Shared Genetic Determinant of Axial Length, Anterior Chamber Depth, and Angle Opening Distance: The Guangzhou Twin Eye Study

Abstract

purpose. To estimate the extent to which common genetic and environmental effects contribute to covariances among axial length, anterior chamber depth, and angle-opening distance.

methods. The study participants were recruited from the Guangzhou Twin Registry. Anterior segment optical coherence tomography and custom software were used to quantify the angle opening distance (AOD) at the location 500 μm anterior to the scleral spur. Anterior chamber depth (ACD) and axial length (AL) were measured using laser interferometry. Cross-trait, cross-twin correlations for monozygotic (MZ) and dizygotic (DZ) twins and the Cholesky model were used to quantify shared genetic and environmental effects for AL, ACD, and AOD after adjusting for age and sex.

results. A group of 459 pairs of twins (304 MZ and 155 DZ) aged 8 to 16 years were available for analysis. The phenotypic correlations among AL, ACD, and AOD ranged from 0.39 to 0.64. Cross-twin, cross-trait correlations for these three phenotypes for MZ twins were consistently greater than the corresponding correlations for DZ twins. The results of the Cholesky model-fitting analyses can be summarized as follows: first, of 70% of additive genetic factors for AOD, 23% and 13% were those shared with ACD and AL, respectively, whereas the remaining 34% were those unique to AOD. Second, of 89% of additive genetic factors for ACD, 25% were those shared with AL, whereas 64% were those unique to ACD. Third, random environmental influences on covariances among AL, ACD, and AOD were very small.

conclusions. Analyses of Chinese children twin data suggest that shared genes are responsible for the significant phenotypic correlations found for AL ACD, and AOD.

Angle closure and myopia have been traditionally considered to be two separate phenotypes with different mechanisms and etiologic determinants. However, the anatomic characteristics involved in these two diseases are closely associated with each other. A shallow anterior chamber, short axial length (AL), small corneal diameter and steep curvature, shallow limbal chamber depth, and a thick relatively anteriorly positioned lens are considered to be risk factors for primary angle closure. ^{1 2 3 4 5} In contrast, myopia is characterized as a longer AL, flattening corneal curvature, and thinner lens. ⁶ Among these biometric traits, shallow anterior chamber and narrow drainage angle width were considered to be the most relevant traits for angle closure. ^{7 8} Increases in AL and vitreous length have been shown to be the best anatomic features for the onset and progression of myopia. ^{9 10} Thus, the angle closure and myopia appear to be phenotypes with somewhat opposite anatomic features. 

Twin studies are widely used to disentangle genetic and environmental contributions to variations in phenotypes. Monozygotic (MZ) twins are genetically identical, whereas dizygotic (DZ) twins have, on average, 50% of their genes in common. The heritability—the proportion of the total phenotypic variance attributable to genetic effects—can be estimated by the use of a univariate model. Furthermore, Cholesky models allow the estimation of genetic and environmental effects on the phenotypic associations among traits. ¹¹ The MZ and DZ twin correlations between trait A in one twin and trait B in his/her co-twin (termed cross-twin, cross-trait correlations) provide information on the extent to which genetic and environmental factors contribute to the phenotypic correlations between traits A and B. If genetic factors influence the association between the two traits, the MZ cross-twin, cross-trait correlation is expected to be larger than the DZ cross-twin, cross-trait correlation. On the other hand, if shared environmental factors are important for the relationship between the two traits, then the DZ cross-twin, cross-trait correlation is expected to be larger than ½ the MZ cross-twin, cross-trait correlation. If the association between the two traits is purely due to random environmental factors, then the MZ and DZ cross-twin, cross-trait correlations would not be significantly different from 0. 

The use of intermediate phenotype (also called endophenotype) has been increasingly adopted for familial aggregation and gene mapping of complex diseases. ¹² Understanding the etiology of intermediate phenotypes may provide critical evidence for genetic and environmental pathways to the complex diseases of interest. As the anterior chamber depth (ACD) and drainage angle width (quantified as angle opening distance [AOD] in anterior segment optical coherence tomography [ASOCT]) are the anatomic basis of angle closure, they were chosen as intermediate phenotypes for angle closure in the present study. Axial length has been considered to be a phenotype closely correlated with myopia. Thus, AL, ACD, and AOD were selected as endophenotypes of myopia and angle closure in the present study. Our previous studies based on the Guangzhou Twin Registry confirmed that ACD, AL, and AOD are all highly heritable traits. ^{13 14} We have shown that additive genetic effects explain 70% to 90% of the total variances of the three phenotypes. Given the substantial phenotypic associations among ACD, AOD, and AL and the high heritability observed for these traits, it is possible that shared genes may be responsible for the phenotypic covariation among ACD, AOD, and AL. The purposes of the present study were to determine whether common genetic and/or environmental effects influence the relationships among AL, ACD, and AOD and to estimate these common genetic and environmental effects. 

Materials and Methods

Participants

The study participants were recruited from the Guangzhou Twin Registry, which has been described in depth elsewhere. ¹⁵ In brief, this registry was established in Guangzhou City, China. All twins born between 1987 and 2000 were identified by using an official Household Registry of Guangzhou, followed by a door-to-door verification. In July and August in 2006, we invited all twins aged 7 to 15 years (defined at the date of July 1, 2006) living in two districts to the baseline data collection. A total of 563 of the 705 pairs invited were enrolled in the study, yielding a participation rate of approximately 80%. All participants were invited for further phenotypic data collection in 2007. The phenotypes described in this manuscript were all collected in the examination in 2007. For all participants, written informed consent was obtained either from parents or legal guardians of the twins after an in depth explanation of the study. Ethics approval was obtained from the Zhongshan University Ethical Review Board and Ethical Committee of Zhongshan Ophthalmic Center, and the study was conducted in accordance with the tenets of the World Medical Association’s Declaration of Helsinki. 

In this twin cohort, the zygosity of all same-sex twin pairs was determined by 16 multiplex short-tandem repeats (STR’s; PowerPlex 16 system, Promega, Madison, WI) ¹⁶ at the Forensic Medicine Department of Sun Yat-Sen University in 2006. Opposite-sex twin pairs were assigned to be dizygotic (DZ) without genotyping. 

Twin pairs were excluded from data analyses if one or both twins had pathologic changes (e.g., retinopathy of prematurity), recent orthokeratology contact lenses correction, or previous laser treatment for myopia. 

Examination and Measurement

The right eye was arbitrarily selected to represent the phenotypic characteristics of the specific individual in the data analysis. 

AL was measured by noncontact partial-coherence laser interferometry (IOLMaster; Carl Zeiss Meditec, Oberkochen, Germany) in a dark room (<5 lux illumination) before pharmacologic dilation of the pupils. The mean result of 10 continuous measurements was used. Poor measurements with signal-to-noise ratio (SNR) < 2.0 (displayed as “Borderline SNR,” or “Error”) or measurements with one result differing by more than 0.1 mm from the others (displayed as “Evaluation!”) were deleted and remeasured. ACD was measured with the same interferometry device. The ACD measurement was made after automatic adjusting for the corneal radius. A mean of five measurements was taken. Measurements with displayed the “Error” message were deleted and remeasured. The ACD was measured as the distance between the anterior lens surface and corneal epithelium illumination and therefore included the thickness of the cornea. 

ASOCT imaging (Visante; Carl Zeiss Meditec, Dublin, CA) was performed with the participants in a seated position without pupil dilation in a dark room with the same illumination as for the ACD and AL measurements. One scan, centered over the pupil, was taken on the horizontal meridian (between 0° and 180°). The scan direction was properly aligned until a full corneal reflex was achieved (indicated by an interference flare on the axis of anterior chamber). The fixation angle was adjusted to align and enable the image horizontal. One image with best quality was selected and saved for each eye of each person. 

The ASOCT images were subsequently analyzed with custom software, the Zhongshan Angle Assessment Program (ZAAP, Guangzhou, China). This program performed noise and contrast reconditioning and allowed automatic measurement after the scleral spurs were identified. AOD, calculated at 500 μm from the scleral spur, was used in the present analysis. AOD in the present study was defined as the length of a line drawn from the anterior iris to the corneal endothelium perpendicular to a line along the trabecular meshwork 500 μm anterior to the scleral spur. ¹⁷  

Data Analysis and Genetic Modeling

To determine the causes of covariances among AL, ACD, and AOD, we computed twin correlations and cross-trait, cross-twin correlations for MZ and DZ twins and performed model-fitting analyses by using the Cholesky model. 

As variances can be divided into genetic and environmental sources, covariances can be decomposed into additive genetic (A), dominance genetic (D), and random environmental (E) factors. ¹¹ The A factors represent the sum of the average effect of all alleles that influence a trait. The A factors correlate at 1.0 for MZ and at 0.5 for DZ twins. The D factors reflect the genetic factors that do not add up across alleles because of interaction. The D factors correlate at 1.0 for MZ and at 0.25 for DZ twins. Finally, the E factors, environmental factors that are unique to each member of a twin pair (e.g., accident, virus infection) and measurement error, do not contribute to the twin similarity. The shared family environmental factors were not considered in the present study, because the patterns of MZ and DZ twin correlations for AL, ACD, and AOD suggested that shared environmental factors are negligible (Table 1) , and because our previous studies have shown no significant shared environmental influences on AL, ACD, and AOD. ^{13 14}  

Figure 1shows that each of the three Cholesky factors for AL, ACD, and AOD are decomposed into A, D, and E. ¹¹ The first Cholesky factors (i.e., A₁, D₁, and E₁) exert influences on all three traits—that is, AL, ACD, and AOD—although they predominantly affect AL. The second Cholesky factors (i.e., A₂, D₂, and E₂) have effects mainly on ACD, although they influence AOD also. The third Cholesky factors (i.e., A₃, D₃, and E₃) are those unique to AOD. The A, D, and E covariance matrices were computed by the product of their respective Cholesky factor loading matrix and its transpose. The genetic and environmental correlations, derived from the variances and covariances, allowed us to determine what percentage of the phenotypic correlation is due to shared genetic and/or shared environmental influences. 

Variations of the full Cholesky model were made to select the best-fitting model. We used the Akaike information criterion (AIC = χ² − 2 df) and the likelihood-ratio χ² test (LRT) to select the best-fitting model. If competing models were not nested, the model that produced the lowest AIC was considered the best-fitting model. If alternative models were nested, then the LRT was used because the change in χ² is distributed as χ². A significant increase in χ² in the reduced model compared with the full model would suggest that the reduced model fit the data less well than the full model. A nonsignificant change in χ² would indicate that the reduction of the model parameter is acceptable. 

Our model-fitting analyses were based on observed variance and covariance matrices computed separately for MZ and DZ twins among AL, ACD, and AOD. We used Mx to fit the Cholesky decomposition model to the observed variance and covariance matrices. ¹⁸  

Results

Descriptive Statistics

The study included 459 twin pairs (304 MZ and 155 DZ pairs) with ACD, AOD, and AL data all available. Age was significantly positively correlated with all three phenotypes (r = 0.26 for AOD, P < 0.01; r = 0.25 for ACD, P < 0.01; r = 0.34 for AL, P < 0.01) in the present sample, suggesting that AL and angle closure increase with age. Means and SDs for AL, ACD, and in the present sample have been reported elsewhere. ^{13 14} In brief, the mean (0.68 in the boys and 0.67 in the girls) or SD (0.23 for the boys and 0.24 for the girls) for AOD was not significantly different between the two sexes. However, the means for AL and ACD were slightly, but significantly higher among the boys than among the girls (23.9 vs. 23.5, P < 0.01 for AL; 3.56 vs. 3.50, P < 0.01 for ACD). ACD showed no significant difference in SD between the boys and the girls (0.25 vs. 0.28, P < 0.17). The SD for AL, however, was significantly different between the boys and the girls (1.02 in the boys and 1.13 in the girls, P < 0.05). The means and SDs for AOD, ACD, and AL were not significantly different between the first- or second-born twins, suggesting that birth order has no effect on these variables. For all three traits, no significant mean or SD differences were found between MZ and DZ twins, fulfilling the basic assumptions of twin analyses. 

Correlational Analyses

Table 1presents phenotypic correlations, twin correlations (the elements in the rectangles on the diagonals), and cross-twin, cross-trait correlations (the off-diagonal elements) for AL, ACD, and AOD. All correlations were corrected for sex and age. The phenotypic correlations among AL, ACD, and AOD were high and significant, ranging from 0.39 to 0.64. MZ twin correlations were significantly greater than DZ twin correlations for all three phenotypes, confirming substantial genetic influences and no shared environmental influences on AL, ACD, and AOD. Cross-twin, cross-trait correlations in MZ were all greater than the corresponding DZ correlations, indicating that shared genetic factors played an important role in covariation among the three phenotypes. The impressions gained from these correlational results were tested using model-fitting analyses. 

Model-Fitting Analyses

Cholesky Model-Fitting.

Table 2provides the results of model-fitting analyses. The full Cholesky model fit the data well (χ² ₂₄ = 23.04, P = 0.52). When we removed nonadditive genetic variances and covariances among AL, ACD, and AOD from the full model, a nonsignificant change in χ² occurred (model 2; Δχ² ₆ = 0.9, P = 0.99), suggesting that genetic effects on the three phenotypes are largely of the additive rather than the nonadditive kind. Eliminating additive genetic variances unique to AL, ACD, and AOD, respectively, significantly worsened the fit (models, 3, 4, and 5). Next, we removed additive genetic covariances between AL and ACD, between AL and AOD, and between ACD and AOD, respectively (models 6, 7, and 8). These procedures also produced significant changes in χ², suggesting the importance of common genetic effects on the covariances of AL, ACD, and AOD. Finally, random environmental covariances between AL and ACD, between AL and AOD, and between ACD and AOD were dropped from model 2, which yielded significant changes in χ² (models 9, 10, and 11). We did not attempt to remove random environmental variances unique to AL, ACD, and AOD, as these variances were confounded with measurement error. Taken together, these results suggested that the best-fitting model was model 2 where additive genetic and random environmental factors exert significant influences on variation and covariation among AL, ACD, and AOD. 

Maximum-Likelihood Parameter Estimates in the Best-Fitting Model.

Figure 2presents standardized factor loadings and their 95% CIs in the best-fitting model. ^{13 14} In Figure 2 , although both additive genetic and random environmental parameters were statistically significant, additive genetic parameters were consistently greater than the corresponding random environmental parameters, suggesting that the phenotypic relationships among AL, ACD, and AOD are largely mediated by common genetic rather than common random environmental factors. 

Heritability for AL, ACD, and AOD were 88%, 89%, and 70%, respectively, confirming our previous findings. ^{13 14} When we decomposed these heritability estimates, of 70% of additive genetic factors for AOD, 23% were shared with those for ACD (A₂ in Fig. 2 ), 13% were in common with those for AL (A₁ in Fig. 2 ), and the remaining 34% were those unique to AOD (A₃ in Fig. 2 ). Likewise, of 89% of additive genetic factors for ACD, 25% were shared with those for AL (A₁ in Fig. 2 ), and 64% were those unique to ACD (A₂ in Fig. 2 ). 

Random environmental influences were modest: 12%, 11%, and 30% for AL, ACD, and AOD, respectively. When we decomposed these estimates, of 30% of random environmental influences on AOD, 26% were those unique to AOD (E₃ in Fig. 2 ), 3% were those shared with ACD (E₂ in Fig. 2 ), and 1% were those in common with AL (E₁ in Fig. 2 ). Likewise, 11% of random environmental influences on ACD consisted of 2% of those shared with AL (E₁ in Fig. 2 ) and 9% of those unique to ACD (E₂ in Fig. 2 ). 

Discussion

In this article, we attempted to determine the cause of associations of angle closure (both anterior chamber and drainage angle width) and myopia-related traits by exploring underlying relationships among AL, ACD, and AOD. Using Cholesky modeling, we demonstrated that shared genetic effects largely determined the relationships among ACD, AOD, and AL. Random environmental covariances were minimal, if any. Correlated measurement errors for the traits may be responsible for the random environmental covariances found in the present study. 

In our study, as we expected, we found high phenotypic associations among AL, ACD, and AOD. ACD and AL are both the distance measurement on the axial direction of the eye globe. AL is equivalent to a sum of ACD, lens thickness, and vitreous length and therefore, is closely associated with ACD. Using a case control design, Lowe ⁵ showed that 35% of the variance of ACD was attributable to lens thickness and the remaining 65%, to the lens position. Population studies have consistently suggested that the anterior chamber tends to be deeper in myopic eyes. ⁶ On the other hand, angle width, usually quantified using either gonioscopy or even better by ASOCT, is an estimation of the proximity between the peripheral iris and trabecular meshwork in the drainage angles of the eye. It is associated with ACD but also reflects a combination of other anatomic components, including iris thickness, iris curvature, and corneal curvature. 

In most quantitative traits, genes are known to influence multiple phenotypes (pleiotropic effects of genes). The cross-twin, cross-trait correlations for AL, ACD, and AOD in the MZ pairs were greater than those in the DZ pairs, suggesting the importance of common genetic effects on the relationship among the three phenotypes. Cholesky model-fitting analyses confirmed the results of twin correlational analyses: Shared genetic variances for AL, ACD, and AOD were significant and high, whereas shared random environmental factors were very modest. Taken together, these results suggest that pleiotropic effects of genes may be operating in AL, ACD, and AOD and that the pleiotropic actions of genes probably contribute to the associations between angle closure and myopia-related traits. 

Molecular genetic findings on angle closure and myopia remain limited and controversial. Pedigree-based studies have reported linkage loci on chromosome 11 for nanophthalmos but were not able to replicate linkage for angle-closure glaucoma or narrow angles. ¹⁹ One recent study reported the MMP-9 gene to be associated with acute angle closure but yet to be confirmed. ²⁰ None of these angle closure–related chromosome regions have been reported to be overlapped with the loci related to myopia, although the linkage loci of myopia are mainly for high myopia and from pedigree-based studies. ²¹ However, one has to be cautious that both angle closure and common myopia are likely polygenic disorders, in which the effect size at any single locus is likely to be small. The pedigree study or even population- and family-based association studies may lack the power to detect genes of such small effect size. Therefore, it would not be surprising, even if no chromosomal overlap has been detected in previous studies. However, the association between angle closure and myopia is biologically plausible. Myopia is characterized as excessive growth of the eye globe (longer eye), whereas angle closure occurs more often in shorter eyes, presumably representing no growth of the eye globe. Therefore, it is possible that there is a common set of genes underlying these two phenotypes that express in different pathways 

In this study, the ACD and AOD are both quantitative traits and chosen as the intermediate phenotypes for the angle closure. However, these two traits are also closely associated with myopia—myopes tend to have longer eyes and deeper anterior chambers. Therefore, one may argue that the ACD and AOD may be phenotypes related to both angle closure and myopia. In our previous study, ¹⁴ we divided the ACD by AL in an attempt to control the influence of myopia on the ACD trait. However, the heritability of ACD remained high even with adjustment for myopic effects. This finding underscores the need to consider the effect of myopia when we intend to use ACD or AOD as quantitative trait loci (QTL) traits of interest for the molecular genetic study of angle closure. 

As our study participants were healthy young people, angle-closure glaucoma was extremely uncommon in the sample. Thus, the results of the present study are only applicable to normal children in China. Further inference on the etiology of angle closure or angle-closure glaucoma may depend on the validity of the assumption that shallow ACD in childhood indicates a propensity to a shallow ACD in later life. However, intrapair environmental factors are likely to be similar in young people, whereas adult or elderly twins may incur more diversified intrapair environmental influences. This may confer advantages for using relatively young twin for the heritability study. 

In summary, we confirm strong, shared additive genetic effects on the traits related to angle closure and myopia. Only very small proportions of the covariances were explained by random environmental factors including measurement errors. This study underscores the need to consider the possible influence of pleiotropic effects of genes in angle closure and myopia in QTL-based association studies. 

 Supported by National Natural Science Foundation of China Grant 30772393; Program for New Century Excellent Talents in University, National Ministry of Education Grant NCET-06-0720; Sino-Australia NSFC-DEST (Chinese State National Science Fund Committee–Australian Department of Education, Science and Training) Special Fund Grant 30371513; and Guangzhou Science and Technology Development Fund Grant 2006Z3-E0061.

 Submitted for publication April 4, 2008; revised June 6, 2008; accepted August 27, 2008.

 Disclosure: M. He, None; Y.-M. Hur, None; J. Zhang, None; X. Ding, None; W. Huang, None; D. Wang, 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.

 

Correlations in MZ and DZ Twins

	AL	ACD	AOD
Phenotypic correlations (n = 918)
AL	1.0
ACD	0.52	1.0
AOD	0.39	0.64	1.0
	AL1	ACD1	AOD1
Twin correlations and cross-twin cross-trait correlations (n = 459)
MZ twins (n = 304 pairs)
AL2	0.88	0.46	0.32
ACD2	0.50	0.90	0.59
AOD2	0.36	0.55	0.70
DZ twins (n = 155 pairs)
AL2	0.39	0.11	0.11
ACD2	0.30	0.46	0.34
AOD2	0.18	0.27	0.35

 Data are adjusted for sex and age. All correlations are significantly different from 0 at P < 0.05; 1 and 2 refer to measures made on first and second twins. The elements on the diagonals (bold) represent twin correlations. The off-diagonal elements are cross-twin, cross-trait correlations.

Figure 1.

 

Decomposition of the three Cholesky factors into A, D, and E factors. For simpler presentation, the D factors are not shown. Double-headed arrows indicate that correlations for MZ and DZ pairs were set at 1.0 and 0.5, respectively.

Table 2.

 

Cholesky Model-Fitting for Three Traits

Model	Description	χ2	df	AIC	P	Δχ2	Δdf	P
1	Full ADE Cholesky	23.0	24	−25.0	.52
2	Drop nonadditive genetic variances and covariances among AL, ACD, and AOD	24.0	30	−36.0	.77	0.9	6	.99
3	Same as model 2, but drop additive genetic variance unique to AL	508.8	31	446.8	.00	485.7	7	.00
4	Same as model 2, but drop additive genetic variance unique to ACD	479.0	31	417.0	.00	456.0	7	.00
5	Same as model 2, but drop additive genetic variance unique to AOD	156.7	31	94.7	.00	133.6	7	.00
6	Same as model 2, but drop additive genetic covariance between AL and ACD	188.2	31	126.2	.00	165.2	7	.00
7	Same as model 2, but drop additive genetic covariance between AL and AOD	109.4	31	47.4	.00	86.4	7	.00
8	Same as model 2, but drop additive genetic covariance between ACD and AOD	215.3	31	153.3	.00	192.3	7	.00
9	Same as model 2, but drop nonshared environmental covariance between AL and ACD	95.0	31	33.0	.00	72.0	7	.00
10	Same as model 2, but drop nonshared environmental covariance between AL and AOD	40.7	31	−21.3	.00	17.7	7	.00
11	Same as model 2, but drop nonshared environmental covariance between ACD and AOD	54.0	31	−8.0	.00	31.0	7	.00

 The best-fitting model is in bold print.

Figure 2.

 

Standardized parameter estimates in the best-fitting Cholesky model for AL, ACD, and AOD.