Skip to main content
PMC home page
American Journal of Human Genetics logoLink to American Journal of Human Genetics
. 2005 Jul 15;77(3):377–388. doi: 10.1086/433195

Identification of Significant Association and Gene-Gene Interaction of GABA Receptor Subunit Genes in Autism

D Q Ma 1, P L Whitehead 1, M M Menold 1, E R Martin 1, A E Ashley-Koch 1, H Mei 3, M D Ritchie 4, G R DeLong 2, R K Abramson 5, H H Wright 5, M L Cuccaro 1, J P Hussman 6, J R Gilbert 1, M A Pericak-Vance 1
PMCID: PMC1226204  PMID: 16080114

Abstract

Autism is a common neurodevelopmental disorder with a significant genetic component. Existing research suggests that multiple genes contribute to autism and that epigenetic effects or gene-gene interactions are likely contributors to autism risk. However, these effects have not yet been identified. Gamma-aminobutyric acid (GABA), the primary inhibitory neurotransmitter in the adult brain, has been implicated in autism etiology. Fourteen known autosomal GABA receptor subunit genes were studied to look for the genes associated with autism and their possible interactions. Single-nucleotide polymorphisms (SNPs) were screened in the following genes: GABRG1, GABRA2, GABRA4, and GABRB1 on chromosome 4p12; GABRB2, GABRA6, GABRA1, GABRG2, and GABRP on 5q34-q35.1; GABRR1 and GABRR2 on 6q15; and GABRA5, GABRB3, and GABRG3 on 15q12. Intronic and/or silent mutation SNPs within each gene were analyzed in 470 white families with autism. Initially, SNPs were used in a family-based study for allelic association analysis—with the pedigree disequilibrium test and the family-based association test—and for genotypic and haplotypic association analysis—with the genotype-pedigree disequilibrium test (geno-PDT), the association in the presence of linkage (APL) test, and the haplotype family-based association test. Next, with the use of five refined independent marker sets, extended multifactor-dimensionality reduction (EMDR) analysis was employed to identify the models with locus joint effects, and interaction was further verified by conditional logistic regression. Significant allelic association was found for markers RS1912960 (in GABRA4; P = .01) and HCV9866022 (in GABRR2; P = .04). The geno-PDT found significant genotypic association for HCV8262334 (in GABRA2), RS1912960 and RS2280073 (in GABRA4), and RS2617503 and RS12187676 (in GABRB2). Consistent with the allelic and genotypic association results, EMDR confirmed the main effect at RS1912960 (in GABRA4). EMDR also identified a significant two-locus gene-gene effect model involving RS1912960 in GABRA4 and RS2351299 in GABRB1. Further support for this two-locus model came from both the multilocus geno-PDT and the APL test, which indicated a common genotype and haplotype combination positively associated with disease. Finally, these results were also consistent with the results from the conditional logistic regression, which confirmed the interaction between GABRA4 and GABRB1 (odds ratio = 2.9 for interaction term; P = .002). Through the convergence of all analyses, we conclude that GABRA4 is involved in the etiology of autism and potentially increases autism risk through interaction with GABRB1. These results support the hypothesis that GABA receptor subunit genes are involved in autism, most likely via complex gene-gene interactions.

Introduction

Autistic disorder (MIM 209850) is a neurodevelopmental disorder characterized by impairments in reciprocal social interaction and communication and the presence of restricted and repetitive patterns of interest or behavior. These impairments are apparent in the first 3 years of life and persist into adulthood. With the improved detection and recognition of autism that has resulted from a broadening of the diagnostic concept and systematic population approaches, a recent prevalence study reported that autistic disorder affects as many as 1 in 300 children in a U.S. metropolitan area (Yeargin-Allsopp et al. 2003). The increase in prevalence has drawn significant attention from scientists, and a rapid increase in the level of interest in the etiology of autism has been seen in the past decade (Fombonne 1999, 2003).

Autism has turned out to be one of the most heritable complex genetic disorders in psychiatry. A strong genetic component in autism is indicated by an increased concordance rate in MZ twins (60% and 91% for the narrow and broad phenotypes, respectively), compared with that in DZ twins (0% and 10% for the narrow and broad phenotypes, respectively) (Steffenburg et al. 1989; Bailey et al. 1995), and by a 75-fold greater risk to siblings of idiopathic cases, in comparison with the risk to the general population (Bolton et al. 1994). Collectively, these studies suggest that autistic disorder involves multiple variants in multiple unlinked loci that interact to cause the autism phenotype. In addition to genetic risk-assessment studies, both direct mapping approaches (chromosomal methods and linkage and association studies) and indirect mapping approaches (the characterization of disorders that share some of the symptoms of autism, such as Rett or fragile X syndrome) have been applied to identify autism-susceptibility genes. These studies have also yielded convincing evidence for the multigenic inheritance and the locus or allelic heterogeneity in autism.

Over 10 genomewide autism screens have been performed (International Molecular Genetic Study of Autism Consortium 1998, 2001; Meyers et al. 1998; Philippe et al. 1999; Risch et al. 1999; Collaborative Linkage Study of Autism 2001; Liu et al. 2001; Auranen et al. 2002; Shao et al. 2002; Yonan et al. 2003). Results from these various screens indicate potential susceptibility genes spread across the entire genome. Estimates of the number of genes involved in autism range from 3–10 (Pickles et al. 1995; Folstein and Rosen-Sheidley 2001) to ⩾15 (Risch et al. 1999) to 100 (Pritchard 2001). Numerous association studies of the candidate genes have been conducted on the basis of location in a linkage peak or potential function, but no single gene has been consistently replicated across studies. One explanation for the lack of consistency in association studies is that there are many contributing genetic and environmental factors in autism. Moreover, multiple interacting genes may be the main causative determinants of autism (Muhle et al. 2004; Veenstra-VanderWeele et al. 2004). With only a modest sample size, a small-to-moderate locus effect is not easily detected. Therefore, tests for joint effects may be more successful in the search for autism-susceptibility genes.

Several lines of research indicate that there are abnormalities in the gamma-aminobutyric acid (GABA) system that may lead to developmental changes similar to those observed in autism. The evidence implicates GABA receptor (GABAR) subunit genes as functional candidates for autism (Blatt et al. 2001; Hussman 2001; Aldred et al. 2003). GABA acts on the GABAR complex, a heteromeric structure, and mediates synaptic inhibition in the adult brain (Hahn et al. 2003; Moore 2003). During development, GABA also acts as an excitatory neurotransmitter because of the high intracellular chloride concentration in immature neurons (Jentsch et al. 2002). Eight GABA classes (α, β, δ, ɛ, γ, π, θ, and ρ) and 18 receptor subunit genes have been characterized in mammals. In addition to providing binding sites for GABA, the GABAR contains sites for several therapeutic agents, including benzodiazepines, barbiturates, anesthetics, and alcohols. Binding studies that used labeled ligands in children indicate that GABAR density is greater early in life and then dramatically decreases to adult levels (Chugani et al. 2001). Subunit composition varies developmentally and across brain structure. It is notable that the studies found a significant decrease in GABAR density in autism (Blatt et al. 2001) and an elevated plasma GABA level in autistic children (Dhossche et al. 2002).

The most promising region identified by autism association studies is on chromosome 15q12, which harbors a set of three GABAR subunit genes (Cook et al. 1998; Martin et al. 2000a; Wolpert et al. 2000; Boyar et al. 2001; Menold et al. 2001; Buxbaum et al. 2002). Chromosome 15q11-q13 duplications and deletions have also been documented in children with autism (Bundey et al. 1994; Smith et al. 2000; Pujana et al. 2002). In addition, several groups have identified this region as an area of interest through linkage studies (Philippe et al. 1999; Liu et al. 2001). Follow-up fine mapping narrowed this 15q region to the GABRB3 gene by use of a phenotypic subtype defined by a high degree of insistence on sameness (Shao et al. 2003). All of these findings from direct or indirect mapping studies strongly suggest that the GABAR subunit genes may play an important role, both independently and interactively, in the etiology of autism.

Epistasis or gene-gene interaction has been widely accepted as an important contributor to the complexity of mapping complex disease genes (Moore 2003). The failure to replicate some single-locus results might be the result of an underlying genetic architecture in which gene-gene interactions are the norm rather than the exception (Moore and Williams 2002). Thus, genetic studies that ignore epistasis or gene-gene interactions are likely to reveal only part of the genetic architecture. Although the term “epistasis” was initially used by William Bateson early in the 20th century to describe the reason for distortions of Mendelian segregation ratios and was later defined by Fisher as deviation from additivity in a linear statistical model (Moore 2005), the methodology for testing epistasis or gene-gene interaction is still in its infancy. The available methods have been thoroughly reviewed recently (Thornton-Wells et al. 2004). In general, a lack of powerful statistical methods and large sample sizes limits the identification and characterization of gene-gene interactions (Moore and Williams 2002). The main issues confronted by traditional methods, such as logistic regression, are insufficient power and flexibility to detect high-order gene-gene interactions. Several newly developed methods, such as the multilocus genotype-pedigree disequilibrium test (geno-PDT) (Martin et al. 2003a) and the multifactor-dimensionality reduction (MDR) method (Ritchie et al. 2001), improve the ability to identify the high-order gene-gene interactions with the use of relatively small sample sizes. However, the methods have difficulty in distinguishing true interactive effects from joint effects. With the data-driven analytic methods that are continually in development to examine complex genetic interactions, it has become increasingly important to emphasize model validation to ensure that significant effects represent true relationships rather than chance findings (Coffey et al. 2004). Thus, a multianalytic approach to the analysis of gene-gene interactions was proposed (Ashley-Koch et al. 2004), an approach that searches for consistency of results and preponderance of evidence to draw the most useful conclusions. In the current study, this new paradigm was applied to test our hypothesis that GABAR subunit genes may contribute to the etiology of autism independently and/or through complex interactions between subunit genes.

Families and Methods

Family Ascertainment

A standard ascertainment protocol was conducted by the clinical groups at the Duke Center for Human Genetics and the William S. Hall Psychiatric Institute. Both sites recruited, enrolled, and sampled individuals with autism and family members per study protocols approved by their respective institutional review boards (IRBs). Participating families were ascertained using clinical referrals and active recruitment through lay organizations that provide services to families with autism. After a full description of the study was given to the families, written informed consent was obtained from parents and from children who were able to give informed consent. For the current study, a total of 470 white families were included, of which 266 were multiplex (i.e., more than one affected individual sampled) and 204 were triads (i.e., only one affected individual sampled). The Collaborative Autism Team from the Duke Center for Human Genetics and the William S. Hall Psychiatric Institute contributed 246 families, and 224 families were from the Autism Genetic Resource Exchange. Probands for the study consisted of individuals between the ages of 3 years and 21 years who received a clinical diagnosis of autism by use of DSM-IV criteria. A consistent set of diagnostic criteria was applied to all families. Qualified individuals and families were those who met best-estimate clinical research diagnoses for autism, as determined by the lead clinicians (H.H.W. and M.L.C.) at each of the research sites. The best-estimate diagnoses were made using all available case material, including clinical records, Autism Diagnostic Interview–Revised (ADI-R) results, and clinical assessment information. All qualified individuals met current DSM-IV diagnostic criteria for autism. The ADI-R (Lord et al. 1997) is a validated, semistructured diagnostic interview, which yields a diagnostic algorithm based on the DSM-IV criteria for autism. All ADI-R interviews were conducted by formally trained interviewers who have established reliability within our group as well as within Dr. Lord’s group. Finally, all participants who met current diagnostic criteria for autism were included only if they had a minimal developmental level of 18 mo for the Vineland Adaptive Behavior Scale score (Sparrow et al. 1984) or had an IQ equivalent >35. These minimal developmental levels assure that ADI-R results are valid and reduce the likelihood of including individuals with severe mental retardation only. Subjects were excluded if they had evidence of developmental disorders with known phenotypic overlap with autism (e.g., Prader-Willi syndrome, Angelman syndrome, tuberous sclerosis complex, Rett syndrome, and fragile X syndrome), neurologic disorders, or severe sensory or motor disorders.

Genotyping

Blood was obtained from patients and other family members in accordance with IRB-approved procedures. DNA was extracted from whole blood by use of standard protocols (Vance 1998). Analysis of the candidate region was performed using data obtained from SNPs. SNPs located within the GABAR subunit genes across chromosomes were analyzed. Between three and seven intronic and silent mutation SNPs within each gene were identified with Applied Biosystems (ABI) Assay-on-Demand products. The selected GABAR subunit genes were GABRG1 (3 SNPs typed), GABRA2 (6), GABRA4 (7), and GABRB1 (7) on 4p12; GABRB2 (6), GABRA6 (4), GABRA1 (5), GABRG2 (3), and GABRP (4) on 5q34-q35.1; GABRR1 (7) and GABRR2 (4) on 6q15; and GABRB3 (5), GABRA5 (4), and GABRG3 (5) on 15q12. SNPs were identified from the National Center for Biotechnology Information SNP database (see dbSNP Web site) and were ordered for use in either Assays-on-Demand or Assays-by-Design (ABI). All SNPs were genotyped using TaqMan. All reactions contained 2.7 ng of total genomic DNA and were run on ABI 9700 GeneAmp PCR systems in accordance with the manufacturer’s instructions. Analysis of the SNP genotypes was performed using an ABI Prism 7900HT Sequence Detection System.

For quality control (QC) procedures, two CEPH standards were included on each 96-well plate, and samples from six individuals were duplicated across all plates as QCs, with the laboratory technicians blinded to their identities. Analysis required that identical QC samples within and across plates have matching genotypes, to identify errors in loading and reading and thus to minimize the error rate in genotype assignments. Meanwhile, a 95% efficiency of genotype is required. Technicians generating the genotypic data were blinded to the clinical status of the patients. After QC verification, genotypes of the samples were uploaded into the PEDIGENE database and were merged into the LAPIS management system for creation of analysis input files (Haynes et al. 1995).

Statistical Analysis

To further check for genotyping error, the program PedCheck was run for detection of Mendelian inheritance inconsistency. The error checking option embedded in Merlin (Abecasis et al. 2002) was run to identify the samples with excess recombinations, and the families were checked further for possible genotyping errors. One affected and one unaffected individual were selected randomly from each family for tests of Hardy-Weinberg equilibrium (HWE), which was assessed using exact tests implemented in the Genetic Data Analysis program (Zaykin et al. 1995). For SNPs found to be not in HWE in the unaffected sample, a sequence of samples at that particular SNP was required to ensure the quality of the SNPs. Pairwise linkage disequilibrium (LD) (D′ and r2) between markers was calculated using the GOLD software package (Abecasis and Cookson 2000). The allelic association analyses were conducted using the pedigree disequilibrium test (PDT) (Martin et al. 2000b) and the family-based association test (FBAT) (Horvath et al. 2004). These two tests are similar in many aspects, but each has distinct advantages. The PDT has the advantage of being valid as a test of both linkage and association in extended pedigrees, whereas the FBAT treats nuclear families within large pedigrees as independent but permits haplotype-based association tests. Both PDT and FBAT are allele-based tests. The geno-PDT (Martin et al. 2003a) is an extension of PDT used to examine the association between marker genotype and disease. The haplotype family-based association test (HBAT) (Horvath et al. 2004) was used for haplotype association analysis for SNPs within each GABAR subunit gene. Tagging SNPs within each gene were selected by use of the confidence-interval function in Haploview (Barrett et al. 2005). Both haplotype-specific P values and global P values (with adjustment for all possible haplotypes) were calculated by the program.

The core program of MDR (Ritchie et al. 2001; Hahn et al. 2003) was employed in this study to test for potential gene-gene interaction and thereby to identify specific locus combinations of interest for further investigation and replication. Some new features were added to MDR through the extended MDR (EMDR) (Mei et al., in press). Basically, EMDR uses the same algorithm as the core MDR program, a data reduction program that tests for interactions (Ritchie et al. 2001; Hahn et al. 2003). EMDR contains several new features. Briefly, these are (1) allowance of missing data in individuals with partial genotype data, (2) use of a χ2 statistic in addition to the prediction error as a test statistic, and (3) introduction and implementation of a nonfixed permutation test to assess the statistical significance of models identified by EMDR. This nonfixed permutation generates an empirical P value for a particular n-locus model, with consideration of all combinations of n loci. For example, for a particular 2-locus combination, the nonfixed permutation test accounts for the search of all possible 2-locus models to decide whether the best model is significant. An empirical P value <.05 was regarded as statistically significant and is inherently adjusted for multiple testing. In this study, a cross-validation option was not used.

For case-control pairs used in EMDR, the proband (or the most completely genotyped affected child) from each multiplex and triad family was selected (n=470), and the untransmitted alleles were generated using parental genotypes as a control. Given the sample size of 470 case-control pairs in this study, we did not test for interactions that were >4-way interactions (Mei et al., in press).

Independent markers (tagging SNPs) were used in four chromosome-by-chromosome models and were selected using the confidence-interval function in Haploview (Barrett et al. 2005). Meanwhile, to retain adequate power to detect a gene-gene effect, the markers with the smallest P values from the four by-chromosome models were selected to build the final cross-chromosome model. The reason for the selection is that the permutation test inherently adjusts for multiple comparisons, and true effects can be overwhelmed when many markers are considered. Therefore, to maintain reasonable power, we judiciously chose a relatively small subset of markers for the MDR analysis. Our approach was staged so that each chromosome was examined and the markers having the smallest P values within each chromosome were selected for the overall analysis.

The significant best models identified by EMDR can only suggest a gene-gene effect, rather than a certain interaction. This holds true especially if a particular locus in a significant n-locus model also presents a significant main effect as the best 1-locus model. In this case, the identified gene-gene effect may be driven by the main effect from the locus, rather than by a true interaction. To verify the interaction between genes in the identified model, conditional logistic regression (by use of COXREG in SPSS, version 11.5 for Windows) was performed. To test for interaction, all variables (i.e., markers in the identified model) and their interaction terms were forced into the model. The genotypes of the markers were recoded in the logistic regression analysis. Genotypes with a case:control ratio >1 were collapsed and were recoded as the high-risk group, and those with a ratio <1 were recoded as the low-risk group. This matched the dimensionality-reduction strategy applied in EMDR, enabling consistent interpretation of the results between EMDR and the logistic regression analysis. In this study, GG was coded as a high-risk group for marker RS1912960, and GG and TT were coded as high-risk groups for RS2351299. Finally, multilocus geno-PDT and association in the presence of linkage (APL) analysis (Martin et al. 2003b) were used to validate the gene-gene interaction from the logistic regression.

Results

No significant deviation from HWE was found in unaffected whites for all SNPs. SNP RS1426217 (in GABRB3) on chromosome 15 presented evidence of deviation from HWE in the affected individuals (P=.019). PDT showed that RS1912960 (in GABRA4) on chromosome 4 had a preferential transmission of the common G allele to the affected offspring (P=.012) (table 1). In addition, FBAT identified a significant association at HCV9866022 (in GABRR2) on chromosome 6 (P=.04), where the PDT results suggested a similar trend (P=.064) (the entire FBAT data are not shown; results were similar to those of PDT). The geno-PDT displayed positive genotype association on chromosome 4 for homozygous common genotypes TT, GG, and GG at HCV8262334 (in GABRA2), RS1912960 (in GABRA4), and RS2280073 (in GABRA4), respectively, and, on chromosome 5, for heterozygous genotypes CT and CG in RS2617503 and RS12187676 (both in GABRB2), respectively (global P values given in table 1). SNPs on the same chromosome did not show LD with each other (LD data not shown).

Table 1.

The PDT and geno-PDT Association Analyses of GABA Genes and Autism

Global P fora
Chromosome,Gene, and SNP Marker Number PDTb geno-PDTc
Chromosome 4:
 GABRG1:
  RS1497571 1 .899 .906
  RS2350439 2 .509 .717
  RS1826923 3 .340 .622
 GABRA2:
  HCV7537166 4 .556 .778
  RS279858 5 .361 .623
  RS279844 6 .138 .294
  HCV8262290 7 .064 .141
  RS4695152 8 .508 .730
  HCV8262334 9 .149 .033
 GABRA4:
  RS7678338 10 .737 .936
  RS1512136 11 .935 .996
  HCV1592545 12 1.000 .968
  RS1912960 13 .012 .003
  RS2280073 14 .072 .034
  RS10517174 15 .738 .142
  RS3792211 16 .677 .391
 GABRB1:
  RS2351299 17 .817 .098
  RS1372496 18 .088 .160
  RS3114084 19 .180 .317
  HCV11353524 20 .115 .243
  HCV2119841 21 .906 .432
  RS6289 22 .544 .276
  RS6290 23 .940 .506
Chromosome 5:
 GABRB2:
  RS253017 24 .774 .317
  RS252965 25 .649 .299
  RS2617503 26 .108 .025
  RS2962425 27 .367 .443
  RS2962407 28 .771 .149
  RS12187676 29 .407 .015
 GABRA6:
  RS3811995 30 .613 .488
  RS6883829 31 .932 .236
  HCV164095 32 .814 .920
  RS3811991 33 .652 .283
 GABRA1:
  RS4340950 34 .426 .650
  HCV11258504 35 .633 .601
  RS6878494 36 .395 .699
  HCV1667770 37 .861 .522
  HCV11814555 38 .294 .576
 GABRG2:
  RS7728001 39 .670 .833
  RS766349 40 .700 .223
  RS211014 41 .655 .815
 GABRP:
  HCV3165046 42 .872 .779
  RS1812910 43 .965 .981
  RS1862242 44 .347 .593
  RS1063310 45 .560 .807
Chromosome 6:
 GABRR1:
  RS404943 46 .619 .674
  RS407206 47 .623 .835
  RS423463 48 .475 .718
  RS3777530 49 .644 .831
  RS2297389 50 .851 .150
  RS881293 51 .832 .978
  RS6902106 52 .829 .712
 GABRR2:
  RS282117 53 .277 .494
  HCV9866022 54 .064 .171
  RS2148174 55 .855 .770
  HCV9865968 56 .780 .962
Chromosome 15:
 GABRB3:
  RS2081648 57 .602 .837
  RS1426217 58 .191 .305
  RS754185 59 .672 .852
  HCV8865209 60 .337 .521
  RS2059574 61 .304 .405
 GABRA5:
  HCV42974 62 .646 .072
  RS7173260 63 .938 .845
  RS140681 64 .886 .762
  RS140683 65 .825 .978
 GABRG3:
  HCV2078506 66 .079 .240
  RS208129 67 .281 .266
  RS897173 68 .240 .451
  HCV428306 69 1.000 .611
  RS140679 70 .410 .589
Open in a new tab
a

P values <.05 are shown in bold italics.

b

P values adjusted for two alleles.

c

P values adjusted for three genotypes.

Haplotype analysis was performed using tagging SNPs within each gene, and it confirmed significant association with autism for specific haplotypes within GABRA2 (P=.027), GABRA4 (P=.025), and GABRR2 (P=.028). However, the global P value was not significant (P>.05) for any of the genes tested.

To test for a gene-gene effect, EMDR was run for by-chromosome and cross-chromosome models. Of all of the by-chromosome models tested (tables 2–6), two significant models were found on chromosome 4. There is a 2-locus model involving RS1912960 in GABRA4 and RS2351299 in GABRB1 (P=.002) and a 3-locus model involving RS2350439 in GABRG1, RS1912960 in GABRA4, and RS3114084 in GABRB1 (P=.03), which suggests a potential gene-gene interaction among GABRG1, GABRA4, and GABRB1 (table 2). (The original MDR under the 10-fold cross-validation option was run, and it confirmed a potential gene-gene effect in a 2-locus model [prediction error = 43%; P=.023].) In the cross-chromosome model (table 6), EMDR identified the same best 1-locus and 2-locus (P=.001) models as in the by-chromosome model for chromosome 4 and confirmed the main effect at RS1912960 (in GABRA4; P=.02). Another 3-locus model (RS282117 and RS2148174 in GABRR2 and RS208129 in GABRG3; P=.008) was also identified, suggesting a potential gene-gene interaction across chromosomes, between GABRR2 (chromosome 6) and GABRG3 (chromosome 15).

Table 2.

Best Gene-Gene Effect Models Identified by EMDR for GABAR Subunit Genes on Chromosome 4

Nonfixed P Valuea
Geneand Marker Marker Number Location Best Model χ2 Misclassification
GABRG1:
 RS1497571 1 65.47 9 .06 .038
 RS2350439 2 65.49 9, 13 .004 .002b
GABRA2: 2, 9, 15 .016 .03
 RS279858 3 65.66 2, 7, 9, 15 .16 .33
 RS279844 4 65.66
 HCV8262290 5 65.67
 RS4695152 6 65.68
 HCV8262334 7 65.68
GABRA4:
 HCV1592545 8 65.85
 RS1912960 9 65.86
 RS2280073 10 65.86
 RS10517174 11 65.87
 RS3792211 12 65.87
GABRB1:
 RS2351299 13 65.92
 RS1372496 14 65.94
 RS3114084 15 65.95
 HCV11353524 16 65.97
 HCV2119841 17 65.99
 RS6289 18 66.00
 RS6290 19 66.00
Open in a new tab
a

Empirical P value derived from nonfixed permutation test by use of χ2 or the misclassification rate as the test statistic in EMDR.

b

The locus with the lowest P value (in bold italics) was selected as the one for the final cross-chromosome model.

Table 3.

Best Gene-Gene Effect Models Identified by EMDR for GABAR Subunit Genes on Chromosome 5

Nonfixed P Valuea
Genesand Marker Marker Number Location Best Model χ2 Misclassification
GABRB2:
 RS253017 1 165.09 6 .412 .282b
 RS252965 2 165.09 3, 6 .657 .762
 RS2617503 3 165.098 4, 8, 18 .710 .551
 RS2962425 4 165.155 3, 10, 13, 17 .810 .800
 RS2962407 5 165.203
 RS12187676 6 165.246
GABRA6:
 HCV164095 7 165.362
 RS3811991 8 165.364
GABRA1:
 RS4340950 9 165.476
 RS6878494 10 165.494
 HCV1667770 11 165.502
 HCV11814555 12 165.505
GABRG2:
 RS169793 13 165.667
 RS7728001 14 165.674
 RS766349 15 165.677
 RS211014 16 165.693
GABRP:
 HCV3165046 17 182.846
 RS1812910 18 182.859
 RS1862242 19 182.871
 RS1063310 20 182.877
Open in a new tab
a

Empirical P value derived from nonfixed permutation test by use of χ2 or the misclassification rate as the test statistic in EMDR.

b

The locus with the lowest P value (in bold italics) was selected as the one for the final cross-chromosome model.

Table 4.

Best Gene-Gene Effect Models Identified by EMDR for GABAR Subunit Genes on Chromosome 6

Nonfixed P Valuea
Genesand Marker Marker Number Location Best Model χ2 Misclassification
GABRR1:
 RS404943 1 94.926 8 .644 .587
 RS407206 2 94.935 5, 11 .579 .452
 RS423463 3 94.946 2, 8, 10 .398 .246b
 RS3777530 4 94.975 7, 8, 10, 11 .645 .665
 RS2297389 5 94.991
 RS881293 6 94.998
 RS6902106 7 95.019
GABRR2:
 RS282117 8 95.171
 HCV9866022 9 95.208
 RS2148174 10 95.238
 HCV9865968 11 95.277
Open in a new tab
a

Empirical P value derived from nonfixed permutation test by use of χ2 or the misclassification rate as the test statistic in EMDR.

b

The locus with the lowest P value (in bold italics) was selected as the one for the final cross-chromosome model.

Table 5.

Best Gene-Gene Effect Models Identified by EMDR for GABAR Subunit Genes on Chromosome 15

Nonfixed P Valuea
Geneand Marker Marker Number Location Best Model χ2 Misclassification
GABRB3:
 RS2081648 1 11.07 10 .219b .706
 RS1426217 2 11.08 5, 10 .494 .56
 RS754185 3 11.23 4, 10, 13 .843 .85
 HCV8865209 4 11.33 4, 5, 10, 13 .623 .875
 RS2059574 5 11.54
GABRA5:
 HCV42974 6 12.06
 RS140681 7 12.12
 RS140683 8 12.14
GABRG3:
 HCV2078506 9 12.38
 RS208129 10 12.81
 RS897173 11 12.94
 HCV428306 12 14.46
 RS140679 13 14.66
Open in a new tab
a

Empirical P value derived from nonfixed permutation test by use of χ2 or the misclassification rate as the test statistic in EMDR.

b

The locus with the lowest P value (in bold italics) was selected as the one for the final cross-chromosome model.

Table 6.

Best Gene-Gene Effect Models Identified by EMDR for All Known Autosomal GABAR Subunit Genes

Nonfixed P Valuea
Marker Gene Location Chromosome Marker Number SNP Number Best Model χ2 Misclassification
RS1912960 GABRA4 65.857 4 9 1 1 .035 .02
RS2351299 GABRB1 65.916 4 13 2 1, 2 .002 .001
RS12187676 GABRB2 165.246 5 6 3 5, 6, 7 .009 .008
RS407206 GABRR1 94.935 6 2 4
RS282117 GABRR2 95.171 6 8 5
RS2148174 GABRR2 95.238 6 10 6
RS208129 GABRG3 12.813 15 10 7
Open in a new tab
a

Empirical P value derived from nonfixed permutation test by use of χ2 or the misclassification rate as the test statistic in EMDR.

To evaluate whether the joint effects identified by the EMDR are the result of interacting genes, conditional logistic regression was conducted; the results supported a significant 2-locus gene-gene interaction between GABRA4 and GABRB1 (odds ratio [OR] of 2.9 for interaction term, high risk vs. low risk; P=.002) but did not detect an interaction in the cross-chromosome 3-locus or chromosome 4 3-locus model.

Consistent with the interaction term in the logistic regression described above (high risk [GG] and high risk [GG+TT] combination), multilocus geno-PDT (table 7) supported a positive cross-marker genotype association with disease between two common variant genotypes at RS1912960 (GG) and RS2351299 (GG). The APL analysis confirmed a positive association of the G allele at RS1912960 (P=.031) and also presented a positive haplotype association with disease for a haplotype with two common variants (G-G) (RS1912960 and RS2351299; P=.014; global P=.014), which indirectly supported the genotype association shown by EMDR.

Table 7.

Results for Multilocus geno-PDT for Two Loci on Chromosome 4

Genotypea
RS1912960 RS2351299 P Valueb
1,1 1,1 .015
1,1 1,2 .330
1,1 2,2 .096
1,2 1,1 .061
1,2 1,2 .635
1,2 2,2 .835
2,2 1,1 .001
2,2 1,2 .046
2,2 2,2 .386
Open in a new tab
a

For RS1912960, 1 = C; 2 = G (common allele); for RS2351299, 1 = G (common allele); 2 = T.

b

P value for each genotype combination. The global P value (after adjustment for all possible genotype combinations) is .0007.

Discussion

This is, to our knowledge, the first comprehensive investigation of the allelic, genotypic, and haplotypic association together with an investigation of potential gene-gene interactions of all known autosomal GABAR subunit genes with autism. These novel findings indicate that GABRA4 is involved in the etiology of autism, both independently and through interaction with GABRA1. These data support the hypothesis that complex interactions account for autism risk and present some of the first evidence for it.

In the present study, several approaches were used to control for false-positive results and thus to protect against incorrect conclusions regarding the etiology of the disease. First, we included only autism-affected white families in the analysis, to avoid biased results due to population stratification. Second, all GABA genes selected have a substantial a priori probability of involvement in autism (Sullivan et al. 2001). Finally, we used a multianalytic approach, as described elsewhere (Ashley-Koch et al. 2004), to interpret our findings. This approach looked for the convergence of results across several methods, rather than relying on results from a single analytic tool. Specifically, several approaches were applied to validate the interaction identified by EMDR, including conditional logistic regression.

To evaluate multilocus effects in a comprehensive way, the results from allelic, genotypic, and haplotypic analyses were integrated for a best estimate. We also used an extended version of MDR (Mei et al., in press), in which several modifications were made to MDR, including allowance for missing data, improved estimation of test statistic distribution, and more-accurate adjustment of multiple testing. These new features in EMDR have been validated elsewhere (Mei et al., in press). In this study, we chose to use the no-cross validation option and to omit the 10-fold cross-validation in each run. This option has shown lower false-positive and false-negative rates than the original MDR (Mei et al., in press).

We previously reported linkage and weak association for SNPs in the cluster GABAR region on chromosome 15q in our autism data set (Bass et al. 2000; Martin et al. 2000a; Menold et al. 2001; Shao et al. 2003). One possible explanation for this finding would be that there are multiple disease variants for autism risk in this region and that any one variant is only weakly associated with an individual haplotype. Similarly, the present study did not find association for a single locus in this region. However, RS1426217, which is located in intron 6 of GABAB3, significantly deviated from HWE only in affected individuals. This does not invalidate the association analysis, since both PDT and FBAT do not require HWE. It has been suggested elsewhere that absence of HWE may be an indication of the presence of association of a susceptibility allele that is in LD with the tested SNP (Nielsen et al. 1998). Thus, this finding might suggest that RS1426217 is in LD with a nearby disease allele. Extending the analysis to chromosome 15 GABAR genes (GABRB3, GABRA5, and GABRG3), we applied a similar genetic analysis paradigm (Ashley-Koch et al. 2004) to look for interactions among these three genes to determine whether these interactions contribute to risk, but no multilocus effects were detected. In a cross-chromosome model, however, we did find a joint effect between GABRR2 (chromosome 6) and GABRG3 (chromosome 15), although conditional logistic regression failed to confirm this interaction. On the basis of the FBAT and HBAT results for GABRR2, this joint effect is most likely driven by the effect from GABRR2 only. Even so, the finding merits further investigation in a larger and/or independent sample.

The most promising finding in this study was the significant allelic and genotypic association that was found at RS1912960 (in GABRA4), both from the common variant G and the common genotype GG. Also, the HBAT identified a significant haplotype within GABRA4, although the global P value showed only marginal significance (P=.06). Moreover, the association remained significant even after adjustment for multiple testing in EMDR. The program generates 1,000 simulated data sets by permuting the status of cases and controls to obtain an empirical P value for the marker while testing the significance for the 1-locus best model. RS1912960 remained the best 1-locus model in both the by-chromosome and the cross-chromosome models. The empirical P value for this marker was .038 for the by-chromosome model and .020 for the cross-chromosome model. Thus, this significant association appeared to be consistent across all analyses, which strongly suggests that GABRA4 is involved in the etiology of autism.

Although no strong linkage or significant association at GABRA4 had previously been reported, two genome screens reported suggestive linkage at 4p (International Molecular Genetic Study of Autism Consortium 1998; Auranen et al. 2000). Two chromosomal studies indicated abnormality in the 4p12 region, which is at the same location as GABRA4: a British group reported a 46,XX, dup (4) p12-p13 chromosome abnormality in an 18-year-old female with autism (Sabaratnam et al. 2000), and a Canadian group reported a paracentric inversion of the short arm of chromosome 4 (46,XY, inv [4] p12-p15.3) in two brothers with autism (Choufani et al. 2003). Given the convergence of all available results, we believe the positive association at RS1912960 (in GABRA4) reflects a true association with autism.

Relatively little is known about the biological function(s) of the GABARA4 subunit. Gene expression is known to be highly variable, depending on brain region, neuronal activity, and development, which suggests complex regulation and involvement in multiple brain activities and functions. Levels of α4 mRNA are found in the hippocampus, the dentate gyrus, the thalamus, the nucleus accumbens, the cerebellum, the outer layers of the cortex, and other regions, and they peak during development. Unlike most GABAA receptor complexes, those containing α4 are not sensitive to modulation by diazepam. It has been suggested that the α4 subunit may be involved in neuronal hyperexcitability. The promoter for α4 has multiple transcription initiation sites, and alternate splicing in mouse brain has been observed (Ma et al. 2004).

We also found a potential interaction between GABRA4 and another clustering GABA gene, GABRB1 (OR=2.9 for interaction term). We identified this potential gene-gene effect model in EMDR from both the by-chromosome and the cross-chromosome models. In addition, the interaction was further confirmed by conditional logistic regression, in which two common GG-GG variant combinations substantially increased autism risk. This finding is also consistent with the results from multilocus geno-PDT (GG-GG) and APL haplotype analysis (G-G). Again, the accumulation of findings across all analyses leads us to conclude that GABRB1 may be involved through interaction with GABRA4.

In conclusion, this study suggests that the GABRA4 gene is involved in the etiology of autism and substantially increases autism risk through interaction with the GABRB1 gene. Furthermore, these findings support the hypothesis that the etiology of autism is complex and that complex interactions may contribute to autism risk.

Acknowledgments

We thank the patients with autism and family members who agreed to participate in this study and the personnel of the Center for Human Genetics at Duke University Medical Center, for their input on this project. This research was supported in part by National Institutes of Health (NIH) program project grant NS26630 and grants R01 NS36768, R01 AG20135, and R01 NS42165; by the National Alliance of Autism Research (NAAR); and by a gift from the Hussman Foundation. The research conducted in this study complies with current U.S. laws. We also gratefully acknowledge the resources provided by the AGRE consortium and the participating AGRE families. The Autism Genetic Resource Exchange (AGRE) is a program of Cure Autism Now (CAN).

Web Resources

The URLs for data presented herein are as follows:

  1. dbSNP, http://www.ncbi.nlm.nih.gov/SNP/
  2. Online Mendelian Inheritance in Man (OMIM), http://www.ncbi.nlm.nih.gov/Omim/ (for autistic disorder)

References

  1. Abecasis GR, Cherny SS, Cookson WO, Cardon LR (2002) Merlin—rapid analysis of dense genetic maps using sparse gene flow trees. Nat Genet 30:97–101 [DOI] [PubMed] [Google Scholar]
  2. Abecasis GR, Cookson WO (2000) GOLD—graphical overview of linkage disequilibrium. Bioinformatics 16:182–183 [DOI] [PubMed] [Google Scholar]
  3. Aldred S, Moore KM, Fitzgerald M, Waring RH (2003) Plasma amino acid levels in children with autism and their families. J Autism Dev Disord 33:93–97 [DOI] [PubMed] [Google Scholar]
  4. Ashley-Koch AE, Martin ER, Jaworski J, Ma DQ, Rabionet R, Skaar DA, Menold MM, Abramson RK, Wright HH, Cuccaro ML, Gilbert JR, Pericak-Vance MA (2004) Evidence for gene-gene interactions influencing susceptibility to autistic disorder. Paper presented at the XIIth World Congress of Psychiatric Genetics, Dublin, Ireland, October 9–13 [Google Scholar]
  5. Auranen M, Nieminen T, Majuri S, Vanhala R, Peltonen L, Jarvela I (2000) Analysis of autism susceptibility gene loci on chromosomes 1p, 4p, 6q, 7q, 13q, 15q, 16p, 17q, 19q and 22q in Finnish multiplex families. Mol Psychiatry 5:320–322 [DOI] [PubMed] [Google Scholar]
  6. Auranen M, Vanhala R, Varilo T, Ayers K, Kempas E, Ylisaukko-Oja T, Sinsheimer JS, Peltonen L, Jarvela I (2002) A genomewide screen for autism-spectrum disorders: evidence for a major susceptibility locus on chromosome 3q25-27. Am J Hum Genet 71:777–790 [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Bailey A, Le Couteur A, Gottesman I, Bolton P, Simonoff E, Yuzda E, Rutter M (1995) Autism as a strongly genetic disorder: evidence from a British twin study. Psychol Med 25:63–77 [DOI] [PubMed] [Google Scholar]
  8. Barrett JC, Fry B, Maller J, Daly MJ (2005) Haploview: analysis and visualization of LD and haplotype maps. Bioinformatics 21:263–265 [DOI] [PubMed] [Google Scholar]
  9. Bass MP, Menold MM, Wolpert CM, Donnelly SL, Ravan SA, Hauser ER, Maddox LO, Vance JM, Abramson RK, Wright HH, Gilbert JR, Cuccaro ML, DeLong GR, Pericak-Vance MA (2000) Genetic studies in autistic disorder and chromosome 15. Neurogenetics 2:219–226 [DOI] [PubMed] [Google Scholar]
  10. Blatt GJ, Fitzgerald CM, Guptill JT, Booker AB, Kemper TL, Bauman ML (2001) Density and distribution of hippocampal neurotransmitter receptors in autism: an autoradiographic study. J Autism Dev Disord 31:537–543 [DOI] [PubMed] [Google Scholar]
  11. Bolton P, Macdonald H, Pickles A, Rios P, Goode S, Crowson M, Bailey A, Rutter M (1994) A case-control family history study of autism. J Child Psychol Psychiatry 35:877–900 [DOI] [PubMed] [Google Scholar]
  12. Boyar FZ, Whitney MM, Lossie AC, Gray BA, Keller KL, Stalker HJ, Zori RT, Geffken G, Mutch J, Edge PJ, Voeller KS, Williams CA, Driscoll DJ (2001) A family with a grand-maternally derived interstitial duplication of proximal 15q. Clin Genet 60:421–430 [DOI] [PubMed] [Google Scholar]
  13. Bundey S, Hardy C, Vickers S, Kilpatrick MW, Corbett JA (1994) Duplication of the 15q11-13 region in a patient with autism, epilepsy and ataxia. Dev Med Child Neurol 36:736–742 [DOI] [PubMed] [Google Scholar]
  14. Buxbaum JD, Silverman JM, Smith CJ, Greenberg DA, Kilifarski M, Reichert J, Cook EH Jr, Fang Y, Song CY, Vitale R (2002) Association between a GABRB3 polymorphism and autism. Mol Psychiatry 7:311–316 [DOI] [PubMed] [Google Scholar]
  15. Choufani S, Vincent JB, Skaug J, Kwasnicka D, Szatmari P, Scherer SW (2003) A report of one family with two autistic siblings having a paracentric inversion of chromosome 4p. Am J Hum Genet 73:A1990 [Google Scholar]
  16. Chugani DC, Muzik O, Juhasz C, Janisse JJ, Ager J, Chugani HT (2001) Postnatal maturation of human GABAA receptors measured with positron emission tomography. Ann Neurol 49:618–626 [PubMed] [Google Scholar]
  17. Coffey CS, Hebert PR, Ritchie MD, Krumholz HM, Gaziano JM, Ridker PM, Brown NJ, Vaughan DE, Moore JH (2004) An application of conditional logistic regression and multifactor dimensionality reduction for detecting gene-gene interactions on risk of myocardial infarction: the importance of model validation. BMC Bioinformatics 5:49 [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Collaborative Linkage Study of Autism (2001) An autosomal genomic screen for autism. Am J Med Genet 105:609–615 [DOI] [PubMed] [Google Scholar]
  19. Cook EH Jr, Courchesne RY, Cox NJ, Lord C, Gonen D, Guter SJ, Lincoln A, Nix K, Haas R, Leventhal BL, Courchesne E (1998) Linkage-disequilibrium mapping of autistic disorder, with 15q11-13 markers. Am J Hum Genet 62:1077–1083 [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Dhossche D, Applegate H, Abraham A, Maertens P, Bland L, Bencsath A, Martinez J (2002) Elevated plasma gamma-aminobutyric acid (GABA) levels in autistic youngsters: stimulus for a GABA hypothesis of autism. Med Sci Monit 8:PR1–PR6 [PubMed] [Google Scholar]
  21. Folstein SE, Rosen-Sheidley B (2001) Genetics of autism: complex aetiology for a heterogeneous disorder. Nat Rev Genet 2:943–955 [DOI] [PubMed] [Google Scholar]
  22. Fombonne E (1999) The epidemiology of autism: a review. Psychol Med 29:769–786 [DOI] [PubMed] [Google Scholar]
  23. ——— (2003) Epidemiological surveys of autism and other pervasive developmental disorders: an update. J Autism Dev Disord 33:365–382 [DOI] [PubMed] [Google Scholar]
  24. Hahn LW, Ritchie MD, Moore JH (2003) Multifactor dimensionality reduction software for detecting gene-gene and gene-environment interactions. Bioinformatics 19:376–382 [DOI] [PubMed] [Google Scholar]
  25. Haynes C, Speer MC, Peedin M, Roses AD, Haines JL, Vance JM, Pericak-Vance MA (1995) PEDIGENE: a comprehensive data management system to facilitate efficient and rapid disease gene mapping. Am J Hum Genet 57:A193 [Google Scholar]
  26. Horvath S, Xu X, Lake SL, Silverman EK, Weiss ST, Laird NM (2004) Family-based tests for associating haplotypes with general phenotype data: application to asthma genetics. Genet Epidemiol 26:61–69 [DOI] [PubMed] [Google Scholar]
  27. Hussman JP (2001) Suppressed GABAergic inhibition as a common factor in suspected etiologies of autism. J Autism Dev Disord 31:247–248 [DOI] [PubMed] [Google Scholar]
  28. International Molecular Genetic Study of Autism Consortium (1998) A full genome screen for autism with evidence for linkage to a region on chromosome 7q. Hum Mol Genet 7:571–578 [DOI] [PubMed] [Google Scholar]
  29. ——— (2001) A genomewide screen for autism: strong evidence for linkage to chromosomes 2q, 7q, and 16p. Am J Hum Genet 69:570–581 [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Jentsch TJ, Stein V, Weinreich F, Zdebik AA (2002) Molecular structure and physiological function of chloride channels. Physiol Rev 82:503–568 [DOI] [PubMed] [Google Scholar]
  31. Liu J, Nyholt DR, Magnussen P, Parano E, Pavone P, Geschwind D, Lord C, Iversen P, Hoh J, Ott J, Gilliam TC (2001) A genomewide screen for autism susceptibility loci. Am J Hum Genet 69:327–340 [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Lord C, Pickles A, McLennan J, Rutter M, Bregman J, Folstein S, Fombonne E, Leboyer M, Minshew N (1997) Diagnosing autism: analyses of data from the Autism Diagnostic Interview. J Autism Dev Disord 27:501–517 [DOI] [PubMed] [Google Scholar]
  33. Ma L, Song L, Radoi GE, Harrison NL (2004) Transcriptional regulation of the mouse gene encoding the α-4 subunit of the GABAA receptor. J Biol Chem 279:40451–40461 [DOI] [PubMed] [Google Scholar]
  34. Martin ER, Bass MP, Gilbert JR, Pericak-Vance MA, Hauser ER (2003a) Genotype-based association test for general pedigrees: the genotype-PDT. Genet Epidemiol 25:203–213 [DOI] [PubMed] [Google Scholar]
  35. Martin ER, Bass MP, Hauser ER, Kaplan NL (2003b) Accounting for linkage in family-based tests of association with missing parental genotypes. Am J Hum Genet 73:1016–1026 [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Martin ER, Menold MM, Wolpert CM, Bass MP, Donnelly SL, Ravan SA, Zimmerman A, Gilbert JR, Vance JM, Maddox LO, Wright HH, Abramson RK, DeLong GR, Cuccaro ML, Pericak-Vance MA (2000a) Analysis of linkage disequilibrium in gamma-aminobutyric acid receptor subunit genes in autistic disorder. Am J Med Genet 96:43–48 [DOI] [PubMed] [Google Scholar]
  37. Martin ER, Monks SA, Warren LL, Kaplan NL (2000b) A test for linkage and association in general pedigrees: the pedigree disequilibrium test. Am J Hum Genet 67:146–154 [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Mei H, Ma DQ, Ashley-Koch AE, Martin ER. Extension of multifactor dimensionality reduction for identifying multi-locus effects in the GAW14 simulated data. BMC Genet (in press) [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Menold MM, Shao Y, Wolpert CM, Donnelly SL, Raiford KL, Martin ER, Ravan SA, Abramson RK, Wright HH, DeLong GR, Cuccaro ML, Pericak-Vance MA, Gilbert JR (2001) Association analysis of chromosome 15 gabaa receptor subunit genes in autistic disorder. J Neurogenet 15:245–259 [DOI] [PubMed] [Google Scholar]
  40. Meyers RM, Risch N, Spiker D, Lotspeich L (1998) A full genome screen for susceptibility genes to autism by linkage analysis in ninety multiplex families. Am J Hum Genet 63:A302 [Google Scholar]
  41. Moore JH (2003) The ubiquitous nature of epistasis in determining susceptibility to common human diseases. Hum Hered 56:73–82 [DOI] [PubMed] [Google Scholar]
  42. ——— (2005) A global view of epistasis. Nat Genet 37:13–14 [DOI] [PubMed] [Google Scholar]
  43. Moore JH, Williams SM (2002) New strategies for identifying gene-gene interactions in hypertension. Ann Med 34:88–95 [DOI] [PubMed] [Google Scholar]
  44. Muhle R, Trentacoste SV, Rapin I (2004) The genetics of autism. Pediatrics 113:e472–e486 [DOI] [PubMed] [Google Scholar]
  45. Nielsen DM, Ehm MG, Weir BS (1998) Detecting marker-disease association by testing for Hardy-Weinberg disequilibrium at a marker locus. Am J Hum Genet 63:1531–1540 [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Philippe A, Martinez M, Guilloud-Bataille M, Gillberg C, Rastam M, Sponheim E, Coleman M, Zappella M, Aschauer H, Van Maldergem L, Penet C, Feingold J, Brice A, Leboyer M, Paris Autism Research International Sibpair Study (1999) Genome-wide scan for autism susceptibility genes. Hum Mol Genet 8:805–812 [DOI] [PubMed] [Google Scholar]
  47. Pickles A, Bolton P, Macdonald H, Bailey A, Le Couteur A, Sim CH, Rutter M (1995) Latent-class analysis of recurrence risks for complex phenotypes with selection and measurement error: a twin and family history study of autism. Am J Hum Genet 57:717–726 [PMC free article] [PubMed] [Google Scholar]
  48. Pritchard JK (2001) Are rare variants responsible for susceptibility to complex diseases? Am J Hum Genet 69:124–137 [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Pujana MA, Nadal M, Guitart M, Armengol L, Gratacos M, Estivill X (2002) Human chromosome 15q11-q14 regions of rearrangements contain clusters of LCR15 duplicons. Eur J Hum Genet 10:26–35 [DOI] [PubMed] [Google Scholar]
  50. Risch N, Spiker D, Lotspeich L, Nouri N, Hinds D, Hallmayer J, Kalaydjieva L, et al (1999) A genomic screen of autism: evidence for a multilocus etiology. Am J Hum Genet 65:493–507 [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Ritchie MD, Hahn LW, Roodi N, Bailey LR, Dupont WD, Parl FF, Moore JH (2001) Multifactor-dimensionality reduction reveals high-order interactions among estrogen-metabolism genes in sporadic breast cancer. Am J Hum Genet 69:138–147 [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Sabaratnam M, Turk J, Vroegop P (2000) Case report: autistic disorder and chromosomal abnormality 46, XX duplication (4) p12-p13. Eur Child Adolesc Psychiatry 9:307–311 [DOI] [PubMed] [Google Scholar]
  53. Shao Y, Cuccaro ML, Hauser ER, Raiford KL, Menold MM, Wolpert CM, Ravan SA, Elston L, Decena K, Donnelly SL, Abramson RK, Wright HH, DeLong GR, Gilbert JR, Pericak-Vance MA (2003) Fine mapping of autistic disorder to chromosome 15q11-q13 by use of phenotypic subtypes. Am J Hum Genet 72:539–548 [DOI] [PMC free article] [PubMed] [Google Scholar]
  54. Shao Y, Wolpert CM, Raiford KL, Menold MM, Donnelly SL, Ravan SA, Bass MP, McClain C, von Wendt L, Vance JM, Abramson RH, Wright HH, Ashley-Koch A, Gilbert JR, DeLong RG, Cuccaro ML, Pericak-Vance MA (2002) Genomic screen and follow-up analysis for autistic disorder. Am J Med Genet 114:99–105 [DOI] [PubMed] [Google Scholar]
  55. Smith M, Filipek PA, Wu C, Bocian M, Hakim S, Modahl C, Spence MA (2000) Analysis of a 1-megabase deletion in 15q22-q23 in an autistic patient: identification of candidate genes for autism and of homologous DNA segments in 15q22-q23 and 15q11-q13. Am J Med Genet 96:765–770 [DOI] [PubMed] [Google Scholar]
  56. Sparrow SS, Balla D, Cicchetti D (1984) Vineland Adaptive Behavior Scales, interview edition. AGS Publishing, Circle Pines, MN [Google Scholar]
  57. Steffenburg S, Gillberg C, Hellgren L, Andersson L, Gillberg IC, Jakobsson G, Bohman M (1989) A twin study of autism in Denmark, Finland, Iceland, Norway, and Sweden. J Child Psychol Psychiatry 30:405–416 [DOI] [PubMed] [Google Scholar]
  58. Sullivan PF, Eaves LJ, Kendler KS, Neale MC (2001) Genetic case-control association studies in neuropsychiatry. Arch Gen Psychiatry 58:1015–1024 [DOI] [PubMed] [Google Scholar]
  59. Thornton-Wells TA, Moore JH, Haines JL (2004) Genetics, statistics and human disease: analytical retooling for complexity. Trends Genet 20:640–647 [DOI] [PubMed] [Google Scholar]
  60. Vance JM (1998) The collection of biological samples for DNA analysis. In: Haines JL, Pericak-Vance MA (eds) Approaches to gene mapping in complex human diseases. Wiley-Liss, New York [Google Scholar]
  61. Veenstra-VanderWeele J, Christian SL, Cook EH Jr (2004) Autism as a paradigmatic complex genetic disorder. Annu Rev Genomics Hum Genet 5:379–405 [DOI] [PubMed] [Google Scholar]
  62. Wolpert CM, Menold MM, Bass MP, Qumsiyeh MB, Donnelly SL, Ravan SA, Vance JM, Gilbert JR, Abramson RK, Wright HH, Cuccaro ML, Pericak-Vance MA (2000) Three probands with autistic disorder and isodicentric chromosome 15. Am J Med Genet 96:365–372 [DOI] [PubMed] [Google Scholar]
  63. Yeargin-Allsopp M, Rice C, Karapurkar T, Doernberg N, Boyle C, Murphy C (2003) Prevalence of autism in a US metropolitan area. JAMA 289:49–55 [DOI] [PubMed] [Google Scholar]
  64. Yonan AL, Alarcon M, Cheng R, Magnusson PKE, Spence SJ, Palmer AA, Grunn A, Juo SHH, Terwilliger JD, Liu JJ, Cantor RM, Geschwind DH, Gilliam TC (2003) A genomewide screen of 345 families for autism-susceptibility loci. Am J Hum Genet 73:886–897 [DOI] [PMC free article] [PubMed] [Google Scholar]
  65. Zaykin D, Zhivotovsky L, Weir BS (1995) Exact tests for association between alleles at arbitrary numbers of loci. Genetica 96:169–178 [DOI] [PubMed] [Google Scholar]

Articles from American Journal of Human Genetics are provided here courtesy of American Society of Human Genetics

RESOURCES