Exp. Eye Res. (1997) 65, 739–745

Cyclooxygenase-1 and Cyclooxygenase-2 in the Human Optic

Nerve Head

ARTHUR H. NEUFELD*, M. ROSARIO HERNANDEZ, MIRIAM GONZALEZ

ARI GELLER

Department of Ophthalmology and Visual Sciences, Washington University School of Medicine,

St. Louis, MO 63110, U.S.A.

(Received Columbia 15 April 1997 and accepted in revised form 7 July 1997)

To investigate the hypothesis that eicosanoids act as cellular mediators in the optic nerve head of normalsand of patients with glaucoma, we have determined the presence of the two cyclooxygenase (COX)isoforms in human tissue. Histological sections of optic nerve heads were studied by immunohisto-chemistry. Age matched normal donors were compared with eyes from glaucoma patients with moderateto severe nerve damage. Polyclonal antibodies to human COX-1 and COX-2 were localized withimmunoperoxidase staining. Specific antibodies for vascular endothelia and microglia were also co-localized. In normal and glaucomatous eyes, COX-1 was localized exclusively to the prelaminar andlamina cribrosa regions of the optic nerve head. No staining for COX-1 was observed in the nerve fiberlayer or the myelinated optic nerve. COX-1 was associated with the astrocytes of the glial columns andthe cribriform plates, but not with the endothelia lining the capillaries. In glaucoma, more astrocytesappeared to be stained with antibody to COX-1 than in normals and staining was intensely perinuclear.There was no staining for COX-2 in normal tissue. A few COX-2 positive cells were found in theprelaminar, lamina cribrosa and postlaminar regions of the glaucomatous optic nerves. Positive stainingfor COX-2 was not associated with microglia. COX-1 is constitutively present in astrocytes that arelocalized exclusively to the prelaminar and lamina cribrosa regions of the human optic nerve head.Eicosanoids, synthesized by COX-1 in this tissue, may have a homeostatic and a neuroprotective rolerelated to the axons of the retinal ganglion cells. The sparse presence of COX-2 in glaucomatous tissueprobably reflects the lack of inflammation associated with glaucomatous optic neuropathy.

# 1997 Academic Press LimitedKey words : cyclooxygenase; glaucoma; human; neuroprotection; optic nerve.

1. Introduction

Cyclooxygenase (COX), or prostaglandin H-synthase,

is the rate limiting enzyme in the synthesis of

eicosanoids from arachidonic acid. The enzyme cata-

lyses the formation of prostaglandin G#and hydrogen

peroxide from the free fatty acid. Prostaglandin G#

is

then reduced to prostaglandin H#

which is the

precursor of several eicosanoids, including prosta-

glandin E#, prostagandin F

#α, prostaglandin D

#, prosta-

cyclin and thromboxane A#.

There are two isoforms of COX, reviewed recently

with respect to selective pharmacological inhibition

(Vane and Botting, 1995a,b) and selective knock out

mice (DeWitt and Smith, 1995). Although expressed

by different genes, the amino acid sequences of the

isoforms are approximately 60% homologous (Wu,

1996). Human COX-1 contains 599 amino acids,

localizes to the endoplasmic reticulum, and is now

known to be a constitutive enzyme, widely distributed

throughout the body in most cells. Human COX-2

contains 604 amino acids, localizes to the nuclear

membrane and is an inducible enzyme, initially

* Correspondence to: Arthur H. Neufeld, Department of Oph-thalmology and Visual Sciences, Box 8096, Washington UniversitySchool of Medicine, 660 South Euclid Avenue, St. Louis, MO 63110,U.S.A.

demonstrated in tissue culture systems in response to

lipopolysaccharide (Fu et al., 1990; Masferrer et al.,

1990). In quiescent cells, there is little or no expression

of COX-2; however, under conditions of inflammation

when cytokines, growth factors and hormones are

present, COX-2 is highly inducible by these factors

(Bazan, Marcheselli and Mukherjee, 1995). COX-2 is

expressed in a variety of cell types, including macro-

phages, endothelial cells, fibroblasts, smooth muscle

cells and astroglia. COX-1 is also inducible but

expression is an order of magnitude less than that of

COX-2. Furthermore, although COX-1 and COX-2 are

similar proteins, the ability of non-steroidal anti-

inflammatory drugs to inhibit the two enzymes can

differ markedly (Mitchell et al., 1993; Vane and

Botting, 1995a,b).

COX-1 serves as a generator of eicosanoids for

physiological functions, for example regulation of

vascular tone, platelet aggregation and mucous

secretion of the stomach, and as such COX-1 provides

eicosanoids for cytoprotective functions (Vane and

Botting, 1995a,b; Goetzl, An and Smith, 1995;

Wallace and Tigley, 1995). Conversely, the induction

of COX-2 during inflammation in selective tissues

suggests a pro-inflammatory role for eicosanoids (Wu,

1996). Induction of COX-2 is an example of expression

of an immediate early gene and is associated with in-

0014–4835}97}120739­07 $25.00}0}ey970394 # 1997 Academic Press Limited

740 A. H. NEUFELD ET AL.

F. 1. Immunohistochemistry for COX-1 in human optic nerve heads from normal and glaucomatous age matched eyes. Atlow power (original magnification¬13), COX-1 is localized to the optic nerve head in normal (A) and glaucomatous (B) eyes.In the glial columns of the prelaminar region of normal eyes (C), there are a few COX-1 positive cells but many more COX-1positive cells can be seen in the disorganized prelaminar region of glaucomatous eyes (D) (original magnification¬163). At highpower (original magnification¬325, E-H), COX-1 localizes to many, but not all, astrocytes (thick arrows) in the cribiform plates

COX-1 AND COX-2 IN OPTIC NERVE HEAD 741

flammation and cellular injury (Bazan, Marcheselli

and Mukherjee, 1995). In the central nervous system,

induction of COX-2 occurs in response to markedly

increased synaptic excitation by glutamate, such as

during seizure activity, and is inhibited by NMDA

antagonists (Adams, Collaco-Moraes and Belleroche,

1996); Kaufman et al., 1996).

In the glaucomatous, human optic nerve, there are

axons undergoing degeneration, reactive astrocytes

and remodeling of the extracellular matrix (Hernandez

and Pena, 1997). In this environment, cellular

mediator pathways are likely to be active and to be

participating in the cell–cell interactions that are

coordinating these changes. We have recently de-

scribed the nitric oxide synthase (NOS) pathways in

the human optic nerve head and hypothesized that

nitric oxide may have both neuroprotective and

neurodestructive roles in the glaucomatous process

(Neufeld, Hernandez and Gonazlez, 1997). COX path-

ways are alternative potential mediator pathways that

can orchestrate cellular responses and often work in

concert with NOS pathways (DiRosa et al., 1996). In

the work presented here, we have used immuno-

histochemistry to demonstrate the presence of the

cyclooxygenase pathways in normal and age-

matched, glaucomatous, human optic nerves.

2. Materials and Methods

Six normal human eyes from six donors, ages 51 to

91 (73±1³14±7, mean³..), and eight eyes from

eight donors with documented primary open angle

glaucoma with moderate to advanced nerve damage,

ages 51 to 91 (79³12±3), were obtained from

eyebanks throughout the United States. Primary open

angle glaucoma was defined by a clinical history of

observation and treatment by an ophthalmologist,

and the presence of optic nerve damage on histological

examination, as evidenced by the presence of a cup

and the disorganization of glial columns and cribriform

plates. The eyes from donors with glaucoma had C}D

ratios of 0±6 to 0±9, demonstrated marked visual field

defects, and were reportedly on medications to lower

their elevated intraocular pressure.

The eyes were enucleated and fixed in 4% para-

formaldehyde within 24 hr after death. The optic

nerve heads were dissected free of surrounding tissues.

Fixed tissue was washed in 0±2% glycine in phosphate

buffered saline (PBS), pH 7±4, embedded in paraffin

and oriented for 6 µm sagittal sections.

Slides containing sections were preincubated with

of the lamina cribrosa of normal eyes (E). In the disorganized lamina cribrosa of glaucomatous eyes (F), COX-1 staining isassociated with presumably reactive astrocytes (thick arrows) that have rounded nuclei. COX-1 staining appears in aperinuclear and cytoplasmic pattern. Using double immunoperoxidase staining in normal eyes, (G) COX-1 (black staining, largearrows) does not colocalize with von Willebrand’s factor (brown staining, small arrow), and (H) COX-1 (black staining, smallarrows) does not colocalize with NOS-1 (brown staining, large arrows). Vit : vitreous ; Pre-L : prelaminar region; LC: laminacribrosa; GC: glial column; NB: nerve bundle ; Cap: capillary ; CP: cribriform plate.

5% milk for 30 minutes, rinsed and then incubated

with primary antibody for 30 minutes. The isoforms of

COX were identified using polyclonal antibodies to

human COX-1, which recognizes a peptide corre-

sponding to amino acids 570–599 and mapping to the

carboxy terminus (working dilution 1:500), and to

human COX-2, which recognizes a peptide corre-

sponding to amino acids 27–46 and mapping to the

amino terminus (working dilution 1:500), both

purchased from Santa Cruz Biotechnology, Inc. Pri-

mary antibodies were also used to co-localize different

cell types, including vascular endothelia with von

Willebrand factor (working dilution 1:400, purchased

from Sigma), cells with neuronal nitric oxide synthase,

NOS-1, Clone R20 (working dilution 1:100, pur-

chased from Santa Cruz Biotechnology, Inc.) and

microglia}macrophages (Diaz-Araya et al., 1995) with

HLA-DR CR3}43 (working dilution 1:50, purchased

from Accurate Chemical and Scientific Corp.).

Primary antibodies were localized by immuno-

peroxidase staining with reagents purchased from

Vector Laboratories. The biotinylated secondary anti-

body was incubated on the sections for 30 minutes,

washed with PBS and reacted with streptavidin-

peroxidase conjugate for 30 minutes. Following

washing, sections were incubated with the substrate

mixture: 1±5 mg 3,3-diaminobenzidine tetrahydro-

chloride (DAB) and 50 µl of 30% hydrogen peroxide in

0±1 Tris, pH 7±6. The sections were reacted in the

dark until brown staining appeared (about 5–7

minutes) washed in PBS, counterstained with hemat-

oxylin, dehydrated and coverslipped with Permount.

For co-localization protocols, double sequential

immunoperoxidase procedures were used. The first set

of primary and secondary antibodies were applied as

described above. After brown color appeared, the first

reaction was stopped by washing extensively in PBS.

Sections were then blocked again by incubating in 5%

dry milk in PBS for 30 min, rinsed, and the second set

of primary and secondary antibodies were applied as

before. After incubating with streptavidin peroxidase

conjugate for 30 min, slides were washed and incu-

bated with the second substrate mixture, which

contained DAB and hydrogen peroxide, as above, and

nickel chloride (100 µl}5 ml), for 3–5 min until black

color appeared. Slides were then dehydrated and

mounted with Permount.

Representative sections of all samples on a given

day were stained simultaneously to control variation

in the reactions. Negative controls were performed by

eliminating primary antibody from the incubation

742 A. H. NEUFELD ET AL.

F. 2. Immunohistochemistry for COX-2 in human optic nerve heads from normal and glaucomatous age matched eyes. Inthe normal optic nerve head, there are no COX-2 positive cells in the prelaminar region (A), lamina cribrosa (C), or postlaminarregion (E). In the glaucomatous optic nerve head, COX-2 positive cells (arrows) are infrequently found in the prelaminar region(B), lamina cribrosa (D) and postlaminar region (F). In the glaucomatous eye, microglia are identified with antibody to HLA-DR (G). Using double immunoperoxidase staining in glaucomatous eyes (H), microglia (black staining, thin arrows) are notassociated with COX-2 positive cells (brown staining, thick arrow). PS: pial septa ; all other abbreviations as in Fig. 1 (A) and(B), original magnification¬163; (C)–(H), original magnification¬325.

COX-1 AND COX-2 IN OPTIC NERVE HEAD 743

medium, or by replacing the primary antibody with

nonimmune serum, followed by immunoperoxidase

staining.

Slides were examined in a Nikon Optiphot-2

microscope and images were recorded using Royal

Gold 400 ASA color print film (Kodak).

3. Results

Immunohistochemical Localization of COX-1 in Human

Optic Nerves

Normal tissues COX-1 antibody stained the cyto-

plasm of positive cells with more intense localization in

the perinuclear region (Fig. 1), reflecting the as-

sociation of this isoform with the endoplasmic reti-

culum. No staining was detected in the extracellular

matrix or in association with non-myelinated or

myelinated axons. Replacement of the primary anti-

body with non-immune serum or PBS resulted in the

absence of specific immunostaining.

COX-1 was localized to the prelaminar and lamina

cribrosa regions of the normal optic nerve [Fig. 1(A)].

COX-1 was detected sparsely in astrocytes of the glial

columns of the prelaminar region [Fig. 1(C)] and in

many, but not all, astrocytes in the cribriform plates

[Fig. 1(E)] and the insertion region of the lamina

cribrosa. Astrocytes inside the nerve bundles did not

stain with COX-1 antibody. COX-1 staining was not

associated with the vascular endothelia throughout

the optic nerve head, as delineated by double immuno-

peroxidase staining for von Willebrand factor [Fig.

1(G)]. Also using double immunoperoxidase staining,

cells that stained positively for COX-1 were not the

same cells as those that stained positively for NOS-1

[Fig. 1(H)]. Few astrocytes, if any, were stained in the

postlaminar glial septa of the myelinated optic nerve.

No staining was associated with glial cells, astrocytes

or oligodendrocytes, inside the myelinated axon

bundles.

Glaucomatous tissues Antibody staining of COX-1

demonstrated intense immunoreactivity localized to

the compressed, prelaminar and lamina cribrosa of the

glaucomatous optic nerve head and did not extend

into the myelinated optic nerve beyond the level of the

lamina cribrosa [Fig. 1(B)]. Intense immunoreactivity

for COX-1 was localized to enlarged, rounded astro-

cytes occupying the diminished prelaminar region

[Fig. 1(D)] and the axon bundles at the level of the

disorganized lamina cribrosa [Fig. 1(F)]. Perinuclear

regions of the astrocytes demonstrated particularly

dense reaction product and the cytoplasmic areas of

many cells were also intensely stained [Fig. 1(F)].

Some remnant astrocytes and astrocytic processes in

the cribriform plates were also stained. Staining for

COX-1 was clearly localized to the optic nerve head,

labeled cells were not found in the postlaminar

myelinated nerve. As in normal tissue, the vascular

walls were not stained positively for COX-1 in

glaucomatous tissue.

Immunohistochemical Localization of COX-2 in Human

Optic Nerves

Normal tissues No staining for COX-2 was detected

in the normal optic nerve. All microscopic fields

examined in the prelaminar region [Fig. 2(A)], lamina

cribrosa [Fig. 2(C)] and postlaminar optic nerve [Fig.

2(E)] tissues did not contain positive staining for COX-

2. These findings are consistent with the absence of

COX-2 in normal tissue.

Glaucomatous tissues COX-2 was detected infrequently

in the glaucomatous optic nerve. Most microscopic

fields did not contain COX-2 positive cells. When a cell

was positive for COX-2, the staining was primarily

localized to the nucleus. Replacement of the primary

antibody with non-immune serum of PBS resulted in

the absence of any specific immunostaining. When

present in the compressed prelaminar region, a few

isolated, rounded cells were noted [Fig. 2(B)]. When

COX-2 positive staining was observed in the lamina

cribrosa, a few cells were in small clusters and

associated with the diminished, non-myelinated, axon

bundles or in the remnant cribriform plates [Fig. 2(D)].

Isolated, COX-2 positive cells were present in the

postlaminar, myelinated nerve but without a clear

regional distribution [Fig. 2(F)]. Microglia were present

in the glaucomatous optic nerve head [Fig. 2(G)].

When COX-2 positive cells were observed, double

immunoperoxidase staining did not co-localize these

cells to microglia in the glaucomatous tissue [Fig.

2(H)]. Overall, there was not a significant presence of

COX-2 positive cells in the glaucomatous optic nerve.

4. Discussion

Within the CNS, cyclooxygenase is associated with

neurons, glial cells and vascular endothelia in many

species (Smith, Gutekunst and Lyons, 1980; Bishai

and Coceani, 1992; Tsubokura et al., 1991). We have

demonstrated that COX-1 is constitutively present in

astrocytes in the glial columns and the cribiform plates

of the human optic nerve head. Neither the nerve

fibers nor the astrocytes within the nerve bundles

stain positively for COX-1. In addition, the vascular

endothelia lining the small vessels of the optic nerve

head do not exhibit COX-1 positive staining.

The constitutive presence of COX-1 in the astrocytes

of the cribriform plates suggests a physiological role for

the synthesis for eicosanoids in the human optic nerve

head. In tissue culture, glial cells have a higher

capacity to synthesize prostaglandins than do neurons

(Keller et al., 1985; Murphy et al., 1988). Glial cells

express prostanoid receptors EP, FP and TP (Inagaki

744 A. H. NEUFELD ET AL.

and Wada, 1994) and can be targets for prosta-

glandins to increase intracellular Ca#+, cyclic AMP and

phosphoinositol (Ito et al., 1992). Prostaglandins can

boost the intracellular energy supply of glia by

inducing the gene expression of creatine kinase,

presumably to maintain ATP levels during periods of

hypoxia or ischemia (Kuzhikndathil and Molloy,

1995). In vitro, prostaglandin E#

downregulates the

neurotoxicity of brain microglia (Thery, Dubbertin

and Mallat, 1994; Minghetti et al., 1997). In neonatal

cultures of glia, prostaglandins inhibit proliferation

(Granger and Kubes, 1994). The influences of the

other products of eicosanoid synthesis from arachi-

donic acid by COX-1 on CNS tissue have not been

extensively studied.

Based on the current concept of the role of COX-1 in

a variety of tissues as a protective mediator pathway

and the in vitro observations on glial cells exposed to

prostaglandins, we hypothesize that COX-1 in astro-

cytes is a neuroprotective pathway in the human optic

nerve head. Eicosanoids synthesized by COX-1 in

normal optic nerve may regulate the homeostasis of

astrocytes, supporting the neurons, and the local

vascular perfusion in this highly oxygen dependent

tissue.

In glaucoma, there may be an increase in the

amount of COX-1 localized to the damaged optic nerve

head. Although immunohistochemistry is not quan-

titative, COX-1 positive staining is clearly, intensely

associated with the astrocytes of the remodeled lamina

cribrosa. The astrocytes in the lamina cribrosa may be

generating eicosanoids from arachidonic acid that are

neuroprotective in the glaucomatous optic nerve head.

Prostaglandins may cause vasodilation to improve

vascular perfusion, inhibit platelet aggregation and

adhesion of lymphocytes, and potentiate the activity of

reactive astrocytes by elevating their ATP supply.

Enhanced generation of prostaglandins in the glau-

comatous optic nerve head may also be suppressing

the invasion of the tissue by hematogenous inflamma-

tory cells (DuBois, Bolton and Cuzner, 1986; Kunkel

et al., 1986) and the release of inflammatory cyto-

kines, such as TNF, IL-1, leukotrienes and PAF (Kundel

and Chensue, 1985; Ham et al., 1983; Haurand and

Floh, 1989). Compounds which increase intracellular

levels of cyclic AMP, such as prostaglandin E#

and

beta-adrenergic agonist, suppress peritoneal macro-

phages and inhibit the neurotoxic activity of activated

microglia (Thery, Dobbertin and Mallat, 1994; Ming-

hetti et al., 1997). Pathological observations of

glaucomatous optic nerves indicate that the neuronal

degeneration and remodeling of the extracellular

matrix are occuring in the absence of inflammation

(Hernandez and Pena, 1997). Perhaps prostaglandin

E#, synthesized by COX-1, is locally suppressing

invasion by circulating macrophages and the local

activity of macrophages and microglia in the glau-

comatous optic nerve head as part of a neuroprotective

role.

COX-2 is not present in normal optic nerves and is

only sparsely present, associated with astrocytes, in

the glaucomatous optic nerve. When present in the

CNS, COX-2 is induced by ischemia (Collaco-Moraes et

al., 1996; Ohtsuki et al., 1996) and cytokines in

neurons (Yamagata et al., 1993) and glia (O’Banion et

al., 1996). The relative absence of COX-2 in the

degenerating optic nerve in glaucoma is consistent

with the absence of inflammation and suggests that

inflammatory cytokines are not abundantly present in

the tissue. If we are correct that prostaglandin

synthesis by COX-1 suppresses the activity of macro-

phages and microglia in the glaucomatous optic nerve,

the suppressed activity of these cells would spare the

tissue from the destruction associated with inflam-

matory responses and the formation of a glial scar.

Nevertheless, the presence of a few astrocytes that

are positive for COX-2 in the glaucomatous optic nerve

head should not be overlooked as possibly contributing

to pathology. Glaucomatous optic neuropathy can be

a progressive disease that advances slowly over many

years. The sporadic appearance of COX-2 positive cells

may contribute eicosanoids to the tissue capable of

cytodestruction.

We conclude that COX-1 is constitutively present

and localized to the astrocytes of the lamina cribrosa of

the human optic nerve head. We hypothesize that in

this tissue, the enzyme generates prostaglandins and

other eicosanoids which serve a neuroprotective role

during optic nerve degeneration. If this role is to

suppress inflammation in the optic nerve, then COX-1

in the lamina cribrosa, and perhaps elsewhere in the

eye, is another component of the blood-ocular barrier

to limit inflammation-induced destruction of intra-

ocular tissue.

Acknowledgements

This work was supported in part by a grant from theGlaucoma Research Foundation, a NIH grant (EY-06416), aNIH Core Grant (EY-02687), and an unrestricted grant tothe Department of Ophthalmology and Visual Sciences fromResearch to Prevent Blindness, Inc.

The valuable technical assistance of Smita Vora andBelinda McMahan is gratefully acknowledged. We thankBernard Becker, for constantly updating our literature files.Human eyes were provided by the National Disease ResearchInterchange, Philadelphia, PA, the Mid America Eye andTissue Bank, St. Louis, MO, U.S.A., the Rochester Eye andHuman Parts Bank, Inc., Rochester, NY, U.S.A. and theGlaucoma Research Foundation, San Francisco, CA, U.S.A.

References

Adams, J., Collaco-Moraes, Y. and de Belleroche, J. (1996).Cyclooxygenase-2 induction in cerebral cortex: Anintracellular response to synaptic excitation. J. Neuro-chem. 66, 6–13.

Bazan, N. G., Marcheselli, V. L. and Mukherjee, P. K. (1995).Inducible prostaglandin synthase in cell injury. In:(Samuelsson et al., Eds) Advances in prostaglandin,thromboxane and leukotriene research. Vol 23. Pp.317–23; Raven Press : New York, U.S.A.

COX-1 AND COX-2 IN OPTIC NERVE HEAD 745

Bishai, I. and Coceani, F. (1992). Eicosanoid formation inthe rat cerebral cortex. Contribution of neurons andglia. Mol. Chem. Neuropathol. 3, 219–38.

Collaco-Moraes, Y., Aspey, B., Harrison, M. and DeBelleroche, J. (1996). Cyclo-oxygenase-2 messengerRNA in focal cerebral ischemia. J. Cereb. Blood FlowMetab. 16, 1366–72.

DeWitt, D. and Smith, W. L. (1995). Yes, but do they still getheadaches? Cell 83, 345–8.

Diaz-Araya, C. M., Provis, J. M., Penfold, P. L. and Billson,F. A. (1995). Development of microglia topography inhuman retina. J. Comp. Neurol. 363, 53–68.

Di Rosa, M. (1996). Interaction between nitric oxide andcyclooxygenase pathways. Prostaglandins, Leukotrienesand Essential Fatty Acids 54, 229–38.

DuBois, J. H., Bolton, C. and Cuzner, M. L. (1986). Theproduction of prostaglandin and the regulation of celldivision in neonate rat primary mixed glial cultures. J.Neuroimmunol. 11, 277–85.

Fu, J.-Y., Masferrer, J. L., Seibert, K., Raz, A. and Needleman,P. (1990). The induction and suppression of prosta-glandin H

#synthase (cyclooxygenase) in human mono-

cytes. J. Biol. Chem. 265, 16737–40.Goetzl, E. J., An, S. and Smith, W. L. (1995). Specificity of

expression and effects of eicosanoid mediators in normalphysiology and human diseases. FASEB J. 9, 1051–8.

Granger, D. N. and Kubes, P. (1994). The microcirculationand inflammation: Modulation of leukocyte-endothelialcell adhesion. J. Leukocyte Biol. 55, 662–675.

Ham, E. A., Soderman, D. D., Zanetti, M. E., Dougherty,H. W., McCauley, E. and Kuehl, F. A. Jr. (1983). Inhi-bition by prostaglandins of leukotrine B4 release fromactivated neutrophils. Proc. Natl. Acad. Sci. U.S.A. 80,4349–53.

Haurand, M. and Floh, L. (1989) Leukotriene formation byhuman polymorphonuclear leukocytes from endogen-ous arachidonate. Physiological triggers and modu-lation by prostanoids. Biochem. Pharmacol. 38, 2129–37.

Hernandez, M. R. and Pena, J. D. O. (1997). The optic nervehead in glaucomatous optic neuropathy. Arch. Ophthal-mol. 115, 389–95.

Inagaki, N. and Wada, H. (1994). Histamine and prostanoidreceptors on glial cells. Glia. 11, 102–9.

Ito, S., Sugama, K., Inagaki, N., Fukui, H., Giles, H., Wada,H. and Hayashi, O. (1992). Type-1 and type-2 astro-cytes are distinct targets for prostaglandins D

#, E

#and

F#α. Glia. 6, 67–74.

Kaufman, W. E., Worley, P. F., Pegg, J., Bremer, M. andIsakson, P. (1996). COX-2, a synaptically inducedenzyme, is expressed by excitatory neurons at post-synaptic sites in rat cerebral cortex. Proc. Natl. Acad. Sci.USA. 93, 2317–21.

Keller, M., Jackisch, R., Seregi, A. and Hertting, G. (1985).Comparison of prostanoid forming capacity of neuronaland astroglial cells in primary cultures. Neurochem. Intl.7, 655–65.

Kunkel, S. L. and Chensue, S. W. (1985). Arachidonic acidmetabolites regulate interleukin-1 production. Biochem.Biophys. Res. Commun. 128, 892–7.

Kunkel, S. L. (1986). Regulation of macrophage tumornecrosis factor production by prostaglandin E

#. Biochem.

Biophys. Acta. 137, 404–10.

Kuzhikandathil, E. V. and Molloy, G. R. (1995). Prosta-glandin E

", E

#, and cholera toxin increase transcription

of the brain creatine kinase gene in human U87glioblastoma cells. Glia. 15, 471–9.

Masferrer, J. L., Zweifel, B. S., Seibert, S. and Needleman, P.(1990). Selective regulation of cellular cyclooxygenaseby dexamethasone and endotoxin in mice. J. Clin. Inves.86, 1375–9.

Minghetti, L., Nicolini, A., Polazzi, E., Creminon, C., Maclouf,J. and Levi, G. (1997). Inducible nitric oxide synthaseexpression in activated rat microglial cultures is down-regulated by exogenous prostaglandin E

#and by cyclo-

oxygenase inhibitors. Glia. 19, 152–60.Mitchell, J. A., Akarasereenont, P., Thiemermann, C.,

Flower, R. J. and Vane, J. R. (1993). Selectivity of non-steroidal antiinflammatory drugs as inhibitors of consti-tutive and inducible cyclooxygenase. Proc. Natl. Acad.Sci. USA. 90, 11693–7.

Murphy, S., Pearce, B., Jeremy, J. and Dandona, P. (1988).Astrocytes as eicosanoid-producing cells. Glia. 1,241–5.

Neufeld, A. H., Hernandez, M. R. and Gonzalez, M. (1997).Nitric oxide synthase in the human glaucomatous opticnerve head. Arch. Ophthalmol. 115, 497–503.

O’Banion, M. K., Miller, J. C., Chang, J. W., Kaplan, M. D.and Coleman, P. D. (1996). Interleukin-1 beta inducesprostaglandin G}H synthetase (cyclooxygenase-2) inprimary murine astrocyte cultures. J. Neurochem. 66,2532–40.

Ohtsuki, T. (1996). Induction of cyclooxygenase-2 mRNA ingerbil hippocampal neurons after transient forebrainischemia. Brain Res. 736, 353–6.

Smith, W. L., Gutekunst, D. I. and Lyons, R. H. (1980).Immunocytochemical localization of the prostaglandin-forming cyclooxygenase in the cerebellar cortex. Prosta-glandins. 19, 61–9.

Thery, C., Dobbertin, A. and Mallat, M. (1994). Down-regulation of in vitro neurotoxicity of brain macro-phages by prostaglandin E

#and a beta-adrenergic

agonist. Glia. 11, 383–6.Tsubokura, S., Watanabe, Y., Ehara, H., Imamura, K.,

Sugimoto, O., Kagamiyama, H., Yamamoto, S. andHayaishi, O. (1991). Localization of prostaglandinendoperoxide synthase in neurons and glia in monkeybrain. Brain Res. 543, 15–24.

Vane, J. R. and Botting, R. M. (1995a). New insights into themode of action of anti-inflammatory drugs. Inflamm.Res. 44, 1–10.

Vane, J. R. and Botting, R. M. (1995b). A better under-standing of antiinflammatory drugs based on isoformsof cyclooxygenase (COX-1 and COX-2). In : Samuelssonet al., eds. Advances in Prostaglandin, Thromboxane, andLeukotriene Research. Vol 23. Raven Press ; 41–8.

Wallace, J. L. and Tigley, A. W. (1995). Review article : newinsights into prostaglandins and mucosal defence.Aliment. Pharmacol. Ther. 9, 227–35.

Wu, K. K. (1996). Cyclooxygenase 2 induction: Molecularmechanism and pathophysiologic roles. J. Lab. Clin.Med. 128, 242–5.

Yamagata, K., Andreasson, K. I., Kaufmann, W. E., Barnes,C. A. and Worley, P. F. (1993). Expression of a mitogen-inducible cyclooxygenase in brain neurons: Regulationby synaptic activity and glucocorticoids. Neurons. 11,371–86.