DRD2 and PPP1R1B (DARPP-32) polymorphisms independently confer increased risk for autism spectrum disorders and additively predict affected status in male-only affected sib-pair families

Subjects

The 112 affected sib-pair families and the comparison group (N = 443) were previously described[5]. Briefly, the study group included 28 families from Canada[18], 5 from the South Carolina Autism Project[39], and 79 families obtained through the Autism Genetic Resource Exchange (AGRE) in the United States[40]. All families have two or more children with either autism or an ASD, including Asperger syndrome and pervasive developmental disorder (PDD) variants. This study was approved by the Queen’s University Research Ethics Board and written informed consent was obtained from parents of all participating families from Canada and South Carolina, and through AGRE[40].

All 443 samples from the comparison group (bloodspots from anonymous newborns taken for the purpose of PKU testing and made available from the Ontario Ministry of Health) were available for the study of the PPP1R1B locus. Two hundred and fifty-three comparison group samples were used for the DRD2 gene studies. There were no significant differences in the allele frequencies for DRD2 (P = 0.57–0.93) and PPP1R1B (P = 0.28–0.91) markers in males and females from the respective comparison cohorts and thus our comparison cohorts included both males and females. Although comprehensive information regarding psychiatric and behavioural disorders is not available for the comparison group, we do not expect the prevalence of ASDs in this comparison group to be greater than that in the general population, or approximately 1/110[41].

Marker amplification and genotyping

DRD2

Four polymorphisms (rs1799732 Ins/Del, rs1079597 G/A, rs1800498 T/C and rs1800497 C/T) were studied at the DRD2 locus (Figure1). The DRD2 gene contains four haplotype blocks [Haploview 4.1; available athttp://hapmap.ncbi.nlm.nih.gov][42]. Three of the blocks are small (<4 kb) and one block is 20 kb in size. One of four polymorphisms used in this study, rs1079597, is part of the HapMap dataset. both this SNP and and rs1800498 are located within the larger haplotype block and were examined in previous studies of other neuropsychiatric disorders. Primer sequences, PCR and digestion conditions are shown in Table1. All PCR reactions were carried out using 5 ng of template DNA; digestion products were separated on either 2% (rs1079597, rs1800498 and rs1800497) or 2.5% (rs1799732) agarose gels and visualized using ethidium bromide and UV illumination. In order to minimize genotyping errors, DNAs from affected individuals and family members were randomly arranged on 96-well plates, and all results were independently scored and tabulated by two persons.

DRD2 Marker (5′ → 3′)1	Primers	[MgCl2]	Annealing temperature	# of cycles	Restriction enzyme (U)2
rs1799732 Ins/Del	F 5′-GAGAAGACTGGCGAGCAGAC-3′	1.5 mM	63°C	35	BstNI (0.05)
	R 5′-CCACCAAAGGAGCTGTACCT-3′
rs1079597 G/A3	F 5′-GATACCCACTTCAGGAAGTC-3′	1.0 mM	55°C	34	TaqI (0.4)
	R 5′-CAGTAAAGAACTAGGAGTCAG-3′
rs1800498 T/C3	F 5′-CCCAGCAGGGAGAGGGAGTA-3′	1.0 mM	55°C	34	TaqI (0.4)
	R 5′-GACAAGTACTTGGTAAGCATG-3′
rs1800497 C/T3	F 5′-CCGTCGACGGCTGGCCAAGTTGTCTA-3′	1.0 mM	58°C	34	TaqI (0.4)
	R 5′-CCGTCGACCCTTCCTGAGTGTCATCA-3′
PPP1R1B Marker (5′ → 3′)	Primers	[MgCl 2 ]	Annealing temperature	# of cycles	Restriction enzyme (U) 2
rs1495099 G/C	F 5′-TTGTTGCTGAGCTGAGATGC-3′	1.0 mM	60°C	35	PvuII (0.3)
	R 5′-CTCCAGGGAAATGCACAAAG-3′
rs907094 T/C	F 5′-ACCTGATTGGGAGAGGGACT-3′	1.0 mM	60°C	34	MseI (0.3)
	R 5′-GTAAGCTGAGGGGCCTGTG-3′
rs3764352 A/G	F 5′-CTGTTTTGGAGGGGTCTCAG-3′	1.0 mM	60°C	35	BccI (0.3)
	R 5′-TGGGAATACTGAAGAGTCAACC-3′

¹Rs1799732 Ins/Del, rs1079597 G/A, rs1800498 T/C, and rs1800497 C/T previously known as −141 C Ins/Del, TaqI B, TaqI D, and TaqI A, respectively.

²Restriction enzymes obtained from New England Biolabs, Pickering, ON, Canada

³Primer sequences have been reported previously[43].

PPP1R1B

Three polymorphisms (rs1495099 G/C, rs907094 T/C and rs3764352 A/G) were examined in the PPP1R1B gene (Figure2). These variants were chosen from the NCBI dbSNP Build 121 database from the Human Genome Project (available athttp://www.ncbi.nlm.nih.gov/SNP/snp_summary.cgi) based on the following criteria: the markers span the PPP1R1B locus, they have minor allele frequencies (MAFs) of approximately 20%, and alleles at rs907094 and rs3764352 are associated with changes in DARPP-32 mRNA expression and measures of cognitive performance[32]. The PPP1R1B locus has a single haplotype block which includes rs907094 and rs3764352 as haplotype-tagged SNPs (htSNPs). PCR reactions were carried out using 5 ng of template DNA and amplicons were digested using conditions shown in Table1. All digestion products were separated on 2% agarose gels and visualized using ethidium bromide and UV illumination. In order to minimize genotyping errors, DNAs from affected individuals and family members were randomly arranged on 96-well plates, and all results were independently scored and tabulated by two persons.

Statistical analyses

Prior to carrying out analyses, Mendelian errors were checked in the family cohort using the FBAT program, v1.5.5[44]. DRD2 marker data on five families and PPP1R1B marker data on two families were excluded from the analyses due to identified Mendelian errors. Single gene analyses were performed as previously described[5]. To avoid allele and genotype frequency distortion from using related individuals in case–control comparisons, one affected individual was selected at random from each family using SPSS v14.0 (SPSS, Chicago, IL) with the same cohort of randomly chosen individuals used for single marker allele and genotype frequency comparisons for all polymorphisms as well as general discriminant analyses. Family-based association tests (FBAT), including quantitative disequilibrium tests (QTDT), were done using FBAT v1.5.5. Because FBAT v1.5.5 can accommodate multiple affected individuals from each family, all affected individuals including those used for case–control comparisons, were included for FBAT and QTDT analyses. The domains, ‘reciprocal social interaction’, ‘communication’ and ‘repetitive stereotyped behaviours’ used for QTDT analyses were derived from the total scores from the ‘Qualitative Abnormalities in Reciprocal Social Interaction’ (A1 to A4), ‘Qualitative Abnormalities in Communication’ (B1, B2(V), B3(V) and B4) and ‘Restricted, Repetitive, and Stereotyped Patterns of Behaviour’ (C1 to C4) subdomains in the ADI-R diagnostic algorithm[45].

General Discriminant Analysis (GDA) was used to evaluate the predictive value of our single gene findings in discriminating between individuals with and without autism, as well as to test for evidence of interaction effects between genes. Genotypes were coded as categorical variables and GDA was performed using Statistica 9.1 [Statsoft, Tulsa, OK, USA].

Corrections for multiple comparisons

The contribution of a single gene to autism susceptibility is predicted to be relatively small and thus difficult to detect statistically. Although corrections for multiple testing must be made in genetic association studies, Bonferroni correction is thought to be too stringent, with a high risk for rejecting true significant findings. The false discovery rate (FDR) approach[46] is a compromise between not correcting for multiple comparisons, which is too lax, and Bonferroni adjustments, which are too strict. FDR has two methods, the Benjamini and Hochberg (BH) method and the Benjamini and Liu method (BL)[46]. The BH method is appropriate for correcting for both independent and positively-dependent comparisons[47], and thus is appropriate for genetic studies using polymorphisms.

FDR corrections were performed separately for single gene case–control and family-based comparisons, as well as GDA findings, using an initial FDR threshold of 0.050.

Linkage disequilibrium of polymorphisms at the DRD2 and PPP1R1B loci

High r² measures of linkage disequilibrium (LD) were observed between DRD2 polymorphisms rs1079597 and rs1800497, with no LD between rs1799732 and rs1079597 and rs1800497 (Figure1). R-squared measures between PPP1R1B markers showed high LD (r² > 0.9) between rs907094 and rs3764352, and lower LD between rs1495099 and rs907094 and rs3764352 (Figure2).

Case–control comparisons

All three PPP1R1B markers were in HWE in the comparison group; none were in HWE in the cohort of affected individuals (P = 0.008, P = 0.033 and P = 0.033, respectively), although all markers were in HWE in the parents (data not shown). As shown in Table2, the rs1495099 CC (P = 0.002), rs907094 CC (P = 0.028) and rs3764352 GG (P = 0.025) genotype frequencies were increased in affected males (22.0%, 14.5% and 14.5%, respectively) relative to the comparison group (9.9%, 6.9% and 6.7%, respectively). Findings on alleles of these polymorphisms were similar, with the minor allele frequencies of all three markers, rs1495099 C, rs907094 C and rs3764352 G, being increased in the affected males relative to the comparison group (P = 0.001, P = 0.014 and P = 0.021, respectively, all remained significant following FDR correction; data not shown).

Because we hypothesize that maternal effects including genetic factors may contribute to autism susceptibility in some autism families[18], we compared frequencies of rs1495099 C alleles and rs1495099 CC genotypes between mothers (33.2% and 12.7%, respectively) and the comparison group (28.7% and 9.9%, respectively) but found no significant differences (P = 0.19 and P = 0.39, respectively; data not shown).

Family-based association tests

For PPP1R1B, a recessive model was used in FBAT because of the increased frequency of rs1495099 CC, rs907094 CC and rs3764352 GG genotypes found in affected individuals compared to the comparison cohort. Family-based association analyses and FDR-based corrections showed significant over-transmission of rs1495099 C (P = 0.00092) but not of rs907094 C (P = 0.11) or rs3764352 G (P = 0.09) (Table3). The rs1495099 C - rs907094 C - rs3764352 G (C-C-G) haplotype was not significantly over-transmitted to affected males (P = 0.031; not significant following FDR correction; data not shown) compared to that of rs1495099 C alone (P = 0.00092).

Genotype-phenotype associations

We next used quantitative transmission disequilibrium tests (QTDT) to determine whether the extent of impairment in the core behaviours was more pronounced in affected males with the risk allele. The DRD2 rs1800498 T allele was associated with more severe impairments in reciprocal social interaction (P = 0.0002), verbal communication (P = 0.0004), and repetitive and stereotyped behaviours (P = 0.0021); these findings remained significant following corrections for multiple comparisons.

General discriminant analyses

Adding all possible two-way interactions between DRD2 rs1800498 and PPP1R1B rs1495099 genotypes, as well as comparisons between DRD1 rs265981–rs4532–rs686 haplotypes and DRD2 rs1800498 genotypes, and DRD1 rs265981–rs4532–rs686 haplotypes and PPP1R1B rs1495099 genotypes, we found no evidence for gene-gene interactions (P = 0.35–0.75; data not shown).

Functional effects of DRD2 and PPP1R1B risk alleles on gene expression

Our findings at the DRD2 and PPP1R1B loci may reflect the functional effects of unidentified risk variants in LD with rs1800498 at DRD2 and rs1495099 at PPP1R1B. Functional analyses of these markers have not been reported but in silico analyses performed using PupaSuite [available athttp://pupasuite.bioinfo.cipf.es/][48] did not identify any putative functional role for these polymorphisms while analyses using FASTSNP [available at http://http://fastsnp.ibms.sinica.edu.tw/pages/input_CandidateGeneSearch.jsp][49] predicted a ‘very low-to-low’ effect for rs1800498 at DRD2 as an intronic enhancer and a ‘very low-to-medium’ effect for rs1495099 at PPP1R1B as a regulatory region/intronic enhancer. Meyer-Lindenberg et al.[32] (2007) identified a common 7-marker PPP1R1B haplotype that was associated with increased DARPP-32 mRNA expression and improved performance on measures of working memory and cognitive flexibility. This haplotype included the T and A alleles of rs907094 and rs3764352 respectively, while a haplotype containing the minor alleles at these loci (i.e. rs907094 C and rs3754352 G) was associated with decreased mRNA expression in post-mortem brain. Houlihan et al.[50] (2009) screened this 7-marker haplotype to test PPP1R1B as a genetic determinant of cognitive ageing and found that rs907094 C and rs3754352 G alleles were associated with decreased cognitive ability. However, our findings do not support an association of either rs907094 C or rs3764352 G with autism in our family cohort. Unfortunately, because rs1495099 was not included in these studies, its functional role is not known. With respect to the DRD2 locus, the rs1800498 polymorphism was found to be in low LD with rs1799732, a functional variant in the DRD2 promoter[51], in our comparison group and parents from families (Figure1). To investigate whether alleles from rs1799732 are contributing as a risk factor for autism susceptibility in our families, comparisons between family-based tests of rs1800498 and rs1799732 alleles considered separately (P = 0.0003 and P = 0.16, respectively), and haplotypes containing alleles from both rs1800498 and rs1799732, showed that the observed over-transmission in families is derived from rs1800498, and not because of the rs1799732 polymorphism (data not shown). However, only 31 families were informative for rs1799732 compared to 73 families for rs1800498, so we cannot determine from these findings whether alleles from the functional variant rs1799732 are contributing to autism susceptibility in our families. Another genetic variant at the DRD2 locus, rs1076560GT, has been associated with altered mRNA isoform expression[52, 53] and differences in striatal post-synaptic D2 receptor abundance[54]. Bertolini et al.[53] (2009) found in control and schizophrenia cohorts (N = 114 and N = 91, respectively) that individuals heterozygous for the minor “T” allele perform worse in the N-back test of working memory but only at a high level of difficulty (2-back) compared to individuals homozygous for the major “G” allele while Zhang et al.[52] (2007) found using healthy subjects (N = 117) that heterozygous individuals performed worse at higher attentional loads in the variable attentional control (VAC) task, and had increased activity as measured using BOLD fMRI in PFC and striatum compared to individuals homozygous for the major allele. However, the true contribution of this variant to DAergic function and cognition is unclear as no genotype effects of this polymorphism to overall working memory performance and fMRI activity were also reported[52, 54]. Nevertheless, both rs1076560 and rs1079597 are in high LD in the HapMap CEU panel (r² > 0.9) and are found in the same 20 kb haplotype block as rs1800498. However, no information is available regarding LD between rs1076560 and rs1800498. It is of interest that low LD (r² < 0.3) was found between rs1800498 and rs1079597 in our comparison group and parents from families (Figure1). Additional families informative at these DRD2 loci are needed to determine whether these functional variants or another functional polymorphism in LD with rs1800498 is responsible for the increased risk for autism. In addition, functional analyses and resequencing of both genes in affected individuals with this risk haplotype at DRD2 or the risk allele at PPP1R1B are required.

Pathophysiological contributions of DRD2 and PPP1R1B to risk for autism

The QTDT results support an association of the DRD2 and PPP1R1B loci with autism (Table4). A role for the DRD2 gene in autism susceptibility is suggested by the fact that antipsychotic medications, which prevent dopamine D2 receptor activation, improve the core symptoms of ASDs[55]. Postsynaptic D2 receptors and presynaptic D2 autoreceptors are involved in the DAergic modulation of cognitive and emotional processes that are impaired in individuals with autism[56, 57]. Thus, functional polymorphisms which affect receptor availability (e.g. altered gene expression), either postsynaptically on DAceptive neurons or presynaptically on DAergic neurons, may contribute to the impairments found in individuals with autism.

DARPP-32 mediates the downstream effects of dopamine receptor activation, and thus plays an important role in the modulation of DA-related processes which are abnormal in individuals with autism. Unlike dopamine receptors, which can be studied using systemic or local administration of ligands, DARPP-32 is found in the cytoplasm of DAceptive neurons and there are few studies which have examined its role in DA-modulated processes and behaviours. One such study by Hotte et al.[58] (2006) found that administration of D1 receptor antagonists in mice caused deficits in working memory which coincided with decreased levels of phosphorylated-DARPP-32 in the PFC. Deficiencies in working memory[59] and impairments in reversal learning[10] are found in individuals with autism. Both Drd2 −/− mice and Ppp1r1b −/− mice exhibit impairments in reversal learning compared to wild-type mice[31, 60].

The role of DARPP-32 in mediating the DA-related changes to neuronal excitability necessary for memory and learning was shown in a study by Calabresi et al.[61] (2000). These authors were unable to induce long-term potentiation (LTP) and long-term depression (LTD), two forms of synaptic plasticity, in the striatum of Ppp1r1b −/− mice. DA has a role in synaptic plasticity in both the striatum[62] and amygdala[63], subcortical structures that are important for regulating emotional behaviours[64] and social interactions[65], and are implicated in the pathophysiology of repetitive behaviours[66].

Effects between dopamine-related genes and risk for ASD

Based on our findings in this study and our previous findings[5], single-gene analyses showed evidence for association of DRD1 DRD2 and PPP1R1B with autism in male-only affected sib-pair families. We performed general discriminant analysis using Statistica v9.1 to predict ASD susceptibility in affected males, and to test for evidence of gene-gene interactions of DA-related genes and ASDs. We found that DRD2 rs1800498 genotypes and PPP1R1B rs1495099 genotypes significantly contributed to prediction of ASDs in our families when tested separately and together (P = 0.0063–0.00011), which accounted for ~7% of the variance. These results were significant following corrections for multiple comparisons (Table5). We also found that 97% of individuals from our comparison cohort and 13% of individuals with autism were correctly classified while 72% of control individuals and 64% of affected individuals were predicted correctly using a weighted additive combination of DRD2 and PPP1R1B genotypes. The inclusion of DRD1 rs265981–rs4532–rs686 haplotypes into the analysis did not significantly change our findings (data not shown), and we found no evidence for gene-gene interactions (data not shown).

Our findings suggest that single genes individually confer risk of autism susceptibility in our families and, although there was no evidence for gene-gene interactions between DA-related genes and autism susceptibility per se, we did find evidence that DRD2 and PPP1R1B together significantly contribute to ASD prediction. Thus, while both genes independently confer risk of autism susceptibility, there is a cumulative effect towards predicting whether an individual has the condition. Furthermore, based on our findings from GDA analysis we found an effect size of ~7%. While this effect size is small, it is comparable to those reported in similar studies[4]. These generally small effect sizes likely reflect the highly heterogeneous nature of ASDs.

Parent-of-origin effects of DRD1 and DRD2 alleles on development

The increased transmission of the DRD1 rs265981 C –rs4532 A –rs686 T (C-A-T) haplotype from mothers (P = 0.029)[5], and of the DRD2 rs1800498 T allele from fathers (P = 0.023), to affected sons suggests that imprinting effects of these genes are also important risk factors for ASDs. A role for imprinting has been proposed for several brain-related disorders, including autism[67]. There is evidence of imprinting of genes from neurotransmitter pathways implicated in ASDs such as the 5-hydroxytryptamine receptor 2A (HTR2A) gene in the serotoninergic pathway[68].

The dopamine D1 and D2 receptors are expressed in human placentae[69, 70] and fetal brains[71, 72]. Placental D1 and D2 receptors are involved in DA-mediated release of opioids[73] and lactogen[74] respectively, which are required for fetal development and growth . Both dopamine D1 and D2 receptors are involved in brain development. Dopamine D2 receptors expressed in fetal brain induce neurite outgrowth and axon elongation while dopamine D1 receptors inhibit neurite outgrowth in cortical neuron differentiation[75], with the opposite effects found in striatal neuron differentiation[76]. There is no evidence for the kind of parent-of-origin-specific DNA methylation at the DRD1 or DRD2 loci in either human placentae or fetal brains that is suggestive of imprinting[77]. Further, a review of the ‘imprinted gene and parent-of-origin effect’ database [available athttp://igc.otago.ac.nz/home.html][78] did not yield any evidence supporting imprinting at either locus. The possibility remains, however, that imprinting of these genes may occur during a very narrow developmental period or in a specific subpopulation of brain cells, as has been demonstrated for the Ube3a gene in mice[79]. It should be noted that the evidence presented for parent-of-origin effects in ASDs is based on three markers in the DRD1 gene and one marker in the DRD2 gene and thus, more polymorphisms need to be studied to confirm our findings.