Autism spectrum disorders (ASDs) are pervasive developmental disorders that include autism, Asperger syndrome, and pervasive developmental disorder-not otherwise specified (PDD-NOS). ASDs are characterized by varying degrees of impairment in communication and social interactions, as well as by the presence of restricted, repetitive and/or stereotyped patterns of behavior, interests, and/or activities, with signs typically manifesting by the age of 3. Recent epidemiological studies report a prevalence of ASD of 3–6 per 1000 children 12. Males have a higher risk than females, with a male to female ratio of ~4:1. Twin studies show a concordance rate of 60%–90% for monozygotic twins and 0–10% for dizygotic twins, and the genetic liability for autism is estimated to be more than 90% 3. In spite of the high heritability, the identification of genetic factors in ASDs has proved difficult, due at least in part to the fact that ASDs are characterized by a high degree of genetic heterogeneity.

In the past, approximately 10% of individuals with ASDs could be diagnosed as having secondary autism, in which a chromosome abnormality, single-gene disorder, or known environmental agent was identified as causative 34. However, more recent studies suggest that chromosome abnormalities and single-gene disorders are more common as causes of ASDs 56. Many children with ASD have some degree of cognitive impairment (CI) and genetic disorders associated with CI have also been associated with ASDs. Thus, ASDs have been reported as possible manifestations of genetic syndromes resulting in CI, including fragile X syndrome, tuberous sclerosis, Angelman syndrome, Prader-Willi syndrome, 22q11 deletion/DiGeorge syndrome, Cohen syndrome, Smith-Lemli-Opitz syndrome, Smith-Magenis syndrome, Williams syndrome, Sotos syndrome, and Klinefelter syndrome 78910.

Chromosomal abnormalities throughout the genome have been reported in cases of ASD 11. In addition to traditional methods like FISH to detect chromosome abnormalities, new array-based techniques allow robust whole genome scanning for copy number variants (CNVs) and have resulted in the identification of new disease genes 121314. For example, using Affymetrix 10 K SNP arrays, the Autism Genome Project Consortium identified deletions in the neurexin 1 gene (NRXN1) in 2p16 as a cause for ASD 15, now described in further studies 131617. More recently several groups have reported deletions in 16p11 in ASD 131418. Multiplex ligation-dependent probe amplification (MLPA) is a recently developed method to accurately quantify levels of diverse DNA sequences based on target-directed ligation followed by quantitative polymerase chain reaction (Q-PCR), and can be used to test for the presence of chromosomal abnormalities that lead to changes in copy number. For each sequence of interest, the two probes are joined by the use of a ligase only if they are both bound to a specified target sequence. The resultant ligated product is then amplified by Q-PCR in a manner proportional to the initial levels. MLPA has been proposed as a rapid and cost-effective method for routine screen of chromosome abnormalities in mental retardation 1920. To determine the degree to which chromosome abnormalities associated with CI contribute to autism, and to evaluate MLPA for genetic counselling, 279 unrelated ASD subjects were screened for CI-associated deletions using MLPA.

We made use of four commercial MLPA panels (MR1, MR2, MRX, and SHOX, which included 84 probes directed at autosomal loci and 46 probes directed at X-linked loci, 4 probes at Y-linked loci, and 27 probes directed at PAR1/PAR2), covering many of the chromosomal regions associated with known, recurrent deletions/duplications associated with CI [see Additional file 1].

MLPA screening identified increased or decreased signal in 28 subjects out of the 279 screened, consistent with potential microduplications or microdeletions. We observed potential duplications in 15q11-q13 (4 subjects), 22q11 (2 subjects), the TM4SF2 gene in Xp11 (2 subjects), and the ASMT gene in Xp22.32 (PAR1) (16 subjects). We observed potential deletions in the ASPA gene in 17p13 (1 subject), the PAX6 gene in 11p13 (1 subject), the EXT1 gene in 8q24 (1 subject), and the ARHGEF6 gene in Xq26 (1 subject).

In this study, 279 unrelated subjects with ASDs were investigated for microdeletions and microduplications associated with CI using MLPA. The cohort was ascertained for genetic studies with exclusion criteria including a prior medical diagnosis. Two hundred and fifty-four of the subjects screened were from multiplex families. We made use of four commercial MLPA probe kits, which captured many of the microdeletions and microduplications previously associated with CI and, in some cases, with ASDs. A focus on known microdeletions and microduplications makes the results most relevant to genetic counselling. In general, our study found ~1–2% of the cases had a chromosomal abnormality previously associated with CI. This yield suggests that directed MLPA, expanded to include additional ASD loci, would represent one option for clinical testing as the cost is substantially smaller than genome-wide arrays. Of course, additional methods will be required for the detection of point mutations in the relevant genes as MLPA will not readily detect point mutations.

We identified 4 duplications in chromosome 15, two typical and two atypical. Chromosome 15q11-q13 duplications are the most common alteration in autism in prior studies 1127282930. The most frequently reported pattern of 15q11-q13 duplications in autism is supernumerary marker chromosome 15 or isodicentric 15 [idic(15)], with two or more extra copies of this region including the PWA critical region and variable in size 2829. In this study, we did not find any individuals with this type of duplication, probably because these abnormalities are identified by karyotyping, and such individuals would have been excluded in the cohort. All four duplications we found were interstitial microduplications. The extent of these four duplications was confirmed by dense MLPA probes. Two of these duplications corresponded to the well-defined duplication covering the PWA critical region between breakpoints 1 and 3 and have been reported in many individuals with autism or ASD 2425. This region is subject to imprinting and maternal duplications are associated with ASD, while paternal duplications usually have a normal phenotype 30. The two novel 15q11-q13 duplications were situated on each side of PWA critical region, on 15q11.2 and 15q12, respectively.

One of the novel duplications (proband 3) is ~1 Mb in size adjacent to the UBE3A gene, which is the gene responsible for Angelman syndrome. The duplication encompasses two potential autism susceptibility genes, GABRB3 and ATP10A. Several association studies have suggested that GABRB3 could be an autism susceptibility gene 313233. For example, we and others previously reported that there is a significant association between a polymorphism of the GABRB3 gene and ASD (and see discussion therein) 33. Another group reported GABRB3 associated with ASD and found evidence for linkage in a subgroup of patients with an elevated score on the "insistence on sameness" factor 34. A recent functional study using ASD brain samples demonstrated that GABRB3 is subject to epigenetic dysregulation: the product of the MECP2 gene, which when mutated can cause Rett syndrome, binds to methylated CpG sites within GABRB3 and acts as a chromatin organizer for optimal expression of both alleles of GABRB3 in neurons 35. ATP10A is a P-type ATPase gene maternally expressed. A linkage disequilibrium study in the 15q11-q13 region found preferential transmission of a haplotype of ATP10A to ASD subjects 36. The de novo origin of this duplication in patient 3 is consistent with a potentially causal role in ASD.

Another novel small duplication (proband 4) was located adjacent to SNRPN, a gene implicated in Prader-Willi syndrome (PWS). Three small genes, MKRN3, MAGEL2 and NDN, are situated within this very circumscribed region. MKRN3 codes a ring finger protein while NDN and MAGEL2 belong to the same NDN/MAGE gene family. All three genes are expressed exclusively from the paternal allele and are thought to be implicated in some clinical features of PWS 373839. MAGEL2 is important in regulating the circadian rhythmicity related to some features of PWS 4041. NDN also plays a role in facilitating the differentiation and specification of GABAergic neurons in cooperation with Dlx homeodomain proteins 42. Again, the de novo origin of the duplication in family 4 is consistent with a potentially causal role in ASD.

In a recent review that diagrammed all the cytogenetic abnormalities reported in cases of autism 11, many cases of 15q11-q13 duplications with variable size overlapped our two novel duplications. Candidate genes in these two causal regions may implicate aberrant glutamate signalling interacting with epigenetic factors in some ASD cases. However, segregation studies complicate simple explanations of causality. In Family 3, the small duplication of ATP10A and GABRB3 was found in one affected male proband but not in his affected sister, whereas in Family 4 the duplication of MKRN3, MAGEL2 and NDN was found in the affected monozygotic twins. However, it is again important to note that both of the novel duplications arose de novo, providing support for a causal or contributory role for these genomic variants to ASD. However, the mechanism by which these novel duplications may contribute to ASD needs further analyses.

Two subjects with a 22q11.2 duplication cognate to the 22q11 deletion syndrome or DiGeorge syndrome were observed. Several studies reported a high rate (20%–50%) of autistic spectrum disorders in children with 22q11.2 deletion syndrome 434445. Recently, the 22q11.2 duplication was found as a highly variable syndrome 2646. The majority of patients with 22q11 duplications have cognitive deficits including speech delay and developmental delay, although the penetrance is variable 47. The 22q11 duplication was also associated with variable presentation in our study. The duplication was found in a male proband with autism but not in an affected brother with broad-spectrum disorder in one case, while it was present in a female proband with autism but not in her autistic brother in the second case. In Family 5, the reportedly healthy father was found as a carrier of duplication. Phenotypically normal parents of 22q11.2 duplication children also were reported in other studies as carriers of this duplication 26. Notably, in Family 5, the duplication was associated with a more severe phenotype (narrow autism versus board spectrum disorder), while in Family 6, a girl (perhaps with a higher threshold for liability) carried the duplication but her autistic brother did not. It is widely recognized that deletion of 22q11 region is related to CI and psychiatric symptoms, including delayed motor and speech-language development, mental retardation, impaired spatial reasoning, attention-deficit hyperactivity disorder, autism spectrum disorders, mood disorders, and/or schizophrenia spectrum disorders 4849. Multiple genes within this causal region have been identified playing a role in neuronal cognitive development 505152, with TBX1 implicated in ASD 52. Thus, the 22q11.2 duplication might also function as a risk factor for ASD and increase the liability or severity in ASD individuals.

ASDs can share characteristics with Williams-Beuren syndrome and a recent study of 128 children with WBS identified nine with ASD 5354. Duplications of the WBS critical region in 7q11.23 have also been described in several patients with ASD, severe language delay and mental retardation 55. Another study reported a case with a diagnosis of ASD and an atypical deletion of WBS interval 56. Similarly, a 17p11 deletion can cause Smith-Magenis syndrome, and 17p11 duplication is now recognized as a new syndrome of mental retardation, often associated with autistic features 5758. We did not observe any deletion/duplication in these regions in our sample, suggesting that these gene dosage anomalies are rare contributors to ASDs.

One subject and his monozygotic twin were found to be partially deleted for one allele of aspartoacylase (ASPA) gene, while the second allele was intact. Mutations of ASPA gene cause Canavan disease [MIM608034], an autosomal recessive disorder. Both subjects in our study have a diagnosis of autism, and their mother, determined as unaffected, also carries this deletion. We assume that this deletion is not associated with the ASD phenotype.

Fragile X syndrome is the most common syndrome related to ASD. The prevalence of autism or autistic features in individuals with fragile syndrome is about 25–33%, and the prevalence of fragile X in autism is estimated at 2% 89. Fragile X is a single-gene disorder typically caused by inactivation of the FMR1 gene at Xq28 typically by trinucleotide repeat expansion. In our study, no gene dosage abnormality was found in FMR1 and FMR2, but note that expansions would not be detected by MLPA.

In addition to the FMR genes, other genes related to X-linked mental retardation were screened in this study. TM4SF2 or tetraspanin 7 (TSPAN7) at Xp11.4 was found partially duplicated in two unrelated subjects with autism in our sample. Q-PCR confirmed that the duplication included exons 2 to the down stream of TM4SF2 gene, but not exon 1. Probes of PQBP1 gene on Xp11.23, and probes of ARX, IL1RAPL1 and RPS6KA3 genes on Xp22 showed normal dosage. This indicates that a microduplication exists in Xp11 with a breakpoint within the TM4SF2 gene. The protein encoded by the TM4SF2 gene mediates signal transduction events that play a role in the regulation of cell development, activation, growth and motility 59. It is known to form a complex with integrins 60, and two integrin genes, integrin beta 3 (ITGB3) and integrin beta 4 (ITGB4), were reported associated with ASDs 6162. There is some evidence for a role for TM4SF2 in mental retardation. The first study identified a translocation in this gene in a female patient with mild mental retardation associated with minor autistic features, as well as mutations (a premature stop codon and a missense mutation, both maternally inherited) in 2 out of 33 small families with X-linked mental retardation 60. Note, however, that a subsequent study questioned the role of the gene and particularly one of the missense mutations (P172H) identified in the first study, because further linkage analysis based on cognitive status appeared to exclude the region of the TM4SF2 gene 63. However, this same variant has subsequently been observed in an additional individual with mental retardation 64, but not in 320 controls with similar ethnic background. An additional deletion in the gene has also been identified in mental retardation 65. A recent study reported TM4SF2 duplication in two patients, one with syndromic and the other with nonsyndromic mental retardation 66, but suggested that this duplication might be a neutral polymorphism as there was no evidence for skewed X-chromosome inactivation in the unaffected mothers carrying the duplication. We have now identified this same duplication in both male and female unscreened control subjects, supporting the interpretation that this duplication is a neutral polymorphism.

In individuals with Klinefelter syndrome (47, XXY), levels of ASD traits are significantly higher across all dimensions of the phenotype, which suggests that gene dosage changes on X chromosome might play a role in ASD behaviours 1067. Similarly, individuals with an XYY karyotype have been reported to have delayed speech and language skills or ASD 68. The pseudoautosomal regions (PAR1 and PAR2) are homologous across the X and Y chromosomes. Deletions of these regions have been reported in three females with ASD 69. In addition, the gene for acetyserotonin O-methyltransferase (ASMT), located in PAR1, has been reported to be associated with abnormal melatonin synthesis in ASD 7071.

The PAR1 and PAR2 regions are not well-represented in some arrays used for genome-wide studies and our MLPA results represent the first extensive analysis of these regions in ASD. We found a duplication in the ASMT gene to be significantly associated with ASD in the current sample. Follow-up genetic studies in additional cohorts would be worthwhile to confirm our finding. Further studies on genetic variation in the ASMT gene, the effects of the CNV on gene expression, and the relationship of this CNV to both ASDs and sleep disorders are warranted.

Finally, our study showed that MLPA can be a useful, inexpensive tool to evaluate clinically significant chromosomal microdeletions and microduplications in ASD and associated disorders. It is likely that with a targeted panel of MLPA probes, clinical genetic diagnoses can now be made in over 10% of ASDs cases. MLPA is considerably less expensive when compared to FISH and to aCGH-based method; and it can be efficient even when compared to Q-PCR. MLPA is also an accepted approach in New York State (NYS), which has not yet approved large-scale aCGH-based tests. MLPA can also be used to validate aCGH or microarray findings.

We have identified one limitation of MLPA, in that there can be a loss of signal in the case of a single base change within a probe-binding site. In the current study, three apparent deletions were later identified to be caused by single-base sequence changes in probe binding regions, indicating that MLPA results indicating a potential deletion in a single probe should be followed-up by additional studies. Although none of these variants that we identified were observed in 49 Caucasian controls (data not shown) (nor were they found in the NCBI SNP database), the variants were non coding (two, in PAX6 and ARHGEF6 resulted synonymous amino acid changes, and one was located in the 3' UTR of EXT1), and are variants of unknown significance.

In summary, we used MLPA as a screen and identified known and novel chromosomal abnormalities in ASD. We observed instances of microduplications in 15q11-q13, including two novel microduplications, and microduplications in 22q11.2. We also identified duplications in ASMT, which may contribute to ASD risk or ASD-associated traits. Compared to array based methods, MLPA, with subsequent validation of potential deletions, is a relatively practical, inexpensive and fast tool for screening for chromosome rearrangements in ASDs. It is important to validate deletions identified with MLPA by the use of multiple probes and/or additional technical approaches.

The authors declare that they have no competing interests.

GC was involved in the design and performance of the study and drafted the manuscript. LE and NC performed the FISH studies. JEG was involved in the MLPA experiments. AN was involved in the Q-PCR experiments. JGR was involved in sample preparation and clinical data management. EJH was involved in clinical data management. DMZ, LS, EA, JMS and EH were involved in family ascertainment, with LS, EA, JMS and EH supervising training and recruitment oversight. CB proposed and analyzed experiments and critically revised the manuscript. JDB was involved in the design and analysis of the study, and edited the manuscript.