Chromosomal Microarray in Patients with Non-Syndromic Autism Spectrum Disorders in the Clinical Routine of a Tertiary Hospital

Autism spectrum disorders (ASD) comprise a group of neurodevelopmental disorders (NDD) characterized by deficits in communication and social interaction, as well as repetitive and restrictive behaviors, etc. The genetic implications of ASD have been widely documented, and numerous genes have been associated with it. The use of chromosomal microarray analysis (CMA) has proven to be a rapid and effective method for detecting both small and large deletions and duplications associated with ASD. In this article, we present the implementation of CMA as a first-tier test in our clinical laboratory for patients with primary ASD over a prospective period of four years. The cohort was composed of 212 individuals over 3 years of age, who met DSM-5 diagnostic criteria for ASD. The use of a customized array-CGH (comparative genomic hybridization) design (KaryoArray®) found 99 individuals (45.20%) with copy number variants (CNVs); 34 of them carried deletions (34.34%) and 65 duplications (65.65%). A total of 28 of 212 patients had pathogenic or likely pathogenic CNVs, representing approximately 13% of the cohort. In turn, 28 out of 212 (approximately 12%) had variants of uncertain clinical significance (VUS). Our findings involve clinically significant CNVs, known to cause ASD (syndromic and non-syndromic), and other CNVs previously related to other comorbidities such as epilepsy or intellectual disability (ID). Lastly, we observed new rearrangements that will enhance the information available and the collection of genes associated with this disorder. Our data also highlight that CMA could be very useful in diagnosing patients with essential/primary autism, and demonstrate the existence of substantial genetic and clinical heterogeneity in non-syndromic ASD individuals, underscoring the continued challenge for genetic laboratories in terms of its molecular diagnosis.


Introduction
Autism spectrum disorders (ASD) are neurodevelopmental alterations characterized by deficits in communication and social interaction, as well as repetitive and restricted behaviors, anxiety, hyperactivity, or experienced arousal [1,2]. It includes typical childhood autism and other clinical phenotypes (such as Asperger syndrome) and generalized developmental disorders not otherwise specified in other categories [2]. According to the International Classification of Diseases (ICD-11.1) and the fifth edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5), ASDs are categorized as generalized developmental disorders. ASD can be non-syndromic (primary, essential) or syndromic (related to more than 100 different genetic disorders) [3]. In terms of health, ASD is the most individual had a genetic examination and clinical evaluation of their cognitive abilities and ASD symptoms, to ensure that only subjects with non-syndromic ASD phenotypes were included. After this evaluation, 232 patients aged >3 years with apparently non-syndromic ASD and who met DSM-5 diagnostic criteria were initially recruited for the study. In the end, other cases were excluded if the DNA quality prevented the CMA assay's performance and there was no possibility of collecting a new sample. The final cohort was composed of 212 individuals. Data collection and sampling were conducted with prior written informed consent from the subject.

patients aged > 3 years with non-syndromic ASD
Informed consent

Karyotyping and FISH Analyses
Cytogenetic analyses were performed on GTG-banded metaphases at a resolution of approximately 550 bands. Analyses were carried out according to standard laboratory protocol using Chromosome Kit P (Euroclone; Siziano, PV, Italy). Fluorescence in situ hybridization (FISH) analyses were performed according to standard laboratory protocols using different probes (Kreatech Biotechnology B.V, Amsterdam, The Netherlands, and Vysis Inc.; Downers Grove, IL, USA).

Chromosomal Microarray Analysis (CMA)
CMA analysis was performed by comparative genomic hybridization (aCGH) in all 212 patients using our custom platform (KaryoArray ® v3.0, [35,36]). The microarray design used for the analysis was SurePrint G3 CGH 8 × 60 K (Agilent Technologies; Santa Clara, CA, USA). The design focused on more than 375 regions of high clinical significance in the human genome. It had a higher resolution than the ISCA 8 × 60 K model (comparable to a 240 K array), which was defined based on the ISCA Consortium and our own clinical experience [35,36]. Data were analyzed using CytoGenomics software (Agilent Technologies) with the default CGH analysis method and interpreted with reference to available databases (DGV, ClinVar, DECIPHER, STRING, etc.). Classification and interpretation of

Karyotyping and FISH Analyses
Cytogenetic analyses were performed on GTG-banded metaphases at a resolution of approximately 550 bands. Analyses were carried out according to standard laboratory protocol using Chromosome Kit P (Euroclone; Siziano, PV, Italy). Fluorescence in situ hybridization (FISH) analyses were performed according to standard laboratory protocols using different probes (Kreatech Biotechnology B.V, Amsterdam, The Netherlands, and Vysis Inc.; Downers Grove, IL, USA).

Chromosomal Microarray Analysis (CMA)
CMA analysis was performed by comparative genomic hybridization (aCGH) in all 212 patients using our custom platform (KaryoArray ® v3.0, [35,36]). The microarray design used for the analysis was SurePrint G3 CGH 8 × 60 K (Agilent Technologies; Santa Clara, CA, USA). The design focused on more than 375 regions of high clinical significance in the human genome. It had a higher resolution than the ISCA 8 × 60 K model (comparable to a 240 K array), which was defined based on the ISCA Consortium and our own clinical experience [35,36]. Data were analyzed using CytoGenomics software (Agilent Technologies) with the default CGH analysis method and interpreted with reference to available databases (DGV, ClinVar, DECIPHER, STRING, etc.). Classification and interpretation of each of the variants was categorized as benign/likely benign (B/LB), pathogenic/likely pathogenic (P/LP), or of uncertain clinical significance (VUS), according to the American College of Medical Genetics (ACMG/AMP, see [37]).

MLPA (Multiplex Ligation-Dependent Probe Amplification)
Several MLPA Salsa kits were used to confirm some of the CNVs found. To rule out subtelomeric rearrangements, we applied MLPA kits P036 and P070, along with Salsa kits (including Salsas P245, P373, P064), for several microdeletion/microduplication syndromes (MRC-Holland; Amsterdam, The Netherlands). Data analyses were performed according to the manufacturer's protocols. The relative probe signals were defined by dividing each measured peak area by the sum of all peak areas of the control probes of that sample. The ratio of the relative probe area of each peak was then compared to that obtained on a DNA control sample using Coffalysser v.9.4; (MRC-Holland; Amsterdam, The Netherlands).

Fragile-X Syndrome Analysis
Repeated CGG in the FXS region of the FMR1 gene involved was quantified by TP-PCR, using a commercial kit following the manufacturer's instructions (LabGscan FRAXA; Diagnostica LongWood; Zaragoza, Spain). The products were migrated into the 3130xl ABIPrism genetic autoanalyzer (Thermo-Fischer, Waltham, MA, USA).

Statistical Analysis
Statistical analysis was performed with SPSS version 25 (IBM Corporation; Armonk, NY, USA). The two groups were compared by Student's t-test for continuous variables and chi-square tests for categorical variables. Comparisons between several groups were carried out by ANOVA analysis and Bonferroni post hoc tests for continuous variables and z-tests of column proportions for categorical variables. A p-value (observed significance level) of less than 0.05 was considered to indicate a statistically significant difference.

Limitation of the Study
One of the difficulties in carrying out this study was finding individuals with primary autism who did not show malformations or any kind of dysmorphia when examined by clinical experts. The limited sample size and lack of functional/segregation analysis in uncertain CNVs may also impact this study.

Results
The cohort was composed of 212 individuals with a mean age of 10.73 ± 6.42 years and a median age of 10. The male/female ratio was 4.04:1 (186/46 cases, respectively), with no statistically significant differences between mean ages by gender, which were 10.58 ± 6.13 years for males and 11.20 ± 6.86 years for females (Student's-t-test: p = 0.571). The patients in the cohort were mostly of pediatric age (<16 years, 189 individuals, 89.15%), and all were from Spain. Although the primary reason for visiting our clinic was autism (173/212, 81%), we also identified 24 individuals with Asperger's syndrome (11.32%), 17 (8%) with a diagnosis of attention deficit and hyperactivity disorder (ADHD), and 15 with non-specific developmental disorders (7.7%). In addition to ASD, we also detected the presence of other comorbidities, with intellectual disability and psychomotor delay being observed most frequently (69 patients, 32%). In addition, 6.6% (14 individuals) were diagnosed with epilepsy but in none of them could any syndromic entity can be recognized.
The familial segregation of CNVs was only studied in 11 of 28 P/LP patients. We found five de novo pathogenic variants, four of maternal origin and two of paternal origin. Tables 1 and 2 summarize the findings and classification of the P/LP and VUS variants using ACMG/AMP criteria [37].    The familial segregation of CNVs was only studied in 11 of 28 P/LP patients. We found five de novo pathogenic variants, four of maternal origin and two of paternal origin. Tables 1 and 2 summarize the findings and classification of the P/LP and VUS variants using ACMG/AMP criteria [37].   (Table 3). Table 3 also summarizes statistically significant size differences between B/LB and LP/P (p < 0.05, Student's t-test) at the three ranges of CNV sizes analyzed (<0.1; 0.1-1 Mb and >1 Mb), regardless of the type of rearrangement (deletions or duplications). VUS CNVs had an intermediate size between B/LB and P/LP variants.

Discussion
Autism spectrum disorders (ASDs) constitute a group of severe neurodevelopmental conditions characterized by an increase in the "informal" inclusion of different aspects (anxiety, experienced arousal [1]; early motor signs [38]) with complex multifactorial etiology. It is widely accepted that evaluations of patients with autism/ASD may involve a CMA as a first-tier test under different recommendations [12,13]. Nevertheless, the situation in Spain during the enrollment period for this study was different. We realized that CMA was not frequently used in our country, at least not in most public hospitals. In 2017, only 16.28% of the network of public hospitals in Spain fully addressed these recommendations (Spanish Human Genetics Association, AEGH; unpublished data). This fact could compromise, in a significant part, the diagnostic yield in ASD individuals, among others. Genetic studies for individuals with ASD may also be compromised by the fact that ASD is not diagnosed through biomarkers but through a process of observation and a systematic cognitive/behavioral examination. In addition, the irruption of other genomic technologies, such as NGS, competing as the first line of testing in patients with ASD [19,20,22,23] can significantly delay the entire diagnostic process, for economic reasons, at least in non-tertiary hospitals in our country. Ironically, none of these patients have a simple genetic study, as they are waiting for a trio of Exome analyses. Therefore, since, for us, the clinical and genetic diagnosis of individuals with ASD is critical, we encourage our colleagues to fight against this circumstance, namely the insufficient use of new genomic technologies such as CMA and NGS (as a first test or combined), in the routine clinical diagnosis of ASD. Data within this work support that CMA could also be very useful to segregate cases of individuals with apparently essential/primary autism.
Presently, only a few CNVs are reported in databases as being associated with essential ASD clinical features (OMIM, https://www.omim.org/, accessed on 27 February 2023), and most of them are characterized by incomplete penetrance and highly variable expressivity. Few studies have been conducted on the interpretation of CNVs in Spanish ASD patients, most of which have been extracted from a heterogeneous group of patients with global neurodevelopmental delay [18][19][20][21][22][23][24]. In this study, we used a previously validated and trained CMA platform [35,36] for the screening of apparently (under clinical geneticists' supervision) non-syndromic ASD individuals, which allowed us to segregate up to 87% of individuals in the cohort or to classify CNVs in approximately 75% of the patients showing CNVs (see Figure 3). These rates are similar to or even higher than those previously published for global ASD in our country or nearby [19,23,24,30,32,34]. Regarding the quick segregation mentioned before, we were able to establish a statistically significant relationship between the size of the CNVs and their clinical classification (under ACM/AMP criteria) in primary ASD patients. In fact, according to previous work, P/LP CNVs are also larger (including more genes) than B/LB variants (Monteiro, Pinto) in essential cases of ASD. Interestingly, VUS had an intermediate size between B/LB and P/LP CNVs. Like other previous studies, CNV gains were also more frequent than deletions (Monteiro, Pinto). Our results also supported the theory that a well-known platform by the user, the selection of the patients, and the use of their databases may help improve CMA results in general, also including individuals with primary ASD.
We found 28 cases carrying P/LP CNVs associated with apparently primary ASD (approximately 13% of the subjects) that were similar to previous reports in global ASD individuals [20,30,34] in the literature and slightly higher than other studies reported in our country [19,22,23], but is difficult to compare because many of the studies do not segregate between isolated and non-isolated ASD subjects. We also observed VUS variants in 26 of 212 cases (12.26%). This number is higher than other studies addressing VUS in ASD individuals in Spain [20,22] or our previous report for this platform in patients with ID, congenital malformations, or ASD, together (estimated in 3%, during its validation [35] or up to 7% within its clinical implementation process; unpublished data). This platform covers clinically relevant regions of genomic imbalances maintaining an acceptable balance between the number of VUS and known/unknown genomic aberrations [35]. Thus, this fact may give us stronger support for these regions, CNVs, or genes to be putatively associated with primary ASD, as the platform used combines genome-wide coverage with targeted clinically validated P/LP enrichment regions.
Overall, CNVs (P/LP or VUS) found may involve regions that play a role in several neurodevelopmental disorders and probably contribute substantially to either ASD or ASD risk/susceptibility. Among them, we highlight some candidate regions that have not been strongly associated with ASD such as (i) duplications at: Xp22.33, 2p25.3, 3p26.3,  3p26.3-p26.2, 6q16.3, 9p24.3, 9q34.13, 15q11.2, 22q11.21 (ii) deletions at: Xp22.33-p22.2,  2q23.1, 3p25.3-p25.2, 5p15.33, 8q21, 9p24.3, 10q21.3, 15q26.1 y 18q21. Interestingly, although the primary focus for selecting these cases was non-syndromic ASD, an important number of affected regions and genes found are related to syndromes, ID, and other comorbidities. They can be grossly divided into four groups: The first group includes genes that are already known to be related to ASD. An example is individual AUT45, who presented a gain in the ZNF44 gene, a gene previously associated with ASD (https://gene.sfari.org/ accessed on 3 November 2022 and [39,40]). This patient also had a duplication in the CNTN4 gene. Deletions and duplications of the CNTN4 gene, which plays an important role in neuronal maintenance and plasticity, have also been described in autism patients [41]. Individual AUT164 presented a deletion including the SLC6A1 gene, which has an established role in ASD [42]. We also found one patient (AUT28) with a duplication in the Xp22.33 region, associated with autism. This characteristic has been already reported in the DECIPHER database in nine patients, (#285674, 284752, 290397, 325813, 351449, 314771, 328499, 289827, and 333223). Another example is the duplication involving the CHL1 gene (3p26.3), also reported associated with ASD [43][44][45]. One of the genes frequently observed in copy number variations in this cohort was CHRNA7. The role of deletions and duplications of CHRNA7 remains controversial, but they have also been associated with ASD in several research articles, and includes singlenucleotide variants (SNVs) within this gene. CHRNA7 deletions are considered pathogenic (AUT230), although it is difficult to establish whether the relationship is causal or involves susceptibility to ASD. On the other hand, duplications in this gene are classified as VUS (patients AUT224, AUT225, AUT227, AUT229, AUT231), because CHRNA7 duplications were observed both in individuals with ASD and in controls (apparently without ASD) [46].
A second group of CNVs involved a significant number of syndromic deletions/gains (12/28 cases; 40.85%) that are already known to be related to ASD [47][48][49][50][51][52][53][54][55][56][57][58][59][60]. They include 16p11.2-p12.2 duplication (reciprocal to OMIM#613604, two patients,), 15q11.2 duplication (including autism susceptibility in region 4, part of OMIM#608636, two patients), 22q13 deletion or Phelan-McDermid syndrome (OMIM#606232, two patients), 2q11.1-q13 duplication (one patient), 2q37 deletion (OMIM#600430, one patient,), 3q29 deletion (OMIM #611936, one patient), 1q21.1 duplication (OMIM #612475, one patient), including only BP3 and a small part of BP4. We also observed that duplications in 15q11.2-q13 (OMIM #608636) seemed to occur in approximately 3% of this cohort, and are known to be maternally inherited (two patients, AUT97 and AUT136). In addition, we found a patient (AUT146) with an intronic duplication affecting GABRG3. However, we consider it to be a VUS because it is a 30 Kb deletion near exon-2, and we cannot rule out the possibility of involving additional probes using a higher-resolution platform. The GABRG3 gene in the 15q12 region has been suggested as a possible candidate to explain ASD in several patients [54]. Regarding other syndromic regions, we highlighted a deletion in 18q21.2 encompassing the TCF4 gene (patient AUT36). Defects in this gene are a cause of Pitt-Hopkins syndrome (OMIM #610954), which seems to correspond with certain clinical features of the findings in this patient after re-evaluating. The deletion in 15q26 (AUT81), a region belonging to the 15q26-qter microdeletion syndrome, had not been previously reported in ASD patients, but it has been associated with ID [55]. However, this region includes CHD2, a gene related to ASD in recent works and some databases (Simons Simplex Collection and SFARI database, with evidence in at least ten articles). This gene is also involved in the development of the central nervous system and belongs to the immunoglobulin superfamily of cell adhesion molecules [45,56]. We also remark additionally a duplication in 9q34.13 (AUT95), which affects the POMT1 gene. This gene has been associated with Walker-Warburg syndrome (OMIM #236670), C1-girdle muscular dystrophy (OMIM #609308), and intellectually impaired muscular dystrophy (OMIM #613155). POMT1 has also been associated with ASD in the SFARI database, a dosage-sensitive gene that appears to contribute to the pathogenesis of primary ASD [57]. We also observed a duplication in 22q11.2 (AUT185) associated with the 22q11.2 microduplication syndrome (OMIM #608363). Patients with ASD have been reported to be associated with duplication in 22q11.2 [58,59]. This region includes the SEPT5, GP1BB, and TBX1 genes. Finally, another interesting example was a deletion in 5p15.33, located in a region that includes 5p-or Cri du Chat syndrome (OMIM# 123450) and is associated with psychiatric behavior (auto-aggression, hyperactivity deficit, sleep disturbances, etc.) [60]. However, this particular patient also had a duplication in the 2p25.3 region, suggesting a derived chromosome, as a result of a balanced translocation in the father. Detecting balanced translocations in parents has important implications for family genetic counseling.
The third group of CNVs includes some of the regions that not only correspond to ASD traits but which may also be related to different comorbidities including psychiatric disorders, aggressive behavior, mood swings, epilepsy, or ID. This is the case of abnormalities in the DOCK8 gene located in 9p24.3, which have also been associated with type-2 mental retardation [61]. We identified two patients affected with pathogenic variations in this region, one carrying a deletion (AUT46, who presented autism) and one with a duplication (AUT194). Several articles correlate deletions and duplications in this region with ASD [62]. In addition, DOCK8 is included in the list of genes associated with ASD in the AutDB database [63]. In the Autism Sequencing Consortium cohort, there are patients with de novo loss of function variants associated with the DOCK8 gene in patients with ASD. Patient AUT35 carried a duplication in the GRIK2 gene region at 6q16.3, which has also been described as associated with autism (DECIPHER#284729). Although this gene has been mainly related to ID and epilepsy [64], there is a literature report of another patient with a deletion in the gene who was also associated with autism [65]. In addition, an MBD5 deletion was found in patient AUT218. MBD5 is a member of the methyl-CpG-binding domain (MBD) family of proteins. Haploinsufficiency of this gene is associated with a syndrome involving microcephaly, ID, severe speech impairment, seizures, short attention span, autistic-like behaviors (gaze avoidance, inattention, and repetitive behaviors), and other stereotypic and repetitive behaviors (e.g., teeth grinding, hand chewing and repetitive hand movements) [66].
The fourth group of CNVs involves chromosomal regions with somewhat more variable phenotypes. For example, some of the clinical characteristics of patient AUT167 (with short stature, ASD, and ID) could be explained by the deletion found in Xp22.33-p22.2. This region includes the Xp22.3 microdeletion syndrome (OMIM # # 300830), which presents a highly variable phenotype depending on the size of the deletion. Nonetheless, it is primarily characterized by X-linked ichthyosis, mild to moderate ID, Kallmann syndrome, short stature, chondrodysplasia punctata, and ocular albinism. Epilepsy, attention deficit hyperactivity disorder, autism, and social communication disorder may also be associated. In patient AUT167, the deleted gene that seems to be associated with autism and ID is NLGN4X (whose protein participates in synaptogenesis and glutamatergic pathways), located in the region of susceptibility for autism and Asperger (OMIM #300497 and OMIM #300495) [67]. In other subgroups of patients, more than one CNV (two deletions) were found to be responsible for the phenotype (for example, patient AUT217). We observed a gene deletion of IMPA1 in one of them and CTNNA3 in the other. Disorders associated with IMPA1 include ID, mental retardation, autosomal recessive 59, and bipolar disorder (https://www.omim.org, accessed on 3 November 2022). Mutations in CTNNA3 are associated with familial arrhythmogenic right ventricular dysplasia-13, which is not ASD-related (https://www.omim.org, accessed on 3 November 2022).
Regarding uncertain CNVs, VUS are a critical aspect in the use of genomic technologies in clinical routine, as well as for appropriate genetic counseling. In the absence of more information, we believe that VUS should not be used to condition the clinical management of these patients, but it may potentially suggest new regions associated with ASD, from a research point of view. These variants will be reclassified as benign or pathogenic in the future, as more information becomes available. Thus, it is necessary to conduct a review of these variants in these particular patients with a consistent frequency. Among the VUS found in our cohort, we noted some examples of CNVs associated with ASD. One patient (AUT175) was found with a duplication in 17p11.2-p11.1 involving the genes MAP2K3, KCNJ12, KCNJ18, C17orf51, FAM27L, and FLJ36000. Recently, another patient with autism and ID showing a similar CNV was reported in the DECIPHER database (#24937). Patients AUT103 and AUT104 had a deletion in 8p11.21.1, affecting the POMK (SGK196) gene. Loss of function compound heterozygous variants in this gene were found in a patient affected by the Walker-Warburg syndrome (WWS) phenotype. We also found some VUS not previously associated with ASD, such as 1p32.3 duplication (AUT100), 1q44 duplication (AUT75), 3p25.3-p25.2 deletion (AUT169), 10q24.32 duplication (AUT129), 12q24.33 duplication (present in two brothers, both with ASD; AUT179 and AUT180), 14q13.2 duplication (AUT62), and 14q24.3 duplication (AUT148).
Finally, the present study has several limitations. One of the difficulties in carrying out this study was finding individuals with primary autism who did not show malformations or any kind of dysmorphia when examined by clinical experts. In addition, given the limited sample size and lack of parental segregation in many of these CNVs, we cannot define whether CNVs are de novo or inherited, at this point. This limitation is especially relevant for VUS variants. Furthermore, this study is focused on a customized aCGH of 60K. The use of a higher-resolution SNP array (850 K or higher) could give us other experimental points of view.

Conclusions
Overall, these data may also support the idea that the etiology of the essential ASD may involve the functional alteration of inter-related genes in neuronal networks (see graphical abstract and [4,33]), and demonstrate substantial genetic and clinical heterogeneity in ASD individuals, underscoring the continued challenges in their molecular diagnosis. These findings also demonstrate that CMA could be very useful for the segregation of cases of individuals with apparently essential/primary autism. We established that approximately over 85% of the individuals in our cohort have been rapidly screened using CMA analysis, a cost-effective first-tier test in most laboratories in the world. This has allowed us to establish clinically relevant CNVs (around 13%) with important implications for genetic counseling for these families. Among them, we mainly found a significant number of recurrent variations [68]. Lastly, we documented several CNVs that have not been previously reported in ASD that may also influence the development of susceptibility to ASD, which needs further experimental approaches (such as NGS) and/or larger cohorts to confirm their role.