Journal of Neurophysiology

Global Gene Expression Analysis of Rodent Motor Neurons Following Spinal Cord Injury Associates Molecular Mechanisms With Development of Postinjury Spasticity

J. Wienecke, A-C. Westerdahl, H. Hultborn, O. Kiehn, J. Ryge

Abstract

Spinal cord injury leads to severe problems involving impaired motor, sensory, and autonomic functions. After spinal injury there is an initial phase of hyporeflexia followed by hyperreflexia, often referred to as spasticity. Previous studies have suggested a relationship between the reappearance of endogenous plateau potentials in motor neurons and the development of spasticity after spinalization. To unravel the molecular mechanisms underlying the increased excitability of motor neurons and the return of plateau potentials below a spinal cord injury we investigated changes in gene expression in this cell population. We adopted a rat tail-spasticity model with a caudal spinal transection that causes a progressive development of spasticity from its onset after 2 to 3 wk until 2 mo postinjury. Gene expression changes of fluorescently identified tail motor neurons were studied 21 and 60 days postinjury. The motor neurons undergo substantial transcriptional regulation in response to injury. The patterns of differential expression show similarities at both time points, although there are 20% more differentially expressed genes 60 days compared with 21 days postinjury. The study identifies targets of regulation relating to both ion channels and receptors implicated in the endogenous expression of plateaux. The regulation of excitatory and inhibitory signal transduction indicates a shift in the balance toward increased excitability, where the glutamatergic N-methyl-d-aspartate receptor complex together with cholinergic system is up-regulated and the γ-aminobutyric acid type A receptor system is down-regulated. The genes of the pore-forming proteins Cav1.3 and Nav1.6 were not up-regulated, whereas genes of proteins such as nonpore-forming subunits and intracellular pathways known to modulate receptor and channel trafficking, kinetics, and conductivity showed marked regulation. On the basis of the identified changes in global gene expression in motor neurons, the present investigation opens up for new potential targets for treatment of motor dysfunction following spinal cord injury.

INTRODUCTION

Spinal cord injury leads to an immediate impairment of motor and sensory functions that changes its manifestation over time. This involves an initial period of spinal shock with reduced reflexes followed by the development of a disturbing hyperreflexia, often referred to as spasticity (Ditunno et al. 2004). Significant effort has been devoted to understand the pathological changes that occur as a response to spinal cord injury as well as the mechanisms behind the subsequent development of spasticity (Hultborn 2003; Little et al. 1999; Nielsen et al. 2007). Research conducted on humans has pointed toward changes in reflex transmission as a mechanism for the hyperreflexia (Nielsen et al. 2007; Pierrot-Deseilligny and Burke 2005). Animal studies suggest that many factors contribute to spasticity, including morphological changes such as axonal sprouting (Bareyre et al. 2004; Raisman 1969), denervation supersensitivity (Stavraky 1961), altered expression of transmitter/receptor systems (Giroux et al. 1999; Tillakaratne et al. 2002), as well as the expression of plateau potentials in motor neurons (Bennett et al. 1999; Eken et al. 1989). Plateau potentials are seen in most vertebrate motor neurons, including man, and their activation leads to enhanced and prolonged muscle contraction (Crone et al. 1988; Heckman et al. 2005; Hultborn 1999; Kiehn and Eken 1998; Schwindt and Crill 1977, 1980). The expression of plateaux in motor neurons is conditional and depends on metabotropic receptor activation, including activation of noradrenergic or serotonergic receptors (Alaburda and Hounsgaard 2003; Conway et al. 1988; Delgado-Lezama et al. 1997; Hounsgaard and Kiehn 1989; Hounsgaard et al. 1988; Hultborn and Kiehn 1992; Lee and Heckman 1999). The ability to generate plateaux disappears in motor neurons located caudal to an acute transection (Crone et al. 1988; Hounsgaard et al. 1988), but reappears after 2 to 3 wk. This reappearance coincides with the development of spasticity, which led to the proposal that the increased extensor tonus and stretch reflexes may be related to the expression of plateau potentials in motor neurons located caudal to the chronic spinal injury (Bennett et al. 1999; Eken et al. 1989). The aim of the present study was to investigate possible molecular and cellular mechanisms underlying the increased motor neuron excitability and return of plateau potentials in the chronic spinal phase.

Plateau potentials in normal motor neurons are generated by persistent inward currents mediated by sodium and/or calcium channels and the neuromodulators that enable the expression of plateaux act either through activation of the persistent inward currents or by reducing opposing outward currents (Carlin et al. 2000; Hounsgaard and Kiehn 1989; Lee and Heckman 2001; Li and Bennett 2003; Li et al. 2004a; Simon et al. 2003; Svirskis and Hounsgaard 1998). Plateaux in chronic spinal animals are also mediated by calcium and sodium persistent inward currents (Harvey et al. 2006c; Li and Bennett 2003; Li et al. 2004a). However, the molecular mechanisms underlying the increased motor neuron excitability remain elusive. To shed new light on this unresolved conundrum we took advantage of recently established methodologies to examine the global transcriptional response of identified motor neurons (Cui et al. 2006; Ryge et al. 2008) in the rat tail-spasticity model developed by Bennett and colleagues (1999). We compare the global gene expression in identified tail motor neurons of transected animals with their sham-operated counterparts in the late chronic phase 21 and 60 days postinjury, where the transected animals show clear signs of spasticity. We find that the motor neurons undergo a very broad transcriptional response to the injury, where 1,502 and 1,784 genes are significantly differentially expressed 21 and 60 days postinjury, respectively. Among these genes we identified several new channel and receptor targets that are related to the increased excitability and reappearance of plateaux.

In summary, the present study focuses on the global transcriptional changes of a specific neuronal cell population in response to spinal cord injury and associates these changes with mechanisms previously related to a distinct pathological state of spinal cord injury—i.e., spasticity and the increased excitability observed in motor neurons in the chronic phase. This is a first step toward understanding how individual cellular components of the spinal cord interact after injury to create the pathophysiological postinjury state including spasticity, which will open up for the possibility of designing more specific treatments of spinal cord injury.

Part of this study was previously published in abstract form (Wienecke et al. 2007).

METHODS

Spinal cord operation

All handling of animals was approved by the Danish Animal Experiments Inspectorate. The handling and experimental procedures of the animals was conducted at University of Copenhagen (Denmark) and the isolated spinal cord tissue was further processed at Karolinska Institutet in Stockholm (Sweden).

Adult male Wistar rats (325–480 g) were used in this study. The animals used for microarray hybridization were separated into four groups: spinalized for 21 days (Spi-21, n = 8), spinalized for 60 days (Spi-60, n = 8), sham-operated controls for 21 days (ShamC-21, n = 6), and sham-operated controls for 60 days (ShamC-60, n = 5). This study also includes data from clinical and electrophysiological quantifications of the reflex response performed on the day of termination. To describe the full time course of the reflex response some rats (n = 15) were tested at several additional time points prior to termination.

Before operation the rats were initially anesthetized with isoflurane (Baxter) and thereafter a xylocaine–marcaine mix (xylocaine, 12.5 μg/ml:marcaine, 2.5 mg/ml, 1:4, AstraZeneca) of 0.2–0.3 ml per animal was injected intramuscularly. The second, and a few times also the third, lumbar vertebrae (L2 and L3) were removed. For ShamC-21 and ShamC-60 the dura was left intact after laminectomy, where after the wound was closed suturing muscles, muscle fascia, and skin separately. For the spinalized animals the dura was opened and spinal cord tissue corresponding to 1–2 mm was gently removed using a glass rod, forceps, and suction—creating a clear separation between the rostral and caudal cord at the sacral S2 segment. Only rats in which the dorsal vein and the ventral artery remained intact across the dissected gap were included in the study. After spinalization, the wound was closed suturing muscles, muscle fascia, and skin separately. Care was taken to relieve pain postoperatively by administration of buprenorfin subcutaneously (0.1 mg/kg, Temgesic, Schering-Plough) three times a day for the first 48 h. Until termination of the experiment the welfare of the rats was routinely checked (e.g., for signs of infections, motor loss, or bladder dysfunction). Rats that showed signs of distress were immediately killed. Since the spinal cord injury was inflicted at S2, only the motor and sensory functions of the tail were affected, leaving the bladder, bowel, and hind limb functions intact.

Spasticity and polysynaptic reflex measures

The development of tail spasticity was evaluated clinically and electrophysiologically acutely after operation (i.e., after 2–3 days) and at 1, 2, 4, and 8 wk (i.e., the intervals of 5–9, 14–15, 27–28, and 58–60 days) postinjury. These assessments were performed on awake rats after they were immobilized in a Plexiglas tube, leaving only the tail hanging out and free to move (see Fig. 1A). A stretch-rub maneuver (as described in Bennett et al. 1999, 2004) was used, together with simple pinching of the tip of the tail, to clinically rate the degree of spasticity on a scale from 0 to 5. Briefly, the stretch-rub maneuver was performed by holding the base of the tail with thumb and index finger while the other thumb and index finger were sliding a 37°C wet piece of gauze down the tail, starting at the base of tail and finishing by sliding off the tip (three times in a row). After the stimulus the tail was released and the tail movements were observed and the tail was subsequently touched/pinched for a rating (Fig. 1B): 0–1 describes no “clonus” (a rapid tremor of the end of the tail during spasms), no coiling, no response to light touch, and little or no hypertonus (no response to pinching); 2–3 describes strong flexor directed coiling spasms of the whole tail in response to the stretch-rub maneuver (lasts ≤10 s), and some coiling, some clonus, hypertonus, and very sensitive to light touch (pinching gives some flicking, coiling, and clonus); 4–5 is similar to 2–3 but with a greater amplitude response and longer duration, showing both extensor and flexor coiling, resulting in an S-shape curvature (lasting > tens of minutes), greater clonus, greater hypertonus, and greater flicking than that in 2–3 (pinching gives flicking and long-lasting coiling and clonus). All animals were assessed clinically by the same person.

Fig. 1.

Spasticity development in the injury model. A: clinical evaluation of spasticity and subsequent quantification of tail reflexes were performed while the rat was pacified in a test tube. The electrodes placed in the tail for reflex quantification are indicated in the drawing: R, recording electrodes; G, ground electrodes; and S, stimulation electrodes. B: clinical evaluation score of the development of spasticity ≤8 wk following transection. The vertical lines indicate SDs. C: tonic reflex component (rectified electromyography [EMG]) evoked by 5 times the motor threshold (5×MT) 2 days (Spi-2, light gray trace), 2 wk (Spi-14, gray trace), and 8 wk (Spi-60, black trace) after transection. The inset shows a magnification of the first 50 ms of the Spi-60 trace illustrating the initial reflex component and the M-wave. This initial component is similar in Spi-60 and Spi-21 (dashed line in D). D: development of the initial (dashed line) and the late tonic (full line) reflex component at 5×MT with time. The size of the reflexes is standardized to the size in acute spinal rats 2 days after transection. Each point illustrates rectified and integrated averaged values of both reflex components. Vertical lines indicate SDs. The values for 2, 4, and 8 wk are significantly different (P < 0.05) from acute spinal values for both the initial and the late tonic reflex components.

After clinical testing the reflex was assessed with electromyographic (EMG) recordings, quantifying muscle activity as a response to electrical stimulation of afferent fibers. The rats were briefly anesthetized with isoflurane while four pairs of sterile steel wire (Ethicon, 4–0, Johnson & Johnson) electrodes were inserted in the tail, each pair passing subcutaneously on either side of the tail. Each wire electrode pair was joined, allowing bilateral stimulation or recording. The electrode configuration for stimulation and recording is illustrated in Fig. 1A. Electrophysiological recordings were performed when the rat had fully recovered from the wire-mounting surgery (∼20 min later).

Stimulus–response curves were obtained by applying single stimulation pulses of 0.2 ms duration delivered every 2–5 s with increasing stimulus strength to cover a range from 0.2 to 10 mA. The maximal motor response (Mmax) was determined as maximal amplitude in the raw signal (measured peak-to-peak) following stimulation of all motor axons during the acquisition of the stimulus–response curve. The stimulus was normalized to the motor threshold (MT). EMG responses were quantified on the basis of a stimulus strength five times the motor threshold (5×MT). The EMG trace was band-pass filtered (100 Hz to 10 kHz), rectified, integrated, and subsequently normalized with respect to Mmax and Spi-2 rats (Fig. 1C). This normalization was performed to produce a standardized measure of the increase in activity/spasticity. We normalized to Spi-2 (“acute spinal”) rats instead of normal animals to avoid inflicting unnecessary pain to animals (which a 5×MT stimulation would otherwise induce in normal rats with intact spinal cord). However, the reflex amplitude in normal and early acute spinal rats has been shown to be more or less identical (Bennett et al. 2004). The reflex response was quantified as the area under the curve in two separate intervals following stimulation: 1) 15–50 ms (i.e., initial reflex component) and 2) 50–500 ms (i.e., late tonic reflex component).

Motor neuron labeling, spinal cord dissection, cryosectioning, and laser microdissection

Motor neurons were retrogradely labeled by injection of Fluoro-Gold dissolved in saline (Fluorochrome LLC) into both the tail muscles (150 μl) and intraperitoneally (30 μl) (Leong and Ling 1990). All animals (spinalized and sham operated) were injected with this dual-injection procedure under anesthesia (isoflurane) 5 to 7 days before the spinal cord was removed. We found that this dual-injection procedure enhanced the labeling of the sacral motor neurons.

Animals were anesthetized with pentobarbital (initially 20 and then 5 mg/kg every 30 min, Mebumal) before the sacrocaudal spinal cords were removed, after which the rats were killed with an overdose of pentobarbital. The spinal cords were immediately frozen in liquid nitrogen and stored at −80°C until cryosectioning.

The isolated and frozen spinal cords (the S3–S4 level, where most of the tail motor neurons are located; Bennett et al. 2004) were cryosectioned into 10 μm sections. The sections were mounted on nuclease and human nucleic acid free 0.9 μm polyester (POL) membrane frame slides (Leica Microsystems) and stored at −80°C until laser microdissection. The Fluoro-Gold–labeled motor neurons were isolated from the spinal cord sections using the Leica AS laser microdissection system (Leica Microsystems) at room temperature (Fig. 2A before and Fig. 2B after laser microdissection). To reduce RNA degradation laser microdissection was performed within 1 h for each slide. Between 70 and 200 Fluoro-Gold–labeled neurons were laser microdissected per rat and the motor neurons were collected in the cap of a eppendorf tube by force of gravity. RNA degradation and contamination from external sources were minimized in all steps of the experimental procedure. Possible contamination of samples with material from surrounding nonmotor neurons during laser microdissection was also minimized as described by Ryge et al. (2008).

Fig. 2.

Motor neuron laser microdissection. A: picture of Fluoro-Gold–labeled motor neurons from 10-μm-thick sections on polyester membrane frame slides ready for laser microdissection. The 3 arrows indicate 3 motor neurons in the ventral horn. B: picture of the same section as in A after laser microdissection.

Total RNA isolation, mRNA amplification, and aRNA biotinylation for microarray hybridization

Total RNA was isolated using the PicoPure RNA Isolation Kit (Arcturus) and the messenger RNA (mRNA) fraction was amplified in a two-round T7 linear amplification process using the RiboAmp HS RNA Amplification Kit (Arcturus). The complementary DNA (cDNA) product from the second round of the amplification process was used to generate biotin-labeled antisense RNA (aRNA) (GeneChip Expression 3′-Amplification Reagents for IVT Labeling; Affymetrix). The integrity and concentration of the amplified and biotinylated aRNA was assessed on an Agilent RNA chip with the Agilent 2100 bioanalyzer (Agilent Technologies) both before and after fragmentation. Only samples of good integrity were further used and 15 μg of the fragmented samples were hybridized to GeneChip Rat Genome 230 2.0 Arrays (Affymetrix, RAT230_2 chip) and subsequently scanned. Each array always originated from a single animal. The Agilent analysis and microarray hybridizations were conducted at the Affymetrix core facility at Novum (Bioinformatics and Expression Analysis Core Facility, Department of Biosciences and Nutrition, Karolinska Institutet, Huddinge, Sweden).

Microarray preprocessing

The microarray normalization and analysis for detection of significantly differentially expressed (DE) genes were adopted from Ryge and colleagues (2008). The RAT230_2 chip contains 31,099 probe sets, whose annotations are continuously updated within the Ensembl genome database (Hubbard et al. 2007). We used the Affymetrix probe sets verbatim, but discarded those not included in the Ensembl database for the RAT230_2 chip prior to the statistical analysis. Ensembl includes Affymetrix probe sets in their database only if more than half of its probes match the updated target gene sequence (http://mar2009.archive.ensembl.org/info/docs/microarray_probe_set_mapping.html). This procedure reduced the set of probes from 31,099 to 12,919. The microarrays were then background compensated, normalized, and RMA (robust multiarray average) expression summaries were calculated (Irizarry et al. 2003; Ryge et al. 2008). To validate the background correction and normalization procedure of all microarrays, the RMA distributions were examined together with the average distribution of all microarrays. Additional background compensation was carried out on the expression summaries since we observed a discrepancy between the RMA distributions at low expression levels that could not be eliminated with the probe level normalization and background compensation. This discrepancy of RMA distributions was manifested as a deviation from the diagonal of the quantile–quantile plots for low expression values between different samples. To compensate for this effect the expression summaries were transformed to normal scale (2^RMA), the average of the first 10% of the RMA values (1,292 probe sets) was calculated for each microarray, and used as a microarray-specific background measure subtracted from their corresponding microarray RMA values. To ensure that no negative expression summaries occurred after this procedure the universal minima of all microarrays (negative) was subtracted from all RMA values. After this procedure the expression summaries were transformed back to log2 scale and inspection of their distributions showed that all microarray RMA profiles followed the average distribution throughout the intensity range, validating the microarray preprocessing steps (Supplemental Fig. S1A).1 To further examine the separation of experimental groups principal components analysis was performed on the RMA profiles, which showed a nice separation into two groups of injured and uninjured animals (Supplemental Fig. S1B).

RMA expression summaries based on alternative probe mappings to Ensembl exons were also used to examine the possible regulation of splice variants (Dai et al. 2005) (Brainarray project, Molecular and Behavioral Neuroscience Institute [MBNI], version 11).

The RMA expression summaries together with the raw CEL files for all microarrays were submitted to the Gene Expression Omnibus (GEO; http://www.ncbi.nlm.nih.gov/geo/) hosted by the National Center for Biotechnology Information and can be accessed under accession number GSE16710.

Differentially expressed genes

Differentially expressed (DE) genes were determined separately for the two time points 21 and 60 days postinjury—i.e., Spi-21 and Spi-60 compared with their respective sham-operated counterparts, ShamC-21 and ShamC-60. To determine the significantly DE genes within each time point, the false discovery rates (FDRs) of three statistical tests [Cyber-T, limma, and SAM described in Baldi and Long (2001), Smyth (2004), Tusher et al. (2001), and Wettenhall and Smyth (2004)] were used to create a conglomerate ranking of each gene, reflecting their degree of significance across all three tests. In short, for each test all genes were ranked on the basis of their test statistics, producing three ordered lists of genes from most to least significantly DE gene. The three rankings obtained from these lists were then used to produce an average rank for each gene, according to which all genes were reordered (Ryge et al. 2008). This procedure was carried out for both the Spi-21 and Spi-60 groups and the top-ranking genes that fall within a 1% FDR were extracted to produce two lists of significantly DE genes, one for Spi-21 and one Spi-60 (these two lists of DE are provided in Supplemental Tables S1 and S2). For each test the number of DE genes that fall within a 1% FDR was calculated and the mean of these was used as the number of genes that on average pass the 1% FDR to be extracted from the ordered list. For Spi-21 a number of 1,502 genes pass a 1% FDR (Cyber-T = 1,305, limma = 1,336, SAM = 1,866), whereas for Spi-60 a number of 1,784 genes pass (Cyber-T = 1,520, limma = 1,516, SAM = 2,317). MA and Volcano plots of the injured groups versus their sham-operated counterparts show that the ratios of expression between the two conditions are centered around 1.0 and well balanced throughout the intensity range (Supplemental Fig. S1, CF), supporting the preprocessing of the data. Furthermore, the DE genes separate nicely from the non-DE genes, supporting the robustness of the regulation of gene expression in the injured state compared with sham. To compensate for multiple testing the P values of Cyber-T and limma were converted to FDRs using the approach of Allison and colleagues (2002), whereas the FDRs of SAM are based on a methodology of permutation and resampling of the data (i.e., the FDRs are output from the SAM analysis directly).

Allen Brain Atlas

The expression of genes in the motor neurons of the normal spinal cord have not previously been described in the literature for all of the significantly DE genes contained in Tables 1, 2, and 3. To examine the presence of their transcripts in motor neurons located in the ventral horn of normal rodents we therefore used the Allen Brain Atlas (http://www.brain-map.org/). At the time of writing the Allen Brain Atlas contained in situ hybridizations of close to 17,000 genes mapped across all anatomical segments of the cord for juveniles (P4) and adults (P56) of the C57BL/6J mouse strain. Of the 68 genes contained in Tables 13, in situ hybridization data were available for 63 genes in the Allen Brain Atlas. These summaries are included in Tables 1, 2, and 3, where the level of expression in the ventral horn is scaled from low to high as: low: *, medium: **, high: ***, very high: ****. Absent: minus (−) and Not Available: NA. We assume that the expression of genes in motor neurons of the uninjured spinal cord of the mouse also provide strong evidence for their expression in the rat, an evolutionary closely related species with very similar locomotor behavior.

View this table:
Table 1.

Significantly differentially expressed genes relating to ion channels

View this table:
Table 2.

Significantly differentially expressed genes relating to neuro-transmitter receptors

View this table:
Table 3.

Significantly differentially expressed genes relating to intracellular pathways affecting motor neuron excitability

Real-time reverse-transcription PCR

Real-time reverse transcription PCR (real-time RT-PCR) was used to validate the regulation of expression for 10 genes detected as significantly DE in the array analysis (from Tables 13): Scn1b, Scn9a, Slc18a3, Cacng2, Grina, Grin3b, Kcnc3, Clcn3, Nrg1, and Calm1. Sodium/potassium-transporting ATPase subunit β-1 (Atp1b1) and 40S ribosomal protein S18 (Rps18) were chosen as normalization genes (i.e., endogenous controls) based on the microarray data, where they showed a consistent and conserved level of expression across all samples. Glyceraldehyde 3-phosphate dehydrogenase (Gapdh) and β-actin (Actb) are both frequently used as endogenous reference housekeeping genes in real-time RT-PCR experiments, but both appear on the list of significantly DE genes (Supplemental Tables S1 and S2). This is not unreasonable because injury not only may induce changes in cytoskeletal structures but also may regulate cellular metabolism.

Real-time PCR was performed on cDNA reverse transcribed from two linear round amplified and biotinylated aRNA (i.e., excess aRNA from microarray experiments not used for hybridization). TaqMan Reverse Transcription Reagents (Applied Biosystems) were used for the RT reaction and the cDNA was primed with random hexamers. TaqMan Gene Expression Assays and TaqMan Universal PCR Master Mix were used in the real-time PCR reactions (Applied Biosystems). The TaqMan primers chosen for the TaqMan gene expression assays all targeted the last exon boundary of their transcript to ensure detection after two rounds of linear amplification and their identifiers are listed in Supplemental Table S3. Three individual samples from each experimental group (Spi-21, Spi-60, ShamC-21, and ShamC-60) were assayed in three replicates and the input amount of the respective cDNA template was consistently 0.1 μg per 50 μl PCR reaction. The real-time PCR reactions were performed in a ABI Prism 7000 Sequence Detection System (Applied Biosystems) under the following thermal cycling conditions: 1) incubation at 50°C for 2 min, 2) incubation at 95°C for 10 min, and 3) amplification for 55 cycles of 95°C for 15 s (denaturation) and 60°C for 60 s (annealing/extension). Real-time RT-PCR with TaqMan MGB probes was performed according to manufacturer's instructions.

The 2(−ΔΔCT) method (Livak and Schmittgen 2001) was used to quantify the difference in mRNA levels between samples and endogenous controls. The analysis was done in R on data exported from the ABI Prism 7000 SDS software (Applied Biosystems).

Software

The microarray and real-time RT-PCR analyses were done using R (http://www.r-project.org/) and Bioconductor (http://bioconductor.org). Cyber-T source code was obtained from the website http://cybert.microarray.ics.uci.edu/. Separate R scripts for filtering, analyses, ranking, and plotting microarray data as well as for the real-time RT-PCR analysis were developed and can be obtained from the authors on request.

RESULTS

Development of spasticity after spinalization in the rat tail model

Clinical evaluation of spasticity as well as reflex quantification were performed in all spinalized rats before removal of the spinal cord while the animals were immobilized in a test tube (Fig. 1A). The clinical evaluation of spasticity was performed using the stretch-rub-maneuver and pinching (see methods). In the acute phase the tail is completely flaccid and does not respond to the stretch-rub maneuver or pinching. After 1 wk the first signs of spasticity are observed, which then gradually develop ≤8 wk following the lesion (Fig. 1B). Spi-21 rats all showed moderate spasticity, characterized by hypertonus and occasionally clonus of the tail, and were rated 2–3. Spi-60 rats showed severe spasticity and were rated 4–5, characterized by severe hypertonus, clonus, and coiling of the tail. After clinical evaluation the animals were tested electrophysiologically. All animals showed two reflex components, an initial reflex with a delay of around 25–30 ms, and a late tonic reflex at 50–500 ms (Fig. 1C). The size of the reflexes was measured at stimulus strengths 5×MT, with the quantification of the reflexes shown in Fig. 1D (data from 15 rats). The two reflex components progress slightly differently. The initial reflex component increases gradually until 14–21 days postinjury, after which it levels off and remains at a constant level. In contrast, the late reflex component continues to develop throughout the recording period. These data show the presence of hyperreflexia at day 21 with a further increase in severity until 60 days postinjury in accordance with previous studies (Bennett et al. 1999, 2004).

Transcriptional response of motor neurons to the injury

To determine the transcriptional response in motor neurons of spinalized rats we compared samples of Spi-21 or Spi-60 with their sham-operated counterparts, ShamC-21 and ShamC-60, respectively. To extract the significantly DE genes we used a conglomerate classifier based on three statistical methods: limma, Cyber-T, and SAM (see methods). Each constitutes a modified t-test with alternative ways of calculating the expression variance and, taken together, they constitute a good estimate of likely differential expression (Ryge et al. 2008). A list of significantly DE genes was thus created for the two time points, each containing the average number of genes that pass a 1% FDR. For Spi-21 and Spi-60 this produces two lists of 1,452 and 1,841 DE genes, respectively (Supplemental Tables S1 and S2).

Differentially expressed genes affecting ionic conductances

From the two lists of DE genes 21 and 60 days postinjury (Supplemental Tables S1 and S2), the candidates that may have a direct effect on the intrinsic membrane excitability were extracted and are displayed in Table 1. These encompass the genes coding for the protein subunits that constitute the calcium channels, sodium channels, chloride channels, and potassium channels. Table 1 contains 31 probe sets representing 26 unique genes, where two genes are represented by two or more probe sets (Clcn3 and Kcnc3). The level of significance within a 1% FDR of either Spi-21 or Spi-60 is indicated by: +, one of the three statistical methods classify the gene as significant; ++, two of the statistical methods classify the gene as significant; or +++, all three methods classify the gene as significant. The absence of + signifies no significant expression of that gene within a 1% FDR at that time point. The table also contains the ratios of regulation, where down-regulation is indicated with a minus prefix. These ratios are furthermore plotted in Fig. 3A for Spi-21 versus ShamC-21 and in Fig. 3B for Spi-60 versus ShamC-60, including error bars to give a visual impression of the significance of regulation. Note that a gene needs only to be classified as significantly DE at one time point to be included in Table 1 (data included for genes only at time points of significant DE). Figure 3, A and B on the other hand includes the ratios for all genes contained in Table 1 irrespective of significance at a give time point. This illustrates that although the magnitude of gene regulation (i.e., ratios of Spi/ShamC) changes over time, most genes relating to ion channels are changing their expression in the same direction 21 and 60 days postinjury, even for genes not reaching a significance level of 1% FDR at one of the two time points.

Fig. 3.

Average expression ratios of genes affecting motor neuron excitability. The average ratios of expression for genes contained in Tables 1, 2, and 3 for Spi-21 and Spi-60 with respect to their sham-operated counterpart. Error bars signify SE of expression ratios and horizontal dotted lines are drawn at ±1.3 ratio of expression. A: expression ratios of genes relating to ion channels (Table 1) for Spi-21 vs. ShamC-21. Color code: red, calcium channels; green, sodium channels; blue, chloride channels; dark blue, potassium channels. B: the same as A but for Spi-60 vs. ShamC-60. C: expression ratios of genes relating to neurotransmitter receptors (Table 2) for Spi-21 vs. ShamC-21. Color code: red, glutamate receptors; yellow, acetylcholine receptors; green, GABA receptors; turquoise, glycine receptor; light blue, serotonin receptors; dark blue. adrenergic receptors; purple: dopamine receptors. D: the same as C but for Spi-60 vs. ShamC-60. E: expression ratios of genes relating to intracellular pathways affecting motor neuron excitability (Table 3) for Spi-21 vs. ShamC-21. Color code: red, calmodulin and CaM kinase; yellow, Ca2+ binding proteins; green, protein kinase A and related genes; light blue, protein kinase C; dark blue, protein tyrosine phosphatases; purple, Cl cotransporters. F: same as E but for Spi-60 vs. ShamC-60.

Inward currents.

Calcium and sodium ion channels conduct inward currents and up-regulation of these channel complexes can thus lead to increased membrane excitability. Calcium voltage-gated channels are classified into five groups according to their distinctive conductance properties, determined by their core pore-forming protein subunits. The five channel families contain the following protein subunits: L-type (Cav1.1–Cav1.4), P/Q-type (Cav2.1), N-type (Cav2.2), R-type (Cav2.3), and T-type (Cav3.1–Cav3.3). Two genes coding for different α subunits undergo regulation as a response to injury—i.e., Cacna1b (Cav2.2) and Cacna1d (Cav1.3). Both genes are significantly DE in Spi-60, where Cacna1b is up-regulated and Cacna1d is down-regulated. Two genes coding for β calcium channel subunits are DE—β2 (Cacnb2) and β4 (Cacnb4)—both of which are down-regulated. Cacnb4 is down-regulated in both Spi-21 and Spi-60, whereas Cacnb2 is significant only in Spi-21. The gene Cacng2 also shows a response to the injury. This gene was initially classified as a calcium channel γ subunit (γ2), but has recently been shown to belong to the TARP family and will therefore be described in the subsequent section Regulation of excitatory synaptic transmission.

For the voltage-gated sodium channels 2 of the 11 known α subunit genes (Scn1a through Scn11a) are DE: Scn2a1 (Nav1.2) and Scn9a (Nav1.7). Both are down-regulated. Two genes of the β subunits are also DE: Scn1b and Scn3b. Scnb1 is up-regulated in both Spi-21 and Spi-60, whereas Scnb3 is down-regulated in Spi-60.

Outward currents.

Chloride and potassium ion channels conduct outward, hyperpolarizing currents on activation, opposing the effect of sodium and calcium channels. The potassium channel family constitutes the largest group in this set (Table 1, 15 of 31 genes).

Among the genes coding for potassium channel subunits that undergo regulation as a response to injury, we find transcripts of calcium-activated, inward-rectifying, voltage-gated, and tandem pore domain (leak) channels. Genes relating to the voltage-gated type family seem to constitute the largest group of DE potassium channels (8 of 15 genes). Genes belonging to the voltage-gated family include two genes of the Shaker subfamily (Kcna1 and Kcna4), two genes of the Shaw subfamily (Kcnc2 and Kcnc3), one gene of the Isk-related subfamily (Kcne2), one gene of the subfamily H (Kcnh2), and one gene of the delayed-rectifier family (Kcns3). One gene that codes for the modulatory β subunits belonging to the Shaker subfamily is also found to be DE (Kcnab1). Of the other types, one gene belongs to the calcium-activated potassium channels of the SK (small conductance Ca2+-activated) subfamily (Kcnn2), two genes belong to the inward-rectifying family (Kcnj2 and Kcnj6), and three genes belong to the tandem domain family (Kcnk2, Kcnk3, and Kcnk12). Finally, one potassium channel tetramerization-domain-containing gene is DE (Kctd11). Note here that 8 of the 15 genes related to K+ channels are clearly down-regulated in either Spi-21 (Kcna1b, Kcnc2, Kcnj6, and Kctd11), Spi-60 (Kcna4 and Kcne2), or both (Kcnk2 and Kcnn2). The rest are up-regulated in either Spi-60 (Kcna1, Kcnj2, Kcnk3, Kcnk12, and Kcns3) or both (Kcnc3 and Kcnh2).

The chloride channels are not as well characterized as the other classes of ion channels. One gene (Clca6) coding for a calcium-sensitive chloride channel is up-regulated in Spi-21 animals. Two other genes coding for voltage-gated chloride channels were DE as a response to the injury, where one is up-regulated (Clcn3) and one is down-regulated (Clcc1).

Differentially expressed genes related to neurotransmitter receptors

In a similar fashion to Table 1, the genes relating to receptors of the most common neurotransmitters and neuromodulators were extracted from the two lists of DE genes and are shown in a separate table (Table 2). This extraction includes receptors of glutamate, acetylcholine, glycine, and γ-aminobutyric acid (GABA) as well as monoamine receptors. Table 2 contains 27 distinct genes belonging to this class, where only one gene is represented with more than one probe set (Nrg1). Levels of significance and expression ratios are indicated as in Table 1 for Spi-21 and Spi-60. The ratios of expression are plotted for all genes in Table 2 and for Spi-21 and Spi-60 in Fig. 3, C and D, respectively. Note again the similar pattern of gene regulation at the two time points, even for genes not significantly DE in both because absolute ratios may differ.

Excitatory neurotransmitter receptors.

Glutamate and acetylcholine receptors belong to the excitatory class of receptor channels. For the DE genes relating to ionotropic glutamate receptors, the gene coding for the α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) glutamate receptor 4 (Gria4) is down-regulated, whereas the gene coding for the kainate glutamate receptor 5 (Grik5) is up-regulated. Up-regulation is also seen for one gene coding for a protein belonging to the TARP family (Cacng2), known to regulate the localization of AMPA receptors as well as modulate their channel conductivity: γ2 (stargazin). Four genes related to the N-methyl-d-aspartate (NMDA) subfamily are DE (Grin1, Grin3b, Grina, and Grinl1a), suggesting that the NMDA complex undergoes strong regulations as a response to the injury. Grinl1a is down-regulated, whereas the other three are up-regulated. Two genes belonging to the metabotropic glutamate receptor subfamily are DE: Grm1 (mGluR1) is down-regulated, whereas Grm4 (mGluR4) is up-regulated.

For acetylcholine receptors, primarily genes coding for the nicotinic receptors undergo regulation in response to the injury. Genes of two α subunits are up-regulated (Chrna4 and Chrna6). One gene possibly involved in the anchoring of acetylcholinesterase in the membrane (Cuta) together with the gene coding for the vesicular acetylcholine transporter (VAChT, Slc18a3) are up-regulated as a response to the injury. One transcript related to maturation and maintenance of acetylcholine receptor system has two probe sets that show opposing direction of regulation (Nrg1), indicating regulation of different splice variants (see regulation of splice variant transcripts). Taken together these DE genes indicate that the nicotinic receptor complex undergoes regulation as well as the machinery for cholinergic synaptic transmission.

Inhibitory neurotransmitter receptors.

GABA and glycine receptors belong to the inhibitory class of receptor channels. Most of the regulated genes coding for GABA-receptor subunits relate to the ionotropic GABAA subfamily, i.e., two genes of the α subunits (Gabra1 and Gabra5) and one gene of the γ subunit (Gabrg2) are down-regulated. Three genes putatively involved in GABAA-receptor trafficking and clustering in the membrane are DE, with two being up-regulated (Gabarapl1 and Trak2) and one being down-regulated (Gabarapl2). Only one gene belonging to metabotropic GABAB subfamily is up-regulated (Gabbr1). One of the four α subunit genes belonging to glycine receptors is up-regulated (Glra1).

Modulatory neurotransmitter receptors.

Monoaminergic receptors are important for modulating motor neuron excitability and were therefore included in Table 2. Four of the five DE genes related to this class of receptors are up-regulated as a response to injury. One gene of the serotonin receptor family (Htr2b) is down-regulated in Spi-21. Two of the DE genes belong to the adrenergic receptor family, which comprise two classes of G-protein–coupled receptors, α and β. One gene belonging to the α receptor subfamily is up-regulated, α1 (Adra1d). A transcript of the receptor kinase that phosphorylates the β-2 adrenergic receptor is also up-regulated (Adrbk1). The D2-like dopamine receptor gene (Drd4) is up-regulated in Spi-60 together with a gene of the D1 receptor interacting protein (Caly) up-regulated in Spi-21.

Differentially expressed genes related to intracellular pathways affecting motor neuron excitability

Channel and receptor functions are subject to modulation by various intracellular pathways. In particular, we focus on DE genes coding for proteins involved in intracellular pathways that have been shown to modulate persistent inward currents as well as NMDA- and GABAA-mediated synaptic transmission, identified as the major receptor targets of the transcriptional injury response (previous section). These genes are included in Table 3. The extracted categories include genes affecting 1) persistent inward calcium currents: calmodulin and Ca2+/calmodulin (CaM)-dependent protein kinase as well as additional Ca2+-binding proteins; 2) persistent inward currents as well as GABAergic transmission: protein kinase A (PKA), protein kinase C (PKC), and their counterbalancing phosphatases; and 3) NMDA and GABA transmission: protein tyrosine kinase and phosphatases. The table also contains DE genes coding for anion transporters, possibly affecting the reversal potential of inhibitory ionotropic receptors, such as GABAA. In total Table 3 contains 16 probe sets representing 15 distinct genes, where expression ratios and levels of significance are indicated as in Tables 1 and 2 for Spi-21 and Spi-60. The ratios of expression are plotted for these probe sets in Fig. 3, E and F for Spi-21 and Spi-60, respectively.

The gene coding for Calmodulin 1 (Calm1) and the gene coding for the CaM kinase II γ subunit (Camk2g) were up-regulated. Two other genes coding for Ca2+-binding proteins were also up-regulated: Caldendrin (Cabp1) and Neuronal calcium sensor 1 (Freq). The gene coding for the β-catalytic subunit for PKA (Prkacb) was down-regulated in both Spi-21 and Spi-60, whereas the gene for the α-catalytic subunit for PKA (Prkaca) was up-regulated at both time points. Three genes coding for phosphatases that may counteract PKA activity were all down-regulated: PP1-β (Ppp1cb), PP1-γ (Ppp1 ml), and PP2-α (Ppp2ca). The gene for PKC beta (Prkcb1) was strongly down-regulated both in Spi-21 and Spi-60. Genes relating to phosphatidylinositol 3-kinase (PI3K) were not differentially expressed.

The expression of the gene coding for the protein tyrosine kinase Src, known to modulate the function of both NMDA- and GABAA-receptor complexes, was not changed after spinal cord injury. Several genes coding for protein tyrosine phosphatase (PTP) receptor subunits were down-regulated both in Spi-21 and Spi-60. These include PTP receptor type α (Ptpra), PTP receptor type O (Ptpro), PTP receptor type N (Ptprn), and PTP receptor type R (Ptprr).

One gene coding for the anion exchange protein 3 (Slc4a3) that may accumulate Cl in the motor neurons was up-regulated in Spi-21 and Spi-60.

Regulation of splice variant transcripts

In many cases the microarray contains several probe sets that target the same gene at various locations. For two genes relating to ion channels (Table 1: Clcn3 and Kcnc3), one gene relating to receptors (Table 2: Nrg1), and one gene relating to intracellular pathways (Table 3: Ppp1cb), multiple probe sets show significant DE. One of these genes has probe sets that exhibit opposing direction of regulation (Nrg1). This apparent discrepancy could be due to a differential regulation of different splice variants. To investigate this possibility we had to use alternative probe set mappings where the probes have been mapped specifically to Ensembl exons (Dai et al. 2005), since the probe sets of the RAT230_2.0 chip are neither transcript nor exon specific. RMA summaries were calculated for these new probe sets. Overall, this analysis confirmed our findings based on the original probe sets showing similar pattern of regulation in Ensembl exons and DE of the genes contained in Tables 1, 2, and 3 (data not shown). This analysis therefore supports the initial filtering conducted on Affymetrix probe sets based on the Ensembl annotations. Two of the exons of Nrg1 were significantly DE within a 1% FDR exhibiting opposing ratio of expression, suggesting that Nrg1 most likely undergo splice variant regulation 21 and 60 days postinjury.

Validation with real-time RT-PCR

To validate the results of the microarray analysis ten genes were chosen for real-time RT-PCR (see also Ryge et al. 2008) among the significantly DE genes of Tables 1 and 2 (Scn1b, Scn9a, Slc18a3, Cacng2, Grina, Grin3b, Kcnc3, Clcn3, Nrg1, and Calm1). For the Nrg1 target transcript as well as Scn9a the microarray analysis showed down-regulation, whereas all the other tested genes showed an up-regulation (see Tables 1 and 2). The real-time RT-PCR validation of the microarray ratios 21 days postinjury is shown in Fig. 4A. The error bars of Fig. 4A indicate SEs of expression ratios. The same genes are shown in similar fashion for Spi-60 compared with ShamC-60 (Fig. 4B). Of the tested genes four exhibited amplification curves that leveled off below the signal threshold set for CT determination, indicating the presence of a low number of transcript copies for these genes. Of the remaining six genes all show the same direction of regulation in the real-time RT-PCR as in the microarray analysis in Spi-60. In Spi-21 the genes Scn1b, Grina, and Nrg1 are not DE because they show rather large variation of expression, whereas the remaining genes Cacng2, Clcn3, and Calm1 confirm the microarray results at this time point.

Fig. 4.

Real-time RT-PCR validation. Ratios of expression for 6 genes (Scn1b, Cacng2, Grina, Clcn3, Nrg1, and Calm1) 21 and 60 days postinjury compared with their sham counterparts. A: gene expression ratios of Spi-21 compared with ShamC-21. B: the same as A but for Spi-60 compared with ShamC-60. Error bars illustrate the SE for each ratio.

DISCUSSION

Spasticity model and gene expression analysis

For the purpose of the present project it was crucial to study the segmental effects of spinal cord injury in isolation without interference of compensatory or plastic effects resulting from any remaining descending fibers. We therefore used an injury model where the spinal cord is fully transected, effectively causing a degeneration of all descending fibers, rather than using the other commonly used contusion or compression spinal cord injury model, in which descending fibers may be spared (see Grill 2005 for a commentary of those models). To analyze the transcriptional regulation of motor neurons we adopted the method of Ryge and colleagues (2008), where it was shown that reliable differences in gene expression could be detected between neuronal populations of the spinal cord based on microarrays originating from laser microdissected cells identified by retrograde labeling and with minimal contamination from neighboring cells such as glia (Cui et al. 2006; Ryge et al. 2008).

Two strategies were used to support the microarray analysis. To validate the presence of transcripts in motor neurons of uninjured animals we examined in situ hybridization data obtained from Allen Brain Atlas. Of the 68 distinct genes shown in Tables 1, 2, and 3, 63 genes were available in the database and 62 of these were present at some level in the ventral horn of the rodent spinal cord of either juvenile or adult animals, suggesting a role for these genes in motor neurons under normal physiological conditions. Since in situ hybridization cannot be used to quantify small changes in gene expression within a given cell population, real-time RT-PCR was used on a small subset of genes to validate their regulation in the injured animals. Six of the 10 genes chosen for real-time RT-PCR analysis validated the microarray results. The four remaining gene transcripts did not amplify above the signal threshold set for CT determination. The four transcripts did exhibit weak signal at late cycles (cycles 35–45, not shown) and we speculate that in our very small samples (100–200 cells) the amplification is limited for transcripts with relative low to medium copy numbers. This included the genes Slc18a3 (VAChT) and Grin3b, which are both expressed ubiquitously in motor neurons (Allen Brain Atlas), thus supporting the specificity and sensitivity of the microarray analysis. The low signal observed with the TaqMan probes of the real-time RT-PCR for these genes therefore seems to indicate that they are expressed in lower levels compared with the other real-time RT-PCR positive genes. This was also reflected in the microarray intensities (not shown). We therefore conclude that the microarrays are sensitive to low expressed genes including ion channels and receptors, enabling us to examine the regulation of these as a response to injury.

Transcriptional response related to postinjury spasticity and the return of plateau potentials

The emergence of plateau potentials and increased motor neuron excitability is likely to be influenced by many factors, which can be divided into a few functional categories relating to 1) ion channel configuration that define the intrinsic membrane properties, 2) receptors mediating fast synaptic transmission or acting on “neuromodulatory” pathways, and 3) intracellular pathways modulating membrane excitability and signal transduction.

The patterns of regulation for the genes associated with these categories (Tables 13) are quite similar between the two time points—i.e., whereas the magnitude of gene regulation might differ, the transcriptional response largely changes in the same direction 21 and 60 days postinjury (Fig. 3). The magnitudes of regulation (ratios) seem to increase progressively for many genes from Spi-21 to Spi-60, signified by a 20% increase of genes being classified as significantly DE in the motor neurons 60 days (1,784 genes) compared with 21 days (1,502 genes) postinjury. This strongly suggests that the molecular mechanism underlying the increased motor neuron excitability is shared 21 and 60 days postinjury, whereas the response is more extensive 60 days postinjury, supporting the observed progressive increase in the late-phase hyperreflexia (Fig. 1).

Ion channels

REGULATION OF INWARD CURRENTS.

Our study shows that the expression of genes coding for both sodium and calcium channels undergoes changes as a response to the injury. It is not known which type of L-type channels is involved in generating the plateau in normal animals, although it has been suggested that Cav1.3 may conduct the persistent Ca2+ current in rat motor neurons (Li and Bennett 2003). Based on immunohistochemistry the Cav1.2 channel has been shown to be up-regulated postinjury, compared with sham-operated animals with open dura (Anelli et al. 2007). Here we find that of the L-type channels the gene coding for Cav1.3 is down-regulated, but we observed no expressional change in the gene coding for Cav1.2. It therefore seems unlikely that it is an increased amount of Ca2+ channel α subunits that are directly responsible for the expression of the Ca2+ part of the plateau in the injury state. Although the reports by the Bennett group focus on the supersensitivity to monoaminergic transmitters and their agonists following transection (Harvey et al. 2006a; Li et al. 2007), they also present evidence that suggests that the returning persistent inward currents following spinal transection reach higher values than what can be achieved in the control animals even after maximal monoaminergic facilitation. This is the case both for the Na+ PIC (Harvey et al. 2006c) and the Ca2+ PIC (Li et al. 2007). However, the conductance kinetics and activation curves of α subunits can be modulated, both through physical modulations such as phosphorylation or through altered composition of the channel complex with its ancillary subunits. In light of this it therefore seems more relevant to focus on the modulatory mechanisms affecting the channels conducting the persistent inward current rather than on the expression of the pore-forming α-subunits themselves.

Interestingly, we find the genes coding for the calcium channel β subunits to be down-regulated, which could indicate both a regulation of L-type channel conductance dynamics as well as an altered expression in the membrane. Calcium β subunits can have facilitatory effect on the persistent inward current by shifting the activation curve to more hyperpolarized potentials. The β subunits also interact with other proteins involved in second-messenger signaling pathways as well as cytoskeletal structures, causing reorganization and incorporation of the target channels into the membrane and thus increasing the density of the channels conducting persistent Ca2+current (Arikkath and Campbell 2003; Catterall et al. 2005b; Dolphin 2003). Both Cabnb22) and Cacnb44) are down-regulated, suggesting a modulation of the Ca2+ channel gating and membrane targeting. Interestingly, it has been shown that Calmodulin (CaM) can induce membrane targeting and gating of Cav1.2 in the absence of the β subunits, effectively resulting in slow inactivation of the channel (Ravindran et al. 2008). CaM is known to be tethered constitutively to the L-type Ca2+ channels Cav1.2 and Cav1.3, where it works as a calcium detector responsible for both Ca2+-dependent facilitation and inactivation (Halling et al. 2005; Perrier et al. 2000; Peterson et al. 1999), balancing a positive and negative feedback regulation of the persistent inward current, probably acting on different timescales to prevent Ca2+ overload as well as refine channel signaling. The functional membrane targeting and gating of Cav1.2 by CaM works through another mechanism independent of Ca2+ binding, but relying on relatively high levels of CaM.

The up-regulation of calmodulin (Calm1) together with the down-regulation of the β subunits could therefore indicate an altered membrane targeting of the Ca2+ channels with a potentially slower inactivation. The up-regulation of a gene coding for an alternative Ca2+ binding protein (Cabp1), which affects the inactivation of L-type channels, could also play a part in the observed Ca2+-mediated persistent inward current seen in the chronic spinal phase. Cabp1 codes for a neuronal calmodulin-like Ca2+ binding protein that has been described to substitute for calmodulin as the Ca2+ sensor of both Cav1.2 and Cav1.3, prolonging Ca2+ currents by preventing Ca2+-dependent inactivation (Cui et al. 2007; Zhou et al. 2004). The up-regulation of the gene coding for CaM kinase II gamma (Camk2n1) further suggests that Cav1.2 and/or Cav1.3 conductivity may be enhanced. CaM kinase II has been shown to cause a negative shift in L-type Ca2+ current activation (Calin-Jageman and Lee 2008; Gao et al. 2006). Persistent inward Ca2+ currents have also been shown to be promoted by stimulation of PI3K (Viard et al. 2004) and PKA pathways (Cav1.3; Dai et al. 2009; Qu et al. 2005), whereas PKC inhibits Cav1.3 (Baroudi et al. 2006). The concomitant up- and down-regulation of genes associated with PKA and PKC and down-regulation of their counterbalancing phosphatases (Table 3) therefore suggest a conglomerate modulatory effect by these on the persistent inward calcium current, effectively increasing conductance without affecting the channel composition.

One other gene coding for an α Ca2+-channel subunit not directly related to persistent Ca2+current is up-regulated: the N-type channel Cav2.2. This channel has been implicated with synaptic release (Pardo et al. 2006; Schenning et al. 2006) and could therefore be related to the apparent increase in the machinery for acetylcholine release (see regulation of excitatory synaptic transmission).

No separate sodium ion channel that solely conducts persistent inward currents has been described. It has therefore been hypothesized that the persistent inward currents conducted by sodium ion channels most likely are mediated through the same channels that conduct transient currents, which are thought to have an additional slowly inactivating state (Ulbricht 2005). This additional conductive state is also modulated by the β sodium channel subunits (Catterall et al. 2005a; Isom et al. 1995; Zhou and Goldin 2004). Several genes pertaining to sodium channels were found to undergo regulation as a response to injury. The genes coding for α channel subunits Nav1.2 and Nav1.7 are both down-regulated. The sodium channel Nav1.6 is the major component of the sodium current of spinal neurons of the rodent, but the expression of the gene coding for Nav1.6 was unaltered. PKA and PKC are known to reduce the current in Nav1.2 but they have no reported regulatory effect on Nav1.6 (Cantrell and Catterall 2001; Dai et al. 2009). The only reported regulation of Nav1.6 is by MAP kinase p38-α (Wittmack et al. 2005), which was not affected by spinalization. Thus any effect on this channel appears to reside in the regulation of its subunits.

The up-regulation of the gene coding for the sodium β1 subunit (Scnb1) together with the down-regulation of the gene coding for the sodium β3 subunit (Scnb3) indicate altered intracellular trafficking of sodium channels as well as modulation of their conductance kinetics. Experiments have shown that coexpression of either β1 or β3 with sodium channel α subunits can shift the steady-state inactivation curve as well as the voltage dependence of channel activation in the hyperpolarizing direction (Isom et al. 1995; Meadows et al. 2002). β1 also causes an additional increase in the expression of the α subunits in the plasma membrane not observed for β3 in mammalian cell lines. The ostensible larger expression of sodium channels in the plasma membrane could therefore increase overall sodium conductances and a concomitant redirection of channel trafficking to soma and dendrites, rather than to axons and terminals, could also increase the role of sodium persistent inward currents.

In summary, from previously published results showing nonconditional expression of persistent inward currents in motor neurons through both Ca2+ and Na+ channels in chronic spinal animals (Li and Bennett 2003), we might have expected to find an up-regulation of the genes coding for the pore-forming α subunits of Cav 1.3 and the Na+ channels that have a persistent component. Our results certainly demonstrate that the changes in these channel complexes are far more complex and not least targeting modulatory subunits and intracellular pathways.

REGULATION OF OUTWARD CURRENTS.

Only one of the DE genes belonging to the family of chloride channels are involved in membrane conductance, i.e., a calcium-sensitive chloride channel Clca6 (Piirsoo et al. 2009); this gene is up-regulated. Little is known about the electrophysiological properties and the functional significance for neuronal signaling of this class of channels (Hartzell et al. 2005), but Ca2+-activated Cl channels have been implicated with the switch of action potential afterhyperpolarization to afterdepolarization in axotomized sympathetic ganglion cells (Sanchez-Vives and Gallego 1994) and vagal ganglion neurons (Lancaster et al. 2002). This depends on the reversal potential of Cl, which could be affected toward a more depolarizing potential in motor neurons postinjury, possibly contributing to the observed increase in excitability (see regulation of inhibitory synaptic transmission). The two remaining chloride channel genes are not known to influence membrane conductance, but are expressed in subcellular organelles (Clcc1 and Clcn3).

The potassium channels comprise the largest ion channel family that was regulated. A little more than half of the genes coding for potassium channels are down-regulated (see Table 1). Among these we find the gene Kcnn2, which codes for a small-conductance calcium-activated potassium channel that generates the afterhyperpolarization in motor neurons. This conductance has been shown to be blocked by serotonin in motor neurons, leading to increased expression of plateaux (Grunnet et al. 2004; Hounsgaard and Kiehn 1989). Its down-regulation could therefore signify an increased ability to generate plateaux or at least contribute to a higher input resistance, which has been reported in Harvey et al. (2006a). The remaining significantly DE genes relating to potassium channels suggest a rather complex regulation of potassium currents in motor neurons postinjury and the appearance of plateaux does not seem to be a result of a clear reduction in overall outward K+ currents.

Neurotransmitter receptors

REGULATION OF EXCITATORY SYNAPTIC TRANSMISSION.

Several genes relating to ionotropic glutamate receptors were regulated after injury. The gene coding for one kainate receptor (Grik5) was up-regulated, whereas the gene coding for the AMPA receptor (Gria4) was down-regulated. Another gene up-regulated as a response to injury could affect the expression of AMPA receptors in the membrane, Cacng22). It codes for a transmembrane AMPA receptor regulatory protein (TARP) (Chen et al. 2007) and its up-regulation suggests an increase of AMPA receptor trafficking as well as in its expression in the plasma membrane. Most interesting among the glutamate receptor regulations is the apparent strong regulation of the NMDA-receptor complex. Four genes relating to this complex are regulated as a response to injury: Grin1 (NR1), Grin3b (NR3B), Grina (NRA1), and Grinl1a (GL1AD). The up-regulation of the gene for NR1 suggests an increase in the expression of the channel itself and promotes increased sensitivity to synaptic glutamate transmission. Grinl1a is part of a complex transcript unit that is known to modulate NR1 and seems to facilitate glutamatergic signal transduction and also possibly to serve a protective role against glutamate-mediated excitotoxicity (Roginski et al. 2008). NR3B is a motor-neuron–specific NMDA-receptor modulatory subunit that reduces the NMDA conductance as well as calcium permeability (Matsuda et al. 2003). NRA1 also interacts with the NMDA-receptor complex, but its role in modulating signal transmission is not known. Src protein tyrosine kinase enhances and potentiates NMDA current, whereas protein tyrosine phosphatases (PTPs) have an opposing action (Salter and Kalia 2004). Although we found no change in Src there was a clear down-regulation of PTPs pathways, which potentially could further enhance the NMDA-receptor–mediated activity. Thus the overall regulation of the NMDA complex suggests a rather intricate regulation, with increased sensitivity to glutamate resulting in increased signal transmission at the same time as the Ca2+ permeability decreases, possibly serving a neuroprotective role.

For acetylcholine receptors the up-regulation of the genes coding for the nicotinic receptors Chrna44) and Chrna66) suggests an increased sensitivity to acetylcholine transmission in the motor neurons. Three other genes related to acetylcholine transmission are DE: Nrg1, Cuta, and Slc18a3 (VAChT). Slc18a3 and Cuta are up-regulated, whereas the two DE probe sets of Nrg1 indicate a regulation of splice variants, where one is down-regulated and another is up-regulated. Nrg1 is known as a gene coding for a transynaptic growth factor that increases acetylcholine receptor activity, stimulating the expression of acetylcholine receptor genes, the insertion of the receptors into the plasma membrane, and their assembly into clusters (Falls 2003; Sandrock Jr et al. 1997). Thus the altered expression of an Ngr1 splice variant might play a role in the up-regulation of Chrna4 and Chrna6. The neuronal Ca2+ sensor NCS-1 (Freq) has been shown to have a facilitating effect on synaptic transmission (Hilfiker 2003) and the up-regulation of its gene together with the increase of the gene coding for the acetylcholine vesicular transporter, VAChT, indicates increased transmitter release both at central synapses and at the neuromuscular junctions. Together, these regulations of genes relating to both pre- and postsynaptic cholinergic transmission suggest an increased acetylcholine signaling at central synapses that may enhance the recurrent excitation among motor neurons, both by motor axon collaterals terminating directly onto motor neurons (Nishimaru et al. 2005) and by a polysynaptic positive feedback loop, which was recently described (Machacek and Hochman 2006). The latter pathway seems to be unmasked by the presence of noradrenaline. Local release of noradrenaline has been described recently in the spinal cord following transection, where a group of cells seems to absorb this transmitter from the bloodstream through the damaged blood brain barrier, subsequently releasing it in the spinal cord (Cohen et al. 2009; Rank et al. 2007, 2008). This might provide a mechanism that could unmask the polysynaptic feedback loop in the injured cord, creating a very powerful positive feedback loop between motor neurons. We also note that this feedback mechanism can work via the peripheral synapses through the reflex loop, where an increased cholinergic release at the neuromuscular junction that results in enhanced muscle tone and activation of Ia afferents projecting directly onto motor neurons could provide a similar powerful source of feedback excitation.

REGULATION OF INHIBITORY SYNAPTIC TRANSMISSION.

Genes relating to GABA receptors underwent a striking regulation that seems to target GABAA receptors and their expression in the membrane. Only one gene not relating to this family is DE, a gene coding for a GABAB receptor (Gabbr1), which is up-regulated. All the DE genes coding for GABAA receptor subunits are down-regulated: two α subunits (Gabra1 and Gabra5) and a γ subunit (Gabg2). There are six members of the GABAA α receptor family (α1–α6), which form functional receptor complexes in combination with β and γ subunits. Their distribution have been described in spinal interneurons (Geiman et al. 2002), but there are presently limited data available on their distribution in spinal motor neurons. The γ2 is ubiquitously expressed throughout the brain and presumably participates in most functional combinations of the GABAA-receptor complex (Sieghart and Sperk 2002). Its down-regulation together with the genes coding for the α1 and α5 GABAA subunits suggest a general down-regulation of GABAA-receptor complexes in the motor neurons postinjury. The GABAA-receptor subunit γ2 also interacts with the GABARAP family (GABAA receptor interacting proteins) that are involved in GABAA-receptor trafficking and clustering (Chen and Olsen 2007) and it has previously been shown to be down-regulated in postural motor neurons postinjury in the neonatal rat (Khristy et al. 2009). The DE of the genes Gabrg2, Gabarapl1, and Gabarapl2 indicates a strong regulation of GABAA location and membrane expression. Trak2 is also involved in GABAA-receptor trafficking, further stressing that trafficking is subject to a substantial regulation in response to the injury for this receptor subfamily. Together this response points to reduced GABAergic transmission and a substantial regulation of GABAA receptors. The intracellular regulation of GABAA receptors is complex and depends on the subunit composition of the receptor complex. The functional effect of PKA and PKC phosphorylation of Ser/Thr residues can either increase or decrease the receptor transmission depending on the identity of the β subunit in the receptor complex, i.e., β1, β2, or β3 (Poisbeau et al. 1999). Phosphorylation of tyrosine residues by Src has been shown to potentiate the GABAA-receptor function (Moss et al. 1995). Since little is known about the exact subunit composition of GABAA-receptor complexes of motor neurons, it seems difficult to predict the conglomerate modulatory effect on the GABAA-receptor function based on the observed complex regulation of the genes associated with these intracellular pathways. It has also been shown in the rat tail model that the reflex response in the chronic phase lacks an initial inhibitory phase present in normal animals, indicating that part of the reason for the pathologically long-lasting motor discharge may be due to a lack of inhibition failing to terminate the activity (Bennett et al. 1999; Li et al. 2004b). In combination with our present findings this implies that regulation of the GABAergic system plays a part in the chronic pathological state. Interestingly, a recent modeling study including a small network of motor neurons and recurrent inhibition through Renshaw cells has shown that a selective decrease in the GABAA inhibition can unmask motor neuron plateau potentials (Venugopal et al. 2009), suggesting a role for the observed regulation in GABAA receptors in the expression of plateaux in the injury state.

The effect of inhibitory synapses can be further altered if the reversal potential for chloride is changed indirectly, leading to increased motor neuron excitability. Recent experiments have shown that the K+-Cl cotransporter 2 (KCC2) responsible for Cl extrusion is down-regulated as a response to spinal cord injury, effectively decreasing the reversal potential for chloride and pushing the effect of chloride channel activation toward more depolarizing potentials (Jean-Xavier et al. 2006; Vinay and Jean-Xavier 2008). The gene for KCC2 is not included in the Ensembl annotations used to filter the probe sets and could therefore not be investigated in the present study. Two other Cl transporters have been shown to accumulate Cl in the embryonic motor neurons, Na-K-Cl cotransporter 1 (NKCC1) and anion exchange protein 3 (EA3), and their up-regulation could therefore also shift the Cl reversal potential (Gonzalez-Islas et al. 2009). We find no change in transcript levels for the gene coding for NKCC1 (Slc2a2), but the gene coding for EA3 (Slc4a3) seems to be subject to up-regulation as a response to the injury. The combined effect of KCC2 down-regulation with AE3 up-regulation could very well shift the reversal potential for Cl increasing the excitability of motor neurons through GABAergic as well as glycinergic synaptic transmission from spinal interneurons.

REGULATION OF NEUROMODULATORY RECEPTORS.

Increased sensitivity to 5-HT (Barbeau and Bedard 1981; Harvey et al. 2006a; Li et al. 2007) and noradrenalin (Rank et al. 2007) have been observed after spinalization. This hypersensitivity is thought to contribute to the expression of plateaux in the chronic spinal phase through serotonergic and noradrenergic mechanisms since 10% of 5-HT and noradrenalin levels remain in the spinal cord in the chronic spinalized state (Cohen et al. 2009; Hadjiconstantinou et al. 1984; Newton and Hamill 1988; Rank et al. 2008). One gene coding for the adrenergic α receptor 1D (Adra1d) is up-regulated together with the gene coding for an adrenergic receptor kinase (Adrbk1), indicating a role for these in the noradrenaline hypersensitivity. It has recently been shown that the activation of the adrenergic α1 receptor result in a strong activation of both Ca2+ and Na+ persistent inward currents in the rat tail model (Harvey et al. 2006b; Rank et al. 2007), suggesting a strong role for the up-regulation of the gene Adra1d in increasing the excitability of motor neurons postinjury. One gene coding for a serotonin receptor, Htr2b, is down-regulated. The 5-HT2b receptor has been described in frog and respiratory rat motor neurons, where it contributes to increased activity (Gunther et al. 2006; Holohean and Hackman 2004). The down-regulation of the gene coding for this receptor seems to suggest a decreased sensitivity to 5-HT, although the posttranscriptional modulation of other genes relating to the serotonin receptor family, possibly remaining undetected in the present work, has been shown to increase motor neuron excitability (Murray et al. 2008) (see Regulation of splice variants). There has been little focus on the role of dopamine transmission in the expression of plateaux in the postinjury phase, but the expression of its receptors in the motor neurons of the spinal cord and its influence on motor neuron excitability have been documented (Han et al. 2007; Madriaga et al. 2004; Zhu et al. 2007). The up-regulation of two genes coding for the dopamine receptor 4 (Drd4) and a D1 dopamine receptor-interacting protein (Caly), could therefore contribute to an increased excitability of the motor neurons.

Regulation of splice variants

One gene, Nrg1, has two probe sets with opposing direction of expression, indicating the differential expression of different splice variant transcripts. Since the probe sets on the RAT230_2.0 chip are not designed to target specific exons of their target transcript, we remapped the probe sets to Ensembl exons and found that Nrg1 may undergo splice regulation. Because probes unfortunately do not cover all exons of their target transcript it is not possible to do an exhaustive analysis of exon regulation with the microarray design used in the present study. However, it is worth noting that many ion channels and their ancillary subunits have several splice variant transcripts and if different splice forms are expressed as a response to injury it might very well remain undetected if these are not targeted specifically and the transcript levels otherwise remain at a constant level. A recent abstract describes such a mechanism in sacral motor neurons as a response to injury, where the expression of two different transcript splice variants of the 5-HT2C receptor resulting in two different receptor isoforms are subject to regulation (Murray et al. 2008). This study demonstrates a significant increase in expression of the major constitutively active isoform of the 5-HT2C receptor paralleled by a simultaneous decrease of a nonconstitutively active isoform following the spinal transection. We did not see any significant change in the gene expression of the 5-HT2C receptor in the present material–thus this report raises the question as to which extent functionally significant changes of different splice variants may hide behind an apparent unchanged expression observed for the Affymetrix probes targeting a gene. This functional change is primarily caused by posttranscriptional modulation, leaving this modulation undetected in the present study if the underlying regulation of transcript expression remains unchanged. However, if the transcript splice variants are subject to expressional regulation, the changes in their transcript levels will be detected, irrespective of the splice form expressed, and we therefore remain assured that we do detect the majority of the interesting changes in transcript regulation as a response to injury.

Conclusion

In the present study the global transcriptional response to spinal cord injury has been examined for a specific cell population in the rat spinal cord, the motor neurons. The focus of the study was to correlate the changes in gene expression of this cell population with the pathological state of injury-induced spasticity—i.e., to elucidate the molecular mechanisms behind the increased motor neuron excitability and the appearance of plateaux in motor neurons following injury. Our analysis shows a complex regulation of genes related to modulation of inward depolarizing currents and a massive regulation of outward currents. Receptors involved in ionotropic synaptic transmission are subject to even more substantial lesion-related regulation. For all of the channels and receptors that are subject to regulation the nonpore-forming modulatory subunits as well as intracellular pathways that affect their function are intricately involved in the injury response. The channel ancillary subunits affect conductivity of existing channels and channel clustering in the plasma membrane and, until now, have not been associated with the enhanced excitability of the motor neurons in the injury state. These subunits can therefore be very potent candidates for some of the observed changes relating to the expression of plateaux and may serve as new targets for future studies of spinal cord injury.

Footnotes

  • 1 The online version of this article contains supplemental data.

REFERENCES

View Abstract