Genome-Wide Association Study of Vitamin D Levels in Children: Replication in the Western Australian Pregnancy Cohort (Raine) Study

SHORT COMMUNICATION Genome-wide association study of vitamin D levels in children: replication in the Western Australian Pregnancy Cohort (Raine) study

D Anderson1, BJ Holt1, CE Pennell2, PG Holt1, PH Hart1 and JM Blackwell1

This genome-wide association study (GWAS) utilises data from the Western Australian Pregnancy Cohort (Raine) Study for 25- hydroxyvitamin D (25(OH)D) levels measured in blood collected at age 6 years (n = 673) and at age 14 years (n = 1140). Replication of signiﬁcantly associated genes from previous GWASs was found for both ages. Genome-wide signiﬁcant associations were found both at age 6 and 14 with single nucleotide polymorphisms (SNPs) on chromosome 11p15 in PDE3B/CYP2R1 (age 6: rs1007392, P = 3.9 × 10 − 8; age14: rs11023332, P = 2.2 × 10 − 10) and on chromosome 4q13 in GC (age 6: rs17467825, P = 4.2 × 10 − 9; age14: rs1155563; P = 3.9 × 10 − 9). In addition, a novel association was observed at age 6 with SNPs on chromosome 7p15 near NPY (age 6: rs156299, P = 1.3 × 10 − 6) that could be of functional interest in highlighting alternative pathways for vitamin D metabolism in this age group and merits further analysis in other cohort studies.

Genes and Immunity (2014) 15, 578–583; doi:10.1038/gene.2014.52; published online 11 September 2014

INTRODUCTION RESULTS AND DISCUSSION In humans, vitamin D is obtained primarily through exposure to Descriptive statistics of the study participants are shown in ultraviolet B radiation from sunlight and secondarily through a Table 1. Differences were seen between estimated 25(OH)D levels diet including vitamin D-rich foods or supplements.1 A sunny for the 12 serum samples measured with both enzyme climate does not ensure immunity to vitamin D deficiency. For immunoassay (EIA) and isotope-dilution liquid chromatography- example, a recent Australian population-based health survey tandem mass spectrometry at the age 6 years follow-up estimated prevalence of vitamin D deficiency (serum 25- (Supplementary Figure S1A) but good agreement is seen between hydroxyvitamin D (25(OH)D)o50 nmol l − 1) amongst adults to the two assays at the age 14 years follow-up (Supplementary be 23%.2 This is within the range of deficiency prevalence found in Figure S1B). In light of this, a larger sample of 50 sera from the age other national population-based studies carried out in the United 6 years follow-up were remeasured with liquid chromatography States, Canada, United Kingdom and New Zealand where separation coupled tandem mass spectrometry and it can be seen estimates of deficiency ranged from 18 to 36%.3 Detrimental that the EIA appears to overestimate 25(OH)D levels at higher effects of prolonged vitamin D deficiency on musculoskeletal concentrations (Supplementary Figure S1C), which was also found health include development of rickets in children and osteoma- to be the case in a previous study of 100 serum samples.13 The lacia in adults.4,5 Associations have also been found between 25 main results presented here for both the age 6 years and age 14 (OH)D levels and many other diseases6 including cancer, years follow-ups are for the GWAS performed on standardised 25 cardiovascular disease, influenza type A, rheumatoid arthritis, (OH)D levels (described in Methods; standardisation equations and both type 1 and type 2 diabetes mellitus. shown in Supplementary Figure S2) with results for the GWAS Previous twin and family-based studies provide evidence performed on 25(OH)D levels determined by EIA presented in the that 25(OH)D levels are determined by a significant genetic supplementary information. component, with estimates of heritability varying widely from 23 Manhattan plots of the GWAS analyses are shown in Figure 1 to 80%.7 A number of genome-wide association studies (GWASs) and Supplementary Figure S3. Quantile-quantile plots of the have identified genetic variants associated with 25(OH)D observed versus expected − log10 P-values did not reveal levels,8–11 with only one of these focussing on children.10 This evidence for systematic bias in either the age 6 year follow-up study presents results from a GWAS utilising data from The (inflation factor λ = 1.01; Supplementary Figure S4A) or age 14 year Western Australian Pregnancy Cohort (Raine) Study12 (hereinafter follow-up (inflation factor λ = 1.01; Supplementary Figure S4B) referred to as the Raine Study) for 25(OH)D levels measured results. Supplementary Tables S1 and S2 list the top 50 associated in blood collected at both the age 6 years and age 14 years single nucleotide polymorphisms (SNPs) for the age 6 years follow-ups. Replication of significantly associated genetic variants follow-up and the age 14 years follow-up, respectively. Table 2 with 25(OH)D levels from previously published GWASs was also summarises results for the top SNP in each of the genes of assessed. functional interest identified by these top 50 SNPs. Regional plots

1Telethon Kids Institute, The University of Western Australia, Subiaco, Western Australia, Australia and 2School of Women's and Infants' Health, The University of Western Australia, Perth, Western Australia, Australia. Correspondence: Professor JM Blackwell, The University of Western Australia, Telethon Kids Institute, 100 Roberts Road, Subiaco, 6008, Western Australia, Australia. E-mail: [email protected] Received 15 May 2014; revised 15 July 2014; accepted 17 July 2014; published online 11 September 2014 Novel pathways for vitamin D metabolism in children D Anderson et al 579 of association are shown in Figure 2 for both the age 6 years Table 1. Characteristics of the Raine Study participants at the age 6 follow-up and the age 14 years follow-up. years and age 14 years follow-upsa Genome-wide significant associations were found for SNPs fi Age 6 years Age 14 years located within intronic regions of the group-speci c component follow-up follow-up (vitamin D binding protein) (GC) gene on chromosome 4q13 for − (n = 673) (n = 1140) both the age 6 years (lowest P = 4.2 × 10 9, rs17467825, Figure 2a) and the age 14 years follow-ups (lowest P = 3.9 × 10 − 9, rs1155563, Male (n (%)) 364 (54) 592 (52) Figure 2c). The GC gene encodes a vitamin D binding protein that Age, years (mean (s.d.)) 5.9 (0.19) 14.1 (0.21) transports vitamin D to various tissues of the body.14 Significant BMI (median (IQR)) 16 (14.8–16.5) 20 (18.5–23.0) 25(OH)D, nmol l − 1 101 (84–123) 81 (68–98) associations were also found for SNPs located on chromosome (median (IQR)) 11p15 within either the phosphodiesterase 3B, cGMP-inhibited (PDE3B) or cytochrome P450, family 2, subfamily R, polypeptide 1 − Caucasian mother and father (n (%)) (CYP2R1) genes for both the age 6 years (lowest P = 3.9 × 10 8, Yes 645 (96) 1073 (94) rs1007392, Figure 2b) and the age 14 years follow-ups (lowest No 28 (4) 67 (6) P = 2.2 × 10 − 10, rs11023332, Figure 2d). The top associated SNPs

b on chromosome 11p15 are located in an intron of PDE3B but the Season of blood collection (n (%)) true functional association is believed to be with CYP2R1 as these Summer 126 (19) 189 (17) Autumn 165 (25) 320 (28) SNPs are in strong linkage disequilibrium with the top associated Winter 205 (30) 308 (27) SNPs in CYP2R1 (Figure 2b and d). The CYP2R1 gene encodes an 15 Spring 177 (26) 323 (28) enzyme that converts vitamin D to 25(OH)D. Nevertheless, it is also plausible that the association could be with PDE3B because Abbreviations: BMI, body mass index; IQR, interquartile range; 25(OH)D, there is evidence that vitamin D and phosphodiesterase 3B are 25-hydroxyvitamin D. BMI is defined as weight(kg)/height(m2). aDescriptive both involved in a common pathway controlling lipid statistics are shown for those participants with genome-wide SNP 16,17 genotyping data. bThe four seasons are defined as Summer (December metabolism. Specifically, it is thought that parathyroid to February), Autumn (March to May), Winter (June to August) and Spring hormone promotes conversion of 25(OH)D to 1α,25-dihydroxyvi- (September to November). tamin D3 (1,25(OH)2D) which then stimulates calcium influx in adipocytes. This influx of calcium inhibits lipolysis through

Figure 1. Manhattan plots showing the association between standardised 25(OH)D levels and SNP allele dosages for the age 6 years follow-up (a) and for the age 14 years follow-up (b). The top SNPs meeting genome-wide signiﬁcance (P ⩽ 5×10− 8) are shown in red and SNPs within 400 kilobases of these top SNPs are highlighted in red.

Table 2. Top associated SNPs in/near genes of functional interest for both the age 6 and age 14 years follow-ups

̂ SNP Chr Locationa Major/Minor allele MAFb Risk allele P-valuec eβ (95% CI)d Neighbouring gene(s)e (SNP functional class)f

Age 6 years follow-up rs17467825 4 72605517 A/G 0.30 A 4.2 × 10 − 9 1.07 (1.05–1.09) GC rs10766192 11 14811686 A/G 0.41 A 4.0 × 10 − 8 1.06 (1.04–1.08) CYP2R1

Age 14 years follow-up rs1155563 4 72643488 T/C 0.28 C 3.9 × 10 − 9 0.94 (0.92–0.96) GC (intron variant) rs10500804 11 14910273 T/G 0.42 G 3.5 × 10 − 10 0.94 (0.92–0.96) CYP2R1 (intron variant) Abbreviations: SNP, single nucleotide polymorphism. aChromosomal location based on NCBI Human Genome Build 37 coordinates. bMinor allele frequency. cP-values were obtained for the linear regression parameter estimate, β^, of each SNP by comparing the Wald test statistic to a χ2 distribution with one degree of ^ freedom. deβ represents the proportionate change in the mean 25(OH)D level for each extra copy of the risk allele. eA flanking distance of 100 kilobases either side of the SNP location was used to define neighbouring genes of functional interest. Genes of functional interest are those genes that are functionally relevant and/or have been found to be associated with vitamin D levels in previous GWASs. Significant associations have been found at the NADSYN1/DHCR7 locus in previous GWASs but SNPs at this locus were not present in our top 50 SNPs (age 6 years follow-up top SNP rs12793607 (P-value 0.0001); age 14 years follow-up top SNP rs4945008 (P-value 0.0002)). fOnly those SNPs residing within the boundary of a gene have a functional class.

Figure 2. Regional LocusZoom association plots for SNPs at GC (age 6 years follow-up (a); age 14 years follow up (c)) and CYP2R1 (age 6 years follow-up (b); age 14 years follow-up (d)). The left-hand y axis shows − log10 P-values of association. Genotyped SNPs are shown as circles and imputed SNPs are shown as triangles. The top associated SNP is shown in purple. Colour coding of the other plotted SNPs in the region represents linkage disequilibrium (r2) with the top associated SNP that was calculated using 1000 Genomes EUR samples. The recombination rate for the 1000 Genomes EUR samples is indicated by the blue line and the right-hand y axis. The x axis shows the chromosomal position of the SNPs and genes based on NCBI Human Genome Build 37 coordinates.

activation of phosphodiesterase 3B, which suppresses the lipolytic We hereafter report the results from our two follow-up analyses response to catecholamines.18 for the top reported SNPs in these previous GWASs. Associations Signiﬁcant associations between 25(OH)D levels and SNPs in GC, with the GC intronic SNP, rs2282679, located on chromosome CYP2R1, NAD synthetase 1 (NADSYN1) and 7-dehydrocholesterol 4q13 were previously found in the GWASs conducted by reductase (DHCR7) have been reported in previous GWASs.8–11 Ahn et al.8 (P = 1.8 × 10 − 49), Wang et al.11 (P = 1.9 × 10 − 109) and

Genes and Immunity (2014) 578 – 583 © 2014 Macmillan Publishers Limited Novel pathways for vitamin D metabolism in children D Anderson et al 581 Lasky-Su et al.10 (P = 2.1 × 10 − 14) and we found a genome-wide association, it seems likely that we should have seen replication of significant association for this SNP in our study for the age 6 years the association in the age 14 years follow-up analysis. follow-up (P = 4.7 × 10 − 9) and a borderline genome-wide signifi- Animal models have shown that NPY suppresses bone − cant association for the age 14 years follow-up (P = 5.7 × 10 8). formation through the Y2 receptor in the hypothalamus and Ahn et al.8 reported intronic rs1993116 to be the top associated locally in the bone through the osteoblastic Y1 receptor.21 Given SNP in the CYP2R1 gene located on chromosome 11p15 that vitamin D is well known to be involved in bone homeostasis, − (P = 2.9 × 10 17) and we found suggestive association for this the association seen between NPY and 25(OH)D levels could be − SNP in our study (age 6 years follow-up P = 2.5 × 10 6; age 14 through the effect of NPY on bone homeostasis. The active form − 7 11 years follow-up P = 3.6 × 10 ). Wang et al. found rs10741657 to of vitamin D, 1,25(OH)2D, stimulates absorption of calcium from be the top associated SNP on chromosome 11p15 which is located the gastrointestinal tract and reabsorption of calcium from the − near CYP2R1 (P = 3.3 × 10 20) and we also found suggestive kidneys into the circulation which is used by osteoblasts to form − association for this SNP (age 6 years follow-up P = 2.6 × 10 6; bone. If NPY suppresses bone formation, then there would − age 14 years follow-up P = 4.5 × 10 7). We did not replicate presumably be less demand from osteoblasts for circulating fi fi ndings of signi cant associations for SNPs on chromosome calcium, which could lead to suppression of 1,25(OH)2D. The fact 11q13 in the region of NADSYN1 and DHCR7 as reported by both that the association was only seen for the age 6 years follow-up 8 11 8 Ahn et al. and Wang et al. The GWASs by Ahn et al. and Wang analysis and not the age 14 years follow-up analysis could be 11 et al. were large meta-analyses performed on cohorts of because of a difference in the primary bone homeostasis European ancestry so the most likely reason for non-replication pathways at age 6 versus age 14. In particular, sex hormones at NADSYN1 and DHCR7 in our GWAS is lack of power due to the are major contributors to bone homeostasis from commencement smaller sample sizes used in our study. of puberty,23 which is relevant for our age 14 years follow-up data When the GWAS was performed on 25(OH)D levels determined but not our age 6 years follow-up data. Pertaining to this, 88% of by EIA for the age 6 years follow-up, the only genome-wide the 14-year-old females had experienced menarche prior to blood significant association was for SNPs located on chromosome 7p15, collection. Interestingly, studies using animal models have found − 8 with the top associated SNP, rs156299 (P = 3.0 × 10 ), being that NPY and sex hormones interact in the bone homeostasis located ≈ 99 kilobases upstream of the neuropeptide Y (NPY) gene pathway24 further highlighting the difference expected in bone (Supplementary Figure S3). The strength of this association homeostasis at age 6 versus age 14. diminished when the GWAS was performed on standardised 25 − 6 In conclusion, we have conducted a GWAS of serum 25(OH)D (OH)D levels (Figure 1) but remains suggestive (P = 1.3 × 10 , levels in children at ages 6 and 14 years. We have replicated rs156299, Supplementary Figure S5). This association was not findings found in previous GWASs and present a possible novel replicated in our age 14 years follow-up analysis and we are finding for the age 6 years follow-up analysis. Replication of this fi unaware of any previous studies reporting such a nding. One novel finding in another cohort of the same age is recommended. possibility is that this association could be a spurious result in light of the disagreement seen between the EIA and LC-MS assays (Supplementary Figures S1A and S1C). It is not surprising to find MATERIALS AND METHODS disagreement between the two assays as it is commonly The Raine Study obtained consent from study participants and ethics acknowledged that there is considerable interassay variability approval for recruitment and subsequent follow-ups from the Ethics when measuring 25(OH)D and it remains unclear which is the Committee at King Edward Memorial Hospital and/or Princess Margaret ‘best’ assay.19 We have presented the GWAS performed on Hospital (Perth, WA, Australia). Measurement of 25(OH)D levels in the age 6 standardised 25(OH)D levels in the main results rather than the years and age 14 years follow-ups has been described previously.25 GWAS performed on 25(OH)D levels measured by EIA as Collection of serum samples took place over a 3-year period; from 1995 to associations were strengthened for genes found in previous 1998 for the age 6 years follow-up and from 2003 to 2006 for the age 14 years follow-up. Serum from the age 6 years (n = 989) and age 14 years GWASs; from borderline for GC and suggestive for PDE3B/CYP2R1 − fi (n = 1 380) follow-ups stored at 80 °C was thawed and 25(OH)D levels to genome-wide signi cant for both genes after standardisation. were measured using the Immunodiagnostic Systems Inc. 25(OH)D EIA kit Notwithstanding this, we considered reasons for a possible (Scottsdale, AZ, USA). Inter-plate coefficients of variation (CV) were association between NPY and 25(OH)D levels. One explanation is determined using control samples for both the age 6 years follow-up that the association between NPY and 25(OH)D levels seen for our (CV = 6.5% at mean 25(OH)D of 32 nmol l − 1; CV = 6.3% at mean 25(OH)D of age 6 years follow-up analysis relates to differences in the way 76 nmol l − 1; CV = 3.3% at mean 25(OH)D of 116 nmol l − 1) and the age 14 vitamin D is metabolised in younger children compared to post- years follow-up (CV = 5.9% at mean 25(OH)D of 33 nmol l − 1; CV = 4.5% at − pubescent children and adults. mean 25(OH)D of 77 nmol l 1; CV = 4.9% at mean 25(OH)D of 117 − 1 NPY is a neurotransmitter found in abundance in the central nmol l ). Measurement of 25(OH)D was also repeated for 12 samples at 20 both the age 6 years and age 14 years follow-ups using isotope-dilution and peripheral nervous systems. The effects of NPY are 26 mediated by NPY receptors with the subtypes Y1, Y2 and Y5 liquid chromatography-tandem mass spectrometry performed by RMIT being the best understood of the receptors.21 The diverse range of Drug Discovery Technologies (Melbourne, VIC, Australia) and for 50 samples at the age 6 years follow-up using two-dimensional ultra physiological processes mediated by these receptors includes performance liquid chromatography separation coupled tandem mass amongst others food intake, bone homeostasis, cardiovascular spectrometry27 performed by Metabolomics Australia (Crawley, WA, 22 regulation, mood and circadian rhythm. We consider food intake Australia). or bone homeostasis to be the most likely processes underlying Genomic DNA was extracted from whole blood or saliva (≈ 5% of the association seen between NPY and 25(OH)D levels in our age 6 samples) collected at the age 14 or 17 years follow-ups. Genotyping of the years follow-up analysis because vitamin D levels can be altered samples (n = 1 593) was performed by the Centre for Applied Genomics through food intake and vitamin D plays a vital role in bone (Toronto, ON, Canada) using the Illumina Human660W-Quad BeadChip (Illumina Inc., San Diego, CA, USA). The SNP genotyping data were homeostasis. 28 Food intake could be an effect mediator of the association seen subjected to a number of quality control checks within PLINK. Samples between NPY and 25(OH)D levels. This makes the assumption that where more than 3% of SNP genotypes were called as missing were excluded (n = 16) and this removed all samples with excessive hetero- increasing food intake is associated with ingestion of greater zygosity rates. Samples with an excessive homozygosity rate were also amounts of dietary vitamin D. We did adjust for body mass index excluded (n = 3; Fixation index 40.07). Exclusions were made because of in our analyses, but this is not a perfect proxy for food intake and gender discrepancies between reported gender and gender inferred from information regarding food intake or vitamin D supplementation the X chromosome data (n = 7). Where pairs of samples showed was not available. However, if this is the sole reason for the relatedness equivalent to second-degree relatives or greater (estimated

© 2014 Macmillan Publishers Limited Genes and Immunity (2014) 578 – 583 Novel pathways for vitamin D metabolism in children D Anderson et al 582 identity by descent 40.1875), the member of the pair with the highest 3 Daly RM, Gagnon C, Lu ZX, Magliano DJ, Dunstan DW, Sikaris KA et al. Prevalence missing SNP genotyping call rate was removed (n = 73). Of the 1494 of vitamin D deficiency and its determinants in Australian adults aged 25 years samples passing quality control, 25(OH)D measurements were available for and older: a national, population-based study. Clin Endocrinol (Oxf) 2012; 77: 673 participants at the age 6 year follow-up and 1140 participants at the 26–35. age 14 year follow-up. 4 Munns CF, Simm PJ, Rodda CP, Garnett SP, Zacharin MR, Ward LM et al. Incidence Quality control of the SNPs included exclusions based on a call rate less of vitamin D deficiency rickets among Australian children: an Australian Paediatric than 95%, evidence of deviation from Hardy–Weinberg equilibrium Surveillance Unit study. Med J Aust 2012; 196:466–468. − 7 (Po5.7 × 10 ) and minor allele frequency less than 1%. Imputation of 5 Reid IR, Bolland MJ, Grey A. Effects of vitamin D supplements on bone mineral 29 untyped SNPs was performed using MaCH with the HapMap Phase II density: a systematic review and meta-analysis. Lancet 2014; 383:146–155. (Release 22) CEU samples as reference genotypes for prediction. Imputed 6 Nowson CA, McGrath JJ, Ebeling PR, Haikerwal A, Daly RM, Sanders KM et al. 2 SNPs with a minor allele frequency less than 1% or a MaCH r quality score Vitamin D and health in adults in Australia and New Zealand: a position state- less than 0.3 were excluded. This increased the number of SNPs available ment. Med J Aust 2012; 196:686–687. for analysis from the 535 632 passing quality control to 2 461 244. 7 Dastani Z, Li R, Richards B. Genetic regulation of vitamin D levels. Calcif Tissue Int Standardisation of 25(OH)D levels for the age 6 year and age 14 year 2013; 92:106–117. follow-ups was performed by fitting a weighted Deming regression model 8 Ahn J, Yu K, Stolzenberg-Solomon R, Simon KC, McCullough ML, Gallicchio L et al. 30 31 in R using the mcreg() function available in the mcr package. The Genome-wide association study of circulating vitamin D levels. Hum Mol Genet two models were fitted using the 50 age 6 year samples remeasured with 2010; 19: 2739–2745. liquid chromatography separation coupled tandem mass spectrometry 9 Engelman CD, Meyers KJ, Ziegler JT, Taylor KD, Palmer ND, Haffner SM et al. and the 12 age 14 year samples remeasured with isotope-dilution liquid Genome-wide association study of vitamin D concentrations in Hispanic Americans: chromatography-tandem mass spectrometry. These models were used to the IRAS family study. J Steroid Biochem Mol Biol 2010; 122:186–192. predict 25(OH)D levels as measured by liquid chromatography separation 10 Lasky-Su J, Lange N, Brehm JM, Damask A, Soto-Quiros M, Avila L et al. Genome- coupled tandem mass spectrometry at age 6 and by isotope-dilution liquid wide association analysis of circulating vitamin D levels in children with asthma. chromatography-tandem mass spectrometry at age 14 for the remainder Hum Genet 2012; 131: 1495–1505. of each cohort based on 25(OH)D levels measured by EIA. 11 Wang TJ, Zhang F, Richards JB, Kestenbaum B, van Meurs JB, Berry D et al. Given that the serum samples were collected throughout all seasons of Common genetic determinants of vitamin D insufficiency: a genome-wide the year, a cosinor model was used to remove the seasonal variability in 25 association study. Lancet 2010; 376:180–188. (OH)D levels. The cosinor model was fitted in R using the cosinor() 32 12 Newnham JP, Evans SF, Michael CA, Stanley FJ, Landau LI. Effects of frequent function available in the season package. Deseasonalisation of 25(OH)D ultrasound during pregnancy: a randomised controlled trial. Lancet 1993; 342: levels was performed by adding the overall mean 25(OH)D level to each 887–891. ’ 33 individual s actual 25(OH)D level minus their predicted 25(OH)D level. 13 Knox S, Harris J, Calton L, Wallace AM. A simple automated solid-phase extraction Imputed SNP allele dosages were tested for association with log- procedure for measurement of 25-hydroxyvitamin D3 and D2 by liquid transformed, deseasonalised 25(OH)D levels under an additive linear 34 chromatography-tandem mass spectrometry. Ann Clin Biochem 2009; 46 (Pt 3): regression model using ProbABEL v.0.3.0. Models were adjusted for 226–230. gender and body mass index. To adjust for population stratification, 14 Chun RF. New perspectives on the vitamin D binding protein. Cell Biochem Funct principal component analysis was carried out on a set of 42 888 pruned 2012; 30: 445–456. SNPs in linkage equilibrium (pairwise r2 ⩽ 0.1) using the smartpca program 15 Shinkyo R, Sakaki T, Kamakura M, Ohta M, Inouye K. Metabolism of vitamin D by (available in the EIGENSTRAT35 suite of programs) and ancestry estimates human microsomal CYP2R1. Biochem Biophys Res Commun 2004; 324:451–457. from the top five principal components were included in the models. 16 McCarty MF, Thomas CA. PTH excess may promote weight gain by impeding Manhattan plots of the GWAS results were generated in R using the catecholamine-induced lipolysis-implications for the impact of calcium, vitamin D, manhattanplot() function available in the postgwas package.36 − and alcohol on body weight. Med Hypotheses 2003; 61:535–542. Genome-wide significant associations were defined as (P ⩽ 5×10 8) and − 17 Zemel MB, Shi H, Greer B, Dirienzo D, Zemel PC. Regulation of adiposity by dietary borderline genome-wide significant associations as (P45×10 8 and − calcium. FASEB J 2000; 14: 1132–1138. P ⩽ 1×10 7).37 Regional plots of association were generated using 18 Reinehr T, de Sousa G, Alexy U, Kersting M, Andler W. Vitamin D status and LocusZoom.38 The SNP functional class and neighbouring genes were parathyroid hormone in obese children before and after weight loss. Eur J obtained from NCBI using the GetSNPInfo() and GetNeigh- Endocrinol 2007; 157:225–232. Genes() functions available in the NCBI2R package.39 19 Lai JK, Lucas RM, Clements MS, Harrison SL, Banks E. Assessing vitamin D status: pitfalls for the unwary. Mol Nutr Food Res 2010; 54: 1062–1071. CONFLICT OF INTEREST 20 Thorsell A, Heilig M. Diverse functions of neuropeptide Y revealed using geneti- cally modified animals. Neuropeptides 2002; 36:182–193. fl The authors declare no con ict of interest. 21 Franquinho F, Liz MA, Nunes AF, Neto E, Lamghari M, Sousa MM. Neuropeptide Y and osteoblast differentiation--the balance between the neuro-osteogenic net- work and local control. FEBS J 2010; 277: 3664–3674. ACKNOWLEDGEMENTS 22 Brothers SP, Wahlestedt C. Therapeutic potential of neuropeptide Y (NPY) We are extremely grateful to all the families who took part in this study and the receptor ligands. EMBO Mol Med 2010; 2: 429–439. whole Raine Study team, which includes data collectors, cohort managers, data 23 Wang Q, Nicholson PH, Suuriniemi M, Lyytikainen A, Helkala E, Alen M et al. managers, clerical staff, research scientists and volunteers. The Raine Study has been Relationship of sex hormones to bone geometric properties and mineral density supported by the National Health and Medical Research Council over the last 20 years in early pubertal girls. J Clin Endocrinol Metab 2004; 89:1698–1703. with funding for core management provided by The University of Western Australia 24 Zengin A, Zhang L, Herzog H, Baldock PA, Sainsbury A. Neuropeptide Y and sex (UWA), Raine Medical Research Foundation, Telethon Kids Institute, UWA Faculty of hormone interactions in humoral and neuronal regulation of bone and fat. Trends Medicine, Dentistry and Health Sciences, Women and Infants Research Foundation, Endocrinol Metab 2010; 21:411–418. Curtin University and Edith Cowan University. The 5-year cohort review was 25 Hollams EM, Hart PH, Holt BJ, Serralha M, Parsons F, de Klerk NH et al. Vitamin D supported by the National Health and Medical Research Council (ID 963209) and The and atopy and asthma phenotypes in children: a longitudinal cohort study. Eur Raine Foundation. The 14-year follow-up was supported by the National Health and Respir J 2011; 38: 1320–1327. Medical Research Council (ID 211912 and ID 003209) and The Raine Foundation. 26 Maunsell Z, Wright DJ, Rainbow SJ. Routine isotope-dilution liquid Genome-wide SNP genotyping data generated from samples collected at either the chromatography-tandem mass spectrometry assay for simultaneous measure- 14- or 17-year follow-ups were supported by the National Health and Medical ment of the 25-hydroxy metabolites of vitamins D2 and D3. Clin Chem 2005; 51: Research Council (ID 572613). 1683–1690. 27 Clarke MW, Tuckey RC, Gorman S, Holt B, Hart PH. Optimized 25-hydroxyvitamin D analysis using liquid-liquid extraction with 2D separation with LC/MS/MS detec- REFERENCES tion, provides superior precision compared to conventional assays. Metabolomics 1 Liu J. Vitamin D content of food and its contribution to vitamin D status: a brief 2013; 9: 1031–1040. overview and Australian focus. Photochem Photobiol Sci 2012; 11: 1802–1807. 28 Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MA, Bender D et al. PLINK: 2 Australian Bureau of Statistics. Australian Health Survey: Biomedical Results for a tool set for whole-genome association and population-based linkage analyses. Nutrients, 2011-12 cat. no. 4364.0.55.006 2014. Am J Hum Genet 2007; 81:559–575.

Genes and Immunity (2014) 578 – 583 © 2014 Macmillan Publishers Limited Novel pathways for vitamin D metabolism in children D Anderson et al 583 29 Li Y, Willer CJ, Ding J, Scheet P, Abecasis GR. MaCH: using sequence and genotype 35 Price AL, Patterson NJ, Plenge RM, Weinblatt ME, Shadick NA, Reich D. Principal data to estimate haplotypes and unobserved genotypes. Genet Epidemiol 2010; 34: components analysis corrects for stratiﬁcation in genome-wide association stu- 816–834. dies. Nat Genet 2006; 38: 904–909. 30 R Core Team. R: A Language and Environment for Statistical Computing In R 36 Hiersche M. postgwas: GWAS Post-Processing Utilities. In: R package version 1.02, Foundation for Statistical Computing: Vienna, Austria, 2013. 2012. 31 Manuilova E, Schuetzenmeister A, Model F. mcr: Method Comparison Regression. 37 Panagiotou OA, Ioannidis JP. What should the genome-wide signiﬁcance In. 2012. threshold be? Empirical replication of borderline genetic associations. Int J 32 Barnett A, Baker P, Dobson A. Season: Analysing Seasonal Data R Functions. In. R Epidemiol 2012; 41:273–286. package version 0.3-2 ed, 2012. 38 Pruim RJ, Welch RP, Sanna S, Teslovich TM, Chines PS, Gliedt TP et al. LocusZoom: 33 van der Mei IA, Ponsonby AL, Dwyer T, Blizzard L, Taylor BV, Kilpatrick T et al. regional visualization of genome-wide association scan results. Bioinformatics Vitamin D levels in people with multiple sclerosis and community controls in 2010; 26:2336–2337. Tasmania, Australia. J Neurol 2007; 254:581–590. 39 Melville S. NCBI2R: NCBI2R-An R package to navigate and annotate genes and 34 Aulchenko YS, Struchalin MV, van Duijn CM. ProbABEL package for genome-wide association analysis of imputed data. BMC Bioinformatics 2010; 11: 134. SNPs. In. R package version 1.4.4 ed, 2012.

Supplementary Information accompanies this paper on Genes and Immunity website (http://www.nature.com/gene)