Epigenetics and Genome Instability

DNA methylation and histone modifications are intrinsically linked to chromatin structure and gene expression. Epigenetic changes during development indicate an essential role in development and in tissue-specific differentiation. Moreover, aberrant epigenetic alterations are a hallmark of most cancers making it increasing clear that the genome is vulnerable to epigenetic as well as genetic alterations that contribute to malignant progression. Cancer cells tend to be globally hypomethylated, resulting in chromosomal instability by possibly altering chromatin structure in repetitive DNA and promoting DNA recombination events that result in DNA deletions and rearrangements.

In contrast, it is estimated that about 50% of the known tumor suppressor genes can be silenced by DNA hypermethylation in gene promoter regions as an alternative to mutation or DNA deletion. However, DNA hypermethylation may also result in increased gene expression. For example, hypermethylation of the H19/IGF2 Imprinting Control Region (ICR) may result in biallelic expression of IGF2 in tumors that promotes cell growth and survival. In contrast, hypomethylation of the KvDMR ICR may result in silencing of CDKN1C, a tumor suppressor gene, which is more than 150 kb away. Thus, DNA hypermethylation and hypomethylation may have different consequences on gene expression in malignant cells.

From the first description of Restriction Landmark Genomic Scanning (RLGS), members of the RPCI Genetics Program have been instrumental in the development and application of this genome-wide screening method for identifying methylation changes. RLGS is a method for the two dimensional display of end-labeled DNA restriction fragments. Using the methylation-sensitive restriction enzyme NotI as the restriction landmark, RLGS targets genes rich in CpG island regions and detects differences in DNA methylation between tissue samples.

Comparison of a RLGS profile between normal liver (top) and tumor (below) shows loss of methylation of specific CpG islands (arrows)

This process has been greatly simplified by the development of “Virtual” RLGS software by the Nagase and Held labs (Matsuyama et al., Nucleic Acid Res 2003; 31:4490) and independently by the Smiraglia laboratory. This development allows the unequivocal in silico identification of genes showing aberrant methylation patterns following gel electrophoresis and virtually eliminates the need for subcloning and sequencing thereby providing a rapid means of proceeding to biological assays.

Unmethylated CpG-islands: A paradigm shift 

With the exception of the inactive X chromosome, imprinted genes, and rare examples of non-imprinted genes (e.g. cancer-testis genes), CpG-islands were thought to be methylation-free.  RLGS, in conjunction with Virtual RLGS, has been used by the Held, Smiraglia, and Nagase labs to define and analyze several unique tissue- and developmental-specific differences in genomic methylation in mouse (Song et al., Proc Natl Acad Sci USA, 2005; 102:3336; Oakes et al., Proc Natl Acad Sci USA 2007; 104:228) and in human (Kitamura et al., Genomics 2007; 89:326).

Tissue-specific differences in methylation (TDMs) are shown to be relatively abundant in the mammalian genome and have a broad distribution within the genome, including 5’ promoter regions, 3’ exons, introns, and intergenic regions, both within and outside of CpG islands. These findings dispel the previous notion that almost all CpG islands are unmethylated, suggesting that DNA methylation may have regulatory functions in addition to gene silencing due to 5’ promoter methylation. This new understanding of tissue-specific differentially methylated regions (TDMs) has important implications regarding tumor-specific DNA methylation and almost certainly impacts on malignant growth.

RLGS has also been used to define and then characterize tumor-specific methylation changes (Smiraglia et al., J Med Genet  2003; 40:25) and resulted in the identification of a new tumor suppressor gene, TCF21, that is frequently inactivated in human malignances (Smith et al., Proc Natl Acad Sci USA, 2006; 103:982). The silencing of TCF21 was associated with an increased mesenchymal phenotype whereas expression of this gene promoted mesenchymal to epithelial transition. 

RLGS also identified a candidate tumor suppressor gene, SLC5A8, that is silenced in human colon aberrant crypt foci and colon cancer (Li et al., Proc Natl Acad Sci USA 2003; 100:8412). Studies indicate that RLGS can be used to develop methylation biomarkers (Smiraglia and Plass, Ann NY Acad Sci 2003; 983:110) because of their ability to identify methylation changes in very small subpopulations of cells.

In addition, the Smiraglia lab has collaborated with Dr Foster (Prostate Program) and Dr Karpf  (Molecular Targets and Experimental Therapeutics Program) to analyze methylation in TRAMP mice (a transgenic mouse model of prostate cancer) using RLGS, and this study revealed a systemic DNA methylation pathway defect reminiscent to what occurs in human prostate cancer (Morey et al., Cancer Res 2006; 66:11659). In parallel studies being conducted in collaboration with James Mohler, MD (Prostate Program), examination of patterns of methylation in androgen dependent and castration resistant human prostate cancer samples is ongoing.

The KvDMR1 ICR, Beckwith-Wiedemann Syndrome and cancer

The major interest of Michael Higgins, PhD over the last few years has been genomic imprinting and cancer. Recently seminal publications provide strong evidence supporting a role for deregulated genomic imprinting in the development of cancer.

For example, LOI of Igf2 increases the number of intestinal tumors observed in Apc/Min mice and virtually 100% of imprint-free mice develop cancer. Of particular interest is the KvDMR1 ICR. This CpG-island like regulatory element, first described by the Higgins lab group 8 years ago, is located in the 10th intron of the Kcnq1 gene on human chromosome 11, is differentially methylated on the maternally derived chromosome, and is the promoter for the paternal-specific noncoding RNA (ncRNA), KCNQ1OT1.  This locus undergoes an epimutation (i.e. loss of maternal-specific methylation [LOM] and consequent biallelic expression of KCNQ1OT1) in Beckwith-Wiedemann syndrome (BWS), which is the most frequent alteration (genetic or epigenetic) in patients with this cancer predisposition condition. As a result of this pioneering work, methylation analysis of KvDMR1 is now performed world wide for the diagnosis of BWS.

Significantly several groups have shown that LOM at KvDMR1 is a negative predictor of cancer-risk in BWS patients, and as such can be used to help dictate the aggressiveness of therapy prescribed for these children.  Others have demonstrated LOM at KvDMR1 occurs in adult tumors including hepatocellular carcinoma, breast carcinomas, cervical, gastric, esophageal cancers, colon tumors and phaeochromocytomas.

To understand the function of the KvDMR1 locus, it was deleted in the mouse by the Higgins lab with the help of the RPCI Gene Targeting and Transgenics Resource. This mutant demonstrated that KvDMR1 regulates imprinted expression of 8 flanking genes by silencing their paternal alleles (Fitzpatrick et al., Nat Genet 2002; 32:426).  Generation of a mutant mouse with a truncation of the 60 kb Kcnq1ot1 transcript, reducing it to only 3 kb, provided proof that the ncRNA or its transcription is critical to imprinted gene silencing in this region. Furthermore, analysis of these mutant mice demonstrated that KvDMR1 silences the tumor suppressor gene Cdkn1c by two distinct tissue-specific mechanisms making this locus the first ICR shown to silence the same gene in more than one way (Shin et al., 2007; EMBO, In Press). Finally, this group has demonstrated that KvDMR1 is a multipartite ICR (i.e. separable promoter and repressive activities) and that it binds the insulator protein CTCF (Fitzpatrick et al., Mol Cell Biol 2007; 27:2636. 

With respect to BWS, LOM at KvDMR1 is associated with down-regulation of the tumor suppressor CDKN1C in human disease (Diaz-Meyer et al., J Med Genet 2003; 40:797); this same relationship has been demonstrated in esophageal cancer.  The Higgins group has also demonstrated that CDKN1C can be silenced in BWS by an additional methylation-independent mechanism, recruitment of repressive chromatin modifications (Diaz-Meyer et al., J Med Genet 2005; 42:648).

Recent evidence suggests that CTCF-modulated regulatory elements may be common in the mammalian genome and that deregulation of these loci is likely responsible for much of the aberrant gene expression in malignant cells. Thus, KvDMR1 serves as paradigm for the study of other imprinted and non-imprinted CTCF-mediated silencers/insulators as well as non-coding RNAs.

Gene-environment interactions involved in epigenetic silencing

To test the hypothesis that epigenetic gene silencing in cancer cells is the consequence of aberrant upstream transcriptional repression due to both genetic and environmental causes, Dr Sacchi and colleagues have been investigating the MTG8 gene which is deranged by the t(8;21) chromosome rearrangement of acute myelogenous leukemia. This gene and other members of the family have been shown to act as chromatin repressors (Rossetti et al., Genomics 2004; 84:1).

Fusion of either MTG8 or MTG16 with the transcription factor AML1, leads to chimeric proteins fusing transcription factor and repressor domains. AML1-MTG16 is a cytogenetic marker of therapy-related leukemia. Dr Sacchi’s group has also been studying whether genetic and environmental factors predisposing nuclear hormone receptor genes to epigenetic silencing in breast and prostate cancer cells. Her group has demonstrated that breast and prostate cancer cells showing resistance to retinoic acid (RA), the bioactive derivative of vitamin A, do not respond to RA-differentiation therapy in association with epigenetic changes in the nuclear RA-receptor beta 2, RARB2-  a known tumor suppressor.

Aberrant chromatin status of RARB2 now seems to be the prerequisite for DNA hypermethylation and heritable silencing of the RARB2 tumor suppressor activity. This work has several translational applications for breast and prostate cancer treatment, because RARB2 silencing can be reverted by using chromatin-remodeling drugs, including histone deacetylase inhibitors. Previous work of this lab demonstrated signs of epigenetic RA-resistance (measured as RARB2 hypermethylation) in ductal carcinoma in situ and phenotypically “normal” epithelial cells of the breast, thus showing that RARB2 hypermethylation is an early change of breast tumorigenesis.

Further, collaborative work with Johns Hopkins researchers has shown significant evidence of epigenetic RA-resistance in metastases of breast cancer to the brain, bone and lung (Mechrotra et al., Clin Cancer Res 2004; 10:3104).  Furthermore, RARB2 reactivation by the histone deacetylase inhibitor MS 275 and RA has been shown in preclinical models of prostate cancer (Qian et al., Prostate 2005; 64:20).