Developmental dyslexia (DD) is known to be a hereditary neurological disorder with unbalanced development between reading ability and age, which cannot be accounted for by low intelligence, inadequate education, visual or auditory acuity deficits or other mental disorders[1]. The prevalence of DD is reported to be around 1.3–17.2% among school-age students varying in different orthographies and depending on the criteria used for diagnosis[2] and its etiology is influenced by both genetic and environmental factors[3].
Numerous studies have also suggested that environmental factors, such as birth weight, born preterm, home-literacy environment (HLE), socioeconomic status (SES) and parental education (PE), have important influences on children reading abilities’s development[4–8] The studies have highlighted the importance of family socioeconomic status in early childhood literacy development, and these findings consistently show that children from high socioeconomic status have higher reading and language skills before and after formal education than children from low socioeconomic status[9–11].
In recent years, a few studies have attempted to identify single nucleotide polymorphisms (SNPs) associated with reading or dyslexia by genome-wide association analysis study (GWAS)[12–18] There were four variants genome-wide significantly associated with reading ability and dyslexia in the previous literature[15]. However, genome-wide association analysis ignored the influence of environment on gene expression, it is essential to consider the effect of environment in the variants.
At present, gene-environment interactions (G×E) on reading ability have been rarely studied, and the sample size of gene-environment interaction studies is usually four times larger than the sample size needed to discover the main effect of genes[19]. Therefore, we wanted to explore the feasibility of the interaction between the set of SNPs in GWAS and the environment.
Although research of G×E on reading ability is little, G×E on behavior have been well-studied in recent years[20, 21]. For example, behavioral genetic studies with identical and fraternal twins have suggested that the degree of genetic influence on, or heritability of, individual differences in cognitive and reading abilities varies with SES[22] or PE[7]. The bioecological model is usually applied in twin studies that explain how does the heritability of certain phenotype vary with the environment. Friend et al. (2008) as the first twin study of G×E on reading disability found that genetic influence was higher among children whose parents had a high level of education, than those children whose parents had a lower level of education[7].
Consistently, molecular genetic studies further reported that dyslexic candidate genes’ influence on reading disability may depend on the environment. Mascheretti et al. (2013) explored five candidate genes’ (DYX1C1, DCDC2, ROBO1, KIAA0319, and TTRAP) interactions with a series of environmental factors. Results showed that children with DYX1C1 risk alleles would perform worse in adversity environment, such as low birth weight, low SES and history of smoking during pregnancy but would not be affected in positive environment[23]. Su et al. (2015) investigated G×E on orthography processing in Chinese children and found that children carrying the minor allele of rs1091047 exhibited a smaller N170 effect in low home-literacy environment than those children from high hone-literacy environment[24]. Both studies were consistent with the diathesis-stress model, which suggests that heritability for a particular behavior would be greater in poor environments while the deleterious gene would not be observed in more supportive environments and this model has been proposed to explain why certain behavioral disorders had a greater association with risk genes in environments where individuals have been exposed to a great deal of stressful life events[25, 26].
Recently, however, a theoretical alternative to the diathesis-stress model has been proposed (i.e., the differential-susceptibility model, and applied to the study of gene-environment interactions[27]. The differential-susceptibility model framework stipulates that some individuals are not only more susceptible to negative environmental factors but also to positive ones as well. According to this model, some ‘risk genes’ might be better conceptualized as ‘plasticity genes’[27]. In early literacy instruction area, Kegel, Bus, & IJzendoorn (2011) have found children had differential susceptibility of environment. Individuals with 7-repeat allele of the dopamine receptor D4 (DRD4-7R) profited most from the positive feedback of computer program, whereas they performed worst of early literacy skills in the absence of such feedback[28]. And more recently, many other molecular genetic studies of gene-environment interactions have confirmed this method[29, 30].
So far, studies on G×E effect have mainly focused on Single Nucleotide Polymorphisms (SNPs), such as SNPs from previous genome-wide association (GWA) studies[20]. GWA study is a theoretical approache to gene discovery that is meant to find phenotype associated genetic variation without regard to biological pathway and function. Therefore, the 9 single-nucleotide polymorphisms (SNPs) from previous genome-wide association (GWA) studies in different cohorts in developmental dyslexia and reading ability were selected in the present study as genetic factors on reading ability of Chinese children. We focused on parental education (PE) as a marker of environmental variable. Parental education (PE) as an environmental factor has been shown to be a strong predictor for a variety of cognitive outcomes of children[31, 32]. We predicted that if genetic and environmental influence were interdependent in reading ability, the effects of genetic signal should vary along the PE distribution. In confirmation analysis, two gene-environment interaction models (the diathesis-stress model vs. the differential-susceptibility model) were compared in a re-parameterized analysis to identify which model was a better fit to our data.