Complex nature of apparently balanced chromosomal rearrangements in patients with autism spectrum disorder

Background Apparently balanced chromosomal rearrangements can be associated with an abnormal phenotype, including intellectual disability and autism spectrum disorder (ASD). Genome-wide microarrays reveal cryptic genomic imbalances, related or not to the breakpoints, in 25% to 50% of patients with an abnormal phenotype carrying a microscopically balanced chromosomal rearrangement. Here we performed microarray analysis of 18 patients with ASD carrying balanced chromosomal abnormalities to identify submicroscopic imbalances implicated in abnormal neurodevelopment. Methods Eighteen patients with ASD carrying apparently balanced chromosomal abnormalities were screened using single nucleotide polymorphism (SNP) arrays. Nine rearrangements were de novo, seven inherited, and two of unknown inheritance. Genomic imbalances were confirmed by fluorescence in situ hybridization and quantitative PCR. Results We detected clinically significant de novo copy number variants in four patients (22%), including three with de novo rearrangements and one with an inherited abnormality. The sizes ranged from 3.3 to 4.9 Mb; three were related to the breakpoint regions and one occurred elsewhere. We report a patient with a duplication of the Wolf-Hirschhorn syndrome critical region, contributing to the delineation of this rare genomic disorder. The patient has a chromosome 4p inverted duplication deletion, with a 0.5 Mb deletion of terminal 4p and a 4.2 Mb duplication of 4p16.2p16.3. The other cases included an apparently balanced de novo translocation t(5;18)(q12;p11.2) with a 4.2 Mb deletion at the 18p breakpoint, a subject with de novo pericentric inversion inv(11)(p14q23.2) in whom the array revealed a de novo 4.9 Mb deletion in 7q21.3q22.1, and a patient with a maternal inv(2)(q14.2q37.3) with a de novo 3.3 Mb terminal 2q deletion and a 4.2 Mb duplication at the proximal breakpoint. In addition, we identified a rare de novo deletion of unknown significance on a chromosome unrelated to the initial rearrangement, disrupting a single gene, RFX3. Conclusions These findings underscore the utility of SNP arrays for investigating apparently balanced chromosomal abnormalities in subjects with ASD or related neurodevelopmental disorders in both clinical and research settings.


Background
Balanced chromosome abnormalities, including translocations and inversions, are structural rearrangements of genetic material with no overall gain or loss detected with conventional karyotyping. Apparently balanced chromosomal rearrangements (ABCR) have an estimated frequency of 0.5% in newborns [1], including 14% de novo. About 6% of de novo ABCR detected at amniocentesis are associated with an abnormal phenotype, including intellectual disability (ID) and multiple congenital anomalies [2]. Because of the insufficient discrimination power of conventional cytogenetics, in which genomic imbalances smaller than 5 to 10 megabases (Mb) are usually not detected, the most common explanation for the clinical abnormalities is cryptic loss or gain of genomic material at or in the vicinity of the breakpoint. When there is no loss or gain of DNA sequences, the rearrangement can disrupt a dosage-sensitive gene, separate a gene from its cis regulatory elements, or generate a functional chimeric gene.
Studies using DNA microarray technologies have demonstrated submicroscopic anomalies related or not to the breakpoint in 46% (range 31% to 100%) of patients with an abnormal phenotype carrying a de novo ABCR [3][4][5][6][7][8][9][10]. Inherited ABCR in patients with an abnormal phenotype have been studied less often, but they can also be associated with cryptic imbalances at the breakpoint or elsewhere in the genome, with a combined frequency of 25% (6/24) in three studies [7][8][9]. Imbalances are more frequent in complex rearrangements involving more than two breakpoints and in patients with a complex phenotype [10]. In contrast, genomic imbalances are unlikely to be detected in phenotypically normal carriers of apparently balanced translocations [6]. Taken together, these findings indicate that a significant proportion of ABCR in phenotypically abnormal individuals are in fact associated with genomic imbalances and that these rearrangements should be systematically investigated by high-resolution microarrays independently of their de novo or inherited origin. The phenotypes of the patients in previous studies were very heterogeneous, including developmental delay, ID, multiple congenital anomalies, and autism spectrum disorder (ASD).
ASD is an etiologically heterogeneous neurodevelopmental disorder characterized by impairments in social communication and by restricted interests and stereotyped behaviors. Hundreds of rare variants, including chromosomal abnormalities, copy number variants (CNVs), and single nucleotide variants have been implicated in ASD [11,12]. However, for about 80% of cases, the underlying genetic determinants remain unknown. The frequency of structural chromosomal imbalances detected by conventional cytogenetics in autism varies between 2% and 6% [13,14], including karyotypically balanced chromosomal abnormalities. To date, only isolated cases with ABCR and ASD have been studied by array technology (for example [15,16]). Here we report a series of 18 patients with ASD carrying de novo or inherited ABCR studied by single nucleotide polymorphism (SNP) arrays to identify cryptic CNVs implicated in abnormal neurodevelopment.

Subjects
Patients with ASD and ABCR were ascertained through two sources: 1) 12 patients from the Paris Autism Research International Study (PARIS) family dataset [17], and 2) 6 patients referred by the network of French cytogeneticists. A summary of the clinical and cytogenetic data of the patients is shown in Table 1. Sixteen patients fulfilled DSM-IV criteria for autistic disorder, while two subjects (patients 9 and 13) had a previous diagnosis of ASD but could not be formally evaluated for ASD for this study. The Autism Diagnostic Interview-Revised (ADI-R) was performed in 14 subjects; in 1 individual, the Diagnostic Interview for Social and Communication Disorders, tenth revision (DISCO-10) was used instead. In addition, five subjects were assessed with the Autism Diagnostic Observation Schedule (ADOS). Fourteen patients had ID. Based on the presence of facial dysmorphism and/or malformations, six subjects were considered to have syndromic ASD (patients 3, 4, 5, 9, 10, and 15). Fragile X molecular testing and metabolic screening were normal in all individuals. All parents were phenotypically normal, including those carrying a rearrangement. The study was performed in accordance with the ethical standards of the responsible institutional and national committees on human experimentation, in compliance with the Helsinki Declaration. Informed consent was obtained from all families participating in the study. For the PARIS patients, the study was approved by the research ethics boards of the collaborating institutions (Comité de Protection des Personnes Ile-de-France VI, Paris, France, and Ethical Review Board in Gothenburg, Sweden).

Conventional cytogenetic analysis
Karyotype (G-banded and/or R-banded, 450 band level) showing an ABCR was available for all patients and in most of the parents before inclusion. The sample included ten autosomal reciprocal translocations, one X-autosome translocation, three rare inversions, and four common pericentric inversions of chromosomes 2, 5, 9, and Y, usually considered as cytogenetic variants (Table 1). Nine chromosomal rearrangements were de novo, four were maternally inherited, three were paternal, and in two cases, the inheritance could not be defined because no blood samples were available from the fathers.
Whole-genome SNP array SNP array analysis was performed in affected individuals and their parents using the Human CNV370-Duo DNA  Analysis BeadChip (Illumina, San Diego, CA, USA). In three cases, parental DNA was not available (patients 8, 9, and 18), and in one case, only the mother was available (patient 17). The array contains more than 370,000 markers, with a mean resolution of 30 kb. Genomic DNA was processed according to the Infinium II assay manual. SNP copy numbers (log R ratio) and B allele frequencies were assessed using the Bead Studio software version 3.2, with the CNV partition algorithm, V 1.3.2 or 2.4.4 (Illumina). All genomic coordinates are based on GRCh37/hg19. The clinical relevance of CNVs was interpreted according to the American College of Medical Genetics guidelines [18].

FISH analyses
Fluorescence in situ hybridization (FISH) analysis was carried out using specific bacterial artificial chromosomes (BACs) to confirm CNVs larger than 1 Mb. BAC clones were hybridized on metaphase spreads of patients and their parents, in order to confirm the inheritance.
qPCR Real-time quantitative PCR (qPCR) was used to confirm and map CNVs smaller than 1 Mb. For patient 2, qPCR was used to confirm a larger CNV because metaphase spreads were not available. We used the Universal Probe Library (UPL) system (Roche, Indianapolis, IN, USA) and a LightCycler 480 real-time PCR system (Roche), as described previously [19].

Results
In this study, 18 patients with ASD and an apparently balanced chromosome rearrangement were analyzed for submicroscopic imbalances using genome-wide SNP arrays ( Table 1). The microarray analysis detected clinically significant de novo CNVs in 4/18 patients (22%), including 3/9 with de novo ABCR and 1/7 with an inherited abnormality ( Table 2). Patients 1 and 4 had both a deletion and a duplication in the chromosome implicated in the rearrangement. Patient 2 had a deletion at one of the translocation breakpoints, whereas patient 3 carried a deletion on a chromosome unrelated to the known rearrangement. The size of the pathogenic CNVs ranged from 3.31 to 4.93 Mb. In addition, patient 5 carried a de novo deletion on a chromosome unrelated to the initial rearrangement and disrupting a single gene, RFX3. In the remaining 13 patients, no clinically relevant rearrangement was detected. None of the patients had long contiguous stretches of homozygosity involving multiple chromosomes.

Patient 1
This boy was referred for evaluation at the age of 2 years and 9 months because of developmental delay and ASD features. He is the second child of healthy nonconsanguineous parents (maternal age 31 years, paternal age 38 years). He was born at 42 weeks after an uneventful pregnancy, weighing 5,280 g (+4.3 SD), probably due to latent diabetes in the mother. He was hypoglycemic during the first days of life, and an inguinal hernia was corrected at the age of two months. There was no delay in motor milestones or language development. He was diagnosed with autism after formal testing at an autism referral center. He had a borderline IQ of 69 (verbal IQ 67, performance IQ 78). There was no history of seizures, and his EEG was normal. His physical and neurological exams were normal except for mild strabismus; in particular, he had no dysmorphic features. At the age of 4 years, he was +2 SD to +3 SD in weight and height. Conventional karyotyping showed a de novo translocation between the telomeric region of the short arm of chromosome 4 and the short arm of an acrocentric chromosome. His karyotype was 46,XY,der (4) Figure 1A), absent from the parents. Both rearrangements were confirmed by qPCR and FISH. The deletion was confirmed by FISH with the 4p subtelomeric probe RP11-2H3. Using probes RP11-296G16 (4p16.3) and RP11-265012 (4p16.2), the FISH findings were consistent with an inverted duplication (data not shown), indicating a chromosome 4p inverted duplication deletion. SNP genotyping indicated that both rearrangements occurred on the paternal chromosome. The deletion includes eight RefSeq genes; the duplication encompasses 70 genes and overlaps the Wolf-Hirschhorn critical region ( Figure 2).

Patient 2
The patient is a 5-year-old boy presenting with nonsyndromic autism and mild ID. He is the second child of healthy non-consanguineous parents (maternal age 26 Figure 1B), originating on the maternally inherited chromosome. The deletion was confirmed by qPCR. Both parents had a normal array profile and normal gene dosage by qPCR.

Patient 3
This boy is the first child of non-consanguineous healthy European parents (maternal age 32 years, paternal age 31 years) with unremarkable family history. Increased nuchal translucency was observed at 12 weeks of gestation. The amniocentesis karyotype was considered normal. At 22 weeks, an ultrasound revealed bilateral ectopic testes and dilated colon, and intrauterine growth retardation was noted at 32 and 36 weeks. He was delivered by cesarean section at 37 weeks because of fetal heart deceleration. He was transferred to the neonatal resuscitation unit, ventilated for 20 min, and placed in an incubator. Birth weight was 1,850 g (<5th centile), length 43 cm (<5th centile), and occipitofrontal circumference (OFC) 32.5 cm (<25th centile). Tests for congenital infections were negative. After 10 days, he was discharged with sequelae of congenital muscular torticollis. Feeding difficulties, axial hypotonia, and hyperlaxity were reported during the first months of life. At 10 months, his weight and length were below −2 SD, and his OFC decreased at −3 SD. His language development and motor milestones were delayed; he was able to stand up with support at 42 months and did not walk or talk at 4 years.  Figure 1C), occurring on the paternal allele. The deletion was confirmed by FISH with BACs targeting 7q22.1 (RP11-44M6 and DBACA-20A02, Integragen, Evry, France) and excluded in the parents.

Patient 4
The clinical and genetic characterization of this 14-year-old male was reported previously [15]. Briefly, he is the second child of a non-consanguineous healthy couple (maternal age 26 and paternal age 31 years) and presented with autism, ID, insulin-dependent diabetes, growth hormone deficiency, and mild dysmorphic features (Table 1). Standard karyotype showed an apparently balanced paracentric inversion of the long arm of chromosome 2 inherited from his healthy mother: 46,XY,inv (2) Figure 1D). Both imbalances occurred on the maternally derived chromosome and were confirmed by qPCR. FISH analysis revealed that the duplicated material was located at the telomeric end of chromosome 2, distal to the inverted region [15]. Both parents had normal SNP array profiles. The terminal deletion results in the 2q37 deletion syndrome, also known as brachydactyly mental retardation syndrome. Brachydactyly, reported in approximately 50% of affected individuals [20], was not observed clinically in patient 4. Genomic position in hg19 coordinates; b in patient 1, the 4p duplication is pathogenic; the deletion is considered benign; c in patient 4, the 2q37 deletion is pathogenic; the clinical significance of the duplication is unknown.

Patient 5
This female was referred at the age of 4 years and 8 months because of developmental delay, autistic behavior, and speech delay and diagnosed with autism after formal testing. She was born at term after an uneventful pregnancy, with weight and height in the normal range and OFC +1.9 SD. The parents were non-consanguineous; parental ages at the time of birth were 27 and 28 years for the mother and father, respectively. She walked at 22 months. At 3 years, she exhibited overgrowth, with height +3 SD and macrocephaly (+3 SD). Assessment of cognitive functioning at 6 years indicated severe ID. When re-evaluated at 11 years, she met criteria for autism according to the ADI-R, and the Vineland Adaptive Behavior Scales indicated developmental ages between 14 months (communication and socialization) and 37 months (motor skills). At the age of 13 years, all her growth parameters were increased (weight +3.5 SD, height +2 SD, OFC +4.2 SD), and she remained non-verbal. She had mild dysmorphic features, including narrow palpebral fissures, and short philtrum; her hands and feet were large, and a café-au-lait spot was present on a toe. A brain MRI performed at 9 years showed a Chiari malformation type 1 and mild hyperintensities of the white matter. The macrocephaly prompted screening for mutations and deletions in PTEN, with normal results [21]. The standard karyotype was 46,XX,inv(2)(p12q14.1)dn, which is considered a polymorphic cytogenetic variant. SNP array analysis revealed a de novo intragenic deletion of 179 kb in the RFX3 gene at 9p24.2 (3,344,203-3,513,286), on the paternal allele. qPCR confirmed a deletion of exons 2 to 4 of RFX3 in the patient and was normal in both parents.

Discussion
We identified de novo genomic imbalances in 5/18 patients with ASD and ABCR, of which four were considered pathogenic (4/18, 22%) and one of unknown clinical significance. If we exclude the four patients carrying common pericentric inversions, which a priori are not expected to be causative of the phenotype, the yield increases to 33% (4/12). This is in accordance with a previous study showing that 24% of patients with abnormal phenotype and a de novo two-point chromosome rearrangement  have a clinically significant cryptic genomic imbalance [10]. Three imbalances were related to the breakpoint regions, and two deletions occurred on unrelated chromosomes. In two patients, the SNP array analysis identified a terminal deletion and a duplication occurring on the same chromosome involved in the ABCR. Four out of five ABCR with a genomic imbalance occurred de novo: one rare pericentric inversion, one pericentric inversion of chromosome 2 considered as a chromosomal variant, and two reciprocal translocations. Two of the five patients with a de novo imbalance have a non-syndromic form of ASD. Our results are concordant with previous studies reporting that ABCR often hide more complex rearrangements regardless of the type of ABCR, particularly when they appear de novo [3,4,9,10], and when they are associated with an abnormal phenotype [3,5,6,[8][9][10]. Of the five de novo imbalances, three originated on the paternally inherited chromosome, in agreement with previous studies showing that male gametogenesis is more susceptible to such rearrangements [4,6,9]. Four of our patients carry common inversions considered variants without phenotypic effects. In all cases, the inversion was indeed balanced, confirming that common pericentric inversions of chromosomes 2, 5, 9, and Y are not implicated in abnormal phenotype by imbalances at the breakpoints, even when they occur de novo [22][23][24]. In a patient with a pericentric inversion of chromosome 2, a de novo deletion occurred on an unrelated chromosome.

Imbalances at the breakpoint region
Duplication of the Wolf-Hirschhorn region Patient 1, carrying a translocation (4;acro p), has two imbalances, a 541 kb subtelomeric 4p16.3 deletion and a directly adjacent inverted duplication spanning 4.24 Mb. The 4p16.3 deletion overlaps two previously reported 4p terminal deletions between 200 and 400 kb in phenotypically normal individuals and can be considered benign [25,26]. The phenotype of our patient is likely related to the proximal duplication, which encompasses the Wolf-Hirschhorn critical region (WHSCR), with two proposed loci, WHSCR1 and WHSCR2 (Figure 2). Partial 4p trisomy, including the WHSCR, is associated with variable clinical manifestations depending on the size of the duplicated segment. Clinical features include growth retardation, ID, dysmorphic features, and heart and renal abnormalities. Most cases reported thus far were identified cytogenetically and are therefore very large [27][28][29]. Six smaller molecularly defined duplications overlapping the WHSCR have been reported in the literature (Figure 2). Roselló et al. [30] described a boy with a 1.1 Mb 4p16.3 duplication presenting with ID, dysmorphic features, language delay, absence seizures, hyperactivity, aggressive behavior, and non-specific anomalies on MRI. He also has a terminal 4p16.3 deletion of 1.3 Mb, interpreted as benign. A terminal 2.3 Mb duplication of 4p16.3 was reported by Carmany et al. [31] in a girl with mild phenotype, including delayed language development, borderline normal IQ, mild hypertelorism, and normal height with advanced bone age. The duplication resulted from a reciprocal translocation and was associated with a 314 kb deletion of 17q25.3, considered probably benign. A 5 Mb terminal 4p16.2 duplication, very similar to the one observed in our patient, was identified by Wu et al. [32] in two individuals from the same family with a t(4;15) translocation, a 2-month-old boy and his paternal uncle with mild ID and poor language. The two subjects shared some characteristics, including similar facial features, widely spaced nipples and small penis. Similarly, Schonewolf-Greulich et al. [33] described a 3 Mb terminal 4p16.3 duplication in a three-generation family with macrocephaly, tall stature, speech delay, mild ID, and minor dysmorphic features. Finally, two smaller interstitial duplications overlapping the WHSCR have been reported. Hannes et al. [34] detected an interstitial 4p16.3 duplication of 560 kb encompassing both WHSCR1 and WHSCR2 in a male toddler with developmental and speech delay, hypotonia, seizures, dysmorphic features, hand malformation, and glaucoma. The duplication occurred at the breakpoint of an inversion between 4p16.3 and 4q22; given the severity of his features, it is possible that the inversion contributes to his phenotype by disrupting a gene at the breakpoint. A 506 kb duplication in 4p16.3 involving only WHSCR2 was reported by Cyr et al. [35] in a 16-month-old boy with some characteristics of the trisomy 4p syndrome, including psychomotor delay, prominent glabella, low-set ears, and short neck. Compared to the patient reported by Hannes et al., the only common features observed in these individuals were developmental delay and seizures. Our patient presented with mild ID and autism, without malformations or seizures. Taken together, these individuals support the observation that duplications of the WHSCR result in variable, usually mild, clinical manifestations. The most common features include developmental delay, speech delay, and mild cognitive deficits; dysmorphic features, when present, are usually mild. ASD was not mentioned in any of the subjects with 4p16.3 duplications including the WHSCR or in patients with cytogenetically detectable duplications. In a large CNV study in ASD, Pinto et al. [36] identified a male with a 4p16.3p16.1 duplication spanning 9.3 Mb, but he also had a 6.8 Mb 8p23.3p23.1 terminal deletion resulting from a de novo translocation. ASD has been reported in at least two patients with Wolf-Hirschhorn syndrome [37,38], suggesting that overexpressed genes within this region could be responsible for ID and ASD in our patient.
Inverted duplications with terminal deletions involving the short arm of chromosome 4 have been rarely reported in the literature [28,[39][40][41]. The majority of rearrangements with terminal deletion associated with a duplication in the same chromosome arm correspond to inverted duplication deletions (inv dup del) [42]. Three mechanisms have been proposed to explain such rearrangements [42]. All involve an intermediate dicentric chromosome which breaks during meiotic division to produce a monocentric duplicated and deleted chromosome. The first mechanism occurs because of a parental paracentric inversion: during meiosis pairing, both homologues create an inversion loop; a crossing over in the loop leads to the formation of a dicentric chromosome and a reciprocal acentric chromosome. A subsequent breakage of the dicentric allows the formation of a monocentric chromosome with the inverted-duplicateddeleted chromosome. The second mechanism occurs through non-allelic homologous recombination between inverted segmental duplications in the same arm. For both mechanisms, the inverted duplicated region and the deleted region appear separated by a single copy region. The third mechanism, which is probably the most frequent, involves U-type exchange between the sister chromatids; in this case, the duplicated and deleted regions are adjacent and not separated by a single copy region [43,44]. In the patient reported here, there is no single copy region between the deletion and duplication, suggesting a U-type exchange mechanism. Moreover, as a telomere capture mechanism is frequently used for stabilizing the broken chromosome ends [44], we suggest that the translocation between the short arm of an acrocentric chromosome and the derivative chromosome 4 was generated to stabilize it.

18p11.22p11.31 deletion
In patient 2, with a de novo translocation (5;18), we identified an interstitial deletion of 4.22 Mb in the 18p11.22p11.31 region. The 18p deletion syndrome, associated with terminal deletions of variable size, is characterized by typical facial features, growth retardation, and ID [45]. Our patient does not have any dysmorphic features or growth retardation. Pure interstitial deletions in this region have not been reported in the literature so far. The DE-CIPHER and ISCA databases include several patients with partially overlapping interstitial deletions of variable sizes ( Figure 3); although several are inherited from a normal parent, two are de novo, a 9 Mb deletion in a patient with autism (290283) and a 3 Mb deletion (286638, no phenotype information). The deletion in patient 2 overlaps 24 known genes; only 1 has a disease-associated OMIM entry, NDUFV2, involved in autosomal recessive mitochondrial complex I deficiency characterized by early-onset cardiomyopathy and encephalopathy. We speculate that haploinsufficiency of one or more genes included in the deletion could be implicated in the ASD phenotype. Two genes in the region have a high haploinsufficiency score, EPB41L3 and ANKRD12, both highly expressed in brain. EPB41L3 encodes band 4.1-like protein 3 (also known as protein 4.1B), which interacts with the synaptic cell adhesion molecule 1 (SynCAM1) to recruit NMDA receptors to synapses [46], and also acts as tumor suppressor. ANKRD12 (ankyrin repeat domain 12) is a member of the ankyrin repeats-containing cofactor family, which can inhibit the transcriptional activity of nuclear receptors through the recruitment of histone deacetylases.

2q37 terminal deletion
Patient 4 carries a maternal inv(2)(q14.2q37.3) with a terminal 2q37.3 deletion and a 2q14.1q14.2 duplication at the inversion breakpoints, which arose from a recombination event between the normal and inverted homologues 2 on maternal meiosis 1, followed by breakage of a dicentric chromosome [15]. The deletion involves the region implicated in the 2q37 deletion syndrome, frequently associated with ASD [38,47]. His phenotype is similar to that reported in patients with terminal 2q deletions and includes autism, ID, language delay, growth retardation, and mild facial dysmorphism. Haploinsufficiency of the HDAC4 gene is responsible for brachymetaphalangy and ID [48], but the contribution of one or more distal genes cannot be excluded since individuals with distal deletions not including HDAC4 have been reported with ID, ASD, and seizures. The 2q14.1q14.2 duplication includes 20 genes; no similar imbalances are reported in DECIPHER, ISCA, or the Database of Genomic Variants (DGV) so its clinical significance is unknown. Five smaller duplications with a common duplicated region of 440 kb (chr2:120,126,884-120,567,392) have been reported in DECIPHER; four are inherited from normal parents (one unknown inheritance), suggesting that they are probably benign.

Imbalances on unrelated chromosomes
In patient 3, the de novo pericentric chromosome 11 inversion was not associated with a genomic imbalance but the SNP array analysis revealed a 7q21.3q22.1 de novo deletion of 4.9 Mb encompassing over 100 genes. In this interval, only AP1S1 has been implicated in a neurological phenotype, a recessive disorder of copper metabolism. No similar deletions have been reported in the literature although several cases in DECIPHER and ISCA have smaller or larger deletions. Two inherited apparently balanced rearrangements mapping to this region have been reported in autism: a paracentric inversion (7q22.1q31.1) in two siblings with autism [49] and a translocation involving chromosome 7q22.1, just distal to NPTX2, in a subject with autism and ID [50].
Patient 5 has a pericentric inversion of chromosome 2, which is considered a benign cytogenetic variant. No CNV was observed in chromosome 2 but the SNP array identified a de novo 9p24.2 deletion interrupting the RFX3 gene. A 440 kb deletion including RFX3 was observed in one individual in a schizophrenia cohort [51]. Following-up on this finding, Sahoo et al. [52] analyzed 38,779 individuals referred for microarray testing and identified seven deletions encompassing the RFX3 gene, with sizes ranging from 177 kb to 3.5 Mb. Two deletions were de novo, two paternal, and three of unknown origin. The clinical features included autism, ID, and behavioral problems. In addition, one DECIPHER patient (248290) has a partial deletion of RFX3 but she also carries two large deletions in other chromosomes (no inheritance or phenotype information available). A similar deletion was reported in a patient from ISCA (nssv580652) with cataract and global developmental delay (unknown inheritance). DGV lists only one coding exon deletion (nsv892090), identified among 6533 Asians from the general population. Distal deletions of 9p of varying sizes are characterized by ID, trigonocephaly, dysmorphic facial features, and genital abnormalities; autism or autistic-like behavior are also common [53]. Two critical regions have been proposed for the 9p deletion syndrome, between 11 Mb and 16 Mb, and the first 2 Mb of 9pter [53], neither of which includes RFX3 (located at 3.2 Mb). RFX3 encodes a transcription factor required for ciliogenesis in mammals [54]. Many ciliopathy genes, such as those involved in Joubert syndrome, can present with neurodevelopmental phenotypes, including ID and ASD. Rfx3−/− mice show several hallmarks of ciliopathies, including left-right asymmetry defects, hydrocephalus, and corpus callosum agenesis; the phenotype of heterozygous mice was not reported [55,56]. Although the clinical significance of the RFX3 deletion identified in our patient is unknown, the results on Rfx3 mutant mice and the presence of rare RFX3 deletions in patients with ID or abnormal behavior suggest that further studies are warranted on the contribution of RFX3 loss of function to abnormal brain development.

Conclusions
In this study, we used genome-wide microarray analysis to elucidate the molecular complexity of microscopically balanced chromosome rearrangements in a series of 18 patients with ASD. We detected de novo imbalances related to the initial ABCR or occurring elsewhere in the genome in 22%. This work provides additional evidence that ABCR are often more complex than they appear at first and can hide cryptic rearrangements involved in the abnormal phenotype regardless of the type and inheritance of the rearrangement. Thus, ABCR associated with abnormal phenotype have to be analyzed with genomewide molecular cytogenetic techniques in order to exclude imbalances either within or outside the initial breakpoints. In the patients where a genomic imbalance was not detected by array analysis, higher resolution techniques based on next-generation sequencing [57,58]  The phenotype is indicated when available. The + sign indicates additional morphological features. Two genes within the deleted region in patient 2 with a high haploinsufficiency score are indicated in orange. Most of the deletions in the distal region are inherited, usually from normal parents, whereas two other de novo deletions as well as one inherited from a parent with a similar phenotype overlap the proximal region, highlighted in peach. DD, developmental delay; dn, de novo; ID, intellectual disability; inh, inherited; mat, maternal; pat, paternal; unk, unknown inheritance. might be necessary to explore the rearrangement at the nucleotide level.

Competing interests
The authors declare that they have no competing interests.
Authors' contributions ACT conceived the study with CB, participated in its design and coordination, carried out the molecular genetic studies, collected the clinical data, and drafted the manuscript. AV contributed clinical data and helped draft the manuscript. MP participated in the molecular genetic studies and prepared the figures. ED participated in the analysis and interpretation of data, prepared the figures, and helped draft the manuscript. RD, GN, FD, MG, SP, DH, JPS, AJ, AD, LP, CD, AA, PB, MC, ML, and CG contributed patient samples and clinical data. BB participated in the design and coordination of the study. CB conceived the study, participated in its design and coordination and wrote the manuscript. All authors read and approved the final manuscript.