Does genetic background moderate the association between parental education and school achievement?

Authors


L. Keltikangas-Järvinen, Department of Psychology, University of Helsinki, PO Box 9, FIN-00014, Finland. E-mail: Liisa.Keltikangas-Jarvinen@helsinki.fi

Abstract

This study was conducted with a purpose to examine whether the T102C polymorphism of the serotonin receptor 2A (HTR2A) gene moderates the association between parental education and children's school achievement across nine compulsory school years. The study was carried out in a population-based sample of Finnish students (aged 9, 12 and 15 years, n = 982). It was found that the HTR2A gene was not related to the school achievement at any school level, but moderated the association between maternal education and the children's grade point averages. The T/T genotype carriers benefited most from high-maternal education, and suffered from a low one more than the carriers of the other variants of the HTR2A gene. The present finding may at least partly answer the important question why academic outcomes of environmental interventions vary even at the same intelligence levels of the students.

A genetic influence on learning disabilities is widely documented, and even specific genes responsible for this influence have been identified (Plomin et al. 2007). Although the heritability of academic achievement has been shown to be almost as high as the heritability of intelligence (Petrill & Wilkerson 2000), genes associated with normal variation of learning abilities are not well understood.

The effect of achievement-predisposing environment is, instead, well documented. Higher educated parents are suggested to create environments that motivate and facilitate learning (Teachman 1987), and to involve themselves in their children's school experiences (Steinberg et al. 1992). Consequently, parental education has been constantly shown to be associated with children's school achievement and cognitive abilities (Alexander et al. 1993; Duncan et al. 1994; Patterson et al. 1990). According to a recent meta-analysis (Sirin 2005), the average effect size of parental socioeconomic position on school performance ranged between r = 0.28 and 0.30.

Indirect genetic effects, i.e. the suggestion that environmental influences on a child's academic achievement may vary as a function of the child's genotype, have been mostly ignored in the previous school achievement literature. Gene polymorphisms that are not directly associated with academic achievement might have an effect on the efficiency by which children are able to take advantage of enriching and achievement-appropriate elements of their environments. There are some studies on gene × environment interaction effects in certain cognitive abilities (van den Oord & Rowe 1997), but their role in the normal variation of school achievement has not been explored. The present study was taken with that purpose. We examined whether the T102C polymorphism of the serotonin receptor 2A (5-HTR2A) gene moderates the association between parental education and students' school achievement across the whole compulsory school attendance in a Finnish population-based sample.

The 5-HT2A receptors have been shown to play a role in a wide range of behaviors and psychopathological conditions. For instance, affective disorders (Mintun et al. 2004) and temperamental trait harm avoidance have been associated with alterations in the 5-HT2A receptors' binding capacity (Moresco et al. 2002). Polymorphic variation in the gene coding for serotonin 2A receptors has been shown to influence memory in humans (de Quervain et al. 2003) and to be related to schizophrenia (Abdolmaleky et al. 2004), anxiety-related disorders (Unschuld et al. 2007) as well as to many personality traits, e.g. impulsivity (Bruce et al. 2005). The gene coding for serotonin 2A receptors is located in the chromosome 13q14–q21 (Sparkes et al. 1991) and contains several polymorphic sites. The most studied polymorphism in the context of psychiatric and psychological outcomes is the T102C variant, which has been associated with the expression of the gene and accordingly may affect the binding potential of serotonin 2A receptors (Polesskaya & Sokolov 2002). The C allele of the T102C polymorphism of the 5HTR2A gene has been associated with lower binding potential of the 5-HTR2A receptors (Turecki et al. 1999).

We have previously shown that the T-allele carriers of T102C polymorphism of the HTR2A appear to be more sensitive to both positive and negative aspects of the environment. For example, the T-allele carriers were responsive to the protective aspects of mother's nurturance in childhood, and accordingly showed lower levels of depressive symptoms in adulthood (Jokela et al. 2007a) and benefited from high-parental socioeconomic status by showing lower levels of harm avoidance in adulthood, whereas the same was not true for C/C genotype carriers (Jokela et al. 2007c).

Based on these findings, we hypothesized that individuals carrying T alleles of the T102C polymorphism are more sensitive to achievement-appropriate atmosphere created by parental education, so that in low-educated families school performance of T/T carriers is lower than that of other allelic variations, whereas in high-educated families, the situation is the opposite. This association should remain even after controlling mother's age and maternal nurturance, which both have been shown to correlate with children's school achievement (Kuusela 2002). Mother's age correlates marginally positively with her child's school achievement, whereas maternal nurturance (a level of emotional support and closeness) modifies an effect of maternal education.

Materials and methods

Participants

The participants were 982 boys (n = 463) and girls (n = 519) participating in the ongoing population-based Cardiovascular Risk in Young Finns study (Åkerblom et al. 1991). In this study, a randomly selected sample of 3596 Finnish healthy children and adolescents from six birth cohorts (aged 3, 6, 9, 12, 15 and 18 years at the baseline in 1980) have been followed for 27 years, and monitored in seven follow-ups. Complete details of the study are given elsewhere (Åkerblom et al. 1991). In the present study, a randomly selected subsample of 1593 participants was derived for genotyping from the original sample, and of these 982 had data on parental education and maternal nurturance, and data on grade point average (GPA) at least once at the ages 9, 12 and 15 (Table 1). Genotyping was completed in 2001 when all the participants gave their written informed consent and blood samples in accordance with the Helsinki Declaration.

Table 1.  Age of participants included in the present study by birth and follow-up year
Birth yearFollow-up yearTotal
1980198319861989
AgenAgenAgenAgen
  1. The number of participants is 982, and they may contribute data at multiple ages, resulting in a total of 2019 person-observations.

1962
196515187187
19681220015200400
197192041220515191600
197491961218415178558
1977910912165274
Total 591 601 484 3432019

Measures

School achievement

School achievement was assessed on the basis of the GPA, which was calculated as the mean of school marks of all the school subjects (range from 4 [fail] to 10 [the best grade]). GPA was reported by the parents at the ages 9 and 12, and by the participants at the age of 15. Questions of school achievement were queried in four follow-up phases in 1980, 1983, 1986 and 1989. Thus, three cohorts had data on Age-9 GPA (a school report of the 3. school grade; of the 9-year primary school n = 509; 286 girls), four cohorts had data on Age-12 GPA (a certificate of the primary school; n = 754; 409 girls) and four cohorts had data on Age-15 GPA (a certificate of the junior high school n = 756; 415 girls). Thus, different numbers of participants were available at each age, as described in Table 1. Means (and SD) of GPA at the ages 9, 12 and 15 were 7.8 (0.6), 7.9 (0.7) and 7.9 (0.9), respectively, that completely fit to the average student GPAs reported by the Finnish national register of The Ministry of Education, indicating that the sample is representative of the Finnish population in terms of school achievement.

Parents' years of education

The mother's and the father's education was assessed on the basis of completed years of education (mother: mean = 10.0, SD = 3.3; father: mean = 9.8, SD = 3.7; between measures' correlation r = 0.68). In illustrating the results, mother's and father's education levels were categorized as primary (< 10 years of education), secondary (between 10 and 12 years) and tertiary (more than 12 years) education levels. In Finland, education is highly correlated with socioeconomic measures, such as occupational status. For example, in our population-based sample the correlation is around 0.6 for mothers and 0.7 for fathers.

Maternal nurturance

Maternal nurturance was self-rated at baseline (1980) by the mothers of the participants using a scale derived from the Operation Family Study addressing the emotional significance of the child for the mother. The scale comprises of four items (‘My child is emotionally important to me’; ‘I enjoy spending time with my child’; ‘I am emotionally important to my child’; ‘My child allows/enables me to fulfill myself’), which were rated on a five-point scale ranging from totally disagree (1) to totally agree (5). The Cronbach's alpha reliability was α = 0.66. The original scale was negatively skewed and was corrected in the analyses by a cubic root transformation and then standardized (mean = 0, SD = 1). At the time of assessment of maternal nurturance, the participants were 3-, 6-, 9-, 12- and 15-year olds depending on the age cohort.

Mother's age

Mother's age at baseline was determined on the basis of mother's self-report (mean = 38.1, SD = 7.8).

HTR2A 102 T > C (Rs 6313) genotyping

Genomic DNA was extracted from peripheral blood leukocytes using QIAamp DNA Blood Minikit and automated biorobot M48 extraction (Qiagen Inc., Hilden, Germany). DNA samples were genotyped by using the 5′ nuclease assay and fluorogenic TaqMan MGB probes (Livak 1999) using the ABI Prism 7900HT Sequence Detection System (Applied Biosystems, Foster City, CA, USA). The nucleotide sequences of primers and allele-specific probes, labeled with the reporter dyes FAM or VIC, were deduced from sequences deposited in the GenBank database and synthesized in conjugation with Applied Biosystems using the TaqMan® Validated SNP Genotyping Assays (assay ID C_3042197_1_). A polymerase chain reaction (PCR) containing genomic DNA, 1× Universal PCR Master Mix, 900 nM of each primer and 200 nM of each probe was performed in 96-well plates using the standard protocol in a total volume of 25 µl. After PCR amplification, end-point reading of the fluorescence signal generated from each probe was measured by the allelic discrimination analysis module, resulting in clear identification of three genotypes. Random duplicates and known control samples were run in parallel with unknown DNA samples.

Statistical analysis

The interaction effect between HTR2A and parents' education was assessed with linear regression analysis where HTR2A was coded as categorical variables (T/T genotype groups as the reference category) and parents' educational levels as continuous variables. Separate regression models were fitted for GPA at the ages 9, 12 and 15. In addition, we pooled all the person-observations from the ages 9, 12 and 15 (2019 observations of 982 unique participants) and fitted a multilevel linear regression predicting GPA at the ages 9, 12 and 15 in a single model. In multilevel modeling, SE of regression coefficients are calculated taking into account the not all observations are independent, i.e. the same participant may contribute several observations to the dataset. All models were adjusted for gender, birth year, mother's age and maternal nurturance. The multilevel model was further adjusted for age at GPA assessment.

Results

In the total sample (n = 982), allele frequencies of the T/T, T/C and C/C genotypes were 9.0%, 44.5% and 46.5%, respectively. The genotype distributions were in Hardy–Weinberg equilibrium (P > 0.05).

Table 2 shows the correlations between study variables. The HTR2A polymorphism was not associated with GPA at the ages 9, 12 or 15. Mother's and father's education was associated with GPA at all ages, and there were no gender differences in these associations (all P values > 0.40 for gender–education interaction effects).

Table 2.  Bivariate correlations between study variables
 123456789
  1. GPA = grade point average (ranging from 4 – fail to 10 – highest grade). The correlations involving the serotonin receptor 2A (HTR2A) genotype are point-biseral correlations for dichotomous contrasts of HTR2A.

  2. *P < 0.05 (minimum).

1 Gender (0 = boys , 1 = girls )1.00        
2 Birth year−0.011.00       
3 Mother's education0.040.28*1.00      
4 Father's education0.020.22*0.65*1.00     
5 Mother's age0.02−0.68*−0.23*−0.22*1.00    
6 Maternal nurturance−0.010.10*0.03−0.02−0.031.00   
7 GPA at 9−0.22*0.20*0.22*0.23*0.060.051.00  
8 GPA at 12−0.26*0.22*0.30*0.33*−0.050.030.77*1.00 
9 GPA at 15−0.29*0.16*0.22*0.29*−0.010.060.63*0.83*1.00
HTR2A: T/T vs. T/C0.000.030.010.000.040.02−0.050.00−0.01
HTR2A: T/T vs. C/C−0.010.050.01−0.010.000.02−0.03−0.020.00
HTR2A: T/C vs. C/C0.000.02−0.01−0.01−0.060.010.02−0.030.01

When mother's and father's education was used to predict GPA in the same model, both predicted GPA at the age of 9 (mother B = 0.02, SE = 0.01, P = 0.01; father B = 0.02, SE = 0.01, P = 0.04) and 12 (mother B = 0.03, SE = 0.01, P = 0.001; father B = 0.04, SE = 0.01, P < 0.001) but only father's education predicted GPA at the age of 15 (mother B = 0.01, SE = 0.01, P = 0.35; father B = 0.06, SE = 0.01, P < 0.001). Adding interaction effects between HTR2A and parents' education indicated that there was a gene–environment interaction between HTR2A and mother's education but not between HTR2A and father's education. This was marginally significant at the age of 9 and statistically significant at the age of 12 and 15 (Table 3). The interaction effect was statistically significant also in the multilevel regression model. To examine whether the strengthening of the HTR2A–maternal education interaction effect over age was statistically significant, we tested for three-way interaction between HTR2A and maternal education and age at GPA assessment in the multilevel regression model. This interaction effect was significant for T/C (B = −0.01, SE = 0.004, P = 0.01) and C/C (B = −0.01, SE = 0.004, P = 0.003) genotypes, indicating that the interaction effect strengthened with age. High-maternal nurturance predicted higher GPA at the age of 15 (Table 3), but there were no interaction effects between HTR2A and maternal nurturance on GPA at any age (all P values > 0.27), so these interaction effects were not included in the models.

Table 3.  Predicting grade point average (GPA) by parental education and serotonin receptor 2A (HTR2A) gene
Independent variablesDependent variable
Model 1: GPA at the age of 9Model 2: GPA at the age of 12Model 3: GPA at the age of 15Model 4: GPA at the ages 9, 12 and 15
  1. Values are regression coefficients (and SE in parentheses) of mutually adjusted linear regression models. Model 4 was fitted using multilevel regression modeling that pools all the participant-observations into one model (further adjusted for GPA measurement age). GPA has been assessed with a scale ranging from 4 – fail to 10 – highest grade. ***P < 0.001, **P < 0.01, *P < 0.05.

Gender (0 = boy, 1 = girl)−0.21***(0.05)−0.34***(0.05)−0.54***(0.06)−0.41***(0.04)
Birth year0.04***(0.01)0.03***(0.01)0.02*(0.01)0.04***(0.01)
Mother's education0.07**(0.03)0.09***(0.02)0.09**(0.03)0.08***(0.02)
Father's education0.00(0.03)0.03(0.02)0.07**(0.03)0.06**(0.02)
Mother's age0.01*(0.00)0.01**(0.00)0.02***(0.01)0.02***(0.00)
Maternal nurturance0.34(0.23)0.24(0.21)0.64*(0.26)0.38*(0.19)
HTR2A main effect
 T/T(Reference)(Reference)(Reference)(Reference)
 T/C0.37(0.33)0.61*(0.28)1.07**(0.34)0.79**(0.25)
 C/C0.46(0.33)0.44(0.28)1.03**(0.33)0.73**(0.25)
HTR2A × mother's education
 T/T(Reference)(Reference)(Reference)(Reference)
 T/C−0.06*(0.03)−0.07*(0.03)−0.09*(0.04)−0.07**(0.03)
 C/C−0.05(0.03)−0.07*(0.03)−0.10**(0.04)−0.07**(0.03)
HTR2A × father's education
 T/T(Reference)(Reference)(Reference)(Reference)
 T/C0.03(0.03)0.01(0.03)−0.02(0.03)−0.01(0.02)
 C/C0.01(0.03)0.03(0.03)0.00(0.03)0.00(0.02)
n5097547562019

The interaction effect between HTR2A and maternal education was illustrated by calculating the predicted GPA at the ages 9, 12 and 15 by HTR2A genotype and mother's education level group (Fig. 1). Carriers of the T/T genotype had higher GPA than that of the T/C or C/C genotype in the group of highly educated mother (Cohen's d effect sizes between T/T carriers and others were d = 0.40, 0.39 and 0.50 at the ages 9, 12 and 15, respectively). In contrast, in the group of least educated mothers the T/T carriers tended to have lower GPA than others (d = −0.15, −0.19 and −0.28 at the ages 9, 12 and 15) although this difference was statistically significant only at the age of 15. T/T genotype carriers did not differ from others in the intermediate group of mother's education (d = 0.02, −0.01 and −0.02 at the age of 9, 12 and 15).

Figure 1.

Predicted GPA at the ages 9, 12 and 15 by serotonin receptor 2A (HTR2A) genotype and mother's educational level. The vertical lines are 95% confidence intervals.

Illustrating the interaction effect the other way around, regression analyses fitted separately by HTR2A genotype group indicated that mother's education predicted GPA more strongly in carriers of the T/T genotype than in others (Table 4).

Table 4.  Linear regressions predicting grade point average (GPA) at the ages 9, 12 and 15 with mother's education separately in different serotonin receptor 2A (HTR2A) genotype groups
HTR2ADependent variable
Model 1: GPA at the age of 9Model 2: GPA at the age of 12Model 3: GPA at the age of 15Model 4: GPA at the ages 9, 12 and 15
  1. Values are regression coefficients (and SE in parentheses) of mother's education predicting GPA in linear regression models. Model 4 was fitted using multilevel regression modeling that pools all the participant-observations into one model (further adjusted for GPA measurement age). GPA has been assessed with a scale ranging from 4 – fail to 10 – highest grade. All models adjust for gender, birth year, mother's age and maternal nurturance. ***P < 0.001, **P < 0.01.

T/T0.07** (0.02)0.10*** (0.02)0.14*** (0.02)0.11*** (0.02)
T/C0.03** (0.01)0.05*** (0.01)0.04*** (0.01)0.04*** (0.01)
C/C0.03** (0.01)0.06*** (0.01)0.05*** (0.01)0.05*** (0.01)

Discussion

Previous research has focused either on genetic or on environmental influences on one's academic achievement, but their interaction effects have been studied less. This study was taken with this purpose. The present findings suggest that the academic benefits of high-parental education may depend on children's genetic backgrounds. Supporting our hypotheses, we showed that even though the T102C polymorphism of the HTR2A gene was not directly related to a school achievement, it moderated the association between mother's education and student's school performance. The T/T-genotype carriers benefited most from high-maternal education, and suffered from a low one more than carriers of the other variants of the HTR2A. This is in agreement with our previous findings suggesting that the T allele may sensitize a person to environmental effects more than the C allele (Jokela et al. 2007a,b,c). The present study indicates that the outcomes of this environmental sensitivity are not restricted to personality traits or psychiatric symptoms but extend to normal variation in school achievement.

Twin studies of parental socioeconomic status and cognitive function in children have shown that genetic influences may emerge most strongly in children with high-parental socioeconomic status (Friend et al. 2008; Harden et al. 2007; Nagoshi & Johnson 2005; Rowe et al. 1999; Turkheimer et al. 2003). The present findings provide molecular genetic evidence for a similar interaction effect, as the HTR2A gene was particularly strongly associated with school achievement among those with high-parental education. Furthermore, heritable influences on cognitive abilities are shown to vary with age, so that the genetic influences on achievement become increasingly important with age (Ceci & Williams 1997; Plomin et al. 1997; Thompson et al. 1991). In line with these findings, we observed that the environment-dependent associations between HTR2A and school achievement became stronger with age, which may reflect the increasing importance of genetic influence with age.

In Finland and many other countries, school performances of girls and boys differ significantly in all school subjects in girls' favor. However, there were no gender differences in the present gene–environment interaction effect. Despite the high correlation between parents' years of education, both the mother's and the father's education had independent associations with school achievement – with the exception of GPA at the age of 15. In Finland, maternal and paternal education has been shown to explain 38% and 32%, respectively, of the variance in students' school performance across all stages of comprehensive school, and its effect is likely to be even higher in high schools and academic education (Kuusela 2002). Finnish studies show systematically that maternal education plays a more important role in children's school performance (Kuusela 2002), suggesting that the stronger effect of paternal education observed at the age of 15 may be a coincidence and result from a high between-parents correlation. Nevertheless, the gene–environment interaction was observed only for maternal education.

The association between maternal education and children's school achievement reflects, of course, not only environmental but also genetic effects (Rowe & Rodgers 1997; Scarr 1993). Maternal education can be seen as a proxy variable reflecting a variety of genetic and environmental differences existing between families. First of all, maternal education may reflect gene–environment correlation. Certain genes may have led the mother to achieve high education. These same genes may also affect child's school performance directly and through the environment that the highly educated mother provides for her child. Secondly, maternal education has been associated with several positive environmental outcomes that may be reflected to child's learning. For example, maternal education has been linked to positive child rearing practices and child's problem solving strategies (Jones et al. 1980). Maternal education is also reflected in lower risk in outcomes such as offspring smoking and alcohol use (Singhammer & Mittelmark 2006) and mother's education has been shown to buffer against low-educational outcomes in very low-birth weight children (Hollomon et al. 1998).

Thus, given the partly shared genetic background of a mother and her child, the HTR2A–parental education interaction may reflect a gene–gene interaction effect between HTR2A and some unidentified gene rather than a gene–environment effect. However, our previous studies of the HTR2A (Jokela et al. 2007a,b,c; Keltikangas-Järvinen et al. 2008) have used several different measures of environmental factors, so it seems more plausible that these interactions reflect gene–environment interaction effects, which stem from individual differences in the sensitivity to environmental influences.

As suggested in the introduction, achievement and ability are highly correlated. Thus, a generalization of the present finding, i.e. an issue whether this gene–environment effect is specific for academic achievement or whether it is true for all intelligence-related constructs needs to be studied more.

The present study has several strengths. It benefits from a population-based sample and repeatedly measured GPAs assessed at three different points of the comprehensive school (preadolescence, early adolescence and adolescence). Several possible confounding factors were adjusted for, including maternal nurturance that has previously been shown to highly interact with the HTR2A gene (Jokela et al. 2007a). Furthermore, the homogeneity of the Finnish school system and comparable school reports offer an ‘ideal’ setting to examine school achievement that is not confounded by factors such as demographical characteristics or differences between public and private schools. Nevertheless, some limitations need to be considered. Longitudinal studies are unavoidably affected by study attrition. However, in terms of GPAs the sample was representative of the Finnish population. GPAs were self-reported, but several studies have shown the validity of self-reports in assessing school performance. Finally, the homogeneity of the Finnish school system – strength of the present study – may limit the generalizability of the results. It is known that the effects of genes are more pronounced in environmentally homogeneous settings, so replications of the present findings in different societies and different school systems are needed.

As emphasized by previous researchers (Grigorenko & Plomin 2007), behavior genetic research does not attempt to refute the role of the environment. On the contrary, findings of behavioral genetics offer significant new possibilities for a better understanding of how environmental factors affect individual differences in students' educational outcomes. As noticed by Petrill and Justice (2007), the treatment resistance, that is a child's lack of response to educational interventions, is a great concern of educational researchers. The extent to which responsiveness to interventions varies among children has received little attention. It is possible, even likely, that genetic differences may directly and indirectly explain a part of the variability in intervention responsiveness: environmental influences are experienced differently by different genotypes.

The present finding may, at least partly, answer the important question of why academic outcomes of environmental interventions vary even when children's intelligence differences are taken into account. Genetic variability should not be ignored when searching for the answer to the question why children with same intelligence levels do not similarly benefit from environmental enrichment.

Acknowledgments

This study was financially supported by the Academy of Finland (Project No. 111056 to L.K.-J.), Signe and Ane Gyllenberg's Foundation (L.K.-J. and M.H.), Niilo Helander foundation (M.H.) the Emil Aaltonen Foundation (T.L.), the Tampere University Hospital Medical Fund (T.L.) and The Finnish Cultural Foundation (T.H.).

Ancillary