Genetic overlap between psychotic experiences in the community across age and with psychiatric disorders

This study explores the degree to which genetic influences on psychotic experiences are stable across adolescence and adulthood, and their overlap with psychiatric disorders. Genome-wide association results were obtained for adolescent psychotic experiences and negative symptom traits (N = 6297–10,098), schizotypy (N = 3967–4057) and positive psychotic experiences in adulthood (N = 116,787–117,794), schizophrenia (N = 150,064), bipolar disorder (N = 41,653), and depression (N = 173,005). Linkage disequilibrium score regression was used to estimate genetic correlations. Implicated genes from functional and gene-based analyses were compared. Mendelian randomization was performed on trait pairs with significant genetic correlations. Results indicated that subclinical auditory and visual hallucinations and delusions of persecution during adulthood were significantly genetically correlated with schizophrenia (rg = 0.27–0.67) and major depression (rg = 0.41–96) after correction for multiple testing. Auditory and visual subclinical hallucinations were highly genetically correlated (rg = 0.95). Cross-age genetic correlations for psychotic experiences were not significant. Gene mapping and association analyses revealed 14 possible genes associated with psychotic experiences that overlapped across age for psychotic experiences or between psychotic experiences and psychiatric disorders. Mendelian randomization indicated bidirectional associations between auditory and visual hallucinations in adults but did not support causal relationships between psychotic experiences and psychiatric disorders. These findings indicate that psychotic experiences in adulthood may be more linked genetically to schizophrenia and major depression than psychotic experiences in adolescence. Our study implicated specific genes that are associated with psychotic experiences across development, as well as genes shared between psychotic experiences and psychiatric disorders.


Supplementary Figures and Supplementary Note
Genetic overlap between psychotic experiences across age and psychiatric disorders   Table S1. Genetic covariance estimates from LD score regression……………………………………………………16 Table S8. MR-Egger intercept test and Cochran Q statistics……………………………………………………………..17 Table S9. Sensitivity analyses: Mendelian randomization with instrumental variables selected at p < 5           Note: Manhattan plot displayed in outer most ring with loci colour coded according to the amount of LD shared with lead independent SNPs as follows: red (r 2 > 0.8), orange (r 2 > 0.6), green (r 2 > 0.4) and blue (r 2 > 0.2). Genomic risk loci are displayed in blue on the chromosome ring (second and third layers). Genes mapped by chromatin interaction are displayed in orange, by eQTLs in green, and by both chromatin interaction and eQTLs in red; Mapped genes that overlapped between phenotypes are listed in Table 3.    Note: ρg = Genetic covariance; GenCov intercept = LD score regression genetic covariance intercept (prior to constraining SNP heritability intercepts) that reflects degree of sample overlap; Genetic covariances are reported in instances where covariance estimates could not be standardised into genetic correlations due to low sample size or SNP heritability. Genetic covariances gives an indication of the presence and direction of genetic overlap but not the magnitude of effect and these results should be interpreted accordingly.   Four schizotypy scales were used to assess psychotic experiences during middle adulthood: Perceptual aberrations were assessed with the Perceptual Aberration Scale (4) with 35 true/false items devised to assess experiences in the general population that resemble clinical features of schizophrenia with an emphasis on body image aberrations including unclear body boundaries, body size and physical attributes being distorted, or feelings of estrangement from one's own body. Items also assessed unusual visual and auditory experiences, for example "My hearing is sometimes so sensitive that ordinary sounds become uncomfortable" and "Sometimes when I look at things like tables and chairs, they seem strange".
Hypomania was from the Hypomanic Personality Scale (5) and consisted of 48 true/false items devised to assess hypomania, gregariousness, grandiosity and euphoria (e.g. "I can usually slow myself down when I want to" and "I have often been so excited about an involving project that I didn't care about eating or sleeping").
Two scales from Chapman's Schizotypia Scales were employed to assess social anhedonia with the Revised Social Anhedonia Scale and physical anhedonia from the Revised Physical Anhedonia Scale (6), devised to assess the inability to take pleasure from physical (61 true/false items, e.g. "One food tastes as good as another to me") and social (40 true/false items, e.g. "I prefer watching television to going out with other people") stimuli respectively.
Summary statistics from linear regression GWAS performed on these four schizotypy scales were obtained from the authors (7). SNPs were identified at p <1 x 10 -5 for PE, at p <1 x 10 -6 for MDD, and at p <1 x 10 -8 for schizophrenia and bipolar disorder (p-value thresholds were set to allow for more than 20 independent SNPs to be analysed) within a 250kb window at r 2 < 0.1 based on LD structure in the 1000 Genomes phase 3 reference panel for individuals of European decent.

Positive psychotic experiences assessed in adults
Annotation of functional consequences associated with independent lead SNPs and SNPs obtained from the reference panel that are in LD with independent SNPs (at r 2 ≥ 0.6) was performed using ANNOVAR (11) (based on Ensembl genes build version 92) whilst excluding the extended MHC region (25,000,000-35,000,000). ANNOVAR is a software tool used to identify whether SNPs are associated with protein coding or amino acid changes.
Annotation are based on several sources of information such as gene or splicing site locations, mRNA sites, genomic region-based information such as conserved regions and predicted transcription factor binding sites, stable RNA secondary structures or microRNA target sites. ANNOVAR offers the utility to use several public databases for a range of functional annotations as well as options on which to filter variants, such as SIFT scores for non-synonymous mutations. Based on user-defined gene definition databases like Ensembl, ANNOVAR annotates each variant to indicate its position in relation to genes (for instance, whether the variant is exonic, intronic, within a splicing site, upstream or downstream from a gene). For non-synonymous single nucleotide variants or indels, amino acid changes are also annotated. Precomputed functional importance scores, such as CADD scores (12), that indicate how likely a variant would have deleterious consequences, can also be annotated to variants. Based on these variant annotations, ANNOVAR offers the option to automate the process of gene mapping according to user-defined parameters.
Mapping of variants to the most likely causal genes was performed by employing a combination of positional mapping, expression quantitative trait loci (eQTL) mapping and 3D chromatin interaction mapping using the following parameters. Gene mapping was performed on lead independent SNPs and SNPs from the 1000 Genomes reference panel for individuals of European descent that were in LD with lead SNPs at r 2 > 0.6. For positional mapping, variants located within 10kb of known gene regions were mapped to genes if likely to be deleterious based on a CADD score ≥ 12.37 (12). eQTL mapping of SNPs to genes were performed based on significant eQTL associations at a false discovery rate (FDR) < 0.05 obtained from 13 brain regions from GTEx v7 brain tissue repository and 10 from GTEx v6 (13,14). SNPs were mapped to genes based on significant chromatin interactions obtained from high-resolution HiC datasets for fetal and adult human brain samples (15) and for the dorsolateral prefrontal cortex and the hippocampus from GSE87112 at the recommended FDR of p < 1 x 10 -6 250kb upstream and 500kb downstream from the transcription start site (16). Promoter and enhancer regions were annotated from the Roadmap 111 epigenomes brain tissue for 13 brain regions (17,18). Additionally, parameters in FUMA was set to map variants within protein-coding regions only.