Update on the Epidemiology and Genetics of Myopic Refractive Error

Justin C Sherwin; David A Mackey


Expert Rev Ophthalmol. 2013;8(1):63-87. 

In This Article

Genetic Etiology of Myopia

Approach to Gene Finding in Myopia

Ocular and systemic syndromes associated with myopia, and results from familial aggregation studies, heritability estimates, segregation analyses, linkage studies and, more recently, genome-wide association studies (GWAS) provide weight to a genetic basis of myopia. Inter-ethnic differences in refractive error support a genetic etiology further (see section 'Prevalence of myopia'). Elucidating the genetic determinants of a disease involves several steps.[136] Until recently, the main approach to identify susceptible chromosomal regions associated with disease was linkage analysis of related individuals (either using parametric or nonparametric approaches), followed by gene localization and sequencing techniques. Candidate gene and GWAS can be performed on unrelated individuals with no specific hypothesis about which single-nucleotide polymorphism (SNP) may be involved, thus necessitating large sample sizes and strict p-value thresholds to determine statistically associations between gene variants and myopia. Once a possible causal gene variant is identified, functional studies can be conducted to identify the consequence(s) of changes in gene expression.

Animal Studies

Samples of ocular tissue that might be of interest in myopia research, such as the retina and sclera, cannot be obtained from living humans thus animal models are required. There are several ways of inducing myopia experimentally: form deprivation (eyelid suturing or placing opaque lenses in the anterior eye causing axial elongation and myopia), lens-induced optical defocus (exposure to optical defocus via plus- or minus-powered lenses leading to compensatory changes in AL and refraction) and restricted visual environment conditions. Knockout, breeding experiments and quantitative-trait loci models in mice and chickens have identified genes involved in eye size and refractive regulation.[137] Animal studies have demonstrated an active emmetropization mechanism that normally ensures coherence between AL and optical power of the eye.[138] Failure of emmetropization could arise from irregular expression of genes in the retina, retinal pigment epithelium, lens, choroid and/or sclera, resulting in axial elongation and myopia.

Familial Aggregation Analysis

Given that families share environments, difficulty exists in differentiating the degree of familial clustering that is due to shared genes and shared environments. The intrapair correlation coefficient for spherical equivalent is significantly higher in monozygotic than dizygotic twins.[35] Several studies, encompassing several ethnic groups, have indicated an increased risk of myopia in offspring when either one or both parents are myopic compared with when both parents are nonmyopic.[139,140] The estimated recurrence risk for siblings of individuals with myopia (λs) varies between 1.5 and 3.0 for low myopia and several-fold higher for high myopia.[141] This increased risk persists even after controlling for environmental risk factors.[121] In a cross-sectional study of individuals aged 17–45 years in Singapore, the odds ratio (OR) for having mild/moderate myopia was between 2.5 and 3.7, and for high myopia was 5.5; there was also a strong association between family history of myopia and having a longer AL (p < 0.001).[142] In a study of Singaporean preschool children, subjects with two myopic parents were more likely to be myopic and to have a more myopic refraction than children without myopic parents.[143] Children with myopic parents may be predisposed to myopia because of inheriting factors associated with a longer AL.[76]

Heritability Studies

The genetic contribution to a trait/disease can be divided into an additive (A) genetic variance component and a nonadditive (D) genetic variance component – dominance and epistasis (where the effects of one gene are modified by one or several other genes). Heritability can be defined as the proportion of variance of a disease or trait due to additive genetic factors. Determination of heritability can be through either regression/correlation methods or variance component equations/structural equation modeling using modern computer software packages. Novel estimation methods of heritability analysis can employ high-density genetic marker technologies.[144] In twin studies, the most reported approximation of heritability using regression methods (Falconer's formula) represents twice the difference in the correlation of monozygotic (MZ) and dizygotic (DZ) twin pairs; to be valid it needs to satisfy the assumptions of a classical twin study.[145] If a genetic basis for a disease or trait exists, MZ twins should be more similar as they share 100% of their genetic material, as opposed to DZ twins who on average share only 50%. Other assumptions of twin studies include the equal environment assumption, trait normality, homoscedasticity (equality) of trait, variances between zygosity groups and accurate zygosity classification. Reduced variation in the range of environmental variation in twin pairs compared with other family relationships may explain why heritability estimates from twin studies are higher than from other study designs. Within a population, heritability is not always constant, and may be altered by changes in measurement methods and environmental factors, as well as effects of migration, selection and inbreeding.[144]

A meta-analysis found that the pooled heritability estimate of both refractive error (six studies) and AL (seven studies) was 0.71 (71%), even though there was very high interstudy heterogeneity.[146] However, as the study designs contributing heritability measurements were different, the interpretation of such findings is unclear even with a random-effects model. In that study,[146] the individual heritability estimates for refractive error were broad, ranging from 0.2 to 0.91, with the highest estimates derived from twin studies. Most large studies, using quantitative data and structural equation techniques, showed that refractive error is largely genetic in origin (heritability >50%). A high heritability of other refractive error endophenotypes, including AL, anterior chamber depth and corneal curvature, has also been demonstrated.[146,147] Investigation of myopia endophenotypes may provide a useful strategy for unraveling the genetic risk factors of myopia.[148] Classical twin studies are potentially biased towards finding a genetic basis of disease or trait as heritability estimates are often inflated due to the failure to model shared environmental effects, following on from the assumption that any excess correlation in MZ twins compared with DZ twins only represents genetic effects. This may explain why heritability estimates from twin studies tend to be higher than those obtained from family-based designs.[147] Even when researchers attempt to model shared environmental effects, studies are often underpowered to detect them.[149]

Segregation Analyses

Segregation analysis, that is, using maximum likelihood analysis to estimate transmission probabilities of the observed data, is a valuable statistical method to determine the way complex disorders such as myopia are inherited. Multiple different inheritance modes for myopia have been proposed, including recessive, dominant and X-linked forms, providing further evidence of disease heterogeneity. Findings from 602 pedigrees encompassing 2138 BDES participants revealed that a multifactorial mode of inheritance was the most parsimonious,[150] a finding that has been shown elsewhere.[151]

Syndromal Myopia

There are many ocular and systemic syndromes that are associated with myopia (Table 2). In syndromal myopia, the degree of myopia is usually severe and characterized by an early age of onset and clearly recognized familial pattern. A search of OMIM[301] and PubMed databases performed in November 2012 revealed more than 250 individual syndromes in which myopia has been described. The probability that the myopia results from the underlying genetic mutation is elevated in cases when the myopia is severe, the myopia is highly penetrant and when the gene(s) involved has a plausible role in refraction. Furthermore, most cases of high myopia are not associated with a syndrome. In a community-based UK population of children with high myopia who were identified from optometric and orthoptic practice records, 56% had no associated ocular or systemic condition,[152] less than in a US study.[153]

Many mutations leading to syndromes that are associated with myopia are in connective tissues or extracellular matrix (ECM) components. This is unsurprising as the majority of the sclera is composed of collagen, and the growth and remodeling of the sclera is increasingly recognized as playing an important role in human refraction.[154] Numerous inherited syndromes are associated with high myopia and abnormal vitreous that predisposes to rhegmatogenous retinal detachment: Stickler's syndrome, Wagner disease and erosive vitreoretinopathy, Knobloch syndrome and Marfan syndrome.[155] Identification of mutations involved with syndromic myopia led many researchers to hypothesize that these genes would be involved in cases of nonsyndromic myopia.[138] Thereafter, a candidate gene approach could be employed. Polymorphisms in the collagen type 2 α 1 (COL2A1) gene, mutations of which are associated with Stickler syndrome,[156] have been associated with nonsyndromic myopia.[157]

Linkage Studies

The role of genetic factors is likely to be stronger in cases of nonsyndromic high myopia of early onset in contrast to cases of low/moderate myopia in which the environmental influence is likely to be more apparent.[158] Indeed, many of the loci found for myopia apply to high myopia only; 20 loci for myopia (MYP1-3; MYP5-21) are currently listed on OMIM (Table 3).[301] MYP4 was originally used for the locus on 7q36, but Paget et al.[159] found no evidence of linkage to 7q36 even though they used the same families as Naiglin et al..[160] Instead, they found significant linkage to 7p15, now referred to as MPY17.[159] Most of these loci have been found through linkage analyses in highly myopic probands with multiple affected relatives, and corroborative findings from replication studies have been limited. Most of these loci have displayed an autosomal dominant inheritance pattern, but other forms of inheritance have included autosomal recessive (MYP18) and X-linked forms (MYP1 and MYP13). One locus, MYP21, represents a mutation in the zinc finger protein 644 isoform 1 (ZNF644) gene, located on chromosome 1p22.2.[161] Shi et al. discovered heterozygosity for a 2091A>G transition in exon 3 of the ZNF644 gene in a Han-Chinese family with high myopia, leading to an ile587val (I587V) substitution.[161]ZNF644 was expressed in human liver, placenta, retina and retinal pigment epithelium, the only tissues examined. Subsequently, two novel single-nucleotide variants in ZNF644 (c.725C>T, c.821A>T) in two high-grade myopia individuals (one Caucasian and one African–American) were identified in a US cohort of individuals with high myopia.[162]

Not all myopia susceptibility loci are formally recognized by HUGO but may be in due course. Researchers in the BDES identified two novel regions of suggestive linkage on chromosome 1q and 7p.[163] Ciner et al. conducted linkage analysis on 96 families containing 493 African–American individuals in the Myopia Family Study (mean SE: -2.87 D).[164] They found significant linkage at 47 centimorgans (cM) on chromosome 7 (logarithm of the odds [LOD] score: 5.87; p = 0.00005). There were also three regions on chromosomes 2p, 3p and 10p showing suggestive evidence of linkage (LOD >2; p < 0.005) for ocular refraction. Using an additional 36 white families in addition to the African–American families, a suggestive linkage at chromosome 20 was found, which became more significant when the scores were combined for both groups.[165] Using refractive error as a continuous variable, two additional potential myopia susceptibility loci at 6q13-16.1 and 5q35.1-35.2 for myopia were found.[166]

Li et al. performed whole-genome linkage scans for high myopia, using 1210 samples from five independent sites.[167] In addition to replicating several previously identified loci, they found a novel region q34.11 on chromosome 9 (max LOD: 2.07 at rs913275). Sequencing of entire coding regions and intron–exon boundaries can be performed after genome-wide linkage analysis. In Bedouin kindred with autosomal recessive high myopia, genome-wide linkage analysis mapped the disease gene to 3q28 (LOD score of 11.5 at marker D3S1314). Six genes lying in the locus were subsequently sequenced and a single mutation (c. 1523G>T) in exon 10 of LEPREL1 was identified that encodes prolyl 3-hydroxylase 2 (PRHS2) which is involved in the hydroxylation of collagens.[168]

Candidate Gene Studies

Candidate genes are chosen for several reasons: presence within a myopia susceptibility locus, relevant structure and/or function or previously identified as having a critical role in refraction in animal studies. Only a few genes have been consistently replicated in myopia. Some candidate genes have been positively associated with both high and lesser severe forms of myopia,[157] thus suggesting that common pathways underpin both forms, although evidence supporting a common pathway is weak. When a myopia susceptibility locus is identified, genes lying within that locus can be sequenced, but this approach does not always result in successful gene finding.[169,170] Next-generation sequencing strategies may improve changes of gene discovery in gene regions of interest. The interplay between different biological classes of refraction-associated genes, including generic binding proteins, transcription factors, metalloproteinases and receptors, may be important.[137]

By adopting a candidate gene approach, several positive associations have been identified, especially with genes involved in ECM growth and remodeling (Table 4). In this group of genes, positive associations have been found for genes encoding collagens,[157,171] proteoglycans,[172,173] matrix metalloproteinases,[174,175] and growth factors and their receptors.[176–180] The vitamin D receptor (VDR) gene is another example of a candidate gene. It is plausible that people who develop myopia have lower levels or function of vitamin D as this nutrient is important for eye growth in animal models.[181] Polymorphisms within the VDR gene appear to be associated with low-to-moderate myopia in white individuals,[182] whereas a polymorphism in the VDR gene start codon (Fok1) has been associated with high myopia.[183] These associations have not been found in recent GWAS.

It has become apparent that most of the (nonhigh myopia) candidate gene associations reported previously represent poor quality, false-positive findings. These studies are typified by small sample sizes and unfeasibly large SNP effect sizes. In addition, even where candidate gene studies have been replicated, the replication often relates to different markers from those identified originally, with the originally associated variants not being replicated. In cases of high myopia, where one would expect a stronger genetic role, replication of initial associations has also been rare. For example, the rs1635529 polymorphism in the COL2A1 gene has been associated with myopia in several Caucasian datasets,[157] but was not replicated in a Chinese population with high myopia.[184] This may suggest that this SNP is only associated with only one ethnicity. However, lack of replication has also been seen in studies of populations of the same ethnicity. Four allele SNPs were previously associated with high myopia (alleles of rs2229336 in TGIF,[185] rs3759223 in lumican,[186] rs1982073 in TGFB1[185] and rs3735520 in HGF[178]) in Chinese people living in southeast China. However, none of these four were replicated in a study of 288 Chinese subjects with high myopia and 208 controls,[187] or other populations.[188–192] Another notable example has been PAX6, for which there have been many failed attempts at replication (see Table 4).

GWAS of Myopia

Because myopia is a complex disease in which multiple genes may contribute small effects on phenotype, case-controlled and family-based association studies represent a powerful approach to identify its associated genetic risk factors. In GWAS, myopia can be expressed as a binary trait (any degree of myopia or a category of myopia severity; e.g., high myopia) at various SE or spherical error cutoffs, or alternatively, refraction or an endophenotype of myopia (e.g., AL) can be used as quantitative traits. The number of GWAS being performed for complex diseases has increased exponentially in recent years, with an accompanying reduction in cost of the SNP-containing chips. The absence of successful replication of previously identified SNPs may reveal a type 1 error. In this type of study, the relative risk of a causal SNP (commonly presented as an OR) is usually modest in contrast to findings from linkage analyses. Most loci identified from GWAS have only been for high myopia, with limited findings being successfully replicated. Furthermore, some variants are home to chromosomal regions, including noncoding regions, without a plausible candidate gene.

Most GWAS have investigated subjects with high myopia (Table 5) but some success has been achieved when looking at subjects with less severe refractive error. The Consortium of Refractive Error and Myopia recently published their first findings.[193] This group included 31 cohorts with a total of 55,177 individuals of Caucasian (81.5%) and Asian (18.5%) ancestry. They performed a fixed-effect meta-analysis of 14 SNPs on 15q14 and five SNPs on 15q25; these regions were previously identified as being associated with human refraction in GWAS.[194,195] Within this consortium, all of the SNPs at chromosome 15q14 were significantly replicated, with the top SNP at rs634990.[193] An increased relative risk of myopia versus hyperopia was identified in homozygotes of the rs634990 risk allele (OR: 1.88; 95% CI: 1.64–2.16; p < 0.001) and for heterozygotes (OR: 1.33; 95% CI: 1.19–1.49; p < 0.001). However, a significant association between myopia and SNPs at locus 15q25 was not found. The 15q14 locus lies close to two genes: gap junction protein Δ 2 (GJD2) and actin α cardiac muscle 1, both of which may have possible roles in myopia development. GJD2 encodes the connexin 36 protein and is expressed in the retina where it is involved in transmission and processing of visual signals.[196] Actin α cardiac muscle 1 may play a role in scleral remodeling.[197] Recently, a very large GWAS study, which used questionnaire-based ascertained data on myopia status, was performed in 43,360 Europeans.[198] The results from the largest myopia GWAS to date were compelling: 19 significant (p-value: <5 × 10−8) associations with myopia were elucidated, 17 out of which were novel, which explained 2.7% of the total variance of myopia. Most of the associations (13; 68.4%) were located within or near genes known to be associated with eye development, neuronal development and signaling, the retinal visual cycle of the retina and general eye morphology: DLG2, KCNMA1, KCNQ5, LAMA2, LRRC4C, PRSS56, RBFOX1, RDH5, RGR, SFRP1, TJP2, ZBTB38 and ZIC2.

In a meta-analysis of two genome-wide datasets of Singapore Chinese subjects, Li et al. identified two variants (rs12716080 and rs6885224) in CTNND2 on 5p15 for high myopia,[199] one of which (rs6885224) was successfully replicated in a Japanese population in the same study[199] and another Chinese population.[200] CTNND2 is a neuronal-specific protein that is involved in retinal development and whose function may be regulated by paired box gene 6 (PAX6).[201] Positive GWAS findings have also been found for myopia endophenotypes. These include variants in FKBP12-rapamycin complex-associated protein (FRAP1) on chromosome 1p36.2 and PDGF receptor α on chromosome 4q12 associated with corneal curvature variation,[202] and variants in 1q41 influencing differences in AL.[79]

Gene–Environment Interaction

Myopia is a complex ocular disorder, in which several genes are likely to act in concert to control eye growth; the expression of these genes may be modified by environmental factors. Some cases of myopia are clearly familial. In these instances, myopia is typically early in onset and severe, and a pattern of inheritance can be identified among family members. In others, the mode of inheritance is heterogeneous. One common misconception is that a high heritability is incompatible with rapid changes in prevalence.[70] The higher heritability estimates for refractive error in twin compared with family-based studies[146] could be explained by gene–environment interaction. In addition, the sharp rise in the prevalence of predominantly mild-to-moderate myopia in only a few decades in east Asia argues for a strong environmental component that is associated with changes in environmental exposures that are associated with urbanization, such as increasing level of educational and decreasing time spent outdoors.[53] In urban China, the prevalence of myopia considerably higher in 15-year-old children than in parents (78.4 vs 19.8%) but children with myopic parents carried an even greater risk of myopia.[203] Evidence to support an inherited basis of school myopia, and to support an inherited susceptibility to environmental risk factors of myopia, is weak.[33] In countries with a high prevalence of myopia, a high proportion of high myopia may be acquired.[4]

Morgan and Rose have identified three questions that need to be asked when investigating the possibility of gene–environment interaction in myopia:[158]

  • Do genetic differences contribute to phenotypic variations in myopia?

  • Do environmental exposures contribute to phenotypic variations in myopia?

  • Is there evidence of differential sensitivity of the different genotypes characterized to environmental changes?

Overall, the evidence supporting gene–environment interaction in acquired myopia in humans is weak, with the strongest evidence attributed to changes in environmental pressures leading to aberrant emmetropization.[158] Support from a recent animal study has provided an important insight into gene–environment interaction. In a study of outbred White Leghorn chicks,[204] after two rounds of selective breeding for high and low susceptibility in response to 4 days of form deprivation, an established animal model for myopia,[205] chicks in the high-susceptibility groups developed approximately twice the myopia as the low-susceptibility group. There was also a significant difference between the groups in AL, with additive genetic effects explaining approximately half of the interanimal variation. These results support a role for gene–environment interaction in myopia by showing that susceptibility to environmentally induced myopia in chickens is strongly genetic in origin. As form deprivation during early life has been shown to occur in humans,[206] results from this work suggest a potentially major role for gene–environment interaction in human myopia.

Most evidence supporting a gene–environment interaction in cases of acquired myopia in humans have used surrogates of genotype, namely parental myopia and ethnicity. Evidence for a gene–environment interaction is supported by a differential effect of outdoor activity and risk of myopia according to levels of family history of myopia.[125] However, the relationship between parental myopia and measures of education or near work is more equivocal.[121,207] Furthermore, some modifiable environmental risk factors, including educational attainment, are strongly influenced by genetic factors.[92] Recently, an association between SNPs close to an MMP gene cluster and refractive error was identified after stratifying by educational attainment;[208] this may represent a gene–environment interaction. As more genetic associations are being discovered in cases of mild-to-moderate myopia, this will provide a better opportunity to investigate gene–environment interactions. Furthermore, the possibility of multiple and varied gene–gene interactions contributing to refractive phenotypes is high, and exploring this will require advanced statistical techniques in conjunction with extremely large sample sizes to permit sufficient power. It will be also necessary to collect data on environmental factors in future GWAS studies, as most current studies have limited environmental data, which could account for the dearth of literature on gene–environment interaction.