Background

Newer genomic technologies like high-resolution microarrays and next generation exome sequencing have enabled the identification of many clinically relevant genetic variants for both Mendelian and complex disorders. Yet for many conditions the identified genes account for only a proportion of heritability. This observation coupled with the recognition of the functional relevance of non-genic regions [1] target these genomic segments as candidates for investigation for a role in disease.

ASD encompasses a range of neurodevelopmental disorders characterised by social impairment, communication difficulties and restricted, repetitive behavioural patterns. ASD, which is clinically and genetically heterogeneous, demonstrates high heritability, familial clustering and ~4:1 male to female bias. While there has been progress identifying risk genes, most are still unknown [2]. Analyses of rare (<1% population frequency) CNVs, insertions and deletions (indels) and point mutations have most convincingly identified synaptic genes such as members of the Neuroligin (NLGN3, NLGN4)[3], Neurexin (NRXN1[4], NRXN2[5], NRXN3[6]), SHANK (SHANK1[7], SHANK2[8], SHANK3[9]) families and Gephyrin [10] as highly-penetrant risk loci [2]. ASD subjects with multiple genetic risk factors for ASD and associated medical conditions are also known [11]. In addition, there are a few examples of mutations in ASD cases identified in non-genic segments of DNA [12] and non-coding RNAs [13]. Similar findings are even better documented in studies of intellectual disability [14, 15], which is observed in ~40% of cases of ASD. Focusing on the intergenic intervals of the genome, we performed a systematic genome-wide investigation to identify rare CNVs enriched in cases compared with controls [16] to identify known and novel ASD susceptibility loci.

Methods

A collection of 1491 unrelated ASD cases were genotyped using either the Illumina 1M (993) or the Affymetrix SNP 6.0 platforms (498). The ASD subjects, all diagnosed using gold-standard instruments including Autism Diagnostic Interview and Autism Diagnostic Observation Schedule, are described elsewhere [16, 17]. Informed written consent was obtained from all participants, as approved by the Research Ethics Boards at The Hospital for Sick Children and McMaster University. For controls, 1287 samples from the SAGE cohort were genotyped on with the Illumina 1M and 1234 samples from the Ottawa Heart Institute (OHI) and 1123 from the POPGEN collections were genotyped on the Affymetrix SNP 6.0. CNV discovery was performed using previously described pipelines [16–18]. Three CNV detection tools were used for each platform (Birdsuite, iPattern and Genotyping Console for Affymetrix 6.0 and iPattern, QuantiSNP & PennCNV for Illumina 1 M). A subset of CNVs in both cases and controls were considered rare if they were present in <1% of the overall dataset and these were further analysed if they failed to intersect or fall within a known gene (according to the NCBI Reference Sequence (RefSeq), August 2011). Rare genic CNVs identified from these data have been reported previously and from these data approximately 10% of cases carry a de novo or rare inherited CNV thought to contribute to ASD in that individual [16, 17, 19, 20]. All CNVs discussed were validated where DNA was available using independent laboratory methods such as long range or quantitative PCR and the mode of inheritance determined (Additional files 1 and 2).

Results and discussion

Microarray data from a cohort of 1491 unrelated ASD probands were analysed for rare copy number variants as described previously [16, 17] and CNVs falling outside of known coding sequence were identified. A total of 212 non-coding genomic regions were determined as harboring overlapping CNVs in two or more unrelated ASD cases that were absent in control samples. Each region was examined for plausible biological function by comparison with multiple databases. Data was collated for evidence of expressed sequences from mRNA or EST data at GenBank or evolutionary conservation as well as functional predictions from the VISTA enhancer browser (http://enhancer.lbl.gov/) and Rfam (http://rfam.sanger.ac.uk/). The Database of Genomic Variants (http://dgvbeta.tcag.ca/dgv/app/home) was used to eliminate additional regions as non-ASD specific CNVs and regions with >80% masked as repetitive sequences were removed. Loci were also prioritised as being of potential clinical significance in ASD due to proximity to genes considered known or candidate ASD risk genes [17].

Fifteen intergenic regions emerged as plausible candidate ASD risk loci and in all instances the defining CNV events were inherited. In one of these regions, an additional case (SK0167-003) was found with an overlapping CNV described by Marshall et al. (2008) [19] (Table 1, Figure 1 and Additional files 1 and 2). In 14 of 15, the intergenic interval identified has not been described before and in three regions the CNV neighboured a known ASD gene, namely, CDH8 [21], C3orf58 [22] and NRXN1 [4]. In the case of the NRXN1 gene, upstream CNVs found in five individuals impact the same mRNA (AK127244) reported elsewhere with a CNV in a family with ASD (Table 1, Figure 1A) [23]. Examples of other intergenic CNVs identified highlight regions at 8q24.12 upstream of ZHX2, 6p21.2 upstream of MOCS1, 11p12 upstream of LRRC4C (Figure 1B) and 7q21.11 upstream of SEMA3C, as putative novel ASD rearrangements. In one case (8-14208-3350), deletions were identified at three separate loci; 4q13.1 upstream of EPHA5, 11p14.3 upstream of LUZP2 and 11p12 upstream of LRRC4C and another case (3-0496-003) carried a 46, XXY sex chromosome imbalance. Other CNVs found in these 41 cases are shown in Additional file 3 and any or all of these may be contributing to the genetic load for ASD [11, 17]. Interestingly, all the CNVs identified through our analysis are inherited events. The significance of this observation is still to be determined but suggests incomplete and/or variable penetrance of phenotype, which is something often observed in ASD [6, 7, 17].

Table 1 ASD specific CNVs in intergenic regions
Full size table
Figure 1
figure 1

Genome browser views of ASD specific CNVs at A) 2p16.3 B) 11p12 C) 8p12 and D) 4q13.1. In each case, representative isoforms of known RefSeq genes, mRNA and/or Expressed Sequence Tags are shown. Deletions and duplications are represented by red and blue bars, respectively. In Figure 1A) a dashed line indicates a diploid region located between two adjacent deletions in the same individual. Additional browser views from other loci shown in Table 1 are included in Additional file 1 A-J. In all cases where parental DNA was available, the CNVs shown were found to be inherited. Additional case SK0167-003 found in Marshall et al.[19].

Full size image

The mechanism of action of these rare CNVs in the pathogenesis of ASD could be (i) through altering the necessary copy number or positional context of key DNA sequence elements required for regulating the proper expression of nearby genes [1], (ii) affecting still undiscovered genes or non-coding RNAs residing in the CNV regions and (iii) disrupting uncharacterized isoforms of the adjacent annotated genes. In the first scenario, we find CNVs both upstream (e.g. UNC5D (Figure 1C), MOCS1, ASTN2, SEMA3C, ZHX2, LUZP2, CDH8) and down-stream (C3orf58, RXRA, MRGPRD) of known ASD risk genes and putative novel loci. For at least three regions (4q13.1, 6p21.2 and 11p12 (shown in Figure 1D, Additional file 1C and Figure 1B respectively)), our CNV mapping data in fact identify two distinct clusters of CNVs at the same locus, all overlapping spliced ESTs and thus with a possible regulatory role. Secondly, three independent CNV deletions interrupting a collection of spliced expressed sequenced tags approximately 330 kb proximal to EPHA5 highlight a potentially newly discovered ASD risk gene (Figure 1D). Finally, longer isoforms of LRRC4C likely exist given the discovery of mRNAs DQ084201 and DQ084202. There are, of course, other functional DNA elements or modifications that need to be considered [24] as the mapping resolution increases.

Conclusions

Given the challenges faced in interpreting the clinical significance of multitudes of genetic variants found in for example, whole genome sequencing [25], accruing evidence across multiple studies will advocate loci outside of known genes or other regulatory elements for further study, particularly for rare variants. In this light, these data provide a useful resource for comparison as new data sets of both CNVs and nucleotide-level variants become available to help fine-map additional discover new ASD risk loci. This general research strategy can also be applied to other disease gene studies.