Neurobiology of Disease 43 (2011) 455–464

Contents lists available at ScienceDirect

Neurobiology of Disease

j ourna l homepage: www.e lsev ie r.com/ locate /ynbd i

Elevation of p-NR2AS1232 by Cdk5/p35 contributes to retinal ganglion cell apoptosisin a rat experimental glaucoma model

Jie Chen, Yanying Miao, Xiao-Han Wang, Zhongfeng Wang ⁎Institutes of Brain Science, Institute of Neurobiology and State Key Laboratory of Medical Neurobiology, Fudan University, Shanghai 200032, China

Abbreviations: BCA, bicinchoninic acid; Bcl-2, B-cecyclin-dependent kinase 5; Co-IP, co-immunoprecipisystem; CTB, cholera toxin B subunit; DMSO, dimecell layer; Glu, glutamate; HRP, horseradish-peroxidaIP, immunoprecipitate; NMDA, N-methyl-D-aspartate;phosphorylation at S1232 site; ONH, optic nerve heaphosphate-buffered saline; PFA, paraformaldehyde; RGterminal dUTP nick end labeling.⁎ Corresponding author at: Institutes of Brain Science

State Key Laboratory of Medical Neurobiology, Fudan UnShanghai 200032, China. Fax: +86 21 5423 7643.

E-mail address: zfwang@fudan.edu.cn (Z. Wang).Available online on ScienceDirect (www.scienced

0969-9961/$ – see front matter © 2011 Elsevier Inc. Aldoi:10.1016/j.nbd.2011.04.019

a b s t r a c t

a r t i c l e i n f o

Article history:Received 25 October 2010Revised 18 April 2011Accepted 22 April 2011Available online 30 April 2011

Keywords:GlaucomaNMDA receptorCdk5/p35ApoptosisInner retinaRoscovitine

Glaucoma, mainly caused by high intraocular pressure (IOP), is characterized by apoptotic death of retinalganglion cells (RGCs). We investigated the possible involvement of cyclin-dependent kinase 5 (Cdk5) and itsactivator p35, which have been implicated in a variety of neurological disorders, in RGC apoptosis in a ratexperimental glaucoma model reproduced by blocking episcleral veins. Cholera toxin B subunit (CTB)retrogradely labeled RGCs displayed a dramatic reduction in number both in the central and peripheral retinaon day 14 (D14) (Pb0.05 vs control), D21 (Pb0.01 vs control) and D28 (Pb0.001 vs control) after operation.Terminal dUTP nick end labeling (TUNEL)-positive cells were detected on D14 both in the central andperipheral regions, and numerous TUNEL-positive cells were found on D21 and D28 in both the regions(P allb0.001 vs control). As compared with the control eyes, the expression level of Cdk5 was significantlyincreased on D21 (Pb0.001), whereas that of p35 displayed a marked increase on D14 (Pb0.01) and D21(Pb0.001). Meanwhile, both NR2A and p-NR2AS1232 increased from D14 onwards (Pb0.01 to 0.001). Co-immunoprecipitation indicated a direct interaction between Cdk5 and p-NR2AS1232. Intraperitoneal injectionof the Cdk5 inhibitor roscovitine remarkably inhibited RGC apoptosis (Pb0.001 vs vehicle group) andincreased the number of CTB-labeled RGCs (Pb0.05 to 0.01 vs vehicle group) in whole flat-mounted retinas,which was accompanied by a significant decrease in expression levels of p35 and p-NR2AS1232 (P allb0.01 vsvehicle group). Our results suggest that elevation of p-NR2AS1232 by Cdk5/p35 contributes to RGC apoptoticdeath in experimental glaucoma rats, which could be effectively ameliorated by inhibiting Cdk5/p35.

ll lymphoma protein 2; Cdk5,tation; CNS, central nervousthyl sulfoxide; GCL, ganglionse; IOP, intraocular pressure;p-NR2AS1232, NR2A receptor

d; PB, phosphate buffer; PBS,C, retinal ganglion cell; TUNEL,

, Institute of Neurobiology andiversity, 138 Yixueyuan Road,

irect.com).

l rights reserved.

© 2011 Elsevier Inc. All rights reserved.

Introduction

Glaucoma, the second leading cause of blindness, is a neurode-generative disease that is characterized by optic nerve degenerationresulting from apoptotic death of retinal ganglion cells (RGCs) (Guoet al., 2005; Hitchings, 2000; Resnikoff et al., 2004). Although elevatedintraocular pressure (IOP) is commonly regarded as a hallmarkrisk factor (Quigley et al., 1995; Weinreb and Khaw, 2004), thepathogenesis of RGC death following intraocular hypertension isstill poorly understood. The role of the excitotoxicity induced by

glutamate (Glu), a major excitatory neurotransmitter in the retina(Thoreson and Witkovsky, 1999; Yang, 2004), in neurodegenerationin glaucoma models is still a controversial issue (Dreyer et al., 1996;Levkovitch-Verbin et al., 2002; Ullian et al., 2004). Nevertheless, lotsof evidence suggest that the excess of extracellular Glu may be apotential risky factor for retinal malfunction in glaucoma (Guo et al.,2006; Levin, 2003; Salt and Cordeiro, 2006). Indeed, expression ofGlu transporters is significantly reduced in rat glaucoma models,indicating frustration of the effective buffering of extracellular Glu(Martin et al., 2002; Vorwerk et al., 2000). Moreover, prolongedinjection of Glu of low-concentration was shown to induce RGC deathin rat (Nucci et al., 2005). Glu-induced apoptotic death of RGCsis known to be primarily mediated by the N-methyl-D-aspartate(NMDA) subtype receptor (Guo et al., 2006; Lipton, 2001; Seki andLipton, 2008), and the NMDA channel blocker MK-801/memantineindeed prevents RGC death in experimental rat glaucoma models(Calzada et al., 2002; Chaudhary et al., 1998; Guo et al., 2006; Hareet al., 2004; WoldeMussie et al., 2002).

Cyclin-dependent kinase 5 (Cdk5), a proline-directed serine/threonine kinase, has multiple roles in neural development andsynaptic plasticity by phosphorylating numerous synaptic substrates(Dhavan and Tsai, 2001; Ko et al., 2001; Lee et al., 2004; Morabitoet al., 2004; Tomizawa et al., 2002). Cdk5 is expressed ubiquitously,

http://dx.doi.org/10.1016/j.nbd.2011.04.019mailto:zfwang@fudan.edu.cnhttp://dx.doi.org/10.1016/j.nbd.2011.04.019http://www.sciencedirect.com/science/journal/09699961

456 J. Chen et al. / Neurobiology of Disease 43 (2011) 455–464

but shows high activity exclusively in the central nervous system(CNS) due to the restricted distribution of its activators (Tang et al.,1995). P35, an essential activator of Cdk5 (Chae et al., 1997; Ko et al.,2001), restricts the expression of active Cdk5 primarily to post-mitoticneurons (Tang et al., 1995; Tsai et al., 1994). The dysregulation ofCdk5/p35 has been implicated in many neurological disorders (Borghiet al., 2002; Bu et al., 2002; Chen and Wang, 2010; Lee et al., 1999;Nguyen et al., 2003; Smith et al., 2003). There is accumulatingevidence suggesting that Cdk5 may be a “Jekyll and Hyde” kinase inthat it takes different responsibilities under different conditions (Cruzand Tsai, 2004). For instance, activation of Cdk5 induces hippocampalCA1 cell death by directly phosphorylating the NR2A subunit at S1232site (p-NR2AS1232) in an ischemic rat model (Wang et al., 2003). Incontrast, Cdk5 is shown to prevent neuronal apoptosis through ERK-mediated upregulation of B-cell lymphoma protein 2 (Bcl-2) in vitro(Wang et al., 2006) and by directly phosphorylating Bcl-2 at Ser70site in rat cortical neurons (Cheung et al., 2008), respectively.Furthermore, Cdk5 may exert a pro-survival role by negativelyregulating c-Jun N-terminal kinase 3 (Bu et al., 2002) and/or byparticipating in the neuregulin-dependent activation of PI3K and Aktpathway (Li et al., 2003). In the present study we investigate thepossible involvement of Cdk5/p35 in the pathogenesis of RGCs in arat experimental glaucoma model. Our results demonstrate thatactivation of Cdk5/p35 causes an elevation of p-NR2AS1232 that likelycontributes to RGC apoptotic death in this model and this defect issignificantly ameliorated by roscovitine, an inhibitor of Cdk5/p35.

Materials and methods

Animals and glaucoma model

All experimental procedures described here were carried out inaccordance with the National Institutes of Health (NIH) guidelinesfor the Care and Use of Laboratory Animals and the guidelines ofFudan University on the ethical use of animals. During this study allefforts were made to minimize the number of animals used and theirsuffering. Male Sprague–Dawley rats weighing 100–300 g obtainedfrom SLAC Laboratory Animal Co. Ltd (Shanghai, China) were housedon a 12 h light/dark schedule, with standard food and water providedad libitum. In this study a total of 176 rats were used, among which44 were for control (24 for whole flat-mounted experiments, 10for Western blot experiments, 4 for co-immunoprecipitation exper-iments and 6 for vertical slices) and 132 for reproduction of theglaucoma model (70 for whole flat-mounted experiments, 28 forWestern blot experiments and 34 for vertical slices). The detailedanimal numbers used in each of the experiments were given in theResults Section.

The rat glaucomamodel used in the present study was reproducedfollowing the procedure previously described (Naka et al., 2010; Wuet al., 2010; Yu et al., 2006), with minor modifications. Briefly, ratswere anesthetized by intramuscular injection of amixture of ketamine(25 mg/kg) and xylazine (10 mg/kg), and the right eyes were furtheranesthetized locally with a topical application of 0.4% oxybuprocainehydrochloride eyedrop (Benoxil, Santen Pharmaceutical Co. Ltd,Osaka, Japan). After minimum conjunctival incision, three episcleralveins of the right eyes near the superior and temporal rectus muscleswere carefully separated, ligated or cauterized under an OPMI VISU140 microscopy (Carl Zeiss, Jena, Germany). IOPs of both eyes weremeasured using a handheld digital tonometer (Tonopen XL, MentorO&O Inc., MA, USA) under general and local anesthesia as describedabove. The average value of five consecutive measurements with adeviation of less than 5% was accepted. All measurements wereperformed in the morning to avoid possible circadian difference. TheIOPs of both eyesweremeasured before surgery as a baseline (D0), andthen on the next day of the operation (D1), days 3, 7, 11, 14 (D3, D7,D11, D14), and weekly afterwards.

Labeling and quantification of RGCs

Retrograde labeling of RGCs in control and experimental eyes wasperformed as previously described in detail (Chen et al., 2004; Zhaoet al., 2010). Briefly, after rats were deeply anesthetized with 25%urethane (1.25 g/kg), 1% cholera toxin B subunit (CTB) (List BiologicalLaboratories, Campbell, CA, USA) was injected into the superiorcolliculus bilaterally (2 μl each site) (6.0 mm posterior and 2.0 mmlateral to the bregma and 4.5–5 mm deep from the cortical surface).After a survival period of 5 days, RGCs could be retrogradely labeledby CTB and detected by the anti-CTB antibody.

CTB immunohistochemistry experiments were performed followingthe procedure described in detail by Zhang and Diamond (2006).Animals were anesthetized and killed by decapitation. The eyes wereremoved quickly and fixed with 4% paraformaldehyde (PFA) in 0.1 Mphosphate buffer (PB, pH 7.4) for 1 h at 4 °C, and then dehydrated withgraded sucrose solutions at 4 °C (1 h each in 15%, 20%, and overnightin 30%). The eyecups were embedded in OCT compound (Tissue Tek,Torrance, CA) and stored at−20 °C for further use. For retinal sections,the tissues were vertically sectioned at 14 μm thickness on a freezingmicrotome (Leica, Nussloch, Germany). After rinsing with 0.01 M PBS(pH 7.4), the sections were blocked for 1 h in 5% normal donkey serum(v/v) (NDS, Sigma, St. Louis, MO, USA) in PBS plus 0.3% Triton X-100 atroom temperature, and then incubated with the primary antibodypolyclonal goat-anti-CTB (1:4000 dilution) overnight at 4 °C. Bindingsites of the primary antibody were visualized by incubating withRhodamine Red-X (RRX)-conjugated donkey anti-goat IgG (1:400dilution, Jackson ImmunoResearch Laboratories, West Grove, PA,USA). The sections were rinsed and coverslipped with Vectashield(Vector, Burlingame, CA, USA). For whole flat-mounted retina exami-nation, the same procedure was conducted except the incubation timein anti-CTB being 5 days and in RRX-conjugated donkey anti-goatIgG being 1 day. CTB-immunoreactivity and terminal dUTP nick endlabeling (TUNEL, see below) signals were visualized with a confocallaser scanning microscope through a 20× objective (Fluroview 1000,Olympus, Monolith, Tokyo, Japan). Counting of CTB-positive RGCs inretinal sections was made using a fluorescence microscope (BX51,Olympus, Monolith, Tokyo, Japan). Final images were assembled usingAdobe Photoshop version 7.0 (Adobe Systems, San Jose, CA, USA).

Quantification of RGCs was performed on vertical retinal sectionsfollowing the procedure previously reported (Qiu et al., 2010), withsome modifications. For a single preparation, the whole retina wassectioned into 256 slices at 14 μm thickness from the nasal side to thetemporal side across the optic disk. The first 8 slices were sequentiallymounted on 8 different glass slides. Again, the next 8 slices weremounted in the sameway and so on, therebymaking the total 32 sliceson each of the 8 slides to cover different zones of the whole retinafrom the nasal side to the temporal side. We randomly selected oneslide from the 8 slides to count all CTB-positive RGCs in the 32 slicesusing the fluorescent microscope through a 20× objective.

Degeneration assay of cells in GCL

To detect cell apoptosis, TUNEL assay (Gavrieli et al., 1992) wasperformed using the DeadEnd Fluorometric TUNEL System G3250 kit(Promega, Madison, WI, USA) and following the manufacturer'sinstructions. TUNEL signals were visualized with the confocal laserscanning microscope. Images of DAPI-positive cells were takensimultaneously to confirm the co-localization of the TUNEL signalswith the cell nuclei.

In order to get a general profile of the distribution of TUNEL-positivesignals and CTB-positive cells in whole flat-mounted retinas, weselected two representative regions, which are named as central andperipheral ones throughout the text as previously described (Neufeldet al., 1999; Sawada andNeufeld, 1999; Yanget al., 2009). The rat retinasused in the present work ranged in diameter from 10.0 to 12.0 mm.

457J. Chen et al. / Neurobiology of Disease 43 (2011) 455–464

Whereas an area located at less than 1.0 mm (commonly between 200and 830 μm) distant from the optic nerve head (ONH) was defined asthe central region, that located between 3.0 and 3.63 mm distant fromthe ONH as the peripheral region. To quantify the number of CTB- orTUNEL-positive cells, two fields, one from the central region and theother from the peripheral region, were selected at each of four angles ofthe retina (0, 90, 180 and 270°). Therefore, a total of eight fields werechosen fromone retina, and thenumbers of CTB- or TUNEL-positive cellswere counted (Neufeld et al., 1999; Sawada and Neufeld, 1999; Yanget al., 2009). The field area is 0.397 mm2 (630×630 μm). All the imageswere acquired and exported using FV10-ASW Viewer 1.7 software andassembled using Adobe Photoshop.

Western blot

Western blot analysis was conducted as previously described withsomemodifications (Chen et al., 2004; Zhao et al., 2010). Retinas werehomogenized in RIPA lysis buffer supplemented with protease andphosphatase inhibitor cocktail (Roche, Mannheim, Germany). Theconcentration of total proteins was measured using a standardbicinchoninic acid (BCA) assay kit (Pierce Biotechnology, IL, USA).The extracted protein samples (20 μg, 15 μl in volume) were resolvedby 8% SDS-PAGE gel and electroblotted onto PVDF membranes(Immobilon-P, Millipore, Billerica, MA, USA) using Mini-PROTEAN 3Electrophoresis System and Mini Trans-Blot Electrophoretic TransferSysterm (Bio-Rad, Hercules, CA, USA). After blocking in 5% non-fatmilk at room temperature for 1 h, the membranes were incubatedovernight at 4 °C with primary antibodies: monoclonal mouse anti-Cdk5 (Millipore, Billerica, MA, USA), monoclonal rabbit anti-p35/25(C64B10, Cell signaling Technology, Danvers, MA, USA), polyclonalrabbit anti-phospho-NR2A (Ser1232), monoclonal mouse anti-NR2A(Invitrogen, Carlsbad, CA, USA) at a 1:1000 dilution, and monoclonalmouse anti-β-actin (Sigma-Aldrich, St. Louis, MO, USA) at a 1:10,000dilution. After washing in Tris-buffered saline-Tween 20 the mem-branes were incubated with horseradish-peroxidase (HRP)-conjugateddonkey anti-mouse or donkey anti-rabbit secondary antibody (ThermoScientific, Rockford, IL, USA) at a 1:2500 dilution for 1 h at roomtemperature. The blots were then incubated with chemofluorescentreagent ECL (Thermo Scientific, Rockford, IL, USA) and exposed to X-rayfilm in the dark. The experiments were performed in triplicate, andthe protein bands were quantitatively analyzed with NIH Image JAnalysis software.

Co-immunoprecipitation

To test possible correlation between Cdk5 and p-NR2AS1232, co-immunoprecipitation (co-IP) experiments were performed by using aPierce co-IP kit (No: 23600; Pierce Biotechnology, Rockford, IL, USA)following the manufacturer's instructions. Briefly, anti-Cdk5 mousemonoclonal antibody, anti-NR2A mouse monoclonal antibody, or thenormal mouse IgG (negative control) was preliminarily incubatedwith an amine-reactive resin respectively at 4 °C for 4 h. Then, proteinsof 100 μg employed from the retinal sample supernatant (four rats)were incubatedwith the antibody-pretreated resin respectively at 4 °Covernight. Afterwashingwith the elution buffer providedby the Pierceco-IP kit, the immunoprecipitated complexes were recovered in 20 μl2× sample buffer and boiled for 5 min, followed by an analysis ofimmunoblotting probed with anti-phospho-NR2A (Ser1232).

Roscovitine administration

From D1 onwards, the animals received intraperitoneal injectionof roscovitine, a specific Cdk5 inhibitor (Biomol, Plymouth Meeting,PA, USA), once daily at a dose of 2.8 mg/kg till D28. Roscovitine wasdissolved in DMSO (Sigma-Aldrich, St. Louis, MO, USA). Glaucoma rats

receiving injections of equal volume of DMSO (80 μl/100 g) wereserved as control animals.

Statistical analysis

The density of Western blot bands was first normalized to theircorresponding β-actins and then to that of left eyes of control. All dataare presented as mean±S.E.M. throughout the text. Statisticalanalysis was performed by using the Graphpad Prism software(version 5.0; Graphpad Software, San Diego, CA, USA). Two-wayanalysis of variance (ANOVA) (protein levels vs IOP elevation andtime) was used to evaluate the effect of IOP elevation for proteinexpression. Both non-parametric test (Kruskal–Wallis test) and one-way ANOVA with Bonferroni post test were used to evaluatedifferences within the operated groups or within the sham-operatedgroups. One-way ANOVA with Bonferroni post test was applied toevaluate IOP changes. Paired t test was applied to evaluate differencesbetween the operated eye and the contralateral eye in the same timepoint. Student's t test was used to evaluate the effects of roscovitinefor protein expression and for changes in number of CTB-labeled RGCand TUNEL-positive signal.

Results

Steady elevation of IOP and progressive apoptotic death of RGCs in ratglaucoma model

Changes of IOP in operated right eyes and contralateral left eyes ofrats were monitored. As reported in previous studies, using glaucomarats produced by similar methods (Naka et al., 2010; Wu et al., 2010;Yu et al., 2006), the average IOP in the operated eyes showed a sharpincrease to 32.0±0.4 mmHg on D1 after operation [vs 16.9±0.2 mmHg before operation (D0), n=73, P allb0.001 vs D0 and thecontralateral left eyes, one-way ANOVA with Bonferroni post testand paired t test]. The IOP declined on subsequent two days andremained thereafter at a higher level, ranging from 23.6±0.4 mmHgto 26.6±0.3 mmHg, from D3 to D28 (n=73 for D3 and D7; n=56for D11 and D14; n=39 for D21 and n=22 for D28, Pb0.001 vs D0and the contralateral left eyes in all cases, one-way ANOVA withBonferroni post test and paired t test). Even on D42, the IOP of theoperated eyes (22.8±1.0 mmHg) was still significantly higher thanthat of the left eyes (17.0±0.4 mmHg) (n=5, Pb0.01 vs D0; Pb0.001vs the contralateral left eyes, one-way ANOVA with Bonferroni posttest and paired t test) (Supplemental Fig. 1A).

Wefirst assayed the survival of RGCs in our glaucomamodel by CTBretrograde labeling technique in whole flat-mounted retinas. Fig.1shows representative micrographs of central and peripheral regionsof the retinas taken on D0, D7, D14, D21 and D28 after operation,illustrating the changes in number of CTB-labeled RGCs at differenttimes. In all themicrographs CTB-labeled RGCs and axonswere clearlyseen. No obvious changes in number of RGCs were observed in bothcentral and peripheral regions on D7 (Fig. 1B1, B2), compared to thecontrol eyes (Fig. 1A1, A2). The average numbers of CTB-labeled cellswere 885±15/field (n=6) and 805±18/field (n=6) for the centraland peripheral regions respectively (P allN0.05 vs the control values of888±20/field and 810±19/field, respectively. n=6, Kruskal–Wallistest and one-way ANOVA with Bonferroni post test). On D14the number of CTB-positive RGCs was significantly reduced in boththe regions (Fig. 1 C1, C2) and the number reduction became moreconspicuous on D21 and D28 (Fig. 1D1, D2, E1, E2). The averagenumbers of CTB-labeled cells on D14 were 816±17/field (n=6) and736±22/field (n=6) for the central and peripheral regions respec-tively (P allb0.05 vs control, Kruskal–Wallis test and one-way ANOVAwith Bonferroni post test). On D21 andD28 the average numberswerereduced to 758±25/field (n=6, Pb0.01 vs control) and318±39/field(n=6, Pb0.001 vs control, Kruskal–Wallis test and one-way ANOVA

Fig. 1. Decrease in the number of CTB retrogradely labeled retinal ganglion cells (RGCs) in rat glaucoma model. (A1, A2) Images of two representative regions, central and peripheralones, taken from normal (Ctr) whole flat-mounted retina. (B1, C1, D1, E1) Images taken from central retinal regions on D7, D14, D21 and D28 after operation, respectively. (B2, C2,D2, E2) Images taken from peripheral retinal regions at the same time points after operation, respectively. Note that the number of CTB-positive RGCs was progressively reducedfrom D14 onwards. All the images were taken at same magnification. Scale bar, 50 μm.

458 J. Chen et al. / Neurobiology of Disease 43 (2011) 455–464

with Bonferroni post test) respectively for the central region, and to568±31/field (n=6, Pb0.001 vs control) and 464±37/field (n=6,Pb0.001 vs control, Kruskal–Wallis test and one-way ANOVA withBonferroni post test) respectively for the peripheral region.

We further counted the total number of CTB-labeled RGCs in all32 retinal sections on each slide. Whereas the average number ofCTB-labeled RGCs was 9909±116 (n=6) in the unoperated lefteyes, the number of labeled RGCs in the operated right eyes declinedto 9110±144 on D14 (n=6, Pb0.05, Kruskal–Wallis test and one-way ANOVA with Bonferroni post test), and further to 6403±116(n=6, Pb0.01) and 5983±111 (n=6, Pb0.001, Kruskal–Wallis testand one-way ANOVA with Bonferroni post test) on D21 and D28,respectively.

We next applied TUNEL staining technique to detect apoptosis ofRGCs in the glaucoma rats. As illustrated in Fig. 2, no visible TUNEL-positive signals were detected in the flat-mounted glaucoma retinason D7 both in the central (B1–B3) and peripheral regions (B4–B6). OnD14 sparse TUNEL-positive signals were detected in both central(Fig. 2 C1–C3) and peripheral regions (Fig. 2 C4–C6) (Fig. 2 C2 and C5:corresponding DAPI images of C1 and C4; C3 and C6: merged imagesof C1, C2 and C4, C5, respectively). The average numbers of TUNEL-positive cells were 12±3/field (n=6) and 21±6/field (n=6) for thecentral and peripheral regions respectively (P allb0.001 vs control,Kruskal–Wallis test and one-way ANOVA with Bonferroni post test).Numerous TUNEL-positive signals were observed on D21 (Fig. 2D1,D4) and D28 (Fig. 2E1, E4) in both the regions. The average numbersof TUNEL-positive cells were increased to 104±13/field (n=6,Pb0.001 vs control) on D21 and 536±25/field (n=6, Pb0.001 vscontrol, Kruskal–Wallis test and one-wayANOVAwith Bonferroni posttest) on D28 respectively for the central region, and to 160±21/field(n=6, Pb0.001 vs control) on D21 and 528±23/field (n=6, Pb0.001vs control, Kruskal–Wallis test and one-way ANOVA with Bonferronipost test) on D28 respectively for the peripheral region. Enlargedimages in the left corner of Fig. 2E1–E3 extracted from the correspond-ing fields (squares) show that the TUNEL signals are well overlappedwith the DAPI images (Fig. 2E3), demonstrating cell apoptosis. Similarresults were observed in all retinas tested (n=6 for each time point).

Changes in expression of Cdk5, p35 and p-NR2AS1232 in rat glaucomamodel

Since Cdk5/p35 are known to be involved in ischemic hippocampalCA1 cell death (Wang et al., 2003), we tested, using Western blots,

whether the expression of Cdk5 and p35 would change in the ratglaucoma model. Fig. 3A shows representative results obtained byWestern blot analysis, presenting the expression levels of Cdk5 andp35 before and on different days (from D7 to D28) after operation,whereas the bar charts in Fig. 3B and C show a statistical analysis ofthe time-dependent changes in Cdk5 and p35 expression. Two-wayANOVA analysis showed that both IOP elevation and time affected theexpression of Cdk5 and p35 (P allb0.01). On the other hand, Kruskal–Wallis test showed that there was difference between the operatedeyes and the sham-operated eyes for both Cdk5 and p35 (Pb0.05 forCdk5 and Pb0.05 for p35). Furthermore, one-way ANOVA withBonferroni post test was used for the multiple groups' comparison.As compared to that of the right eyes of control (Ctr), the averageCdk5 level in the operated right eyes showed an initial decrease to60.6±9.4% (n=6, Pb0.001; Pb0.01 vs the contralateral left eyes,paired t test) on D7, but then returned to the control level on D14(105.3±17.8%, n=6, PN0.05). On D21 the level significantlyincreased to 161.3±17.9% (n=6, Pb0.001; Pb0.01 vs the contralat-eral left eyes, paired t test) that was followed by a return to the controllevel on D28. The changes in p35 expression were basically similar.That is, it was at higher level of 176.8±23.2% (n=6, Pb0.01) on D14and 224.2±15.4% (n=6, Pb0.001; Pb0.001 vs the contralateral lefteyes, paired t test) on D21, but returned to the control level onD28. Unlike the Cdk5 level, however, the p35 level did not decreaseon D7 and showed a moderate increase on D14 (154.3±28.6%, n=6,PN0.05). It should be noted here that no band corresponding tothe molecular weight of p25, a truncated form of p35, that is cleavedfrom p35 commonly by calpain in many types of neurodegenerativediseases (Lee et al., 2000; Patrick et al., 1999), was observed on allthe days.

Activation of Cdk5 is known to phosphorylate NR2A at S1232site and p-NR2AS1232 is involved in ischemic hippocampal CA1 celldeath (Li et al., 2001; Wang et al., 2003). We therefore examinedthe expression of NR2A and p-NR2AS1232 on different days afteroperation. Similarly, Two-way ANOVA analysis showed that both IOPelevation and time affected the expression of NR2A and p-NR2AS1232

(P allb0.01), and Kruskal–Wallis test showed that there wasdifference between the operated eyes and the sham-operatedeyes for both NR2A and p-NR2AS1232 (Pb0.05 for NR2A and Pb0.05for p-NR2AS1232). One-way ANOVA with Bonferroni post test wasalso used for the multiple groups' comparison. As shown in Fig. 4A–C,the protein level of p-NR2AS1232 in the operated eyes remarkablyincreased to 165.8±17.3% (n=6, Pb0.01; Pb0.01 vs the contralateral

Fig. 2. TUNEL staining detection of RGC apoptotic changes in control and operated eyes. (A1, A4) Microphotographs show representative confocal images of TUNEL signals (green) inthe central and peripheral regions taken from whole flat-mounted retinas in normal (Ctr) eye. (B1, B4) Images of TUNEL signals on D7 after operation; (C1, C4) on D14; (D1, D4) onD21; (E1, E4) on D28. (A2, B2, C2, D2, E2, A5, B5, C5, D5, E5) Counterstained images with DAPI (blue). (A3, B3, C3, D3, E3, A6, B6, C6, D6, E6) Merged images of corresponding TUNELand DAPI images. Note that sparse TUNEL-positive signals were detected in both central and peripheral regions on D14, and the signals became numerous on D21 and D28 afteroperation. Inset images in the left corner of E1–E3 are, respectively, the enlarged images of the corresponding fields (squares), showing that the TUNEL signals are well overlappedwith the DAPI images. All the images were taken at same magnification except the enlarged images in E1–E3. Scale bars, 50 μm.

459J. Chen et al. / Neurobiology of Disease 43 (2011) 455–464

left eyes, paired t test), 174.7±24.3% (n=6, Pb0.001; Pb0.01 vs thecontralateral left eyes, paired t test) and 184.8 21.8% (n=6, Pb0.001;Pb0.01 vs the contralateral left eyes, paired t test) of that in theright eyes of control (Ctr) on D14, D21 and D28, respectively(Fig. 4A, B). Similarly, the average NR2A level in the operatedright eyes tended to increase after operation and reached levels of191.0±13.2% (n=6, Pb0.001; Pb0.01 vs the contralateral left eyes,paired t test), 224.4±17.2% (n=6, Pb0.001) and 224.0±14.2%(n=6, Pb0.01; Pb0.05 vs the contralateral left eyes, paired t test) ofthat in the right eyes of Ctr on D14, D21 and D28 respectively.

Interaction between Cdk5 and p-NR2AS1232 in normal retinalextracts was further examined. No binding was observed between theantibody against p-NR2AS1232 and the immunoprecipitate (IP)derived using the antibody against NR2A. But a band at the locationcorresponding to p-NR2AS1232 (180 KD) was detected in the IPderived using the antibody against Cdk5 (Fig. 4D), suggesting aspecific interaction between Cdk5 and p-NR2AS1232. Similar resultswere observed in other three preparations.

Decrease of RGC apoptotic death by roscovitine

The above results suggest a possibility that activation of Cdk5mightincrease phosphorylation of NR2A, which is involved in apoptoticdeath of RGCs in the glaucoma model. To test this possibility, thepotent inhibitor of Cdk5/p35 roscovitine was administrated intraper-itoneally from D1 till D28. Roscovitine did not change the elevatedIOP in the operated eyes (Supplemental Fig. 1B). For example, onD28 the average IOP of operated eyes in roscovitine-injected rats was26.3±0.7 mmHg (n=15), which was not significantly different fromthe value in DMSO-injected rats (24.6±0.7 mm Hg) (n=15, PN0.05,Student's t test). We evaluated the changes in the number of CTB-positive RGCs after roscovitine injection in the glaucoma model.As shown in Fig. 5, in the DMSO-injected rat (D28+DMSO) CTB-labeled cells in the whole flat-mounted retina obviously decreasedin number on D28 in both central (B1) and peripheral (B2) regions,as compared to the normal retina (A1, A2), just like observed in theglaucoma retina (Fig. 1). The average numbers of CTB-labeled cells

image of Fig.�2

Fig. 3. Expression of Cdk5 and p35 proteins in retinas of control and operated eyes. (A) Representative immunoblots showing the changes of Cdk5 and p35 levels in normal (Ctr) andglaucoma retinal extracts at different time points (D7, D14, D21 and D28) after operation. At each time point, unoperated (left, L) and operated (right, R) eyes were taken from thesame rat. (B) Bar chart showing the average densitometric quantification of immunoreactive bands of Cdk5 on different days after operation. (C) Bar chart showing the averagedensitometoric quantification of immunoreactive bands of p35. All data are first normalized to their corresponding β-actins (ratio) and then to the left eyes of control (Ctr). Averageratios of density in (B) and (C) were taken from six rats at each time point. Note the significant increase in expression of both Cdk5 and p35 on D21. ⁎⁎Pb0.01, ⁎⁎⁎Pb0.001 vsright eyes of Ctr; #Pb0.05 vs left eyes of Ctr (one-way ANOVA with Bonferroni post test); &&Pb0.01, &&&Pb0.001 vs the contralateral left eyes (paired t test). All data are presentedas mean±S.E.M.

460 J. Chen et al. / Neurobiology of Disease 43 (2011) 455–464

were 320±26/field (n=6) and 400±28/field (n=6) for the centraland peripheral regions respectively in the D28+DMSO group(P allb0.001 vs normal values of 816±16/field and 801±19/fieldrespectively, Student's t test). However, CTB-labeled cells on D28 inthe roscovitine-injected rat (D28+Ros) were muchmore numerous inboth the regions (Fig. 5C1, C2) than that observed in the D28+DMSOrat. The average numbers of CTB-labeled cells were significantlyincreased to 552±21/field (n=6) and 656±20/field (n=6) for thecentral and peripheral regions respectively in the D28+Ros group(P allb0.01 vs D28+DMSO; P allb0.05 vs normal, Student's t test).On the other hand, the total number of CTB-labeled RGCs in all the32 retinal slices on each slide derived from the roscovitine-treatedrats on D28 was 9448±177 (n=5) that was significantly higherthan the level (5890±140) in the DMSO-treated rats (n=5) (Pb0.01,Student's t test), but close to the valueof normal rats (9909±116, n=6,PN0.05).

Consistent with the CTB-labeling experiments, numerous TUNEL-positive signals were clearly seen in both central (Fig. 6B1–3)and peripheral retinal regions of the DMSO-injected rat on D28(Fig. 6B4–6), which is similar to that observed in glaucoma rats(Fig. 2), but these signals were much less in the roscovitine-injectedrat (Fig. 6C1–3 for central region and Fig. 6C4–6 for peripheral region).The average numbers of TUNEL-positive cells were 600±26/field(n=5) and 688±24/field (n=5) for the central and peripheralregions respectively in the D28+DMSO group (P allb0.001 vs normal,Student's t test). The average numbers of TUNEL-positive cells weresignificantly decreased to 296±18/field (n=5) and 96±12/field(n=5) for the central and peripheral regions respectively in theD28+Ros group (P allb0.001 vs D28+DMSO group, Student's t test).

To explore possible mechanisms underlying the roscovitine-induced reduction of RGC apoptotic death, we examined theexpression of Cdk5, p35, NR2A and p-NR2AS1232 in retinal extractsof vehicle-injected (DMSO) and roscovitine-treated (Ros) groupson D28. As shown by the representative results in Fig. 7A, thereseemed no significant change in expression of Cdk5 and NR2A in theroscovitine-treated retina, as compared to the DMSO-injected retina,but the expression of both p-NR2AS1232 and p35 clearly decreased.A statistical analysis (Fig. 7B) revealed that no significant changein expression of both Cdk5 and NR2A was found in the Ros group[99.4±16.3% of the DMSO-injected group for Cdk5 (n=4, PN0.05,Student's t test) and 93.8±14.0% for NR2A (n=4, PN0.05, Student'st test)], but the average densities of p35 and p-NR2AS1232 were only56.9±11.7% (n=4, Pb0.01, Student's t test) and 51.9±6.6% (n=4,Pb0.01, Student's t test) of the DMSO-injected group respectively.Accordingly, the ratio p-NR2AS1232/NR2A decreased to 55.4±14.0%of the DMSO-injected group (n=4, Pb0.01, Student's t test) (Fig. 7B).These results suggest that the neuroprotective effect of roscovitinein experimental glaucoma may be, at least in part, mediated bya decrease of the production of p-NR2AS1232 through inhibiting theCdk5/p35.

Discussion

In our rat glaucoma model, RGC apoptosis was first detected onD14 (Fig. 2), which was in parallel with the decrease of RGC number,suggesting that apoptotic death may be a main cause for the RGCnumber decrease (Doh et al., 2010; Ju et al., 2008). It should be notedthat blocking of the episcleral veins may unavoidably lead to retinal

image of Fig.�3

Fig. 4. Expression of NR2A and p-NR2AS1232 proteins in retinas of control and operated eyes. (A) Representative immunoblots showing the changes of p-NR2AS1232 and NR2A levelsin normal (Ctr) and glaucoma retinal extracts at different time points (D7, D14, D21 and D28) after operation. At each time point, unoperated (left, L) and operated (right, R) eyeswere taken from the same rat. (B) Bar chart showing the average densitometric quantification of immunoreactive bands of p-NR2AS1232 on different days after operation. (C) Barchart showing the average densitometoric quantification of immunoreactive bands of NR2A. All data are first normalized to their corresponding β-actins (ratio) and then to the lefteyes of control (Ctr). Average ratios of density in (B) and (C) were taken from six rats at each time point. Note the significant increased in expression of both p-NR2AS1232 and NR2Aon D14, D21 and D28. ⁎⁎Pb0.01, ⁎⁎⁎Pb0.001 vs right eyes of Ctr; ##Pb0.01 vs left eyes of Ctr (one-way ANOVA with Bonferroni post test); &Pb0.05, &&Pb0.01, &&&Pb0.001 vs thecontralateral left eyes (paired t test). All data are presented as mean±S.E.M. (D) Co-immunoprecipitation (co-IP) analysis of Cdk5 and p-NR2AS1232 using normal retinas. A band ofp-NR2AS1232 at the location corresponding to 180 KD was detected in the immunoprecipitates derived using the antibody against Cdk5, whereas no band was observed between theantibody against p-NR2AS1232 and the immunoprecipitates derived using the antibody against NR2A. Two retinal samples, 100 μg proteins for each, from different animals were usedfor the left and right experiments respectively.

461J. Chen et al. / Neurobiology of Disease 43 (2011) 455–464

ischemia, in addition to elevating IOP, which would also result in RGCdeath (Chen et al., 2007; Schmidt et al., 2004), just as it occurred inall other IOP elevated glaucoma models (Naka et al., 2010; Schmidtet al., 2004; Wu et al., 2010). We therefore could not exclude the lossof RGCs directly induced by ischemia in our condition. Nevertheless,it seems to be reasonable to assume that elevated IOP may be amajor factor for RGC apoptosis observed in the present study (Dohet al., 2010).

Lots of evidence demonstrate that Cdk5/p35 play important rolesin both pro-survival and pro-apoptosis in CNS development andneurodegenerative diseases (Borghi et al., 2002; Bu et al., 2002;Dhavan and Tsai, 2001; Lee et al., 1999; Nguyen et al., 2003; Smithet al., 2003). Cdk5/p35 are expressed in rat RGCs and may be involvedin RGC degeneration (Hirooka et al., 1996; Lefevre et al., 2002).Involvement of activated Cdks (most likely Cdk5) was found inaxotomy-induced death of RGCs, and the Cdk5 inhibitor indolinone Awas shown to be highly effective against apoptosis of axotomizedRGCs (Lefevre et al., 2002). Furthermore, there is a work in a ratischemic model showing that Cdk5 activation induces CA1 neuron

death in hippocampus by directly phosphorylating NMDA receptors(Wang et al., 2003). In the present work we demonstrated that theexpression of both p35 and Cdk5 showed a significant increase onD21, and the expression of both p-NR2AS1232 and NR2A increasedmarkedly from D14 onwards, which concurred in parallel withsignificant apoptosis and cell number decrease of RGCs from D14onwards. It is therefore speculated that the apoptotic death of RGCsmay be a consequence of the upregulated phosphorylation of NR2Ainduced by Cdk5/p35.

In our model the increase of p35 expression (on D14) appeared toprecede that of Cdk5 expression (on D21) (Fig. 3). This result suggeststhat the elevated expression of p35 during IOP elevation may be acause leading to the upregulation of Cdk5 expression. Mechanismsunderlying the upregulation of p35 expression by elevated IOPremain to be further explored. A puzzle is that the expression ofboth p-NR2AS1232 and NR2A remained at a higher level and thenumber of CTB-labeled RGCs showed a further decrease on D28,while the expression of both p35 and Cdk5 almost declined to thebasal level (Fig. 3). One of the possible explanations may be that

image of Fig.�4

Fig. 5. Administration of roscovitine (Ros) promotes RGC survival in the glaucomamodel. (A1, A2) Confocal images of CTB-labeled RGCs taken from a whole-flat mountednormal retina. (B1, B2) Images of CTB-labeled RGCs taken from a DMSO-injected retinaon D28 (D28+DMSO). (C1, C2) Images of CTB-labeled RGCs taken from a Ros-injectedretina on D28 (D28+Ros). Note that Ros administration significantly promotedsurvival of RGCs in the glaucoma rat. All the images were taken at same magnification.Scale bar, 50 μm.

462 J. Chen et al. / Neurobiology of Disease 43 (2011) 455–464

the increase of p35 and Cdk5 expressions by elevated IOP might betransient, but activation and phosphorylation of NR2A leading to theincrease in p-NR2AS1232 expression may last for a longer period oftime. Another explanation may be that the sustained loss of RGCswould in turn result in a decrease of expression of p35 and Cdk5 (Luoet al., 2009).

In many CNS neurodegenerative diseases p35 is cleaved intop25, which activates Cdk5 more efficiently and results in deleterious

Fig. 6. Administration of roscovitine (Ros) prevents RGCs from apoptosis in glaucoma rats.(green) in central and peripheral regions taken from whole flat-mounted retinas in normalrats (C1, C4) on D28, respectively. (A2, A5, B2, B5, C2, C5) Counterstained images with DAPI (respectively. Note that numerous TUNEL-positive signals were detected in both central andAll the images were taken at same magnification. Scale bars, 50 μm.

effects on neurons (Lee et al., 2000; Patrick et al., 1999). However, wecould not detect the expression of p25 after operation. Similarly,conversion of p35 to p25 is also not detected in the rat retinafollowing optic nerve transaction (Weishaupt et al., 2003). Probably,such conversion occurs transiently at the early stage of IOP elevationso that the p25 level could be too low to be detected (Oka et al., 2006).Another possibility is that the elevated Cdk5 activity upregulatesphosphorylation of p35, thus suppressing both proteasome-mediateddegradation of p35 and calpain-mediated cleavage of p35 (Hosokawaet al., 2006; Kerokoski et al., 2002; Wei et al., 2005). In this context, itwas indeed reported that Cdk5 could be activated without p25 andactivation of Cdk5/p35 was correlated with cell death under differentpathological conditions (Hayashi et al., 1999; Zhang et al., 1997).Furthermore, we indeed detected p25 protein in Glu-treated culturedneurons using our antibody (unpublished data), thus, we can excludethe possibility that the antibody used in the present work could notrecognize p25 protein.

NR1/NR2A and NR1/NR2B are the main NMDA receptor combina-tions in rat RGCs (Chen and Diamond, 2002; Fletcher et al., 2000; Luoet al., 2009; Shen et al., 2006). Glu-induced apoptotic death of RGCsis primarily mediated by NMDA receptors, and delivery of NMDAchannel blockers has a neuroprotective effect on RGCs in experimen-tal rat glaucoma models (Choi, 1987; Guo et al., 2006; Luo et al., 2009;Novelli et al., 1988). The fact that the broad-spectrum NMDA receptorantagonist MK-801 was more effective than ifenprodil, a NR2B-selective antagonist, in reducing RGCs apoptosis induced by IOPelevation, suggesting the involvement of both NR2A and NR2B (Guoet al., 2006). Consistent with this, we found a significant increase oftotal NR2A in our glaucoma model, suggesting that NR2A-containingreceptors are also responsible for RGC apoptotic death in glaucoma.Moreover, the concurrent increase in p-NR2AS1232 level and Cdk5expression on D21 suggests that p35-induced activation of Cdk5 maydirectly phosphorylate NR2A at S1232 (Li et al., 2001; Wang et al.,

(A1, A4) Microphotographs are representative confocal images showing TUNEL signalseye, and eyes of DMSO-injected (D28+DMSO) (B1, B4) and Ros-injected (D28+Ros)blue). (A3, A6, B3, B6, C3, C6) Merged images of corresponding TUNEL and DAPI images,peripheral retinal regions of the DMSO-injected eye, but not of the Ros-injected one.

image of Fig.�6image of Fig.�5

Fig. 7. Pharmacological interference of Cdk5 activity by roscovitine (Ros) decreasesthe expression of p35 and p-NR2AS1232. (A) Representative immunoblots showingthe changes of p35, Cdk5, NR2A and p-NR2AS1232 in Ros- or vehicle (DMSO)-injectedretinas on D28. (B) Bar chart summarizing the average densitometric quantificationof the immunoreactive bands of p-NR2AS1232, NR2A, Cdk5, p35 and the ratio p-NR2AS1232/NR2A in Ros- and DMSO-injected retinas. The values of p-NR2AS1232, NR2A,Cdk5, p35 and p-NR2AS1232/NR2A in the Ros-injected group were all compared to thosein the DMSO-injected group. n=4 for each group. **Pb0.01. Student's t test. All data arepresented as mean±S.E.M.

463J. Chen et al. / Neurobiology of Disease 43 (2011) 455–464

2003). Co-IP experiments confirmed such direct interaction betweenCdk5 and p-NR2AS1232 (Fig. 4D).

Administration of roscovitine largely prevented RGCs fromapoptosis. This result further strengthens the notion that Cdk5 isinvolved in IOP elevation-induced apoptosis and loss of RGCsobserved. Although the increase in CTB-labeled RGCs followingroscovitine administration does not mean that all of these labeledcells function normally, the present work raises an intriguingpossibility that application of Cdk5 inhibitors, in combination ofagents of reducing IOP, may be beneficial for an effective medicationof glaucoma. It is of interest that there are several studies showingthat roscovitine may also prevent neurons from injury by directlyinhibiting cell apoptosis, microglial-induced inflammation andgliosis (Hilton et al., 2008; Leitch et al., 2009), or by modulating theneurotransmitter release (Tomizawa et al., 2002; Yan et al., 2002). Ofcourse, it should still be cautious to use this strategy to treat glaucomabecause activation of Cdk5 is known to induce a diversity of effectsthat are mediated by different signaling pathways (Bu et al., 2002;Cheung et al., 2008; Li et al., 2003; Wang et al., 2006). For instance,it is recently reported in mouse RGC cultures that Cdk5 contributesto the maintenance of neuronal survival through phosphorylation ofBcl-2 (Cheung et al., 2008).

In conclusion, we demonstrate for the first time that Cdk5/p35-inducedphosphorylation of NR2A subunit of NMDA receptors at S1232

site contributes to RGC apoptotic death in a rat glaucoma model.In addition, systemic administration of roscovitine is effective inpreventing RGCs from apoptotic death through inhibiting the activityof Cdk5/p35 and decreasing the expression of p-NR2AS1232.

Supplementarymaterials related to this article can be found onlineat doi:10.1016/j.nbd.2011.04.019.

Acknowledgments

Wewould like to thank Dr. Xiong-Li Yang for his helpful discussionand critical comments on the manuscript. We also thank Dr. Xiang-Tian Zhou atWenzhouMedical College and Dr. Xiang-Ge He in DapingHospital affiliated to the Third Military Medical University for theirkind help in reproduction of the rat glaucoma model. This work wassupported by grants from the National Program of Basic Researchsponsored by the Ministry of Science and Technology of China(2007CB512205; 2011CB504602), the Natural Science Foundationof China (31070966; 30900427; 30870803; 30930034), PujiangTalent Project of the Shanghai Science and Technology Committee(08PJ14016), and the211Project sponsoredby theMinistryof Educationof China.

References

Borghi, R., et al., 2002. Increase of cdk5 is related to neurofibrillary pathology inprogressive supranuclear palsy. Neurology 58, 589–592.

Bu, B., et al., 2002. Deregulation of cdk5, hyperphosphorylation, and cytoskeletalpathology in the Niemann–Pick type C murine model. J. Neurosci. 22, 6515–6525.

Calzada, J.I., et al., 2002. Glutamate-induced excitotoxicity in retina: neuroprotectionwith receptor antagonist, dextromethorphan, but not with calcium channel blockers.Neurochem. Res. 27, 79–88.

Chae, T., et al., 1997. Mice lacking p35, a neuronal specific activator of Cdk5, displaycortical lamination defects, seizures, and adult lethality. Neuron 18, 29–42.

Chaudhary, P., et al., 1998. MK801-a neuroprotectant in rat hypertensive eyes. BrainRes. 792, 154–158.

Chen, J., Wang, Z., 2010. Roles of cycline-dependent kinase 5 in central nervous systemdevelopment and neurodegenerative diseases. Acta Physiol. Sin. 62, 295–308.

Chen, L., et al., 2004. Inwardly rectifying potassium channels in rat retinal ganglion cells.Eur. J. Neurosci. 20, 956–964.

Chen, S., Diamond, J.S., 2002. Synaptically released glutamate activates extrasynapticNMDA receptors on cells in the ganglion cell layer of rat retina. J. Neurosci. 22,2165–2173.

Chen, Y.N., et al., 2007. Hypoxia-induced retinal ganglion cell death and theneuroprotective effects of beta-adrenergic antagonists. Brain Res. 1148, 28–37.

Cheung, Z.H., et al., 2008. Cyclin-dependent kinase 5 supports neuronal survivalthrough phosphorylation of Bcl-2. J. Neurosci. 28, 4872–4877.

Choi, D.W., 1987. Ionic dependence of glutamate neurotoxicity. J. Neurosci. 7, 369–379.Cruz, J.C., Tsai, L.H., 2004. A Jekyll and Hyde kinase: roles for Cdk5 in brain development

and disease. Curr. Opin. Neurobiol. 14, 390–394.Dhavan, R., Tsai, L.H., 2001. A decade of CDK5. Nat. Rev. Mol. Cell Biol. 2, 749–759.Doh, S.H., et al., 2010. Retinal ganglion cell death induced by endoplasmic reticulum

stress in a chronic glaucoma model. Brain Res. 1308, 158–166.Dreyer, E.B., et al., 1996. Elevated glutamate levels in the vitreous body of humans and

monkeys with glaucoma. Arch. Ophthalmol. 114, 299–305.Fletcher, E.L., et al., 2000. Synaptic localization of NMDA receptor subunits in the rat

retina. J. Comp. Neurol. 420, 98–112.Gavrieli, Y., et al., 1992. Identification of programmed cell death in situ via specific

labeling of nuclear DNA fragmentation. J. Cell Biol. 119, 493–501.Guo, L., et al., 2005. Retinal ganglion cell apoptosis in glaucoma is related to intraocular

pressure and IOP-induced effects on extracellular matrix. Invest. Ophthalmol. Vis.Sci. 46, 175–182.

Guo, L., et al., 2006. Assessment of neuroprotective effects of glutamate modulationon glaucoma-related retinal ganglion cell apoptosis in vivo. Invest. Ophthalmol. Vis.Sci. 47, 626–633.

Hare, W.A., et al., 2004. Efficacy and safety of memantine treatment for reductionof changes associated with experimental glaucoma in monkey, II: Structuralmeasures. Invest. Ophthalmol. Vis. Sci. 45, 2640–2651.

Hayashi, T., et al., 1999. DNA fragmentation precedes aberrant expression of cell cycle-related protein in rat brain after MCA occlusion. Neurol. Res. 21, 695–698.

Hilton, G.D., et al., 2008. Roscovitin reduces neuronal loss, glial activation andneurological deficits after brain trauma. J. Cereb. Blood FlowMetab. 28, 1845–1859.

Hirooka, K., et al., 1996. Developmental alteration of the expression and kinase activityof cyclin-dependent kinase 5 (Cdk5)/p35nck5a in the rat retina. J. Neurochem. 67,2478–2483.

Hitchings, R.A., 2000. Selective ganglion cell death in glaucoma. Br. J. Ophthalmol. 84,678–679.

Hosokawa, T., et al., 2006. Enhanced activation of Ca2+/calmodulin-dependent proteinkinase II upon downregulation of cyclin-dependent kinase 5-p35. J. Neurosci. Res.84, 747–754.

image of Fig.�7http://dx.doi.org/10.1016/j.nbd.2011.04.019

464 J. Chen et al. / Neurobiology of Disease 43 (2011) 455–464

Ju, W.K., et al., 2008. Glutamate receptor activation triggers OPA1 release and inducesapoptotic cell death in ischemic rat retina. Mol. Vis. 14, 2629–2638.

Kerokoski, P., et al., 2002. Influence of phosphorylation of p35, an activator of cyclin-dependent kinase 5 (cdk5), on the proteolysis of p35. Brain Res. Mol. Brain Res. 106,50–56.

Ko, J., et al., 2001. P35 and p39 are essential for cyclin-dependent kinase 5 functionduring neurodevelopment. J. Neurosci. 21, 6758–6771.

Lee, K.Y., et al., 1999. Elevated neuronal Cdc2-like kinase activity in the Alzheimerdisease brain. Neurosci. Res. 34, 21–29.

Lee, M.S., et al., 2000. Neurotoxicity induces cleavage of p35 to p25 by calpain. Nature405, 360–364.

Lee, S.Y., et al., 2004. Regulation of synaptojanin 1 by cyclin-dependent kinase 5 atsynapses. Proc. Natl. Acad. Sci. USA 101, 546–551.

Lefevre, K., et al., 2002. Involvement of cyclin-dependent kinases in axotomy-inducedretinal ganglion cell death. J. Comp. Neurol. 447, 72–81.

Leitch, A.E., et al., 2009. Cyclin-dependent kinase inhibitor drugs as potential novel anti-inflammatory and pro-resolution agents. Br. J. Pharmacol. 158, 1004–1016.

Levin, L.A., 2003. Retinal ganglion cells and neuroprotection for glaucoma. Surv.Ophthalmol. 48 (Suppl 1), S21–S24.

Levkovitch-Verbin, H., et al., 2002. Measurement of amino acid levels in the vitreoushumor of rats after chronic intraocular pressure elevation or optic nerve transection.J. Glaucoma 11, 396–405.

Li, B.S., et al., 2003. Cyclin-dependent kinase-5 is involved in neuregulin-dependentactivation of phosphatidylinositol 3-kinase and Akt activity mediating neuronalsurvival. J. Biol. Chem. 278, 35702–35709.

Li, B.S., et al., 2001. Regulation of NMDA receptors by cyclin-dependent kinase-5. Proc.Natl. Acad. Sci. USA 98, 12742–12747.

Lipton, S.A., 2001. Retinal ganglion cells, glaucoma and neuroprotection. Prog. BrainRes. 131, 712–718.

Luo, X.G., et al., 2009. The selective vulnerability of retinal ganglion cells in rat chronicocular hypertension model at early phase. Cell. Mol. Neurobiol. 29, 1143–1151.

Martin, K.R., et al., 2002. Retinal glutamate transporter changes in experimentalglaucoma and after optic nerve transection in the rat. Invest. Ophthalmol. Vis. Sci.43, 2236–2243.

Morabito, M.A., et al., 2004. Cyclin-dependent kinase 5 phosphorylates the N-terminaldomainof thepostsynaptic densityproteinPSD-95 inneurons. J.Neurosci. 24, 865–876.

Naka, M., et al., 2010. Reduced expression of aquaporin-9 in rat optic nerve headand retina following elevated intraocular pressure. Invest. Ophthalmol. Vis. Sci. 51,4618–4626.

Neufeld, A.H., et al., 1999. Inhibition of nitric-oxide synthase 2 by aminoguanidineprovides neuroprotection of retinal ganglion cells in a rat model of chronic glaucoma.Proc. Natl. Acad. Sci. USA 96, 9944–9948.

Nguyen, M.D., et al., 2003. Cell cycle regulators in the neuronal death pathway ofamyotrophic lateral sclerosis caused bymutant superoxide dismutase 1. J. Neurosci.23, 2131–2140.

Novelli, A., et al., 1988. Glutamate becomes neurotoxic via the N-methyl-D-aspartatereceptor when intracellular energy levels are reduced. Brain Res. 451, 205–212.

Nucci, C., et al., 2005. Neurochemical evidence to implicate elevated glutamate in themechanisms of high intraocular pressure (IOP)-induced retinal ganglion cell deathin rat. Neurotoxicology 26, 935–941.

Oka, T., et al., 2006. Presence of calpain-induced proteolysis in retinal degeneration anddysfunction ina ratmodel of acuteocularhypertension. J. Neurosci. Res. 83, 1342–1351.

Patrick, G.N., et al., 1999. Conversion of p35 to p25 deregulates Cdk5 activity andpromotes neurodegeneration. Nature 402, 615–622.

Qiu, W., et al., 2010. A glycine site-specific NMDA receptor antagonist protects retinaganglion cells from ischemic injury by modulating apoptotic cascades. J. Cell.Physiol. 223, 819–826.

Quigley, H.A., et al., 1995. Retinal ganglion cell death in experimental glaucoma andafter axotomy occurs by apoptosis. Invest. Ophthalmol. Vis. Sci. 36, 774–786.

Resnikoff, S., et al., 2004. Global data on visual impairment in the year 2002. Bull. WorldHealth Organ. 82, 844–851.

Salt, T.E., Cordeiro, M.F., 2006. Glutamate excitotoxicity in glaucoma: throwing the babyout with the bathwater? Eye 20, 730–731 author reply 731–732.

Sawada, A., Neufeld, A.H., 1999. Confirmation of the rat model of chronic, moderatelyelevated intraocular pressure. Exp. Eye Res. 69, 525–531.

Schmidt, K.G., et al., 2004. Ischemia and hypoxia, An attempt to explain the differentrates of retinal ganglion cell death in glaucoma. Ophthalmologe 101, 1071–1075.

Seki, M., Lipton, S.A., 2008. Targeting excitotoxic/free radical signaling pathways fortherapeutic intervention in glaucoma. Prog. Brain Res. 173, 495–510.

Shen, Y., et al., 2006. N-methyl-D-aspartate receptors in the retina. Mol. Neurobiol. 34,163–179.

Smith, P.D., et al., 2003. Cyclin-dependent kinase 5 is a mediator of dopaminergicneuron loss in a mouse model of Parkinson's disease. Proc. Natl. Acad. Sci. USA 100,13650–13655.

Tang, D., et al., 1995. An isoform of the neuronal cyclin-dependent kinase 5 (Cdk5)activator. J. Biol. Chem. 270, 26897–26903.

Thoreson, W.B., Witkovsky, P., 1999. Glutamate receptors and circuits in the vertebrateretina. Prog. Retin. Eye Res. 18, 765–810.

Tomizawa, K., et al., 2002. Cdk5/p35 regulates neurotransmitter release throughphosphorylation and downregulation of P/Q-type voltage-dependent calciumchannel activity. J. Neurosci. 22, 2590–2597.

Tsai, L.H., et al., 1994. P35 is a neural-specific regulatory subunit of cyclin-dependentkinase 5. Nature 371, 419–423.

Ullian, E.M., et al., 2004. Invulnerability of retinal ganglion cells to NMDA excitotoxicity.Mol. Cell. Neurosci. 26, 544–557.

Vorwerk, C.K., et al., 2000. Depression of retinal glutamate transporter function leads toelevated intravitreal glutamate levels and ganglion cell death. Invest. Ophthalmol.Vis. Sci. 41, 3615–3621.

Wang, C.X., et al., 2006. Cyclin-dependent kinase-5 prevents neuronal apoptosisthrough ERK-mediated upregulation of Bcl-2. Cell Death Differ. 13, 1203–1212.

Wang, J., et al., 2003. Cdk5 activation induces hippocampal CA1 cell death by directlyphosphorylating NMDA receptors. Nat. Neurosci. 6, 1039–1047.

Wei, F.Y., et al., 2005. Control of cyclin-dependent kinase 5 (Cdk5) activity byglutamatergic regulation of p35 stability. J. Neurochem. 93, 502–512.

Weinreb, R.N., Khaw, P.T., 2004. Primary open-angle glaucoma. Lancet 363, 1711–1720.Weishaupt, J.H., et al., 2003. Inhibition of CDK5 is protective in necrotic and apoptotic

paradigms of neuronal cell death and prevents mitochondrial dysfunction. Mol.Cell. Neurosci. 24, 489–502.

Woldemussie, E., et al., 2002. Neuroprotective effect of memantine in different retinalinjury models in rats. J. Glaucoma 11, 474–480.

Wu, J., et al., 2010. Neuroprotective effect of upregulated sonic Hedgehog in retinalganglion cells following chronic ocular hypertension. Invest. Ophthalmol. Vis. Sci. 51,2986–2992.

Yan, Z., et al., 2002. Roscovitine: a novel regulator of P/Q-type calcium channels andtransmitter release in central neurons. J. Physiol. Lond 540, 761–770.

Yang, H., et al., 2009. Neuroprotective effects of angiotensin II type 1 receptor blocker ina rat model of chronic glaucoma. Invest. Ophthalmol. Vis. Sci. 50, 5800–5804.

Yang, X.L., 2004. Characterization of receptors for glutamate and GABA in retinalneurons. Prog. Neurobiol. 73, 127–150.

Yu, S., et al., 2006. A rat model of glaucoma induced by episcleral vein ligation. Exp. EyeRes. 83, 758–770.

Zhang, J., Diamond, J.S., 2006. Distinct perisynaptic and synaptic localization ofNMDA and AMPA receptors on ganglion cells in rat retina. J. Comp. Neurol. 498,810–820.

Zhang, Q., et al., 1997. Cyclin-dependent kinase 5 is associated with apoptotic cell deathduring development and tissue remodeling. Dev. Biol. 183, 222–233.

Zhao, W.J., et al., 2010. Melatonin potentiates glycine currents through a PLC/PKCsignaling pathway in rat retinal ganglion cells. J. Physiol. Lond 588, 2605–2619.

Elevation of p-NR2AS1232 by Cdk5/p35 contributes to retinal ganglion cell apoptosis in a rat experimental glaucoma modelIntroductionMaterials and methodsAnimals and glaucoma modelLabeling and quantification of RGCsDegeneration assay of cells in GCLWestern blotCo-immunoprecipitationRoscovitine administrationStatistical analysis

ResultsSteady elevation of IOP and progressive apoptotic death of RGCs in rat glaucoma modelChanges in expression of Cdk5, p35 and p-NR2AS1232 in rat glaucoma modelDecrease of RGC apoptotic death by roscovitine

DiscussionAcknowledgmentsReferences