Common Variants at 9p21 and 8q22 Are Associated with Increased Susceptibility to Optic Nerve Degeneration in Glaucoma

1Glaucoma is a leading cause of blindness worldwide. Primary open angle glaucoma (POAG), the most common form of glaucoma in the Western world, is an age-related, complex disease characterized by progressive irreversible degeneration of the optic nerve due to apoptotic retinal ganglion cell death. In addition to age, epidemiologic studies have revealed multiple risk factors for the condition including elevated intraocular pressure (IOP), African-American race, family history and low ocular perfusion pressure –. Of these, elevated intraocular pressure (IOP) is the only treatable risk factor; however, many individuals have IOP elevation without optic nerve disease, and at least 33% of affected individuals have progressive retinal ganglion cell loss despite IOP measurements in the normal range (less than 22 mmHg), a condition defined as normal-pressure glaucoma (NPG). Preventative or neuro-protective therapies for glaucoma are not yet available and little is known about the molecular events that influence susceptibility to glaucomatous optic nerve degeneration.

2POAG is genetically complex. Linkage studies have identified over 20 genomic regions likely to contain POAG-related genes. Importantly, genes that influence POAG risk overall may specifically contribute to separate independent biological processes affecting the disease outcome including regulation of IOP and retinal ganglion cell physiology. Individuals with elevated IOP without optic nerve disease may only carry genetic variants that influence IOP regulation, while individuals with NPG may primarily carry genetic variants that predispose to retinal ganglion cell death as the nerve degenerates in these patients without the added stress of elevated IOP.

3Recent genome-wide association studies (GWAS) have identified several genetic risk factors for POAG overall, including single-nucleotide polymorphisms (SNPs) located in the CAV1/CAV2 intergenic region –, and in the genomic regions containing the TMCO1 and CDKN2BAS genes in a study of glaucoma patients with advanced optic nerve disease. Linkage studies have identified two genes that contribute to rare familial forms of NPG, OPTN (optineurin) and TBK1 (TANK-binding kinase), and an intronic SNP in SRBD1 (S1 RNA binding domain) has been associated with NPG in a GWAS of 355 Japanese cases. Genome-wide association studies have not yet identified genes commonly associated with normal-pressure glaucoma or with optic nerve disease in glaucoma. The identification of genes that influence glaucoma-related optic nerve degeneration is an important step toward the development of neuro-protective therapies that could substantially reduce the morbidity caused by this common disease, and such therapies could be relevant to optic nerve degeneration occurring in many chronic forms of glaucoma.

4To identify genes that predispose to glaucomatous optic nerve disease, we completed two GWAS for POAG: the GLAUGEN (Glaucoma Genes and Environment) GWAS that is part of the GENEVA (GENEVA Genes Environment Association) studies and the NEIGHBOR (NEI Glaucoma Human genetics collaBORation) GWAS. We then performed a meta-analysis as well as normal-pressure and high-pressure subgroup analyses of the combined dataset. To determine if the observed associations were specific to POAG or could generalize more broadly to optic nerve disease in other forms of glaucoma we selected the lead SNPs showing significant association in the normal-pressure glaucoma subgroup analysis, and tested them for association with glaucoma in a population of unrelated individuals with exfoliation syndrome glaucoma.

5After data cleaning, 976 cases and 1,140 controls collected from three study sites [Nurses' Health Study, Health Professionals Follow-up Study and the Genetic Etiologies of Primary-Open angle Glaucoma Study (GEP)] were analyzed for the GLAUGEN study, and 2,170 cases and 2,347 controls collected from 12 sites (Table S1) were analyzed for the NEIGHBOR study. All cases and controls for both studies were residents of the continental United States and were of mainly European ancestry, which was confirmed by principal component analysis. The general characteristics for the cases and controls are shown in Table 1. For POAG overall in the GLAUGEN GWAS, there were no SNPs with p-values that reached the genome-wide significance level of p = 5×10−8 (Figure S1 and Figure S2).

6In the NEIGHBOR overall POAG GWAS, there were seventeen SNPs that reached the genome-wide significance level of p = 5×10−8. Sixteen of the significant SNPs were found in the CDKN2BAS gene region on chromosome 9p21.3 (top SNP rs4977756 (OR = 0.66 [95% CI 0.59–0.73], p value = 7.4×10−16). The remaining significant SNP rs10483727 was located in the SIX1/SIX6 region on chromosome 14 (OR = 1.32 [95% CI 1.20–1.46], p value = 3.1×10−8). (Figure S3 and Figure S4; Table S2).

7Using METAL, we conducted a meta-analysis of the GLAUGEN and NEIGHBOR datasets that included a total of 3,146 cases and 3,487 controls and found nineteen SNPs that achieved genome-wide significance for POAG overall (Figure S5): seventeen of them in the CDKN2BAS region, with the most significant SNP being rs2157719 (OR = 0.69 [95%CI 0.63–0.75], p = 1.86×10−18, and two SNPs in the SIX1/SIX6 region on chromosome 14q23 (most significant SNP: rs10483727, OR = 1.32 [95%CI 1.21–1.43], p = 3.87×10−11) (Table S3). A number of genomic regions, including the CDKN2BAS and SIX1/SIX6 regions have been previously associated with optic nerve quantitative parameters (cup-to-disc ratio (CDR) and optic nerve area) –. Of these, only the CDKN2BAS and SIX1/SIX6 regions were significantly associated with glaucoma in the meta-analysis, although several other previously identified gene regions demonstrated suggestive associations including SALL1, LRP1B and SIRPA (Table 2, Table S4).

8To identify genetic risk factors primarily associated with optic nerve disease in glaucoma, we performed separate meta-analyses for NPG (720 cases) and high-pressure glaucoma (HPG, 1669 cases with IOP ≥22 mm Hg), (757 cases did not have untreated IOP data available). In the HPG meta-analysis, no SNPs, even those in the CDKN2BAS and SIX1/SIX6 regions, reached genome-wide significance (Figure S6). In the NPG meta-analysis, SNPs in two regions reached genome-wide significance, (Figure 1, Table 3): thirteen SNPs in the CDKN2BAS region on 9p21 (most significant SNP was rs2157719, (OR = 0.58 [95% CI 0.50–0.67], p = 1.17×10−12) (Figure 2, Figure S7) and three SNPs in an evolutionarily conserved region on chromosome 8q22 (most significant SNP was rs284489, OR = 0.62 [95% CI 0.53–0.72], p = 8.88×10−10) (Figure 3, Figure S8).

9After imputation 90 SNPs in the 9p21 region were significantly associated with NPG (Table S5), although conditional analysis showed that none of these were independent of rs2157719, the most significantly associated SNP in the meta-analysis. The minor allele for rs2157719 showed a protective effect, and is located within the genomic region coding for the noncoding RNA, CDKN2BAS, also known as ANRIL, which regulates expression of CDKN2B as well as other genes. rs2157719 falls within a DNase I hypersensitivity site that has a significant signal in a variety of cell types (Figure S9). Targeted re-sequencing of the CDKN2BAS gene in a subset of 11 cases carrying a protective haplotype that includes rs2157719 (Table S6) identified two rare variants (rs7341786 [A]; rs7341791 [A]) in 8 of the 11 individuals. After imputation we found that these variants are significantly associated with glaucoma (p = 2.73×10−9, OR = 0.75; p = 3.09×10−9, OR = 0.75 respectively). These SNPs have been previously shown to affect RNA splicing, suggesting that altering CDKN2BAS expression may contribute to glaucoma.

10The three SNPs located in the 8q22 region that were significantly associated with NPG in the subgroup analysis are in strong linkage disequilibrium (r2>0.8) (Figure S8), and imputation followed by conditional analysis did not reveal additional independent associations. The genomic region containing the associated SNPs includes highly conserved DNA sequences and several DNaseI hypersensitivity sites, (Figure S10). An imputed SNP (rs284492) in strong LD with the lead SNP (rs284489) falls within or proximal to DNaseI hypersensitivity sites with significant activity in choroid plexus epithelial cells, as well as two ocular tissues (nonpigmented ciliary epithelial cells, and iris pigment epithelial cells) (Figure S10, Table S5). Haplotypes significantly associated with NPG (Table S7) include these DNaseI hypersensitivity sites. This putative regulatory region is located within a 630 Kb gene desert flanked by LRP12 (Low-density lipoprotein receptor related protein 12) and ZFPM2 (Zinc finger protein multitype 2) and could influence expression of either of these genes.

11Using RT-PCR, we found that both LRP12 and ZFPM2 are expressed in the human optic nerve, as well as in other ocular tissues relevant to glaucoma (Figure S11). In mouse experiments using laser microdissection to selectively study specific layers of the retina, both LRP12 and ZFPM2 were expressed in the retinal ganglion cell and inner nuclear layers (Figure S11), making them good candidates for optic nerve susceptibility genes.

12We performed targeted resequencing of the LRP12 and ZFPM2 genes in 16 NPG patients and also analyzed the exomes of 50 POAG cases and 18 controls for DNA sequence variants in both genes (Table S8). None of the variants identified are expected to significantly affect gene expression or protein function.

13To determine if SNPs associated with NPG were associated with optic nerve degeneration in other types of glaucoma, we examined the association of the top SNPs in both the 9p21 and 8q22 genomic regions with optic nerve degeneration in exfoliation glaucoma, another form of open-angle glaucoma. Exfoliation syndrome, a condition predisposing to elevated intraocular pressure, is associated with genetic variants in LOXL1, however, these variants appear to contribute to the development of the exfoliation syndrome and not to optic nerve disease in POAG. We found that rs2157719 in 9p21 was significantly associated with exfoliation glaucoma (Table 4), and rs284489 in 8q22 was also nominally associated with the same directionality of the association observed for NPG.

14Both the 9p21 and 8q22 regions could contribute to TGF-beta signaling. The 9p21 SNPs could effect expression of CDKN2BAS, a noncoding RNA that influences expression of CDKN2B, a member of the TGF-beta signaling pathway. Using the SCAN database, we found that the three 8q22 SNPs associated with NPG (rs284489, rs284495 and rs1521774) influence the expression of TSC22 (TGF-beta stimulated clone 22), which also modulates signaling by TGF-beta. Using the PARIS (Pathway Analysis by Randomization Incorporating Structure) algorithm, we found that the TGF-beta pathway (KEGG, hsa04350) Kyoto Encyclopedia of Genes and Genomes) overall was associated with NPG in our combined dataset (permuted p = 0.009).

15Using two large POAG case-control datasets, we have completed a meta-analysis for POAG followed by subgroup analyses for NPG and HPG. We have identified two genomic regions that are associated with NPG and showed that the lead SNP in each region is also associated with optic nerve disease in a second type of open-angle glaucoma (exfoliation syndrome related glaucoma).

16The first region includes the CDKN2BAS gene on chromosome 9p21, previously associated with cup-to-disc ratio (CDR) an optic nerve quantitative parameter, as well as POAG in candidate gene studies,, and more recently in a GWAS using a sample of severely affected POAG patients. In this study, we show that this gene region is associated with NPG suggesting that CDKN2BAS contributes to optic nerve degeneration in glaucoma. CDKN2BAS codes for an antisense RNA that regulates the expression of CDKN2B, which is an inhibitor of cyclin-dependent kinase 4 (CDK4), a protein kinase that has a pivotal role in cell cycle progression. The minor alleles of the NPG associated SNPs are protective, suggesting that these SNPs, or other variants in linkage disequilibrium with these SNPs, could influence CDKN2BAS and CDKN2B expression with corresponding changes in cyclin-dependent kinase activity that could result in retinal ganglion cell apoptosis. CDKN2BAS expression can also be influenced by interferon alpha. As inflammation and autoimmunity may be factors contributing to glaucoma pathogenesis, the regulation of CDKN2BAS expression by interferon could suggest a direct role for CDKN2BAS in glaucoma pathogenesis. CDKN2BAS also regulates the expression of CDKN2A, a gene previously shown to be down-regulated in other neurodegenerative disorders, including Alzheimer's disease, suggesting that regulation of CDKN2A expression by CDKN2BAS could also contribute to degeneration of the optic nerve in glaucoma.

17The second region associated with NPG in this study is an evolutionarily conserved DNA segment on chromosome 8q22 with predicted regulatory function. This region partially overlaps with a POAG locus defined by a linkage study of a single large family with low intraocular pressure (GLC1D), and a second linkage study using multiple affected families stratified by age of disease onset. The conserved region falls within a 630 Kb gene desert flanked by the LRP12 and ZFPM2 genes. ZFPM2, also known as FOG2, codes for a zinc finger protein that appears to contribute to cardiovascular development, and may have a role in ocular development. ZFPM2 is expressed in the eye, and expression was observed to increase after injury to the optic nerve. The LRP12 gene is a member of the LRP (low-lipoprotein receptor) gene family, with several members previously implicated in glaucoma. LRP1 is decreased in the optic neuropathy associated with Alzheimer's disease, LRP1B is associated with optic nerve area, a knock-out of LRP2 causes a glaucoma-like phenotype in zebrafish, LRP4 expression is increased in response to retinal ischemia, and expression of LRP10 and LRP11 is up-regulated in a rat model of glaucoma,. Importantly, LRP12, previously known as ST7, is a receptor for SMAD4, a major regulator of the TGF-beta signaling pathway. Interestingly, SMAD4 influences activity of TSC22.

18The 8q22 genomic region associated with NPG contains putative regulatory sites that appear to be active in two cell types that could contribute to glaucoma-related optic nerve disease, choroid plexus epithelial cells and non-pigmented ciliary body epithelial cells. The choroid plexus is responsible for formation of cerebral spinal fluid, and recent studies have suggested that low cerebral spinal fluid pressure, which could be caused by decreased formation of cerebral spinal fluid, may create a deleterious gradient across the lamina cribrosa in NPG mimicking a similar gradient induced by higher IOP in HPG,. The 8q22 regulatory region is also active in non-pigmented ciliary body epithelial cells, which affect ocular intraocular pressure through the production of aqueous humor (a fluid that is similar in composition to cerebral spinal fluid).

19Both genomic regions identified in this study could contribute to regulation of TGF-beta signaling and our pathway analysis provides additional support for a role for TGF-beta signaling in glaucomatous optic nerve disease and retinal ganglion cell death. TGF-beta and other members of the TGF-beta signaling pathway have been previously implicated in glaucoma,, and genes participating in TGF-beta signaling are expressed in ocular structures that are involved in glaucoma, including the anatomic structures regulating IOP and the optic nerve –. TGF-beta1 is up-regulated in optic nerve tissue in an animal model of glaucoma, and mutations in LTBP2 (latent transforming growth factor beta binding protein 2) cause a rare autosomal recessive form of congenital glaucoma. Collectively, these results suggest that therapies directed toward the regulation of TGF-beta signaling could protect the optic nerve from degeneration in glaucoma. The discovery of POAG genetic risk factors, especially the factors that predispose to glaucoma-related optic neuropathy, is a critical first step toward understanding the pathophysiology of POAG and the development of gene-based screening tests and neuro-protective therapies for this common blinding disease.

20The institutional review boards of the Massachusetts Eye and Ear Infirmary, Harvard School of Public Health, the Brigham and Women's Hospital, University of Pittsburgh, Johns Hopkins University, Duke University, University of West Virginia, University of Miami, University of Michigan, Stanford University, Marshfield Clinic, and the University of California, San Diego. approved this study.

21POAG cases were defined as individuals for whom reliable visual field (VF) tests show characteristic VF defects consistent with glaucomatous optic neuropathy. Individuals were classified as affected if the VF defects were reproduced on a subsequent test or if a single qualifying VF was accompanied by a cup-disc ratio (CDR) of 0.7 or more in at least one eye. The examination of the ocular anterior segment did not show signs of secondary causes for elevated IOP such as exfoliation syndrome or pigment dispersion syndrome and the filtration structures were deemed to be open based on clinical measures. Elevation of IOP was not a criterion for inclusion; however, 67% of cases did have a history of elevated IOP (≥22 mm Hg) measured in a clinical setting (typically between the hours of 8AM and 5PM) and were classified as high-pressure glaucoma (HPG). Cases with IOP <22 mm Hg measured in the clinic at the time of study enrollment (without treatment) were classified as normal-pressure glaucoma (NPG). Cases undergoing IOP-lowering therapy at the time of enrollment were included in the HPG group if they had a documented history of IOP >22 prior to treatment and cases undergoing IOP-lowering therapy at the time of enrollment were included in the NPG if they did not have recorded pressures >22 mmHg before treatment. As glaucoma patients are long-term patients with several clinic visits each year, for most of the glaucoma cases in this study the IOP measurements were made at least twice on multiple occasions. Exfoliation glaucoma cases had evidence of characteristic fibrillar deposits as previously described in addition to glaucoma as defined above. Controls had normal optic nerves (cup-disc ratios ≤0.6) and normal intraocular pressure (≤21 mm Hg).

22The GLAUGEN dataset included 976 cases and 1,140 controls drawn from three different studies: the Genetic Etiologies of Primary-Open angle Glaucoma (GEP), the Nurses' Health Study (NHS) and the Health Professionals Follow-up study (HPFS). The GEP is a clinic-based case-control set, and the NHS and HPFS are case-control sets nested within population-based studies. Additional information about the case identification and control selection process in GLAUGEN can be found at The Primary Open-Angle Glaucoma Genes and Environment (GLAUGEN) Study. Study Accession: phs000308.v1.p1. www.ncbi.nlm.nih.gov/projects/gap. December 21, 2010. The majority of DNA samples were prepared using Qiagen extraction kits (Invitrogen). More than half of the samples were derived from buccal cells, and we previously demonstrated the feasibility of genotyping buccal cell DNA on the Illumina 660W Quad platform.

23Cases and controls for the NEIGHBOR study were collected from 12 sites (Table S1). Additional information about the NEIGHBOR consortium can be found in reference 17.

24Genotyping for the GLAUGEN and NEIGHBOR samples was completed using the Illumina Human660W_Quad_v1 array. (Illumina, San Diego, CA) at the Broad Institute (GLAUGEN) and the Center for Inherited Diseases Research (CIDR) (NEIGHBOR). At both of the genotyping centers, samples were plated to allow equal representation of cases and controls per plate from each study site in order to minimize batch effects.

25For the GLAUGEN samples genotyping calls were generated using Illumina's BeadStudio, GenomeStudio and Autocall software along with genotype cluster definitions based on study samples. SNPs with a GenTrain score <0.636, cluster separation score <0.4 and call rate <97% were considered technical failures at the genotyping center and were automatically deleted before release for further quality control. Data was released for 2,241 study samples (95% of attempted samples). Subsequent data quality control measures consisted of identifying and removing samples with gender misidentification, unexpected duplicates and unexpected relatedness. Analysis of connectivity removed samples that appeared to be related to other samples and/or suggestive of contamination. Any SNP with missing call rate >2% or with Hardy Weinberg p-value<10−4 in the control population was excluded. Logistic regression analysis indicated that study site (GEP, NHS or HPFS), DNA source (blood or cheek cell), and DNA extraction method (DNAzol, Qiagen or GENTRA) were independent predictors of genotyping call rate for the GLAUGEN samples. Hence, these variables along with age and gender were used in the logistic regression models.

26For the NEIGHBOR samples data from CIDR was released for 5,155 study samples (97% of attempted samples). Study samples, including 117 study duplicates, were plated and genotyped together with 227 HapMap controls (208 CEU; 11 YRI, 4 JPT, 4 CHB). Genotyping was performed using Illumina Human660W-Quadv1_C BeadChips (Illumina, San Diego, CA, USA) and the Illumina Infinium II assay protocol. Allele cluster definitions for each SNP were determined using Illumina GenomeStudio Genotyping Module version 1.7.4, GenTrain version 1.0 and the combined intensity data from 99.9% of the samples. The resulting cluster definitions were used on all samples. Genotypes were not called if the quality threshold (Gencall score) was below 0.15. Genotypes were released by CIDR for 557,029 SNPs (99.58% of attempted). Genotypes were not released for SNPs that had call rates less than 85%, more than 1 HapMap replicate error, cluster separation less than 0.2, more than a 3% (autosomal) or 2.2% (X chromosome) difference in call rate between genders, more than 0.4% (X chromosome) male heterozygosity, or more than a 8% (autosomal) difference in AB frequency. XY, Y and mitochondrial SNPs were manually reviewed and clusters adjusted or genotypes dropped as appropriate. Intensity data was released for all attempted SNPs. The mean non-Y SNP call rate and mean sample call rate was 99.9% for the released CIDR dataset. Study duplicate reproducibility was 99.99%. After applying quality control filters, 495,132 SNPs were analyzed in 976 cases and 1140 controls for GLAUGEN and 523,528 SNPs were analyzed in 2170 cases and 2347 controls for NEIGHBOR.

27Logistic regression to assess the association between individual SNPs and POAG was done using PLINK v1.07. For GLAUGEN, the logistic regression model included age, gender, study site, DNA source, DNA extraction method and 3 eigenvectors (EV 1, 2 and 6). For NEIGHBOR, the logistic regression model included age, gender, study site and 2 eigenvectors (EV1 and 2). Quantile-quantile plots (Figures S1 and S3) were used to estimate genomic inflation factors which were 1.009 for GLAUGEN and 1.034 for NEIGHBOR.

28Combined meta-analysis of the GLAUGEN and NEIGHBOR datasets was done using the METAL software package. We analyzed each study using logistic regression as described previously. Then, we combined the results using the inverse weighted variance method based on the regression coefficients and standard errors estimated from each study as implemented in the program METAL. The GENOMICCONTROL option was set to ON to adjust for genomic inflation differences between the studies.

29Haplotype analysis (logistic regression) was performed with PLINK 1.07 using sliding windows of 2–6 SNPs across the associated regions. Covariates in the model were the same as for the single-allele analyses.

30Pathway analysis was done using the Pathway Analysis by Randomization Incorporating Structure (PARIS) pathway analysis software package. Genes comprising the transforming growth factor beta (TGF-beta) pathway were identified using the KEGG database (hsa:04350). SNPs were considered to reside in a pathway gene if the SNP fell within the ENSEMBLE genomic interval+/−50 kb to either side of the gene. If the overlap included another gene, the overlapping SNP(s) were counted once. A single-allele p-value of <0.05 was considered to be nominally significant and included in the PARIS analysis.

31Focused regional imputation was used to infer genotypes for SNPs not directly genotyped using MACH 1.0, to impute to the 1000 genomes data, release 2010–06. SNPs with a quality score (Rsq) of <0.5 were discarded before analysis. The resulting genotypes were analyzed using PLINK with the same covariates for each dataset as the non-imputed analyses. The meta-analysis procedure was also identical to that described above.

32Retinas from adult mice were isolated, processed, and subjected to laser capture microdissection. cDNA purified from the respective cell layers was synthesized and analyzed by quantitative PCR as previously described.

33Total RNA was extracted from dissected tissues (cornea, trabecular meshwork, retina, optic nerve) from normal human donor eyes as previously described. Primer sequences were designed to specifically amplify the LRP12 and ZFPM2 genes. Amplification products were visualized by gel electrophoresis.

34Targeted resequencing using Sanger methods as described was carried out on genes from the associated loci on chromosome 8 and chromosome 9. All exons of LRP12 and ZFPM2 were sequenced in 16 NPG patients, and all exons of CDKN2BAS were sequenced in 11 cases carrying a protective haplotype including rs2157719 (Table S6). Genomic DNA was sequenced using primers designed to amplify the coding exons as well as the adjacent splice sites for both genes. PCR products were directly sequenced on the ABI PRISM 3100 Genetic Analyzer (Applied Biosystems) with BigDye Terminators (Applied Biosystems) according to standard protocols. Methods and analysis for exome sequencing are described in Text S1.

35Exfoliation cases were collected from the Massachusetts Eye and Ear infirmary glaucoma clinic (cases) and the comprehensive eye service (controls in GLAUGEN or NEIGHBOR). All cases and controls were examined by a board certified ophthalmologist prior to study enrollment. The lead SNPs for each region associated with NPG were analyzed in 196 Caucasian patients with exfoliation syndrome, 104 of who also had glaucoma as defined as above. Single-SNP associations were analyzed using chi-square and logistic regression models adjusting for age.

36PLINK (http://pngu.mgh.harvard.edu/~purcell/plink/); EIGENSOFT (http://genepath.med.harvard.edu/~reich/Software.htm); ENCODE (http://genome.ucsc.edu/ENCODE/); SCAN (http://www.scandb.org/newinterface/about.html); KEGG (http://www.genome.jp/kegg/); UCSC genome (http://genome.ucsc.edu/); ENSEMBL (http://useast.ensembl.org/index.html).