Autism spectrum disorder (ASD) is a neurodevelopmental disorder distinguished by irregular social interaction and communication associated with repetitive behaviors and interests (Marini et al., 2020). ASD can be associated with different comorbidities, such as epilepsy, intellectual disability (ID), and attention deficit hyperactivity disorder (ADHD) (Freitag, 2007; Hyman et al., 2020). The worldwide prevalence of ASD is estimated to be approximately 1%, affecting males about four times more frequently than females (Levy et al., 2009; Hyman et al., 2020). Although there are ASD studies published from the Middle East (Amr, 2012; Alallawi et al., 2020; Yousef et al., 2021), the number is significantly disproportionate to the magnitude of the problem. Recently, a cross-sectional study of Qatari children aged 6–11 years surveyed between 2015 and 2018 revealed an ASD prevalence of 1.14% (95% CI: 0.89–1.46) (Alshaban et al., 2019). The Diagnostic and Statistical Manual of Mental Disorders (DSM-V) identifies ASD to include all social communication/interaction impairments and a minimum of two criteria in the restricted and repetitive behaviors (American Psychiatric Association, 2013).

ASD is one of the most heritable neuropsychiatric disorders, where the estimated recurrence rate is 5%–36% (Ozonoff et al., 2011; Werling and Geschwind, 2015). The high contribution of genetic factors to the etiology of ASD is suggested by a concordance rate of 60%–92% among monozygotic twins compared to 0%–10% among dizygotic twins (Bailey et al., 1995). Nevertheless, the recurrence rates among siblings born after two children affected by ASD per family revealed higher recurrence in males (47.5%) than in females (21.1%) and in siblings of females (44.3%) versus siblings of male probands (30.4%) (Werling and Geschwind, 2015).

ASD is clinically and genetically heterogeneous with several known monogenic disorders presenting with ASD symptoms, such as Fragile X syndrome (FMR1), Tuberous Sclerosis (TSC1, TSC2), and Rett syndrome (MECP2), contributing 1%–5% to the ASD etiology (Betancur and Coleman, 2013). Moreover, high throughput genetic testing identified genetic defects in ∼25% of the ASD cases (Yoo, 2015). These include genes involved in the synaptic formation, remodeling, and maintenance (NRX1, CNTN4, DCLK2, CNT- NAP2, TRIM32, ASTN2, CTNTN5, SYN1), neurotransmission (SYNGAP1, GABRG1, CHRNA7), and DNA methylation and chromatin remodeling (MBD5) (Lo-Castro and Curatolo, 2014). Pathogenic variants were primarily recurrent in TSC1, TSC2, NF1, UBE3A, and MECP2 (Betancur, 2011). The vast majority of known ASD genes have a high incidence of de novo pathogenic variants. The accumulating number of distinct rare genetic causes of ASD (Durand et al., 2007; Morrow et al., 2008; Glessner et al., 2009; Wang et al., 2009) suggests a complex and heterogenous genetic architecture, similar to that of intellectual impairment and epilepsy (Jiang et al., 2004).

The genome-wide microarray studies in ASD patients have reported rare microdeletions or microduplications, collectively called copy number variation (CNV), that contribute to the increased risk of ASD and its associated comorbidities (Marshall et al., 2008; Sanders et al., 2011). Several studies of de novo events confirmed multiple CNVs in a considerable fraction (3%–20%) of ASD cases (Christian et al., 2008; Marshall et al., 2008; Pinto et al., 2010; Sanders et al., 2012). It has now been demonstrated that CNVs explained 3%–20% of the ASD genetic etiology over the past 5 years (Roberts et al., 2014; Zarrei et al., 2019; Yap et al., 2021). Our group described 3 CNV regions that were located in genomic regions implicated in ASD. We reported a patient with distal trisomy 10q syndrome presenting with a few previously undescribed physical features, as well as ASD (Al-Sarraj et al., 2014). Also, we described a case report of a male patient affected by ASD with a mosaic trisomy of the pericentromeric region of chromosome 8 and maternal uniparental disomy of the same chromosome (Ahram et al., 2016). Besides, we reported an ASD female patient with intellectual disability (ID) and epilepsy presenting with recurrent microdeletion [del(15) q24.1-q24.2]) (Ahram et al., 2017).

The genome-wide association studies (GWAS) of ASD conducted over the past 15 years were primarily performed on European populations (Ma et al., 2009; Wang et al., 2009; Weiss et al., 2009; Anney et al., 2010; Anney et al., 2012; Connolly et al., 2013; Autism Spectrum Disorders Working Group of The Genomics Consortium, 2017; Grove et al., 2019), and a few were on the Chinese and Korean populations (Cho et al., 2011; Kuo et al., 2015; Xia et al., 2020). Our group reported on GWAS of ASD in the Middle Eastern population of Qatar using a family-based approach (Al-Sarraj et al., 2021). Results showed that common single nucleotide polymorphisms (SNP) are associated with ASD. Although the identified loci did not reach genome-wide significance, many of the top associated SNPs are located within or near genes that have been implicated in ASD or related neurodevelopmental disorders (Al-Sarraj et al., 2021).

There is a rapid rise in studies utilizing genomic sequencing to identify ASD genetic risk factors (Chapman et al., 2015; Yuen et al., 2015; Satterstrom et al., 2020; Kim et al., 2021). Most of these studies investigated and identified rare de novo variants in multiple genes, such as SCN2A, CHD8, ADNP, SHANK3, PTEN, DEAF1, and ANKRD11 (Yuen et al., 2015; Turner et al., 2017; Satterstrom et al., 2020). Lately, there has been a striking increase in the studies aimed at identifying ASD susceptibility genes in multiple populations using genomic sequencing (Besenbacher et al., 2015; Ye et al., 2015; Chang et al., 2019; Tran et al., 2020; Kim et al., 2021). Nevertheless, improved genomic techniques and analyses have expedited the detection of genes involved in ASD (Shen et al., 2014).

In this study, we inspected the contribution of CNVs in 113 patients with ASD, then we investigated the mutational spectrum of ASD patients using whole exome/genome sequencing in the families of 88 ASD cases. This study represents the first comprehensive genetic study of ASD in Qatar’s Middle Eastern population using simplex (trio) and complex families (families with multiple affected individuals).

We present a comprehensive study for ASD in the Middle Eastern population of Qatar using WGS/WES of simplex (trio) and multiplex families. We identified 40 candidate genes for ASD in 39 ASD subjects. Many of the identified genes had previously been implicated in ASD or related neurodevelopmental disorders. We identified three novel de novo variants in genes not previously linked to ASD (DTX4, ARMC6, and B3GNT3). All three de novo mutations were predicted to be damaging and rare, with either never reported in gnomAD and dbSNP or having allele frequencies less than 0.001%. We have defined these genes as highly conceivable suggested novel genes not reported in the SFARI database. Still, the literature showed that these genes are expressed in the human brain and indirectly illustrate a link to ASD or neurodevelopmental disorders (NDD). However, it necessitates future functional studies to ascertain their potential role in ASD or NDD pathogenesis.

Moreover, recent studies investigating exome or genome sequencing data aggregation of ASD cases have identified de novo mutations in DTX4 and B3GNT3 (Fu et al., 2022; Zhou et al., 2022). DTX4 (Deltex E3 ubiquitin ligase 4) is mainly expressed in the brain and is denoted to be involved in the Notch signaling pathway and protein ubiquitination. NLRP4 recruits DTX4, and through their interaction, it promotes the degradation of TBK1(Cui et al., 2012). Recent studies highlighted ubiquitin’s role in neurodevelopmental disorders and reported their monogenic forms to provoke lesions in gene coding, including ubiquitin proteins. These studies hypothesize a possible autoinflammation involvement in neurodevelopmental disorders pathogenesis (Ebstein et al., 2021). ARMC6 (armadillo repeat containing 6), the function of this gene’s protein product has not been determined. However, this gene is expressed in the brain. ARMC6 belongs to the Armadillo repeat-proteins (ARMCs) family which are distributed in eukaryotes and play prominent roles in cell-cell adhesion, intracellular signaling, and cytoskeletal regulation (Song et al., 2019). Also, in situ hybridization revealed a high expression of Armcx6 in mouse neuronal tissue in the developing and adult nervous system (López-Doménech et al., 2012). B3GNT3 (UDP-GlcNAc:betaGal beta-1,3-N-acetylglucosaminyltransferase3) gene is involved in the biosynthesis of poly-N-acetyllactosamine chains and the biosynthesis of the backbone structure of dimeric sialyl Lewis a (Zhang et al., 2015). This gene is mainly expressed in the intestine, salivary gland, and stomach. B3GNT3 is a member of GlcNAc transferases and part of glycobiology enzymes. Several studies have implicated glycobiology-related genes in autism spectrum disorders (Dwyer and Esko, 2016). Also, mutations in N acetylglucosaminyltransferase 1 have been associated with inherited forms of ASDs (Yuen et al., 2017).

Nevertheless, we found de novo variants in 15 genes that have been previously associated with or reported in patients with ASD or neurodevelopmental disorders. These include PHF21A (Kim et al., 2019), WASF1 (Ito et al., 2018), TCF20 (Vetrini et al., 2019), DEAF1 (Rajab et al., 2015), MED13 (Snijders Blok et al., 2018), CREBBP (Zheng et al., 2016), KDM6B (Stolerman et al., 2019), SMURF1 (De Rubeis et al., 2014), ADNP (Helsmoortel et al., 2014), CACNA1G (Chemin et al., 2018), MYT1L (Blanchet et al., 2017), KIF13B (Chen et al., 2020), GRIA2 (Salpietro et al., 2019), CHM (Tran et al., 2020), and KCNK9 (Delgado et al., 2014). Our results are consistent with many previous reports. For example, we identified a de novo variant (c.215G>A/p.R72H, CADD score: 29.6) in WASF1 in an ASD subject who had epilepsy comorbidity from non-consanguineous parents. This variant was confirmed by Sanger sequencing (Figure 2).

Notably, a recent study has reported three nonsense WASF1 de novo mutations in five unrelated cases with Intellectual Disability and Seizures (Ito et al., 2018). Also, WASF1 is considered a Rho family GTPases protein member, which is known to be involved in the pathogenesis of ASD (Guo et al., 2020). ADNP is another example of an essential and well-defined gene associated with ASD. Our study revealed a validated de novo variant in ADNP (c.2157C>A/p.Y719*, CADD score: 22.8), a gene in which multiple studies have reported variants in ASD patients (Helsmoortel et al., 2014; Vandeweyer et al., 2014; Rossi et al., 2017; Arnett et al., 2018). A study by Helsmoortel et al. (Helsmoortel et al., 2014) listed ten ASD patients with different clinical presentations, mainly ID and facial dysmorphism due to ADNP mutations. Besides, they estimated that nearly 1 in 10,000 ASD cases could have ADNP mutations. A recent in vivo study using Adnp knock-out mice showed a reduction in dendritic spine density and a change in gene expression due to Adnp deficiency (Hacohen-Kleiman et al., 2018). Our sequencing data revealed a deleterious de novo variant in TCF20. Consistent with our study, pathogenic variants in this gene have been reported in patients with ASD, intellectual impairment, behavioral abnormalities, and epilepsy (Babbs et al., 2014; Vetrini et al., 2019). Notably, about twenty-five inherited or de novo pathogenic variants in TCF20 were recently observed in affected patients with clinical presentations similar to the phenotype observed in Smith–Magenis syndrome (Babbs et al., 2014; Vetrini et al., 2019). The novel de novo variant in DEAF1 is predicted to be damaging and positioned in a highly conserved area of the genome and is not reported in the gnomAD database. A family-based study in a consanguineous Omani family using WES reported a homozygous variant in DEAF1 in three affected siblings with autism and ID (Rajab et al., 2015). Also, recent studies have reported many deleterious de novo and recessive variants in DEAF1 implicated with ASD genetic risk (Chen et al., 2017; Satterstrom et al., 2020).

Our study provides additional support for inherited genetic risk in ASD. We have identified three multiplex families with affected individuals who shared the same variants. One interesting case is a consanguineous family with identical triplets’ children (three monozygotic male siblings) who had ASD with variable clinical presentation, but all suffered from epilepsy. WES identified a homozygous missense variant in the VPS13B (c.8441G>A/p.R2814Q, CADD score: 32.0) in all three children (Table 3). Both parents were heterozygous for this variant, and all three ASD patients were homozygotes. The variant was infrequent in the gnomAD database (MAF = 0.00002, homozygous count = 0) and predicted to be pathogenic by functional prediction tools. Current evidence indicates a vital role for VPS13B in the normal growth and development of neurons. Pathogenic variants in VPS13B have been reported in patients with Cohen syndrome, a rare autosomal recessive disease characterized by intellectual disability, dysmorphism, and microcephaly (Yu et al., 2013). Sequencing results also revealed a homozygous missense variant in the PIGN (c.364G>C/p.E122Q, CADD score: 26.7) (Table 3) in the three ASD patients from this family. This mutation was rare in the gnomAD database (MAF = 0.0007, homozygous count = 1) and predicted to be pathogenic. The PIGN protein is involved in glycosylphosphatidylinositol (GPI)-anchor biosynthesis, and homozygous pathogenic variants in PIGN have been reported in patients with multiple congenital anomalies, including delayed psychomotor development, hypotonia, and seizures (Maydan et al., 2011; Tian et al., 2022). Our data reveal the complex genetic architecture of ASD and identify two homozygous variants in genes involved in neuronal development. Notably, expression of VPS13B and PIGN varies significantly during differentiation of wild-type iPSC to neurons suggesting a role for these two genes in neuronal development (Lizio et al., 2019). The variant in PIGN could explain the observed seizures in the patients, but further functional studies will be required to confirm this finding.

Also, our exome data presented another extended family with two affected boys (only one of them has seizures at 30 months) and one healthy sister. Interestingly, a homozygous missense variant in C12orf57 (Table 3) was detected in both affected brothers but not in their healthy sister. This variant has been previously reported (c.A1G/p.M1V, CADD score: 21.9) in several consanguineous Saudi and Kuwaiti patients who displayed diverse clinical characteristics such as a profound global development delay, autistic features, and epilepsy (Akizu et al., 2013; Salih et al., 2013; Zahrani et al., 2013). Besides, C12orf57 is associated with Temtamy syndrome comprised of hypotonia, moderate to severe ID with ASD features, corpus callosum hypoplasia, seizures, and microphthalmia (Akizu et al., 2013; Salih et al., 2013; Zahrani et al., 2013; Wang et al., 2019). Notably, our homozygosity mapping analysis showed that C12orf57 falls within a 22 Mb run of homozygosity on chromosome 2 in both affected brothers. WES for the affected ASD patient with epilepsy revealed a deleterious X-linked variant in WNK3 (c.236A>G/p.E79G, CADD score: 22.9) (Table 3) but this variant was not detected in his brother who had ASD without epilepsy. WNK3 was reported to be partially deleted in two brothers with autism, ID, and cleft lip/palate and a maternally inherited microdeletion (Qiao et al., 2008). WNK3 protein was co-localized with Cl₂ transporters (NKCC1 and KCC) and regulate the chloride (Cl₂) ion transportation that revealed a critical role in the excitability of gamma-Aminobutyric acid (GABA) receptors in neurons (Kahle et al., 2005).

The third multiplex family includes two affected boys from a consanguineous marriage who share a likely pathogenic variant in TSPYL2 (c.1668G>C/p.Q556H) on the X chromosome. TSPYL2 was described in many patients with Xp11.2 microduplications and neurodevelopmental disorders (Moey et al., 2016). Furthermore, one affected individual had an additional rare variant (c.9565delG/p.V3189fs*7) in dynein axonemal heavy chain 3 (DNAH3) (Table 3). Loss of function variants in DNAH3 has been described in large-scale studies of patients with ASD (the Simons Simplex Collection (Krumm et al., 2015) and in a Chinese study (Guo et al., 2017)).

Collectively, the evidence from this study suggests that 75% of multiplex families (6 out of 8) showed at least two siblings sharing the same variants. This is higher than previous studies (that reported about 30% of affected siblings carrying same variants relevant to ASD risk) which is likely due to the prevalent consanguinity. Besides, there were novel genetic findings that have an additional individual risk effect to ASD.

Furthermore, data from one ASD trio showed a likely pathogenic homozygous variant in CADPS2. The identified variant was in a highly conserved genomic area with no recorded allele frequencies in gnomAD or dbSNP databases. This gene is located in the 7q31 region that is has been reported to be deleted in patients with ASD (Okamoto et al., 2011; Grabowski et al., 2017). Studies using animal models showed that Cadps2-knockout mice exhibited a brain impairment and autistic-like behavioral phenotypes (Sadakata et al., 2007). Besides, CADPS2 variant screening in Italian patients with ID and with ASD subjects have shown parental inheritance for deleterious variants with a suggestion of parent-of-origin effect (Bonora et al., 2014).

It is well known that ASD complexities may be subjected to the multiple heterogeneity threshold or dosage (Schaaf et al., 2011; Cristina et al., 2013), where affected subjects may inherit different genetic risks (either novel or de novo mutations) in known genes, as well as chromosomal anomalies (CNV). Together these genetic risks may contribute as additive interacting factors to ASD-threshold (Cook and Scherer, 2008). Significantly, we recognized such multifactorial effects in several families (Table 4). For instance, one simplex family had variants in two candidate genes; a de novo p.G727R in SMURF1 which was validated by Sanger Sequencing (Supplementary Figure S1), and a recessive p.P635L in NAALAD2. SMURF1 encodes a ubiquitin ligase protein that is essential as a regulatory receptor for the SMAD proteins in the bone morphogenetic protein (BMP) pathway (Cui et al., 2011). Also, SMURF1 was found to play a role in controlling cell motility, signaling, and polarity (Zhu et al., 1999). Moreover, SMURF1 was significantly associated with ASD risk as a de novo splice site variant, as recently described (De Rubeis et al., 2014) by the Autism Sequencing Consortium patient in the Hartwell Autism Research and Technology Initiative (iHART) cohort (Ruzzo et al., 2019). Regarding the NAALAD2 gene, the NAALAD2 protein is considered one of the N-acetylated alpha-linked acidic dipeptidase (NAALADase) gene families. NAALADase was proposed to have an associated role with N-acetyl-L-aspartate-L-glutamate (NAAG) in the central nervous systems. But there is no clear link between NAALAD2 and ASD or any neurodevelopmental disorders (Pangalos et al., 1999).

Another example of oligo-heterogeneity was observed in a consanguineous family with a child affected with ASD and early onset epilepsy (9 months). The WES data identified a de novo variant in KCNK9 (c.479T>C/p.V160A, CADD score: 25.7) (Table 4). KCNK9 (potassium channel subfamily K member 9) is a channelopathy protein associated with Birk-Barel syndrome. This gene is imprinted with paternal silencing, and the mutation entirely abolishes the potassium channel activity when acting as a homodimer and when functioning as a heterodimer (Barel et al., 2008). Two recent studies have described patients with developmental delay and central hypotonia with de novo variants in KCNK9 by exome sequencing (Graham et al., 2016; Šedivá et al., 2020). Nevertheless, the child had also a likely pathogenic homozygous variant in TRAF3IP1 (c.1064G>A/p.G355E, CADD score: 29.6). The variant is rare, with reported allele frequencies of 0.01% in gnomAD and 0.02% in the 1,000 genomes. TRAF3IP1 is an autosomal recessive ciliopathy gene associated with Senior-Loken syndrome nine that is represented by early-onset nephronophthisis and pigmentary retinopathy (Bizet et al., 2015). Besides, a study reported five consanguineous families with four homozygous recessive and one compound heterozygous mutations using exome sequencing, where few families had affected subjects with global developmental delay and severe cognitive impairment.(Bizet et al., 2015).

Our study also identified a novel missense mutation in SUOX (in case 33; c.1156G>A/p.V386M, CADD score: 26.7 Supplementary Table S4). This gene encodes an enzyme, Sulfite oxidase, that catalyzes the oxidation of sulfite to sulfate (Irreverre et al., 1967). The deficiency of this enzyme has been associated with hereditary metabolic disorders and lethal neurological complications that occur at an early age (Rupar et al., 1996). Nevertheless, several previously published variants of SUOX were reported in patients with seizures, feeding difficulties, microcephaly, and brain atrophy.(Rupar et al., 1996; Johnson et al., 2002). Further work needs to be done, including the follow-up of the subject with this identified inborn metabolic errors for re-examining and the testing of sulfite oxidase deficiency or other metabolic profiling.

This study highlighted many genes that warrant further investigation. Future studies should aim at functional studies to characterize the identified genes deeply using animal models and human-induced pluripotent stem cells (hiPSC). The functional studies for the identified putative ASD genes using the known animal models in mice (Ergaz et al., 2016), Drosophila (Bellosta and Soldano, 2019), or zebrafish (Meshalkina et al., 2018); will help us understand the etiology, pathogenesis, and treatment of human ASD. Furthermore, taking advantage of patient-derived stem cell research, the researchers are now using the human-induced potent stem cell technology to generate neurons in vitro to model ASDs (Russo et al., 2019; Pensado-López et al., 2020).

We are aware that our project has limitations since our analysis focused on the coding regions for most cases. These constraints were due to the sequencing-platform variabilities. Another limitation of this study is that it was not possible to validate all identified de novo mutations using Sanger sequencing. However, all those picked variants with a minimum depth coverage were validated. Additionally, the combined sequencing depth for most of the remaining mutations was very high.

In summary, in our study, the known syndromic neurodevelopmental disorders indicated that ∼7% (7/102) were similar to the reported contribution of classical ASD etiology (1%–5%). Additionally, we identified ∼15% (15/102) of the ASD families due to possessing potential CNV regions. On the other hand, our study showed that 44.32% (39 out of 88 ASD cases) of studied cases had variants in genes that are relevant to ASD or related neurodevelopmental disorders. This finding is consistent with the power of clinical NGS as a diagnostic tool for neurodevelopmental (Álvarez-Mora et al., 2022). However, this rate is slightly higher than what has been reported previously for ASD (Stefanski et al., 2021) because it was estimated for genes related to ASD from the literature and the SAFARI database. However, some of these genes warrant functional studies to confirm their involvement in ASD. After crossing the reported genes in this study with the SAFARI database and the recent large SPARK study (Zhou et al., 2022), the rate was approximately ∼28% (25/88).

Furthermore, we crossed our findings with the most recent study from the region that shared a similar scope to our project (Abdi et al., 2023). Regarding the patient background, our cohort had a broader representation of the Middle East and North Africa, with ∼89% versus approximately 50% of Arab in Abdi et al. study. There was no overlap in top candidate variants between the two studies which is expected in highly heterogenous disorders such as ASD. Our study has revealed a significant homozygous burden within our ASD cohort, with an estimated rate of 63% (12/19) among consanguineous ASD cases and 12% (2/17) among non-consanguineous cases. Notably, Our findings are in alignment with the homozygous burden observed in consanguineous versus non-consanguineous families in the Abdi et al. Study.

Our data have revealed several genetic factors associated with either known or rare syndromes that have neurodevelopmental impacts only or have multiple body system effects combining with the nervous system (Birk-Barel syndrome (KCNK9), Temtamy syndrome (C12orf57), Rubinstein-Taybi syndrome 1 (CREBBP), Chediak-Higashi syndrome (LYST), Deafness (TECTA), Cohen syndrome (VPS13B), Escobar syndrome (CHRNG), and Arrhythmogenic right ventricular dysplasia 2 (RYR2)). We acknowledge that in our study, the information provided to our research team was limited to ASD-positive cases and the reported consanguinity within families. Although we recognize the importance of comprehensive clinical characterization, the constraints on available clinical data necessitated our reliance on standardized assessment tools and limited information provided by the families. However, the replication of these findings resulted from various applied molecular genetic testing, which showed high confidence in explaining the multifactorial condition of ASD.

Finally, despite the ASD multifactorial etiology that complicates the discovery of ASD risk variants, the number of identified novel or known putative ASD genes from our study was appreciable. Moreover, this study represents the first large-scale characterization of the ASD genetic risk in Qatar’s Middle Eastern population.