Dominant inheritance of retinal ganglion cell resistance to optic nerve crush in mice

Inbred mice exhibit quantitative differences in ganglion cell loss after optic nerve crush

Retinal ganglion cell loss after optic nerve crush occurs over a period of 3 weeks, with the most rapid period of loss occurring between the first and second weeks [26]. We chose to examine the period of 2 weeks after optic nerve crush in order to assess the process of cell death while still at an intermediate stage (see Methods section). A bar graph of the mean (± SD) of cells remaining in the crushed retina for 10 mice in 15 different lines is shown in Figure 1. Representative micrographs of control and crushed Nissl-stained retinas from mice at each end and the middle of this distribution are shown in Figure 2. Mice exhibited a range of cell loss after 2 weeks from 26.3% to 47.4% of the total neurons in the ganglion cell layer. A matrix of P values comparing each of the means against each other (t-tests) is shown in Table 1. Both sets of inbred lines with similar genetic backgrounds (DBA and C57BL mice) had comparable amounts of cell loss based on statistical testing. There were no significant correlations between coat color and susceptibility or known neurosensory defects and susceptibility (data not shown).

Ganglion cell prevalence in DBA/2J and BALB/cByJ mice

Dominant inheritance of ganglion cell resistance to crush

To confirm these as resistant and susceptible strains, we performed optic nerve crush on increased numbers of mice from each strain (28 DBA/2J and 31 BALB/cByJ mice). Consistent with the initial screen, DBA/2J mice exhibited significantly less cell loss than BALB/cByJ mice (Figures 3A and 3B, P = 3.8 × 10^-8, t test). To determine whether the susceptibility to optic nerve crush was genetically controlled, crosses were performed between the two strains. F1 mice (n = 28) inherited the optic nerve crush-resistance phenotype exhibited by DBA/2J mice (Figure 3C, P = 0.178 vs DBA/2J and P = 0.001 vs BALB/cByJ mice). N2 mice (n = 55), exhibited a shift toward a more susceptible phenotype (Figure 3D, P = 2.2 × 10^-6 vs DBA/2J and P = 0.15 vs BALB/cByJ). F2 mice (n = 56) yielded an intermediate phenotype after optic nerve crush with a bias toward the resistance phenotype (Figure 3E, P = 0.017 vs DBA/2J and P = 0.006 vs BALB/cByJ). By assigning a mean value for cell loss for the resistant and susceptible phenotypes, cell loss in each of the crosses was accurately predicted by the expected distribution of 2 dominant alleles using a Punnett Square method of analysis (data not shown).

A statistical estimation of the number of affecting loci was made using a modified Wright Formula, which takes into account the means and variances of the F1 mice and the recurrent backcross parental strain [27, 28]. This formula yielded a value of 1.3, suggesting that only a few genes were contributing to the inheritance pattern of this phenotype, consistent with the Punnett Square analysis.

Elimination of Bax as a contributing allele to the resistance phenotype

The proapoptotic gene Bax is essential for ganglion cell death after optic nerve crush [24] and in chronic spontaneous glaucoma [25]. Reduced Bax expression in mice heterozygous for a Bax knock-out allele has varying effects on neurons, depending on genetic background [25, 29], an effect that appears to be related to the level of latent Bax mRNA from the endogenous wild type gene (S. Semaan and R. Nickells, unpublished observation) [30]. Bax^+/- DBA/2J mice are completely resistant to optic nerve crush [25] and have relatively low levels of neuronal Bax expression. Interestingly, DBA/2J mice were the most resistant to the same crush protocol of all the inbred mice examined.

To examine Bax as a candidate gene for the resistant phenotype, we first quantified Bax mRNA levels in the brains of DBA/2J and BALB/cByJ mice using RNase Protection Assays (RPA). Both lines express a BAX protein with the same predicted amino acid sequence (data not shown), so if Bax was associated with the resistance phenotype, we predicted that BALB/cByJ mice would have higher levels of latent neuronal Bax mRNA. Quantitative examination of Bax mRNA levels in BALB/cByJ and DBA/2J mice showed no statistical difference between the two lines (P = 0.49, Figure 4). In a second experiment, we mapped the segregation of DBA/2J (D2) and BALB/cByJ (B) Bax alleles in N2 and F2 mice used in this study. A scatter plot of phenotype as a function of Bax genotype is shown in Figure 5. Evaluation of these data was made comparing mice with the Bax^B/B background against Bax^D2/B and Bax^D2/D2 mice (animals with the D2 allele were either pooled together as one group or compared as separate groups). None of the 4 possible comparisons yielded a significant association between the Bax genotype and the resistance phenotype (ANOVA, P = 0.150 for Bax^B/B vs. Bax^D2/Band Bax^D2/D2 N2 and F2 mice combined, n = 106; P = 0.356 for Bax^B/B vs. Bax^D2/B vs. Bax^D2/D2 N2 and F2 mice, n = 106; P = 0.987 for Bax^B/B vs. Bax^D2/B and Bax^D2/D2 F2 mice only, n = 53; and P = 0.575 for Bax^B/B vs. Bax^D2/B vs. Bax^D2/D2 F2 mice only, n = 53).

The Influence of Genetic Background on Neurodegeneration in Inbred Mice

Several studies have reported that inbred strains of mice exhibit variable susceptibility to neurodegenerative stimuli, but there has been no consistent correlation between strain and susceptibility. C57BL/6 mice, which were relatively crush resistant in our study, are more susceptible to MPTP exposure, spinal chord injury, and experimental autoimmune encephalitis (EAE) than other strains. Conversely, BALB/c mice (crush susceptible) are resistant to EAE and MPTP, but exhibit a similar susceptibility as C57BL/6 mice to spinal chord injury. C57BL/10 mice (crush resistant) and B10.PL mice, which are EAE resistant and susceptible, respectively, both exhibit less severe injury after spinal chord lesion than other strains [9]. Part of the differences exhibited by some inbred strains may reside in their variable abilities to mount an immune response, which can affect the progression and severity of certain neurodegenerative diseases [9, 31]. The lack of a correlation between mouse strains and their susceptibility to different degenerative challenges is not surprising. Each challenge is potentially unique to different subsets of neurons (such as cells in the substantia nigra being susceptible to MPTP). It is also likely that different diseases elicit different damaging stimuli, which in turn can activate alternative pathways leading to neuronal death [24, 32].

Genes that affect Retinal Ganglion Cell Death

Recently, Libby et al [14] outlined two broad classes of genes that can affect ganglion cell susceptibility to damage in glaucoma. The first class modulates extrinsic susceptibility of ganglion cells. This set of genes is predicted to influence the response of non-ganglion cells (e.g., glia) to ocular hypertension, the composition and modulation of the extracellular matrix, the density and regulation of retinal vasculature of the optic nerve head, and possibly the immune responses in the eye. The second class of genes modulates the intrinsic susceptibility of ganglion cells. These latter genes are predicted to directly affect ganglion cell survival. By the nature of our screen, which directly activates apoptosis by axonal lesion, we predict that we are most likely selecting for intrinsic susceptibility genes.

Many genes could be intrinsic susceptibility factors. Retinal ganglion cell death follows the basic molecular pathways identified in other neurons and several of the key genes involved in cell death could affect ganglion cell susceptibility to damaging stimuli. We examined the Bax gene as a candidate based on the evidence that this gene was critical for ganglion cell death in a variety of situations [23–25]. Several other genes, however, are also known to influence ganglion cell death. Proapoptotic genes active in ganglion cells include c-Jun [33, 34] and c-JUN terminal Kinase (Jnk) [35], p53 [36], Bim [37], Apaf-1 [34], poly(ADP-ribose) polymerase (Parp) [38], and several caspases [39–42]. Interestingly, a significant association has been reported between p53 (TP53) haplotypes and primary open angle glaucoma in humans [43].

In addition to proapoptotic genes, ganglion cells are also significantly affected by the expression of antiapoptotic genes. Ganglion cells express BclX, whose expression decreases during the initial stages of apoptosis [44]. Upregulation of this gene in ganglion cells enhances their survival [45]. Similarly, transgenic mice that overexpress the BclX homolog, Bcl2, exhibit nearly complete abrogation of ganglion cell death during development and after optic nerve transection [46, 47]. Since downregulation of endogenous BclX, along with other normally expressed genes [48, 49], is an early event in ganglion cell apoptosis, any variation that reduces this process and allows these proteins to persist could affect the normal apoptotic program.

Damaged ganglion cells in mammals also exhibit an early stress response that includes the up-regulation of heat shock proteins [50, 51] and iron-regulating genes including transferrin and ceruloplasmin [52, 53]. These latter genes may also act as antioxidants by scavenging free radicals formed from mitochondria during the cell death process [54].

Although Bax has been eliminated as a candidate gene in this population of mice, clearly there are other genes that could affect the cell death phenotype. We are currently in the process of conducting a genome wide screen to identify potential loci that affect the resistance phenotype.

Using Mice to Identify Susceptibility Alleles for Glaucoma

In this study, we used inbred mice to assess if genetic background affects the susceptibility of retinal ganglion cells to a lesion of the optic nerve. These experiments do not directly address whether the same genes that affect ganglion cells after optic nerve crush will also be susceptibility alleles for ganglion cells exposed to elevated levels of IOP, the principal damaging insult in glaucoma. The actual mechanism linking high IOP with ganglion cell damage is still unresolved, but some studies suggest that elevated pressure causes increased sheer forces on the optic nerve at the level of the scleral canal and lamina cribrosa [55, 56]. This has lead to the hypothesis that increased IOP causes mechanical damage to axons at this region [57, 58], and it is becoming more evident that this is the region of initial damage in glaucoma [59–61]. Although the concept of sheer stress acting directly on the axons is attractive, and could be analogous to our crush insult, there are several other mechanisms by which IOP can induce damage in this region. These include reducing the perfusion of microvessels in the nerve head and the focal activation of astrocytes and microglia. Consequently, it is possible that optic nerve crush may act differently on ganglion cells than elevated IOP.

Regardless of the nature of the insult, there is some evidence, however, both optic nerve crush and elevated IOP result in a similar cell death response within the ganglion cells themselves. Ganglion cell apoptosis is Bax-dependent [24, 25] and leads to the activation of caspases [39, 41, 42] in both conditions. In addition, ganglion cells undergo early losses of gene expression before the activation of detectable ganglion cell loss [48] after crush and in glaucoma. In view of the similarities of the cell death pathways activated by both conditions, we predict that the susceptibility alleles that affect ganglion cell death after crush will also affect ganglion cell loss in glaucoma. A critical test of this hypothesis is to first identify the loci that affect the crush phenotype and then assess how the same gene(s) affect ganglion cell loss in glaucoma. For example, DBA/2J mice naturally develop a form of secondary chronic inherited glaucoma [62–65]. Although DBA/2J mice are the most resistant strain assayed in our study, they still exhibit variable susceptibility to elevated IOP as they age, which eventually leads to ganglion cell loss. Breeding the susceptibility alleles inherent in BALB/cByJ mice onto the DBA/2J background, however, could result in mice that develop a more aggressive form of glaucoma. Thus, in response to the same levels of elevated IOP, these modified mice may develop earlier onset of ganglion cell loss or perhaps an accelerated rate of retinal degeneration.

Inbred mice

Animals were handled in accordance with the Association for Research in Vision and Ophthalmology Statement on the use of animals for research. All inbred mice were purchased from the Jackson Laboratory (Bar Harbor, ME). Five female and 5 male mice, aged 6 weeks, were obtained from each of the following lines: 129X1/2J, A/J, BALB/cByJ, C3H/HeJ, C57BL/6J, C57BL/10J, CBA/CaJ, DBA/1J, DBA/2J, FVB/NJ, LG/J, MRL/MpJ, NOD/LtJ, NZB/BINJ, SJL/J. Lines were chosen to obtain a cross section of mice with different coat colors, varying degrees of neurosensory defects (including 8 lines with hearing loss due to homozygosity for the Cdh23 ^ahl allele and 3 lines with photoreceptor degeneration due to homozygosity for the Pde6b ^rd1 allele), and at least 2 pairs of lines with similar genetic backgrounds (C57BL/6J – C57BL/10J and DBA/1J – DBA/2J). DBA/2J mice also exhibit spontaneous chronic glaucoma associated with the loss of ganglion cells [62, 65]. This disease affects older mice and was not expected to influence the response of young mice to an acute lesion.

Mice were aged to 8 weeks, at which time they underwent optic nerve crush on the left eye (see below). Mice were allowed to recover and were maintained for another 2 weeks before being euthanized for cell counting. This protocol was strictly adhered to for all the mice examined in this study. New animals were obtained from the Jackson Laboratory for all matings that required the original inbred parental strains. A reciprocal breeding strategy was used for subsequent matings to generate N2 backcrossed mice (F1 × susceptible parental strain) and F2 mice (F1 intercross).

Optic nerve crush and ganglion cell counting

For optic nerve crush surgery, mice were anesthetized with ketamine (6 mg/mL) and xylazine (0.4 mg/mL). The crush protocol was performed as described previously [26] using an intraorbital approach. Only the left eye of each mouse underwent the procedure. The loss of ganglion cells is a continuous process over a 3–4 week period, therefore we examined the time point at which there was the earliest maximum difference between the resistant and susceptible strains. Both DBA/2J and BALB/cByJ mice exhibit minimal cell loss at one week, followed by a period of rapid cell loss between 1 and 2 weeks. Each strain also exhibits modest cell loss after this point, but the difference between the two is not different than that observed at 2 weeks after crush (data not shown). For this reason, we examined all populations of mice at 2 weeks after optic nerve crush.

To quantify cell loss, mice were euthanized and the superior region of each eye was marked with an ophthalmic cautery prior to enucleation. Each eye was then enucleated and fixed for 1 h at 22°C in 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4) containing 100 mM NaCl (PBS) followed by rinsing in PBS. An eyecup of the posterior pole was isolated and incubated in PBS containing 0.3% TritonX-100 (v/v), overnight at 22°C. Following incubation, the retina was dissected and mounted (ganglion cell layer up) on a glass "Plus" slide (Fisher Scientific, Chicago, IL), dried, and flattened under a coverslip weighted with a 10 g weight [66]. Dry retinas were stained by painting them with 1% cresyl violet acetate in 0.25% acetic acid [67]. Stained retinas were then differentiated and dehydrated in 100% ethanol, cleared in xylene, and coverslipped. Each slide contained the control and experimental retinas from a single mouse.

For counting cells in the ganglion cell layer, we captured digital images taken at 200× magnification using an Olympus BX40 light microscope (Olympus, Mellville, NY) with a SONY DXC-390 video camera attachment (Sony, New York, NY) and imported them into Image Pro Plus v4.5 (Media Cybernetics, Inc., Silver Springs, MD) quantification software. Parameters were set to automatically identify cells with round nuclei greater than 4 μm in diameter and exclude vascular endothelial cells (with spindle-shaped nuclei) and small densely staining cells that are common in the mouse retina [68]. The selection process did not distinguish between ganglion cells and large amacrine cells, but previous studies have estimated the proportion of ganglion cells in this layer to be 40–60% of the neurons present [26, 68, 69]. The number of cells present in a minimum of 4 microscopic fields, each encompassing an area of 0.33 mm² and taken in the peripheral to midperipheral region around the 4 quadrants of each retina, was counted. In non-experimental retinas, the total numbers of cells ranged from 2055 – 2913 per field, consistent with previous reports of neuronal cell density in this layer [69]. The numbers of cells present in each retina was taken as the average of these counts and cell loss was calculated in each mouse as the percent change between the control and experimental eyes. Overall, we counted approximately 10% of the neurons present in each retina, which previous studies have demonstrated provides an accurate estimate of the loss of retinal ganglion cells after optic nerve crush [25, 26, 48].

Ganglion cell number was also estimated in untreated eyes of DBA/2J and BALB/cByJ mice by axon counting. Briefly, the optic nerves of 5 eyes from each strain were harvested and fixed overnight in 2.5% glutaraldehyde in PBS and then embedded in glycol methacrylate (JB-4 Plus, PolyScience, Worthington, PA). Transverse sections of 2 μm thick were cut approximately 1 mm from the globe and stained by silver impregnation to identify axons [67]. Axon counting was conducted as described previously [36] using 3 sections from each nerve.

Ribonuclease protection assays (RPAs)

Total RNA was isolated from the brains of DBA/2J (n = 24) or BALB/cByJ (n = 23) mice. Radiolabeled antisense riboprobes for RPA were synthesized from linearized pBK-CMV plasmid (Stratagene, La Jolla, CA) templates containing the terminal 127 nucleotides of the murine Bax coding region or nucleotides 55–252 of the murine S16 ribosomal protein coding region. Probe synthesis, purification, hybridization, and RNase digestion conditions were as described previously [48]. Because retinas were used for the quantification of cell loss, RPA assays were performed on total RNA isolated from brain tissue of mice. In some mice, however, we confirmed that the level of Bax mRNA, and the ratio of Bax mRNA to S16 mRNA, was consistent between the 2 CNS tissues (data not shown). RPAs were initiated by precipitating 20 μg brain total RNA for Bax or 1 μg brain total RNA for S16 with 5 × 10⁵ acid precipitable cpm of probe. After hybridization and RNase digestion, protected probes were separated on 0.4 mm 8 M urea/6% polyacrylamide gels, and exposed to BioMax MR film (Kodak, Rochester, NY). Exposure times ranged from overnight to 10 d at -80°C with an intensifying screen. Band intensities were quantified from densitometric scans of multiple exposures of each gel using NIH Image software, and confirmed by liquid scintillation counting of the excised bands.

Genotyping mice

Genomic DNA was isolated from the spleens of N2 backcrossed mice (n = 55) and F2 mice (n = 56) for genotyping. Bax is located at position 23.0 on chromosome 7. PCR genotyping analysis was performed for the polymorphic marker D7Mit27 using the primers: 5' TGA ACT GGG GAG GAA AGT TG and 5' AAC ATG AAA AGA CAT TCC CCC. These primers yielded a 218 bp fragment for DBA/2J mice and a 182 bp fragment for BALB/cByJ mice and map within 1.5 Mbp of the mouse Bax gene. PCR products were separated on 3% Meta Phor Agarose (Cambrex, Rockland, ME) gels and the bands were identified by ethidium bromide staining.