Gene expression profiling of experimental traumatic spinal cord injury as a function of distance from impact site and injury severity
Andrea De Biase1,*,
Susan M. Knoblach1,*,
Simone Di Giovanni2,*,
Chenguang Fan1,
Annamaria Molon1,
Eric P. Hoffman1 and
Alan I. Faden2
1 Children's National Medical Center, Center for Genetic Medicine
2 Department of Neuroscience, Georgetown University School of Medicine, Washington, District of Columbia
 |
ABSTRACT
|
---|
Changes in gene expression contribute to pathophysiological alterations following spinal cord injury (SCI). We examined gene expression over time (4 h, 24 h, 7 days) at the impact site, as well as rostral and caudal regions, following mild, moderate, or severe contusion SCI in rats. High-density oligonucleotide microarrays were used that included
27,000 genes/ESTs (Affymetrix RG-U34; A, B and C arrays), together with multiple analyses (MAS 5.0, dChip). Alterations after mild injury were relatively rapid (4 and 24 h), whereas they were delayed and prolonged after severe injury (24 h and 7 days). The number and magnitude of gene expression changes were greatest at the injury site after moderate injury and increased in rostral and caudal regions as a function of injury severity. Sham surgery resulted in expression changes that were similar to mild injury, suggesting the importance of using time-linked surgical controls as well as naive animals for these kinds of studies. Expression of many genes and ESTs was altered; these were classified functionally based on ontology. Overall representation of these functional classes varied with distance from the site of injury and injury severity, as did the individual genes that contributed to each functional class. Different clustering approaches were used to identify changes in neuronal-specific genes and several transcription factors that have not previously been associated with SCI. This study represents the most comprehensive evaluation of gene expression changes after SCI to date. The results underscore the power of microarray approaches to reveal global genomic responses as well as changes in particular gene clusters and/or families that may be important in the secondary injury cascade.
trauma; microarray; central nervous system; mRNA; rat; contusion
 |
INTRODUCTION
|
---|
TRAUMATIC SPINAL CORD INJURY (SCI) induces widespread, coordinated changes in gene expression (3, 11, 1517, 19, 23, 64). These may be associated with pathogenic auto-destructive events such as hemorrhage, metabolic failure, inflammatory/immune activation, loss of ionic homeostasis, lipid degradation, production of free radicals, and neurotransmitter/neuromodulator imbalances (for review see Refs. 8, 26). Such alterations may contribute to death of neurons and oligodendroglial cells, glial proliferation (scarring), demyelination, and axonal loss. Alternatively, reactive biochemical and expression changes that are neuroprotective and/or promote recovery (regeneration/plasticity) may also be induced (17, 18, 26, 51).
Recently, many investigators have utilized high-throughput methods such as differential-display PCR, subtractive PCR, or microarrays to delineate global changes in gene expression patterns that may underlie the complex response to SCI (3, 11, 22, 58, 62, 64, 67). Indeed, several groups have used spotted cDNA or high-density oligonucleotide microarray platforms to evaluate expression of from
1,200 to 15,000 genes. Such studies have revealed global changes in expression of multiple genes involved in inflammatory/immune (3, 11) and wound-healing responses to injury (64). Others have used microarrays to identify expression clusters that led to studies of specific novel pathways that have not previously been associated with SCI, including cell cycle genes involved in neuronal cell death (17) as well as genes involved in axonal plasticity and remodeling (15, 16). The majority of such studies have focused on mild and/or moderate levels of SCI. Of these, only a few have examined regions distant from the site of impact (3, 11). No study has evaluated the potential effects of injury severity.
We describe an extensive gene expression profiling study of rat experimental SCI that assessed changes in gene expression after three levels of injury (mild, moderate, and severe) and across several spinal cord regions (injury site and adjacent rostral and caudal regions). The analysis utilized the Affymetrix complete rat genome GeneChip set (
27,000 genes; GeneChips RG-U34A, RG-U34B, RG-U34C), which includes not only full-length or annotated genes (A array), but also thousands of expressed sequence tags (ESTs) (B and C arrays). In addition, the data were analyzed with two different probe set algorithms.
 |
MATERIALS AND METHODS
|
---|
Experimental SCI and expression profiling.
SCI was performed in rats as previously described (17). Briefly, male Sprague-Dawley rats (275325 g) were anesthetized with intraperitoneal pentobarbital sodium (65 mg/kg). Injury was induced using a weight drop method (10-g weight dropped from 17.5 mm, 30 mm, and 50 mm, respectively, for the three levels of injury) that included stabilization of the vertebral column before impact. The site of impact was vertebral T9, which corresponds to spinal T10T11. At the appropriate time after injury, animals were deeply anesthetized with pentobarbital sodium (100 mg/kg) and decapitated. Complete laminectomy of the entire vertebral canal was rapidly performed, and the cord was dissected into injury, rostral, or caudal sections (each 1 cm long) (see Fig. 1). These were immediately removed and frozen on dry ice. The injury site and both rostral and caudal regions were collected from four injured and two sham-operated rats (that received only laminectomy surgery) for mild and severe injury levels at 4, 24, and 168 h (7 days) after injury. Only the injury site was collected from rats subjected to moderate SCI, and samples were taken at 0.5, 4, 12, 24, 48, 72, and 168 h after injury. Two naive controls (rats that did not undergo any surgical procedure or any anesthesia at the time of death) were also included.

View larger version (69K):
[in this window]
[in a new window]
|
Fig. 1. Functional and morphological outcomes after graded weight-drop contusion injury in the rat. A: representative histological images from the site of impact 28 days after mild, moderate, or severe contusion injury. B: recovery of locomotor function observed after mild (n = 10), moderate (n = 12), or severe (n = 11) injury as assessed over time on the Basso, Beattie, Bresnahan (BBB) scale (7). C: neuroanatomical depiction of the vertebral and spinal segments (with respect to injury) represented in the "rostral," "injury," and "caudal" samples. Vertebral segments are indicated in white type, and spinal segments are in yellow. An asterisk marks the site of impact. Three adjacent spinal cord segments, each 1 cm long (indicated on ruler), were sampled. Arrows indicate boundaries for each sample, where "rostral" included spinal segments T5T8, and "injury" or "caudal" spinal segments T9T12 and T12L2, respectively.
|
|
In parallel, additional animals were prepared (mild, n = 10; moderate, n = 12; severe, n = 11) to examine the effects of the injury parameters used on locomotor recovery and histopathology. Performance on the Basso, Beattie, and Bresnahan (BBB) locomotor test was assessed from 128 days after injury of each severity, as previously detailed (7). At 28 days after injury animals were anesthetized and perfused, and their cords were removed and processed using standard paraffin embedding and hemotoxylin-eosine staining, to assess pathology. All animal procedures and experiments complied fully with the principles set forth in the "Guide for the Care and Use of Laboratory Animals" prepared by the Committee on Care and Use of Laboratory Animals of the Institute of Laboratory Animal Resources, National Research Council [DHEW Pub. No. (NIH) 8523, 1985].
Microarray (GeneChip) quality control.
Expression profiling was performed as described previously (17). Briefly, RNA was extracted from each cord sample individually using TRIzol reagent (Invitrogen). Seven micrograms of total RNA was used for cDNA and biotinylated cRNA synthesis and fragmentation as detailed in the manufacturer's protocol (Affymetrix). The same amount of total RNA (7 µg) was used to generate gene expression profiles regardless of injury level or spinal cord location. Affymetrix rat RG-U34A, RG-U34B, and RG-U34C arrays were used. Each spinal cord region from each individual animal was hybridized to all three RG-U34 arrays (A, B, or C).
Stringent quality control methods were employed as previously published (17) and detailed on our Public Expression Profiling Resource web site: http://pepr.cnmcresearch.org/home.do. Each array fulfilled the following quality control measures: cRNA fold changes from 5 to 10, scaling factor from 0.3 to 1.5, percentage of "present" (P) calls from 40 to 55%, and average signal intensity levels between 900 and 1,100. Housekeeping genes and internal probe set controls showed >80% present calls, consistent values, and 5'/3' ratios <3.
Experimental normalization, data filtering, and statistical analysis of gene expression profiles generated with MAS 5.0 and dChip.
Prior to any type of experimental-based normalization, array normalization was applied to generate expression values using two different algorithms: dChip and MAS 5.0. When we used MAS 5.0, an array-to-array normalization using a scaling factor of 800 was employed, generating MAS 5.0-type CHP files that were later loaded into GeneSpring for experimental normalization and analysis. In parallel, the same arrays were analyzed with the dChip probe set algorithm (34, 35) to establish a project-based normalization, rather than single-array normalization, generating new "dChip-type" expression values "homologous" to the CHP files generated with MAS 5.0. These new dChip-type CHP files were also loaded into GeneSpring for experimental normalization and analysis. Finally, data from both algorithms were analyzed using GeneSpring software (Silicon Genetics) and normalized to the average of the naive signal intensities for each gene (experimental normalization). Experimental normalization to naive was entirely arbitrary, and additional types of normalization can be performed online on our publicly available SAS server database (http://sas.cnmcresearch.org/). Genes that were significantly altered under both algorithms were retained for functional analysis, whereas genes that were altered only in one or the other algorithm were not included in this study.
Experiment normalization was performed, prior to any type of statistical or fold change filtering, by normalizing arrays from injured and sham controls to the mean of the two arrays from naive animals considered as the baseline gene expression level. Normalized data were then compared for differential gene expression analysis across time points between sham and injured groups. We have recently shown that use of signal intensity values, together with a "present call" noise filter achieves an excellent signal/noise balance relative to other probe set analysis methods (56). Thus, data analyses were limited to probe sets that showed one or more "present" (P "calls") across the complete data set. Initial data analysis also included a fold change filter of
1.5 (increase or decrease) relative to sham-operated animals (Affymetrix MAS 5.0).
Statistical analysis was performed on GeneSpring software, using a Welch ANOVA t-test P value of <0.01 between sham and injured time-matched groups. Although a P value of <0.01 alone would give many false positives, the combination of present call filters, fold change thresholds, and P value thresholds eliminates most false positives that are obtained with only P < 0.01.
Genes having at least one "present" call across the entire data set, a fold change ratio >1.5 and a significant expression (P < 0.01) were subsequently sorted into functional groups, using annotations provided in the Database for Annotation, Visualization and Integrated Discovery (DAVID) (National Institute for Allergy and Infectious Diseases; http://david.niaid.nih.gov/david) and SwissProt/TrEMBL (http://www.expasy.org). In cases of genes with multiple functions/activities, assignment was based on the most frequently attributed activity or to any central nervous system (CNS)- or SCI-specific activity, if stated.
Real-time RT-PCR.
In previous reports, real-time PCR data was described that validated selected transcripts from different clusters derived from this data set (1517). For the present study, additional real-time RT-PCR was performed to validate additional genes from several major functional classes altered by injury. Three injured and two sham rats (biological replicates) were used in triplicate (experimental replicates) for RT-PCR. Animals were different from the ones used in the array analysis. Genes included in the analysis were selected from a wide range of expression changes (small to large) to assess the accuracy of such validation across this variable. RT-PCR was performed for the following functional classifications/genes: inflammatory genes (Tgfb1), neurotransmission (KCh, Scn1a, P2rx4), transcription factors (c-fos), extracellular matrix (Col1a1), and growth factors/differentiation (Syngr1). The following primers were used: Tgfb1 754F 5'-CACCAACCCGCGTGCTAATGG[FAM]G-3', Tgfb1 856R 5'-GCTCTGCACGGGACAGCAAT-3'; KCh 2469 5'-CTACGTGCTTCCAGCATTCTTTGACACG[FAM]AG-3', KCh 2546R 5'-CCCATCCTTAGTCCCTGACCT-3'; Scn1a 7395F 5'-GACAGCGAGAAACGTAGAGTGCAAGCTG[FAM]C-3', Scn1a 7493 5'-TGAGATTCTGAGTGGTGGGAGAA-3'; c-fos 1393F 5'-CACGCGGGAGGACCTTATCTGTGCG[FAM]G-3', c-fos 1446R 5'-CCGGAAGAGGTGAGGACTGG-3', Col1a1 395FL 5'-CACAGTGGGGTCAGATATGGGGTCACTG5G-3', Col1a1 479R 5'-GGGCACCTGCTCATCCATGT-3'; P2rx4 114FL 5'CACTTTCATCCGCAGCCGTAAAG5G-3', P2rx4 237R 5'-TGTTGGTCACAGCCACACCT-3'; synaptogyrin 78FL 5'-CACCACGCATGAGTGAGCCAGTGG5G-3', synaptogyrin 151R 5'-gggagtccacactgtcagcag-3'.
Fluorophore-labeled LUX primers (forward) and their unlabeled counterparts (reverse) were purchased from Invitrogen. LUX primers were designed within Affymetrix probe set sequences for each gene, and all primers were designed using LUX Designer software (http://www.invitrogen.com/lux). A multiplex PCR method that combined each experimental gene with the housekeeping gene GAPDH in each PCR mix was used. For each sample, 20 µl PCR mix contained 2 µl cDNA (first diluted 1:10 after reverse transcription), 200 nM of each gene-specific primer (two pairs for multiplex PCR), 1x Platinum Quantitative PCR SuperMix-UDG (Invitrogen), and 1x ROX reference dye (500 nM in SuperMix). PCR conditions were standard (Invitrogen, http://www.invitrogen.com/lux), and reactions were conducted in a 96-well spectrofluorometric thermal cycler (ABI Prism 7700 Sequence detector system, Applied Biosystems). Fluorescence was monitored during every PCR cycle at the annealing or extension step and during the post-PCR temperature ramp. Fold changes were measured according to manufacturer's instructions. GADPH was used as a reference and subtracted for net changes. Student's t-test was used to calculate P values.
 |
RESULTS
|
---|
Motor scores and histopathology.
In the parallel dose-response study, lesion size at the site of impact and surrounding white matter damage was increased with injury severity (Fig. 1A). In addition, severity-dependent locomotor differences on the BBB rating scale were observed over 28 days after injury (Fig. 1B). The dissection scheme for rostral, caudal, and injury sites is shown (Fig. 1C).
Database, Public Expression Profiling Resource (PEPR), and SAS visualization tool.
All microarray data are available on the Public Expression Profiling Resource web site (http://pepr.cnmcresearch.org/home.do). This website contains .DAT, .CEL, and .TXT files for every microarray as well as online time query and visualization tools to assist with analysis of the database. In addition to Supplemental Tables and Supplemental Figures for this report, the web site contains data from additional acute time points (0.5, 12, 48, and 72 h) (RG-U34; A, B, and C arrays). (The Supplemental Material for this article is also available at the Physiological Genomics web site.)1
In addition to PEPR, all biological transcripts on the RG-U34 A, B, and C arrays can be visualized across time, injury levels, and spinal cord locations in three-dimensional graphical format in our SAS server and visualization tool (http://sas.cnmcresearch.org/). Moreover, normalization of the entire data set to either naive or sham samples is made optional, and graphical outputs can be zoomed and rotated for better visualization. In addition to Supplemental Tables and Figures for this report (PEPR), both web sites contain data from all time points (RG-U34; A, B, and C arrays) and regions as detailed in MATERIALS AND METHODS. We have previously reported gene changes induced by mild injury at selected time points (17) and specific clusters involved in cell cycle, plasticity, and regeneration (15, 16). However, here we focus on a global analysis of selected time points after injury (4, 24, 168 h) that are widely studied with respect to the pathophysiology of secondary injury, and we focus on a few potentially related clusters.
Global changes in expression after SCI: effect of injury severity, spinal cord region, and laminectomy.
Hierarchical clustering and intensity maps revealed broad expression patterns related to injury severity, spinal cord region, and laminectomy alone (sham). Sham surgery, mild, moderate, and severe injury resulted in marked overall changes in expression profiles compared with naive animals (Fig. 2, AC). After mild injury, expression of many genes was increased, with the highest fold changes observed relatively early (4 and 24 h), followed by gradual reductions (although still significantly elevated compared with sham surgery) over time (Fig. 2A, boxed genes). After severe injury, the highest fold changes were comparatively delayed (24 h and 7 days) (Fig. 2B, solid boxes). Severe and moderate injuries generally showed the greatest number and magnitude of alterations; indeed, expression of some genes was directly proportional to injury severity (Fig. 2C). Expression of many genes was reduced at the injury site, but increased in rostral and caudal regions (Fig. 2B, dashed boxes).

View larger version (68K):
[in this window]
[in a new window]
|
Fig. 2. Gene tree illustrations of global changes in gene expression related to severity and region. In AC, time after injury, spinal cord region and intensity of fold changes are indicated on the right. Baseline/naive data are arbitrarily set at 1 (yellow at time 0), and fold changes for all other groups are respectively shown in red and blue according to the color bar on the right. Values <1 indicate decreased expression, and values >1 increased expression. A: significantly altered genes (Welch ANOVA, P < 0.01) for injured and sham samples relative to naive samples in the rostral, injury site, and caudal regions at 4 h, 24 h, and 7 days following mild injury. Ovals highlight examples of altered expression in both sham and injured groups, relative to naive animals. Boxes show examples of lower fold and fewer changes in rostral and caudal regions compared with the injury site. Boxes also highlight that highest fold changes were observed relatively early (4 and 24 h) after mild injury. B: significantly altered genes (Welch ANOVA, P < 0.01) for injured and sham samples relative to naive animals in the rostral region, injury site, and caudal region at 4 h, 24 h, and 7 days following severe injury. Dashed boxes highlight examples of genes with reduced expression at the injury site, but increased expression in the rostral and caudal regions, compared with respective shams. Solid boxes show examples of lower fold changes in rostral and caudal regions, compared with the injury site. They also illustrate that highest fold changes were mainly delayed (4 h and 7 days) after severe injury. C: comparison of significantly altered genes (Welch ANOVA, P < 0.01) at the injury site, across all 3 injury levels (mild, moderate, severe). Boxes highlight examples of genes whose expression (up or down, magnitude of change) varies with injury severity. Note that both moderate and severe injury resulted in greater changes (number and magnitude) than did mild injury. Also, changes in gene expression were greatest (in number and magnitude) after moderate injury and decreases in gene expression were greatest after severe injury.
|
|
Spinal cords from animals that received laminectomy only (sham injured controls) also showed overall changes in expression compared with naive animals. These alterations in expression induced by laminectomy alone extended not only from 4 h to 7 days after surgery, but also to rostral and caudal regions away from the site of impact (shown as example in mild injury Fig. 2A, ovals). Thus, laminectomy alone induced widespread relatively prolonged effects on gene expression that resembled a mild injury effect.
Table 1 shows data presented in Fig. 2 in terms of gene numbers. Since the intensity maps indicated that laminectomy alone altered gene expression, separate analyses compared injury-induced changes to either naive or to sham controls for each respective time point and region (Table 1, A). As expected from the intensity maps, comparison of injuries to sham-operated animals rather than naive animals reduced the number of genes that met criteria for significant changes in expression, particularly at the injury site (Table 1, A). Generally, more genes were altered after moderate or severe injury than after mild injury (Table 1, A), although such changes differed between regions. At the injury site, the number of changes increased, and then decreased, with progressive injury severity (from mild to moderate to severe). In rostral and caudal regions, the greatest number of changes was observed after severe injury. The direction of changes (up or down) depended on injury severity, with increased expression of most genes observed after mild or moderate injury and with decreased expression of genes predominating after severe injury (Table 1B).
View this table:
[in this window]
[in a new window]
|
Table 1. Number of significantly altered probe sets (along with significant increases and decreases) in injured samples, compared with naive or sham samples, for each injury severity, time, and spinal cord region
|
|
Functional classifications.
Aproximately 50% of all gene changes were significant by both the MAS 5.0 and dChip algorithms. Injury-induced changes in gene expression that met all significance criteria (significant by MAS 5.0 and dChip, significant vs. sham,
1.5-fold change) were classified into functional categories as detailed in MATERIALS AND METHODS. These were then grouped into gene lists with functional groups ranked by overall expression level and actual fold change values (Supplemental Tables S1S7). Functional classes with the largest number of genes altered by injury included those associated with metabolism, inflammation/immune responses, signal transduction, transcription/protein synthesis, and the cytoskeleton (Fig. 3).

View larger version (34K):
[in this window]
[in a new window]
|
Fig. 3. Schematic overview showing the distribution of significantly altered genes from each region and injury level compared with respective shams, after functional classification as defined in the text. Individual diagrams and numbers represent the total number of altered genes (compiled across time points) within each region/injury severity. Individual genes in each functional category are identified in the functional lists (Supplemental Tables S1S7), and on the Public Expression Profiling Resource web site (http://pepr.cnmcresearch.org/home.do). The area of each subdivision/"slice" within a pie diagram represents the number of genes associated with each functional category. Unclassified genes/ESTs are not included in this representation scheme.
|
|
There were no marked alterations in global representation of the functional classes (% individual functional group/all changes) that could be consistently associated with location relative to the injury site or to injury severity. However, these factors did alter representation of the classes such that they were not precisely identical. They varied slightly between different regions and injury severities, with the actual number and identity of genes within each class fluctuating widely with these variables (Fig. 3 and Supplemental Tables S1S7). Further sorting of the initial functional classes into subgroups of two or more functionally similar genes also failed to reveal any marked trends or patterns related to distance from the impact or injury severity (Supplemental Tables S1S7).
ESTs and RG-U34B/RG-U34C arrays.
Enough information was available on public databases to allow functional classification of
65% of the ESTs that were altered by SCI (see Supplemental Materials). After the functional classification was completed, we calculated how many classified genes/ESTs were represented on RG-U34B or RG-U34C arrays to gauge what level of information was obtained by including them in the analysis. We found that
50% of functionally classified ESTs were derived from RG-U34B or RG-U34C arrays (Table 2).
View this table:
[in this window]
[in a new window]
|
Table 2. Number of significantly altered, functionally classified genes from RG-U34A (A) vs. RG-U34B/RG-U34C (B, C) arrays across injury severity and region
|
|
Genes common to SCI profiling studies.
Table 3 summarizes published studies of expression profiling of traumatic SCI. Many gene expression changes that were initially described in these studies were also observed in the present experiment. Examples include specific genes associated with inflammation (i.e., cathepsins, interleukin-1ß, intercellular adhesion molecule-1) (3, 23), secretory/synaptic processes (SNAP proteins, synaptogyrin) (3, 11), and oxidative and/or stress-responses (Hsp70, Fos, metallothioniens) (11) (see Supplemental Material). Additional genes in each of these groups were identified with the RG-U34B and RG-U34C arrays (for example of genes that clustered with IL-1ß, see Supplemental Fig. S1). Several other functionally related gene subclasses were also identified that have not previously been a focus of reported SCI array data. These included genes involved in the ubiquitin cycle, G protein-coupled signal transduction, translation initiation/elongation, and RNA splicing (Supplemental Material).
Identification of neuronal-specific genes and a transcription factor binding cluster.
Highlighted below are some gene clusters that were significantly altered after injury but have not previously been associated with SCI. They serve as examples of genes that may represent potential candidates for further study.
An association was derived by initially comparing expression profiles of two potassium channel genes [Kcncl2 (M59980), Kcnk1 (AF022819)] that were previously associated with CNS injuries in multiple array studies (25) with that of two potassium channel genes that showed marked downregulation in our data set [Kcnc2 (X62829), Kcnc1 (X62840); Kcnc1 was also selected for RT-PCR analysis, described below] (Fig. 4). Next, we used the average signal of these related channel genes as an anchor for a new gene cluster (Fig. 5). The resultant cluster (correlation coefficient = 0.993) contained a large group of genes whose expression was inversely correlated with injury severity (Fig. 5). Many of these genes have important neuronal-specific functions (synuclein, brain glycogen phosphorylase, visinin-like 1, neurofilament, HuD, Ntab)(4, 5, 21, 30, 47, 48, 52). Some are already well-described (synuclein, neurofilament, GABA-B receptor), others are not (visinin-like, Ntab, HuD), and most have not previously been associated with SCI (synuclein, visinin-like, Ntab, HuD).

View larger version (18K):
[in this window]
[in a new window]
|
Fig. 4. Expression of potassium channels at the injury site decreased with increasing injury severity over time after mild, moderate, or severe injury. Time after injury at each severity level is represented on the x-axis, and normalized intensity of fold change ± SE is on the y-axis. Expression levels are color coded to individual channels (Kcncl2, Kcnk1, Kcnc2, Kcnc1) as indicated in key.
|
|

View larger version (33K):
[in this window]
[in a new window]
|
Fig. 5. Gene cluster that nucleated with alterations of potassium channels at the injury site over time after mild, moderate, or severe injury (correlation coefficient of 0.993). Normalized intensity of fold change ± SE vs. time after injury at each severity level is graphed, and the genes represented in the cluster are identified below, where bold indicates probe sets on RG-U34B or RG-U34C arrays. Eighteen members of the cluster are ESTs that could not be functionally classified. Probe sets for seven of these were located on RG-U34A arrays, and 11 were on RG-U34B or RG-U34C arrays.
|
|
A second cluster focused on transcription factors, since alterations in these might affect expression of multiple downstream target genes. The transcription factor Egr-2/Krox-20/SCIP/Oct-6 (U78102) was significantly altered after mild and severe injury and is involved in myelination, plasticity, and ischemic CNS injury (46, 57, 59). Therefore, we selected this gene to nucleate (R2 = 0.99) a cluster of transcription factors after mild injury. The cluster included Cebp
(M65149), Cited4 (AA900875), and Copeb (AI044674) (Fig. 6). Of these, it seemed that Cebp
might be the best candidate for further analysis, since it was the most responsive to injury and is one of a group of CCAAT enhancer binding proteins that are involved in inflammation, growth arrest, and stress response (Fig. 6). Overexpression of Cebpß in neuroblastoma cells increased expression of a specific group of target genes [ornithine decarboxylase (ODC), GRO, lipocalcin 2, decorin, and xanthine dehydrogenase] that may be involved in CNS injury (14). Cebpß is not present on the RG-U34 arrays, but it shows pleiotropic activity with Cebp
, so we looked at the potential relationship between Cebp
and these genes in our data set. Elevations in Cebp
correlated with Gro and preceded elevations of ODC and lipocalin 2 after mild and moderate injury (Fig. 7, A and B), but neither decorin nor xanthine dehydrogenase were increased after injury (Fig. 7, A and B).

View larger version (22K):
[in this window]
[in a new window]
|
Fig. 6. Expression of selected transcription factors not previously associated with spinal cord trauma increased at the injury site and in rostral and caudal regions following mild injury (correlation coefficient of 0.998). Alterations were still significant although less marked in the rostral and caudal regions. Accession numbers for each factor: Cebp (M65149), Cited4 (AA900875), Copeb (AI044674); Egr2 (AB032419).
|
|
Technical confirmation of microarray data with multiplex RT-PCR.
Using quantitative and multiplex real-time PCR, we have extensively confirmed our mild injury site data set in several previously published papers (1517). In the present study, we performed additional confirmation with multiplex PCR of seven genes from different functional groups (cell growth/differentiation, neurotransmission, inflammation, extracellular matrix, and transport) that were expressed over a range of levels (low to high). Fold changes were statistically significant (P < 0.01). We then compared the fold changes declared after analysis of the array data with each probe set algorithm to those detected by RT-PCR, and we found that fold changes in array data derived with either algorithm were confirmed by RT-PCR, although actual values were scaled differently (Table 4). Fold changes of dChip values correlated linearly with RT-PCR values (P = 0.0478), whereas MAS 5.0 values did not (P = 0.521) (Table 4), which suggests a closer alignment between dChip values and those from RT-PCR.
 |
DISCUSSION
|
---|
Comprehensive evaluation of changes in gene expression that contribute to secondary injury, tissue loss, and neurological dysfunction may lead to the development of novel interventions that promote recovery. To that end, we have undertaken an extensive study to assess temporal and spatial changes in expression after injuries of different severities. Our results indicate that both laminectomy and impact trauma result in prolonged changes in gene expression in the rostral and caudal regions, as well as the impact site. Laminectomy showed a profile consistent with mild injury. This unexpected observation depended upon the use both sham and naive controls, the combination of which has not been previously published in SCI microarray studies. The greatest number of gene changes was observed at the injury site after moderate injury. Perhaps surprisingly, both severe and mild injury led to relatively fewer changes in gene expression in this region. However, the number and magnitude of alterations in the rostral and caudal regions generally increased with injury severity. Moreover, decreases in gene expression correlated with severity in all regions. Major contributors to the global expression response included genes involved in metabolism, inflammation/immune response, signal transduction, transcription/protein synthesis, and cytoskeleton. Nevertheless, there was considerable heterogeneity in the proportional representation of these classes, as well as for other functional groups and individual genes, when responses to injury severity were compared spatially. Last, several specific clusters of genes were identified that may correlate with cell death and/or neuronal injury. In previous reports, we have focused on subsets of these data that are involved in cell cycle, synaptic plasticity, and regeneration (1517).
Regional changes as a function of injury severity.
At the site of impact, we observed that the number of genes altered by severe injury decreased from moderate to severe injury. This might reflect substantial cell death that reduced the number of surviving cells immediately adjacent to the epicenter (37). Thus, the decrease in gene expression at the injury site over time after severe injury is consistent with a substantially larger lesion, as verified histologically.
Gene expression changes in rostral and caudal regions following severe injury were robust, reflecting the fact that injury can cause substantial biochemical changes at a distance from the primary impact site. It is unclear why the region caudal to the lesion showed far more altered transcripts than the rostral region after severe injury. However, such changes may be attributed to differences in gray-to-white matter ratios and/or differences in blood flow and/or metabolic responses (38, 65).
It should be underscored that sham injury (laminectomy surgery alone) produced temporally extended (7 days) responses in the rostral and caudal regions that resembled mild injury. The fact that sham surgery alters gene expression emphasizes the importance of study designs that include sham controls that are temporally yoked to experimental injury, as well as the inclusion of naive controls. Not only do the naive controls help determine the magnitude of change in shams, but they can protect against false negatives arising from comparisons of experimental groups to sham groups, in cases where the sham response is robust.
Functional classifications and novel clusters.
Consistent with findings from other groups (3, 11, 64), transcripts from genes with roles in inflammation, transcription and protein synthesis, metabolism, and the cytoskeleton were well represented after injury; however, changes in genes involved in oxidative stress, apoptosis, cell growth/differentiation, proteolysis, cell adhesion/extracellular matrix, and transport were less common. Representation of most of the functional classifications was broadly similar but also showed considerable severity and spatially dependent heterogeneity, rather than a specific injury or spatially dependent pattern. Our analysis rejected all but the most robust differences (see later DISCUSSION) and therefore is less sensitive to smaller yet potentially physiologically meaningful alterations. It is unlikely that the functional classifications were too broad to reveal important changes, as further sorting into additional subgroups did not suggest any distinct trends. However, the classifications themselves are subjective; therefore, a different sorting scheme may reveal different expression patterns. In addition, pathophysiological mechanisms typically involve groups of genes that cross several different functional groups and thus may not be as readily apparent from functional gene lists. Pathway driven or temporal clustering approaches may assess such changes more effectively (see Supplemental Fig. S2 for pathway analysis).
We applied temporal clustering approaches in search of novel clusters that might be related to mechanisms or markers of injury and/or cell death. The first cluster was anchored with potassium channel expression levels; therefore, it was not unexpected that many of the genes in the cluster were neuronal-specific. However, it is interesting that several genes in this cluster are particularly associated with neurites/synapses and plasticity. For example, HuD is a developmentally regulated neuronal-specific Hu/ELAV-like protein that stabilizes GAP-43 and promotes neurite outgrowth and/or differentiation of PC12 cells (6, 13, 42). Ntab is a noncoding transcript that is only expressed during later stages of development and in adult mammalian brain (21). Although its presence has not previously been demonstrated in spinal cord, in brain it is apparently transported to neuronal processes; however, its function is unknown (21). Synucleins (
and ß) are also synaptic proteins; their decreased expression contrasts to findings in chronic neurodegeneration or traumatic brain injury, where synucleins accumulate and/or increase (12, 44, 63). Visinin-like 1 is a neuronal calcium sensor/binding protein that is reduced in Alzheimers disease and translocated to the extracellular space, where it is associated with plaques, tangles, and dystrophic neurites (9, 53). Because many of these genes are "neuronal-specific" and are both decreased over time and with increasing injury severity, it is possible that these changes may reflect cell death within the epicenter, rather than downregulated expression. These issues require additional study for clarification.
A second cluster included a group of transcription factors (Cited4, Coped, Cebp
) that may be related to injury. Cited4/Cbp/p300-interacting transactivator participates in the regulation of hypoxia-inducible factor 1
(HIF-1
) as well as downstream HIF-1
-controlled hypoxia-mediated gene activation (36, 40); thus, it may be involved in hypoxia-induced signaling in the CNS after injury. Coped/Zf9/CPBP/Kruppel-like factor 6 (Klf6) is one of several Kruppel-like transcription factors that have roles in angiogenesis and tumor suppression (glioblastoma) (28, 33). Increases in Kfl6 have previously been associated with tissue injury or systemic stress, including hemorrhagic shock, radiation-induced microvessel damage, and balloon-injury to endothelial cells, where it may be involved in fibrosis and deposition of extracellular matrix (27, 29, 32).
Within the latter cluster, we focused primarily on Cebp
, which has been associated with neuronal differentiation, plasticity, and acute-phase and inflammatory responses (10, 41, 54, 60, 66).
In our data set, expression of Cebp
was similar to that of ODC, Gro, and lipocalin 2, which are potential downstream Cebp
target genes (14). ODC is a rate-limiting enzyme in the synthesis of polyamines (putrescine, spermidine, spermine) (61). Polyamines and ODC are altered after various types of brain injury, and they may be associated with antioxidant activity, altered ionic and/or excitatory amino acid signaling, blood-brain barrier breakdown/vasogenic edema, and astrogliosis (1, 24, 50, 68). Gro is an inflammatory chemokine, and lipocalin 2 is associated with the acute phase response, two features that have been associated with Cebp
in non-CNS tissue (49, 55). Interestingly, TNF and IL-1ß were significantly increased early after SCI, and both induce Cebp
synthesis (see Supplemental Material)(10). Therefore, both upstream and downstream gene changes are consistent with the concept that Cebp
may be important in the secondary injury response. Additional studies will be necessary to clarify the role of the CCAAT/enhancer-binding proteins in SCI and to determine which specific cell-types may upregulate expression of this transcription factor family in response to injury.
Comparison to previous studies.
Many studies have recently utilized microarray analysis to profile changes in gene expression in a variety of rodent models of SCI (3, 16, 17, 20, 22, 23, 43, 64). The present study differs from most previous approaches (summarized in Table 2) not only in focus (injury severity, rostral/caudal examination, inclusion of sham and naive animals) but also in other differences in sampling, platform, and analysis. Our sampling/analysis utilized individual (unpooled) animals, time-yoked sham animals, and groups with relatively high numbers (n = 4/group). In addition, although we presently report data from selected time points within the acute postinjury period, the entire study includes temporally repetitive sampling [0.5, 4, 12, 24, 48, 72, 168 h (7 day)] over this time frame (data are available at http://pepr.cnmcresearch.org/home.do), which is convenient for temporal validation of results, as well as alternative statistical approaches to analysis (31, 45). In addition, we have now implemented a publicly available SAS server database and visualization tool for better visualization across injuries, locations, and time, offering three different project normalization options (naive animals, time-matched sham animals, and absolute expression values). The present experiments are also unique in that we used the complete set of rat RG-U34 Affymetrix arrays, which covers
27,000 genes/ESTs, whereas previous studies utilized either spotted arrays, RG-U34A, or Affymetrix rat neurobiology arrays. RG-U34A or neurobiology arrays cover
7,000 genes, the majority of which are full-length and/or annotated (only
1,000 are ESTs) (http://Affymetrix.com). However, the entire set of RG-U34 arrays (A, B, C) contains over 16,000 ESTs; therefore, it was important to show that the majority of genes whose expression was altered by injury have been studied and annotated well enough to be classified and thus represented in the global assessment of the functional genomic response. Nearly half of all genes that were altered by injury were identified through the use of RG-U34B and RG-U34C arrays, which have not previously been used in SCI studies. In many cases, ESTs from the RG-U34B and RG-U34C arrays also increased the number of genes associated with clusters that have already been verified and/or are under currently under investigation (3, 11, 1517, 23).
Our study also used both MAS 5.0 and dChip for analysis. Most previous SCI profiling studies have employed the MAS 4.0 algorithm, which is empirically based, whereas MAS 5.0 was developed with a Latin Square test of many transcripts over a large range of concentrations (2). Despite this difference, concordance between MAS 5.0 and MAS 4.0 is relatively high (2). However, dChip is substantially different in that it provides both a project-based normalization and an outlier detection algorithm that detects array outliers for any probe set in any array across all arrays within an experiment (34, 35). Our dChip analysis detected 14 outliers out of 270 arrays (these were subsequently removed); moreover, the majority of the outliers found by dChip had satisfied all initial quality control (QC) parameters, which suggests that dChip may be particularly useful as a project-based quality control tool. Nevertheless, only about 50% of genes were altered significantly using both MAS 5.0 and dChip, whereas others were significantly altered only with either dChip or MAS 5.0. Therefore, our requirement that significant changes must survive both algorithms independently was a fairly specific (stringent) analysis approach, which was probably relatively insensitive to all but the most robust changes. Indeed, many more expression changes were significant when either algorithm was used alone (data not shown), which suggests that the combined analysis may have reduced the rate of potential false positives at the cost of increasing false negatives. It should also be noted that even relatively small but significant alterations (i.e., less than 2-fold) in expression that were detected with this analysis were confirmed by RT-PCR. In addition, internal validation via temporal replication of fold changes was remarkably consistent across the data set (see Supplemental Material).
Nevertheless, due to the stringency of this particular analysis, genes that are expressed in very low abundance (i.e neurotrophins, genes related to apoptosis) may be underrepresented in Tables 1 4. Thus, the PEPR and SAS resources allow analyses with user selected stringency criteria that can be manipulated to evaluate genes expressed at low levels.
In light of recent studies that raise concern about the lack of concordance between microarray platforms and findings derived from separate array studies (39), it is encouraging to note that many genes and clusters represented in previous expression profiling studies of SCI were also significantly altered in the current study. Examples include not only the prevalence of major functional groups (e.g., inflammation, transcription factors, etc.), but also individual clusters (cathepsins, interleukins, etc.) (3, 11, 23). Indeed, such concurrence across laboratories and analyses ultimately serves as the best means to verify the reproducibility of data obtained with arrays (39).
Conclusion.
The present data constitute the most definitive evaluation of the acute functional genomics of SCI to date. Continued analysis of these data should contribute to our understanding of SCI and hopefully will ultimately lead to improvements in treatment and recovery.
 |
GRANTS
|
---|
This work was supported by National Institutes of Health contract NIH-NINDS-01 (NS-1-2339).
 |
ACKNOWLEDGMENTS
|
---|
We thank Andres Lerner, Jorge Garay, and Cinzia Brandoli for expert technical assistance.
 |
FOOTNOTES
|
---|
Address for reprint requests and other correspondence: A. Faden, Dept. of Neuroscience, Georgetown Univ. School of Medicine, 3900 Reservoir Road, NW, Washington, DC 20057 (e-mail: fadena{at}georgetown.edu)
10.1152/physiolgenomics.00081.2005
* A. De Biase, S.M. Knoblach, and S. Di Giovanni contributed equally to this work. 
1 The Supplemental Material (Supplemental Tables S1S7 and Supplemental Figs. S1 and S2) for this article is available online at http://physiolgenomics.physiology.org/cgi/content/full/00081.2005/DC1. 
 |
REFERENCES
|
---|
- Adibhatla RM, Hatcher JF, Sailor K, and Dempsey RJ. Polyamines and central nervous system injury: spermine and spermidine decrease following transient focal cerebral ischemia in spontaneously hypertensive rats. Brain Res 938: 8186, 2002.[CrossRef][ISI][Medline]
- Affymetrix New statistical algorithms for monitoring gene expression on GeneChip probe arrays. In: Technical Notes. Santa Clara: Affymetrix, 2001.
- Aimone JB, Leasure JL, Perreau VM, and Thallmair M. Spatial and temporal gene expression profiling of the contused rat spinal cord. Exp Neurol 189: 204221, 2004.[CrossRef][ISI][Medline]
- Akamatsu W, Fujihara H, Mitsuhashi T, Yano M, Shibata S, Hayakawa Y, Okano HJ, Sakakibara S, Takano H, Takano T, Takahashi T, Noda T, and Okano H. The RNA-binding protein HuD regulates neuronal cell identity and maturation. Proc Natl Acad Sci USA 102: 46254630, 2005.[Abstract/Free Full Text]
- Al-Chalabi A and Miller CC. Neurofilaments and neurological disease. Bioessays 25: 346355, 2003.[CrossRef][ISI][Medline]
- Anderson KD, Morin MA, Beckel-Mitchener A, Mobarak CD, Neve RL, Furneaux HM, Burry R, and Perrone-Bizzozero NI. Overexpression of HuD, but not of its truncated form HuD I+II, promotes GAP-43 gene expression and neurite outgrowth in PC12 cells in the absence of nerve growth factor. J Neurochem 75: 11031114, 2000.[CrossRef][ISI][Medline]
- Basso DM, Beattie MS, and Bresnahan JC. A sensitive and reliable locomotor rating scale for open field testing in rats. J Neurotrauma 12: 121, 1995.[ISI][Medline]
- Beattie MS, Hermann GE, Rogers RC, and Bresnahan JC. Cell death in models of spinal cord injury. Prog Brain Res 137: 3747, 2002.[ISI][Medline]
- Braunewell K, Riederer P, Spilker C, Gundelfinger ED, Bogerts B, and Bernstein HG. Abnormal localization of two neuronal calcium sensor proteins, visinin-like proteins (vilips)-1 and -3, in neocortical brain areas of Alzheimer disease patients. Dement Geriatr Cogn Disord 12: 110116, 2001.[CrossRef][ISI][Medline]
- Cardinaux JR, Allaman I, and Magistretti PJ. Pro-inflammatory cytokines induce the transcription factors C/EBPbeta and C/EBPdelta in astrocytes. GLIA 29: 9197, 2000.[CrossRef][ISI][Medline]
- Carmel JB, Galante A, Soteropoulos P, Tolias P, Recce M, Young W, and Hart RP. Gene expression profiling of acute spinal cord injury reveals spreading inflammatory signals and neuron loss. Physiol Genomics 7: 201213, 2001.[Abstract/Free Full Text]
- Clayton DF and George JM. The synucleins: a family of proteins involved in synaptic function, plasticity, neurodegeneration and disease. Trends Neurosci 21: 249254, 1998.[CrossRef][ISI][Medline]
- Clayton GH, Perez GM, Smith RL, and Owens GC. Expression of mRNA for the elav-like neural-specific RNA binding protein, HuD, during nervous system development. Brain Res Dev Brain Res 109: 271280, 1998.[ISI][Medline]
- Cortes-Canteli M, Wagner M, Ansorge W, and Perez-Castillo A. Microarray analysis supports a role for ccaat/enhancer-binding protein-beta in brain injury. J Biol Chem 279: 1440914417, 2004.[Abstract/Free Full Text]
- Di Giovanni S, De Biase A, Yakovlev A, Finn T, Beers J, Hoffman E, and Faden A. In vivo and in vitro characterization of novel neuronal plasticity factors identified following spinal cord injury. J Biol Chem 280: 20842091, 2004.[CrossRef][ISI][Medline]
- Di Giovanni S, Faden A, Yakovlev A, Duke-Cohan J, Finn T, Thouin M, Knoblach S, De Biase A, Bregman B, and Hoffman E. Neuronal plasticity after spinal cord injury: identification of a gene cluster driving neurite outgrowth. FASEB J 19: 153154, 2004.[Medline]
- Di Giovanni S, Knoblach SM, Brandoli C, Aden SA, Hoffman EP, and Faden AI. Gene profiling in spinal cord injury shows role of cell cycle in neuronal death. Ann Neurol 53: 454468, 2003.[CrossRef][ISI][Medline]
- Dumont RJ, Okonkwo DO, Verma S, Hurlbert RJ, Boulos PT, Ellegala DB, and Dumont AS. Acute spinal cord injury. I. Pathophysiologic mechanisms. Clin Neuropharmacol 24: 254264, 2001.[CrossRef][ISI][Medline]
- Dutcher SA and Michael DB. Gene expression in neurotrauma. Neurol Res 23: 203206, 2001.[CrossRef][ISI][Medline]
- Fan M, Mi R, Yew DT, and Chan WY. Analysis of gene expression following sciatic nerve crush and spinal cord hemisection in the mouse by microarray expression profiling. Cell Mol Neurobiol 21: 497508, 2001.[CrossRef][ISI][Medline]
- French PJ, Bliss TV, and O'Connor V. Ntab, a novel non-coding RNA abundantly expressed in rat brain. Neuroscience 108: 207215, 2001.[CrossRef][ISI][Medline]
- Gris P, Murphy S, Jacob JE, Atkinson I, and Brown A. Differential gene expression profiles in embryonic, adult-injured and adult-uninjured rat spinal cords. Mol Cell Neurosci 24: 555567, 2003.[CrossRef][ISI][Medline]
- Hashimoto M, Koda M, Ino H, Yoshinaga K, Murata A, Yamazaki M, Kojima K, Chiba K, Mori C, and Moriya H. Gene expression profiling of cathepsin D, metallothioneins-1 and -2, osteopontin, and tenascin-C in a mouse spinal cord injury model by cDNA microarray analysis. Acta Neuropathol (Berl), 2004.
- Johnson TD. Polyamines and cerebral ischemia. Prog Drug Res 50: 193258, 1998.[Medline]
- Kaiser S and Nisenbaum LK. Evaluation of common gene expression patterns in the rat nervous system. Physiol Genomics 16: 17, 2003.[Abstract/Free Full Text]
- Kalb RG and Strittmatter SM. Neurobiology of Spinal Cord Injury. Totowa, NJ: Humana, 2000.
- Kiang JG, Bowman PD, Wu BW, Hampton N, Kiang AG, Zhao B, Juang YT, Atkins JL, and Tsokos GC. Geldanamycin treatment inhibits hemorrhage-induced increases in KLF6 and iNOS expression in unresuscitated mouse organs: role of inducible HSP70. J Appl Physiol 97: 564569, 2004.[Abstract/Free Full Text]
- Kimmelman AC, Qiao RF, Narla G, Banno A, Lau N, Bos PD, Nunez Rodriguez N, Liang BC, Guha A, Martignetti JA, Friedman SL, and Chan AM. Suppression of glioblastoma tumorigenicity by the Kruppel-like transcription factor KLF6. Oncogene 23: 50775083, 2004.[CrossRef][ISI][Medline]
- Kojima S, Hayashi S, Shimokado K, Suzuki Y, Shimada J, Crippa MP, and Friedman SL. Transcriptional activation of urokinase by the Kruppel-like factor Zf9/COPEB activates latent TGF-beta1 in vascular endothelial cells. Blood 95: 13091316, 2000.[Abstract/Free Full Text]
- Konkle AT and Bielajew C. Tracing the neuroanatomical profiles of reward pathways with markers of neuronal activation. Rev Neurosci 15: 383414, 2004.[ISI][Medline]
- Krajewski P and Bocianowski J. Statistical methods for microarray assays. J Appl Genet 43: 269278, 2002.[Medline]
- Kruse JJ, te Poele JA, Russell NS, Boersma LJ, and Stewart FA. Microarray analysis to identify molecular mechanisms of radiation-induced microvascular damage in normal tissues. Int J Radiat Oncol Biol Phys 58: 420426, 2004.[CrossRef][ISI][Medline]
- Lelievre E, Lionneton F, Soncin F, and Vandenbunder B. The Ets family contains transcriptional activators and repressors involved in angiogenesis. Int J Biochem Cell Biol 33: 391407, 2001.[CrossRef][ISI][Medline]
- Li C and Hung Wong W. Model-based analysis of oligonucleotide arrays: model validation, design issues and standard error application. Genome Biol 2: RESEARCH0032, 2001.[Medline]
- Li C and Wong WH. Model-based analysis of oligonucleotide arrays: expression index computation and outlier detection. Proc Natl Acad Sci USA 98: 3136, 2001.[Abstract/Free Full Text]
- Linke S, Stojkoski C, Kewley RJ, Booker GW, Whitelaw ML, and Peet DJ. Substrate requirements of the oxygen-sensing asparaginyl hydroxylase factor-inhibiting hypoxia-inducible factor. J Biol Chem 279: 1439114397, 2004.[Abstract/Free Full Text]
- Liu XZ, Xu XM, Hu R, Du C, Zhang SX, McDonald JW, Dong HX, Wu YJ, Fan GS, Jacquin MF, Hsu CY, and Choi DW. Neuronal and glial apoptosis after traumatic spinal cord injury. J Neurosci 17: 53955406, 1997.[Abstract/Free Full Text]
- Magnuson DS, Trinder TC, Zhang YP, Burke D, Morassutti DJ, and Shields CB. Comparing deficits following excitotoxic and contusion injuries in the thoracic and lumbar spinal cord of the adult rat. Exp Neurol 156: 191204, 1999.[CrossRef][ISI][Medline]
- Marshall E. Getting the noise out of gene arrays. Science 306: 630631, 2004.[Abstract/Free Full Text]
- Mazure NM, Brahimi-Horn MC, Berta MA, Benizri E, Bilton RL, Dayan F, Ginouves A, Berra E, and Pouyssegur J. HIF-1: master and commander of the hypoxic world. A pharmacological approach to its regulation by siRNAs. Biochem Pharmacol 68: 971980, 2004.[CrossRef][ISI][Medline]
- Menard C, Hein P, Paquin A, Savelson A, Yang XM, Lederfein D, Barnabe-Heider F, Mir AA, Sterneck E, Peterson AC, Johnson PF, Vinson C, and Miller FD. An essential role for a MEK-C/EBP pathway during growth factor-regulated cortical neurogenesis. Neuron 36: 597610, 2002.[CrossRef][ISI][Medline]
- Mobarak CD, Anderson KD, Morin M, Beckel-Mitchener A, Rogers SL, Furneaux H, King P, and Perrone-Bizzozero NI. The RNA-binding protein HuD is required for GAP-43 mRNA stability, GAP-43 gene expression, and PKC-dependent neurite outgrowth in PC12 cells. Mol Biol Cell 11: 31913203, 2000.[Abstract/Free Full Text]
- Nesic O, Svrakic NM, Xu GY, McAdoo D, Westlund KN, Hulsebosch CE, Ye Z, Galante A, Soteropoulos P, Tolias P, Young W, Hart RP, and Perez-Polo JR. DNA microarray analysis of the contused spinal cord: effect of NMDA receptor inhibition. J Neurosci Res 68: 406423, 2002.[CrossRef][ISI][Medline]
- Newell KL, Boyer P, Gomez-Tortosa E, Hobbs W, Hedley-Whyte ET, Vonsattel JP, and Hyman BT. Alpha-synuclein immunoreactivity is present in axonal swellings in neuroaxonal dystrophy and acute traumatic brain injury. J Neuropathol Exp Neurol 58: 12631268, 1999.[ISI][Medline]
- Pan W. A comparative review of statistical methods for discovering differentially expressed genes in replicated microarray experiments. Bioinformatics 18: 546554, 2002.[Abstract/Free Full Text]
- Parkinson DB, Dickinson S, Bhaskaran A, Kinsella MT, Brophy PJ, Sherman DL, Sharghi-Namini S, Duran Alonso MB, Mirsky R, and Jessen KR. Regulation of the myelin gene periaxin provides evidence for Krox-20-independent myelin-related signalling in Schwann cells. Mol Cell Neurosci 23: 1327, 2003.[CrossRef][ISI][Medline]
- Perrone-Bizzozero N and Bolognani F. Role of HuD and other RNA-binding proteins in neural development and plasticity. J Neurosci Res 68: 121126, 2002.[CrossRef][ISI][Medline]
- Plonquet A, Garcia-Pons F, Fernandez E, Philippe C, Marquet J, Rouard H, Delfau-Larue MH, Kosmatopoulos K, Lemonnier F, Farcet JP, Gherardi RK, and Langlade-Demoyen P. Peptides derived from the onconeural HuD protein can elicit cytotoxic responses in HHD mouse and human. J Neuroimmunol 142: 93100, 2003.[CrossRef][ISI][Medline]
- Poli V, Mancini FP, and Cortese R. IL-6DBP, a nuclear protein involved in interleukin-6 signal transduction, defines a new family of leucine zipper proteins related to C/EBP. Cell 63: 643653, 1990.[CrossRef][ISI][Medline]
- Rao AM, Hatcher JF, Baskaya MK, and Dempsey RJ. Simultaneous assay of ornithine decarboxylase and polyamines after central nervous system injury in gerbil and rat. Neurosci Lett 256: 6568, 1998.[CrossRef][ISI][Medline]
- Resnick DK, Schmitt C, Miranpuri GS, Dhodda VK, Isaacson J, and Vemuganti R. Molecular evidence of repair and plasticity following spinal cord injury. Neuroreport 15: 837839, 2004.[CrossRef][ISI][Medline]
- Saitoh S, Takamatsu K, Kobayashi M, and Noguchi T. Immunohistochemical localization of neural visinin-like Ca(2+)-binding protein 2 in adult rat brain. Neurosci Lett 171: 155158, 1994.[CrossRef][ISI][Medline]
- Schnurra I, Bernstein HG, Riederer P, and Braunewell KH. The neuronal calcium sensor protein VILIP-1 is associated with amyloid plaques and extracellular tangles in Alzheimer's disease and promotes cell death and tau phosphorylation in vitro: a link between calcium sensors and Alzheimer's disease? Neurobiol Dis 8: 900909, 2001.[CrossRef][ISI][Medline]
- Sekine O, Nishio Y, Egawa K, Nakamura T, Maegawa H, and Kashiwagi A. Insulin activates CCAAT/enhancer binding proteins and proinflammatory gene expression through the phosphatidylinositol 3-kinase pathway in vascular smooth muscle cells. J Biol Chem 277: 3663136639, 2002.[Abstract/Free Full Text]
- Sekine O, Nishio Y, Egawa K, Nakamura T, Maegawa H, and Kashiwagi A. Insulin activates CCAAT/enhancer binding proteins and proinflammatory gene expression through the phosphatidylinositol 3-kinase pathway in vascular smooth muscle cells. J Biol Chem 277: 3663136639, 2002.[Abstract/Free Full Text]
- Seo J, Bakay M, Chen YW, Hilmer S, Shneiderman B, and Hoffman EP. Optimizing signal/noise ratios in expression profiling: project-specific algorithm selection and detection p value weighting in affymetrix microarrays. Bioinformatics 20: 25342544, 2004.[Abstract/Free Full Text]
- Smith PM, Sim FJ, Barnett SC, and Franklin RJ. SCIP/Oct-6, Krox-20, and desert hedgehog mRNA expression during CNS remyelination by transplanted olfactory ensheathing cells. Glia 36: 342353, 2001.[CrossRef][ISI][Medline]
- Song G, Cechvala C, Resnick DK, Dempsey RJ, and Rao VL. GeneChip analysis after acute spinal cord injury in rat. J Neurochem 79: 804815, 2001.[CrossRef][ISI][Medline]
- Soriano MA, Tessier M, Certa U, and Gill R. Parallel gene expression monitoring using oligonucleotide probe arrays of multiple transcripts with an animal model of focal ischemia. J Cereb Blood Flow Metab 20: 10451055, 2000.[ISI][Medline]
- Sung YJ, Povelones M, and Ambron RT. RISK-1: a novel MAPK homologue in axoplasm that is activated and retrogradely transported after nerve injury. J Neurobiol 47: 6779, 2001.[CrossRef][ISI][Medline]
- Tabor CW and Tabor H. Polyamines. Annu Rev Biochem 53: 749790, 1984.[CrossRef][ISI][Medline]
- Tachibana T, Noguchi K, and Ruda MA. Analysis of gene expression following spinal cord injury in rat using complementary DNA microarray. Neurosci Lett 327: 133137, 2002.[CrossRef][ISI][Medline]
- Uryu K, Giasson BI, Longhi L, Martinez D, Murray I, Conte V, Nakamura M, Saatman K, Talbot K, Horiguchi T, McIntosh T, Lee VM, and Trojanowski JQ. Age-dependent synuclein pathology following traumatic brain injury in mice. Exp Neurol 184: 214224, 2003.[CrossRef][ISI][Medline]
- Velardo MJ, Burger C, Williams PR, Baker HV, Lopez MC, Mareci TH, White TE, Muzyczka N, and Reier PJ. Patterns of gene expression reveal a temporally orchestrated wound healing response in the injured spinal cord. J Neurosci 24: 85628576, 2004.[Abstract/Free Full Text]
- Wallace MC, Tator CH, and Frazee P. Relationship between posttraumatic ischemia and hemorrhage in the injured rat spinal cord as shown by colloidal carbon angiography. Neurosurgery 18: 433439, 1986.[ISI][Medline]
- Yukawa K, Tanaka T, Tsuji S, and Akira S. Expressions of CCAAT/Enhancer-binding proteins beta and delta and their activities are intensified by cAMP signaling as well as Ca2+/calmodulin kinases activation in hippocampal neurons. J Biol Chem 273: 3134531351, 1998.[Abstract/Free Full Text]
- Zhang KH, Xiao HS, Lu PH, Shi J, Li GD, Wang YT, Han S, Zhang FX, Lu YJ, Zhang X, and Xu XM. Differential gene expression after complete spinal cord transection in adult rats: an analysis focused on a subchronic post-injury stage. Neuroscience 128: 375388, 2004.[CrossRef][ISI][Medline]
- Zini I, Zoli M, Grimaldi R, Pich EM, Biagini G, Fuxe K, and Agnati LF. Evidence for a role of neosynthetized putrescine in the increase of glial fibrillary acidic protein immunoreactivity induced by a mechanical lesion in the rat brain. Neurosci Lett 120: 1316, 1990.[CrossRef][ISI][Medline]