) changes. This gene, also called NOGO, is a neurite outgrowth inhibitor [31].

Genes related to proinflammatory cytokines and chemokines

In Tables  3 4 and 5, in the section headed “Genes with special disc relevance” we have listed many proinflammatory cytokines and chemokines identified in our analyses. These deserve special mention here because a large number of proinflammatory cytokines are well recognized as products of disc cells themselves in vivo (see [32–36] for an introduction to this field). Many chemokines are also produced by disc cells (our unpublished data). It is very important to note in the present study that many proinflammatory cytokines and chemokines are now known to induce or exacerbate inflammatory and neuropathic pain and hyperalgesia [37–39]. We suggest here that this is an exceptionally important aspect of low back pain that has here-to-fore been little recognized. The degenerating human disc is primarily an avascular tissue site into which disc cells have contributed high levels of proinflammatory cytokines which are not cleared from the tissue and remain there over time. We suggest that as nerves grow into the human annulus, they encounter a proinflammatory cytokine-rich milieu which exacerbates pain production.

Kim et al. using in vitro work has suggested that disc cells themselves are involved in inflammatory activities, and suggested that interactions between annulus cells and nerve cells enhances the production of growth factors responsible for neovascularization and nerve ingrowth into the disc [40]. Previous research by Aoki et al. showed that disc inflammation potentially promoted axonal regeneration of dorsal root ganglion neurons innervating the disc in a rat model [41]. Using gene correlations, recent work by Lee et al. suggests that IL-1ß is generated during degeneration of the disc, and this stimulates expression of agents such as nerve growth factor, which result in nerve in growth into the disc [42].

It is important to recognize that proinflammatory cytokine production within the degenerating disc can also be exacerbated by repeated disc injury, which may lead to persistent proinflammatory cytokine elevation [43]; thus repeated disc injury may also influence neuroinflammation and pain.

A final important comment from the joint literature on proinflammatory mediators concerns the fact that release of these agents in damaged tissue and in the spinal cord is known to sensitize the peripheral terminals of nociceptors [44]. It is possible that similar hyper excitability of pain transmitting neurons results from proinflammatory cytokines in the disc matrix during degeneration; proinflammatory cytokines likely to be at play here are IL-1ß and TNF-α.

Genes which share high importance to the disc itself

It was interesting that many genes were identified in our analyses which have relevance to the nerve, neurotrophin and pain ontology and to disc biology itself (as shown in section C of Tables  3 4 and 5). These included extracellular matrix (ECM) components, such as collagens, fibronectin, laminin, thrombospondin, brevican, proteoglycans, and versican, and also genes related to matrix degradation (metallopepdiases and TIMP metallopeptidase inhibitors), growth factors (connective tissue growth factor), nitric oxide synthase 3 (ENOS), SOD, and hypoxia inducible factor 1 alpha subunit. Also important to disc biology were the vitamin D receptor gene, growth arrest specific genes (important to cell senescence), apoptosis-related genes, and BMP receptor type II. Although some of these gene products may be resident only in sites of neurovascular ingrowth, many may influence the disc ECM itself and disc cell functions. Such findings point to the importance of future research directed towards identifying functional interactions between disc and nerve cells in vivo and in vitro. For example, it has long been recognized that there is an accumulation of fibronectin fragments in the aging/degenerating disc, and these fragments initiate signaling pathways which can increase MMP expression causing a cycle of matrix destruction [45–50]. Strong up regulation of fibronectin was present in our analyses (16.8 fold upregulation in more degenerated vs. healthier discs (Table 3), 37.0 fold upregulation in surgical vs. control discs (Table 4), and a 52.4 fold up regulation in herniated vs. control discs (Table 5).

Possible limitations to the present analyses include the fact that some specimens, noted as “controls” in Table 1, were obtained from the Cooperative Human Tissue Network (CHTN). Although these were shipped quickly to our lab post-autopsy as quickly as possible via CHTN, delays might result in potential alterations in mRNA levels during our study. The reader should note that in order to carefully investigate this issue, in Table 4 we compared findings in surgical specimens vs. those obtained from CHTN. As would be expected in different sized microarray group comparisons with ontology searchers, Table 4 does differ in some respects from data presented in Table 3. It is also worth noting that in our laboratory, CHTN specimens are also routinely used for derivation of disc cells in vitro, and in our hands over 98% of CHTN specimens yielded viable cells.

Another point for the reader to note is that our tables have reported gene expression fold changes which in some cases were less than 2.0, a level which is commonly used. Since we feel that smaller changes in important genes may be clinically relevant, we have reported these changes in our data tables.

Conclusions

Even though our three study groups were not large, the present analysis showed that microarray analysis could successfully be used to examine key pain-, neurotrophin- and nerve-related genes in specimens of human disc tissue. Many genes were found in these ontology searches which held significance not only for nerves, pain and neurotrophins, but also for disc ECM, signaling and functional components. Key findings included confirmation of the presence of calcitonin gene-related peptide, catechol-0-methyltransferase and bradykinin receptor B1 at the protein level in the human annulus using immunohistochemistry, and identification of significant changes in a number of proinflammatory and chemokine genes identified from nerve, neurotrophin and pain ontology searches. Since the disc is primarily avascular, and since disc cells themselves produce proinflammatory cytokines and chemokines which are not removed from the tissue, we hypothesize that as nerves grow into the human annulus, they encounter a proinflammatory cytokine-rich milieu which may sensitize nociceptors and exacerbate pain production. Findings reported here point to the importance of future studies of the functional interactions between disc and nerve cells in vivo and in vitro.

Methods

Clinical study population

Experimental study of human disc specimens was approved prospectively by the authors’ Human Subjects Institutional Review Board. The need for informed consent was waived by the ethical board since disc tissue was removed as part of routine surgical practice. Scoring of disc degeneration utilized a modification of the Thompson scoring system [51] incorporating author ENH’s radiologic, MRI and surgical findings. The Thompson system scores disc degeneration over the spectrum from a healthy disc (Thompson grade I) to discs with advanced degeneration (grade V, the most advanced stage of degeneration) [51]. Patient specimens were derived from surgical disc procedures performed on individuals with herniated discs and degenerative disc disease. Surgical specimens were transported to the laboratory in sterile tissue culture medium. Non-surgical, control donor disc specimens were obtained via the National Cancer Institute Cooperative Human Tissue Network (CHTN); they were shipped overnight to the laboratory in sterile tissue culture medium and processed as described below. Specimen procurement from the CHTN was included in our approved protocol by our human subjects Institutional Review Board.

Microarray analysis

Disc tissue was snap frozen in liquid nitrogen, pulverized (BioPulverizer, BioSpec Products, Inc., Bartlesville, OK, USA), and homogenized via the FastPrep-24 instrument (MP Biomedicals L.L.C., Santa Ana, CA, USA). Total RNA was isolated via a modified version of TRIzol Reagent (Life Technologies: Invitrogen, Carlsbad, CA, USA), and prepared for microarray hybridization using the GeneChip 3’ IVT Express Kit (Affymetrix, Santa Clara, CA). In brief, total RNA was reverse transcribed to synthesize cDNA, converted to double stranded DNA, subjected to transcription generating biotin-labeled amplified RNA (cRNA) and hybridized to the DNA microarray in the Affymetrix Fluidics Station 400. Affymetrix human U133 X3P arrays were used. The GCOS Affymetrix GeneChip Operating System (version 1.2, Affymetrix, Santa Clara, CA 95051) was used for determining gene expression levels. mRNA from annulus tissue from each subject was analyzed separately (i.e., samples were not pooled).

Statistical analysis

The GCOS Affymetrix GeneChip Operating System (version 1.2, Affymetrix, Santa Clara, CA) was used for determining gene expression levels. GeneSifterTM web-based software (VizX Labs, Seattle, WA, USA) was used to analyze all microarray data. Using GC-RMA (Robust multi-array average), Affymetrix ‘.cel’ files were uploaded to the GeneSifterTM web site and normalized, and corrected for false discovery rate (FDR). Using the student t-test (2 tailed, unpaired), statistical significance was determined (p < 0.05).

Gene Ontology (GO) searches were employed in our analyses to select genes of interest and groups of critically important genes. This approach lets one avoid searching through results gene by gene, and provides a controlled vocabulary of search terms for gene characteristics. In our analyses, Gene Ontologies (GO) were generated by GeneSifterTM based on the Gene Ontology Consortium. Searches were performed in the present study on “pain” and “nerve”; for each, ontologies were searched under “biological process”, “molecular function” (the activities of the gene product at the molecular level), and “cellular component” (parts or cells or the extracellular milieu). To aid the reader in visualizing the key terms covered in these ontology grouping, details are provided in Table 2.

Gene array data for the human disc specimens analyzed here have been uploaded to the Gene Expression Omnibus (GEO) website [GEO:GSE23130] and may be accessed there.

Immunohistochemistry

Bradykinin receptor B1 (BDKRB1) and calcitonin gene related peptide (CGRP) Immunohistochemistry: Disc specimens were fixed in 10% neutral buffered formalin, embedded undecalcified, paraffin sections cut at 4 μm, collected on PLUS slides(Cardinal Health, Dublin, OH) and dried at 60°C. Sections were deparaffinized in xylene (Cardinal) and rehydrated through graded alcohols (AAPER, Shelbyville, KY) to distilled water. Antigen retrieval was performed using Biocare Antigen Decloaker Solution, pH 6.0 (Biocare Medical, Concord, CA) for 20 minutes at 95°C followed by cooling for 20 minutes. The remainder of the procedure was performed using the Dako Autostainer Plus (Dako, Carpenteria, CA) automated stainer. Endogenous peroxidase was blocked using 3% H202 (Sigma, St Louis, MO). Slides were incubated for 30 minutes with Serum-Free Protein Block (Dako); blocking solution was drained from slides and primary antibody applied. Sections were incubated for one hour with anti-Bradykinin receptor B1 (BDKRB1) (Novus Biologicals, Littleton, CO) at a 1:50 dilution, or with for one hour with anti-calcitonin gene related peptide (CGRP) (Abcam, Cambridge, MA) at a 1:100 dilution. Secondary antibody was 4 + Biotinylated Universal Goat Link (Biocare) for 10 minutes followed by 4+ streptavidin HRP Label (Biocare) for 10 minutes and DAB (Dako) for 5 minutes. Slides were removed from stainer, rinsed in water, counterstained with light green, dehydrated, cleared and mounted with resinous mounting media. Universal Rabbit Negative (Dako, Carpinteria, CA) was used as a negative control.

Catechol-O-methyltransferase (COMT) immunohistochemistry did not require antigen retrieval. Sections were prepared as described above, and incubated for one hour with anti-catechol-O-methyltransferase (COMT) (Lifespan Biosciences, Seattle, WA) at a 1:200 dilution. The secondary antibody and negative control utilized were as described above.

Positive control human tissues were also included with each immunolocalization run; for bradykinin receptor B1 this was brain, for calcitonin gene related peptide this was thyroid, and for catechol-O-methyltransferase, adrenal.