Cyclin-Dependent Kinase Inhibitor 2b Mediates Excitotoxicity-Induced Death of Retinal Ganglion Cells

Abstract

Purpose: Glutamate excitotoxicity seems to contribute to retinal ganglion cell (RGC) death in various eye diseases, but the underlying molecular mechanisms are not fully understood. We studied the roles of cyclin-dependent kinase inhibitors Cdkn2a and Cdkn2b, known as cellular stress-related senescence markers, in N-methyl-d-aspartate (NMDA)-induced RGC death.

Methods: Gene expression was analyzed using quantitative reverse transcription (qRT)-PCR, in situ hybridization, and immunochemistry. Cdkn2a and Cdkn2b gain- and loss-of-function experiments were performed using the adeno-associated virus type 2 (AAV2)-mediated gene delivery system. AAV2-CRISPR-Cas9–mediated knockout of Cdkn2a or Cdkn2b was validated using cultured cells by T7 endonuclease I assay and Western blot analysis. The effects of altered expression of Cdkn2a and Cdkn2b on NMDA-induced RGC death were evaluated by quantification of RNA binding protein with multiple splicing (Rbpms)-immunoreactive RGCs.

Results: Intravitreal NMDA injection resulted in upregulation of Cdkn2a and Cdkn2b expression in RGCs of the mouse retina. AAV2-mediated overexpression of Cdkn2b led to increased expression of Cdkn2a in RGCs, but not vice versa. Overexpression of Cdkn2b, but not Cdkn2a, resulted in a further reduction in RGC viability in NMDA-injected retinas. However, excessive levels of Cdkn2a or Cdkn2b had no effect on RGC viability in healthy mice. AAV2-CRISPR-Cas9–mediated knockout of either Cdkn2a or Cdkn2b attenuated NMDA-induced RGC death.

Conclusions: Cdkn2a and Cdkn2b have pivotal roles in the regulation of excitotoxic RGC degeneration under NMDA-induced pathologic conditions. Our findings imply that Cdkn2a and Cdkn2b are novel therapeutic targets for ocular diseases displaying excitotoxicity-induced neuronal degeneration.

Retinal ganglion cell (RGC) death is an irreversible symptom observed in eye diseases, including ischemia-reperfusion injury, diabetic retinopathy, and glaucoma, leading to visual impairment and blindness. According to animal experimental models, neuronal degeneration in these diseases is attributable at least partly to glutamate receptor-mediated excitotoxic neuronal cell death.^1–8 Ischemia-reperfusion injury leads to RGC loss,^9–11 which is attenuated by treatment with glutamate-receptor antagonists.^12–14 Excitatory neurotransmitter-induced toxicity is a pathomechanism of RGC degeneration in the hyper- and normal-tension glaucoma. Administration of the N-methyl-d-aspartate (NMDA) receptor antagonist memantine prevents RGC loss in animal models of glaucoma.^4,15,16 Therefore, molecular events leading to glutamate-induced RGC death must be clarified to prevent RGC loss in eye diseases associated with neuronal excitotoxicity. 

Cyclin-dependent kinase inhibitors Cdkn2a and Cdkn2b (also known as p16^INK4a and p15^INK4b, respectively) inhibit the activity of cyclin-dependent kinases (Cdks) 4 and 6,^17–19 which cooperate with D-type cyclins to activate retinoblastoma family proteins, and consequently regulate cell cycle progression.^20,21 Primary cells derived from normal tissues undergo irreversible cell cycle arrest, known as senescence, after a limited number of cell divisions. Overexpression of Cdkn2a or Cdkn2b results in senescence transition in established cell lines,²² and Cdkn2a expression is upregulated in aged cultured cells and tissues.^23–25 The effect of altered Cdkn2a expression on RGC survival has been investigated using animal models of glaucoma. In mice with ocular hypertension, retinal Cdkn2a expression was upregulated and senescence-associated beta-galactosidase (SA-ß-gal)–positive cell number increased in the retinal ganglion cell layer (GCL).²⁶ High IOP resulted in reduced RGC numbers; however, Cdkn2a deficiency rescued the RGC loss.^26,27 The human CDKN2B-AS gene,located within the CDKN2B-CDKN2A gene cluster, encodes a noncoding RNA that regulates the expression of neighboring genes by interacting with polycomb proteins.²⁸ Single-nucleotide polymorphisms in the CDKN2B-AS locus are associated with human glaucoma.^29–31 Partial deletion of the CDKN2B-AS locus in mice led to increased Cdkn2a mRNA expression in the optic nerve.³² Furthermore, CDKN2B-AS mutant mice displayed increased RGC vulnerability to microbead-mediated ocular hypertension, leading to fewer surviving RGCs.³² These findings implicated that Cdkn2a is a key regulator of RGC death in pathologic conditions; however, little is known about the role of Cdkn2b in RGC death induction. 

Glutamate excitotoxicity seems to contribute to RGC death in various eye diseases.^1–8 However, the underlying molecular mechanisms are not fully understood. We elucidated the roles of Cdkn2b and Cdkn2a in NMDA receptor-related RGC death using an animal experimental model established by intravitreal NMDA injection.^33,34 

Materials and Methods

Animals

Eight- to 12-week-old male C57BL/6J mice were obtained from SLC (Shizuoka, Japan) or CLEA (Tokyo, Japan) and were maintained at animal facilities in Tohoku University Graduate School of Medicine under a 12-hour light/dark cycle. All animal experiments and animal care were conducted according to procedures and guidelines approved by the committee on animal research at Tohoku University, and the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. 

NMDA Administration

Mice were anesthetized by intraperitoneal injection of sodium pentobarbital dissolved in normal saline (78 mg/kg body weight), and then administered with 10 nmol NMDA (Sigma-Aldrich Corp., St. Louis, MO, USA) dissolved in PBS (2 μL 5 mM NMDA solution per eye) intravitreally using a microsyringe with a sharp 32-gauge needle (Ito, Shizuoka, Japan). 

Quantitative Reverse-Transcription (qRT)-PCR

Total RNA was extracted from mouse retinas using the miRNeasy Mini Kit (Qiagen, Hilden, Germany) and reverse-transcribed into cDNA using the SuperScript III First-Strand Synthesis System (Thermo Fisher Scientific, Waltham, MA, USA). qRT-PCRs were run in a 7500 Fast Real-Time PCR System (Thermo Fisher Scientific) using TaqMan Fast Universal PCR Master Mix (Thermo Fisher Scientific) and a mixture of predesigned TaqMan primers and probes (Thermo Fisher Scientific or Integrated DNA Technologies; Supplementary Table S1). 

AAV Plasmid Construction

Cdkn2a and Cdkn2b expression plasmids were constructed by ligating hemagglutinin (HA) tagged-DNA fragments containing the entire coding region of mouse Cdkn2a (NM_001040654: 82–588) or Cdkn2b (NM_007670: 236–628), amplified from cDNA prepared from adult mouse retina using KOD DNA polymerase (Toyobo, Osaka, Japan), into a mammalian expression vector (Cell Biolabs, San Diego, CA) containing the adeno-associated virus (AAV) inverted terminal repeat, cytomegalovirus (CMV) promoter, internal ribosome entry site (IRES), and green fluorescent protein gene (gfp; pAAV-CMV-Cdkn2a-IRES-GFP or pAAV-CMV-Cdkn2b-IRES-GFP). For clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated protein 9 (Cas9)-mediated loss-of-function assays of Cdkn2a and Cdkn2b, single-guide (sg)RNAs were designed at the 208th and 353rd nucleotide of mouse Cdkn2a (NM_001040654) or Cdkn2b (NM_007670) in the reverse strand, using an online tool (CRISPOR; available in the public domain at http://crispor.tefor.net/), and ligated into the BsaI restriction recognition site of pX601-AAV-CMV::NLS-SaCas9-NLS-3xHA-bGHpA;U6::BsaI-sgRNA (pX601; a gift from Feng Zhang; Addgene plasmid #61591).³⁵ Sequences of the constructed plasmids were confirmed by Sanger sequencing. Information on the oligos used for plasmid construction is shown in Supplementary Table S2. 

AAV2 Production and Administration

Production of AAV2 for intravitreal injection and determination of AAV2 vector genome titers were conducted as reported previously.³⁶ To infect RGCs with exogenous protein-encoding AAV2, ten billion genome copies of AAV2 were injected with 2.0 μL Ca²⁺ and Mg²⁺-free PBS (Nacalai Tesque, Kyoto, Japan) into the vitreous humor of mice anesthetized with a mixture of ketamine (180 mg/kg) and xylazine (90 mg/kg). 

In Situ Hybridization (ISH)

Dissected eyes were fixed with 4% paraformaldehyde dissolved in PBS (4% PFA/PBS) at 4°C overnight, incubated in 20% sucrose in PBS at 4°C overnight, and embedded in O.C.T. compound (Sakura Finetek Japan, Tokyo, Japan). The eyes were sliced at 10-μm thickness using a cryostat (Leica, Wetzlar, Germany). Retinal sections containing the optic nerve head were collected and used in following experiments. To prepare complementary RNA probes, mouse Cdkn2a (NM_001040654.1: bases 82-585) and Cdkn2b (NM_007670.4: bases 236-625) were amplified by PCR using primers containing the T3 or T7 promoter sequence (Supplementary Table S2), and pAAV-CMV-Cdkn2a-IRES-GFP or pAAV-CMV-Cdkn2b-IRES-GFP as templates. Amplicons were purified using the NucleoSpin Gel and PCR Clean-up Kit (Macherey-Nagel, Düren, Germany). Purified DNA fragments were used to synthesize digoxigenin-labeled antisense probes for mouse Cdkn2a and Cdkn2b. ISH was conducted as reported.³⁷ 

Immunohistochemistry

Dissected retinas were fixed with 4% PFA/PBS at 4°C overnight. After washing, the retinas were incubated in radioimmunoprecipitation assay (RIPA) buffer for 4 hours, and then incubated with 10% normal donkey serum in PBS containing 0.1% Tween-20 (PBST) for 1 hour at room temperature to block nonspecific antibody binding. The specimens were incubated with primary antibodies at 4°C under shaking for 3 days. After extensive washing, the specimens were incubated with appropriate fluorophore-conjugated secondary antibodies. Retinal sections were prepared as described above. After blocking with 10% normal donkey serum in PBST, retinal sections were treated with primary antibodies at 4°C overnight, and then treated with secondary antibodies at room temperature for 2 hours. Antibodies used for immunohistochemistry are listed in Supplementary Table S3. 

Cell Culture

NIH3T3-L1 cells (JCRB, Osaka, Japan) were maintained in Dulbecco's modified Eagle's medium (DMEM; Thermo Fisher Scientific) supplemented with 10% fetal bovine serum (Cosmo Bio), and were subcultured upon reaching 80% confluency. 

Plasmid Transfection and Sorting of Transfectants

NIH3T3-L1 cells were seeded into 6-well cell culture plates, incubated for 1 day, and then transfected with the CRISPR-Cas9 plasmids using Avalanche-Everyday Transfection Reagent (EZ Biosystems, College Park, MD) according to the manufacturers' instructions. On the third day after transfection, cells were harvested in 0.25% trypsin-EDTA solution (Thermo Fisher Scientific) and sorted by fluorescence-activated cell sorting (FACS) on a FACSAria III (BD Biosciences, Franklin Lakes, NJ, USA) to obtain GFP-expressing transfectants. Separated cells were used in T7 endonuclease I (T7E1) assays and Western blot analysis. 

T7E1 Assay

Genomic DNA was purified from FACS-sorted bulk transfectants using the QIAamp DNA Mini Kit (Qiagen) according to the manufacturer's instructions. Genomic regions containing CRISPR-Cas9 targets were amplified using KOD DNA polymerase (Toyobo) with Cdkn2a- and Cdkn2b-specific primers (Supplementary Table S2). After purification with the NucleoSpin Gel and PCR Clean-up Kit (Macherey-Nagel, Düren, Germany), PCR products were denatured at 95°C for 5 minutes, and then reannealed by decreasing the temperature gradually from 95°C to 60°C. Reannealed DNA fragments were treated with T7E1 (New England Biolabs, Ipswich, MA) at 37°C for 30 minutes, and then electrophoresed on 1.5% agarose gels. 

Western Blotting

FACS-sorted single transfectants were harvested after expansion and sonicated using a Bioruptor UCD-300 (Cosmo Bio, Tokyo, Japan) in RIPA buffer containing protease inhibitor cocktail (Cayman Chemical, Ann Arbor, MI, USA). Sonicated cells were centrifuged to obtain lysates, and protein concentrations were determined using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific). Total protein (30–50 μg) was analyzed by Western blotting according to routine protocols. Antibodies used for Western blotting are listed in Supplementary Table S3. 

Image Acquisition

Images were captured using a BZ-9000 fluorescence microscope (Keyence, Osaka, Japan). Contrast and brightness adjustment and photo trimming were performed in Adobe Photoshop Elements (Adobe Systems, San Jose, CA, USA). 

Quantification of RNA Binding Protein With Multiple Splicing (Rbpms)-Immunopositive RGCs

Images were captured with a ×20 objective lens at different microscopic fields near the optic nerve head of each flat-mounted retina immunostained for Rbpms. Rbpms⁺ cells in each image were counted using ImageJ software (National Institutes of Health, Bethesda, MD, USA) and expressed as the number of RGCs per mm². The average was calculated from images obtained from the same retina, and for five to eight retinas per experimental group. 

Statistics

Quantitative data were analyzed using Student's t-test for two experimental groups or ANOVA followed by Tukey-Kramer post hoc tests for more than two experimental groups, using JMP Pro 14 software (SAS Institute, Cary, NC, USA). P < 0.05 was considered significant. 

Results

Intravitreal NMDA Injection Alters Cdkn2a and Cdkn2b Gene Expression in RGCs

We first investigated the effects of intravitreal NMDA injection (10 nmol per eye) on the expression of Cdkn2a and Cdkn2b, as well as the RGC marker Rbpms, in the mouse retina by qRT-PCR. Retinas were collected at 10, 20, and 30 hours after injection. Rbpms mRNA expression was unchanged at 10 hours, but significantly decreased at 20 and 30 hours, suggesting that RGCs underwent NMDA-induced cell death 20 hours after injection (Fig. 1A). Retinal Cdkn2b mRNA was upregulated at 20 hours (Fig. 1A). Thus, the timing of Cdkn2b mRNA upregulation coincided with the reduction in Rbpms mRNA. Cdkn2b mRNA expression further increased at 30 hours (Fig. 1A). The expression level of Cdkn2a tended to increase at 20 hours; however, this increase was not significant (Fig. 1A). Furthermore, NMDA injection had no effect on the expression Six6, known as a Cdkn2a upstream gene (Fig. 1A).²⁶ Cdkn2a and Cdkn2b mRNA expression also was analyzed using ISH. Weak ISH signals for Cdkn2a were detected in Rbpms⁺ RGCs, but not other cells, at 10 and 20 hours (Fig. 1B). Cdkn2b signals were detected in RGCs at 10 hours, and the signal intensity was substantially increased at 20 hours. Cdkn2b expression also was detected in the inner nuclear layer (INL) cells at 20 hours (Fig. 1B). No obvious signals for Cdkn2a and Cdkn2b were detected in any layer of PBS-injected control retinas (Fig. 1B). 

Figure 1

 

Gene expression changes in mouse retinas after intravitreal NMDA injection. (A) mRNA expression profiles of Cdkn2a, Cdkn2b, Six6, and Rbpms determined by qRT-PCR. Gene expression levels in NMDA-injected retinas are shown as a percentage of the control. Error bars: SEM. **P < 0.01 vs. PBS (Student's t-test; n = 5–8). (B) ISH analysis of Cdkn2a and Cdkn2b, and double-staining with Rbpms immunohistochemistry. Localization of Cdkn2a and Cdkn2b mRNAs was examined on retinal sections prepared from eyes untreated (control) and treated with NMDA for 10 and 20 hours (NMDA10h and NMDA20h, upper column). The GCL is focused on in the lower column. Cdkn2a and Cdkn2b mRNAs can be detected in Rbpms⁺ RGCs. Arrowheads indicate RGCs coexpressing Rbpms and Cdkn2a or Cdkn2b. Arrows indicate INL cells expressing Cdkn2b. Scale bars: 50 μm.

Overexpression of Cdkn2a or Cdkn2b Does Not Affect RGC Survival in Healthy Mice

We next performed Cdkn2a and Cdkn2b gain-of-function experiments to investigate their role in RGC death induction using AAV2-mediated gene transfer. The AAV2 vectors contained CMV promoter-driven expression cassettes for mouse Cdkn2a or Cdkn2b tagged with the HA epitope, followed by IRES and GFP sequences (AAV2-Cdkn2a-IRES-GFP or AAV2-Cdkn2b-IRES-GFP; Fig. 2A). AAV2-IRES-GFP vector was used as a negative control. qRT-PCR analysis confirmed that the expression of both genes was drastically increased on the 32nd day after infection compared to the levels in control retinas (Fig. 2B). Unexpectedly, overexpression of Cdkn2b also led to increased Cdkn2a expression, but not vice versa (Fig. 2B). We next performed ISH and immunohistochemical analysis on retinal sections prepared from mice on the 32nd day following intravitreal injection of AAV2-Cdkn2a-IRES-GFP or AAV2-Cdkn2b-IRES-GFP, to confirm exogenous Cdkn2a and Cdkn2b expression. ISH detected abundant Cdkn2a and Cdkn2b mRNAs in Rbpms⁺ RGCs and INL cells (Fig. 2C). Cells coexpressing GFP and Cdkn2a or Cdkn2b were detectable in the GCL of retinas injected with AAV2-Cdkn2a-IRES-GFP or AAV2-Cdkn2b-IRES-GFP, but not AAV2-IRES-GFP (Fig. 2D). This indicated that protein expression levels of endogenous Cdkn2a and Cdkn2b probably are less than the limit of detection in the normal retina, consistent with ISH data. Furthermore, double-staining of Cdkn2b immunohistochemistry and Cdkn2a ISH indicated that AAV2-mediated Cdkn2b overexpression resulted in upregulated Cdkn2a expression in the GCL (Fig. 2E), as suggested by qRT-PCR analysis. We next quantified Rbpms-immunoreactive RGCs in flat-mounted retinas prepared from mice on days 16, 24, and 32 following AAV2 infection. No significant differences were observed in the number of Rbpms⁺ RGCs among AAV2-Cdkn2a-IRES-GFP-, AAV2-Cdkn2b-IRES-GFP-, and AAV2-IRES-GFP-infected retinas at all time points investigated (Figs. 2F, 2G). 

Figure 2

 

Effects of AAV2-mediated overexpression of Cdkn2a or Cdkn2b on RGC survival in healthy mice. (A) Schematic representation of the AAV2 vectors encoding Cdkn2a or Cdkn2b and GFP (AAV2-Cdkn2a-IRES-GFP or AAV2-Cdkn2b-IRES-GFP) used for gain-of-function experiments. (B) Upregulated expression of Cdkn2a and Cdkn2b in the retina on day 32 after AAV2 infection determined by qRT-PCR. The number above each bar indicates the actual relative fold change in the expression level compared to the control. Note that infection of the AAV2 vectors expressing Cdkn2b significantly increased the expression levels of Cdkn2b and Cdkn2a. (C) ISH analysis of Cdkn2a and Cdkn2b, and double-staining with Rbpms immunohistochemistry. Localization of Cdkn2a and Cdkn2b mRNAs was examined on retinal sections prepared from eyes infected with AAV2-Cdkn2a-IRES-GFP or AAV2-Cdkn2b-IRES-GFP (upper column). The GCL is focused on in the lower column. Arrowheads and arrows indicate Cdkn2a- or Cdkn2b-expressing Rbpms⁺ RGCs and INL cells, respectively. (D) Expression of Cdkn2a and Cdkn2b proteins on day 32 after AAV2 infection according to immunohistochemistry. Protein localization of exogenous or endogenous Cdkn2a (left) and Cdkn2b proteins (right) is shown in the GCL of retinas infected with AAV2-Cdkn2a-IRES-GFP or AAV2-Cdkn2b-IRES-GFP, and AAV2-IRES-GFP. Arrowheads indicate AAV2-infected cells coexpressing exogenous Cdkn2a or Cdkn2b and GFP. Note that the expression level of endogenous Cdkn2a and Cdkn2b is below the detection limit. (E) Double-staining of Cdkn2b immunohistochemistry and Cdkn2a ISH. Cdkn2a mRNA localization was examined in the GCL of the retina on day 32 after AAV2-Cdkn2b-IRES-GFP infection. Cdkn2b-overexpressing GCL cells co-express Cdkn2a (arrowheads). (F) Immunostaining of the RGC marker Rbpms and GFP in the retina on days 16, 24, and 32 after AAV2 infection. (G) Quantification of Rbpms⁺ RGCs in AAV2-injected retinas. Error bars: SEM. n.s., not significant. **P < 0.01 versus control (Tukey-Kramer test; n = 6 in [B] and 5–9 in [G]). Scale bar: 50 μm.

Overexpression of Cdkn2b, But Not Cdkn2a, Increases the Susceptibility of RGCs to NMDA-Induced Death

We explored the possibility that Cdkn2a and Cdkn2b are implicated in RGC death under the pathologic condition induced by intravitreal NMDA injection. To this end, we first monitored the number of Rbpms⁺ RGCs following NMDA injection. The number of surviving Rbpms⁺ RGCs decreased significantly at the first time point; that is, 5 hours after NMDA injection, and further decreased at 10 and 15 hours after injection (Fig. 3A). We then examined the effects of AAV2-mediated overexpression of Cdkn2a or Cdkn2b on the survival of RGCs in the retinas 10 hours after NMDA injection using Rbpms immunohistochemistry. Fewer RGCs survived in the retinas infected with AAV2-Cdkn2b-IRES-GFP, but not AAV2-Cdkn2a-IRES-GFP, than in the retinas infected with AAV2-IRES-GFP (Figs. 3B, 3C). Taken together, these results indicated that Cdkn2b increases the susceptibility of RGCs to NMDA-induced death. 

Figure 3

 

Effect of AAV2-mediated overexpression of Cdkn2a or Cdkn2b on RGC survival in NMDA-injected mice. (A) Time-course changes in the number of Rbpms⁺ RGCs following intravitreal injection of NMDA (10 nmol). (B) Immunostaining for the RGC marker Rbpms and GFP in flat-mounted retinas infected with AAV2-Cdkn2a-IRES-GFP or AAV2-Cdkn2b-IRES-GFP. (C) Quantification of Rbpms⁺ RGCs in AAV2-infected retinas after NMDA injection. A decrease in the number of Rbpms⁺ RGCs was observed upon overexpression of Cdkn2b, but not Cdkn2a. Error bars: SEM. *P < 0.05, **P < 0.01 versus control (Student's t-test and Tukey-Kramer test; n = 5 and 5–6 in [A] and [C], respectively). Scale bar: 50 μm.

In Vitro Validation of CRISPR-Cas9-Mediated Cdkn2a and Cdkn2b Knockout

To perform loss-of-function experiments of Cdkn2a and Cdkn2b, the CRISPR-Cas9 system consisting of Staphylococcus aureus-derived Cas9 and a sgRNA was used in this study (Fig. 4A). We first verified the effectiveness and accuracy of genome editing by the sgRNAs designed in the first exons of Cdkn2a (sgCdkn2a) and Cdkn2b (sgCdkn2b), and containing the translation start site, using an established mouse cell line NIH3T3-L1 (Fig. 4B). A T7E1 assay indicated successful editing of the genome of NIH3T3-L1 cells transfected with CRISPR-Cas9 plasmids encoding sgCdkn2a (CRISPR-Cas9-sgCdkn2a) or sgCdkn2b (CRISPR-Cas9-sgCdkn2b, Fig. 4C). We further investigated on- and off-target sequences of sgCdkn2a and sgCdkn2b in the genomes extracted from single-cell clones of the CRISPR-Cas9 transfectants. On-target genome editing was verified in three of 21 and 26 clones of CRISPR-Cas9-sgCdkn2a and CRISPR-Cas9-sgCdkn2b transfectants, respectively. The on-target editing effect was a deletion of several nucleotides leading to a frame shift mutation (Fig. 4D). Western blot analysis confirmed that no intact Cdkn2a and Cdkn2b proteins were detectable in each of three clones of CRISPR-Cas9-sgCdkn2a and CRISPR-cas9-sgCdkn2b transfectants (Fig. 4E). Off-target sites for sgCdkn2a and sgCdkn2b, as predicted in silico (CRISPOR; available in the public domain at http://crispor.tefor.net/; Supplementary Table S4), had no detectable modification in these clones. 

Figure 4

 

In vitro validation of CRISPR-Cas9-mediated knockout of Cdkn2a and Cdkn2b. (A) Schematic representation of the AAV-CRISPR-Cas9 construct. (B) The location of CRISPR-Cas9 targets. The entire genome structures of the Cdkn2a and Cdkn2b genes are shown in (a) and (b), respectively. The first exons of Cdkn2a and Cdkn2b in (a) and (b) are magnified in (a') and (b'), respectively. Black triangles in (a') and (b') indicate the locations of CRISPR-Cas9 on-targets on exon 1 of each gene (see Materials and Methods for details). White and gray boxes indicate untranslated and coding regions, respectively. (C) Genome editing effects of CRISPR-Cas9-sgCdkn2a (sg2a), CRISPR-Cas9-sgCdkn2b (sg2b), and CRISPR-Cas9 control (Cont) in NIH3T3-L1 cells. A T7E1 assay was performed on FACS-sorted bulk transfectants of CRISPR-Cas9 plasmids. Asterisks indicate T7E1-mediated digestion of PCR amplicons. M: 100-bp DNA ladder. (D) Modification of the target sequences of CRISPR-Cas9-sgCdkn2a (upper) and CRISPR-Cas9-sgCdkn2b (lower) in each of three FACS-sorted clones obtained from plasmid transfectants. (E) Immunochemical detection of endogenous Cdkn2a (upper) or Cdkn2b (lower), with β-actin as an internal control in the same clones as described in (D). UTR, untranslated region; PC, parental cell.

Loss of Cdkn2a or Cdkn2b Attenuates NMDA-Induced RGC Death

Intravitreal injection of AAV2 vectors encoding CRISPR-Cas9 and sgRNA into the mouse retina led to successful expression of Cas9 proteins in RGCs in our previous study.³⁶ In our study, AAV2 vectors encoding CRISPR-Cas9-sgCdkn2a (AAV2-Cas9-sgCdkn2a) or CRISPR-Cas9-sgCdkn2b (AAV2-Cas9-sgCdkn2b), and AAV2-IRES-GFP as an infection control, were coinjected intravitreally to knockout Cdkn2a or Cdkn2b in RGCs in vivo. NMDA was intravitreally administered on the day 32 following virus injection to examine the effect of loss of Cdkn2a or Cdkn2b on NMDA-induced RGC death (Fig. 5A). Loss of Cdkn2a or Cdkn2b significantly prevented NMDA-induced cell death in Rbpms⁺ RGCs compared to control cells infected with nonsense sgRNA-encoding AAV2 (Fig. 5B). Furthermore, a higher number of RGCs was observed in the retinas deficient for Cdkn2b than in those deficient for Cdkn2a. 

Figure 5

 

Impact of CRISPR-Cas9-mediated loss of Cdkn2a or Cdkn2b on NMDA-induced RGC reduction. (A) Immunostaining for Rbpms and GFP in NMDA-treated retinas co-infected with AAV2 vectors containing the SaCas9 gene and an sgRNA sequence (AAV2-SaCas9-sgRNA), or containing the gfp gene, as an infection control (AAV2-GFP). (B) Quantification of Rbpms⁺ RGCs in AAV2-injected retinas after NMDA injection. The number of surviving RGCs increased when Cdkn2a or Cdkn2b was knocked out. Error bars: SEM. *P < 0.05, **P < 0.01 (Tukey-Kramer test; n = 5–8). Scale bar: 50 μm.

Discussion

In our study, we characterized the phenotypic changes in NMDA-induced RGC death in mouse retinas, in which either Cdkn2a or Cdkn2b was overexpressed or knocked out using AAV2-mediated transgene expression. Results showed that loss of Cdkn2a or Cdkn2b attenuated NMDA-induced cell death in RGCs; moreover, Cdkn2b overexpression enhanced NMDA-induced RGC death. Intravitreal AAV2 injection induced transgene expression under control of the CMV promoter in RGCs as well as in other retinal cells.^38–40 We believe that the phenotypic changes observed in AAV2-infected retinas could be attributed primarily to altered expression of Cdkn2a or Cdkn2b in RGCs, as shown by the following results. First, a higher infection efficiency was found in RGCs than in other cell types in AAV2-infected retinas (Supplementary Fig. S1). Second, endogenous Cdkn2a and Cdkn2b expression was undetectable in the retinas of the control group, but markedly upregulated in RGCs in response to NMDA (Fig. 1B). Finally, NMDA-induced RGC death could occur due to a direct effect of NMDA on RGCs. Previous studies have reported that retinal cells, such as amacrine, bipolar, and horizontal cells, as well as RGCs express NMDA receptor subunits, and are potentially sensitive to NMDA.^38–40 Thus, NMDA-mediated loss of the retinal cells, other than RGCs, could induce RGC death as its secondary effect. However, intravitreal NMDA injection increased the number of TUNEL⁺ cells in the GCL earlier than in the INL and outer nuclear layer (ONL; Supplementary Fig. S2); most of the TUNEL⁺ cells in the GCL were Rbpms-expressing RGCs (92.1%; Supplementary Fig. S2). Taken together, we concluded that attenuation and enhancement of RGC death in AAV2-infected retinas were due to the primary effect of altered expression of Cdkn2a and Cdkn2b in RGCs. 

Expression of Cdkn2a and Cdkn2b was altered in experimental animal models of glaucoma and ischemia-reperfusion injury, and Cdkn2a deficiency results in attenuation of cell death.^26,30,32,41 However, the molecular mechanism underlying Cdkn2a-mediated cell death has not been completely elucidated. Cdkn2a can form a molecular complex with Cdk4 and the tumor suppressor p53, enhancing the binding efficiency of p53 to the promoter region of its target genes related to cell cycling and apoptosis.⁴² Thus, proapoptotic genes, such as Bax, p63, and p73, were downregulated following knockout of Cdkn2a.⁴² Another study demonstrated that overexpression of p53 and Cdkn2a, but not p53 alone, triggered apoptosis in tumor cells.⁴³ Given that NMDA-induced RGC death is partially mediated by p53,⁴⁴ the prevention of RGC death following Cdkn2a knockout observed in our study may be attributable, at least in part, to attenuation of p53 signaling. In contrast to Cdkn2a, little is known about the molecular mechanism underlying Cdkn2b-mediated cell death induction. Cdkn2b has similar binding properties as Cdkn2a in terms of interacting with common members of the cyclin-dependent kinase family (i.e., CDK4 and 6).^17–19 Therefore, it is possible that Cdkn2b induces RGC death via a molecular mechanism similar to that of Cdkn2a. 

Cdkn2a and Cdkn2b have been shown previously to exert identical effects on tumor cells, such as inhibition of cell growth and telomerase activity, and senescence induction.⁴⁵ Our study indicated that knockout of either Cdkn2a or Cdkn2b rescued NMDA-induced RGC loss, suggesting their similar roles in the context of RGC death induction. However, this study also revealed potential differences in physiologic functions between Cdkn2a and Cdkn2b: knockout of Cdkn2b resulted in stronger inhibition of NMDA-induced RGC death than did knockout of Cdkn2a. This phenomenon can be explained partially by the novel finding that Cdkn2b can stimulate Cdkn2a expression, but not vice versa. CRISPR-Cas9-mediated knockout of Cdkn2b led to a reduction in not only Cdkn2b, but also Cdkn2a proteins in RGCs. Thus, Cdkn2b knockout would be more effective in preventing NMDA-induced RGC loss. Another striking difference between Cdkn2a and Cdkn2b is that overexpression of Cdkn2b facilitated RGC loss in NMDA-treated retinas, whereas overexpression of Cdkn2a did not. Endogenous Cdkn2b expression was upregulated gradually in RGCs after NMDA injection, coincident with an increase in RGC death (Fig. 1B). This implies that there is a critical correlation between Cdkn2b expression and NMDA-mediated RGC death. In contrast, Cdkn2a expression increased in response to NMDA; however, its expression level did not increase over time (Figs. 1A, 1B). Because Cdkn2a expression was undetectable in the control retinas untreated with NMDA, the role of Cdkn2a in NMDA-induced RGC death may depend on the presence of Cdkn2a, but be independent of its expression level. Taken together, we hypothesized that Cdkn2b has the ability to facilitate NMDA-induced RGC death by coexistence with Cdkn2a. Thus, the loss of either Cdkn2a or Cdkn2b can attenuate RGC death in NMDA-treated retinas. 

Intravitreal NMDA administration results in RGC degeneration due to neuronal excitotoxicity.^33,34 The degeneration process would be triggered by excessive influx of extracellular calcium into cells via NMDA-type glutamate receptors.^46–48 Calcium accumulation hyperactivates intracellular enzymes, including proteases, kinases, phospholipases, and neuronal nitric oxide synthase, leading to oxidative stress.^49–57 Oxidative stress-induced cellular damage drives senescence, coincident with increased expression of senescence markers, such as SA-ß-gal, Cdkn2a, and Cdkn2b.^58,59 Consistent with these, increased expression of senescence markers Cdkn2a and Cdkn2b was observed in RGCs suffering NMDA-mediated oxidative stress. Cellular senescence is found initially as a unique state of normal mitotic cells displaying the limited replicative property after repeated proliferation.⁶⁰ Accumulating evidence has indicated that differentiated postmitotic neurons as well as normal mitotic cells show senescence-like state, characteristic of senescence marker expression, with age, and also in response to various cellular stresses and damages.⁶¹ This could lead to pathogenesis of senescence-related diseases.⁵⁹ In an animal model of glaucoma, high IOP induced upregulated expression of Six6 and Cdkn2a, and genetic reduction of Six6 and Cdkn2a led to decreased expression of Cdkn2a and rescue of the RGC loss phenotype, respectively.²⁶ On the other hand, our study indicated that NMDA-induced excitotoxity upregulated the expression of Cdkn2a and Cdkn2b, but not Six6. Considering these findings, expression of senescence genes, such as Cdkn2a and Cdkn2b, appears to be triggered in response to different extrinsic and intrinsic stresses, suggesting that these proteins function as downstream cell death signal transducers. 

In conclusion, we elucidated the roles of Cdkn2a and Cdkn2b in RGCs using in vivo gain- and loss-of-function approaches, and found that these molecules, especially Cdkn2b, are essential for NMDA-induced cell death. Our results suggested that Cdkn2a and Cdkn2b are potential therapeutic targets to prevent RGC degeneration observed in eye diseases related to neuronal excitotoxicity, including ischemia-reperfusion injury and glaucoma. 

Acknowledgements

The authors thank Junko Sato and Mayumi Suda for technical assistance, and the Biomedical Research Unit of Tohoku University Hospital for technical support. 

Supported by JSPS KAKENHI Grant Number JP17H04349 (T.N.) from the Ministry of Education, Science and Technology of Japan. 

Disclosure: H. Tawarayama, None; Q. Feng, None; N. Murayama, None; N. Suzuki, None; T. Nakazawa, None