Linking obesity-associated genotype to child language development: the role of early-life neurology-related proteomics and brain myelination

Abstract

Background: The association between childhood obesity and language development may be confounded by socio-environmental factors and attributed to comorbid pathways.

Methods: In a longitudinal Singaporean mother-offspring cohort, we leveraged trans-ancestry polygenic predictions of body mass index (BMI) to interrogate the causal effects of early-life BMI on child language development and its effects on molecular and neuroimaging measures. Leveraging large genome-wide association studies, we examined whether the link between obesity and language development is causal or due to a shared genetic basis.

Findings: We found an inverse association between polygenic risk for obesity, which is less susceptible to confounding, and language ability assessed at age 9. Our findings suggested a shared genetic basis between obesity and language development rather than a causal effect of obesity on language development. Interrogating early-life mechanisms including neurology-related proteomics and language-related white matter microstructure, we found that EFNA4 and VWC2 expressions were associated with language ability as well as fractional anisotropy of language-related white matter tracts, suggesting a role in brain myelination. Additionally, the expression of the EPH-Ephrin signalling pathway in the hippocampus might contribute to language development. Polygenic risk for obesity was nominally associated with EFNA4 and VWC2 expression. However, we did not find support for mediating mechanisms via these proteins.

Interpretation: This study demonstrates the potential of examining early-life proteomics in conjunction with deep genotyping and phenotyping and provides biological insights into the shared genomic links between obesity and language development.

Funding: Singapore National Research Foundation and Agency for Science, Technology and Research.

Keywords: Language development; Neurology-related protein; Obesity; Polygenic risk score.

Fig. 1

Overarching framework for investigating the relationships between obesity-associated genotype to child language development. a) Genetic correlation and colocalisation for obesity-related traits and language-related skills. Childhood obesity-related traits include childhood BMI and childhood obesity from the Early Growth Genetics (EGG) Consortium and self-reported childhood body size from the UK Biobank. Language-related skills include word reading, nonword reading, spelling, phoneme awareness, and nonword repetition. b) Construction of polygenic risk score (PRS) for obesity based on genome-wide association studies (GWAS) of body mass index (BMI) in European (EUR) and East Asian (EAS) populations using a trans-ancestry Bayesian polygenic method, PRS-CSx; c) Longitudinal analysis of the associations of child PRS for obesity and language development assessed by Peabody Picture Vocabulary Test, Fourth Edition (PPVT-4, age 4) and the Wechsler Individual Achievement Test, Third Edition (WIAT-III, age 9), and mediation analysis via protein expression. d) Cross-sectional analysis of the associations between candidate proteins (age 8) and language-related neuroimaging measures (age 7.5). Candidate proteins were prioritised based on their association with child PRS for obesity (MSR1) and WIAT-III scores (CNTN5, EFNA4, and VWC2). For neuroimaging measures, we focused on fractional anisotropy (FA) of language-related white matter tracts; e) Construction of expression-based PRS (ePRS) for the EPH-Ephrin signalling pathway. Autosomal genes on the pathway were obtained from the REACTOME database, and the brain-region-specific genetic associations of gene expression (GEx) were obtained from the Genotype-Tissue Expression (GTEx) project. Longitudinal analysis was performed for the associations between ePRS for the EPH-Ephrin signalling pathway and WIAT-III scores (age 9).

Fig. 2

The association of child polygenic risk score (PRS) for obesity with child language scores assessed by PPVT-4 and WIAT-III from multiple linear regression (PPVT-4: N_Overall = 448, N_Boys = 224, N_Girls = 224; WIAT-III: N_Overall = 235, N_Boys = 121, N_Girls = 114). Line segments indicate 95% confidence intervals. (PPVT-4: Peabody Picture Vocabulary Test, Fourth Edition; WIAT-III: Wechsler Individual Achievement Test, Third Edition; P-values were obtained from multiple linear regression).

Fig. 3

Clusters of protein-coding genes based on protein–protein interaction. Larger nodes represent coding genes of neurology-related proteins measured in the GUSTO cohort. Smaller nodes represent coding genes of proteins that are associated with the neurology-related proteins based on the STRING database. Each edge represents the interaction between two proteins (only interactions with a STRING combined score ≥0.4 are presented).

Fig. 4

Cluster-specific forest plot for the associations of neurology-related proteins with WIAT-III language-related composite score (N = 215). The clusters with a posterior inclusion probability (PIP) higher than 0.5 and the protein with the highest conditional PIP within those clusters are highlighted in yellow. The dots and segments indicate the point estimates and 95% confidence intervals from the linear regression model.

Fig. 5

Mediation analysis for child PRS for obesity, neurology-related protein (YR8), and WIAT-III language score (YR9). Line segments indicate 95% confidence intervals. (N_Overall = 150, N_Boys = 77, N_Girls = 73).