SMARCA4 inactivating mutations cause concomitant Coffin–Siris syndrome, microphthalmia and small‐cell carcinoma of the ovary hypercalcaemic type

Abstract SMARCA4 chromatin remodelling factor is mutated in 11% of Coffin–Siris syndrome (CSS) patients and in almost all small‐cell carcinoma of the ovary hypercalcaemic type (SCCOHT) tumours. Missense mutations with gain‐of‐function or dominant‐negative effects are associated with CSS, whereas inactivating mutations, leading to loss of SMARCA4 expression, have been exclusively found in SCCOHT. We applied whole‐exome sequencing to study a 15‐year‐old patient with mild CSS who concomitantly developed SCCOHT at age 13 years. Interestingly, our patient also showed congenital microphthalmia, which has never previously been reported in CSS patients. We detected a de novo germline heterozygous nonsense mutation in exon 19 of SMARCA4 (c.2935C > T;p.Arg979*), and a somatic frameshift mutation in exon 6 (c.1236_1236delC;p.Gln413Argfs*88), causing complete loss of SMARCA4 immunostaining in the tumour. The immunohistochemical findings are supported by the observation that the c.2935C > T mutant transcript was detected by reverse transcription polymerase chain reaction at a much lower level than the wild‐type allele in whole blood and the lymphoblastoid cell line of the proband, confirming nonsense‐mediated mRNA decay. Accordingly, immunoblotting demonstrated that there was approximately half the amount of SMARCA4 protein in the proband's cells as in controls. This study suggests that SMARCA4 constitutional mutations associated with CSS are not necessarily non‐truncating, and that haploinsufficiency may explain milder CSS phenotypes, as previously reported for haploinsufficient ARID1B. In addition, our case supports the dual role of chromatin remodellers in developmental disorders and cancer, as well as the involvement of SMARCA4 in microphthalmia, confirming previous findings in mouse models and the DECIPHER database. Finally, we speculate that mild CSS might be under‐recognized in a proportion of SCCOHT patients harbouring SMARCA4 mutations. © 2017 The Authors. The Journal of Pathology published by John Wiley & Sons Ltd on behalf of Pathological Society of Great Britain and Ireland.


Introduction
Coffin-Siris syndrome (CSS) (OMIM 135900) is a rare and clinically heterogeneous congenital disorder caused by germline mutations in different subunits of the ATP-dependent SWI/SNF (SWItch/sucrose non-fermentable) chromatin remodelling complex involved in transcription, lineage specification, and maintenance of stem cell pluripotency. SMARCA4/BRG1, one of the catalytic subunits of the complex, is mutated in a proportion (∼11%) of CSS patients [1][2][3][4][5][6]. In addition, SMARCA4 mutations are associated with small-cell carcinoma of the ovary hypercalcaemic type (SCCOHT), a malignant undifferentiated tumour of the ovary, characterized by a dismal prognosis and poor response to chemotherapy [7]. Because of the high frequency of mutation of the SWI/SNF chromatin remodelling complex in all cancer types, an increased cancer risk in CSS patients has been suggeested, but this has never definitively proven, as 10 E Errichiello et al malignancies have been only very rarely reported in these patients [8].

Materials and methods
We studied a 15-year-old female patient of Italian descent, and her healthy relatives, showing peculiar features of CSS as compared with previously reported cases, such as developmental delay, distinctive facial dysmorphisms (e.g. coarse facies, irregular dentition, and abnormal ears), remarkable hirsutism/hypertrichosis, sparse scalp hair, hearing impairment, and hypoplasia of the distal toenails (supplementary material, Figure S1 and Table S1). At age 13 years, she developed SCCOHT (stage IIIC) with severe hypercalcaemia (17.5 mg/dl). The patient was diagnosed by her attending oncologist and clinical geneticist/dysmorphologist. The study was performed in accordance with the Code of Ethics of the University of Pavia, and parental consent was obtained for publication of images of the patient.

Germline analysis
Conventional karyotyping on peripheral blood metaphases revealed a normal female chromosomal pattern (46,XX), and array comparative genomic hybridization (CGH) and EXCAVATOR did not identify any pathogenic constitutional copy number variation (CNV) (supplementary material, Figure  S2). The whole-exome sequencing (WES) analysis revealed a germline heterozygous nonsense mutation (chr19:11,134,269C > T, GRCh37/hg19; NM_001128849.1:c.2935C > T; NP_001122321.1:p. Arg979*) in exon 19 of SMARCA4, which truncated the SMARCA4 protein upstream of both the highly evolutionarily conserved helicase C-terminal domain (amino acids 1084-1246) and the bromodomain (amino acids 1477-1547). Parental DNA examination revealed that the variant arose de novo in the proposita and was absent in the DNA of the 18-year-old healthy sister ( Figure 1A-C). The c.2935C > T substitution, which is not reported in any of the dbSNP, ExAC, NHLBI ESP and 1000 Genome Project databases, caused the interruption of the reading frame by a premature stop codon, suggesting either a severely truncated translation product or nonsense-mediated mRNA decay as possible consequences. Interestingly, this variant was previously identified by Ramos et al. [9] in the germline DNA of a 9-year-old European patient with SCCOHT (stage IA) and hypercalcaemia, but without any reported extratumoural manifestation, whose tumour was not available for investigation of a second somatic hit.

Somatic analysis
Array CGH and EXCAVATOR analyses excluded the presence of copy number loss in the 19p13.2 region surrounding the SMARCA4 locus in the formalin-fixed paraffin-embedded (FFPE) tumour sample as well as any other causative chromosomal aberration elsewhere, thus confirming the high genomic stability of SCCOHT (supplementary material, Figure S2). By WES, we detected a second somatic frameshift mutation in exon 6 of SMARCA4 (chr19:11,100,110delC, GRCh37/hg19; NM_001128849.1:c.1236_1236delC; NP_001122321.1:p.Gln413Argfs*88) that caused ablation of the downstream helicase ATP-binding (amino acids 766-931) and C-terminal (amino acids 1084-1246) domains, along with the last bromodomain (amino acids 1477-1547) ( Figure 1B, C). This 1-bp deletion, which is not recorded in any database of somatic variants (e.g. COSMIC and ICGC) but has been previously reported by Witkowski et al. [7], is expected to trigger nonsense-mediated decay of the corresponding mRNA transcript, as in the case of the germline mutation. Interestingly, we did not detect any additional pathogenic variant in the SWI/SNF genes or other tumour-associated genes (supplementary material, Table S2), confirming mutual exclusion of SWI/SNF mutations as well as the peculiar genomic stability of SCCOHT lesions also at the nucleotide level.
The localization of our two variants across conserved amino acids is in line with previous findings, SMARCA4 mutations being widely spread along the entire length of the protein in SCCOHT (with the sole exception of the C-terminal bromodomain), and being clustered within the three central domains (helicase/SANT-associated domain, helicase ATP-binding domain, and helicase C-terminal domain) in CSS ( Figure 2; supplementary material, Table S3).

Functional validation
The c.2935C > T mutant transcript was detected by reverse transcription polymerase chain reaction (RT-PCR) at a much lower level than the wild-type allele in the patient's whole blood and B-lymphoblastoid cell line (B-LCL). Moreover, cycloheximide (CHX) treatment confirmed that the germline c.2935C > T nonsense mutation resulted in an unstable transcript that was substantially subject to nonsense-mediated decay ( Figure 3). Accordingly, immunoblotting under basal conditions or after CHX stabilizing treatment demonstrated the presence of only the wild-type full-length form of the protein (1647 amino acids, ∼184.68 kDa), and conversely, the absence of a possible truncated polypeptide (expected to be composed of 979 amino acids, ∼107.78 kDa) that may theoretically exert a dominant-negative effect ( Figure 4A; supplementary material, Figure S3). The proband's B-LCL showed half as much protein as the father's B-LCL and the SK-OV-3 control cell line, only one allele being translated in the patient ( Figure 4B).
In addition, the immunohistochemical analysis of FFPE tumour tissue revealed tumour-specific complete loss of SMARCA4 nuclear staining, in keeping with the presence of biallelic inactivating mutations in SMARCA4. Notably, as the primary antibody recognizes an N-terminal epitope located upstream of both p.Arg979* and p.Gln413Argfs*88 truncating mutations (amino acids 200-300), the complete absence of detectable protein expression is consistent with nonsense-mediated decay of both germline and somatic mutant transcripts ( Figure 4C).

SMARCA4 and microphthalmia
Ophthalmological anomalies have been reported in up to 62% of CSS patients, consisting of severe myopia (up to −18D) or, less often, strabismus, astigmatism, spherophakia, amblyopia, and optic disk coloboma [3]. We observed congenital microphthalmia in the proband, a symptom that has never been reported before. Interestingly, it has been demonstrated that Brg1/Smarca4 regulates both terminal differentiation of lens fibre cells and organized degradation of their nuclei (denucleation) in mouse embryos [12], and a Brg1 nonsense mutation (Y390X) abrogated retinal cell differentiation in zebrafish [13]. Moreover, N-ethyl-N-nitrosourea-mutagenized C57BL/6 J mice carrying a mutation in the helicase ATP-binding domain of Smarca4 (NM_001174078.1:c.2381C > T; NP_001167549.1:p.Thr794Ile) showed multiple physical defects, including anophthalmia (MGI:5313991), and a de novo germline heterozygous deletion of 1.23 Mb spanning SMARCA4 (chr19:10,640,379-11,873,382, GRCh37/hg19) has been reported in a female patient with microphthalmos (DECIPHER #250826) (supplementary material, Figures S4 and S5). Surprisingly, this patients showed other clinical features (such as intellectual disability, short toes, recurrent infections, ear abnormalities, and other facial dysmorphisms) highly suggestive of CSS. The deleted region also includes DOCK6, which is associated with Adams-Oliver syndrome 2 (OMIM 614219). Although, in rare cases, this condition may include ocular anomalies, such as microphthalmia, retinal detachment, and visual impairment, it is caused by homozygous or, less frequently, compound heterozygous mutations. Importantly, a focused next-generation sequencing data analysis considering all known microphthalmia-associated genes failed to identify additional causative mutations   in our patient (supplementary material, Table S4 and Figure S6). Taken together, these findings suggest that microphthalmia might be considered to be the result of SMARCA4 deregulation.

Discussion
We have identified, for the first time, a SMARCA4 lossof-function mutation, resulting in haploinsufficiency, in a patient with CSS and concomitant early development of SCCOHT (supplementary material, Table S5). There is growing evidence that chromatin remodelling factors may simultaneously cause developmental disorders (with various amounts of intellectual disability) and cancer [14]. For instance, frameshift mutations and loss of expression of chromodomain helicase DNA-binding (CHD) genes have been commonly found in gastric and colorectal cancers [15], and more recently in small-cell lung cancer [16]. Importantly, loss-of-function mutations of CHD7 are also associated with CHARGE syndrome (OMIM 214800), which is characterized, as in the case of CSS, by multiple congenital anomalies, including microphthalmia. Recently, de novo missense substitutions in the chromatin remodeller gene CHD4 have been associated with Sifrim-Hitz-Weiss syndrome (OMIM 617159), and mutations in the same positions have been reported in malignant tumours [17], thus supporting the hypothesis of common alterations shared between intellectual disability syndromes with distinctive dysmorphisms and cancer. Similarly to what has been reported for ARID1B-haploinsufficient individuals, the CSS phenotype in our patient was milder than in all previously reported cases with missense substitutions, which may exert a dominant-negative effect on other complex proteins. According to this hypothesis, individuals with SMARCA4 truncating mutations and deletions might share less severe clinical features: in fact, DECIPHER case #303790 with an approximately 247-kb microdeletion surrounding SMARCA4 showed mild neurodevelopmental and facial anomalies. Moreover, although further studies are needed, this report broadens the phenotypic spectrum of CSS by including microphthalmia among the clinical features.
In conclusion, our and previous findings suggest that clinical follow-up should be undertaken in CSS patients to monitor potential neoplasm development and, conversely, that careful investigation of CSS features in patients with SCCOHT should be strongly recommended. In fact, we tentatively speculate that mild CSS might be under-recognized in a proportion of SMARCA4-positive SCCOHT patients. Furthermore, patients with biallelic somatic mutations but without germline assessment, especially those with an early age of onset, might also harbour constitutional monoallelic alterations, underlying an undiagnosed CSS.