Parental and child genetic contributions to obesity traits in early life based on 83 loci validated in adults: the FAMILY study

Transcription

1 doi: /ijpo Parental and child genetic contributions to obesity traits in early life based on 83 loci validated in adults: the FAMILY study A. Li 1, S. Robiou-du-Pont 1, S. S. Anand 1,2, K. M. Morrison 2,3, S. D. McDonald 1,4, S. A. Atkinson 3,K.K.Teo 1,2 and D. Meyre 1,5 1 Department of Clinical Epidemiology and Biostatistics, McMaster University, Hamilton, Ontario, Canada; 2 Department of Medicine, McMaster University, Hamilton, Ontario, Canada; 3 Department of Pediatrics, Hamilton Health Sciences and McMaster University, Hamilton, Ontario, Canada; 4 Department of Obstetrics and Gynecology, McMaster University, Hamilton, Ontario, Canada; 5 Department of Pathology and Molecular Medicine, McMaster University, Hamilton, Ontario, Canada Address for correspondence: David Meyre, McMaster University, Michael DeGroote Centre for Learning & Discovery, Room 3205, 1280 Main Street West, Hamilton, Ontario L8S 4L8, Canada. meyred@mcmaster.ca Received 1 June 2016; revised 16 November 2016; accepted 18 November 2016 Summary Background: The genetic influence on child obesity has not been fully elucidated. Objective: This study investigated the parental and child contributions of 83 adult body mass index (BMI)-associated single-nucleotide polymorphisms (SNPs) to obesity-related traits in children from birth to 5 years old. Methods: A total of 1402 individuals were genotyped for 83 SNPs. An unweighted genetic risk score (GRS) was generated by the sum of BMI-increasing alleles. Repeated weight and length/height were measured at birth, 1, 2, 3 and 5 years of age, and age-specific and sex-specific weight and BMI Z-scores were computed. Results: The GRS was significantly associated with birthweight Z-score (P = 0.03). It was also associated with weight/bmi Z-score gain between birth and 5 years old (P = 0.02 and , respectively). In longitudinal analyses, the GRS was associated with weight and BMI Z-score from birth to 5 years (P = and , respectively). The maternal effects of rs in DMXL2 on weight and BMI variation from birth to 5 years were significantly greater compared with the paternal effects by Z test (P = and , respectively). Conclusions: SNPs contributing to adult BMI exert their effect at birth and in early childhood. Parent-of-origin effects may occur in a limited subset of obesity predisposing SNPs. Keywords: Birthweight, body mass index, genetic risk score, genome-wide association study, parent-of-origin effect. Introduction Almost one-third of Canadian children aged 5 to 17 years were overweight or with obesity in Childhood obesity is linked to early puberty, type 1 and type 2 diabetes, poor mental and physical health during childhood, as well as adult obesity and its associated comorbidities (1). The total health costs were 21% higher in children with obesity compared with their normal weight counterparts in Canada (2). Therefore, it is urgent to understand the determinants of obesity in early life to manage the burden of childhood obesity and prevent adult obesity more efficiently. Childhood obesity results from environmental, behavioural and genetic influences and their interactions (1). Parental BMI, socioeconomic status, mother s gestational weight gain, gestational smoking status, child s birthweight, weight gain during infancy and infant nutrition are all predictors of childhood/adolescent obesity (1). Temporal effects also influence susceptibility to the development of obesity, the critical windows including gestation, early infancy, adiposity rebound and adolescence (3). Rare deleterious mutations and chromosome abnormalities lead to early onset monogenic obesity (4). Until recently, 114 common single-nucleotide polymorphisms (SNPs) have been identified to be 2016 World Obesity Federation Pediatric Obesity 13, , March 2018

2 134 A. Li et al. associated with adult BMI at genome-wide significance level (P < ) (5). Evidence from crosssectional studies and longitudinal cohorts indicates that some of the adult BMI loci affect BMI in childhood and adolescence (6 10). Meta-analyses of genomewide association studies (GWAS) also identified four loci (FAM120AOS, ELP3, RAB27B and ADAM23) for childhood BMI (8,11). However, the effects of BMIassociated SNPs/genetic risk score on the variation of birthweight and weight gain during early childhood are still unclear (6,12). These studies greatly improved our knowledge of genetic contribution to childhood obesity, but many recently identified adult BMI risk variants have not been investigated in children (previous studies examined at most 32 SNPs identified until 2010). Furthermore, several important maternal variables during pregnancy such as maternal prepregnancy BMI, gestational age, gestational weight gain, gestational diabetes, parity and smoking status were not accounted for when examining the associations between genetic variants and birthweight. The variation of childhood adiposity may also depend upon the parent from whom the variant is inherited, which is called parent-of-origin effects (13,14). Parent-of-origin effects are involved in foetal and placental growth and function, and they have also been reported to be associated with the development of obesity (13,15). But such effects have not been thoroughly investigated. Among different mechanisms that regulate differential maternal and paternal expression, DNA methylation is the most common type of parent-specific epigenetic modifications (14). DNA methylation adds a methyl group to cytosine residues of DNA, and it varies in different cell types and different genotypes. We aimed to investigate the effects of a genetic risk score (GRS) combining 83 SNPs robustly associated with adult BMI on birthweight, weight/bmi gain and growth trajectory from birth to 5 years of age in children using the longitudinal Family Atherosclerosis Monitoring In early life (FAMILY) birth cohort. We also hypothesized that parental risk alleles in specific genes contributed to the variation of child s weight and BMI in early life. Methods Statistical analysis The risk allele for each of the 83 SNPs was determined as previously reported in the literature (Table S1). An additive mode of inheritance was applied to code three genotypes as 0, 1 and 2 designating the number of the BMI-increasing allele. A GRS was calculated by adding up the risk alleles of 83 SNPs, and we used an unweighted GRS as recommended by Dudbridge (16). The values for the missing genotypes (<0.1%) were imputed with arithmetic average of the coded genotypes from individuals who were successfully genotyped. The cross-sectional associations between the GRS and weight or BMI Z-score at birth, 1, 2, 3 and 5 years of age were tested by using multiple linear regression models. Each model was adjusted for the first 10 principal component analysis axes for ethnicity (Supplementary Information), and the associations between the GRS and birthweight/bmi Z-score at birth were adjusted for additional covariates of gestational age, maternal gestational weight gain, parity, gestational diabetes and smoking status. Although these covariates do not confound the association between the BMI GRS and birthweight, they contribute to the variance of birthweight. A sensitivity analysis of the association between the GRS and birthweight/bmi Z-score was performed without adjustment of these covariates. The GRS was also tested for associations with weight and BMI Z-score gain between birth and 5 years old by using multiple linear regression models. Linear mixed-effects models were used to investigate the effect of the GRS on the overall change of weight and BMI Z-score during 0 5 years. This approach was selected because it takes the correlations between repeated measures on the same individual into account and allows for missing measurement data and measurement at different time points, assuming that the missing events are random (17). The weight and BMI Z-score at birth of each individual (intercept) and correlation among measurements on the same subject (slope of age) were modelled as random effects, and time (in years), GRS and ethnicity were modelled as fixed effects. An interaction term, GRS time that tests if the effects of the GRS on weight and BMI Z-score change over time, was also initially added to the linear mixed-effects regression model. Due to its non-significance, this term was removed from the final analysis to produce the most parsimonious model. Predicted trajectories of weight Z-score and BMI Z-score by tertile of the GRS were based on linear mixed-effects models. One of the underlying mechanisms of parent-oforigin effects is imprinting. Imprinting has been suggested to contribute to phenotypic variation; however, it has been confirmed to occur in less than 1% of human genes (14). Theoretically, 1 out of the 83 BMI-associated genes may have a parent-of-origin effect, and its effect in the GRS as an overall measurement of 83 genes will not be detected. We therefore performed an exploratory analysis of the parent-oforigin effect of each SNP. To assess the effects of Pediatric Obesity 13, , March World Obesity Federation

3 Adult BMI SNPs in early life 135 maternal/paternal SNPs on the child weight and BMI Z-score patterns, linear mixed-effects regression was also performed by using the maternal/paternal SNP and child s SNP and ethnicity as fixed effects and the weight/bmi Z-score at birth of each individual (intercept) and correlation among measurements on the same subject (slope of age) as random effects. Existence of a parent-of-origin effect on a specific SNP was tested by a Z-test that compared the effect sizes on child weight/bmi Z-score between maternal and paternal SNPs. Considering the low statistical power of this exploratory test, we did a sensitivity analysis in which only families with complete data of mother, father and child were included (N = 222). This approach would substantially reduce the bias (18). Because of 50% sharing of genetic information between mother/father and offspring, the inclusion of the offspring genotype in the model would likely represent overadjustment for the effect of parental transmitted obesity susceptibility variants. The analyses of the associations between the maternal/paternal genotype and offspring weight and BMI Z-score longitudinally from birth to 5 years old were repeated without adjustment of the offspring genotype. Bonferroni corrected P values are routinely applied to exploratory genetic association studies. However, they are overly conservative given the high prior likelihood of association in post-gwas experiments. Previous studies reported the associations between GWAS BMI-associated genetic variants and weight and BMI during childhood without applying a Bonferroni correction (6,7,12). Therefore, a two-tailed α level of 0.05 was considered significant for the analyses of associations between the GRS and child obesity-related traits. A P value less than the threshold corrected for Bonferroni procedures was considered statistically significant (P < = 0.05 / (4 83)) when exploratory analyses of parent-of-origin effects were applied. All the statistical analyses were performed by using PLINK (version 1.07) and R (version ) (19,20). Additional information on the study participants, anthropometric measurements and DNA genotyping are provided in the Supplementary Information. Results The characteristics of children included in the analysis are shown in Table 1. Among the 816 singletons, genotyping data were available in 541 children. The average birthweight in the analyzed individuals was 3.4 kg (SD = 0.5). The mean BMI at birth was kg m 2. The average gestational age at birth was weeks. The GRS based on 83 adult BMI-associated SNPs ranged from 59 to 94 (mean SD, ). Cross-sectional analyses indicated that one additional risk allele in the GRS increased birthweight Z- score by units (P = 0.03) after adjustment of gestational age, maternal gestational weight gain, parity, gestational diabetes and smoking status. This association still remained without adjustment of these covariates (β SE: /allele; P = 0.02). It was also associated with weight Z-score at 1, 2, 3 and 5 years of age (P 0.02) (Fig. S2-A). Each additional risk allele was associated with weight Z-score gain between birth and 5 years by units (P = 0.02). In the longitudinal analysis, each additional risk allele was significantly associated with unit increase in weight Z-score from birth to 5 years of age (P = ) (Table 2). Predicted weight Z- score by the GRS tertiles between birth and 5 years old showed the growth trajectories at different level of genetic predisposition (Fig. S3-A). Cross-sectional analyses showed that the GRS was not associated with birth BMI Z-score after adjustment of gestational age, maternal gestational weight gain, parity, gestational diabetes and smoking status, but each additional risk allele was significantly associated with unit increase in BMI Z-score at birth without adjustment of these covariates (P = 0.03). It was not associated with BMI Z-score at 1 year, but associations of the GRS with BMI Z-score at 2, 3 and 5 years were observed (P 0.02) (Fig. S2- B). Each additional risk allele was also associated with BMI Z-score gain between birth and 5 years old by units (P = ). In the longitudinal analysis, each additional risk allele was significantly associated with unit increase in BMI Z-score from birth to 5 years of age (P = ) (Table 2). Predicted BMI Z-score by GRS tertiles between birth and 5 years old showed the growth trajectories at different level of genetic predisposition (Fig. S3-B). After controlling for the contribution of child genotype, the longitudinal analyses showed that only maternal risk allele A of rs in DMXL2 was significantly positively associated with weight Z-score (β SE: /allele, P = ), whereas paternal risk allele A for rs was nominally negatively associated with weight Z-score (β SE: /allele, P = )in children from birth to 5 years of age. The difference in the effects of rs between mothers and fathers was statistically significant (P = ) after Bonferroni correction, indicating a parent-of-origin effect. The sensitivity analysis with complete family data showed that the parent-of-origin effect was more significant (P = ). The sensitivity analysis 2016 World Obesity Federation Pediatric Obesity 13, , March 2018

4 136 A. Li et al. Table 1 Characteristics of participants in FAMILY study Measurements Boys (n = 273) Girls (n = 268) Total (n = 541) n Mean SD Z-score n Mean SD Z-score n Mean SD Z-score Weight (kg) Birth year years years years Length/height (cm) Birth year years years years BMI (kg m 2 ) Birth year years years years Weight gain from birth to 5 years (kg) BMI gain Z-score from birth to 5 years (kg m 2 ) Gestational age at birth (week) Maternal gestational weight gain (kg) Mother prepregnancy BMI (kg m 2 ) Father BMI (kg m 2 ) Table 2 Linear mixed modelling of the associations between the GRS and overall changes in weight and BMI Z-score from birth to 5 years of age Trait N β SE P Weight 540 (2263*) BMI 539 (2237*) In the linear mixed-effects models, the weight and BMI Z-scores at birth of each individual (intercept) and correlation among measurements on the same subject (slope of age) were modelled as random effects, and the GRS and ethnicity were modelled as fixed effects. Each additional risk allele in the GRS was significantly associated with unit increase in weight Z-score and unit increase in BMI Z-score. *The numbers indicated the total number of measurements in the longitudinal data. with complete family data showed that the parent-oforigin effect still remained without adjustment of offspring genotype (P = ) (Table 3). Similarly, after controlling for the contribution of child genotype, the longitudinal analyses showed that maternal risk allele A of rs in DMXL2 was nominally positively associated with BMI Z-score (β SE: /allele, P = ), whereas paternal risk allele A for rs was nominally negatively associated with BMI Z-score (β SE: /allele, P = ) in children from birth to 5 years of age. The maternal association of rs with increased child BMI Z-score was the only one to be statistically significant after Bonferroni correction compared with the paternal effect (P = ), indicating a parent-of-origin effect of rs (DMXL2) in the development of childhood obesity. The sensitivity analysis showed that the parent-of-origin effect became more significant (P = ) with complete family data. Pediatric Obesity 13, , March World Obesity Federation

5 Adult BMI SNPs in early life 137 Table 3 Parent-of-origin effects of risk allele A for rs in DMXL2 on childhood obesity traits (overall effects from birth to 5 years of age) Obesity traits Maternal Paternal N β SE P value N β SE P value Z-test P value Weight 422 (1796*) (1051*) BMI 421 (1776*) (1040*) Sensitivity analysis with adjustment of offspring genotype Weight 222 (970*) (970*) BMI 222 (960*) (960*) Sensitivity analysis without adjustment of offspring genotype Weight 222 (970*) (970*) BMI 222 (960*) (960*) Linear mixed-effects regression was performed to assess the overall effects of maternal/paternal SNPs on the child s weight and BMI Z-score from birth to 5 years old, by using the maternal/paternal genotypes and child s genotypes and ethnicity as fixed effects and the weight and BMI Z-score at birth of each individual (intercept) and correlation among measurements on the same subject (slope of age) as random effects. The comparison of the effect sizes between maternal and paternal genetic variants was tested using Z-test. The sensitivity analysis included only families with complete data of mother, father and child (N = 222). *The total number of measurements in the longitudinal analyses. The sensitivity analysis with complete family data showed that the parent-of-origin effect still remained without adjustment of offspring genotype (P = ) (Table 3). Discussion In the present study, we demonstrate that the GRS is associated with birthweight, suggesting that SNPs previously associated with adult BMI have an effect on body composition during foetal growth. These data may support, to some extent, the genetic link between high birthweight and subsequent increased risk of obesity (21). It is important to note that this relationship was observed after adjustment for age, sex, gestational age at birth, ethnicity, gestational weight gain, parity, gestational diabetes and smoking status. This result aligns with observations by Elks et al. of a borderline association (P = 0.05) between the GRS of 11 BMI-associated variants and birthweight using the 1946 British birth cohort (N = 2537) (7). Whereas alterations in the intrauterine environment play important roles in birthweight, 10 40% of the variation in birthweight may be explained by inherited factors (22). The absence of association between adult BMI genetic variants and birthweight in other studies may be explained by a small subset of adult BMI SNPs being examined (6,12,23). GWAS for birthweight have revealed that genetic variants associated with obesity among other different biological pathways influence foetal growth (24,25). The GRS examined in our study covered more of the currently identified BMI adult variants (N = 83 SNPs) compared with previous studies (N 32 SNPs). The increasing association between the GRS and BMI from birth to 5 years (Fig. S2) is consistent with the observation that the genetic influence on BMI becomes stronger with age over childhood (26). Epidemiologic studies have shown that rapid early weight gain predicts later obesity and metabolic diseases (27). Rapid weight gain may occur at any stage, but the greatest variation is commonly seen in the first 1 2 years of life (27). This may be related to influences from the intrauterine environment, and then, children return to their genetic trajectory around 2 years old (27). In our study, the GRS was associated with both weight and BMI gain between birth and 5 years, which has been shown in other studies from 1 to 3 years after birth (6,9). Obesity typically develops over a long period of time. Using longitudinal analyses of repeated measurements of weight and height may identify some specific developmental windows during which the GRS is associated with child growth. In our study, longitudinal analysis of five measurements from birth to 5 years showed that children carrying a high number of obesity risk alleles had higher overall weight and BMI. This is in line with other independent longitudinal studies. Elks et al. found that the GRS of 11 adult BMI loci was positively associated with weight and BMI from birth to 11 years (7). Belsky et al. reported that the GRS derived from 32 BMI loci predicted higher BMI across childhood (ages 3 through 13 years) and adulthood (ages 13 through 38 years) (6). Wermter et al. analyzed the transmission pattern of rs in DLK1 from parents to the offspring with extreme obesity and revealed a significant parent-oforigin effect with a known silencing of maternally 2016 World Obesity Federation Pediatric Obesity 13, , March 2018

6 138 A. Li et al. inherited copy by imprinting (28). Liu et al. found that several SNPs in intron 1 of FTO showed suggestive parent-of-origin effects on BMI variation in Sorbs, a self-contained German population (29). Recently Hoggart et al. developed a novel method and detected two paternal risk alleles in SLC2A10 and KCNK9 that increased child BMI compared with the respective maternal alleles using genome-wide genotype data of unrelated individuals (13). Our study was the first to identify that rs in DMXL2 influenced weight and BMI in early childhood in a parent-of-origin-specific manner. Maternal rs risk allele A increased BMI, whereas the same allele was protective when inherited from the father. The opposite direction of maternal and paternal effects has also been observed in SNPs having strong parent-of-origin effects with BMI (13). Different levels of epigenetic modifications of the genes between mothers and fathers may be caused by different environmental exposures, leading to differential maternal and paternal effects on offspring. At early age of childhood, the influence from intrauterine environment and similar diet between mother and offspring might have contributed to dominant risk effect from mothers (13). Dmx-like 2 or rabconnectin-3, encoded by DMXL2,is involved in the regulation of the Notch signalling pathway that may inhibit brown adipocytes in white adipose tissue and therefore decrease energy expenditure and promote obesity (30). DMXL2 is located on chromosome 15q21.2, not close to any known imprinted genes ( Therefore, this parent-of-origin effect warrants further replication in independent studies. Because a potential mechanism underlying allele-specific expression is DNA methylation, investigating the association between the rs SNP and methylation levels in DMXL2 in relevant tissues would be an exciting perspective (14). Our study had several strengths. First, our longitudinal birth cohort, having weight and BMI at birth, 1, 2, 3 and 5 years of age and critical confounding factors influencing intrauterine environment, enabled us to more accurately investigate the effect of the GRS on weight and BMI at birth and in early childhood (17). Second, the GRS derived from an updated list of 83 SNPs provided the most current state of evidence. Third, a family-based design with genotypes from both parents and child permitted us to investigate the potential parent-of-origin effects of genetic variants on weight and BMI in early childhood. Limitations of this study included the relatively modest sample size, which restricted us from investigating the impact of individual SNPs on BMI in early life, and fewer fathers than mothers having genotype data, which may overestimate the parentof-origin effects. Furthermore, BMI is an imperfect tool to measure fatness in adults and young children, especially before 2 years of age. For etiological studies on obesity, actual measurements of the body fat mass would be preferable. We analyzed BMI in this study because we examined the effect of BMI-associated SNPs. GWAS for BMI and body fat mass showed a partial genetic overlap (5). The association between fat mass-associated SNP or GRS and fat mass in children is beyond the scope of this study but is definitely a relevant topic for upcoming investigations. In conclusion, we provide evidence that the GRS derived from 83 adult BMI SNPs is associated with birthweight. The GRS also predicts higher levels of overall weight and BMI from birth and 5 years of age, suggesting that adult BMI SNPs exert effect in early childhood. Parent-of-origin effects may occur in a limited subset of obesity predisposing SNPs, providing new insights into the mechanisms underlying susceptibility to childhood obesity. Conflict of interest statement No conflict of interest was declared. Author contributions A.L. and D.M. designed the study. S.S.A., K.M.M., S. D.M., S.A.A. and K.K.T. contributed to data collection. D.M. performed the experiments. A.L., S.R.P. and D. M. generated and cleaned up the database and performed the data analysis and interpretation. A.L., S. R.P. and D.M. wrote the manuscript. S.S.A., K.M.M., S.D.M., S.A.A. and K.K.T. provided a critical review of the manuscript. Acknowledgements This study was funded by grants from CIHR (grant no. MOP ), Heart and Stroke Foundation of Ontario (grant no. NA 7293) and the Population Health Research Institute (PHRI), Hamilton Health Science (HHS) and McMaster University. A.L. is funded by the Ontario Graduate Scholarship. S.S.A. holds the Heart and Stroke Foundation of Ontario; Michael G. DeGroote endowed Chair in Population Health and a Canada Research Chair in Ethnicity and Cardiovascular Disease. D.M. holds a Canada Research Chair in Genetics of Obesity, and S.D.M. is supported by a Canada Research Chair in Maternal and Child Obesity Prevention and Intervention. We would also like to thank Hudson Reddon for his comments and help in proofreading of this manuscript. Pediatric Obesity 13, , March World Obesity Federation

7 Adult BMI SNPs in early life 139 References 1. Lakshman R, Elks CE, Ong KK. Childhood obesity. Circulation 2012; 126: Kuhle S, Kirk S, Ohinmaa A, Yasui Y, Allen AC, Veugelers PJ. Use and cost of health services among overweight and obese Canadian children. Int J Pediatr Obes 2011; 6: Dietz WH. Critical periods in childhood for the development of obesity. Am J Clin Nutr 1994; 59: Choquet H, Meyre D. Molecular basis of obesity: current status and future prospects. Curr Genomics 2011; 12: Locke AE, Kahali B, Berndt SI, et al. Genetic studies of body mass index yield new insights for obesity biology. Nature 2015; 518: Belsky DW, Moffitt TE, Houts R, et al. Polygenic risk, rapid childhood growth, and the development of obesity: evidence from a 4-decade longitudinal study. Arch Pediatr Adolesc Med 2012; 166: Elks CE, Loos RJ, Hardy R, et al. Adult obesity susceptibility variants are associated with greater childhood weight gain and a faster tempo of growth: the 1946 British Birth Cohort Study. Am J Clin Nutr 2012; 95: Felix JF, Bradfield JP, Monnereau C, et al. Genome-wide association analysis identifies three new susceptibility loci for childhood body mass index. Hum Mol Genet 2016; 25: Makela J, Lagstrom H, Pitkanen N, et al. Genetic risk clustering increases children s body weight at 2 years of age the STEPS Study. Pediatr Obes 2016; 11: Zandona MR, Sangalli CN, Campagnolo PD, Vitolo MR, Almeida S, Mattevi VS. Validation of obesity susceptibility loci identified by genome-wide association studies in early childhood in South Brazilian children. Pediatr Obes 2016 [Epub ahead of print]. 11. Warrington NM, Howe LD, Paternoster L, et al. A genome-wide association study of body mass index across early life and childhood. Int J Epidemiol 2015; 44: Elks CE, Heude B, de Zegher F, et al. Associations between genetic obesity susceptibility and early postnatal fat and lean mass: an individual participant meta-analysis. JAMA Pediatr 2014; 168: Hoggart CJ, Venturini G, Mangino M, et al. Novel approach identifies SNPs in SLC2A10 and KCNK9 with evidence for parent-of-origin effect on body mass index. PLoS Genet 2014; 10: e Lawson HA, Cheverud JM, Wolf JB. Genomic imprinting and parent-of-origin effects on complex traits. Nat Rev Genet 2013; 14: Frost JM, Moore GE. The importance of imprinting in the human placenta. PLoS Genet 2010; 6: e Dudbridge F. Power and predictive accuracy of polygenic risk scores. PLoS Genet 2013; 9: e Thompson WK, Hallmayer J, O Hara R. Design considerations for characterizing psychiatric trajectories across the lifespan: application to effects of APOE-epsilon4 on cerebral cortical thickness in Alzheimer s disease. Am J Psychiatry 2011; 168: Laird NM, Lange C. Family-based designs in the age of large-scale gene-association studies. Nat Rev Genet 2006; 7: Purcell S, Neale B, Todd-Brown K, et al. PLINK: a tool set for whole-genome association and population-based linkage analyses. Am J Hum Genet 2007; 81: Ihaka R, Genteman R. R: a language for data analysis and graphics. J Comput Graph Stat 1996; 5: Zhao Y, Wang SF, Mu M, Sheng J. Birth weight and overweight/obesity in adults: a meta-analysis. Eur J Pediatr 2012; 171: Lunde A, Melve KK, Gjessing HK, Skjaerven R, Irgens LM. Genetic and environmental influences on birth weight, birth length, head circumference, and gestational age by use of population-based parent-offspring data. Am J Epidemiol 2007; 165: Kilpelainen TO, den Hoed M, Ong KK, et al. Obesitysusceptibility loci have a limited influence on birth weight: a meta-analysis of up to 28,219 individuals. Am J Clin Nutr 2011; 93: Freathy RM, Mook-Kanamori DO, Sovio U, et al. Variants in ADCY5 and near CCNL1 are associated with fetal growth and birth weight. Nat Genet 2010; 42: Horikoshi M, Yaghootkar H, Mook-Kanamori DO, et al. New loci associated with birth weight identify genetic links between intrauterine growth and adult height and metabolism. Nat Genet 2013; 45: Llewellyn CH, Trzaskowski M, Plomin R, Wardle J. From modeling to measurement: developmental trends in genetic influence on adiposity in childhood. Obesity (Silver Spring) 2014; 22: Ong KK, Ahmed ML, Emmett PM, Preece MA, Dunger DB. Association between postnatal catch-up growth and obesity in childhood: prospective cohort study. BMJ 2000; 320: Wermter AK, Scherag A, Meyre D, et al. Preferential reciprocal transfer of paternal/maternal DLK1 alleles to obese children: first evidence of polar overdominance in humans. Eur J Hum Genet 2008; 16: Liu X, Hinney A, Scholz M, et al. Indications for potential parent-of-origin effects within the FTO gene. PLoS One 2015; 10: e Sethi N, Yan Y, Quek D, Schupbach T, Kang Y. Rabconnectin-3 is a functional regulator of mammalian Notch signaling. J Biol Chem 2010; 285: Supporting information Additional Supporting Information may be found in the online version of this article at the publisher s web-site: Table S1. Characteristics of the 83 SNPs associated with adult BMI variation. Figure S1. Flow chart for quality control. Figure S2. Cross-sectional associations between the genetic risk score (GRS) and (A) weight Z-score and (B) BMI Z-score at five different time points from birth to 5 years old. Regression coefficients and 95% CI are 2016 World Obesity Federation Pediatric Obesity 13, , March 2018

8 140 A. Li et al. shown from multiple linear regression models (adjusted for child ethnicity). Each additional risk allele increased weight Z-score by 0.019, 0.019, 0.020, and units at birth, 1, 2, 3 and 5 years old, respectively. Each additional risk allele also increased BMI Z- score by 0.020, 0.010, 0.019, and units at birth, 1, 2, 3 and 5 years old, respectively. Figure S3. Predicted weight Z-score (A) and BMI Z- score (B) by tertiles of the BMI-associated GRS from birth to 5 years old. The predicted values are shown from linear mixed-effects model. The solid line represents the mean weight Z-score in those in the highest tertile of the GRS, the dashed line the middle tertile and the dotted line the lowest tertile. Pediatric Obesity 13, , March World Obesity Federation