A total of 522 patients with ASD belonging to 430 families were studied. Subjects were recruited by the Paris Autism Research International Sibpair (PARIS) study at specialized clinical centers in Europe and the United States. The patients included 393 males and 129 females (ratio 3:1), with a mean age at the last evaluation of 11±7.5 years (range 2.5–43); 187 belonged to 95 multiplex families (with two or more affected siblings) and 335 were sporadic cases. All patients were evaluated by psychiatrists or child neurologists and diagnosed based on clinical evaluation and DSM-IV criteria. Patients were assessed with the Autism Diagnostic Interview-Revised (ADI-R) (28) and the Asperger Syndrome Diagnostic Interview (29). Four-hundred seventy-two individuals met criteria for autism, 26 for Asperger syndrome and 24 for PDD-NOS. Three-hundred fifty-six patients (68%) had mental retardation, 261 (50%) had very limited or no language, and 66 (13%) had a history of epilepsy. Most patients were Caucasian (89%). Laboratory tests included karyotype, fragile X and metabolic screening; brain imaging and EEG were performed when possible. Patients with known genetic disorders were excluded. Chromosome analysis identified one female harboring a supernumerary isodicentric derivative chromosome 15; we report here the cytogenetic, molecular and phenotypic characterization of this patient. No other patients had been excluded from the sample prior to this study because of cytogenetic abnormalities of chromosome 15q11-q13. The study was approved by the research ethics boards of the collaborating institutions. Informed consent was obtained from all families participating in the study.

Patients were screened for rearrangements involving the 15q11-q13 region using the ME028 PWS/AS and/or the P064 MR1 MLPA kits (n=351 and 522, respectively) (MRC-Holland). The ME028 kit contains 25 probes specific for sequences in or near the PWS/AS critical region, as well as 5 probes to assess methylation status (25). The P064 kit contains 5 probes in the PWS/AS region (one probe in MKRN3, NDN, and GABRB3, and two in UBE3A). MLPA data were analyzed using GeneMarker 1.70 software (SoftGenetics). After population normalization, the peak height from each sample was compared to a synthetic control, which represents the median of all normal samples in each experiment. Peak heights below 0.75 were considered as deletions and values above 1.3 as duplications. Cases with apparent deletions or duplications were confirmed with quantitative PCR (qPCR) and fluorescent in situ hybridization (FISH). Apparent deletions of a single probe were sequenced to rule out single-base changes within the probe-binding region. For further details see Supplement 1.

Five microsatellite markers in the 15q11-q13 region (D15S11, D15S817, D15S1513, D15S1019 and D15S815) were used to assess DNA copy number and follow the transmission of the alleles (Fig. 1). PCR conditions (in particular the number of PCR cycles) were set up so as to perform allele dosage. PCR products were quantified on an ABI3730 sequencer (Applied Biosystems).

Patient 1 had an idic(15) revealed by the karyotype and confirmed by FISH, described as 47,XX,+mar.ish der(15)(D15Z1++, SNRPN++, UBE3A++) (Fig. 2A). Microsatellite analysis of the patient and her parents showed that she had a de novo copy number gain of D15S11, D15S817 and D15S815 (Fig. 2B). Methylation-sensitive MLPA revealed that the idic(15) was maternally-derived, as indicated by the increased dosage of the methylated probes in SNRPN and NDN (Fig. 3). MLPA confirmed that the idic(15) involved all the probes in the 15q11-q13 region (between BP1 and BP3) but did not affect two genes located distally, at 15q26, BLM and IGF1R (Fig. 2C). Further analysis of the 15q11-q13 interval with 15 microsatellites showed six triallelic markers, with one allele of paternal origin and two of maternal origin (Fig. 2D). One informative microsatellite between BP4 and BP5, D15S1031, had a single maternal allele, suggesting the breakpoint lies proximal to this marker (Fig. 2D). Detailed investigation of the distal breakpoint with qPCR showed increased dosage of MTMR10, TRPM1 and KLF13, located between BP4 and BP5, but two distal genes in the same interval, OTUD7A and CHRNA7, showed a normal copy number (Fig. 2E). These findings suggest that the distal breakpoint of the idic(15) lies between BP4 and BP5, in agreement with the microsatellite results. The KLF13 probeb showed an increase in dosage to the hexasomy range, suggesting a complex rearrangement at the site of the breakpoint. We could not differentiate between one or two breakpoints, because of the inability to estimate the precise copy number of individual probes on the basis of qPCR (Fig. 2E).

Patient 2 had three alleles for both D15S11 and D15S817. He was homozygous for D15S1513 and D15S1019, which were therefore uninformative, and heterozygous without imbalance for D15S815, located between BP3 and BP4 (not shown). MLPA confirmed the presence of a ~4.6 Mb duplication between BP2 and BP3, corresponding to the critical region deleted in PWS/AS (Fig. 4). Methylation-sensitive MLPA showed normal dosage of the methylated maternal probes in SNRPN and NDN, indicating that the duplication was paternally derived (Fig. 3). MLPA of the parents showed that the duplication was inherited from the unaffected father (Fig. 4). Methylation analysis of the father showed a paternal origin of the duplication (not shown), but no DNA was available from the grandfather to determine if the duplication was inherited or had arisen de novo. FISH using the UBE3A probe confirmed the presence of an increased signal in Patient 2 (not shown).

Patients 3 and 4 had only one allele for three markers located in the PWS/AS region (D15S11, D15S817 and D15S1513), suggesting that they could be either homozygous or deleted in this region (Fig. 5A). Examination of the parents’ haplotypes showed that both patients had only one allele from the father and none from the mother (Fig. 5A). MLPA confirmed that Patient 3 had a maternal class I deletion extending ~5.6 Mb from BP1 to BP3 that arose de novo (Fig. 3, 5C). FISH with probe D15S10 confirmed the deletion (not shown). In Patient 4, MLPA showed a normal copy number of the 15q11-q13 region (Fig. 5C) but methylation-sensitive MLPA revealed a complete absence of maternal methylated DNA sequences, indicating a paternal uniparental disomy (Fig. 3). FISH showed normal hybridization of probes SNRPN and D15S10 (not shown), consistent with the diagnosis of uniparental disomy.

MLPA also detected three patients with a deletion (Patients 5, 6 and 7) and two with a duplication (Patients 8 and 9) of the BP1-BP2 interval, which spans 254 kb and encompasses four genes, TUBGCP5, CYFIP1, NIPA1, and NIPA2 (Fig. 1). All the rearrangements were inherited from parents who were unaffected or had unrelated psychiatric disorders (Table 1).

The cytoplasmic FMR1 interacting protein 1 (CYFIP1) interacts with FMRP, encoded by the fragile X mental retardation 1 (FMR1) gene. Since fragile X syndrome is often associated with autism, CYFIP1 constitutes a candidate gene for autism. In order to test the possibility that point mutations in this gene, either inherited from the other parent or de novo, could lead to autism in association with the BP1-BP2 imbalance, we sequenced the whole coding region of CYFIP1 in the five probands with abnormal dosage of this region. In Patients 5 and 8 we extended the search of point mutations to the coding regions of NIPA1, NIPA2 and TUBGCP5, as well as those of FMR1, FXR1 and FXR2, which encode proteins of the FMRP complex. No mutations were identified in any of the genes screened.

We identified one patient with a paternally-inherited deletion of exon 1 in one of the transcript variants of the UBE3A gene. Because the deleted exon is noncoding and UBE3A is maternally expressed, this genomic imbalance is unlikely to have clinical consequences. In another patient, MLPA showed an apparent deletion of one probe in the SNRPN promoter/exon1 region, due to a maternally-inherited nucleotide change previously reported as a rare neutral variant (see Supplement 1 for further details on these two patients).

Aside from the uniparental disomy identified in Patient 4, no other methylation abnormalities were observed in 331 patients screened with methylation-sensitive MLPA. In the remaining subjects, uniparental disomy was ruled out with microsatellite analysis.

Supernumerary marker chromosomes are a relatively common cytogenetic finding, with an estimated incidence of 0.8/1000 prenatal diagnoses; those derived from chromosome 15 account for about half of all marker chromosomes (30). The frequency is higher in children with developmental delay, with 10 idic(15) identified in 2000 such cases in one study (11). Idic(15) that do not include the PWS/AS critical region have no clinical effect (31), whereas those including the PWS/AS region lead to a neurobehavioral phenotype including autism or autistic-like behavior, cognitive deficits, hypotonia, mild dysmorphic features, and seizures (32). There have been numerous reports of idic(15) associated with autism (9–13). As in Patient 1, the vast majority of idic(15) that encompass the PWS/AS locus are maternally derived and arise de novo (33). Almost two-thirds extend beyond BP3 and the majority are asymmetrical, with two distinct breakpoints, at BP4 and BP5 (14, 33). Analysis of the number of copies across the 15q11-q13 region in Patient 1 with qPCR and microsatellites predicted an atypical distal breakpoint between BP4 and BP5. A similar breakpoint was reported in a patient with an idic(15), between D15S1013 andD!S1031 (33).

Although genotype/phenotype correlation studies show that the segmental tetrasomy of idic(15) is associated with a more severe phenotype in terms of developmental outcome compared to interstitial triplications or duplications, indicating a dosage effect, no clear correlation has been observed between gene dosage and autism phenotype (34). Furthermore, no consistent relationship has been observed between rearrangement size and clinical severity (11, 34).

At least 33 cases of interstitial duplications of the 15q11-q13 region have been reported in association with ASDs, according to the Autism Chromosome Rearrangement Database (http://projects.tcag.ca/autism/). Interstitial duplications of proximal 15q that do not include the PWS/AS critical region have no clinical effect, are usually familial and may be considered normal variants (18). Patients with interstitial duplications of the PWS/AS locus have an abnormal phenotype that includes developmental delay, particularly affecting speech and language, varying degrees of mental retardation, autism or autistic features, motor coordination difficulties, and mild or no dysmorphic features (16). The phenotype is highly variable, even among members of the same family carrying identical rearrangements (16) and can in some cases manifest as developmental language disorder and dyspraxia, without autism (35). The clinical findings in Patient 2 are similar to those described in patients with maternal 15q11-q13 duplications, although he lacks the hypotonia, motor coordination problems and joint laxity observed in other cases (16, 17).

The frequency of 15q11-q13 interstitial duplications is estimated at 1:600 individuals referred with developmental delay (36). Although previous studies in small samples of patients had suggested a frequency of 1% for maternally-derived interstitial duplications in autism (10, 24), we did not detect any such cases in our sample but found one patient harboring a paternal 15q11-q13 duplication. In agreement with our findings, Bolton et al. screened 181 patients with autism and found only one interstitial duplication of paternal origin (37). Similarly, Thomas et al. studied 327 samples referred for a suspected ASD as well as 87 patients with confirmed autistic disorder and found no 15q duplications (36). More recently, a genome-wide microarray study of 427 subjects with ASD identified two maternally-derived 15q11-q13 duplications (7). Collectively, these results suggest that maternal interstitial duplications are less frequent than previously assumed, with an estimated ~0.3% frequency (4/1369) in subjects with ASD. Based on the same studies, the estimated frequency of idic(15) is ~0.2% (3/1369) and that of paternal duplications ~0.1% (2/1369). Note however, that the frequency of idic(15) may be underestimated because previous studies included an unknown proportion of subjects that had been screened for cytogenetic abnormalities prior to inclusion. It also should be noted that none of these studies were based on epidemiological samples and the possibility of ascertainment biases cannot be excluded.

Interstitial duplications of 15q11-q13 of maternal origin are associated with developmental delay and/or autistic behavior, whereas paternally-derived duplications usually lead to a normal phenotype (10, 15, 16, 18). However, as shown in Table 2, there have been at least six reports of paternal duplications encompassing the PWS/AS region associated with an abnormal phenotype, including mental retardation, delayed or absent speech, and ASD or autistic behavior (37–41). In addition, two patients with an interstitial 15q11-q13 triplication of paternal origin have been reported (42, 43). Together with the patient with the paternal duplication described here, these findings suggest that paternally-derived duplications may lead to phenotypic effects, including autism. Thus, there may be other genetic and/or environmental/epigenetic factors operating to modify the penetrance and expressivity of paternal duplications. Similar as yet unknown factors could come into play to modify the phenotypic expression of maternal 15q11-q13 duplications, which are also associated with a wide range of developmental problems and show marked clinical variability among members of the same family, including individuals who appear unaffected (16).

It is interesting to note that several of the patients with paternally-inherited duplications or triplications described previously had Prader-Willi syndrome-like features (Table 2) (38, 40, 43), suggesting a role of paternally-expressed genes mapping within the PWS/AS region (e.g., MKRN3, MAGEL2, NDN, and SNRPN-SNURF) in the phenotypic manifestations. Several recent studies have shown that ASD is more prevalent in Prader-Willi syndrome patients than was previously thought (44), in agreement with a contribution of paternally-expressed genes to the ASD phenotype. Further research is needed to examine the effect of parental origin on phenotypic outcomes of 15q11-q13 duplications and their contribution to ASD. In particular, studies of gene expression in tissues from patients with paternal duplications may improve our understanding of the effect of these rearrangements on phenotype.

The prevalence of Angelman syndrome is estimated at 1/12,000 (45). Angelman syndrome is characterized by severe mental retardation, profound speech impairment, ataxia, and typical behavior including happy disposition with frequent laughter/smiling and hand-flapping. Seizures, microcephaly and distinctive physical traits (large open mouth, widely-spaced teeth, prognathism) are also common (46). It is caused by deficiency of the maternally-inherited UBE3A gene, resulting from deletion of the maternal 15q11.2-q13 region (70%), paternal uniparental disomy (2%), UBE3A mutations (10%), or imprinting defects (5%), with no molecular abnormality identified in the remaining patients (13).

The association between Angelman and autism has been known for more than a decade (45). According to recent data, 62% (38/61, range 50%–81%) of subjects with Angelman syndrome meet criteria for autism or ASD (19–21), suggesting that ASD may be part of the behavioral phenotype of Angelman syndrome. Although the diagnosis of ASD in Angelman syndrome is complicated by the severe intellectual disability usually observed in these patients, there is clear evidence for ASD in a subgroup of Angelman patients who exhibit social and communication deficits that are disproportionate to their overall cognitive function. There is little information about the prevalence of Angelman syndrome in ASD. Schroer et al. found one maternal 15q11-q13 deletion among 100 patients with autism (10). In this study, we identified two subjects with Angelman syndrome, one with a maternal deletion and the second with a uniparental disomy, for a ~0.5% frequency.

Because certain features of autism overlap with those seen in Angelman syndrome, including severe mental retardation, absence of speech or very limited language, epilepsy and/or abnormal EEG, and stereotyped behaviors, the diagnosis of Angelman syndrome may be overlooked in patients with ASD, as happened with our two patients. Although neither patient exhibited the classical phenotype, their clinical presentation is within the normal range of phenotypic variability observed in Angelman syndrome. In common with the majority of patients with Angelman, they exhibited severe cognitive impairment and limited or absent language, but had no ataxia or microcephaly, and seizures and unprovoked laughter (during infancy only) were observed only in one.

We identified three deletions and two duplications of the region between BP1 and BP2, all inherited from unaffected parents. A BP1-BP2 deletion was reported recently in a boy with mental retardation, speech delay and neurological deficits, but the interpretation was complicated by the presence of the same deletion in the father, who also showed cognitive impairment (47). Copy number gains and losses of this region have also been identified in genome-wide array analyses of control individuals (Database of Genomic Variants, http://projects.tcag.ca/variation/), with a population frequency estimated at 1% (48), suggesting that this may be a copy number polymorphism. In agreement with this interpretation, individuals carrying small idic(15) with extra copies of the four genes included in the BP1-BP2 interval but that do not carry extra genomic material distal to BP2 are phenotypically normal (31), suggesting that these genes do not exert clinically significant dosage effects.

Nevertheless, it is interesting to note that NIPA1, NIPA2, CYFIP1 and TUBGCP5 have been suggested to modulate the PWS/AS phenotype. Angelman patients with class I deletions (including the BP1-BP2 interval) are more likely to meet criteria for autism and lack vocalizations compared with patients having smaller class II deletions (20). Similarly, in Prader-Willi syndrome, class I deletions are associated with more obsessive-compulsive behaviors and lower intellectual ability (49), and mRNA levels of these four genes are positively correlated with better behavioral outcomes (50). More recently, BP1-BP2 deletions were shown to be significantly associated with schizophrenia (51), suggesting that these variants may increase risk for various neuropsychiatric phenotypes, albeit with low penetrance and/or variable expressivity.

One of the genes situated in this interval, CYFIP1, is particularly interesting in the context of ASD. CYFIP1 interacts with fragile X mental retardation protein (FMRP) as well as with the Rho GTPase Rac1, which plays a role in the regulation of axonal migration and dendritic spine morphology. CYFIP1 has been proposed as a potential molecular link between fragile X syndrome and 15q11-13 duplication, since both disorders result in excess free CYFIP1 (52). In order to test the possibility that the BP1-BP2 deletion unmasked a recessive mutation on the other chromosome we sequenced CYFIP1, as well as NIPA1, NIPA2, TUBGCP5, and genes encoding fragile X-related proteins (FMR1, FXR1 and FXR2), in patients carrying BP1-BP2 deletions, identifying no mutations. Clearly, more research is required to determine the phenotype correlations of BP1-BP2 deletions and duplications and the role of NIPA1, NIPA2, CYFIP1 and TUBGCP5, if any, in neurodevelopmental disorders.