The present address of Dr. Y. Taketani is Department of Clinical Nutrition, Tokushima University, School of Medicine, Tokushima 770-8503, Japan.
Abbreviations used : COX, cyclooxygenase ; IL interleukin ; PG, prostaglandin ; TX, thromboxane.
Address correspondence and reprint requests to Dr. T. Tonai at Department of Orthopedic Surgery and Clinical Research Institute, National Zentsuji Hospital, 2-1-1 Sen-yu-cho, Zentsuji, Kagawa 765-0001, Japan.
Abstract : A standardized compression injury of rat spinal cord brought about a time-dependent biphasic production of thromboxane A2 (detected as thromboxane B2) and prostaglandin I2 (detected as 6-ketoprostaglandin F1α. Thromboxane B2 was predominant during the first 1 h, whereas the 6-ketoprostaglandin F1α level exceeded that of thromboxane B2 at 8 h postinjury. As examined by inhibitor experiments and northern blotting, cyclooxygenase-1 was responsible for the first phase, and cyclooxygenase-2 was involved in the second phase. On compression injury the levels of interleukin-1α and -1β detected as mRNA and protein increased and peaked at 2-4 h. Injection of exogenous interleukin-1 α into the spinal cord resulted in an increase of cyclooxygenase-2 mRNA content and a predominant production of 6-ketoprostaglandin F1α resembling the second phase of eicosanoid production. Concomitantly, extravascular migration of polymorphonuclear leukocytes was enhanced after the interleukin-1α injection. These cells together with vascular endothelial cells and glial cells were stained positively with an anti-cyclooxygenase-2 antibody. The results suggest that the immediate eicosanoid synthesis after spinal cord injury was due to the constitutive cyclooxygenase-1 and the delayed synthesis of eicosanoids was attributable to the induction of cyclooxygenase-2 mediated by interleukin-1 α.
Several pathologic changes, including hemorrhage, ischemia, edema, leukocyte infiltration, and neuronal death, occur sequentially after spinal cord injury (Tator and Fehlings, 1991). Because arachidonate metabolites (eicosanoids) possess various relevant biologic activities, their possible involvement in the pathologic events in CNS injury has been investigated extensively (Siesjö, 1992 ; Anderson and Hall, 1993).
Cyclooxygenase (COX) is a key enzyme initiating the formation of proaggregatory and vasoconstrictive thromboxane (TX)A2, as well as antiaggregatory and vasodilating prostaglandin (PG)I2 (Samuelsson et al., 1978 ; Moncada and Vane, 1979). Both compounds are very labile and are degraded rapidly to their respective stable metabolites, TXB2 and 6-keto-PGF1α, which are actually detected in biologic tissues. Two isoforms of COX are now well known : COX-1 is expressed constitutively in most tissues, whereas COX-2 is not detectable under physiologic conditions but is induced in various cells by cytokines, hormones, mitogens, and lipopolysaccharide (Herschman, 1996). Moreover, dexamethasone suppresses the expression of COX-2 without exerting a significant effect on COX-1 (Masferrer et al., 1992). Recently COX-2 induction in the CNS has been demonstrated in response to seizures and excitotoxin injection (Yamagata et al., 1993 ; Adams et al., 1996).
Enhanced eicosanoid synthesis has been reported in various pathologic processes in the CNS, such as brain infarction (Fagan et al., 1986), subarachnoid hemorrhage (O'Neil et al., 1992), and spinal cord injury (Demediuk et al., 1985 ; Hsu et al., 1985 ; Mitsuhashi et al., 1994 ; Nishisho et al., 1996). Hsu et al. (1985) demonstrated an increase in TXB2 content exceeding that of 6-keto-PGF1α and suggested that a TXA2-PGI2 imbalance is involved in the pathogenesis of microvascular thromboembolism in experimental spinal cord injury. The same investigators described biphasic profile of eicosanoid synthesis, with PGI2 predominating over TXA2 in the second phase (Hsu et al., 1988). However, the mechanisms underlying this biphasic synthesis and its pathophysiologic significance are still unknown.
Interleukin (IL)-1 is a proinflammatory cytokine produced by activated macrophages, fibroblasts, vascular smooth muscle cells, and endothelial cells. IL-1 exists in two forms, IL-1α and IL-1β, which exhibit amino acid sequence homology of only 26% but show similar biologic activities (Dinarello, 1989, 1991). IL-1 exerts various neurologic effects such as fever, induction of slow-wave sleep, analgesia, astroglial proliferation, and stimulation of corticotropin-releasing hormone secretion (Blatteis, 1990 ; Rothwell, 1991). Recently, IL-1 was detected in rat brain after percussion trauma (Taupin et al., 1993), and more recently IL-1β was found in rat spinal cord after contusion injury (Wang et al., 1997). In addition to the above effects, IL-1 is also active in inflammation of the CNS, disrupting the blood-brain barrier (Quagliarello et al., 1991) and inducing brain edema (Yamasaki et al., 1992). Thus, IL-1 may mediate pathophysiologic processes following CNS trauma. Furthermore, IL-1 stimulates eicosanoid synthesis by inducing COX-2 in various cell types (Herschman, 1996).
In this study we reinvestigated the mechanism of the time-dependent changes in eicosanoid synthesis after rat spinal cord injury observed earlier by Hsu et al. (1988) and found the induction of COX-2 mediated by IL-1 in the second phase.
MATERIALS AND METHODS
The radiolabeled compounds [α-32P]dCTP (110 TBq/mmol), [5,6,8,11,12,14,15-3H(N)]TXB2 (> 120 Ci/mmol), and 6-keto-[5,6,8,11,12,14,15-3H(N)]PGF1α (130-200 Ci/mmol), as well as Hybond-N+ nylon membrane and a Megaprime DNA labeling system, were purchased from Amersham International (Bucks, U.K.). Indomethacin, dexamethasone, 0-phenylenediamine tablets, 3,3′ -diaminobenzidine-4HCl tablets, and gum arabic were purchased from Sigma (St. Louis, MO, U.S.A.), 96-well immunoplates (type I) from Nunc (Roskilde, Denmark), rat IL-1β ELISA kits from Biosource International (Camarillo, CA, U.S.A.), and horseradish peroxidase-labeled goat anti-rabbit IgG from Bio-Rad Laboratories (Hercules, CA, U.S.A.). A DNA amplification kit was obtained from Perkin-Elmer Cetus (Norwalk, CT, U.S.A.), a DNA sequencing kit from U.S. Biochemical Co. (Cleveland, OH, U.S.A.), Isogen from Nippon Gene (Tokyo, Japan), glyceraldehyde 3-phosphate dehydrogenase cDNA probe from Clontech Laboratories (Palo Alto, CA, U.S.A.), and RAV-2 reverse transcriptase from Takara Shuzo (Kyoto, Japan). G-1 glass tubes were purchased from Narishige Scientific Instrument Laboratory (Tokyo), paraformaldehyde from Merck (Darmstadt, Germany), the Vectastain ABC kit, goat biotinylated anti-rabbit IgG, and normal goat serum from Vector Laboratories (Burlingame, CA, U.S.A.), and Wistar rats from Kagawa Science Co. (Kagawa, Japan). TXB2 and 6-keto-PGF1α were kindly provided by Ono Pharmaceutical Co. (Osaka, Japan), and NS-398 was from Taisho Pharmaceutical Co. (Saitama, Japan). Anti-TXB2 and 6-keto-PGF1α antibodies were prepared as described previously (Yamamoto et al., 1987).
Spinal cord injury model and intramedullary spinal cord injection
Wistar rats weighing 200-250g were anesthetized by intraperitoneal injection of sodium pentobarbital (40 mg/kg). The rats were positioned on the operating table, and an operating microscope was used. As previously described (Mitsuhashi et al., 1994), the spinal cord was compressed at T13 by placing a 20-g square compressor on the dura mater for 10 min. In sham-operated animals the same dural exposure procedure was performed, but the spinal cord was not compressed.
A partial laminectomy was performed at the T13 level for intraparenchymal injection of test solution. A microtube made from a G-1 glass tube was inserted into the parenchyma using a micromanipulator (model 151-P ; Narishige), and solution was injected into the spinal cord using a microinjector (model IM 5B ; Narishige). One microliter of test solution was injected over at least 1 min. Solutions of purified recombinant rat IL-1α (107 growth inhibitory factor units/mg of protein) at concentrations of 0, 0.02, 0.2, and 2 ng/μl of saline were injected.
Indomethacin (2 mg/ml) and dexamethasone (10 mg/ml) were dissolved in ethanol and administered intraperitoneally to rats. NS-398 (1 mg/ml) was dissolved in 5% gum arabic for oral administration. The animal experiments followed the guidelines of the ethical committee on the care and use of laboratory animals at National Zentsuji Hospital.
Radioimmunoassay for TXB2 and 6-keto-PGF1α
Radioimmunoassay of TXB2 and 6-keto-PGF1α in rat spinal cord was performed as described previously (Mitsuhashi et al., 1994). In brief, the standard solution or the sample extracted by ethanol from the spinal cord (T13 segment of 45-55 mg wet weight from one rat) was dissolved in sodium phosphate buffer (pH 7.3) and mixed with tritiated eicosanoid. Then the mixture was incubated with anti-eicosanoid antibody at 37°C for 1 h, followed by incubation at 4°C for 12 h. Radioactivity in the immunoprecipitate collected by addition of bovine γ-globulin and polyethylene glycol-6,000 was counted in a liquid scintillation counter, and the results were expressed as picomoles per gram of tissue. Cross-reactivity of the antiserum with other eicosanoids was <5%. The minimal detectable quantities of TXB2 and 6-keto-PGF1α were 7.4 and 11 pg, respectively.
Preparation of anti-IL-1α and anti-COX-2 antibodies
Female New Zealand white rabbits and goats were immunized with 50 (for rabbit) and 100 μg (for goat) of recombinant rat IL-1α emulsified with Freund's complete adjuvant (for the first injection) or incomplete adjuvant (for subsequent injections). Immunization was performed every 2 weeks for a total of seven times. Antisera were titrated by enzyme immunoassay. Rabbit anti-murine COX-2 antibody prepared with a synthetic peptide corresponding to amino acids 579-594 of the murine enzyme (Yamamoto et al., 1995) was provided by Dr. Yoko Hayashi of Tokushima University School of Medicine.
Enzyme immunoassay for IL-1
Spinal cord (T13 segment of 45-55 mg wet weight from one rat) was homogenized in 15 volumes of phosphate-buffered saline using a Sonifier (model 250 ; Branson). The homogenate was centrifuged at 11,000 g for 5 min. IL-1α in the supernatant was measured, and the result was expressed as nanograms per gram of tissue. Sandwich-type enzyme immunoassays for rat IL-1 α were performed as previously reported for human IL-1β (Tanaka et al., 1987). An immunoplate coated with goat anti-rat IL-1α antibody was incubated for 12 h with rat IL-1α or the samples to be tested and incubated further with rabbit anti-IL-1α antiserum for 2 h. After washing, horseradish peroxidase-labeled goat anti-rabbit IgG was added to the well and incubated for 2 h. After two washings, the peroxidase reaction was carried out with o-phenylenediamine for 3 min and stopped by addition of 1 M H2SO4. Absorbance at 492 nm was measured using an immunoplate reader. Concentrations of IL-1β in rat spinal cords were measured according to the manufacturer's instructions. The minimal detectable quantities of IL-1α and IL-1β were 10 and 3 pg, respectively.
Eight hours after injection of 1 μl of IL-1α (2 ng) or saline (as a control) into the T13 segment of spinal cord, all rats were anesthetized by pentobarbital (40 mg/kg) and killed by intracardiac perfusion with 0.1 M sodium phosphate buffer (pH 7.4), followed by the same buffer containing 4% paraformaldehyde for tissue fixation. One myelomere of the T13 segment was removed, immersed in the same buffered fixative for 8 h, and then sequentially rinsed with 10, 15, and 20% sucrose in phosphate-buffered saline.
Immunohistochemical staining was performed by the avidinbiotin complex method as reported previously (Tomlinson et al., 1994). The frozen samples were sectioned at 6 μm in the transverse plane. The specimen was incubated with rabbit anti-COX-2 antiserum (dilution, 1:300) for 12 h. Peroxidase was detected with 0.02% diaminobenzidine and 0.035% hydrogen peroxide. The sections were counterstained with hematoxylin. Negative control sections were prepared with nonimmune serum. For an immunoabsorption study the anti-COX-2 antiserum was incubated for 12 h with a synthetic COX-2 peptide (2 mg/ml) as described above.
Preparation of cDNA probes for COX-1, COX-2, IL-1α, and IL-1β
To isolate cDNA for rat COX-1 and COX-2, we synthesized oligonucleotide primers according to the published cDNA sequences (Feng et al., 1993). The sequences of the COX-1 and COX-2 primers were as follows : 5′-TGCTGGATGGAGAG-GTGTAC-3′ (sense, bases 773-792) and 5′-TTGCGATACT-GGAACTGGGC-3′ (antisense, bases 1,111-1,130) and 5′-TCCAGTATCAGAACCGCATTGC-3′ (sense, bases 1,070-1,091) and 5′-AGTAGACTCTTACAGCTC-3′ (antisense, bases 1,807-1,824), respectively. The position of the adenine of the initiation methionine codon was referred to as 1. Total RNA was prepared from injured rat spinal cord using Isogen and was allowed to react with 10 units of reverse transcriptase (Newman et al., 1988) in the presence of the PCR downstream primer. DNA amplification was carried out as described previously (Endo et al., 1995). The cDNA inserts of the isolated clones were sequenced by the dideoxynucleotide method (Sanger et al., 1977) using a modified T7 DNA polymerase (Sequenase ; U.S. Biochemical).
Rat IL-1β cDNA was prepared from the brain of lipopolysaccharide-injected rats, and IL-1α cDNA was prepared from peritoneal exudate cells of rats treated with 10% proteose peptone as described previously (Nishida et al., 1988). The cDNA fragments were labeled with [α-32P]dCTP by the Megaprime DNA labeling system according to the manufacturer's instructions.
Northern blot analysis
Total RNA was extracted from rat spinal cord (T13 segment of 45-55 mg wet weight from one rat) using Isogen as described above. Denatured RNA (20 μg) was separated by electrophoresis with 1% agarose gel containing formaldehyde. After electrophoresis, the RNAs were transferred to a Hybond N+ nylon membrane in 20× SSC (1× SSC = 0.15 M NaCl and 15 mM trisodium citrate) and then cross-linked to the membrane by ultraviolet light exposure. The membrane was prehybridized with rapid hybridization buffer (Amersham) at 65°C for 2 h and then hybridized with buffer containing the 32P-labeled probe at 65°C for 4 h. Then the membrane was washed twice in 0.2× SSPE (1× SSPE = 0.15 M NaCl, 10 mM NaH2PO4, and 1 mM EDTA) containing 0.1% sodium dodecyl sulfate at 65°C for 15 min. Radioactivity on the membrane was detected by a Fujix bioimaging analyzer (model BAS 2000). Stripping for rehybridization was carried out by washing the membrane in a buffer containing 10 mM Tris-HCl (pH 8.0), 1 mM EDTA, and 0.5% sodium dodecyl sulfate twice for 10 min at 95°C.
Data are expressed as mean ± SD values. Statistical comparisons were performed using the nonparametric Mann-Whitney U test for uncorrelated pairs. A p value < 0.05 was considered statistically significant.
Eicosanoid synthesis in spinal cord injury
After compression injury with a 20-g compressor for 10 min, the rat spinal cord was removed at various intervals, and TXB2 and 6-keto-PGF1α levels were measured by radioimmunoassay. TXB2 and 6-keto-PGF1α levels changed time-dependently with two peaks (Fig. 1). Starting from nearly undetectable levels (<2 pmol/g of tissue), the first peak of each metabolite was observed at 5-15 min (148.2 ± 14.9 and 30.0 ± 2.8 pmol/g of tissue, respectively). Then, levels of both eicosanoids decreased to basal values (<5 pmol/g of tissue) 2 h after injury. Subsequently, their levels increased again to reach the second peaks (16.3 ± 1.8 and 40.0 ± 4.7 pmol/g of tissue, respectively) at ~8 h, followed by a gradual decrease to the basal level by 36 h. The level of TXB2 was much higher than that of 6-keto-PGF1α in the first phase, whereas 6-keto-PGF1α was produced in a larger amount than TXB2 in the second phase. In contrast, the level of neither eicosanoid increased in sham-operated animals.
COX-1 and COX-2 gene expressions in spinal cord injury
We followed the time-dependent changes in amounts of COX-1 and COX-2 mRNAs by northern blottings. The level of COX-2 mRNA, which was essentially undetectable, started to increase on spinal cord injury and reached a maximal value (about ninefold increase at 4 h), followed by a subsequent decrease. COX-1 mRNA content increased only twofold at 8-12 h even if the increase was statistically significant (Fig. 2).
To determine which enzyme (COX-1 or COX-2) was responsible for the first and second increases in amounts of TXB2 and 6-keto-PGF1α, we used two COX inhibitors : NS-398, specific for COX-2 (Futaki et al., 1993), and indomethacin, inhibiting both COX-1 and COX-2 (Fig. 3). When NS-398 or indomethacin was administered to rats 2 h before spinal cord injury, indomethacin inhibited the first increase of TXB2 by 86% and that of 6-keto-PGF1α by 69%, respectively. NS-398 failed to inhibit the first increase in amount of either eicosanoid. In contrast, the second increases in amounts of both eicosanoids were inhibited when either inhibitor was administered 30 min before spinal cord injury. These results indicate that the first increase of eicosanoid synthesis is attributed to COX-1 and the second phase to COX-2.
Effect of dexamethasone on eicosanoid synthesis and COX-2 mRNA expression
Dexamethasone (10 mg/kg) was injected intraperitoneally in rats before spinal cord injury, and TXB2 and 6-keto-PGF1α levels were measured. As shown in Fig. 4, the first increases in content of TXB2 and 6-keto-PGF1α were slightly but not significantly decreased by dexamethasone treatment. In contrast, dexamethasone treatment significantly reduced the second increase in level of TXB2 to 62% and that of 6-keto-PGF1α to 64% of those in vehicle-injected controls. The relative COX-2 mRNA level of dexamethasone-treated animals was also decreased to 62 and 59% in two separate experiments, respectively, 8 h after the injury compared with vehicle-treated animals.
IL-1 synthesis and its mRNA expression in spinal cord injury
As shown in Fig. 5A, on rat spinal cord injury the IL-1α and -1β levels increased up to a peak at 4 h (1.49 ± 0.33 and 6.5 ± 1.3 ng/g of tissue, respectively) and then gradually decreased. IL-1α content returned to the basal level by 24 h after injury, but IL-1β was still above the basal level at that time. Sham-operated animals showed no increase in either IL-1α or -1β synthesis. By northern analysis IL-1α and -1β mRNAs were detected as early as 1 h after injury, reached a maximum with a 23- and 18.3-fold increase, respectively at 2 h, and declined to the baseline levels by 8 h (Fig. 5B).
COX-2 mRNA expression and eicosanoid production enhanced by injection of IL-1α into rat spinal cord
On the basis of these findings with IL-1, COX-2 expression in rat spinal cord was examined after intraparenchymal injection of various doses of recombinant IL-1α. On northern blotting, COX-2 mRNA levels increased in a dose-dependent manner, showing a 7.8-fold increase with injection of 0.2 ng of IL-1α and a 12-fold increase with 2 ng of IL-1α (Fig. 7A). When rat spinal cord was removed at various intervals after intraparenchymal injection of 2 ng of IL-1α, the COX-2 mRNA level increased dramatically, reaching a peak at 2 h, and then decreased (Fig. 7B).
As shown in Table 1, intraparenchymal injection of 2 ng of IL-1α resulted in enhanced synthesis of both TXB2 and 6-keto-PGF1α at a time corresponding to the second phase of the eicosanoid response described above. 6-Ke-to-PGF1α was produced in a larger amount than TXB2.
Table 1. Eicosanoid synthesis after IL-1α injection into spinal cordSpinal cord was removed 8 h after intraparenchymal injection of IL-1 α (2 ng) or saline (control). TXB2 and 6-keto-PGF1α were extracted from the spinal cord and quantified by radioimmunoassay. Data are mean ± SD values (n = 5).
When the spinal cord was examined immunohistochemically at 8 h after injury, the tissue and individual cells were already severely damaged, and no clear expression of COX-2 was detected in any particular cell. When the spinal cord was stained 8 h after intraparenchymal injection of 2 ng of IL-1α, COX-2 antigen was detected in polymorphonuclear leukocytes, glial cells, and vascular endothelial cells (Fig. 8). Adjacent sections processed with normal rabbit serum were not stained. In control rats, which received saline instead of IL-1α, COX-2 antigen was not detected. The absorption test, where anti-COX-2 antiserum was preincubated with synthetic COX-2 peptide (2 mg/ml), gave negative staining.
Eicosanoids have been implicated in the pathogenesis of spinal cord injury. Several investigators have reported a predominant production of TXA2 as opposed to PGI2 shortly after injury (Demediuk et al., 1985 ; Hsu et al., 1985 ; Mitsuhashi et al., 1994), suggesting that an imbalanced production of these two eicosanoids results in a secondary process causing spinal cord injury such as ischemia. Following the changes in levels of TXB2 and 6-keto-PGF1α for a longer period than other investigators, Hsu et al. (1988) found two peaks of eicosanoid production representing an immediate phase and also a delayed phase. However, the mechanism producing a biphasic pattern of eicosanoid synthesis had not been fully elucidated.
In the present study, we reproduced the biphasic profile of eicosanoid productions previously shown by Hsu et al. (1988). Our data attributed the first increase in levels of eicosanoids to constitutively present COX-1 and the second increase to inducible COX-2 induction. By northern analysis, COX-2 mRNA was not detectable before spinal cord injury but then increased in level to peak at 4 h, corresponding to the second phase of eicosanoid production, followed by a gradual decline. Dexamethasone inhibited both COX-2 mRNA expression and the delayed increase of eicosanoid production. These findings were in agreement with the other results that indomethacin, a dual inhibitor of COX-1 and -2, suppressed both phases, whereas NS-398, a selective inhibitor of COX-2, suppressed only the delayed eicosanoid peak.
Because IL-1 is well known as an inflammatory cytokine and an inducer for COX-2, we examined whether or not COX-2 induction was associated with IL-1α and -1β gene expression. Both IL-1α and -1β mRNAs were detected ~1 h after injury, with their levels peaking at 2 h. IL-1α and -1β could be detected by enzyme immunoassay before the second peak of eicosanoid production, with a maximal level 4 h after injury. Moreover, intraparenchymal injection of IL-1α induced COX-2 mRNA accompanied by predominant production of 6-keto-PGF1α rather than TXB2. This is a profile similar to that seen in the second phase of eicosanoid production following spinal cord injury. Thus, IL-1 may mediate the COX-2 induction, but cooperation of other cytokines cannot be ruled out.
COX-2 induction by IL-1α injection into the spinal cord was also shown immunohistochemically with strong staining of migrating polymorphonuclear leukocytes, vascular endothelial cells, and glial cells. This observation is consistent with previous reports that COX-2 expression is detected in various cell types, including carrageenin-treated polymorphonuclear leukocytes (Tomlinson et al., 1994), and IL-1α-treated vascular endothelial cells (Ristimäki et al., 1994) and glial cells (Nam et al., 1995). Vascular endothelial cells therefore might be the main site of the 6-keto-PGF1α production in the second phase in spinal cord injury. Alternatively, it seems likely that neutrophil action such as migration in response to IL-1 plays an important role in the acute inflammatory reaction caused by spinal cord injury.
Recently, much attention has been focused on the role of COX-2 in inflammation. Glucocorticoids have potent antiinflammatory effects and are known to attenuate IL-1 formation and arachidonate metabolism by suppressing COX-2 induction. However, in our experiments dexamethasone suppressed the production of eicosanoids and the expression of IL-1 and COX-2 genes by only 40-50%. A role of a still unknown factor (s) other than IL-1 remains to be elucidated.
Other proinflammatory effects of IL-1 include stimulated proliferation of fibroblast (Postlethwaite et al., 1984) and astroglial (Giulian and Lachman, 1985) cells, and a role of IL-1 is implicated in the promotion of inflammation and glial scar formation. PGE2 and PGI2 potently inhibit IL-1 synthesis (Knudsen et al., 1986). In our study the induction of COX-2 may be mediated by IL-1 and leads to the increase in PGI2 production, which may limit IL-1 synthesis by a negative feedback mechanism. PGI2 also is effective in preventing fibrosis in liver injury (Itoh et al., 1992) and attenuating neuronal damage (Cazevieille et al., 1993). Therefore, the predominant PGI2 production in the second synthetic phase after spinal cord injury may limit the extension of inflammation and glial scaring caused by IL-1 and may also protect the nervous system from neuronal damage.
Expression of IL-1β in spinal cord injury was reported previously (Wang et al., 1997). The present study is the first report to associate the expression of IL-1α with the induction of COX-2 resulting in a biphasic production of eicosanoid. However, it is unknown whether the association between IL-1 and COX-2 extends the inflammation or facilitates the wound repair. It was reported in the Second National Acute Spinal Injury Study (Bracken et al., 1990) that very high doses of methylprednisolone improved neurologic outcome in a randomized controlled clinical trial. However, the clinical use of glucocorticoids may cause various adverse reactions as well as inhibiting induction of various neurotrophic factors (Foreman et al., 1992 ; Cosi et al., 1993). Suppression of the detrimental effects and enhancement of the beneficial effects of IL-1 and COX-2 may improve outcome in spinal cord injury. Hamada et al. (1996) suppressed detrimental effects of IL-1 by the use of a monoclonal antibody for intercellular adhesion molecule-1, reducing motor disturbance and enhancing neurologic recovery in rat spinal cord injury. Further studies on the roles of IL-1, COX-2, and other biofactors, including adhesion molecules, are required for a better clinical treatment of patients with spinal cord injury.