The clinical assessment of the patients comprised a thorough medical history also attentive to investigating systemic symptoms and sleep, neurological-behaviour examination, developmental/cognitive assessment, and basic metabolic screening. Prior to array-CGH analysis, all patients were first evaluated by using different genetic tests for the common genetic causes of NDDs, including karyotype analysis, and tests for Fragile X syndrome. Written informed consent was obtained from the patient’s parents or legal representative. This study was reviewed and approved by the Ethics Committee of the Italian Regione Liguria (R. P. 001/2019).

Array-CGH analysis was performed on DNA samples, extracted from peripheral blood, using a whole-genome 180 K Agilent array with ∼13 Kb overall median probe spacing (Human Genome CGH Microarray, Agilent Technologies, Santa Clara, CA, USA). Data were analyzed using Agilent CytoGenomics and genomic positions reported according to the human genome assembly (GRCh37/hg19). All detected CNVs were tested for inheritance by hybridization of the parental DNA with the same array platform. De novo CNVs were confirmed either by FISH analysis or by quantitative real-time PCR as already reported (Vaccari et al., 2014; Tassano et al., 2015). To assess the clinical significance of the detected CNVs, we used several tools. Aberration segments were reviewed using GRCh37 hg19 of UCSC Genome Browser (http://genome.ucsc.edu/index.html). We annotated all detected CNVs and CNV-encompassed genes across public databases: Genomic Variants Database (DGV) (http://dgv.tcag.ca/dgv/app/home), DECIPHER (https://decipher.sanger.ac.uk/), Clinical Genome Resource (ClinGen) (https://clinicalgenome.org/), the Online Mendelian Inheritance in Man (OMIM) (http://www.omim.org), Simons Foundation Autism Research Initiative (SFARI) database (https://gene.sfari.org/), databases of mouse models (Mouse Genome Informatics, MGI) (http://www.informatics.jax.org). Literature mining was also performed. We considered only variants present in <0.1% of control individuals. We checked CNVs for the presence of dosage-sensitive genes (haploinsufficiency/triple sensitivity score) by using Clinical Genome Resource (ClinGen) consortium resource (https://dosage.clinicalgenome.org/). We also considered probability of intolerance to loss-of-function (LoF) variants (pLI) as reported in the Genome Aggregation Database (gnomAD, (https://gnomad.broadinstitute.org/). Genes with pLI scores of 0.9 or higher are extremely intolerant to heterozygous LoF variation, and thus haploinsufficient (Lek et al., 2016).

We then investigated deleted/duplicated regions for encompassing topologically associating domains (TADs), or their boundaries (TDBs), using Hi-C (http://3dgenome.fsm.northwestern.edu/). Indeed, TADs data from genome-wide higher order chromatin interaction data in human embryonic stem cells (h-ESC), h-ESC derived neural progenitor cells (H1NPC) (Dixon et al., 2015), and brain cortex (Schmitt et al., 2016), were downloaded (http://chromosome.sdsc.edu/mouse/download.html) (see Supplementary file) and mapped to hg19 coordinates using the UCSC browser. We then investigated identified CNV-genes, not previously associated with genetic diseases, for relevance to NDDs by considering different criteria: expression in the brain; reported mutations in the mouse associated with a neurodevelopmental phenotype; interaction with genes known to be associated with a neurodevelopmental disorder. To identify CNV disrupted genes with high brain expression, we used the BrainSpan database (http://www.brainspan.org/static/download.html) (Huckins et al., 2019) and extracted the list of genes with average log2 RPKM >4.5 (the top 18%), as also reported (Addis et al., 2018; Cerminara et al., 2021). We further investigated brain expression of genes at single gene isoforms using GTEx data (https://www.gtexportal.org/home/). We searched for gene-specific mouse models presenting a neurodevelopmental phenotype using MGI database. CNV-genes coding for microRNAs (miRNAs) were further investigated to detect the presence, among their possible targets, of any genes known to be associated with NDDs (Supplementary file) or to be involved in brain development and/or function by using miRDB (http://mirdb.org/), Target Scan (http://www.targetscan.org/vert_72/) and DIANA TOOLS (http://diana.imis.athena-innovation.gr/DianaTools/index.php).

We used GeneCodis4 (Tabas-Madrid et al., 2012) to unveil enrichment of annotations, as already reported (Cerminara et al., 2021). Briefly, all the genes obtained from either duplicated or deleted regions were used as input in GeneCodis4 together with known NDD-associated genes, as those related to ASD and reported in SFARI database and those reported in OMIM as associated with intellectual disability (see Supplementary file). This tool allows the classification of genes according to their putative biological function by screening the Gene Ontology (GO), OMIM, Panther, and KEGG Pathways. In the analysis, the hypergeometric test was applied followed by the false discovery rate correction (FDR) with a cut-off of 5% to determine which annotations were significantly enriched. For GO analysis, various hierarchical levels of the annotation data structure were used. All genes of the top gene ontology terms, including candidate and known NDD genes, were then projected onto the STRING network (v11) (Szklarczyk et al., 2019). Edges within the STRING network were thresholded at 0.4, according to the authors’ recommendation. A graphical representation of the GeneCodis4 and STRING results was obtained by using Cytoscape tool (Shannon et al., 2003).

To evaluate the pathogenicity of variants we focused on the following criteria: 1) patients with multiple CNVs, enclosing NDD genes, which could have additive effects on the patients’ phenotype on the basis of results from gene enrichment and protein-protein interactions analyses; 2) patients with CNVs encompassing regions that control the expression of known NDD genes, including those with potential long-range effects; 3) patients whose CNVs encompass genes not yet reported as NDD-associated, but whose expression profile and function suggest a possible involvement in the disease, when at least another patient is reported with a similar CNV and phenotype in publically available databases. Results of this analysis were graphically reported using UCSC and specific Custom Tracks. In particular we used RefSeq Curated for gene representation and OMIM genes to underline genes already associated with disease. Decipher track was manually modified; we selected patients with CNVs similar to those of our patients, that is either deletions or duplications according to those found in our patients and overlapping approximately the same region and genes. We selected Decipher CNVs only if their related patients had a NDD phenotype (we kept out patients for whom we had no information about their phenotype). The CNVs thus selected were further investigated for being present in the same patient with other secondary CNVs overlapping additional NDD genes, and in this case reported in figures (marked with letters). Developmental Delay case track includes manually selected CNVs similar in size and gene content to those of our cases; we proceeded analogously also for Developmental Delay Control CNVs. A track reporting SFARI genes was added to underline the presence of a known ASD associated gene in the genomic region shown in figures. Brain-expressed gene track represents genes with a prevalent expression in the brain according to the BrainSpan data processed as detailed above, useful to highlight possible new NDD candidate genes.

The 12 patients we report were all born to non-consanguineous parents. After clinical evaluation and follow up, the employment of specific tests and instruments, all subjects showed a complex phenotype, each of them characterized by different neurological features and other associated impairments and comorbidities. The main clinical features were represented by a variable degree of intellectual disability, in three cases associated with a diagnosed ASD, and epilepsy. Language disorder, self-aggressive behaviours and sleep disturbances could be also observed (see Table 1).

Four patients presented with two CNVs, overlapping known NDD genes, and inherited, respectively, by each of their healthy parents. We thought that in these cases, CNVs could additively contribute to the patient’s phenotype (Table 2).

The first case, IGGAC13, is a male patient with a maternal deletion of chromosome 7 involving CNTNAP2 and a paternal deletion of chromosome 11 involving LRRC4C (Figures 1A,B). Both genes are present in the SFARI database. Few CNVs are reported to overlap these two genes in control individuals and in NDD patients according to Developmental Cases and Decipher databases. Of note, NDD patients reported in Decipher and represented in Figure 1 carried one additional CNV involving other NDD-associated genes. Indeed, a case has been previously reported with two deleterious variants, one of which disrupting LRRC4C and one affecting another SFARI-reported gene DPP6 (Maussion et al., 2017). The two variants, as in our case, were inherited one from the mother and one from the father, both unaffected.

CNTNAP2 encodes a member of the neurexin family which functions in the nervous system as cell adhesion molecule. LRRC4C (NGL1) is a specific binding partner for netrin G1, which is a member of the netrin family of axon guidance molecules. In fact, both molecules, CNTNAP2 and LRRC4C, are implicated in synapse formation (Um and Ko, 2017). We hypothesized that, while half dosage of each of these two molecules is compatible with a normal development, half dosage of both genes can impair synapse correct development and function. In fact, gene enrichment and String analysis showed that both genes are members of the KEGG cell adhesion molecule pathway and interact with other known NDD genes (Figure 1C; Supplementary Table S1).

The second case is a male patient (IGGAC14) that inherited from his healthy father a deletion of chromosome 17 encompassing a SFARI gene, DNAH17, and its antisense gene DNAH17-AS1. DNAH17 encodes a heavy chain associated with axonemal dynein, a member of microtubule-associated motor protein complexes. From his healthy mother, the patient inherited a duplication involving SHOX on chromosome X (Figure 2), also reported in the SFARI database. SHOX belongs to the paired homeobox family and is located in the pseudoautosomal region 1 (PAR1) of X and Y chromosomes. Enrichment of SHOX microduplications in NDD cases, particularly in those with ASD, has been described (Tropeano et al., 2016). Gene enrichment evidenced a function of SHOX in the regulation of transcription by RNA polymerase II, and of DNAH17 in microtubule-based movement, two important biological processes in neurodevelopment disorder (Figure 2; Supplementary Table S1). As shown in Figure 2A, other patients with NDD phenotype are reported in Decipher database to have duplications of SHOX and a secondary CNV involving neurodevelopmental genes. Indeed, a chromosome 17 deletion similar to that of our patient was reported in Decipher database in a case with NDD phenotype having a secondary CNV also overlapping developmental genes (Figure 2B).

We observed a family with two affected siblings, a male (IGGAC15) and a female (IGGAC16), showing discordant phenotypes. They inherited from their unaffected mother a deletion encompassing a syndromic region at 15q11.2 (Figure 3A). The siblings also shared a duplication involving the SFARI gene SYNGAP1 (Figure 3B). Of note, the SYNGAP1 duplication was not observed in parents, at least in their blood-extracted DNA, thus leading to suppose mosaicism in either father or mother. Patient IGGAC16 also inherited a paternal, recurrent deletion involving the SFARI gene RBFOX1 (Figure 3C). Interestingly, loss of function variants of SYNGAP1 have been identified in patients with ASD and intellectual disability with or without epilepsy. However, to the best of our knowledge, no cases with duplication or gain of function variants have been reported so far. Overexpression of SynGAP was reported to block neurite outgrowth by a mechanism that involves Ras-like GTPase cascade (Tomoda et al., 2004), to produce a depression of AMPAR-mediated excitatory postsynaptic currents in neurons (Rumbaugh et al., 2006), and to regulate decreasing and increasing miniature excitatory synaptic currents in hippocampal neurons (McMahon et al., 2012). In fact, some phenotypic differences between the two siblings were observed. The male patient showed a mild phenotype characterized by a borderline cognitive functioning associated with atypical behaviours and psychomotor delay in early infancy, his sister had a more complex phenotype with mild cognitive impairment, irregular sleep rhythms, hyperactivity, coordination, and language disorders. We could speculate that the two CNVs shared by the two siblings can explain the phenotypic features shared by both of them, while the third CNV could have worsened the phenotype of the female patient.

Four patients reported double CNVs involving genes or genomic regions not yet described as NDD-associated but that could act in concert, similarly to the four cases described above, with a potential impact on the patients’ phenotype (Table 2).

The first case (IGGAC06) carried a maternal deletion of chromosome 9 and a non-maternal deletion of chromosome 11; as genetic analysis of the patient’s father was not available, we cannot determine if the chromosome 11 deletion was de novo or paternally inherited (Figures 4A,C). The chromosome 9 deletion encompassed PTPRD, which encodes a protein-tyrosine phosphatase receptor, and PTPRD-AS2, a long non-coding RNA (lncRNA) gene known as PTPRD antisense RNA 2 (head-to-head) predicted to make expression of mRNA more stable and thus enhancing PTPRD protein production. Thus, a reduced expression of PTPRD can be hypothesized because of the deletion of the first two untranslated exons of the gene and of PTPRD-AS2. A total of 7 deletions encompassing exonic regions of PTPRD are reported in controls, while several CNVs and, in particular, 16 deletions and one intragenic duplication involving coding regions of the gene are present among cases in Developmental Delay and Decipher databases with NDD phenotypes. Five among Decipher cases have an additional CNV involving brain expressed genes that, analogously to those reported in patient IGGAC06, could have an additive effect in causing patient’s phenotype. Interestingly, PTPRD is predicted to be intolerant to loss of function variants (pLI = 1) and to haploinsufficiency mechanisms (score = 0.75; value above 0.5 are predicted to be haploinsufficient) (Shihab et al., 2017). PTPRD has a role in synaptic adhesion and synapse organization and can bidirectionally induce pre- and postsynaptic differentiation of neurons by trans-synaptically binding to interleukin-1 receptor accessory protein (IL1RAP) (Yoshida et al., 2012; Yamagata et al., 2015). In the mouse, PTPRD was shown to regulate neurogenesis (Tomita et al., 2020) and mice lacking PTPRD showed impaired learning with enhanced hippocampal long-term potentiation (Uetani et al., 2000), thus suggesting a role for PTPRD in neurodevelopmental disorders. In OMIM database, PTPRD is not associated with a disorder, and its implication in human disease is still to be completely clarified. Homozygous deletions of PTPRD were observed in a patient with intellectual disability and trigonocephaly (Choucair et al., 2015) and his family members carrying the same deletions were reported to be unaffected. Instead, a heterozygous de novo splicing variant of the gene was reported in a girl with moderate nonsyndromic developmental delay (Yan et al., 2019). Furthermore, genetic studies showed an oligogenic association of PTPRD variants with obsessive-compulsive disorder (Mattheisen et al., 2015) and with restless leg syndrome (Schormair et al., 2008). Overall, results from various genetic studies are consistent with both major/oligogenic and modest/polygenic contributions of common and rare PTPRD variations in neurological, behavioural, and neurodevelopmental disorders (Uhl and Martinez, 2019). Gene enrichment and String analysis showed PTPRD interacting with many NDD genes playing a role in synaptic membrane adhesion and neuronal development (Figure 4B; Supplementary Table S1).

The chromosome 11 deletion encompassed BUD13, which encodes a component of the retention and splicing (RES) complex of the spliceosomal complex. In zebrafish, the lack of BUD13 caused defects in intron splicing specifically in genes with neurodevelopmental regulatory functions, thus resulting in a decrease of differentiated neurons and brain developmental defects (Fernandez et al., 2018). GeneCodis analysis showed an enrichment of BUD13 with various NDD genes in the GeneOntology category of mRNA splicing, via spliceosome, a biological process with an important role in neurodevelopment (Figure 4D, Supplementary Table S1).

Patient IGGAC08 has a complex phenotype mainly characterized by language disorder, hyperactivity, learning difficulties, borderline cognitive level, motor skill impairment, and short stature (Table 1). His mother also had school difficulties and showed dyslexia in childhood. Mother and son share a deletion of chromosome 4q34.1 encompassing GLRA3 which encodes the alpha-3 subunit of the neuronal glycine receptor, a ligand-gated ion channel. In addition, patient IGGAC08 had a de novo deletion involving hsa-mir-4465 and a de novo duplication including INPP5A (Table 2 and Figures 5A–C). GLRA3 receptors are expressed in spinal cord, brainstem, hippocampus, amygdala, striatum, and cortex. The glycine receptor is a glycoprotein composed of 5 subunits, three α and two β subunits. The alpha subunits bind the glycine ligand while the beta subunits bind to gephyrin, a cytosolic protein required for a regulated synaptic aggregation and clustering of these receptors.

Gephyrin has functional links with several synaptic proteins, mutations of which have been reported in various neurodevelopmental disorders (Choii and Ko, 2015; Kim et al., 2021). Half dosage of GLRA3 may contribute to impair gephyrin-mediated aggregation and post-synaptic clustering.

Loss of PTEN, a SFARI gene, can cause postsynaptic and presynaptic changes in excitatory and inhibitory connectivity. In vitro analyses have recently demonstrated that a miRNA (microRNA-301a) can downregulate PTEN after glycine receptor activation (Chen et al., 2016). Tao et al. (2019) showed that miR-4465 significantly inhibited the expression of PTEN, upregulated phosphorylated AKT and ultimately inhibited autophagy by activating mTOR in HEK293, HeLa, and SH-SY5Y cells (Tao et al., 2019). Half dosage of hsa-mir-4465 in our patient is expected to cause upregulation of PTEN, contributing to postsynaptic impairment. MiRNAs are important regulators of brain development and neuronal function, and have been associated with a variety of nervous system diseases, including ASD (Cheng et al., 2018; Wu et al., 2020).

In mice, the deletion of Inpp5a causes perinatal lethality in 90% of the homozygous mutants, early onset ataxia and relatively small stature in surviving mutants. Heterozygotes do not exhibit obvious motor coordination impairment unless challenged in motor skill test (Yang et al., 2015). Inpp5a is a downstream effector of signalling from mGlu1 receptor, one of the two members of group I metabotropic glutamate receptors, that regulates synaptic plasticity, particularly in cerebellar Purkinje cells. The mGlu1 receptor activity needs to be fine-tuned and balanced for normal motor coordination. Alteration of expression, either enhancement or decrease, of group 1 mGlu metabotropic receptors in mice is known to alter their activity, which results in defects of motor coordination (Rossi et al., 2013; Bossi et al., 2018). The deletion of INPP5A in this patient may be in part responsible for his impairment in motor skills.

IGGAC10 patient showed a complex phenotype mainly characterized by intellectual disability and epilepsy (Table 1). He carried a maternal deletion encompassing SYNCRIP and a paternal duplication involving the upstream regulatory genomic region of ADCY5. Both CNVs overlap a TDB between two flanking TADs described in different tissues, including brain cortex, which could also contribute to dysregulating the expression of implicated genes (Figures 6A,B).

SYNCRIP encodes the synaptotagmin-binding cytoplasmic RNA-interacting protein, a member of the cellular heterogeneous nuclear ribonucleoprotein (hnRNP) family that plays a role in multiple aspects of mRNA maturation. Missense and truncating variants of SYNCRIP were found in ASD patients with more severe phenotypes (Guo et al., 2019) and in patients with severe non-syndromic sporadic intellectual disability (Rauch et al., 2012), respectively. ADCY5 encodes a member of the membrane-bound adenylyl cyclase enzymes which mediate G-protein-coupled receptor signalling through the synthesis of the second messenger cAMP. According to SFARI database, missense and frameshift variants of ADCY5 have been reported in ASD patients. Gain of function variants of ADCY5 have been associated with a broad range of movement disorders, most notably chorea, dystonia and myoclonus (Vijiaratnam et al., 2019).

Thus, the two CNVs could cause, by a mechanism also implicating a modification of chromatin conformation, a dysregulation of both SYNCRIP and ADCY5 contributing to the complex phenotype observed in this patient.

IGGAC07 patient carried a de novo duplication of chromosome 14, and two maternally inherited duplications of chromosome 12 and chromosome 22, respectively (Table 2; Figures 7 A–C).

Interestingly, chromosome 14 duplication encompassed ABHD4, a gene recently shown to be involved in developmental anoikis, a mechanism of cell death in the prenatal brain preventing from survival of misplaced cells (Laszlo et al., 2020). Indeed, ABHD4 is involved in N-acylethanolamine biosynthesis, including the endocannabinoid molecule anandamide which has various physiological functions such as the regulation of synaptic plasticity and apoptosis. ABHD4 expression is tightly controlled spatially and temporally in the developing brain and a misregulation of its expression due to the duplication could contribute to the patient’s phenotype. Duplications encompassing ABHD4 were annotated in only one control individual, while various cases were reported both in Decipher and in Developmental Delay databases with duplications larger than that observed in our patient and a smaller one sharing with our patient ABHD4 duplication.

The maternal duplication on chromosome 22 involved a region with triplosensitivity including GNAZ, the Guanine Nucleotide-binding protein (G protein) Alpha Z polypeptide encoding gene. In this region, many duplications are reported in both Decipher and Developmental Delay databases, most of them extending over the short region containing GNAZ. Only a total of 4 control individuals are reported in either Developmental Delay Control or in DGV databases with duplications overlapping the GNAZ gene. GNAZ is a member of the Gαi subfamily of heterotrimeric G proteins, and couples to G-protein-coupled receptors triggering important pathways in neuronal development (Hultman et al., 2014). The duplication of chromosome 12 entirely encompassed WSCD2, which encodes a WSC Domain-containing protein with a still unclear function. Interestingly, genome wide association analyses found significant results for variants in WSCD2 in patients with psychiatric disorders (Lo et al., 2017). Finally, we could hypothesise that all the de novo and the inherited duplications could have jointly contributed to the NDD phenotype of patient IGGCA07.

We found 4 patients with CNVs that could affect expression of NDD genes throughout an indirect and/or long-range effect (Table 1 and Table 2).

One case (IGGAC04) presented with very severe psychomotor delay, hypotonia, and epilepsy (Table 1), carrying a de novo heterozygous deletion of chromosome 5 (Figure 8A). This deletion encompasses 6 genes, two of them expressed in the nervous system: MEF2C-AS1, which encodes a lncRNA annotated as antisense RNA of MEF2C, and ADGRV1, which encodes a member of the G-protein-coupled receptor superfamily associated with familial febrile seizures and Usher syndrome 2 (OMIM 602851). MEF2C (myocyte enhancer factor-2, OMIM 600662) is a transcription factor playing a role during development, mostly in myogenesis and neurogenesis. MEF2C haploinsufficiency has been shown to be responsible for severe cognitive deficit with stereotyped movements, epilepsy, and/or cerebral malformations (Le Meur et al., 2010). It has been recently demonstrated that, during myogenic differentiation a lncRNA, the OIP5-AS1, through its complementarity with MEF2C 3′UTR, enhances MEF2C mRNA stability thus promoting MEF2C expression and ultimately enhancing myogenesis (Yang et al., 2020). We hypothesized that during neurogenesis, analogously to OIP5-AS1, MEF2C-AS1 could enhance MEF2C mRNA stability and promote MEF2C expression and neurogenesis. In this view, the heterozygous deletion observed in patient IGGAC04 could reduce the expression of MEF2C-AS1 causing a decrease of MEF2C, thus interfering with neuronal differentiation and contributing to the patient’s phenotype. Although as a consequence of MEF2C-AS1 deletion a gene loss-of-function is likely, functional experiments would be required to exclude some compensatory mechanisms, as up-regulation of the healthy allele. The deletion of ADGRV1 could be responsible for the epilepsy phenotype. Of note, we found one patient in the Decipher database (Patient 251716) carrying a deletion overlapping the one of our patient and having similar clinical features characterized by intellectual disability, muscular hypotonia, and seizures. No deletions overlapping the same region were present among controls (Developmental Delay and DGV Controls). The chromosome 5 region deleted in these patients overlaps a TDB described in different tissues, including those relevant for neurodevelopment (e.g., Cortex), according to their relative Hi-C maps (Figure 8A), which could further complicate the fine regulation of expression of the overlapped genes, including MEF2C-AS.

One case (IGGAC01) showed a complex phenotype mostly characterized by psychomotor delay, autistic features, hypotonia, epilepsy, gastrointestinal and sleep disturbances (Table 1), and carried a de novo heterozygous duplication of chromosome 16 (Figure 8B). The duplication involved, in particular, two OMIM reported genes, USP7, encoding the Ubiquitin Specific Peptidase 7, and GRIN2A, encoding the NR2 subunit of the N-methyl-D-aspartate (NMDA) glutamate receptor. USP7 has been shown to regulate the ubiquitination of a variety of proteins with functions in different biological processes as oxidative stress response, histone modification and regulation of chromatin remodelling (Khoronenkova et al., 2011). Deletions, truncating and missense variants of USP7 have been associated with the Hao-Fountain syndrome (OMIM 616863), characterized by speech delay, autistic spectrum disorder, attention-deficit hyperactivity disorder, sleep disturbances and gastroesophageal reflux disease (Fountain et al., 2019). All these clinical features are present in our case who has the complete duplication of USP7. Notably, in syndromic forms of ASD, duplications in addition to deletions of USP7 have been reported (Sanders et al., 2011), thus supporting the idea of dosage sensitivity, with both low and high USP7 expression causing imbalances of neuronal homeostasis. The duplication present in patient IGGAC01 partially overlapped GRIN2A. Truncating as well as missense, activating variants of GRIN2A have been associated with an idiopathic form of focal epilepsy, EEG continuous spike-and-wave during sleep (CSWS), and speech disorder (OMIM 245570) (Lemke et al., 2013) whose clinical aspects are present in our patient. Different molecular alterations of NMDA receptor subunits seem to result in a deleterious dysregulation of NMDA receptor function (Balu and Coyle, 2011). Although the effect of partial duplications on gene function are not easily predictable, some clinical features of the IGGAC01 patient suggested the occurrence of a dysregulation of GRIN2A that could have contributed to the patient’s phenotype. Of note, a patient with phenotype resembling that of patient IGGAC01 was reported in Decipher database with a similar duplication, including both USP7 and GRIN2A (Figure 8B). We thus hypothesised that the phenotype of these two patients could result from the impairment of both USP7 and GRIN2A.

IGGAC24 had a de novo duplication of chromosome 19 that encompasses various genes, 11 of which are expressed in the brain. The duplication also overlaps a TDB described in the brain cortex and other different tissues, which could contribute to dysregulating gene expression of implicated genes (Figure 9A and Table 1). Attempting to prioritize enclosed genes and define possible candidate genes, we performed a gene enrichment analysis as described above (see Methods). Two genes emerged for their functions known to play a role in neurodevelopmental disorders, namely SHISA7 and U2AF2. Indeed, in Decipher at least one NDD case with a duplication that partially overlaps the TDB and the SHISA7 gene (Patient 338729) was reported while no similar duplications, implicating brain-expressed genes, were reported among controls. We focused on SHISA7, a transmembrane protein that interacts with GABAA receptors (Han et al., 2019) and, by acting as auxiliary protein for the trafficking of AMPA receptors, regulates synaptic plasticity (Schmitz et al., 2017). Notably, spine loss has been observed in SHISA7 knockout mice and in wild type mice with SHISA7 overexpression (Schmitz et al., 2017; von Engelhardt, 2019). Thus, we hypothesised that the duplication could cause a dysregulation of SHISA7 expression, possibly also by altering chromatin conformation, and ultimately causing a dysregulation of synaptic plasticity.

Patient IGGAC23 showed a de novo duplication of chromosome 10 overlapping 7 brain-expressed genes and a TDB described in the brain cortex and other tissues (Figure 9B and Table 1). After gene enrichment analysis of these 7 genes with known NDD genes, one emerged as the most interesting, CTBP2. This gene produces two alternative transcripts for two different proteins, one of them is a component of specialized synapses (ribbon synapses) and the other one is a negative regulator of transcription. CTBP2 knock-out mice present a neurological phenotype characterized by abnormalities of brain development. A recent study reported that the overexpression of CTBP2 inhibits neuronal differentiation (Zhang et al., 2020). Of note, while no duplications similar to that of this patient were reported among controls, in Decipher a patient with autistic behaviour and a duplication shorter than the one of patient IGGC23, but still enclosing CTBP2, was reported. Thus, the duplication present in patient IGGAC23 could have caused overexpression and/or dysregulated expression of overlapped genes, among which CTBP2, whose overexpression has already been associated with impaired neuronal defects.