Common obesity risk alleles in childhood attention-deficit/hyperactivity disorder†
How to Cite this Article: Albayrak, Z, Pütter C, Volckmar A-L, Cichon S, Hoffmann P, Nöthen MM, Jöckel K-H, Schreiber S, Wichmann H-E, Faraone SV, Neale BM, Herpertz-Dahlmann B, Lehmkuhl G, Sinzig J, Renner TJ, Romanos M, Warnke A, Lesch K-P, Reif A, Schimmelmann BG, Scherag A, Hebebrand J, Hinney A, Psychiatric GWAS Consortium: ADHD Subgroup. 2013. Common Obesity Risk Alleles in Childhood Attention-Deficit/Hyperactivity Disorder. Am J Med Genet Part B 162B:295–305.
Children with attention-deficit/hyperactivity disorder (ADHD) have a higher rate of obesity than children without ADHD. Obesity risk alleles may overlap with those relevant for ADHD. We examined whether risk alleles for an increased body mass index (BMI) are associated with ADHD and related quantitative traits (inattention and hyperactivity/impulsivity). We screened 32 obesity risk alleles of single nucleotide polymorphisms (SNPs) in a genome-wide association study (GWAS) for ADHD based on 495 patients and 1,300 population-based controls and performed in silico analyses of the SNPs in an ADHD meta-analysis comprising 2,064 trios, 896 independent cases, and 2,455 controls. In the German sample rs206936 in the NUDT3 gene (nudix; nucleoside diphosphate linked moiety X-type motif 3) was associated with ADHD risk (OR: 1.39; P = 3.4 × 10−4; Pcorr = 0.01). In the meta-analysis data we found rs6497416 in the intronic region of the GPRC5B gene (G protein-coupled receptor, family C, group 5, member B; P = 7.2 × 10−4; Pcorr = 0.02) as a risk allele for ADHD. GPRC5B belongs to the metabotropic glutamate receptor family, which has been implicated in the etiology of ADHD. In the German sample rs206936 (NUDT3) and rs10938397 in the glucosamine-6-phosphate deaminase 2 gene (GNPDA2) were associated with inattention, whereas markers in the mitogen-activated protein kinase 5 gene (MAP2K5) and in the cell adhesion molecule 2 gene (CADM2) were associated with hyperactivity. In the meta-analysis data, MAP2K5 was associated with inattention, GPRC5B with hyperactivity/impulsivity and inattention and CADM2 with hyperactivity/impulsivity. Our results justify further research on the elucidation of the common genetic background of ADHD and obesity. © 2013 Wiley Periodicals, Inc.
Twin, family, and adoption studies implicate a substantial genetic contribution to the variance of body mass index (BMI) [Maes et al., 1997; Hinney et al., 2010]. A meta-analysis of genome-wide association studies (GWAS) pertaining to BMI revealed 32 gene loci that are genome-wide significantly associated with increased BMI (α = 5 × 10−8) [Speliotes et al., 2010].
Attention-deficit/hyperactivity disorder (ADHD) is among the most common mental health conditions in childhood with prevalence rates at approximately 3–9% in the general population [(NICE), 2008]. According to the DSM-IV TR diagnostic criteria [APA, 2000] ADHD is characterized by symptoms of inattention, hyperactivity and impulsivity. Heritability has repeatedly been estimated within the range of 0.8 [for reviews see: Albayrak et al., 2008; Franke et al., 2009; Faraone and Mick, 2010; Freitag et al., 2010]. So far, five independent GWAS for ADHD [Mick et al., 2008; Neale et al., 2008; Lantieri et al., 2010; Neale et al., 2010a; Hinney et al., 2011] comprising 2,059 family trios and 1,391 independent cases and 3,755 controls and one GWAS meta-analysis based on 2,064 trios, 896 cases and 2,455 controls [Neale et al., 2010b] have not, however, revealed genome-wide significant findings (at α = 5 × 10−8).
A growing number of clinical and epidemiological studies suggest a link between ADHD and obesity [Cortese and Vincenzi, 2012]. ADHD was found to be significantly more prevalent in clinically ascertained obese children [Agranat-Meged et al., 2005], adolescents [Erermis et al., 2004] and adults [Altfas, 2002; Fleming et al., 2005]. Similarly, obesity is more prevalent in pediatric clinically [Spencer et al., 1996; Biederman et al., 2003; Holtkamp et al., 2004; Curtin et al., 2005; Faraone et al., 2005; Hubel et al., 2006; Spencer et al., 2006; Swanson et al., 2006; Ptacek et al., 2009] and epidemiologically [Anderson et al., 2006; Lam and Yang, 2007; Waring and Lapane, 2008; Chen et al., 2010; Fuemmeler et al., 2011; Erhart et al., 2012] ascertained subjects with ADHD, as well as in epidemiologically ascertained adult probands with ADHD [Pagoto et al., 2009; de Zwaan et al., 2011]. However, the evidence for the association of ADHD and obesity is not completely consistent [Mustillo et al., 2003; Braet et al., 2007; Dubnov-Raz et al., 2011]. Psychostimulants used for the treatment of ADHD commonly cause side effects with disordered eating signaling, such as appetite reduction and weight loss [Schertz et al., 1996]. In this respect, the comorbidity of ADHD and obesity might seem surprising, since the data is derived in part from patients receiving psychostimulants. The common pathophysiology of ADHD and obesity remains to be elucidated, although there is some evidence that hypoactivity in dopaminergic neurocircuits relevant for impulsivity, deficits in inhibitory control and reward deficiency link ADHD and obesity [Cortese et al., 2007; Pauli-Pott et al., 2010; Cortese and Vincenzi, 2012; Volkow et al., 2013].
Epidemiological studies give rise to the assumption that the clinically and epidemiologically observed co-morbidity of ADHD and obesity could be mediated by familial factors. Offspring exposed to maternal obesity and high fat diet consumption in utero as well as postnatally are more susceptible to develop mental health and behavioral disorders such as anxiety, depression, attention deficit hyperactivity disorder, and autism spectrum disorder [Van Lieshout et al., 2011; Sullivan et al., 2012]. Two epidemiological studies [Rodriguez et al., 2008; Rodriguez, 2010] suggest that pre-pregnancy overweight and obesity as well as weight gain during pregnancy significantly increases the risk of ADHD symptoms in the offspring. Latent class analysis adjusted for maternal smoking during pregnancy, maternal education, maternal age, gestational age, birth weight, infant sex, family structure at follow-up and cohort country revealed that children of mothers, who were both overweight and gained a large amount of weight during gestation, had a twofold risk of ADHD symptoms [Rodriguez et al., 2008].
In light of the possible patho-physiological link between ADHD and obesity, we hypothesize that genetic risk factors overlap between the two conditions. Thus, we assessed the association of the 32 loci robustly associated with increased BMI (α = 5 × 10−08) [Speliotes et al., 2010] in our ADHD GWAS [Hinney et al., 2011] and in the currently largest meta-analysis dataset available for childhood ADHD [Neale et al., 2010b]. Further, we explored the BMI single nucleotide polymorphisms (SNPs) for associations to the quantitative traits hyperactivity/impulsivity and inattention.
MATERIALS AND METHODS
The analysis is based on a GWAS for childhood ADHD [Hinney et al., 2011]. Briefly, ADHD cases comprised 495 children and adolescents with ADHD (age range 6–18 years) who were recruited in six psychiatric in- and outpatient units (Aachen, Cologne, Essen, Marburg, Regensburg, and Würzburg). Patients received a diagnosis of ADHD according to DSM-IV-TR [APA, 2000]. Ascertainment strategy and inclusion criteria have been described previously [see Hebebrand et al., 2006; Schimmelmann et al., 2007; Hinney et al., 2011]. For quantitative analyses, inattention- and hyperactivity/impulsivity scores are defined as number of fulfilled DSM-IV-TR ADHD criteria of the respective type (nine inattention criteria and nine hyperactivity/impulsivity criteria under no medication) [Taylor et al., 1986a, b; APA, 2000].
We used 1,300 controls of German ancestry who were drawn from three German population-based epidemiological studies: (a) the Heinz Nixdorf Recall (Risk Factors, Evaluation of Coronary Calcification, and Lifestyle) study (initial n = 383) [Schmermund et al., 2002], (b) PopGen (initial n = 490) [Krawczak et al., 2006], (c) KORA (initial n = 488) [Wichmann et al., 2005]. The recruitment areas were Essen, Bochum, and Mülheim (Ruhr area) for (a), Schleswig-Holstein (Northern Germany) for (b) and Augsburg (Southern Germany) for (c), respectively. Controls were not screened for ADHD, a small proportion of controls was obese.
Written informed consent was given by all participants and in case of minors by their parents. Minors gave their consent, in addition to their parents, if they were aged ≥12 years. Studies were approved by the respective institutional ethics committees and conducted in accordance with The Declaration of Helsinki.
The cases and controls were genotyped as delineated in Hinney et al. . In short, Controls were genotyped on HumanHap550v3 (Illumina, Inc., San Diego, CA) and the cases on Human660W-Quadv1 BeadArrays (Illumina). Genotyping was performed by (i) the Department of Genomics, Life & Brain Center, University of Bonn, Germany (all 495 ADHD cases and 383 Heinz Nixdorf Recall study controls) [Schmermund et al., 2002]; (ii) Illuminas customer service (all 490 PopGen controls) [Krawczak et al., 2006]; and (iii) the Helmholtz Zentrum Munich, Germany (all 488 KORA controls) [Wichmann et al., 2005].
We applied a quality control (QC) protocol to filter genotypes and individuals to the overlapping genotype content of all GWAS data sets. This QC protocol has been previously described in detail [Cichon et al., 2011]. Briefly, it accounts for call rates (CR), heterozygosity, cross-contamination, population stratification, relatedness, deviations from Hardy–Weinberg equilibrium (HWE) and minor allele frequencies (MAF). Within the GWAS we analyzed the 32 SNPs, whose alleles were recently described to be robustly associated with increased BMI [Speliotes et al., 2010] (see Table I). The controls were already used for various GWAS studies [Cichon et al., 2008, 2011; Hinney et al., 2011] so that the risk of incorrect genotypes is minimal; additionally our strict QC leads to the inclusion of high quality SNP data only.
Table I. Association to ADHD of 32 obesity risk alleles [Speliotes et al., 2010] in a German GWAS Study and in a large scale meta analysis [Neale et al., 2010b]
|6, rs206936, NUDT3||G||0.21||0.26||0.20||1.39||3.4 × 10−4||0.01||+ (imp)||−||0.17||1|| |
|3, rs9816226, ETV5||T||0.82||0.80||0.83||0.81||0.03||0.96||− (imp)||n.a.||n.a.||n.a.||n.a.|
|15, rs2241423, MAP2K5||G||0.78||0.75||0.78||0.84||0.04||1||−||−||0.57||1|| |
|18, rs571312, MC4R||A||0.24||0.26||0.24||1.17||0.07||1||+ (imp)||+||0.64||1|| |
|4, rs13107325, SLC39A8||T||0.07||0.08||0.06||1.29||0.08||1||+||n.a.||n.a.||n.a.||n.a.|
|11, rs3817334, MTCH2||T||0.41||0.43||0.40||1.14||0.09||1||+ (imp)||+||0.34||1|| |
|14, rs11847697, PRKD1||T||0.04||0.05||0.04||1.36||0.10||1||+ (imp)||−||0.03||0.96||rs1440983/0.744|
|4, rs10938397, GNPDA2||G||0.43||0.46||0.43||1.11||0.18||1||+ (imp)||n.a.||n.a.||n.a.||n.a.|
|2, rs887912, FANCL||T||0.29||0.28||0.26||1.11||0.20||1||+ (imp)||n.a.||n.a.||n.a.||n.a.|
|1, rs1514175, TNNI3K||A||0.43||0.43||0.41||1.08||0.34||1||n.a.||n.a.||0.94||1|| |
|16, rs12444979, GPRC5B||C||0.87||0.86||0.85||1.10||0.39||1||+ (imp)||+||7.2 × 10−4||0.02||rs6497416/1.000|
|11, rs4929949, RPL27A||C||0.52||0.51||0.53||0.93||0.42||1||n.a. (imp)||n.a.||0.38||1||rs4929942/0.966|
|5, rs4836133, ZNF608||A||0.48||0.48||0.49||0.95||0.49||1||n.a. (imp)||n.a.||n.a.||n.a.||n.a.|
|1, rs543874, SEC16B||G||0.19||0.19||0.18||1.06||0.51||1||n.a. (imp)||n.a.||0.86||1|| |
|16, rs1558902, FTO||A||0.42||0.42||0.41||1.05||0.56||1||n.a. (imp)||+||0.57||1||rs1421085/1.000|
|5, rs2112347, FLJ35779||T||0.63||0.64||0.63||1.05||0.56||1||n.a. (imp)||n.a.||n.a.||n.a.||n.a.|
|19, rs29941, KCTD15||G||0.67||0.68||0.67||1.04||0.59||1||n.a.||−||0.93||1|| |
|9, rs10968576, LRRN6C||G||0.31||0.30||0.29||1.04||0.63||1||n.a.||−||0.35||1|| |
|1, rs1555543, PTBP2||C||0.59||0.58||0.57||1.04||0.64||1||n.a. (imp)||−||4.5 × 10−2||1|| |
|2, rs713586, RBJ||C||0.47||0.46||0.48||0.97||0.66||1||n.a. (imp)||n.a.||0.68||1|| |
|11, rs10767664, BDNF||A||0.78||0.79||0.780||1.04||0.67||1||n.a. (imp)||−||0.59||1||rs7103411/1.000|
|3, rs13078807, CADM2||G||0.20||0.20||0.21||0.96||0.70||1||n.a.||+||0.04||1|| |
|16, rs7359397, SH2B1||T||0.40||0.40||0.40||0.97||0.74||1||n.a. (imp)||n.a.||0.64||1|| |
|6, rs987237, TFAP2B||G||0.18||0.18||0.18||1.03||0.74||1||n.a.||n.a.||0.65||1|| |
|12, rs7138803, FAIM2||A||0.38||0.40||0.40||1.03||0.75||1||n.a.||−||0.73||1|| |
|14, rs10150332, NRXN3||C||0.21||0.22||0.22||0.98||0.84||1||n.a. (imp)||n.a.||n.a.||n.a.||n.a.|
|1, rs2815752, NEGR1||A||0.61||0.61||0.61||0.99||0.87||1||n.a. (imp)||n.a.||0.82||1|| |
|2, rs2867125, TMEM18||C||0.83||0.82||0.82||0.99||0.90||1||n.a.||−||0.21||1|| |
|2, rs2890652, LRP1B||C||0.18||0.18||0.19||0.99||0.94||1||n.a. (imp)||−||0.32||1||rs1523702/0.702|
|19, rs3810291, TMEM160||A||0.67||0.71||0.71||1.00||0.98||1||n.a. (imp)||−||0.31||1||−|
In our data only 11 of the 32 BMI risk SNPs were directly genotyped, so that we decided to impute the data using MACH [Li et al., 2009, 2010] and the 1000 Genomes Project (release 2010-06) as reference data sets. The imputation quality scores ranged between 0.79 and 0.99 (on average 0.97). We used mach2dat and mach2qtl to analyze the imputed data [Li et al., 2009]. We report the odds ratios for the obesity risk alleles within our ADHD GWAS. The significance level α was set to 0.0016 (=0.05/32) applying a Bonferroni-correction for the multiple testing of 32 loci. We used QUANTO (http://hydra.usc.edu/GxE) for power considerations. The power was calculated for a level α = 0.0016 (two-sided) under a (log-) additive inheritance mode for MAF between 5% and 90% and an assumed average effect size (per risk allele odds ratio) of 1.3. For a sample size of 495 ADHD cases and 1,300 controls, the power ranged between 6% and 63%. Thus, our study was sufficiently powered to detect larger genetic effects for more frequent variants only.
Additionally, we performed in silico analyses of imputed GWAS data from the currently largest meta-analysis of the international ADHD GWAS data available in the Psychiatric GWAS Consortium (2,064 trios, 896 independent cases, and 2,455 controls) [Neale et al., 2010b]. For details on this meta-analysis we refer to the original publication. Finally, we also explored the obesity risk alleles with regard to their association to the quantitative inattentive and hyperactivity/impulsivity scores and report beta estimates and the respective nominal two-sided P-values. For the meta-analysis data, nominal P-values are given for the association of the obesity risk alleles (or of SNPs in strong linkage disequilibrium (LD) with the originally reported SNP, which were called “proxies”) with inattention and hyperactivity/impulsivity.
32 BMI SNPs and Childhood ADHD Risk
We detected association with an increased risk for ADHD for the obesity risk allele G at SNP rs206936 (P = 3.4 × 10−4, adjusted for multiple-testing P = 0.01) with an odds ratio (OR) of 1.39 per G risk allele. SNP rs206936 is located in intron 1 of the NUDT3 gene on chromosome 6p21.31 (see Table I). We checked LD of rs206936 with flanking GWAS SNPs. Only for SNPs within NUDT3 and the ribosomal protein S10 gene (RPS10) pairwise r2 were above 0.8. Hence, this association signal is likely to point to (one of) these two genes. Results of the association analyses for all 32 BMI SNPs [Speliotes et al., 2010] and childhood ADHD are displayed in Table I.
Three further BMI SNPs resulted in directionally consistent, uncorrected P-values below 0.10. The SNPs are located (1) approximately 200 kb downstream of the melanocortin 4 receptor gene (MC4R), (2) 16 kb upstream of the gene solute carrier family 39 (zinc transporter), member 8 (SLC39A8), and (3) in the intronic region of mitochondrial carrier 2 (MTCH2; Table I).
For 10 out of the 32 SNPs implicated in obesity [Speliotes et al., 2010] it was possible to compare directions of obesity effect allele and (potential) ADHD risk allele; for 8/10 the direction was consistent (one-sided binomial sign test P = 0.01). The remaining 22 SNPs were not assessed, because minor risk allele frequencies were ∼50% or obesity allele effect for ADHD was <1.10. If we assume that for the remaining 22, half would have the same direction (i.e., 19 of the total 32 in total) the binomial test would change to P = 0.11. Hence, the directional effect would need to be confirmed in an independent sample to assess if our initial finding is spurious.
Psychiatric GWAS Consortium; ADHD subgroup meta-analysis data
For 24 out of the 32 BMI SNPs [Speliotes et al., 2010] data was available in the large meta-analysis [Neale et al., 2010b] (Table I). Directionally consistent effects with Speliotes et al.  [Speliotes et al., 2010] were estimated for 6 out of 24 SNPs, which is less than expected by chance (see Table I). Of these six, one SNP was significantly associated with ADHD risk upon correction for multiple testing (rs6497416 in GPRC5B: G protein-coupled receptor, family C, group 5, member B; P = 7.2 × 10−4; Pcorr = 0.02).
32 BMI SNPs and Inattention and Hyperactivity/Impulsivity
Five BMI SNP risk alleles were associated with the quantitative trait inattention with uncorrected P-values below 0.05 (Table II). Among these five BMI SNPs were SNP alleles at the NUDT3, GNPDA2, and RBJ loci, for which we observed consistent directions of obesity effect allele and (potential) ADHD risk allele for inattention.
Table II. Effect Estimates of 32 SNPs With Risk Alleles for Increased BMI [Speliotes et al., 2010] on Inattention and Hyperactivity in Our German “ADHD GWAS Sample” [Hinney et al., 2011] and in the Meta-Analysis Data [Neale et al., 2010b]
|6, rs206936, NUDT3||G||3.4 × 10−4||+||0.25||0.02||−0.17||0.28||0.09||0.70||0.54|| |
|3, rs9816226, ETV5||T||0.03||−||0.01||0.96||0.08||0.63||n.a.||n.a.||n.a.|| |
|15, rs2241423, MAP2K5||G||0.04||−||>−0.01||0.99||0.12||0.47||0.19||2.1 × 10−2||0.37|| |
|18, rs571312, MC4R||A||0.07||+||0.10||0.38||−0.19||0.23||n.a.||n.a.||n.a.|| |
|4, rs13107325, SLC39A8||T||0.08||+||0.23||0.19||0.27||0.27||n.a.||n.a.||n.a.|| |
|11, rs3817334, MTCH2||T||0.09||+||−0.09||0.35||−0.18||0.19||0.47||0.64||0.80|| |
|14, rs11847697, PRKD1||T||0.10||+||−0.15||0.50||−0.443||0.15||n.a.||n.a.||n.a.||rs1440983/0.744|
|4, rs10938397, GNPDA2||G||0.18||+||0.20||0.04||0.07||0.63||n.a.||n.a.||n.a.|| |
|2, rs887912, FANCL||T||0.20||+||0.10||0.36||−0.01||0.97||n.a.||n.a.||n.a.|| |
|1, rs1514175, TNNI3K||A||0.34||n.a.||0.15||0.13||−0.04||0.78||0.68||0.10||0.12|| |
|16, rs12444979, GPRC5B||C||0.39||+||−0.17||0.23||0.17||0.39||0.17||0.08||0.04||rs6497416/1.000|
|11, rs4929949, RPL27A||C||0.42||n.a.||0.07||0.50||0.14||0.30||n.a.||n.a.||n.a.||rs4929942/0.966|
|5, rs4836133, ZNF608||A||0.49||n.a.||0.04||0.66||0.09||0.50||n.a.||n.a.||n.a.|| |
|1, rs543874, SEC16B||G||0.51||n.a.||−0.07||0.58||−0.02||0.92||n.a.||n.a.||n.a.|| |
|16, rs1558902, FTO||A||0.56||n.a.||>−0.01||0.99||−0.01||0.97||n.a.||n.a.||n.a.||rs1421085/1.000|
|5, rs2112347, FLJ35779||T||0.56||n.a.||−0.14||0.16||−0.09||0.52||n.a.||n.a.||n.a.|| |
|19, rs29941, KCTD15||G||0.59||n.a.||0.07||0.50||0.07||0.62||n.a.||n.a.||n.a.|| |
|9, rs10968576, LRRN6C||G||0.63||n.a.||>0.01||1.00||>0.01||1.00||0.74||0.98||0.66|| |
|1, rs1555543, PTBP2||C||0.64||n.a.||0.06||0.55||0.16||0.24||n.a.||n.a.||n.a.|| |
|19, rs2287019, QPCTL||C||0.65||n.a.||0.08||0.51||−0.01||0.96||n.a.||n.a.||n.a.|| |
|2, rs713586, RBJ||C||0.66||n.a.||0.20||0.03||0.10||0.44||0.09||0.41||0.73|| |
|11, rs10767664, BDNF||A||0.67||n.a.||−0.21||0.07||−0.32||0.05||n.a.||n.a.||n.a.||rs7103411/1.000|
|3, rs13078807, CADM2||G||0.70||n.a.||0.08||0.55||−0.09||0.63||0.92||0.04||0.12|| |
|16, rs7359397, SH2B1||T||0.74||n.a.||0.17||0.09||−0.01||0.94||n.a.||n.a.||n.a.|| |
|6, rs987237, TFAP2B||G||0.74||n.a.||−0.27||0.04||−0.13||0.46||n.a.||n.a.||n.a.|| |
|12, rs7138803, FAIM2||A||0.75||n.a.||−0.05||0.60||−0.05||0.71||0.33||0.53||0.26|| |
|13, rs4771122, MTIF3||G||0.78||n.a.||−0.25||0.04||−0.01||0.97||0.31||0.28||0.19||rs1006353/0.740|
|14, rs10150332, NRXN3||C||0.84||n.a.||−0.03||0.83||−0.33||0.05||n.a.||n.a.||n.a.|| |
|1, rs2815752, NEGR1||A||0.87||n.a.||−0.08||0.39||0.07||0.60||n.a.||n.a.||n.a.|| |
|2, rs2867125, TMEM18||C||0.90||n.a.||0.02||0.86||0.21||0.25||n.a.||n.a.||n.a.|| |
|2, rs2890652, LRP1B||C||0.94||n.a.||−0.05||0.72||−0.24||0.17||n.a.||n.a.||n.a.||rs1523702/0.702|
|19, rs3810291, TMEM160||A||0.98||n.a.||>−0.01||1.00||−0.09||0.57||n.a.||n.a.||n.a.|| |
Psychiatric GWAS Consortium; ADHD subgroup meta-analysis data
Ten out of 23 SNPs retrieved in silico were available for the analysis of a possible association with inattention and hyperactivity/impulsivity. Among these, we observed four with a directionally consistent effect to Speliotes et al. . Three of these had a nominal P-value below 0.05 (Table II). These were loci at MAP2K5 for inattention, GPRC5B for the combination of hyperactivity/impulsivity and inattention and CADM2 for hyperactivity/impulsivity.
We identified directionally consistent association of two obesity risk alleles with ADHD: (1) rs206936 in our German sample comprising 495 young patients and 1,300 controls [Hinney et al., 2011] and (2) rs6497416 in the Psychiatric GWAS Consortium ADHD meta-analysis sample comprising 2,064 trios, 896 independent ADHD cases, and 2,455 controls [Neale et al., 2010b]. These findings suggest an overlap in the polygenic predisposition between obesity and ADHD. At the same time, our study documents that most of the investigated 32 obesity risk alleles are seemingly not involved in ADHD.
- (1)SNP rs206936 is located in intron 1 of NUDT3. NUDT3 is a member of the Nudix protein family. The function of this protein family is to clean the cell of potentially deleterious endogenous metabolites that promote transversions of nucleotides. Thus, an active NUDT3 prevents somatic gene mutations [Bessman et al., 1996; Safrany et al., 1998]. The role of NUDT3 is currently neither clear for obesity nor for ADHD. We also had a look at RPS10 (ribosomal protein S10), because variation in this gene and in NUDT3 forms a haplotype block. RPS10 belongs to the S10E family of 40S subunit ribosomal proteins and contains 6 exons [Boria et al., 2010]. RPS10 is necessary for the production of 18S ribosomal subunit protein [Doherty et al., 2010]. To our knowledge, there is no evidence available indicating that RPS10 is involved in obesity or ADHD.The NUDT3 allele had been shown to have a very small effect on BMI: According to Speliotes et al.  this allele explains 0.01% of the BMI variance in adults. Because this same allele was picked up in our comparatively small sized German ADHD sample, we speculate that the NUDT3 variant might be more relevant for ADHD than for obesity. However, we were not able to confirm the association to ADHD in the in silico analysis of the large ADHD meta-analysis comprising 2,064 trios, 896 cases, and 2,455 controls [Neale et al., 2010b].
- (2)SNP rs6497416 is located in an intronic region of GPRC5B, that belongs to the superfamily of G protein-coupled receptors (GPCR) [Brauner-Osborne and Krogsgaard-Larsen, 2000] and shows homology to the family C receptors, which includes eight metabotropic glutamate receptors. GPRC5B was found to display 20–25% protein sequence identity in the 7 transmembrane regions with metabotropic glutamate receptors 1–3 (GRM1, GRM2, GRM3) [Brauner-Osborne and Krogsgaard-Larsen, 2000]. Rare copy number variations (CNV) in different subtypes of metabotropic glutamate receptors (e.g., GRM5, GRM7) are associated with ADHD [Elia et al., 2010, 2012].
Two BMI SNPs with nominal statistical significance for an association with ADHD did not reveal directionally consistent effects: rs9816226 (ETV5) and rs2241423 (MAP2K5). Although we cannot rule out that a risk allele for obesity is at the same time a protective allele for ADHD this explanation seems rather unlikely.
We hypothesized, that the association of BMI SNPs with ADHD could well be mediated via the ADHD clinical endophenotypes inattention and hyperactivity/impulsivity [Fuemmeler et al., 2011]. Our post hoc analyses in the German sample revealed that the SNP with the lowest P-value in the initial analysis (rs206936 in NUDT3) was primarily associated with inattention, and not with hyperactivity/impulsivity (Table II). In addition to NUDT3, we also observed that alleles of rs10938397 in GNPDA2, rs713586 in RBJ and rs4771122 in MTIF3 were nominally associated with inattention, too. However, there is no prior evidence suggesting that either of these four genes or variants therein are related to attention processes. These results could not be confirmed in our in silico analysis of the meta-analysis dataset [Neale et al., 2010b]. Instead, we found variants in MAP2K5 associated to inattention, in GPRC5B to the combination of hyperactivity/impulsivity and inattention and in CADM2 to hyperactivity.
There is some evidence, currently relying mainly on case reports, that suggests a common neurobiological link between ADHD and obesity [see Cortese and Vincenzi, 2012]. The brain-derived neurotrophic factor (BDNF) has been implicated to mediate both phenotypes. Rodent models of reduced or deleted bdnf expression result in an obese, hyperphagic, and hyperactive phenotype [Kernie et al., 2000; Rios et al., 2001; Xu et al., 2003]. In humans, a girl with a de novo chromosomal inversion encompassing the BDNF gene was also reported to be both severely obese and hyperactive [Gray et al., 2006]. Recently, a case study demonstrated a genotype–phenotype relation in three pediatric patients with obesity, ADHD and mild to borderline cognitive impairment with an overlapping segment of micro-deletions spanning 2.3 MB at chromosome 11p14.1. The locus covers the BDNF gene [Shinawi et al., 2011]. However, evidence for association of single BDNF alleles with ADHD is conflicting [Friedel et al., 2005]. Our results from the meta-analysis do not suggest a significant role for BDNF in ADHD [Neale et al., 2010b]. Both, a family and two case studies suggest a link between a reduced melanocortinergic tone, due to melanocortin 4 receptor (MC4R) gene mutations and ADHD [Agranat-Meged et al., 2008; Albayrak et al., 2011; Pott et al., 2013] in addition to the well-known risk for development of obesity. It is of interest to note, that the MC4R risk allele was nominally associated with a P-value of 0.07 with ADHD risk in the German sample (Table I).
The lack of confirmation of the results obtained in our own ADHD GWAS sample and those of the meta-analysis is of course a clear limitation of our findings. On the other hand heterogeneity between the two samples might have contributed to difficulties in confirmation. Facing both false positives and false negatives it should be up to the reader to carefully weigh the evidence.
We analyzed if obesity risk alleles are associated with ADHD risk. Hence, we only corrected for the number of tests (n = 32) in this hypothesis driven approach. We are aware of the fact that there are no ADHD risk alleles that reach genome wide significance in ADHD GWAS [e.g., Neale et al., 2010b; Hinney et al., 2011]. Thus, the P-values we have obtained are not genome wide significant indeed much larger (hence, they are not below or equal to 5 × 10−8).
It has repeatedly been shown, that GWAS proved to be useful in identifying association signals that account for an increased risk in two or three conditions, that were previously not suspected to share common pathways [see Manolio, 2010], like for example variants near the leucine-rich repeat kinase 2 gene (LRRK2) for Parkinson's disease [Paisan-Ruiz et al., 2004] and Crohn's disease [Barrett et al., 2008]. Given the methodological strength of GWAS based cross-disorder analyses in revealing shared genetic factors between different phenotypes, our results justify future scientific analyses of the possible genetic relationship between single genetic risk factors for both obesity and ADHD. Our finding in the in silico analysis of the ADHD meta-analysis data set highlights the role of the glutamate receptor, metabotropic homologue GPRC5B in potentially both ADHD and obesity.
The authors express their gratitude to the children and their families for participation. Financial disclosure: Prof. Herpertz-Dahlmann receives industry research funding from Vifor Pharma, Switzerland. Prof. Lehmkuhl receives industry research funding from Lilly Deutschland GmbH, Germany. The other authors declare they have no conflict of interest relevant to this work. We thank the following sources for funding or research: the German Ministry for Education and Research (National Genome Research Net plus 01GS0820 and 01KU0903), the German Research Foundation (DFG; HE 1446/9-1, KFO 125/1-1, SCHA 542/10-2, ME 1923/5-1, ME 1923/5-3) and the European Community's Seventh Framework Programme (FP7/2007-2013) under grant agreement no. 245009. The START-Program EK 119/05 of the Medical Faculty, RWTH Aachen, Germany. The ADHD meta-analysis was supported by the following grants: US National of Institute of Health Grants R13MH059126, R01MH62873 and R01MH081803 to S.V. Faraone and K23MH066275 to J. Elia, The University of Pennsylvania CTR grant UL1-RR-024134 to J. Elia and H. Hakonarson and Institutional Development Award to the Center for Applied Genomics from the Children's Hospital of Philadelphia to H. Hakonarson; Affymetrix Power Award, 2007 to B. Franke; NHMRC (Australia) and Sidney Sax Public Health Fellowship (443036) to S.E. Medland; Wellcome Trust, UK for sample collection to L Kent. MH58277 to S. Smalley. UMC Utrecht Genvlag Grant, an Internal Grant of Radboud University Nijmegen Medical Centre to J. Buitelaar. These sources had no further role in study design, in the collection, analysis and interpretation of data, in the writing of the report, and in the decision to submit the paper for publication.