A Genetic Contribution to Intraocular Pressure: The Beaver Dam Eye Study

Abstract

purpose. To investigate a potential genetic contribution to intraocular pressure (IOP), we performed a complex segregation analysis on 2337 individuals in 620 extended pedigrees ascertained through a population-based cohort, the Beaver Dam Eye Study (BDES). IOP is a principal risk factor for primary open-angle glaucoma (POAG) a leading cause of blindness worldwide.

methods. Segregation analysis is an analytical method that provides statistical evidence supporting the involvement of a major gene or polygenes in a particular phenotype. Detailed medical histories and eye examinations were performed on all participants. From the two eyes, the higher IOP measurement was used as a continuous trait after adjustment for covariates. A genome-wide scan (GWS) using affected sib pair linkage analysis was performed on 218 sibling pairs.

results. In this segregation analysis the model that allowed for an unmeasured major environmental effect plus a polygenic/multifactorial effect provided the best fit and was the most parsimonious model. The lack of an adequate fit for the Mendelian single-gene models is consistent with a multifactorial model of inheritance that may include multiple genes and environmental factors that contribute to IOP. The results of the GWS yielded two novel loci as potential linkage regions for IOP on chromosomes 6 (P = 0.008) and 13 (P = 0.0007). Neither of these regions has previously been identified in GWS of POAG.

conclusions. The segregation and familial correlation analyses of IOP suggest a polygenetic component with environmental influences. The pilot linkage study further confirms the heterogeneity of IOP with the identification of two novel genetic loci.

Primary open-angle glaucoma (POAG) is a leading cause of blindness in the world and affects 1% to 2% of the population over age 40. ¹ Although the disease etiology of POAG is unclear, studies suggest that there is a genetic component to the disease. Numerous genetic loci for POAG have been mapped through linkage analyses, although they have not all been confirmed. These include regions 1q23, 2cen-q13, 3q21-q24, 8q23, 10p15-p14, and 7q35-q36, ^{2 3 4 5 6} identified in extended families with primarily autosomal dominant inheritance. A genome-wide sib pair linkage analysis of POAG identified suggestive linkage (lod score ≥ 2.0) to loci on chromosomes 2, 14, 17 (short arm), and 19, using 113 sibling pairs. ⁷ Recently, a genome-wide scan of an Afro-Caribbean population provided evidence for linkage to POAG on 2q and 10p. ⁸ Two loci, TIGR/myocilin (1q23) and OPTN/optineurin (10p14) have been identified ^{9 10} and >43 mutations have been identified in myocilin in patients with POAG. ¹¹ These studies support POAG as a genetically heterogeneous disorder. 

Elevated intraocular pressure (IOP) is a principal risk factor for POAG and results in optic nerve damage. In the United States, it is estimated that 3 to 6 million people have elevated IOP and are at an increased risk for development of glaucoma. ¹² The goal of this study was to investigate a potential genetic contribution to IOP, a predictor of POAG. We performed a complex segregation analysis on 2336 individuals in 620 extended pedigrees ascertained through a population-based cohort, the Beaver Dam Eye Study (BDES) in Beaver Dam, Wisconsin. In addition, we performed a nonparametric genome-wide linkage analysis of 218 sibling pairs from the BDES. 

Methods

Study Population

A full description of the methods used to identify the population and a description of the participants in the BDES has been published. ^{13 14} From 1988 to 1990, the BDES investigators conducted baseline examinations for recruitment into the study. Eligibility requirements for participation in the study included living in the town of Beaver Dam, Wisconsin, and being 43 to 84 years of age. Of the 5924 individuals eligible to enroll, 4926 individuals participated in the baseline examination. Family relationships were ascertained from participants, and pedigrees were constructed. From 1993 to 1995, the first follow-up visit was conducted and family relationships were confirmed. Of the 5924 eligible individuals, 2783 individuals had available information on familial relationships. Of these 2783 individuals, 2331 had complete information for age, sex, systolic blood pressure measurements, treatment for IOP and complete IOP measurements. The 2331 individuals comprised 602 extended pedigrees, but due to limitations of the analysis programs, some of the pedigrees were divided to establish 620 distinct pedigrees. Institutional Review Board approval at the University of Wisconsin Medical School was granted for each phase of the study, and informed consent was obtained from study participants. The study adhered to the tenets of the Declaration of Helsinki. 

Data Collection

IOP was measured with a Goldmann applanation tonometer after instillation of a drop of fluorescein (Fluoress; Barnes-Hind Armour Pharmaceutical Co., Kankakee, IL) in each eye. Before measuring the IOP in the right eye, the tonometer was set to 10 mm Hg and the pressure was recorded only after the tonometer was moved back from the cornea. The tonometer was then reset to 10 mm Hg for measurement of the left eye. Any measurement the examiner thought was unreliable was excluded from the analysis. In addition to a full eye examination each participant also provided a detailed medical history and examination, which included information about hypertension, diabetes, and other medical conditions as well as a history of medication use. Although longitudinal eye measurements are collected, only baseline examination IOP measurements were used in this analysis. 

Statistical Methods

Preliminary Analysis.

Linear regression was used to assess potential confounding variables (SAS, ver. 8.1; SAS, Cary, NC). IOP was defined as the higher IOP measurement between each eye and was the dependent variable. Age, sex, systolic blood pressure (SBP), and treatment of IOP were considered potential confounders and were either included as covariates or adjusted for before further analysis. Familial correlations including parent-offspring, sibling, cousin, avuncular, and spouse pairs were estimated with the FCOR program of S.A.G.E. (1987; Statistical Analysis for Genetic Epidemiology). We estimated the heritability (h ²) from the adjusted sibling correlations (r), by the equation h ² = 2r. 

Segregation Analysis.

The REGC program of S.A.G.E. (1987) version 3.0 was used to perform complex segregation analysis. In addition, the REGCHUNT program ¹⁵ which allows REGC to run multiple times with random starting values was used. Using the regressive models proposed by Bonney ^{16 17} (S.A.G.E., 1994), REGC performs segregation analysis of continuous traits. Because the BDES is a population-based survey, no ascertainment correction was necessary. 

Mendelian (autosomal dominant, autosomal codominant, and autosomal recessive), sporadic, and environmental models were used to test for a locus that influences IOP. Maximum-likelihood methods were used to estimate the parameters for each model. Each model was compared to a general, unrestricted model by using the likelihood ratio test. Models can include different discrete factors that may affect a person’s phenotype or IOP. These are called “type” ¹⁸ or “ousiotype.” ¹⁹ In the Mendelian models, type is represented by genotype, and three types are possible: AA (homozygote), AB (heterozygote), and BB (homozygote). However, in the non-Mendelian models, type may represent any transmission, including shared exposures to environmental risk factors within families. If the type frequencies are in Hardy-Weinberg equilibrium, then they are defined as the frequency of allele A, qA. The probability of a parent’s transmitting these discrete factors to an offspring is represented by a transmission parameter (τ). Under Mendelian transmission the τ’s are interpreted as the probability that a parent transmits the A allele to an offspring. Thus, τ_AA = 1, τ_AB = 0.5, and τ_BB = 0. However, when these transmission factors are fixed equal to each other, they represent an environmental model in which parental type (environmental exposure) has no effect on offspring type. 

Using the S.A.G.E. statistical program, we performed all analyses with class D models that assume, after conditioning on the parents, that the correlations among siblings are equal but that they are not completely due to common parentage. Additional familial correlations can therefore be estimated that may explain the effects of other genes (polygenic inheritance) or common environmental factors. These correlations may include spousal (ρ_fm), parent-offspring (ρ_po), and sibling (ρ_ss). In addition, age, sex, treatment of IOP, and SBP measurements at baseline were included as covariates in the analysis because they were considered potential confounders. The IOP measurements were initially analyzed both with and without transformation to normality by using the Box-Cox transformation. ²⁰ However, the IOP measurements followed a normal distribution, and the data were analyzed without the need for transformation. 

Six hypotheses were tested against the likelihood of a general or unrestricted model and likelihood ratio tests, and Akaike’s Information Criterion (AIC) was used to select the most parsimonious model. The six hypotheses of transmission included (1) no major gene, (2) Mendelian dominant, (3) Mendelian codominant, (4) Mendelian recessive (arbitrary major gene), (5) environmental mixed model with a major environmental effect (no parent-offspring transmission) which estimates the transmission probabilities to be equal to the allele frequency, and (6) a semigeneral or τ_AB estimated model that estimates the parameter τ_AB when τ_AA and τ_BB are fixed to the Mendelian values of 1.0 and 0.0, respectively. Likelihood ratio tests were computed by −2 times the difference in lnLikelihood of the general model and a nested model. AIC is defined as −2lnL + 2(number of parameters estimated). ²¹ The model with the minimum AIC is the most parsimonious model. 

Pilot Linkage Analysis.

Previously, 325 individuals (263 sibling pairs) from 102 pedigrees in the BDES were selected for genotyping and linkage analysis of 345 genome-wide markers (Weber panel 8 marker set) because some members of these families were affected with age-related maculopathy (ARM). ²² Although ARM is unrelated to IOP, many individuals genotyped (218 sibling pairs within 87 pedigrees) had complete IOP and covariate measurements and could be used in a pilot genome-wide linkage analysis. 

We determined the allele frequencies of all the markers by using maximum-likelihood methods in the S.A.G.E. program FREQ (S.A.G.E., ver. 4.5). MARKERINFO (S.A.G.E., ver. 4.5) identified two sibships with Mendelian inheritance inconsistencies. These sibships were then excluded from the remaining analyses. RELTEST (S.A.G.E., ver. 4.5), RELCHECK, ^{23 24} and PREST ²⁵ were used to identify potential misspecified family relationships. Based on these results, data from an additional 13 sibships were excluded. IOP was treated as a quantitative trait, and we used the higher untransformed IOP measurement in either eye, adjusted for age, sex, IOP treatment, and SBP as the dependent variable in a Haseman-Elston regression. Using GENIBD, we estimated the single point and multipoint identity by descent-sharing for the study subjects. Linkage analysis was performed with the modified Haseman-Elston regression models and the option that allows for the nonindependence of sibling pairs and uses a weighted combination of the squared trait difference and squared mean corrected trait sum (option W4) ^{26 26} in the model-free linkage program SIBPAL (S.A.G.E., ver. 4.5). For each marker, P-values were obtained using the asymptotic distribution of the likelihood-ratio test statistics. Each chromosome was analyzed separately. 

Results

Familial Correlation Analysis

Table 1describes the adjusted familial correlations for spousal, parent-offspring, sibling, avuncular, and cousin pairs. There was modest positive correlation between both parent-offspring (0.13) and siblings (0.15). No spousal correlation was detected (−0.09), which is consistent with an inherited disease, in which unrelated individuals are not expected to correlate. 

Segregation Analysis

The parameter estimates from the complete segregation analysis are presented in Table 2 . In our segregation analysis, the results indicate that a single distribution model with polygenic effects did not provide an adequate fit (model 1). The models that incorporated a major gene effect with additional polygenic effects (models 2, 3, and 4) also failed to provide an adequate fit to the data, and these Mendelian models were rejected when compared with the general transmission model (model 7; P < 0.0001) and the semigeneral transmission model (τ_AB free model 6; P < 0.0001). The environmental mixed model with τ’s equal to the allele frequency and multiple distributions (model 5) with polygenic effects provided a better fit to the data and could not be rejected (P = 0.12). However, both the environmental mixed model τ’s equal to the allele frequency and the τ_AB free model were highly parsimonious according to the AIC (11974.0469 and 11974.2469, respectively). 

Linkage Analysis

The results of the two-point whole-genome scan yielded two novel loci on chromosomes 6 and 13, as potential linkage regions with a threshold lod score of 1.45. Figure 1graphically depicts the two-point linkage evidence across the genome. Table 3identifies the chromosome 6 and 13 markers and the associated two-point asymptotic probabilities. We observed the strongest linkage peaks at markers D6S1027 (P = 0.008) and D13S317 (P = 0.00071), the chromosome 13 locus reaches the threshold for suggestive linkage of a complex trait. ²⁷ Neither locus has been identified previously in genome-wide screens for POAG. We performed multipoint linkage analysis on chromosomes 6 and 13 (Figs. 2 3) . However, multipoint linkage analysis weakened any suggestion of linkage for both regions. Although two regions of interest were identified, neither region reached genome-wide significance. 

Discussion

The results of our familial correlation analysis for IOP are consistent with a genetic model of inheritance in which first-degree relatives (parent-offspring and siblings) have stronger positive correlations than extended family members. However, this is also suggestive of familial-shared environment. We estimate 30% heritability for IOP after adjustment for age, sex, IOP treatment, and SBP. Previously, we determined 36% heritability for IOP based on familial correlations adjusted for age and sex, and much greater heritability for optic cup and disc diameters (0.55 and 0.57, respectively). ²⁸ The overall conclusions of both familial correlations are similar, although the correlations and heritability declined after the inclusion of IOP treatment and SBP, consistent with some role for common environment on the trait values and/or of genes that influence both IOP and blood pressure. These heritability estimates are similar to those for nuclear cataract (36%), ²⁹ adjusted for age, sex, race, and smoking status, and less than those for ARM (45%). ³⁰  

In the segregation analysis, the model that allowed for an unmeasured major environmental effect plus a polygenic/environmental effect provided the best fit to the data and was the most parsimonious model. Another version of the general model (semigeneral) commonly used to test for the presence of a major locus effect in a mixed model is one in which the τ of the A allele for AA and BB genotypes are fixed to 1.0 and 0, and the remaining parameters are unrestricted. This model was marginally rejected (P = 0.02) and highly parsimonious and yielded an estimate of the τ_AB transmission of 0.70, close to its Mendelian value of 0.5. All the Mendelian models tested for only a single major gene to control IOP across the spectrum of IOP values. The lack of an adequate fit for these models is consistent with a multifactorial or complex model of inheritance that may include multiple genes (at least more than one gene) and environmental factors that contribute to IOP. The conclusions of the segregation analysis are not surprising, because IOP has been associated with numerous physiological and environmental factors including SBP and diastolic (DBP) blood pressure, body mass index, hematocrit, serum glucose, glycohemoglobin, cholesterol levels, pulse, nuclear sclerosis, and season and time of day of measurement. ³¹ As with most genetic analyses it is difficult to account for all potential confounders because of the complexity of the genetic model and current limitations of analytical programs. However, we tried to account for those covariates that may have the greatest effect on IOP, but there may be some confounding by unaccounted for covariates. 

The inclusion of SBP as a covariate is important, because a recent study of the BDES by Klein et al. ³² determined that IOPs correlated with SBP and DBP at both baseline and 5-year follow-up. Specifically, they found that reduced systemic blood pressure was associated with reduced IOP. 

In addition, whole-genome scans for hypertension including both SBP and DBP have also indicated a polygenic inheritance with heterogeneity. ^{22 33 34 35 36 37 38 39 40 41 42} Our analysis of IOP appears to be consistent with blood pressure and is suggestive that we are looking at a similar pattern of inheritance. It is plausible that IOP, like blood pressure, is heterogeneous, and there may be loci that control both the pressure in the eye and blood. 

The two regions we identified by our pilot linkage analysis, chromosomes 6 and 13, warrant further attention, especially since neither of these loci has been identified in previous genome-wide scans of POAG. The identification of two new regions for POAG may be due to the definition of IOP and the ascertainment of cases. This study analyzed the quantitative trait, IOP, compared with the dichotomous POAG previously studied in linkage analyses of glaucoma. Although elevated IOP is a key characteristic in the diagnosis of glaucoma, IOP may represent only one aspect of development of glaucoma. Further localization of these novel regions will provide greater insight into the mechanism of IOP and the contribution to POAG. In addition, the ascertainment of elevated IOP individuals in the Beaver Dam population is in contrast to some of the previous linkage studies involving individuals with POAG from highly aggregated families. The previous linkage studies selected families with well-defined cases of POAG and may have identified rare highly penetrant susceptibility alleles. The BDES involved a predominantly European-American community-based sample which ascertained all age-eligible individuals in Beaver Dam, Wisconsin. Thus, it is possible this sample lacked the power to replicate the initial linkage findings for rare and highly penetrant alleles, and instead we identified more common susceptibility alleles at two potential loci that may control the variation in IOP. Finally, it is important to note that many of the previous linkage findings have not yet been replicated, nor has a gene been identified for each identified locus, and so it is likely some of these regions are not true disease loci. However, the replication of a true linkage finding can be difficult if there are multiple genes that contribute to a disease ⁴³ as with POAG, which may help to explain the lack of replication in our study and others. 

Although we identified a strong two-point linkage signal, the strength of the signal decreased in the multipoint analysis. The weakened signal for linkage in the multipoint analysis is not surprising, since, for both chromosome 6 and 13, there is a wide distance surrounding the marker that provided the greatest linkage signal (25 cM and 28 cM, respectively). Further fine mapping in this region may identify a narrower linkage peak. A whole-genome linkage analysis is currently under way and involves the complete family structure from the BDES (2231 individuals). This should give us additional power to identify new genes or further localize novel genes that contribute to the complex traits IOP and glaucoma. 

 Supported by National Eye Institute Grants R03 EY13438 (BEKK) and UIO EY06594 (RK, BEKK) and by National Human Genome Research Institute Intramural Funds (JEB-W). The results of this paper were obtained by using the program package S.A.G.E., which is supported by a U.S. Public Health Service Resource Grant (RR03655) from the National Center for Research Resources, Bethesda, Maryland.

 Submitted for publication June 21, 2004; revised October 13, 2004; accepted October 19, 2004.

 Disclosure: P. Duggal, None; A.P. Klein, None; K.E. Lee, None; S.K. Iyengar, None; R. Klein, None; J.E. Bailey-Wilson, None; B.E.K. Klein, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Table 1.

 

Familial Correlation Analysis of IOP

Familial Pair	Number of Pairs	Correlation with Equal Weight to Pairs	95% CI
Spousal	80	−0.09	−0.31, 0.13
Parent-offspring	510	0.13	0.04, 0.22
Sibling	1125	0.15	0.08, 0.22
Avuncular	715	0.02	−0.07, 0.11
Cousin	1646	0.09	0.03, 0.15

 IOP values adjusted for age, sex, IOP treatment and SBP.

Table 2.

 

Sex-Specific Segregation Analysis of Intraocular Pressure (IOP) with Eye Drops, Age, and Systolic Blood Pressure

	qA	τAA	τAB	τBB	βAA	βAB	βBB	TX	Age	SBP	PO	SS	DF	−2lnL	P	AIC
1. No major gene (no transmission)+ familial correlations (sporadic)*
Male	1.0	—	—	—	11.49	11.49	11.49	4.33	0.03	0.002	0.14	0.13	6	12084.8617	<0.001	12096.8617
Female					11.86	11.86	11.86
2. Mendelian dominant+ familial correlations*
Male	0.21	1	0.5	0	21.37	21.37	13.99	4.14	0.01	0.005	0.17	0.12	4	12039.5707	<0.001	12047.5707
Female					21.42	21.42	13.58
3. Mendelian codominant+ familial correlations*
Male	0.03	1	0.5	0	−21.9	17.84	11.75	4.09	0.03	−0.001	0.13	0.10	3	12004.2755	<0.001	12010.2755
Female					3.76	18.18	11.32
4. Mendelian recessive+ familial correlations*
Male	0.21	1	0.5	0	21.37	13.99	13.99	4.14	0.01	0.005	0.17	0.12	4	12039.5707	<0.001	12047.5707
Female					21.42	13.58	13.58
5. Environmental mixed model τ’s = qA+ familial correlations*
Male	0.96	0.96	0.96	0.96	11.79	17.46	28.45	4.07	0.03	−0.002	0.22	0.14	3	11968.0469	0.12	11974.0469
Female					11.38	18.36	2.56
6. τAB estimated (semigeneral)+ familial correlations*
Male	0.93	1.0	0.70	0	11.86	17.56	28.59	3.90	0.03	−0.004	0.18	0.12	2	11970.2469	0.02	11974.2469
Female					11.45	18.48	4.05
7. General+ familial correlations*
Male	0.94	0.98	0.74	0.56	11.75	17.31	27.9	3.89	0.03	−0.002	0.19	0.13	—	11962.1624	—	11964.1624
Female					11.35	18.19	2.91

 Gray boxes denote non-estimated parameters for each model. TX, treatment with drops for IOP; SBP, systolic blood pressure; PO, parent-offspring correlation coefficient; SS, sibling correlation coefficient; DF, degrees of freedom; AIC, Akaike’s information criterion; the lowest value represents the most parsimonious model; −2lnL, −2 log to the base nLikelihood.

*  Additional familial correlations may explain multifactorial effects including polygenic and common environmental effects.

Figure 1.

 

Whole-genome scan for the quantitative trait, IOP, from the baseline examination. Marker positions are from Marshfield Laboratories (Marshfield, WI).

Table 3.

 

Probabilities Associated with Model Free-Linkage Analyses of Markers at Chromosomes 6 and 13

Marker Name	Distance (cM)	P
Chromosome 6
f13a1	0	0.568698
ATA50C05	16	0.037721
D6S1959	29	0.384235
GATA163B10	40	0.82304
GGAA15B08	49	0.856078
D6S1017	56	0.769203
D6S1053	78	0.844592
D6S1031	86	0.838587
D6S1056	100	0.269214
D6S1021	111	0.071395
D6S474	117	0.168925
D6S1040	126	0.667139
D6S1009	137	0.040428
GATA184A08	147	0.546271
GATA165G02	156	0.20618
D6S305	168	0.792633
D6S1277	175	0.219341
D6S1027	192	0.00885
GATA24F03	200	0.180632
Chromosome 13
D13S787	0	0.925463
D13S1493	19	0.434573
D13S894	28	0.643878
D13S325	34	0.264386
D13S788	41	0.040309
D13S800	53	0.222854
D13S317	66	0.00071
D13S793	81	0.161194
D13S285	126	0.373975

Figure 2.

 

Multipoint linkage analysis of IOP in a region of chromosome 6.

Figure 3.

 

Multipoint linkage analysis of IOP in a region of chromosome 13.