SWI/SNF complex in disorder: SWItching from malignancies to intellectual disability.

We reasoned that examining the mutational spectrum of patients with intellectual disability and the mutations related to carcinogenesis may shed light on whether the patients with intellectual disability might be predisposed to tumor development. As shown in Table 2 for ARID1A , SMARCB1 and possibly ARID1B, the same type of mutations causing intellectual disability syndromes in the germline has also been observed as somatic mutations in tumors. Therefore, an increased risk of malignancies in CSS-patients with mutations in these genes should be considered a possibility. Unfortunately, since SWI/SNF components are expressed in almost every tissue ( Table 2 ), it seems impossible to predict what type of malignancies might be expected. For SMARCA4 the mutational spectrum differs between the intellectual disability and tumor predisposition syndrome cases. For intellectual disability syndromes, missense mutations specific to the ATPase domain are seen, whereas haploinsufficiency causes a tumor predisposition syndrome. However, as we have speculated above, one could easily conceive that missense mutations in the ATP binding domain observed in CSS-patients might also increase the probability of malignancies as they probably impair the function of the SWI/SNF complex. Therefore, with the data available to us, we cannot rule out an increased susceptibility to cancer in CSS patients carrying SMARCA4 mutations. The same applies to a lesser extent for SMARCE1 , since two truncating somatic mutations were identified in breast tumor tissue of a single patient whereas a germline missense mutation was identified as a cause in CSS. Lastly, as no mutations in SMARCA2 have been described in malignancies, we can tentatively infer that patients with SMARCA2 mutations might not have an increased tumor risk. Contrary to what might be expected, there have been no reports of increased susceptibility to malignancies in CSS or NBS so far. However, as these patients are often not followed up after their childhood, we cannot exclude that some of these patients might have developed cancer. An alternative explanation is that the tumor risk is not significantly increased, e.g., because a plethora of mutations elsewhere in the genome contribute to tumor development in the reported tumors with SWI/SNF mutations.

In 2011, mutations in several genes coding for key components of the SWI/SNF complex were implicated in intellectual disability syndromes. First, haploinsufficiency of ARID1B was implicated in non-syndromic intellectual disability, with the common features being severe intellectual disability, speech delay and in many cases agenesis of the corpus callosum. The first reports were of patients with translocations through ARID1B or deletions of ARID1B and several other genes. 14 , 15 In March 2012, Hoyer et al. 16 published a study in which they identified 8 truncating mutations in ARID1B in an unselected group of 887 patients with intellectual disability (0.9%). Again, the phenotype of these patients appeared to be non-syndromic, with moderate to severe intellectual disability, speech delay and hypotonia as the main overlapping features. In this paper, MRI scanning did not identify agenesis of the corpus callosum in these patients. The 2012 April issue of Nature Genetics included three papers on germline mutations in SWI/SNF complex proteins leading to intellectual disability syndromes: our group identified de novo truncating mutations in ARID1B in three patients with Coffin Siris syndrome (CSS, MIM 135900), 17 Tsurusaki and colleagues 18 reported mutations in ARID1B, ARID1A, SMARCA4, SMARCE1 and SMARCB1 in 19/22 (86%, excluding one patient as explained below) of patients with CSS. Van Houdt and colleagues 19 reported mutations in SMARCA2 in patients with Nicolaides-Baraitser syndrome (NBS, MIM 601358), which was later confirmed by Wolff et al. 20 Tsurusaki and colleagues also reported a patient with a SMARCA2 mutation and CSS; however, the phenotype of this patient resembles NBS more than CSS (N. Matsumoto, personal communication). CSS is characterized by severe intellectual disability, often accompanied by speech delay, as well as hypertrichosis and hypoplastic or absent fifth fingernails or toenails. Although mutations in ARID1B can cause CSS, both our data and the literature seem to indicate that not all patients with ARID1B mutations have the specific characteristics of CSS. Indeed, if the percentages of mutations in the different subunits (about 30% in ARID1B ) 18 is the same in CSS as in non-syndromic intellectual disability, one could hypothesize that up to 3% of unexplained intellectual disability is caused by mutations in genes encoding SWI/SNF components (as 0.9% of unexplained intellectual disability seems to be caused by ARID1B mutations). 16 Interestingly, mutations in SMARCA2 seem to cause a slightly different phenotypic spectrum. NBS has some overlap with CSS, but differences are the observed frequency of early-onset seizures in NBS, digital anomalies and the typical sparse hair, although this has also been reported in CSS patients.

At first glance, there is no obvious link between tumor suppression and intellectual disability. However, the involvement of the SWI/SNF complex in tissue differentiation may provide a clue. Problems in tissue differentiation and neuronal differentiation are a possible cause of intellectual disability. Likewise, it does not require much imagination to assume that erroneous tissue differentiation may lead to aberrant growth and, therefore, tumor formation. Lessard et al. 33 elegantly showed that the change in subunit composition of the SWI/SNF complex is a required event in the differentiation of neural stem and progenitor cells. This indicates that subunit alteration of the SWI/SNF complex could play an important role in tissue differentiation in general. The involvement of epigenetic factors in the origin of both tumorigenesis and intellectual disability was recently also described by Gilissen et al. and does not seem to be unique to SWI/SNF. They noted that mutations in epigenetic modifiers seem to have a dual role in human disease. 34 For example, somatic mutations in NSD1 (a histone methyltransferase that has been shown to be involved in transcriptional regulation) and ASXL1 (involved in transcriptional regulation mediated by ligand-bound nuclear hormone receptors) have been identified in hematological malignancies, whereas germline mutations cause Sotos syndrome and Bohring-Opitz syndrome, respectively. 35 , 36 Recently, germline mutations in EZH2 , the catalytic component of the histone methyltransferase polycomb repressive complex 2 (PRC2) were identified in Weaver syndrome, 37 , 38 a syndrome which has clinical overlap with Sotos syndrome. There is evidence that EZH2 may function both as oncogene and tumor suppressor gene. 39 Similar to what we noted for SMARCB1, some of the germline mutations identified in Weaver syndrome are the same as the somatic mutations described in tumor cells. Interestingly, although tumors have been described in Weaver syndrome, they are not as common a feature as might be expected based on the similarity in mutations.

Although it is too early to tell if a phenotypic distinction may be made among the CSS patients with mutations in different components of SWI/SNF, it is apparent that the type of mutation identified depends on the specific component. In cases involving ARID1B, haploinsufficiency seems to be the pathophysiological mechanism. Only nonsense and frame shift mutations have been identified thus far, as well as whole-gene deletions. ARID1B was recently linked to cancer, as truncating mutations (and one small in-frame deletion) in ARID1B were identified in breast cancer tissue and designated as driver mutations. 21 Likewise, only truncating mutations have been described in ARID1A , pointing to haploinsufficiency as the pathophysiological mechanism. Zhang et al. 22 recently showed that inactivating mutations in ARID1A are a common feature of gastric cancer. Similarly, others previously identified (mostly truncating) somatic mutations in endometriosis-associated ovarian clear cell and endometrioid carcinomas. 23 As the ARID1A and ARID1B proteins are mutually exclusive in SWI/SNF complexes, 24 it seems that a distortion of the balance of these proteins may be the driver of further pathophysiological processes. Flores-Alcantar et al. 25 show that this ratio is dynamic and tissue specific, suggesting that the balance between ARID1A and ARID1B might be important for tissue differentiation. The mutations found in subunits other than ARID1A and ARID1B are highly conserved missense mutations. In the case of SMARCA2 , all missense mutations identified thus far are in the SNF2_N ATPase and HELICASE_C domain of the protein. For SMARCA4 the same holds, and in addition a three-base pair deletion located 13 aminoacids outside of the small helicase/SANT-associated (HAS) domain (p.Lys546del) has been reported. 18 Interestingly, as van Houdt et al. 19 pointed out, haploinsufficiency of SMARCA2 does not result in an NBS phenotype. Larger deletions including SMARCA2 as well as many of its neighboring genes have been documented ( http://decipher.sanger.ac.uk ), but to our knowledge no patient with isolated SMARCA2 haploinsufficiency has been described. Likewise, SMARCA4 haploinsufficiency does not result in CSS, but in rhabdoid tumor predisposition syndrome (MIM 613325). 9 , 26 At this point, we can only speculate about the reasons why missense mutations cause severe intellectual disability and haploinsufficiency gives rise to an increased tumor risk. Truncating mutations may result in insufficient amounts of functional SWI/SNF complexes, thereby reducing the ability to control differentiation and cell cycle processes (and thus lead to tumorigenesis). A possible explanation for the more severe phenotype of the missense mutations is that these mutations result in a gain of function, switching the SMARCA2/4 proteins to a constantly “on” state. However, it seems unlikely that so many different missense mutations as well as in-frame deletions could cause the same gain-of-function effect. A more likely scenario is that the missense phenotype results from defective SMARCA2/4 protein being incorporated into the SWI/SNF complex, thereby blocking binding sites for functional SWI/SNF complexes (competitive inhibition). Thus, it is conceivable that missense mutations will also result in an increased risk of malignancies, as the ability of the SWI/SNF complex to control cell cycle processes may be diminished. Recently, missense mutations in the helicase domain of SMARCA4 were identified in medulloblastoma, lending support to this hypothesis. 27 The missense mutations in SMARCB1 causing CSS have until now only been described in its last two exons (8 and 9). Only a SNF5 domain (unique to SMARCB1), which was shown to be important for promoter targeting and chromatin remodeling, is reported by Uniprot. 28 29 Interestingly, the p.Arg377His mutation that occurred de novo in one of the patients with CSS 18 was also identified as somatic mutations in four meningiomas. 30 Truncating as well as missense mutations in SMARCB1 in the germline have also been described in patients diagnosed with Schwannomatosis, a tumor suppressor syndrome. 10 The last subunit in which a mutation was identified in CSS patients is SMARCE1. In one patient a de novo missense mutation was identified in the high mobility group (HMG) domain of this protein. 18 The HMG domain has a high affinity for alternative DNA-conformations and is often found in subunits of chromatin remodeling complexes. 31 SMARCE1 mutations have not been unequivocally linked to tumor formation, although two truncating mutations which are thought to lead to haploinsufficiency were reported in the breast cancer tissue of 1/95 patients. 32 However, to our knowledge, this latter finding has not yet been replicated.

The literature on the SWI/SNF complex is expanding rapidly, as can be seen by the number of citations in PubMed [76 in 2012 (queried on August 29th) and 105 in 2011, compared with 39 in 2000]. Consequently, we expect that we will learn much more about the function of the SWI/SNF complex in the coming years. However, whereas in past years the focus has been on elucidating the mechanisms that lead to an increased risk of malignancies, we expect this shift to also include the investigation of possible pathophysiological mechanisms leading to intellectual disability syndromes. Furthermore, for (the parents of) patients with germline mutations in SWI/SNF components it is important to know if they have an increased susceptibility to malignancies. Future studies may shed some light on this issue by investigating how often malignancies are observed in these patients. To this end, it will be important to keep following up on the patients with germline mutations and record any malignancy that is detected. Since inactivating mutations in SWI/SNF have been shown to lead to DNA repair defects and impair cell survival in response to genotoxic insult, 1 we may expect tumor sensitivity after local radiation therapy (which kills cancer cells by damaging their DNA) in SWI/SNF patients. However, whole-body exposure of these patients to, for example, DNA-damaging chemotherapies should probably be avoided until more research has been done on this subject. Hopefully, the following years will provide us with enough information to be able to inform the patients and their parents on prognosis and better treatment of SWI/SNF-related malignancies.

The SWI/SNF (switch/sucrose nonfermenting) chromatin remodeling complex plays an important role in several distinct processes, such as transcription, cell differentiation and DNA repair. Moreover, it has also been implicated in tumor suppression and intellectual disability. 1 , 2 Here, we will summarize what is known about the function of the SWI/SNF complex, its role in tumor formation and its involvement in intellectual disability, prior to discussing the comparison of the mutational spectrum between intellectual disability and tumor development.