Gpr88 knockout mice (or Gpr88-/- mice) are impaired in the Five-Choice Serial Reaction Time Task (5-CSRTT).
The 5-CSRTT (Suppl. Fig. S1) evaluates all three features of ADHD, namely hyperactivity, inattention and impulsivity37, and is considered most translatable between humans and rodents38. In this test, Gpr88-/- mice were able to learn the task normally (Suppl. Fig. S2). However, at testing (Fig. 2A and statistics in Suppl. Table S1), mutant mice displayed reduced accurate responding, which requires spatial and temporal attention, and this was observed across the four stimulus presentation times. Further, more omission errors were found at shorter intervals, indicating that attention deficits relate to task difficulty. Gpr88-/- mice also made more responses prior to signal presentation (premature responses) during the 5s-period preceding stimulus presentation, an index of higher motor impulsivity. Finally, Gpr88-/- mice showed increased beam breaks (activity), confirming hyperactivity in these mice. Otherwise, reward collection latency was higher or unchanged, suggesting intact motivation to retrieve the food. The absence of GPR88 in mice therefore decreases attention, increases motor impulsivity, and produces hyperactivity in the 5-CSRTT.
Gpr88 knockout mice are impaired in the Attentional Set Shifting task (ASST).
The ASST, based on human cognitive testing39, is also used in rodents40. The paradigm involves seven successive steps of increasing difficulty, where an odor and/or a medium are associated with a reward (Suppl. Fig. S3). Data for three parameters are shown in Fig. 2B and statistics are in Suppl. Table S1. In the first three steps of the procedure, mutant mice made more wrong trials in compound discrimination (CD) and reversal 1 (R1) steps, and showed higher average time to choose in the simple discrimination (SD) task, suggesting lower capability to maintain attention. Importantly, Gpr88-/- mice were slower in acquiring the last task (extra-dimensional shift, EDS) considered most meaningful, and this was observed for all the parameters (number of trials, wrong trials and latency to choice). Lack of GPR88 in mice therefore reduces attentional performance in the ASST.
Gpr88 knockout mice are impaired in the Go/No-Go task
This task41 was designed to detect deficient inhibitory processes, which characterize exacerbated impulsivity in children with ADHD, and was successfully transposed to rodents42 (Suppl. Fig. S4). In the conditioning phase (Suppl. Fig. S5), Gpr88-/- mice showed a learning pattern similar to Gpr88+/+ controls, suggesting that the motivational state was intact. In the Go phase (Fig. 3A and statistics in Suppl. Table S1), the two groups similarly used fewer nosepokes to earn more rewards between first and last trials, indicating normal learning of the task. However, Gpr88-/- mice showed significantly more premature responses, demonstrating lower ability to refrain from initiating an action. Despite higher premature responding, mutant mice eventually reached the criterion and moved to the next Go/No-Go step. In this phase success rates for the Go trials remained stable (Suppl. Fig. S6), however, Gpr88-/- mice showed higher number of premature responses, more commission errors (inability to refrain from inappropriate responding) and a higher impulsivity index (Fig. 3B and statistics in Suppl. Table S1). Thus, the absence of GPR88 in mice increases motor impulsivity in the Go/No-Go task.
We also treated Gpr88-/- mice and their controls with atomoxetine (Suppl. Fig. S7), a noradrenaline reuptake blocker used clinically to treat both adult and childhood ADHD43. In the saline groups, as expected, the percent of premature responses was increased in Gpr88-/- mice compared to Gpr88-/- controls. Atomoxetine (10 mg/kg) caused a drastic reduction in the percentage of premature responses in both genotypes, and the increased impulsivity detected in Gpr88-/- mice was totally abolished. The drug had no effect on accuracy indicating a preserved instrumental memory performance in atomoxetine treated mice. These data confirm predictive validity of the Go/No-Go task under our experimental conditions, and supports the notion that lack of GPR88 activity produces an ADHD-like impulsive phenotype.
A2A-Gpr88 and D1-Gpr88 knockout mice are impaired in waiting and stopping impulsivity, respectively
We previously showed dissociable roles of GPR88 at the level of MSN of the direct and indirect pathways for a number of behaviors11. To investigate whether GPR88 in D2R- and D1R-MSNs plays differential roles in the regulation of impulsive behaviors, we tested A2AR-Gpr88 (D2R pathway), and D1R-Gpr88 mice and their corresponding controls (A2AR-Ctl and D1R-Ctl) in the Go/No-Go task. A2AR-Gpr88, and D1R-Gpr88 mice did not differ from their controls in the conditioning phase (Suppl Fig. S8), suggestive of intact motivational state in the conditional mutants. In the Go phase, all the groups used fewer nosepokes to earn more rewards between first and last trials, indicating normal learning of the task (not shown). During testing however (Fig. 4A and Supp.l Table S1), A2AR-Gpr88 mice showed significantly more premature responses over all sessions, as well as between first and last sessions, indicating lower ability to wait during the pre-trial period. This was not the case for D1R-Gpr88 mice.
Despite higher premature responding, A2AR-Gpr88 mutant mice eventually reached the criterion of accuracy and moved to the next Go/No-Go step. In this phase (Fig. 4B-D and statistics in Suppl. Table S1), accuracy for the Go trials remained stable (not shown). However again, A2AR-Gpr88, but not D1R-Gpr88, mice showed higher number of premature responding (Fig. 4B), confirming lower inhibition control. Interestingly, the percentage of commission errors decreased normally along sessions in A2AR-Gpr88 mice, similar to their controls, but remained stable in D1R-Gpr88 mice. Thus, commission errors were higher than controls in several sessions, over the last 5 sessions and when comparing first and last session, indicating failure for D1R-Gpr88 mice to refrain their impulsive behavior. This was confirmed by the impulsivity index (Fig. 4D) that was higher across sessions for D1R-Gpr88 but not A2AR-Gpr88 mice.
Overall therefore, A2AR-Gpr88 mice were unable to withhold from starting the task (higher premature anticipatory responding, waiting impulsivity) while D1R-Gpr88 mice were unable to refrain from performing the task (higher commission errors, stopping impulsivity). Together these results clearly differentiate two distinct GPR88-mediated mechanisms regulating impulsivity.
The human GPR88 gene is associated with ADHD in children
Because ADHD-like behaviors were detectable in Gpr88-/- mice across three gold standard animal models, we next moved to humans and investigated whether genetic variants in the GPR88 gene show an association with ADHD dimensions in a clinical sample of children with ADHD (clinical characteristics in Methods).
We tested a selected panel of 6 single nucleotide polymorphisms (SNPs) for association with overall DSM-IV diagnosis, quantitative measures of behavior and cognition, and response to treatment with a fixed dose of methylphenidate (MPH) using family-based association tests (see Methods). The study included 567 nuclear families having one or more child with a DSM-IV diagnosis of ADHD44. Of the total number of children included in this family-based study, 78.1% were male; 52.9% were diagnosed with the combined subtype, while 38.1% and 9% had the inattentive and hyperactive subtypes of ADHD respectively. Among comorbid disorders, 41.9% had oppositional defiant disorder, 17.3% had conduct disorder, 43% had anxiety disorder (including phobias), and 7% had a mood disorder.
In the total sample, a significant but minor association was observed with one of the 6 SNPs in the selected panel (rs2036212, 5’gene16) with overall DSM-IV diagnosis and dimensions of behaviour, treatment response and cognition (Table 1 and statistics in Table S3).
More interestingly, however, when the total sample was stratified based on maternal stress during pregnancy, a highly significant association was observed with tag SNP rs2809817 (3’UTR16) in the group where mothers experienced moderate to severe stress during pregnancy. Here, a significant over-transmission of the T allele was observed with categorical DSM-IV diagnosis (Z=3.41, P=0.0006), suggesting that this allele carries risk for ADHD. Further association was observed in the quantitative FBAT analysis. The T allele showed association with: (a) total number of ADHD items (Z=2.74, P=0.006); (b) number of inattention (Z=2.67, P=0.008), hyperactivity (Z=2.73, P=0.006), and impulsivity (Z=2.70, P=0.007) items on DISC-IV; (c) Conners’-P (Z=2.25, P=0.02) and Conners’-T (Z=2.85, P=0.004) scores at baseline; (d) dimensional scores on the Child Behaviour Checklist (CBCL). This suggests that this allele is associated with a more severe symptom profile as assessed in the home (Conners’-Parents, CBCL, DISC), school (Conners’-T), and clinic (DSM-IV diagnosis).
In terms of cognitive function, association was observed with several dimensions, with the T allele over-transmitted to higher number of errors: (a) spatial working memory, planning and interference control, as measured by the SOPT; (b) commission errors on the CPT; (c) non-perserverative errors on the Wisconsin Card Sorting Test. It is interesting that this tag SNP shows association not only with clinical dimensions of ADHD, but also with Research Domain Criteria (RDoC) within the domain of cognitive function. Interestingly, significant association was also observed with response to treatment with MPH (Table), as noted in clinic.
It is also interesting to note that when family-based analysis was conducted in the group where the mothers experienced none/mild stress during pregnancy, the T allele of tag SNP rs2809817 showed significant association with: (a) lower number of inattention items on DISC-IV (under-transmission of T allele; Z= -1.99, P=0.05); (b) lower scores on specific dimensions on CBCL, particularly withdrawn behaviour (Z= -2.84, P=0.004); (c) better performance on the SOPT (Z= -2.37, P=0.02). These results suggest that not having the T allele at this locus may be protective on some dimensions of ADHD.