Human Molecular Genetics, 2012, Vol. 21, No. 14 3275–3282 doi:10.1093/hmg/dds136 Advance Access published on April 10, 2012 Meta-analysis of two genome-wide association studies identifies four genetic loci associated with thyroid function Rajesh Rawal1,∗,{, Alexander Teumer5,{, Henry Vo¨ lzke6,{, Henri Wallaschofski7, Till Ittermann6, Bjørn O. A˚ svold8,9, Trine Bjøro10,11, Karin H. Greiser12,13, Daniel Tiller13, Karl Werdan14, Henriette E. Meyer zu Schwabedissen15, Angela Doering2,3, Thomas Illig4,16, Christian Gieger1, Christa Meisinger3,{ and Georg Homuth5,{ 1 2 3 4 Institute of Genetic Epidemiology, Institute of Epidemiology I, Institute of Epidemiology II and Research Unit of Downloaded from Molecular Epidemiology, Helmholtz Zentrum Mu¨nchen, 85764 Neuherberg, Germany 5Interfaculty Institute for Genetics and Functional Genomics, 6Institute for Community Medicine and 7Institute of Clinical Chemistry and Laboratory Medicine, Ernst Mortiz-Arndt-University Greifswald, 17475 Greifswald, Germany 8Department of Public Health, Norwegian University of Science and Technology, 7491 Trondheim, Norway 9Department of Endocrinology, Trondheim University Hospital, 7491 Trondheim, Norway 10Department of Medical Biochemistry, Division of http://hmg.oxfordjournals.org/ Diagnostics and Intervention, Olso University Hospital, 0104 Oslo, Norway, 11Faculty of Medicine, University of Olso, 0104 Oslo, Norway, 12Division of Cancer Epidemiology, German Cancer Research Center (DKFZ), 69126 Heidelberg, Germany, 13Institute of Medical Epidemiology, Biostatistics, and Informatics and 14Department of Medicine III, University Clinics Halle, Martin-Luther-University Halle-Wittenberg, 06108 Halle, Germany, 15Department of Pharmacology, Ernst Mortiz-Arndt-University Greifswald, 17491 Greifswald, Germany and 16Hannover Unified Biobank, Hannover Medical School, 30625 Hannover, Germany at GSF Zentralbibliothek on July 27, 2012 Received November 22, 2011; Revised March 29, 2012; Accepted April 2, 2012 Thyroid hormones play key roles in cellular growth, development and metabolism. Although there is a strong genetic influence on thyroid hormone levels, the genes involved are widely unknown. The levels of circulat- ing thyroid hormones are tightly regulated by thyrotropin (TSH), which also represents the most important diagnostic marker for thyroid function. Therefore, in order to identify genetic loci associated with TSH levels, we performed a discovery meta-analysis of two genome-wide association studies including two cohorts from Germany, KORA (n 5 1287) and SHIP (n 5 2449), resulting in a total sample size of 3736. Four genetic loci at 5q13.3, 1p36, 16q23 and 4q31 were associated with serum TSH levels. The lead single- nucleotide polymorphisms of these four loci were located within PDE8B encoding phosphodiesterase 8B, up- stream of CAPZB that encodes the b-subunit of the barbed-end F-actin-binding protein, in a former ‘gene desert’ that was recently demonstrated to encode a functional gene (LOC440389) associated with thyroid volume, and upstream of NR3C2 encoding the mineralocorticoid receptor. The latter association for the first time suggests the modulation of thyroid function by mineral corticoids. All four loci were replicated in three additional cohorts: the HUNT study from Norway (n 5 1487) and the two German studies CARLA (CARLA, n 5 1357) and SHIP-TREND (n 5 883). Together, these four quantitative trait loci accounted for ∼3.3% of the variance in TSH serum levels. These results contribute to our understanding of genetic factors and physiological mechanisms mediating thyroid function. ∗ To whom correspondence should be addressed at: Institute of Genetic Epidemiology, Helmholtz Zentrum Mu¨nchen, German Research Center for En- vironmental Health, Ingolstaedter Landstr. 1, 85764 Neuherberg, Germany. Tel: +49-8931873101; Fax: +49-8931873380; Email: rajesh.rawal@ helmholtz-muenchen.de †These authors contributed equally to this work. # The Author 2012. Published by Oxford University Press. All rights reserved. For Permissions, please email: [email protected] 3276 Human Molecular Genetics, 2012, Vol. 21, No. 14 INTRODUCTION independence of the four SNPs from each other. The distribu- tion of the mean log TSH levels depending on the number of Thyroid hormones [thyroxine (T4) and triiodothyronine (T3)] TSH increasing alleles for both discovery cohorts is shown in are essentially involved in metabolic regulation, and minor var- Supplementary Material, Table S2. Additional adjustment for iations in serum thyroid hormone levels can influence the basal the participant’s thyroid peroxidase (TPO) antibody status in metabolic rate, protein synthesis, fat and carbohydrate metabol- the regression model did not influence the P-values (Table 1), ism and cellular response to catecholamines (1,2). Accordingly, suggesting that the associations of the four SNPs with TSH serum thyroid hormone levels are tightly regulated by thyro- levels are independent of the TPO antibody status. tropin (TSH) from the pituitary gland (3). Serum TSH concen- Subsequently, the four locus-specific lead SNPs were taken trations in healthy individuals show considerable forward for replication in HUNT, CARLA and SHIP-TREND. inter-individual variation (4), and twin and family studies The significant associations of all four lead SNPs with TSH have indicated that 58–71% of this variation is genetically levels were confirmed by the three replication studies determined (5–7). Thus, it is plausible that inter-individual dif- (Table 1, Supplementary Material, Table S3). ferences in regulatory mechanisms of the pituitary–thyroid axis Finally, a combined meta-analysis including all the samples are responsible for a large proportion of the inter-individual dif- from the two discovery and the three replication cohorts was ferences in TSH concentrations. However, the causal genes carried out, encompassing a total number of 7463 individuals. have not been well established so far (5–7). Until now, only Downloaded from The P-values obtained for the lead SNPs rs2046045, two genetic loci associated with circulating TSH levels have rs10917477, rs10028213 and rs3813582 of the four addressed been identified and convincingly replicated. The lead single- loci 5q13.3 (PDE8B), 1p36 (CAPZB), 4q31 (NR3C2) and nucleotide polymorphisms (SNPs) of these two loci were 2 2 16q23 (LOC440389) were 2.79 × 10 27, 1.54 × 10 8, located in PDE8B encoding phosphodiesterase (PDE) 8B × 210 × 210 + 2.88 10 and 5.63 10 , respectively. Since PDE8B, (OMIM 603390) (8) and in CAPZB (9), which encodes the http://hmg.oxfordjournals.org/ CAPZB and NR3C2 showed the moderate-to-high heterogen- b-subunit of the barbed-end F-actin binding protein. These eity measured by the I2 quantity (11) in the combined two loci account for only 2.3 and 1.0–1.3% of TSH variance, meta-analysis, we additionally meta-analyzed the four lead respectively, and it is very likely that other genes are involved SNPs using a random-effects model to take into account the in the regulation of TSH levels. To investigate this further, we between-study heterogeneity. Although rs10917477 and performed a meta-analysis of two genome-wide association rs10028213 did not reach genome-wide significance using studies (GWASs) to identify additional genetic loci influencing this model, all P-values remained highly significant (Table 1), serum TSH levels in two population-based samples (KORA and whereas the two known TSH-associated loci PDE8B and SHIP) and replicated the results in three further independent CAPZB showed the highest quantity of heterogeneity. These cohort studies (HUNT, CARLA and SHIP-TREND). at GSF Zentralbibliothek on July 27, 2012 four quantitative trait loci together accounted for 3.3% of the variance in the TSH serum levels. RESULTS The discovery GWAS meta-analysis identified several genomic DISCUSSION regions associated with TSH levels (Fig. 1). Of the 2 524 918 SNPs tested by regression analysis, three distinct genetic loci The present meta-analysis confirmed that the common genetic reached genome-wide significance: 5q13.3 with the lead SNP variation in PDE8B was associated with TSH levels in the rs2046045 localized within PDE8B encoding PDE 8B general population as reported by a previous GWAS (8) and a (Fig. 2A), 1p36 with its lead SNP rs10917477 located the study including pregnant women (12). In our discovery upstream of CAPZB, which encodes the b-subunit of the meta-analysis, rs4704397 located within the first intron of barbed-end F-actin binding protein (Fig. 2B), and 4q31 where PDE8B exhibited a P-value of 1.28 × 1028. This result is in the lead SNP rs10028213 was identified the upstream of line with the study of Arnaud-Lopez et al. (8), which found NR3C2, which encodes the mineralocorticoid receptor the strongest significant associations between rs4704397 and (Fig. 2C). Furthermore, an additional locus exhibited associ- TSH levels. In a meta-analysis, Taylor et al.(13) also described ation P-values of ,1027 and thus showed the suggestive a pronounced association between the rs4704397 SNP in evidence of association: 16q23 with the lead SNP rs3813582 PDE8B and serum TSH levels. For the replication analyses, (Fig. 2D). The latter locus represents a former ‘gene desert’ rs2046045, also located within the first intron of PDE8B and 110-kb upstream of MAF. However, it was recently demon- in strong linkage disequilibrium (LD) with rs4704397 (r2 ¼ strated that it is actually directly located the downstream of a 0.94), was selected as a proxy SNP for the latter due to its