No Sex Differences in the Frequencies of Common Single Nucleotide Polymorphisms Associated with Age-Related Macular Degeneration

Abstract

Purpose:

Since some studies have reported differences in the association of age-related macular degeneration (AMD) with biological sex, we set out to determine whether the difference in the disease susceptibility is afforded by common single nucleotide polymorphisms (SNPs) associated with AMD.

Methods:

We genotyped 2067 Caucasian subjects from the Age-Related Eye Disease Study cohort for commonly associated AMD SNPs, including those in CFH (rs1061170, rs1410996, and rs3766404), ARMS2 (rs10490924), and C3 (rs2230199) using either a Sequenom MassARRAY MALDI-TOF mass spectrometer or using Taqman genotyping reagents. A Cox proportional hazards model was used to determine the effect of genotype, age, sex, and smoking status on the development of AMD.

Results:

All tested SNPs genotyped are associated strongly with AMD (p < 0.0001), in concordance with previous studies. However, we found no observable differences in any of the SNPs studied when categorized by sex. Interactions between SNPs and sex were found to be not statistically significant (p = 0.38–0.79).

Conclusions:

The difference between male and female incidence of AMD is not explained by the most commonly AMD-associated SNPs, though it does not exclude the possibility that other, less common SNPs contribute to this difference.

Introduction

Age-related macular degeneration (AMD) is the leading cause of central irreversible blindness in the elderly, though estimates of the prevalence, or percent of a population with AMD, have varied widely with a range of approximately 2–10% of the overall population.¹ The most current estimates based on a meta-analysis of 39 studies in varying ethnic groups show a prevalence of 8.69% with a predicted 196 million people affected worldwide by 2020.² For people of European/Caucasian descent, AMD prevalence increases to 12.3%, greater than people of Hispanic (10.4%), Asian (7.4%), or African descent (7.5%).² Many past population-based studies have noted a difference in the distribution of AMD prevalence according to sex; that is, females seem to have a slightly higher chance of developing the disease than males, particularly for the late-stage neovascular form of the disease.^1,3,4 A recent meta-analysis of 14 publications on AMD prevalence rates in Caucasians used a Bayesian risk model to show that women were 38% more likely to develop late AMD than men when taking mortality rates into account.⁵ However, other studies have found either no difference or even higher risk in males than females.^2,6

Multiple cofounders known to affect AMD pathogenesis have been suggested for explanations to the discrepancy between studies on AMD prevalence in females and males, including increased longevity and healthcare usage by females, increased smoking in males, underpowered studies unable to detect true differences in prevalence, and the role of exogenous estrogen in AMD pathogenesis.^1,3,4,7–11 A potential unexplored source of sex differences may lie in the associations of different single nucleotide polymorphisms (SNPs) with sex.

Many SNPs have been previously associated with AMD, but they do not explain the entire variance of the disease, as environmental and epigenetic effects have also been shown to play a role in pathogenesis.^12,13 Most notably, SNPs in certain genes, notably CFH, ARMS2/HTRA1, C2/CFB, and C3, strongly associate with AMD pathogenesis and contribute a large risk to developing the disease.^14–20 SNPs with smaller effects and rare allele SNPs in other genes have also been described through both case-control and genome-wide association studies.^21–28 We hypothesized that, due to the sex discrepancy in AMD prevalence and incidence, that there may be a SNP–sex interaction wherein certain SNPs lead to AMD in one sex more frequently than the other. SNP–sex interactions have been described previously for other disease-associated SNPs.^29–32 We chose the most common SNPs associated with AMD to evaluate whether or not SNP–sex interactions might play a role in AMD pathogenesis and prevalence.

Materials and methods

Study subjects

All research followed the tenets of the Declaration of Helsinki. Participants were recruited from 11 retinal specialty clinical centers from the United States for a longitudinal study on the development and progression of AMD in a general population.³³ All center IRBs approved the protocol. Informed consent was given by all subjects, whose group information is summarized in Table 1. Methods for participant selection and evaluation in the Age-Related Eye Disease Study (AREDS) have been previously described.^34,35 Because few ethnicities other than Caucasian were represented in large numbers in AREDS and the genetic diversity of AMD between different ethnic groups, we limited our analysis to subjects of Caucasian descent. Briefly, AMD patients and control subjects from both studies were evaluated clinically using the AREDS AMD Classification.³⁴ AMD cases [both geographic atrophy (GA) and neovascular/exudative (CNV)] and controls were determined by centralized grading of stereoscopic fundus photographs obtained at baseline and annual study visits. All non-AMD participants (AMD category 1) presented with fewer than five small drusen (less than 63 μm in diameter) with no signs of other retinal diseases. AMD category 2 consisted of participants with medium sized drusen (63–125 μm in diameter), extensive small drusen, or pigmentary abnormalities. AMD category 3 participants showed signs of late AMD, including at least one large druse (greater than 125 μm in diameter), extensive medium sized drusen, central GA, or CNV. AMD category 4 participants showed extensive signs of late AMD.

Table 1.

Baseline demographics in Age-Related Eye Disease Study (AREDS) participants (N = 2067) by the development of late age-related macular degeneration (AMD).

	Development of late AMD during follow-up
	No		Yes
	N	%	N	%
Total	1643	100.0	424	100.0
Sex
Female	935	56.9	258	60.8
Male	708	43.1	166	39.2
Smoking status
Never	850	51.7	182	42.9
Former	730	44.4	213	50.2
Current	63	3.8	29	6.8
AREDS AMD category 1	594	36.2	9	2.1
2	538	32.7	31	7.3
3	496	30.2	337	79.5
4	15	0.9	47	11.1
Median age, y	67.8		69.9

Abbreviations: late AMD, late age related macular degeneration, either neovascular or central geographic atrophy.

Subject genotyping

DNA isolated from the whole blood of Caucasian AREDS participants was genotyped for five SNPs in three genes previously associated with AMD: rs1061170, rs1410996, and rs3766404 in CFH, rs10490924 in ARMS2, and rs2230199 in C3.^{16,18,25,28,36,37} All SNPs except for rs1061170 were genotyped using the MassARRAY MALDI-TOF mass spectrometer and software (Sequenom, San Diego, CA, USA). rs1061170 was genotyped using Taqman SNP Genotyping Assay (Applied Biosystems, Foster City, CA, USA) under standard cycling conditions in a PTC-225 Thermal Cycler (MJ Research, St. Bruno, Quebec, Canada).

Statistical analysis

SNP association and sex interaction analysis was carried out using the PHREG procedure in SAS version 9.3 (Cary, NC, USA). The progression to late AMD was analyzed by genotypes and sex using a Cox proportional hazards model and was adjusted for age and smoking status at baseline. The late AMD subtype analysis (GA or CNV) was performed in the same manner. A separate association of genotype on AMD progression when separated by sex was also performed. Type 3 tests to determine the statistical significance of a contributing effect were used, and were considered significant if p < 0.005 after correction for multiple comparisons.

Results

Five genotyped SNPs are associated with AMD, but do not show sex interactions

Median age, sex, smoking status, and AMD subtypes of the 2067 AREDS participants categorized by the presence of AMD at follow-up are listed in Table 1. During the follow-up, 424 participants (21.64% of females vs. 19.01% of males) developed late AMD (Table 1). All participants were genotyped; however, only 1998 of the 2067 participants yielded readable genotypes for all five tested SNPs and were used in further analysis.

The distribution of CFH rs1061170, rs1410996, and rs3766404; ARMS2 rs10490924; and C3 rs2230199 SNPs stratified by sex are summarized in Table 2. No large differences in the percentage of males vs. females with a particular genotype were seen. After performing a Cox regression analysis on the data as a whole, as expected, each SNP continued to associate strongly with AMD, as did age and smoking status (p < 0.001, Table 3). However, none of the sex-SNP interaction terms were significant, indicating no difference between males and females for each SNP in the development of AMD (p > 0.005, Table 3). Further, the computed hazard ratios for each SNP effect on males vs. females showed no significant differences (p > 0.005, Figure 1). To ensure that each SNP associated with AMD in males and females, we then analyzed the data separately for sex; nearly every genotype of the five SNPs tested showed an association with progression of AMD in both males and females (p < 0.005, Table 4). However, the CG genotype for C3 rs2230199 and CC genotype for CFH rs3766404 were not significantly associated with AMD progression in either males or females (p > 0.005, Table 4).

Table 2.

Distribution of the five commonly associated SNPs among participants stratified by progression to late AMD and sex.

		Female				Male
		AMD during follow-up				AMD during follow-up
		No		Yes		No		Yes
	Total	N	%	N	%	N	%	N	%
Total	1998	906	78.6	246	21.4	684	80.9	162	19.1
CFH rs1061170
CC	440	155	61.0	99	39.0	120	64.5	66	35.5
CT	917	413	78.8	111	21.2	316	80.4	77	19.6
TT	641	338	90.4	36	9.6	248	92.9	19	7.1
CFH rs1410996
AA	273	161	97.0	5	3.0	102	95.3	5	4.7
AG	857	416	85.2	72	14.8	323	87.5	46	12.5
GG	868	329	66.1	169	33.9	259	70.0	111	30.0
CFH rs3766404
TT	1547	679	75.4	221	24.6	506	78.2	141	21.8
CT	418	212	89.8	24	10.2	163	89.6	19	10.4
CC	33	15	93.8	1	6.3	15	88.2	2	11.8
ARMS2 rs10490924
GG	1063	527	87.7	74	12.3	403	87.2	59	12.8
GT	752	315	71.8	124	28.2	236	75.4	77	24.6
TT	183	64	57.1	48	42.9	45	63.4	26	36.6
C3 rs2230199
CC	1180	548	82.5	116	17.5	428	82.9	88	17.1
GC	719	320	74.9	107	25.1	233	79.8	59	20.2
GG	99	38	62.3	23	37.7	23	60.5	15	39.5

Abbreviations: AMD, age related macular degeneration; CFH, complement factor H; ARMS2, age-related maculopathy susceptibility 2; C3, complement factor 3.

Table 3.

The Cox proportional hazards regression (Type 3 p-values) for each of the five tested SNPs describing the overall effect of each covariate on development of late AMD.

			Type 3 P-value
	Age	Sex	Smoking status	SNP	SNP–Sex interaction
CFH rs1061170	<0.0001	0.2560	<0.0001	<0.0001	0.7896
CFH rs1410996	<0.0001	0.5786	<0.0001	<0.0001	0.6360
CFH rs3766404	<0.0001	0.0599	<0.0001	<0.0001	0.6563
ARMS2 rs10490924	<0.0001	0.7429	<0.0001	<0.0001	0.6793
C3 rs2230199	<0.0001	0.9581	<0.0001	<0.0001	0.3847

Abbreviations: SNP, single nucleotide polymorphism; CFH, complement factor H; ARMS2, age-related maculopathy susceptibility 2; C3, complement factor 3.

Figure 1.

Computed hazard ratios with 95% confidence interval for males vs. females after Cox proportional hazards regression for each level of the five tested SNPs.

Table 4.

The Cox proportional hazards regression for each of the five tested SNPs done separately for females and males for the development of late AMD.

			Females			Males
		Hazard ratio (95% confidence interval)		p	Hazard ratio (95% confidence interval)		p
CFH rs1061170 (ref = CC)	CT	0.52	(0.40–0.67)	<0.0001	0.53	(0.38–0.73)	0.0001
	TT	0.22	(0.15–0.31)	<0.0001	0.18	(0.11–0.30)	< 0.0001
CFH rs1410996 (ref = AA)	AG	5.37	(2.17–13.27)	0.0003	2.84	(1.13–7.15)	0.0266
	GG	13.28	(5.45–32.32)	<0.0001	7.14	(2.92–17.50)	< 0.0001
CFH rs3766404 (ref = TT)	CT	0.38	(0.25–0.58)	<0.0001	0.47	(0.29–0.76)	0.0020
	CC	0.22	(0.03–1.60)	0.1360	0.59	(0.15–2.40)	0.4632
ARMS2 rs10490924 (ref = GG)	GT	2.53	(1.91–3.35)	<0.0001	2.10	(1.50–2.95)	< 0.0001
	TT	4.06	(2.84–5.81)	<0.0001	3.32	(2.10–5.24)	< 0.0001
C3 rs2230199 (ref = CC)	CG	0.67	(0.44–1.01)	0.0583	0.50	(0.28–0.88)	0.0169
	GG	0.42	(0.28–0.64)	<0.0001	0.41	(0.24–0.71)	0.0016

Variables included in model: age, smoking status, and genetic variant.

Abbreviations: SNP, single nucleotide polymorphism; CFH, complement factor H; ARMS2, age-related maculopathy susceptibility 2; C3, complement factor 3.

The genetic association profiles of GA and CNV, the two subtypes of late AMD, differ in the associated genes and conferred risk. We performed supplementary subtype analysis of the five SNPs for GA and CNV separately. For CNV, like in the late AMD analysis, the Type 3 p-values for sex-SNP interactions were not significant for any of the SNPs tested (p > 0.005), suggesting that sex did not differentially influence CNV development. Notably, no females presented with the CC variant for CFH rs3766404, so subtype analysis could not be performed on that SNP (Table 5). Further, with the CNV cases, most alleles associated equally with AMD across the sexes, except for the AG genotype of CFH rs1410996 (female: p = 0.0018, male: p = 0.2607) and the CG genotype of C3 rs2230199 (female: p = 0.3674, male: p = 0.0030, Table 5).

Table 5.

The Cox proportional hazards regression for each of the five tested SNPs done separately for females and males for the development of CNV.

			Females			Males
		Hazard ratio (95% confidence interval)		p	Hazard ratio (95% confidence interval)		p	SNP–Sex interaction
CFH rs1061170 (ref = CC)	CT	0.59	(0.44–0.80)	0.0005	0.43	(0.30–0.63)	< 0.0001	0.3620
	TT	0.22	(0.14–0.34)	< 0.0001	0.16	(0.09–0.30)	< 0.0001
CFH rs1410996 (ref = AA)	AG	6.39	(2.00–20.42)	0.0018	1.73	(0.67–4.47)	0.2607	0.2364
	GG	16.79	(5.35–52.74)	< 0.0001	5.09	(2.07–12.54)	0.0004
CFH rs3766404 (ref = TT)	CT	–	–	–	0.40	(0.21–0.74)	0.0037	0.9948
	CC	–	–	–	0.85	(0.21–3.46)	0.8240
ARMS2 rs10490924 (ref = GG)	GT	2.18	(1.59–2.98)	< 0.0001	2.59	(1.72–3.90)	<0.0001	0.8020
	TT	3.57	(2.40–5.31)	< 0.0001	3.88	(2.25–6.71)	<0.0001
C3 rs2230199 (ref = CC)	CG	0.79	(0.48–1.32)	0.3674	0.40	(0.22–0.73)	0.0030	0.1960
	GG	0.51	(0.31–0.84)	0.0086	0.32	(0.18–0.58)	0.0001

Variables included in model: age, smoking status, and genetic variant.

Abbreviations: SNP, single nucleotide polymorphism; CNV, choroidal neovascularization; CFH, complement factor H; ARMS2, age-related maculopathy susceptibility 2; C3, complement factor 3.

For GA, the Type 3 p-values for sex-SNP interactions were similarly insignificant for all SNPs tested (p > 0.0388, Table 6). Of note, the AG genotype of CFH 1410996, CC genotype of CFH rs3766404, and both the CG and GG genotypes of C3 rs2230199 did not associate with GA in either males or females (p > 0.005, Table 6).

Table 6.

The Cox proportional hazards regression for each of the five tested SNPs done separately for females and males for the development of GA.

			Females			Males
		Hazard ratio (95% confidence interval)		p	Hazard ratio (95% confidence interval)		p	SNP–Sex interaction
CFH rs1061170 (ref = CC)	CT	0.47	(0.35–0.65)	< 0.0001	0.55	(0.39–0.78)	0.0009	0.5585
	TT	0.29	(0.19–0.43)	< 0.0001	0.23	(0.13–0.39)	< 0.0001
CFH rs1410996 (ref = AA)	AG	4.21	(1.53–11.64)	0.0055	3.92	(1.21–12.66)	0.0224	0.9963
	GG	9.44	(3.49–25.52)	< 0.0001	8.67	(2.75–27.36)	0.0002
CFH rs3766404 (ref = TT)	CT	0.44	(0.27–0.72)	0.0010	0.45	(0.26–0.76)	0.0031	0.9909
	CC	0.39	(0.05–2.76)	0.3428	0.36	(0.05–2.59)	0.3109
ARMS2 rs10490924 (ref = GG)	GT	4.03	(2.77–5.88)	< 0.0001	2.00	(1.37–2.91)	0.0003	0.0388
	TT	6.01	(3.91–9.24)	< 0.0001	3.95	(2.48–6.28)	< 0.0001
C3 rs2230199 (ref = CC)	CG	0.97	(0.56–1.68)	0.9204	0.73	(0.38–1.40)	0.3417	0.4876
	GG	0.60	(0.35–1.04)	0.0708	0.60	(0.32–1.13)	0.1103

Variables included in model: age, smoking status, and genetic variant.

Abbreviations: SNP, single nucleotide polymorphism; GA, geographic atrophy; CFH, complement factor H; ARMS2, age-related maculopathy susceptibility 2; C3, complement factor 3.

Discussion

Sex-based differences refer to X chromosomal gene dosage differences, hormonal changes, disease susceptibility, epigenetics, environment, and other physiological differences between women and men. Differences in disease incidence, presentation, and progression between males and females are common, particularly with respect to the immune system. Many etiologies have been proposed to explain this parity, including differential effects of the microbiome on immune system priming, the effect of incomplete X inactivation on immune gene dosage, and altered immune responses in men compared to women.^32,38–42 Other factors, such as epigenetic modifications (DNA and histone methylation, chromatin remodeling, and non-coding RNAs), are responsible for promoting sexual dimorphism in the brain and could contribute to various neurological diseases.^43,44 With respect to blindness, women are disproportionally affected according to a meta-analysis of 44 prevalence studies.⁴⁵ In total, 64.5% of blind people surveyed were female, and the overall hazard ratio was 1.43. In older populations, the discrepancy was more pronounced, potentially suggesting increased loss of vision in females due to aging-related diseases. In general, women are thought to be at a greater risk of developing AMD.^1,3–5 Indeed, a recent 15-year study on the incidence of AMD in Australia found that women are more likely to develop early AMD than males (OR: 1.52, 95% CI: 1.11–2.08) after multivariable logistic regression, though they are not more likely to advance to later stages of AMD.⁴⁶ Two of the SNPs tested in that study, CFH rs1061170 and ARMS2 rs10490924 were also independently associated with AMD.⁴⁶ The discrepancy in AMD incidence for females may contribute to the sex differences in blindness seen in other studies.⁴⁵

Another important factor in AMD development is ethnicity. Generally, Caucasians are at the highest risk of developing AMD in their lifetime.^2,47 As well, female Caucasians are at higher risk than their male counterparts; however, this is not necessarily the case in all ethnicities. Three large cohort studies of Japanese participants and the Los Angeles Latino Eye Study found that men in these populations were more likely to develop late AMD than women.^48–51 However, it is unclear to what extent genetic, behavioral, or environmental differences between Caucasians and other ethnicities contribute to the sex differences in AMD prevalence found in these populations. Further, while environmental factors, such as increased smoking rate, play a significant role on the effects of different SNPs on the development of AMD, SNP effects are also modulated by ethnicity.⁵² Some SNPs associated with AMD in Caucasians do not replicate or contribute less risk in other ethnic groups.^52,53 These findings suggest multiple complementary genetic contributions may interact with environmental factors to converge on AMD pathology.⁵⁴

In our study, we replicated the previous findings that SNPs in CFH (rs1061170, rs1410996, and rs3766404), ARMS2 (rs10490924), and C3 (rs2230199) associate strongly with AMD in a Caucasian cohort.^{14–20,25,26,36,37} However, we did not find a significant SNP–sex interaction in our cohort for these common variations with either development of late AMD or either of its subtypes, GA or CNV. This finding suggests a lack of sex-dependent associations of major SNP variations in CFH, ARMS2, and C3 with AMD globally. On more detailed analysis, two genotypes showed differential significance between males and females with CNV (CFH rs1410996 AG and C3 rs2230199 CG); however, because the hazard ratios overlap between males and females in these genotypes, it is likely that this is artifact of the low number of patients with these genotypes who develop CNV and is a product of a lack of power to determine significance in this subtype analysis. Our negative findings do not preclude the potential of other SNPs to have sex interactions, particularly from moderate risk or rare alleles that would require larger sample sizes to detect a significant SNP–sex interaction. Further, because the effects of genetic, epigenetic, and environmental factors in AMD are not fully understood, further research to explain the difference in AMD prevalence and incidence according to sex is warranted.