Modifiable pathways and epigenetic markers for the effects of childhood maltreatment on neuropsychiatric disorders : a mendelian randomization study

Abstract

Childhood maltreatment (CM) is prevalent globally with long-term consequences. Traditional observational studies are unable to draw causal associations between exposure to CM and long-term outcomes or delineate the causal mediators. This study aims to confirm the causal relationships between CM and neuropsychiatric disorders, explore the role of biological, behavioral, cognitive, and socioeconomic traits as mediators, and identify the potential influence of DNA methylation (DNAm).

This study leveraged the Mendelian randomization (MR) framework. Summary-level data obtained from genome-wide association studies based on European ancestry from the UK Biobank (N = 143,473), the Psychiatric Genomics Consortium (up to N = 500,199), and other large studies were used for analyses (up to N = 766,435). Univariable MR was performed to evaluate the causal effect of genetically predicted CM on the risk of neuropsychiatric disorders. Mediation MR was conducted to obtain the mediating effects of 27 mediators: two-step MR and the product-of-\coefficients method obtained the indirect effect of CM acting through each mediator; and multivariable MR to determine the direct effects of CM following adjustment of all significant mediators. The inverse variance weighted method was the main estimate. Sensitivity analyses (weighted median, MR-PRESSO, Qhet MVMR) were conducted to evaluate MR assumptions. Epigenetic MR was conducted to determine the potential role of DNAm, with colocalization as a sensitivity analysis.

CM was significantly associated with higher risks of attention-deficit/hyperactivity disorder (ADHD) (OR: 10.093, 95% CI: 4.761 to 21.395), major depressive disorder (MDD) (OR: 1.888, 95% CI: 1.321 to 2.698), and schizophrenia (OR: 5.889, 1.458 to 23.782). Mediation analysis revealed three addictive behaviors (smoking initiation, leisure screen time, and substance abuse), three cognitive traits (executive functioning, intelligence, and risk tolerance), and one socioeconomic trait (educational attainment) as significant mediators. Significant mediators mediated 4.969% to 20.375% of the relationship between CM and ADHD, 6.514% to 9.903% of the relationship between CM and MDD, and 4.135% to 22.174% of the relationship between CM and schizophrenia. Following adjustment for all significant mediators, the direct effect of CM on ADHD attenuated substantially (OR: 2.574, 95% CI: 1.759 to 3.766), while the direct effect of CM on schizophrenia was slightly smaller than the total effect (OR 5.100, 95% CI: 2.709 to 9.604). Contrastingly, the direct effect on MDD remained unchanged despite adjusting for mediators (OR 1.948, 95% CI: 1.676, 2.264). DNAm levels at three, four, and 19 CpG sites were associated with ADHD, MDD and schizophrenia, respectively. Of the 11 CpG sites mapped to the nearest genes, three (CLU, MAPT, HNRNPK) were involved in the formation of neurofibrillary tangles.

In this study, genetically predicted CM had robust causal effects on the risks of ADHD, MDD, and schizophrenia. Significant mediators span behavioral, cognitive, and socioeconomic traits. As such, multi-modal interventions targeting these factors may reduce the long-term consequences of CM. However, the direct effect of CM on psychiatric outcomes, especially on MDD, is indisputable. Therefore, preventative strategies for reducing CM prevalence should still be of top priority.