Genetic association of FMRP targets with psychiatric disorders

Abstract

Genes encoding the mRNA targets of fragile X mental retardation protein (FMRP) are enriched for genetic association with psychiatric disorders. However, many FMRP targets possess functions that are themselves genetically associated with psychiatric disorders, including synaptic transmission and plasticity, making it unclear whether the genetic risk is truly related to binding by FMRP or is alternatively mediated by the sampling of genes better characterised by another trait or functional annotation. Using published common variant, rare coding variant and copy number variant data, we examined the relationship between FMRP binding and genetic association with schizophrenia, major depressive disorder and bipolar disorder. High-confidence targets of FMRP, derived from studies of multiple tissue types, were enriched for common schizophrenia risk alleles, as well as rare loss-of-function and de novo nonsynonymous variants in schizophrenia cases. Similarly, through common variation, FMRP targets were associated with major depressive disorder, and we present novel evidence of association with bipolar disorder. These relationships could not be explained by other functional annotations known to be associated with psychiatric disorders, including those related to synaptic structure and function. This study reinforces the evidence that targeting by FMRP captures a subpopulation of genes enriched for genetic association with a range of psychiatric disorders.

Fig. 1. Schizophrenia association of gene sets ranked by FMRP binding confidence in four tissue types.

Data of FMRP binding were derived from crosslinking immunoprecipitation studies of mRNA targets in mouse cortex, mouse hippocampal CA1 pyramidal neurons, healthy human frontal cortex and human embryonic kidney (HEK) 293 cells. Genes were ranked by FMRP binding confidence and grouped into bins of 400 genes. Presented are −log₁₀(P), where the P value is derived from gene set association analysis using the genetic variant type shown. CNV analyses were corrected for P value inflation using random size-matched sets of expressed genes. Rare coding variants were derived from case-control exome sequencing studies of schizophrenia and defined as variants observed once in all sequenced samples and never in the non-psychiatric component of ExAC. Loss-of-function (LoF) variants include nonsense, splice site and frameshift mutations. Nonsynonymous (NS) variants include loss-of-function and missense variants. Dotted lines indicate a threshold for statistical significance after correction for the number of bins. SNPs single nucleotide polymorphisms, CNVs copy number variants.

Fig. 2. Pathway analysis workflow.

Predominant functional subsets of FMRP targets were tested for genetic association with psychiatric disorders. GO gene ontology, MP mammalian phenotype, FDR false discovery rate.

Fig. 3. Common variant association of FMRP target bins with schizophrenia, bipolar disorder, major depressive disorder and Alzheimer’s disease.

Data of FMRP binding in four tissue types were derived from the sources shown. Presented is −log₁₀(P value) following common variant gene set association analysis of bins of 400 genes ranked by FMRP binding confidence. Dotted lines represent a threshold for statistical significance after correction for the number of bins. Data presented in the top row of panels (a, e, i, m) are duplicated from Fig. 1.