Mammalian SWI/SNF (SWitch/Sucrose Non-Fermentable) complexes are ATP-dependent chromatin remodelers. Many SWI/SNF members, such as ARID1A and SMARCA4, have emerged among the most frequently mutated genes in certain diseases, especially cancer. Overall, the SWI/SNF complex is the most mutated chromatin remodeling complex across multiple cancers, highlighting its central role in tumorigenesis.
SWI/SNF (SWitch/Sucrose Non-Fermentable) complexes are ATP-dependent chromatin remodeling complexes that control gene expression by altering the positioning of nucleosomes and recruiting other chromatin binding factors[1][2]. SWI/SNF complexes were discovered in yeast as modulators of mating-type switching (“SWI”) and of growth on sucrose media (“SNF”)[3][4]. Today, advances in proteomics, biochemistry, molecular biology, and structural biology have substantially improved the knowledge of SWI/SNF composition, assembly, and function in mammals, including humans. In the latter, SWI/SNF complexes can be constituted by different combinations of 10-15 subunits out of 29 possible subunits[5]. The specific combinations of subunits that form SWI/SNF complexes in a cell at a given time point depend on the cell type and on the context[6]. Overall, three major SWI/SNF complexes have been identified in mammals: canonical BAF (cBAF), polybromo-associated BAF (PBAF), and noncanonical BAF (ncBAF), each of which is constituted by a combination of common and complex-specific subunits. In all cases, complexes contain one catalytic ATPase subunit, which can be either SMARCA4 (previously known as BRG1) or SMARCA2 (previously known as BRM). In addition, certain groups of SWI/SNF subunits are paralogs that have similar functions so that, when one subunit is missing, a paralog subunit can take its place. However, these changes may alter some of the functions of the complexes[7].
SWI/SNF complexes can exert their functions through various mechanisms. The main mechanism is using the energy from the hydrolysis of ATP to eject nucleosomes and, thus, to reorganize chromatin[8]. However, other mechanisms have been suggested for certain SWI/SNF functions. For example, certain SWI/SNF subunits may interact with other chromatin-modifying enzymes to activate promoters or enhancers[9]. Moreover, SWI/SNF complexes are recruited to DNA damage sites, where they participate in DNA damage response pathways[10][11][12]. Finally, SWI/SNF complexes participate in the 3D organization of chromosomes by interacting with CTCF binding sites, cohesins, lamin, and replication origins[13].
As a consequence of the diversity of SWI/SNF complexes and their genome-wide mechanisms of action, their biological functions are highly heterogeneous, dynamic, and context-dependent. During development, certain SWI/SNF subunits have been associated with embryogenesis and with differentiation into various cell types (reviewed in [14]). In differentiated cells, SWI/SNF complexes keep playing an essential role in the dynamic regulation of gene expression, genome organization, and DNA repair. Accordingly, deficiencies or alterations in SWI/SNF subunits are associated with disease, especially developmental disorders and cancer.
SWI/SNF complexes are essential for development, and specific SWI/SNF complexes guide the differentiation of embryonic stem cells into neurons[15]. Alterations in SWI/SNF subunits can lead to neurodevelopmental disorders (NDDs), which manifest as intellectual disabilities that may or may not be accompanied by other symptoms. For example, heterozygous germline mutations in various SWI/SNF subunits can lead to a rare NDD called Coffin-Siris syndrome (CSS; OMIM #135900). The most commonly mutated subunit in CSS patients is ARID1B, followed by ARID1A, SMARCB1, SMARCA4, SMARCA2, and SMARCE1[16]. Another example is the Nicolaides-Baraitser syndrome (OMIM #601358), which is caused by missense mutations in SMARCA2[17].
During tumor development, cancer cells accumulate somatic mutations in a wide variety of genes. The mutations that contribute to the tumor phenotype are named ‘driver mutations’. While different cancer types tend to accumulate mutations in different genes, some genes or groups of genes are recurrently mutated across multiple cancer types. The latter is the case for the SWI/SNF complex, which is mutated in >25% of human cancers, and as such it is the most recurrently mutated chromatin remodeling complex across all cancer types[18][19]. In particular, SWI/SNF subunits such as ARID1A and SMARCA4 are recurrently identified as pan-cancer drivers in next-generation sequencing studies in thousands of patients[20][21]. However, other SWI/SNF subunits are preferentially mutated in specific tumors. Mutations in SWI/SNF subunits are usually loss-of-function and they often involve single nucleotide substitutions or small insertions or deletions. However, certain SWI/SNF subunits can also be affected by large-scale deletions or translocations.
The first evidence of an association between SWI/SNF and cancer was discovered in rhabdoid tumors, which are rare pediatric tumors that are almost exclusively mutated (by point mutations, small indels, or large deletions) in the SWI/SNF subunit SMARCB1[22]. Later studies revealed other SWI/SNF subunits that are recurrently mutated in other cancer types. In lung adenocarcinoma (LUAD), we discovered that SMARCA4 is recurrently mutated in more than 10% of the patients (manuscript under submission) and 40% of cell lines[23]. In renal cell clear cell carcinoma, up to 41% of the cases harbor mutations or loss of PBRM1[24]. In diffuse large B-cell lymphoma and other B cell hematological malignancies, BCL7A is recurrently mutated, especially at its first splice donor site[25]. Overall, the diversity and recurrence of mutations in SWI/SNF subunits in cancer reflect the central role of SWI/SNF complexes in cells as well as their tissue-specific or context-dependent composition and function.
Chromosomal translocations are large-scale rearrangements that frequently occur in certain cancers. Translocations can have various effects, such as the fusion of two or more distant protein-coding genes, leading to an aberrant fusion protein, or the fusion of a promoter and a cancer gene, resulting in an aberrant control of the transcription of the gene. Translocations that affect SWI/SNF genes are highly recurrent in certain cancers. For example, a hallmark of synovial sarcoma is a t(X;18) translocation that involves the SS18 gene and which creates a SS18-SSX fusion protein[26]. In addition, certain aggressive chronic lymphocytic leukemias harbor a t(2;14)(p13;q32.3) translocation that results in the BCL11A locus being under transcriptional control of regulatory elements from the IGH locus, leading to aberrant overexpression of BCL11A[27].
SWI/SNF subunits can be aberrantly expressed in cancer in the absence of mutations in their genes. Recently, we reported that the mRNAs of all SWI/SNF subunits are recurrently downregulated in LUAD compared to matched normal lung tissue, and only a small proportion of the cases of downregulation could be explained by the presence of mutations in SWI/SNF genes (manuscript under submission). Similar observations were made for the protein levels of SMARCA4 and SMARCA2 measured by immunohistochemistry[28]. There may be multiple causes for the downregulation of SWI/SNF subunits in LUAD, including promoter hypermethylation and regulation by non-coding RNAs. For example, SMARCA2 is downregulated in some LUAD samples as a consequence of promoter hypermethylation[29]. In squamous cell carcinomas of various tissues, ARID1A is recurrently downregulated as a consequence of promoter hypermethylation[30]. In primary cutaneous T cell lymphoma, BCL7A is often epigenetically silenced by promoter hypermethylation[31]. In rhabdoid tumors and small cell carcinoma of the ovary, hypercalcemic type (SCCOHT), SMARCA2 is either epigenetically silenced or transcriptionally inactive[32].
Regarding the regulation of the expression of SWI/SNF subunits by non-coding RNAs, we reported that hsa-miR-155 modulates SMARCA4 in LUAD[33]. In addition, in various tumor types, SMARCA2 expression is modulated by hsa-miR-199a-5p and -3p[34]. In colorectal carcinoma, SMARCC1 has an oncogenic role and it is a direct target of hsa-miR-202-5p, a tumor suppressor microRNA[35]. In turn, the SWI/SNF complex can also modulate the expression of cancer-related microRNAs. In lung adenocarcinoma, SMARCA4-containing SWI/SNF complexes control the expression of hsa-miR-222, which contributes to a tumor suppressor phenotype (manuscript under submission).
Importantly, changes in the expression of certain SWI/SNF subunits are associated with patient survival in certain tumor types. For example, in squamous cell carcinomas and cervical cancer, low expression of ARID1A is associated with poor prognosis[36]. In hepatocellular carcinoma, low expression of SMARCA2 is associated with poor survival[37]. However, in other cases, low expression of SWI/SNF subunits may be associated with a better prognosis. This is the case of SMARCA4 and ARID1A in breast cancer and bladder cancer[38][39]. Finally, high SMARCA4 expression is associated with unfavorable prognosis in hepatocellular carcinoma[40].
SWI/SNF alterations in tumors are usually loss-of-function. Because of the functional redundancy of certain SWI/SNF subunits, when one subunit is inactivated, it can often be partially or totally replaced by a paralog subunit. Therefore, the functional effect of the inactivation of SWI/SNF subunits is usually not a complete loss of function of the complex, but an aberrant activity as a result of alternative residual SWI/SNF complexes[41]. This situation can lead to synthetic lethalities: if SWI/SNF subunits “X” and “Y” are partly or completely interchangeable, and a tumor has inactivated “X”, then pharmacological inhibition of “Y” can lead to tumor regression. On the other hand, in cells that are wild type for “X”, inhibition of “Y” should not have such a strong functional effect because “X” can replace the function of “Y”. Therefore, synthetic lethalities open the possibility of tumor-specific therapies. Synthetic lethalities have also been observed between pairs of non-paralog SWI/SNF subunits. For example, SMARCB1-mutant cell lines have an increased dependence on the non-paralogous subunit BRD9[42]. In any case, because SWI/SNF aberrations are highly context-dependent, further studies are required to prove that cancers that lack one subunit are dependent on its paralog and that they have not evolved to become paralog-independent. However, even if a tumor becomes paralog-independent, other functional dependencies may arise. For example, lung and ovarian cancers lacking both ATPase subunits of the SWI/SNF become sensitive to bromodomain inhibitors[43].
Synthetic lethalities can also arise between SWI/SNF and non-SWI/SNF genes. For example, certain SWI/SNF mutant cancers are sensitive to inhibition of EZH2, the catalytic subunit of the polycomb repressive complex 2 (PRC2)[44]. Further dependencies can be assessed using drug screens coupled with functional genomic depletion screens. One of such screens was performed in rhabdoid tumors, SCCOHT, and non-small cell lung carcinoma cell lines, and it identified that depletion of SWI/SNF genes leads to sensitivity to inhibitors of cell cycle regulators such as cyclin-dependent kinase 4 or 6 (CDK4/6) and Aurora A[45][46][47][48]. In another screen, rhabdoid tumors and ARID1A-mutant ovarian clear cell carcinomas were found to be dependent on several receptor tyrosine kinases, including platelet-derived growth factor receptors, FGFR1, and MET[49][50][51]. Finally, loss of PBRM1 expression may improve the response to immune checkpoint inhibitors because of transcriptional changes in genes from immune signaling pathways, such as JAK-STAT[52].
In conclusion, SWI/SNF subunits are especially attractive as therapeutic targets in the context of synthetic lethalities. Finding the correct inhibitors for specific contexts and tumor types may lead to highly specific and effective cancer therapies.