Background

The dorsal horn of the spinal cord plays important roles in sensory information processing. The dorsal horn contains the neural circuitry conveying nociceptive information, including pain and temperature, from the periphery by the primary afferents [1–3]. Nociceptive afferent fibers terminate predominately in the dorsal horn of the spinal cord. Activation of the nociceptors transmits afferent messages to the spinal cord dorsal horn through neurotransmitters such as glutamate. Initial processing of nociceptive information occurs in the spinal cord dorsal horn by excitatory and inhibitory interneurons. The projecting neurons, spinalthalamic tract cells, then convey primary nociceptive information to higher centers, signaling localization and encoding the character of the nociceptive input. Other inputs related to the subjective components of pain and related to motor and autonomic control are also relayed to higher centers. The projection neurons of the dorsal horn also activate the descending control system, which in turn controls the gain of dorsal horn neurons either through excitatory or inhibitory mechanisms. In this manner, the initial nociceptive information may be further modulated by signals descending from higher centers [1–3].

The dorsal horn can be subdivided into six distinct laminae on the basis of the cytological features of its resident neurons [2, 4]. Classes of primary afferent neurons that convey distinct modalities terminate in distinct laminae of the dorsal horn. Thus, there is a close correspondence between the functional and anatomical organization of the neurons in the dorsal horn of the spinal cord. Nociceptive neurons are mostly located in the superficial dorsal horn, in the marginal layer (lamina I) and in the substantia gelatinosa (lamina II), and receive direct synaptic input from Aδ and C fibers. Laminae III and IV are located ventral to the substantia gelatinosa and contain neurons that receive monosynaptic input from Aβ fibers. Lamina V primarily contains wide-dynamic-range neurons that project to the brain stem and to regions of the thalamus. These neurons receive monosynaptic input from Aδ and Aβ fibers. They also receive input from C-fibers, either directly on their dendrites, or indirectly via interneurons. Many neurons in lamina V also receive nociceptive input from visceral structures. Neurons in lamina VI receive input from large-diameter afferents from muscles and joints and respond to nonnoxious manipulations of joints. These neurons are thought not to contribute to the transmission of nociceptive messages.

Consistent with the important role of the dorsal horn of the spinal cord in pain transmission and modulation, neurochemical studies have implicated an enriched expression of neurotransmitters, neuropeptides, ion channels, and receptors in these neurons, including substance P, enkephalin, CGRP, somatostatin, and GABA. The morphology of primary afferent central terminals, dorsal horn neurons, and descending systems, together with their chemical neuroanatomy, synaptic arrangements, transmitter systems, and functional properties have been extensively documented [1–5]. However, the molecular/chemical neuroanatomy of the dorsal horn is a subject that is constantly being revised and updated as improved techniques reveal new insights into classical pathways or substances that are localized in the spinal cord. In this study, we performed a large-scale screening for genes that are enriched in the dorsal spinal cord. In addition to molecules that are known to be highly expressed in the dorsal spinal cord, we identified other neuropeptides, ion channels, and signaling molecules enriched in the dorsal spinal cord. We then further characterized the cellular localization of a subset of these genes in the spinal cord, as well as the regulation of a subset of the genes in a neuropathic pain model.

Results

Global identification of genes that are enriched in the dorsal spinal cord using DNA microarray analysis

In order to identify genes that are enriched in the dorsal spinal cord, we took advantage of DNA microarray technology which is powerful in identifying regional-specific gene expression. We screened globally for genes that are expressed at higher levels in the dorsal spinal cord as compared to the ventral spinal cord. A total of 9 rats were used in this study. Dorsal spinal cord and ventral spinal cord are separated by cutting through the central cannel. Dorsal spinal cord samples from 3 rats were pooled and ventral spinal cord samples from the same 3 rats were pooled to form each pair of samples. RNA was prepared from 3 such pairs of samples, each was analyzed by hybridization with a Affymetrix chip RG-U34A, which contains 8799 probesets. Using the SAFER analysis approach (see methods), we compared gene expression between the dorsal and ventral spinal cord samples using a paired t-test and analysis of variance (ANOVA). Using a paired t-test we found 31 probesets with p-value smaller than 0.05 and greater than 2-fold higher expression in the dorsal spinal cord (Table 1). An ANOVA analysis detected 30 of these 31 probesets along with 21 additional probesets that met these criteria. We found evidence in the literature for enriched expression in the dorsal spinal cord of 23 genes among the 51 genes detected by ANOVA (Table 1). In addition, we independently tested the expression of 7 genes by QRT-PCR, 2 genes by in situ hybridization, and 2 genes by immunohistochemistry. All but one of these genes were confirmed to be enriched in the dorsal spinal cord. Both paired t-test and ANOVA analysis detected genes that are enriched in the dorsal spinal cord as revealed by independent confirmation. As ANOVA analysis provides more candidate genes, we chose to subsequently present the results based on ANOVA analysis.

Table 1 Confirmation and comparison of ANOVA and paired t-test for microarray analysis.
Full size table

The 51 genes that are expressed higher in the dorsal spinal cord represent several functional categories including neuropeptides, ion channels, receptor and signaling molecules, calcium/calmodulin binding proteins, transcription factors, synaptic proteins, and others (Table 2). Neuropeptide genes are among the category that have the highest number of genes expressed specifically in the dorsal spinal cord. These include somatostatin, protachykinin, beta-tachykinin, neuropeptide Y (NPY), neurokinin B, and nociceptin. In addition, we also detected another peptide, C-type natriuretic peptide (CNP), which is expressed at 2-fold higher levels in the dorsal spinal cord. A number of ion channels were found to be expressed higher in the dorsal spinal cord, including, voltage-gated sodium channel type VI (SCAN6A), a putative four repeat ion channel, NMDA receptor NR1, and potassium channel Kv3.1. A few G-protein coupled receptors such as cannabinoid receptor, GABA-B R2 receptor, and olfactory receptor-like protein (SCRD-9), as well as molecules important in signaling such as Galpha(olf), Guanine nucleotide releasing factor, and CAM kinase I beta 2, CAM kinase II beta, and PKC beta were also identified to be expressed more in the dorsal spinal cord. We found that the dorsally enriched genes also include several members of the calcium sensor protein family, such as vesicle associated calcium calmodulin binding protein (CAMK IV homolog), PKCγ substrate neurogranin, chromogranin, neurovisinin-like protein, secretogranin, and hippocalcin. In addition, several transcription factors and synaptic proteins were also expressed higher in the dorsal spinal cord (Table 2).

Table 2 Genes that are enriched in the dorsal spinal cord based on microarray analysis.
Full size table

QRT-PCR confirmation of genes enriched in the dorsal spinal cord

We sought to confirm the differential gene expression of a selected subset of genes in the spinal cord along the dorsal-ventral axis using QRT-PCR. The genes selected showed expression differences between 2 to 7-fold by DNA microarray. We compared gene expression between dorsal and ventral spinal cord samples. And the results of these comparisons are depicted in Fig. 1. In general, higher expression was confirmed for 6 out of the 7 genes tested in the dorsal spinal cord, despite some degree of variation among the 3 pairs of samples from different animals (Fig. 1).

Figure 1
figure 1

Differential expression of genes in the dorsal versus ventral spinal cord as revealed by QRT-PCR. The ratio of expression between dorsal and ventral spinal cord for each gene in each dorsal and ventral spinal cord sample pair is shown. Data were from QRT-PCR assays performed in triplicate for a pair of RNA samples, each of which was pooled from the dorsal or ventral samples of the 3 pairs of RNA samples used for microarray analysis (from a total of 9 animals). The length of bars represent 95% confidence intervals and the symbol on each bar is the mean of ratios estimated from the assay performed in triplicate.

Full size image

We confirmed the enriched expression of the following genes: a transcription factor NGFI-A, a putative four repeat ion channel, a vesicle associated calcium/calmodulin binding protein that is homologous to CAMK IV, a neuronal calcium sensor protein hippocalcin, synuclein beta, and an EST gene. The Zinc finger protein mRNA was found to be only marginally enriched in the dorsal spinal cord (Fig. 1).

Cellular expression of dorsal spinal cord-enriched genes

Using in situ hybridization and immunohistochemistry, we sought to investigate whether differences in gene expression as detected by DNA microarray reflect differences at the cellular level. In situ hybridization revealed that PKCγ substrate neurogranin and vesicle-associated calcium/calmodulin binding protein (CAMK IV homolog) mRNA are both highly enriched in the superficial layers of the dorsal horn of the spinal cord (Fig. 2). Neurogranin expression seems to be enriched in subsets of the cells of lamina II and III (Fig. 2A), similar to PKCγ localization (Fig 3C). CAMK IV homolog seems to be broadly expressed in layers I-III of the dorsal horn (Fig. 2B). Immunohistochemistry with antibodies against Galpha(olf), syntaxin 1A, and PKCγ revealed that these proteins are also enriched in the dorsal spinal cord (Fig. 3). Galpha(olf) and Syntaxin 1A appear to be localized in the neuropil of the dorsal horn (Fig. 3A and 3B) whereas, PKCγ appears to be expressed in layer IIi of the dorsal horn (Fig. 3C) as previously described [6].

Figure 2
figure 2

mRNA localization of neurogranin and CAMK IV homolog. Neurogranin (A) and CAMK IV homolog (B) are located in the superficial dorsal horn and exhibit different patterns with the antisense riboprobes. Pictures were pieced together from 4 images which were taken using 4X objectives.

Full size image
Figure 3
figure 3

Cellular localization in the spinal cord as revealed by immunohistochemistry. Galpha(olf) (A), Syntaxin 1A (B), and PKCγ (C) are localized primarily in the dorsal spinal cord. Images were taken from sections of spinal cord from the same animal although similar patterns of staining were found when two animals were examined.

Full size image

Regulation of dorsally enriched genes in a chronic neuropathic pain model

We sought to study whether the genes that are enriched in the dorsal spinal cord are regulated during a persistent pain state such as the spinal nerve ligation (SNL) model of neuropathic pain. For a subset of genes enriched in the dorsal spinal cord, we used QRT-PCR to compare the gene expression in the lumbar spinal cord between spinal nerve ligation rats and rats subjected to sham surgery, and to compare expression between ipsilateral and contralateral sides (separated by cutting through the midline in the segment of lumbar spinal cord) in spinal nerve ligation rats. In general, we found that genes that are enriched in the dorsal spinal cord are not among those found to be up-regulated in the spinal nerve ligation model of neuropathic pain (Table 3).

Table 3 Relative gene expression in the spinal cord of SNL (Chung) neuropathic pain model as revealed by QRT-PCR.
Full size table

Discussion

The dorsal spinal cord is a region implicated in sensory perception, receiving, transmitting, and modulation of signals from peripheral sensory system. It is important in the integration of computational and neuromodulatory functions. In a pathological state, the changes in the dorsal horn of the spinal cord may contribute to prolonged abnormal pain. The aim of this study was to find genes that are enriched in the dorsal spinal cord that can potentially play important roles in pain transmission, pain modulation, and pathophysiological conditions. Since microarray technology represents a potentially powerful method for identifying cell type- and regionally restricted genes expressed in the nervous system [7, 8], we conducted a large-scale screening using this technique for genes that are expressed higher in the dorsal spinal cord. Genes found in this screen can then be further studied for their cellular localization in the spinal cord.

Validation of microarray data

The reliability of these DNA microarray results is demonstrated by the following three observations: (1) A subset of the genes was observed to be consistently regulated by multiple probesets on the microarray; (2) We detected 21 genes (by 23 probesets) which have previously been described to be enriched in the dorsal spinal cord; (3) Genes that we chose to study further were confirmed to be expressed higher in the dorsal spinal cord by QRT-PCR (6 out of 7), in situ hybridization (2 out of 2), and immunohistochemistry (2 out of 2). Based on the consistency and the rate of independent confirmation, many of the genes listed in the tables are likely to be true positives.

We note that our list of dorsal spinal cord-enriched genes is likely to be incomplete since the Affymetrx DNA microarrays do not represent the entire rat genome and our study may not have been sensitive enough to discover all the genes, especially those that are expressed at low levels.

Toward a molecular anatomy of the spinal cord, particularly in the pain sensory pathways

We found that some genes are expressed in a lamina-specific manner, while others may have a gradient of expression. Neurogranin is highly enriched in the superficial laminae. PKCγ has been demonstrated to be expressed in lamina IIi. The colocalization of neurogranin and PKCγ in similar regions in the spinal cord suggests that neurogranin may be an endogenous substrate of PKCγ in the spinal cord. Neurogranin has been shown to be phosphorylated by PKCγ, and this phosphorylation is greatly decreased in PKCγ knockout mice, suggesting neurogranin is a PKCγ-specific substrate [9]. Evidence that strongly suggests the phosphorylation of neurogranin plays an important role in neural plasticity comes from the studies that demonstrated neurogranin knockout mice show essentially the same deficits in behaviors related to learning and memory as that of the PKCγ knockout mice [10]. Neurogranin is not only a substrate of PKCγ, but also plays a important role in regulating PKC signaling [11]. Interstingly, PKCγ knockout mice demonstrate reduced neuropathic pain [12]. Similar to neurogranin, we found that a calmodulin-binding, vesicle-associated, CAM kinase IV-like protein is highly expressed and enriched in the superficial layer of the dorsal spinal cord. The function of this gene is not yet known. However, the protein was also found to be enriched in forebrain neurites [13].

Using immunostaining, we found that Syntaxin 1A and Galpha(olf) proteins were enriched in the dorsal spinal cord. Syntaxin 1A has previously been shown to be preferentially expressed in the dorsal spinal cord neuropil. Galpha(olf) is a G protein that was initially found to be expressed in the olfactory epithelium. The Galpha(olf) has been shown to be specifically expressed in the striatum in the brain and was found to colocalize with and activated by adenosine A2A receptors [14]. It is possible that the specific expression of Galpha(olf) in the dorsal spinal cord implies its mediation of functions for specific G-protein-coupled receptors in these neurons.

Conclusions

1. We conducted a large-scale screening using DNA microarray for genes that are specifically expressed in the dorsal spinal cord. We found additional neuropeptides, receptors, ion channels, and signaling molecules to be enriched in the dorsal spinal cord.

2. The regulation of a subset of the genes was confirmed by QRT-PCR. Six out of the seven genes were confirmed to be enriched in the dorsal spinal cord.

3. In situ hybridization and immunohistochemistry revealed that neurogranin, CAMK IV homolog, Galpha(olf), and Syntaxin 1A, are indeed enriched in the dorsal spinal cord.

4. Through the detection of a large number of genes which were previously determined to have enriched expression in the dorsal spinal cord and our QRT-PCR, in situ hybridization and immunohistochemical confirmations, this study provides a level of validation for the case of Affymetrix DNA microarrays in conjunction with SAFER algorithm to detect differences in gene regulation.

Materials and Methods

Animals

Male Sprague-Dawley rats (Taconic, Germantown, NY.) weighing 200–300 g at the time of testing, were maintained in a climate-controlled room on a 12 h light/dark cycle (lights on at 06:00) with food and water available ad libitum. All of the handling of the animals and testing was performed in accordance with the policies and recommendations of the International Association for the Study of Pain [15] and received approval from the Institutional Animal Care and Use Committee of MRL, West Point, PA.

Spinal nerve ligation (SNL) injury was induced using the procedure of Kim and Chung [16]. Anesthesia was induced with 2% gaseous isofluorane (For induction 3–5% and O2 500–700 μl, for maintenance 2–3% and O2 400–500 μl). Following dorsal skin incision and muscle separation, the posterior interarticular transverse process of L/S1 was exposed and carefully removed with a micro Rongeur. The L5 and L6 spinal nerves were tightly ligated by a square knot with 6–0 silk thread. The muscles were closed with 4–0 absorbable sutures and the skin was closed with wound clips. Rats that exhibited motor deficiency (such as paw dragging) or failure to exhibit subsequent tactile allodynia were excluded from further testing (less than 5% of the animals were excluded). Sham control rats underwent the same operation and handling as the experimental animals but without spinal nerve ligation.

Behavioral testing

The assessment of tactile allodynia (i.e. decreased threshold to paw withdrawal following probing with non-noxious mechanical stimuli) consisted of measuring the withdrawal threshold of the paw ipsilateral to the site of nerve injury in response to probing with a series of calibrated von Frey filaments. Each filament was applied perpendicularly to the plantar surface of the ligated paw of rats kept in suspended wire-mesh cages. The withdrawal threshold was determined by sequentially increasing and decreasing the stimulus strength ("up-down" method), analyzed with a Dixon non-parametric test [17] and expressed as the mean withdrawal threshold. Animals were tested before surgery and only animals with a paw withdrawal threshold greater than 10 grams were used for the subsequent study. Surgically treated animals were then tested on postoperative days 3 and 12. Only those animals that showed allodynia (paw withdrawal threshold smaller than 3 g) on both days were used for tissue collection on postoperative day 13 (less then 10% of the animals were excluded for tissue collection).

Tissue dissection and RNA preparation

Total RNA from each sample was prepared using Trizol™ (Life Technologies, Gaithersburg, MD), followed by RNEasy™ (Qiagen, Hilden Germany). RNA samples were analyzed by denatured gel electrophoresis. In addition, total RNA quality was assessed by capillary electrophoresis (Bioanalyzer 2100 Agilent, Palo Alto, CA) to ensure that the 28S:18S rRNA ratio was >1.0 for each sample.

Affymetrix microarray hybridization and staining

Hybridization probes were prepared according to Affymetrix instruction [18]. 5 μM primer encoding the T7 RNA polymerase promoter linked to oligo-dT24 primer was used to prime double-stranded cDNA synthesis from each total RNA sample (25 μg). cDNA synthesis reactions were carried out at 42°C using Superscript II RNAseH- reverse transcriptase (Life Technologies, Rockville MD). Second strand cDNA synthesis was finished using DNA polymerase I and T4 DNA ligase. Each double-stranded cDNA sample was purified by sequential phenol/chloroform extraction (Ambion, Austin, TX) and adsorption to silica (Qiaquick™ kit, Qiagen, Hilden, Germany) according to manufacturers' instructions. Half of each cDNA sample was transcribed in vitro into the copy RNA (cRNA) labeled with biotin-UTP and biotin-CTP using the BioArray High Yield RNA Transcript Labeling Kit (Enzo Biochemicals, New York, NY). These cRNA transcripts were purified using RNeasy™ columns (Qiagen, Hilden Germany) and quantitated by measuring absorption at 260 nm/280 nm. 15 μg aliquots of each cRNA sample were fragmented at 95°C for 35 min in 40 mM Tris-acetate, pH8.0, 100 mM KOAc, and 30 mM MgOAc to a mean size of ~50–150 nucleotides. Hybridization buffer (0.1 M MES, pH6.7, 1 M NaCl, 0.01% Triton, 0.5 mg/ml BSA, 0.1 mg/ml H. Sperm DNA, 50 pM Control Oligo B2, and 1X Eukaryotic hybridization Control) was added to each sample. Samples were then hybridized to RG-U34A microarrays (Affymetrix) at 45°C for 16 h. Microarrays were washed and sequentially incubated with streptavidin phycoerythrin (Molecular Probes, Eugene, OR), biotinylated anti-streptavidin antibody (Vector Laboratories, Inc., Burlingame, CA), and streptavidin phycoerythrin on the Fluidic Station (Affymetrix, Santa Clara, CA). Finally, the microarrays were scanned with a dedicated Gene Array Scanner (Hewlett Packard Instruments, TX) to capture a fluorescence image.

Affymetrix microarray data analysis

A total of 9 animals was divided into 3 groups, and the dorsal and ventral spinal cord were pooled for each group to give rise to 6 samples. Each sample was analyzed on one Affymetrix microarray RG-U34A. For each probeset (an array of 16 pairs of oligonucleotides for a specific gene) an index of gene expression was calculated and analyzed using the SAFER algorithm [19] for all chip analysis. The SAFER gene expression index is a robust and resistant measure of gene expression which is an alternative to the 'average difference' calculated by the Affymetrix analysis software and the model-based expression index proposed by Li and Wong [20]. Like Li and Wong's procedure, the procedure for calculating the SAFER gene index involves both between-array normalization and an adjustment for probe-specific biases.

To analyze gene expression in the spinal cord, differences in mean level of the gene expression index between dorsal and ventral samples were assessed for each probeset using a paired t-test and ANOVA. These models facilitated estimation of ratios comparing the dorsal and ventral samples and calculation of p-values testing whether the ratios are different from one (i.e. a ratio of one implies no change between the means for the experimental conditions). By fitting separate models for each probeset, differences were assessed using an error term that included biological variability between samples and did not assume that this variability was the same for all genes.

Quantitative Real-Time PCR (QRT-PCR)

Total RNA was treated with DNase I, Amplification Grade (Invitrogen, Carlsbad, CA) to remove DNA contamination before cDNA synthesis. cDNA was synthesized with oligo (dT)12–18 using Superscript First-Strand Synthesis System for RT-PCR (Invitrogen, Carlsbad, CA). Real-time PCR analysis was performed on a Applied Biosystems ABI Prism7700 Sequence Detection System. Matching primers and fluorescence probes were designed for each of the genes using the Primer Express program provided by Applied Biosystems. Both forward and reverse primers were used at 900 nM. In all cases, the final probe concentration was 250 nM. The PCR reaction was performed in a final volume of 50 μl using TaqMan Universal PCR Master Mix containing AmpliTaq GoldR DNA Polymerase, AmpEraseR UNG, dNTPs (with dUTP), Passive Reference 1, optimized buffer components (proprietary formulation) and 1 μl of cDNA template.

Primers and probes for GAPDH and IL-18 were obtained from Applied Biosystems. The sequence of the primers and probes used is listed in Table 4.

Table 4 Primers and probes used for QRT-PCR. The sequences of forward, reverse and specific probes are listed sequentially for each gene (accession #).
Full size table

QRT-PCR Data Analysis

Average Ct values from triplicate PCR reactions were normalized to average Ct values for GAPDH RNA from the same cDNA preparations. The ratio of expression of each gene between dorsal and ventral samples was calculated as: 2-(meanΔΔCt). C t represents the threshold cycle and ΔΔC t represents the difference Ct(test gene) - Ct(GAPDH RNA) for dorsal sample minus ventral sample. Using the ANOVA method, 95% confidence intervals were determined for each ratio as:

where t0.975 is the 97.5th percentile of the t-distribution with N-m degrees of freedom, N is the total pooled sample size for a gene, m is the number of treatments including control, s is the pooled standard deviation, n i and n j are the number of dorsal and ventral samples, respectively, being compared. Similarly, expression between ipsilateral and contralateral samples were analyzed.

Immunocytochemistry

Rat spinal cords were dissected from rats which had been perfused with 4% paraformaldehyde and post fixed for 4 hours. After cryoprotection in 30% sucrose overnight and 30% sucrose/OCT (1:1) mixture for 8 hours, the tissue was frozen and sections of 30 um were cut with cryostat. Tissue sections were floated, washed several times with PBS, then treated with 0.5% H2O2 for 30 minutes followed by washing with PBS 3 times. The sections were then incubated with blocking buffer (3% BSA + 3% donkey serum + 0.1% Triton) for 1 hour, followed by incubation with primary antibodies for 2 hours at room temperature. After washing with PBS 10 times, the sections were incubated with secondary antibodies, AB enzyme reagent (ABC kit, Vector) and developed using a Vector DAB staining Kit according to manufacturer's recommendations. The antibody to syntaxin-1 and Galpha(olf) were purchased from Santa Cruz Biotechnology, Inc. and used at 1: 20 dilution.

In situhybridization

Twenty micron sections were collected on gelatin-coated slides, dried and then stored at -80°C in desiccated boxes. At the time of processing, the slides were warmed to room temperature, postfixed in paraformaldehyde, treated with acetic anhydride and then delipidated and dehydrated. Processed section-mounted slides were hybridized with antisense or sense (control) riboprobes (8–12 × 106 DPM/slide) in 50% formamide hybridization mix and incubated overnight at 55°C in an open-air humidified slide chamber. In the morning, the slides were immersed in 2 × SSC (0.3 M NaCl, 0.03 M sodium citrate; pH 7.0)/10 mM DTT, treated with RNase A (20 mg/ml) and washed 2 × 30 min at 65°C in 0.1× SSC to remove nonspecific label. After dehydration, the slides were apposed to BioMax (BMR-1; Kodak) X-ray film for 1–2 days and then dipped in NTB2 nuclear emulsion (Eastman Kodak; diluted 1:1 with 600 mM ammonium acetate). The slides were exposed for 21 days in light-tight black desiccated boxes, photographically processed, stained in Cresyl violet and coversliped.