Abstract
Purpose:
To test the hypothesis that genes known to cause clinical syndromes featuring myopia also harbor polymorphisms contributing to nonsyndromic refractive errors.
Methods:
Clinical phenotypes and syndromes that have refractive errors as a recognized feature were identified using the Online Mendelian Inheritance in Man (OMIM) database. One hundred fifty-four unique causative genes were identified, of which 119 were specifically linked with myopia and 114 represented syndromic myopia (i.e., myopia and at least one other clinical feature). Myopia was the only refractive error listed for 98 genes and hyperopia and the only refractive error noted for 28 genes, with the remaining 28 genes linked to phenotypes with multiple forms of refractive error. Pathway analysis was carried out to find biological processes overrepresented within these sets of genes. Genetic variants located within 50 kb of the 119 myopia-related genes were evaluated for involvement in refractive error by analysis of summary statistics from genome-wide association studies (GWAS) conducted by the CREAM Consortium and 23andMe, using both single-marker and gene-based tests.
Results:
Pathway analysis identified several biological processes already implicated in refractive error development through prior GWAS analyses and animal studies, including extracellular matrix remodeling, focal adhesion, and axon guidance, supporting the research hypothesis. Novel pathways also implicated in myopia development included mannosylation, glycosylation, lens development, gliogenesis, and Schwann cell differentiation. Hyperopia was found to be linked to a different pattern of biological processes, mostly related to organogenesis. Comparison with GWAS findings further confirmed that syndromic myopia genes were enriched for genetic variants that influence refractive errors in the general population. Gene-based analyses implicated 21 novel candidate myopia genes (ADAMTS18, ADAMTS2, ADAMTSL4, AGK, ALDH18A1, ASXL1, COL4A1, COL9A2, ERBB3, FBN1, GJA1, GNPTG, IFIH1, KIF11, LTBP2, OCA2, POLR3B, POMT1, PTPN11, TFAP2A, ZNF469).
Conclusions:
Common genetic variants within or nearby genes that cause syndromic myopia are enriched for variants that cause nonsyndromic, common myopia. Analysis of syndromic forms of refractive errors can provide new insights into the etiology of myopia and additional potential targets for therapeutic interventions.
There is growing recognition that myopia represents a significant public health issue.
1–3 In relation to developing preventive public health strategies and potential treatments, this makes understanding the etiology of myopia all the more important. There has been a long-running debate about whether myopia is predominantly genetic or environmental.
4,5 The rising prevalence of myopia over recent decades has led to recognition that environmental effects must play an important role. Yet there remains strong evidence that genetic factors also make a significant contribution. The first clear evidence for a genetic role in myopia came from twin studies, with identical (monozygotic) twins showing a higher similarity (concordance) in refractive error than nonidentical (dizygotic) twins.
6,7 Such studies have estimated heritability for myopia to be as high as 91%. Observed associations between parent refraction and that of their children may also have a genetic contribution.
8–11 Conventional genetic linkage analysis studies revealed a number of potential loci for myopia-related genes, but the causative genes have mostly proved elusive.
12–14
More recently, genome-wide association studies (GWAS) have attempted to identify genes associated with the risk of developing myopia. Early GWAS studies that focused on a handful of plausible candidate genes reported association between common gene variants and myopia or high myopia, using sample sizes of a few hundred participants; however, the results showed limited replication in later studies.
14,15 In contrast, a move to sample sizes in the thousands, coupled with a much more systematic examination of common genetic variants across the whole genome, has proved more successful, with more than 100 gene loci identified and excellent concordance between studies.
16–19 However, these loci together explain only a small percentage of the observed variation in refractive error (e.g., 2.3% in a study of children
20 and 10% in a sample of adults
17). The difference between the heritability estimates from twin and GWAS studies has been described as the “missing heritability” or “heritability gap” in myopia.
2,21 This difference indicates either that twin studies have greatly overestimated the genetic contribution to refractive error, or a large number of refractive error–related genes have yet to be identified.
To date, one potential source of genetic variation in myopia has largely been overlooked, namely syndromic myopia, that is, myopia associated with at least one other medical condition. Many medical syndromes are genetic in origin, with clearly defined Mendelian inheritance patterns and positively identified genes. Some of the better-known forms of syndromic myopia are linked with genes that have a plausible role in myopia pathogenesis. For example Stickler's syndrome is most commonly associated with defects in the gene for collagen type II alpha-1 (
COL2A1), which is expressed in the sclera, a structure that displays significant alterations in myopia.
22,23 What has been lacking to date is a comprehensive analysis of the genetic basis of syndromic myopia and its potential broader relevance to refractive error development.
The primary aim of the present study was to examine the hypothesis that a proportion of human myopia results from polymorphisms in genes known to cause myopic syndromes. To achieve this aim, we compared the locations of identifiable syndromic myopia genes with data from two large GWAS studies that examined refractive errors and myopia (specifically, those carried out by CREAM
17 and 23andMe
16). Our second aim was to determine whether any specific biological processes that might be associated with the genetic control of refractive error could be identified from analyzing the functions of genes linked to syndromic myopia.
We identified clinical phenotypes that have refractive errors as a recognized feature using the Online Mendelian Inheritance in Man database (OMIM, Johns Hopkins University, Baltimore, MD, USA). OMIM is an expert-curated, comprehensive database of human genetic disorders and provides information about both the clinical features and specific causative genes where identified.
The entire OMIM database was downloaded in text format for off-line analysis following registration of this project with Johns Hopkins University (
http://www.omim.org; in the public domain). Purpose-written software (predominantly written in AWK, a Unix scripting language) was used to process this text file to extract clinical syndromes that included any form of refractive error within the clinical features. The extraction of syndromes with refractive errors was performed on the basis of the phenotypic descriptors within OMIM. The latter are based on a clinical categorization from the wide variety of the cited papers for each condition, and therefore no consistent or specific dioptric thresholds could be derived. The list of terms used is shown in
Table 1. The resulting set of phenotypic entries was then reviewed to ensure the reliability of the automated extraction process.
Table 1 Search Terms Used to Detect Refractive Error Within OMIM Entries
Table 1 Search Terms Used to Detect Refractive Error Within OMIM Entries
Two additional OMIM database files (morbidmap and mim2gene.txt) were then used to determine if the phenotypes were associated with a specific gene and to extract the relevant NCBI Gene ID, Ensembl Gene ID, and the HGNC Approved Gene Symbol. The final dataset of OMIM-derived ametropia syndromes included OMIM entry number, clinical synopsis, type of refractive error, cytogenetic locus, Ensembl Gene ID, and HGNC (HUGO Gene Nomenclature Committee) nomenclature. All gene names referred to in this paper are the HGNC descriptors.
A total of 219 clinical entities were identified within the OMIM database as conditions featuring ametropia, and in 167 cases at least one causative gene had been identified at the time of analysis. Of these 167 genetically defined, ametropia-featuring phenotypes, 162 were syndromic forms of refractive error, with at least one other phenotypic feature, and 5 were known Mendelian forms of high myopia. Myopia was the most commonly listed refractive error (n = 130), followed by hyperopia (n = 43) and astigmatism (n = 23). Myopia was the only type of refractive error described in 107 phenotypes. In 23 phenotypes, myopia was listed along with one or more other types of refractive error. Overall, 154 unique genes linked to various types of refractive error were identified within this group of conditions. The phenotypic subset associated with myopia (n = 130) was found to represent a set of 119 unique genes since 7 genes (COL2A1, COL11A1, FBN1, LTBP2, POMT1, POMT2, POMGNT1) had two or more associated phenotypes, leading to 11 duplicates. Hyperopia was associated with 42 unique syndromic genes, and astigmatism with 23 unique syndromic genes. Of the 119 myopia-associated genes, 107 were autosomal and 12 were on the X chromosome.
A comparison of the chromosomal locations of the complete set of 154 refractive error–associated genes with the genes identified in the CREAM Consortium GWAS analysis
17 is shown in
Figure 1. The complete list of ametropia-associated genes identified within OMIM is given in Supplementary Table S1.
The biological processes associated with the 154 OMIM genes linked to refractive errors were examined using the PANTHER Overrepresentation Test, with Bonferroni correction for multiple tests. This approach estimates both the overrepresentation of a specific category of genes and the associated
P value. For the complete group of ametropia genes, a wide range of biological processes was identified, as shown in
Table 2. This table lists the identified biological process, the observed number of genes associated with that process, the expected number of such genes, the enrichment ratio, and the associated corrected
P value. It includes all biological processes that were found to be statistically significant and where the level of gene overrepresentation was at least a factor of 10.
Table 2 Biological Processes Overrepresented Among 154 OMIM Genes Known to Cause Syndromic Ametropia (PANTHER Analysis)
Table 2 Biological Processes Overrepresented Among 154 OMIM Genes Known to Cause Syndromic Ametropia (PANTHER Analysis)
A similar analysis for genes specifically associated with myopia (rather than any type of ametropia) also revealed a wide range of biological processes (see
Table 3). Several of these processes are highly plausible candidates for a role in the pathogenesis of myopia, such as collagen organization and metabolism, while others have not previously been associated with myopia development, such as protein mannosylation, glycosylation, and Schwann cell differentiation.
Table 3 Biological Processes Overrepresented Among 119 OMIM Genes Known to Cause Syndromic Myopia (PANTHER Analysis)
Table 3 Biological Processes Overrepresented Among 119 OMIM Genes Known to Cause Syndromic Myopia (PANTHER Analysis)
There is considerable overlap between the overrepresented biological processes found in myopia-related genes and the broader group of ametropia-related genes, as might be expected since myopic syndromes are the largest subgroup. Nevertheless, several processes either are unique to the myopic syndromes, or show greater overrepresentation when compared to the ametropic group. The four processes that are significantly overrepresented in myopia but not in the ametropia group are protein O-linked mannosylation (GO process: 0035269), collagen fibril organization (GO process: 0030199), lens development in camera-type eye (GO process: 0002088), and gliogenesis (GO process: 0042063). Thirteen processes are highly overrepresented (>10-fold) within the myopic syndromes to a greater degree than is observed within the ametropic syndromic genes: protein O-linked glycosylation (GO process: 0006493), Schwann cell differentiation (GO process: 0014037), collagen metabolism processes (GO processes: 0030574 and 0032963), visual processing (GO processes: 0009584, 0007601, 0050953), extracellular matrix organization and structure (GO processes: 0030198 and 0043062), and nonspecific metabolic processes (GO processes: 0044243, 0044259, 0044236).
Figure 2 shows the individual myopia-related genes associated with these overrepresented biological processes and their links to those processes. As shown, several genes are linked to more than one biological process.
A different pattern of biological processes was identified in syndromes associated with hyperopia. There were just four processes showing a greater than 10-fold overrepresentation (P < 0.001 for each process) over that expected from an equally random sample of human genes: sensory organ morphogenesis, camera-type eye development, eye development, and sensory organ development. All four processes were also identified in myopic and ametropic syndromes, but with a lower level of overrepresentation. For syndromes that had astigmatism as a feature there were 23 unique genes, but no significant enrichment was found using the PANTHER analysis.
An analysis using DAVID identified enrichment of genes involved in extracellular matrix (ECM) receptor interaction (P = 5.4E-03), as well as those involving O-mannosyl glycan biosynthesis (P = 8.9E-03) and focal adhesion (P = 3.5E-02). Finally, pathway analysis using REACTOME identified enrichment for signaling by PDGF (platelet-derived growth factor; P = 6.3E-06), axon guidance (P = 2.8E-04), and integrin cell surface interactions (P = 2.1E-03).
By their very nature, GWAS cannot determine whether a particular gene has a causal role in the condition being studied. Instead, when a genetic locus is identified by GWAS analysis, there may be one or more causative genes nearby, for example, within 100 kb. Often, the risk-determining polymorphism has a regulatory influence on the expression of one or more nearby genes.
34 Nonetheless, for several of the genes identified by the CREAM Consortium and 23andMe, there is supporting evidence for a causal role in refractive error development; for example,
PRSS56 is associated with microphthalmia in humans and reduced eye size in knockout mice.
37,38 SIX6 is another gene involved in ocular morphogenesis.
39 RDH5 is involved in retinoic acid metabolism and all-trans retinoic acid has been identified as a potential intraocular signaling molecule in a mammalian model of myopia.
40 LAMA2 and
BMP2 are involved in ECM remodeling, and
BMP2 also appears to be an important intraocular signaling molecule for controlling eye growth.
41 In contrast, all of the OMIM syndromic genes found to be statistically linked with myopia in the CREAM Consortium and 23andMe GWAS datasets (
Tables 4,
5) have already been determined to have a direct role in producing myopia in humans.
The large number of genes identified from syndromic forms of myopia and other refractive errors also provided opportunity to study the biological mechanisms involved in refractive error by looking for patterns of overrepresentation of pathways within this set of genes. We analyzed GO terms overrepresented among OMIM syndromic ametropia and myopia genes (
Tables 2,
3) to gain insight into underlying biological processes. GO terms are used to annotate sequences, genes, or gene products in biological databases, and represent a controlled vocabulary designed to facilitate the integration of information across species and biological disciplines. Although OMIM is not one of the databases directly contributing to the GO project, it is conceivable that researchers developing databases that do contribute to the GO project have utilized OMIM in their own work. This could potentially have led to some circularity in our pathway analysis, whereby one biological database (e.g., OMIM) is used to harness information from another database (e.g., the GO database). However, since OMIM is not a direct contributor to the GO project, we consider that any circularity would have little impact on the majority of our findings.
Many of the identified pathways (as shown in
Fig. 2 and
Tables 2,
3), such as connective tissue composition and ocular morphogenesis fit with prior hypotheses related to myopia development.
42,43 Indeed, for the pathway analysis using REACTOME, all of the enriched pathways—namely, PDGF signaling, axon guidance, and integrin cell surface interactions—have already been implicated in myopia pathogenesis.
17,44–49 Using other bioinformatics tools (PANTHER), several additional novel pathways were identified as being very overrepresented, including protein glycosylation, protein mannosylation, lens development, gliogenesis, and Schwann cell differentiation. This form of analysis also identified well-recognized pathways that may be involved in myopia development, such as those affecting collagen metabolism and other ECM components.
The largest OMIM analysis group related to genes linked to myopia (119 out of 154 ametropia genes). Thus the ametropia analysis is heavily weighted toward myopia, and there is considerable overlap in the biological processes that reached statistical significance when comparing myopia and all forms of refractive error. It is therefore doubtful that we can reliably distinguish pathways relating to ametropia per se from myopia. Nonetheless, in relation to whether the observed enrichment patterns in hyperopia and myopia make biological sense, the hyperopia-related genes showed a very different pattern of process enrichment from myopia-related genes, with emphasis on processes involved in early growth and organogenesis. This mirrors a fundamental difference between hyperopia and myopia in that hyperopia in adults usually represents a persistence of a refractive error that has existed since infancy, as compared to myopia, which, in the vast majority of cases, is acquired during childhood and early adult years. The smaller number of pathways found for hyperopia, as compared to myopia, is most likely to be the result of the smaller number of genes available for analysis. A sensitivity analysis of random subsamples of 46 out of the 119 myopes (to match the number of hyperopia-associated genes) showed a mean number of significant processes (P < 0.05 and >10× enrichment) reduced to 8.8 as compared to 22 with the full set of myopia genes (n = 119).
There are some identifiable, albeit speculative, links between these novel pathways and ocular development. Glycosylation and mannosylation are processes whereby macromolecules, such as proteins, are modified by the addition of simple sugars (glucose or mannose). In relation to glycosylation, high glucose levels in diabetes appear to promote myopia, perhaps by disrupting such pathways by nonenzymatic glycosylation.
50 Primary glycosylation deficits (such as congenital disorder of glycosylation type Ia) cause significant brain dysfunction as well myopia, suggesting that a neuronal pathway may be involved.
51 A similar association is seen with deficits in O-mannosylation, as demonstrated in Walker-Warburg syndrome (OMIM phenotype 253280). While the link with myopia remains unexplained, mannosylation is important for cell–matrix interactions and communication and therefore has a plausible role in the process by which neural elements within the eye influence the growth of its connective tissue components, including the sclera.
52
There is experimental evidence that lens plays a role in eye growth. Experimental lensectomy in infant monkeys has been demonstrated to slow eye growth in monkeys less than 7.5 months of age, but not in older animals.
53 It is therefore possible that genes influencing lens development may have an impact on early ocular growth and hence final refraction. Changes in lens thickness during later childhood have also been proposed to be an important factor in the development of myopia.
54 Glial cells have received relatively little attention in relation to the development of refractive errors, but certainly merit more detailed attention. One of the CREAM-identified myopia risk genes (
PRSS56) has recently been found to be expressed in retinal Müller cells and to have a role in refractive error development in mice (Paylakhi S, et al.
IOVS 2017;58:ARVO E-Abstract 5635). Abnormal glial cell–derived myelination of retinal ganglion cells (i.e., myelinated retinal nerve fibers) is also associated with myopia development, though the mechanisms remain unexplained.
55 Finally, within the eye Schwann cells are associated with the intrinsic neurons of the choroid,
56 a structure implicated in ocular growth regulation.
Compared to studies of the genetic basis of myopia, hyperopia has received far less attention. A smaller number of syndromic hyperopia genes were identified in this analysis, but they were linked to different biological processes, principally involved in organogenesis. This fits in with the general clinical picture that most hyperopia is congenital and most myopia develops during postnatal growth.
The authors thank members of the CREAM Consortium for providing GWAS summary statistics.
Supported by a Wellcome Trust ISSF Populations Pilot Award (Grant 508353/509506).
Disclosure: D.I. Flitcroft, None; J. Loughman, None; C.F. Wildsoet, None; C. Williams, None; J.A. Guggenheim, None