Mechanisms by which SMARCB1 loss drives rhabdoid tumor growth

Abstract

SMARCB1 (INI1/SNF5/BAF47), a core subunit of the SWI/SNF (BAF) chromatin-remodeling complex, is inactivated in the large majority of rhabdoid tumors and germline heterozygous SMARCB1 mutations form the basis for rhabdoid predisposition syndrome. Mouse models validated Smarcb1 as a bona fide tumor suppressor as Smarcb1 inactivation in mice results in 100% of the animals rapidly developing cancer. SMARCB1 was the first subunit of the SWI/SNF complex found mutated in cancer. More recently, at least seven other genes encoding SWI/SNF subunits have been identified as recurrently mutated in cancer. Collectively, 20% of all human cancers contain a SWI/SNF mutation. Consequently, investigation of the mechanisms by which SMARCB1 mutation causes cancer has relevance not only for rhabdoid tumors, but also potentially for the wide variety of SWI/NSNF mutant cancers. Here we discuss normal functions of SMARCB1 and the SWI/SNF complex as well as mechanistic and potentially therapeutic insights that have emerged.

Eukaryotes carry genetic information in a highly compacted organizational structure termed chromatin. The basic unit of chromatin consists of DNA wrapped around histones, collectively termed nucleosomes. Nucleosomes undergo progressive coiling to generate compact higher order structures. In order for cells to carry out processes such as transcriptional regulation, DNA repair, and DNA replication, chromatin and nucleosomal structures must be remodeled to generate accessible DNA. Three mechanisms involved in this process are covalent histone modification, incorporation of histone variants, and energy dependent remodeling of chromatin structure. With respect to the last category, there are at least five families of chromatin remodeling complexes; SWI/SNF, ISI, NuRD/Mi-2/CHD, INO80, and SWR1. Of these, the SWI/SNF complex, which is evolutionarily conserved from yeast to humans, represents a novel link between ATP-dependent chromatin remodeling and tumor suppression. The complex consists of 10–15 subunits, including core subunits present in all variants of the complex, SMARCB1 (SNF5), SMARCC1 (BAF155), SMARCC2 (BAF170), and two mutually exclusive ATPase subunits, SMARCA4 (also known as BRG1) or SMARCA2 (also known as BRM). Additionally, the complex contains a number of related lineage-restricted subunits that vary by cell type (1). For example the BAF45 subunit is encoded by four different, but highly related genes, BAF45A, BAF45B, BAF45C, and BAF45D. Similarly, the BAF60 (SMARCD) subunit is encoded by three different genes and the BAF53 (ACTL6) by two genes. In addition, groups of subunits can be exchanged for others. For instance, BAF variants of the complex can contain ARID1A or ARID1B, while in the PBAF variants of the complex these are replaced by ARID2 (BAF200), PBRM1 (BAF180), and BRD7. Consequently, while core remodeling complex subunits are ubiquitous, up to several hundred variants of the complex have been proposed that may contribute in a highly specific way to the regulation of lineage identity (2–5).

Nomenclature

First, a word on the complex, and at times confusing, nomenclature for the gene mutated in rhabdoid tumor. Four different names are in use for this gene, with use of the name varying by field of investigation. Those who came to know of the gene as a consequence of its original discovery in yeast often use the name “SNF5”, a reference to its discovery in a screen for genes required to metabolize sucrose. Such genes were called “Sucrose Non-Fermenting” or SNF genes. Those who discovered and learned of the human homolog from it’s interaction with the integrase of HIV typically refer to the gene as Integrase Interactor 1, or INI1. This latter name has often been in the human pathology and rhabdoid tumor literature. Separate from these two, the gene has been given an official HUGO name of SWI/SNF related, matrix associate, actin dependent regulator of Chromatin, Subfamily B, Member 1 (SMARCB1). Since SMARCB1 is the “official” name, it has often been used in the cancer genome sequencing literature when lists of mutated genes are reported. Further, as the name has some official sanction, there has been some movement toward it. However, others prefer alternate nomenclature and refer to subunits of the complex as Brg1 associated factors (BAFs) followed by the mass of the protein in kilodaltons, thus resulting in related names for each subunit. The gene mutated in rhabdoid tumors is then referred to as BAF47. Thus far a consensus on terminology has yet to emerge. Hopefully, in the future, investigators from all of these fields can reach consensus on the nomenclature agreement. For the purposes of this review, we have chosen to use HUGO nomenclature but have previously, and recently, used other variations as well.

Rhabdoid tumors: linking SWI/SNF chromatin remodeling complexes to cancer

SMARCB1 (SNF5) is a core subunit present in all variants of the SWI/SNF complex. The first link between SMARCB1 and oncogenesis emerged from the study of rhabdoid tumors. Two laboratories discovered that specific, biallelic, inactivating mutations in SMARCB1 are found in rhabdoid tumors (RTs) (6, 7), and further that heterozygous Smarcb1 mutations are the basis of a familial cancer syndrome (7, 8). As described in more detail elsewhere in this issue, these cancers are aggressive and highly lethal pediatric tumors commonly found in the kidney, the central nervous system, where it is also known as atypical teratoid/rahbdoid tumor (AT/RT), and also less frequently in other soft tissues. Despite the use of intensive chemotherapy and radiotherapy, outcomes remain poor.

Recent data emerging from whole-exome sequencing of human cancers demonstrates that SMARCB1 is not the only subunit of the SWI/SNF complex mutated in cancer. Indeed, at least six genes encoding SWI/SNF subunits, including ARID1A, ARID1B, ARID2, SMARCA4 (BRG1), SMARCB1 (SNF5), PBRM1, and others, have been identified as recurrently mutated in wide spectrum of cancers (4) including ovarian (9, 10), lung (11, 12), liver (13, 14), pancreas (15), breast (16, 17), bladder (18), kidney (19), endometrial (20–22), skin (23), and brain cancers (24, 25), among others (26, 27) (Figure 1). As multiple SWI/SNF subunits are recurrently mutated in cancer, this suggests at least some degree of a shared mechanism of oncogenesis for these cancers. Consequently, findings and therapeutic insights from the study of the role of SMARCB1 mutations in RT may have implications for the variety of other SWI/SNF mutant cancers.

Figure 1. The SWI/SNF ATPase subunit genes are frequently mutated in specific types of human cancers.

The cancers in which genes encoding SWI/SNF subunits have been reported mutant.

SMARCB1 as a tumor suppressor: protecting the genome or epigenome?

Study of genetically engineered mouse models has demonstrated that homozygous Smarcb1 deficiency results in early embryonic lethality while heterozygous mice are predisposed to aggressive cancers that are histologically quite similar to human RT, including the presence of classic rhabdoid cells (28–30). In the mice, as in humans, these tumors are aggressive, locally invasive, and frequently metastatic to regional lymph nodes and/or lung. In contrast, the location of Smarcb1 deficient cancers in mice differs somewhat from those seen in humans. In mice, the tumors occur most commonly on the face and occasionally in brain, but never in kidney. Conditional, biallelic inactivation of Smarcb1 using the interferon inducible Mx1-Cre transgene results in profound cancer predisposition. All of these mice develop aggressive cancer including mature T cell lymphomas and rhabdoid-like tumors at a median onset of only 11 weeks (31). This is quite rapid compared to other tumor suppressors. For example, p53 inactivation leads to cancer at 20 weeks, p19Arf loss at 38 weeks, and p16Ink4a loss at 60 weeks. Thus, the rapid onset and complete penetrance of cancer following inactivation of Smarcb1 established this gene as a potent and bona fide tumor suppressor.

SMARCB1 and the SWI/SNF complex have been implicated in several forms of DNA repair including DNA double-strand break repair (32), UV-induced DNA damage repair (33), homologous recombinational repair (34), DNA decatenation (35), and nucleotide excision repair (36). Given this and the rapidity and full penetrance by which Smarcb1 loss causes cancer, we and others initially hypothesized that SMARCB1 loss drives cancer by leading to the rapid accumulation of DNA mutations and/or chromosomal instability. However, when testing this hypothesis we failed to detect major effects of SMARCB1 loss upon DNA repair and found that RT were diploid and largely lacked gene amplifications or deletions detectable at the level of SNP arrays aside from deletions at the SMARCB1 locus. Subsequently, via exome sequencing of RT we and others found that these cancers have a remarkably low rate of mutation (37–40) with loss of SMARCB1 being essentially the sole recurrent event detected. Given the absence of genomic instability despite the rapidity by which SMARCB1 deficient cancers arise both in humans and mice, these findings collectively suggest that the initiation and progression of aggressive cancers caused by SMARCB1 loss may not arise due to effects upon genome integrity, but rather epigenetically via perturbation of transcriptional regulation.

It is worthy of note that the function of SMARCB1 within the SWI/SNF complex remains largely unknown. SMARCB1 contains no obvious domains that clearly indicate function, but does contain two imperfect repeats to which most protein-protein interactions map (41) and has been implicated in non-specific DNA binding. It has been reported dispensable for formation of the SWI/SNF complex (42) but implicated in contributing to the targeting of the SWI/SNF complex to promoters (43).

Pathways & mechanisms

Since the identification of SMARCB1 mutations in RT, substantial efforts have sought to identify pathways and mechanisms (summarized in Figure 2) that underlie tumor formation. As many of these efforts are covered elsewhere in this issue, here we principally highlight a few findings to which we have contributed, at least in part.

Figure 2.

The SWI/SNF complex regulates expression of numerous target genes and pathways.

SMARCB1 regulates Cyclin D1/CDK4 activation

In a search for genes and pathways regulated by SMARCB1, it was found that reintroduction of SMARCB1 into an RT cell line resulted in G0–G1 cell cycle arrest associated with transcriptional repression of Cyclin D1, induction of P16(INK4A), and hypophosphorylation of RB (44, 45). Cyclin D1 is overexpressed in many human tumors and plays important roles during cell cycle. It interacts with cyclin-dependent kinase (cdk) 4/6 facilitating the phosphorylation and inactivation of RB to mediate G₁ to S progression. It also regulates activity of several transcription factors in a cdk-independent manner. Subsequently, re-expression of SMARCB1 was found to repress Cyclin D1 transcription by directly recruiting HDAC activity to the Cyclin D1 promoter (44–46). The requirement of Cyclin D1 for the genesis of rhabdoid tumors was then established in vivo when it was shown that genetic ablation of Cyclin D1 blocked formation of rhabdoid tumors (47). In collaboration with others, we found that levels of the Cyclin D1 transcript were high in ATRT compared to medulloblastomas or normal cerebellum. Notably, however, later studies of both renal rhabdoid tumors and ATRT did not detect elevated levels of Cyclin D1 (48, 49). As described in greater detail elsewhere in this issue, efforts are now focused on determining whether RT are sensitive to pharmacological disruption of Cyclin D1/CDK4 (50), and a phase I clinical trial is now being conducted to test the response of RT patients to CDK4 inhibition (http://clinicaltrials.gov/show/NCT01747876).

SMARCB1 loss activates the Sonic Hedgehog pathway

The Hedgehog (Hh) pathway plays important roles modulating patterning and differentiation during development. A clear genetic contribution of Hh pathway activity to oncogenesis was first realized when loss-of function mutations in the Hh pathway in familial and sporadic basal cell carcinoma (mutations in the PTCH1 gene), and medulloblastoma (mutations in the SMO gene) were identified (51, 52). Downstream, activation of glioma-associated oncogene homolog (GLI) transcription factors leads to transcription of Hh pathway target genes, including GLI1, GLI2, and PTCH1, and amplifications in the GLI1 transcription factor also occurs in gliomas and medulloblastomas. A role for SMARCB1 and the SWI/SNF complex in the control of Hh signaling was studied when SMARCB1 was identified as one of the top interactors of GLI1 and when human ATRT were found to have transcriptional profiles with similarities to Hh mutant medulloblastomas (53). Findings from this study demonstrated that SMARCB1 is directly recruited to the upstream regions of the transcription start sites of GLI1 target genes, GLI1 and PTCH1, and the expression of these genes were upregulated followed by shRNA-mediated knockdown of SMARCB1, supporting GLI1 and PTCH1 as direct targets. Conversely, re-expression of SMARCB1 in RT cells reduced GLI1 expression substantially, revealing a novel function of SMARCB1 as a mediator of Hh signaling. Notably, inactivation of SMARCB1 was found to result in direct activation of GLI1 mediated transcription and upstream inhibitors of the pathway, such as via inhibition of Smoothened (SMO), had no effect on expression of GLI1 targets in SMARCB1-deficient RT cell lines.

SMARCB1 loss activates the WNT/β-Catenin Pathway

Numerous studies have implicated the SWI/SNF complex in the control of lineage specification. For example, variations in the composition of SWI/SNF complexes have been shown to regulate transition from neural stem cell to post-mitotic neuron (3). Similarly, SMARCB1 is essential for hepatocyte differentiation (54) and likely for many other tissues as well (31). In mouse models, Smarca4 and Smarcb1 knockout mice fail to develop past the blastocyst stage and in lineage specific experiments, SMARCA4 and SMARCB1 are essential for control of T-cell development (55). Additionally, SWI/SNF complexes have been shown required for proper development of blood, liver, heart, muscle, melanocyte, adipocyte, and bone lineages (56–59), suggesting a role for the complex in regulation of developmental pathways.

Canonical Wnt signaling (β-Catenin dependent) regulates genes that are involved in diverse cellular functions such as differentiation and proliferation and aberrant activation of Wnt signaling is found in several types of human cancer. Studies have implicated the SWI/SNF complex in regulation of the Wnt signaling pathway. For instance, SMARCA4 can directly modulate transcription of multiple Fzd receptors and a subset of Wnt target genes in yolk sac endothelial cells (YSECs) (60). Also, SMARCB1 loss causes aberrant activation of the Wnt pathway and results in phenotypic defects consistent with Wnt/β-Catenin overexpression (61).

In addition, studies have implicated SMARCB1 in regulating numerous cancer-related genes, several of which we mention here. SMARCB1 has been implicated in the regulation of the cell cycle through RB-mediated repression of E2F target genes including Cyclin A, E2F1, and CDC6 (62, 63) and in both SMARCB1-deficeint RTs and Smarcb1-deficient mouse embryonic fibroblasts, the expression of E2F target genes was increased (45, 64). In addition, several studies found either interactions of SMARCB1 with c-MYC and/or a role for SMARCB1 in the control of expression of both c-MYC and its targets (65, 66). Given that MYC is highly expressed in RTs (48, 67), the aberrant expression of MYC and its target genes may have a role in the growth of these cancers. SMARCB1 has also been implicated in the control of P16(INK4A) and both in vitro and in vivo data suggest that P16(INK4A) is a direct downstream target of SMARCB1 (44, 64, 68, 69). Additional downstream targets that may play a role in oncogenesis include Polo-like kinase1 (PLK1), Aurora A kinase (70), RhoA mediated control of migration (71).

Collectively, these and other findings have implicated subunits of the SWI/SNF complex in the transcriptional regulation of the cell cycle, development, proliferation, and differentiation. Interestingly, upon mutation of the complex, alteration in the activity of these pathways seems to occur directly at the level of chromatin. For example, activation of both the Hh and WNT pathways in the absence of SMARCB1 is not dependent upon activation of the canonical pathways. Indeed, upstream blockade of the Hh pathway via inhibition of Smoothened (SMO) had no effect upon the high level of pathway activity in the absence of SMARCB1 (53). Similarly, blockade of the upstream WNT pathway had no effect upon the high level expression of pathway targets (61). Collectively, these findings suggest that mutation of SMARCB1 leads to alterations in chromatin structure that result in direct change to transcription and uncoupling of pathways from upstream canonical control.

Epigenetic antagonism between Polycomb and SWI/SNF complex during oncogenic transformation

So, how might changes in chromatin structure disrupt lineage specification? Enhancer of Zeste 2 (EZH2) is a histone-lysine N-methyltransferase that is the catalytic subunit of the Polycomb complex 2 (PRC2). It mediates the trimethylation of lysine 27 of histone H3 (H3K27me3), a repressive chromatin mark, playing a vital role in epigenetic gene silencing and oncogenic transformation. EZH2 has also been implicated in the regulation of pluripotency, differentiation, and proliferation (72, 73). Importantly, its deletion is embryonic lethal, suggesting that EZH2 is critical for normal development (74). Recently, EZH2 has been found to be over-expressed in numerous cancer types and associated with tumor cell proliferation, invasive growth, and poor prognosis (75, 76). Given that EZH2 has been implicated in development and maintenance of cancers, the mechanism of action of EZH2 has been evaluated in several malignancies. For instance, recent studies have demonstrated that disruption of aberrant EZH2 activity inhibited prostate cancer tumorigenicity (77) and EZH2 serves a crucial role for tumor stem cell maintenance in glioblastoma (78). Interestingly, the roles of EZH2 in oncogenesis may be highly context dependent. For instance, gain-of-function somatic mutations of EZH2 have been found in B-cell non-Hodgkin’s lymphoma while inactivating mutations of EZH2 occur in myleodysplastic syndrome (79–81). Thus appropriate control of EZH2 is important for normal development such that both over- and under-expression may be capable of promoting transformation in certain contexts.

Potential links between Polycomb and the SWI/SNF complex first emerged from genetic studies in Drosophila, from which mutations in SWI/SNF subunits were found act as positive regulators of Hox genes and to partially suppress the effects of Polycomb mutations (82). Proteins encoded by trithorax group genes activate expression of Hox genes while proteins encoded by polycomb genes mediate repression. Several members of the trithorax group of genes were subsequently found to encode subunits of the SWI/SNF complex including Brm, moira (BAF155 ortholog) and Osa (BAF270 ortholog). Mechanistic insight revealing that polycomb proteins can directly repress chromatin remodeling by SWI/SNF complexes was shown in vitro (83). This balance between Polycomb and SWI/SNF activity was found to extend to mammals and to be essential for tumor suppression by SMARCB1 as inactivation of the polycomb gene Ezh2 using a conditional mouse model completely blocked tumor formation caused by Smarcb1 inactivation (63). Mechanistically, deletion of SMARCB1 led to elevated expression and recruitment of EZH2 to Polycomb targets that became broadly H3K27-trimethylated and repressed in SMARCB1-deficient fibroblasts and cancers. Importantly, these effects were highly specific, as knockout of EZH2 completely blocked the growth of SMARCB1 mutant cancers, but had no effect on osteosarcomas driven by p53/Rb loss. In follow up to this work, a small molecule inhibitors (EPZ-6438), which block the methyltransferase active site of EZH2 was shown to specifically inhibit cellular H3K27 methylation and selectively induce apoptotic death of SMARCB1 mutant RT cells both in culture and in xenograft experiments in vivo (84). Consequently, there is interest in testing EZH2 inhibitors in patients who have recurrent/progressive RT, as such disease is nearly uniformly fatal.

The Swi/Snf complex contributes to transcriptional regulation by mobilizing nucleosomes

Appropriate regulation of gene expression requires the interplay of complexes that modulate chromatin compaction in conjunction with sequence-specific transcription factors and basal transcription machinery. Nucleosome occupancy and position correlate with transcription rate as, for example, gene activation correlates with additional nucleosome depletion (85, 86). In addition, the promoter regions of active genes are enriched with nucleosomes containing specific post-translationally modified histones that carry correlate with modulation of transcription rate.

The role of the SWI/SNF complex in cell fate specification has been postulated to arise in large part from nucleosome mobilization driven by the activity of its ATPase subunits (SMARCA4 and SMARCA2). The molecular mechanism by which the SWI/SNF complex alters chromatin structure has been an active area of investigation. Evidence supports a role for the SWI/SNF complex in controlling nucleosome position and transcription regulation. For instance, it has been reported that purified SWI/SNF complex is able to mobilize chromatinized nucleosomes in vitro and the complex can bind preferentially to promoters and other regulatory regions to move nucleosomes in vivo (87–91).

Recent studies have used murine embryonic fibroblasts (MEFs) to demonstrate perturbation in SWI/SNF-driven regulation of chromatin structure and transcription when core subunits, Smarcb1 or Smarca4, are conditionally knocked out. In this study, the function of each subunit and its contribution to chromatin structure alteration were directly compared on a uniform genetic background. The analysis showed that Smarcb1 and Smarca4 are essential for the establishment of appropriate nucleosome structures surrounding transcription start sites (TSS) genome wide (92). The complex may also serve roles in the regulation of enhancers (93, 94). Despite marked alterations in nucleosome occupancy, Smarcb1 and Smarca4 deficient MEFs did not have large differences in gene expression levels when compared to wild-type MEFs. These findings were perhaps surprising given previous findings that open promoter states are established when they are activated. However, the relationship between nucleosome occupancy and gene expression is complex as it has been demonstrated that even promoters of infrequently transcribed genes can stably form open states and nucleosome depletion is not sufficient to induce transcription activation (95, 96). In addition, there are more factors to consider at promoters, such as histone modifications, histone variants, and proteins that can modify post-translational modification of histone tails, not to mention at distal enhancers, in order to fully understand how nucleosome position contributes to transcriptional regulation. Consequently, nucleosome occupancy can affect both activation and repression of genes and there is a great need for further study of the contributions of SWI/SNF mediated nucleosome regulation to transcriptional regulation.

Conclusion

Rhabdoid tumors are rare cancers that strike young children and result in poor survival rates. SMARCB1 is mutated in the vast majority of RT, cancers that otherwise have remarkably simple genomes. Although mutation of SMARCB1 was the first SWI/SNF subunit mutation identified in human cancers, recent findings from cancer genome sequencing studies show that other SWI/SNF subunits are also mutated at a high frequency in a variety of human cancers (Fig 1).

Cancer is largely thought to be a disease dependent upon aberrant expression of genes involved in key cellular pathways that regulate proliferation, survival, and apoptosis. Recent cancer genome sequencing data have identified frequent mutations of genes that are involved in regulation of chromatin structure, with mutations in subunits of the SWI/SNF complex perhaps being the most frequent. These discoveries highlight the important interrelationships between genetic and epigenetic changes in the genesis of cancer. However, the specific mechanisms by which mutation of chromatin regulators drive cancer remain largely unclear.

Although it has been established that the SMARCB1 and the SWI/SNF complex contribute to regulation of some canonical cancer-related pathways, given that the complex binds to the promoters of roughly one-third of all genes, as well as to active enhancers, the fundamental mechanisms tumor suppression seem unlikely to be derived from a single downstream target pathway but more likely the result of disruption of coordinated pathways of lineage specific signaling. Consequently, pursuit of further insight into the fundamental role of the SWI/SNF complex in transcription regulation should help to understand the role of SMARCB1 and the SWI/SNF complex in tumor suppression. Given the simple genomes of RT and the rapidity with which cancer forms in Smarcb1 mutant mice, study of SMARCB1 offers potential benefit not only to patients with RT but can serve as a useful model with which to potentially identify therapeutic vulnerabilities conferred by SWI/SNF mutation across the wide variety of SWI/SNF mutant cancers.