In this paper, we aimed to uncover biological mechanisms underlying the observed genetic associations between ADHD and obesity measures. Based on known and self-derived genetic correlation estimates for ADHD and BMI/obesity obtained from the world-wide largest data sets for each phenotype, we first applied a hypothesis-driven testing approach of two selected gene sets (DOPA and CIRCA), which showed that the dopaminergic neurotransmission system partially explains the genetic overlap between ADHD and BMI. Our data-driven, genome-wide approach subsequently showed that dopaminergic signaling, specifically Dopamine-DARPP32 Feedback in cAMP Signaling, was significantly enriched in both the ADHD–BMI and the ADHD–obesity gene-based meta-analysis results.
Both ADHD and obesity measures have been linked to disturbances in dopaminergic signaling. Alterations of the brain’s executive and reward circuits—modulated by mesocortical and mesolimbic dopamine, respectively—have been postulated as the basis of the deficient inhibitory control and impaired reward processing characteristics of ADHD . The ability to resist the impulse to eat desirable foods, and an appropriate reward response to those, also require proper functioning of these dopamine-regulated processes [23, 24]. For example, impulsive eating, as a result of a high arousal response to a potential reward and impaired inhibitory control, can lead to weight gain and obesity . Eating behavior is also dependent on the hypothalamic homeostatic system, which comprises hormonal regulators of energy balance—such as insulin, leptin, and gut hormones—and controls hunger, satiety, and adiposity . Increasing evidence suggests that such metabolic hormones also affect food-related sensitivity of the dopaminergic reward system , pointing to an overlap between the homeostatic and reward/reinforcement systems related to obesity .
Also confirming our hypothesis, the CIRCA gene set was associated with BMI, but the absence of a significant association with ADHD was unexpected. ADHD has previously been associated with altered circadian rhythmicity at molecular, endocrine, and behavior levels . Furthermore, zebrafish mutants for per1b, a key gene in circadian rhythm regulation, and Per1-knockout mice display hyperactive, impulsivity-like, and attention deficit-like behaviors . The lack of a significant association between ADHD and the CIRCA gene set in our study may be due to a true lack of effect of the circadian rhythm pathway on ADHD. However, given that some of the CIRCA genes are among the cross-disorder(/trait) overlapping genes, it is also possible that there is a true (unobserved) effect but that the gene set we assembled was not appropriate/informative enough to detect such association.
Going beyond candidate gene-set analyses, we conducted data-driven, genome-wide ADHD–BMI and ADHD–obesity gene-based meta-analyses. Cross-disorder(/trait) overlapping genes were carried forward into two follow-up approaches: one testing the association of (ADHD–BMI) overlapping genes with specific subcortical brain volumes previously linked to these phenotypes and the other aimed at identifying enriched biological pathways underlying the shared heritability. Both follow-up approaches again pointed to a role of the dopaminergic system. Through the first, we observed a significant association of ADHD–BMI overlapping genes with putamen volume in two independent samples (Table 4). This finding is of particular interest given the strong role of the dopaminergic system in this brain region and the prominent involvement of the putamen in inhibitory control functioning, one of the key neurobiological features suggested to be altered both in ADHD and obesity [18, 19]. The second follow-up approach showed several pathways significantly enriched in the ADHD–BMI and ADHD–obesity results. Dopamine signaling was at the heart of the pathway that was significantly enriched in both analyses, i.e., the Dopamine-DARPP32 Feedback in cAMP Signaling pathway. This postsynaptic pathway centers around the Dopamine- and cAMP-regulated neuronal phosphoprotein (DARPP-32; also known as Protein phosphatase 1 regulatory subunit 1B (PPP1R1B)), the phosphorylation state of which modulates dopaminergic neurotransmission (see Fig. 1 and description in Supplementary Material for details).
DARPP-32 is primarily expressed in postsynaptic dopaminergic neurons in the dorsal striatum (i.e., brain structure that includes, in addition to the caudate, the putamen; see results above for the association of ADHD–BMI overlapping genes and brain volumes), which is involved in certain executive functions, such as inhibitory control, and in the ventral striatum, which is the main brain region responsible for reward processing (https://gtexportal.org/home/gene/PPP1R1B). As described above, poor inhibitory control and altered reward processing, in the form of steeper delay discounting, are key neurobiological circuitries implicated in both ADHD and obesity [21, 23]. Further evidence linking dopamine DARPP-32 signaling, reward processing, and the brain comes from findings in animal models. Upon investigation of the consequences of frustrated expected reward of palatable food on gene expression in the mouse brain, Dopamine-DARPP32 Feedback in cAMP Signaling pathway was found to be enriched among differentially expressed genes, both the ventral striatum and in frontal cortex .
DARPP-32 modulates the effects of dopamine on cAMP/PKA-dependent gene transcription through transcription factors of the cyclic AMP-responsive element-binding (CREB) complex (Fig. 1), and CREB dysregulation has been linked to both ADHD  and obesity . Of note, the CREB Signaling in Neurons pathway was also significantly enriched in our ADHD–BMI gene-based meta-analysis, along with two other partially overlapping pathways involved in synaptic plasticity processes (namely, the Synaptic Long Term Depression and the Synaptic Long Term Potentiation pathways; Table 2), which are also closely related to dopamine DARPP-32 signaling.
Additional evidence for an involvement of DARPP-32 signaling to the ADHD–BMI/obesity overlap comes from the study of rare variants. The most common form of monogenic obesity is caused by mutations in the melanocortin 4 receptor (MC4R) gene , and MRC4 signaling is known to activate DARPP-32 . In addition to early-onset obesity, a higher prevalence of ADHD has been reported in MC4R mutation carriers . It has been hypothesized that such co-occurrence may be, in part, underpinned by reward processing deficits , and animal studies provide further support regarding the involvement of MC4R signaling and dopaminergic-dependent reward processing .
Our study has strengths and limitations. A clear strength is that we make use of the largest GWAS results available for each of the phenotypes being investigated. The sample sizes used to generate the (European ancestry) summary statistics used here were, in total, >53,000 for the iPSYCH-PGC ADHD GWAS, up to 700,000 for the GIANT-UK Biobank BMI GWAS, and almost 99,000 for the GIANT obesity GWAS. Obesity measures were, therefore, assessed both as a trait and a state. Although we performed the (categorical) obesity analysis using GWAS data from the obesity class with the largest sample size (obesity class I N = 32,858 cases, N = 65,839 controls; class II N = 9889 cases, N = 62,657 controls; class III N = 2896 cases, N = 47,468 controls; ), it is possible that the quantitative nature of BMI and the much larger sample size of the BMI GWAS provide more powerful analyses/results than with the obesity GWAS class I, which may account, at least in part, for some of the differences observed between the BMI and obesity results. All GWAS summary statistics used here are derived from individuals with European ancestry; the homogeneous background can be a strength given that genetic analyses can be sensible to population stratification, but we also would like to highlight the need of large studies on more diverse populations. Another strength is that we did not restrict our gene set assembly to single GO terms or KEGG pathways but applied a more inclusive approach regarding the processes involved. For dopaminergic neurotransmission, we thus assembled a gene set (DOPA) that was subsequently found to be significantly associated with ADHD and BMI. This contrasts with the approach adopted in the iPSYCH-PGC ADHD GWAS paper, which tested dopaminergic candidate genes and GO term pathways only individually, failing to detect significant associations with ADHD . The large difference in sample sizes between the phenotypes imposed some difficulties when analyzing them together. We minimized such limitations by carrying out gene-based cross-disorder(/trait) meta-analyses in MAGMA, which allows sample sizes to vary between and within samples and accounts for such variation by weighting the effects accordingly. We also opted for performing gene-based—rather than SNP-based—cross-disorder(/trait) meta-analyses. Apart from assuming that the (combined effect of SNPs within) genes represent entities closer to the biological mechanisms, this approach has a reduced statistical burden compared to SNP-based analyses and seems most suitable for these data given the difference in SNP density between the ADHD and the BMI and obesity GWASs (the latter ones being restricted to about 2.4 million SNPs present in HapMap 2). An additional advantage of using gene-based approach when meta-analyzing different phenotypes is that it does not rely on a priori expectations of concordance of the direction of effects, which avoids information on loci with discordant direction of effects from being lost. Another limitation we addressed was the presence of overlapping samples, since Welcome Trust participants had been included both in the iPSYCH-PGC ADHD GWAS and the GIANT BMI and obesity GWASs. The reduction in sample size reduced power of our analysis, but findings from the canonical pathway enrichment analysis remained stable. Finally, despite the undeniable genetic component of these complex disorders/traits, the current available sample sizes and techniques applied in genome-wide studies still only allow for a small proportion of the phenotypic variance to be accounted for by common variants genome-wide. However, we strongly believe that identifying the biological pathways shared between disorders represents a promising way forward to a better understanding of comorbidity, which goes far beyond the observed effect sizes of specific genes/pathways and their variance explained. Given the limitations stated above, our results should be interpreted with caution and considered as exploratory until more adequately powered samples and methods are available.
Overall, the findings of the present study identify dopaminergic neurotransmission as a key player underlying the shared heritability of ADHD and BMI/obesity, implicating mechanisms involving DARPP-32 signaling in particular and possibly involving neurobiological features related to putamen, such as inhibitory control. This is especially interesting since DARPP-32 has been directly implicated in the mechanism of action of ADHD medication , which has been suggested to attenuate the increased risk for obesity in people with ADHD . The fact that we observe a convergence between the results from hypothesis-driven and hypothesis-free approaches provides extra support to the robustness of our findings. Uncovering critical aspects of the shared etiology underlying the prevalent ADHD–obesity comorbidity may have important implications for clinical outcome, preventive interventions, and/or efficient treatment of these conditions.