48 I.G. Morgan and K.A. Rose

School or acquired myopia

The situation is different with regard to schoolor acquired myopia, which is generally later in onset and mild to moderate in severity. This form of myopia could be heterogeneous in etiology, but there is, at present, no real evidence for this. This form often appears during the school years; hence its name. In populations with little formal schooling it may hardly be present,11 while it may affect most of the school population in urban East Asia.12,13 In these parts of the world, the prevalence of myopia has increased at a rapid rate over the past few decades, which is incompatible with simple ideas of genetic determinism.5

It has been argued that this form of myopia is genetic in origin, because children with myopic parents are more likely to become myopic,14,15 and there are clear patterns of associated sibling risk.16,17 The problem with this argument is that families share environments, and it is not clear how much of the familial clustering is due to shared genes and shared environments.5,18 There is certainly evidence that when there has been rapid social change, such that the environments in which children and parents grew up were very different, as with the Eskimo and Inuit of North America during the period in which the indigenous populations were moved into settlements, parent-offspring correlations can almost completely disappear, although sibling correlations remain substantial.5,19 This shows that, at least in certain circumstances, environmental variation can overshadow genetic relationships.

It should be noted that distinguishing forms of myopia on the basis of severity alone is problematic. The massive change in prevalence of myopia in urban East Asia seems to be associated with earlier onset and more rapid progression, leading to severe myopia by the end of schooling.12 Thus high myopia in a low prevalence country may be predominantly genetic, but in a high prevalence environment such as that of Taiwan, much of the high myopia may itself be acquired, showing the same features of rapid change in prevalence over the past few decades.

Misunderstandings of Heritability and Twin Studies

Heritability from twin studies has generally been taken as strong evidence for a genetic basis to variation in myopia and refractive error, but the limitations of such studies are often not appreciated. Modern twin studies examine correlations in refractive status between genetically identical

49 Gene-Environment Interactions in the Aetiology of Myopia

monozygotic (MZ) twins, as compared to those for dizygotic (DZ) twins, which share half their genes. For additive genetic effects, the correlations between MZ twins are expected to be around twice those for DZ twins. If the MZ correlations are significantly greater than twice the DZ correlations, then this may be an indication of dominant genetic effects. Studies using this approach have produced consistently high heritability values for myopia as a category or for refractive error as a quantitative trait, of around 80–90%.20,21 This does not mean that 80–90% of myopia is genetically determined, as is sometimes erroneously stated, but that 80–90% of the variance in myopia may be cautiously attributed to genetic variation, provided that certain assumptions, such as the common environment assumption, are met.

Unfortunately, the limitation of twin studies, and heritability analysis, have often been ignored, despite the emphasis in textbooks of statistical genetics on the fact that heritability values are specific to particular populations at particular times, and cannot be easily generalized. It is easy to understand why — because in an analysis of the contribution of genetic and environmental effects to phenotypic variation, if the range of relevant environmental variation increases, then the genetic contribution must decrease, and vice versa.

To give just one example of the dogmatic interpretation of twin studies, on the basis of one of Sorsby’s early studies of twins,22 the British Medical Research Council concluded that:

“It may therefore be taken as established that the dimensions of the optical components, the efficiency of the mechanism co-ordinating the growth of the components and thus the refraction of the eye are all genetically determined. The modes of inheritance and the possibility that environmental factors have a minor modifying influence are the principal problems now awaiting clarification.”

Despite the greater sophistication of current data analysis relative to the simple analysis of Sorsby’s era, heritability analysis still depends on certain assumptions, followed by mathematical modeling, and, as with all modelling, does not, and in principle, cannot prove that the phenotype has a genetic basis. Unfortunately, this sort of claim is often made. It is important to recognize that heritability analysis is about modeling the sources of variation in the phenotype, and in the case of twin studies, the sources of variation within twin pairs. This poses a major barrier to generalization,

50 I.G. Morgan and K.A. Rose

since the characteristic of twin pairs is extremely restricted environmental variation. This probably explains why more extended family studies, where there is almost always greater environmental variation, almost always result in lower estimates of heritability.23–25

Hopper,26 a leading statistical geneticist and Director of the Australian Twin Register, has expressed some of the limitations of twin study analysis in the following way:

“Therefore, the typical lack of evidence for shared or common environment effects resulting from textbook application of classical biometrical modeling may not be a proper interpretation of reality. Such statistical analysis, biased towards a genetic explanation and then interpreted as evidence (if not proof!) that genetic factors alone are causing familial aggregation, may be misleading … It is suggested that twin researchers might benefit from taking a more critical approach to model fitting, and in particular, trying to falsify genetic hypotheses.”

One of the critical assumptions of the classical modeling is the common environment assumption, namely that the effects of environmental factors shared by twins are independent of zygosity, and hence that higher correlations within MZ pairs as compared with DZ pairs must be explained in terms of genetic differences. Attempts to test this assumption have rarely been made for any phenotype, and in any case must be trait-specific. However, recently, heritability of one of the common environmental associations of myopia, namely level of educational attainment, has been examined in the context of a study of myopia.27 The authors reported that correlations in educational achievements were lower in DZ twins than in MZ twins. But instead of interpreting this result as a falsification of the common environment assumption in the case of refractive error, the authors suggested that this was “proof” that educational outcomes were also genetically determined. This provides a logical trap, because this logic means that unless the correlation in a trait is lower in MZ twins than in DZ twins, a highly unlikely outcome, either the common environment assumption will be confirmed, or another source of genetically determined impact will be “discovered.”

There is no basis to the common misinterpretations of high heritability, such as the idea that high heritability values imply tight genetic determination and resistance to environmental impacts. There is, in fact, no

51 Gene-Environment Interactions in the Aetiology of Myopia

incompatibility between a high heritability of a trait and rapid and massive environmental change.28 Genotypes are expressed within particular environmental conditions, and even with simple Mendelian inheritance, relevant variations in environmental conditions can lead to variations in phenotype, changing the balance of genetic and environmental contributions. This is simply more likely with complex multifactorial diseases, with smaller changes in environmental exposures.

But Heritability has Its Uses

A more positive view of heritability is that high heritability values in twin studies do not prove that there is a genetic basis to any condition, but they validate the search for a genetic basis using modern molecular population genetic tools. They also provide a privileged population for genetic studies, DZ twins, where environmental variation is minimized, and average genetic variation is understood. In the absence of a reasonable heritability value from such studies, the highly expensive search for relevant genes would arguably not be justified. The proof of a genetic basis will, however, only come if genetic variation associated with variation in the trait is detected.

Evidence for Genetic Associations of School Myopia

So far, there has been little success, with little replication of studies. In a follow-up molecular study on DZ twins from the Twin Eye Study, Hammond and colleagues found a chromosomal localization suggestive of the involvement of Pax6.29 However, subsequent analysis on the British 1958 Birth Cohort did not find any association.30 In addition, most of the genes associated with syndromic and non-syndromic high myopias do not seem to contribute to more moderate myopia.31

Studies on socially isolated populations such as the Old Order Amish and orthodox Ashkenazi Jews have so far given shared chromosomal localizations associated with refractive status in the two populations, despite what would appear to be markedly different prevalence rates in the two populations. As the authors have pointed out,32 this could be explained if the genetic and environmental factors operated independently, with the environmental factors shifting the overall distribution, while the genetic factor determined the position within the distribution — in other words, if

52 I.G. Morgan and K.A. Rose

there were separate genetic and environmental risk factors, but no geneenvironment interactions. It should be noted that as myopia prevalences have changed, not only the mean, but also the shape of the distribution has changed. This could be evidence of differential sensitivity in some individuals, but it could also indicate differential exposure to whatever the environmental effects are. The authors also cautioned that because of their selection criteria (only one parent being myopic and more than one sibling being myopic), they may be looking at autosomal dominant myopia, which may make only a small contribution to the overall prevalence of school myopia.

Baird and colleagues33 have reported linkage of common myopia at 2q37.1 in three families from the GEM study. While the families were identified from probands with mild-moderate myopic, 35/49 members of the families were myopic and over 10% were highly myopic, making this a very unusual sample. It is therefore questionable whether these families are representative of most cases of common myopia, and indeed a case-con- trol association study failed to find significant associations between myopia and this locus.34

Polymorphisms in COL2A1 have also been identified as associated with syndromic high myopia.8 Mutti et al.35 reported that COL2A1 polymorphisms were also associated with myopia, but a recent analysis by Metlapally et al.36 suggests that the association is with high, rather than mild to moderate forms of myopia.

Two recent reports using candidate gene approaches are also of interest. Analysis of the SCORM data has suggested associations between genetic variation in the hepatocyte growth factor receptor c-Met37 and myopia. In this case, it is possible to calculate the prevalence of myopia in children without, or with the susceptibility allele. In children without the mutation, the prevalence is 61%; in children with one copy of the mutation, the prevalence is 68%; and in children with two copies, the prevalence rate is 66%. None of the progression rates reported in the paper are statistically different between the mutation groups. Thus, myopia develops in children in Singapore irrespective of whether they carry the mutant form of c-Met or not, but those with the mutation in c-Met become slightly more myopic. It should be noted that the variation is associated with different corneal powers, and not with different axial lengths. It will certainly be interesting to see if this variation is replicated in other populations, and what its relationship is to axial myopia. Strong impact on the prevalence myopia of polymorphisms of matrix metalloproteases has also been reported,38 with a strong dose effect.