Background
Autism spectrum disorders (ASD) are a constellation of neurodevelopmental disorders characterized by the deficits in social reciprocity and language/communication ability, and the presence of restricted interests and repetitive behaviors [
1]. The prevalence of ASD was estimated approximately as 1 per 110 children, with a male-to-female ratio of approximately 4:1 [
2,
3]. Genetic factors have been found to play an important role in the etiology of ASD [
4‐
6].
Chromosome 15q11-q13 is a hot region of occurrence of genomic DNA deletions and duplications that are usually associated with developmental disorders including ASD [
7‐
9]. For example, deletion of paternal segment 15q11.2-q12 is associated with Prader-Willi syndrome that is characterized by obesity, short stature, and hypotonia, while deletion of maternal segment of 15q11-13 is associated with Angelman syndrome which is characterized by mental retardation, movement disorder, and impaired language and speech development. Both Angelman and Prader-Willi syndromes are liable to have ASD [
10]. Moreover, maternal duplication of 15q11-q13 was found in approximately 1 to 3% of patients with ASD [
11]. Hence, genes located at this region have been considered to be potential candidate genes of ASD.
Gamma-aminobutyric acid (GABA) is the main inhibitory neurotransmitter in the brain. A cluster of GABA
A receptor subunit genes, including
GABRB3,
GABRA5, and
GABRG3, which encode subunits β3, α5, and γ3, respectively, were mapped to chromosome 15q12 [
8]. Several lines of study indicate that an altered GABAergic signaling pathway is associated with the pathogenesis of autism [
12]. For example, reduced expression of GABA
A receptor subunits including
GABRB3[
13‐
17] and the GABA synthesizing enzymes, glutamic acid decarboxylase (GAD) 65 and 67 were found in several brain regions of patients with autism [
18‐
20]. Mori and colleagues reported dramatically reduced GABA
A receptor binding in the superior and medial frontal cortex of patients with ASD using
123I-iomazenil (IMZ) single photon emission computed tomography [
21]. These data render GABA
A receptor subunit genes potential candidate genes of autism.
Several genetic linkage and association studies have investigated the association of the three GABA
A receptor subunit genes at 15q11-13 with ASD. Cook and colleagues first reported linkage disequilibrium (LD) between autism and a genetic marker at
GABRB3[
22]. This finding was replicated by some studies [
23‐
25], but not by others [
26‐
29]. Menold and colleagues found two genetic markers in the
GABRG3 associated with autism [
30]. McCauley and colleagues conducted a LD analysis of genetic markers spanning the 1-Mb of 15q12; they found six markers across
GABRB3 and
GABRA5 nominally associated with autism [
31]. In view of the clinical heterogeneity of patients with ASD, several groups studied the genetic association of these GABA
A receptor subunit genes with subsets of ASD patients based on particular phenotypes. For examples, Shao and colleagues reported increased linkage of
GABRB3 locus with autism in families sharing the high insistence-on-sameness scores [
32]. Similarly, Nurmi and colleagues reported improved linkage of
GABRB3 with autism in subset families based on savant skills [
33]. Warrier and colleagues examined the association between 45 SNPs in
GABRB3 and Asperger syndrome; they found significant association of three SNPs with Asperger syndrome and multiple related endophenotypes of ASD [
34]. Furthermore, Ma and colleagues investigated the genetic association of 14 known GABA receptor subunit genes and their interaction with autism. They concluded that the genetic interaction between
GABRA4 and
GABRB1 increased the risk of autism [
35]. Investigating the interaction between the markers in four GABA
A receptor subunit genes in an Argentinean sample of ASD, Sesarini and colleagues found that the genetic interaction between
GABRB3 and
GABRD was associated with an increased risk of autism [
36]. However, Ashley-Koch and colleagues investigated the multi-locus effect of three GABA
A receptor subunit genes at 15q12 on autism but they did not find any positive association [
37]. Atypical sensory sensitivity is one of the core features of patients with autism [
38] and Tavassoli and colleagues found an association between genetic markers of
GABRB3 and tactile sensitivity in typically developing children, implicating the involvement of
GABRB3 in the atypical sensory sensitivity in autism spectrum conditions [
39]. In addition, postmortem studies showed reduced
GABRB3 expression in patients with autism [
15,
16]. Taken together, converging evidence from these studies supports the idea that
GABRB3 may be an important candidate gene of ASD.
One study reported that mice deleted for all three subunit genes (Gabrg3, Gabra5, and Gabrb3) mostly died at birth with a cleft palate, and approximately 5% that survived exhibited neurological abnormalities. However, mice lacking the expression of Gabra5 or Gabrg3 did not show the neurological symptoms found in the mice lacking the three genes [
40]. Furthermore, mice with deletion or reduction of Gabra5 showed enhanced learning and memory [
41,
42]. Mice lacking the
Gabrb3 had epilepsy phenotype and many behavioral abnormalities such as deficits in learning and memory, poor motor skills, hyperactivity, and a disturbed rest-activity cycle [
43].
Gabrb3 deficient mice also manifested a wide range of neurochemical, electrophysiological, and behavioral abnormalities that overlapped with the traits observed in ASD [
44]. DeLorey and colleagues found that
Gabrb3 deficient mice exhibited impaired social and exploratory behaviors, deficits in non-selective attention and hypoplasia of cerebellar vermal lobules [
45]. Hence, the phenotype of
Gabrb3 deficient mice was considered to represent a potential model of ASD [
45]. Duplication of chromosome 15q11-q13 accounts for approximately 1 to 3% of autism cases [
11]. A mouse model of 15q11-13 duplication showed several behavioral abnormalities that replicate various aspects of human autistic phenotypes [
46]. However, the relevance of
Gabrb3 to the behavioral phenotypes has not yet been addressed in this animal model.
Prompted by these findings, we were interested to know whether
GABRB3 was associated with ASD in our population. The study specifically aimed to investigate whether there are variants at
GABRB3 that may confer increased risk to ASD in our population. To address this issue, we conducted deep sequencing of 1.6 Kb of the 5′ region and all exons and their flanking sequences of
GABRB3 in a sample of patients with ASD and control subjects from Taiwan using Sanger-sequencing. The
GABRB3 [GenBank: NG_012836.1] spans approximately 230 Kb at chromosome 15q12, and contains 10 exons. The first two exons (exon 1a and exon 1) are alternatively spliced exons of
GABRB3 that encode an open reading frame of 473 amino acids of isoform 2 and isoform 1, respectively. Although both isoforms have the same amino acid numbers, they have different amino acid sequences in their N-terminus. The sequencing approach basically followed the study conducted by Urak and colleagues with slight modification in the primer sequences [
47].
Methods
Subjects
All the subjects enrolled into this study were Han Chinese from Taiwan. Patients with a clinical diagnosis of ASD made by board-certificated child psychiatrists according to the
Diagnostic and Statistical Manual of Mental Disorders-IV (DSM-IV) [
1] were recruited from the Children’s Mental Health Center, National Taiwan University Hospital (NTUH), Taipei, Taiwan; the Department of Psychiatry, Chang Gung Memorial Hospital (CGMH), Taoyuan, Taiwan; and Taoyuan Mental Hospital (TMH), Taoyuan, Taiwan. The clinical diagnosis of ASD was further confirmed by interviewing the caregivers (mainly parents) by qualified child psychiatrists using the Chinese version of the Autism Diagnostic Interview-Revised (ADI-R) [
48,
49]. The Chinese version of the ADI-R, translated into Mandarin by Gau and colleagues [
48,
49], was approved by Western Psychological Services in 2007. Their parents also reported their autistic symptoms on the Chinese version of the Social Responsiveness Scale (SRS), a 65-item rating scale with each item rated from 1 to 4 [
50], and the Chinese version of the Social Communication Questionnaire (SCQ), a 40-item rating scale with each item rated as ‘yes’ or ‘no’ [
49].
A total of 356 patients (312 boys and 44 girls, mean age: 8.84 ± 4.05 years) were recruited into this study. The ADI-R interviews revealed the 356 patients scored 20.65 ± 5.97 in the ‘qualitative abnormalities in reciprocal social interaction’ (cut-off = 10), 14.99 ± 4.21 in the ‘qualitative abnormalities in communication, verbal’ (cut-off = 8), 8.31 ± 3.25 in the ‘qualitative abnormalities in communication, nonverbal’ (cut-off = 7), and 14.99 ± 4.21 in the ‘restricted, repetitive and stereotyped patterns of behaviors’ (cut-off = 3). Ninety-five point one percent of the patients had abnormal development evident before 30 months of age. Among the 356 subjects with ASD, 17 have been diagnosed with epilepsy (4.78%), 5 have been suspected of seizure (1.40%), and 22 have had febrile convulsions (6.18%). Thirteen of them (3.65%) had been currently diagnosed with epilepsy.
The parents of patients with ASD were clinically assessed by well-trained interviewers who major in psychology and those whose children with ASD had genetic findings received psychiatric interviews by the corresponding author to confirm whether they had ASD. They also reported on the Chinese version of the Autism Spectrum Quotient (AQ) [
51] to screen for any autistic trait and the Chinese version of Adult Self-Report Inventory-IV (ASRI-IV) [
52] to screen for any psychiatric symptoms according to the DSM-IV diagnostic criteria.
A total of 386 unrelated control subjects (189 males, 197 females, mean age: 45.8 ± 12.5 years) were recruited from the Department of Family Medicine of Buddhist Tzu Chi General Hospital, Hualien, Taiwan. The current mental status and history of mental disorders of the control subjects were evaluated by a senior psychiatrist using the Mini-International Neuropsychiatric Interview [
53]. Those with current or past history of mental disorders were excluded from the study. We did not check whether they had a family history of mental disorders.
The study protocol was approved by the Ethics Committee of each hospital (NTUH: 9561709027; CGMH: 93-6244; TMH: C20060905; ClinicalTrials.gov number, NCT00494754) and written informed consent was obtained from the participants and/or parents after full explanation of the protocol and reassurance of confidentiality and voluntary participation. Due to ethical consideration and human subject protection, we were not able to recruit gender- and age-matched control subjects into this study.
PCR-based direct sequencing
Genomic DNA was prepared from the peripheral blood of each participant for PCR amplification. Thirteen amplicons that cover 1.6 Kb of the 5′ region and 10 exons of the
GABRB3 were generated from each individual and these amplicons were subjected to PCR-based autosequencing using ABI autosequencer 3730 (PerkinElmer Applied Biosystems, Foster City, CA, USA). Approximately 30 to 60 bp of the intronic region flanking the exon-intron junction of each exon were sequenced. All mutations identified in this study were confirmed by repeating PCR and sequencing. The primer sequences, optimal PCR conditions, and the size of amplicons are listed in the Additional file
1, and the locations of these amplicons are illustrated in the Additional file
2. The allele frequency of the variant greater than 1% was defined as common variation, whereas that less than 1% was defined as rare variation in this study.
Reporter gene activity assay
Genomic DNA prepared from peripheral blood cells was used to construct the inserts for the reporter gene assay. For common and rare variants at the 1.6 Kb of the 5′ region, a sense primer containing the KpnI recognition sequence and an antisense primer containing the XhoI recognition sequence were used to PCR amplify the fragment from nucleotide positions −1,646 to −46 upstream to the ATG starting site of GABRB 3 exon 1a (genomic DNA positions: chromosome 15: 27018917-27020517). The 1.6 Kb amplicon was first cloned into the pCR-Blunt II-TOPO vector (Invitrogen, Carlsbad, CA, USA), then subcloned into the pGL3-enhancer vector (Promega, Madison, WI, USA) using In-Fusion HD cloning kit (Clontech, Mountain View, CA, USA). For the mutation g.-53G > T at exon 1, a sense primer containing the KpnI recognition sequences and an antisense primer containing the XhoI recognition sequences were used for PCR amplification of a fragment of 597 bp from genomic DNA nucleotide positions: chromosome 15: 27017866-27018462. The amplicon was first cloned into the pCR-Blunt II-TOPO vector (Invitrogen, Carlsbad, CA, USA), then subcloned into the pGL3-enhancer vector (Promega, Madison, WI, USA). The authenticity of these clones was verified by Sanger sequencing.
Plasmids were transfected into an HEK293 or SKNSH cell line in 24-well plates using Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA) following the manufacturer’s protocol manufacturer. At 30 hours after transfection, cells were lysed and the luciferase activities were measured and normalized against Renilla luciferase using the Dual-Luciferase Reporter Assay System (Promega, Madison, WI, USA). Transfection of each plasmid construct was conducted in quadruplicate in each reporter gene experiment, and the reporter gene experiment was repeated three times. The ratio of firefly luciferase to Renilla luciferase was used to represent the normalized reporter gene activity of each construct. Comparison of reporter gene activity among different expression constructs was conducted using one-way analysis of variance (ANOVA) with post-hoc Tukey test. P < 0.05 was considered statistically significant.
Statistical analysis
Deviation from the Hardy-Weinberg equilibrium was assessed by the Chi-square test. Genetic Power Calculator (
http://pngu.mgh.harvard.edu/~purcell/gpc/) was used to perform a
post-hoc power analysis. Linkage disequilibrium (LD) analysis was performed using Haploview version 4.2 [
54] in which D’ was calculated using the method reported by Lewontin [
55] and haplotype block was defined by the method described by Gabriel and colleagues [
56]. Differences in allele, genotype, and estimated haplotype frequencies between patients and controls were evaluated using SHEsis (
http://analysis.bio-x.cn/SHEsisMain.htm) [
57]. Comparison of reporter gene activity among different expression constructs was conducted using one-way ANOVA with
post-hoc Tukey test.
Discussion
In this study, we identified a total of six common known SNPs in this sample; however, no association of these SNPs with ASD was detected. Due to the small sample size of this study, the study has only approximately 34 to 37% power to detect the association under the assumptions of multiplicative inheritance mode and the genotype relative risk of 1.2. But in view of the clinical heterogeneity of ASD, the possible association of these SNPs with some subsets of patients cannot be ruled out.
During the sequencing experiment, we detected only three common SNPs (rs4906902, rs8179184, rs20317) at the 1.6 Kb of the 5′ regulatory region in our sample. Urak and colleagues screened a sample of patients with childhood absence epilepsy for mutations in the 10 exons and the 5′ regulatory sequences of
GABRB3. They found a total of 13 SNPs from the 5′ regulatory region to the beginning of intron 3. Four different haplotypes derived from these SNPs. Further, they found a SNP at the 5′ regulatory region which changed the binding of a neuron-specific transcriptional activator N-Oct-3 and showed different promoter activities in a reporter gene assay [
47]. Several SNPs at this region, such as rs4273008, rs4243768, rs7171660, and rs4906901, as reported by Urak and colleagues in Austrians [
47], were not detected in our sample. In addition, the rs4363842 as reported in Mexican and people in Honduras [
58] was not found in the study of Urak and colleagues [
47] and in our study. The three SNPs found in our sample also showed variations in frequency in different populations according to the dbSNP. Taken together, these findings suggest that there might be a population stratification of SNPs at the regulatory region of
GABRB3, which should be taken into consideration when conducting a case-control association study of this gene. Among these three SNP, the rs20317 was shown to be located at the core promoter region of the
GABRB3 and the rs4906902 was located at the enhancer region. The C allele of rs20317 has been shown to have a significantly increased luciferase activity in a reporter assay [
58].
A total of 22 rare variants were detected in this sample, including four variants (IVS1a + 10G > A, IVS1a + 17C > T, IVS1-3C > T, and IVS2-13G > C) located at intronic regions (Figure
1). The functional impact of these variants was predicted using bioinformatic analysis, and the results are listed in Additional file
3. The frequency of rare variants in the patient group was significantly higher than that in the control group, suggesting that rare mutations of
GABRB3 might be associated with ASD. Of note, all the rare variants at the 5′ regulatory region were detected in the patient group only in this study, with a significantly higher frequency in the patient group than that in the control group. This finding also implies that altered
GABRB3 gene expression might be involved in the neurobiology of ASD. The idea was partly supported by the reporter gene activity assay in this study. Four rare variants (g-1528T > C, g-1442G > A, g.-142G > T, and g-140A > T) at the 5′ regulatory region were shown to have elevated reporter gene activity compared to the wild type alleles. The results are compatible with our bioinformatic analysis that showed gain of transcriptional binding sites of these four variants except g.-142G > T. In addition to the gain of a LBP-1/(+)TCTGG, the g.-142G > T was also predicted to lead to a loss of an AP-2/(+)CCGCCACGGC binding site (Additional file
3). Our data are also compatible with the increased Gabrb3 expression seen in the chromosomal-engineered mouse model for human 15q11-13 duplication of autism [
46].
GABRB3 is biallelically expressed in control brain tissue samples. One study showed that the expression of
GABRB3 was subject to epigenetic alterations that resulted in monoallelic expression in a subset of autism [
59]. A recent study reported significantly increased variance of
GABRB3 expression in the brain of patients with 15q11-13 duplication compared to control subjects and patients with autism [
15]. Nevertheless, our findings are different from several postmortem studies that showed reduced
GABRB3 expression in the brains of patients with autism [
14‐
16,
60]. Hence, the clinical relevance of our findings in this study remains to be clarified.
In this study, we identified three missense mutations (T186M, P336S, and Y402H) and two synonymous mutations (T17T and F314F). The T17T is a new variant that was detected in one patient and two controls in this study, and appears to be neutral according to the bioinformatic analysis. The F314F was detected in only one patient but not in the controls in this study. Lachance-Touchette and colleagues reported the detection of the F314F synonymous variant in 1 out of 183 patients with idiopathic generalized epilepsy (IGE), but not in the 190 controls [
61]. The authors suggested this variant might be pathological to IGE. This variant may influence the binding of the serine/arginine-rich splicing factor 6 (SRSF6) to an exonic splicing enhancer according to bioinformatic analysis conducted in this study, but the clinical relevance of this variant to ASD remains to be explored. The three missense mutations identified in this study, which occurred in the evolutionarily conserved region of
GABRB3, may affect the secondary protein structure and alter a phosphorylation-based substrate motif or a phosphorylation-based binding motif. These findings suggest that these three missense mutations might have a functional impact on
GABRB3.
Family studies revealed that almost all the rare variants found in patients were transmitted from their parents, and not all the carriers of rare variants met the clinical diagnosis of ASD, suggesting these rare variants are more likely to be risk factors rather than causative factors of ASD. Notably in the family of Figure
3g, the affected male patient (U985) carried two variants (g-142G > T and g.-140A > T) at the 5′ region of the
GABRB3 that were transmitted from his mother and father respectively, lending the evidence to support the multi-hit model of ASD [
62,
63].We also identified a novel mutant allele that harbored multiple mutations at the 5′ regulatory region and the T186M at exon 6 in the patient U1452 and his father (Figure
3b). Although the father did not meet the DSM-IV diagnostic criteria of ASD, he manifested autistic trait, suggesting this mutant allele might have a contributing effect to the pathogenesis of ASD. The T186M was predicted to be probably damaging in bioinformatic analysis. As the T186M was linked to the multiple variants at the 5′ regulatory region, and the reporter gene assay showed no significant differences in the reporter gene activity between the multiple regulatory variant alleles and the wild type constructs, the T186M alone might be the risk variant of ASD.
Missense mutations of
GABRB3 have been reported to be associated with childhood absence epilepsy (P11S, S15F, G32R) [
64,
65], insomnia (R192H) [
66], and autism (P11S) [
67]. In this study, we did not detect these missense mutations in our samples. In addition, among our patients with missense mutations, only the patient U1452 who carried the T186M mutation had a history of abnormal electroencephalography (EEG) in his childhood. The other patients did not have a history of seizure. These findings indicate that different mutations of
GABRB3 may confer different clinical presentations of neurodevelopmental disorders in their carriers [
68].
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made.
The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.
Competing interests
The authors declare no competing interests.
Authors’ contributions
SSG is the principle investigator in this project. CHC and SSG designed the study and wrote the protocol. SSG trained the clinical research team, supervised in research execution, and collected all the clinical data of the ASD cases. SSG and YYW were responsible for the ADI-R training and interviews. SSG, YYW, YNC, WCT and SKL helped recruit and evaluate the patients and CHC screened for mental disorders in the controls. CCH and MCC conducted the experimental works and CHC supervised the experimental works and analyzed the data. CCH prepared the first draft, CHC and SSG critically revised the manuscript. All authors reviewed the article and approved its publication.