Rhabdoid Tumors: An Initial Clue to the Role of Chromatin Remodeling in Cancer

Abstract

The discovery of biallelic, inactivating SMARCB1 mutations in rhabdoid tumors (RTs) over a decade ago represented the first recognized link between chromatin remodeling and tumor suppression. SMARCB1 is a core subunit of the SWI/SNF chromatin remodeling complex, and the recent emergence of frequent mutations in genes that encode subunits of this complex across a wide variety of cancers suggests that perturbation of this chromatin remodeling complex constitutes a key driver of cancer formation. Despite the highly aggressive nature of RTs, they are genetically simple cancers that appear to lack chromosomal instability and contain very few mutations. Indeed, the mutation rate in RTs is among the lowest of all cancers sequenced, with loss of SMARCB1 as essentially the sole recurrent event. Given the genetic simplicity of this disease, understanding the chromatin dysregulation caused by SMARCB1 loss may provide more general insight into how epigenetic alterations can contribute to oncogenic transformation and may reveal opportunities for targeted therapy not only of RT but also the variety of other SWI/SNF mutant cancers.

Chromatin and Cancer

History has revealed a theme that genes mutated in early onset pediatric cancers and familial cancer predisposition syndromes are often of fundamental importance to cancer in general. For example, retinoblastoma is a rare pediatric cancer but served as the initial clue to the importance of the RB gene in tumor suppression. In 1998, the causative mutation of rhabdoid tumors (RTs) was identified as biallelic inactivation of the gene SMARCB1 (SNF5/INI1/BAF47) 5, 6, 71. SMARCB1 encodes a core subunit of SWI/SNF chromatin remodeling complexes 53, 76. This finding represented the first identified link between ATPase chromatin remodeling complexes and tumor suppression—a link that has since blossomed with findings of SWI/SNF subunit mutations in many types of cancer, both pediatric and adult. Indeed, recent cancer genome sequencing efforts have further highlighted the role of chromatin regulators in cancer as a variety of genes whose products covalently modify histones and chromatin have also been found to be mutant in cancer. Here we review and discuss RT and describe mechanistic and potential therapeutic insights that have resulted from subsequent studies.

RTs are aggressive pediatric cancers that arise most frequently in the brain, where they are referred to as atypical teratoid/rhabdoid tumor (AT/RT), and the kidney, but can also arise in soft tissues throughout the body 3, 8, 59, 78. These tumors tend to have a very early onset, with roughly half of cases arising in children less than 1 year of age and the large majority of cases occurring in children under the age of 3 3. These tumors are highly malignant and confer a particularly poor prognosis, although the recent use of highly intensive multimodality therapeutic regimens may offer some improvement in outcome 14. Although an origin from neural progenitors has been hypothesized, analyses of gene expression have yielded conflicting conclusions, and the cell(s) of origin remains unclear 11, 20.

After cytogenetic analysis identified the deletion of chromosome 22q11.2 region as the only frequent copy number change in RT cell lines and primary tumors, further mapping showed that the gene SMARCB1 was almost universally missing in these cancers 4, 6, 31, 71. These findings provided the first clear evidence that AT/RT, renal RT and soft‐tissue RT are all genetically related 6, 64. Analogous to retinoblastoma, germ line mutations in SMARCB1 strongly predispose to RTs and the rare cases of familial rhabdoid predisposition syndrome are predominantly due to inherited heterozygous mutations in SMARCB1 65, 68. Indeed, biallelic loss of SMARCB1, particularly as evidenced by negative immunostaining with the INI1 antibody, is now used in the diagnosis of RT 3. It should be noted, however, that in a very small percentage of RT cases, the tumors lack SMARCB1 mutations but carry mutations in another SWI/SNF subunit, SMARCA4 (BRG1/SNF2‐beta) 24, 61, further establishing mutation of the SWI/SNF complex as a signature event of RT.

SWI/SNF Complex Subunits are Frequently and Widely Mutated in Cancer

SWI/SNF complexes were originally characterized in yeast as factors being essential for mating type switching (SWI) or for sucrose metabolism (mutants were sucrose non‐fermenting, SNF). Mammalian versions of these complexes consist of 10–12 subunits and are transcriptional modulators that possess ATP‐dependent nucleosome remodeling activity 9, 16. SWI/SNF complexes are comprised of one of two ATPase subunits (SMARCA4 or SMARCA2/BRM), a set of highly conserved “core” subunits (SMARCB1, SMARCC1/BAF155 and SMARCC2/BAF170), and variant subunits thought to contribute to lineage‐specific functions of the complexes 26, 27, 36, 51, 53, 75. While SMARCB1 is constitutively present in SWI/SNF complexes, its biochemical contribution remains unclear.

Although SWI/SNF complexes were first linked to cancer when SMARCB1 was found to be recurrently mutated in RT, recent findings from cancer genome sequencing studies reveal that at least seven SWI/SNF subunits are mutated, often at high frequency, across a wide variety of cancers (Table 1) 82. For example, the ARID1A (BAF250a) subunit is mutated in 50% of ovarian clear cell carcinomas, 30% of endometrioid carcinomas, 27% of gastric cancers, 13% of bladder carcinomas, 10% of colorectal and lung adenocarcinomas, 10% of hepatocellular carcinomas, and occasionally in breast, pancreatic, and prostate cancers and neuroblastomas 10, 15, 19, 29, 30, 34, 35, 54, 62, 66, 80, 84. The PBRM1 (BAF180) subunit is mutated in 40% of renal carcinomas, as well as occasionally in breast, gastric and pancreatic cancers 18, 52, 66, 70, 85. The BRD7 subunit is lost in up to 20% of p53 wild‐type breast cancers 17, and the ARID2 (BAF200) subunit is mutated in 18% of hepatitis C‐associated hepatocellular carcinomas, 7% of melanomas, and occasionally in some pancreatic cancers 19, 28, 43, 46. The SMARCA4 subunit is mutated in 35% of non‐small cell lung cancers and in medulloblastomas, particularly the WNT (26%) and group 3 (11%) subtypes 44, 49, 50, 54, 55, 58. Finally, mutations in SWI/SNF subunits were identified in one‐third of pancreatic cancers 2, 66. It remains unclear why different SWI/SNF subunits are mutated in different cancers, although it raises the possibility of lineage‐specific contributions of individual subunits. Collectively, a broad role for the SWI/SNF complex as a tumor suppressor has recently emerged, suggesting substantial relevance for this complex in tumor suppression beyond RT.

Table 1.

SWI/SNF subunits are mutated in a wide variety of cancers. Selected studies revealing SWI/SNF subunit mutations in cancer

Common SWI/SNF subunit mutations discovered by cancer genome sequencing
SWI/SNF complex subunit	Cancer type	Cell line or primary tumor	Mutation type	% Mutated	Reference
SMARCB1	Rhabdoid tumor	Primary tumors	Deletions and copy neutral loss of heterozygosity	94	25
		Primary tumors	Homozygous inactivation: deletions, nonsense and other null mutations	100	42
		Primary tumors	Homozygous inactivation: deletions, nonsense and other null mutations	98	31
SMARCA4	Medulloblastoma	Primary tumors	Missense mutations	4	54
		Primary tumors	Null and missense mutations	5	33
	Lung adenocarcinoma	Both	Unknown point mutation	32	44
		Primary tumors	Null and missense mutations	11	30
	Pancreatic ductal adenocarcinoma	Both	Heterozygous deletions	9.6	66
SMARCA2	Pancreatic ductal adenocarcinoma	Both	Heterozygous deletions	2.6	66
PBRM1	Renal clear cell carcinoma	Primary tumors	Null mutations	53	52
		Both	Null mutations	60	18
		Primary tumors	Null and missense mutations	21	23
		Primary tumors	Truncating mutations	41	70
	Pancreatic ductal adenocarcinoma	Both	Null mutations and deletions	9.6	66
		Cell lines	Truncating mutations	27	74
ARID2	Hepatocellular carcinoma	Primary tumors	Indel, nonsense and null mutations	5.8	19
		Primary tumors	Truncating mutations	14–18	43
	Non‐small cell lung carcinoma	Both	Homozygous deletions, null mutations	7.3	46
	Pancreatic ductal adenocarcinoma	Primary tumors	Null and missense mutations, copy number changes	12	2
	Melanoma	Primary tumors	Null and missense mutations	9	28
ARID1A	Ovarian clear cell carcinoma	Both	Null and missense mutations	46	80
		Primary tumors	Null mutations	57	35
	Endometrioid carcinoma	Primary tumors	Null and missense mutations	30	80
	Gastric adenocarcinoma	Primary tumors	Null mutations	8	86
		Both	Indel, missense and nonsense mutations	10	34
		Primary tumors	Truncating mutations	29	73
	Bladder transitional cell carcinoma	Primary tumors	Null and missense mutations	13	21
	Colorectal carcinoma	Primary tumors	Indel and nonsense mutations	10	10
		Both	Indel and nonsense mutations	10	34
	Hepatocellular carcinoma	Primary tumors	Indel, missense and nonsense mutations	13	29
		Primary tumors	Indel, missense and nonsense mutations	10	19
	Pancreatic ductal adenocarcinoma	Primary tumors	Indel and nonsense mutations	10	2
		Both	Null mutations and deletions	8.3	66
		Both	Indel and nonsense mutations	8	34
	Neuroblastoma	Primary tumors	Null and missense mutations	6	60
	Lung adenocarcinoma	Primary tumors	Indel, missense and nonsense mutations	8	30
		Both	Indels	2	34
	Breast cancer	Primary tumors	Indel, missense and nonsense mutations	4	67
		Primary tumors	Deletions (point mutations)	37 (3)	15
		Both	Indel, missense and nonsense mutations	4	34
		Primary tumors	Deletion	13	45
	Prostate cancer	Both	Indels	8.7	34
ARID1B	Neuroblastoma	Primary tumors	Deletions, null and missense mutations	7	60
	Breast cancer	Primary tumors	Indel, missense and nonsense mutations	5	67
	Pancreatic ductal adenocarcinoma	Both	Null mutations and deletions	3.9	66
	Hepatocellular carcinoma	Primary tumors	Indel, missense and nonsense mutations	6.7	19

RT: Remarkably Simple Genomes

Since first being linked to RT, studies have sought to identify additional genetic mechanisms and mutations that cooperate with SMARCB1 loss to drive RT. While SMARCB1 was initially implicated in DNA repair 72, a subsequent evaluation found no clear role and further found that 16 out of 18 primary RT samples were diploid and indistinguishable from normal peripheral blood DNA by SNP array, other than SMARCB1 loss, suggesting an unusual degree of genomic stability in RT 47. To characterize the genomes of RT and search for genetic mutations that cooperate with SMARCB1 loss to drive RT formation, SNP analysis and exome sequencing were recently performed on 32 frozen primary RT samples paired with matched normal genomic DNA. RT genomes were found to be remarkably simple with no recurrent copy number variations other than focal deletions of the SMARCB1 locus or monosomy of chromosome 22, where SMARCB1 is located 42, 47. Whole exome sequencing revealed a particularly low mutation rate, with an average of four non‐synonymous somatic mutations detected per tumor exome 42. This mutation rate is among the lowest of all sequenced cancer genomes, and indeed two tumors contained no somatic mutations other than biallelic loss of SMARCB1. Across the 32 samples, SMARCB1 was the only gene found to have significant recurrent mutations. Two additional studies focused upon known cancer‐related genes also found an essential absence of cooperating mutations 25, 39. Collectively, these data raise the possibility that, at least within the protein coding exome, SMARCB1 loss might be the sole genetic driving event in RT, although it remains possible that germ line variants or mutations located outside the exome could be required to cooperate with SMARCB1 loss to drive RT tumor formation.

The surprisingly low mutation rate seen in RT suggests that widespread genomic instability might not be a necessary hallmark for the genesis of even highly aggressive cancers such as RT. Rather, at least in certain contexts, perhaps only a few, or one, genetic perturbation can be sufficient to drive cancer formation. Other pediatric cancers such as infant leukemia, neuroblastoma and retinoblastoma have also been found to possess mutation rates several log‐fold lower than highly mutagenic adult cancers such as melanoma 28, 60, 87. Some adult cancers, such as acute myeloid leukemia (AML), have also been found to have low mutation rates, and occasional individual cases of most common cancer types also have very low mutation rates 79. While 5–15 somatic driver mutations had been predicted as being essential for the genesis of cancer 7, it now appears that substantially lower numbers are required, at least in some contexts. But how could a single genetic lesion be sufficient for cancer? The mathematics would seem to suggest, were this true, that everyone would develop cancer at a young age. This raises the possibility that cooperating events, rather than being genetic mutations, might occur as epigenetic chromatin alterations.

Functional Insights into SMARCB1

Insights come from studies regarding the normal function of SMARCB1 and the mechanisms by which its loss drives cancer formation. Several mouse models of Smarcb1 inactivation have been developed 22, 40, 56, 57, 69. Homozygous deletion of Smarcb1 results in early embryonic lethality while heterozygous knockout mice develop tumors comparable to human RT 22, 40, 56. Analysis revealed that these tumors have spontaneously lost the second Smarcb1 allele, a requirement that likely explains the 30% penetrance and 11‐month median onset. In comparison, in conditional mice in which Smarcb1 is inducibly inactivated in some cells, 100% of mice develop lymphomas or RTs with a median onset of only 11 weeks, markedly faster than that occurs following inactivation of tumor suppressors such as p53, p16 or p14 37, 57, 63, 81. Of note, as cancers driven by Smarcb1 loss occur in multiple genetic backgrounds, this could suggest that a specific germ line variant is not essential for the genesis of cancers driven by Smarcb1 loss. Collectively, these experiments demonstrated a potent and bona fide role for Smarcb1 as a tumor suppressor.

Providing important insight into transformation driven by Smarcb1 loss, while Smarcb1 inactivation results in rapid cancer onset, its loss is detrimental to the vast majority of cells. Inactivation in embryonic stem cells blocks development and causes arrest at the blastocyst stage 40. Deletion of Smarcb1 in primary fibroblasts results in cell cycle arrest and a block in proliferation 41. Indeed, even in a T‐cell lymphoma model where Smarcb1 deletion in vivo results in rapid cancer onset in all mice 57, the vast majority of T cells do not tolerate Smarcb1 loss, resulting in markedly reduced T‐cell numbers in these animals. From this aberrant T‐cell environment, cancer arises like a phoenix from the ashes. As the human tumor data suggest that this phenotypic variation is not explained by the acquisition of genetic mutations, epigenetic differences and alterations likely play a key role in determining susceptibility to transformation upon SMARCB1 loss. Notably, while SWI/SNF subunit mutations are collectively found in a wide variety of cancers, SMARCB1 loss itself has only been associated with RT and a few other rare cancers—it is not generally mutated in common adult cancers. In mouse lymphomas, Smarcb1 loss drives lymphoma formation from mature CD8+ memory‐like T cells, but not from immature T cells or B cells 77. Collectively, this suggests that only certain cell types, or epigenetic states, are susceptible to SMARCB1 loss. Other cell types/states might be susceptible to the loss of other specific SWI/SNF subunits.

Alterations of the epigenetic landscape caused by SMARCB1 loss affect several pathways that might be important in the genesis of RT in susceptible cell types. Studies of mouse models and human tumors revealed the existence of epigenetic antagonism between SMARCB1 and the Polycomb repressor complex subunit EZH2, such that the loss of SMARCB1 results in unopposed activity of EZH2 that serves an essential role in tumor formation 38, 83. In addition, the loss of SMARCB1 in RT has been found to have effects upon a wide spectrum of cellular processes. SMARCB1 and SWI/SNF complexes appear to modulate the cell cycle through transcriptional regulation of the p16^INK4a‐cyclinD/CDK4‐pRb‐E2F mitotic checkpoint 1, 13, 72. Similarly, the absence of SMARCB1 increases cell motility by posttranslational deregulation of RhoA, possibly through epigenetically driven changes in expression of G‐protein guanine exchange factors and activating proteins 12. These various effects might be modulated by adaptor proteins such as BIN1 48. Evidence has also linked SMARCB1 directly to GLI1, an effector in the hedgehog signaling pathway 32. While it remains unknown whether these same pathways are activated in cancers linked to mutations in other SWI/SNF subunits, the understanding gleaned from studying RT suggests that SMARCB1 loss results in the deregulation of multiple pathways that contribute to oncogenesis in a lineage‐specific fashion.

Understanding the mechanisms by which epigenetic changes contribute to oncogenesis offers the promise of informing development of therapeutics. Recent sequencing efforts have revealed a theme that genes encoding chromatin modifiers, not just SWI/SNF, are recurrently mutated across a wide spectrum of cancer types, suggesting that changes in chromatin structure might be important in oncogenesis. Indeed, epigenetic alterations have been implicated in many cancer types, although it has often been challenging to define potential epigenetic contributions when they occur in the setting of high mutation rates and genomic instability. Perhaps in part due to a low incidence of age‐related and environmentally induced passenger mutations, early onset pediatric cancers such as RT often have simpler genomes. Consequently, these cancers might constitute outstanding models with which to mechanistically characterize the roles of chromatin‐based changes in driving cancer growth and, potentially, with which to identify specific mutation‐driven therapeutic susceptibilities.