Phenotypic and molecular spectra of patients with switch/sucrose nonfermenting complex-related intellectual disability disorders in Korea

The switch/sucrose nonfermenting (SWI/SNF) complex is an adenosine triphosphate-dependent chromatin-remodeling complex associated with the regulation of DNA accessibility. Germline mutations in the components of the SWI/SNF complex are related to human developmental disorders, including the Coffin–Siris syndrome (CSS), Nicolaides–Baraitser syndrome (NCBRS), and nonsyndromic intellectual disability. These disorders are collectively referred to as SWI/SNF complex-related intellectual disability disorders (SSRIDDs). Whole-exome sequencing was performed in 564 Korean patients with neurodevelopmental disorders. Twelve patients with SSRIDDs (2.1%) were identified and their medical records were retrospectively analyzed. ARID1B, found in eight patients, was the most frequently altered gene. Four patients harbored pathogenic variants in SMARCA4, SMARCB1, ARID2, and SMARCA2. Ten patients were diagnosed with CSS, and one patient without a typical phenotype was diagnosed with ARID1B-related nonsyndromic intellectual disability. Another patient harboring the SMARCA2 pathogenic variant was diagnosed with NCBRS. All pathogenic variants in ARID1B were truncating, whereas variants in SMARCA2, SMARCB1, and SMARCA4 were nontruncating (missense). Frequently observed phenotypes were thick eyebrows (10/12), hypertrichosis (8/12), coarse face (8/12), thick lips (8/12), and long eyelashes (8/12). Developmental delay was observed in all patients, and profound speech delay was also characteristic. Agenesis or hypoplasia of the corpus callosum was observed in half of the patients (6/12). SSRIDDs have a broad disease spectrum, including NCBRS, CSS, and ARID1B-related nonsyndromic intellectual disability. Thus, SSRIDDs should be considered as a small but important cause of human developmental disorders.

the SWI/SNF complex were first recognized as tumorsuppressor genes implicated in oncogenesis [2]. The association between this chromatin-remodeling complex and human developmental disorders was discovered and studied with remarkable progress using next-generation sequencing [3][4][5].
The Nicolaides-Baraitser syndrome (NCBRS, MIM #601358) overlaps with the CSS, with more severe ID associated with a dysmorphic coarse face, microcephaly, seizures, and prominence of the interphalangeal joints. This syndrome is caused by SMARCA2, which is also one component of the SWI/SNF complex [12].
As pathogenic variants in the SWI/SNF complex are continuously detected in more patients with ID, these conditions are considered as manifestations of one clinical continuum, with ARID1B-related ID and mild CSS at one end, more severe forms of CSS in the middle, and NCBRS at the other end of the spectrum [13]. Therefore, the concept of SWI/SNF complex-related intellectual disability disorders (SSRIDDs) was introduced to explain this clinical spectrum [13,14].
This study analyzed 12 unrelated Korean patients with SSRIDDs confirmed via genetic testing while evaluating the cause of neurodevelopmental delay in these patients. Clinical information and the result of molecular analysis were analyzed to better characterize the phenotypic spectrum of SSRIDDs among Asian populations.

Subjects and clinical assessment
Whole-exome sequencing (WES) was used to evaluate 564 patients with neurodevelopmental disorders, such as developmental delay (DD), ID, epilepsy, neuromuscular disease, and central nervous system (CNS) anomalies, at the Medical Genetic Center of the Asan Medical Children's Hospital, Seoul, Korea, from March 2018 to October 2020. If any candidate variants were found, parental genetic testing using Sanger sequencing was performed to verify the pathogenicity of the identified variants. Patients harboring pathogenic variants or microdeletions in the components of the SWI/SNF complex were analyzed in this study.
Clinical data were retrospectively collected to describe the detailed phenotypes of SSRIDDs. Standard deviation scores (SDSs) of the height and body weight were calculated based on the Korean National Growth Charts for children and adolescents [15]. Short stature was defined as the height SDS below − 2.0 SDS for age-and sex-matched normative data [15]. The degree of ID was assessed with an intelligence quotient (IQ) test in patients aged ≥ 5 years. IQ scores of 50-70 were considered to indicate mild ID, IQ scores of 35-50 were considered to indicate moderate ID, and IQ scores < 35 were considered to indicate severe ID. Developmental status indicated by the developmental quotient (DQ) was evaluated using the Korean infant and child development test (KICDT) [16], which was developed by the Development Education Enacting Subcommittee of the Korean Pediatrics Academy. KICDT was designed to assess development in five functional domains: gross motor, fine motor, social-personal, language, and cognitive-adaptive skills. DQ [DQ = (developmental age/chronological age) × 100] lower than 80 was regarded as abnormal development.
All subjects were born from nonconsanguineous Korean parents. Blood or buccal smear samples were obtained with the informed consent of the patients' parents. This study was approved by the Institutional Review Board for Human Research of the Asan Medical Center (2021-0347).

Molecular analysis
WES was performed using genomic DNA isolated from either whole blood or buccal epithelial cells. Exons of human genes (approximately 22,000) were captured using a SureSelect kit (version C2; Agilent Technologies, Inc., Santa Clara, CA, USA). The captured genomic regions were sequenced using a NovaSeq platform (Illumina, San Diego, CA, USA). Raw genome-sequencing data analyses involved alignment to the reference sequence [National Center for Biotechnology Information genome assembly GRCh37; accessed in February 2009]. Mean read depth was 100-fold, with 99.2% coverage higher than tenfold. Variant calling, annotation, and prioritization were performed as previously described [17].
Chromosomal microarray (CMA) was performed using the CytoScan 750 K assay platform (Thermo Fisher Scientific, Waltham, MA, USA). The genomic DNA (250 ng) extracted from the peripheral blood was digested using   NspI and amplified using ligation-mediated polymerase chain reaction (PCR). The PCR product was purified, quantified, fragmented using DNase I, labeled with biotin, and hybridized overnight (16-18 h) in a CytoScan 750 K array. After hybridization, the sample was washed and stained with streptavidin using GeneChip Fulidics Station 450. Moreover, the array was scanned using GeneChip Scanner 3000 to generate a CEL file. The CEL file was analyzed using Chromosome Analysis Suite (Thermo Fisher Scientific) and converted to a CYCHP file to visualize the status of the genomic copy number and absence of heterozygosity.

Clinical features of patients with SSRIDDs
The clinical features of the 12 patients (7 females and 5 males) are described in Tables 1 and 2.
Ten patients were clinically diagnosed with CSS, and one patient with subtle dysmorphic features and mild ID was diagnosed with ARID1B-related nonsyndromic ID. Moreover, one patient harboring a SMARCA2 mutation was diagnosed with NCBRS.
DD/ID was a cardinal feature (Table 3). Hypotonia during infancy associated with gross motor delay was noted in all patients (12/12 patients, 100%). The mean age at walking without assistance was 20.4 ± 3.7 months. All patients had a delay in language development, including four patients with no meaningful speech at all (4/12 patients, 33.3%). The degree of ID was assessed in patients aged > 5 years. Two patients had mild ID (2/12 patients, 16.7%), whereas three had moderate ID (3/12 patients, 25%). Seizure and hyperactivity were documented in five (5/12 patients, 41.7%) and four patients (4/12 patients, 33.3%), respectively.

Molecular analysis of patients with SSRIDDs
WES identified 10 pathogenic variants in 10 patients, which neither parent carried. All of these ten patients had a confirmed de novo mutation origin. No pathogenic variants were observed using WES in the remaining two patients (subjects 5 and 11), whereas further analysis using CMA revealed microdeletions at regions encompassing the genes of the SWI/SNF complex (Table 4).
The remaining three patients harbored mutations in the other components of the SWI/SNF complex (i.e., SMARCA4, SMARCB1, and SMARCA2).
A novel variant in SMARCA4 (p.Arg1043Leu) was identified in subject 9, which was absent from the    general population database (gnomAD). This variant was predicted to be "disease causing" in MutationTaster, "damaging" in SIFT, and "deleterious" in PROVEAN. A missense change at this amino acid residue, SMARCA4 p.Arg1043Trp, was previously reported as a likely pathogenic variant in ClinVar. Therefore, SMARCA4 c.3128G > T (p.Arg1043Leu) was interpreted as a likely pathogenic variant based on the evidence of PS2, PM2, PM5, and PP3. SMARCB1 p.Lys363Glu observed in subject 10 was previously reported (https:// www. ncbi. nlm. nih. gov/ clinv ar/ varia tion/ 212263/). Consequently, it was considered as a pathogenic variant following the addition of PS2 after confirming the de novo mutation origin (PS2, PM2, PP2, PP3, and PP5). SMARCA2 p.Ala1160Gly observed in subject 12, who was diagnosed with NCBRS, was located in a mutational hotspot (C-terminal helicase domain) and absent from the general population database. In silico analysis predicted this variant to be "probably damaging" in PolyPhen-2, "disease causing" in MutationTaster, and "damaging" in SIFT. Thus, SMARCA2 p.Ala1160Gly was classified as a pathogenic variant (PS2, PM1, PM2, PP2, and PP3).

Discussion
This study provided clinical and molecular information on 12 Korean patients with SSRIDDs. These 12 patients were recruited from the neurodevelopmental disorder cohort who underwent WES or CMA for elucidating the genetic cause of their condition. ARID1B, identified in eight patients, was the most frequently altered gene in this study. The remaining four patients harbored pathogenic variants or microdeletions in SMARCA4, SMARCB1, SMARCA2, and ARID2. The clinical diagnoses were CSS for 10 patients, ARID1B-related nonsyndromic ID for one patient, and NCBRS for one patient.
Among the patients in the neurodevelopmental disorder cohort, 2.13% had SSRIDDs (12/564, 2.13%). Unexplained ID due to SWI/SNF complex mutations was estimated to be up to 3%, and the data (2.13%) of this study supported this idea [20]. Hoyer et al. [3] reported that ARID1B mutations were identified in 0.9% of unexplained ID cases.
A definite genotype-phenotype correlation could not be established owing to the small number of patients. However, several phenotypic differences were found among various genotypes.
ARID1B mutations are considered to be the leading cause of CSS (68-83%) [7,8,21]. In this study, the pathogenic variants in ARID1B were identified in 66.7% of patients (8/12 patients). Clinical phenotypes associated with ARID1B alterations have been reported to be highly variable and not severe compared to phenotypes of other genotypes [22]. As the use of broad genetic tests such as WES is becoming widespread, individuals who may not fit the diagnosis of classic CSS but rather present with more inconclusive phenotypes are now being discovered. These patients with ARID1Bassociated ID are expanding the phenotypic spectrum of the ARID1B-related disorder. The major differences between ARID1B-ID and ARID1B-CSS are the presence of typical dysmorphic features, including thick eyebrows, long eyelashes, hypoplastic/absent nail or distal phalanx of the fifth finger, and hypertrichosis [23]. For example, subject 3 was incidentally found to have a pathogenic variant in ARID1B during the evaluation of his mild DD. At the first examination, no dysmorphic features were noted in subject 3. However, the patient was reevaluated after identifying a pathogenic variant in ARID1B, and thick eyebrows and long eyelashes were noted. However, his phenotype was not sufficient to make a clinical diagnosis of CSS.
The patients with ARID1B-associated CSS in this study were likely to have a coarse face, hypertrichosis, thick eyebrows, large mouths, thick lips, long eyelashes, and micrognathia. Nail hypoplasia and/or a short distal phalanx of the fifth finger, which are known as cardinal CSS features, were identified in three patients (subjects 4, 6, and 7).
Previous studies [7,8,21] reported that a hypoplastic nail or a short distal phalanx of the fifth finger are present in 50%-68% of patients. According to a web-based survey (www. arid1 bgene. com), which is an open collection of clinical information on patients with ARID1B mutations, the incidences of a hypoplastic fifth fingernail and short distal phalanx of the fifth finger were estimated to be 24.6% (42/171 patients) and 22.0% (37/168 patients), respectively. Previously reported high incidences (50-68%) of these abnormalities may reflect an ascertainment bias because ARID1B mutations were preferentially sought after among those with clinically diagnosed CSS [7,8,21]. In the present study, three (subjects 4, 6, and 7) out of seven patients with ARID1B-associated CSS (3/7 patients, 42.9%) exhibited nail and/or distal phalanx abnormalities, which corroborated the previously reported data (48%) [22].
The position of the pathogenic variants in ARID1B may not influence the severity of the clinical phenotypes. Santen et al. [24] found no relationship between the variant position on cDNA and clinical severity. For example, patients who had pathogenic variants in exon 20, at the 3′ terminal region of the gene, had severe ID [24]. Among the present cases, subject 8, who had a variant in exon 20, had short stature, moderate ID, and classical features of CSS.
Almost all patients with genetic alterations in SMARCA4 were reported to have hirsutism, thick eyebrows, long eyelashes, and a less coarse face [25]. Subject 9, with a pathogenic variant in SMARCA4, also exhibited these typical features.
The pathogenic variants in SMARCB1 lead to a severe form of CSS with various CNS anomalies and severe growth retardation [7,8]. Subject 10 harbored a SMARCB1 variant in exon 8, which is a highly-conserved region and well-established causative domain for CSS [7,8]. Considered small for gestational age at birth, the patient underwent gastrostomy due to severe feeding difficulties. Severe growth retardation and microcephaly were also observed. Brain magnetic resonance imaging at 6 months revealed partial agenesis of the corpus callosum.
Subject 11 had mild ID with a profound short stature. As previously described [19], the patient exhibited both RASopathy-related features (e.g., profound short stature, epicanthal folds, down slanting palpebral fissures, and webbed neck) and CSS-like phenotypes (e.g., thick eyebrows, thick upper lips, and a large mouth). CMA revealed a 3.7-Mb deletion at chromosome 12q12-13.11 causing complete deletion of ARID2. As one of the components in the SWI/SNF complex, ARID2 haploinsufficiency has been shown to be associated with CSSlike phenotypes [10]. A previous study demonstrated increased extracellular signal-regulated kinase (ERK) activation in ARID2 haploinsufficiency, suggesting an association between the SWI/SNF complex and RAS-MAPK pathway [19].
Subject 12, with the SMARCA2 variant, displayed typical features of NCBRS (e.g., coarse face with hypertrichosis, thick eyebrows, thick lips, long eyelashes, nail hypoplasia, and microcephaly), but did not have prominent interphalangeal joints. Cognitive dysfunction was more severe in this patient than in those with other types of SSRIDDs. Differential diagnosis is sometimes confusing because CSS and NCBRS are overlapping syndromes that share similar phenotypes. Moreover, the clinical diagnosis may change according to the results of molecular analysis [8,13]. Molecular confirmation is thus required to make an accurate diagnosis between these two overlapping syndromes.
Similar to previous studies [13,21], variants in ARID1B in this study were truncating (nonsense or splicing-site mutations), whereas those in SMARCA4, SMARCB1, and SMARCA2 were nontruncating (missense mutation). The ARID1B haploinsufficiency is a pathogenic mechanism that leads to CSS or ARID1B-related ID. Subject 5 with an exon 10-18 deletion in ARID1B also showed a CSS phenotype. AIRD2 haploinsufficiency seems to have caused a CSS-like phenotype as well as ID in subject 12.
All variants in SMARCA4, SMARCB1, and SMARCA2 were missense mutations, implying that they may exert a gain-of-function or dominant-negative mechanism of pathogenicity [13,21].
The SWI/SNF complex components were initially recognized as tumor-suppressor genes associated with oncogenesis. Inactivating mutations in several SWI/ SNF components have recently been identified in a wide variety of tumors, including rhabdoid and lung cancer tumors [26]. Furthermore, truncating and missense germline mutations in SMARCB1 and truncating germline mutations in SMARCA4 have been shown to lead to a cancer predisposition syndrome [27,28]. Several cases with tumor formation were found among patients with SSRIDDs. Papillary thyroid cancer was reported in a patient with an interstitial 6q25 deletion, including ARID1B [29]. Moreover, a patient carrying an ARID1A pathogenic variant with hepatoblastoma was described previously in the literature [6]. van der Sluijs et al. [23] reported a boy with an ARID1B variant diagnosed with a Sertoli-Leydig cell tumor and a temporal glioneuronal tumor at 3 and 12 years, respectively. Longer observational periods are needed to conclude whether there is an association between SSRIDDs and cancer predisposition.
The limitation of this study should be noted. As a retrospective study, some clinical information was not available for some patients. The phenotypes among the patients were variable because of their varying ages. Thus, a longer observational period and larger patient population are needed to determine the complete clinical features and disease courses of these patients.

Conclusions
SSRIDDs can be found in a small but considerable proportion of the neurodevelopmental disorder patient cohort. Some common clinical features (e.g., hypertrichosis, coarse face, thick eyebrows, long eyelashes, and thick lips) and agenesis or hypoplasia of the corpus callosum can be clues suggesting SSRIDDs. Moreover, SSRIDD seems to be a disorder spectrum with ARID1Brelated ID on one end, classic CSS in the middle, and NCBRS on the other end [22]. The phenotypic spectrum of SSRIDDs will be more clearly documented as more individuals with SSRIDDs are identified with large-scale genomic analysis of unselected patient cohorts and followed up for a longer term.