ARTICLE https://doi.org/10.1038/s41467-019-12283-6 OPEN Associations of autozygosity with a broad range of human phenotypes David W Clark et al.# In many species, the offspring of related parents suffer reduced reproductive success, a phenomenon known as inbreeding depression. In humans, the importance of this effect has 1234567890():,; remained unclear, partly because reproduction between close relatives is both rare and frequently associated with confounding social factors. Here, using genomic inbreeding coefficients (FROH) for >1.4 million individuals, we show that FROH is significantly associated (p < 0.0005) with apparently deleterious changes in 32 out of 100 traits analysed. These changes are associated with runs of homozygosity (ROH), but not with common variant homozygosity, suggesting that genetic variants associated with inbreeding depression are predominantly rare. The effect on fertility is striking: FROH equivalent to the offspring of first cousins is associated with a 55% decrease [95% CI 44–66%] in the odds of having children. Finally, the effects of FROH are confirmed within full-sibling pairs, where the variation in FROH is independent of all environmental confounding. *email: jim.wilson@ed.ac.uk. #A full list of authors and their affiliations appears at the end of the paper. NATURE COMMUNICATIONS | (2019)10:4957 | https://doi.org/10.1038/s41467-019-12283-6 | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12283-6 G iven the pervasive impact of purifying selection on all FROH is associated with reduced reproductive success (decreased populations, it is expected that genetic variants with large number and likelihood of having children, older age at first sex deleterious effects on evolutionary fitness will be both rare and first birth, decreased number of sexual partners), as well as and recessive1. However, precisely because they are rare, most of reduced risk-taking behaviour (alcohol intake, ever-smoked, self- these variants have yet to be identified and their recessive impact reported risk taking) and increased disease risk (self-reported on the global burden of disease is poorly understood. This is of overall health and risk factors including grip strength and heart particular importance for the nearly one billion people living in rate). We show that the observed effects are predominantly populations where consanguineous marriages are common2, and associated with rare (not common) variants and, for a subset of the burden of genetic disease is thought to be disproportionately traits, differ between men and women. Finally, we introduce a due to increased homozygosity of rare, recessive variants3–5. within-siblings method, which confirms that social confounding Although individual recessive variants are difficult to identify, the of FROH is modest for most traits. We therefore conclude that net directional effect of all recessive variants on phenotypes can inbreeding depression influences a broad range of human phe- be quantified by studying the effect of inbreeding6, which gives notypes through the action of rare, recessive variants. rise to autozygosity (homozygosity due to inheritance of an allele identical-by-descent). Results Levels of autozygosity are low in most of the cohorts with Cohort characteristics. As expected, cohorts with different genome-wide data7,8 and consequently very large samples are demographic histories varied widely in mean FROH. The within- required to study the phenotypic impact of inbreeding9. Here, we cohort standard deviation of FROH is strongly correlated with the meta-analyse results from 119 independent cohorts to quantify mean (Pearson’s r = 0.82; Supplementary Fig. 3), and the most the effect of inbreeding on 45 commonly measured complex traits homozygous cohorts provide up to 100 times greater per-sample of biomedical or evolutionary importance, and supplement these statistical power than cosmopolitan European-ancestry cohorts with analysis of 55 more rarely measured traits included in UK (Supplementary Data 5). To categorise cohorts, we plotted mean Biobank10. FROH against FIS (Fig. 2). FIS measures inbreeding as reflected by Continuous segments of homozygous alleles, or runs of non-random mating in the most recent generation, and is cal- homozygosity (ROH), arise when identical-by-descent haplotypes culated as the mean individual departure from Hardy–Weinberg are inherited down both sides of a family. The fraction of each equilibrium (FSNP; see Methods). Cohorts with high rates of autosomal genome in ROH > 1.5 Mb (FROH) correlates well with consanguinity lie near the FROH = FIS line, since most excess SNP pedigree-based estimates of inbreeding11.We estimate FROH using homozygosity is caused by ROH. In contrast, cohorts with small standard methods and software6,12 for a total of 1,401,776 indi- effective population sizes, such as the Amish and Hutterite iso- viduals in 234 uniform sub-cohorts. The traits measured in each lates of North America, have high average FROH, often despite cohort vary according to original study purpose, but together avoidance of mating with known relatives, since identical-by- cover a comprehensive range of human phenotypes (Fig. 1, descent haplotypes are carried by many couples, due to a Supplementary Data 7). The five most frequently contributed restricted number of possible ancestors. traits (height, weight, body mass index, systolic and diastolic blood pressure) are measured in >1,000,000 individuals; a further 16 traits are measured >500,000 times. Traits affected by FROH. To estimate the effect of inbreeding on We find that FROH is significantly associated with apparently each of the 100 phenotypes studied, trait values were regressed on deleterious changes in 32 out of 100 traits analysed. Increased FROH within each cohort, taking account of covariates including Trait group 1,250,000 Anthropometry Blood pressure Cognition 1,000,000 Haematology Electrocardiology Sample size Fertility 750,000 Blood lipids Behavioural Well-being 500,000 Lung function Female reproductive Renal function 250,000 Liver enzymes Inflammatory Glycaemic 0 Ocular Height Body mass index Weight Waist-Hip ratio Grip strength Birth weight Diastolic blood pressure Education attained Reaction time Cognitive g Haemoglobin White blood cell count Platelet count Lymphocytes (%) Monocytes (%) Heart rate QT interval PR interval QRS duration Number of children Ever had children Age at first birth Total cholesterol HDL cholesterol Triglycerides LDL cholesterol Ever smoked Ever married Self-reported risk taker Driving speeding Age at first sex Number of sexual partners Self-reported overall health Walking pace Frequency of vigorous activity Facial ageing Hearing acuity FEV1 FEV1/FVC Age at menarche Age at menopause eGFR Uric acid Alanine transaminase Gamma-Glutamyl transferase Hs-CRP Fibrinogen Interleukin−6 Tumour necrosis factor alpha Fasting plasma glucose Haemoglobin A1c Fasting insulin Spherical equivalent refraction Visual acuity Systolic blood pressure Mean platelet volume Alcohol units per week Fig. 1 Census of complex traits. Sample sizes are given for analyses of 57 representative phenotypes, arranged into 16 groups covering major organ systems and disease risk factors. HDL high-density lipoprotein, LDL low-density lipoprotein, hs-CRP high-sensitivity C-reactive protein, TNF-alpha tumour necrosis factor alpha, FEV1 forced expiratory volume in one second, FVC forced vital capacity, eGFR estimated glomerular filtration rate 2 NATURE COMMUNICATIONS | (2019)10:4957 | https://doi.org/10.1038/s41467-019-12283-6 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12283-6 ARTICLE autozygosity also decreases the likelihood of having children in Cohort ancestry the subset of individuals who are, or have been, married, Polynesia east African Hispanic S & W Asian (OR0.0625 = 0.71 ± 0.09, p = 3.8 × 10−8) it appears that the cause East Asian Japanese is a reduced ability or desire to have children, rather than reduced 0.06 European Mixed opportunity. Consistent with this interpretation, we observe no Small effective population BiB Pakistani (UK born) significant effect on the likelihood of marriage (OR0.0625 = 0.94 ± Consanguinity 0.07, p = 0.12) (Fig. 3b). All effect size, odds ratios and 95% CI are stated as the difference between FROH = 0 and FROH = 0.0625. BiB Pakistani (baby) Hutterites The effects on fertility may be partly explained by the effect of Mean FROH 0.04 Amish FROH on a second group of traits, which capture risky or addictive Polynesia west BiB Pakistani behaviour. Increased FROH is associated with later age at first sex SAUDI (β0.0625 = 0.83 ± 0.19 years, p = 5.8 × 10−17) and fewer sexual partners (β0.0625 = −1.38 ± 0.38, p = 2.0 × 10−12) but also UKB Pakistani reduced alcohol consumption (β0.0625 = −0.66 ± 0.12 units per week, p = 1.3 × 10−22), decreased likelihood of smoking 0.02 TCGS BBJ (OR0.0625 = 0.79 ± 0.05, p = 5.9 × 10−13), and a lower probability Silk road UKB Others of being a self-declared risk-taker (OR0.0625 = 0.84 ± 0.06, p = MESA Hispanic 3.4×10−5) or exceeding the speed limit on a motorway (p = 4.0 × UKB British Population structure 10−8). Conservative beliefs are likely to affect these traits, and are 0.00 known to be confounded with FROH in some populations14, 0.00 0.02 0.04 0.06 however, fitting religious participation as a covariate in UKB FIS reduces, but does not eliminate the reported effects (Supplemen- tary Fig. 10b, Supplementary Data 20). Similarly, fitting Fig. 2 Mean FROH and FIS for 234 ROHgen sub-cohorts. Each cohort is educational attainment as an additional covariate reduces 16 of represented by a circle whose area is proportional to the approximate 25 significant effect estimates, but actually increases 9, including statistical power (Nσ 2FROH ) contributed to estimates of βFROH . Mean FROH can age at first sex and number of children (Supplementary Fig. 10a, be considered as an estimate of total inbreeding relative to an unknown Supplementary Data 20). This is because reduced educational base generation, approximately tens of generations past. FIS measures attainment is associated with earlier age at first sex and increased inbreeding in the current generation, with FIS = 0 indicating random mating, number of children, which makes it an unlikely confounder for FIS > 0 indicating consanguinity, and FIS < 0 inbreeding avoidance46. In the effects of FROH, which are in the opposite directions. cohorts along the y-axis, such as the Polynesians and the Anabaptist A third group of traits relates to cognitive ability. As previously isolates, autozygosity is primarily caused by small effective population size reported, increased autozygosity is associated with decreased rather than preferential consanguineous unions. In contrast, in cohorts general cognitive ability, g6,15 and reduced educational attain- along the dotted unity line, all excess SNP homozygosity is accounted for by ment6. Here, we also observe an increase in reaction time ROH, as expected of consanguinity within a large effective population. A (β0.0625 = 11.6 ± 3.9 ms, p = 6.5 × 10−9), a correlate of general small number of cohorts along the x-axis, such as Hispanic and mixed-race cognitive ability (Fig. 3a, Supplementary Data 10). groups, show excess SNP homozygosity without elevated mean FROH, A fourth group relates to body size. We replicate previously indicating population genetic structuring, caused for instance by admixture reported decreases in height and forced expiratory volume6 and known as the Wahlund effect. A few notable cohorts are labelled. BBJ (Supplementary Data 21) and we find that increased FROH Biobank Japan, BiB Born in Bradford, UKB UK Biobank, MESA Multiethnic is correlated with a reduction in weight (β0.0625 = 0.86 ± 0.12 kg, Study of Atherosclerosis, TCGS Tehran Cardiometabolic Genetic Study p = 3.4 × 10−28) and an increase in the waist to hip ratio (β0.0625 = 0.004 ± 0.001, p = 1.4 × 10−11). age, sex, principal components of ancestry and, in family studies, The remaining effects are loosely related to health and frailty; a genomic relationship matrix (GRM) (Supplementary Data 3). higher FROH individuals report significantly lower overall Cross-cohort effect size estimates were then obtained by fixed- health and slower walking pace, have reduced grip strength effect, inverse variance-weighted meta-analysis of the within- (β0.0625 = −1.24 ± 0.19 kg, p = 6.9 × 10−24), accelerated self- cohort estimates (Supplementary Data 10). Twenty-seven out of reported facial ageing, and poorer eyesight and hearing. Increased 79 quantitative traits and 5 out of 21 binary traits reach FROH is also associated with faster heart rate (β0.0625 = 0.56 ± 0.24 experiment-wise significance (0.05/100 or p < 0.0005; Fig. 3a, b). bpm, p = 5.9 × 10−6), lower haemoglobin (β0.0625 = 0.81 ± 0.24 gL Among these are replications of the previously reported effects on −1, p = 1.6 × 10−11), lymphocyte percentage, and total cholesterol reduction in height13, forced expiratory lung volume in one (β0.0625 = −0.05 ± 0.015 mmol L−1, p = 5.2 × 10−10). second, cognition and education attained6. We find that the 32 phenotypes affected by inbreeding can be grouped into five broader categories: reproductive success, risky behaviours, cog- Sex-specific effects of FROH. Intriguingly, for a minority of traits nitive ability, body size, and health. (13/100), the effect of FROH differs between men and women Despite the greater individual control over reproduction in the (Fig. 3c, Supplementary Data 12). For example, men who are the modern era, due to contraception and fertility treatments, we find offspring of first cousins have 0.10 mmol L−1 [95% CI 0.08–0.12] that increased FROH has significant negative effects on five traits lower total cholesterol on average, while there is no significant closely related to fertility. For example, an increase of 0.0625 in effect in women; LDL shows a similar pattern. More generally, for FROH (equivalent to the difference between the offspring of first these traits, the effect in men is often of greater magnitude than cousins and those of unrelated parents) is associated with having the effect in women, perhaps reflecting differing relationships 0.10 fewer children [β0.0625 = −0.10 ± 0.03 95% confidence between phenotype and fitness. interval (CI), p = 1.8 × 10−10]. This effect is due to increased FROH being associated with reduced odds of having any children Associations most likely caused by rare, recessive variants. The (OR0.0625 = 0.65 ± 0.04, p = 1.7 × 10−32) as opposed to fewer use of ROH to estimate inbreeding coefficients is relatively new in children among parents (β0.0625 = 0.007 ± 0.03, p = 0.66). Since inbreeding research11,16–19. Earlier frequency-based estimators NATURE COMMUNICATIONS | (2019)10:4957 | https://doi.org/10.1038/s41467-019-12283-6 | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12283-6 a p-value b Height 7e−149 p-value Weight 3e−28 Ever had children 2e−32 Forced expiratory volume 8e−22 Ever married 0.1 Grip strength 7e−24 Ever had children (married) 4e−08 Cognitive g 3e−17 Education attained 7e−27 Ever had children (unmarried) 2e−06 Reaction time 6e−09 Self-reported infertility 6e−04 Number of children 2e−10 Self-reported risk taker 3e−05 Age at first birth (men) 2e−18 Ever smoked 6e−13 Age at first sex 6e−17 Number of sexual partners 2e−12 −0.5 0 0.5 1 1.5 Driving speeding 4e−08 Log odds−ratio for FROH = 0.0625 Alcohol units per week 1e−22 c Self-reported overall health 3e−11 1 Weight Walking pace 7e−18 4 2 Body mass index 13 3 Age at first birth Frequency of vigorous activity 8e−05 4 Haemoglobin Effect in WOMEN (trait sd /FROH) 5 Total cholesterol Heart rate 6e−06 6 Triglycerides 2 7 Alanine transaminase Waist-Hip ratio 1e−11 8 Facial ageing 1e−16 7 2 10 3 Visual acuity 1e−06 4 5 6 0 11 12 1 Hearing acuity 1e−09 9 Haemoglobin 2e−11 8 Gamma-glutamyl transferase Lymphocytes (%) 2e−11 −2 9 Grip strength 10 Systolic blood pressure Total cholesterol 5e−10 11 LDL cholesterol 5e−04 12 White blood cell count LDL cholesterol −4 13 Age at first sex −4 −3 −2 −1 0 1 2 3 4 5 −4 −2 0 2 4 Effect size (trait sd per FROH) Effect in MEN (trait sd /FROH) Fig. 3 Scope of inbreeding depression. a Effect of FROH on 25 quantitative traits. To facilitate comparison between traits, effect estimates are presented in units of within-sex standard deviations. Traits shown here reached Bonferroni-corrected significance of p = 0.0005 (=0.05/100 traits). Sample sizes, within-sex standard deviations, and effect estimates in measurement units are shown in Supplementary Data 9. FEV1 forced expiratory volume in one second. Traits are grouped by type. b Effect of FROH on eight binary traits with associated p values. Effect estimates are reported as ln(Odds-Ratio) for the offspring of first cousins, for which E(FROH) = 0.0625. Self-declared infertility is shown for information, although this trait does not reach Bonferroni corrected significant (OR0:0625 = 2.6 ± 1.1, p = 0.0006). Numbers of cases and controls and effect estimates for all binary traits are shown in Supplementary Data 10. c Sex-specificity of ROH effects. The effect of FROH in men versus that in women is shown for 13 traits for which there was evidence of significant differences in the effects between sexes. For 11 of these 13 traits the magnitude of effect is greater in men than in women. Traits such as liver enzymes levels (alanine transaminase, gamma-glutamyl transferase) show sex-specific effects of opposite sign (positive in women, negative in men), which cancel out in the overall analysis. BMI body mass index, LDL low-density lipoprotein. All errors bars represent 95% confidence intervals such as FSNP and FGRM20, made use of excess marker homo- lengths have similar effects per unit length (Fig. 4a, Supplemen- zygosity21–23 and did not require physical maps. We performed tary Fig. 11a), consistent with their having a causal effect on traits both univariate and multivariate regressions to evaluate the and not with confounding by socioeconomic or other factors, as effectiveness of FROH against these measures. The correlations shorter ROH arise from deep in the pedigree are thus less between them range from 0.13 to 0.99 and are strongest in correlated with recent consanguinity. cohorts with high average inbreeding (Supplementary Data 6, Supplementary Fig. 6). Significantly, univariate regressions of traits on both FSNP and FGRM show attenuated effect estimates Quantifying the scope of social confounding. Previous studies relative to FROH (Supplementary Data 13). This attenuation is have highlighted the potential for FROH to be confounded by non- greatest in low autozygosity cohorts, suggesting that FROH is a genetic factors6,24. We therefore estimated the effect of FROH better estimator of excess homozygosity at the causal loci within various groups, between which confounding might be (Fig. 4c). expected either to differ, or not be present at all. To explore this further, we fit bivariate models with FROH and For example, the effect of FROH on height is consistent across FGRM as explanatory variables. For all 32 traits that were seven major continental ancestry groups (Supplementary Fig. 1, significant in the univariate analysis, we find that b βFROH jFGRM is Supplementary Data 18), despite differing attitudes towards consanguinity, and consequently different burdens and origins of of greater magnitude than β b in the conditional analysis FGRMjFROH ROH. Similarly, grouping cohorts into consanguineous, more (Fig. 4b, Supplementary Data 22). This suggests that inbreeding cosmopolitan, admixed and those with homozygosity due to depression is predominantly caused by rare, recessive variants ancient founder effects also shows consistent effects (Supplemen- made homozygous in ROH, and not by the chance homozygosity tary Fig. 2, Supplementary Data 19). Equally, categorising samples of variants in strong LD with common SNPs (Fig. 4d, into bins of increasing FROH shows a dose-dependent response of Supplementary Note 5). We also find that ROH of different the study traits with increased FROH (Supplementary Data 17 and 4 NATURE COMMUNICATIONS | (2019)10:4957 | https://doi.org/10.1038/s41467-019-12283-6 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12283-6 ARTICLE a b ROH < 5 Mb FGRM ROH > 5 Mb FHET outside ROH FROH –0.175 –0.15 –0.125 –0.1 –0.075 –0.05 –0.025 0 –0.175 –0.15 –0.125 –0.1 –0.075 –0.05 –0.025 0 Effect of F on height (m/F ) Effect of F on height (m/F ) c d 2.0 1.2 FGRM FROH 1.5 Slope = 1.01 1.0 1.0 0.8 FGRM/FROH Beta 0.5 0.6 0.0 0.4 –0.5 0.2 –1.0 0.0 –0.2 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.1 0.2 0.3 0.4 0.5 var(FROH)/var(FGRM) MAF Fig. 4 Inbreeding depression caused by ROH. a Effect of different ROH lengths on height, compared with the effect of SNP homozygosity outside of ROH. The effects of shorter (<5 Mb) and longer (>5 Mb) ROH per unit length are similar and strongly negative, whereas the effect of homozygosity outside ROH is much weaker. The pattern is similar for other traits (Supplementary Fig. 11a; Supplementary Data 14). b F ROH is more strongly associated than F GRM in a bivariate model of height. Meta-analysed effect estimates, and 95% confidence intervals, are shown for a bivariate model of height (Height  FROH þ FGRM ). The reduction in height is more strongly associated with FROH than FGRM , as predicted if the causal variants are in weak LD with the common SNPs used to calculate FGRM (Supplementary Note 5). The pattern is similar for other traits (Supplementary Fig. 15a, b; Supplementary varðFROH Þ Data 22). c FROH is a lower variance estimator of the inbreeding coefficient than FGRM. The ratio of βFGRM : βFROH is plotted against varðF Þ for all traits in all GRM ^ cohorts. When the variation of FGRM which is independent of FROH has no effect on traits, β varðFROH Þ FGRM is downwardly biased by a factor of varðF Þ (Supplementary GRM Note 4). A linear maximum likelihood fit, shown in red, has a gradient consistent with unity [1.01; 95% CI 0.84–1.18], as expected when the difference between FGRM and FROH is not informative about the excess homozygosity at causal variants (Supplementary Note 5). d FROH is a better predictor of rare variant homozygosity than FGRM. The excess homozygosities of SNPs, extracted from UK Biobank imputed genotypes, were calculated at seven discrete minor allele frequencies (FMAF ), and regressed on two estimators of inbreeding in a bivariate statistical model (see Supplementary Note 5). The homozygosity of common SNPs is better predicted by FGRM , but rare variant homozygosity is better predicted by FROH . The results from real data (Fig. 4b, Supplementary Figs 15a, b and Supplementary Data 22) are consistent with those simulated here, if the causal variants are predominantly rare. All errors bars represent 95% confidence intervals Fig. 5a, b show the response for height and ever having children; reach replication significance (p < 0.004). In addition, a meta- Supplementary Figs 9a–f for all significant traits). The propor- analysis of the ratio bβFROH ADOPTEE : b βFROH across all traits differs tionality of these effects is consistent with a genetic cause, while it significantly from zero (Fig. 5c; average = 0.78, 95% CI 0.56–1.00, is difficult to envisage a confounder proportionally associated p = 2 × 10−12). across the entire range of observed FROH. In particular, the Finally, the effect of FROH was estimated in up to 118,773 highest FROH group (FROH > 0.18), equivalent to the offspring of first-degree relatives, are found to be, on average, 3.4 [95% CI individuals in sibships (full-sibling pairs, trios, etc.: bβFROH wSibs ). 2.5–4.3] cm shorter and 3.1 [95% CI 2.5–3.7] times more likely to FROH differences between siblings are caused entirely by be childless than an FROH = 0 individual. Mendelian segregation, and are thus independent of any reason- Next, we estimated βF ROH for 7153 self-declared adopted able model of confounding. The variation of FROH among siblings individuals in UK Biobank, whose genotype is less likely to be is a small fraction of the population-wide variation11 (Supple- confounded by cultural factors associated with the relatedness of mentary Data 5); nevertheless, 23 out of 29 estimates of b βFROH wSibs their biological parents. For all 26 significant traits measured in b are directionally consistent with β , and two (self-reported FROH this cohort, effect estimates are directionally consistent with the meta-analysis and 3 (height, walking pace and hearing acuity) overall health and ever having children) reach replication NATURE COMMUNICATIONS | (2019)10:4957 | https://doi.org/10.1038/s41467-019-12283-6 | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12283-6 a b 0.00 0.0 Log odds-ratio for ever had children −0.5 −0.01 −1.0 Height (m) −0.02 −1.5 −0.03 −2.0 −2.5 −0.04 0.00 0.05 0.10 0.15 0.20 0.00 0.05 0.10 0.15 0.20 FROH FROH c d Ever had children Height Height Age at first birth (men) Grip strength Ever had children Education attained Alcohol units per week Ever smoked Age at first sex Weight Forced expiratory volume Haemoglobin Walking pace Forced expiratory volume Facial ageing Alcohol units per week Cognitive g Cognitive g Self-reported overall health Grip strength Reaction time Lymphocytes (%) Hearing acuity Number of children Total cholesterol Education attained Walking pace Ever smoked Age at first sex Number of sexual partners Waist-Hip ratio Driving speeding Facial ageing Lymphocytes (%) Number of sexual partners Haemoglobin Self-reported overall health Visual acuity Reaction time Waist-Hip ratio Heart rate Number of children LDL cholesterol Driving speeding Weight Hearing acuity Self-reported risk taker Self-reported risk taker Frequency of vigorous activity Visual acuity Heart rate Frequency of vigorous activity Average Average −5 −4 −3 −2 −1 0 1 2 3 4 5 −5 −4 −3 −2 −1 0 1 2 3 4 5 Ratio Ratio Fig. 5 Evidence ROH effects are un-confounded. a Linear decrease in height with increasing FROH. Average heights (in metres) is plotted in bins of increasing FROH. The limits of each bin are shown by red dotted lines, and correspond to the offspring of increasing degree unions left-to-right. The overall estimate of βFROH is shown as a solid black line. Subjects with kinship equal to offspring of full-sibling or parent–child unions are significantly shorter than those of avuncular or half-sibling unions who in turn are significantly shorter than those of first-cousin unions. b Linear decrease in odds of ever having children with increasing FROH. Linear model approximations of ln(Odds-Ratio) for ever having children (1 = parous, 0 = childless) are plotted in bins of increasing FROH. A strong relationship is evident, extending beyond the offspring of first cousins. c ROH effects are consistent in adoptees. The ratios of effect estimates, βFROH , between adoptees and all individuals are presented by trait. All traits are directionally consistent and overall show a strongly significant difference from zero (average = 0.78, 95% CI 0.56–1.00, p = 2 × 10−12). FEV1 forced expiratory volume in one second. d ROH effects are consistent in full siblings. The ratios of effect estimates within full siblings to effects in all individuals (βFROH wSibs : βFROH ) are presented by trait. Twenty-three of 29 estimates are directionally consistent and overall show a significant difference from zero (average = 0.78, 95% CI 0.53–1.04, p = 7 × 10−10). BMI body mass index. All errors bars represent 95% confidence intervals 6 NATURE COMMUNICATIONS | (2019)10:4957 | https://doi.org/10.1038/s41467-019-12283-6 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12283-6 ARTICLE significance. A meta-analysis of the ratio b βFROH wSibs : b βFROH for all The alternative hypothesis of overdominance, whereby positive traits is significantly greater than zero (Fig. 5d; average = 0.78, selection on heterozygotes has brought alleles to intermediate 95% CI 0.53–1.04, p = 7 × 10−10), indicating a substantial fraction frequencies, would predict that more common homozygous SNPs of these effects is genetic in origin. However, for both adoptees outside long ROH would also confer an effect. The differential and siblings, the point estimates are less than one, suggesting that provides evidence in humans that rare recessive mutations non-genetic factors probably contribute a small, but significant, underlie the quantitative effects of inbreeding depression. fraction of the observed effects. Previous studies have shown that associations observed between FROH and traits do not prove a causal relationship14,24. Traditional Genome-wide Association Studies (GWAS) can infer Discussion causality because, in the absence of population structure, genetic Our results reveal inbreeding depression to be broad in scope, variants (SNPs) are randomly distributed between, and within, influencing both complex traits related to evolutionary fitness and different social groups. However, this assumption does not hold others where the pattern of selection is less clear. While studies of in studies of inbreeding depression, where, even within a couples show optimal fertility for those with distant kinship25,26, genetically homogeneous population, social groups may have fewer have examined reproductive success as a function of indi- differing attitudes towards consanguinity, and therefore different vidual inbreeding. Those that did are orders of magnitude smaller average FROH and, potentially, different average trait values. We in size than the present study, suffer the attendant drawbacks of therefore present a number of analyses that discount social pedigree analysis, and have found mixed results27–29. Our geno- confounding as a major factor in our results. Firstly, we show that mic approach also reveals that in addition to socio-demographic the effects are consistent across diverse populations, including factors and individual choice, recessive genetic effects have a those where ROH burden is driven by founder effects rather than significant influence on whether individuals reproduce. The dis- cultural practices regarding marriage. Effects are also consistent cordant effects on fertility and education demonstrate that this is across a 20-fold range of FROH: from low levels, likely unknown to not just a result of genetic correlations between the two the subject, to extremely high levels only seen in the offspring of domains30. first-degree relatives. Secondly, we show that the effects of ROH The effects we see on fertility might be partially mediated are consistent in direction and magnitude among adopted indi- through a hitherto unknown effect of autozygosity on decreasing viduals, and also for short ROH which are not informative about the prevalence of risk-taking behaviours. Significant effects of parental relatedness. Finally, we introduce a within-siblings autozygosity are observed for self-reported risk taking, speeding method, independent of all confounders, that confirms a on motorways, alcohol and smoking behaviour, age at first sexual genetic explanation for most of the observed effects. Variation in intercourse and number of sexual partners. Independent evidence FROH between siblings is caused entirely by random Mendelian for a shared genetic architecture between risk-taking and fertility segregation; we show that higher FROH siblings experience poorer traits comes from analysis of genetic correlations using LD-score overall health and lower reproductive success, as well as other regression in UKB (Supplementary Table 1). The core fertility changes consistent with population-wide estimates. Nevertheless, traits, ever had children and number of children, are strongly average effect sizes from both adoptees and siblings are 20% genetically correlated (rG = 0.93; p < 10−100). Genetic correlations smaller than population-wide estimates, confirming the impor- with ever-smoking and self-reported risk-taking are lower, but tance of accounting for social confounding in future studies of also significant: 0.23–0.27, p < 10−10. Age at first sex is strongly human inbreeding depression. genetically correlated both with the fertility traits, (rG = Our results reveal five large groups of phenotypes sensitive to 0.53–0.57), and number of sexual partners, ever-smoking and inbreeding depression, including some known to be closely linked risk-taking30 (rG = 0.42–0.60). to evolutionary fitness, but also others where the connection is, Reproductive traits are understandable targets of natural with current knowledge, more surprising. The effects are medi- selection, as might be walking speed, grip strength, overall health, ated by ROH rather than homozygosity of common SNPs, cau- and visual and auditory acuity. While we cannot completely sally implicating rare recessive variants rather than exclude reverse causality, whereby a less risk-taking, more con- overdominance as the most important underlying mechanism. servative, personality is associated with greater likelihood of Identification of these recessive variants will be challenging, but consanguineous marriage, we note that the effects are consistent analysis of regional ROH and in particular using whole-genome for ROH < 5 Mb, which are less confounded with mate choice, sequences in large cohorts with sufficient variance in autozygosity due to their more distant pedigree origins (Supplementary will be the first step. Founder populations or those which prefer Fig. 11a). This group of traits also shows similar evidence for un- consanguineous marriage will provide the most power to confounded effects in the analysis of adoptees and full siblings understand this fundamental phenomenon. (Fig. 5c, d; Supplementary Data 16) and the signals remained see Supplementary Data. after correcting for religious activity or education. On the other hand, for some traits that we expected to be influenced by ROH, we observed no effect. For example, birth Methods weight is considered a key component of evolutionary fitness in Overview. Our initial aim was to estimate the effect of FROH on 45 quantitative traits and to assess whether any of these effects differed significantly from zero. mammals, and is influenced by genomic homozygosity in deer31; Previous work7,11 has shown that inbreeding coefficients are low in most human however, no material effect is apparent here (Supplementary populations, and that very large samples are required to reliably estimate the Data 10). Furthermore, in one case, ROH appear to provide a genetic effects of inbreeding13. To maximise sample size, a collaborative con- beneficial effect: increasing FROH significantly decreases total and sortium (ROHgen6) was established, and research groups administering cohorts with SNP chip genotyping were invited to participate. To ensure that all partici- LDL-cholesterol in men, and may thus be cardio-protective in pants performed uniform and repeatable analyses, a semi-automated software this regard. pipeline was developed and executed locally by each research group. This software Our multivariate models show that homozygosity at common pipeline required cohorts to provide only quality-controlled genotypes (in plink SNPs outside of ROH has little influence on traits, and that the binary format) and standardised phenotypes (in plain-text) and used standard software (R, PLINK12,32, KING33) to perform the analyses described below. Results effect rather comes from ROH over 1.5 Mb in length. This sug- from each cohort were returned to the central ROHgen analysts for meta-analysis. gests that genetic variants causing inbreeding depression are During the initial meta-analysis, genotypes were released for >500,000 samples almost entirely rare, consistent with the dominance hypothesis1. from the richly phenotyped UK Biobank (UKB)10. It was therefore decided to add a NATURE COMMUNICATIONS | (2019)10:4957 | https://doi.org/10.1038/s41467-019-12283-6 | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12283-6 further 34 quantitative phenotypes and 21 binary traits to the ROHgen analysis. Calculating estimators of F. For each sample, two estimates of the inbreeding Many of these additional traits were unique to UKB, although 7 were also available coefficient (F) were calculated, FROH and FSNP. We also calculated three additional in a subset of ROHgen cohorts willing to run additional analyses. In total, the effect measures of homozygosity: FROH<5Mb, FROH>5Mb and FSNP_outsideROH. of FROH was tested on 100 traits and therefore experiment-wise significance was FROH is the fraction of each genome in ROH >1.5 Mb. For example, in a sample defined as 5 × 10−4 (=0.05/100). for which PLINK had identified n ROH of length li (in Mb), i ϵ {1..n}, then FROH was then calculated as Pn i¼1 i ; ð1Þ l Cohort recruitment. In total, 119 independent, genetic epidemiological study FROH ¼ 3Gb cohorts were contributed to ROHgen. Of these, 118 were studies of adults and contributed multiple phenotypes, while 1 was a study of children and contributed where FROH<5Mb and FROH>5Mb are the genomic fractions in ROH of length >5 Mb, only birth weight. To minimise any potential confounding or bias caused by and in ROH of length <5 Mb (but >1.5 Mb), respectively, and the length of the within-study heterogeneity, studies were split into single-ethnicity sub-cohorts autosomal genome is estimated at 3 gigabases (Gb). It follows from this definition wherever applicable. Each sub-cohort was required to use only one genotyping that array and be of uniform ancestry and case-status. For example, if a study contained FROH ¼ FROH>5Mb þ FROH<5Mb : ð2Þ multiple distinct ethnicities, sub-cohorts of each ancestry were created and ana- lysed separately. At minimum, ancestry was defined on a sub-continental scale (i.e. Single-point inbreeding coefficients can also be estimated from individual SNP European, African, East Asian, South Asian, West Asian, Japanese, and Hispanic homozygosity without any reference to a genetic map. For comparison with FROH, were always analysed separately) but more precise separation was used when a method of moments estimate of inbreeding coefficient was calculated34, referred deemed necessary, for example, in cohorts with large representation of Ashkenazi to here as FSNP, and implemented in PLINK by the command–het. Jews. In case-control studies of disease, separate sub-cohorts were created for cases and controls and phenotypes associated with disease status were not analysed in FSNP ¼ OðHOM ÞEðHOMÞ NEðHOMÞ ; ð3Þ the case cohort: for example, fasting plasma glucose was not analysed in Type 2 where O(HOM) is the observed number of homozygous SNPs, E(HOM) is the diabetes case cohorts. Occasionally, cohorts had been genotyped on different SNP P expected number of homozygous SNPs, i.e. Ni¼1 ð1  2pi qi Þ, and N is the total genotyping microarrays and these were also separated into sub-cohorts. There was number of non-missing genotyped SNPs. one exception (deCODE) to the single microarray rule, where the intersection FROH and FSNP are strongly correlated, especially in cohorts with significant between all arrays used exceeded 150,000 SNPs. In this cohort the genotype data inbreeding, since both are estimates of F. To clarify the conditional effects of FROH from all arrays was merged since the correspondence between FROH for the indi- and FSNP, an additional measure of homozygosity,FSNPoutsideROH, was calculated to vidual arrays and FROH the intersection dataset was found to be very strong describe the SNP homozygosity observed outside ROH. (βmerged;hap = 0.98, r2 = 0.98; βmerged;omni = 0.97, r2=0.97). Dividing studies using these criteria yielded 234 sub-cohorts. Details of phenotypes contributed by each FSNPoutsideROH ¼ O′ðHOM ÞE′ðHOMÞ N′E′ðHOMÞ ; ð4Þ cohort are available in Supplementary Data 4. where Ethical approval. Data from 119 independent genetic epidemiology studies were O′ðHOMÞ ¼ OðHOMÞ  NSNP ROH ; ð5Þ included. All subjects gave written informed consent for broad-ranging health and   NNROH genetic research and all studies were approved by the relevant research ethics E′ðHOMÞ ¼ N  EðHOMÞ ; ð6Þ committees or boards. PubMed references are given for each study in Supple- mentary Data 2. N′ ¼ N  NROH ð7Þ And NSNP_ROH is the number of homozygous SNPs found in ROH. Note that: Genotyping. All samples were genotyped on high-density (minimum 250,000 markers), genome-wide SNP microarrays supplied by Illumina or Affymetrix. FSNPoutsideROH  FSNP  FROH ð8Þ Genotyping arrays with highly variable genomic coverage (such as Exome chip, A further single point estimator of the inbreeding coefficient, described by Yang Metabochip, or Immunochip) were judged unsuitable for the ROH calling algo- bIII , is implemented in PLINK by the parameter –ibc (Fhat3) and was et al.20 as F rithm and were not permitted. Imputed genotypes were also not permitted; only also calculated for all samples. called genotypes in PLINK binary format were accepted. Each study applied their   own GWAS quality controls before additional checks were made in the common XN xi2  ð1 þ 2pi Þxi þ 2p2i analysis pipeline: SNPs with >3% missingness or MAF <5% were removed, as were bIII ¼ 1 FGRM ¼ F ; ð9Þ N i¼1 2pi ð1  pi Þ individuals with >3% missing data. Only autosomal genotypes were used for the analyses reported here. Additional, cohort-specific, genotyping information is where N is the number of SNPs, pi is the reference allele frequency of the ith SNP in available in Supplementary Data 2. the sample population and xi is the number of copies of the reference allele. Effect size estimates for quantitative traits. In each cohort of n samples, for Phenotyping. In total, results are reported for 79 quantitative traits and 21 binary each of the quantitative traits measured in that cohort, trait values were modelled traits. These traits were chosen to represent different domains of health and by reproductive success, with consideration given to presumed data availability. Many of these traits have been the subject of existing genome-wide association meta- y ¼ βFROH  FROH þ Xb þ ε ; ð10Þ analyses (GWAMA), and phenotype modelling, such as inclusion of relevant covariates, was copied from the relevant consortia (GIANT for anthropometry, where y is a vector (n × 1) of measured trait values, βF ROH is the unknown scalar EGG for birth weight, ICBP for blood pressures, MAGIC for glycaemic traits, effect of FROH on the trait, FROH is a known vector (n × 1) of individual FROH, b is a CHARGE-Cognitive, -Inflammation & -Haemostasis working groups for cognitive vector (m × 1) of unknown fixed covariate effects (including a mean, μ), X in a function, CRP, fibrinogen, CHARGE-CKDgen for eGFR, CHARGE-ReproGen for known design matrix (n × m) for the fixed effects, and ε is an unknown vector (n × ages at menarche and menopause, Blood Cell & HaemGen for haematology, 1) of residuals. GUGC for urate, RRgen, PRIMA, QRS & QT-IGC for electrocardiography, GLGC The m fixed covariates included in each model were chosen with reference to for classical lipids, CREAM for spherical equivalent refraction, Spirometa for lung the leading GWAMA consortium for that trait and are detailed in Supplementary function traits, and SSGAC for educational attainment and number of children Data 8. For all traits, these covariates included: age (and/or year of birth), sex, and ever born). Further information about individual phenotype modelling is available at least the first 10 principal components of the genomic relatedness matrix in Supplementary Note 1 and Supplementary Data 8. (GRM). Where necessary, additional adjustments were made for study site, medications, and other relevant covariates (Supplementary Data 3). For reasons of computational efficiency, it was decided to solve Eq. (10) in two ROH calling. Runs of homozygosity (ROH) of >1.5 Mb in length were identified steps. In the first step, the trait (y) was regressed on all fixed covariates to obtain the using published methods6,11. In summary, SNPs with minor allele frequencies maximum likelihood solution of the model: below 5% were removed, before continuous ROH SNPs were identified using PLINK with the following parameters: homozyg-window-snp 50; homozyg-snp 50; y ¼ Xb þ ε′ : ð11Þ homozyg-kb 1500; homozyg-gap 1000; homozyg-density 50; homozyg-window- All subsequent analyses were performed using the vector of trait residuals ε′, missing 5; homozyg-window-het 1. No linkage disequilibrium pruning was per- which may be considered as the trait values corrected for all known covariates. formed. These parameters have been previously shown to call ROH that corre- In cohorts with a high degree of relatedness, mixed-modelling was used to spond to autozygous segments in which all SNPs (including those not present on correct for family structure, although, because ROH are not narrow-sense heritable, the chip) are homozygous-by-descent, not chance arrangements of independent this was considered less essential than in Genome-Wide Association Studies. homozygous SNPs, and inbreeding coefficient estimates calculated by this method Equation (11) becomes (FROH) correlate well with pedigree-based estimates (FPED)11. Moreover, they have also been shown to be robust to array choice6. y ¼ Xb þ u þ ε′; ð12Þ 8 NATURE COMMUNICATIONS | (2019)10:4957 | https://doi.org/10.1038/s41467-019-12283-6 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-12283-6 ARTICLE where u is an unknown vector (n × 1) of polygenic effects with multivariate normal height deficit has been inherited from genetically shorter parents, but this would distribution of mean 0 and covariance matrix σg2A, where A is the genomic require an implausibly strong relationship between short stature and a propensity relationship matrix (GRM). In these related cohorts, a GRM was calculated using to marry a very close relative. Thirdly, the sex-specific effects we observe could only PLINK v1.9 and Grammar+ residuals of Eq. (12) were estimated using be explained by assortative mating if the additive heritability of these traits also GenABEL35. These Grammar+ residuals (ε′) were used in subsequent analyses. differed by gender. To estimate βF ROH for each trait, trait residuals were regressed on FROH to obtain the maximum likelihood (ML) solution of the model Average trait values in groups of similar FROH. In each cohort individuals were ε′ ¼ μ þ βFROH  FROH þ ε: ð13aÞ allocated to one of ten groups of similar FROH. The bounds of these groups were the same for all cohorts, specifically {0, 0.002, 0.0041, 0.0067, 0.0108, 0.0186, 0.0333, The sex-specific estimates of βF ROH (Supplementary Data 12) were obtained 0.06, 0.10, 0.18 and 1.0}. Within each group the mean trait residual (ε′) and mean from Eq. (13) applied to the relevant sex. FROH were calculated, along with their associated standard errors. Within each For all traits, a corresponding estimates of βF SNP and βFGRM were obtained from cohort the expectation of ε′ is zero at the mean FROH, however as mean FROH varies the models between cohorts (Fig. 2, Supplementary Data 5) it was necessary to express ε′ relative to a common FROH before meta-analysis. Hence, for this analysis only, the ε′ ¼ μ þ βFSNP  FSNP þ ε; ð13bÞ trait residuals (ε′) were expressed relative to the FROH = 0 intercept, i.e. by sub- tracting μ from Eq. (13). ε′ ¼ μ þ βFGRM  FGRM þ ε ð14Þ and the effects of different ROH lengths and of SNP homozygosity (Fig. 4b) were Effect of FROH within adoptees. We compared βF ROH ADOPTEE to cross-cohort obtained from the model βF ROH , not that from UKB alone, as we consider the latter to be a noisy estimate     of the former; estimates in UKB are consistent with those from meta-analysis. ε′ ¼ μ þ β1  FSNPoutsideROH þ β2  FROH<5Mb   : ð15Þ þ β3  FROH>5Mb þ ε Effect of FROH within full-sibling families. In a subset of cohorts, with substantial numbers of related individuals, further analyses were performed to investigate the effect of FROH within full-sibling families. In each of these cohorts, all second- Effect size estimates for binary traits. Binary traits were analysed by two degree, or closer, relatives were identified using KING (parameters:–related–degree methods. The primary estimates of βF ROH (Fig. 3b and Supplementary Data 10) 2). Full-siblings were then selected as relative pairs with genomic kinship >0.175 were obtained from full logistic models: and IBS0 >0.001. This definition includes monozygotic twins, who were inten- tionally considered as part of full-sibling families. Although monozygotic twins are g ðE½yÞ ¼ Xb ; ð16Þ expected to have identical FROH, they may not have identical trait values, and where g() is the link function (logit), and where FROH and all applicable covariates including additional trait measurements decreases the sampling error of the within- (Supplementary Datas 3, 8) were fitted simultaneously. Mixed modelling for family family variance estimate, hence increasing statistical power. Dizygotic twins were structure was not attempted in the logistic models since an accepted method was also included. not apparent. For each individual (j) with identified siblings, the values of FROH and trait For all subsequent results, y was scaled by 1=σ 2y and analysed by linear models, residual (ε′) were calculated relative to their family mean (and called FjROH_wSibs and εjwSibs, respectively), i.e. for individual j with n full-siblings Sk where k ϵ {1..n} as for quantitative traits, including mixed-modelling where appropriate for family P studies. This method of estimating binary traits with simple linear models gives FjROHwSibs ¼ FjROH  ðnþ1 1 Þ ROH iϵfj;Sk g Fi ; ð17Þ asymptotically unbiased estimates of βF ROH and se(βF ROH ) on the ln(Odds-Ratio) scale36. For all significant binary traits, a comparison of bβ from the full model P FROH εwSibs j ¼ ε′j  ðnþ1 1 Þ iϵfj;Sk g ε′i : ð18Þ with bβFROH from the linear model approximation is presented in Supplementary The effect of FROH within-full-siblings (βF ROH wSibes ) was estimated by linear Fig. 8. regression of εwSibs on FROH_wSibs. To give b βFROH a more tangible interpretation, effect estimates are frequently Importantly, the variation of FROH within full-siblings is entirely caused by quoted in the text as β0.0625, i.e. the estimated effect in the offspring of first cousins, differences in Mendelian segregation, and is therefore completely independent of where 6.25% of the genome is expected to be autozygous. all possible confounders. Hence, the effect estimates obtained by this method are estimates of the genetic effects of FROH, unbiased by any possible confounder. Since Religiosity and educational attainment as additional covariates. To assess the confounding by social factors is a major concern in this field, methods that can importance of potential social confounders, proxy measures of socio-economic definitively exclude this possibility are of critical importance. status and religiosity were separately included in Eq. (13) as additional covariates. The modified effect estimates (β′b ) were tested for significance (Supplementary Between-cohort meta-analysis. As is typical in genome-wide association meta- FROH Data 20) and compared to the uncorrected estimates (βF ROH ) (Supplementary analyses (GWAMA), genetic effects were estimated within single-ethnicity sub- Fig. 10a, b). cohorts, and meta-analysis of the within-cohort effect sizes was used to combine Since Educational Attainment (EA) was measured in many cohorts, this was results38. This established method eliminates any potential confounding caused by chosen as the most suitable proxy for socio-economic status. However, since FROH between-cohort associations between FROH and traits. is known to affect EA directly6 any change in βF ROH when conditioning on EA Each cohort returned estimates and standard errors of: βF ROH , cannot be assumed to be entirely due to environmental confounding. βF SNP ; βF ROH>Mb ; βF ROH