1 Brain Tumor Research Center, Department of Neurological Surgery and UCSF Comprehensive Cancer Center, University of California, San Francisco, California, USA. jcostello@cc.ucsf.edu
2 Department of Radiation Oncology and the Winship Cancer Institute, Emory University, Atlanta, Georgia, USA. pvertin@emory.edu
The theory that DNA methylation controls tissue-specific gene expression has remained controversial despite being frequently cited for more than 25 years. A new study shows a pristine example of methylation and gene expression that is cell-type−restricted, although questions about causality persist.
Experimental results tend to be ambiguous, reflecting nature's indifference to scientific theories. Rarely does one see a result that is so black-and-white as to be untainted by interpretation. On page 175 of this issue, Bernard Futscher and colleagues1 report one of these rare instances. They address a two-decade−old theory, often cited but never proven, that methylation plays a role in tissue-specific gene expression. The theory strikes at the core of a fundamental biological question—what determines the gene expression pattern that uniquely defines different tissues?
Landmark publications in the 1970s expounded the growing suspicion that differential methylation of cytosines controls tissue-restricted gene expression during development and in differentiated adult tissues2,
3. Although methylation patterns are unequivocally different from tissue to tissue, the methylation of tissue-specific genes has generally been in the wrong place or at the wrong time, or more often, there is a complete lack of correlation with gene expression. More recent studies of all CpG motifs (most methylation occurs at cytosine residues within CpG dinucleotides) within defined promoters of tissue-specific genes have shown a similar lack of definitive correlation with gene expression4. The finding that tissue-specific genes previously suspected to be regulated by methylation are unaffected by demethylation in methyltransferase-deficient mouse embryos5 seemed to drive the final nail in the coffin. However, absence of evidence is not necessarily evidence of absence, and the study by Futscher et al.1 has now imbued this old theory with new life.
Spotlight on SERPINB5 It is not trivial that this long-awaited example involves SERPINB5 (encoding maspin). The gene SERPINB5 is a member of the serine protease inhibitor (serpin) gene cluster on chromosome 18q21.3 (refs 6,7). These proteins have a number of important functions, including the regulation of cell adhesion and differentiation. Maspin is unique in the serpin superfamily in that it is a serine protease substrate, rather than an inhibitor. Regulation of maspin is critical to normal development and the tissue-specific functions in the adult. For example, targeted overexpression of maspin in the mouse disrupts development of the mammary gland and inhibits that of lobular-alveolar structures during pregnancy8.
The implications of the study by Futscher et al.1 extend beyond normal gene regulation into clinically relevant aspects of cancer. SERPINB5 was first discovered as a gene that is downregulated in breast-cancer cells6, and in mouse models, it is a potent tumor suppressor that inhibits cell motility, invasion, angiogenesis and metastasis7. Loss of maspin expression correlates with increased metastatic potential in breast-cancer and other human tumors7.
Earlier work by Domann et al.9 showed that silencing of SERPINB5 in breast cancer cells is associated with aberrant methylation. The first clue that methylation might also be important in the normal expression of SERPINB5 stemmed from the observation that 'control' peripheral blood lymphocytes also have dense methylation of the SERPINB5 promoter9.
Maspin is expressed in a cell-type−restricted manner and is expressed in epithelial cells of the airway, breast, skin and prostate, but not in skin fibroblasts, lymphocytes, bone marrow, heart and kidney1. At the transcriptional level, SERPINB5 is regulated by AP-1 and p53-binding sites, and can be negatively regulated by a hormone-responsive element recognized by the androgen receptor7. Given that these proteins are ubiquitous in their expression, the question arises: what dictates the cell-type−specificity of maspin expression?
Methylation does it Futscher et al.1 use primary human cultures to show that cell-type−specific expression of maspin is inversely correlated with methylation of the SERPINB5 promoter. Not only are the critical AP1 and p53 promoter elements differentially methylated, but so is the entire CpG-rich island surrounding the promoter and first exon of the gene, and it is strictly all-or-none (see figure). They further show that SERPINB5 expression can be reactivated by 5-aza-2'-deoxycytidine treatment of immortalized fibroblasts, suggesting that methylation is the only impediment to SERPINB5 expression in a non-expressing tissue. Taken together, these results satisfy two criteria for a role for methylation in tissue-specific gene repression—its promoter is densely methylated in non-expressing tissues and induced demethylation results in ectopic activation of the gene.
Model of the cell-type−specific control of SERPINB5 expression by methylation. The epithelial cells in skin have an unmethylated SERPINB5 promoter that is occupied by the transcriptional regulators AP1 and p53. In addition, the histones (blue) are acetylated (Ac), thus limiting histone−histone interactions and providing an open chromatin structure that is required for SERPINB5 expression. In contrast, the promoter in skin fibroblasts is completely methylated (CH3), associated with hypoacetylated histones and adopts an inaccessible transcriptionally inactive state. Methylation of DNA allows the binding of methyl CpG−binding proteins (MeCP), which can attract histone deactylases (HDAC) and chromatin remodelling complexes to direct a local change in chromatin organization. In this model, methylation is a primary impediment to SERPINB5 expression and thus determines the cell type−specificity.
One issue raised by this study is the presumed methylation status of CpG islands in normal tissues. CpG islands are regions of DNA characterized by an unusually high C+G content and CpG frequency relative to the remainder of the genome10,
11. Most CpG islands encompass the promoter and first exons of genes and are found in nearly all housekeeping genes and more than 50% of tissue-restricted genes. There is a widely held assumption that CpG islands are unmethylated in all normal tissues, regardless of expression status. Well-documented exceptions include genes on the inactive X chromosome and those subject to parental imprinting. In these cases, CpG island methylation is a developmentally regulated event that occurs in the early embryo and is necessary to maintain repression in differentiated cells of the adult12,
13.
Is it possible that CpG island methylation has a similar role in restricting the expression of some tissue-specific genes? Albeit smaller and less CpG-rich than some CpG islands, the promoter of the maspin gene qualifies as a CpG island based on the established definitions11,
14; yet, it is heavily methylated in normal human cell types that lack maspin expression. And, as with other genes that undergo permanent methylation-dependent repression (for example, genes on the inactive X chromosome, imprinted genes, and aberrantly methylated genes in cancer), Futscher et al. show that, when methylated, the SERPINB5 promoter is in a transcriptionally repressive chromatin conformation (see figure)1. The SERPINB5 methylation pattern is not entirely without precedent. De Smet et al.15 have shown that the MAGE genes, which also have promoters of intermediate CpG-richness, are unmethylated and expressed in male germ cells but are heavily methylated and silent in somatic tissues. Importantly, ectopic activation of MAGE occurs in human melanomas and is associated with hypomethylation of the promoter.
To be continued The maspin example underscores the necessity of distinguishing between normal methylation cues and abnormal ones. Although SERPINB5 expression is masked by methylation in many normal cell types, the promoter is unmethylated in breast epithelial cells. Thus, the methylation and silencing of maspin in human breast cancers, which are primarily derived from epithelial cells, would clearly be an aberrant event. Perhaps cancer cells have inappropriately activated a silencing mechanism that is otherwise operative only during a particular stage in cellular differentiation.
Cautious scientists must agree that there is much work to be done before we will come to know the full impact of this study. Because it necessarily relied on primary human cells in short-term culture, the results must be confirmed in vivo. It will also be of interest to determine at what developmental stage the cell-type−specific methylation of SERPINB5 is set, and whether it establishes silencing or ensures the propagation of a stably repressed state. Finally, is SERPINB5 unique or are there other genes that rely on methylation as a means of restricting expression to different tissues or cell types? If such studies support the present findings, the publication by Futscher et al.1 will be inextricably linked across two decades to one of the most influential and tenacious theories in the field.
