Article Text

Download PDFPDF

Phenomic determinants of genomic variation in autism spectrum disorders
  1. Y Qiao1,2,
  2. N Riendeau1,
  3. M Koochek1,
  4. X Liu3,5,
  5. Chansonette Harvard2,
  6. M J Hildebrand1,
  7. J J A Holden3,4,5,
  8. E Rajcan-Separovic2,
  9. M E S Lewis1
  1. 1
    Department of Medical Genetics, University of British Columbia (UBC), BC Child and Family Research Institute (CFRI), C234-4500 Oak Street, Vancouver, V6H 3N1, British Columbia, Canada
  2. 2
    Department of Pathology, UBC, CFRI, 4480 Oak Street, Vancouver, V6H 3V4, British Columbia, Canada
  3. 3
    Department of Psychiatry, Queen’s University, 191 Portsmouth Avenue, Kingston, K7M 8A6, Ontario, Canada
  4. 4
    Department of Physiology, Queen’s University, 191 Portsmouth Avenue, Kingston, K7M 8A6, Ontario, Canada
  5. 5
    Autism Research Program, Ongwanada, 191 Portsmouth Avenue, Kingston, K7M 8A6, Ontario, Canada
  1. Correspondence to Professor M E S Lewis, Department of Medical Genetics, The University of British Columbia, and BC Child and Family Research Institute, C234-4500 Oak Street, Vancouver, BC, V6H 3N1, Canada; slewis{at}


Background: Autism spectrum disorders (ASDs) are common, heritable neurobiologic conditions of unknown aetiology confounded by significant clinical and genetic heterogeneity.

Methods: This study evaluated a broad categorisation of phenotypic traits (or phenome) for 100 subjects with Autism Diagnostic Interview-Revised/Autism Diagnostic Observation Schedule-Generic (ADI-R/ADOS-G) confirmed idiopathic ASD undergoing 1 Mb bacterial artificial chromosome (BAC) array comparative genomic hybridisation (CGH).

Results and conclusions: Array CGH uncovered nine different pathogenic copy number variants (pCNVs) in 9/100 ASD subjects having complex phenotypes (ASD± intellectual disability (ID; IQ<70)) and/or physical anomalies), normal karyotype, fragile X analysis, and comprehensive evaluation by a clinical geneticist. Unique pCNVs in our cohort included del(5)(p15.2p15.31) (2.4 Mb), del(3)(p24.3) (0.1 Mb) and dup(18)(p11.3)(0.9 Mb). Five pCNVs were recurrent in our cohort or were previously described in subjects with ASD±ID: (dup(7)(q11.23)(1.5 Mb); del(2)(p15p16.1) (6.1 Mb and 7.9 Mb); del(14)(q11.2) (0.7 Mb) and dup(15)(q11q13) (10 Mb), including del(X)(p11.22) (470 Kb) in two autistic brothers. Male: female distribution in subjects with pCNVs was reduced to 1.25:1 from 3.2:1 in the original cohort. The authors stratified the study population according to a broad spectrum of clinical features and correlated specific phenotypes with respect to CNV load and pathogenicity. The findings indicate increased prevalence of pCNVs in subjects with microcephaly (<2nd centile; n = 2 of 4 ASD subjects with microcephaly; p = 0.04), and ID (n = 9 of 64 subjects with ASD and ID; p = 0.02). Interestingly, in the absence of ID co-morbidity with an ASD, no pCNVs were found. The relationship between parental ages at delivery and CNV load and pathogenicity was also explored.

View Full Text

Statistics from

Autism spectrum disorders (ASDs) are a group of common neurodevelopmental conditions characterised by impairments in communication, social interaction, and behaviour. At least one of every 166 children is likely to be affected by this spectrum of disorders, with a male: female (M:F) ratio of 4:1,1 and it is the most heritable of all complex neuropsychiatric conditions.2 Since ASDs are considered complex genetic disorders, resulting from the interaction of several genes and environmental factors, the lumping together of all cases of ASD, with no subgrouping based on phenotypic characteristics, makes the identification of contributory genes extremely difficult. The fact that autism is known to be associated with several distinct medical/genetic disorders further highlights its genetic heterogeneity. Thus, comprehensive “whole body” phenotyping and more accurate diagnostic methods are necessary to clarify the underlying comorbidities, causes and symptoms of ASDs—inclusive of neurobehavioral, medical and morphologic traits.

A variety of different approaches to identify genes for ASDs have been undertaken, including cytogenetic assessment for chromosome abnormalities, genome scan linkage studies, and association studies (see Freitag3 for review). The overlap of results among different study methods is limited, largely attributed to the significant clinical and genetic heterogeneity of ASDs based on variable behavioural indices, variations among study populations, and limitation of methods for detecting ASD susceptibility genes of mild to moderate effect.

Most recently, array comparative genomic hybridisation (CGH) technology has been used to screen rapidly the genome for pathogenic copy number variants (pCNVs) associated with ASD.4 5 6 7 8 pCNVs are more prevalent in “complex” syndromic ASDs (27.5%) manifesting intellectual disability (ID; IQ<70) and/or dysmorphic features4 than in non-syndromic (simple or idiopathic) cases (7.2%)6 manifesting neurobehavioral features alone. Relatively consistent frequencies have been noted in studying simplex (single incidence; SPX) (7–10%)6 7 8 and multiplex (multiple incidence; MPX) families (2–12%) across different studies.5 6 7 8 While current data suggest that pCNVs contribute to ASD pathogenesis, their role within a growing constellation of ASD microdeletion and microduplication syndromes remains poorly understood, due to the absence of consistent, standardised and comprehensive somatic, medical and neurobehavioral phenotyping of ASD subjects.

To address this, we selected 100 subjects with “complex” ASD scores of ⩾3 based on criteria modified from de Vries et al9 for array CGH screening for CNVs, and here summarise our systematic categorisation of the broad spectrum of clinical features present in those individuals with and without pCNVs. We stratified our findings with CNV type (putatively pathogenic or benign) and total CNV load and reviewed detailed prenatal, medical, developmental and multi-generation family histories, assigning subjects to specific phenotypic subgroups based on comorbidity with: (1) ID (IQ<70); (2) presence of prenatal (ultrasound detected) and/or postnatal growth anomalies; (3) micro- or macrocephaly; (4) epilepsy; (5) craniofacial dysmorphisms; (6) congenital physical or systemic anomalies, (7) pregnancy complications; and (8) neonatal complications. We believe that characterisation of the sum phenomic and genomic determinants of ASD will make the identification of contributory genes possible and the correlation of genetic changes with clinical features more meaningful.

Subjects and methods


In the course of systematic medical genetic evaluation of subjects with an ASD recruited through the research registry of the Autism Spectrum Disorders-Canadian American Research Consortium (ASD-CARC;, 100 subjects with complex features of idiopathic ASD (76 males and 24 females) scoring ⩾3 using phenotype criteria modified from de Vries et al9 were randomly selected for array CGH analysis.

All subjects underwent comprehensive screening of medical systemic and morphological features. ASD diagnoses for all subjects were based on standardised Diagnostic and statistical manual, 4th ed (DSM-IV-TR) criteria using Autism Diagnostic Interview-Revised (ADI-R) and/or Autism Diagnostic Observation Schedule-Generic (ADOS-G) standards,10 11 including a variety of measures of cognitive and adaptive function to assess the presence or absence of ID (Leiter International Performance scale-revised; Stanford Binet Intelligence Scale, 4th ed; Weschler Intelligence Scale, 4th ed; Vineland Adaptive Behaviour Scales, etc).

A standardised metric of clinical characteristics9 was assigned by the evaluating clinical geneticist (MESL) blinded to the CNV findings. Among the 100 subjects studied, 31 were from SPX families, 45 from MPX-immediate families (MPX-I: sharing an ASD with another family member via a 1st degree co-relationship), and 24 from MPX-extended (MPX-E: sharing an ASD with another family member via ⩾2nd degree co-relationship) families. All subjects received routine karyotype testing, including targeted 15q11, 22q11, 22q13, subtelomeric FISH (fluorescence in situ hybridisation) studies, fragile X, and clinical chemistry screening.

Array CGH

Genomic DNA was extracted from peripheral blood using PUREGENE DNA Isolation Kits (Gentra, Minneapolis, Minnesota, USA). A pool of normal male or female control DNAs (Promega, Madison, Wisconsin, USA) was used as reference DNA to match the sex of the samples studied. The 1 Mb bacterial artificial chromosome (BAC) array CGH (Spectral Genomics (SG), Houston, Texas, USA) was performed as previously described.12 High resolution genome-wide human single nucleotide polymorphism (SNP) array 6.0 was performed by Affymetrix for subjects 1 and 2 to refine physical breakpoints further.

Databases cataloguing putatively benign CNVs (Database of Genomic Variants, DGV, and pathogenic CNVs (Decipher, were used to interpret the significance of CNVs. CNVs that are reported in at least two independent studies of healthy control subjects are typically referred to as benign copy number variants (bCNVs).


FISH analyses were performed using the array CGH identified BAC DNA clones (SG-PerkinElmer: or The Centre for Applied Genomics, TCAG: to confirm the deletions and duplications, as described previously.13 Slides were viewed on a Zeiss Axioplan 2 fluorescence microscope and images captured using MacProbe software (Applied Imaging, Santa Clara, California, USA).

Real-time quantitative PCR

Real-time quantitative PCR (RT-qPCR) with a minimum of three non-polymorphic markers within each CNV was used to validate CNVs. Primer sequences are available upon request. The RT-qPCR products were assessed via an ABI Prism 7900HT system (Applied Biosystems, Foster City, California, USA) using SYBR Green I detection as previously described.14

Statistical analyses

Exact Fisher tests were used for all categorical sample and phenotype comparisons, while the non-parametric Mann–Whitney test was used to evaluate numerical data such as ages and phenotype scores across categories. All statistical tests were performed using the Statistical Computation Web Site from Vassar College (

Gene prioritisation

We applied computational candidate gene prioritisation software, Endeavor,15 to help pinpoint potential ASD related genes within the deleted/duplicated genomic regions.

The training (reference) gene set includes 19 putative and known autism related genes summarised from recent publications: FMR1, NLGN3, NLGN4, NRXN1, SHANK3, CNTNAP2, PCDH10, CNTN3, NHE9 (SLC9A9), NHE6 (SLC9A6), DIA1 (c3orf58), A2BP1, MECP2B, UBE3A, EN2, SLC6A4, MET, SCN7A, RNF8.16

Based on the reference set, the Endeavor software prioritises genes in each altered genomic region, which have similar functions or biological processes/pathways with the reference genes.


Cohort and clinical demographics

To reduce clinical heterogeneity, we stratified our 100 subjects with ASDs into phenomic subgroups based on a broad spectrum of clinical comorbidities (table 1).

Table 1

Cohort and clinical demographics

Our cohort had an overall M:F ratio of 3.2:1 with noted comorbidities including epilepsy (21%), ID (64%), minor craniofacial anomalies (86%), and various medical or congenital systemic anomalies (77%) (table 1). Subjects in our study were almost 10 times more likely to present with macrocephaly (occipitofrontal circumference (OFC) >98th centile), as opposed to microcephaly (OFC <2nd centile).

Interestingly, 14% and 21% of subjects had birth weights (BW) ⩽5th centile or ⩽10th centile, respectively, as compared to the general population (both p = 0.05). However, our results do not confirm an association of low BW with the presence of pCNVs, even when <5th and <10th centile BW cohorts are combined.

Also of note is a significantly higher percentage of individuals with height measuring ⩾98th centile (11%, p = 0.02) or weight measuring ⩾98th centile (16%, p = 0.0008). However, no association was found between enlarged growth parameters and pCNV frequency.

A detailed list of these parameters and their respective prevalence in our cohort is shown in table 1.

Characterisation of CNVs in ASD subjects

Among our 100 subjects with idiopathic complex ASDs, we identified 67 non-redundant CNV loci with nine pCNVs in nine cases (9%) and three CNVs of unknown significance in three subjects (table 2).

Table 2

Abnormal copy number variants (CNVs) detected in 100 subjects with autism spectrum disorders (ASDs) using array comparative genomic hybridisation (CGH)

In concordance with current criteria, we consider the nine CNVs to be pathogenic because: (1) they were not reported in our control cohorts14 or in the DGV (; (2) they were either de novo, or maternally inherited X-linked CNVs in male offspring (that is, 470 Kb del (X)(p11.22))17; or (3) occurred within reported ASD loci (that is, 1.5 Mb dup(7)(q11.23))18).

The nine pCNVs were detected in 9.7% (3/31) of SPX families, 6.7% (3/45) of MPX-I and 12.5% (3/24) of MPX-E families. They included one maternally inherited CNV (del (X)(p11.22)) in two brothers,17 one female with dup(7)(q11.23)of unknown parental origin, and seven de novo CNVs including del(2)(p15p16.1)(6.1 Mb and 7.9 Mb) in two unrelated cases13; a maternally transmitted cryptic translocation causing del(14)(q11.2)(0.7 Mb) and dup(15)(q11q13)(10 Mb) in a niece and her maternal aunt19; a male with dup(18)(p11.3)(0.9 Mb); and a male with del(5)(p15.2p15.31)(2.4 Mb) and del(3)(p24.3)(0.1 Mb)20 for whom karyotype analysis showed inv(3)(p24q24), t(5;7)(p15.1;p12.2). Additional information about these CNVs is shown in table 2.

We also identified three CNVs of unknown clinical significance in three subjects, including dup(4)(q32.3)mat, del(13)(q32.3q33.1)pat, and dup(Y)(q11.22) (father unavailable for study), all of which were 0.2 Mb in size (table 1). They were not reported in our control cohorts14 and overlapped with CNVs reported 0–1 times in the DGV.

The clinical features of these subjects with nine pCNVs and three CNVs of unknown clinical significance are listed in table 3.

Table 3

Clinical features of subjects with abnormal copy number variants (CNVs)

Each of the 12 CNV loci identified, except dup(Y)(q11.22), contain gene(s), and four loci are single-gene containing regions [del(3)(p24.3), dup(18)(p11.3), dup(4)(q32.3) and del (13)(q32.3q33.1).

In order to further specify potential candidate genes, we applied the Endeavor gene prioritisation software ( to pinpoint the top two most likely (potentially ASD related) candidate gene(s) in each affected region, as listed in table 1.

In addition to recognised pCNVs, we identified 212 putative bCNVs including 109 deletions and 103 duplications among our 100 ASD subjects. These CNVs were found to be either completely overlapping with normal variants in the DGV (reported in at least two independent studies), or present in our own normal controls.14

Phenome-genome comparisons

Correlation of pCNVs with specific phenotypes

We compared the presence of pCNVs with standardised phenotypic subtypes utilising a scoring metric modified from de Vries et al9 to compare cases with and without pCNVs. Of all clinical features interrogated, microcephaly (OFC <2nd centile) and ID (IQ<70) severity were found to be most significantly associated with the identification of pCNVs (fig 1, table 1). Two of the nine cases with pCNVs (22.2%) had microcephaly versus only 2/91 (2.2%) cases without pCNVs (p = 0.04) (fig 1). All subjects with pCNVs had comorbidity with ID (9/9); 55/86 (64%) subjects without pCNVs had ID (p = 0.02), whereas no pCNVs were found in the absence of ID comorbidity (n = 31/86) (fig 1, table 1). Seven of nine cases with pCNVs (78%) were found to have moderate to severe ID (IQ <50) versus 35/86 (41%) for patients without pCNVs (p = 0.04; data not shown). A relative risk (RR) for harbouring pCNVs in patients with ID versus those without could not be computed, since no pCNVs were identified in the 31 subjects without ID (thus rendering an RR equal to infinity). Microcephaly was not a sensitive indicator for pCNVs (22%, 95% confidence interval (CI) 4% to 60%), but was highly specific (98%, 95% CI 92% to 100%). The presence of ID revealed low specificity for pCNVs (36%, 95% CI 26% to 47%) but was highly sensitive (100%, 95% CI 63% to 100%). All cases with microcephaly also presented with ID. No other clinical features were found to be more prevalent in subjects with pCNVs.

Figure 1

Comparison of clinical features between cases with and without pathogenic copy number variants (pCNVs). The star indicates the pair of values in each group that show significant differences (p<0.05; two tailed Fisher exact test). The numbers in brackets indicate the total number of ASD subjects with the clinical phenotype. The specific definitions of the clinical parameters examined are provided in table 1.

Correlation of total CNV load with clinical phenotypes

Total CNV load, including benign, pathogenic and CNVs of unknown significance, was determined for each ASD proband and correlated with clinical phenotypes. Subjects with a clinical phenotype score ⩾4 had a significantly higher tendency to harbour at least one CNV (76/83, 91.6%) compared to those with cut-off scores of 3 (12/17, 70.6%) (p = 0.03), as originally described by de Vries.9 There were no significant differences between any clinical parameter listed in table 1 and the presence or absence of CNVs. No significant difference in the number of CNVs was found in groups of ASD subjects divided by family type (SPX or MPX), and we did not find any correlation between the number of CNVs/person and the presence of pCNVs.

Correlation of CNVs with gender

A decreased M:F ratio was found in pCNV cases compared to the initial sample set (1.25:1 versus 3.2:1) (table 2). The nine pCNVs were found in five of 76 male (6.6%) and four of 24 (16.7%) female ASD subjects (p = 0.21). No significant association was found between gender and phenotype scores (p = 0.25). However, there was a higher percentage of cases with ⩾2 craniofacial dysmorphisms (100%) in females, compared to 62/76 (81.6%) males (p = 0.04). We examined whether gender was associated with co-occurrence of ID and found an increased M:F sex ratio in 31 ASD subjects without ID (6.75:1), but only a trend towards association between ID and female gender (p = 0.06). No other major somatic phenotype was associated with one specific gender. Finally, no significant difference in the total number of CNVs was found between males and females (p = 0.28).

Correlation of CNVs with parental age

We applied different statistical methods to delineate if there was any association between the presence of pCNVs in autistic offspring according to parental age at the time of birth. Surprisingly, lower paternal age was associated with increased pCNV frequency in children of fathers under 30 years of age (n = 20) (RR 2.66, 95% CI 1.17 to 6.04; p = 0.02). The same was true for children born to fathers under age 35 years, but the relationship was weaker (n = 52) (RR 1.65, 95% CI 1.19 to 2.30). Although lower maternal age was not as strongly associated with pCNVs in their offspring, the relative risk was significantly above 1 for mothers under age 30 years (n = 40) (RR 1.92, 95% CI 1.19 to 3.10). Next we examined total CNV load in relation to parental ages and found that offspring of both fathers and mothers under the age of 30 and 35 years, respectively, had significantly more CNVs (benign and pathogenic) than children of older parents (p = 0.02 and p = 0.05, respectively).


To our knowledge, this is the first report describing detailed correlation analyses between standardised “whole body” phenotypes and CNVs in ASD subjects. Of the indices examined, only microcephaly and ID were significantly associated with the presence of pCNVs. Furthermore, we also found that subjects with pCNVs have a much higher percentage of cases with moderate to severe ID (IQs between 20–50) than subjects without pCNVs. A recent study of ID subjects found that those with moderate to severe ID were twice as likely to harbour pCNVs as those with borderline to mild ID.21 Our data agree with this determination as 7/42 (16.7%) of our ASD probands with moderate to severe ID had detectable pCNVs in comparison to 2/22 (9%) of subjects with mild ID (IQs between 50–70). Our results thus suggest that microcephaly and IDs may represent a strong phenomic predictor of pCNV risk in persons with ASDs.

It is interesting to note that macrocephaly or seizures, both commonly described with ASDs, did not associate with the presence of pCNVs, even if subjects in our cohort were almost 10 times more likely to manifest macrocephaly as compared to microcephaly. Other reports have also showed that macrocephaly and seizures are not good aetiologic indicators of ASDs.22 23 Extremely low (<3rd centile) or high (>98th centile) birth weights did not associate with significant pCNV risk, although smaller babies (<5th and <10th centile birth weight for gestational age) were overrepresented in our cohort. Prenatal and postnatal growth abnormalities, craniofacial dysmorphisms, systemic anomalies, pregnancy and neonatal complications also did not associate with significant pCNV risk.

Our study indicates that the standardised phenotype scoring system of de Vries et al9 does not help better predict the presence of pCNVs in persons with ASDs, as there was no correlation between increased phenotype scores and pCNV presence. Our results may be confounded by our population cohort, which focused on the analysis of persons with more complex forms of idiopathic ASD. However, when considering CNV load we found that the group with greater phenotypic complexity had a greater chance of harbouring at least one CNV. Increased CNV burden may be relevant in contributing to a “network” effect of CNVs, and/or “additive aneuploidy effect” where the sum of genes that are lost or gained contributes to a more severe clinical phenotype.24 These findings were obtained using the 1 Mb resolution SG array and warrant further studies using higher resolution arrays and a different template for phenotypic evaluations incorporating ID and microcephaly as key components.

Interestingly, we observed that children born to younger fathers had a higher prevalence of pCNVs. A similar relationship for maternal age was present, albeit weaker. Consistent with this is our finding that children of younger parents (fathers under age 30 and mothers under age 35 years) also have a significantly higher total CNV load than children born to older parents. A number of epidemiological studies have shown different results for the distribution of parental ages in ASD cohorts in comparison with cohorts of typically developing individuals.25 26 While some studies have found higher maternal and/or paternal ages in ASD cohorts,25 other studies have showed a “U shaped” distribution of parental ages relative to the offspring’s risk for ASDs or other genetic conditions.26 These findings are counterintuitive to the expected paradigm underscoring increased age with markers of genomic instability including point mutations and altered DNA methylation.27 28 Increased pCNV risk to offspring of younger fathers could possibly result from the immaturity of spermatids or low activity of DNA repair or antioxidant enzymes in younger parents.29 Our findings compare individuals with and without pCNVs from a population cohort of persons with phenotypically complex ASDs and thus should be interpreted with caution until larger ASD and control cohorts can be similarly studied.

In concordance with other reported array based pCNVs studies in ASDs, we found a similar pCNV detection rate (9%) and a decreased M:F sex ratio among subjects with pCNVs from the original sample cohort (1.25:1 vs the original cohort 3.2:1) (table 4).4 5 6 7 8 The observed rates of pCNVs in SPX (6.5%) and MPX-I (2.2%) families are in general agreement with those previously reported (∼7–10% and 2–3% in respective family types.6 7 8). Probands from MPX-E families had a higher pCNV frequency (12.5%) than observed in our own or previously reported SPX or MPX-I families.5 6 7 8 pCNV frequency in probands from MPX-E families have not been previously reported, since most studies have focused on affected members (that is, sibling pairs) from the immediate family. Failure to record information about extended family histories could lead to the misclassification of MPX-E cases as SPX individuals.

Table 4

Comparison of present autism spectrum disorder (ASD) array comparative genomic hybridisation (CGH) study with published reports

Although female gender did not directly associate with the presence of detectable pCNVs, we did find a decreased M:F ratio in subjects with pCNVs (1.25:1) versus those without pCNVs (3.6:1), a trend that has been reported in every other large scale CNV study of ASDs (table 4).4 5 6 7 8 This strongly suggests that pCNVs may play a substantial role in determining the ASD phenotype and more equally contribute to ASD risk in both genders, due to increased penetrance.6 These pCNV changes could also encompass genes for one or more sex limited ASD traits within the heterogeneous ASD population. Furthermore, our findings of an increased M:F ratio in subjects without ID (6.75:1), and of females with an ASD more commonly manifesting two or more craniofacial dysmorphisms, are consistent with clinical morphology research that shows that more simple forms of ASD have a significantly higher male-to-female ratio than complex subgroups with abnormal physical phenotypes.30

It is reasonable to assume that the genes within or close to the pCNVs are the most likely ASD related candidate genes. Close to 200 genes are contained within the pCNV loci identified in our study (table 2). We applied the computational candidate gene prioritisation software, Endeavor, to assist with further pinpointing potential ASD related genes that share similarity to selected autism related reference genes.15 Different from other gene prioritisation tools, the publicly available Endeavour software can access many more data sources (currently up to 20 for Homo sapiens) including categories of functional annotations, protein interactions, expression profiles, regulatory information, sequence based and text mining data31 with all referenced information validated by in vivo experiments.15 However, the rank of prioritised genes in a specific locus is mainly dependent on the training set of genes and the selection of models in the database. We used a list of 19 autism related genes16 as our training set and selected all models in our analysis. Among the top two prioritised genes in each loci in table 2, most were involved in either mammalian nervous system development and/or neuronal excitability—for example, WNK3,32 and NALCN33—or have been reported to be associated with neurodevelopmental disorders including NIPA1 (involved in hereditary spastic paraplegia),34 and CTNND2 (severe ID in cri-du-chat syndrome).35 Some have been reported in ASD related studies, such as PHF8,17 WNK317 SEMA5A,36 GTF2I,37 STX1A,38 NIPA139 and UBE3A.2

Three pCNV loci (del(14)(q11.2), dup(15)(q11q13) and del (X)(p11.22)) were recurrent in being detected in multiple individuals within their respective families. One recurrent locus (del (2)(p15p16.1) was identified in two unrelated probands with overlapping but non-identical deletions,13 with additional recurrences described in two other reports confirming striking concordance in clinical phenotype and refinement of the genes involved40 41 (DECIPHER: The genotype–phenotype analysis of these collective cases will help us further understand this newly recognised syndrome and its relationship to ASD/ID pathophysiology and facilitate the identification of the disease genes in the syndrome.

This study is, to our knowledge, the first to apply standardised, comprehensive phenomic analyses in correlation with genomic CNV findings. The advantages of identifying “whole body” phenomic and genomic biomarkers as diagnostic tools for ASD include the potential for better standardisation of behaviour based diagnoses, evaluations of treatment response, and earlier identification allowing earlier treatment. Moreover, the pursuit of aetiologic clues that define ASDs physiologically are more likely to improve our understanding of natural history and causal pathways leading to autism, while enriching the evidence base for genetic counselling that awaits refinement from the growing spectrum of ASD associated microdeletion and microduplication syndromes.


The authors are very grateful to the ASD-CARC families who participated in this research as members of our on-line Research Registry We appreciate the assistance of Liza and Levina Kasmara, Lindsay Swinton and Dr Elena Lopez in reviewing and cataloguing the phenomic findings of all study subjects.


View Abstract


  • Funding This work was supported by funding from the Canadian Institutes for Health Research (CIHR) (RT-64217; PI: MESL), Autism Speaks (PI: MESL), Michael Smith Foundation for Health Research (PI: MESL) a CIHR Interdisciplinary Health Research Team grant (RT-43820) to ASD-CARC (PI: JJAH) [], and the Ontario Mental Health Foundation (PI: JJAH). YQ, NR and MK are trainees with the CIHR/NAAR-Autism Speaks STIHR Inter-Institute Autism Spectrum Disorders Training Program (PI: JJAH). ERS is supported by a CIHR Institute of Genetics Clinician Investigator Award (2005–09). MESL and ERS are Career Scholars supported by the Michael Smith Foundation for Health Research.

  • Competing interests None.

  • Patient consent Not required.

  • All authors are members of the Autism Spectrum Disorders-Canadian American Research Consortium (ASD-CARC:,

  • Provenance and peer review Commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.