Induction of endothelial nitric-oxide synthase in rat brain astrocytes by systemic lipopolysaccharide treatment.

In the brain, three isoforms of nitric oxide (NO) synthase (NOS), namely neuronal NOS (nNOS, NOS1), inducible NOS (iNOS, NOS2), and endothelial NOS (eNOS, NOS3), have been implicated in biological roles such as neurotransmission, neurotoxicity, immune function, and blood vessel regulation, each isoform exhibiting in part overlapping roles. Previous studies showed that iNOS is induced in the brain by systemic treatment with lipopolysaccharide (LPS), a Gram-negative bacteria-derived stimulant of the innate immune system. Here we found that eNOS mRNA is induced in the rat brain by intraperitoneal injection of LPS of a smaller amount than that required for induction of iNOS mRNA. The induction of eNOS mRNA was followed by an increase in eNOS protein. Immunohistochemical analysis revealed that eNOS is located in astrocytes of both gray and white matters as well as in blood vessels. Induction of eNOS in response to a low dose of LPS, together with its localization in major components of the blood-brain barrier, suggests that brain eNOS is involved in early pathophysiologic response against systemic infection before iNOS is induced with progression of the infection.

In the brain, three isoforms of nitric oxide (NO) synthase (NOS), namely neuronal NOS (nNOS, NOS1), inducible NOS (iNOS, NOS2), and endothelial NOS (eNOS, NOS3), have been implicated in biological roles such as neurotransmission, neurotoxicity, immune function, and blood vessel regulation, each isoform exhibiting in part overlapping roles. Previous studies showed that iNOS is induced in the brain by systemic treatment with lipopolysaccharide (LPS), a Gram-negative bacteria-derived stimulant of the innate immune system. Here we found that eNOS mRNA is induced in the rat brain by intraperitoneal injection of LPS of a smaller amount than that required for induction of iNOS mRNA. The induction of eNOS mRNA was followed by an increase in eNOS protein. Immunohistochemical analysis revealed that eNOS is located in astrocytes of both gray and white matters as well as in blood vessels. Induction of eNOS in response to a low dose of LPS, together with its localization in major components of the blood-brain barrier, suggests that brain eNOS is involved in early pathophysiologic response against systemic infection before iNOS is induced with progression of the infection.
Nitric oxide (NO) is a gaseous messenger molecule functioning mainly in vascular regulation, immunity, and neurotransmission (for recent reviews, see Refs. 1 and 2). NO is formed from L-arginine by NO synthase (NOS). 1 At least three isoforms of NOS have been identified, namely, neuronal NOS (nNOS, NOS1), inducible NOS (iNOS, NOS2), and endothelial NOS (eNOS, NOS3). In the brain, all three NOS isoforms are expressed constitutively or inducibly, and implicated in a number of physiologic and pathophysiologic functions.
nNOS is expressed constitutively in specific neurons of various brain regions (3)(4)(5). Expression of iNOS is under a detectable level in the normal brain, but is induced in neurons, glias, and other cells in various pathological conditions such as ischemia (6, 7), trauma (8), multiple sclerosis (9, 10), Parkinson's disease (11), Alzheimer's disease (12), tumors (13), acquired immunodeficiency syndrome encephalitis (14), infection (15), and lipopolysaccharide (LPS) treatment (16 -18). The locus of eNOS expression in the brain is controversial. In addition to blood vessels, eNOS-like protein was detected immunohistochemically in pyramidal neurons of the hippocampus, exhibiting profiles different between the rat (19) and human (20). eNOS has been also described to be present in astrocytes (20 -22), and several groups reported that eNOS was detected preferentially in astrocytes rather than in neurons (21,22).
Gene disruption studies have revealed that the three NOS isoforms in the brain play interdigitated biomedical roles, functioning cooperatively or antagonistically. While mice deficient in nNOS displayed aggressive behavior and inappropriate sexual behavior (23), long term potentiation in the hippocampal cornu ammonis (CA) 1 region of the mutant mice was almost normal (24), which was also the case for eNOS-deficient mice (25). On the other hand, mice doubly deficient in nNOS and eNOS showed a significant reduction of long term potentiation in the stratum radiatum of the CA1 region, suggesting a reciprocally compensatory role for nNOS and eNOS (25). In brain ischemia following occlusion of the middle cerebral artery, infarct volume was reduced in mice deficient in nNOS (26) or iNOS (27). In contrast, eNOS-deficient mice showed an increase in infarct volume (28). Apparently, nNOS-and iNOSderived NO is toxic, while eNOS-derived NO is protective, to the ischemic brain.
In order to clarify roles of NOS isoforms in the nervous system, we previously examined precise distribution of nNOS mRNA in the rat brain, by using high resolution non-radioisotopic in situ hybridization (5). We have also studied regulation of the iNOS gene in response to LPS in peripheral organs, being concerned with coordination with genes for L-arginine-metabolizing enzymes (29,30). In the course of examining the effects of LPS on expression of the three NOS isoforms, we by chance found that a relatively low dose of LPS induces mRNA for eNOS, but not for iNOS, in the rat brain.

EXPERIMENTAL PROCEDURES
Animals and LPS Treatment-Specific pathogen-free male Wistar rats (5-6 weeks old) were injected intraperitoneally with Escherichia coli LPS (serotype 0127:B8, Sigma), and after appropriate times brain and other organs were excised from the rats anesthetized in ether. Regional dissection of the rat brain was done essentially according to the division of Glowinski and Iversen (31).
RNA Blot Analysis-Total RNA from rat tissues was prepared by the guanidinium thiocyanate-phenol-chloroform extraction procedure (32). After electrophoresis in formaldehyde-containing agarose gels, RNA was transferred to nylon membranes. The digoxigenin-labeled antisense RNA probe was synthesized using the DIG-RNA labeling kit (Roche Molecular Biochemicals, Mannheim, Germany) with cDNA templates for rat eNOS (corresponding to nucleotides 44 -1164 of the human sequence of Ref. 33), rat iNOS (nucleotides 2529 -3211 of Ref. 34), and rat nNOS (nucleotides 296 -2150 of Ref. 35). eNOS and iNOS cDNAs were amplified using reverse transcription-polymerase chain reaction, and cloned into the EcoRV site of pcDNAII and the HincII site of pGEM3Zf(ϩ), respectively. nNOS cDNA was isolated as described previously (5). After hybridization and washing as recommended by Roche Molecular Biochemicals, chemiluminescent detection of hybridized probes on x-ray films was done using the alkaline phosphataseconjugated DIG antibody (Roche Molecular Biochemicals) and CDP-Star™ (Tropix Inc., Bedford, MA). Densitometric quantification was performed using the MacBas bioimage analyzer (Fuji Photo Film Co., Tokyo, Japan).
Immunoblot Analysis-Rat tissues were homogenized in nine volumes of 20 mM potassium HEPES buffer (pH 7.4) containing 1 mM dithiothreitol, 50 M antipain, 50 M leupeptin, 50 M chymostatin, and 50 M pepstatin. The homogenate was centrifuged at 25,000 ϫ g for 30 min at 4°C, and supernatant was used as tissue extract. Protein in the tissue extract was determined with the protein assay reagent (Bio-Rad) using bovine serum albumin as standard. The tissue extracts (15 g of protein/lane) were subjected to SDS-6% polyacrylamide gel electrophoresis, and protein was electrotransferred to nitrocellulose membranes. As a primary antibody, a rabbit polyclonal anti-human eNOS antibody (1:50 dilution; N30030; Transduction Laboratories, Lexington, KY) in Fig. 3 or a monoclonal anti-human eNOS antibody (1 g/ml IgG; N30020; Transduction Laboratories) in Fig. 4 was incubated with the membranes. Immunodetection was performed using the ECL kit (Amersham Pharmacia Biotech, Buckinghamshire, UK). Chemiluminescent signals were detected on x-ray films and quantified using the MacBas bioimage analyzer.
Assay of NOS Activity-The tissue extracts were prepared as described above. NOS catalytic activity was assayed by determining the conversion of L-[ 3 H]arginine to L-[ 3 H]citrulline essentially as described (36,37). Ca 2ϩ -dependent NOS activity was measured by incubating 50 In the Ca 2ϩ -independent NOS assay, EGTA (1 mM) was substituted for CaCl 2 . The reaction was stopped by adding 100 l of 3% trichloroacetic acid. After 30 min on ice, 250 l of 1.5 M potassium HEPES (pH 7.5) were added, and the precipitates were removed by centrifugation at 15,800 ϫ g for 5 min at 4°C. The supernatants were applied to 1-ml columns of Dowex 50W-X8 (Na ϩ form), and eluted L-[ 3 H]citrulline was measured by liquid scintillation counting. Enzyme activity was expressed as picomoles of L-[ 3 H]citrulline produced/mg of protein in 1 min. Data are represented as mean Ϯ range.
Immunohistochemistry-Excised rat brains were embedded in OCT Compound™ (Miles, Elkhart, IN) and frozen in the liquid nitrogen. Ten-m sections were cut, air-dried, washed in Dulbecco's phosphatebuffered saline, and fixed with methanol for 10 min on ice. After inhibition of endogenous peroxidase activity by the method of Isobe et al. (38), the sections were subjected to incubation with the primary antibody diluted with phosphate-buffered saline containing 0.5% bovine serum albumin (fraction V) overnight at 4°C. A primary antibody used was a rabbit polyclonal anti-human eNOS antibody (1:200 -5000 dilution) or a mouse monoclonal anti-human glial fibrillary acidic protein (GFAP) antibody (28 g/ml IgG; Dako Corp., Carpinteria, CA). After incubation with the biotinylated secondary antibody (1:200; Vector Laboratories Inc., Burlingame, CA) for 2 h, the sections were further incubated with a mixture of avidin and horseradish peroxidase-conjugated biotin for 1 h. Peroxidase activity was visualized using 3,3Ј-diaminobenzidine as a substrate. Double-staining analysis with the rabbit polyclonal anti-eNOS antibody and the mouse monoclonal anti-GFAP antibody was performed by the peroxidase-antiperoxidase and alkaline phosphatase-antialkaline phosphatase procedure using the Doublestain kit (Dako Corp.). For control, sections were incubated with rabbit serum or non-immunized mouse IgG1 instead of each specific primary antibody. Sections were slightly counterstained with hematoxylin, and the slides were mounted in Aquatex (E. Merck, Darmstadt, Germany).

Induction of eNOS mRNA in the Brain by Systemic LPS
Treatment-Rats were injected intraperitoneally with Escherichia coli LPS at 2.5 g/g body weight, and 12 h later eNOS mRNA levels in various organs were examined by RNA blot analysis (Fig. 1A). In the heart and kidney, eNOS mRNA levels were lowered by the LPS treatment, concordant with a previous report (17). On the other hand, in the brain, eNOS mRNA was increased in response to LPS. No apparent induction of iNOS mRNA was observed in the brain with this amount of LPS, while the induction was obvious in the lung (Fig. 1B). The LPS treatment had no effect on nNOS mRNA levels in the brain (Fig. 1C).
Differential Dose Dependence in Induction of Brain eNOS and iNOS mRNAs by LPS-Previously, it has been reported that mRNA for iNOS, rather than eNOS, is induced in the brain as well as in other organs by LPS treatment (16 -18). Since the discrepancy can result from differences of experimental conditions, we examined changes in eNOS and iNOS mRNA levels with various amounts of LPS (Fig. 2). eNOS mRNA was substantially induced with LPS at 2.5 g/g, and reached a plateau level at 10 g/g. On the other hand, only a slight increase in iNOS mRNA was detected at 10 g/g, and a maximum increase at 50 g/g. In previous studies (16 -18), induction of iNOS mRNA in the brain has been detected with comparable amounts of LPS (15-50 g/g). mRNA levels for both eNOS and iNOS decreased with LPS at 100 g/g, compared with at 50 g/g. The exact reason for these decreases is not known, but a possible cause is the nonspecific catastrophic damage of the brain, since the LPS treatment at this concentration occasionally resulted in death of the animals. In conclusion, in the pathophysiologic dose range, higher amounts of LPS are required for induction of iNOS mRNA than in that for eNOS mRNA.
Accumulation of eNOS Protein following Its mRNA-Time course of changes in mRNA and protein levels for eNOS in the brain after treatment with LPS (2.5 g/g) are shown in Fig. 3. eNOS mRNA levels were raised 6 h after the treatment, and reached a maximum at 12 h. A delayed increase in eNOS protein levels was apparent at 12 h, and an additional increase at 24 h. Therefore, induction of eNOS mRNA in response to LPS leads to accumulation of eNOS protein.
We also examined changes in NOS catalytic activity in the brain 24 h after treatment with LPS (2.5 g/g). Total Ca 2ϩ -dependent NOS activity, which is derived mainly from nNOS and eNOS isoforms, slightly increased to 5.28 Ϯ 0.22 pmol/mg/min compared with 4.50 Ϯ 0.25 pmol/mg/min for the control (mean Ϯ range, n ϭ 2). On the other hand, iNOS-derived Ca 2ϩ -independent activity was below the detectable level in both control and LPS-treated rats. We further tried to measure eNOS activity specifically, by using nNOS inhibitors such as N -propyl-L-arginine (39) and 1-(2-trifluoromethylphenyl)imidazole (40), and by using antibodies against eNOS and nNOS (for each isoform, two commercially available antibodies) for immunoselection of eNOS and immunodepletion of nNOS, respectively. However, these attempts were unsuccessful, presumably at least in part because of relatively high nNOS ac-

eNOS Induction in Brain by LPS
tivity that amounts to more than 90% of total NOS activity in the normal rodent brain (41).
Distribution of eNOS in Brain Regions-Distribution of NOS isoforms in various regions of the rat brain was examined by RNA blot analysis (Fig. 4A). nNOS mRNA was detected in all brain regions examined, exhibiting higher levels in the olfactory bulb and cerebellum than in other regions, concordant with previous reports (5,35). iNOS mRNA was under the detectable level in any region of the normal rat brain. eNOS mRNA was detected in all brain regions, and showed more equal distribution than nNOS mRNA. By using immunoblot analysis (Fig. 4B), the eNOS protein of about 135 kDa was also detected in all brain regions, and showed rather uniform distribution resembling that of eNOS mRNA.
Localization of the eNOS Protein in Astrocytes-To determine the precise localization of the eNOS protein, immunohistochemical staining was performed on brain sections (Fig. 5,  A-H). As expected, blood vessels showed strong eNOS immunoreactivity in almost all brain regions examined (Fig. 5, A and  D). No apparent signal was detected with a control non-immune antibody (data not shown). Strong staining was also detected in cells morphologically resembling astrocytes in the white matter of the striatum (Fig. 5A), anterior commissure (Fig. 5B), corpus callosum, fimbria of the hippocampus, internal capsule (Fig. 5C), and cerebellum (Fig. 5H). Less intense but obvious staining was detected in astrocyte-like cells in the gray matter of the cerebral cortex (Fig. 5, D and E), hippocampus (Fig. 5, F and G), and cerebellum (Fig. 5H). Some astrocytelike cells were seen to radiate the processes coming in contact with blood vessels (Fig. 5A; see also Fig. 5M). In the hippocampus, immunolabeling was located in the stratum oriens and stratum radiatum (Fig. 5G). On the other hand, in contrast to the previous reports (19,24), no obvious signal was detected in pyramidal neurons under the present condition. In the cerebellum (Fig. 5H), the deep white matter showed strong staining. Moderate and weak signals were detected in the granular layer and molecular layer, respectively, while no apparent signal in the Purkinje cell layer.

FIG. 2. Differential dose-dependent induction of brain eNOS and iNOS mRNAs by LPS.
Northern blot analysis for detection of eNOS and iNOS mRNAs was performed with total RNAs isolated from the rat brain 12 h after intraperitoneal injection of LPS of indicated amounts. Representative chemiluminograms are shown with ethidium bromide staining of rRNAs. At the bottom, the chemiluminograms were quantified densitometrically, and mRNA levels relative to the maximum value (100%) are represented by mean Ϯ S.E. (solid bar, n ϭ 3 or 4) or mean Ϯ range (dotted bar, n ϭ 2).

FIG. 3. Time course of induction of eNOS mRNA and protein in
the LPS-treated rat brain. Total RNAs and tissue extracts were prepared from the rat brain at indicated times after intraperitoneal injection of LPS (2.5 g/g body weight), and subjected to Northern and Western analysis, respectively. Representative chemiluminograms are shown. Below the Northern chemiluminogram, ethidium bromide staining of rRNAs is presented. At the bottom, the chemiluminograms were quantified densitometrically, and mRNA or protein levels relative to the maximum value (100%) are represented by mean Ϯ S.E. (solid bar, n ϭ 4) or mean Ϯ range (dotted bar, n ϭ 2).

FIG. 4. Distribution of NOS isoforms in the rat brain regions.
A, Northern blot analysis for detection of mRNAs for nNOS, iNOS, and eNOS was performed with total RNAs from various brain regions: Ob, olfactory bulb; Cx, cerebral cortex; St, striatum; Hp, hippocampus; Hy, hypothalamus; MT, midbrain and thalamus; Cb, cerebellum; PM, pons and medulla oblongata. As positive controls (P.C.) for iNOS and eNOS mRNAs, total RNAs prepared from the LPS-treated rat lung and the normal rat heart, respectively, were coelectrophoresed. Below, ethidium bromide staining of 28 and 18 S rRNAs is presented. B, Western blot analysis. Distribution of eNOS protein was examined for tissue extracts from the brain regions. As a positive control (P.C.), a human endothelial cell lysate was coelectrophoresed.

eNOS Induction in Brain by LPS
Expression of eNOS protein in astrocytes was confirmed by examining colocalization with GFAP as a marker for astrocytes (Fig. 5, I-K). Immunostaining of serial sections with the anti-eNOS antibody (Fig. 5I) or anti-GFAP antibody (Fig. 5J) exhibited a similar pattern of distribution of positive cells. Furthermore, double staining for eNOS and GFAP (Fig. 5K) detected overlapping immunoreactive cells, verifying colocalization of these two proteins.
The effect of LPS treatment on eNOS immunostaining was examined (Fig. 5, L and M). Concordant with the results of protein blotting analysis (Fig. 3), the strength of immunolabeling increased by the LPS treatment for 24 h, while precise FIG. 5. Histochemical analysis for distribution and induction of eNOS protein in the rat brain. A-H, immunodetection of eNOS protein in various regions of the normal rat brain. A, the striatum. In addition to the blood vessel (v), astrocyte-like cells (arrows) radiating processes to the vessel exhibit immunostaining. B, the anterior commissure (ac). C, the corpus callosum (cc), fimbria of the hippocampus (fi), and internal capsule (ic). D, the cerebral cortex. The blood vessels (arrows) are immunopositive. E, higher magnification view of the cerebral cortex reveals staining of astrocyte-like cells. F, the CA1-3 and dentate gyrus (DG) of the hippocampus. G, higher magnification view of the CA1 region reveals staining of astrocyte-like cells in the stratum oriens (so) and stratum radiatum (sr), but not of neurons in the pyramidal cell layer (py). H, the cerebellum. Abbreviations: gr, granular layer; mol, molecular layer; Pur, Purkinje cell layer; w, deep white matter. I-K, colocalization of eNOS with GFAP as a marker for astrocytes. Serial sections through the corpus callosum and caudate putamen were subjected to immunostaining with the anti-eNOS antibody (I, brown), anti-GFAP antibody (J, bright red), and both antibodies (K, dark red). L and M, augmentation of eNOS immunolabeling by LPS treatment. Sections through the hippocampal CA1 region from rats non-treated (L) and treated with LPS (10 g/g body weight) for 24 h (M) were subjected to eNOS immunostaining. quantitative analysis on increase in the number of positive astrocytes and/or in the intensity of labeling in each cell remains to be performed. DISCUSSION This study showed that intraperitoneal injection of LPS causes an increase in eNOS expression specifically in the brain, making a sharp contrast to decreases in the heart and kidney (Fig. 1A). This can be at least in part because of the difference of cells expressing eNOS in the brain and other organs. As reported previously (20 -22) and confirmed here (Fig. 5), astrocytes as well as endothelial cells are a major source of eNOS in the brain. It remains to be examined if the increase in eNOS expression in astrocytes is directly triggered by LPS itself or mediated by LPS-induced agents such as cytokines. The latter possibility is especially noteworthy, taking the blood-brain barrier into account. Astrocytes themselves constitute the outer component of the blood-brain barrier, and are separated from blood flow by tightly connected endothelial cells and the basement membrane. Therefore, it should be carefully examined if LPS of low concentrations in blood flow can directly reach astrocytes. Possible involvement of the blood-brain barrier in the LPS response of the brain was also suggested by the fact that higher amounts of LPS were necessary for iNOS induction in the brain than in the lung (Figs. 1B and 2).
What are the roles of astroglial eNOS? Barna et al. (21) reported that infection of the mouse brain by vesicular stomatitis virus led to increases in eNOS protein in astrocytes, implying a role of astroglial eNOS in inhibition of local brain infection. On the other hand, the present demonstration that astroglial eNOS was induced in response to intraperitoneal injection of LPS suggests involvement of astroglial eNOS in protection of the brain against systemic bacterial infection. Localization of eNOS in endothelial cells and astrocytes, namely two major components of the blood-brain barrier that is the defensive front of the brain, is also concordant with this notion. The amount of LPS required for induction of eNOS was smaller than that required for induction of iNOS, suggesting that eNOS can be induced in relatively early stages following bacterial infection. In summary, astroglial eNOS is likely to be involved in early precaution and defense against systemic infection.
Studies with eNOS knockout mice revealed that eNOS is involved in protection of neurons on brain ischemia (28), in addition to regulation of systemic blood pressure (42,43). It remains to be determined whether eNOS with this protective effect is attributable to astrocytes or endothelial cells. Gene knockout and related studies have also revealed that eNOS is involved in neuronal functions such as long term potentiation in the hippocampus (24,25) and N-methyl-D-aspartate-stimulated ␥-aminobutyric acid release in various brain regions (44). These functions of eNOS have been attributed to the enzyme located in neurons, on the assumption that brain eNOS is mainly distributed in neurons (19). However, since the present and other studies (21,22) demonstrated preferential distribution of eNOS in astrocytes rather than in neurons, it is tempting to speculate that NO derived from astroglial eNOS modulates these neuronal functions. Possible involvement of eNOS in neuronal functions suggests that a part of anti-microbial effects of astroglial eNOS can be mediated also through neuronal activities such as driving the autonomic nervous system that controls both brain and peripheral organs. Conditional astrocyte-specific disruption of the eNOS gene will provide useful tools to evaluate these intriguing roles of astroglial eNOS.