Expression of Glial Fibrillary AcidicProtein and Glutamine Synthetase by

Muller Cells After Optic NerveDamage and Intravitreal Applicationof Brain-Derived Neurotrophic Factor

HAO CHEN1 AND ARTHUR J. WEBER2*1Department of Pharmacology, University of Tennessee at Memphis, Memphis, Tennessee

2Department of Physiology and the Neuroscience Program, Michigan State University,East Lansing, Michigan

KEY WORDS retina; degeneration; neurotrophin; immunocytochemistry; ganglioncells

ABSTRACT Muller glia play an important role in maintaining retinal homeostasis,and brain-derived neurotrophic factor (BDNF) has proven to be an effective retinalganglion cell (RGC) neuroprotectant following optic nerve injury. The goal of thesestudies was to investigate the relation between optic nerve injury and Muller cellactivation, and to determine the extent to which BDNF affects the injury response ofMuller cells. Using immunocytochemistry and Western blot analysis, temporal changesin the expression of glial fibrillary acidic protein (GFAP) and glutamine synthetase (GS)were examined in rats after optic nerve crush alone, or in conjunction with an intra-vitreal injection of BDNF (5 �g). GFAP protein levels were normal at 1 day post-crush,but increased �9-fold by day 3 and remained elevated over the 2-week period studied.Muller cell GS expression remained stable after optic nerve crush, but the proteinshowed a transient shift in its cellular distribution; during the initial 24-h periodpost-crush the GS protein appeared to translocate from the cell body to the inner andouter glial processes, and particularly to the basal endfeet located in the ganglion celllayer. BDNF alone, or in combination with optic nerve crush, did not have a significanteffect on the expression of either GFAP or GS compared with the normal retina, or afteroptic nerve crush alone, respectively. The data indicate that although BDNF is a potentneuroprotectant in the vertebrate retina, it does not appear to have a significant influenceon Muller cell expression of either GS or GFAP in response to optic nerve injury. GLIA 38:115–125, 2002. © 2002 Wiley-Liss, Inc.

INTRODUCTION

The mammalian CNS contains two general types ofcells: neurons and glia. While neurons provide the ba-sis for complex electrophysiological activities, the pri-mary role of glia is support. However, in addition toproviding structural support to the nervous system,glia also have been shown to play an important role inneurotransmitter metabolism, to serve as modulatorsof synaptic transmission, and to help maintain a ho-

meostatic environment for neurons (see Kettenmann,1996, for review).

Grant sponsor: National Institutes of Health; Grant number: EY11159; Grantsponsor: Strategic Partnership Fund of Michigan State University.

*Correspondence to: Arthur J. Weber, Department of Physiology, B-512 WestFee Hall, Michigan State University, East Lansing, MI 48824.E-mail: weberar@msu.edu

Received 25 July 2001; Accepted 14 January 2002

DOI 10.1002/glia.10061

Published online 00 Month 2002 in Wiley InterScience (www.interscience.wiley.com).

GLIA 38:115–125 (2002)

© 2002 Wiley-Liss, Inc.

In the vertebrate retina, the predominant glia arethe Muller cells. These radial glia have cell bodies thatreside near the middle of the inner nuclear layer (INL);they then extend processes vertically across the fullthickness of the retina. The distal border of Muller cellsis marked by the outer limiting membrane, which con-sists of gap junctional complexes between the processesof different Muller cells, as well as Muller cells and thephotoreceptors. The inner (basal) endfeet of Mullercells do not form gap junctions, but they do form ad-herent junctions among themselves and with neighbor-ing astrocytes, and contribute to the inner limitingmembrane of the retina. Both the inner and outerprocesses of Muller cells are highly complex structuresthat interdigitate extensively with the photoreceptorsof the outer retina, and ganglion cells and optic nerveaxons of the inner retina (Bussow, 1980; Dowling,1987; Hollander et al., 1991). In contrast with retinalastrocytes, which vary in density as a function of nervefiber layer thickness, Muller cells have a relativelyuniform distribution (Bussow, 1980; Dreher et al.,1992), dividing the retina into distinct compartments,the microenvironments of which they control via theirion-gated channels, neurotransmitter receptors, andvarious transporter systems. More specifically, Mullercells play an important role in modulating neuronalactivity and maintaining retinal homeostasis by regu-lating intra- and extracellular levels of calcium, potas-sium, glutamate, and �-aminobutyric acid (GABA).Muller cells also regulate retinal pH levels by facilitat-ing the removal of carbon dioxide, and provide meta-bolic support via glycogenolysis (see Newman andReichenbach, 1996, for review).

Previous studies have demonstrated that brain-de-rived neurotrophic factor (BDNF) reduces, but does noteliminate, retinal ganglion cell death after optic nerveinjury (Mey and Thanos, 1993; Cohen et al., 1994; DiPolo et al., 1998; Isenmann et al., 1998; Peinado-Ramon et al., 1996; Chen and Weber, 2001). One goal ofthe present studies was to determine whether Mullercell activation might be influenced by the neuroprotec-tive effects of BDNF, as was demonstrated previouslyin the cat (Lewis et al., 1999). A second goal was todetermine whether its inability to prevent ganglion celldeath completely might be due, in part, to a decrease inMuller cell number or to their ability to respond to theinjury. To examine the relation between BDNF andMuller cell activation, the temporal expression of twoglia-specific proteins was examined after optic nervecrush alone, and in conjunction with the application ofBDNF. The two proteins examined were glial fibrillaryacidic protein (GFAP) and glutamine synthetase (GS).GFAP is a 51-kDa intermediate filament protein foundin the Muller cell endfeet and processes. Although thefunctional significance of GFAP remains unclear, thefact that it is upregulated by glial cells in response toinjury has led to its acceptance as an indication of anongoing injury response. GS is found in the Muller cellsoma. In the rat, GS is an octamer with eight identical45-kDa subunits. GS plays an important role in con-

trolling the level of extracellular glutamate, which, leftunchecked, can be excitotoxic to ganglion cells (Li et al.,1999; Kawasaki et al., 2000). Muller cells take up theexcess glutamate and, via GS, convert it to glutamine.They then release the glutamine into the extracellularspace, where it is taken up by ganglion cells and isreused to synthesize glutamate.

A decrease in GS or GFAP levels after optic nervedamage might signal either a decrease in Muller cellnumbers or an inability of these critical support cells toperform their normal maintenance role. These data areimportant for understanding the extent to which reti-nal ganglion cell death results from events unrelated todirect damage to the neurons themselves.

MATERIALS AND METHODSAnimals and Surgical Procedures

Adult Sprague-Dawley rats (n � 58) weighing 250–300 g were obtained from Charles River Laboratories(Wilmington, MA). Food and water were provided adlibitum, and the animals were maintained on a 12-hlight–dark cycle. All procedures used adhered to theNIH guidelines for the use of laboratory animals, andwere approved by the Animal Use Committee at Mich-igan State University.

Optic Nerve Crush Procedures

In preparation for the optic nerve crush, all rats wereanesthetized with buprenorphine HCl (0.4 mg/kg) andpentobarbital sodium (30 mg/kg). The right optic nervewas exposed by making a small incision in the lateralcanthus, and the nerve was crushed 1–2 mm posteriorto the globe for 15 s, using a pair of cross-action forceps.Care was taken to ensure that the crush did not com-promise the ophthalmic artery.

Intravitreal BDNF Treatment

Human recombinant BDNF (5 �g in 5 �l sterilesaline; Regeneron Pharmaceuticals, Tarrytown, NY)was injected into the vitreal chamber of each eye im-mediately after the optic nerve crush, using a Hamiltonsyringe with a 30-gauge needle. The injections weremade over a 30 s period, and the needle was left inposition for an additional 2 min, to allow for diffusion ofBDNF from the injection site and to minimize back-flow. Sham injections consisted of 5 �l of sterile saline.In all cases, care was taken to avoid the lens and ciliarybody, as these structures are potential sources of en-dogenous neurotrophic factors (Bennett et al., 1999;Leon et al., 2000) that can enhance retinal ganglionsurvival in the small rat eye (Mansour-Robaey et al.,1994).

116 CHEN AND WEBER

Immunohistochemistry

A total of 15 rats were used for GS and GFAP immu-nohistochemistry analyses. Rats receiving unilateraloptic nerve crush were allowed to survive for 1 day, 3days, 7 days, and 14 days. Three animals were exam-ined at the conclusion of each survival period. An ad-ditional three normal (unoperated) rats were used ascontrols, as previous work has shown that damage toone eye can result in the upregulation of injury-respon-sive genes in the contralateral eye (Bodeutsch et al.,1999). After their respective survival periods, all ani-mals were deeply anesthetized with pentobarbital so-dium (�35 mg/kg) and perfused transcardially with150 ml of saline, followed by 300 ml of a solutioncontaining 4% paraformaldehyde in 0.1 M phosphatebuffer (PB, pH 7.4). The eyes then were removed andpostfixed for an additional hour at room temperature(RT). The globes were bisected, and the complete pos-terior eye cup containing the retina, choroid, and sclerawas soaked in 20% sucrose in 0.1 M PB at 4°C. Thesucrose-saturated tissue then was embedded in HistoPrep (Fisher Chemical, Fair Lawn, NJ). For histologi-cal analysis, the eyecups were sectioned at 20 �m,using a cryostat, and the sections were mounted ontosubbed slides. To visualize GFAP, the sections werefirst incubated for 1 h at RT with CAS™solution(Zymed, San Francisco, CA) to block nonspecific bind-ing. Incubation in primary polyclonal anti-GFAP anti-body (1:1,000, 4°C, overnight; DAKO, Carpinteria, CA)was followed by incubation in a secondary goat-anti-rabbit IgG (1:200, 30 min, RT; Vector, Burlingame,CA). Texas Red Avidin D (1:2,000, 20 min, RT; Vector)was used for the final visualization. Similarly, GS wasdetected by first incubating the retinal tissue with theCAS™blocker (1 h, RT), followed by treatment with amonoclonal anti-GS antibody (1:1,000, 4°C, overnight;Chemicon, Temecula, CA). The sections then were in-cubated in horse-anti-mouse IgG (1:200, 30 min, RT;Vector). A diaminobenzidine staining kit (Scytek, Lo-gan, UT) was used to visualize the location of GSwithin the tissue. For all procedures, the sections werewashed 3 times (5 min each) in 0.01 M phosphate-buffered saline (PBS) between each step.

Glutamine Synthetase-Positive Cell Counts

The number of GS-positive cells was counted using amicroscope and �20 objective. Three adjacent cross-sections of the entire eye were obtained from each rat,and a total length of 2,240 �m of the inner nuclearlayer of the retina, representing all quadrants of theeye, was sampled randomly for each section. Cellcounts were not restricted to a specific quadrant of theretina because a systematic sampling of the entire ret-inal cross-section yielded a consistent Muller cell den-sity of 13.6 � 0.96 cells/100 �m. In all retinae, however,the cell counts were made at locations of equal distancefrom the optic nerve head. Differences in the number of

GS-positive Muller cells for normal and experimentaleyes were derived on the basis of comparisons of thenumber of cells per 100 �m of retinal tissue examined.

Western Blot Analysis

Forty-three rats were used for quantitativeGFAP/GS protein analysis, with the animals dividedrandomly into 3 groups of 12 animals each; the firstgroup received only an optic nerve crush, the secondgroup received only a single intravitreal injection ofBDNF (5 �g), and the third group received both anoptic nerve crush and a BDNF injection at the time ofthe nerve crush. Each group of animals then was di-vided into four survival period subgroups (3 animals/subgroup) that corresponded with those used for theimmunocytochemical analysis (1 day, 3 days, 7 days, 14days). Five additional rats were used as normal con-trols, and two received sham injections consisting of 5�l of sterile saline. No change in either GFAP or GSexpression was observed for any of the control or sham-injected animals. For Western blot analysis, the reti-nae were dissected in chilled 0.01 M PBS and werehomogenized in 500 �l of 20 mM Tris buffer (pH 7.5),using a TissueMite homogenizer (Tekmar, Cincinnati,OH). The buffer contained 1 mM sodium orthovana-date (Sigma, St. Louis, MO), 150 mM sodium chloride,10 mM sodium fluoride (Sigma), 1 mM phenylmethyl-sulfonyl fluoride (PMSF) (Sigma), 5 �g/ml leupeptin(Sigma), 10 �g/ml aprotinin (Sigma), and 1% NonidetP-40 (NP-40) (Sigma). Samples of the homogenized tis-sue then were microcentrifuged at 10,000 rpm for 10min at 4°C. Protein concentrations in the homogenizedtissue lysate were determined using the Bradford as-say kit (Bio-Rad, Hercules, CA). In this study, 1 �g ofprotein from each sample was mixed with equal vol-umes of a buffer containing 30% glycerol, 6% sodiumdodecyl sulfate (SDS), 0.075% bromphenol blue, 15%�-mercaptoethanol, and 187.5 mM Tris (pH 7.0). Themixtures then were boiled for 3 min and subjected to10% SDS-polyacrylamide gel electrophoresis (PAGE).The separated proteins were transferred to ImmobilonP membranes (Millipore, Bedford, MA) by electroblot-ting in a solution containing 25 mM Tris, 192 mMglycine, and 20% methanol (v/v, pH 8.3). To detectGFAP, the membranes first were incubated with 5%nonfat dry milk in Tris-buffered saline (TBS, 20 mMTris, 150 mM NaCl, pH 7.5) to block nonspecific bind-ing, then treated with a 1:10,000 dilution of a poly-clonal anti-GFAP antibody (DAKO), followed by a1:3,000 dilution of horseradish peroxidase (HRP)-con-jugated goat-anti-rabbit IgG (Santa Cruz Technology).For GS detection, the membranes also were firstsoaked in a solution of 5% nonfat dry milk in TBS, thenincubated with a 1:10,000 dilution of a monoclonalanti-GS antibody (Chemicon), followed by a 1:3,000dilution of HRP-conjugated horse-anti-mouse IgG(Santa Cruz). For both proteins, the immunoblots werewashed 3 times (10 min each) in TBS containing 0.1%

117GFAP AND GS EXPRESSION WITH ON CRUSH AND BDNF

Triton X-100 between each step. The blots were visu-alized by chemiluminescence detection using ECL re-agent (Amersham, Arlington Heights, IL) and the pro-tein information then was transferred to Hyperfilm MP(Amersham) for visualization. The films were scannedat 600 dpi and the resulting digital images analyzedquantitatively using Un-Scan-It gel software (Silk Sci-entific, Orem, UT). In all cases, protein samples ob-tained from normal and each group of experimentalanimals were presented on the same blot so that directcomparisons could be made between normal and exper-imental animals.

Quantitative changes in protein expression were as-sessed from the Western blots. Accuracy of these mea-surements was established by first determining that,for increasing amounts of total protein from the sameretinal sample (2 �g, 4 �g, 8 �g, 16 �g, 32 �g, 64 �g),there was a strong linear relation between total proteinloaded and signal intensity (r2 � 0.98, P 0.001, datanot shown).

Statistics

All protein analysis data were derived and are pre-sented as “percentage of normal.” These data are ex-pressed as mean � 1 SD. A one-sample t-test withBonferroni’s adjustment for multiple comparisons wasperformed using SPSS software (Chicago, IL), and thelevel of significance was adjusted to P � 0.05/number ofcomparisons.

RESULTSGFAP Expression

Immunocytochemistry

Muller cells from normal retinae showed a low levelof GFAP immunocytochemical staining, restricted pri-marily to the inner processes and basal endfeet of thesecells, located within the inner plexiform (IPL) and gan-glion cell (GC) layers of the retina, respectively (Fig.1A). The pattern and level of GFAP staining changedonly slightly over the initial 24 h post-crush (Fig. 1B)but increased significantly by day 3, when heavily la-beled processes could be seen forming parallel arraysthat now extended across the entire width of the retina(Fig. 1C). A few processes appeared to form extensivebranches, suggesting lateral interactions among neigh-boring Muller cells. Not only was dense GFAP-positivestaining seen over the remaining 2-week post-crushperiod examined, but there also appeared to be anincrease in the thickness of the individual processeswith longer post-crush durations (Fig. 1D,E). As ex-pected, the somata of the Muller cells within the INLdid not show positive GFAP staining at any of the timeperiods studied.

Western blot analysis

All the GFAP Western blots displayed distinct pro-tein bands with an anticipated molecular weight of�51 kDa (Fig. 2A). Intravitreal injection of BDNFalone had little effect on GFAP expression; over the2-week period studied there was, on average, only a1.7 � 0.33-fold increase in GFAP levels (range: 1.4–2.1). In agreement with the immunocytochemicalstaining, animals that received only a unilateral crushof the optic nerve showed little change in GFAP expres-sion over the initial 24 h post-crush. By 72 h, however,GFAP expression had increased �9.5-fold, and reacheda maximum of 10.5-fold at 2 weeks post-crush. WhenBDNF was provided at the time of the crush, GFAPexpression 24 h post-crush was only marginally higherthan that seen over the same time period followingeither procedure alone (2.5-fold vs 1.4-fold and 1.6-fold,respectively). As with optic nerve crush alone, however,GFAP expression showed a significant increase rela-tive to normal over the next 3–14 days, reaching amaximum of 18.4 � 13.1-fold in the 2 week animals. Asindicated by the high standard deviation (Fig. 2B), thisgroup of animals showed the greatest level of variabil-ity in protein expression. With the exception of thoseanimals receiving a 1-day survival period, both opticnerve crush and optic nerve crush combined withBDNF application resulted in levels of GFAP expres-sion that were significantly higher than normal (Fig.2B). No significant differences in GFAP expressionwere found among any of the BDNF and optic nervecrush-BDNF groups of animals.

Glutamine Synthetase Expression

Immunocytochemistry

In normal retinae, immunocytochemical staining forGS was heaviest within the Muller cell bodies, al-though weak label also could be seen among some ofthe Muller cell endfeet in the ganglion cell layer (Fig.3A). At 24 h after optic nerve crush, the pattern of GSlabeling was basically reversed; now there was strongGS immunoreactivity within the basal Muller cell pro-cesses within the ganglion cell layer, but almost no GSlabel associated with Muller cell bodies (Fig. 3B). Amodest increase in GS label also could be seen in theouter retina. GS immunoreactivity remained elevatedin the inner and outer retina on day 3 post-crush, andit also reappeared within the Muller cell somata (Fig.3C). By 1–2 weeks post-crush, the pattern of GS stain-ing had essentially returned to normal (Fig. 3D,E). Inorder to determine whether optic nerve injury resultsin a loss of Muller cells, the number of GS-positive cellbodies within the INL was counted in the whole eyeretinal cross-sections of normal and optic nerve crushanimals. There were 13 � 1 GS-positive cells/100 �m ofretina in the INL of the normal retinae, but only 7 � 2GS-positive cells/100 � m present 1 day after the nerve

118 CHEN AND WEBER

crush. The number of GS-positive cells per 100 � m ofINL measured at 3, 7, and 14 days post-crush were 8 �1, 13 � 2, and 12 � 2 cells, respectively. Since thereduction in GS-positive cells coincides with the shift inGS staining between the Muller cell bodies and theirinner processes, optic nerve damage does not appear toresult in an actual loss of Muller cells.

Western blot analysis

Western blot analyses of the retinae from the normaland optic nerve crush animals showed GS as a singleprotein band with a molecular weight of �45 kDa (Fig.4A). In general, the level of GS protein expression wasnot affected by any of the experimental conditions:optic nerve crush, intravitreal BDNF injection, or com-bination of the two (Fig. 4B).

DISCUSSION

Since its discovery in the 1970s, the function ofGFAP has remained elusive (Eng and Shiurba, 1988).Recent genetic advances have made it possible to gen-erate mice that carry a defective GFAP gene, and thusare incapable of expressing GFAP. There are at leastfour different lines of GFAP-null mice (Gomi et al.,1995; Pekny et al., 1995; Liedtke et al., 1996; McCall etal., 1996). Surprisingly, these mice undergo normaldevelopment and achieve adulthood without showingany anatomical or behavioral abnormalities. In addi-tion, the astrocytes of these mutant mice continue torespond to injury, although quantification of the re-sponse is difficult without GFAP as a marker (Pekny etal., 1995). Further, the absence of an obvious pheno-type does not appear to be attributable to a compensa-tory upregulation of other glial intermediate filaments,

Fig. 1. Immunofluorescent staining showed limited amounts of glialfibrillary acidic protein (GFAP) within the inner processes of Muller cellsin the normal retina (A). GFAP expression increased only modestly overthe first 24 h post-crush (B) but was significantly greater than normal at3 days after nerve crush (C). GFAP expression reached a peak at 7 dayspost-crush (D) and remained elevated over the entire 14-day period exam-ined (E). Muller cell bodies did not stain positive for GFAP. GCL, ganglioncell layer; IPL, inner plexiform layer; INL, inner nuclear layer; OPL, outerplexiform layer; ONL, outer nuclear layer; IS/OS, inner and outer seg-ments. Scale bar � 50 �m.

119GFAP AND GS EXPRESSION WITH ON CRUSH AND BDNF

such as vimentin or nestin (Gomi et al., 1995; Pekny etal., 1995; McCall et al., 1996). Subsequent studies ofGFAP-null mice have demonstrated enhanced long-term potentiation in the hippocampus (McCall et al.,1996) and deficient long-term depression in cerebellarPurkinje cells (Shibuki et al., 1996). Although thesefindings suggest that GFAP might play a role in regu-lating synaptic transmission, the mechanism remainsunknown. Nevertheless, these data support the gen-eral view that increased expression of GFAP repre-sents a sign of glial plasticity in response to a changingneuronal environment.

In the visual system, increased expression of GFAPby Muller cells has been reported after a wide variety ofretinal injuries, including mechanical damage (Big-nami and Dahl, 1979), lensectomy-vitrectomy (Yoshidaet al., 1993), photoreceptor degeneration (Eisenfeld etal., 1984; Burns and Robles, 1990; DiLoreto Jr. et al.,1995; de Raad et al., 1996), experimental retinal de-tachment (Hiscott et al., 1984; Erickson et al., 1987;Lewis et al., 1999), diabetic retinopathy (Lieth et al.,1998; Mizutani et al., 1998; Rungger-Brandle et al.,2000), retinal ischemia (Fitzgerald et al., 1990; Os-borne et al., 1991), and both experimentally-inducedglaucoma (Tanihara et al., 1997; Wang et al., 2000) andhuman glaucoma (Hiscott et al., 1984). Elevated levelsof GFAP in Muller cells also have been observed afteroptic nerve transection (Scherer and Schnitzer, 1991;Huxlin et al., 1995). Taken together, these data haveled to the general acceptance that increased expressionof GFAP by Muller cells is a reliable indicator of retinalneuropathy. To the best of our knowledge, this study isthe first to demonstrate that compressive damage tothe optic nerve also results in an increase in GFAPexpression by Muller cells. Further, examination ofGFAP levels over different time periods indicates thatalthough the initial injury response (0–24 h) is rela-

tively mild (�1.7-fold), Muller cell production of GFAPincreases dramatically (�10-fold) over the next 48 hand remains elevated for at least 2 weeks. Equally highlevels of GFAP expression also were present 3 weeksafter optic nerve crush alone (data not shown).

While our immunohistochemical analysis identifiedMuller cells as the primary source of increased retinalGFAP after optic nerve crush (Fig. 1), it is important tonote that they are not the only source of GFAP in thevertebrate retina. Previous studies have shown that apopulation of astrocytes within the nerve fiber layeralso produce GFAP (Eisenfeld et al., 1984; Vaughan etal., 1990). Although these cells are not easily distin-guishable from the endfeet of Muller cells when viewedusing retinal cross-sections, they are clearly visible onretina whole mounts. It is unlikely, however, that theseglia contributed significantly to the increases in GFAPseen here, as these retinal astrocytes do not increasetheir levels of GFAP expression in response to opticnerve transection (Scherer and Schnitzer, 1991; Huxlinet al., 1995) This lack of a response to nerve injurymight result from the fact that these retinal astrocytesappear to associate most closely with the retinal vas-culature, and not with ganglion cell axons (Scherer andSchnitzer, 1991; Huxlin et al., 1995).

The initial change in GFAP expression by Mullercells after optic nerve crush coincides with the onset ofRGC degeneration, but it appears to increase at a rategreater than that of actual ganglion cell loss. Whileprevious studies in the adult rat have indicated nosignificant decrease in ganglion cell number until 4–5days post-optic nerve injury (Mansour-Robaey et al.,1994; Berkelaar et al., 1994; Peinado-Ramon et al.,1996), our immunocytochemical and Western blot dataindicate significantly higher than normal (�10-fold)levels of Muller cell GFAP expression within 3 dayspost-crush. Although we cannot rule out the possibility

Fig. 2. A: Western blot analysis of glial fibrillary acidic protein(GFAP) expression under each of the experimental conditions used.Equal amount of total retina protein were separated by SDS-PAGEand detected using a GFAP-specific antibody. The molecular weight ofall bands was �51 kDa. Data are from three eyes under each condi-tion. B: Semiquantitative analysis of GFAP expression showing that

although brain-derived neurotrophic factor (BDNF) alone has littleeffect on GFAP levels, optic nerve crush and nerve crush combinedwith BDNF treatment results in a significant increase in GFAP pro-tein compared with normal. The difference between these two condi-tions, however, was not significant. (*P 0.05, vs normal; error barsare �1 SD; N, normal; 1d, 3d, 7d, 14d, post-crush survival in days).

120 CHEN AND WEBER

that even a minor loss of ganglion cells induces a fullgliotic response from Muller cells, it is conceivable thatMuller cells respond to retinal abnormalities that pre-cede actual ganglion cell death. One factor in thismight be a reduction in the level of neuronal activitywithin the injured retina. Reduced afferent input hasbeen shown to result in increased GFAP expression atthe frog neuromuscular junction (Georgiou et al.,1994), within the chick cochlear nucleus (Canady andRubel, 1992; Rubel and MacDonald, 1992), and the ratlateral geniculate nucleus (Canady et al., 1994). Withinthe retina, a decrease in ganglion cell activity afteroptic nerve crush might occur as a result of a reductionin afferent input from bipolar cells due to degenerativeabnormalities in ganglion cell dendritic field structure(Mey and Thanos, 1993; Watanabe et al., 1995; Weberet al., 1998). In addition, ganglion cell activity mightalso be reduced due to a decrease in the level of trophic

materials transported retrogradely to the retina fromtarget neurons located outside the eye. Previous stud-ies have shown that BDNF can evoke rapid excitationof cortical neurons via their high-affinity tyrosine re-ceptor kinase B (TrkB) receptors (Kafitz et al., 1999),and it is well known that damage to the optic nervedisrupts retrograde transport of the receptor–ligandcomplex from retinal target nuclei in the brain (Peaseet al., 2000). Because Muller cell processes are inti-mately associated with synapses within the retinalplexiform layers, and since they are capable of detect-ing changes in neuronal activity via their neurotrans-mitter receptors (Peng et al., 1995; Lopez et al., 1997),it is plausible that a reduction in ganglion cell activitymay also signal Muller cells to increase GFAP expres-sion. Long-term increases in GFAP expression, how-ever, most likely result from soluble factors released byneurons and glia. In particular, transforming growth

Fig. 3. Immunohistochemistry showing glutamine synthetase (GS)-positive staining in association with Muller cells. In normal retinae (A),GS immunoreactivity is strongest within the cell soma, while at 1 daypost-crush (B), the highest levels of protein appear within the basalendfeet, adjacent to ganglion cells. The pattern of label shifts backtoward the INL and cell somata at around 3 days post-crush (C), and by1–2 weeks post-crush (D,E) has returned to normal. GCL, ganglion celllayer; IPL, inner plexiform layer; INL, inner nuclear layer; OPL, outerplexiform layer; ONL, outer nuclear layer; IS/OS, inner and outer seg-ments). Scale bar � 50 �m.

121GFAP AND GS EXPRESSION WITH ON CRUSH AND BDNF

factor (Reilly et al., 1998; Zhou and Skalli, 2000), fibro-blast growth factor (Reilly et al., 1998), and ciliaryneurotrophic factor (Levison et al., 1998; Gomes et al.,1999), have been shown to be present at increasedlevels after neuronal injury, and these factors also areknown to modulate GFAP expression.

The fact that BDNF alone did not produce an in-crease in GFAP expression in the normal retina wasnot unexpected (Fig. 2). Previous studies have shownthat BDNF has no effect on GFAP expression in cul-tured fetal rat astrocytes (Hoglinger et al., 1998) orwhen injected in high concentration (100 �g) into thenormal cat eye (Lewis et al., 1999). Similarly, Wahlinet al. (2000) found that although intravitreal injectionsof both BDNF and CNTF (ciliary neurotrophic factor)into the mouse eye produce increases in the immediateearly gene c-fos and phosphorylated extracellular sig-nal-regulated kinase (pERK), only CNTF yields an in-crease in Muller cell GFAP immunoreactivity. How-ever, because the increase in GFAP expression isdelayed, starting at least 6 h post-injection, long afterc-fos and pERK levels have returned to normal, it ap-pears most likely that CNTF does not have a directeffect on Muller cell GFAP expression. This is not sur-prising, as Muller cells have been shown to expressCNTF in response to injury, but they do not appear tocontain the CNTF receptor (Ju et al., 1999; Chun et al.,2000). Similarly, in preliminary studies in which wecombined GS and TrkB immunocytochemistry, wewere unable to demonstrate co-localization of TrkB andGS in Muller cells from normal retinae (Chen andWeber, 1999), again indicating that GFAP expressionwithin the retina is not directly linked to the signalingcascades of BDNF.

We did, however, expect to see a reduction in GFAPexpression after optic nerve injury and BDNF applica-tion. This expectation was based on our recent work in

the cat (Chen and Weber, 2001), as well as our own andthat of others in the rat (Mey and Thanos, 1993; Cohenet al., 1994; Peinado-Ramon et al., 1996; Di Polo et al.,1998; Isenmann et al., 1998), showing that BDNF pro-motes RGC survival and reduces neuronal degenera-tion after optic nerve injury; optic nerve crush aloneproduces a 36% reduction in ganglion cell number 7days post-crush, while treatment with BDNF using themethods of this study results in only a 6% loss ofganglion cells over the same survival period (Chen,2001). Thus, in agreement with Lewis et al. (1999) whofound that BDNF attenuates GFAP expression in catsafter retinal detachment, we expected that by reducingthe level of ganglion cell death, BDNF also would leadto a decrease in GFAP expression in rats with opticnerve damage. In contrast, we found that application ofBDNF at the time of nerve crush resulted in GFAPexpression levels that were either similar or slightlyhigher than those measured after optic nerve crushalone. The different results of these studies might beexplained by several factors, including the animal spe-cies used (cats vs rats), model of injury (optic nervecrush vs retinal detachment), and GFAP quantificationmethod (Western blot vs image analysis). In addition, itis important to note that the dose of BDNF used byLewis et al. (1999) was considerably higher than thatwhich we found optimal for preserving cat retinal gan-glion cells after optic nerve crush (100 �g vs 30 �g).Although the present study is based on rats, when onetakes into account the species difference in vitreal vol-ume (rat: 50 �l; cat: 3 ml), the optimal dose of BDNFused in our cat and rat studies are similar (0.01 �gBDNF/�l vitreal volume) (Chen and Weber, 2001).Thus, it is possible that dose is an important factor foridentifying BDNF-related effects on GFAP expression.

Although the overall level of GFAP was not affectedby BDNF, there is some evidence that BDNF may

Fig. 4. Western blot analysis of glutamine synthetase (GS). A:Equal amount of total retina protein were separated by SDS-PAGE.GS was visualized by specific antibody. The molecular weight of thedetected bands was �45 kDa. Top: brain-derived neurotrophic factor(BDNF) application only; middle: optic nerve crush only; bottom: opticnerve crush and BDNF application. Data are from three eyes under

each condition. B: Semiquantitative analyses of glutamine synthetaselevels showed that the amount of glutamine synthetase protein wasnot affected by BDNF application, optic nerve crush, or a combinationof the two procedures. Error bars are � 1 SD. N, normal; 1d, 3d, 7d,14d, post-crush survival in days).

122 CHEN AND WEBER

affect the turnover rate of GFAP. In a pilot study, wenoted the presence of multiple, rather than a single,protein band, on each GFAP Western blot (Fig. 2).Adjusting the concentration of �-mercaptoethanol inthe gel-loading buffer reduced the intensity of thelower band, but did not eliminate it. While a weaklower band could be seen under most conditions, in-cluding some normal subjects, it was most prominentin the crush plus BDNF animals. One possible expla-nation for the multiple bands is post-translationalmodification of GFAP. Because phosphorylation or gly-cosylation leads to a slight increase in protein molecu-lar weight, it is possible that the upper band reflectsphosphorylated/glycosylated GFAP. Similarly, becausestepwise proteolytic breakdown of GFAP has beenshown to produce multiple bands with lower molecularweights (DeArmond et al., 1983), it is possible that thelower band represents the result of this degradativeprocess. Although Western blot analysis does not per-mit precise determination of molecular weight, com-parison of the positions of the different bands with thenormal data suggests that the lower bands in our gelsmost likely represent proteolytic cleavage of the 51-kDa GFAP protein (upper band). The presence of asecond band in normal animals was variable but notunexpected, as one would expect some basal level ofGFAP production and degradation even in these eyes.Because the overall GFAP content in the crush-BDNFgroup was not statistically different from that of thecrush only group (upper bands of each), the very prom-inent lower band (degraded GFAP) for the BDNF-crushanimals suggests that in actuality more GFAP wasproduced in these animals.

One of the more interesting findings of these studieswas the change in the pattern of GS labeling thatoccurred as a result of the optic nerve injury. Becausethe shift from dense labeling within the Muller cellsomata to the inner, and to a lesser degree the outer,retina did not involve a net increase in GS proteinlevels, it seems reasonable to assume that the initialresponse of Muller cells to optic nerve injury involves arapid translocation of GS from the soma to that regionof the cell located closest to the primary site of injury.This response is not surprising, since Muller cells playan important role in the control of retinal glutamate,and glutamate has been shown to be excitotoxic in theretina, as well as other areas of the central nervoussystem; direct application of glutamate to the retinaresults in retinal ganglion cell death (Li et al., 1999),and the degenerative effects can be alleviated by theapplication of glutamate receptor antagonists (Kapinet al., 1999). The modest increase in GS immunoreac-tivity in the outer retina most likely reflects either aglobal injury response by Muller cells, or a low levelresponse to optic nerve injury-induced changes at thephotoreceptor level (Nork et al., 2000). In addition, wecannot rule out the possibility that some of the GSimmunoreactivity might result from astrocytes; GS-positive astrocytes have been identified in the rat brain(Norenberg and Martinez-Hernandez, 1979) and retina

(Derouiche and Rauen, 1995), and astrocytic GS ex-pression has been shown to be upregulated after spinalcord injury (Benton et al., 2000). However, becauseretinal astrocytes are most closely associated with thevasculature (Huxlin et al., 1995; Scherer andSchnitzer, 1991; Lehre et al., 1997), and not with gan-glion cells, it seems unlikely that astrocyte-derived GScontributed significantly to the transient increase ininner retina GS immunoreactivity noted here (Fig. 3).Nevertheless, a more thorough investigation that com-bines double label immunocytochemistry and ultra-structural analysis is needed to define more clearly theGS contribution of each cell type (Muller vs astrocytes).

Unlike the case with GFAP, we found the level ofglutamine synthetase to remain relatively stable undereach of the experimental conditions tested. The lack ofa GS response might be due to the fact that photore-ceptors, and not ganglion cells, are a major source ofglutamate within the retina. Thus, any small addi-tional amount of glutamate released by degeneratingRGCs may not increase overall glutamate levelsenough to induce a change in GS production. Increas-ing the availability of GS, however, has been shown tobe beneficial for the survival of RGCs. Vardimon (2000)reported that application of purified GS into retinalexplants promoted the survival of RGCs. Because ourresults show that BDNF reduces RGC death after opticnerve injury without increasing GS expression, itwould be interesting to know whether a combination ofBDNF and GS might result in even greater ganglioncell survival than either treatment alone. That GSlevels did not increase in a manner similar to thoseseen for GFAP is not surprising, as a similar dissocia-tion between GS and GFAP expression has been dem-onstrated in human diabetic retinopathy (Mizutani etal., 1998).

In summary, the central goal of these studies was toexamine the relationship between optic nerve injuryand Muller cell activation, and to determine the extentto which BDNF might affect the injury response, eitherby direct interaction with these radial glia or indirectlyby means of its neuroprotective properties. The datademonstrate that optic nerve crush, like other forms ofoptic nerve/retinal neuropathy, induces Muller cell ac-tivation, as indicated by an increase in GFAP expres-sion. Application of BDNF alone, or in conjunction withan optic nerve crush, however, had no significant effecton GFAP expression, suggesting that, at least withrespect to this component of the Muller cell injuryresponse, BDNF does not have a clear influence. Unlikeganglion cells, optic nerve injury produces little or nochange in the number of retinal Muller cells. Althoughupregulation of GS by Muller cells does not appear tobe a necessary part of the injury response after opticnerve damage, the transient translocation of proteinfrom the cell soma to regions of neuronal injury indi-cates a clear neuroprotective response. Nevertheless,additional studies are needed for a better understand-ing of the complex relationship between Muller cellsand retinal neurons. In particular, a better under-

123GFAP AND GS EXPRESSION WITH ON CRUSH AND BDNF

standing is needed with respect to the spatial andtemporal relations of the various injury response com-ponents (e.g., early gene expression, second messen-gers, GFAP/GS expression, trophic factor response/pro-duction) of Muller cells and their neuronal neighbors.

ACKNOWLEDGMENT

The authors acknowledge the technical assistance ofMrs. Judy McMillan.

REFERENCES

Bennett JL, Zeiler SR, Jones KR. 1999. Patterned expression of BDNFand NT-3 in the retina and anterior segment of the developingmammalian eye. Invest Ophthalmol Vis Sci 40:2996–3005.

Benton RL, Ross CD, Miller KE. 2000. Glutamine synthetase activi-ties in spinal white and gray matter 7 days following spinal cordinjury in rats. Neurosci Lett 291:1–4.

Berkelaar M, Clarke DB, Wang YC, Bray GM, Aguayo AJ. 1994.Axotomy results in delayed death and apoptosis of retinal ganglioncells in adult rats. J Neurosci 14:4368–74.

Bignami A, Dahl D. 1979. The radial glia of Muller in the rat retinaand their response to injury. An immunofluorescence study withantibodies to the glial fibrillary acidic (GFA) protein. Exp Eye Res28:63–69.

Bodeutsch N, Siebert H, Dermon C, Thanos S. 1999. Unilateral injuryto the adult rat optic nerve causes multiple cellular responses in thecontralateral site. J Neurobiol 38:116–128.

Burns MS, Robles M. 1990. Muller cell GFAP expression exhibitsgradient from focus of photoreceptor light damage. Curr Eye Res9:479–486.

Bussow H. 1980. The astrocytes in the retina and optic nerve head ofmammals: a special glia for the ganglion cell axons. Cell Tissue Res206:367–378.

Canady KS, Rubel EW. 1992. Rapid and reversible astrocytic reactionto afferent activity blockade in chick cochlear nucleus. J Neurosci12:1001–1009.

Canady KS, Olavarria JF, Rubel EW. 1994. Reduced retinal activityincreases GFAP immunoreactivity in rat lateral geniculate nucleus.Brain Res 663:206–214.

Chen H. 2001. Brain-derived neurotrophic factor promotes retinalganglion cell survival and reduces tyrosine receptor kinase B pro-tein and mRNA in vivo. Doctoral dissertation. p 1–25.

Chen H, Weber AJ. 1999. Response of Muller cells to optic nerve crushin the rat. Soc Neurosci Abs 25:1794.

Chen H, Weber AJ. 2001. BDNF enhances retinal ganglion cell sur-vival in cats with optic nerve damage. Invest Ophthalmol Vis Sci42:966–974.

Chun MH, Ju WK, Kim KY, Lee MY, Hofmann, HD, Kirsch M, Oh SJ.2000. Upregulation of ciliary neurotrophic factor in reactive Mullercells in the rat retina following optic nerve transection. Brain Res868:358–362.

Cohen A, Bray GM, Aguayo AJ. 1994. Neurotrophin-4/5 (NT-4/5)increases adult rat retinal ganglion cell survival and neurite out-growth in vitro. J Neurobiol 25:953–959.

de Raad S, Szczesny PJ, Munz K, Reme CE. 1996. Light damage in therat retina: glial fibrillary acidic protein accumulates in Muller cellsin correlation with photoreceptor damage. Ophthalmic Res 28:99–107.

DeArmond SJ, Fajardo M, Naughton SA, Eng LF. 1983. Degradationof glial fibrillary acidic protein by a calcium dependent proteinase:an electroblot study. Brain Res 262:275–282.

Derouiche A, Rauen T. 1995. Coincidence of L-glutamate/L-aspartatetransporter (GLAST) and glutamine synthetase (GS) immunoreac-tions in retinal glia: evidence for coupling of GLAST and GS intransmitter clearance. J Neuroci Res 42:131–143.

Di Polo A, Aigner LJ, Dunn RJ, Bray GM, Aguayo AJ. 1998. Prolongeddelivery of brain-derived neurotrophic factor by adenovirus-infectedMuller cells temporarily rescues injured retinal ganglion cells. ProcNatl Acad Sci USA 95:3978–3983.

DiLoreto DA Jr, Martzen MR, del Cerro C, Coleman PD, del Cerro M.1995. Muller cell changes precede photoreceptor cell degeneration

in the age-related retinal degeneration of the Fischer 344 rat. BrainRes 698:1–14.

Dowling JE. 1987. The retina: an approachable part of the brain.Belknap Press/Harvard University Press: Cambridge, MA.

Dreher Z, Robinson SR, Distler C. 1992. Muller cells in vascular andavascular retinae: a survey of seven mammals. J Comp Neurol323:59–80.

Eisenfeld AJ, Bunt-Milam AH, Sarthy PV. 1984. Muller cell expres-sion of glial fibrillary acidic protein after genetic and experimentalphotoreceptor degeneration in the rat retina. Invest OphthalmolVis Sci 25:1321–1328.

Eng LF, Shiurba RA. 1988. Glial fibrillary acidic protein: A review ofstructure, function, and clinical application. In: Marangos PJ,Campbell IC, Cohen RM, editors. Neuronal and glial proteins. SanDiego, CA: Academic Press. p 339–359.

Erickson PA, Fisher SK, Guerin CJ, Anderson DH, Kaska DD. 1987.Glial fibrillary acidic protein increases in Muller cells after retinaldetachment. Exp Eye Res 44:37–48.

Fitzgerald ME, Vana BA, Reiner A. 1990. Evidence for retinal pathol-ogy following interruption of neural regulation of choroidal bloodflow: Muller cells express GFAP following lesions of the nucleus ofEdinger-Westphal in pigeons. Curr Eye Res 9:583–598.

Georgiou J, Robitaille R, Trimble WS, Charlton MP. 1994. Synapticregulation of glial protein expression in vivo. Neuron 12:443–455.

Gomes FC, Paulin D, Moura NV. 1999. Glial fibrillary acidic protein(GFAP): modulation by growth factors and its implication in astro-cyte differentiation. Braz J Med Biol Res 32:619–631.

Gomi H, Yokoyama T, Fujimoto K, Ikeda T, Katoh A, Itoh T, ItoharaS. 1995. Mice devoid of the glial fibrillary acidic protein developnormally and are susceptible to scrapie prions. Neuron 14:29–41.

Hiscott PS, Grierson I, Trombetta CJ, Rahi AH, Marshall J, McLeodD. 1984. Retinal and epiretinal glia—an immunohistochemicalstudy. Br J Ophthalmol 68:698–707.

Hoglinger GU, Sautter J, Meyer M, Spenger C, Seiler RW, Oertel WH,Widmer HR. 1998. Rat fetal ventral mesencephalon grown as solidtissue cultures: influence of culture time and BDNF treatment ondopamine neuron survival and function. Brain Res 813:313–322.

Hollander H, Makarov F, Dreher Z, van-Driel D, Chan-Ling T-L,Stone J. 1991. Structure of the macroglia of the retina: sharing anddivision of labour between astrocytes and Muller cells. J CompNeurol 313:587–603.

Huxlin KR, Dreher Z, Schulz M, Dreher B. 1995. Glial reactivity inthe retina of adult rats. Glia 15:105–118.

Isenmann S, Klocker N, Gravel C, Bahr M. 1998. Protection of axo-tomized retinal ganglion cells by adenovirally delivered BDNF invivo. Eur J Neurosci 10:2751–2756.

Ju WK, Lee MY, Hofmann HD, Kirsch M, Chun MH. 1999. Expressionof CNTF in Muller cells of the rat retina after pressure-inducedischemia. NeuroReport 10:419–422.

Kafitz KW, Rose CR, Thoenen H, Konnerth A. 1999. Neurotrophin-evoked rapid excitation through TrkB receptors. Nature 401:918–921.

Kapin MA, Doshi R, Scatton B, DeSantis LM, Chandler ML. 1999.Neuroprotective effects of eliprodil in retinal excitotoxicity and isch-emia. Invest Ophthalmol Vis Sci 40:1177–1182.

Kawasaki A, Otori Y, Barnstable CJ. 2000. Muller cell protection ofrat retinal ganglion cells from glutamate and nitric oxide neurotox-icity. Invest Ophthalmol Vis Sci 41:3444–3450.

Kettenmann H. 1996. Beyond the neuronal circuitry. TINS 19:305–306.

Lehre KP, Davanger S, Danbolt NS. 1997. Localization of the gluta-mate transporter protein GLAST in rat retina. Brain Res 744:129–137.

Leon S, Yin Y, Nguyen J, Irwin N, Benowitz LI. 2000. Lens injurystimulates axon regeneration in the mature rat optic nerve. J Neu-rosci 20:4615–4626.

Levison SW, Hudgins SN, Crawford JL. 1998. Ciliary neurotrophicfactor stimulates nuclear hypertrophy and increases the GFAPcontent of cultured astrocytes. Brain Res 803:189–193.

Lewis GP, Linberg KA, Geller SF, Guerin CJ, Fisher SK. 1999. Effectsof the neurotrophin brain-derived neurotrophic factor in an exper-imental model of retinal detachment. Invest Ophthalmol Vis Sci40:1530–1544.

Li Y, Schlamp CL, Nickells RW. 1999. Experimental induction ofretinal ganglion cell death in adult mice. Invest Ophthalmol Vis Sci40:1004–1008.

Liedtke W, Edelmann W, Bieri PL, Chiu FC, Cowan NJ, KucherlapatiR, Raine CS. 1996. GFAP is necessary for the integrity of CNS whitematter architecture and long-term maintenance of myelination.Neuron 17:607–615.

124 CHEN AND WEBER

Lieth E, Barber AJ, Xu B, Dice B, Ratz MJ, Tanase D, Strother JM.1998. Glial reactivity and impaired glutamate metabolism in short-term experimental diabetic retinopathy. Diabetes 47:815–820.

Lopez T, Lopez-Colome AM, Ortega A. 1997. NMDA receptors incultured radial glia. FEBS Lett 405:245–248.

Mansour-Robaey S, Clarke DB, Wang YC, Bray GM, Aguayo AJ.1994. Effects of ocular injury and administration of brain-derivedneurotrophic factor on survival and regrowth of axotomized retinalganglion cells. Proc Natl Acad Sci USA 91:1632–1636.

McCall MA, Gregg RG, Behringer RR, Brenner M, Delaney CL, Gal-breath EJ, Zhang CL, Pearce RA, Chiu SY, Messing A. 1996. Tar-geted deletion in astrocyte intermediate filament (GFAP) altersneuronal physiology. Proc Natl Acad Sci USA 93:6361–6366.

Mey J, Thanos S. 1993. Intravitreal injections of neurotrophic factorssupport the survival of axotomized retinal ganglion cells in adultrats in vivo. Brain Res 602:304–317.

Mizutani M, Gerhardinger C, Lorenzi M. 1998. Muller cell changes inhuman diabetic retinopathy. Diabetes 47:445–449.

Newman E, Reichenbach A. 1996. The Muller cell: a functional ele-ment of the retina. Trends Neurosci 19:307–312.

Norenberg MD, Martinez-Hernandez A. 1979. Fine structural local-ization of glutamine synthetase in astrocytes of rat brain. Brain Res161:303–310.

Nork TM, VerHoeve JN, Poulsen GL, Nickells RW, Davis MD, WeberAJ, Vaegan, Sarks SH, Lemley HL, Millecchia LL. 2000. Swellingand loss of photoreceptors in chronic human and experimentalglaucomas. Arch Ophthalmol 118:235–245.

Osborne NN, Block F, Sontag KH. 1991. Reduction of ocular bloodflow results in glial fibrillary acidic protein (GFAP) expression inrat retinal Muller cells. Vis Neurosci 7:637–639.

Pease ME, McKinnon SJ, Quigley HA, Kerrigan-Baumrind LA, ZackDJ. 2000. Obstructed axonal transport of BDNF and its receptorTrkB in experimental glaucoma. Invest Ophthalmol Vis Sci 41:764–774.

Peinado-Ramon P, Salvador M, Villegas-Perez MP, Vidal-Sanz M.1996. Effects of axotomy and intraocular administration of NT-4,NT-3, and brain-derived neurotrophic factor on the survival of adultrat retinal ganglion cells. A quantitative in vivo study. Invest Oph-thalmol Vis Sci 37:489–500.

Pekny M, Leveen P, Pekna M, Eliasson C, Berthold CH, WestermarkB, Betsholtz C. 1995. Mice lacking glial fibrillary acidic proteindisplay astrocytes devoid of intermediate filaments but develop andreproduce normally. EMBO J 14:1590–1598.

Peng YW, Blackstone CD, Huganir RL, Yau KW. 1995. Distribution ofglutamate receptor subtypes in the vertebrate retina. Neuroscience66:483–497.

Reilly JF, Maher PA, Kumari VG. 1998. Regulation of astrocyte GFAPexpression by TGF-beta1 and FGF-2. Glia 22:202–210.

Rubel EW, MacDonald GH. 1992. Rapid growth of astrocytic processesin N. magnocellularis following cochlea removal. J Comp Neurol318:415–425.

Rungger-Brandle E, Dosso AA, Leuenberger PM. 2000. Glial reactiv-ity, an early feature of diabetic retinopathy. Invest Ophthalmol VisSci 41:1971–1980.

Scherer J, Schnitzer J. 1991. Intraorbital transection of the rabbitoptic nerve: consequences for ganglion cells and neuroglia in theretina. J Comp Neurol 312:175–192.

Shibuki K, Gomi H, Chen L, Bao S, Kim JJ, Wakatsuki H, Fujisaki T,Fujimoto K, Katoh A, Ikeda T, Chen C, Thompson RF, Itohara S.1996. Deficient cerebellar long-term depression, impaired eyeblinkconditioning, and normal motor coordination in GFAP mutant mice.Neuron 16:587–599.

Tanihara H, Hangai M, Sawaguchi S, Abe H, Kageyama M, Naka-zawa F, Shirasawa E, Honda Y. 1997. Up-regulation of glial fibril-lary acidic protein in the retina of primate eyes with experimentalglaucoma. Arch Ophthalmol 115:752–756.

Vardimon L. 2000. Neuroprotection by glutamine synthetase. Isr MedAssoc J 2(Suppl):46–51.

Vaughan DK, Erickson PA, Fisher SK. 1990. Glial fibrillary acidicprotein (GFAP) immunoreactivity in rabbit retina: effect of fixation.Exp Eye Res 50:385–392.

Wahlin KJ, Campochiaro PA, Zack DJ, Adler R. 2000. Neurotrophicfactors cause activation of intracellular signaling pathways in Mul-ler cells and other cells of the inner retina, but not photoreceptors.Invest Ophthalmol Vis Sci 41:927–936.

Wang X, Tay SS, Ng YK. 2000. An immunohistochemical study ofneuronal and glial cell reactions in retinae of rats with experimen-tal glaucoma. Exp Brain Res 132:476–484.

Watanabe M, Sawai H, Fukuda Y. 1995. Number and dendritic mor-phology of retinal ganglion cells that survived after axotomy inadult cats. J Neurobiol 27:189–203.

Weber AJ, Kaufman PL, Hubbard WC. 1998. Morphology of singleganglion cells in the glaucomatous primate retina. Invest Ophthal-mol Vis Sci 39:2304–2320.

Yoshida A, Ishiguro S, Tamai M. 1993. Expression of glial fibrillaryacidic protein in rabbit Muller cells after lensectomy-vitrectomy.Invest Ophthalmol Vis Sci 34:3154–3160.

Zhou R, Skalli O. 2000. TGF-alpha differentially regulates GFAP,vimentin, and nestin gene expression in U-373 MG glioblastomacells: correlation with cell shape and motility. Exp Cell Res 254:269–278.

125GFAP AND GS EXPRESSION WITH ON CRUSH AND BDNF