Heritability and genetic correlations explained by common SNPs for MetS traits. Shashaank Vattikuti, Juen Guo and Carson Chow LBM/NIDDK

Transcription

1 Heritability and genetic correlations explained by common SNPs for MetS traits Shashaank Vattikuti, Juen Guo and Carson Chow LBM/NIDDK

2 The Genomewide Association Study. Manolio TA. N Engl J Med 2010;363:

3 Nature Genetics 42, (2010) Association analyses of 249,796 individuals reveal 18 new loci associated with body mass index Obesity is globally prevalent and highly heritable, but its underlying genetic factors remain largely elusive. To identify genetic loci for obesity susceptibility, we examined associations between body mass index and ~2.8 million SNPs in up to 123,865 individuals with targeted follow up of 42 SNPs in up to 125,931 additional individuals. We confirmed 14 known obesity susceptibility loci and identified 18 new loci associated with body mass index (P < ), one of which includes a copy number variant near GPRC5B. Some loci (at MC4R, POMC, SH2B1 and BDNF) map near key hypothalamic regulators of energy balance, and one of these loci is near GIPR, an incretin receptor. Furthermore, genes in other newly associated loci may provide new insights into human body weight regulation. Obesity is a major and increasingly prevalent risk factor for multiple disorders, including type 2 diabetes and cardiovascular disease 1,2. Although lifestyle changes have driven its prevalence to epidemic proportions, heritability studies provide evidence for a substantial genetic contribution (with heritability estimates (h 2 ) of ~40% 70%) to obesity risk 3,4. BMI is an inexpensive, non-invasive measure of obesity that predicts the risk of related complications 5. Identifying genetic determinants of BMI could lead to a better understanding of the biological basis of obesity. 19 loci associated with BMI at P < (Table 1, Fig. 1a and Supplementary Table 1). These 19 loci included all ten loci from previous GWAS of BMI 6 10, two loci previously associated with body weight 10 (at FAIM2 and SEC16B) and one locus previously associated with waist circumference 14 (near TFAP2B). The remaining six loci, near GPRC5B, MAP2K5-LBXCOR1, TNNI3K, LRRN6C, FLJ HMGCR and PRKD1, have not previously been associated with BMI or other obesity-related traits. heritability 40% - 70% explained by GWAS 1.45%

4 Missing Heritability and GWAS Utility Clifton Bogardus* doi: /oby The environment is largely responsible for differences in BMI between populations; genetics is largely responsible for differences in BMI within populations. Evidences in support of the environmental effect are many. The prevalence of obesity differs greatly between countries, even among developed countries (1). On a smaller scale, differences in the built environment within a city are associated with differences in BMI. In New York City a greater density of bus and subway stops and greater variety in uses of land within a neighborhood are associated with lower BMIs (2). Closely genetically related populations living in different environments differ considerthe melanocortin 4 receptor (MC4R). About two and a half percent of severely obese individuals have causative variants in this gene (8). These variants are associated with obesity largely due to increased food intake but also due to reduced energy expenditure (9). GWASs are the most commonly used agnostic approach to identify susceptibility genes for common disease, including obesity. These have been done in populations of European descent and used ~350,000 SNPs, covering more than 75% of the genome (10,11). The two major findings were common noncoding SNPs in an intron of the FTO gene and a common SNP within a pre- OBESITY VOLUME 17 NUMBER 2 FEBRUARY 2009

5 New York Times

6 Epstasis Rare variants Many variants of small effect

7 BMI SNPs Nature Genetics 42, (2010)

8 Heritability vs. Penetrance

9 Mendel Galton Discrete trait, e.g. obesity penetrance Quantitative trait, e.g. BMI heritability

10 Penetrance is the proportion of individuals carrying an allele that express the phenotype

11 Heritability is the variance of phenotype explained by genotype

12 Heritability is the variance of phenotype explained by genotype Phenotype P = G + E + I

13 Heritability is the variance of phenotype explained by genotype P = G + E + I

14 Heritability is the variance of phenotype explained by genotype Genotype (additive) P = G + E + I

15 Heritability is the variance of phenotype explained by genotype P = G + E + I

16 Heritability is the variance of phenotype explained by genotype Environment P = G + E + I

17 Heritability is the variance of phenotype explained by genotype P = G + E + I

18 Heritability is the variance of phenotype explained by genotype Interactions (nonadditive) P = G + E + I

19 Heritability is the variance of phenotype explained by genotype Interactions (nonadditive) P = G + E + I h 2 = var(g) var(p ) narrow sense heritability

20 Regress offspring against parent to estimate heritability

21 Regress offspring against parent to estimate heritability h 2 is slope/relationship

22 Regress offspring against parent to estimate heritability h 2 is slope/relationship rel. = cor(gp,go)

23 Heritability in SNPs Phenotype QQ qq qq q = minor Q=major Biallelic SNP Genotype

24 Heritability in SNPs Phenotype additive effect a QQ qq qq q = minor Q=major Biallelic SNP Genotype

25 Heritability in SNPs Phenotype additive effect a h 2 = 2p(1-p) a 2 p = frequency of allele QQ qq qq q = minor Q=major Biallelic SNP Genotype

26 Heritability in SNPs Phenotype nonadditive additive effect a h 2 = 2p(1-p) a 2 p = frequency of allele QQ qq qq q = minor Q=major Biallelic SNP Genotype

27 Common SNPs explain a large proportion of the heritability for human height Jian Yang 1, Beben Benyamin 1, Brian P McEvoy 1, Scott Gordon 1, Anjali K Henders 1, Dale R Nyholt 1, Pamela A Madden 2, Andrew C Heath 2, Nicholas G Martin 1, Grant W Montgomery 1, Michael E Goddard 3 & Peter M Visscher 1 ture America, Inc. All rights reserved. SNPs discovered by genome-wide association studies (GWASs) account for only a small fraction of the genetic variation of complex traits in human populations. Where is the remaining heritability? We estimated the proportion of variance for human height explained by 294,831 SNPs genotyped on 3,925 unrelated individuals using a linear model analysis, and validated the estimation method with simulations based on the observed genotype data. We show that 45% of variance can be explained by considering all SNPs simultaneously. Thus, most of the heritability is not missing but has not previously been detected because the individual effects are too small to pass stringent significance tests. We provide evidence that the remaining heritability is due to incomplete linkage disequilibrium between causal variants and genotyped SNPs, exacerbated by causal variants having lower minor allele of variation that their effects do not reach stringent significance thresholds and/or the causal variants are not in complete linkage disequilibrium (LD) with the SNPs that have been genotyped. Lack of complete LD might, for instance, occur if causal variants have lower minor allele frequency (MAF) than genotyped SNPs. Here we test these two hypotheses and estimate the contribution of each to the heritability of height in humans as a model complex trait. Height in humans is a classical quantitative trait, easy to measure and studied for well over a century as a model for investigating the genetic basis of complex traits 9,10. The heritability of height has been estimated to be ~0.8 (refs. 9,11 13). Rare mutations that cause extreme short or tall stature have been found 14,15, but these do not explain much of the variation in the general population. Recent GWASs on tens of thousands of individuals have detected ~50 variants that are associated with height in the population, but these in total account 45% Nature Genetics 42, (2010)

28 Use GWAS SNPs to estimate relationship Common SNPs explain a large proportion of the heritability for human height Jian Yang 1, Beben Benyamin 1, Brian P McEvoy 1, Scott Gordon 1, Anjali K Henders 1, Dale R Nyholt 1, Pamela A Madden 2, Andrew C Heath 2, Nicholas G Martin 1, Grant W Montgomery 1, Michael E Goddard 3 & Peter M Visscher 1 ture America, Inc. All rights reserved. SNPs discovered by genome-wide association studies (GWASs) account for only a small fraction of the genetic variation of complex traits in human populations. Where is the remaining heritability? We estimated the proportion of variance for human height explained by 294,831 SNPs genotyped on 3,925 unrelated individuals using a linear model analysis, and validated the estimation method with simulations based on the observed genotype data. We show that 45% of variance can be explained by considering all SNPs simultaneously. Thus, most of the heritability is not missing but has not previously been detected because the individual effects are too small to pass stringent significance tests. We provide evidence that the remaining heritability is due to incomplete linkage disequilibrium between causal variants and genotyped SNPs, exacerbated by causal variants having lower minor allele of variation that their effects do not reach stringent significance thresholds and/or the causal variants are not in complete linkage disequilibrium (LD) with the SNPs that have been genotyped. Lack of complete LD might, for instance, occur if causal variants have lower minor allele frequency (MAF) than genotyped SNPs. Here we test these two hypotheses and estimate the contribution of each to the heritability of height in humans as a model complex trait. Height in humans is a classical quantitative trait, easy to measure and studied for well over a century as a model for investigating the genetic basis of complex traits 9,10. The heritability of height has been estimated to be ~0.8 (refs. 9,11 13). Rare mutations that cause extreme short or tall stature have been found 14,15, but these do not explain much of the variation in the general population. Recent GWASs on tens of thousands of individuals have detected ~50 variants that are associated with height in the population, but these in total account 45% Nature Genetics 42, (2010)

29 GWAS SNP genotypes SNP Subject

30 Mixed-effects linear model Phenotype residual effects P = G + e

31 Mixed-effects linear model Phenotype residual effects P = G + e = Zu

32 Mixed-effects linear model Phenotype residual effects P = G + e = SNPs Zu effect sizes

33 Mixed-effects linear model Phenotype residual effects P = G + e = SNPs Zu effect sizes PP T = var(u)zz T + var(e)i

34 Mixed-effects linear model Phenotype residual effects P = G + e = SNPs Zu effect sizes relationship PP T = var(u)zz T + var(e)i

35 Mixed-effects linear model Phenotype residual effects P = G + e = SNPs Zu effect sizes relationship PP T = var(u)zz T + var(e)i Estimate

36 Mixed-effects linear model Phenotype residual effects P = G + e = SNPs Zu effect sizes relationship PP T = var(u)zz T + var(e)i h 2 = var(u) var(p ) Estimate

37 Linkage Disequilibrium = correlation between markers

38 Linkage Disequilibrium = correlation between markers Related individuals: Common SNPS in LD with rest of genome

39 Linkage Disequilibrium = correlation between markers Related individuals: Common SNPS in LD with rest of genome Difference in h 2 between related and unrelated gives fraction of h 2 explained by common SNPs

40 Genetic correlations SNP variants shared by two traits P 1 = G 1 + e 1 P 2 = G 2 + e 2

41 Genetic correlations SNP variants shared by two traits P 1 = G 1 + e 1 P 2 = G 2 + e 2 cov(p 1,P 2 )=cov(g 1,G 2 )+cov(e 1,e 2 )

42 Genetic correlations SNP variants shared by two traits P 1 = G 1 + e 1 P 2 = G 2 + e 2 cov(p 1,P 2 )=cov(g 1,G 2 )+cov(e 1,e 2 ) r G = cov(g 1,G 2 ) var(g1)var(g2)

43 12771 subjects, ~ 500K SNPS

44 Population structure PC PC1

45 Population structure AA PC PC1

46 Population structure AA Europeans PC PC1

47 12771 subjects, ~ 500K SNPS

48 12771 subjects, ~ 500K SNPS Pruned to 5684 subjects

49 Vattukuti, Guo, Chow. PLoS Genetics 2012 Heritability Table 1. h 2 and h g 2 estimates (ARIC population). BMI WHR GLU INS TG HDL SBP h (0.12) 0.28 (0.12) 0.33 (0.12) 0.23 (0.12) 0.47 (0.11) 0.48 (0.11) 0.30 (0.12) h g (0.05) 0.13 (0.05) 0.10 (0.05) 0.09 (0.05) 0.16 (0.05) 0.12 (0.05) 0.24 (0.05) Mean and standard error estimates from univariate models. doi: /journal.pgen t001

50 Vattukuti, Guo, Chow. PLoS Genetics 2012 Heritability Table 1. h 2 and h g 2 estimates (ARIC population). BMI WHR GLU INS TG HDL SBP h (0.12) 0.28 (0.12) 0.33 (0.12) 0.23 (0.12) 0.47 (0.11) 0.48 (0.11) 0.30 (0.12) h g (0.05) 0.13 (0.05) 0.10 (0.05) 0.09 (0.05) 0.16 (0.05) 0.12 (0.05) 0.24 (0.05) Mean and standard error estimates from univariate models. doi: /journal.pgen t001 Gives cap on how many SNPs to expect

51 Phenotypic correlations related Table 4. Phenotypic correlation coefficients between MetS traits in the ARIC population. BMI WHR GLU INS TG HDL SBP BMI 0.59 (0.04)* 0.20 (0.04)* 0.49 (0.04)* 0.24 (0.04)* (0.04)* 0.25 (0.04)* WHR 0.51 (0.01)* 0.21 (0.04)* 0.43 (0.04)* 0.23 (0.04)* (0.04)* 0.23 (0.04)* GLU 0.24 (0.01)* 0.17 (0.01)* 0.34 (0.04)* 0.21 (0.04)* (0.04)* 0.07 (0.04) INS 0.52 (0.01)* 0.39 (0.01)* 0.35 (0.01)* 0.42 (0.04)* (0.04)* 0.25 (0.04)* TG 0.30 (0.01)* 0.33 (0.01)* 0.19 (0.01)* 0.40 (0.01)* (0.04)* 0.14 (0.04)* HDL (0.01)* (0.01)* (0.01)* (0.01)* (0.01)* (0.04) SBP 0.23 (0.01)* 0.18 (0.01)* 0.15 (0.01)* 0.21 (0.01)* 0.16 (0.01)* (0.01)* Mean and standard error of the Pearson correlation coefficient. Coefficients among related individuals shown in the upper triangle. Coefficients among unrelated individuals shown in the lower triangle. An asterisk indicates significance with p,0.05 adjusted for 21 hypotheses using the two-tailed hypothesis test and normal distribution of the Fisher transformed correlation coefficient. doi: /journal.pgen t004 unrelated

52 Table 2. Genetic and residual correlation coefficients between MetS traits in the ARIC population among related individuals from the bivariate REML model. BMI WHR GLU INS TG HDL SBP related BMI 0.75 (0.16)* 0.23 (0.24) 0.17 (0.27) 0.19 (0.20) (0.21) 0.55 (0.24) WHR 0.52 (0.08)* 0.35 (0.26) 0.67 (0.26)* 0.10 (0.22) (0.22) 0.37 (0.26) GLU 0.19 (0.12) 0.14 (0.12) 0.69 (0.25)* 0.21 (0.21) (0.21) 0.13 (0.27) INS 0.64 (0.08)* 0.35 (0.09)* 0.22 (0.11) 0.76 (0.21)* (0.23) 0.29 (0.29) TG 0.29 (0.12) 0.34 (0.12) 0.21 (0.13) 0.27 (0.11) (0.13)* 0.21 (0.22) HDL (0.12)* (0.12) (0.13) (0.11)* (0.11)* (0.23) SBP 0.11 (0.12) 0.18 (0.11) 0.05 (0.12) 0.24 (0.11) 0.10 (0.13) (0.13) Mean and standard error of the Pearson correlation coefficient for genetic correlations (upper triangle) and residual correlations (lower triangle). An asterisk indicates significance with p,0.05 adjusted for 21 hypotheses using the two-tailed hypothesis test and normal distribution of the Fisher transformed correlation coefficient. doi: /journal.pgen t002 Table 3. Genetic and residual correlations between MetS traits in the ARIC population among unrelated individuals from the bivariate REML model. BMI WHR GLU INS TG HDL SBP unrelated BMI 0.91 (0.18)* 0.01 (0.32) 0.57 (0.24) 0.20 (0.24) (0.28) 0.16 (0.20) WHR 0.44 (0.03)* 0.09 (0.32) 0.33 (0.31) 0.32 (0.23) (0.30) 0.17 (0.21) GLU 0.27 (0.04)* 0.18 (0.04)* 0.05 (0.40) 0.07 (0.30) (0.34) 0.11 (0.24) INS 0.51 (0.03)* 0.40 (0.04)* 0.39 (0.04)* 0.22 (0.29) (0.36) 0.20 (0.25) TG 0.31 (0.04)* 0.33 (0.04)* 0.20 (0.04)* 0.43 (0.04)* (0.19)* (0.19) HDL (0.04)* (0.04)* (0.04)* (0.04)* (0.03)* (0.22) SBP 0.25 (0.05)* 0.18 (0.05)* 0.17 (0.05)* 0.22 (0.04)* 0.21 (0.05)* (0.05) Mean and standard error of the Pearson correlation coefficient for genetic correlations (upper triangle) and residual correlations (lower triangle). An asterisk indicates significance with p,0.05 adjusted for 21 hypotheses using the two-tailed hypothesis test and normal distribution of the Fisher transformed correlation coefficient. doi: /journal.pgen t003 Vattukuti, Guo, Chow. PLoS Genetics 2012