Keywords

FormalPara Key Points

  • While the recent global rise of myopia prevalence is primarily attributable to environmental changes, within populations inherited factors play a large role in explaining why some individuals are affected by myopia while others are not.

  • Early efforts to identify the specific genes underlying the heritability of refractive error used linkage and candidate gene designs to identify up to 50 loci and genes, although most remain unconfirmed.

  • As the sample size in genome-wide association studies (GWAS) has increased, the number of implicated loci has risen steadily, with 161 variants reported in the latest meta-analysis.

  • Interrogation of loci uncovered by GWAS offers insight into the molecular basis of myopia—for example, pathway analysis implicates the light induced retina-to-sclera signaling pathway in myopia development.

  • Although many loci have been uncovered by GWAS, statistical modelling shows there are many more genes to find—identifying these will further illuminate the molecular pathways leading to myopia and open up new avenues for intervention.

5.1 Introduction

This chapter addresses the scientific exploration of the genetic architecture of myopia. Myopia is the most common eye condition worldwide and its prevalence is increasing. Changes in environmental conditions where time spent outdoors has reduced relative to previous generations are the main hypothesized culprit. Despite these environmental trends, within populations, myopia is highly heritable; genes explain up to 80% of the variance in refractive error. Initial attempts to identify myopia genes relied on family studies using linkage analysis or candidate gene approaches with limited progress. For the last decade, genome-wide association study (GWAS) approaches have predominated, ultimately resulting in the identification of hundreds of genes for refractive error and myopia, providing new insights into its molecular machinery. Thanks to these studies, it was revealed that myopia is a complex trait, with many genetic variants of small effect influencing retinal signaling, eye growth and the normal process of emmetropization. However, the genetic architecture and its molecular mechanisms are still to be clarified and while genetic risk score prediction models are improving, this knowledge must be expanded to have impact on clinical practice.

Some sections of this report follow the framework described in a recent International Myopia Institute Genetics report by Tedja et al. [1]

5.2 Heritability

The tendency for myopia to run in families has long been noted, suggesting genetic factors play a role in determining risk [2]. While family studies show familial aggregation, twin studies are required to reliably separate the effects of genes and familial environment [3,4,5,6]. Benchmarking of the relative contribution of genetics and environment is done by computation of heritability, the proportion of the total trait variance (here, spherical equivalent) due to additive genetic factors. Since the contributions of genes and environment can vary across human populations, heritabilities are population and even time specific [7, 8]. The influence of environmental variance is well illustrated in the case of the heritability study in Alaskan Eskimos, where the rapid introduction of the American school system dramatically increased the contribution of the environment. As a result heritability estimates, computed based on families where the parents are less educated relative to their offspring, were very low at this time (10%) [7].

Across most human populations, environment is fairly constant and the estimates of spherical equivalent heritability are usually high (~80%) [9,10,11]. Although the aggregate contribution of genetic factors to variation in refractive error is high, initial studies were unable to determine the genetic architecture of myopia—that is, is myopia caused by rare mutations of large effect? Or is most variation driven by common variants, each with individually small effect on risk? With the advent of genotyping arrays, it became possible to estimate the aggregate effect of all common variants, with “array heritability” estimates of 35% from the ALSPAC study. Such estimates place a lower bound on the proportion of the heritability that is attributable to genetic variants which are common in the population. The remaining 45% (80%–35%) is likely attributable to rare genetic variants, to variants not covered by genotyping arrays or to non-additive genetic effects.

5.3 Syndromic Myopia

Syndromic myopia is generally monogenic and can occur within a wide spectrum of clinical presentations. This type of myopia is usually accompanied by other systemic or ocular disorders. Table 5.1 summarizes all syndromic and ocular conditions that present with myopia [12]. We are able to learn about myopia development by investigating these syndromes. For instance, several types of heritable syndromes result in extreme axial elongation, due to abnormalities in the development of connective tissue (i.e. Marfan syndrome, OMIM #154700; Stickler syndrome, OMIM #,108300 #604841, #614134, #614284 and Ehlers–Danlos syndrome, OMIM #225400, #601776). Similarly, inherited retinal dystrophies lead to myopia due to defects in photoreceptors, for instance, in X-linked retinitis pigmentosa (mutations in RPGR-gene) and congenital stationary night blindness [13].

Table 5.1 Overview of syndromic forms of myopia
Full size table
Table 5.1A Ocular conditions associated with myopia
Full size table

Interestingly, several syndromic myopia genes were found in association to other ocular traits, such as CCT (ADAMTS2, COL4A3, COL5A1, FBN1) [14], and Fuchs’s dystrophy (TCF4) [15]. However, the majority of the genes causing syndromic myopia have not been linked to common forms of myopia, except for COL2A1 [16, 17] and FBN1 [18, 19]. Nevertheless, a recent study found an overrepresentation for syndromic myopia genes in GWAS studies on refractive error and myopia [20], implying their important role in myopia development.

5.4 Linkage Studies

Linkage studies have been successfully applied for many Mendelian disorders, although the success has been much more limited in complex traits. The linkage approach searches for cosegregation of genetic markers with the trait of interest in pedigrees [21]. Families with genetic variants which show an autosomal dominant inheritance pattern were also most successful for myopia linkage studies. Up to now, 20 MYP loci [22,23,24,25] and several other loci [26,27,28,29,30,31] are identified for (high) myopia. Fine mapping led to candidate genes, such as the IGF1 gene located in the MYP3 locus [32]. Linkage using a complex inheritance design found five additional loci [33,34,35,36,37].

Validation of candidate genes often resulted in no association, but other variants appeared associated with the non-Mendelian, common form of myopia. This hints towards a genetic overlap between Mendelian and common myopia [38]. As the GWAS era progressed, linkage studies fell by the wayside. Nevertheless, segregation analyses combined with linkage and next generation sequencing (i.e. whole exome sequencing) of regions in pedigrees with high myopia are, in theory, expected to facilitate the discovery of rare variants with large effects; an aspect which cannot be distilled from GWAS.

5.5 Candidate Gene Studies

In candidate gene studies the focus is on a gene with suspected biological, physiological or functional relevance to myopia, in particular high myopia. Although sometimes effective, candidate gene studies are limited by their reliance on this existing knowledge. Table 5.2 summarizes candidate gene studies on (high) myopia. Particularly notable are genes encoding extracellular matrix-related proteins (COL1A1, COL2A1 [16, 17] and MMP1, MMP2, MMP3, MMP9, MMP10 [59, 60]). For candidates such as PAX6 and TGFB1, the results were replicated in multiple independent extreme/high myopia studies and validated in a large GWAS meta-analysis in 2018, respectively [18, 76]. However in most other cases, the results were not independently validated: LUM and IGF1 failed to confirm an association [77, 78]. Interestingly, in a few cases the candidates were subsequently implicated in GWAS of other ocular traits: TGFβ2 and LUM for central corneal thickness (CCT), a glaucoma and keratoconus endophenotype [14], PAX6 with optic disc area [79] and HGF [80].

Table 5.2 Summary of candidate gene studies reporting positive association results with myopia
Full size table

5.6 Genome-Wide Association Studies

Generally, linkage studies are limited to identification of genetic variants with a large effect on myopia [81]. Given the limited number of genes identified by linkage, it became apparent in the 2000s that identifying large numbers of additional myopia genes was more practical with genome-wide association studies (GWASes), since it has dramatically higher statistical power. GWASes have greatly enhanced our knowledge of the genetic architecture of (complex) diseases [82]. Most of the variants found via GWAS reside in non-exonic regions and their effect sizes are typically small [82, 83]. For GWAS, 200 k–500 k genetic markers are usually genotyped and a further >10 million “imputed”, taking advantage of the correlation structure of the genome. This approach is most effective for common variants (allele frequencies >0.01 in the population, although with larger reference panels, rarer alleles can also be detected).

Initially, GWASes for myopia were performed as a dichotomous outcome (i.e. case-control, Table 5.3). Since myopia constitutes a dichotomization of the quantitative trait spherical equivalent, considering the quantitative trait should be more informative for gene mapping. The first GWASes for spherical equivalent were conducted in 2010 [96, 97], with ~4000 individuals required to identify the first loci. The first loci to reach the genome-wide significance threshold (P < 5 × 10−8, the threshold reflecting the large number of statistical tests conducted genome-wide) were markers near the RASGFR1 gene on 15q25.1 (P = 2.70 × 10−9) and markers near GJD2 on 15q14 (P = 2.21 × 10−14). A subsequent analysis combining five cohorts (N = 7000) identified another locus at the RBFOX1 gene on chromosome 16 (P = 3.9 × 10−9) [98].

Table 5.3 Summary of case-control design GWASs and their highest associations with myopia
Full size table

These early efforts made it clear that individual groups would have difficulty in mapping many genes for spherical equivalent, motivating the formation of the Consortium for Refractive Error and Myopia (CREAM) in 2010, which included researchers and cohorts from the USA, Europe, Asia and Australia. They replicated SNPs in the 15q14 loci [99], which was further affirmed by other studies on both spherical equivalent and axial length alongside with the replication of the 15q25 locus [100, 101].

In 2013, two major GWAS meta-analyses on refractive error traits (spherical equivalent and age of spectacle wear) identified 37 novel loci (Table 5.4), with robust replication of GJD2, RFBOX1 and RASGFR1 in both meta-analyses. The first was the collaborative work of CREAM based on a GWAS meta-analysis on spherical equivalent, comprising 35 individual cohorts (N European = 37,382; N SoutheastAsian = 12,332) [108]. 23andMe, a direct-to-consumer genetic testing company, performed the second major GWAS, replicating 8 of the novel loci found by CREAM and identifying another 11 novel loci based on a GWAS survival analysis of age of spectacle wear in 55,177 participants of European ancestry. To the surprise of some in the field, the effect sizes and direction of the effects of the loci found by these two groups were concordant despite the difference in phenotype definition and in scale: dioptres for CREAM and hazard ratios for 23andMe [109]. Subsequently, replication studies provided validation for the associated loci and highlighted two other suggestive associations. At this point, the implicated loci explained 3% of the phenotypic variance in refractive error [108, 110].

Table 5.4 Overview of the 37 novel loci found in 2013 by CREAM and 23andMe and subsequent replication
Full size table

The CREAM and 23andMe studies represented a large increase in sample size over the initial GWASs. Their meta-analysis approach was very effective in discovering new loci. This motivated joined efforts of CREAM and 23andMe, which resulted in a GWAS meta-analysis including 160,420 participants. Moreover, a denser imputation reference set was used (1000G phase 1 version 3), enabling better characterization of genetic variations. Although CREAM and 23andMe used different phenotypes (spherical equivalent and age at first spectacle wear, respectively) again the results were concordant and the new findings were replicated in an independent cohort with refractive error available (UK biobank, comprising 95,505 participants). Overall, this GWAS increased the number of risk loci to 161, explaining 7.8% of the phenotypic variance in refractive error. Very large sample sizes (millions) will be required to identify all of the loci contributing to myopia risk.

The genetic correlation was estimated to be 0.78 between European and Asian ancestry, suggesting that despite (1) large differences in the rate of myopia between these groups and (2) differences in the genetic ancestry of these groups, most of the genetic variation is in common. Figure 5.1 provides the chronological discovery of all associated loci and Fig. 5.2 shows the effect sizes of the established 161 loci.

Fig. 5.1
figure 1

Historic overview of myopia gene finding. Overview of myopia gene finding in historic perspective. Genes identified using whole exome sequencing are marked as purple. Other loci (linkage studies, GWAS) are marked as red

Full size image
Fig. 5.2
figure 2

Effect sizes of common and rare variants for myopia and refractive error. Overview of SNPs and annotated genes found in the most recent GWAS meta-analysis [18]. X-axis displays the minor allele frequency of each SNP; y-axis displays the effect size of per individual SNP. The blue dots represent the novel loci discovered by Tedja et al. [18] and the pink dots represent the loci found by Verhoeven et al. [108], which now have been replicated

Full size image

Several “endophenotypes” have been considered for myopia: spherical equivalent, axial length, corneal curvature and age of diagnosis of myopia. Axial length is a well-studied “endophenotype” which correlates strongly with refractive error. The first GWAS of axial length considered 4944 Asian ancestry individuals and identified a locus at 1q41. A subsequent meta-analysis combining data on 12,531 European and 8216 Asian ancestry individuals uncovered a further eight genome-wide significant loci at RSPO1, C3orf26, LAMA2, GJD2, ZNRF3, CD55, MIP, ALPPL2, as well as confirming the 1q41 locus. Five of the axial length loci were also associated loci for refractive error. GWASs performed for corneal curvature [104, 111,112,113,114] identified the loci FRAP1, CMPK1, RBP3 and PDGFRA; in the case of PDGFRA, associations have also been found with eye size. A study in 9804 Japanese individuals and replication in Chinese and European ancestry cohorts analysed three myopia-related traits (refractive error, axial length and corneal curvature). They replicated the association of GJD2 and refractive error as well as the association of SNPs in WNT7B for axial length and corneal curvature [114, 115]

5.7 Pathway Analysis Approaches

GWAS approaches improve our understanding of the molecular basis of traits by mapping individual loci. However, it is possible to place such loci into a broader context by applying pathway analysis approaches. In myopia, a retina-to-sclera signaling cascade has been postulated, but the specific molecular components were unclear. Recent GWASs have uncovered genes which lie along this pathway [108, 110, 116]—genetic changes at individual loci only make small changes to phenotype but collectively these perturbations are responsible for larger changes in the retina-to-sclera signaling cascade, ultimately explaining differences in refractive error from individual to individual.

Pathways inferred from the first large-scale CREAM GWAS [108, 110] included neurotransmission (GRIA4), ion transport (KCNQ5), retinoic acid metabolism (RDH5), extracellular matrix remodelling (LAMA2, BMP2) and eye development (SIX6, PRSS56). The 23andMe GWAS identified an overlapping set of pathways: neuronal development (KCNMA1, RBFOX1, LRRC4C, NGL-1, DLG2, TJP2), extracellular matrix remodelling (ANTXR2, LAMA2), the visual cycle (RDH5, RGR, KCNQ5), eye and body growth (PRSS56, BMP4, ZBTB38, DLX1) and retinal ganglion cell (ZIC2, SFRP1) [117]. When considered in the context of known protein–protein interactions, many genes in these pathways are related to growth and cell cycle pathways, such as the TGF-beta/SMAD and MAPK pathways [118].

The most recent meta-analysis combining data from CREAM and 23andMe data taken together confirmed previous findings and offered additional insights [18]. In a gene-set analysis, several pathways were highlighted including “abnormal photoreceptor inner segment morphology” (Mammalian Phenotype Ontology (MP) 0003730); “thin retinal outer nuclear layer” (MP 0008515); “nonmotile primary cilium” (Gene Ontology (GO) 0031513); “abnormal anterior-eye-segment morphology” (MP 0005193) and “detection of light stimulus” (GO 0009583). The genes implicated in this large-scale GWAS were distributed across all cell types in the retina-to-sclera signaling cascade (neurosensory retina, RPE, vascular endothelium and extracellular matrix, Fig. 5.3). The larger GWAS also suggested novel mechanisms, including angiogenesis, rod-and-cone bipolar synaptic neurotransmission and anterior-segment morphology. Interestingly, a novel association was found at the DRD1 gene, supporting previous work linking the dopamine pathway to myopia.

Fig. 5.3
figure 3

Schematic overview of known function in retinal cell types of refractive error and syndromic myopia genes according to literature. Bold: genes identified for both common refractive error and in syndromic myopia

Full size image

5.8 Next Generation Sequencing

GWAS approaches have been highly effective in assessing the role of common variants in myopia but such methods cannot effectively characterize very rare genomic variants. Whole exome sequencing (WES) allows investigation of rare variants in exonic regions; due to cost, applications to date have primarily been in family studies or studies of early onset high myopia.

Studies employing WES to date have either focused on family designs (e.g. particular inheritance patterns such as X-linkage or conditions such as myopic anisometropia) or case-control studies of early onset high myopia [119,120,121,122]. The WES-based approaches identified several novel mutations in known myopia genes (Table 5.5). For instance, Kloss et al. [131] performed WES on 14 families with high myopia, identifying 104 genetic variants in both known MYP loci (e.g. AGRN, EME1 and HOXA2) and in new loci (e.g. ATL3 and AKAP12) [131]. In the family studies, most variants displayed an autosomal dominant mode of inheritance [119, 123, 124, 130] although X-linked heterozygous mutations were found in ARR3 [126].

Table 5.5 Overview of genes and their mutations found by next generation sequencing
Full size table

Both retinal dystrophies and ocular development disorders coincide with myopia. Sun et al. [132] investigated if there was a genetic link by evaluating a large number of retinal dystrophy genes in early onset high myopia. They examined 298 unrelated myopia probands and their families, identifying 29 potentially pathogenic mutations in COL2A1, COL11A1, PRPH2, FBN1, GNAT1, OPA1, PAX2, GUCY2D, TSPAN12, CACNA1F and RPGR with mainly an autosomal dominant pattern.

5.9 Environmental Influences Through Genetics

Although myopia is highly heritable within specific cohorts, dramatic changes in environment across many human populations have led to large changes in prevalence over time [133,134,135,136]. The role of changes in socioeconomic status, time spent outdoors, education and near-work are now well established as risk factors for myopia, based on observational studies [137,138,139]. Education has proven the most influential and consistent factor, with a doubling in myopia prevalence when attending higher education compared to finishing only primary education [140,141,142]. There are two main areas where genetic studies can inform our understanding of the role of environment. Firstly, gene–environment studies can highlight where interactions exist. Secondly, observational studies only establish association and not causation—in some circumstances genetic data can be used to strengthen the case for an environmental risk factor causally (or not) influencing myopia risk (Mendelian randomization).

Gene–environment (GxE) interaction analyses examine whether genes operate differently across varying environments. GxE studies in myopia have focused primarily on education. An early study in North American samples examined GxE for myopia and the matrix metalloproteinases genes (MMP1-MMP10): a subset of SNPs were only associated with refraction in the lower education level [58, 59]. A subsequent study in five Singapore cohorts found variants in DNAH9, GJD2 and ZMAT4, which had a larger effect on myopia in a high education subset [143]. Subsequent efforts to examine GxE considered the aggregate effects of many SNPs together. A study in Europeans found that a genetic risk score comprising 26 genetic variants was most strongly associated with myopia in individuals with a university level education [144]. A study examining GxE in children considered near work and time outdoors in association with 39 SNPs and found weak evidence for an interaction with near work [144, 145]. Finally, a CREAM study was able to identify additional myopia risk loci by allowing for a GxE approach [19].

Mendelian randomization (MR) infers whether a risk factor is causally associated with a disease. MR exploits the fact that germline genotypes are randomly assigned at meiosis, to enable a “natural” randomized controlled trial. Since the assigned genotypes are independent of non-genetic confounding and are unmodified by disease processes, MR offers a better assessment of causality than that available from observational studies [146, 147].

Two MR studies found a causal effect of education on the development of myopia. One of the MR studies tested for causality bi-directionally [148]. Both found a larger effect through MR than that estimated from observational studies suggesting that confounding in observational studies may have been obscuring the true relationship [149]. As expected, there was little evidence of myopia affecting education (−0.008 years/dioptre, P = 0.6). Another study focused on the causality of low vitamin D on myopia due to controversy in the literature [150]. The estimated effects of vitamin D on refractive error were so small (Caucasians: −0.02 [95% CI −0.09, 0.04] dioptres (D) per 10 nmol/l increase in vitamin D concentration; Asians: 0.01 [95% CI −0.17, 0.19] D per 10 nmol/l increase) that the authors concluded that the true contribution of vitamin D levels to degree of myopia is probably zero and that previous observational findings were likely confounded by the effects of time spent outdoors.

5.10 Epigenetics

Epigenetics in refractive error and myopia is postulated to be important due to the known effects of environmental factors on refractive error and myopia development. Nevertheless, this field is still developing and some characteristics of epigenetics render it a difficult issue to unravel. Epigenetic features can be influenced by environmental factors and are time dependent and tissue specific. This complicates the study of these effects, since myopia and refractive errors develop during childhood and young adolescence and obtaining eye tissue, preferably retinal and scleral would be unethical. Furthermore, although some epigenetic processes are conserved across species, this is not always the case: making animal studies not always translational to humans.

Non-Coding RNAs and Myopia

The latest GWAS meta-analysis found 31 of 161 loci residing in or near regions transcribing (small) noncoding RNAs, thus hinting towards the importance of epigenetics [18, 151]. MicroRNAs, or miRNAs, are the best-characterized family of small non-coding RNAs. They are approximately 19–24 nucleotides in length in their mature form. They are able to bind to 3′ UTR regions on RNA polymers by sequence-specific post-transcriptional gene silencing; one miRNA can regulate the translation of many genes. MiRNAs have been a hot topic in the last years due to their potential clinical application: the accessibility of the retina for miRNA-based therapeutic delivery has great potential for the prevention and treatment of retinal pathologies [152]. Up to now, there have only been a handful of studies on miRNA and its role in myopiagenesis in humans, these are summarized in Table 5.6.

Table 5.6 Overview of microRNAs associated with myopia
Full size table

5.11 Implications for Clinical Management

Due to the high polygenicity of myopia and low explained phenotypic variance by genetic factors (7.8%), clinical applications derived from genetic analyses of myopia are currently limited. Risk predictions for myopia in children are based on family history, education level of the parents, the amount of outdoor exposure and the easily measurable refractive error and axial length.

Currently, we are able to make a distinction between high myopes and high hyperopes based on the polygenic risk scores derived from CREAM studies: persons in the highest decile for the polygenic risk score had a 40-fold-greater risk of myopia relative to those in the lowest decile.

A prediction model including age, sex and polygenic risk score achieved an AUC of 0.77 (95% CI = 0.75–0.79) for myopia versus hyperopia in adults (Rotterdam Study I–III) [18]. This AUC is similar to that achieved by modelling environmental factors only; the AUC for myopia incidence in a European child cohort was 0.78 considering parental myopia, 1 or more books read per week, time spent reading, no participation in sports, non-European ethnicity, less time spent outdoors and baseline AL-to-CR ratio [156]. To date, one study has assessed both environmental and genetic factors together and showed that modelling both genes and environment improved prediction accuracy [157]. Although these efforts to improve prediction are promising, a prediction-based approach will only be beneficial if randomized controlled trials of atropine therapy show that children with persistent myopic progression benefit from an earlier and higher dose of atropine administration. The additional costs of genetic testing and potentially invasive regime (collecting blood from children) also need to be taken into account.

5.12 Concluding Remarks

The scientific community has discovered more than 200 loci associated with myopia and its endophenotypes with a variety of approaches (linkage, candidate gene, GWAS, post-GWAS gene-based associations, next generation sequencing approaches, gene environment interactions and epigenetic approaches). With the rise of large biobanks, such as the UK Biobank [158], further GWAS meta-analyses between large consortia and companies will enable identification of many more genes. This will allow full elucidation of the molecular mechanism of myopiagenesis. Whole genome sequencing approaches will replace both GWAS and WES, and will elucidate the genetic structure which regulates the function of the myopia risk variants.

To fully understand the underlying mechanisms, the focus should lie on unraveling the genetic and epigenetic architecture of myopia by exploring interactions and effects of other “omics” in relevant tissue, i.e. multi-omics. This concept includes incorporation of methylomics, transcriptomics, proteomics and metabolomics. Future projects should focus on gathering more omics data on eye tissue. Next to the multi-omics approach, modelling gene–environment effects will tell us more about the genetic key players which are also susceptible to the environment. Furthermore, future functional studies interrogating the candidate genes and loci will point us to therapeutic solutions for myopia management.