A study of the possible association between adenosine A2A receptor gene polymorphisms and attention-deficit hyperactivity disorder traits

Authors


Corresponding author: Y. Molero, Karolinska Institutet, MEB, Box 281, SE-17177 Stockholm, Sweden. E-mail: yasmina.molero.samuelson@ki.se

Abstract

The adenosine A2A receptor (ADORA2A) is linked to the dopamine neurotransmitter system and is also implicated in the regulation of alertness, suggesting a potential association with attention-deficit hyperactivity disorder (ADHD) traits. Furthermore, animal studies suggest that the ADORA2A may influence ADHD-like behavior. For that reason, the ADORA2A gene emerges as a promising candidate for studying the etiology of ADHD traits. The aim of this study was to examine the relationship between ADORA2A gene polymorphisms and ADHD traits in a large population-based sample. This study was based on the Child and Adolescent Twin Study in Sweden (CATSS), and included 1747 twins. Attention-deficit hyperactivity disorder traits were assessed through parental reports, and samples of DNA were collected. Associations between six single nucleotide polymorphisms (SNPs) and ADHD traits were examined, and results suggested a nominal association between ADHD traits and three of these SNPs: rs3761422, rs5751876 and rs35320474. For one of the SNPs, rs35320474, results remained significant after correction for multiple comparisons. These results indicate the possibility that the ADORA2A gene may be involved in ADHD traits. However, more studies replicating the present results are warranted before this association can be confirmed.

Adenosine is the breakdown product of adenosine triphosphate, and is always present in all extracellular fluids, conferring tonic activation on receptors with adenosine as natural ligand. Adenosine is a neuromodulator with diverse effects, e.g. on sleep regulation, pain, blood pressure, neuronal survival and psychomotor behavior (Fredholm et al. 2011). The adenosine A2A receptors (ADORA2A) are located mainly in the nucleus accumbens, the olfactory tubercle and the striatum, where they play an important role in regulating synaptic transmission of glutamate and dopamine (Cunha 2001; Svenningsson et al. 1999).

Adenosine A2A receptors are closely intertwined with the dopamine neurotransmitter system, and are colocalized with dopamine D2 receptors, with opposing effects intracellularly (Svenningsson et al. 1997). Adenosine A2A receptors are also suggested to moderate the signaling of the dopamine D2 receptor in the striatum (Canals et al. 2003). Research on the neurobiological mechanisms underlying attention-deficit hyperactivity disorder (ADHD) has to a great extent focused on dopamine receptor signaling (Faraone & Mick 2010), as drugs enhancing dopamine neurotransmission (i.e. methylphenidate and amphetamine) exert a symptom-reducing effect (Faraone et al. 2006). Support for a dopaminergic dysfunction in ADHD has also been reported in studies of mice (Chudasama & Robbins 2006), in neuroimaging studies (Arnsten 2006) and in genetic studies (Faraone et al. 2005). The influence of ADORA2A on dopamine may hence be regarded as indirect support for an involvement of this receptor in ADHD.

Moreover, the role of adenosine in sleep–wake homeostasis, with adenosine levels in the brain increasing with prolonged wakefulness and decreasing with sleep (Porkka-Heiskanen 1999), lends further support to the notion of an involvement of adenosine in ADHD (Pires et al. 2010), given the empirical support for an association between ADHD and sleep/alertness alterations (Cortese et al. 2006).

Studies on spontaneously hypertensive rats have been used as an animal model of ADHD as they display hyperactivity, impulsivity and attention problems (Takahashi et al. 2008). The blockade of ADORA2A in these animals confers improvement in learning impairments (Pires et al. 2009, 2010). In the same vein, genetic knockout of ADORA2A has been linked to modified psychomotor behavior in mice (Ledent et al. 1997). The available evidence from animal studies suggests that the gene encoding the ADORA2A may represent an important target for studies of the genetic influences on ADHD. While polymorphisms in this gene previously have been associated with several psychiatric conditions, including anxiety, depression, panic disorder, autism and schizophrenia (Cunha et al. 2008; Freitag et al. 2010; Hamilton et al. 2004; Hohoff et al. 2005), no studies of the association between ADHD and the ADORA2A gene have, to our knowledge, been carried out in humans. The aim of this study was thus to examine to what extent polymorphisms of the ADORA2A gene are associated with ADHD traits in a large population-based sample.

Methods

Subjects

This study is part of The Child and Adolescent Twin Study in Sweden (CATSS), an ongoing study since July 2004 that targets all twins born in Sweden since July 1992. Twin pairs were traced through the Swedish Twin Registry and parents were asked to participate in the study when the twins were 9 or 12 years old (Anckarsäter et al. 2011). The sample in this study originally included 1771 twins; however, 20 individuals were excluded because of brain damage and another 4 individuals were excluded because of large chromosome aberrations. The final sample included 1747 twins [50% (873) girls, and 50% (874) boys] with data on parent-reported psychiatric symptoms and genomic DNA. In this sample, 75.7% (1305) of the individuals were 9 years old, and 25.3% (442) of the individuals were 12 years old. Consenting parents were interviewed about the twins over the telephone using the Autism-Tics, ADHD and other Comorbidities inventory (A-TAC), an interview instrument designed to evaluate a wide range of psychiatric symptoms in children (Hansson et al. 2005; Larson et al. 2010). After the interview, twins were asked to give saliva samples for DNA extraction, and a kit was sent home to the twins for DNA collection with saliva (Oragene®, Genotek, Kanata, Canada). Zygosity was determined either by a panel of 46 single nucleotide polymorphisms (SNPs) derived for zygosity analyses (Hannelius et al. 2007) or by an algorithm based on five questions on twin similarity from the telephone interview. Of the twins, 41.7% (728) were monozygotic, and 58.3% (1019) were dizygotic. Permission for the study was granted by the Ethics Committee of Karolinska Institute (No. 02-289).

Measures

Attention-deficit hyperactivity disorder score was assessed with the A-TAC interview, and was defined to include 18 DSM-IV criteria of this disorder, added with one supplementary symptom characteristic for the disorder (boredom proneness). This definition has been validated in previous studies (Hansson et al. 2005; Larson et al. 2010). Response categories for each item were scored as 0 (no), 0.5 (yes, to some extent) and 1 (yes). In this study, we wanted to study ADHD traits in a population-based sample rather than clear-cut diagnoses; thus, cutoff values for ADHD scores were lower than cutoff values for clinical diagnoses. Cutoff values used in this study have, however, been identified in previous studies by the optimal inflection point on the receiver operating characteristic (ROC) curve yielding sensitivity and specificity >0.89 (Hansson et al. 2005; Larson et al. 2010).

Attention-deficit hyperactivity disorder traits were calculated as the sum of scores of both the inattention and hyperactivity–impulsivity scales, yielding between 0 and 19 points. A dichotomous variable was created with a cutoff value of 8 points. In addition, inattention and hyperactivity–impulsivity were also assessed separately. The scales ranged between 0 and 9 points and 0 and 10 points, respectively, with a dichotomous cutoff value of 6 points for each scale.

Genotyping

Genomic DNA was isolated from saliva samples collected from each subject and genotyping of SNPs was performed by KBioscience (http://www.kbioscience.co.uk) with the KASPar chemistry, which is a competitive allele-specific polymerase chain reaction SNP genotyping system using fluorescence resonance energy transfer (FRET) quencher cassette oligos (http://www.kbioscience.co.uk/reagents/KASP/KASP.html, KBioscience, Hoddesdon, UK, ). Six SNPs covering the long form of ADORA2A (25 kb, UCSC Genome Browser, Human Mar.2006 Assembly) were included in this study. Five SNPs were within the gene (from 5′ to 3′: rs2298383, rs3761422, rs2236624, rs5751876 and rs35320474) and one was in the ±10 kb flanking region (rs4822492). Of the participants, 23 displayed missing data on genotyping of rs2236624 and were thus excluded from the analyses of this SNP. For the same reason, 29 participants were excluded from the analyses of rs2298383, 27 participants were excluded from the analyses of rs3761422, 43 participants were excluded from the analyses of rs4822492, 25 participants were excluded from the analyses of rs5751876 and 39 participants were excluded from the analyses of rs35320474.

Statistical analyses

Statistical associations between SNPs and dichotomous measures of ADHD traits, inattention and hyperactivity–impulsivity, respectively, were estimated by using generalized linear mixed-effects models (GLMMs). The GLMM is an extension of the generalized linear model where random effects (in this case, zygosity) can be added to the model to account for the correlated data structure. Two different variance–covariance matrices were modeled: (1) for monozygotic twins and (2) for dizygotic twins. Correlations between individuals in these groups were calculated by using R-side random effects with an unstructured variance–covariance matrix. Dichotomous outcomes were assumed to be binary distributed with a logit link function. The parameters were estimated based on the residual log pseudo-likelihood (RSPL), which is equivalent to restricted maximum likelihood. Generalized linear mixed-effects models were performed with the PROC GLIMMIX procedure in the Statistical Analysis System (SAS), Version 9.1.3. All analyses were performed on the whole sample. Analyses were also run separately for boys and girls; however, no sex differences were found (data not shown).

As ADHD represents a quantitative trait in the population, sensitivity analyses were also conducted with ADHD, inattention and hyperactivity–impulsivity traits, respectively, as a continuous measure. In these analyses, GLMMs were performed only for SNPs nominally associated with dichotomous measures of ADHD, inattention and hyperactivity–impulsivity, respectively. The proportion of variance was also calculated using the coefficient of determination (R2).

Statistical significance was set at P < 0.05. Because of strong linkage disequilibrium between the observed SNPs, adjustment for multiple testing was obtained using permutation tests based on 10 000 permutations.

To further explore permutation-based significant associations, new GLMM analyses were performed for rs35320474 and ADHD traits and inattention, respectively, with dichotomized genotypes (TT vs. -T/--). Odds ratios (ORs) with 95% confidence intervals (CIs) are presented.

Results

Table 1 shows prevalence rates of each genotype for all six SNPs. All SNPs included in this study were in Hardy–Weinberg equilibrium.

Table 1. Genotype frequency (n = 1747)
 Genotype frequency
rs2236624T:T 8.6% (148)
 C:T 41.4% (713)
 C:C 50.1% (863)
rs2298383C:C 20.0% (343)
 C:T 48.4% (832)
 T:T 31.6% (543)
rs3761422T:T 15.8% (272)
 C:T 48.8% (840)
 C:C 35.3% (608)
rs4822492C:C 19.4% (331)
 C:G 48.7% (830)
 G:G 31.9% (543)
rs5751876T:T 16.4% (283)
 C:T 49.3% (849)
 C:C 34.3% (590)
rs35320474-:- 16.2% (276)
 -:T 49.2% (841)
 T:T 34.6% (591)

In Table 2, the percentages of individuals exhibiting ADHD traits above cutoff value within each genotype are presented, as well as mean scores of ADHD traits per genotype.

Table 2. Associations of genotypes with ADHD traits above cutoff value (n = 1747)
 % (n) of individuals with ADHD traits above cutoff value per genotypeMean score (SD) of ADHD symptoms per genotypeGLMM Fdf (unadj. P) (adj. P)aORb (CI) (unadj. P) (adj. P)a
  1. df, degrees of freedom; F, F-value; P, P-value.

  2. a

    Adjusted P-value based on 10 000 permutations.

  3. b

    Dichotomized genotypes.

  4. *P < 0.05; **P < 0.005.

rs2236624T:T 7.5% (11)T:T 2.20 (3.49)1.721,1206 (0.190) 
C:T 8.3% (59)C:T 2.27 (3.36)(0.342)
C:C 10.0% (86)C:C 2.52 (3.75) 
rs2298383C:C 7.0% (24)C:C 2.06 (3.25)2.961,1192 (0.086) 
C:T 9.0% (75)C:T 2.42 (3.58)(0.168)
T:T 10.7% (58)T:T 2.56 (3.76) 
rs3761422T:T 7.4% (20)T:T 2.09 (3.22)4.951,1180 (0.026)* 
C:T 7.6% (64)C:T 2.26 (3.40)(0.060)
C:C 11.5% (70)C:C 2.66 (3.87) 
rs4822492C:C 6.7% (22)C:C 2.03 (3.21)3.601,1167 (0.058) 
C:G 8.7% (72)C:G 2.39 (3.52)(0.121)
G:G 10.7% (58)G:G 2.56 (3.76) 
rs5751876T:T 7.1% (20)T:T 2.11 (3.19)5.121,1188 (0.024)* 
C:T 8.1% (69)C:T 2.28 (3.44)(0.055)
 C:C 11.5% (68)C:C 2.67 (3.90)  
rs35320474-:- 6.9% (19)-:- 2.10 (3.18)5.891,1178 (0.015)*1.67 (1.16–2.29)
-:T 7.7% (65)-:T 2.27 (3.40)(0.037)*(0.005)**
T:T 11.5% (68)T:T 2.65 (3.90) (0.015)*

The GLMM was performed to test for associations between genotypes and ADHD traits above cutoff value. Results from the GLMM showed that three SNPs, rs3761422, rs5751876 and rs35320474, were nominally associated with ADHD traits above cutoff value. After correction for multiple testing by 10 000 permutations, only the association for rs35320474 remained statistically significant. For rs35320474, carrying the T allele conveyed a higher risk (OR 1.67; CI 1.16–2.39) for displaying ADHD traits above cutoff value (Table 2).

Sensitivity analyses of the three SNPs nominally associated with ADHD traits above cutoff value were also carried out (Table 3). New GLMMs were conducted with a continuous measure of ADHD traits, and showed nominal associations for all three SNPs (P < 0.05). After correction for multiple testing by 10 000 permutations, none of the associations remained statistically significant.

Table 3. Sensitivity analyses of SNPs associated with ADHD traitsa (n = 1747)
 GLMM Fdf (unadj. P) (adj. P)b
  1. R-squared: rs3761422, R2 = 0.004; rs5751876, R2 = 0.003; rs35320474, R2 = 0.003.

  2. df, degrees of freedom; F, F-value; P, P-value.

  3. a

    Continuous measure of ADHD traits.

  4. b

    Adjusted P-value based on 10 000 permutations.

  5. *P < 0.05.

rs37614225.231,1232 (0.022)*
(0.052)
rs57518764.851,1234 (0.028)*
(0.063)
rs353204744.601,1228 (0.032)*
(0.070)

To further explore associations between SNPs and ADHD, additional examinations of the two ADHD subscales, inattention and hyperactivity–impulsivity, were carried out with GLMM (Table S1). SNPs rs3761422, rs5751876, and rs35320474 were nominally associated with both inattention and hyperactivity–impulsivity traits (P < 0.05). After correction for multiple testing by 10 000 permutations, only the association between rs35320474 and inattention traits remained statistically significant (P < 0.05). For rs35320474, again, carrying the T allele conveyed a higher risk (OR 1.81; CI 1.19–2.75) for displaying inattention symptoms above cutoff value (Table S1). However, after controlling for hyperactivity–impulsivity, this association was no longer statistically significant when corrected for multiple testing by 10 000 permutations. Furthermore, new GLMMs with a continuous measure of inattention and hyperactivity–impulsivity traits, respectively, were carried out (Table S2). Results showed nominal associations for all three SNPs and inattention traits (P < 0.05); however, after correction for multiple testing by 10 000 permutations, none of these associations remained statistically significant.

Discussion

In this study, we found nominal associations between three SNPs in the ADORA2A gene, rs5751876, rs3761422 and rs35320474, and ADHD traits. However, associations remained significant only for rs35320474 after adjustment for multiple testing. Results showed that individuals carrying the T allele in rs35320474 had higher odds of presenting ADHD traits above cutoff value. In an attempt to determine if this association was confined to one of the ADHD subtype traits only, we could only observe a significant association with the dichotomous measure of inattention traits after correction for multiple testing. However, this association was not significant after controlling for hyperactivity–impulsivity, suggesting that neither of the symptom dimensions is uniquely associated with rs35320474.

Prior studies have related rs35320474 to other psychiatric disorders, including autism, schizophrenia and anxiety (Freitag et al. 2010; Hohoff et al. 2005; Ottoni et al. 2005). It has been proposed that SNP rs35320474 may change the expression of the ADORA2A (Alsene et al. 2003; Childs et al. 2008), which could explain the associations found in this study. Altering the expression of the adenosine receptor expression could result in an altered tone in dopamine D2 receptor signaling, because of the antagonistic functions of the two receptors. However, the question whether the rs35320474 is functional in itself, or whether it is just in linkage disequilibrium with a truly functional SNP, remains to be resolved.

Our results suggested that rs5751876 and rs3761422 too appeared to be associated with ADHD traits, although these associations did not withstand correction for multiple testing. Interestingly, rs5751876, rs3761422 and rs35320474 have all previously been found to be associated with phenotypic variability in autism spectrum disorder (ASD) symptoms among individuals with diagnosed ASD (Freitag et al. 2010). As these three SNPs are in high linkage disequilibrium (R2 > 0.97), the lack of significant associations for rs5751876 and rs3761422 after correction for multiple testing could be due to insufficient statistical power in our sample. Furthermore, the notion that rs5751876 could be associated with ADHD is plausible as several studies have shown associations between this SNP and psychiatric conditions, including anxiety and panic disorders (Andreassi et al. 2011; Hamilton et al. 2004; Hohoff et al. 2005; Rogers et al. 2010). However, rs5751876 is a silent SNP that does not modify the coding sequence and therefore does not alter the function of the ADORA2A. Silent SNPs have nonetheless been suggested to exert influence on gene function through changing mRNA splicing and/or stability, or through protein folding/activity (Komar 2007). It is possible that such a mechanism could be involved in rs5751876, thereby affecting the risk of several psychiatric conditions.

Our findings on an association between the ADORA2A gene and ADHD traits are in line with data from animal models on ADHD. Taken together, our results and previous findings on animal models propose the possible participation of the adenosinergic system in ADHD traits, and the ADORA2A may thus represent an important target for pharmacological treatment of ADHD. This is in line with the proposal of adenosine receptor antagonist caffeine as an alternative pharmacological treatment of ADHD (Pires et al. 2010), which could be plausible as the arousing effects of caffeine are due to ADORA2A antagonism (Fredholm et al. 2011). Furthermore, it has been argued that the ADORA2A gene might account for individual variation in responses to stimulant drugs (i.e. methylphenidate and amphetamine) targeting the dopaminergic dysfunction that occurs in individuals with ADHD (Hohoff et al. 2005). In this vein, rs5751876 and rs35320474 polymorphisms have been shown to be associated with increased feelings of anxiety after intake of amphetamine or caffeine in prior studies (Hohoff et al. 2005; Rogers et al. 2010).

In genome-wide association studies of ADHD, no significant associations have been reported for the ADORA2A gene (Lesch et al. 2008; Neale et al. 2008, 2010; Zhou et al. 2008). This could be due to insufficient statistical power, as genome-wide association studies require very large samples to detect many genes, or to small effect sizes of this gene (Neale et al. 2010). However, even though the individual contribution of the ADORA2A gene is small, it could implicate a biological pathway of potential importance for pharmacological treatment.

This study had a number of strengths and limitations that should be taken into account when interpreting the results. Strengths included a large population-based cohort of both girls and boys with a high response rate (80%; Anckarsäter et al. 2011). Limitations included parental assessment of ADHD traits using the A-TAC interview, which is not equivalent to clinically validated diagnoses. However, A-TAC has been shown to be a reliable and valid method for identifying ADHD symptoms (Hansson et al. 2005; Larson et al. 2010). Furthermore, the use of a cohort of twins could tentatively affect the generalizability of results, although evidence points to little or no differences between twins and singletons (Evans & Martin 2000).

Concluding, to our knowledge, this study was the first to examine the relationship between the ADORA2A gene and ADHD traits in a human sample, thus extending previous knowledge based on animal models. Our results pointed to the possibility that the ADORA2A gene may be involved in ADHD. This notion is plausible as adenosine plays a role in ADHD-related neurobiological mechanisms such as dopamine neurotransmission and regulation of alertness. Furthermore, previous research has shown elevated occurrence of ADHD in individuals with chromosome 22q11 deletion (Jolin et al. 2012), suggesting that genes located in this region, including the ADORA2A located on chromosome 22q11.23, could conceivably exert an influence on the development of ADHD traits. However, these results should be considered as preliminary, as this was the first study to examine the associations between the ADORA2A gene and ADHD traits in a human sample. Replications of the present results are needed before suggesting that the ADORA2A gene is implicated in the etiology of ADHD traits.

Acknowledgments

Funding for this study was provided by Karolinska Institutet, Karolinska Institutet Center of Neurodevelopmental Disorders (KIND), the Swedish Department of Higher Education, the Swedish Council For Working Life and Social Research, the Swedish Research Council, Söderberg's foundation, Hållstens Foundation and the Swedish Brain Foundation. Dr Staffan Nilsson is gratefully acknowledged for valuable statistical advice, and Dr Patrik Magnusson is gratefully acknowledged for valuable advice on methodological aspects of genetic association studies. The authors declare no conflicts of interest.

Ancillary