Assortative mating and parental genetic relatedness contribute to the pathogenicity of variably expressive variants

Abstract

We examined more than 97,000 families from four neurodevelopmental disease cohorts and the UK Biobank to identify phenotypic and genetic patterns in parents contributing to neurodevelopmental disease risk in children. We identified within- and cross-disorder correlations between six phenotypes in parents and children, such as obsessive-compulsive disorder (R = 0.32-0.38, p < 10^-126). We also found that measures of sub-clinical autism features in parents are associated with several autism severity measures in children, including biparental mean Social Responsiveness Scale scores and proband Repetitive Behaviors Scale scores (regression coefficient = 0.14, p = 3.38 × 10^-4). We further describe patterns of phenotypic similarity between spouses, where spouses show correlations for six neurological and psychiatric phenotypes, including a within-disorder correlation for depression (R = 0.24-0.68, p < 0.001) and a cross-disorder correlation between anxiety and bipolar disorder (R = 0.09-0.22, p < 10^-92). Using a simulated population, we also found that assortative mating can lead to increases in disease liability over generations and the appearance of "genetic anticipation" in families carrying rare variants. We identified several families in a neurodevelopmental disease cohort where the proband inherited multiple rare variants in disease-associated genes from each of their affected parents. We further identified parental relatedness as a risk factor for neurodevelopmental disorders through its inverse relationship with variant pathogenicity and propose that parental relatedness modulates disease risk by increasing genome-wide homozygosity in children (R = 0.05-0.26, p < 0.05). Our results highlight the utility of assessing parent phenotypes and genotypes toward predicting features in children who carry rare variably expressive variants and implicate assortative mating as a risk factor for increased disease severity in these families.

Keywords: assortative mating; autism; copy-number variant; family history; liability; neurodevelopment; pedigree; phenotype; polygenic; risk; variable expressivity.

Graphical abstract

Figure 1

Parental phenotypes predict clinical outcomes in children (A) Tetrachoric correlations between self-reported psychiatric disorders in mothers and fathers and male (n = 13,972–34,142) and female (n = 3,632–9,388) probands within the SPARK autism cohort. (B) Regression coefficients from linear models incorporating effects of both genetic factors and parental phenotypes toward quantitative proband autism and cognitive phenotypes (n = 1,240–1,688 for each model). RBS, Repetitive Behaviors Scale; SRS, Social Responsiveness Scale; CBCL int, Child Behavior Checklist 6–18 internalizing subscore; FSIQ, full-scale IQ. Selected models shown here; full data are presented in Table S6. Blue dots and lines indicate nominal significance (uncorrected p < 0.05). (C) Prevalence of psychiatric disorders in male and female probands within the SPARK autism cohort, stratified by their number of affected parents (n = 11–10,984). Two sample proportion tests. n.s. non-significant, ^∗ Bonferroni-corrected p ≤ 0.05, ^∗∗p ≤ 0.01, ^∗∗∗p ≤ 0.001.

Figure 2

Patterns of assortative mating in multiple cohorts (A) Percentage of 16p12.1 deletion-affected adults and their spouses who manifest five neurological and psychiatric disorders (n = 50–57). (B–D) Tetrachoric correlations of neurological and psychiatric phenotypes between (B) individuals with 16p12.1 deletion and their spouses in the 16p12.1 deletion cohort (n = 50–54), (C) mothers and fathers in the SPARK autism cohort (n = 15,094), and (D) female and male spousal partners in the UK Biobank (n = 47,636–47,670). (E and F) Pearson correlations of (E) Social Responsiveness Scale (SRS) (n = 1,822) and (F) Broad Autism Phenotype Questionnaire (BAPQ) (n = 1,814) scores corrected for 10 genetic PCs in parents of probands in the SSC autism cohort. ^∗ Bonferroni-corrected p ≤ 0.05, ^∗∗p ≤ 0.01, ^∗∗∗p ≤ 0.001.

Figure 3

Assortative mating drives increases in genetic risk variance (A) Schematic illustrating increases in disease liability variance over generations resulting from assortative mating. (B) Example families (from left to right, probands SP0006442, SP0180456, and SP0104179) from the SPARK autism cohort where inheritance of multiple variants in disease-associated genes from affected parents is associated with greater phenotypic complexity in children. (C) Average disease liability in families carrying rare variants with large effect size from a simulated population. Large white circles indicate individuals with large-effect rare variants, while small white circles indicate ∼6% of individuals with these variants. Color indicates average liability for the group. (D) Disease liability in individuals carrying large-effect variants over generations in a simulated population. Data shown from simulation with h² = 0.5, partner correlation of 0.29 (derived from the average within-disorder phenotype correlation from the SPARK and UK Biobank cohorts), 9,000 common variant loci and 1,000 rare variant loci, equal contributions from rare and common variants toward genetic liability, and a population of 100,000 individuals. ^∗∗∗One tailed t test, p < 0.001. (E) Example pedigree of assortative mating and compounding disease severity in a 16p12.1 deletion-affected family, GL_007. White indicates no overt phenotype.

Figure 4

Parental relatedness correlates with large runs of homozygosity in children Pearson correlations of (A) the top two genetic principal components (PCs) in spouses (n = 1,830), (B) Euclidean distance between spouses in the top two genetic PCs and spousal kinship coefficients (n = 1,829), and (C) parental kinship coefficients and the number of large (≥1 Mb) runs of homozygosity among male and female probands (male n = 1,678, female n = 266) in the SSC cohort.

Figure 5

Parental relatedness is a risk factor for neurodevelopmental disorders (A) Comparison of parental relatedness among SSC probands with loss-of-function mutations in Tier S SFARI genes (n = 597) and without mutations in SFARI genes (n = 1,246). (B–D) Comparisons of parental relatedness among probands ascertained for neurodevelopmental disease with (B) de novo (n = 129) and inherited (n = 116) CNVs, (C) deletions (n = 249) and duplications (n = 97), and (D) syndromic (n = 47) and variably expressive (n = 299) CNVs. Lines indicate median. One-tailed t test.