One of the many remarkable findings of the cancer genome sequencing projects that have been published in this and other journals is the repeated discovery of somatic driver mutations in genes that encode chromatin remodeling factors, which regulate the epigenome. De novo mutations in this same family of genes also cause several developmental syndromes, whose various features all include intellectual disability. Surprisingly, a few de novo mutations in these genes have recently been reported in autism. How do these mutations (which at least in some cases appear to be loss-of-function in both cancer and in the developmental syndromes) in the same class, and in some cases, the same exact genes cause these different diseases?
Much effort and attention is being paid toward developing drugs that target the epigenome and the proteins that regulate it. The $95M partnership between Constellation Pharmaceuticals and Genentech and MIT’s Koch Institute’s recent symposium–Epigenetics, plasticity and cancer–are just two of many examples. But most intriguingly, can understanding the molecular mechanisms and risk factors of one of these diseases inform the biology and treatment of the others?
The epigenetic state of a genome has a vast influence over gene regulation, cellular activity and cell fate. Protein families that “write,” “erase,” and “read” the major histone marks (i.e. acetyl and methyl groups) ultimately regulate accessibility of the chromatin to transcriptional machinery. (A useful review on epigenetic protein families and recent progress to pharmacologically modulate these proteins was published recently in Nature Reviews Drug Discovery here. )
New driver mutations in cancer
In the past few years, cancer sequencing efforts have identified driver mutations in genes that regulate histone and DNA modification in various types of cancers. In 2012, this journal published many cancer sequencing papers, four of which reported driver mutations in chromatin remodelers. While ARID1A mutations have been found in other cancers (e.g. ovarian carcinoma in Science and the New England Journal of Medicine, Patrick Tan and colleagues reported here in April that ARID1A is mutated in 8% of gastric adenocarcinomas, and that frequent mutations in chromatin remodeling genes (ARID1A, MLL3 and MLL) were found in 47% of the gastric cancers they screened.
In May, Bin Tean Teh and colleagues reported inactivating mutations in MLL3 in 14.8% of cases of liver fluke-associated cholangiocarcinoma, a fatal cancer that occurs in the liver bile ducts that is common in parts of Southeast Asia infested with O. viverrini. MLL3 encodes a histone-lysine N-methyltransferase and was already known to be mutated in several other cancer types. Teh and colleagues found that 75% of MLL3 mutations were likely loss-of-function, with mutation patterns reminiscent of tumor suppressor genes.
In June, Jessica Zucman-Rossi and colleagues reported that inactivating mutations in ARID1A and ARID2 were found in 16.8% and 5.6% respectively, of 125 cases of hepatocellular carcinoma. ARID1A and ARID2 are SWI/SNF-related chromatin remodelers that control the accessibility of transcriptional machinery to promoter regions of DNA. Overall, 24% of HCC’s that they screened had at least one mutation in a chromatin remodeling gene (2 cases with SMARCA4 mutations, and single cases of many genes that encode chromatin remodelers, including SMARCA2, SMARCB1, SMARCA1, ARID4A, PBRM1, CHD3 and CHD4). Interestingly, ARID1A mutations were more often found in tumors related to alcohol intake compared to tumors related to other etiological sources (e.g. hepatitis B or C virus).
In July, Hidewaki Nakagawa and colleagues published another hepatocellular carcinoma sequencing study, which analyzed HCC tumors associated with hepatitis B or C virus infections. They initially sequenced 27 tumors and identified 2 frameshift and one missense mutation in ARID1A. Analysis of 120 more tumors identified 12 more mutations in ARID1A. Altogether, the authors found that 50% (14/27) of the tumors had recurrent mutations in genes that encode chromatin regulators. They also knocked down these chromatin regulators in a panel of 5 HCC cell lines, and found that knockdown of MLL3 led to increased cell proliferation. Knockdown of 11 genes, many of which are chromatin regulators, increased cell proliferation in at least one cell line. These results support the hypothesis that loss-of-function mutations in chromatin regulators promote cell growth in hepatocellular carcinoma.
De novo mutations in intellectual disability and autism
De novo mutations in this family of genes were recently reported in autism in Nature [see last paragraph of previous post on this blog entitled ‘Autism exomes arrive’ ], a pervasive developmental disorder that sometimes includes intellectual disability, and typically presents with cognitive and social dysfunction. In addition to the two cases of CHD8 mutations, single cases of de novo nonsense, missense and frameshift mutations were reported in those autism sequencing papers in ARID1B, CHD3, CHD7, MLL3, SETBP1 and SETD2.
As noted in that post, de novo mutations in this class of genes have also been found in recent years to cause various developmental syndromes, including mutations in MLL2 in Kabuki syndrome. This discovery was published in the landmark paper that was the first application of exome sequencing to define the cause of an autosomal dominant disorder. This year, this journal published 5 other papers that report de novo mutations in genes that encode chromatin remodelers in developmental syndromes that vary in presentation but all include intellectual disability.
In April, Naomichi Matsumoto and colleagues reported that 20 of 23 individuals with Coffin-Siris syndrome (CSS) carried missense and truncating mutations in one of the six genes that encode SWI/SNF subunits, including SMARCB1, SMARCA4, SMARCA2, SMARCE1, ARID1A and ARID1B. CSS is a rare congenital syndrome (MIM 135900) that includes growth deficiency, intellectual disability, severe speech impairment, microcephaly, and coarse facial features. In all cases where parental samples were available, mutations occurred de novo. Notably, only one of the 23 cases presented with a hepatoblastoma. All of the mutations in ARID1A and ARID1B were truncating, and the authors suggest that haploinsufficiency causes CSS. At the same time, Gijs Santen and colleagues also reported truncating de novo mutations in ARID1B in CSS.
In the same issue, Joris Vermeesch and colleagues reported heterozygous mutations in SMARCA2 in Nicolaides-Baraitser syndrome (NBS, MIM 601358). The features of NBS include sparse hair, distinctive facial features, microcephaly, epilepsy and intellectual disability with marked language impairment. Altogether, the authors identified missense mutations in SMARCA2 in 36 of 44 individuals analyzed. In 15 of these patients, parental DNA was available and the mutations in each case were verified to occur de novo. None of the mutations were truncating, and the authors suggest that these mutations act in a dominant-negative or gain-of-function manner.
In June, Marcella Zollino and Bert de Vries and their respective colleagues independently reported mutations in the chromatin regulator KANSL1 in 17q21.31 microdeletion syndrome. KANSL1 is a subunit of a histone acetylatransferase (HAT) protein complex, and is required for its HAT activity. The mutations identified by both papers occur de novo and include nonsense and frameshift mutations. Interestingly, a common feature of this microdeletion syndrome is a “happy, friendly disposition.”
Biological meaning?
What is the biological meaning of these findings? Many of the somatic mutations in these chromatin remodeler proteins in cancer appear to be loss-of-function, but further in vivo and in vitro experiments will determine if these proteins act as tumor suppressors or oncogenes. It is also possible that different proteins can have either effect, depending on the type of cancer. Mouse models of the developmental disorders noted here should bring further insights into how this class of proteins leads to these particular diseases.
Why does a loss-of-function somatic mutation in ARID1A cause so many types of cancer, and how can loss-of-function in this same gene also cause Coffin-Siris syndrome? How does loss of this class of proteins cause cancer in one context and intellectual disability, language impairment or autism in another? Is the sole case of a de novo frameshift indel mutation in ARID1B in autism relevant? If so, how does loss-of-function of ARID1B cause autism? It is clear that much is still to be learned on how the epigenome regulates gene activity and how its misregulation in particular times and spaces can cause radically different severe diseases. Nevertheless, the most intriguing questions are these: 1) Where, and how, do these chromatin remodelers functionally overlap in these disparate diseases? And 2) Would a treatment for one disease be effective for another?