Introduction

Developmental delay (DD)/intellectual disability (ID), characterized by a significant impairment of cognitive and adaptive functions, affects 1–3 % of the general population (Harris 2006; Maulik et al. 2011), with the majority of affected individuals remaining without proper diagnosis (Rauch et al. 2006; Pfundt and Veltman 2012). The etiology of DD/ID is heterogeneous, with both genetic and environmental contribution (Grayton et al. 2012). In addition to Mendelian DD/ID, one of the most common causes are microscopically visible chromosomal aberrations and submicroscopic copy-number variants (CNVs) (Morrow 2010; Regier et al. 2010). A high-resolution G-banded karyotype reveals chromosome abnormalities in 3–5 % of patients with idiopathic DD/ID (Shevell et al. 2003), and molecular cytogenetic analyses, e.g., fluorescent in situ hybridization (FISH) in the subtelomeric regions, provides diagnosis in an additional 3–6 % of cases (Koolen et al. 2004; Ravnan et al. 2006). In the past few years, application of chromosomal microarray analysis (CMA), including array comparative genomic hybridization (array CGH) and single-nucleotide polymorphism (SNP) arrays, has revolutionized the clinical diagnostics in patients with idiopathic DD/ID (Cooper et al. 2011; Kaminsky et al. 2011; Girirajan et al. 2012). Recently, it has been proven that CMA should be a first-tier clinical diagnostic test for individuals with DD, ID, autism spectrum disorders, and dysmorphic features (Miller et al. 2010; Battaglia et al. 2013). Clinically relevant CNVs, ranging in size from megabases to kilobases (Rodriguez-Revenga et al. 2007), have been detected in ∼10–20 % of cases (Menten et al. 2006; Stankiewicz and Beaudet 2007; Koolen et al. 2009) and have led to the identification of several novel microdeletions and microduplications associated with DD/ID (Slavotinek 2008; Vissers and Stankiewicz 2012), e.g., involving chromosomes 1q41q42, 9q22.3, 15q13.3, 15q24, 16p11.2, and 17q21.31 (Koolen et al. 2006; Redon et al. 2006; Sharp et al. 2006; Shaw-Smith et al. 2006; Ballif et al. 2007; Shaffer et al. 2007; Sharp et al. 2007, 2008). Refinement of the critical region of a known syndrome by the identification of atypical deletion (Cooper et al. 2011) may facilitate the detection of a dosage-sensitive gene(s) related to ID (Vissers et al. 2010); however, there are many CNVs for which the clinical significance may still remain unknown (Rodriguez-Revenga et al. 2007). Recently, clinical and biological interpretation of those variants and their genotype–phenotype correlation enabled the generation of a human genome morbid map (Cooper et al. 2011).

Here, we present the results of the application of genome-wide array CGH in a cohort of 256 patients with DD/ID, dysmorphic features, congenital malformations, or additional neurodevelopmental abnormalities. We identified 84 non-polymorphic CNVs in 69 patients, including 41 clinically relevant CNVs, 15 potentially novel pathogenic genetic loci for DD/ID, and, additionally, 28 CNVs of unknown clinical significance, likely to be non-pathogenic changes.

Materials and methods

Patients

We studied 256 patients with DD/ID, dysmorphic features, congenital malformations, or additional neurodevelopmental abnormalities. Among them, 234 patients had normal GTG banding analysis with at least 550-band resolution and 46 patients had a negative fragile X testing.

We provide a detailed clinical description of five patients (Supplementary material) and discuss their genotype–phenotype correlations. Patients 23 and 34 have de novo CNVs that we believe are pathogenic, and patients 45, 50, and 51 have potentially pathogenic CNVs.

DNA isolation

Genomic DNA was extracted from peripheral blood cells using a Puregene DNA Blood Kit (Qiagen, Gentra Systems, Minneapolis, MN), according to the manufacturer’s protocol. The reference DNA samples were obtained from phenotypically normal male and female controls.

Array CGH

Custom-designed exon-targeted clinical array CGH was performed using 105K V7.4 and 180K V8.0 or V8.1 microarrays designed in the Medical Genetics Laboratories at Baylor College of Medicine (BCM) (http://www.bcm.edu/geneticlabs/cma/tables.html) in cooperation with the Department of Medical Genetics at the Institute of Mother and Child and manufactured by Agilent Technologies (Santa Clara, CA). V7.4 consisted of 105,000 oligonucleotides having genome-wide coverage with an average resolution of 30 kb, while V8.0 and V8.1 OLIGO (180K) arrays have genome-wide coverage as well as exon coverage for over 1,700 known or candidate disease genes with an average of 4.2 oligos per exon and intronic gaps no larger than 10 kb (Boone et al. 2010). The microarray used in this study does not contain SNP probes and it does not detect regions of absence of heterozygosity (AOH). The results of studies using the updated version of this array were recently published by the BCM (Wiszniewska et al. 2013). Digestion, labeling, and hybridization were performed following the manufacturer’s instructions. The BCM web-based software platform and the homebrew IMiD-web2py software were used for chromosomal microarray analysis. All genomic coordinates are based on the March 2006 assembly of the reference genome (NCBI36/hg18). To verify the rearrangements identified by array CGH, depending on CNVs size, we used GTG banding and FISH. When available, blood samples were obtained from patients’ parents, and CNV inheritance was investigated.

Karyotype analysis

GTG banding analysis was performed according to the standard protocol in peripheral blood lymphocytes. The metaphases with 550-band resolution were analyzed in cases with CNVs 5 Mb or greater in size.

Fluorescent in situ hybridization (FISH) analysis

Confirmatory FISH experiments were performed to verify the presence of CNVs ranging in size from 150 kb to 5 Mb. FISH analyses were performed by standard procedures in phytohemagglutinin-stimulated peripheral blood lymphocytes using probes derived from bacterial artificial chromosomes. When available, blood samples were obtained from the patient’s parents and the origin of the identified CNVs was studied using FISH with the same probes.

Results

A total of 84 non-polymorphic CNVs were found in 69 of 256 patients studied. We divided the detected CNVs into three groups. The first group contains 41 CNVs considered to be clinically relevant (i.e., pathogenic for DD/ID) (Table 1). This group includes 18 imbalances greater than 5 Mb in size identified in 17 patients (not seen in standard cytogenetic studies) and 23 submicroscopic CNVs. In three cases, two pathogenic CNVs were detected: patient 9 had an unbalanced translocation der(10)t(10;20)(q26.2;q13.33); patient 10 had a terminal duplication dup(11)(p15) and a terminal deletion del(11)(q24), most likely a recombination product of a pericentromeric inversion inv(11)(p15q24); and patient 12 had a large 15q13.3q14 deletion in addition to the well-known recurrent microdeletion 16p13.11. We also identified two mosaic trisomies of chromosome 9 (pt 15 and pt 16) and monosomy of chromosome 7 in a 1-year-old patient (pt 17) with DD and combined immunodeficiency. Among 23 submicroscopic CNVs, we identified known recurrent rearrangements, e.g., deletions 15q11.2, 16p13.3, 17q11.2, and 17q21.31 and different-sized non-recurrent CNVs, e.g., deletions 1p36.31p36.33, 1q43q44, 5q14.3, 6q25, and 10q24.32, as well as duplications 3p21.1 and 19p13.3. In this group, we also identified two recently described microdeletions at chromosomes 4q21 and 17q24.2.

Table 1 CNVs clinically relevant for DD/ID

The second group consists of 15 CNVs potentially pathogenic for DD/ID (Table 2). In three cases, we identified rare CNVs: two de novo deletions at 13q12.11 and 5q35.3, and in one patient, a deletion at 10q21.3 and a duplication at 19p13.42p13.43.

Table 2 CNVs potentially pathogenic for DD/ID

In the third group, we classified 28 deletions and duplications of unknown clinical significance, some of which are likely non-pathogenic for DD/ID (Tables 1, 2, and 3). We found four individuals with BP1/BP2 duplication at 15q11.2 that is frequently found in the general population and considered as a benign event.

Table 3 CNVs of unknown clinical significance (likely non-pathogenic) for DD/ID

Moreover, we present the cumulative data of the studied cohort (Table 4), including patients with normal array CGH results (Supplementary Table 1).

Table 4 Summary of the studied cohort of 256 patients with DD/ID

Discussion

To determine the phenotypic consequences of the 84 identified non-polymorphic CNVs, we considered their type (deletion vs. duplication), size, gene content, inheritance pattern, and the available clinical or genomic database information. We classified them into three groups: 41 (16.0 %) known well-recognized causative for DD/ID, 15 (5.85 %) novel potentially pathogenic, and 28 (10.9 %) variants of unknown clinical significance (Tables 1, 2, and 3). The first group includes CNVs responsible for the well-known diseases and syndromes or published genomic imbalances that were considered as clinically relevant for DD/ID. The second group consists of novel CNVs potentially causative for DD/ID and includes changes that were not previously reported to be associated with DD/ID but contain genes that may contribute to our patients’ phenotypes. Moreover, we compared CNVs classified as the second group with the publically available databases. The same or similar sized aberrations were not found in the Database of Genomic Variants (DGV, http://projects.tcag.ca/variation/), Database of Chromosomal Imbalance and Phenotype in Humans Using Ensembl Resources (DECIPHER, http://decipher.sanger.ac.uk), or the International Standards for Cytogenomic Arrays database (ISCA, https://www.iscaconsortium.org). In the third group of variants of unknown clinical significance, we classified CNVs that contain genes that were not previously associated with DD/ID, are frequently found in the general population, or were inherited from a healthy parent.

From the second group of potentially pathogenic CNVs, we selected for the detailed description three novel rare, probably pathogenic, deletions containing candidate genes that may contribute to the abnormal phenotypes observed in our patients. Furthermore, in two of those cases (pt 51 and pt 45), deletions arose de novo that support their potential pathogenicity.

A de novo ∼250-kb deletion in 13q12.11, encompassing only two genes, PSPC1 (paraspeckle component 1; MIM 612408) and ZMYM5 (zinc finger, MYM-type 5), was identified in pt 51 with remarkably delayed psychomotor development, muscle hypotonia, unilateral microphtalmia with ptosis, congenital eye malformation, and facial dysmorphic features (Fig. 1a, d). Interstitial deletions of 13q12.11 are very rare (Der Kaloustian et al. 2011; Tanteles et al. 2011). Tanteles et al. (2011) described a de novo 2.9-Mb 13q12.11 deletion of 19 genes, including PSPC1 and ZMYM5, in a patient with near normal development and intellect but with scaphocephaly and facial dysmorphism. PSPC1 encodes a nucleolar protein that localizes to punctate subnuclear structures close to splicing speckles, known as paraspeckles. Paraspeckles may function in the control of gene expression via an RNA nuclear retention mechanism. Functionally, PSPC1 was proposed to participate in the regulation of transcription (Passon et al. 2011). ZMYM5 belongs to the zinc finger MYM-type family and probably has molecular functions, such as metal ion binding or zinc ion binding (Pastorcic and Das 2007). Although the phenotype–genotype correlation remains unclear and one smaller deletion within this region was reported in the control group (DGV; nsv455826) (Itsara et al. 2009), we propose that the identified de novo 13q12.11 deletion could contribute to the clinical features observed in our patient.

Fig. 1
figure 1

Results of array CGH analyses: a patient 51, showing a de novo ∼250-kb deletion in 13q12.11; b patient 45, demonstrating a de novo ∼720-kb deletion at 5q35.3; and c patient 50, showing a rare ∼1.4-Mb deletion in 10q21.3. The red dots denote the deleted region. Gene content in the deleted region on: d chromosome 13q12.11; e chromosome 5q35.3 (compared with the critical region of the Sotos syndrome and 5q35.3 subtelomeric deletion syndrome); and f chromosome 10q21.3 (UCSC genome browser, http://genome.ucsc.edu/)

In a girl (pt 45) with an atypical autism spectrum disorder and moderate mental impairment with absent expressive speech development, we detected a de novo ∼720-kb deletion at 5q35.3 that partially overlaps a common-sized ∼1.9-Mb recurrent deletion found in patients with Sotos syndrome (MIM 117550) but leaves the dosage-sensitive NSD1 gene (set domain protein 1; MIM 606681) intact (Fig. 1b, e). Rauch et al. (2003) described a novel 5q35.3 subtelomeric deletion distally adjacent to the Sotos common deletion region, and characterized by pronounced muscular hypotonia, postnatal short stature, and bell-shaped thorax with pectus carinatum. The deletion identified in our patient harbors 15 genes, including DBN1, B4GALT7, PROP1, and NHP2. DBN1 (drebrin E; MIM 126660) is a cytoplasmic actin-binding protein thought to play a role in neuronal growth and dendritic spine formation. It is a member of the drebrin family of proteins that are developmentally regulated in the brain. Shim and Lubec (2002) and Dun and Chilton (2010) suggested that a decreased amount of drebrin may lead to loss of spine plasticity and impaired dendritic arborization, which may underlie cognitive dysfunction. PROP1 (paired-like homeodomain transcription factor; MIM 601538) has both DNA-binding and transcriptional activation ability. Its expression leads to ontogenesis of pituitary gonadotropes, as well as somatotropes, lactotropes, and caudomedial thyrotropes. Heterozygous mutations in PROP1 have been reported in patients with pituitary hormone deficiency-2 (CPHD2; MIM 262600). Homozygous mutations of B4GALT7 (galactosyltransferase I; MIM 604327) have been described in patients with the progeroid form of Ehlers–Danlos syndrome (MIM 130070) characterized by an aged appearance, DD, short stature, craniofacial disproportion, generalized osteopenia, defective wound healing, hypermobile joints, hypotonic muscles, and loose but elastic skin (Okajima et al. 1999). Lastly, homozygous or compound heterozygous mutations in the NHP2 gene (nucleolar protein family A, member 2; MIM 606470) have been found in individuals with autosomal recessive dyskeratosis congenita-2 (DKCB2; MIM 613987). Given that the 5q35.3 deletion in our patient arose de novo and the fact that the deleted genes are associated with neurodevelopmental disorders and play a role in neuronal processes, we suggest that the identified CNV is potentially pathogenic for the described clinical features.

In patient 50 with short stature, moderate ID, and history of Hirschsprung disease and congenital heart defect, we found a rare ∼1.4-Mb deletion in 10q21.3 encompassing 16 genes (Fig. 1c, f) and a small ∼400-kb duplication in 19q13.42q13.43. Deletions in 10q21.3 have never been reported in patients with DD/ID, thus, it is challenging to assess its clinical pathogenicity. However, the size and gene-rich content suggest that the 10q21.3 deletion could be pathogenic. Analysis of the maternal sample revealed normal result; unfortunately, the paternal sample was unavailable. Five genes, including SIRT1 (sirtuin 1; MIM 604479), DNA2 (DNA replication helicase 2; MIM 601810), TET1 (Tet oncogene family, member 1; MIM 607790), CCAR1 (cell division cycle and apoptosis regulator 1; MIM 612569), and DDX50 (dead/h box 50; MIM 610373) represent dosage-sensitive genes (Huang et al. 2010); however, to date, no phenotype associated with haploinsufficiency of any of these genes has been reported. In addition, SLC25A16 (solute carrier family 25, member 16; MIM 139080) encodes a protein that is localized in the inner membrane and facilitates the rapid transport and exchange of molecules between the cytosol and the mitochondrial matrix space. This gene was proposed to play a role in Graves disease (MIM 275000). Of note, other SLC family genes, SLC9A9 (MIM 608396), SLC6A4 (MIM 182138), and SLC25A12 (MIM 603667), are causative for autism. Other genes mapping in the deleted region, including DNAJC12 (MIM 606060), HERC4 (MIM 609248), PBLD (MIM 612189), HNRNPH3 (MIM 602324), RUFY2 (MIM 610328), STOX1 (MIM 609397), and CTNNA3 (MIM 607667), have been linked to late-onset Alzheimer’s disease (AD6; MIM 605526). CTNNA3, encoding the alpha-3 catenin and likely responsible for the formation of stretch-resistant cell–cell adhesion complexes (Weiss et al. 2009), has been associated with late-onset Alzheimer’s disease in females (Miyashita et al. 2007). Bradley et al. (2010) suggested that, if CTNNA3 is involved in Alzheimer’s disease, it is not through a loss-of-function mechanism. Moreover, we suggest that CTNNA3 and/or MYPN (myopalladin, MIM 608517), deleted 10q21 region, can be responsible for the congenital heart defect observed in our patient. CTNNA3 has been considered as a candidate for the form of dilated cardiomyopathy linked to 10q21q23 (CMD1C; 601493) because of its high expression in the heart. MYPN is a component of the sarcomere that tethers nebulette in cardiac muscle to alpha-actinin and may play signaling roles in targeting and orienting nebulin to the Z line during sarcomere assembly. The associated ∼400-kb duplication in 19q13.42q13.43 harbors seven genes (NLRP8; MIM 609659, NLRP5; MIM 609658, ZNF787, ZNF444; MIM 607874, GALP; MIM 611178, ZSCAN5B, and ZSCAN5A), none of which represent a dosage-sensitive gene. Therefore, we classified this duplication as a variant of unknown clinical significance.

Among our 41 CNVs known as being pathogenic for DD/ID, we elected to discuss two deletions, 4q21 and 17q24.2, which have been described recently as responsible for microdeletion syndromes associated with DD/ID. Moreover, in those patients (pt 23 and pt 34), previously performed genetic tests were negative (Table 1), which demonstrates the usefulness of array CGH as a great tool for identifying the etiology of idiopathic DD/ID.

A de novo ∼3.1-Mb deletion at 4q21.21q21.22 identified in patient 23 overlaps CNVs reported in the literature (Harada et al. 2002; Friedman et al. 2006; Bonnet et al. 2010; Lipska et al. 2011). Bonnet et al. (2010) defined a 1.37-Mb critical region containing five genes, PRKG2 (MIM 601591), RASGEF1B (MIM 614532), HNRNPD (MIM 601324), HNRPDL (MIM 607137), and ENOPH1, and proposed that PRKG2 and RASGEF1B are the best candidate genes responsible for the 4q21 deletion syndrome (MIM 613509) characterized by severe ID, lack of speech, hypotonia, significant growth restriction, and distinctive facial features. Our patient enables a better delineation of a novel 4q21 microdeletion syndrome and further supports the proposed contribution of haploinsufficiency of PRKG2 and RASGEF1B.

We also identified a de novo ∼1.9-Mb deletion at 17q24.2 (pt 34), harboring the smallest region of overlap of four deletions recently reported by Vergult et al. (2012). The shared clinical features include ID, speech delay, truncal obesity, and similar facial gestalt. The ∼713-kb critical region contains five genes, including PRKCA (MIM 176960), a cluster of three CACNG genes encoding the gamma subunit of a voltage-dependent calcium channel, CACNG5 (MIM 606405), CACNG4 (MIM 606404), CACNG1 (MIM 114209), and HELZ (MIM 606699). The PRKCA gene encodes the serine- and threonine-specific protein kinase C alpha that plays an important role in many different cellular processes and is the most important candidate gene for many of the observed clinical features. Investigations of more patients with 17q24.2 deletions would strengthen the genotype–phenotype correlation in this newly reported microdeletion syndrome.

Moreover, in two patients (pt 15 and pt 16) with normal standard cytogenetic results, we detected chromosomal aneuploidy in the form of a mosaic trisomy 9 using array CGH. In both cases, chromosomal mosaicism was confirmed by retrospective GTG banding analysis that revealed low-level mosaicism for trisomy 9 of 12.5 % and 8 %, respectively. Additionally, in patient 17 with DD and combined immunodeficiency and normal karyotype, we found a monosomy of chromosome 7 in his peripheral blood leukocytes. Chromosome 7 monosomy and 7q deletion have been frequently found in patients with myelodysplastic syndrome (MDS). MDSs are a heterogeneous group of stem cell disorders characterized by ineffective hematopoiesis, dysplastic changes in bone marrow and peripheral blood, and the risk of transformation to acute myeloid leukemia (AML) (Cordoba et al. 2012). We suggest that the identified aberration is responsible for the phenotypic abnormalities observed in our patient.

Additionally, our analyses of the DD/ID cohort revealed patients with more than one clinically relevant CNVs. Apart from patients 9, 10, and 12 with two pathogenic CNVs, we identified two patients (pt 39 and pt 46) with two rare potentially pathogenic changes. A complex clinical presentation in those patients supports a second-hit model, in which the compound effect of multiple CNVs, including those of unknown pathogenic significance, contributes to the phenotypic heterogeneity (Girirajan et al. 2012).

These results further demonstrate that array CGH, in addition to the identification of CNVs and chromosomal aneuploidies, also enables the detection and estimation of their low-level mosaicism that may remain undetected by conventional cytogenetic methods (Cheung et al. 2007). Furthermore, our results support the usefulness of array CGH in the identification of the aneuploidy of cells under-represented in the T-cell population missed by routine chromosome analysis. However, apart from balanced aberrations (e.g., translocations, inversions), array CGH without SNP oligonucleotides cannot detect uniparental disomy.

In summary, we found that 69 of 256 patients with DD/ID carry one or more CNVs, in 38 cases responsible for the observed clinical features and in 13 patients potentially pathogenic for DD/ID. Our results further confirm the usefulness of array CGH in the detection of pathogenic CNVs in patients with idiopathic neurodevelopmental disorders.