Coffin–Siris Syndrome and BAFopathies

Neurodevelopmental disorders caused by variants in subunits of the BAF (mSWI/SNF) complex, including the ARID1B-related spectrum beyond the classical syndrome and Nicolaides–Baraitser syndrome

Literature and regulatory information updated: 10 October 2026.

0. Initial Reference Table

The term “BAFopathies” encompasses neurodevelopmental disorders caused by germline variants in genes encoding subunits, or associated proteins, of the BAF chromatin-remodelling complex (BRG1/BRM-associated factor, the mammalian form of the SWI/SNF complex) [1, 2]. The table brings together Coffin–Siris syndrome (CSS) and its gene-defined subtypes, the spectrum of ARID1B-related intellectual disability (ID) without the classical phenotype, Nicolaides–Baraitser syndrome (NCBRS) and allelic diseases with tumour predisposition, which are mentioned because of their relevance to counselling. Inheritance is indicated as autosomal dominant (AD), autosomal recessive (AR) or X-linked, and the reference transcript is MANE Select (Matched Annotation from NCBI and EMBL-EBI). The designations CSS1 to CSS12 and the OMIM (Online Mendelian Inheritance in Man) numbers follow GeneReviews, Orphanet and the primary articles cited [3–7].

1. Introduction

1.1. Original description and clinical era

In 1970, Coffin and Siris described three patients with ID and absence of the nail and terminal phalanx of the 5th digit, a combination that gave the founding article its title and led, for a time, to the designation “fifth-digit syndrome” [19–21]. Over the following decades, the syndrome was defined solely by phenotype: Carey and Hall (1978), Levy and Baraitser (1991) and the series of 15 patients reported by Schrier and colleagues (2012) delineated the presentation of developmental delay, hypotonia, coarse facial features, hypertrichosis, sparse scalp hair and distal digital hypoplasia, while DeBassio and colleagues (1985) described the neuropathology [20, 22]. In 1993, Nicolaides and Baraitser described a 16-year-old girl with ID, sparse hair, a prominent lower lip and brachydactyly with cone-shaped epiphyses; Morin and colleagues proposed the eponym Nicolaides–Baraitser syndrome in 2003 [9, 23, 24].

1.2. Molecular discovery

In 2012, two groups simultaneously published the molecular basis of CSS in Nature Genetics. Tsurusaki's Japanese group identified de novo SMARCB1 variants by exome sequencing in 2 of 5 typical patients and, on screening 15 SWI/SNF complex genes in 23 patients, found germline variants in 20 (87%), distributed across SMARCB1, SMARCA4, SMARCA2, SMARCE1, ARID1A and ARID1B [25]. The Leiden group, led by Santen, found de novo truncating ARID1B variants in 3 patients and, in a copy-number variation analysis of 2,000 individuals with ID, three deletions encompassing ARID1B with a partially overlapping phenotype [26]. In the same year, Hoyer and colleagues, starting from a 2.5 Mb 6q25 microdeletion in a patient with severe ID, screened 887 patients with unexplained ID and found 8 de novo truncating ARID1B variants (0.9%), establishing haploinsufficiency of this gene as a frequent cause of non-syndromic ID [27]. Van Houdt and colleagues identified SMARCA2 by exome sequencing in 8 of 10 patients with NCBRS and, in an expanded series, missense variants in 36 of 44 patients, clustered in the ATPase region [28].

The following years added SOX11 (2014–2016), DPF2 (2018), ARID2 (2018–2019), SMARCC2 (2019), SOX4 (2019), SMARCD1 (2019) and BICRA (2020) [2, 29–35]. The review by Kosho and colleagues, including 109 patients in 2014, showed the predominance of ARID1B (65%), followed by SMARCB1 (12%), SMARCA4 (11%), ARID1A (7%), SMARCE1 (2%) and PHF6 (2%) [36]. In 2018, Aref-Eshghi and colleagues described DNA methylation episignatures shared by CSS and NCBRS, reinforcing the concept of a functional continuum between them [37].

1.3. Nomenclature and reclassifications

The term “BAFopathies” was proposed collectively for CSS, NCBRS and CSS-like conditions [1]; Barish and colleagues introduced the alternative designation SWI/SNF-related intellectual disability disorders (SSRIDD) [2]. Three reclassifications affect the interpretation of the literature. First, some patients with an initial clinical diagnosis of CSS and variants in SMARCA2 or PHF6 had, on reassessment, a phenotype closer to NCBRS or the female form of Börjeson–Forssman–Lehmann syndrome [3, 20]. Second, SOX11 produces an episignature distinct from the other BAFopathies and a presentation with hypogonadotropic hypogonadism, leading some authors to consider it a separate entity [38]. Third, ARID1B-related intellectual disability without the classical phenotype forms a continuous spectrum with CSS: in the cohort of 143 patients reported by van der Sluijs and colleagues, differences between patients with and without previous clinical suspicion of CSS were confined to dysmorphic features, and 91.5% of patients without previous suspicion were retrospectively considered by their clinician to be consistent with the spectrum [8, 39].

1.4. Definition of the entities covered

According to GeneReviews, CSS is defined by a combination of suggestive findings (coarse facial features, hypoplasia of the nail or distal phalanx of the 5th digit, hypertrichosis, sparse hair, hypotonia and developmental delay) and a heterozygous pathogenic or likely pathogenic variant in one of the 14 listed genes (ARID1A, ARID1B, ARID2, BICRA, DPF2, PHF6, SMARCA2, SMARCA4, SMARCB1, SMARCC2, SMARCD1, SMARCE1, SOX4, SOX11); there are no published consensus clinical criteria [3]. Vergano and Deardorff proposed diagnostic criteria and management recommendations in 2014, and the first parent-reported registry (Mannino and colleagues, 81 individuals, 2018) added surveillance recommendations; neither document constitutes a formal international consensus [21, 40]. ARID1B-RD is the clinical continuum ranging from classical CSS to ID with or without non-specific dysmorphic features, and includes the 6q25.2q25.3 microdeletion [8, 41]. NCBRS is the disorder caused by SMARCA2 missense variants in the C-terminal helicase region of the ATPase domain, with sparse hair, prominent interphalangeal joints, coarse facial features, acquired microcephaly and epilepsy [9]. Rhabdoid tumour predisposition syndromes, SMARCB1-related schwannomatosis, SMARCE1-related clear cell meningioma predisposition and SMARCA4-associated small cell carcinoma of the ovary, hypercalcaemic type (SCCOHT) are allelic diseases with a different mechanism and are only mentioned [7, 10].

2. Molecular Genetics and Pathophysiology

2.1. The BAF complex and its role in neural development

BAF complexes are ATP-dependent chromatin remodellers, homologous to the SWI/SNF complexes identified in yeast and Drosophila; DNA-stimulated ATPase activity destabilises histone–DNA interactions and alters chromatin accessibility to the transcription machinery [3, 42]. The catalytic subunit is SMARCA4 (BRG1) or, alternatively, SMARCA2 (BRM); SMARCB1 (INI1/BAF47) is a nuclear protein in the core of the complex; ARID1A (BAF250a) and ARID1B (BAF250b) are the largest, mutually exclusive alternative subunits, defining two forms of the canonical complex (cBAF); ARID2 (BAF200) defines and stabilises the PBAF complex; BICRA belongs to the non-canonical complex (ncBAF); DPF2 reads histone H3 tails through two PHD fingers [2, 7, 8, 31]. In the nervous system, subunit switching between the neural progenitor complex (npBAF) and the neuronal complex (nBAF) accompanies cell-cycle exit during differentiation; proper functioning of these complexes is necessary to establish and maintain neuronal fates [7, 20]. SOX11 and SOX4 are transcription factors acting downstream of BAF in neurogenesis, in a partially redundant manner [3, 7].

The analysis by Valencia and colleagues (2023) showed that mSWI/SNF genes harbour the largest number of de novo missense and truncating variants among nuclear protein complexes in neurodevelopmental disorders. Non-truncating variants mainly disrupt cBAF and cluster in four structural regions associated with greater severity: nucleosome interfaces, the site of insertion of the ARID-ARM module into the ATPase core, the Arp module and DNA-binding domains. More than 70% of residues affected in neurodevelopmental disorders are also mutated in cancer, but approximately 60% of amino acid substitutions are specific to neurodevelopment [43].

2.2. Two mechanisms per gene

The mechanism differs between genes and influences variant interpretation, choice of testing method and tumour risk [3, 15].

Table sources: [2, 3, 7–9, 16, 31, 44, 45].

The comparison by Borja and colleagues between germline CSS variants in ClinVar and somatic variants in cBioPortal illustrates this dual role: truncating variants predominate in ARID1A, ARID1B and ARID2 in both settings, whereas in SMARCA4 and SMARCB1, CSS variants are mainly missense and splice variants, and truncating variants are associated with hereditary cancer risk, particularly RTPS [15]. In SMARCB1-related CSS, variants cluster in the C-terminal domain and interfere with the protein's nucleosomal interactions without abolishing its tumour-suppressor function; rhabdoid tumours require complete biallelic loss during an early embryonic window in neural crest cells [46]. This contrast explains why patients with CSS-SMARCB1 do not usually develop SMARCB1-deficient tumours [10, 46].

2.3. ARID1B and the ARID1B-RD spectrum

ARID1B (6q25.3) encodes a 2,249-amino-acid protein in the historical transcript NM_020732.3/NP_065783.3, with 20 exons [8]. The MANE Select transcript is NM_001374828.1, and the numbering difference relative to NM_020732.3 is not constant along the gene (249 nucleotides in the exon 1 region and 210 nucleotides further downstream, according to VariantValidator conversions and Table 1 of van der Sluijs and Santen) [47, 48]. Conversions between transcripts must therefore always use a normalisation tool. ClinGen classifies ARID1B haploinsufficiency as level 3 (sufficient evidence) and has classified the gene–disease relationship with CSS as definitive since December 2019 [7].

In the cohort of 143 patients, all variants were truncating: 118 frameshift or nonsense variants, 18 multiexon or whole-gene deletions and 7 canonical splice-site variants; they were de novo in 107 of 107 cases with parental testing, and two sisters with the same variant indicated gonadal mosaicism [39]. The literature also describes 4 translocations and 3 duplications involving the gene [39]. Variant position does not correlate with severity, and variants in the last exon were not associated with a more severe presentation [39]. Missense variants are a much less frequent cause: 11 ARID1B missense VUS without strong clinical suspicion all had episignature scores below 0.05 [37], and a de novo missense variant in a mother with mild ID, transmitted to a son with agenesis of the corpus callosum, is one of the rare exceptions described [8, 49]. Bosch and colleagues showed that non-truncating variants in the EHD2 (SMARCA4 interaction) and ARID (DNA binding) domains form cytoplasmic aggregates (aggresomes in a vimentin cage) and, in the ARID domain, also nuclear aggregates, with transcriptome and methylome findings consistent with haploinsufficiency; the few patients with EHD2 missense variants have mild CSS [50].

Two regions require caution. Exon 3 of NM_020732.3 is a small in-frame exon of 14 amino acids, absent from NM_017519.2; no pathogenic variants confined to this exon are known, and one truncating variant was inherited from a parent without the phenotype [8, 39]. No pathogenic variants had been described in the first 849 bases (a GC-rich, low-complexity region) by 2019 [39]. That observation gave rise to a controversy that remains unresolved. Beaumont and colleagues (UK Biobank, 2024) proposed not routinely reporting predicted loss-of-function (pLoF) variants before Met584 (NM_001374828.1), because of a presumed alternative translation start [51]. Van der Sluijs and Santen responded in October 2025 with 19 truncating variants before Met584 in patients with a compatible phenotype, several inherited from mildly affected parents and four with a positive BAFopathy episignature; the earliest, NM_001374828.1:c.312_322del p.(Glu105GlnfsTer206) (NM_020732.3:c.63_73del), was inherited from the mother and reclassified as pathogenic after episignature testing [48, 52]. In their reply, Wright and Beaumont acknowledged that these variants can be pathogenic, but maintained that they should not be routinely reported without additional evidence: only approximately 10% of pathogenic pLoF variants in DECIPHER and ClinVar lie in the first quarter of the protein, compared with approximately 82% of pLoF variants in UK Biobank carriers, with no burden association with relevant phenotypes [53]. Both positions converge on treating these variants cautiously and using the episignature and parental phenotyping as additional evidence.

The 2019 GeneReviews identified no genotype–phenotype correlations distinguishing ARID1B-ID from ARID1B-CSS [8]. The 2025 GeneReviews and the adult cohort add that exon 1 variants may produce a milder phenotype: among the 87 adults, these variants were associated with more borderline or absent ID (19% versus 4%; P=0.07), less fine motor delay (P=0.01), fewer feeding difficulties (P=0.03) and absence of renal abnormalities (P=0.01), and all variants inherited from non-mosaic parents were located in exon 1 [3, 52]. Data from the gnomAD (Genome Aggregation Database) population database analysed in 2019 (744.4 expected versus 555 observed missense variants; z=3.39; 10 LoF variants, half at the start of exon 1 or in exon 3) suggest very high but possibly incomplete penetrance [39]. Apart from large microdeletions encompassing neighbouring genes, the absence of a clear correlation points to phenotypic modifiers [8].

Sources: [3, 6, 8, 15, 33, 37, 47, 54–58]. The MANE transcript for SMARCA4 is NM_003072.5; the historical convention in the RTPS literature uses the longer isoform NM_001128849 (36 exons), and the numbering systems are not interchangeable [10, 47].

2.4. ARID1A, ARID2, SMARCB1, SMARCA4 and SMARCE1

ARID1A variants in living patients are frequently mosaic, presumably because constitutional variants are embryonically lethal; in the Leiden fetal cohort, all ARID1A cases had the variant in approximately 50% of reads and a severe or lethal phenotype [12, 15]. GeneReviews recommends considering mosaicism when analysing sequencing data [3]. In 2026, Schuhmann and colleagues described the first pathogenic non-coding alteration in ARID1A: a de novo 24 kb intronic deletion, identified by short-read genome sequencing after a negative trio exome, which activates a pseudoexon containing a premature termination codon in a fraction of transcripts [59].

In ARID2, the series by Houdayer and colleagues (27 new individuals, 53 published in total) included 12 frameshift, 4 nonsense and 2 splice variants and 6 interstitial deletions, all consistent with loss of function, without a clear genotype–phenotype correlation; 21 of 27 were de novo, 2 inherited from a healthy mosaic parent and 1 from an affected mother [44].

In SMARCB1, three germline entities arise largely according to variant class. RTPS1 is associated with truncating alleles or deletions, particularly in exons 2 to 9, and approximately half of probands have a whole-gene or multiexon deletion [7, 60]. SMARCB1-related schwannomatosis is associated with non-truncating hypomorphic alleles at the 5′ or 3′ ends (particularly exons 1 or 8) and with exon 1 frameshift variants with downstream reinitiation [16, 60]. CSS3 is associated with in-frame deletions in exon 9 and missense variants in the last two exons [16]. Saad and colleagues (31 individuals, 2026) observed that C-terminal variants (αC helix) were associated with more severe speech delay, microcephaly and cleft palate, and greater gestalt resemblance than N-terminal variants (winged-helix/SNF5 domain); an XGBoost classifier predicted variant location from the phenotype with 96.7% accuracy [61]. A single individual with CSS, schwannomatosis and the p.(Arg374Gln) variant has been described in several studies [3, 62].

In SMARCA4, heterozygous missense variants cause CSS and truncating variants cause RTPS2 [63]. There are relevant exceptions: a patient with mild CSS, congenital microphthalmia and SCCOHT at 13 years of age had the de novo truncating variant p.(Arg979Ter), with a second somatic event in the tumour, suggesting that haploinsufficiency may cause milder CSS [57]; and three individuals with loss-of-function variants had microphthalmia or coloboma, developmental delay and brain abnormalities, without classical CSS or tumours [63]. In SMARCE1, loss-of-function variants cause clear cell meningioma predisposition and dominant-negative variants cause CSS5; the two presentations have not been described in the same individual [7, 64].

2.5. Other genes

In DPF2, the 8 variants in the founding series (1 frameshift, 2 splice, 5 missense) lie in the PHD1 and PHD2 motifs, near zinc-binding sites; they alter binding to histone H3 tails and form nuclear aggregates that recruit normal DPF2 and BRG1, supporting a dominant-negative mechanism [31]. In SMARCC2 (BAF170), Machol and colleagues described 15 individuals with 13 variants, most de novo, with a more severe phenotype in the 9 with variants in the SANT and SWIRM domains [33]; the largest cohort (65 individuals) showed that N-terminal missense variants reduce protein expression, like loss-of-function alleles, whereas non-truncating variants clustered in functional domains suggest another mechanism [45]. A de novo 233 kb inversion with a breakpoint between exons 16 and 17 of SMARCC2, detected by long-read genome sequencing in a 12-year-old boy from Qatar, was the first structural variant described in CSS8 [65]. SOX11 and SOX4 act through loss of DNA binding in the HMG domain; in SOX4, mutant proteins neither bind DNA nor transactivate [30, 34, 66]. BICRA (GLTSCR1) acts through haploinsufficiency, with models in zebrafish and Drosophila [2]. The Drosophila orthologue of SMARCD1 is Bap60, which is required for long-term memory [35].

2.6. SMARCA2 and Nicolaides–Baraitser syndrome

SMARCA2 (9p24.3; 34 exons) encodes the BRM ATPase. All pathogenic NCBRS variants are non-truncating and lie in the SNF2 ATPase domain (exons 15 to 25); the mutant protein incorporates into the complex and binds its targets but does not remodel, constituting a dominant-negative or gain-of-function effect [9, 28]. Whole-gene deletions do not cause NCBRS, and mice lacking functional Smarca2 have no major abnormalities [9, 67]. GeneReviews also attributes approximately 3% of variants to targeted deletion and duplication analysis (Tsurusaki 2012, Wolff 2012) [9]; this proportion is consistent with intragenic rearrangements rather than complete deletions, and the SMARCA2 duplications described had a phenotype closer to CSS [37, 68]. All patients with a variant in the C-terminal helicase region of the ATPase domain have severe ID and epilepsy, and more than half of cases have a variant in this region [9]. Sousa and colleagues (61 molecularly confirmed cases) suggested that domain VI variants are associated with severe ID and seizures, and that Pro883, Leu946 and Ala1201 are associated with mild phenotypes, based on small numbers [69]. SMARCA2 variants outside the helicase region cause blepharophimosis–ID syndrome (BIS), distinct from NCBRS in clinical presentation, yeast assays, transcriptome and methylation [70]. ClinGen does not recognise haploinsufficiency for SMARCA2 [7].

2.7. Methylation episignatures

BAF dysfunction leaves a methylation signature in leucocyte DNA, possibly through direct interaction between remodelling and the methylation machinery or through downstream effects during development [37]. In the series by Aref-Eshghi and colleagues, differentially methylated probe sets comprised 146 CpGs for CSS1, 135 for CSS3 and 356 for NCBRS, with nominal overlap of 10–20%; clustering separated CSS1 from CSS3+NCBRS, and CSS3 was closer to NCBRS than to CSS1 [37]. Some variants have their own sub-signatures that do not match the general signature: missense variants in the last exon of ARID1A/ARID1B in the c.6200 region (BAF250_C domain) and SMARCA4 p.(Met886Val) [6]. SOX11 and ARID2 have distinct episignatures; the ARID2 signature is globally hypermethylated [6, 38, 44].

2.8. Allelic diseases and modifiers

Allelic diseases are summarised in the table, together with the mechanism distinguishing them from CSS [3, 7, 9, 10, 16, 71].

Somatic SMARCA4, SMARCB1 and ARID1B variants in sporadic tumours (rhabdoid tumours, neuroblastomas) do not indicate hereditary predisposition when absent from the germline [3, 8, 72]. The distal 22q11.2 type 3 deletion includes SMARCB1, and young age at identification is associated with a poorer outcome [71]. SMARCB1-deficient renal medullary carcinoma, somatic SMARCA4 alterations in WNT-activated medulloblastoma and somatic SMARCE1 and ARID1/ARID2 alterations in meningiomas belong to somatic oncology [73–75]. In the heart, the murine orthologue of SMARCA4 (Brg1) participates in neural crest cell proliferation and epithelial–mesenchymal transition, but bicuspid aortic valve is not part of the spectrum of SMARCA4-related diseases [76].

3. Clinical Spectrum and Manifestations by System

3.1. Data sources and what each measures

Published frequencies vary according to ascertainment, and figures must be read alongside their denominators. Table 2 of the 2025 GeneReviews summarises the clinical literature, enriched for cases with the classical phenotype [3]. The international CSS/BAF registry, maintained in the USA since 2015 and completed by parents or caregivers with medical-record verification, analysed 208 individuals with a molecular diagnosis (Vasko and colleagues, 2021) and, later, 376 participants; parental reporting underestimates frequencies [1, 15]. The Leiden cohorts recruited patients with an ARID1B variant through clinician questionnaires, including patients without previous suspicion of CSS (143 patients aged 0 to 51 years, median 10 years; 87 adults aged 18 to 69 years) [39, 52]. Phenotype databases based on Human Phenotype Ontology (HPO) terms, such as DECIPHER, underestimate frequencies, and age-dependent features (speech, seizures) are better interpreted through survival analysis [39]. All these data describe patients receiving supportive care; there is no untreated natural history cohort in the strict sense, because there is no disease-modifying treatment.

Sources: [1, 3, 39]. Internal discrepancies in GeneReviews (epilepsy, nails, hearing) arise because the table and text cite different cohorts; none of these figures is a population prevalence.

The 2021 French PNDS (National Diagnosis and Care Protocol), drafted by the Reference Centre for Intellectual Disabilities of Rare Causes with the participation of the Leiden group, compiles the literature published up to 2020 and states three limitations: the spectrum is still being delineated, the frequencies may change and most published patients have an ARID1B variant, so the figures cannot automatically be transposed to the other genes [11]. Many of its figures reproduce the 2019 ARID1B cohort (recurrent infections 57%, hearing impairment 22%, cryptorchidism 55%, hypothyroidism 19%, GH deficiency 13%, diabetes 7%) and do not constitute an independent source [11, 39]. It does, however, add data not contained in the table above: ID in more than 95% of patients, whatever the gene; absence of speech in 20–30%; infantile hypotonia in 80%; hyperactivity in approximately 25%, aggression in approximately 10% and autistic traits in up to 50%; seizures in approximately 25% of patients with ARID1B, half before 4 years of age, described as of low activity (a single seizure or one per year) and responsive to treatment [11]; the source cohort gives a less uniform picture, with seizures in 27.5% of 142 patients, a Kaplan–Meier estimate of approximately 40% over a lifetime, onset between 0 and 14 years (median 4 years) and, among the 17 patients with a recorded frequency, 13 with one seizure per year or fewer and 4 with monthly or more frequent seizures [39]; agenesis of the corpus callosum in approximately one-third; eczema in one-quarter to one-third and vitiligo in some patients; hypertrichosis in more than 90%, mainly on the back, shoulders and face; hypoplasia of the distal phalanx and nail of the 5th digit in 40% of patients with ARID1B; short stature (below −2 SD) in one-third of patients with ARID1B, with normal head circumference except in some minor genes; nail hypoplasia in 76% of CSS overall (Santen series of 2014), in 68.5% of CSS-ARID1B and in 34% of so-called non-specific ID due to ARID1B, with hypoplasia of all fingernails or toenails in 20% of patients with ARID1B; sucking and swallowing difficulties in more than 70%, with a temporary nasogastric tube in 17%, sometimes beyond the first year; constipation in half of patients with ARID1B; refractive errors in 20–30%, with severe myopia in some patients; heart defects in 20–35%, mainly VSD and atrial septal defect, but also hypoplastic left ventricle, tetralogy of Fallot, PDA, mitral or pulmonary atresia and dextrocardia; renal anomalies in 12.5% of patients with ARID1B (mainly hydronephrosis, rarely horseshoe kidney) and rare hypospadias; scoliosis in 25%, joint hypermobility in 60%, delayed bone age in 40% and delayed eruption of the primary (45%) and permanent (48%) dentition [11]. For the face, the frequencies given are: coarse features 95%, thick eyebrows 90%, long eyelashes 85%, wide mouth 80%, thick lower lip 80%, broad nasal tip 75%, thick alae nasi 70%, long and broad philtrum 70%, and flat and broad nasal bridge, short nose, anteverted nares and thin upper lip in approximately 50% each [11]. The document itself gives two figures for low hairline (95% in the facial list and 75% in the dermatological section), and sparse hair is located, according to the PNDS, in the parietal regions (60%), whereas GeneReviews describes it as mainly temporal [3, 11].

3.2. Neurodevelopment, language and behaviour

ID is typically moderate to severe (intelligence quotient, IQ, between 40 and 69), but IQ values up to 97 have been reported, and the degree depends on the gene and comorbidities; there are no formal IQ studies [3]. On average, children sit at 12 months, walk at 30 months and say their first words at 24 months [3]. Expressive language is more affected than receptive language. In the registry, 26% had normal speech development (words at 18 months), fewer than 1% mild delay (19–21 months), 13% moderate delay (22–25 months) and 60% severe delay (over 25 months); approximately 12% do not speak, with a median age of 10 years at assessment [3, 77]. In the registry of 208, mean ages were 13 months for sitting, 27 months for the first word and 26 months for walking, with SMARCB1 the gene associated with the greatest delay (sitting at 20 months, first word at 41 months), and a significant difference after Bonferroni correction only for age at sitting [1]. In the ARID1B cohort, Kaplan–Meier analysis suggests that approximately 25% never develop speech [8, 39]. Fabre and colleagues (Necker, Paris; 12 patients with ARID1B, mean age 13.8 years) described marked deficits in language, motor skills and social cognition, with memory far exceeding other abilities and changes in perfusion and grey matter density in limbic and visual structures, supporting interventions based on visual cues and memory [78].

Autism spectrum disorders (ASD) affect up to 44%, probably an underestimate; hyperactivity and aggression each occur in approximately 10%, and attention deficit hyperactivity disorder (ADHD) in 25% [3]. In the ARID1B cohort, behavioural disorders affected 83.1%, autistic traits 57.1%, ADHD 33.3% and hyperactivity 42.9%; a high pain threshold was common [39]. The 2019 ARID1B GeneReviews found no increase in self-injury, aggression or sleep disorders, whereas the 2025 CSS GeneReviews describes disturbed sleep, particularly with concomitant autism, and aggression in approximately 10% [3, 8]. The difference may reflect the period and genetic composition of the cohorts.

3.3. Epilepsy

Seizure frequency depends mainly on ascertainment and cohort age, as shown in the table.

Sources: [1, 3, 9, 39, 52, 79]. In CSS, most seizures respond to conventional antiseizure medicines, and tics are part of the presentation [3, 20]. Tonic–clonic seizures predominate in ARID1B-RD, and some patients have an abnormal EEG without clinical seizures [8]. In 2026, Valentine and colleagues drew attention to focal cortical dysplasia as an under-recognised neuroimaging finding in ARID1B disorders [80].

3.4. Growth, feeding and endocrinology

Growth impairment is mild to moderate and begins prenatally; weight and height are below the 50th centile in most patients and below the 5th centile in 20–35%, and bone age is delayed by 2 to 3 years in approximately 40% [3, 20]. Registry-specific growth charts (99 individuals, excluding measurements after starting somatropin) show normal parameters at birth and, at 10 years, near-normal weight and head circumference, with height still reduced [81]. Among children with ARID1B, 30.3% had height below −2 standard deviations (SD), and head circumference was distributed around 0 SD, with only 2.9% below −2 SD [39]. In the registry of 208, macrocephaly was more frequent with ARID1B (24%) and microcephaly with SMARCA4 (38%), differences that were only nominal [1].

Infant feeding difficulties result from oral aversion or difficulty sucking without an evident intestinal malformation; 25–50% used a feeding tube or gastrostomy, and some no longer needed it [3, 21]. In ARID1B-RD, difficulties beginning neonatally are more severe and are associated with tube feeding (16.9% of the cohort), and constipation (30.5%) and gastro-oesophageal reflux are common [8, 39]. SMITH also lists occasional gastrointestinal abnormalities, such as gastric and duodenal ulceration, neonatal intussusception, malrotation and gastric outlet obstruction from redundant mucosa [20].

The 2019 ARID1B cohort documented hypothyroidism in 12 of 63 assessed patients (19.0%), almost all before 15 years of age, with a Kaplan–Meier estimate of approximately 25% (wide interval); diabetes in 5 of 71 (7.0%; 4 with type 2 and high weight, between 18 and 46 years, and 1 with type 1); and growth hormone (GH) deficiency in 7 of 51 (13.7%), 6 of whom were treated [39]. The authors consider GH deficiency potentially underdiagnosed, given that 30% have short stature [39]. In SOX11, hypogonadotropic hypogonadism is an established feature: in a review of 33 cases with hypogonadism, 13 met criteria for Kallmann syndrome, and in a 13-year-old girl, a gonadotropin-releasing hormone (GnRH) pump restored gonadotropins within 72 hours [38, 82]. There are isolated reports of transient hypercalcaemia and premature thelarche [20, 83].

3.5. Facial features, skin, hair and skeleton

Facial features include a wide mouth with thick, everted upper and lower vermilion, a broad nasal root and tip, anteverted nares with thick alae, a long philtrum, thick eyebrows, long eyelashes and periorbital fullness [3, 20]. Approximately 30% have evident dysmorphism at birth, and facial features coarsen over time, sometimes becoming apparent only later in childhood [3]. Hypertrichosis (95% in the literature) affects atypical areas, such as the back and shoulders; scalp hair is sparse in 60%, particularly in the temporal region, and improves with age; a low anterior hairline occurs in 75% [3, 20]. A machine-learning model applied to photographs distinguished patients with CSS from controls with 90.0% accuracy (95% confidence interval 73.5% to 97.9%), without differences in facial shape between genotypes [84].

Hypoplasia or aplasia of the distal phalanx and nail classically affects the 5th finger, but may involve other fingers and, in the feet, several toes; as the phenotype has expanded, many molecularly diagnosed patients have little or no digital involvement, and up to 25% of CSS cases have no distal abnormality [3, 37]. Joint laxity affects 66%, with radial head dislocation, coxa valga and small patellae reported; scoliosis (30–40%) is mainly neuromuscular [3, 20]. In adults with ARID1B, patellar dislocation affected 27–32% (12 of 45, recurrent in 8, median age 27.9 years, with patients undergoing three or more operations), pes planovalgus 67% and scoliosis 30%, with 7 of 22 undergoing surgery [52]. In ARID2, hip dysplasia is more frequent than with other genes [3, 85]. Primary and permanent dentition is delayed in approximately 40% [3]. Hernias occur in 10% [3].

3.6. Vision and hearing

Ptosis and strabismus affect approximately 50%, and myopia 20% [3]. In the ARID1B cohort, median myopia was −6.35 dioptres (range −24 to −1), and a non-functioning or absent lacrimal duct occurred in 15.1% [39]. Among 35 adults with CSS, 29 had ophthalmological problems [3, 86]. Primary congenital glaucoma was described in a 6-month-old infant with a de novo ARID1B variant, with epiphora, photophobia and megalocornea, controlled after surgery, and in an 11-year-old boy with NCBRS; in a 27-year-old man with CSS, glaucoma was diagnosed only on examination under anaesthesia, with no external signs [87–89]. Hearing loss (15–45%) is conductive or sensorineural; conductive loss is more frequent and is associated with recurrent respiratory infections and small auditory canals [3]. Congenital sensorineural loss predominates in ARID1B-RD (63.6% of hearing-loss cases were congenital and 51.9% bilateral) [8, 39].

3.7. Malformations, infections and neuroimaging

Congenital heart defects (approximately one-third) include ventricular septal defect (VSD), atrial septal defect (ASD), tetralogy of Fallot and patent ductus arteriosus (PDA) [3]. Genitourinary abnormalities include cryptorchidism, horseshoe kidney and hypospadias; in the registry, renal abnormalities were more frequent with SMARCB1, SMARCE1 and ARID1A and less frequent with ARID1B (P=1.3×10⁻⁶) [1, 3]. A 15-year-old girl with CSS3 and primary amenorrhoea had Müllerian aplasia, the first molecularly confirmed case [58]. Recurrent infections, mainly viral upper respiratory tract infections, are poorly characterised; immunological studies have shown no consistent defects, and hypotonia and difficulty clearing secretions contribute [3]. A patient with CSS2 and recurrent infections had quantitative deficits in innate and adaptive immunity and suppression of interferon-γ-dependent pathways, a single-case finding [90]. Laryngomalacia (19.8%) and tracheomalacia (14% in the registry) are common [1, 39].

Neuroimaging shows agenesis or hypoplasia of the corpus callosum, Dandy–Walker variant, simplified gyration, delayed myelination, mega cisterna magna and enlarged Virchow–Robin spaces [3, 8, 91]. In the ARID1B cohort, 42.6% of patients with imaging had complete or partial agenesis of the corpus callosum, 17.0% delayed myelination and 14.9% mega cisterna magna [39].

3.8. Phenotype by gene

Gene-specific correlations are weaker than previously assumed: ARID genes are associated with more delay and fewer malformations, SMARC genes with more organ malformations and SOX genes with predominantly neurological involvement, but any patient may have any aspect of the phenotype [1].

The PNDS adds gene-specific details, based on the series published up to 2020 [11]. In ARID2, Wormian bones are noted as a particular feature, and not all carriers meet the criteria for CSS. In DPF2, ID is less severe than in the other genes (among 9 patients, 1 without ID and 2 with borderline functioning), and 3 had craniosynostosis, which is rare in CSS; one had brachymetatarsia of the 4th toes. In BICRA, among the 12 patients then described with a deletion or a de novo variant (7 truncating variants and 2 missense), all with a neurodevelopmental disorder, 4 had microcephaly and 3 hypoplasia of the 5th fingernail, and the dysmorphism resembled neither that of CSS nor that of NCBRS. In SMARCE1 (7 patients, 6 with missense and 1 with a splice variant), nail hypoplasia is more frequent than in CSS overall, as are heart defects (PDA, valvular stenoses, dextrocardia), intrauterine growth restriction and microcephaly, and 2 patients with missense variants had a picture resembling Angelman syndrome or autism, without CSS. In SMARCB1 (13 patients, 9 with missense), the phenotype is more severe than in CSS overall, with more profound ID, more frequent epilepsy, microcephaly, growth deficiency and scoliosis, and more malformations (horseshoe kidney, dextrocardia, VSD); the recurrent variant NM_003073.5:c.110G>A p.(Arg37His) causes, in 4 described patients, a phenotype distinct from CSS, with profound ID and hydrocephalus due to choroid plexus hyperplasia. In ARID1A, with fewer than 20 published patients (mainly loss of function, some missense), the severity of ID varied between series, which Santen and colleagues attributed in part to mosaicism; agenesis of the corpus callosum and nail anomalies are usual, with additional heart defects (coarctation of the aorta, VSD, atrial septal defect) and urogenital anomalies (cryptorchidism, hypospadias, abnormal ureteric insertion). In ARID1B, more than 200 pathogenic variants, almost all truncating, and deletions of the gene have been described, and the variant type determines neither the phenotype nor the distinction between CSS and non-syndromic ID. In SMARCA4, with more than 25 patients (just over 10% of cases), almost all with missense variants and some with a heterozygous deletion of the gene, the picture is that of classical CSS, with moderate to severe ID, heart defects (VSD, complex malformations) and microcephaly in some patients. In SMARCC2, the 15 patients of Machol and colleagues could not be distinguished from patients with CSS due to other genes; in SMARCD1, only some of the described patients had a picture overlapping with CSS (the PNDS mentions 6 patients, but the original series has 5, with a proven de novo variant in 4); in SOX4, 4 patients with de novo missense variants had mild to severe ID and some CSS features (anomalies of the 5th digits, facial appearance); and in SOX11, 9 patients (6 missense, 3 truncating) had, in part, facial dysmorphism or limb anomalies suggestive of CSS, such as nail hypoplasia or clinodactyly of the 5th finger [11, 33, 35]. Microduplications including ARID1A cause a distinct picture, with ID and microcephaly, and duplication of SMARCA2 gives a CSS rather than an NCBRS phenotype. SOX4 and SOX11 do not encode subunits of the BAF complex, and the mechanism by which they cause a CSS-like picture remains unknown; the PNDS considered that the association of SOX4 with CSS required more patients [11].

3.9. ARID1B spectrum: classical CSS and ID without a CSS phenotype

The cohort of 143 patients compared 79 with clinical suspicion of CSS before testing and 64 without such suspicion. Significant differences after correction for multiple comparisons (P<0.001) were exclusively dysmorphic, and some extracranial findings were nominally more frequent in the group without a CSS phenotype [39].

Source: [39]. In the overall cohort, ID was distributed across normal–mild (3.1%), mild (28.3%), mild–moderate (15.7%), moderate (22.0%), moderate–severe (16.5%) and severe (13.4%); hypotonia affected 81.0%, gross motor delay 99.0%, delayed bone age 47.5%, scoliosis 26.0%, joint hypermobility 60.2% and cleft or submucous cleft palate 6.7% [39]. Two patients died, one at 9 years from cerebral oedema and cardiorespiratory arrest and the other from sudden death of unknown cause after 24 years of age [39]. The authors' conclusion, adopted by GeneReviews, is that ARID1B-CSS and ARID1B-ID constitute a single entity requiring the same management [8, 39]. Microdeletions of 6q25.3 encompassing ARID1B produce similar presentations, sometimes more complex because of contiguous gene involvement; in small deletions, the facial gestalt, hypertrichosis, aplasia or hypoplasia of the 5th nail, vertebral defects and cystic hygroma correlate with CSS [8, 95]. ORPHA:251056 (6q25.2q25.3 microdeletion, 4 cases or families) describes developmental delay, facial dysmorphism and hearing loss [41].

3.10. Prenatal phenotype

The fetal cohort from Leiden and an international consortium brought together 44 cases with an ultrasound abnormality and a pathogenic variant in a CSS gene: ARID1A 16 (36%), ARID1B 19 (43%), SMARCA4 3 (7%) and SMARCB1 6 (14%), compared with 5–9% for ARID1A in postnatal cohorts [12]. The first abnormality appeared at a median gestational age of 14 weeks and 5 days for ARID1A, 20 weeks for ARID1B and SMARCA4, and 24 weeks for SMARCB1 [12]. Among the ARID1A cases, there were 11 terminations of pregnancy for medical reasons out of 16, one fetal death and one neonatal death; cerebral (73%; hydrocephalus, agenesis of the corpus callosum, Dandy–Walker, vermian hypoplasia) and cardiovascular abnormalities (80%; hypoplastic left heart syndrome, persistent left superior vena cava, double-outlet right ventricle) predominated, with diaphragmatic hernia in 34%, anal abnormalities in 43%, a single umbilical artery in 47% and increased nuchal translucency in 47% [12]. In ARID1B, 76% had brain abnormalities, but only 29.4% were detected before birth, because agenesis of the corpus callosum often escapes ultrasound detection; intrauterine growth restriction was more frequent than in postnatal cohorts (P=0.04) [12]. The association between CSS and congenital diaphragmatic hernia was also described by Gofin and colleagues [96].

3.11. Adulthood

The first exclusively adult cohort (35 individuals aged 18 to 37 years, a predominantly European consortium) showed overweight or obesity in 19 of 35, visual impairment, scoliosis and behavioural changes more prevalent than in paediatric cohorts, cognition ranging from profound ID to low-normal IQ, with most having moderate ID, only one adult without ID and no malignancy [3, 86]. The cohort of 87 adults with ARID1B (median age 23.3 years) documented ID in 93% (borderline 7%, mild 28%, moderate 42%, severe 20%), loss of skills in 25% (16 of 64), with documented onset between 31 and 54 years, affecting walking in 7 and speech in 5 of the 16 patients, height below −2 SD in 54%, body mass index above 25 in 56%, cataract in 7% and type 2 diabetes associated with obesity; recurrent infections decreased to 39% [52]. Regarding autonomy reported by parents, 88% eat and drink independently, 73% dress themselves, 30% shop, 8% prepare dinner, 65% stay home alone for 30 minutes, 16% use public transport independently and 67% live with their parents [52]. The most commonly used medicines were laxatives, antiseizure medicines, antidepressants and antipsychotics [52]. Two adults died, one from renal abscesses at 24 years and the other from asphyxia due to choking on food at 47 years [52].

3.12. Tumour risk in CSS

The review by Borja and colleagues (literature from 1970 to November 2022) found 12 CSS cases with malignancy, 9 in children and 3 in adults [15].

Source: [15], with the original cases in [57, 62, 97, 98]. The pooled prevalence of hepatoblastoma in CSS-ARID1A was 3 in 84 (3.6%; 95% confidence interval 0.79% to 10.4%), excluding 14 cases without perinatal survival; in CSS-ARID1B, it was 0 in 487 (upper confidence limit 0.94%) [15]. In the registry of 376 participants, with a median age of 9 years and 42 adults, no neoplasm was reported, reflecting short follow-up and not demonstrating absence of risk [15]. GeneReviews considers a strong relationship between the deletion encompassing ARID1B and thyroid carcinoma unlikely, given the large number of carriers without cancer [3]. Among adults with ARID1B, there was one neoplasm in 73 (a probably benign lung tumour) [52]. For SMARCB1 in CSS, no rhabdoid tumour has been described; in SMARCA4, in a series of 15 patients with CSS, 13 with missense and 2 with nonsense variants had no tumours at 9 and 15 years [16]. The 2021 French PNDS concluded that the data did not allow an increased tumour risk in CSS to be asserted and that longer cohort follow-up was needed, although such a risk could not be excluded in patients with SMARCA4 or SMARCB1 missense variants [11]. Two of the cases in the table have an alternative explanation: the anaplastic astrocytoma of the patient with SMARCE1 had a MYBL1-MAML2 fusion sufficient to explain it, and the tumour of the patient with schwannomatosis and SMARCB1 had, in addition to the constitutional variant, a chromosome 22 deletion removing one copy of SMARCB1 and NF2 and a truncating variant in the remaining NF2 allele, so the causal role of the SMARCB1 variant is uncertain [11].

3.13. Nicolaides–Baraitser syndrome

NCBRS has been described in at least 80 individuals with an SMARCA2 variant, and Orphanet records 61 cases or families [9, 69, 99].

Source: [9]. Approximately 80% have severe speech delay and almost one-third never speak; children sit at a mean age of 9 months and walk at 21 months (range 10 months to 5 years) [9]. Psychomotor regression is not typical, but progressive worsening of seizures is associated with loss of speech; seizures initially respond to antiseizure medicines, particularly levetiracetam and valproic acid, but complete remission is very rare [9]. Behaviour includes perseveration, hyperacusis, oral sensitivity and attention deficit, with an ASD diagnosis in a minority [9]. Half have low birth weight and the same proportion short stature [9]. The skin shows apparently reduced pigmentation and prominent veins; in one series, widely spaced teeth reached 58.2% and hypodontia 18.6% [9, 69]. Radiology shows cone-shaped epiphyses, metaphyseal widening and shortening of the phalanges, metacarpals and metatarsals (mainly the 4th and 5th rays), abnormal bone age, platyspondyly, a small pelvis and a short femoral neck [9]. Hearing loss (5–10%) is mainly conductive; myopia affects 15%, astigmatism 6%, and glaucoma has been described in two patients [9]. Heart defects have been described in six individuals, as well as obstructive hypertrophic cardiomyopathy with a partial SMARCA2 deletion and hashitoxicosis in adolescence [9, 100, 101]. The oldest known individual lived to 33 years, and accelerated epigenetic age suggests early biological ageing [9, 102]. The severity of ID correlates with seizures, absence of speech, short stature and microcephaly [69]. The French PNDS summarises the same series (61 patients with a clinical diagnosis of NCBRS and a SMARCA2 variant) with some figures not contained in the table: severe ID in 45%, moderate in 36% and mild in 18%; head circumference below −2 SD in 65%; epilepsy in 64%, with mean onset at 2 years; growth deficiency in half, with delayed bone age in 41%; feeding difficulties in 46%, usually without the need for nutritional support; enlargement of the interphalangeal joints in 85% and broad distal phalanges in 68%; cryptorchidism in 59%; umbilical and inguinal hernias in 46%; and minor heart defects in 10% [11, 69]. It concludes that the neurodevelopmental phenotype is more marked in NCBRS than in CSS, particularly than in CSS-ARID1B [11].

4. Clinical Recognition and Warning Signs

4.1. When to suspect the diagnosis

GeneReviews recommends suspecting CSS in the presence of a combination of coarse facial features (sparse hair, especially temporally in infants, thick eyebrows, long eyelashes, a broad nasal root and tip, anteverted nares with thick alae, a wide mouth with thick, everted vermilion), hypoplasia or aplasia of the nail or distal phalanx of the 5th digit, hypertrichosis in atypical areas, central hypotonia and developmental delay of variable degree, usually moderate to severe [3]. For ARID1B-RD without a classical phenotype, GeneReviews proposes considering the diagnosis in any child with developmental delay or ID, from mild to profound, associated with generalised infantile hypotonia, feeding difficulties, spasticity, predominantly tonic–clonic epilepsy, ADHD or autistic traits, cryptorchidism, laryngomalacia, myopia, speech delay or suggestive dysmorphic features [8].

Classical signs are often absent. Digital hypoplasia is absent in up to 25% of CSS cases and is usually absent in patients with BICRA and SMARCD1 [2, 35, 37]. In the ARID1B cohort, 64 of 143 patients had no previous clinical suspicion of CSS, and only 22.8% of adults with ARID1B had been considered to have possible CSS before testing [39, 52]. In a Turkish series of 8 children (5 months to 6 years), hypertrichosis occurred in 5 and mild hypoplasia of the 5th distal phalanx in only 3 [83]. A BAFopathy should therefore feature in the differential diagnosis of any ID with disproportionately impaired expressive language, hypotonia, feeding difficulties, hypertrichosis, sparse hair, thick eyebrows or agenesis of the corpus callosum.

The French PNDS (National Diagnosis and Care Protocol) organises suspicion around three entry points [11]. The first is agenesis of the corpus callosum: CSS, and not NCBRS, is the most frequent cause of syndromic forms with ID, which agrees with the French series of Mignot and colleagues, and in the fetus the agenesis may appear isolated on ultrasound [11, 49]. The second is the hypotonic infant, sometimes with laryngomalacia and feeding difficulties requiring nutritional support. The third is the child with delayed psychomotor milestones or difficulties adapting at school compatible with ID; growth delay may also be the reason for referral. At any age, including adulthood, the typical morphological features are enough to raise the hypothesis, and some diagnoses arise without prior suspicion, from an exome requested for unexplained ID [11]. NCBRS is rarely suspected prenatally, because agenesis of the corpus callosum is not part of the picture and nutritional support is less often needed; it is diagnosed mainly in the presence of developmental delay or learning difficulties with morphological features and, sometimes, epilepsy [11].

4.2. By age and context

4.3. Warning signs and emergencies

The following signs require targeted assessment. Ophthalmologically, epiphora, photophobia and megalocornea suggest congenital glaucoma, described in CSS-ARID1B and NCBRS, and glaucoma may occur in adults without external signs [87–89]. From a respiratory perspective, choking, unexplained food refusal, poor weight gain and recurrent respiratory infections suggest dysphagia with aspiration; aspiration pneumonia and seizures are reported causes of childhood death, and one adult died from asphyxia during a meal [3, 52]. In adults with ARID1B, loss of motor or language skills (25% of the cohort) and recurrent patellar dislocation require neurological and orthopaedic assessment [52]. In a child with ARID1A, an abdominal mass or distension suggests hepatoblastoma [3, 15]. In an infant with a truncating variant or deletion of SMARCB1 or SMARCA4 (RTPS), an abdominal or soft-tissue mass, or new neurological signs, constitute an oncological emergency; in a girl or woman with a truncating SMARCA4 variant, pelvic pain or a mass and hypercalcaemia suggest SCCOHT [10, 16, 57]. In a person with SMARCB1, localised neuropathic pain or a mass along a nerve suggests schwannoma, and rapid growth with loss of function suggests a malignant peripheral nerve sheath tumour [3, 18].

4.4. Newborn screening

There is no newborn screening for BAFopathies. Prenatal diagnosis, when made, results from genomic sequencing targeted at fetuses with ultrasound abnormalities [12].

5. Diagnosis: Criteria, Tests and Algorithms

5.1. Criteria

There are no published consensus clinical criteria for CSS or for NCBRS [3, 9]. Vergano and Deardorff proposed diagnostic criteria in 2014, based on the clinical features known at the time [40]. The diagnosis is established by the combination of suggestive findings with a heterozygous pathogenic or likely pathogenic (P/LP) variant in a gene on the GeneReviews list, classified according to the standards of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology (ACMG/AMP), which consider the two classes equivalent for clinical decisions; a VUS neither establishes nor excludes the diagnosis [3, 103]. For ARID1B-RD, the definition is molecular: a heterozygous pathogenic or likely pathogenic variant in ARID1B [8]. The French PNDS (National Diagnosis and Care Protocol) considers molecular confirmation necessary even when the clinical diagnosis is certain, but a negative test does not rule out clinically convincing CSS; when an ARID1B variant is found in a person with ID without the morphological features of CSS, the diagnosis retained is non-syndromic ID due to ARID1B, which entails the same initial evaluation as CSS [11]. The PNDS gene list includes SMARCA2 duplication among the causes of CSS, and the pathogenicity of each variant should be discussed between the molecular biologist and the clinician before the results consultation [11].

5.2. Testing strategy

GeneReviews distinguishes two situations. When the phenotype suggests CSS, a multigene panel including CSS and differential-diagnosis genes is preferred, limiting VUS and unrelated findings; composition and sensitivity vary between laboratories, and the panel should include deletion and duplication analysis [3]. When the phenotype is atypical and CSS has not been considered, comprehensive genomic testing (exome or genome) is the option; deletion and duplication detection by exome or genome may be less sensitive than targeted analysis [3]. In ID without a CSS phenotype, the usual approach begins with chromosomal microarray (CMA), followed by an ID panel with deletion and duplication analysis or exome sequencing [8]. ACMG recommends exome or genome sequencing as a first- or second-line test in children with ID or multiple congenital abnormalities [9, 104]. The CSS/BAF registry recommends broad-spectrum testing to define the specific BAFopathy [1].

Source: [3]. The proportion column indicates the fraction of CSS attributed to each gene, and the method columns indicate the fraction of that gene's pathogenic variants detected by each technique; none represents the clinical sensitivity of the test. For ARID1B, sequencing detected the variant in 71 of 80 patients with CSS and deletion and duplication analysis in 9 of 80; in patients with ID, sequencing detected 54 of 63 and CMA 9 of 63 [8]. For SMARCA2 in NCBRS, sequencing detects approximately 97% of variants [9]. SMITH estimates that a variant is identified in 60–90% of patients with clinical CSS [20]; in the 2019 Japanese and international series, pathogenic variants were found in 78 of 214 patients with clinical suspicion (approximately 36%) [68].

5.3. DNA methylation episignature

The methylation episignature in leucocyte DNA is a clinically validated complementary test. Stored blood DNA may be used [3]. GeneReviews indicates its use when findings are suggestive and no variant has been found, or when there is a VUS in a gene on the list [3, 105]. EpiSign version 2 includes a common BAFopathy episignature (CSS1 to CSS4, CSS8 and NCBRS) and a separate one for BIS; in validation, concordance was 207 of 211 archived samples, and in the clinical cohort (n=207), 27.6% of tests were positive, 35.3% in the cohort targeted at ambiguous findings or VUS and 11.3% in the screening cohort [5]. In the Aref-Eshghi series, all 8 test patients had scores above 0.90 and all 356 healthy controls below 0.2, with no positives among 531 patients with 20 other syndromes [37]. At Erasmus MC (298 patients, 2019–2023), targeted analysis was positive in 18% of VUS, 91% of likely pathogenic variants and 89% of pathogenic variants; in unresolved cases, including after exome sequencing, the signature appeared in 9.0% and the confirmed yield was 4.0% (9 of 223) [106]. In Würzburg, 43% of patients with variants in chromatinopathy genes had a concordant episignature; only truncating ARID2 and SMARCB1 variants gave positive results [107].

The French PNDS accepts methylation analysis in the same two settings: deciding the pathogenicity of a VUS and identifying the BAFopathy signature in patients without a variant after sequencing of the known genes [11].

The episignature serves four purposes: reclassifying VUS in genes with a signature (ARID1B, ARID1A, SMARCB1, SMARCA4, SMARCA2, SMARCC2, ARID2, SOX11), screening unresolved ID, supporting early truncating ARID1B variants and detecting mosaic cases [5, 6, 37, 48]. Examples include a de novo SMARCA2 VUS (NM_003070.5:c.2647C>G p.(Pro883Ala), score 0.83) reclassified as likely pathogenic, a de novo SMARCA2 variant without an NCBRS phenotype and with a score below 0.01 reclassified as likely benign under functional criterion BS3, and a 6-year-old boy with two mosaic single-nucleotide deletions in ARID1B (33% and 16% of reads), identified from the episignature [37]. The limitations are several. Only peripheral blood is validated. Low-level mosaicism may give a negative result, as in the ARID2 mosaic parents [44]. Genes without their own signature (BICRA, SMARCD1, DPF2, SOX4, SMARCE1) cannot be excluded by a negative result [5, 107]. Some variants have distinct sub-signatures, and two Gabriele–de Vries syndrome (YY1) samples had a BAFopathy score above 0.5 [6]. The common signature does not distinguish CSS from NCBRS; the gene is the discriminating test [37].

5.4. Test limitations and special situations

A negative result does not exclude the diagnosis. Approximately 40% of patients with clinical CSS have no variant in a known gene [3]. The exome does not detect deep intronic variants or most structural variants: the 24 kb intronic deletion in ARID1A was found by genome sequencing and confirmed by transcriptome analysis, and the 233 kb inversion in SMARCC2 by long-read genome sequencing [59, 65]. In SMARCB1-related schwannomatosis, deep intronic variants have been described, and RNA-seq is useful when DNA testing is negative and suspicion is strong [60]. In ARID1A, mosaicism requires attention to low allele fractions and, if necessary, testing of another tissue [3, 15]. In ARID1B, truncating variants in exon 3 of NM_020732.3 and at the start of exon 1 (before Met584) require caution, supported by the episignature and by parental phenotyping [8, 48, 53]. ARID1B missense variants should be interpreted with parental testing and, if possible, episignature analysis [37, 39]. In SOX11, sequencing detects only approximately 40% of variants, and analysis of 2p25 deletions is essential [3]. The PNDS attributes some of the negative results to the difficulty of fully studying the introns, the 5′ and 3′ untranslated regions and the GC-rich regions of these genes, and points to genome sequencing as a way of overcoming this limitation [11].

The report should state the gene, the versioned reference transcript (preferably MANE Select, with correspondence to the historical transcript when used in the literature), HGVS nomenclature at DNA and protein levels, zygosity, allele fraction when suggestive of mosaicism, and ACMG/AMP class with the criteria applied [3, 47, 103]. For SMARCB1, SMARCA4 and SMARCE1, the variant class (missense or in-frame versus truncating or deletion) must be stated explicitly, because it changes tumour risk and surveillance [7, 16]. A purely somatic SMARCB1 or SMARCA4 finding in a tumour should not be reported as a predisposition syndrome [7, 10].

5.5. Algorithm

Sources: [3, 5, 8, 9, 12, 59].

5.6. Prenatal and preimplantation diagnosis

Once the familial variant has been identified, prenatal diagnosis and preimplantation genetic testing are technically possible [3, 8]. Without a known variant, prenatal genomic sequencing in fetuses with abnormalities may identify a BAFopathy, particularly ARID1A, but interpretation is limited by the still limited knowledge of the prenatal phenotype and the low ultrasound detection of agenesis of the corpus callosum in ARID1B cases [12]. The description of an SMARCA2 variant inherited through low-level mosaicism in paternal sperm shows that a negative parental blood result does not eliminate recurrence risk and justifies offering prenatal diagnosis in subsequent pregnancies [9, 108].

6. Differential Diagnosis

6.1. Comparative table

Sources: [3, 8, 9, 11, 20, 37, 38, 70, 71, 109–114].

6.2. CSS and NCBRS

CSS and NCBRS share sparse hair, coarse facial features with a wide mouth and thick lips, long eyelashes, ID with marked language impairment and epilepsy. The digits discriminate: NCBRS has prominent interphalangeal joints and distal phalanges due to paucity of subcutaneous tissue and lacks hypoplasia of the nail or phalanx of the 5th digit, whereas CSS has distal hypoplasia and hypertrichosis [9, 20]. A markedly everted lower lip and translucent skin with visible vessels are more characteristic of NCBRS [20]. The boundary is not clear-cut: up to 25% of patients with CSS have no distal abnormality, prominent interphalangeal joints occur in CSS3 and CSS4, SMARCA2 duplications produce a phenotype closer to CSS, and intermediate phenotypes have been described [9, 37, 115]. Patients with CSS3 are more similar in methylation to patients with NCBRS than to patients with CSS1 [37]. Since the episignature is shared, gene identification is the discriminating test.

6.3. Börjeson–Forssman–Lehmann in females

Females with de novo PHF6 variants causing complete loss of function may present with ID of variable severity, hypoplastic nails, sparse hair, brachydactyly, clinodactyly particularly of the 4th and 5th digits, dental abnormalities and linear hyperpigmentation, with strong overlap with CSS [3, 109]. The two patients with a CSS phenotype and a PHF6 variant acquired the facial features of Börjeson–Forssman–Lehmann syndrome with age, which is considered a distinct entity [3]. In males, the syndrome evolves from a hypotonic infant with large ears and small genitalia to truncal obesity, gynaecomastia, hypogonadism and a heavy facial appearance with deep-set eyes in late adolescence [7, 116].

6.4. Other differentials with nail hypoplasia or coarse facial features

In the presence of coarse facies with hypertrichosis, the OXF recommends examining the hands (contractures suggest a storage disease; nail hypoplasia suggests CSS; broad thumbs suggest Rubinstein–Taybi), carefully excluding chromosomal abnormalities and performing baseline screening with thyroid function, mucopolysaccharides, oligosaccharides, urinary amino acids and organic acids, plasma amino acids, creatine kinase, glucose, alkaline phosphatase, phosphate and a search for vacuolated lymphocytes, with serial photography [110, 111]. Hypertrichosis over the spine requires exclusion of a spinal anomaly [111]. In Cantú syndrome, thick ribs are a radiological clue [111]. BONE includes CSS among the syndromes with hypoplasia of the distal phalanges, alongside acrodysostosis, Albright hereditary osteodystrophy, Hajdu–Cheney and Zimmermann–Laband [113]. The cognitive level distinguishes BOD, in which the deficit is mild, from CSS, in which it is almost always moderate to severe [3]. ARID1B-ID without distinctive features requires the differential of all syndromic and non-syndromic ID, which justifies recourse to the exome [8]. The French PNDS (National Diagnosis and Care Protocol) organises the differential by mode of presentation: with fetal agenesis of the corpus callosum, CSS is one among many causes and the aetiological work-up is the responsibility of the genetics clinic; the infant with hypotonia, feeding difficulties and dysmorphism may suggest a RASopathy, particularly Noonan syndrome; nail hypoplasia, which may affect all nails, leads to consideration of DOORS, Mabry and Temple–Baraitser; and hypertrichosis with a neurodevelopmental disorder leads to consideration of Cornelia de Lange, Wiedemann–Steiner and FHEIG [11].

7. Treatments and Risk Management Strategies

7.1. Principles and guidance documents

There is no disease-modifying treatment and no medicine authorised by the US Food and Drug Administration (FDA) or the European Medicines Agency (EMA) for CSS, ARID1B-RD or NCBRS [3, 9]. Management is supportive, multidisciplinary and guided by each patient's problems. The published recommendations derive from expert consensus and registry data, not from trials: the parental registry of Mannino and colleagues (2018), GeneReviews CSS (2025), GeneReviews ARID1B (2019), the recommendations of the 2019 ARID1B cohort, updated by the 2024 adult cohort for children and adults, the study of 17 individuals with ARID2 (2024), GeneReviews NCBRS (2025), which states that there are no published clinical guidelines for this syndrome, and the French HAS National Diagnosis and Care Protocol (PNDS, 2021), the only national diagnosis and care document dedicated to CSS and NCBRS, prepared using the 2012 HAS method [3, 8, 9, 11, 21, 39, 52, 85]. Given the variability, GeneReviews recommends individual clinical judgement [3].

Sources: [3, 8, 9, 11, 15, 16, 18, 21, 39, 40, 52, 60, 117].

The PNDS diverges from the other sources on seven points. In ophthalmology, it proposes an annual screening consultation and, given the frequency of refractive errors and strabismus, an examination of refraction under cycloplegia with atropine by a paediatric ophthalmologist at least once and as early as possible; GeneReviews CSS proposes an annual examination without specifying cycloplegia, and the 2024 adult ARID1B cohort an examination every 2 years [3, 11, 52]. For hearing, after the initial screening, the PNDS requests a new test only if hearing loss is suspected, whereas GeneReviews CSS proposes an annual evaluation [3, 11]. In endocrinology, the PNDS allows evaluation at any time in the presence of growth delay, with bone age, thyroid function and GH secretion, without a fixed interval; the 2024 cohort proposes blood glucose and thyroid function every 3 years when risk factors are present [11, 52]. In oncology, the PNDS considers the data insufficient to assert an increased tumour risk and proposes no surveillance, whereas GeneReviews 2025 and Borja and colleagues (2023) propose quarterly AFP and liver ultrasound until 4 years of age in ARID1A, under shared decision-making; the difference stems mainly from the date, because the 2022 and 2023 hepatoblastoma cases and the pooled estimate of 3.6% postdate the PNDS [3, 11, 15]. For the recurrence risk after a de novo variant, the PNDS indicates approximately 1% due to germline mosaicism, a value close to the 1–2% of GeneReviews ARID1B [8, 11]. For preimplantation genetic testing, the PNDS reserves it for couples in whom one parent has proven germline mosaicism, whereas GeneReviews presents it as possible whenever the familial variant is known [3, 11]. On transition, the PNDS includes the move to adulthood among its objectives but sets no age, unlike the 2024 ARID1B cohort (16 years) [11, 52]. The PNDS contains no pharmacological recommendations or doses, so the doses in the following subsections retain the sources already indicated; nor does it mention off-label medicines in CSS [11].

7.2. Evaluations after diagnosis

Sources: [3, 8, 9, 39]. Routine ophthalmological surveillance should include screening for glaucoma, given the cases of congenital glaucoma and of adult glaucoma without external signs [87–89]. In patients with SOX11, Chu and colleagues propose baseline endocrine evaluation and magnetic resonance imaging of the olfactory tracts and pituitary [82]. In patients with ARID2, the skeletal evaluation should include the hip [7, 85]. The French PNDS defines as the minimum initial evaluation an estimate of the degree of ID from the interview, the examination and the documents of the various professionals involved, an assessment of the needs for school or vocational adaptation and for rehabilitation (psychomotor therapy, speech and language therapy, occupational therapy), visual and hearing screening, echocardiography, abdominal ultrasound and an age-appropriate dental examination; it adds, depending on the context, neurology if there is epilepsy, endocrinology with bone age, thyroid function and GH secretion if there is growth delay, gastroenterology if there are feeding or growth difficulties, physical and rehabilitation medicine or orthopaedics if there is joint or spinal deformity, child psychiatry or psychiatry if there are behavioural disorders and immunology if there is a relevant infectious history [11]. The initial evaluation also serves to assess the family environment, organise parental support and inform the family about social benefits [11].

7.3. Treatment of manifestations

Physiotherapy aims to maximise mobility and reduce late orthopaedic complications (contractures, scoliosis, hip dislocation), using wheelchairs, walkers, orthoses and adapted pushchairs; occupational therapy addresses fine motor skills for feeding, hygiene and dressing [3, 8]. Oromotor dysfunction is assessed at every visit, and choking, poor weight gain, frequent respiratory infections or unexplained food refusal warrant swallowing assessment before starting or continuing oral feeding [3, 8]. The organisation of educational and social services described in GeneReviews relates to the USA and varies between countries [3]. In adults, the emphasis shifts from acquiring to preserving skills, preventing functional loss, physical activity and weight control [52]. Relative strengths in memory and visual cues in ARID1B-RD support visual teaching strategies [78].

The French PNDS adds some practical indications [11]. There is no specific diet, and vaccines are not contraindicated. In stable patients, an annual consultation with the developmental specialist (genetics or paediatric neurology) suffices, in liaison with the attending physician, who is responsible for local follow-up and for detecting epilepsy, scoliosis, behavioural disorders, hypothyroidism and diabetes. Child psychiatric evaluation before school entry helps to detect ADHD, behavioural disorders and ASD. Gastroenterological evaluation may include swallowing studies, upper gastrointestinal imaging and endoscopy, and early dental evaluation identifies crowding and delayed dentition. The list of professionals includes neonatology and paediatric intensive care (malnutrition, cardiac complications), paediatric visceral surgery, dermatology for anomalies of the skin appendages, hyperkeratosis and hyperhidrosis, nephrology or urology for renal and urinary tract malformations and vesicoureteral reflux, orthodontics, physiotherapy, orthoptics and hearing-aid fitting, as well as the social services, which link the family with the teams and direct it to the support available; care should start early and continue without interruption. Therapeutic education has no CSS-specific content and follows that for developmental disorders with ID, according to the 2016 Inserm collective expert report on intellectual disabilities. In France, support is provided through the medico-social structures (CAMSP for early intervention, SESSAD for home-based educational support, medico-educational institutes and, for adults, MAS and FAM), through the Maison Départementale des Personnes Handicapées (MDPH) for benefits and through exemption from patient co-payment (exonération du ticket modérateur) as an off-list long-term condition [11].

7.4. Epilepsy

No antiseizure medicine has been studied specifically in BAFopathies, and GeneReviews recommends standard treatment by an experienced neurologist [3, 9]. In CSS and ARID1B-RD, the response is usually good: all 18 treated patients in the 2019 ARID1B cohort responded, and in the international registry seizures were controlled in all patients but one [39, 79]. In NCBRS, an initial response to levetiracetam and valproic acid is frequent, but complete remission is very rare, and many patients need several medicines [9]. The sources consulted do not indicate specific doses for these syndromes, which follow the general rules for each medicine. Educating parents about the forms in which seizures present is part of treatment [3]. The PNDS considers a paediatric neurology or neurology consultation indispensable when there is epilepsy and describes it, in patients with ARID1B, as of low activity and usually responsive to treatment [11]; in the 2019 cohort, however, 4 of 17 patients with a recorded frequency had monthly or more frequent seizures, and the lifetime risk estimated by Kaplan–Meier (approximately 40%) exceeds the observed proportion [39].

7.5. Somatropin (recombinant growth hormone)

Somatropin is not authorised for CSS. In these patients, it is used only when documented GH deficiency is present, which is an approved indication; without such deficiency, use is off-label and unsupported by trials [13, 39]. The biological rationale comes from the Arid1b+/− mouse model, with IGF-1 (insulin-like growth factor 1) deficiency, inadequate compensation by the GHRH/GH axis, and growth delay and muscle weakness corrected by GH supplementation [118]. In humans, evidence is limited to case reports and small series with documented deficiency: in the 2019 ARID1B cohort, 7 of 51 assessed patients had GH deficiency and 6 were treated, without response data [39]; a Greek girl with CSS-ARID1B, severe scoliosis and epilepsy, with deficiency diagnosed at 9 years, achieved a growth velocity of 5.4 cm/year on treatment [56]; a 4-year-old Romanian girl with height of −3.2 SD, pituitary hypoplasia and GH deficiency had a favourable response [119]; a patient with CSS7 (DPF2) had a sustained response for over 3 years [120]; a man with CSS12 (BICRA) had GH deficiency in childhood [94]. No guideline recommends somatropin in CSS without GH deficiency [3, 8].

Source: [13]. The label warnings apply directly to BAFopathies: scoliosis affects 26–30% of patients with ARID1B, hypothyroidism and diabetes are more frequent than in the general population, and tumour-predisposition genotypes (ARID1A in early childhood, truncating SMARCB1 and SMARCA4) require special consideration [13, 15, 39, 52]. In the case reported by Tao and colleagues, the reported dose (0.12–0.15 mg/kg/day) is incompatible with the label units and should not be used as a dosing reference [13, 55]. IGF-1, blood glucose and thyroid function were monitored every 3 months in that case [55].

Sources: [3, 13].

7.6. Clonazepam: the only controlled trial

Arid1b+/− mice have reduced inhibitory GABAergic synapses and parvalbumin interneurons, and the benzodiazepine improved tone, attention to tasks and anxiety in the model [3, 121]. On this basis, a Leiden group (Centre for Human Drug Research and Leiden University Medical Center) conducted a repurposing trial with two sequential designs between September 2021 and May 2022 (EUCTR 2019-003558-98; ISRCTN11225608) [14].

Source: [14]. Adverse effects occurred in 9 patients (56%) with clonazepam and 6 (38%) with placebo, mainly fatigue (8), somnolence (2) and nausea; there was one paradoxical reaction with agitation on day 4, leading to exclusion, and 5 of 15 required dose reduction [14]. Parents reported an effect on sleep [14]. The article abstract reports CGI-I worsening in 7 of 16 (44%), counting the excluded patient, and the text in 6 of 15 [14]. The study had 67% power to detect a 1-point CGI-I difference with 15 patients, and included only patients able to complete the test battery, potentially selecting less severe forms [14]. The article does not provide confidence intervals for CGI-I. The authors conclude that clonazepam offers no additional therapeutic value in ARID1B-RD, and that combining a randomised trial with N-of-1 studies separates the true effect from expectation bias [3, 14]. The result applies to ARID1B and does not constitute evidence for other genes [7].

7.7. Interventions without evidence, ongoing trials and recent developments

A retrospective Turkish series of 3 children with CSS-ARID1B treated with intranasal exosomes alongside rehabilitation described a mean rise in the Vineland-3 composite score from 83.0 to 90.7, without adverse events; the series has no control group, and the intervention is neither authorised nor recommended [122]. A ClinicalTrials.gov search in October 2026 (terms Coffin-Siris, Nicolaides-Baraitser, ARID1B and BAF-complex ID) identified no interventional therapeutic trials in BAFopathies, only observational registries such as CoRDS and Simons Searchlight, which includes ARID1B [123]. A negative search does not establish that studies do not exist. The Coffin-Siris Syndrome Foundation lists no clinical trials, and the Foundation for ARID1B Research states that its goal is to develop disease-modifying therapies, without named programmes [124, 125].

Sources: [48, 53, 58, 59, 61, 63, 65, 78, 82, 87–90, 94, 106, 107, 122]. No opinions, orphan designations or regulatory authorisations for BAFopathies were identified in the sources consulted.

7.8. Tumour risk management in allelic diseases

In RTPS carriers, treatment of rhabdoid tumours is multimodal and intensive (surgery, radiotherapy and chemotherapy), with a high complication rate; strategies to reduce radiotherapy, such as high-dose chemotherapy or proton therapy, are considered [10]. The international SIOPE ATRT01 trial compares three cycles of high-dose chemotherapy with focal radiotherapy as consolidation in a randomised phase III non-inferiority study [10]. Tazemetostat, an EZH2 inhibitor, has been studied in rhabdoid tumours in phase 1 and 2 trials; the FDA label for Tazverik® is limited to advanced epithelioid sarcoma in patients aged 16 years or older and relapsed follicular lymphoma in adults, and does not cover rhabdoid tumours or germline SMARCB1 carriers [7]. In women with SMARCA4-related RTPS, prophylactic bilateral salpingo-oophorectomy may be discussed after completion of reproductive plans, through an interdisciplinary decision, and remains an open question [10, 16]. In schwannomatosis, surgery is the most effective treatment, and pain and psychosocial needs are assessed annually [18].

7.9. Agents to avoid, pregnancy, perioperative care and transition

In RTPS carriers, limiting exposure to DNA-damaging agents (ionising radiation, including radiographs and CT scans, tobacco and ultraviolet radiation) and using radiopharmaceutical imaging only when essential are recommended to reduce secondary cancers; the recommendation is based on risk in young children rather than a demonstrated variant effect [10]. Surveillance in these patients therefore favours magnetic resonance imaging and ultrasound [16]. In BAFopathies without tumour predisposition, the sources consulted identify no contraindicated medicines; clonazepam may cause fatigue, somnolence and paradoxical reactions in these patients [14].

The sources consulted do not describe pregnancies in women with CSS or NCBRS; six adults with ARID1B transmitted the variant to children, two of them mosaic, and reproductive counselling and contraception are part of the transition to adulthood [9, 52]. The sources also do not describe specific anaesthetic recommendations, and the French PNDS, although it cites reports of difficult airway in CSS in its bibliography, makes no recommendations on anaesthesia [11]; the laryngomalacia, tracheomalacia, obstructive sleep apnoea, aspiration risk and heart defects documented in these patients are factors to consider in the pre-anaesthetic assessment [1, 3, 39]. Tracheostomy and respiratory support are associated with higher early mortality [3]. Transition to adult care should be planned from 16 years of age in ARID1B-RD, according to the adult cohort, and GeneReviews ARID1B 2019 proposes discussing financial, vocational and medical plans from 12 years of age; in NCBRS, GeneReviews proposes realistic plans from 10 years of age, modelled on the transition from paediatric to adult epilepsy care [8, 9, 52].

8. Reproductive Options and Family Counselling

8.1. Inheritance and recurrence risks

All autosomal BAFopathies follow dominant inheritance, and most cases result from de novo variants; rarely, a parent is affected [3, 20]. The X-linked form caused by PHF6 is the exception. Recurrence risk depends on parental status, the gene and variant location.

In the cohort of 143 patients with ARID1B, the variant was de novo in all 107 cases with parental testing [39]. Among the 87 adults, 6 transmitted the variant to children (2 were mosaic) and all variants inherited from non-mosaic parents were located in exon 1; the parents identified through testing of the child were mosaic or carriers of an exon 1 variant [52]. Detailed phenotyping of these parents often shows a mild picture, and early truncating ARID1B variants may have reduced penetrance [3, 48, 130]. In SOX11, a mother with mild learning difficulties transmitted the variant to two daughters with ID; all three had hypoplastic nails and similar facial features [3, 127]. In ARID2, 21 of 27 variants were de novo, 2 inherited from a healthy mosaic parent and 1 from an affected mother [44]. The French PNDS (National Diagnosis and Care Protocol) summarises the same reasoning: parental testing detects somatic mosaicism but does not exclude germline mosaicism, which determines a recurrence risk of the order of 1%; if the variant comes from a parent with an attenuated form, the risk is 50% for siblings, as it is for the proband's offspring; and unaffected siblings have the general population risk for their own offspring [11].

8.2. Parental and cascade testing

Parental testing is recommended in all molecularly diagnosed cases, preferably before a further pregnancy, because ARID1B penetrance is very high but possibly incomplete [3, 39]. Leucocyte DNA may not detect all somatic mosaicism and does not detect exclusively gonadal mosaicism, and a parent with somatic and gonadal mosaicism may be minimally or mildly affected [3, 128]. Mosaicism-sensitive methods are used, particularly for ARID1A and ARID2; the episignature of mosaic ARID2 parents was negative and therefore cannot exclude them [15, 44]. An apparently negative family history is confirmed only after clinical and molecular assessment of the parents [3]. In RTPS, testing and counselling, including prenatal diagnosis, are offered to all first-degree relatives, and to siblings even when parents test negative; relatives without the variant have population risk [10, 16]. In NCBRS, finding the variant in an asymptomatic parent is very unlikely, but inheritance from an unaffected mosaic parent has been described [9, 129].

8.3. Prenatal and preimplantation diagnosis

Once the familial variant has been identified, prenatal diagnosis and preimplantation genetic testing are possible, and differences in perspective between professionals and families should be respected, particularly when testing is aimed at termination of pregnancy [3, 8]. Even with a de novo variant and negative parents, the residual risk due to gonadal mosaicism justifies offering prenatal diagnosis [8, 9]. In the absence of a known variant, prenatal genomic sequencing of fetuses with compatible anomalies may identify a BAFopathy, particularly ARID1A, but agenesis of the corpus callosum in fetuses with ARID1B is often invisible on ultrasound [12]. The French PNDS proposes discussing prenatal diagnosis with the parents of a patient, given the risk of germline mosaicism and the degree of disability, and with a proband who has an attenuated form; it reserves preimplantation diagnosis for cases with proven germline mosaicism in a parent [11]. The identification of a variant in a CSS gene during pregnancy, increasingly frequent with prenatal genomic tests requested for malformations such as agenesis of the corpus callosum, should be followed by a genetics clinic results consultation explaining the possible consequences of the variant; in France, it may lead to termination of pregnancy for medical reasons [11].

8.4. Content of counselling

The ideal time for counselling on risks and reproductive options is before pregnancy, and it should be offered to all couples with an affected child [3]. Banking of DNA from probands without a molecular diagnosis is recommended, for future reanalysis [3, 131]. Counselling should explain the variability of expression, the possibility of mildly affected parents with ARID1B exon 1 variants and the meaning of mosaicism [3, 52]. For genes with allelic tumour disorders, the variant class determines whether there is oncological surveillance; a carrier of a SMARCB1 missense variant with CSS does not have the same risk as a carrier of a deletion [16, 46]. In adults with reproductive capacity, the 50% risk for offspring and contraception are part of transition planning [52]. In RTPS, testing of siblings and other minor relatives is justified because it allows surveillance to begin from birth [10, 16]. The French PNDS describes disclosure of the diagnosis: in a private setting, with time, preferably with both parents, by the physician who requested the molecular test and, depending on the context, another member of the team (psychologist, genetic counsellor, nurse); it includes explanation of the diagnosis and results, of the disease and the mode of transmission for siblings, of the need for regular multidisciplinary follow-up and of the care plan; the presence of a psychologist, during or immediately after disclosure or during follow-up, is considered essential, and the contact details of patient associations are given to the family, who decide whether to use them [11].

9. Special Considerations: Population Diversity and Epidemiology

9.1. Prevalence and what each figure measures

There is no population-based measurement of the prevalence or incidence of CSS. The available estimates measure different things. The GeneReviews CSS estimate of approximately 1/100,000 births is an approximation based on known cases and the registry, which the authors consider an underestimate [3]. The estimate of 1/10,000 to 1/100,000 for ARID1B-RD derives from the frequency of ARID1B variants in ID cohorts, such as the 0.9% (8 of 887) in the German cohort of Hoyer and colleagues and the approximately 1% of unexplained ID in large exome studies [8, 27, 39]. The Orphanet category below 1/1,000,000 rests on 190 published cases or families and measures the literature, not the population [4]. For NCBRS, the prevalence is unknown and very low (fewer than 100 described; 61 cases or families in Orphanet) [9, 99]. For RTPS, 25–35% of patients with a newly diagnosed rhabdoid tumour have a germline SMARCB1 variant, and the annual incidence of rhabdoid tumour in the USA below 15 years of age is 0.89 per million for atypical teratoid/rhabdoid tumour, 0.32 per million for extracranial malignant rhabdoid tumour and 0.2 per million for rhabdoid tumour of the kidney [10]. The 2021 French PNDS (National Diagnosis and Care Protocol) presents a fourth estimate, of approximately 2/10,000, obtained by multiplying an ID prevalence of 2% by the proportion of up to 1% of unexplained ID that it attributes to CSS; from this derive more than 100 births per year in France and more than 800 in Europe [11]. It is twenty times higher than the GeneReviews CSS estimate and double the upper limit of the estimate for ARID1B-RD (1/10,000); it includes the non-syndromic forms detected by exome, and it is an extrapolation, not a measurement [3, 8, 11]. The same document reports identification of the genetic cause in 50–70% of cases and more than 300 cases published at the time [11]. Adults are probably underdiagnosed, because many have not undergone genetic testing [3].

9.2. Gene distribution in cohorts

Gene proportions vary with the recruitment method and are not prevalences.

Sources: [1, 3, 12, 15, 36, 68]. The difference between the predominance of ARID1B in clinical cohorts (51–75%) and approximately 37% in GeneReviews reflects the denominator: cohorts count molecularly diagnosed patients, whereas GeneReviews includes the approximately 40% without an identified gene [3, 39].

9.3. Geography and populations

Data come mainly from the USA (CSS/BAF registry, since 2015), the Netherlands (Leiden and Erasmus MC), Germany (Erlangen, Würzburg, Düsseldorf), France (Necker and the AnDDI-Rares network), Japan, China and Turkey [1, 25, 27, 39, 44, 83, 86, 106, 107, 132]. In China, 6 of 8 patients with a clinical CSS diagnosis had ARID1B variants, five of them novel [132]. In Turkey, the series of 8 children included 7 with ARID1B and 1 with SMARCA4 [83]. The facial recognition model trained in Paris was validated in patients from Bangalore and Tbilisi [84]. The sources consulted describe no founder variants: variants are almost all de novo and private, with recurrent exceptions such as SMARCB1 p.(Lys364del) and SMARCC2 c.1926+1G>T, which arise de novo in different populations [3]. The only Latin American reference in the local sources is a Chilean series of two cases cited in SGRT [114]; the Mexican series of prenatally diagnosed 6q25.1q27 deletions provides another Latin American observation [95]. The sources consulted contain no data from Portugal or Brazil, which does not establish that such data do not exist.

9.4. Special populations

Adults are a growing and poorly characterised population: the adult cohort of 35 individuals and the cohort of 87 adults with ARID1B show that obesity, scoliosis, loss of skills and patellar dislocation become more important with age [52, 86]. Fetuses with non-mosaic ARID1A have a severe and often lethal phenotype, whereas survivors are frequently mosaic [12, 15]. The registry of 208 individuals was 60% male, and the 2019 ARID1B cohort 48.3% female; the sources describe no sex-specific clinical differences other than genital findings [1, 39].

9.5. Reference networks and patient organisations

ERN ITHACA (rare malformation syndromes, ID and other neurodevelopmental disorders) is the relevant European network, with neurodevelopment and fetal medicine working groups, the ILIAD registry and the SysNDD database; on the access date, its guideline pages did not include a CSS document [17]. ERN GENTURIS covers RTPS and schwannomatosis [18]. The Coffin-Siris Syndrome Foundation (USA) provides information for newly diagnosed families and publicises the ethics-approved International Registry of CSS and Related Disorders at Seattle Children's Hospital (eligible carriers of variants in ARID1A, ARID1B, ARID2, BICRA, DPF2, SMARCA4, SMARCB1, SMARCC2, SMARCD1, SMARCE1, SOX4 or SOX11), the CSS clinic at the same hospital, an ARID1B neurodevelopmental study and the CARE4ARID1B natural history study (2 to 18 years; seven centres in Sweden, Germany, Turkey, Italy, France, Canada and Israel) [3, 124]. The Foundation for ARID1B Research supports families and is preparing a global registry [125]. The Leiden ARID1B registry (Human Disease Genes), CoRDS and Simons Searchlight collect natural history data [3, 8, 123]. For NCBRS, there are the NCBRS Worldwide Foundation and the Nicolaides-Baraitser Syndrome International Consortium [9].

In France, the organisation is based on the Reference Centre for Intellectual Disabilities of Rare Causes, coordinated at the Pitié-Salpêtrière Hospital (AP-HP, Paris), with ten constituent centres (Lyon, Marseille, Trousseau, two at Necker, Robert Debré, Dijon, Brest, Rennes and Strasbourg) and thirteen competence centres (Tours, Kremlin-Bicêtre, Lille, Amiens, Besançon, Nancy, Nantes, Nice, Pointe-à-Pitre, Reims, Toulouse, Bordeaux and Montpellier) in the 2021 version, integrated into the DéfiScience network (rare diseases of brain development and ID); the AnDDI-Rares network (developmental anomalies) is also involved in diagnosis [11]. The associations listed are the Association Coffin-Siris France, the Alliance Maladies Rares and UNAPEI (federation of associations of people with intellectual disability and their families), together with the Coffin-Siris Syndrome Foundation and a Dutch website dedicated to ARID1B (arid1bgene.com) [11]. The PNDS describes the associations as partners of the centres, sources of information and of peer support between families, and funders of research projects [11].

10. Illustrative Clinical Case

In 2022, Tao and colleagues published in BMC Pediatrics the case of a Chinese girl with CSS caused by an ARID1B variant, short stature due to GH deficiency and multiple melanocytic naevi [55]. She was the second daughter of healthy, non-consanguineous parents, born at term weighing 2,500 g and measuring 49.0 cm (3rd centile), with a black congenital naevus extending from the left temporal region and forehead to the entire left scalp, which did not grow. She sat at 9 months, walked unsupported at 18 months and said her first words at 24 months; later she spoke in incomplete sentences. Cervical vitiligo appeared at 4 years, ADHD was diagnosed at 7 years, with an IQ of 70, and generalised tonic–clonic seizures began at 9 years [55].

At 10 years and 9 months, her height was 123 cm (below the 3rd centile) and bone age 8 years and 10 months (Greulich–Pyle method). Peak GH on stimulation testing was 3.17 ng/mL and IGF-1 was 136 ng/mL (reference 117 to 771), leading to a diagnosis of GH deficiency and 12 months of somatropin treatment, with a gain of 13 cm and quarterly monitoring of IGF-1, blood glucose and thyroid function; the dose was reported in units incompatible with the label and is not reproduced [13, 55]. At 12 years, referred for ID and short stature, she measured 137 cm (below the 3rd centile) and weighed 36.5 kg (25th–50th centile). Examination showed coarse facial features, a low hairline, thick eyebrows, a broad nasal root, a high palate, micrognathia, a thick lower lip, multiple melanocytic naevi on the face and trunk, a single right palmar crease and, on radiography, dysplasia of the middle phalanx of the 5th digit [55]. Full blood count, biochemistry, tumour markers, thyroid function and brain and pituitary magnetic resonance imaging were normal [55].

Exome sequencing with Sanger confirmation identified the de novo variant NM_001374828.1:c.3099delT p.(Phe1034SerfsTer3), classified as pathogenic (ClinVar SCV002549918) [55]. VariantValidator normalisation, shifting to the most 3′ position, describes the same deletion as NM_001374828.1:c.3101del p.(Phe1034SerfsTer4), corresponding to NM_020732.3:c.2891del [47]. The diagnosis was revised to ARID1B-related CSS. Somatropin was continued for a further 14 months, with a final height of 150 cm (10th centile) and a total gain of 28 cm over 3 years according to the authors. Over 4 years of follow-up, the naevi did not grow or show malignant transformation, and dermatological and oncological assessments considered the giant naevus probably benign [55]. Figure 1 of the article (https://pmc.ncbi.nlm.nih.gov/articles/PMC9375425/figure/Fig1/) documents, according to the caption, coarse facial features with copper-coloured hair and thick eyebrows, a small chin and high palate, facial naevi, the giant congenital scalp naevus, the single palmar crease and the short middle phalanx of the 5th digit; Figure 2 (https://pmc.ncbi.nlm.nih.gov/articles/PMC9375425/figure/Fig2/) shows the Sanger electropherogram [55].

The case illustrates four points. First, digital hypoplasia may be subtle and confined to the middle phalanx, and the diagnosis emerged only in adolescence after years of follow-up for short stature. Second, somatropin was used with documented GH deficiency, the only circumstance in which its use is on-label; use in this patient cannot be generalised as routine CSS treatment [13, 39]. Third, the multiple naevi and vitiligo, proposed by the authors as CSS features on the basis of three previous cutaneous cases, represent an association not yet confirmed in cohorts. Fourth, the article contains internal inconsistencies (the dose in mg/kg/day and the reference to Sanger testing “of the patient and father” for a variant declared de novo), reinforcing the need to read case reports critically [55].

11. Prognosis and Long-Term Follow-Up

11.1. Prognosis

There are no long-term follow-up studies, and life expectancy is unknown [3]. Some children have died from aspiration pneumonia or seizures, although this is not frequent, and the need for tracheostomy or respiratory support is associated with higher early mortality [3, 22]. The 2019 ARID1B cohort recorded two deaths among 143 patients (at 9 years and after 24 years of age), and the adult cohort two deaths among 87 (renal abscesses at 24 years and asphyxia from choking at 47 years) [39, 52]. GeneReviews CSS 2025 cites an individual alive at 69 years (Määttänen 2018), and GeneReviews ARID1B 2019 refers to the same publication as describing a 60-year-old woman and an individual alive at 51 years [3, 8, 133]; the 2024 adult cohort included patients up to 69 years of age [52]. The French PNDS (National Diagnosis and Care Protocol), which cites the same case diagnosed at 69 years, stresses that no particular complications of ageing have been described and that prospective follow-up studies are under way [11]. In NCBRS, the oldest individual described lived to 33 years [9].

Functional prognosis depends mainly on the degree of ID and communication. Most adults with ARID1B live with their parents and need support with instrumental activities, and 25% lose skills from the fourth decade onwards, shifting the goal of follow-up to preservation of function [52]. Adults with ARID1B exon 1 variants have, on average, greater independence [52]. Sparse hair tends to improve with age in CSS and worsen in NCBRS, and facial features coarsen in both [9, 20]. In NCBRS, worsening seizures are accompanied by loss of speech, and accelerated epigenetic age suggests early biological ageing [9, 102].

Regarding tumour risk, the absence of neoplasms among the 376 registry participants and the 35 adults with CSS, and the single probably benign neoplasm among 73 adults with ARID1B, indicate low risk in most genotypes, but follow-up is short and these numbers do not establish zero lifetime risk [15, 52, 86]. In ARID1A, the pooled hepatoblastoma prevalence of 3.6% (95% confidence interval 0.79% to 10.4%) exceeds the 1% threshold used in the USA to recommend surveillance, but the interval includes lower values, and the usual European threshold is 5% [15]. The authors recommend shared decision-making [15]. Van der Sluijs, Vergano, Roeder, Jongmans and Santen published a revised surveillance recommendation for children with a pathogenic ARID1A variant in 2023 [98]. The most recent international consensus for Beckwith–Wiedemann syndrome recommends ultrasound alone, without AFP, according to the review by Borja and colleagues; this difference explains why the GeneReviews summary mentions AFP with consideration of ultrasound, whereas the surveillance table states AFP and ultrasound [3, 15].

11.2. Surveillance in BAFopathies

Routine oncological surveillance is not recommended in CSS outside hepatoblastoma in ARID1A, nor in ARID1B (risk below 1%) [3, 8, 15]. The 2021 French PNDS, which predates the pooled data on ARID1A, proposes no oncological surveillance [11]. In SMARCB1 carriers with CSS, clinicians should be alert to schwannomas and meningiomas in adulthood [3].

11.3. Surveillance in allelic diseases with tumour predisposition

SIOPE presents these protocols as provisional recommendations, with limited evidence and ideally applied in a research setting; the protocol note indicates that carriers of missense variants may need different surveillance, and AACR considered the value of surveillance unknown in 2017 [16, 117]. RTPS surveillance therefore does not apply to patients with CSS caused by SMARCB1 missense variants or in-frame deletions, or SMARCA4 missense variants [15, 16]. Women with a truncating SMARCA4 variant, including those with mild CSS, warrant SCCOHT surveillance [16, 57]. ClinGen classifies RTPS actionability as limited and assigns no actionability to CSS [7]. In RTPS1, overall survival is 10–20%, and 6 of 7 patients with atypical teratoid/rhabdoid tumour and RTPS2 died within 20 months [16].

12. Summary and Final Considerations

Coffin–Siris syndrome is the most recognisable form of a group of neurodevelopmental disorders caused by variants in genes of the BAF chromatin-remodelling complex. The presentation combines developmental delay and ID, generally moderate to severe with expressive language more affected, hypotonia, feeding difficulties, coarse facial features, hypertrichosis with sparse scalp hair and hypoplasia of the nail and distal phalanx of the 5th digit. None of these signs is mandatory, and expansion of the phenotype through molecular diagnosis has shown that many patients, particularly those with ARID1B, BICRA or SMARCD1, lack classical digital findings. ARID1B is the most frequent gene, and ARID1B-related ID without a CSS phenotype forms part of the same spectrum, with the same management.

The molecular mechanism separates loss-of-function genes (ARID1A, ARID1B, ARID2, BICRA, SMARCC2, SOX4, SOX11) from genes in which CSS results from dominant-negative or gain-of-function variants (SMARCA4, SMARCB1, SMARCE1, DPF2, and SMARCA2 in Nicolaides–Baraitser syndrome). In the three SMARC genes, loss-of-function variants cause allelic diseases with tumour predisposition, and variant class determines surveillance: rhabdoid tumour protocols apply to truncating variants and deletions, not to CSS missense variants. In CSS itself, documented tumour risk is limited to hepatoblastoma in ARID1A, at approximately 3.6%, for which quarterly surveillance until 4 years is proposed through shared decision-making.

Diagnosis relies on a multigene panel or exome sequencing with copy-number variation analysis, supplemented by DNA methylation episignature testing when there are VUS or when testing is negative despite strong suspicion. Approximately 40% of patients with clinical CSS still have no identified gene, and mosaicism, deep intronic variants and structural variants explain some of these cases. Early truncating ARID1B variants have been debated since 2025 and require additional evidence.

There is no disease-modifying treatment. The only controlled trial, with clonazepam in ARID1B, was negative, and somatropin is indicated only when GH deficiency is documented. Follow-up is multidisciplinary and organised around age-specific problems: feeding, growth, epilepsy, vision, hearing, teeth, spine and behaviour in childhood; weight, endocrinology, patella, independence and loss of skills in adulthood. Genetic counselling starts from autosomal dominant inheritance, almost always de novo, but must consider gonadal mosaicism and mildly affected parents with ARID1B exon 1 variants. Population prevalence estimates, long-term cohorts and data from Portugal and Brazil are lacking.

13. Bibliographic References

1. Vasko A, Drivas TG, Schrier Vergano SA. Genotype-Phenotype Correlations in 208 Individuals with Coffin-Siris Syndrome. Genes (Basel). 2021;12(6):937. doi:10.3390/genes12060937. PMID: 34205270. PMC8233770.
2. Barish S, Barakat TS, Michel BC, Mashtalir N, Phillips JB, Valencia AM, et al. BICRA, a SWI/SNF Complex Member, Is Associated with BAF-Disorder Related Phenotypes in Humans and Model Organisms. Am J Hum Genet. 2020;107(6):1096-1112. doi:10.1016/j.ajhg.2020.11.003. PMID: 33232675. PMC7820627.
3. Schrier Vergano S, Santen G, Wieczorek D, Matsumoto N. Coffin-Siris Syndrome. In: Adam MP, Feldman J, Mirzaa GM, et al., editors. GeneReviews® [Internet]. Seattle (WA): University of Washington, Seattle; 1993–. Available from: https://www.ncbi.nlm.nih.gov/books/NBK131811/ PMID: 23556151.
4. Orphanet. Coffin-Siris syndrome (ORPHA:1465). Available from: https://www.orpha.net/en/disease/detail/1465
5. Sadikovic B, Levy MA, Kerkhof J, Aref-Eshghi E, Schenkel L, Stuart A, et al. Clinical epigenomics: genome-wide DNA methylation analysis for the diagnosis of Mendelian disorders. Genet Med. 2021;23(6):1065-1074. doi:10.1038/s41436-020-01096-4. PMID: 33547396. PMC8187150.
6. Levy MA, McConkey H, Kerkhof J, Barat-Houari M, Bargiacchi S, Biamino E, et al. Novel diagnostic DNA methylation episignatures expand and refine the epigenetic landscapes of Mendelian disorders. HGG Adv. 2022;3(1):100075. doi:10.1016/j.xhgg.2021.100075. PMID: 35047860. PMC8756545.
7. GeneGuidance — local gene-curation database (geneguidance.db; records ARID1B, ARID1A, ARID2, SMARCA4, SMARCB1, SMARCE1, SMARCC2, SOX11, DPF2, SOX4, SMARCA2, SMARCD1, PHF6; source llm_curated).
8. Vergano SA, van der Sluijs PJ, Santen G. ARID1B-Related Disorder. In: Adam MP, Feldman J, Mirzaa GM, et al., editors. GeneReviews® [Internet]. Seattle (WA): University of Washington, Seattle; 1993–. Available from: https://www.ncbi.nlm.nih.gov/books/NBK541502/ PMID: 31132234.
9. Abdul-Rahman O, May F. SMARCA2-Related Nicolaides-Baraitser Syndrome. In: Adam MP, Feldman J, Mirzaa GM, et al., editors. GeneReviews® [Internet]. Seattle (WA): University of Washington, Seattle; 1993–. Available from: https://www.ncbi.nlm.nih.gov/books/NBK321516/ PMID: 26468571.
10. Nemes K, Bens S, Bourdeaut F, Johann P, Kordes U, Siebert R, et al. Rhabdoid Tumor Predisposition Syndrome. In: Adam MP, Feldman J, Mirzaa GM, et al., editors. GeneReviews® [Internet]. Seattle (WA): University of Washington, Seattle; 1993–. Available from: https://www.ncbi.nlm.nih.gov/books/NBK469816/ PMID: 29215836.
11. Haute Autorité de Santé. Protocole National de Diagnostic et de Soins (PNDS): Syndromes de Coffin-Siris et de Nicolaides-Baraitser (BAFopathies). Centre de Référence Déficiences Intellectuelles de causes rares, filière DéfiScience. Saint-Denis La Plaine: HAS; 4 November 2021 (text dated September 2021). Available from: https://www.has-sante.fr/jcms/p_3295021
12. van der Sluijs PJ, Joosten M, Alby C, Attié-Bitach T, Gilmore K, Dubourg C, et al. Discovering a new part of the phenotypic spectrum of Coffin-Siris syndrome in a fetal cohort. Genet Med. 2022;24(8):1753-1760. doi:10.1016/j.gim.2022.04.010. PMID: 35579625. PMC9378544.
13. U.S. Food and Drug Administration. GENOTROPIN (somatropin) for injection — prescribing information. Pfizer Laboratories Div Pfizer Inc; BLA 020280; SPL set id ffebf88b-d257-4542-9808-74d9b7167765, effective version 2026-08-04. Available from: https://api.fda.gov/drug/label.json?search=openfda.brand_name:%22GENOTROPIN%22 (DailyMed: https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=ffebf88b-d257-4542-9808-74d9b7167765).
14. van der Sluijs PJ, Safai Pour K, Berends CL, Kruizinga MD, Müller AR, van Eeghen AM, et al. Clonazepam repurposing in ARID1B patients through conventional RCT and N-of-1 trials: an experimental strategy for orphan disease development. J Med Genet. 2025;62(3):210-218. doi:10.1136/jmg-2024-109951. PMID: 39740803. PMC11877031.
15. Borja NA, Schrier Vergano SA, Tekin M. Coffin-Siris syndrome and cancer susceptibility. Genet Med Open. 2023;1(1):100818. doi:10.1016/j.gimo.2023.100818. PMID: 39669229. PMC11613554.
16. Frühwald MC, Nemes K, Boztug H, Cornips MCA, Evans DG, Farah R, et al. Current recommendations for clinical surveillance and genetic testing in rhabdoid tumor predisposition: a report from the SIOPE Host Genome Working Group. Fam Cancer. 2021;20(4):305-316. doi:10.1007/s10689-021-00229-1. PMID: 33532948. PMC8484234.
17. ERN ITHACA — European Reference Network for Rare Malformation Syndromes, Intellectual and Other Neurodevelopmental Disorders [Internet]. Available from: https://ern-ithaca.eu/
18. Evans DG, Mostaccioli S, Pang D, Fadzil O Connor M, Pittara M, Champollion N, et al. ERN GENTURIS clinical practice guidelines for the diagnosis, treatment, management and surveillance of people with schwannomatosis. Eur J Hum Genet. 2022;30(7):812-817. doi:10.1038/s41431-022-01086-x. PMID: 35361920. PMC9259735.
19. Coffin GS, Siris E. Mental retardation with absent fifth fingernail and terminal phalanx. Am J Dis Child. 1970;119(5):433-9. doi:10.1001/archpedi.1970.02100050435009. PMID: 5442442.
20. Jones KL, Jones MC, del Campo M. Smith's Recognizable Patterns of Human Malformation. 8th ed. Philadelphia: Elsevier; 2022. Coffin-Siris syndrome; p. 816–819.
21. Mannino EA, Miyawaki H, Santen G, Schrier Vergano SA. First data from a parent-reported registry of 81 individuals with Coffin-Siris syndrome: Natural history and management recommendations. Am J Med Genet A. 2018;176(11):2250-2258. doi:10.1002/ajmg.a.40471. PMID: 30276971.
22. Schrier SA, Bodurtha JN, Burton B, Chudley AE, Chiong MA, D'avanzo M G, Lynch SA, Musio A, Nyazov DM, Sanchez-Lara PA, Shalev SA, Deardorff MA. The Coffin-Siris syndrome: a proposed diagnostic approach and assessment of 15 overlapping cases. Am J Med Genet A. 2012;158A:1865–76. PMID: 22711679.
23. Nicolaides P, Baraitser M. An unusual syndrome with mental retardation and sparse hair. Clin Dysmorphol. 1993;2(3):232-6. PMID: 8287185.
24. Morin G, Villemain L, Baumann C, Mathieu M, Blanc N, Verloes A. Nicolaides-Baraitser syndrome: confirmatory report of a syndrome with sparse hair, mental retardation, and short stature and metacarpals. Clin Dysmorphol. 2003;12:237-40. PMID: 14564210.
25. Tsurusaki Y, Okamoto N, Ohashi H, Kosho T, Imai Y, Hibi-Ko Y, et al. Mutations affecting components of the SWI/SNF complex cause Coffin-Siris syndrome. Nat Genet. 2012;44(4):376-8. doi:10.1038/ng.2219. PMID: 22426308.
26. Santen GW, Aten E, Sun Y, Almomani R, Gilissen C, Nielsen M, et al. Mutations in SWI/SNF chromatin remodeling complex gene ARID1B cause Coffin-Siris syndrome. Nat Genet. 2012;44(4):379-80. doi:10.1038/ng.2217. PMID: 22426309.
27. Hoyer J, Ekici AB, Endele S, Popp B, Zweier C, Wiesener A, et al. Haploinsufficiency of ARID1B, a member of the SWI/SNF-a chromatin-remodeling complex, is a frequent cause of intellectual disability. Am J Hum Genet. 2012;90(3):565-72. doi:10.1016/j.ajhg.2012.02.007. PMID: 22405089. PMC3309205.
28. Van Houdt JK, Nowakowska BA, Sousa SB, van Schaik BD, Seuntjens E, Avonce N, et al. Heterozygous missense mutations in SMARCA2 cause Nicolaides-Baraitser syndrome. Nat Genet. 2012;44(4):445-9, S1. doi:10.1038/ng.1105. PMID: 22366787.
29. Tsurusaki Y, Koshimizu E, Ohashi H, Phadke S, Kou I, Shiina M, Suzuki T, Okamoto N, Imamura S, Yamashita M, Watanabe S, Yoshiura K, Kodera H, Miyatake S, Nakashima M, Saitsu H, Ogata K, Ikegawa S, Miyake N, Matsumoto N. De novo SOX11 mutations cause Coffin-Siris syndrome. Nat Commun. 2014a;5:4011. PMID: 24886874.
30. Hempel A, Pagnamenta AT, Blyth M, Mansour S, McConnell V, Kou I, et al. Deletions and de novo mutations of SOX11 are associated with a neurodevelopmental disorder with features of Coffin-Siris syndrome. J Med Genet. 2016;53(3):152-62. doi:10.1136/jmedgenet-2015-103393. PMID: 26543203. PMC4789813.
31. Vasileiou G, Vergarajauregui S, Endele S, Popp B, Büttner C, Ekici AB, et al. Mutations in the BAF-Complex Subunit DPF2 Are Associated with Coffin-Siris Syndrome. Am J Hum Genet. 2018;102(3):468-479. doi:10.1016/j.ajhg.2018.01.014. PMID: 29429572. PMC5985265.
32. Gazdagh G, Blyth M, Scurr I, Turnpenny P, Mehta S, Armstrong R, McEntagart M, Newbury-Ecob R, Tobias E, Joss S, et al. Extending the clinical and genetic spectrum of ARID2 related intellectual disability. A case series of 7 patients. Eur J Med Genet. 2019;62:27–34. PMID: 29698805.
33. Machol K, Rousseau J, Ehresmann S, Garcia T, Nguyen TTM, Spillmann RC, et al. Expanding the Spectrum of BAF-Related Disorders: De Novo Variants in SMARCC2 Cause a Syndrome with Intellectual Disability and Developmental Delay. Am J Hum Genet. 2019;104(1):164-178. doi:10.1016/j.ajhg.2018.11.007. PMID: 30580808. PMC6323608.
34. Zawerton A, Yao B, Yeager JP, Pippucci T, Haseeb A, Smith JD, et al. De Novo SOX4 Variants Cause a Neurodevelopmental Disease Associated with Mild Dysmorphism. Am J Hum Genet. 2019;104(2):246-259. doi:10.1016/j.ajhg.2018.12.014. PMID: 30661772. PMC6369454.
35. Nixon KCJ, Rousseau J, Stone MH, Sarikahya M, Ehresmann S, Mizuno S, et al. A Syndromic Neurodevelopmental Disorder Caused by Mutations in SMARCD1, a Core SWI/SNF Subunit Needed for Context-Dependent Neuronal Gene Regulation in Flies. Am J Hum Genet. 2019;104(4):596-610. doi:10.1016/j.ajhg.2019.02.001. PMID: 30879640. PMC6451697.
36. Kosho T, Okamoto N. Genotype-phenotype correlation of Coffin-Siris syndrome caused by mutations in SMARCB1, SMARCA4, SMARCE1, and ARID1A. Am J Med Genet C Semin Med Genet. 2014;166C(3):262-75. doi:10.1002/ajmg.c.31407. PMID: 25168959.
37. Aref-Eshghi E, Bend EG, Hood RL, Schenkel LC, Carere DA, Chakrabarti R, et al. BAFopathies' DNA methylation epi-signatures demonstrate diagnostic utility and functional continuum of Coffin-Siris and Nicolaides-Baraitser syndromes. Nat Commun. 2018;9(1):4885. doi:10.1038/s41467-018-07193-y. PMID: 30459321. PMC6244416.
38. Al-Jawahiri R, Foroutan A, Kerkhof J, McConkey H, Levy M, Haghshenas S, et al. SOX11 variants cause a neurodevelopmental disorder with infrequent ocular malformations and hypogonadotropic hypogonadism and with distinct DNA methylation profile. Genet Med. 2022;24(6):1261-1273. doi:10.1016/j.gim.2022.02.013. PMID: 35341651. PMC9245088.
39. van der Sluijs PJ, Jansen S, Vergano SA, Adachi-Fukuda M, Alanay Y, AlKindy A, et al. The ARID1B spectrum in 143 patients: from nonsyndromic intellectual disability to Coffin-Siris syndrome. Genet Med. 2019;21(6):1295-1307. doi:10.1038/s41436-018-0330-z. PMID: 30349098. PMC6752273.
40. Vergano SS, Deardorff MA. Clinical features, diagnostic criteria, and management of Coffin-Siris syndrome. Am J Med Genet C Semin Med Genet. 2014;166C(3):252-6. doi:10.1002/ajmg.c.31411. PMID: 25169447.
41. Orphanet. 6q25.2q25.3 microdeletion syndrome (ORPHA:251056). Available from: https://www.orpha.net/en/disease/detail/251056
42. Ronan JL, Wu W, Crabtree GR. From neural development to cognition: unexpected roles for chromatin. Nat Rev Genet. 2013;14:347–59. PMID: 23568486.
43. Valencia AM, Sankar A, van der Sluijs PJ, Satterstrom FK, Fu J, Talkowski ME, et al. Landscape of mSWI/SNF chromatin remodeling complex perturbations in neurodevelopmental disorders. Nat Genet. 2023;55(8):1400-1412. doi:10.1038/s41588-023-01451-6. PMID: 37500730. PMC10412456.
44. Houdayer C, Rooney K, van der Laan L, Bris C, Alders M, Bahr A, et al. ARID2-related disorder: further delineation of the clinical phenotype of 27 novel individuals and description of an epigenetic signature. Eur J Hum Genet. 2025;33(11):1422-1431. doi:10.1038/s41431-025-01798-w. PMID: 40044822. PMC12583565.
45. Bosch E, Popp B, Güse E, Skinner C, van der Sluijs PJ, Maystadt I, Pinto AM, Renieri A, Bruno LP, Granata S, Marcelis C, Baysal Ö, Hartwich D, Holthöfer L, Isidor B, Cogne B, Wieczorek D, Capra V, Scala M, De Marco P, Ognibene M, Jamra RA, Platzer K, Carter LB, Kuismin O, van Haeringen A, Maroofian R, Valenzuela I, Cuscó I, Martinez-Agosto JA, Rabani AM, Mefford HC, Pereira EM, Close C, Anyane-Yeboa K, Wagner M, Hannibal MC, Zacher P, Thiffault I, Beunders G, Umair M, Bhola PT, McGinnis E, Millichap J, van de Kamp JM, Prijoles EJ, Dobson A, Shillington A, Graham BH, Garcia EJ, Galindo MK, Ropers FG, Nibbeling EAR, Hubbard G, Karimov C, Goj G, Bend R, Rath J, Morrow MM, Millan F, Salpietro V, Torella A, Nigro V, Kurki M, Stevenson RE, Santen GWE, Zweier M, Campeau PM, Severino M, Reis A, Accogli A, Vasileiou G. Elucidating the clinical and molecular spectrum of SMARCC2-associated NDD in a cohort of 65 affected individuals. Genet Med. 2023;25:100950. PMID: 37551667.
46. Kehrer-Sawatzki H, Cooper DN. SMARCB1-related schwannomatosis and other SMARCB1-associated phenotypes: clinical spectrum and molecular pathogenesis. Fam Cancer. 2025;24(3):64. doi:10.1007/s10689-025-00486-4. PMID: 40794298. PMC12343709.
47. Freeman PJ, Hart RK, Gretton LJ, Brookes AJ, Dalgleish R. VariantValidator: Accurate validation, mapping, and formatting of sequence variation descriptions. Hum Mutat. 2018;39(1):61-68. doi:10.1002/humu.23348. Tool: https://rest.variantvalidator.org (GRCh38; gene2transcripts_v2 and variantvalidator).
48. van der Sluijs PJ, Santen GWE. Predicted loss-of-function variants before Met584 in ARID1B in population cohorts likely reflect reduced penetrance and should be reported diagnostically. Genome Med. 2025;17(1):111. doi:10.1186/s13073-025-01545-2. PMID: 41053857. PMC12502158.
49. MignotC MoutardML RastetterA BoutaudL HeideS BilletteT DoummarD GarelC AfenjarA JacquetteA LacombeD VerloesA Bole-FeysotC NitschkeP MassonC FaudetA LesneF BienvenuT AlbyC Attie-BitachT DepienneC NavaC HeronD ARID1B mutations are the major genetic cause of corpus callosum anomalies in patients with intellectual disability. Brain. 2016 139 e64. PMID: 27474218.
50. Bosch E, Güse E, Kirchner P, Winterpacht A, Walther M, Alders M, et al. The missing link: ARID1B non-truncating variants causing Coffin-Siris syndrome due to protein aggregation. Hum Genet. 2024;143(8):965-978. doi:10.1007/s00439-024-02688-9. PMID: 39028335. PMC11303441.
51. Beaumont RN, Hawkes G, Gunning AC, Wright CF. Clustering of predicted loss-of-function variants in genes linked with monogenic disease can explain incomplete penetrance. Genome Med. 2024;16(1):64. doi:10.1186/s13073-024-01333-4. PMID: 38671509. PMC11046769.
52. van der Sluijs PJ, Gösgens M, Dingemans AJM, Striano P, Riva A, Mignot C, et al. ARID1B-related disorder in 87 adults: Natural history and self-sustainability. Genet Med Open. 2024;2:101873. doi:10.1016/j.gimo.2024.101873. PMID: 39669611. PMC11613905.
53. Wright CF, Beaumont RN. Reply to: 'Predicted loss-of-function variants before Met584 in ARID1B in population cohorts likely reflect reduced penetrance and should be reported diagnostically'. Genome Med. 2025;17(1):114. doi:10.1186/s13073-025-01544-3. PMID: 41057944. PMC12505673.
54. SmithJA HoldenKR FriezMJ JonesJR LyonsMJ A novel familial autosomal dominant mutation in ARID1B causing neurodevelopmental delays, short stature, and dysmorphic features. Am J Med Genet A. 2016 170 3313 8. PMID: 27570168.
55. Tao DY, Niu HH, Zhang JJ, Zhang HQ, Zeng MH, Cheng SQ. Short stature and melanocytic nevi in a girl with ARID1B-related Coffin-Siris syndrome: a case report. BMC Pediatr. 2022;22(1):486. doi:10.1186/s12887-022-03535-4. PMID: 35964110. PMC9375425.
56. Mouskou S, Leka-Emiri S, Korona A, Mastroyanni S, Manolakos E, Papoulidis I, et al. Growth Hormone Deficiency due to p.(Gln467Argfs*64) Mutation in the ARID1B Gene in a Girl with Coffin-Siris Syndrome. Mol Syndromol. 2022;13(5):425-432. doi:10.1159/000522532. PMID: 36588753. PMC9801320.
57. Errichiello E, Mustafa N, Vetro A, Notarangelo LD, de Jonge H, Rinaldi B, et al. SMARCA4 inactivating mutations cause concomitant Coffin-Siris syndrome, microphthalmia and small-cell carcinoma of the ovary hypercalcaemic type. J Pathol. 2017;243(1):9-15. doi:10.1002/path.4926. PMID: 28608987. PMC5601212.
58. Bolund ACS, Vestergaard ET, Marinovskij E, Lildballe DL, Blechingberg J, Herlin MK. Müllerian Duct Aplasia in a Girl With SMARCB1-Related Coffin-Siris Syndrome: A Rare Co-Occurring Anomaly. Clin Genet. 2026;. doi:10.1111/cge.70257. PMID: 42805767.
59. Schuhmann S, Bosch E, Fink A, Schüssler S, Uebe S, Wiesener A, et al. A multi-omics approach to characterize a deep intronic ARID1A deletion in Coffin-Siris syndrome. Eur J Med Genet. 2026;82:105084. doi:10.1016/j.ejmg.2026.105084. PMID: 42173440.
60. Perrino MR, Jongmans MCJ, Tomlinson GE, Greer MC, Scollon SR, Mitchell SG, et al. Update on Cancer and Central Nervous System Tumor Surveillance in Pediatric NF2-, SMARCB1-, and LZTR1-Related Schwannomatosis. Clin Cancer Res. 2025;31(8):1400-1406. doi:10.1158/1078-0432.CCR-24-3278. PMID: 39937237. PMC12002630.
61. Saad R, Cobolli Gigli C, van der Sluijs PJ, Wilson JR, Hsieh TC, McConnell VPM, et al. Multimodal genotype-phenotype analysis in SMARCB1-associated developmental disorders. Genet Med. 2026;28(9):102614. doi:10.1016/j.gim.2026.102614. PMID: 42206491.
62. Gossai N, Biegel JA, Messiaen L, Berry SA, Moertel CL. Report of a patient with a constitutional missense mutation in SMARCB1, Coffin-Siris phenotype, and schwannomatosis. Am J Med Genet A. 2015;167A(12):3186-91. doi:10.1002/ajmg.a.37356. PMID: 26364901.
63. Chesneau B, Willems M, Bouazzaoui A, Lequeux L, Plaisancié J, El Chehadeh S, et al. Expanding the Phenotypic Spectrum Associated With Loss-of-Function SMARCA4 Variants to Eye Developmental Anomalies. Clin Genet. 2026;109(6):1064-1069. doi:10.1111/cge.70143. PMID: 41568967. PMC13167633.
64. Smith MJ, Wallace AJ, Bennett C, Hasselblatt M, Elert-Dobkowska E, Evans LT, et al. Germline SMARCE1 mutations predispose to both spinal and cranial clear cell meningiomas. J Pathol. 2014;234(4):436-40. doi:10.1002/path.4427. PMID: 25143307.
65. Ibrahim AA, Aamer W, Aliyev E, Al-Maraghi A, Elhag SF, Abdi M, et al. Long-read sequencing identifies a novel de novo inversion in SMARCC2 in a pediatric patient with Coffin-siris syndrome 8: a case report. BMC Med Genomics. 2025;18(1):204. doi:10.1186/s12920-025-02269-3. PMID: 41291750. PMC12752245.
66. Angelozzi M, Karvande A, Molin AN, Ritter AL, Leonard JMM, Savatt JM, et al. Consolidation of the clinical and genetic definition of a SOX4-related neurodevelopmental syndrome. J Med Genet. 2022;59(11):1058-1068. doi:10.1136/jmedgenet-2021-108375. PMID: 35232796. PMC9433470.
67. Gao F, Elliott NJ, Ho J, Sharp A, Shokhirev MN, Hargreaves DC. Heterozygous mutations in SMARCA2 reprogram the enhancer landscape by global retargeting of SMARCA4. Mol Cell. 2019;75:891-904.e7. PMID: 31375262.
68. Sekiguchi F, Tsurusaki Y, Okamoto N, Teik KW, Mizuno S, Suzumura H, et al. Genetic abnormalities in a large cohort of Coffin-Siris syndrome patients. J Hum Genet. 2019;64(12):1173-1186. doi:10.1038/s10038-019-0667-4. PMID: 31530938.
69. Sousa SB, Hennekam RC. Phenotype and genotype in Nicolaides-Baraitser syndrome. Am J Med Genet C Semin Med Genet. 2014;166C(3):302-14. doi:10.1002/ajmg.c.31409. PMID: 25169058.
70. Cappuccio G, Sayou C, Tanno PL, Tisserant E, Bruel AL, Kennani SE, et al. De novo SMARCA2 variants clustered outside the helicase domain cause a new recognizable syndrome with intellectual disability and blepharophimosis distinct from Nicolaides-Baraitser syndrome. Genet Med. 2020;22(11):1838-1850. doi:10.1038/s41436-020-0898-y. PMID: 32694869.
71. Jones KL, Jones MC, del Campo M. Smith's Recognizable Patterns of Human Malformation. 8th ed. Philadelphia: Elsevier; 2022. 22q11.2 microdeletion syndrome; p. 386–391 (cited p. 388).
72. SausenM LearRJ JonesS WuJ ReynoldsCP LiuX BlackfordA ParmigianiG DiazLAJr PapadopoulosN VoeglsteinB KinzlerKW VelculescuVE HogartyMD Integrated genomic analyses identify ARID1A and ARID1B alterations in the childhood cancer neuroblastoma. Nat Genet. 2013 45 12 7. PMID: 23202128.
73. ESMO Handbook — Cancer Genomics. Lugano: ESMO Press. Urological cancers — genomics; p. 61–71 (cited p. 65).
74. ESMO Handbook — Cancer Genomics. Lugano: ESMO Press. Brain tumours — gliomas, glioneuronal and medulloblastomas; p. 97–99 (cited p. 99).
75. ESMO Handbook — Cancer Genomics. Lugano: ESMO Press. Rare cancers — sarcoma, GIST, CUP, neuroblastoma; p. 100–106 (cited p. 100).
76. Verstraeten A, Roos-Hesselink J, Loeys B. Bicuspid aortic valve. In: Baars HF, van der Smagt JJ, Doevendans PAFM, van Tintelen JP, editors. Clinical Cardiogenetics. 3rd ed. Cham: Springer; 2020. p. 345–361 (cited p. 351). doi:10.1007/978-3-030-45457-9_20
77. Vasko A, Schrier Vergano SA. Language Impairments in Individuals With Coffin-Siris Syndrome. Front Neurosci. 2021;15:802583. doi:10.3389/fnins.2021.802583. PMID: 35126043. PMC8811135.
78. Fabre A, Aljabali K, Boisgontier J, Fillon L, Guida L, Saitovitch A, et al. Multimodal imaging reveals resilient memory networks in carriers of pathogenic ARID1B variants. Transl Psychiatry. 2026;16(1). doi:10.1038/s41398-026-04090-7. PMID: 42156711. PMC13354570.
79. Ciliberto M, Skjei K, Vasko A, Schrier Vergano S. Epilepsy in Coffin-Siris syndrome: A report from the international CSS registry and review of the literature. Am J Med Genet A. 2023;191(1):22-28. doi:10.1002/ajmg.a.62979. PMID: 36177969.
80. Valentine R, Mahida S, Smith L, LaFortune L, Harini C, Achkar CME, et al. Focal cortical dysplasia: An under-recognized neuroimaging finding in ARID1B-related disorders. Epileptic Disord. 2026;. doi:10.1002/epd2.70369. PMID: 42579510. PMC13485335.
81. McCague EA, Lamichhane R, Holt N, Schrier Vergano SA. Growth charts for individuals with Coffin-Siris syndrome. Am J Med Genet A. 2020;182(10):2253-2262. doi:10.1002/ajmg.a.61823. PMID: 32851773.
82. Chu S, Yuan X, Niu Q, Gu W. A de novo SOX11 mutation causing hypogonadotropic hypogonadism: a case report and literature review. BMC Pediatr. 2026;26(1). doi:10.1186/s12887-026-06974-5. PMID: 42168980. PMC13471292.
83. Kolkiran A, Ataseven Kulalı M, Daşar T, Kablan A, Taşdelen E, Atay G, et al. Expanding the Coffin-Siris syndrome spectrum: genetic, dysmorphic, and endocrine findings in eight cases. Eur J Pediatr. 2026;185(3). doi:10.1007/s00431-026-06830-9. PMID: 41795723.
84. Hennocq Q, Lienhard O, Rao D, Amiel J, Benichou L, Bongibault T, et al. Next Generation Phenotyping and Synthetic Faces in Coffin Siris Syndrome. Clin Genet. 2025;107(5):495-504. doi:10.1111/cge.14682. PMID: 39726094.
85. Schrier Vergano SA. ARID2, a milder cause of Coffin-Siris syndrome? Broadening the phenotype with 17 additional individuals. Am J Med Genet A. 2024;194:e63540. PMID: 38243407.
86. Schmetz A, Lüdecke HJ, Surowy H, Sivalingam S, Bruel AL, Caumes R, et al. Delineation of the adult phenotype of Coffin-Siris syndrome in 35 individuals. Hum Genet. 2024;143(1):71-84. doi:10.1007/s00439-023-02622-5. PMID: 38117302.
87. Ismedon M, Smirnov V, Hamerstehl A, David TTD, Denis D, Ramtohul P, et al. Primary congenital glaucoma in a patient with Coffin-siris syndrome type 1 due to an ARID1B mutation: a novel association. Ophthalmic Genet. 2026;47(5):497-499. doi:10.1080/13816810.2026.2652012. PMID: 41911953.
88. Alyahya YA, Alsirhy EY, Alwadani SF, Al-Anazi TA, Kondkar AA, Azad TA. Bilateral congenital glaucoma in a child with Nicolaides-Baraitser syndrome: a case report. Ann Med Surg (Lond). 2026;88(1):902-904. doi:10.1097/MS9.0000000000004499. PMID: 41497020. PMC12768013.
89. Chen AC, Miller M, Kapamajian M, Khitri M. Delayed diagnosis of glaucoma in Coffin-Siris syndrome. Am J Ophthalmol Case Rep. 2025;39:102396. doi:10.1016/j.ajoc.2025.102396. PMID: 40791652. PMC12336519.
90. Peng Q, Yang Y, Zhang Y, Li S, Zhong B, Luo Q, et al. Impaired IFN-γ-mediated innate and adaptive immunity in Coffin-Siris syndrome type 2: immunological insights from a patient with a recurrent ARID1A mutation. Front Immunol. 2026;17:1803570. doi:10.3389/fimmu.2026.1803570. PMID: 42495632. PMC13391312.
91. Firth HV, Hurst JA. Oxford Desk Reference: Clinical Genetics and Genomics. 2nd ed. Oxford: Oxford University Press; 2017. Structural intracranial anomalies; p. 320–323 (cited p. 320).
92. Zarate YA, Bhoj E, Kaylor J, Li D, Tsurusaki Y, Miyake N, et al. SMARCE1, a rare cause of Coffin-Siris Syndrome: Clinical description of three additional cases. Am J Med Genet A. 2016;170(8):1967-73. doi:10.1002/ajmg.a.37722. PMID: 27264197. PMC5870868.
93. Milone R, Gnazzo M, Stefanutti E, Serafin D, Novelli A. A new missense mutation in DPF2 gene related to Coffin Siris syndrome 7: Description of a mild phenotype expanding DPF2-related clinical spectrum and differential diagnosis among similar syndromes epigenetically determined. Brain Dev. 2020;42:192–8. PMID: 31706665.
94. Wang H. Beyond Neurodevelopmental Delay: BICRA-Related Coffin-Siris Syndrome 12 with Severe Intestinal Dysmotility and Recurrent Pneumothorax. Genes (Basel). 2026;17(1). doi:10.3390/genes17010081. PMID: 41595501. PMC12841204.
95. Peña-Padilla C, Martínez-Ceccopieri DA, García-Hernández EM, Bobadilla-Morales L, Corona-Rivera JR. Prenatal Diagnosis of 6q Terminal Deletion Associated with Coffin-Siris Syndrome: Phenotypic Delineation and Review. Genes (Basel). 2025;16(11). doi:10.3390/genes16111365. PMID: 41300817. PMC12652033.
96. Gofin Y, Zhao X, Gerard A, Scaglia F, Wangler MF, Schrier Vergano SA, et al. Evidence for an association between Coffin-Siris syndrome and congenital diaphragmatic hernia. Am J Med Genet A. 2022;188(9):2718-2723. doi:10.1002/ajmg.a.62889. PMID: 35796094. PMC9378577.
97. Cárcamo B, Masotto B, Baquero-Vaquer A, Ceballos-Saenz D, Zapata-Aldana E. "Cancer in ARID1A-Coffin-Siris syndrome: Review and report of a child with hepatoblastoma". Eur J Med Genet. 2022;65(11):104600. doi:10.1016/j.ejmg.2022.104600. PMID: 36049608.
98. van der Sluijs PJ, Vergano SA, Roeder ER, Jongmans MCJ, Santen GWE. Recommending revised hepatoblastoma surveillance in children with a pathogenic ARID1A variant. Reply to "Cancer in ARID1A-Coffin-Siris syndrome: Review and report of a child with hepatoblastoma" by Cárcamo et al. 2022. Eur J Med Genet. 2023;66(2):104694. doi:10.1016/j.ejmg.2022.104694. PMID: 36592690.
99. Orphanet. Nicolaides-Baraitser syndrome (ORPHA:3051). Available from: https://www.orpha.net/en/disease/detail/3051
100. Foley R, Duignan S, McArdle L, Betts DR, Green A, McMahon CJ. Nicolaides-Baraitser syndrome in a patient with hypertrophic cardiomyopathy and SMARCA2 gene deletion. Cardiol Young. 2022;32:821-3. PMID: 34521483.
101. Brčić L, Barić A, Benzon B, Brekalo M, Gračan S, Kaličanin D, Škrabić V, Zemunik T, Barbalić M, Novak I, Pešutić Pisac V, Punda A, Boraska Perica V. AATF and SMARCA2 are associated with thyroid volume in Hashimoto's thyroiditis patients. Sci Rep. 2020;10:1754. PMID: 32019955.
102. Shinko Y, Okazaki S, Otsuka I, Horai T, Kim S, Tanifuji T, Hishimoto A. Accelerated epigenetic age and shortened telomere length based on DNA methylation in Nicolaides-Baraitser syndrome. Mol Genet Genomic Med. 2022;10:e1876. PMID: 35092358.
103. Richards S, Aziz N, Bale S, Bick D, Das S, Gastier-Foster J, Grody WW, Hegde M, Lyon E, Spector E, Voelkerding K, Rehm HL; ACMG Laboratory Quality Assurance Committee. Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology. Genet Med. 2015;17:405-24. PMID: 25741868.
104. Manickam K, McClain MR, Demmer LA, Biswas S, Kearney HM, Malinowski J, Massingham LJ, Miller D, Yu TW, Hisama FM; ACMG Board of Directors. Exome and genome sequencing for pediatric patients with congenital anomalies or intellectual disability: an evidence-based clinical guideline of the American College of Medical Genetics and Genomics (ACMG). Genet Med. 2021;23:2029-37. PMID: 34211152.
105. Trajkova S, Kerkhof J, Rossi Sebastiano M, Pavinato L, Ferrero E, Giovenino C, Carli D, Di Gregorio E, Marinoni R, Mandrile G, Palermo F, Carestiato S, Cardaropoli S, Pullano V, Rinninella A, Giorgio E, Pippucci T, Dimartino P, Rzasa J, Rooney K, McConkey H, Petlichkovski A, Pasini B, Sukarova-Angelovska E, Campbell CM, Metcalfe K, Jenkinson S, Banka S, Mussa A, Ferrero GB, Sadikovic B, Brusco A. DNA methylation analysis in patients with neurodevelopmental disorders improves variant interpretation and reveals complexity. HGG Adv. 2024;5:100309. PMID: 38751117.
106. Smits DJ, Debuy C, Brooks AS, Schot R, Ferraro F, Rots D, et al. Clinical utility of DNA-methylation signatures in routine diagnostics for neurodevelopmental disorders. Eur J Hum Genet. 2025;33(10):1281-1289. doi:10.1038/s41431-025-01919-5. PMID: 40730689. PMC12480637.
107. Koparir A, Kerkhof J, Rzasa J, Metzger E, Bahena Carbajal P, Kolokotronis K, et al. Chromatinopathies: clinically overlapping disorders, revealing novel variants and their DNA methylation signatures. Clin Epigenetics. 2026;18(1). doi:10.1186/s13148-026-02120-1. PMID: 41957673. PMC13088768.
108. Pan N, Chen S, Cai X, Li J, Yu T, Huang HF, Zhang J, Xu C. Low-level germline mosaicism of a novel SMARCA2 missense variant: expanding the phenotypic spectrum and mode of genetic transmission. Mol Genet Genomic Med. 2021;9:e1763. PMID: 34296532.
109. Jones KL, Jones MC, del Campo M. Smith's Recognizable Patterns of Human Malformation. 8th ed. Philadelphia: Elsevier; 2022. Börjeson-Forssman-Lehmann syndrome; p. 820–821.
110. Firth HV, Hurst JA. Oxford Desk Reference: Clinical Genetics and Genomics. 2nd ed. Oxford: Oxford University Press; 2017. Coarse facial features; p. 106–109.
111. Firth HV, Hurst JA. Oxford Desk Reference: Clinical Genetics and Genomics. 2nd ed. Oxford: Oxford University Press; 2017. Hypertrichosis; p. 178–181.
112. Firth HV, Hurst JA. Oxford Desk Reference: Clinical Genetics and Genomics. 2nd ed. Oxford: Oxford University Press; 2017. Unusual hair, teeth, nails, and skin; p. 330–332.
113. Spranger JW, Brill PW, Hall C, Nishimura G, Superti-Furga A, Unger S. Bone Dysplasias: An Atlas of Genetic Disorders of Skeletal Development. 4th ed. Oxford University Press; 2018. Brachydactyly B; p. 455–456.
114. Ribeiro EM, coord. Síndromes Genéticas: Reconhecer para Tratar. Fortaleza: EdUnichristus; 2021. Síndrome de Langer-Giedion; p. 62–64 (cited p. 63–64).
115. Pascolini G, Valiante M, Bottillo I, Laino L, Fleischer N, Ferraris A, Grammatico P. Striking phenotypic overlap between Nicolaides-Baraitser and Coffin-Siris syndromes in monozygotic twins with ARID1B intragenic deletion. Eur J Med Genet. 2020;63:103739. PMID: 31421289.
116. Turner G, Lower KM, White SM, Delatycki M, Lampe AK, Wright M, et al. The clinical picture of the Börjeson-Forssman-Lehmann syndrome in males and heterozygous females with PHF6 mutations. Clin Genet. 2004;65(3):226-32. doi:10.1111/j.0009-9163.2004.00215.x. PMID: 14756673.
117. Foulkes WD, Kamihara J, Evans DGR, Brugières L, Bourdeaut F, Molenaar JJ, et al. Cancer Surveillance in Gorlin Syndrome and Rhabdoid Tumor Predisposition Syndrome. Clin Cancer Res. 2017;23(12):e62-e67. doi:10.1158/1078-0432.CCR-17-0595. PMID: 28620006. PMC7309678.
118. Celen C, Chuang JC, Luo X, Nijem N, Walker AK, Chen F, et al. Arid1b haploinsufficient mice reveal neuropsychiatric phenotypes and reversible causes of growth impairment. Elife. 2017;6. doi:10.7554/eLife.25730. PMID: 28695822. PMC5515576.
119. Bilha SC, Teodoriu L, Velicescu C, Caba L. Pituitary hypoplasia and growth hormone deficiency in a patient with Coffin-Siris syndrome and severe short stature: case report and literature review. Arch Clin Cases. 2022;9(3):121-125. doi:10.22551/2022.36.0903.10216. PMID: 36176497. PMC9512126.
120. Li Y, Wang Q, Gong C. More than three years' treatment response of recombinant human growth hormone in a patient with Coffin-Siris syndrome 7. Endokrynol Pol. 2024;75(4):450-451. doi:10.5603/ep.100454. PMID: 39279315.
121. Jung EM, Moffat JJ, Liu J, Dravid SM, Gurumurthy CB, Kim WY. Arid1b haploinsufficiency disrupts cortical interneuron development and mouse behavior. Nat Neurosci. 2017;20:1694-707. PMID: 29184203.
122. Ekici B, İnce T, Kamer İ, Turgut A, Altuğ AS, Demir S, et al. Intranasal exosome therapy in Coffin-Siris syndrome: Clinical evaluation of three children. Surg Neurol Int. 2026;17:156. doi:10.25259/SNI_1198_2025. PMID: 41952703. PMC13054311.
123. Simons Searchlight. Online Study of People Who Have Genetic Changes and Features of Autism. ClinicalTrials.gov identifier: NCT01238250. Available from: https://clinicaltrials.gov/study/NCT01238250
124. Coffin-Siris Syndrome Foundation. Engage in research opportunities — International CSS Registry and clinic [Internet]. Available from: https://coffinsiris.org/current-research-registry-and-clinic/
125. Foundation for ARID1B Research (FAR). Families [Internet]. Chevy Chase (MD). Available from: https://arid1b.org/families
126. Ben-Salem S, Sobreira N, Akawi NA, Al-Shamsi AM, John A, Pramathan T, Valle D, Ali BR, Al-Gazali L. Gonadal mosaicism in ARID1B gene causes intellectual disability and dysmorphic features in three siblings. Am J Med Genet A. 2016;170A:156–61. PMID: 26395437.
127. Hanker B, Gillessen-Kaesbach G, Hüning I, Lüdecke HJ, Wieczorek D. Maternal transmission of a mild Coffin-Siris syndrome phenotype caused by a SOX11 missense variant. Eur J Hum Genet. 2022;30:126-32. PMID: 33785884.
128. Cheng SSW, Luk HM, Mok MT, Leung SS, Lo IFM. Genotype and phenotype in 18 Chinese patients with Coffin-Siris syndrome. Am J Med Genet A. 2021;185:2250-61. PMID: 33768696.
129. Liu X, Yang Q, Shi C, Hao H, Xiao X, Li S. [Analysis of clinical characteristics and genetic variant in a child with Nicolaides-Baraitser syndrome due to maternal mosaicism]. Zhonghua Yi Xue Yi Chuan Xue Za Zhi. 2022;39:1366-9. PMID: 36453960.
130. van der Sluijs PJ, Alders M, Dingemans AJM, Parbhoo K, van Bon BW, Dempsey JC, Doherty D, den Dunnen JT, Gerkes EH, Milller IM, Moortgat S, Regier DS, Ruivenkamp CAL, Schmalz B, Smol T, Stuurman KE, Vincent-Delorme C, de Vries BBA, Sadikovic B, Hickey SE, Rosenfeld JA, Maystadt I, Santen GWE. A case series of familial ARID1B variants illustrating variable expression and suggestions to update the ACMG criteria. Genes (Basel). 2021;12:1275. PMID: 34440449.
131. Huang SJ, Amendola LM, Sternen DL. Variation among DNA banking consent forms: points for clinicians to bank on. J Community Genet. 2022;13:389-97. PMID: 35834113.
132. Ge Y, Zhang XY, Han X, Zhang JT, Ma WM, Yang HC, et al. Identification of novel variants in the ARID1B gene causing Coffin-Siris syndrome. Eur J Pediatr. 2026;185(2):86. doi:10.1007/s00431-025-06729-x. PMID: 41545737. PMC12811165.
133. Määttänen L, Hietala M, Ignatius J, Arvio M. A 69-year-old woman with Coffin-Siris syndrome. Am J Med Genet A. 2018;176:1764-7. PMID: 30055038.
134. Gerkes EH, Fock JM, den Dunnen WF, van Belzen MJ, van der Lans CA, Hoving EW, et al. A heritable form of SMARCE1-related meningiomas with important implications for follow-up and family screening. Neurogenetics. 2016;17(2):83-9. doi:10.1007/s10048-015-0472-y. PMID: 26803492. PMC4794526.