Gene-Environment Pathways to Cognitive Development and Psychotic-Like Experiences in Children

Junghoon Park
Eunji Lee
Gyeongcheol Cho
Heungsun Hwang
Bogyeom Kim
Gakyung Kim
Yoonjung Yoonie Joo author has email address
Jiook Cha author has email address

Interdisciplinary Program in Artificial Intelligence, College of Engineering, Seoul National University, Seoul, South Korea
Department of Psychology, College of Social Sciences, Seoul National University, Seoul, South Korea
Department of Psychology, McGill University, Montréal, Canada
Department of Brain and Cognitive Sciences, College of Natural Sciences, Seoul National University, Seoul, South Korea
Institute of Data Science, Korea University, Seoul, South Korea

https://doi.org/10.7554/eLife.88117.1

Open access
Copyright information

Figures and data

Demographic characteristics of the study participants.
Of the initial 11,878 ABCD samples, we obtained data for the variables of interest for 6,602 multiethnic children. For multiethnic subjects (main analyses, n=6,602), 47.15% were female, and the parents of 70.21% were married. In European ancestry samples (sensitivity analyses, n=5,211), 46.71% were female, and the parents of 77.47% were married. Children of European ancestry had significantly different marital status (p<0.0001), lower BMI (p<0.0001) and family history of psychiatric disorders (p<0.0001) than children of other genetic ancestry. There were no significant differences in other characteristics between the two ancestry groups.
The 6,602 multiethnic participants consisted of 890 African-ancestry (13.48%), 229 Native American ancestry (3.47%), 91 East Asian ancestry (1.38%), 181 not specified (2.74%), and 5,211 European ancestry (78.93%) children. Age is rounded to chronological month. Family History of Psychiatric Disorders represents the proportion of first-degree relatives who experienced mental illness. Differences between genetic ancestry groups were calculated using χ² tests for categorical variables and t tests for continuous variables.

Demographic characteristics of the study participants.
Of the initial 11,878 ABCD samples, we obtained data for the variables of interest for 6,602 multiethnic children. For multiethnic subjects (main analyses, n=6,602), 47.15% were female, and the parents of 70.21% were married. In European ancestry samples (sensitivity analyses, n=5,211), 46.71% were female, and the parents of 77.47% were married. Children of European ancestry had significantly different marital status (p<0.0001), lower BMI (p<0.0001) and family history of psychiatric disorders (p<0.0001) than children of other genetic ancestry. There were no significant differences in other characteristics between the two ancestry groups.
The 6,602 multiethnic participants consisted of 890 African-ancestry (13.48%), 229 Native American ancestry (3.47%), 91 East Asian ancestry (1.38%), 181 not specified (2.74%), and 5,211 European ancestry (78.93%) children. Age is rounded to chronological month. Family History of Psychiatric Disorders represents the proportion of first-degree relatives who experienced mental illness. Differences between genetic ancestry groups were calculated using χ² tests for categorical variables and t tests for continuous variables.

Linear models testing genetic, socioeconomic, and environmental factors associated with cognitive intelligence and PLEs.
Standardized coefficients of a linear mixed model with CP PGS (A, B) and EA PGS (C, D). The analyses included 6,602 samples of multiethnicity. Cognitive intelligence and PLEs correlated with the PGSs, residential disadvantage, positive environment, and family SES in the opposite directions. Error bars indicate 95% bootstrapped confidence intervals with 5,000 iterations. CP and EA denote cognitive performance and education attainment, respectively; PGS, polygenic scores; SES, socioeconomic status; PLEs, psychotic-like experiences; ADI, Area Deprivation Index; Poverty, percentage of individuals below -125% of the poverty level; Years, years of residence. (*p-FDR<0.05)

Linear models testing genetic, socioeconomic, and environmental factors associated with cognitive intelligence and PLEs.
Standardized coefficients of a linear mixed model with CP PGS (A, B) and EA PGS (C, D). The analyses included 6,602 samples of multiethnicity. Cognitive intelligence and PLEs correlated with the PGSs, residential disadvantage, positive environment, and family SES in the opposite directions. Error bars indicate 95% bootstrapped confidence intervals with 5,000 iterations. CP and EA denote cognitive performance and education attainment, respectively; PGS, polygenic scores; SES, socioeconomic status; PLEs, psychotic-like experiences; ADI, Area Deprivation Index; Poverty, percentage of individuals below -125% of the poverty level; Years, years of residence. (*p-FDR<0.05)

Pathway diagram of the structural equation model of gene-environment-intelligence-PLEs pathway: Cognitive capacity PGSs, family SES, neighborhood SES, positive environment, intelligence, and PLEs.
A. Standardized path coefficients are indicated on each path as direct effect estimates (significance level *p<0.05). B. Indirect pathways to PLEs via intelligence were significant for polygenic scores, family SES, neighborhood SES, and positive environment, indicating the significant mediating role of intelligence. C. Relative proportions of direct and indirect pathways within the total effects on PLEs. The standardized effect sizes of direct pathways are colored within each bar. In A-B, child sex, genetic ethnicity, BMI, marital status, family history of psychiatric disorders, and ABCD research sites were included as covariates. CP PGS and EA PGS denote polygenic scores of cognitive performance and education attainment, respectively; SES, socioeconomic status; PLEs, psychotic-like experiences; Crystallized and Fluid, crystallized and fluid intelligence; ADI, Area Deprivation Index; Poverty, percentage of individuals below -125% of the poverty level; Years, years of residence.
Note: * indicates a statistically significant parameter estimate at α=0.05.

Pathway diagram of the structural equation model of gene-environment-intelligence-PLEs pathway: Cognitive capacity PGSs, family SES, neighborhood SES, positive environment, intelligence, and PLEs.
A. Standardized path coefficients are indicated on each path as direct effect estimates (significance level *p<0.05). B. Indirect pathways to PLEs via intelligence were significant for polygenic scores, family SES, neighborhood SES, and positive environment, indicating the significant mediating role of intelligence. C. Relative proportions of direct and indirect pathways within the total effects on PLEs. The standardized effect sizes of direct pathways are colored within each bar. In A-B, child sex, genetic ethnicity, BMI, marital status, family history of psychiatric disorders, and ABCD research sites were included as covariates. CP PGS and EA PGS denote polygenic scores of cognitive performance and education attainment, respectively; SES, socioeconomic status; PLEs, psychotic-like experiences; Crystallized and Fluid, crystallized and fluid intelligence; ADI, Area Deprivation Index; Poverty, percentage of individuals below -125% of the poverty level; Years, years of residence.
Note: * indicates a statistically significant parameter estimate at α=0.05.

IGSCA analysis of multiethnic samples.
Sex, age, genetic ethnicity, BMI, parental education, marital status of the caregiver, household income, and family’s financial adversity based on parents’ self-report, family history of psychiatric disorders, and ABCD research sites were included as covariates. Family socioeconomic status was included to confirm that the aassociations of PGS, neighborhood disadvantage, and positive environment are meaningful. SE and CI represent standard error and confidence intervals, respectively. Significant effects are marked with a star (*).

IGSCA analysis of multiethnic samples.
Sex, age, genetic ethnicity, BMI, parental education, marital status of the caregiver, household income, and family’s financial adversity based on parents’ self-report, family history of psychiatric disorders, and ABCD research sites were included as covariates. Family socioeconomic status was included to confirm that the aassociations of PGS, neighborhood disadvantage, and positive environment are meaningful. SE and CI represent standard error and confidence intervals, respectively. Significant effects are marked with a star (*).

Linear mixed model analysis of CP and EA PGS from multiethnic samples.
The variable indicating 21 ABCD research sites was set as random effect. The following covariates were included as fixed effects: sex, age, genetic ethnicity, BMI, marital status of the caregiver, and family history of psychiatric disorders. PLEs, psychotic-like experiences; CP, cognitive performance; EA, educational attainment; PGS, polygenic score; Poverty, percentage of individuals below -125% of the poverty level. 95% confidence intervals were obtained from 5,000 bootstrapping iterations. The p-values are based on two-tailed tests and false discovery rate (FDR) corrected for multiple comparisons.

Linear mixed model analysis of CP and EA PGS from multiethnic samples.
The variable indicating 21 ABCD research sites was set as random effect. The following covariates were included as fixed effects: sex, age, genetic ethnicity, BMI, marital status of the caregiver, and family history of psychiatric disorders. PLEs, psychotic-like experiences; CP, cognitive performance; EA, educational attainment; PGS, polygenic score; Poverty, percentage of individuals below -125% of the poverty level. 95% confidence intervals were obtained from 5,000 bootstrapping iterations. The p-values are based on two-tailed tests and false discovery rate (FDR) corrected for multiple comparisons.

Factor loadings for each component / factor variables of IGSCA analysis.
Factor loadings of each component / factor variable. CP PGS and EA PGS denote polygenic scores of cognitive performance and education attainment, respectively; Crystallized and Fluid, crystallized and fluid intelligence; ADI, Area Deprivation Index; Poverty, percentage of individuals below -125% of the poverty level; Years, years of residence. Significance of each indicator is marked with a star (*).

Factor loadings for each component / factor variables of IGSCA analysis.
Factor loadings of each component / factor variable. CP PGS and EA PGS denote polygenic scores of cognitive performance and education attainment, respectively; Crystallized and Fluid, crystallized and fluid intelligence; ADI, Area Deprivation Index; Poverty, percentage of individuals below -125% of the poverty level; Years, years of residence. Significance of each indicator is marked with a star (*).

Adjustment for ethnic confounding: Linear mixed model analysis of CP and EA PGS from European ancestry.
The variable indicating 21 ABCD research sites was set as random effect. The following covariates were included as fixed effects: sex, age, BMI, marital status of the caregiver, and family history of psychiatric disorders. PLEs, psychotic-like experiences; CP, cognitive performance; EA, educational attainment; PGS, polygenic score; Poverty, percentage of individuals below -125% of the poverty level. 95% confidence intervals were obtained from 5,000 bootstrapping iterations. The p-values are based on two-tailed tests and false discovery rate (FDR) corrected for multiple comparisons.

Adjustment for ethnic confounding: Linear mixed model analysis of CP and EA PGS from European ancestry.
The variable indicating 21 ABCD research sites was set as random effect. The following covariates were included as fixed effects: sex, age, BMI, marital status of the caregiver, and family history of psychiatric disorders. PLEs, psychotic-like experiences; CP, cognitive performance; EA, educational attainment; PGS, polygenic score; Poverty, percentage of individuals below -125% of the poverty level. 95% confidence intervals were obtained from 5,000 bootstrapping iterations. The p-values are based on two-tailed tests and false discovery rate (FDR) corrected for multiple comparisons.

Adjustment for ethnic confounding: IGSCA analysis of samples of European ancestry.
Sex, age, BMI, parental education, marital status of the caregiver, household income, and family’s financial adversity based on parents’ self-report, family history of psychiatric disorders, and ABCD research sites were included as covariates. Confidence intervals were derived with 5,000 bootstrap iterations. SE and CI represent standard error and confidence intervals, respectively. Significant effects are marked with a star (*).

Adjustment for ethnic confounding: IGSCA analysis of samples of European ancestry.
Sex, age, BMI, parental education, marital status of the caregiver, household income, and family’s financial adversity based on parents’ self-report, family history of psychiatric disorders, and ABCD research sites were included as covariates. Confidence intervals were derived with 5,000 bootstrap iterations. SE and CI represent standard error and confidence intervals, respectively. Significant effects are marked with a star (*).

Gene x Environment interactions analyses: Linear mixed model of CP and EA PGS from multiethnic samples.
Gene x Environment interactions between CP, EA PGS and family SES, neighborhood SES, and positive environment were tested. The variable indicating 21 ABCD research sites was set as random effect. The following covariates were included as fixed effects: sex, age, genetic ethnicity, BMI, marital status of the caregiver, and family history of psychiatric disorders. PLEs, psychotic-like experiences; CP, cognitive performance; EA, educational attainment; PGS, polygenic score; Poverty, percentage of individuals below - 125% of the poverty level. 95% confidence intervals were obtained from 5,000 bootstrapping iterations. The p-values are based on two-tailed tests and false discovery rate (FDR) corrected for multiple comparisons.

Gene x Environment interactions analyses: Linear mixed model of CP and EA PGS from multiethnic samples.
Gene x Environment interactions between CP, EA PGS and family SES, neighborhood SES, and positive environment were tested. The variable indicating 21 ABCD research sites was set as random effect. The following covariates were included as fixed effects: sex, age, genetic ethnicity, BMI, marital status of the caregiver, and family history of psychiatric disorders. PLEs, psychotic-like experiences; CP, cognitive performance; EA, educational attainment; PGS, polygenic score; Poverty, percentage of individuals below - 125% of the poverty level. 95% confidence intervals were obtained from 5,000 bootstrapping iterations. The p-values are based on two-tailed tests and false discovery rate (FDR) corrected for multiple comparisons.

Adjustment for schizophrenia PGS: Linear mixed model of CP and EA PGS from multiethnic samples.
The variable indicating 21 ABCD research sites was set as random effect. The following covariates were included as fixed effects: sex, age, genetic ethnicity, BMI, marital status of the caregiver, and family history of psychiatric disorders. PLEs, psychotic-like experiences; CP, cognitive performance; EA, educational attainment; SCZ, schizophrenia; PGS, polygenic score; Poverty, percentage of individuals below -125% of the poverty level. 95% confidence intervals were obtained from 5,000 bootstrapping iterations. The p-values are based on two-tailed tests and false discovery rate (FDR) corrected for multiple comparisons.

Adjustment for schizophrenia PGS: Linear mixed model of CP and EA PGS from multiethnic samples.
The variable indicating 21 ABCD research sites was set as random effect. The following covariates were included as fixed effects: sex, age, genetic ethnicity, BMI, marital status of the caregiver, and family history of psychiatric disorders. PLEs, psychotic-like experiences; CP, cognitive performance; EA, educational attainment; SCZ, schizophrenia; PGS, polygenic score; Poverty, percentage of individuals below -125% of the poverty level. 95% confidence intervals were obtained from 5,000 bootstrapping iterations. The p-values are based on two-tailed tests and false discovery rate (FDR) corrected for multiple comparisons.

Adjustment for unobserved confounders using null treatment approach: CP and EA PGS from multiethnic samples.
The variable indicating 21 ABCD research sites was set as random effect. The following covariates were included as fixed effects: sex, age, genetic ethnicity, BMI, marital status of the caregiver, and family history of psychiatric disorders. PLEs, psychotic-like experiences; CP, cognitive performance; EA, educational attainment; PGS, polygenic score; Poverty, percentage of individuals below -125% of the poverty level. 95% confidence intervals were obtained from 5,000 bootstrapping iterations. The p-values are based on two-tailed tests and false discovery rate (FDR) corrected for multiple comparisons.

Adjustment for unobserved confounders using null treatment approach: CP and EA PGS from multiethnic samples.
The variable indicating 21 ABCD research sites was set as random effect. The following covariates were included as fixed effects: sex, age, genetic ethnicity, BMI, marital status of the caregiver, and family history of psychiatric disorders. PLEs, psychotic-like experiences; CP, cognitive performance; EA, educational attainment; PGS, polygenic score; Poverty, percentage of individuals below -125% of the poverty level. 95% confidence intervals were obtained from 5,000 bootstrapping iterations. The p-values are based on two-tailed tests and false discovery rate (FDR) corrected for multiple comparisons.

Component correlation matrix of IGSCA analysis.
Correlation between all component / factor variables and their indicators. CP PGS and EA PGS denote polygenic scores of cognitive performance and education attainment, respectively; PLEs, psychotic-like experiences; Crystallized and Fluid, crystallized and fluid intelligence; ADI, Area Deprivation Index; Poverty, percentage of individuals below -125% of the poverty level; Years, years of residence.

Component correlation matrix of IGSCA analysis.
Correlation between all component / factor variables and their indicators. CP PGS and EA PGS denote polygenic scores of cognitive performance and education attainment, respectively; PLEs, psychotic-like experiences; Crystallized and Fluid, crystallized and fluid intelligence; ADI, Area Deprivation Index; Poverty, percentage of individuals below -125% of the poverty level; Years, years of residence.

Sign up for email alerts