Genetic Association at the 9p21 Glaucoma Locus Contributes to Sex Bias in Normal-Tension Glaucoma

Abstract

Purpose: Many genome-wide association studies have identified common single nucleotide polymorphisms (SNPs) at the 9p21 glaucoma locus (CDKN2B/CDKN2B-AS1) to be significantly associated with primary open-angle glaucoma (POAG), with association being stronger in normal tension glaucoma (NTG) and advanced glaucoma. We aimed to determine whether any observed differences in genetic association at the 9p21 locus are influenced by sex.

Methods: Sex was assessed as a risk factor for POAG for 2241 glaucoma participants from the Australian and New Zealand Registry of Advanced Glaucoma, the Glaucoma Inheritance Study in Tasmania, and the Flinders Medical Centre. A total of 3176 controls were drawn from the Blue Mountains Eye Study and South Australia: 1523 advanced POAG and 718 nonadvanced POAG cases were genotyped along with 3176 controls. We selected 13 SNPs at the 9p21 locus, and association results were subanalyszd by sex for high-tension glaucoma (HTG) and NTG. Odds ratios (ORs) between sexes were compared.

Results: A sex bias was present within advanced NTG cases (57.1% female versus 42.9% male, P = 0.0026). In all POAG cases, the strongest associated SNP at 9p21 was rs1063192 (OR, 1.43; P = 4 × 10⁻¹⁸). This association was stronger in females (OR, 1.5; P = 5 × 10⁻¹³) than in males (OR, 1.35; P = 7 × 10⁻⁷), with a statistically significant difference in female to male OR comparison (P = 1.0 × 10⁻²). An NTG to HTG subanalysis yielded statistically significant results only in females (OR, 1.63; P = 1.5 × 10⁻⁴) but not in males (OR, 1.15; P = 2.8 × 10⁻¹), with a statistically significant difference in female to male OR comparison (P = 1.4 × 10⁻⁴).

Conclusions: This study demonstrated that female sex is a risk factor for developing advanced NTG. The stronger genetic signals at the 9p21 locus among females may contribute at least in part to the observed sex bias for NTG.

Primary open-angle glaucoma (POAG) is the most common type of glaucoma and is characterized pathologically by a progressive loss of retinal ganglion cells with corresponding loss of visual field. The prevalence of POAG in the age group over 40 years is estimated to be 2%–3% among Caucasian populations,^1–3 approximately 6%–7% among Black populations,^4,5 and 3.9% among the Japanese.⁶ 

Although demographic factors such as older age and black race are well known to be associated with an increased risk for POAG, there is no consensus with regards to sex as a risk factor. Results from previous large population-based studies have been inconsistent, with some studies reporting higher prevalence in females,^5,7 whereas others showed higher prevalence in males^8,9 or no association at all.^10–13 The Collaborative Normal-Tension Glaucoma Study Group reported female sex as an independent risk factor for disease progression in normal-tension glaucoma (NTG), a subtype of POAG with no recorded intraocular pressure (IOP) elevation (≤21 mm Hg).¹⁴ A recent meta-analysis, however, reported greater POAG prevalence among males in comparison to age-matched females.¹⁵ 

There are obvious biological and physiological differences between males and females, and these differences are known to affect the incidence and progression of various common diseases in human, notably cardiovascular and autoimmune diseases.¹⁶ These sex differences have typically been attributed to the differences in the sex hormone levels between males and females and the genetic contribution of the sex chromosomes (chromosome X).¹⁷ In glaucoma, there is evidence suggesting that the risk of POAG among females may be influenced by estrogen metabolism and estrogen exposure, both endogenously and exogenously.^18–20 

More recently, the autosomes, shared by both males and females, were also shown to contribute significantly to sex-specific disease differences due to sexual dimorphism in gene regulation and expression between sexes.²¹ The dimorphism in the regulation and expression of genes is likely to explain part of the difference in a gene–environment interaction and also influence phenotypic traits in terms of sex-specific susceptibility to disease.¹⁶ 

With the advent of genome-wide association studies (GWASs), single nucleotide polymorphisms (SNPs) associated with a disease, particularly known to have sex-specific differences, can be systematically analyzed in depth to detect whether disease association is stronger in one sex than the other.²² For instance, certain SNPs in the RELN gene were shown to have significant association for schizophrenia and bipolar disorder in females but not in males.^23,24 POAG is a genetically complex disease, and recent GWASs identified several SNPs within the CDKN2B/CDKN2B-AS1 genes on chromosome 9p21 to have strong and reproducible association especially with the NTG subtype.^25–27 These previous studies did not specifically analyze the association of sex among the SNPs relevant to POAG. In this study, we aim to investigate this locus for the existence of sex effect and differences in association to POAG. 

Patients and Methods

Participants

All participants provided written informed consent, and approval was obtained from the Human Research Ethics Committees of Southern Adelaide Health Service/Flinders University, University of Tasmania and University of Sydney. The study adhered to the tenets of the Declaration of Helsinki. 

Participants were drawn from the Australian & New Zealand Registry of Advanced Glaucoma, the Glaucoma Inheritance Study in Tasmania, the Blue Mountains Eye Study (BMES, a population-based study of residents 49 years of age and older living in the Blue Mountains region west of Sydney), and patients attending eye clinics at Flinders Medical Centre, Adelaide, Australia. All participants were Australian of European descent. The cohorts and clinical definitions are as described in detail in earlier reports.^25,28,29 

Briefly, advanced glaucoma was defined by severe visual loss resulting from POAG. This included best-corrected visual acuity worse than 6/60 resulting from POAG or a reliable 24-2 Humphrey Visual Field with a mean deviation (MD) of worse than −22db or at least two of four central fixation squares affected with a pattern standard deviation of less than 0.5%. The field loss had to be the result of POAG, and the less severely affected eye also was also required to have signs of glaucomatous disc damage. Less severe or nonadvanced glaucoma was defined by concordant findings of typical glaucomatous visual field defects on the Humphrey 24-2 test, with corresponding optic disc rim thinning, including an enlarged vertical cup-to-disc ratio (VCDR) (≥0.7) or VCDR asymmetry (≥0.2) between the two eyes. The age at glaucoma diagnosis, highest recorded IOP, central corneal thickness (CCT), and MD in each eye were obtained from the medical records. For each variable (IOP, CCT, MD, and VCDR), the data from the worse eye were used. Any participants without sex or genotype data were excluded. Participants with any form of secondary glaucoma or mutations in the myocilin gene were also excluded. 

Controls were drawn from the BMES and unaffected participants from South Australia.²⁸ All controls were examined and found to have no sign of glaucoma. A total of 2742 elderly participants from the BMES and 434 participants from South Australia were included. Parameters obtained from the controls included age, genotype data, IOP, CCT, and VCDR. 

Genotyping and Association Analysis

Genotyping of the 13 SNPs at the 9p21 locus has been described previously.^25,28 The SNPs chosen were those from our previous GWASs and from targeted genotyping at the 9p21 locus. Briefly, samples used in the discovery phase of the reported GWAS were genotyped on the Human1M-Omni array (Illumina, Inc., San Diego, CA, USA), and samples used in the replication phase of the GWAS were typed on the MassArray platform (Sequenom, Inc., San Diego, CA, USA). The controls were genotyped on Illumina HumanHap 610W Quad and Illumina Human670Quad Bead arrays (BMES) or by MassArray (others). 

Data including sex, age at diagnosis, highest recorded IOP, POAG subtype (NTG or HTG), and CCT were gathered from each participant where possible. Every SNP was analyzed for genetic association. The genetic association analyses were conducted using Plink (Plink version 1.07, 10 August 2009. Purcell S. Available at: http://pngu.mgh.harvard.edu/purcell/plink/. Accessed August 2015).³⁰ Initial analyses were conducted comparing POAG to controls in males, females, and both sexes combined. The same analyses were then conducted in advanced POAG cases only. We then ran separate association analyses in advanced cases of NTG (IOP ≤ 21 mm Hg) and HTG (IOP > 21 mm Hg) for each sex and combined. To test for the effect of sex on the association at the 9p21 SNPs, the obtained odds ratios (ORs) were compared between the sexes by computing  

 

 

P values were computed based on T following a distribution. Using a Bonferroni correction, a P value of 0.004 was required to account for the multiple testing of the 13 SNPs (in practice, due to the correlation between these SNPs, this threshold may be overconservative).  

Results

Overall, there were a total of 2241 cases of POAG and 3176 controls with sex and genotype data available. Among the POAG cases, 1180 (52.66%) were females and 1061 (47.34%) were males, whereas there were 1793 (56.5%) females to 1383 (43.5%) males among the controls (Table 1). The POAG cohort had a mean age of glaucoma diagnosis of 60.6 ± 14.3 years. The mean highest documented IOP was 27.1 ± 11.2 mm Hg, with 66.8% having the highest recorded IOP of >21 mm Hg. A total of 1523 (68%) were classified as having advanced disease, with 744 (48.85%) males and 779 (51.15%) females (P = 0.37). Details of the demographic data for both the POAG cohort and controls are shown in Table 1. A notable sex bias was present within advanced NTG cases (57.1% female versus 42.9% male, P = 0.0026), but not in HTG cases (48.1% female versus 51.9% male, P = 0.24; Table 2). 

Table 1

 

Demographics and Clinical Characteristics of the POAG Cases and Controls

 

Table 2

 

Nonadvanced and Advanced POAG

 

On association analysis conducted for all 13 SNPs of 9p21 among all POAG cases, 4 SNPs reached genome-wide significance (P < 5 × 10⁻⁸): rs1063192 (P = 3.76 × 10⁻¹⁸), rs4977756 (P = 1.97 × 10⁻¹⁶), rs10120688 (P = 6.99 × 10⁻¹¹), and rs3731239 (P = 5.63 × 10⁻¹⁰) (Supplementary Table). Table 3 shows the association results stratified by sex for the top four SNPs. The top three SNPs, namely rs1063192, rs4977756, and rs10120688, reached genome-wide significance only in females but not in males. The OR difference between females and males was statistically significant for rs1063192 (P = 1.04 × 10⁻²), rs4977756 (P = 1.37 × 10⁻⁴), and rs3731239 (P = 2.40 × 10⁻³) (Table 3). 

Table 3

 

Sex Comparison for Top Four SNPs in Association Analyses for 2232 POAG Cases

 

A strong association was observed when the analyses were conducted comparing only advanced cases to the controls (Table 4). In females, the observed association was stronger in the advanced cases than in overall POAG. This trend, however, was not observed among the males, which showed comparable ORs and significance levels in both advanced and overall POAG cases (Table 4). 

Table 4

 

Association Analyses Comparing All POAG and Advanced POAG by Sex

 

The NTG subgroup was then compared directly to the HTG subgroup within advanced POAG, as this locus is known to be more strongly associated with NTG than HTG.^25,31 Three SNPs (rs1063192, rs4977756, and rs10120688) showed statistically significant association to NTG (when applying a conservative Bonferroni correction for 13 tests at 9p21) when both sexes were analyzed together. The risk allele A of SNP rs1063192 carries an OR of 1.40 (P = 2.46 × 10⁻⁴) for developing NTG (Table 5). Marked sex differences were again observed when the analyses were conducted separately for females and males (Table 5). Among the females, these SNPs were significantly associated with NTG, yielding ORs of 1.63 for rs1063192, 1.60 for rs4977756, and 1.62 for rs10120688 (Table 5). On the other hand, in males with NTG, these same SNPs carried weaker ORs of 1.15 for rs1063192 and 1.22 for both rs4977756 and rs10120688 and did not reach statistical significance. The OR difference between females and males was statistically significant (Tables 3 and 5). 

Table 5

 

Association Analyses Comparing NTG With HTG by Sex Among Advanced POAG Cases Only.

 

Discussion

The association between SNPs at chromosome 9p21 and POAG has been widely established in multiple populations.^25–28,31 As previously noted, the association was significantly stronger among the NTG subgroup and also among the advanced blinding cases.^25,32,33 Our current study highlights that the strength of the association also varies markedly depending on sex. The data consistently showed that the association of these known glaucoma risk alleles at chromosome 9p21 with POAG is stronger in females than in males. These sex differences in the strength of association have not been previously reported among SNPs at the 9p21 locus. 

The differences in the strength of association between sexes noticeably increased among the advanced POAG cases. It is well recognized that POAG progresses with increasing age, and therefore, the advanced cases are more frequently documented among the older age group.¹⁵ Nevertheless, the underlying reason for the observed stronger association of the chromosome 9p21 risk alleles particularly among female advanced POAG cases in comparison to the male counterpart is unclear. Whether females with these risk alleles have higher risk of progression to advanced POAG remains to be elucidated. Also, the risk alleles of the three main SNPs, rs1063192, rs4977756, and rs10120688, conferred a statistically significant OR ranging from 1.60 to 1.63 with NTG among females only. The association with NTG did not reach statistical significance among males, and the ORs were substantially weaker (Table 5). The observation also suggests that this locus may confer a greater risk for NTG among females. 

Sex-specific differences in disease susceptibility, course, and severity have long been known, notably in cardiovascular and autoimmune diseases. Several SNPs within chromosome 9p21 namely rs2383207, rs4977574, rs10757274, rs10116277, and rs1333040 are also known to have significant association with coronary heart disease.^34,35 In our current analyses (Supplementary Table), the SNP rs2383207 conferred an OR of 1.39 (P = 9.1 × 10⁻⁷) among females and 1.19 (P = 1.4 × 10⁻²) among males for advanced POAG, almost reaching the level of genome-wide significance among females alone. The other four SNPs, however, did not show any suggestive association with POAG, consistent with previous studies.^25–27 The lack of association between relevant SNPs at chromosome 9p21 that confer risk for POAG and those for cardiovascular diseases is in accordance with the fact that there have been no clear association between POAG and cardiovascular diseases.^36,37 

In POAG overall, sex influences have not been strongly established previously. A recent meta-analysis by Tham et al. reported greater prevalence among males, with an OR of 1.36 (95% confidence interval [CI], 1.23–1.52).¹⁵ Several studies have also highlighted the protective effect of estrogen and linked the risks of POAG among females with estrogen metabolism and exposure.^18–20 Lower IOP was shown to be associated with postmenopausal hormonal use and pregnancy, during which estrogen levels are elevated.^18,38,39 Early menopause (age less than 45 years)¹⁹ was associated with increased risk of POAG, whereas later onset of menopause (age more than 54 years) was associated with reduced risk.²⁰ In the BMES, later age of menarche was also found to be associated with POAG.⁴⁰ Pasquale et al., however, reported that later age of menarche was associated with NTG only.⁴¹ Overall evidence suggests that the risk of POAG is somehow inversely related to the cumulative exposure of estrogen. Meta-analyzed GWAS data also suggested an association between the estrogen SNPs pathways and POAG among females.⁴² Although there are several proposed hypotheses, the exact pathophysiology of any putative protective effect of estrogen against POAG is still unknown. 

There may be other biological mechanisms underlying the observed sex specificity such as epigenetic differences between the sexes. Sexual dimorphism in gene regulation and expression possibly mediates the differences in genotype–environmental interactions, which could subsequently lead to sex-specific susceptibility to POAG.^16,21 Sex-genetic specificity has been observed in several other diseases.^43,44 In hypertension, angiotensin-converting enzyme DD genotypes were significantly associated with hypertension in males only.^43,44 In schizophrenia and bipolar disorder, SNPs in the RELN gene were shown to have a significant association in females only.^23,24 The protein product of RELN, reelin is implicated in neuronal migration and has been shown to be essential for retinogeniculate targeting by retinal ganglion cells.^45,46 Also, the SNP rs7865618 on chromosome 9p21, known to have significant association with coronary artery disease, was recently shown to be male specific,²² explaining at least in part the male bias in the incidence of coronary artery disease.⁴⁷ 

One of the limitations of this study is that the result was obtained from a single population cohort (Australian of European descent). Our current study, however, comprises a large number of both POAG cases (2241) and controls (3176), and the analyses showed a notable sex difference in the strength of association of glaucoma risk alleles at chromosome 9p21, especially in the NTG and advanced disease. Future replication studies will confirm and strengthen these findings. Second, some phenotypic data were not available from the nonadvanced POAG group, and for a subset of the samples, we used different genotyping methods that could potentially introduce artifacts. However, the call rates (a good proxy for genotyping accuracy) for the SNPs were high irrespective of genotyping platform, and we believe our results to be robust. Our previous publications on POAG used a mixture of genotyping methods, and those findings have now been replicated by other groups,⁴⁸ giving us confidence that our findings here are robust. Third, the definition and diagnosis of NTG is often difficult in a cross-sectional population. Many of the diagnosed NTG patients would not have been subjected to phasing; hence, the highest recorded IOP in a clinical setting may not necessarily reflect the highest IOP. 

In summary, the results of this study demonstrate a stronger association of the POAG relevant SNPs at chromosome 9p21 in females compared with males, particularly in the NTG and advanced disease. This genetic association would at least in part contribute to the observed significant sex bias for advanced NTG. Although the exact reason underlying such observations remains to be determined, we prompt other researchers performing GWAS to conduct additional analyses to specifically test for potential sex effects. 

Acknowledgments

Supported by the National Health and Medical Research Council (NHMRC) of Australia (Grants 535074, 577074, 974159, 211069, and 457349); an NHMRC career development award (KPB); an NHMRC practitioner fellowship (JEC); and a Royal Australian and New Zealand College of Ophthalmology (RANZCO) Eye Foundation grant. The Blue Mountain Eye Study was supported by the NHMRC (Grants 974159, 211069, 457349, 512423, 475604, and 529912). 

The sponsor or funding organization had no role in the design or conduct of this research. 

Disclosure: S.K. Ng, None; K.P. Burdon, None; J.T. Fitzgerald, None; T. Zhou, None; R. Fogarty, None; E. Souzeau, None; J. Landers, None; R.A. Mills, None; R.J. Casson, None; B. Ridge, None; S.L. Graham, None; A.W. Hewitt, None; D.A. Mackey, None; P.R. Healey, None; J.J. Wang, None; P. Mitchell, None; S. MacGregor, None; J.E. Craig, None