Address correspondence and reprint requests to Dr Klaus Dinkel, Department of Biological Sciences, Gilbert Hall, Stanford University, Stanford, CA 94305-5020, USA. E-mail: email@example.com
The CNS can mount an inflammatory reaction to excitotoxic insults that contributes to the emerging brain damage. Therefore, anti-inflammatory drugs should be beneficial in neurological insults. In contrast, glucocorticoids (GCs), while known for their anti-inflammatory effects, can exacerbate neurotoxicity in the hippocampus after excitotoxic insults. We investigated the effect of GCs on the inflammatory response after a neurological insult. Intact control (INT; intact stress response GC profile), adrenalectomized/GC-supplemented (ADX; low basal GC profile) and GC-treated (COR; chronically high GC profile) rats were injected with kainic acid into the hippocampal CA3 region. Lesion size was determined 8–72 h later. The inflammatory response was characterized using immunohistochemistry, RNAse protection assay and ELISA. The INT and COR rats developed larger CA3 lesions than ADX rats. We found that GCs surprisingly caused an increase in relative numbers of inflammatory cells (granulocytes, monocytes/macrophages and microglia). Additionally, mRNA and protein (IL-1β and TNF-α) levels of the pro-inflammatory cytokines IL-1α, IL-1β and TNF-α were elevated in COR rats compared with INT and ADX rats. These data strongly question the traditional view of GCs being uniformly anti-inflammatory and could further explain how GCs worsen the outcome of neurological insults.
Brain damage as a consequence of a neurological insult, such as stroke, trauma or seizure, develops from a complex series of pathophysiological events over time. Excessive activation of neurons by excitatory neurotransmitters (e.g. glutamate), which are massively released as a consequence of energy depletion, results in excitotoxic neuron death (Beal 1992; Lipton and Rosenberg 1994; Whetsell 1996). Accumulating evidence during the last decade has shown that brain injury is also accompanied by a marked inflammatory reaction characterized by infiltration of granulocytes and monocytes/macrophages into the respective brain parenchyma, activation of resident brain cells (e.g. microglia and astrocytes) and expression of pro-inflammatory cytokines, adhesion molecules and other inflammatory mediators (Perry and Gordon 1991; Feuerstein et al. 1998b; Dirnagl et al. 1999). There is considerable evidence that the inflammation contributes significantly to the developing neuronal damage by mechanisms like the release of neurotoxic substances, such as free radicals or cytokines (Barone and Feuerstein 1999; Rothwell et al. 1996; McGeer and McGeer 1999). Accordingly, anti-inflammatory drugs should have a beneficial effect in the context of a neurological insult.
Glucocorticoids should ameliorate the damage during brain injury by suppressing the inflammatory reaction but instead have been found to increase neuron loss. Therefore, we investigated in this study the effect of GCs on: (i) lesion size, (ii) cellular inflammatory infiltrate and (iii) cytokine pattern after excitotoxic insult in vivo over the course of time. The glutamate agonist kainic acid (KA) was injected into the CA3 area of the hippocampus of untreated control (INT; intact stress response; GC levels in the normal range of physiological stress response), adrenalectomized/GC-supplemented (ADX; no stress response; low basal GC levels) and corticosterone-treated (COR; chronically high GC levels in the upper stress level) rats.
Materials and methods
Animals and materials
Male Sprague–Dawley rats (weighing 250–300 g; Simonsen Laboratories, Gilroy, CA, USA) were housed under a 12-h light–dark cycle with free access to food and water. Corticosterone (CORT), KA, metyrapone (met; 2-methyl-1,2-di-3-pyridyl-1-propanone), paraformaldehyde, bovine serum albumin (BSA) and FASTTM-diaminobenzidine (FAST-DAB) were purchased from Sigma (St Louis, MO, USA). All antibodies were purchased from BD PharMingen (San Diego, CA, USA).
All experiments were conducted under the guidelines described in the US Public Health Service Policy on Human Care and Use of Laboratory Animals. Four groups of rats (INT, ADX, COR and INT + met) were used in the experiments, each group displaying a different GC level. In the control group (INT) no GC manipulations were performed; thus animals had basal GC levels (approximately 1–10 µg/dL) throughout except for the physiological stress response to surgery and the subsequent KA-induced seizures; this involved an increase of approximately 30 µg/dL above baseline by 30 min after seizure onset and levels 10–15 µg/dL above baseline for the next 6 h before returning to baseline (Stein and Sapolsky 1988). In another group, rats were bilaterally adrenalectomized (ADX) under halothane anesthesia and a 15% CORT pellet was implanted s.c. to generate basal GC levels of approximately 6 µg/dL (Stein-Behrens et al. 1994). Rats were given 3 days for hormone levels to stabilize prior to stereotaxic surgeries. In a subset of experiments, intact rats were treated 1 h before KA injection with the steroid synthesis inhibitor metyrapone (200 mg/kg in 0.9% NaCl, s.c.). In these rats (INT + met), GC levels were locked to basal levels of approximately 6–7 µg/dL (Stein and Sapolsky 1988). In the fourth group (COR) of rats, chronically high levels of CORT of approximately 28 µg/dL were generated for approximately 20 h/day by injecting rats daily for 3 days with 3 mg CORT/day s.c. in sesame oil (Stein-Behrens et al. 1994). These injections continued after microinfusion until the indicated timepoint. As GCs are steroid hormones, they readily penetrate the blood–brain barrier (BBB) and neuronal membranes, then interacting with their intracellular receptors.
Rats were anesthetized with an i.p. injection (1 mL/kg bodyweight) of ‘rodent cocktail’ (76 mg/mL ketamine, 1.5 mg/mL promace and 7.7 mg/mL xylazine) and placed in a stereotaxic frame. The rats were microinfused stereotaxically into the CA3 area of the hippocampus (A/P − 3.85, L/M ± 3.35 from bregma and D/V − 2.55 from dura). The CA3 area was defined as the pyramidal cell layer beginning in between the blades of the dentate gyrus, continuing around Ammon's horn until the thickening of the cell field that characterizes CA2. Animals were injected either with 0.06 µg KA dissolved in phosphate-buffered saline (PBS) or with PBS alone as a control.
Quantification of lesion size
At indicated times after microinfusion, rats were anesthetized by halothane inhalation and then, under deep anesthesia, perfused intracardially with a 0.1% heparin/0.9% saline solution followed by 3% paraformaldehyde solution. The brains were post-fixed in 3% paraformaldehyde (PFA) for 24 h and cryoprotected with PBS 15% sucrose. On a cryostat microtome, 30-µm coronal sections were cut, dried and stained with cresyl violet using the Nissl method. Lesion size in the CA3 area of the hippocampus was measured using a 10 × 10 optical grid in the ocular of the microscope at 40× and counting the number of pixels with damaged/missing cells. As an internal control, the non-lesioned hemisphere of each section was used to measure the size of the intact CA3 area, which was set to 100%. Counts were converted into area measurements and lesion size was expressed as percentage damage of the intact contralateral CA3 area. Starting from the visible needle track, counts were taken from at least four coronal sections at ≈ 0.1 mm increment. All sections were scored blind using a light microscope. This technique produces assessments of lesion size which correlate significantly with more arduous cell counting (Sapolsky and Stein 1989).
Immune cell immunohistochemistry
At specific times after microinfusion, animals (n = 6 rats/group and time) were killed by inhalation of excess halothane and then decapitated. The brains were removed immediately and quick frozen in 2-methylbutane at −42°C for 3 min. Cryostat sections (15 µm) were cut, mounted on slides, dried and kept at −70°C until use. Slides were fixed in icecold acetone at −20°C for 3 min, treated with 0.03% H2O2 solution for 10 min at room temperature (rt) to block endogenous peroxidase activity and then blocked with 5% normal goat serum or 3% BSA solution for 15 min at rt. Sections were incubated with the respective primary antibody (see Table 1) diluted in PBS 3% BSA (anti-CD11b/c, 3 µg/mL; anti-CD3, 2 µg/mL; anti-CD45RA, 5 µg/mL; anti-granulocyte, 3 µg/mL; anti-ED1-like, 1 µg/mL and anti-NKR-P1A, 3 µg/mL) for 2 h at rt in a humid chamber. After rinsing the slides in three changes of PBS they were incubated with the respective biotinylated secondary antibody for 40 min at 37°C in a humid chamber. The slides were rinsed again and then treated with a horseradish peroxidase-streptavidin solution (1 : 400 in PBS 3% BSA) for 45 min at rt. Peroxidase labeling was visualized by incubation with FAST-DAB solution as a substrate for 2–4 min. The details for the primary antibodies employed in this study are given in Table 1. The respective positively stained immune cells in the CA3 area were counted in at least three coronal sections (starting at the needle track) with an increment of ≈ 0.1 mm using 40× and 100× magnification. All sections were scored blind and manually to exclude infiltrating cells in the area of the needle track.
Table 1. Antibodies used for immunohistochemical staining of different types of immune cells
1C7 (anti-ED-1 like)
Frozen coronal hippocampal sections (15 µm) were fixed in icecold acetone : methanol (3 : 1) at −20°C for 10 min, treated with 0.03% H2O2 solution for 10 min at rt to block endogenous peroxidase activity, rinsed in 50 mm Tris 0.4% Triton X-100 for 5 min at rt and then blocked with 10% normal goat serum for 15 min at rt. Sections were then incubated for 48 h at 4°C with polyclonal anti-rat IL-1α antibody (Pierce-Endogen, Rockford, IL, USA) diluted 1 : 100 in PBS (0.3% Triton-X100, 5% BSA). After three 5-min washes in PBS (0.1% Triton-X100), sections were incubated with the biotinylated anti-rabbit Ig secondary antibody for 40 min at 37°C in a humid chamber. The slides were washed again and immunoreactivity was tested by the avidin-biotin-peroxidase technique (Vectastain ABC kit; Vector Laboratories, Burlingame, CA, USA). Peroxidase labeling was visualized by incubation with FAST-DAB solution as a substrate for 2–4 min.
RNase protection assay
At indicated times (4, 8, 12 and 24 h after microinfusion), the hippocampus was dissected and quick frozen. Tissue was homogenized/disrupted with a 1.5-mL Pellet Pestle® grinder (Kimble-Kontes, Vineland, NJ, USA) and total RNA was extracted using a RNA isolation kit (BD PharMingen). RNA concentrations were determined by spectrophotometry.
Detection and semiquantitation of a variety of rat cytokine mRNAs were performed on 10 µg of total RNA utilizing the Riboquant Multiprobe Rnase Protection Assay System and the rCK-1 template set (BD PharMingen) following the manufacturer's instructions. Briefly, [α-32P] UTP-labeled antisense RNA probes were synthesized by in vitro transcription of cDNA templates for 11 rat cytokines (IL-1α, IL-1β, TNF-β, IL-3, IL-4, IL-5, IL-6, IL-10, TNF-α, IL-2 and IFN-γ) and two housekeeping genes (L32 and GAPDH) as internal controls. DNA templates were degraded by DNaseI digestion and probes were purified by phenol : chloroform extraction and ethanol precipitation with subsequent hybridization to total RNA at 56°C overnight. Samples were treated with RNase A+T1 and protected double-stranded RNA was purified by phenol : chloroform extraction and ethanol : salt precipitation. Samples were resuspended in Quickpoint loading dye and resolved on a denaturing 6% polyacrylamide 7M urea gel using the Quickpoint Rapid Nucleic Acid Separation System (Invitrogen, Carlsbad, CA, USA). After overnight exposure at − 80°C to Biomax X-ray films (Kodak, Rochester, NY, USA), bands were quantified by densitometric analysis using the Digital Science 1D software (Kodak). The mRNA levels are expressed as the ratio of the respective cytokine band and the corresponding L32 band in densitometric units.
ELISA for IL-1β and TNF-α
IL-1β and TNF-α protein levels were determined by commercially available rat-specific ELISA (Pierce-Endogen) according to the manufacturer's instructions. Animals were killed 8 h after KA or PBS injection, the respective hippocampi were dissected free on ice and immediately snap frozen in liquid nitrogen. Brain samples were placed in sterile icecold PBS containing a protease inhibitor cocktail [0.2 mm aminoethyl-benzenesulfonyl-fluoride (AEBSF), 1 µg/mL aprotinin, 1 mm benzamidine, 1 mm EDTA, 10 µm leupeptin and 10 µg/mL pepstatin] following homogenization using a 1.5-mL handheld homogenizer (Kimble-Kontes). Samples were then centrifuged (12 000 r.p.m., 20 min, 4°C), the supernatant fluid removed and divided into 50-µL aliquots that were stored at − 70°C until used. Samples were run in duplicate in the ELISA; protein concentrations of all samples were measured by the bicinchoninic acid (BCA) method (Pierce-Endogen) and cytokine levels expressed as pg/mg protein.
Data are given as mean values ± SD. Analysis of results was performed using linear regression analysis or anova followed by all pairwise multiple comparison post-hoc procedure (Tukey test or SNK test, respectively); p-values < 0.05 were considered statistically significant. For correlation analysis (linear regression), the data (lesion and infiltrate, respectively) of the different treatment groups (ADX, INT and COR) were combined for the comparison of the respective timepoints.
Effect of glucocorticoids on kainic acid-induced hippocampal damage
We first investigated whether different GC levels had a significant effect on the lesion size in the CA3 area 8–72 h after KA-induced excitotoxic insult (Fig. 1). Kainic acid was injected locally into the hippocampus. Damage to the CA3 region of the hippocampus was quantified after cresyl violet staining. As would be expected, KA caused significant lesion damage in control rats (INT). Adrenocortical status significantly modified the extent of damage ( p < 0.001, F = 88.6, d.f. = 2/49). In ADX animals, with their reduced GC exposure, the maximum lesion size had developed after 8 h and did not change significantly thereafter ( post-hoc comparison after anova). This resulted in significantly smaller lesions at 24 and 72 h after KA, as compared with INT rats ( p < 0.001 for both times), as well as when compared with COR rats ( p < 0.001 for both times). Increased GC exposure (in COR rats) had no effect on maximal lesion size, but accelerated the emergence of damage, since lesion size at 8 and 15 h was significantly greater than in INT ( p < 0.001 for each time by post-hoc test). Thus, when comparing INT and ADX animals, GCs exacerbate damage and, when comparing COR and INT animals, GCs accelerate the emergence of damage.
Effect of glucocorticoids on cellular inflammation after kainate injection
In the periphery GCs are known to reduce the number of inflammatory cells of the innate immune system infiltrating the injured tissue area. We, therefore, examined how the cellular inflammatory response to KA-induced injury in the CNS is affected by GCs over the course of time. Granulocytes, macrophages/monocytes and microglial cells were detected in the hippocampal CA3 region of all three groups. Immunostaining for T-cells, B-cells and NK-cells was negative. After PBS injection, no immune cells were detected in the CA3 area (data not shown).
The number of granulocytes detected in the KA-injected CA3 region was plotted as a function of time and is shown for the three groups in Fig. 2(a). In the intact animals (INT) group, the neutrophils peaked at 24 h after KA administration. Adrenocortical status significantly modified the extent of granulocyte cellular infiltrate ( p < 0.001, F = 47.2, d.f. = 2/67). In the ADX group, the peak shifted from 24 to 15 h. At 8 and 15 h after KA, granulocyte numbers were significantly greater in ADX rats than in INT animals ( p < 0.001 for each, respectively comparing INT with ADX by post-hoc test). At 24–72 h, however, neutrophil numbers were lower in ADX rats ( p < 0.001 for each time, respectively comparing INT and ADX by post-hoc test). Therefore, removal of the adrenal glands seemed to cause an accelerating effect on the onset of cellular infiltration but did not result in a net increase in neutrophil numbers. In fact, the area under the curve analysis of the three groups including all timepoints (Fig. 2d) revealed a lower relative number of neutrophils in ADX rats than in INT animals ( p < 0.05).
Increased GC exposure (COR group) did not alter neutrophil profiles as compared with INT. However, COR rats had elevated neutrophil numbers as compared with ADX rats (peaking at 15 h) at 15–40 h ( p < 0.001 for each time point, respectively comparing COR and ADX by post-hoc test). Moreover, the relative number of neutrophils in the COR group was significantly higher ( p < 0.001 as compared with ADX by post-hoc test) than in the ADX group (Fig. 2d). These data show that GCs caused an overall increase in inflammatory neutrophils after KA injection thus, surprisingly, having the opposite effect to that seen in the periphery (reduction in cellular inflammation).
The number of macrophages detected in the KA-injected CA3 region was plotted as a function of time and is shown for the three groups in Fig. 2(b). In INT rats, macrophage numbers peaked 24 h after KA. Adrenocortical status significantly modified the extent of macrophage infiltrate ( p < 0.003, F = 109.6, d.f. = 2/67). In ADX rats, macrophage cell numbers were significantly lower than in INT rats ( p < 0.042 for 8 h, p < 0.001 for 24 and 40 h) except at 15 h, where ADX macrophages were slightly higher ( p < 0.004 for 15 h by post-hoc test). Thus, as with neutrophils, reduced GC exposure accelerated the cellular inflammation.
Elevated GC exposure (in COR rats) increased macrophage numbers ( p < 0.001, < 0.005, < 0.013 and < 0.001 for 15, 24, 40 and 72 h, respectively as compared with INT by post-hoc test). Numbers were also elevated relative to ADX rats ( p < 0.001 for all time points). The area under the curve analysis (Fig. 2d) showed a positive correlation between GC level and the accumulation of macrophages ( p < 0.003 for COR compared with ADX and INT and p < 0.005 for INT compared with ADX by post-hoc test). These data suggest that GCs had an unexpected pro-inflammatory effect by increasing the number of macrophages in the CA3 area.
Microglial cells (CD11b/c+)
The number of microglial cells detected in the KA-injected CA3 region was plotted as a function of time and is shown for the three groups in Fig. 2(c). The results are quite similar to those found with macrophages. Microglia numbers peaked at 24 h after KA treatment in INT and, as before, adrenocortical status significantly modified the extent of microglial infiltrate ( p < 0.001, F = 40.4, d.f. = 2/46). At all but one time point ( p < 0.001 for 15 h comparing INT and ADX by post-hoc test), the number of microglia in ADX animals was lower ( p < 0.011 for 40 h comparing INT and ADX by post-hoc test) or not significantly different (8, 24 and 72 h) compared with INT animals.
Raising GC exposure in COR rats increased microglia number as compared with INT rats ( p < 0.001 for 15, 40 and 72 h) and also as compared with ADX rats ( p < 0.011, < 0.001, < 0.001 for 24, 40 and 72 h, respectively). The area under the curve analysis showed that microglia were elevated in COR animals compared with INT and ADX rats (Fig. 2d; p < 0.001 for comparing COR and INT, COR and ADX by post-hoc test). There was also an increase in relative microglia numbers between the ADX and INT group but it was not statistically significant. Like the neutrophil and macrophage data, these results showed that GCs surprisingly caused an increase instead of a decrease in inflammatory cells in the CA3 area after KA injection. Because of the unexpected nature of these findings, we tested whether the extent of damage at any given time point was significantly correlated with the extent of inflammatory infiltration at that time or at the next time point (i.e. if damage predicted the subsequent extent of inflammatory infiltration). The correlation analysis revealed that the pro-inflammatory effects of GCs were not a mere consequence of them causing more hippocampal damage (Table 2). On the contrary, the extent of damage at any given time correlated significantly with the inflammatory infiltration at the previous timepoint (Table 2), thus indicating that neuronal damage was caused by the preceding inflammation.
Table 2. Lack of correlation between extent of damage and simultaneous or following inflammatory infiltration (same timepoint/next timepoint)
Same timepoint (r2)
Next timepoint (r2)
Previous timepoint (r2)
Numbers indicate the r2 value from a linear regression analysis of lesion size (Fig. 1) against number of infiltrating cells (Fig. 2). Groups (adrenalectomized, intact controls and GC treated) were combined for the comparison of the different timepoints. p-values were > 0.05. A significant correlation (*p < 0.05) was found between the extent of damage and the preceding inflammatory infiltration (previous timepoint).
The increase in neuronal damage and cellular inflammatory infiltrate after kainic acid injection are dependent on steroid synthesis
The differences seen between the intact control (INT) and the COR group were definitely caused by GCs alone. However, since removing the adrenal glands removes many more circulating factors than corticosterone, it is possible that the differences seen between adrenalectomized and non-adrenalectomized animals might be due to factors other than GCs. Therefore, we compared intact animals (adrenal glands present and GC levels in the high stress range of 28–32 µg/dL) with intact animals which were treated with metyrapone 1 h before KA insult. Metyrapone is a potent inhibitor of steroid synthesis which locks GC secretion at basal levels of 6–7 µg/dL (INT + met). We found that 40 h after KA administration neuronal damage (Fig. 3a) as well as inflammatory cellular infiltration of neutrophils (Fig. 3b) and macrophages (Fig. 3c) in the hippocampus was significantly reduced in metyrapone-treated animals compared with intact animals (p < 0.01 by SNK post-hoc test). These results clearly showed that the exacerbation of neuronal damage and increase in cellular inflammatory infiltrate was caused by GCs and not by other factors as a consequence of the presence or absence of adrenal glands.
Effect of glucocorticoids on cytokine mRNA pattern
In the periphery, GCs are known to suppress inflammation by inhibiting synthesis of pro-inflammatory cytokines. We examined the effect of GCs on the cytokine pattern over the course of time after KA injection using multiple template RNase protection assay (mRPA). Animals were killed at indicated times (4, 8, 12 and 24 h) after KA or PBS, respectively and total RNA was isolated from the hippocampus, pooled (n = 4–6 rats/group and time point) and subjected to mRPA (Fig. 4a shows one of three mRPAs performed using different mRNA pools with similar results; Fig. 4b–e show the densitometric analysis of the cumulative data of the three RPAs performed). The template set included the pro-inflammatory cytokines IL-1α, IL-1β, TNF-β, IL-3, IL-4, IL-5, IL-6, IL-10, TNF-α, IL-2 and IFN-γ.
Elevated mRNA levels of IL-1α, IL-1β, IL-6 and TNF-α mRNA could be detected 4–12 h after KA injection in all three groups (Fig. 4a, left panel). Quantification of the respective cytokine bands was done by densitometry (Fig. 4b–e) and band intensity was given as the ratio of cytokine : L32 mRNA in densitometric units. Baseline cytokine mRNA synthesis after PBS injection was the same in all three groups at the respective times (Fig. 4a, right panel, representative 8 h PBS controls shown).
Adrenocortical status significantly modified respective mRNA cytokine synthesis ( p < 0.001, F = 260.3, d.f. = 2/35 for IL-1α; p < 0.001, F = 357.6, d.f. = 2/35 for IL-1β; p < 0.001, F = 20.8, d.f. = 2/35 for IL-6 and p < 0.001, F = 85.9, d.f. = 2/35 for TNF-α) 4–8 h after KA. IL-1α (Fig. 4b) and IL-1β (Fig. 4c) mRNA levels in the ADX rats were significantly higher than in INT rats ( p < 0.001 for 4 and 8 h), indicating an anti-inflammatory effect of GCs. In agreement with the data regarding cellular inflammation, the COR group (Fig. 4b, 4–8 h) showed significantly higher IL-1α mRNA levels than in INT rats ( p < 0.001 for 4 and 8 h) as well as when compared with ADX rats ( p < 0.001 for 4 and 8 h).
The same effects could also be observed for IL1-β (Fig. 4c;p < 0.001 for 4 and 8 h compared with ADX and p < 0.001 for 4 and 8 h compared with INT by post-hoc test) and TNF-α (Fig. 4e;p < 0.001 for 4 h compared with ADX and p < 0.001 for 4, 8 and 12 h compared with INT by post-hoc test) mRNA levels. At the indicated time points, the INT group always showed lower mRNA levels of the respective cytokine, whereas the COR group showed higher mRNA levels compared with the ADX group. It is noteworthy that the magnitude of the inhibitory effect (65–183% of control) caused by GC levels in the INT group was smaller compared with the enhancing effect (173–413% of control) caused by the GC levels in the COR group. There were no significant differences detected in IL-6 mRNA levels between the ADX and INT group at any time point (Fig. 4d, 4–24 h by post-hoc test). The COR group, however, showed increased IL-6 mRNA compared with ADX and INT animals 4 h after KA (p < 0.001 for 4 h as compared with ADX and p < 0.001 for 4 h as compared with INT by post-hoc test). Thus, GCs did not suppress but enhanced IL-6 mRNA synthesis at these times. These results indicate that, depending on the dosage, GCs can have anti- or until now unknown pro-inflammatory effects on cytokine mRNA synthesis after KA injury in the hippocampus.
Effects of glucocorticoids on cytokine protein expression
In order to determine if the observed increases/differences in mRNA levels were translated to protein expression, IL-1β(Fig. 5a) and TNF-α (Fig. 5b) protein levels were determined 8 h after KA or PBS injection using a rat-specific ELISA. Compared with the respective control (PBS injection), ADX, INT and COR animals showed an increase in IL-1β and TNF-α expression in response to kainate injection. In line with the mRNA data, KA-induced IL-1β protein levels were much higher (52–122 pg/mg) than TNF-α levels (12–18 pg/mg). Interestingly, the IL-1β levels in the PBS-injected control group were significantly higher (p < 0.05) in COR animals compared with ADX and INT animals (Fig. 5a), thus pointing towards a pro-inflammatory effect of chronically high GCs even in the absence of an insult.
Further confirming the mRNA data, adrenocortical status significantly modified respective cytokine synthesis (p < 0.001, F = 28.5, d.f. = 5/21 for IL-1β and p < 0.05, F = 11.6, d.f. = 5/21 for TNF-α); 8 h after KA IL-1β mRNA levels in ADX rats were significantly higher than in INT rats (p < 0.05). The COR group showed significantly higher IL-1β expression than INT rats (p < 0.001), as well as when compared with ADX rats (p < 0.001). TNF-α levels were not significantly different between ADX and INT animals as well as between ADX and COR animals. However, there was a statistically significant difference in TNF-α levels between COR and INT animals (p < 0.02). Thus, the obtained protein data for IL-1β and TNF-α clearly confirm the effects observed at the mRNA level. The most likely source of these cytokine mRNAs and proteins is neurons or glial cells since 4–8 h after KA only a few inflammatory cells could be detected (see Figs 2a–c). This was further supported by immunohistochemical staining of IL-1β in hippocampal sections of COR animals 8 h after KA injection (Figs 6a–e). Compared with PBS control rats (Figs 6d and e) IL-1β immunoreactivity was enhanced in response to kainate (Figs 6a–c). Intense IL-1β staining was detected in the pyramidal neuron layer of the CA3 region of the hippocampus as well as in the stratum oriens and stratum lucidum which contain mostly glial cells. The morphology and localization of IL-1β-positive cells suggest that glial cells and neurons are the primary source of this cytokine in response to KA.
There is abundant evidence that an inflammatory reaction is mounted within the CNS following trauma, stroke and seizure (Dirnagl et al. 1999; Lee et al. 1999), which can augment the brain damage (Feuerstein et al. 1998a; Barone and Feuerstein 1999). While GCs, with their anti-inflammatory action, should be protective they instead exacerbate neuron loss during excitotoxic insults. To understand this paradox, we investigated the effect of GCs on acute inflammation after KA injection in the hippocampus. We first tested whether GCs did indeed worsen KA neurotoxicity. Rats with an intact stress response (INT) served as controls and had elevated circulating GC levels for about 10 h after the KA exposure. In contrast, adrenalectomized/basal GC-supplemented rats (ADX) were locked at basal GC levels; corticosterone-treated rats (COR) had high levels throughout the study period. We observed that elevated GCs accelerate the emergence of and exacerbate KA-induced neurotoxicity. Acceleration of damage occurred only in COR rats; it is unclear which facet of their enhanced GC exposure caused this acceleration. These results agree with reports that GCs worsen the neurotoxicity of necrotic insults or accelerate their emergence (Sapolsky and Pulsinelli 1985; Stein-Behrens et al. 1992), including a specific demonstration of this acceleration effect (Morse and Davis 1990). This includes worsening the neurotoxicity induced by excitotoxic seizure without changing the intensity of epileptiform activity induced by the seizure (Stein and Sapolsky 1988; Stein-Behrens et al. 1992).
Glucocorticoids are known to reduce the acute cellular inflammatory response and suppress pro-inflammatory cytokines (DeRijk et al. 1997; Sternberg 2001). We found that the inflammatory infiltrate in CA3 post-KA consisted of microglia plus blood-borne granulocytes and macrophages. In PBS-injected rats, microglial staining showed a fine reticular homogenous pattern throughout the brain. In a KA-injected rat, the pattern was altered with reduced cellular processes, swollen cell bodies and increased microglial cell density in area CA3. Compared with normal GC levels (INT), lower levels of GCs (ADX) accelerated the early onset of cellular inflammation but, interestingly, blunted the later phases of inflammation, suggesting an initially delaying but eventually pro-inflammatory effect of GCs. More surprisingly, chronically high GC levels (COR) caused rapid infiltration and further increase in numbers of inflammatory cells. The metyrapone results indicated that the ADX effect was due to the loss of GCs, rather than some other adrenal factor. Correlation analysis showed that overall damage did not predict the subsequent extent of inflammatory infiltration, thus excluding that higher GC levels were associated with more inflammatory infiltration merely because higher GC levels caused a larger lesion. In fact, the extent of damage significantly correlated with the preceding inflammation, thus clearly supporting the concept that inflammatory cells were not just secondary bystanders but instead caused neuronal damage. These cells could, depending on priming and respective microenvironment, contribute to brain injury in several ways, such as oxygen radical generation, excessive glutamate release or secretion of neurotoxic cytokines (Banati et al. 1993; Kreutzberg 1996; Rothwell and Luheshi 1996). However, even in ADX rats, which developed the smallest lesions, high numbers of inflammatory cells were detectable at certain times. An inflammatory response at the right time and of the right extent might potentially serve important functions in tissue reconstruction and remodeling just as it does in peripheral wound healing (Rapalino et al. 1998; Kerschensteiner et al. 1999). In the periphery, GCs reduce inflammatory infiltration by suppressing ICAM-1, an adhesion molecule crucial for migration of neutrophil granulocytes and macrophages from the bloodstream into inflamed tissue (Cato and Wade 1996; Perretti and Flower 1994). The GCs and a neurological insult might alter ICAM-1 expression on endothelial cells of the BBB or alter BBB permeability itself to facilitate recruitment of blood-borne cell types.
Regardless of the mechanisms mediating our demonstrated anti- and pro-inflammatory GC actions, they are both likely to be mediated by the same corticosteroid receptor. The hippocampus contains both high-affinity mineralocorticoid receptors, which are heavily occupied by GCs basally (and are predominately responsible for GC effects in ADX rats), and low-affinity GC receptors, which are only heavily occupied by stress levels of GCs. Thus, both INT and COR rats were likely to have sustained heavy GC receptor occupancy, albeit for different durations.
CNS expression of pro-inflammatory cytokines, normally low (Vitkovic et al. 2000), is greatly increased by necrotic neurological insults. Kainic acid elevated mRNA levels of the pro-inflammatory cytokines IL-1α, IL-1β, IL-6 and TNF-α in all three groups. These mRNAs were probably produced by neurons and glia and provide a certain cytokine microenvironment that affects the actions of the later infiltrating inflammatory cells (‘priming’). Excessive granulocyte priming by pro-inflammatory cytokines, for example, causes spontaneous degranulation and generation of reactive oxygen species (Labro 2000). Acutely high GC levels mildly inhibited IL-1α, IL-1β and TNF-α but not IL-6 mRNA expression compared with basal GC levels. This is similar to what has been reported in the periphery, although of a smaller magnitude of effect (Wilckens and De Rijk 1997). However, in contrast to expected GC effects, chronically high GCs increased expression of these four cytokines. For IL-1β and TNF-α, this could be shown at the mRNA and the protein level.
Among our three groups, the lowest levels of expression of pro-inflammatory cytokines were associated with intermediate lesion size (in INT rats). Controversial literature suggests that a marked excess or paucity of such cytokines can be damaging to the nervous system (Yoles et al. 2001). Supporting this idea of homeostatic inflammation, mice overexpressing IL-6 and TNF-α in glia develop neurotoxic seizures (Akassoglou et al. 1997) while mice lacking TNF-α receptors sustain larger infarcts than wild-type mice (Bruce et al. 1996; Gary et al. 1998). Therefore, a goal is to understand how best to modulate cytokine responses to maximize beneficial effects and minimize neurotoxic effects of inflammation in neurological disorders.
The aim of our study was to reconcile the anti-inflammatory and potentially beneficial effects of GCs with their worsening of KA neurotoxicity. In comparing INT and ADX rats, we found that the endogenous GC response to KA has transient anti-inflammatory but overall pro-inflammatory effects on cellular infiltrate and anti-inflammatory effects on cytokine mRNA synthesis. Chronically high GC levels had uniformly pro-inflammatory effects. These novel pro-inflammatory GC effects provide a surprising answer to our initial question and could further explain how GCs worsen the outcome of neurological insults. Commensurate with this, GCs are of no therapeutic benefit in containing post-stroke edema (Millikan et al. 1987) and can even worsen outcome (Fishman 1982; Tominaga et al. 1988; Goldstein et al. 1989).
This work was supported by grants from BASF and the German National Merit Foundation. Funding was also provided to RMS by NIH grant RO1 MH53814. We thank T. Dumas, S. Brooks and E. Cheng for assistance.