Transcriptional profiling of spinal cord injury-induced central neuropathic pain

Transcriptional profiling of spinal cord injury-induced centralneuropathic pain

Olivera Nesic,* Julieann Lee,* Kathia M. Johnson,� Zaiming Ye,� Guo-Ying Xu,�Geda C. Unabia,� Thomas G. Wood,� David J. McAdoo,� Karin N. Westlund,�Claire E. Hulsebosch� and J. Regino Perez-Polo*

*Department of Human Biological Chemistry and Genetics, �Department of Neuroscience and Cell Biology and �The Molecular

Genomics Core Laboratory, University of Texas Medical Branch, Galveston, Texas, USA

Abstract

Central neuropathic pain (CNP) is an important problem

following spinal cord injury (SCI), because it severely affects

the quality of life of SCI patients. As in the patient popula-

tion, the majority of rats develop significant allodynia (CNP

rats) after moderate SCI. However, about 10% of SCI rats

do not develop allodynia, or develop significantly less allo-

dynia than CNP rats (non-CNP rats). To identify transcrip-

tional changes underlying CNP development after SCI, we

used Affymetrix DNA microarrays and RNAs extracted from

the spinal cords of CNP and non-CNP rats. DNA microarry

analysis showed significantly increased expression of a

number of genes associated with inflammation and astrocytic

activation in the spinal cords of rats that developed CNP. For

example, mRNA levels of glial fibrilary acidic protein (GFAP)

and Aquaporin 4 (AQP4) significantly increased in CNP rats.

We also found that GFAP, S100b and AQP4 protein eleva-

tion persisted for at least 9 months throughout contused

spinal cords, consistent with the chronic nature of CNP.

Thus, we hypothesize that CNP development results, in part,

from dysfunctional, chronically ‘‘over-activated’’ astrocytes.

Although, it has been shown that activated astrocytes are

associated with peripheral neuropathic pain, this has not

previously been demonstrated in CNP after SCI.

Keywords: astrocytes, DNA microarrays, glial fibrilary acidic

protein, inflammation, pain, spinal cord injury.

J. Neurochem. (2005) 95, 998–1014.

Spinal cord injury (SCI) and central neuropathic pain

(CNP)

Traumatic spinal cord injury results in pathophysiologicalchanges that can be loosely divided into two time windows:the acute phase including secondary tissue loss (Tator 1995;Bareyre and Schwab 2003) and the chronic phase (Hul-sebosch 2002; Bareyre and Schwab 2003). In the acutephase, which extends over the first few days, mechanicallesions induce immediate damage to neuronal tracts, bloodflow is disrupted creating substantial ischemia in the injuredspinal cords (Sandler and Tator 1976) followed by increasedproduction of free radicals, and an excessive release ofexcitatory neurotransmitters. In addition, activation of astro-cytes and microglia, in parallel with the recruitment ofperipherally derived immune cells such as neutrophils(6–24 h), macrophages (24 h to 2 weeks) and T cells(Bethea and Dietrich 2002), indicate a strong neuroimmuneresponse during the acute phase and secondary injury, whichmay be either beneficial or deleterious, depending on itstiming, duration and extent. During the chronic phase,

extending from days to years after the trauma, channel andreceptor functions are impaired and astrogliotic scarring anddemyelination accompany Wallerian degeneration (Bareyreand Schwab 2003). All of these processes (underlined) resultin conduction deficits and hyperexcitability in pain pathways,

Received June 1, 2005; revised manuscript received July 6, 2005;accepted July 7, 2005.Address correspondence and reprint requests to Olivera Nesic-Taylor

PhD, Department of HBC&G, University of Texas Medical Branch,MRB 7.138G, 301 University Blvd., Galveston, TX 77555-1072, USA.E-mail: [email protected] used: AQP4, Aquaporin 4; BBB, blood–brain barrier;

BSCB, blood–spinal cord barrier; CNP, central neuropathic pain; FGR,protein tyrosine kinase; GFAP, glial fibrilary acidic protein; IL-1, inter-leukin 1; LCA, leukocyte common antigen; PACAP, pituitary adenylatecyclase-activating peptide; PBS, phosphate-buffered saline; SAM,statistical analysis of microarrays; SCI, spinal cord injury; SDS, sodiumdodecyl sulfate; SSeCKS, src-suppressed C-kinase substrate; SOCS-3,suppressor of cytokine signaling 3; TBS-T, Tris-buffered saline withTween-20; TIMP3, tissue inhibitor of metalloproteinase 3; VEGF,vascular endothelial growth factor.

Journal of Neurochemistry, 2005, 95, 998–1014 doi:10.1111/j.1471-4159.2005.03462.x

998 � 2005 International Society for Neurochemistry, J. Neurochem. (2005) 95, 998–1014

contributing to the development of chronic central neuro-pathic pain (CNP), usually within months following injury(Richards et al. 1980). Neuropathic or exaggerated pain ismaladaptive, chronic and results in a physically andemotionally debilitating condition for which there is noadequate treatment. The pain so greatly affects the quality oflife that depression and suicide frequently result (Segatore1994; Cairns et al. 1996).

The CNP syndromes or dysesthesias (disturbing somaticsensations that may not be painful) can be divided intotwo broad categories based upon their dependence orindependence on peripheral stimuli: (i) spontaneous pain –which occurs independently of peripheral stimuli, is per-sistent, waxes and wanes intermittently, and is describedas numbness, burning, cutting, piercing or electric-like(Davidoff and Roth 1991); (ii) peripherally evoked pain –which occurs in response to either normally non-noxious ornoxious stimuli. In addition, some chronic spinal cordinjured patients experience a band or girdle of hyperpathiaand/or allodynia at the level of the sensory loss (Tasker andDostrovsky 1989). Siddall et al. (2002) defined threecategories of pain that result from SCI: (i) above-level painwhich occurs at dermatomes cranial to the injury site inareas where normal sensation persists following injury, (ii)at-level pain which occurs in dermatomes near the spinalinjury, develops shortly after SCI, and is often characterizedas either stabbing pain or as stimulus independent accom-panied by allodynia, and (iii) below-level pain which islocalized to dermatomes distal to the injury site, developsmore gradually than at-level pain, and is often classified as astimulus-independent, burning, continuous pain (Viercket al. 2000; Sjolund 2002).

Activated astrocytes/microglia and neuropathic pain

Neuropathic pain results from alterations in pain processingneurons within spinal cords (Woolf and Salter 2000).However, as Wieseler-Frank et al. (2004), pointed out,‘drugs developed to control such neuronal deregulationsshould likewise control chronic pain. But, by-and-large, theydon’t’. In the past decade, it has become evident that glialcells (primarily astrocytes and microglia) may substantiallycontribute to the development of pathological pain (Ji andStrichartz 2004). For example, pathological pain afterperipheral nerve injury is regulated by the activation ofastrocytes and microglia (for a review see Wieseler-Franket al. 2004). Because SCI causes a more robust glialactivation than that reported for peripheral nerve injury, weexpected that glial activation after SCI substantially contri-butes to the CNP. Glia were first considered as contributingto pathological pain by Garrison et al. (1994) who found thatmanipulations inducing pathological pain also activateastrocytes, and a drug that blocks pathological pain alsoblocks astrocyte activation. Since then, animal models ofperipheral neuropathic pain have shown a positive correla-

tion between pathological pain and activation of spinal cordastrocytes. However, astrocytic activation has not beencharacterized in SCI-induced CNP.

Several studies have shown that inhibition of glialactivation diminishes hyperalgesia in different models ofperipheral neuropathies. For example, minocycline selec-tively inhibited microglial activation to prevent facilitation,but did not affect existing pain (Raghavendra et al. 2003),suggesting that microglial activation is more important forthe initial, but not the maintenance, phase of exaggeratedpain conditions. The most successful attempt to attenuatechronic pain is the inhibition of both astrocytic andmicroglial activation with propentophylline (Sweitzer et al.2001), which significantly diminished peripheral nerveinjury-induced mechanical allodynia.

Although, it is well documented that glial activationcontributes to the development of exaggerated pain inperipheral neuropathies, the mechanism is still poorlyunderstood. Watkins’ group proposes that substancesreleased by neurons after injury (e.g. fractalkine, sub-stance P, nitric oxide, glutamate or others) activate glia(Watkins et al. 2003; Wieseler-Frank et al. 2004). Activatedglia, in turn, release pro-inflammatory cytokines – criticalmediators of exaggerated pain. Numerous studies show thatblocking pro-inflammatory cytokine activity inhibits diverseexaggerated pain states, including those arising from tissueinflammation, peripheral nerve inflammation, peripheral andspinal nerve trauma, spinal cord inflammation, spinal cordtrauma, and spinal dynorphin (Watkins et al. 1994, 1997;Laughlin et al. 2000; Milligan et al. 2001; Plunkett et al.2001; Sweitzer et al. 2001; Milligan et al. 2003; Raghaven-dra et al. 2003). Pathological pain responses are also inducedby spinal administration of pro-inflammatory cytokines(DeLeo et al. 1996; Reeve et al. 2000), suggesting thatactivated glia and increased release of pro-inflammatorycytokines, including interleukin 1 (IL-1), may also contributeto the development of CNP after SCI.

Because of the large number of biochemical cascades andcellular reactions initiated after SCI, the use of DNAmicroarrays for broad analyses of gene transcription isappropriate. Our group (Nesic et al. 2002) and several othergroups (Carmel et al. 2001; Song et al. 2001; Tachibanaet al. 2002; Di Giovanni et al. 2003) have described globaltranscriptional changes associated with acute phase andsecondary tissue damage after SCI, but no investigationsexamine the chronic post-injury phase. As CNP persists foryears after trauma, it is very important to understand thechronic phase of SCI. Therefore, we performed DNAmicroarray analyses of injured spinal cords 4 weeks afterSCI in rats that developed CNP that had also beenmeasured 28 days after SCI, that is, before killing themand extracting RNA from spinal cord tissue for DNAmicroarray analysis. We hypothesize that many of the geneexpression changes observed are associated with increased

Reactive astrocytes and SCI-induced CNP 999

� 2005 International Society for Neurochemistry, J. Neurochem. (2005) 95, 998–1014

and persistent activation/proliferation of astrocytes andchronic pro-inflammatory processes, and are likely targetsfor therapeutic interventions.

Materials and methods

Rat model of spinal cord injury

Sprague–Dawley male rats (220–230 g) were anesthetized prior to

surgery by administering pentobarbital (35 mg/kg) intraperitoneally.

Anesthesia was deemed complete when there was no response to a

foot pinch. We used an infinite-horizons device with 200-kdyn (2N)

force to inflict moderate contusion spinal cord injury in adult rats in

the thoracic region, at T10. The wound was closed by suturing the

muscle and fascia and the skin closed with surgical staples. Detailed

description of the SCI model was given in Nesic-Taylor et al.(2005). All procedures complied with the recommendations in the

NIH Guide for the Care and Use of Laboratory Animals and were

approved by the UTMB Animal Care and Use Committee.

Behavioral measurements that were performed at 7, 14, 21 and

28 days after SCI, included: (i) measurements of the responses to

mechanical stimuli (von Fray filaments) delivered on the glabrous

surfaces of both the hindpaw and the forepaw (detailed description

in Hulsebosch et al. 2000); and (ii) locomotor assessments using

open field tests with the Basso, Beattie and Bresnahan rank scores

(Basso et al. 1996) assigned as measures of hindlimb activity to

document plantar placement with weight-bearing stance for accurate

somatosensory testing. However, mechanical allodynia measure-

ments reliably detected statistically significant pain-like conditions

only at 28 days (or later) after SCI, but not earlier. In some

experiments, we kept rats longer than 1 month (here are presented

results of analysis of 3 months (Fig. 11), 5 months (Figs 9d and e)

and 9 months (Fig. 10), after SCI, so allodynia was measured at

those later time points, before the rats were killed. We divided rats

into three experimental groups: sham-treated (sham treatment was

performed at the same time as SCI), injured without pain (non-CNP

group) and injured rats that showed significant allodynia (CNP

group). Each figure presented here represents results from separate

experiments in which a different set of rats was used for the DNA

microarray analysis, for the western blot assays or the immunoh-

istochemistry.

The rats used in the DNA microarray analysis, western blot or

immunohistochemistry were chosen based on their (i) CNP score

and (ii) Basso, Beattie and Bresnahan score. Only rats that showed

the same level of locomotor recovery (± 1 Basso, Beattie and

Bresnahan score) but with significant differences in the pain

thresholds measured were used for further biochemical and/or

immunohistochemical measurements. Our analyses of all sham,

naive and injured rats tested for mechanical allodynia showed that

only those rats displaying a significant decrease in pain thresholds

larger than 50% could be reliably categorized as CNP rats.

DNA microarrays

Animals were perfused transcardially with 120–150 mL of 0.9%

NaCl containing heparin (1000 units/L) to eliminate blood from the

spinal cord tissue. The isolated length of cord was immediately

placed in dry ice and allowed to freeze fully. Total RNA was

prepared from frozen spinal cord segments using TRI-Reagent

(Molecular Research Center, Cincinnati, OH, USA). Spinal seg-

ments were homogenized in TRI-Reagent, and total RNA extracted

in chloroform, ethanol precipitated, and stored at )80�C. Total

RNAs was assayed for integrity on 1% denaturing agarose gels.

Approximately 15 lg of total RNAwas used for each target. The rat

RGA microarray from Affymetrix was used in all hybridizations.

Results were analyzed with Affymetrix (Santa Clara, CA, USA)

GeneChip Analysis Suite 5.0 software. Individual microarrays were

scaled to produce mean signal intensity (average difference) of

2500, excluding the top and bottom 2%ile to remove outliers.

To analyze the genomic data we performed: (i) clustering of

arrays, (ii) clustering of genes and (iii) detection of genes with

statistically significantly changed expression levels in the three

experimental groups (methods explained in Nesic et al. 2002;

Svrakic et al. 2003). Finding genes with significantly changed

expression levels in any group being compared was performed using

the statistical analysis of microarrays (SAM), a robust statistical

method devised specifically for the analysis of microarray data

(Tusher et al. 2001), that was used on the data set that resulted from

filtering out (i) genes absent in all samples and (ii) all expressed

sequence tags (ESTs). We accepted only those mRNA values with a

fold change that was higher than 1.5 for up-regulated genes and

lower than 0.66 for down-regulated genes. This pre-filtering

procedure decreased the number of genes that are being analyzed,

and thus decreased the number of false positives from the SAM

analysis.

Protein extraction

Cytosolic buffer (200 lL) was added to each spinal cord segment

for homogenization on ice with a clear grinder that tightly fit inside

the 1.5-mL tubes. Samples were vortexed for 5 s, and then

incubated on ice for 30 min. The lysate was centrifuged at 704 gfor 20 min at 4�C, to pellet the nuclei, cellular debris, and intact

cells. The supernatant was then centrifuged at 704 g for 10 min at

4�C for further purification. The resulting supernatant was centri-

fuged at 9469 g for 20 min at 4�C to pellet the mitochondria. The

new supernatant was centrifuged at 15 339 g for 30 min at 4�C to

remove any residual mitochondria. The resulting supernatant

contained the cytosolic proteins. Seventy microliters of nuclear

extraction buffer (10 mM HEPES, 10 mM KCL, 0.1 mM EDTA,

0.1 mM EGTA, 1 mM dithiothreitol, 0.5 mM PMSF, 2 lg/mL anti-

pain, 2 lg/mL chymostatin, 2 lg/mL pepstatin, 2 lg/mL leupeptin)

was then added during trituration to the nuclear pellet obtained from

the first centrifugation step. This solution was then mixed on a

vortex plate at 1400 rpm for 20 min in 4�C, and subsequently

centrifuged at 15 339 g for 10 min in 4�C. The supernatant

contained the nuclear protein extract, and the pellet was discarded.

Protein concentrations of all extracts was determined using the Bio-

Rad Protein Assay, according to the manufacturer’s instructions

(500-0006; Bio-Rad Laboratories, Hercules, CA, USA).

Electrophoresis and western blotting

Samples containing 40 lg of protein were boiled for 10 min at

100�C with an appropriate volume of 6 · sample buffer [350 mM

Tris HCl, pH 6.8, 1 M Urea, 1% 2-mercaptoethanol, 9.3% dithio-

threitol, 13% sodium dodecyl sulfate (SDS), 0.06% bromophenol

blue, 30% glycerol]. Samples were then placed on ice to cool, and

subsequently centrifuged at 313 g for 30 s before being loaded onto

1000 O. Nesic et al.

� 2005 International Society for Neurochemistry, J. Neurochem. (2005) 95, 998–1014

a SDS–polyacrylamide gel. The stacking gel was 4% acrylamide,

0.04% ammonium persulfate, 0.2% TEMED, and 25% 4 · Tris/CL/

SDS (0.5 M Tris base, .4% SDS, pH 6.8). The separating gel was

10% acrylamide, 0.03% ammonium persulfate, 0.15% TEMED, and

25% 4 · Tris/CL/SDS (1.5 M Tris base, 0.4% SDS, pH 8.8).

Samples were separated by electrophoresis in a Tris–glycine buffer

(25 mM Tris, 192 mM glycine, 0.1% SDS) at 150 V for 4 h. Proteins

were transferred overnight to an Immobilon-P� polyvinylidene

difluoride membrane (IPVH00010; Millipore Corporation, Bedford,

MA, USA) at 4�C and 25 V. The transfer buffer contained 20%

methanol, 20 mM Tris, and 149 mM glycine, at pH 8. Membranes

were reversibly stained with Ponceau S to confirm the transfer of

proteins, and de-stained in water. The Ponceau S solution was 0.5%

w/v Ponceau S and 1% acetic acid. Membranes were incubated for

1 h at room temperature (21�C) in fresh blocking buffer, or

overnight at 4�C. Blocking solutions contain 5% non-fat dry milk in

Tris-buffered saline with Tween-20 (TBS-T) 12.5 mM Tris HCL,

3.7 mM Tris base, 137 mM NaCl, 0.1% Tween-20, pH 7.6. Primary

antibodies were then diluted in the blocking solution, and

membranes were incubated with the appropriate antibody for 1 h

at room temperature, or overnight at 4�C. Membranes were then

washed three times; 10 min for each wash, in TBS-T. Membranes

were subsequently incubated in a horseradish peroxidase-conjugated

secondary antibody for 1 h at room temperature. Membranes were

again washed three times; 10 min each in TBS-T. Peroxidase

activity was detected using the Amersham enhanced chemilumi-

nescence lighting system (ECL) (RPN2209; Amersham Bioscienc-

es, Piscataway, NJ, USA). Antibodies used: GFAP [Chemicon

(Temecula, CA, USA) MAB360, 1 : 60.000 and Dako (Carpinteria,

CA, USA) Cytomation no. Z0334 1 : 1000; both antibodies gave

identical results]; S100b (mouse anti S100 (b subunit) Sigma

(St Louis, MO, USA) S2532, 1 : 1000); Aquaporin 4 (Chemicon

AB3594, 1: 3000); OX-42 (Serotec Raleigh, NC, USA, 1 : 200);

OX-1 (Serotec, 1 : 200). GFAP (0.25 lg; Chemicon) positive

control was also loaded to ensure specificity of bands. b-Actinwas used as a control for loading (mouse anti-b-actin, Sigma;

1 : 10 000).

Double immunofluorescence staining

Animals were perfused with 4% paraformaldehyde in 0.1 &MGR;

phosphate buffer via the aorta. The thoracic spinal cord segments

were removed, post-fixed for 2–4 h at 4�C, cryoprotected in 30%

sucrose in phosphate buffer overnight, and embedded in OCT

compound. Tissue sections were cut transversely at 30 lm on a

sliding microtome. Immunofluorescence staining proceeded accord-

ing to the procedures below. Floating tissue sections were rinsed for

3 · 10 min with phosphate-buffered saline (PBS) and incubated

with of 5% normal goat serum in PBS containing 0.3% Triton-X

(PBST) for 30 min. Sections were then incubated with primary

antibodies overnight at room temperature. Sections were rinsed with

PBST 3 · 10 min and incubated with anti-rabbit IgG Alexa-

Fluor 568 and anti-mouse IgG AlexaFluor 488 diluted in PBST

1 : 200 for 4 h at room temperature (Molecular Probes, Eugene,

OR, USA). After rinsing three times for 10 min, the double-stained

tissue sections were mounted on gelatin-coated glass slides, dried,

and coverslipped with non-fade media. Omission of the primary

antibodies or use of non-specific secondary IgGs in the immuno-

staining process resulted in negative staining of the tissue.

Antibodies used: GFAP (mouse monoclonal, Chemicon; 1 : 500),

Aquaporin 4 (AQP4; rabbit polyclonal Chemicon; 1 : 1000) and

OX-42 (mouse monoclonal, 1 : 200).

Confocal laser scanning microscopy

Stained sections were scanned with a confocal laser scanning system

(Bio-Rad Radiance 2100, K-2 system). For double immunofluores-

cent staining, data from two channels were collected by sequential

scan to avoid bleeding through between two channels. Images were

collected with Krypton lasers of 488 and 568 nm excitation, Green

images for AlexaFluor 488, while red images for AlexaFluor 568.

The overlay of the two channels showing the co-localization of the

two antigens was indicated in yellow. Digital images were saved and

processed with Adobe Photoshop for final editing.

Statistical analysis

All statistical tests were evaluated at the alpha level of

significance of 0.05, two-tailed. All of the experiments have

similar structures in that the effects of a manipulation on the level

of the factors were measured. We used parametric methods (t-test)for our analyses. However, if the assumptions for these tests were

not met, we proceeded with non-parametric analyses (Mann–

Whitney). Likewise, we used non-parametric methods to check all

parametric results as a safeguard of assumptions. If the results

were not consistent, we reported the results from non-parametric

tests.

Results

Not all rats develop significant CNP

Given that 10–15% of moderately injured rats do not developCNP or that their CNP levels are significantly lower, we usedthe K-means clustering method (SPSS program; SAS Insti-tute, Cary, NC, USA) to assign all injured rats (n ¼ 20) totwo groups according to their pain thresholds (Fig. 1). Onegroup (n ¼ 4) showed statistically significant decreases(�90% decrease) in pain thresholds after SCI (Figs 1b andc) and increases (�50%) in mechanical allodynia (Fig. 1d)(CNP group), while the other injured group (n ¼ 4) did notshow such changes (non-CNP group).

The extent of SCI-induced damage is a key determinantof CNP that might contribute to the variability of nocicep-tive sensitivity in SCI rats. Because we used spinal cordsfrom CNP and non-CNP groups for transcriptional profilingand RNA extractions, we could not morphologicaly deter-mine the extent of injury in those rats. Therefore, wemeasured locomotor recovery using the Basso, Beattie andBresnahan score (Basso et al. 1996), which providesindirect measures of injury magnitude, as the amount ofspared matter significantly correlates with locomotor ability(Scheff et al. 2003). Thus, for these experiments weselected only CNP and non-CNP rats that showed thesame level of locomotor recovery (Fig. 2). That is why wehave chosen only four rats in each group, out of 20 rats

Reactive astrocytes and SCI-induced CNP 1001

� 2005 International Society for Neurochemistry, J. Neurochem. (2005) 95, 998–1014

initially contused – the other contused rats showed differentlevels of locomotor recovery. As shown in Fig. 2, Basso,Beattie and Bresnahan scores for both groups of rats were

indistinguishable, evidence that the injuries to the spinalcords were equivalent (Fig. 2).

Gene expression profiling of spinal cords of rats that

developed CNP versus spinal cords of rats that did not

develop CNP suggests increased astrocytic activation in

the former

To characterize the unique transcriptional changes underlyingCNP after SCI, we compared expression profiles of spinalcords of injured rats that developed CNP 28 days after SCIwith expression profiles of spinal cords of injured rats thatdid not develop significant CNP, as dramatic transcriptionalchanges induced by SCI itself may mask subtle changesunderlying CNP alone. We used Affymetrix DNA microar-rays containing 8799 gene-specific probes to analyze tran-scriptional profiles of the injured spinal cords rats (n ¼ 3) inthe CNP group and compared them with expression profilesof injured spinal cords of rats in the non-CNP group (n ¼ 3).We collected five segments above and five spinal segmentsbelow the site of injury (T10) to analyze transcriptionalchanges underlying CNP over a wide spinal cord region.

Hierarchical cluster analysis of all mRNAs in individualspinal cord samples (cluster analysis of arrays) showed thatthere was a distinct difference in the genomic expressionpatterns at the gross transcriptional levels of spinal cords

Fig. 1 (a) K – means clustering divided pain thresholds (g of force

used to determine pain threshold) of all injured rats (n ¼ 8) into two

groups: one group (CNP group) whose threshold decreased by 90% at

28 days after SCI compared with the pre-injury baseline value (e.g. the

mean value of percentage decrease for CNP group is at 10% of pre-

surgery value, as presented on the x-axis), while the other showed a

mean decrease in thresholds at 50% (non-CNP group). (b) Pain

thresholds for these two groups were indistinguishable before the

injury. (c) A graph showing a decrease in pain thresholds after injury

compared with the baseline: there is no statistical significance for the

non-CNP group (p ¼ 0.1), while p ¼ 0.02 for the CNP group. (d) The

number of paw withdrawals significantly increased in the CNP group,

but not in the non-CNP group, when compared with their pre-injury,

baseline values.

Fig. 2 Basso, Beattie and Bresnahan scores measured in CNP (n ¼4) and non-CNP groups (n ¼ 4) of rats showed no difference, indi-

cating that the injury is similar in both groups. This suggests that dif-

ferent processes determine development of CNP. We hypothesize

that individual genomic responses to SCI determine CNP variability.

Data presented as mean ± SD.

1002 O. Nesic et al.

� 2005 International Society for Neurochemistry, J. Neurochem. (2005) 95, 998–1014

belonging to the CNP-group versus the non-CNP-groups ofrats (Fig. 3), implying that there was a large number ofchanges in gene expression, in agreement with the proposedmultiplicity of pathological changes underlying CNP devel-opment. As shown in Fig. 3(b and c), CNP and non-CNP ratschosen for DNA microarray analysis showed significant anddistinct differences in the pain threshold decreases 28 daysafter SCI.

After eliminating ESTs, genes that were absent in allsamples, and filtering out all genes for which mRNA foldchange in CNP versus non-CNP samples was below 1.5, theresulting database contained 988 genes in rostral and 997 incaudal parts. The 988 and 997 genes were then analyzedusing the SAM program (p < 0.01).

Cluster analysis of individual gene expression levels(clustering of genes) revealed groups of genes that areconsistently up- or down-regulated (Fig. 4) in the spinalcords of CNP versus non-CNP-groups. After SAM weidentified 310 genes significantly up- or down-regulated insegments above the injury site and 427 in the segmentsbelow (p < 0.01).

About 30% of these 310 or 427 genes are associated withbroad intracellular signaling; such as kinases, G-proteins andtranscription factors. However, a number of the significantlyup-regulated genes, in both rostral and caudal spinal cords of

the CNP group are associated with inflammation (Table 2)and astrocytes and/or astrocytic activation (Table 1).

Increased activation of astrocytes in injured rats

displaying CNP

As shown in Table 1 and Fig. 5, we found significantlyincreased expression of glial fibrilary acidic protein (GFAP)mRNAs in both rostral and caudal CNP spinal cords. GFAPis commonly used as an astrocytic marker (Ludwin et al.1976), which is up-regulated after astrocytic activation.

Western blot analysis of GFAP protein expression levels(Fig. 6) confirmed that injured spinal cords isolated fromrats experiencing significant CNP (CNP, Fig. 6c) had higherexpression levels of GFAP when compared with injuredspinal cords of rats (non-CNP, Fig. 6c) that did not developsignificant decreases in pain thresholds (the percentage ofthe threshold decreases as shown in Fig. 6c). Thus, whileGFAP levels increased in the spinal cords of injured rats(compared with uninjured), the increases in the spinal cordsof injured rats that develop CNP were significantly higher(Fig. 6).

Quantitative analysis of GFAP immunolabeling (Fig. 7)confirmed that there were significant increases in GFAPexpression in both gray and white matter in the spinal cordsof rats that developed CNP.

Fig. 3 (a) Hierarchical cluster analysis of

12 Affymetrix arrays, e.g. 12 spinal cord

samples isolated from CNP group (P) and

non-CNP group (NP): three spinal cord

samples comprising five segments above

the site of injury (R, rostral) and five seg-

ments below (C, caudal). Euclidian and

Manhattan hierarchical analysis showed

clear differences in genomic responses

between pain and non-CNP group. They

also showed distinct transcriptional differ-

ences between injured rostral and caudal

spinal cord segments. (b) The pain thresh-

old decreases in three CNP and three non-

CNP rats used for DNA microarray analysis.

(c) Those rats showed distinct difference in

the percentage decrease of the pain

thresholds.

Reactive astrocytes and SCI-induced CNP 1003

� 2005 International Society for Neurochemistry, J. Neurochem. (2005) 95, 998–1014

Interestingly, we found not only an increased expression ofintact GFAP, but also a marked degradation of GFAP(Figs 6c, 8b and 9a); see bands with molecular weightssmaller than 37 kDa that may also indicate the presence ofdifferent GFAP isoforms.

We used another astrocytic marker that recognizes 100bprotein to validate GFAP results. S-100 is a 20-kDa Ca2+-binding protein composed of a and b subunits (S-100a andS-100b) primarily expressed by astrocytes (Hu et al. 1996).As shown in Fig. 8(c and d), SCI-induced changes in S100bmonomer expression (10-kDa band) paralleled SCI-inducedchanges in GFAP (Figs 8a and b). Higher expression wasfound in spinal cords of rats that showed robust decrease inpain thresholds after SCI (70 or 80%, lanes 5 and 6), whencompared with injured spinal cord of a rat with smalldecrease in pain threshold (lane 4). The GFAP and S100expression levels were measured in the spinal cord segmentsbelow the site of injury (T11), 4 weeks after SCI.

Given that only a small number of injured rats do notdevelop CNP after SCI, and that in those experiments wecould use only non-CNP rats with the same level oflocomotor recovery as CNP rats (± 1 Basso, Beattie andBresnahan score unit), we did not obtain enough non-CNPsamples to perform statistical analysis of S100b expressionchanges in CNP versus non-CNP rats. However, we showedthat there is a significant (p < 0.05) up-regulation of S100bin CNP rats (n ¼ 6) compared with sham values (n ¼ 3),that was comparable with the GFAP changes in the samesham and CNP spinal cords (Fig. 8e). Similar results wereobtained on sham (n ¼ 3) and CNP spinal cords (n ¼ 4)3.5 months after SCI (data not shown).

Additionally, AQP4 was also found significantlyup-regulated in CNP spinal cords (Table 2; Figs 9b and

c). AQP4 is an astrocytic water channel protein (seeFig. 9d), that is typically up-regulated in activated astro-cytes. A representative western blot shows low AQP4expression in uninjured samples, and robustly increasedAQP4 protein levels in the CNP spinal cords, consistentwith GFAP expression changes (Fig. 9a); this was con-firmed with quantitative analysis of AQP4 immunolabelingin both gray and white matter of CNP spinal cords (n ¼ 3)(Figs 9d and e). We found significant (p < 0.05) up-regulation of immunolabeled AQP4 in CNP rats comparedwith sham values, at 4 weeks (data not shown) and5 months after SCI (Figs 9d and e). Quantitative analysisof immunolabeled AQP4 shows robust increase, partic-ularly in the white mater of CNP rats 5 months after SCI,consistent with increased GFAP immunolabeling in thewhite matter of the same CNP spinal cords.

As shown in Fig. 9(a–c) GFAP and AQP4 are not onlyincreased at the site of injury (T10) 4 weeks after SCI, butalso in the spinal segments away from the site of injury; inlumbar (L4/L5 combined) and in cervical (C7/C8 combined)segments in the spinal cords of rats with different level ofCNP (11 and 54.5% decrease in the pain thresholds). Bargraphs represent quantitative analysis of GFAP and AQP4western blots in cervical and lumbar segments of three shamand seven CNP spinal cords. This result suggests chronic andwidespread astrocytic activation, consistent with mechanicalallodynia observed in both hindlimbs and forelimbs of SCIrats (Hulsebosch et al. 2000).

As shown in Fig. 10(a and b), increased GFAP expressionpersists for months after SCI. Significant increases in GFAPexpression, that start early after SCI (4 h), persists for weeks(4 weeks), and months (4 months, 9 months) after SCI inrats that develop CNP. AQP4 paralleled all SCI-induced

Fig. 4 Representative example of a cluster

analysis of genes of six spinal cord samples

isolated from CNP rats and non-CNP rats

comprising five segments above the site

of injury (R, rostral). The diagram

shows a number of genes that are consis-

tently up- or down-regulated in CNP versus

non-CNP spinal cords. A blue–green color

depicts low levels of mRNA expression,

while red–yellow depicts high levels of

mRNA expression.

1004 O. Nesic et al.

� 2005 International Society for Neurochemistry, J. Neurochem. (2005) 95, 998–1014

increases in GFAP at different time points (data not shown).As shown in Fig. 9(d and e), AQP4 expression stayedelevated for months (5 months) after SCI.

Discussion

Individual variability of CNP after SCI

The prevalence of chronic SCI pain varies considerably, butaverages 65% with approximately one-third of SCI patientsrating their pain as severe. It is not known why some SCIpatients fail to develop pain. We used a rat model ofmoderate spinal contusion (Gruner 1992; Basso et al. 1995;Hulsebosch et al. 2000) similar to the contusion/cyst type ofinjury (Basso et al. 1996) most prevalent in clinical settings(Bunge et al. 1993; van de Meent et al. 1996). With thismodel, there is a smaller, but variable percentage ofmoderately contused rats that did not develop CNP. Theonset and level of spontaneous pain behavior correlate withthe level of the damage after SCI (Lindsey et al. 2000;Yezierski 2000). Consistent with this, neuroprotective drugshave been shown to have beneficial effects on CNPdevelopment (Yezierski 2000), but it is also noteworthy tomention that all active substances shown to diminish tissuedamage after SCI do not alleviate CNP (Mills et al. 2002).Thus, other factors, in addition to the extent of damage,contribute to the development of variable CNP levels in SCIrats, most probably genetic factors.

The underlying genetic bases for the individual variabilityin pain-related traits have only recently been acknowledged(Mogil 1999; Diatchenko et al. 2005). Although the relativeimportance of genes versus experience in human painperception remains unclear, rodent populations display largeheritable differences in both nociceptive and analgesicsensitivity. For example, Mills et al. (2001) showed thatthree strains of rats (Long–Evans, Wistar, and Sprague–Dawley) demonstrate different responses to SCI, includingCNP development, indicating a genetic contribution. Giventhat inflammatory reactions are well-established initiators oflong-term hyperexcitability in pain pathways (Sommer2003), we hypothesize that the inflammatory response toSCI, that depends on individual variations in geneticbackground, determines variable sensitivity to pain stimuliexhibited after SCI. For example, it has been shown that theinflammatory responses induced by bacterial lipopolysac-charide (LPS) differ among genetically distinct strains ofinbred mice (De Maio et al. 1998). B6 mice showed higherlevels of circulating IL-1b and IL-6, as well as higher mRNAlevels of hepatic b-fibrinogen (an acute-phase gene) andmetallothionein compared with A/J mice after LPS admin-istration. As B6 and A/J mice are bred and raised in identicalenvironments and received the same LPS challenge, thecontrasting inflammatory responses observed can be attrib-uted to genetic differences between these two strains. These

Fig. 5 GFAP mRNA expression levels are significantly up-regulated

both rostral and caudal to the site of injury (T10) (p < 0.05; n ¼ 3).

Table 1 Genes typically expressed in astrocytes whose expression

levels are increased in rats experiencing pain (CNP-group) 4 weeks

after SCI

Rostral/fold Caudal fold

GFAP 1.5 2.9

Urea transporter (UT3) 2.8 5.7

Glutamate/aspartate transporter 3.6 5.2

Amyloid precursor-like protein 2 3.4 25.1

Amyloid precursor A4 1.76 3.6

NGF 1.52 2.94

GDNF R b 1.63 2.05

Bone morphogenetic protein type 2.25 3.2

PACAP R 3.6 2.7

Neuroglycan 2.49 2.63

Inwardly rectifying K channel 1 13.8 14.4

K channel Kv2 4.4 2.74

Na/K ATPase a1 subunit 2.4 6.6

Na-K-Cl co-transporter 5 6

Vacuolar ATPase 1.7 2.5

GluRB (AMPA) 2 3

NMDAR2A 1.7 1.6

PSD-93 1.6 1.5

MGLU R 7 2.5 12

GABA B R1c 1.6 4

GABA A R 2.3 3.4

A2b adenosine R 1.6 1.5

ATP ligand-gated ion channel 2.4 2

5-HT2C R 6.3 7.3

Snap 25A 3 8.5

Snap 23 2 2

Syntaxin 6 2 4

Syntaxin 12 1.9 6.5

Fold change in pain versus non-pain spinal cords, in five segments

above (rostral) and five segments below (caudal) the site of injury

(n ¼ 3); p < 0.05, SAM; 1.5 cut-off fold change.

Reactive astrocytes and SCI-induced CNP 1005

� 2005 International Society for Neurochemistry, J. Neurochem. (2005) 95, 998–1014

Fig. 6 GFAP protein expression levels are significantly up-regulated

in spinal cords of rats experiencing significant decreases in pain

threshold (Von Frey tests). (a) A representative example in which nine

samples (three sham, three non-CNP and three injured with CNP) are

immunobloted for GFAP. A band at 50 kDa showed significant

increases in both the non-CNP and CNP groups, but to a higher extent

in the CNP-group. (b) Quantitative western blot analysis (n ¼ 3)

showed significantly higher expression levels of GFAP in the CNP

group in samples consisting of three pooled segments [site of injury

(T1) and two adjacent segments (T9 and T11)]. p ¼ 0.003 is the

p-value for the up-regulation of GFAP in non-CNP samples versus

sham samples, while p ¼ 0.026 is the p-value for the up-regulation of

GFAP in CNP samples versus non-CNP (mean ± SD). (c) Significant

degradation of GFAP is detected in SCI rats, but to a much higher

extent in rats that developed CNP – see bands with molecular weight

lower than 37 kDa; marked with an arrow. Percentages of threshold

decreases for individual rats are depicted above names of samples.

Basso, Beattie and Bresnahan scores for CNP and non-CNP groups

of rats are indistinguishable (not shown). Actin expression is used as a

control for loading.

Fig. 7 Quantitative analysis of GFAP

immunolabeling. (a) A representative

example of immunolabeled astrocytes in

white (lower panels) and gray matter (upper

panels) of three spinal cords: sham, injured

that did not develop pain (non-CNP) (n ¼ 3)

and injured that developed CNP (CNP)

(n ¼ 3). Longitudinal sections are taken

above the site of injury, spanning over two

segments T8 and T9. (b) Quantitative ana-

lysis of GFAP immunolabeling in gray and

white matter showed higher GFAP expres-

sion in injured spinal cords of rats that did

not develop CNP, but significantly elevated

in spinal cords of rats that developed CNP

(n ¼ 3; mean ± SD, p < 0.05). Basso,

Beattie and Bresnahan scores for rats used

in this experiment were not statistically dif-

ferent.

1006 O. Nesic et al.

� 2005 International Society for Neurochemistry, J. Neurochem. (2005) 95, 998–1014

data illustrate that the genetic background of the individualcan modulate the inflammatory responses to injury (De Maioet al. 2005). This is consistent with the data presented inTables 1 and 2, which show that individual differencesamong inflammatory responses to SCI (reflected at the levelof astrocytic activation and expression of inflammatorygenes) contribute to, or determine, individual nociceptivesensitivity after SCI; a novel hypothesis.

Pro-inflammatory genes are up-regulated in SCI rats that

develop CNP

The increased expression of genes presented in Table 2indicate that there are chronic inflammatory conditions in theCNP spinal cords. For example, a-2-macroglobulin, a wide-range protease inhibitor, is an indicator of inflammation as itis only expressed in the adult liver during the acute phase ofthe inflammatory response. In the CNS, a-2-macroglobulin isexpressed in astrocytes that are activated by pro-inflamma-tory cytokines (Gao et al. 2002) and, as shown in Table 2,a)2-macroglobulin mRNA levels increased significantly ininjured CNP versus injured non-CNP spinal cords.

The increased expression of disintegrin metalloproteinase(MMP3, Table 2) and the tissue inhibitor of metalloprotein-ase 3 (TIMP3) have been typically detected under inflamma-tory conditions in CNS, when thesemolecules are expressed inastrocytes (Muir et al. 2002; Kieseier et al. 2003). It has beenshown that atrocytes activated by LPS increase the expressionof signal transducer and activator of transcription 3 (STAT3,Table 2) and TIMP3 (Pang et al. 2001).

STAT3 is a transcription factor activated by a number ofcytokines (most notably, by interleukin 6) that initiatesexpression of suppressor of cytokine signaling 3 (SOCS-3).

Thus, increased expression of SOCS-3 indicates bothstimulation and autoinhibition of various cytokines, and isa hallmark of ongoing inflammatory reactions. IncreasedSOCS-3 mRNA has been detected in astrocytes stimulatedby IFNc (Stark et al. 2004).

The inflammatory response in the CNS includes partici-pation of different cellular types of the immune system(macrophages, mast cells, T and B lymphocytes, dendriticcells) that can, under different pathological conditions, whenthe blood–brain barrier (BBB) or blood–spinal cord barrier(BSCB) become more permeable, enter the CNS paren-chyma. BBB/BSCB permeability depends of the normalastrocytic function, e.g. astrocyte foot processes in closeapposition to the abluminal surface of the microvascularendothelium of the BBB/BSCB contribute to both thestructural and functional integrity of the BBB (Abbott2002). The importance of astrocytes in maintaining BBB isillustrated by the study of Bush et al. (1999) in whichastrocyte ablation led to failure of BBB repair and prolongedand increased infiltration of leukocytes, suggesting thatactivated astrocytes have an integral part in preventingblood-borne immune cell infiltration across the BBB.

Reactive astrocytes also migrate to the injured area in theCNS, where they proliferate and produce extracellular matrixelements (N-cadherin, laminin, fibronectin; see Table 2),thereby reconstituting the BBB. Here, we propose thatchronic over-activation of astrocytes in CNP spinal cordsleads to, and/or results from, BSCB breakdown. There areseveral genes involved in BBB/BSCB function that changedexpression in CNP spinal cords, as reported in Table 2. Forexample, we found significantly up-regulated water channelAquaporin 4 (AQP4) at the mRNA (Table 2) and protein

Fig. 8 The astrocytic marker, S100b (c)

also increased in the CNP spinal cords,

paralleling GFAP changes (a). A represen-

tative western blot of GFAP (a) and S100b

(C) shows three sham samples (lanes 1–3),

an injured spinal cord from a rat that

showed small decreases in pain threshold

(10.4%; lane 4), and two spinal cords

extracted from rats that demonstrated

marked decreases in pain thresholds (80.8

and 70%, lanes 5 and 6). Spinal cords are

isolated 4 weeks after SCI, and samples

taken from the segment below the site of

injury (T11). (b) Quantitative representation

of GFAP western blots: all GFAP bands

(upper graph) or just degraded ones (lower

than 37 kDa) (lower graph). (C) S100b

expression changes. (d) Quantitative pres-

entation of S100b protein bands.

Reactive astrocytes and SCI-induced CNP 1007

� 2005 International Society for Neurochemistry, J. Neurochem. (2005) 95, 998–1014

levels (Fig. 9). AQP4 up-regulation is significantly correlatedwith blood–brain barrier opening in brain trauma (Saadounet al. 2002). AQP4 is primarily expressed in astrocytic footprocesses where it forms the BBB/BSCB with endothelialcells.

Furthermore, increased levels of vascular endothelialgrowth factor (VEGF) mRNAs (Table 2), if translated intoincreased VEGF proteins, might also influence BBB/BSCBpermeability, as it is well established that VEGF increasesleakiness of the BBB (Croll et al. 2004).

Another molecule involved in the regulation of BBBpermeability is the src-suppressed C-kinase substrate(SSeCKS), also found increased in CNP spinal cords

compared with non-CNP spinal cords (Table 2). Lee et al.(2003) showed that over-expression of SSeCKS in ratastrocytes triggers vessel maturation and barriergenesis.Thus, a significant increase in SSeCKS transcription in CNPspinal cords may indicate an incomplete barriergenesis thatpersists for weeks after SCI. This hypothesis is supported byMRI studies showing that newly formed vessels after SCI donot form physiologically functional, impermeable BSCB, butleaky and dysfunctional BSCB (Bilgen et al. 2002).

Taken together, these results suggest the likely presence ofa dysfunctional BSCB in those rats that develop CNP thatmight contribute to prolonged inflammatory conditions andconsequent over-activation of astrocytes, with perturbed

Fig. 9 (a) GFAP is up-regulated, not only at

the site of injury, but away from the impact

site as well: in lumbar segments (combined

L4 and L5) and cervical (combined C7 and

C8). Note low molecular weight bands

indicating GFAP degradation in T10, but

also in lumbar and cervical spinal cord

segments. Bar graphs represent quantita-

tive analysis of GFAP western blots per-

formed with sham (n ¼ 3) and CNP spinal

cords (n ¼ 7). (b) Astrocytic protein,

Aquaporin 4 (AQP4) increases in injured

spinal cord segments, especially in a rat

that showed decreases in pain threshold by

54.5%. AQP4 expression changes were

consistent with GFAP changes, in both

lumbar and cervical segments. (c) Quanti-

tative representation of AQP4 protein bands

in sham (n ¼ 3) and CNP samples (n ¼ 7),

28 days after SCI at the site of injury (T10),

in L4/L5 and C7/C8 segments. (d) Double

immunolabeling (yellow) of GFAP (green)

and AQP4 (red), confirms astrocytic local-

ization of AQP4 in spinal cords 5 months

after SCI (magnification · 40, white matter

in the spinal segment above the site of

injury). The photograph also illustrates

increased GFAP and AQP4 immunolabe-

ling in CNP spinal cords. (e) Quantitative

analysis of AQP4 and GFAP immunolabe-

ling in three sham and three CNP rat spinal

cords 5 months after SCI. Three CNP rats

had the pain threshold decreased by 40, 70

and 36.4%, respectively, with similar loco-

motor recovery scores. AQP4 immuno-

labeling significantly increases in the CNP

spinal cords; more in the white than in the

gray matter. As shown in the bar graph,

AQP4 changes paralleled changes in white

matter GFAP levels.

1008 O. Nesic et al.

� 2005 International Society for Neurochemistry, J. Neurochem. (2005) 95, 998–1014

cellular functionality. Chronically activated and dysfunc-tional astrocytes may in turn fail to regulate BSCB, leukocyteinfiltration and therefore cause chronic inflammation in theCNP rats.

In support of this hypothesis, we found increased levels ofthose mRNAs typically expressed in activated microglia and/or infiltrated blood borne immune cells. For example, theexpression of the major histocompatibility complex II(MHC-II, Table 2) is increased under a wide range ofpathologic conditions and plays multiple roles duringinflammation in the CNS (Neumann 2001), similarly to theleukocyte common antigen (LCA, Table 2). As shown inAlzheimer brain, leukocytes and reactive microglia expressLCA, contrary to astrocytes (Itagaki et al. 1988). We alsofound LCA immunolabeling robustly increased in CNPspinal cords (n ¼ 3) 3 months after SCI, as shown inFig. 11(c). LCA is almost absent in sham spinal cords (n ¼3), while spinal segments above the site of injury (T10),spanning from T7-T9 showed striking up-regulation of LCAin microglia, 3 months after SCI. MHC-II are unique toantigen-presenting cells such as microglia and monocytes. Asshown in Table 2, FGR (protein tyrosine kinase) mRNAs areup-regulated in the CNP spinal cords. FGR is expressed inactivated microglia, T and B cells, but not in astrocytes(Krady et al. 2002). This supports the idea of a post-SCIchronic inflammation, in part as a result of the infiltration ofimmune cells into the spinal cords of rats that developedCNP. To quantitatively analyze the presence of activated

Table 2 Genes whose products are involved in regulating inflamma-

tory reactions

Rostral/fold Caudal/fold

Alpha 2 – macroglobulin 12 2.5

Disintegrin metalloproteinase 3 (MMP3) 3.5 4

TIMP-3 3 3.6

STAT3 2 4.7

Aquaporin 4 2.4 3

VEGF 2.4 2.3

N-cadherin 1.8 1.6

Laminin c 2 2

Fibronectin 4 5

SSeCKS 3 7

Glia maturation factor 3 3.2

MHC-II (B a chain) 1.8 4

LCA 2.3 4

FGR 1.7 1.7

Fibroblast growth factor-2 (FGF-2) R 3 9.3

bFGF 1.9 1.5

NonO/p54nrb (FGF response element) 1.5 1.5

TGFb type I R 9.8 34.4

TGFb type II R 2.3 2.5

TGFb binding protein 3 3.2 1.8

ICE 2 2.3

IL-1rAcP 2 2

Fold changes in CNP versus non-CNP spinal cords, in five segments

above (rostral) and five segments below the site of injury (caudal)

(n ¼ 3); p < 0.05, SAM; 1.5 cut-off fold change.

Fig. 10 Time course of SCI-induced GFAP

increases. (a) GFAP significantly increased

at 4 and 24 h, reaching a peak at 4 weeks

and remaining elevated for 9 months. (b)

Bar graphs represent quantitative analysis

of western blot analyses, normalized to

sham values (n ¼ 3 for all time points

except 4 weeks, when n ¼ 6).

Reactive astrocytes and SCI-induced CNP 1009

� 2005 International Society for Neurochemistry, J. Neurochem. (2005) 95, 998–1014

microglia/macrophages in the spinal cords of uninjured rats,and injured rats with CNP, we used the OX-42 antibodywhich does not differentiate between microglia and blood-borne macrophages (Fig. 11a). Consistent with our hypothe-sis, we found OX-42 immunolabeling significantly increasedin the CNP spinal cords (Fig. 11b). Microglial activation inCNP spinal cords may partly be the result of astrocyticactivation, as glia maturation factor mRNAs were alsosignificantly increased in CNP rats (Table 2). Zaheer et al.(2002) showed that over-expression of glia maturation factorin astrocytes leads to activation of microglia, but this has tobe tested in future experiments.

Astrocytic over-activation in CNP spinal cords

The mRNA changes presented in Tables 1 and 2 are similarto those reported for neonatal astrocytes in vitro (Nakagawaand Schwartz 2004). Reactive adult atsrocytes are thought toundergo dedifferentiation processes, and resemble neonatalastrocytes, especially in terms of their gene expressionpatterns (Nakagawa and Schwartz 2004). For example, DNAmicroarray analysis of CNP versus non-CNP spinal cordsidentified mRNAs whose expression was also found to beelevated in reactive astrocytes (Table 1): amyloid protein(Rossner 2004), pituitary adenylate cyclase-activating pep-tide (PACAP) receptor (Suzuki et al. 2003), inwardly

Fig. 11 (a) A representative example of

immunolabled OX-42-positive cells in spinal

cords of uninjured and injured rat that

developed CNP. (b) Quantitative analyses

of immunolabeled OX-42 showed signifcant

increases in OX-42 staining in rats that

developed CNP, in both gray and white

matter 28 days after SCI, after allodynia

measurements that confirmed that CNP

rats showed more than 50% decrease in the

pain thresholds. (n ¼ 3) (mean ± SD,

p < 0.05). (c) The expression of LCA was

observed using the monoclonal antibody

OX-1. OX-1 immunolabeling in sham (n ¼3) and CNP spinal cords (n ¼ 3) 3 months

after SCI, in spinal segments spanning from

T7–T9, above the site of injury (T10), shows

robust up-regulation of LCA in CNP spinal

cords, in activated microglia [round cells

with morphological features typical of acti-

vated microglia/macrophages, see (a)].

1010 O. Nesic et al.

� 2005 International Society for Neurochemistry, J. Neurochem. (2005) 95, 998–1014

rectifying K channels (Bordey et al. 2000), urea transporter(Berger et al. 1988), etc.

In response to CNS pathology, astrocytes become hyper-trophic and start expressing genes encoding intermediatefilament protein GFAP, whose up-regulation is the mainmarker of astrogliosis. Therefore, persistently increasedexpression of GFAP protein levels in the CNP spinal cordsversus non-CNP spinal cords, with already-elevated GFAPlevels compared with sham spinal cords, is a directconfirmation of astrocytic over-activation (Figs 6, 7 and 8).This increase lasts for at least 9 months after SCI, andspreads away from the site of injury throughout injuredspinal cords (Fig. 9a), consistent with the chronic nature ofCNP spreading over the wide body regions (Siddall et al.2002).

This result was additionally supported by increases inS100b (Fig. 8c) and AQP4 proteins (Figs 9b and c) that wereconsistent with changes in GFAP. S-100 is a 20-kDa Ca2+-binding protein composed of a and b subunits (S100a andS100b), primarily expressed by astrocytes in the brain (Huet al. 1996). In a variety of pathologic conditions, such asischemic brain damage (Kim et al. 1996), cerebrospinaltrauma (Hinkle et al. 1997), and Alzheimer’s disease (Shenget al. 1996), accumulation of the S100 in and around thelesion has been reported. AQP4 is an astrocytic waterchannel protein that is typically up-regulated in activatedastrocytes, as shown in brain trauma (Saadoun et al. 2002),and parallels GFAP expression changes in cultured astrocytes(Yoneda et al. 2001).

Especially intriguing was the observation of increasedGFAP protein degradation (Figs 6 and 8–10) which has alsobeen shown in aging and demented brain (Sloane et al. 2000;Porchet et al. 2003). This phenomenon is not well under-stood, but is believed to result from the activation ofCa-dependent enzymes, such as calpain, and may have a rolein neurodegeneration, not yet characterized. Given that a lackof GFAP in GFAP (GFAP–/–) null mice results in behavioraland functional deficits (Pekny and Pekna 2004), we canspeculate that chronically activated astrocytes with increasedGFAP degradation (Fig. 8b) might be dysfunctional and thuscontribute to CNP development, as hypothesized here.

Among the mRNAs that were up-regulated in the CNPversus non-CNP spinal cords, we found a group of immu-nologically relevant genes whose products are knowninitiators of astrocytic proliferation and activation (Table 2).Several studies indicate that transforming growth factor-b(TGF-b) could be a potent initiator of astrocytic activation.Injections of TGFb into the injured cortex elicit strongastrogliotic reactions, with increased expression of GFAP,laminin, and fibronectin (the same molecules are found up-regulated in CNP rats, see Table 2). Blocking TGFbinhibitied astrogliosis and expression of these molecules(Logan et al. 1994, 1999). We found that mRNA levels ofseveral components of TGFb signaling were significantly

up-regulated in spinal cords of rats that developed CNP (themost robustly up-regulated genes in the list). Fibroblastgrowth factor-2 (FGF-2) is a member of a multigene familyof growth factors that most prominently affects astroglialproliferation, maturation and transition to a reactive pheno-type in vitro and, after exogenous administration, in vivo.FGF-2 is apparently the major regulator of GFAP, as micedeficient for FGF-2 have reduced GFAP expression in thebrain (Reuss et al. 2003).

Activated astrocytes produce cytokines and chemokinesthat might be involved in initiating sensitization in painprocessing neurons, most notably pro-inflammatory cytok-ines (Watkins et al. 1994, 1997; Laughlin et al. 2000;Milligan et al. 2001; Plunkett et al. 2001; Sweitzer et al.2001; Raghavendra et al. 2003; Milligan et al. 2003). DNAmicroarray analyses have not detected increased mRNAlevels of the main pro-inflammatory cytokines: IL-1, TNFaor IL-6, in CNP versus non-CNP rats. However, that does notrule out the possibility that differential cytokine expressionlevels exist between these two groups, as cytokines aretypically expressed below the detection levels of DNAmicroarrays (Nesic et al. 2002), except in the acute phase ofinflammatory reactions to SCI (Nesic et al. 2001, 2002). Wedid find changes in the expression of mRNAs necessary forIL-1 signaling: Caspase 1 (ICE) and interleukin 1 receptoraccessory protein (IL-1racP) (Table 2). Caspase 1 is anenzyme that cleaves the inactive pro-IL-1b, thus enablingactive IL-1b to bind to its cognate receptor. When activatedby IL-1, the interleukin receptor type I (IL-1RI) transducessignals in cooperation with the IL-1racP. IL-1RAcP isexpressed widely in the CNS (Liu et al. 1996) on astrocytes(Zetterstrom et al. 1998) and microglia (Pinteaux et al.2002). It has been shown that mutant strains IL-1racPKOmice display lower pain sensitivity compared with the parentstrains, using the hot-plate test (Woolf and Salter 2000),suggesting an important role for IL-1racP in pain processingthat has not been characterized in any model of CNP. Theincreased expression of IL-1RAcP and ICE mRNAs that wedetected in the spinal cords of CNP rats might indicate anincreased efficacy of IL-1 signaling, even in the presence ofan unchanged production of IL-1R ligands, a novel hypo-thesis not tested before.

Another important result of the DNA microarray analysespresented here is that such significant and uniformlyorchestrated transcriptional reactions of enlarged and/orproliferating astrocytes and microglia/macrophages mostlikely mask CNP-inducing transcriptional changes in other,smaller neuronal populations, that may be most critical topain processing. Astrocytes are the largest cell population inthe normal CNS, and after SCI-induced neuronal andoligodenrocyte depletion, the percantage of astrocytes inthe overall cell population increases. Thus, transcriptionalchanges detected in two groups of injured spinal cord(homogenates) reflect primarily gene expression changes in

Reactive astrocytes and SCI-induced CNP 1011

� 2005 International Society for Neurochemistry, J. Neurochem. (2005) 95, 998–1014

reactive astrocytes. For example, expression changes ingenes involved in regulation of cell excitability, such asneurotransmitter receptors (GABA, glutamate or serotoninreceptors), or SNARE molecules involved in exocytosis(SNAP, syntaxin) (Table 1), that are detected by DNAmicroarray analysis of spinal cord homogenates are mostlikely taking place in astrocytes. It has already been shownthat GABA, glutamate or serotonin receptors are not onlyexpressed but also up-regulated in reactive astrocytes in vitro(Nakagawa and Schwartz 2004), similarly to the SNAREexocytotic machinery (Wilhelm et al. 2004).

Thus, we hypothesize that significant SCI-induced chan-ges in astrocytic functions reflected in the gene expressionchanges listed in Tables 1 and 2 (water and ion transport,K buffering, glutamate uptake and release, or responsivenessto neurotransmitters) might be critically important for CNPdevelopment. Dysfunctional astrocytes may contribute to thedevelopment of chronic inflammatory conditions, in part byaffecting BSCB properties (Table 2) that, in turn, maintainastrocytes in a persistently activated state. Such a self-amplifying feedback loop between dysfunctional astrocytesand persistent inflammation would result in chronic produc-tion of pain-inducing molecules that maintain hyperexcita-bility in pain processing pathways and ensure generation ofchronic CNP after SCI. Identifying drugs that can intervenein this feedback loop will be critically important in devel-oping therapy for SCI patients with chronic CNP.

Acknowledgements

Supported in part by grants from the Mission Connect (TIRR) and

NINDS.

References

Abbott N. J. (2002) Astrocyte–endothelial interactions and blood–brainbarrier permeability. J. Anat. 200, 629–638.

Bareyre and Schwab (2003) Inflammation, degeneration and regener-ation in the injured spinal cord: insights from DNA microarrays.Trends Neurosci. 26, 555–563.

Basso D. M., Beattie M. S. and Bresnahan J. C. (1995) A sensitive andreliable locomotor rating scale for open field testing in rats.J. Neurotrauma 12, 1–21.

Basso D. M., Beattie M. S. and Bresnahan J. C. (1996) Graded histo-logical and locomotor outcomes after spinal cord contusion usingNYU weight drop device versus transection. Exp. Neurol. 139,244–256.

Berger U. V., Tsukaguchi H. and Hediger M. A. (1988) Distribution ofmRNA for the facilitated urea transporter UT3 in the rat nervoussystem. Anat. Embryol. (Berl.) 197, 405–414.

Bethea J. R. and Dietrich W. D. (2002) Targeting the host inflammatoryresponse in traumatic spinal cord injury. Curr. Opin. Neurol. 15,355–360.

Bilgen M., Dogan B. and Narayana P. A. (2002) In vivo assessment ofblood–spinal cord barrier permeability: serial dynamic contrastenhanced MRI of spinal cord injury. Magn. Reson. Imaging 20,337–341.

Bordey A., Hablitz J. J. and Sontheimer H. (2000) Reactive astrocytesshow enhanced inwardly rectifying K+ currents in situ. Neurore-port 11, 3151–3155.

Bunge R. P., Puckett W. R., Beccera J. L., Marcillo A. and Quencer R.M. (1993) Observations on the pathology of human spinal cordinjury. A review and classification of 22 new cases with detailsfrom a case of chronic cord compression with extensive focaldemyelination, in Advances in Neurology Vol. 59 (Seil F. J. ed.),pp. 75–89. Raven Press, New York.

Bush T. G., Puvanachandra N., Horner C. H., Polito A., Ostenfeld T.,Svendsen C. N., Mucke L., Johnson M. H. and Sofroniew M. V.(1999) Leukocyte infiltration, neuronal degeneration, and neuriteoutgrowth after ablation of scar-forming, reactive astrocytes inadult transgenic mice. Neuron 23, 297–308.

Cairns D. M., Adkins R. H. and Scott M. D. (1996) Pain and depressionin acute traumatic spinal cord injury: origins of chronic problematicpain? Arch. Phys. Med. Rehab. 77, 329–335.

Carmel J. B., Galante A., Soteropoulos P., Tolias P., Recce M., Young W.and Hart R. P. (2001) Gene expression profiling of acute spinalcord injury reveals spreading inflammatory signals and neuronloss. Physiol. Genomics 7, 201–213.

Croll S. D., Ransohoff R. M., Cai N. et al. (2004) VEGF-mediatedinflammation precedes angiogenesis in adult brain. Exp. Neurol.187, 388–402.

Davidoff G. and Roth E. J. (1991) Clinical characteristics of central(dysesthetic) pain in spinal cord injury patients, in Pain andCentral Nervous System Disease: the Central Pain Syndromes(Casey K. L. ed.) pp. 77–83. Raven Press, New York.

De Maio A., Mooney M. L., Matesic L. E., Paidas C. N. and Reeves R.H. (1998) Genetic component in the inflammatory responseinduced by bacterial lipopolysaccharide. Shock 10, 319–323.

De Maio A., Torres M. B. and Reeves R. H. (2005) Genetic determinantsinfluencing the response to injury, inflammation, and sepsis. Shock23, 11–17.

DeLeo J. A., Colburn R. W., Nichols M. and Malhotra A. (1996)Interleukin-6-mediated hyperalgesia/allodynia and increased spinalIL-6 expression in a rat mononeuropathy model. J. InterferonCytokine Res. 16, 695–700.

Di Giovanni S., Knoblach S. M., Brandoli C., Aden S. A., Hoffman E. P.and Faden A. I. (2003) Gene profiling in spinal cord injury showsrole of cell cycle neuronal death. Annals of Neurology 53, 454–468.

Diatchenko L., Slade G. D., Nackley A. G. et al. (2005) Genetic basisfor individual variations in pain perception and the development ofa chronic pain condition. Hum. Mol. Genet. 14, 135–143.

Gao F., Bales K. R., Dodel R. C., Liu J., Chen X., Hample H., Farlow M.R., Paul S. M. and Du Y. (2002) NF-jB mediates IL-1b-inducedsynthesis/release of a2-macroglobulin in a human glial cell line.Brain Res. Mol. Brain Res. 105, 108–114.

Garrison C. J., Dougherty P. M. and Carlton S. M., (1994) GFAPexpression in lumbar spinal cord of naive and neuropathic ratstreated with MK-801. Exp. Neurol. 129, 237–243.

Gruner J. A. (1992) A monitored contusion model of spinal cord injuryin the rat. J. Neurotrauma 9, 123–128.

Hinkle D. A., Baldwin S. A., Sheff S. W. and Wise P. M. (1997) GFAPand S100[beta] expression in the cortex and hippocampus inresponse to mild cortical contusion. J. Neurotrauma 14, 729–738.

Hu J., Castets F., Guevara J. L. and Van Eldik L. J. (1996) S100[beta]stimulates inducible nitric oxide synthase activity and mRNAlevels in rat cortical astrocytes. J. Biol. Chem. 271, 2543–2547.

Hulsebosch C. E. (2002) Pharmacology of chronic pain after spinal cordinjury: novel acute and chronic intervention strategies, in SpinalCord Injury Pain: Assessment, Mechanisms, Management. Pro-

1012 O. Nesic et al.

� 2005 International Society for Neurochemistry, J. Neurochem. (2005) 95, 998–1014

gress in Pain Research and Management, Vol. 23 (Yezierski R. P.and Burchiel K. J. eds), pp. 189–204. IASP Press, Seattle.

Hulsebosch C. E., Xu G.-Y., Perez-Polo J. R., Westlund K. N., Taylor C.P. and McAdoo D. J. (2000) Rodent model of chronic central painafter spinal cord contusion injury and effects of gabapentin.J. Neurotrauma 17, 1205–1217.

Itagaki S., McGeer P. L. and Akiyama H. (1988) Presence of T-cytotoxicsuppressor and leucocyte common antigen positive cells in Alz-heimer’s disease brain tissue. Neurosci. Lett. 91, 259–264.

Ji R. R. and Strichartz G. (2004) Cell signaling and the genesis ofneuropathic pain. Sci. STKE. 252, reE14.

Kieseier B. C., Pischel H., Neuen-Jacob E., Tourtellotte W. W. andHartung H. P. (2003) ADAM-10 and ADAM-17 in the inflamedhuman CNS. Glia 42, 398–405.

Kim J. S., Yoon S. S., Kim Y. H. and Ryu J. S. (1996) Serial measure-ment of interleukin-6, transforming growth factor-beta, and S-100protein in patients with acute stroke. Stroke 27(9), 1553–1557.

Krady J. K., Basu A., Levison S. W. and Milner R. J. (2002) Differentialexpression of protein tyrosine kinase genes during microglialactivation. Glia 40, 11–24.

Laughlin T. M., Bethea J. R., Yezierski R. P. and Wilcox G. L. (2000)Cytokine involvement in dynorphin-induced allodynia. Pain 84,159–167.

Lee S. W., Kim W. J., Choi Y. K., Song H. S., Son M. J., Gelman I. H.,Kim Y. J. and Kim K. W. (2003) SSeCKS regulates angiogenesisand tight junction formation in blood–brain barrier. Nat. Med. 9,900–906.

Lindsey A. E., LoVerso R. L., Tovar C. A., Hill C. E., Beattie M. S.and Bresnahan J. C. (2000) An analysis of changes in sensorythresholds to mild tactile and cold stimuli after experimentalspinal cord injury in the rat. Neurorehabil. Neural Repair 14,287–300.

Liu W., Shafit-Zagardo B., Aquino D. A., Zhao M. L., Dickson D. W.,Brosnan C. F. and Lee S. C. (1994) Cytoskeletal alterations inhuman fetal astrocytes induced by interleukin-1b. J. Neurochem.63, 1625–1634.

Logan A., Berry M., Gonzalez A. M., Frautschy S. A., Sporn M. B. andBaird A. (1994) Effects of transforming growth factor b1 on scarproduction in the injured central nervous system of the rat. Eur. J.Neurosci. 6, 355–363.

Logan A., Green J., Hunter A., Jackson R. and Berry M. (1999)Inhibition of glial scarring in the injured rat brain by a recombinanthuman monoclonal antibody to transforming growth factor-beta2.Eur. J. Neurosci. 11(7), 2367–2374.

Ludwin S. K., Kosek J. C. and Eng L. F. (1976) The topographicaldistribution of S-100 and GFA proteins in the adult rat brain: animmunohistochemical study using horseradish peroxidase-labelledantibodies. J. Comp. Neurol. 165, 197–207.

van de Meent H., Hamers F. P., Lankhorst A. J., Buise M. P., Joosten E.A. and Gispen W. H. (1996) New assessment techniques forevaluation of post-traumatic spinal cord function in the rat.J. Neurotrauma 12, 741–754.

Milligan E. D., O’Connor K. A., Nguyen K. T., Armstrong C. B.,Twining C., Gaykema R. P., Holguin A., Martin D., Maier S. F. andWatkins L. R. (2001) Intrathecal HIV-1 envelope glycoproteingp120 induces enhanced pain states mediated by spinal cord pro-inflammatory cytokines. J. Neurosci. 21, 2808–2819.

Milligan E. D., Twining C., Chacur M., Biedenkapp J., O’Connor K.,Poole S., Tracey K., Martin D., Maier S. F. and Watkins L. R.(2003) Spinal glia and proinflammatory cytokines mediate mirror-image neuropathic pain in rats. J. Neurosci. 23, 1026–1040.

Mills C. D., Hains B. C., Johnson K. M. and Hulsebosch C. E. (2001)Strain and model differences in behavioral outcomes after spinalcord injury in rat. J. Neurotrauma 18, 743–756.

Mills C. D., Johnson K. M. and Hulsebosch C. E. (2002) Group Imetabotropic glutamate receptors in spinal cord injury: roles inneuroprotection and the development of chronic central pain.J. Neurotrauma 19, 23–42.

Mogil J. S. (1999) The genetic mediation of individual differences insensitivity to pain and its inhibition. Proc. Natl Acad. Sci. USA 96,7744–7751.

Muir E. M., Adcock K. H., Morgenstern D. A., Clayton R., von StillfriedN., Rhodes K., Ellis C., Fawcett J. W. and Rogers J. H. (2002)Matrix metalloproteases and their inhibitors are produced byoverlapping populations of activated astrocytes. Brain Res. Mol.Brain Res. 100, 103–117.

Nakagawa T. and Schwartz J. P. (2004) Gene expression patternsin in vivo normal adult astrocytes compared with culturedneonatal and normal adult astrocytes. Neurochem. Int. 45, 203–242.

Nesic O., Xu G.-Y., McAdoo D., Westlund High K., Hulsebosch C. andPerez-Polo R. J. (2001) IL-1 receptor antagonist prevents apoptosisand caspase-3 activation after spinal cord injury. J. Neurotrauma.18, 947–956.

Nesic O., Svrakic N., Xu G.-Y., McAdoo D., Westlund High K.,Hulsebosch C. and Regino Perez-Polo J. (2002) DNA microarrayanalysis of the contused spinal cord: the effect of NMDA receptorinhibition. J. Neurosci. Res. 68, 406–423.

Nesic-Taylor O., Cittelly D., Ye Z. et al. (2005) Exogenous Bcl-x(l)fusion protein spares neurons after spinal cord injury. J. Neurosci.Res. 79, 628–637.

Neumann H. (2001) Control of glial immune function by neurons. Glia36, 191–199.

Pang Y., Cai Z. and Rhodes P. G. (2001) Analysis of genes differentiallyexpressed in astrocytes stimulated with lipopolysaccharide usingcDNA arrays. Brain Res. 914, 15–22.

Pekny M. and Pekna M. (2004) Astrocyte intermediate filaments in CNSpathologies and regeneration. J. Pathol. 204, 428–437.

Pinteaux E., Parker L. C., Rothwell N. J. and Luheshi G. N. (2002)Expression of interleukin-1 receptors and their role in interleukin-1actions in murine microglial cells. J. Neurochem. 83, 754–763.

Plunkett J. A., Yu C.-G., Easton J. M., Bethea J. R. and Yezierski R. P.(2001) Effects of interleukin-10 (IL-10) on pain behavior and geneexpression following excitooxic spinal cord injury in the rat. Exp.Neurol. 168, 144–154.

Porchet R., Probst A., Bouras C., Draberova E., Draber P. and RiedererB. M. (2003) Analysis of glial acidic fibrillary protein in the humanentorhinal cortex during aging and in Alzheimer’s disease. Pro-teomics 8, 1476–1485.

Raghavendra V., Tanga F. and DeLeo J. A. (2003) Inhibition ofmicroglial activation attenuates the development but not existinghypersensitivity in a rat model of neuropathy. J. Pharmacol. Exp.Ther. 306, 624–630.

Reeve A. J., Patel S., Fox A., Walker K. and Urban L. (2000) Intra-thecally administered endotoxin or cytokines produce allodynia,hyperalgesia and changes in spinal cord neuronal responses tonociceptive stimuli in the rat. Eur. J. Pain 4, 247–257.

Reuss B., Dono R. and Unsicker K. (2003) Functions of fibroblastgrowth factor (FGF)-2 and FGF-5 in astroglial differentiation andblood–brain barrier permeability: evidence from mouse mutants.J. Neurosci. 23, 6404–6412.

Richards J. S., Meredith R. L., Nepomuceno C., Fine P. R. and BennettG. (1980) Psycho-social aspects of chronic pain in spinal cordinjury. Pain 8, 355–366.

Rossner S. (2004) New players in old amyloid precursor protein-pro-cessing pathways. Int. J. Dev. Neurosci. 7, 467–474.

Saadoun S., Papadopoulos M. C., Davies D. C., Krishna S. and Bell B.A. (2002) Aquaporin-4 expression is increased in oedematous

Reactive astrocytes and SCI-induced CNP 1013

� 2005 International Society for Neurochemistry, J. Neurochem. (2005) 95, 998–1014

human brain tumours. J. Neurol. Neurosurg. Psychiatry 72, 262–265.

Sandler and Tator C. H. (1976) Review of the effect of spinal cordtrauma on the vessels and blood flow in the spinal cord. J. Neuro-surg. 45, 638–646.

Scheff S.W., RabchevskyA.G., Fugaccia I.,Main J. A. and Lumpp J. E. Jr(2003)Experimentalmodeling of spinal cord injury: characterizationof a force-defined injury device. J. Neurotrauma 20, 179–193.

Segatore M. (1994) Understanding chronic pain after spinal cord injury.J. Neurosci. Nursing 26, 230–236.

Sheng J. G., Mrak R. E., Rvnaghi C. R., Koxlowska E., Van Eldik J. L.and Griffin W. S. (1996) Human brain S100-[beta] and S100-[beta]mRNA expression increases with age: pathogenic implications forAlzheimer’s disease. Neurobiol. Aging 17, 359–363.

Siddall P. J., Yezierski R. P. and Loeser J. D. (2002) Taxonomy andepidemiology of spinal cord injury pain, in Spinal Cord InjuryPain: Assessment, Mechanisms, Management. Progress in PainResearch and Management, Vol. 23 (Yezierski R. P. and BurchielK. J, eds), pp. 9–24. IASP Press, Seattle.

Sjolund B. H. (2002) Pain and rehabilitation after spinal cord injury: thecase of sensory spasticity? Brain Res. Brain Res. Rev. 40, 250–256.

Sloane J. A., Hollander W., Rosene D. L., Moss M. B., Kemper T. andAbraham C. R. (2000) Astrocytic hypertrophy and altered GFAPdegradation with age in subcortical white matter of the rhesusmonkey. Brain Res. 862, 1–10.

Sommer C. (2003) Painful neuropathies. Curr. Opin. Neurol. 16, 623–628.

Song G., Cechvala C., Resnick D. K., Dempsey R. J. and Rao V. L.(2001) GeneChip analysis after acute spinal cord injury in rat.J. Neurochem. 79, 804–815.

Stark J. L., Lyons J. A. and Cross A. H. (2004) Interferon-c produced byencephalitogenic cells induces suppressors of cytokine signaling inprimary murine astrocytes. J. Neuroimmunol. 151, 195–200.

Suzuki R., Arata S., Nakajo S., Ikenaka K., Kikuyama S. and Shioda S.(2003) Expression of the receptor for pituitary adenylate cyclase-activating polypeptide (PAC1-R.) in reactive astrocytes. Brain Res.Mol. Brain Res. 115, 10–20.

Svrakic N. M., Nesic O., Dasu M. R., Herndon D. and Perez-Polo J. R.(2003) Statistical approach to DNA chip analysis. Recent Prog.Horm. Res. 58, 75–93.

Sweitzer S., Martin D. and DeLeo J. A. (2001) Intrathecal interleukin-1receptor antagonist in combination with soluble tumor necrosisfactor receptor exhibits an anti-allodynic action in a rat model ofneuropathic pain. Neuroscience 103, 529–539.

Tachibana T., Noguchi K. and Ruda M. A. (2002) Analysis of geneexpression following spinal cord injury in rat using complementaryDNA microarray. Neurosci. Lett. 327, 133–137.

Tasker R. R. and Dostrovsky J. O. (1989) Deafferentation and centralpain, in Textbook of Pain, 2nd edn. (Wall P. D. and Melzack R.,eds), pp. 154–180. Churchill Livingstone, New York.

Tator C. H. (1995) Update on the pathophysiology and pathology ofacute spinal cord injury. Brain Pathol. 5, 407–413.

Tusher V. G., Tibshirani R. and Chu G. (2001) Significance analysis ofmicroarrays applied to the ionizing radiation response. Proc. NatlAcad. Sci. USA 98, 5116–5121.

VierckC. J. Jr, Siddall P. and Yezierski R. P. (2000) Pain following spinalcord injury: animal models and mechanistic studies. Pain 89, 1–5.

Watkins L. R., Wiertelak E. P., Goehler L. E., Smith K. P., Martin D. andMaier S. F. (1994) Characterization of cytokine-induced hyperal-gesia. Brain Res. 654, 15–26.

Watkins L. R., Martin D., Ulrich P., Tracey K. J. and Maier S. F. (1997)Evidence for the involvement of spinal cord glia in subcutaneousformalin induced hyperalgesia in the rat. Pain 71, 225–235.

Watkins L. R., Milligan E. D. and Maier S. F. (2003) Glial proinflam-matory cytokines mediate exaggerated pain states: implications forclinical pain. Adv. Exp. Med. Biol. 521, 1–21.

Wieseler-Frank J., Maier S. F. and Watkins L. R. (2004) Glial activationand pathological pain. Neurochem. Int. 45, 389–395.

Wilhelm A., Volknandt W., Langer D., Nolte C., Kettenmann H. andZimmermann H. (2004) Localization of SNARE proteins andsecretory organelle proteins in astrocytes in vitro and in situ.Neurosci. Res. 48, 249–257.

Woolf C. J. and Salter M. W. (2000) Neuronal plasticity: increasing thegain in pain. Science 288, 1765–1769.

Yezierski R. P. (2000) Pain following spinal cord injury: pathophysiol-ogy and central mechanisms. Prog. Brain Res. 129, 429–49.

Yoneda K., Yamamoto N., Asai K. et al. (2001) Regulation of aquapo-rin-4 expression in astrocytes. Brain Res. Mol. Brain Res. 18,94–102.

Zaheer A., Mathur S. N. and Lim R. (2002) Overexpression of glia mat-uration factor in astrocytes leads to immune activation of microgliathrough secretion of granulocyte-macrophage-colony stimulatingfactor. Biochem. Biophys. Res. Commun. 294, 238–244.

Zetterstrom M., Lundkvist J., Malinowsky D., Eriksson G. and Bartfai T.(1998) Interleukin-1-mediated febrile responses in mice andinterleukin-1b activation of NFjB in mouse primary astrocytes,involves the interleukin-1 receptor accessory protein. Eur. CytokineNetw. 9, 131–138.

1014 O. Nesic et al.

� 2005 International Society for Neurochemistry, J. Neurochem. (2005) 95, 998–1014