Connections between Epigenetic Gene Silencing and Human Disease (original) (raw)

. Author manuscript; available in PMC: 2007 Jun 17.

Abstract

Alterations in epigenetic gene regulation are associated with human disease. Here, we discuss connections between DNA methylation and histone methylation, providing examples in which defects in these processes are linked with disease. Mutations in genes encoding DNA methyltransferases and proteins that bind methylated cytosine residues cause changes in gene expression and alterations in the patterns of DNA methylation. These changes are associated with cancer and congenital diseases due to defects in imprinting. Gene silencing is also controlled through histone methylation. Altered levels of methyltransferases that modify lysine 27 of histone H3 (K27H3) and lysine 9 of histone H3 (K9H3) correlate with changes in Rb signaling and disruption of the cell cycle in cancer cells. The K27H3 mark recruits a Polycomb complex that is involved in regulating stem cell pluripotency, silencing of developmentally regulated genes, and controlling cancer progression. The K9H3 methyl mark recruits HP1, a structural protein that plays a role in heterochromatin formation, gene silencing, and viral latency. Cells exhibiting altered levels of HP1 are predicted to show a loss of silencing at genes regulating cancer progression. Gene silencing through K27H3 and K9H3 can involve histone deacetylation and DNA methylation, suggesting cross talk between epigenetic silencing systems through direct interactions among the various players. The reversible nature of these epigenetic modifications offers therapeutic possibilities for a wide spectrum of disease.

Keywords: chromatin, DNA methylation, epigenetic gene silencing, histone methylation, human disease

1. DNA methylation

Altered gene expression can play a causal role in human disease. In many cases, altered expression results from genetic lesions within the gene or regulatory sequences. However, in some cases genetic lesions are absent from the gene. In such instances, aberrant epigenetic modifications of the chromatin surrounding the gene are the cause of altered expression. There are two major epigenetic gene silencing mechanisms that account for a growing number of diseases: cytosine DNA methylation and covalent histone modification.

The 5' cytosine of CpG dinucleotides within mammalian genomes can be methylated by de novo DNA methyltransferases such as DNMT3A and DNMT3B [1]. Maintenance of DNA methylation is performed by DNMT1, utilizing hemimethylated DNA as a substrate. This provides a mechanism to propagate the epigenetic mark following DNA replication. The methyl groups serve as docking sites for gene silencing proteins [2]. In general, DNA methylation correlates with increased chromatin condensation and gene silencing [1].

There are several ways in which altered patterns of DNA methylation lead to disease (Table 1). CpG dinucleotides are generally methylated in normal cells, with the exception of hypomethylation at CpG “islands” located upstream of many active genes [3]. In contrast, cancer cells exhibit a global hypomethylation and CpG island hypermethylation [3]. This shift in the pattern of DNA methylation frequently results in inappropriate silencing of genes, especially tumor suppressor genes, leading to numerous types of cancer. For example, expression of the serine protease inhibitor family member maspin is reduced due to methylation of promoter sequences in many advanced forms of cancer [4-6].

Table 1.

Gene silencing proteins and disease.

Protein Cellular Defect / Disease References
DNMT1 Developmental abnormalities [110-112]
Igf2 imprinting [9]
Colon cancer [113-115]
Lymphoma [116-118]
Pancreatic cancer [119]
DNMT3B Developmental abnormalities [120]
ICF [1,121-123]
Bladder cancer [124]
Breast cancer [124,125*]
Colon cancer [124]
Hepatocellular carcinoma [126*,127*]
Lung cancer [124,128*,129*]
MeCP2 Chromosome instability/ cell cycle defects [49,54,55,57, 130]
Breast cancer [131]
Rett syndrome [132, 133], RETTBase
EZH2 Cell cycle defects [35-37]
Barrett's esophagus [134]
Bladder cancer [135,136]
Breast cancer [41,42,45,137]
Colorectal cancer [138]
Melanoma [137]
Myeloma/ lymphoma [29,43] [44,46,47,139-141]
Hepatocellular carcinoma [142]
Prostate cancer [34,39,40,137]
Wilms tumor [143]
Suv39h1 Blood cell defects (RBC and WBC) [54,56,59]
Chromosome instability/ cell cycle defects [49,54,55,57] [130]
Chromosome instability [68-73,75-77]
HP1 Breast cancer [78,81,82,144]
Medulloblastoma [86]
Papillary thyroid carcinoma [85]
Viral latency [87-92]

Alterations in methylation patterns are responsible for several congenital diseases that affect growth through the misregulation of imprinted genes. Mammalian genomes contain dispersed clusters of genes in which the expression state of each allele is determined by the parent of origin [7]. Transcription within these clusters is regulated by Imprinting Centers (ICRs), DNA regions that are typically 1-2 kb in size and enriched with CpG dinucleotides [8]. ICRs exhibit allele-specific DNA methylation and histone modifications. An ICR positioned between the insulin-like growth factor IGF2 and H19 genes is methylated only on the paternal allele, presumably by DNMT1 [9]. This methylation blocks the association of the zinc finger protein CTCF [10,11]. On the unmethylated maternally derived allele, ICR is bound by CTCF, which functions as an insulator by blocking interactions between IGF2 enhancers located upstream of the H19 gene. Altered expression of IGF2 due to changes in the imprinted status at ICR result in two diseases with different clinical characteristics, Beckwith-Weidemann Syndrome (BWS, OMIM I30650) and Silver-Russel Syndrome (SRS, OMIM 180860) [7] (Table 1). BWS is primarily identified by macroglossia, umbilical abnormalities and gigantism. In a subset of BWS individuals, DNA methylation at the ICR occurs on both the maternal and paternal alleles, resulting in loss of H19 expression and activation of IGF2 on both alleles. SRS is identified by low birth weight, slow postnatal growth, characteristic facial abnormalities and body asymmetry. In a subset of SRS individuals, DNA methylation does not occur within the ICR on either allele, resulting in the expression of H19 from both alleles and complete loss of IGF2 expression.

In addition to alterations in the patterns of DNA methylation, loss of DNA methyl transferase also leads to disease (Table 1). Immunodeficiency-centromeric instability-facial anomalies (ICF, OMIM 24860) is an autosomal recessive disorder caused by defects within the catalytic domain of DNMT3B [1]. Phenotypes of ICF include instability of chromosomes 1, 9 and 16, which are enriched for pericentric satellite II and III sequences containing CpG dinucleotides. In addition, ICF is associated with immune system malfunctions, facial abnormalities, and short life expectancy. Given that ICF individuals lack the function of a de novo DNA methyltransferase, it was anticipated that changes in patterns of DNA methylation would lead to alterations in gene expression. Expression profiling studies showed alterations in gene expression as expected, however, changes in the DNA methylation patterns at these genes were minor [1]. In contrast, the satellite sequences were found to be hypomethylated [12]. Thus, methylation of repetitive sequences, such as the satellites, might be important for establishing the spatial positioning of chromosomes within the nucleus that is necessary for proper gene regulation [13].

Given that mutations within DNA methyltransferase genes are associated with disease, it follows that mutations within genes encoding proteins that bind to methylated cytosines also result in disease. MeCP2 (methyl-CpG-binding protein 2) is a member of a class of DNA methyl binding proteins (MBDs) that specifically recognize methylated cytosine residues [14]. These binding proteins function by recruiting histone deacetylases (HDACs) to silence target genes. Mutations in the X-linked gene encoding MeCP2 are responsible for approximately 95% of classic Rett syndrome cases (RTT, OMIM 312750) [15] (Table 1). RTT is the second most common form of mental retardation in females, estimated to affect 1 in 10,000 females. Overt disease phenotypes are not observed until after 6 – 8 months of age, at which time mental abnormalities and motor function impairment become obvious. The delayed onset of the syndrome suggests problems with neuronal differentiation. One curious aspect of the disease is that MeCP2 is broadly expressed, yet only neuronal function appears to be affected. Thus, it was hypothesized that loss of MeCP2 would result in alterations of gene expression within the brain. To address this hypothesis, transcriptional profiling experiments were performed using brain tissue from a Mecp2 knock-out mouse model in comparison to control mice [16]. Surprisingly, very subtle alterations in brain-specific gene expression were observed. However, changes in gene expression within a subset of brain cells might not have been detected, as a mixture of cell types was analyzed. The lack of global changes in gene expression might also be due to redundancy in function among the DNA methyl binding proteins as suggested by MBD knock-out studies [14]. Thus, the phenotypic consequences resulting from RTT might be due to subtle changes in gene expression or mis-regulation of gene expression in a limited number of cells within the brain. To date, several MeCP2 target genes have been identified, including brain-derived neurotrophic factor (Bdnf) [15]. Interestingly, BDNF protein levels are decreased in the Mecp2 mouse knock-out model, rather than an anticipated increase if MeCP2 were playing a silencing role. This decrease could reflect complex issues related to neuronal differentiation [17]. Satisfying features that support Bdnf as a key gene involved in the RTT phenotype are a role in neural differentiation, a knock-out mouse that exhibits overlapping phenotypes with RTT syndrome mouse models, and the fact that over-expression of BDNF in the forebrain rescues some of the phenotypes associated with the Mecp2 mutant mice [17]. While these data argue for a connection between MeCP2 and BDPF, questions as to the mechanistic defects associated with RTT remain unresolved [18]. Complicating this issue, MeCP2 appears to play a role in mRNA splicing, through interactions with the Y box binding protein [19], suggesting post-transcriptional events could contribute to the RTT phenotype. Collectively, it is clear that defects in multiple steps of the DNA methylation pathway cause disease, however, the molecular defects that contribute to disease development remain to be elucidated.

2. Histone methylation

Modifications such as phosphorylation, acetylation and methylation frequently occur on histones tails that extend from the nucleosome core [20]. These modifications serve to alter charge interactions of the histone tails with DNA, thereby influencing chromatin packaging. In addition, these modifications serve as binding sites for specific factors that “read” a proposed histone code [21]. In most cases, specific modifications correlate with biological functions such as chromatin condensation, transcriptional regulation and DNA replication. Disruption of the epigenetic modifications on histones throughout the genome is a universal feature observed in cancer cells [22]. Here, we will focus on connections between two histone methyltransferases involved in gene silencing and cancer.

2.1 E(z) and disease

Enhancer of Zeste (EZH2) is one of two mammalian homologues of Drosophila E(Z), a SET domain protein that methylates K27H3 [23]. E(Z) is a Polycomb group (PcG) protein involved in homeotic gene repression in Drosophila [24]. Two Polycomb complexes have been elucidated, Polycomb Repressor Complex 1 and 2 (PRC1 and PRC2) [25]. While components differ among species, the core PRC1 components in mammals are BMI-1, Ring-1, HPH and HPC. The core PRC2 components are EED, EZH2, SU(Z)12. PRC2 is recruited to target genes for the initiation of silencing. PRC2 associates with Type I HDACs, such as HDAC1 and 2, through interactions with EED [26] (Figure 1a). In addition, immunoprecipitation of EZH2 and EED show interactions with DNMT1, 3A and 3B, suggesting a link between histone methylation and DNA methylation [27] (Figure 1a). This was supported by experiments in which cells treated with 5'-aza-deoxycytidine or RNAi knock-down of any of the DNMTs, showed loss of silencing at EZH2 target genes [27]. Following PRC2 silencing, PRC1 is recruited for maintenance of the silent state. Collectively, these data support a model whereby Polycomb silencing occurs through histone deacetylation, subsequent histone methylation and DNA methylation initiated by PRC2, and maintained by PRC1[28] (Figure 1a).

Figure 1.

Figure 1

Diagram showing the connections between epigenetic gene silencing factors. (a) Gene silencing associated with K27H3. HDACs deacetylate nucleosomes, allowing for methylation of K27H3 by EZH2, providing a bindings site for Polycomb. Components of Polycomb complexes interact with DMNTs involved in methylating cytosine residues (small filled circles). Lines indicate interactions between the gene silencing factors. (b) Gene silencing associated with K9H3 methylation. HDACs deacetylate nucleosomes, allowing for methylation of K9H3 by SUV39h, providing a binding site for HP1. DNA methyltransferases modify cytosine residues (small filled circles) to repress gene expression. Lines indicate interactions between the gene silencing factors. In cancer cells hypermethylation typically occurs within CpG islands of promoter regions, while isolated CpGs at other locations are hypomethylated (small open circles).

Homozygous EZH2 knock-out mice die during embryogenesis, while tissue-specific knock-out of EZH2 caused defects in B cell maturation [29,30]. A role for EZH2 in developmentally regulated gene expression has been corroborated by chromatin immunoprecipitation experiments coupled with microarray analysis [31]. PRC1 and PRC2 components were mainly associated with silenced genes involved in differentiation in embryonic stem cells; upon differentiation, epigenetic marks associated with these genes changed [31,32]. Interestingly, both methylated K4 (an activating mark) and K27 (a silencing mark) were found in “bivalent domains” of the promoters of the embryonic silenced genes encoding transcription factors that govern pluripotency [32,33]. This duality of epigenetic marking was resolved upon differentiation, as the appropriate expression or silencing marks on cell lineage genes were retained upon commitment to a particular cellular fate. Surprisingly, this combination of marks results in silencing similar to that mediated by methylated K27 in differentiated cells. This suggests that Polycomb silencing is initiated in embryogenesis, and that retention of the epigenetic mark upon differentiation results in a continuation of silencing later in development [32]. Perhaps it is not surprising that defects of this gene silencing system within differentiated cells (see below) correlate with a poorer prognosis for many cancers, as loss of silencing can provide increased plasticity for cancer cells.

Levels of EZH2 are important for control of cell cycle regulation, with increased levels leading to cancer [34]. One mechanism for controlling EZH2 levels is through transcriptional regulation by p53 [35]. EZH2 and Rb potentially compete for the binding of HDAC1, a component in E2F-mediated gene silencing [36]. Activated p53 silences EZH2 transcription, allowing for repression of E2F-regulated cell cycle genes by Rb through HDACs [37]. In general, EZH2 up-regulation through increased transcription or gene amplification correlates with cancer [38]. EZH2 over-expression is observed in many types of cancer including prostate, breast, and lymphomas (Table 1). In prostate cancer, EZH2 is up-regulated with metastasis, while EED levels remain unchanged [39]. Cellular proliferation is dependent on the SET domain of EZH2, and RNAi knock-down of EZH2 led to cellular growth inhibition and arrest in G2/M. In fact, a combination of high EZH2 expression with moderate to low E-cadherin is one of the best markers for prostate cancer progression [38,40].

Similar data for EZH2 exists for breast cancer and lymphoma (Table 1), where increased levels of expression correlate with tumor proliferation and poor prognosis. Importantly, the methyltransferase activity of the SET domain plays an important role in cell invasion [41-44]. In normal, differentiated cells an almost mutually exclusive expression pattern exists for BMI-1 (a component of PRC1) and EZH2 (a component of PRC2). It appears that BMI-1 and EZH2 are only highly co-expressed in cancer cells [45,46]. While the pattern may be slightly different in the germinal center, in general, B cells lack EzH2 expression and highly expressed BMI-1 in resting or well-differentiated cells. In cycling or poorly differentiated cells the opposite is true [46,47]. One possible model generated by these data is that PRC1 and PRC2 exchange silencing duties when cells enter and exit the proliferation stage during differentiation and cancer formation. These transitions are accomplished through deacetylation of histones at target genes via HDACs, recruitment of PRC2 with subsequent histone and DNA methylation, with final localization of PRC1 to maintain the silenced state [28,47,48]. This model is supported by studies in Drosophila and may be a general paradigm of epigenetic regulation by Polycomb [25].

2.2 SUV39h and disease

The SUV39h1 and SUV39h2 proteins are homologues of Drosophila SU(VAR)3-9, a SET domain histone methyltransferase involved in heterochromatin formation and gene silencing [49]. In mice, these two proteins are co-expressed during embryogenesis, but are differentially expressed in the adult, where SUV39h1 is present in multiple tissues and SUV39h2 is restricted to the testes [50]. The methylation of K9H3 provides a binding site for Heterochromatin Protein 1 (HP1), a structural protein enriched in heterochromatin (see below) [51,52] (Figure 1b). Loss of both SUV39h1 and SUV39h2 or HP1 leads to reduction of methylation at K20H4, a conserved hallmark of heterochromatin, suggesting sequential histone modifications occur in the process of heterochromatin formation [53].

The role of SUV39h proteins in development has been studied using mouse knock-out models. Mice lacking either Suv39h1 or Su39vh2 showed normal viability and fertility, suggesting redundancy of function between the two histone methyltransferases during the embryonic stage [54]. In Suv39h1 and Suv39h2 double knock-out mice, decreased viability was observed with death at embryonic day 12.5. Surviving mice exhibited slow growth and developed B cell lymphomas similar to non-Hodgkin's lymphoma [54]. In addition, males were infertile with a delay in meiosis. At the cellular level, chromosomal segregation defects and abnormally long telomeres containing reduced levels of trimethylated K9H3 and HP1 were observed. Similar phenotypes were apparent in mice upon over-expression of SUV39h1, including slow growth, cell cycle progression defects, chromosome missegregation, and mislocalization of HP1 [55,56]. Taken together, these studies demonstrate a critical role for proper levels of SUV39h1 in chromosome dynamics and development.

On a cellular level, increased K9H3 methylation is observed in differentiated, but not cycling cells [57]. In addition, over-expression of SUV39h1 in red blood cell progenitors caused abnormal cell cycle profiles and immortalization similar to that caused by loss of p53 [56]. Interestingly, p53 loss is normally accompanied by changes in karyotype. Despite the fact that SUV39h has been implicated in chromosome segregation, immortalized Suv39h1 over-expressing cells possessed normal karyotypes. However, these cells did show changes in the Retinoblastoma protein (Rb) signaling pathway, including down-regulation of p21 and up-regulation of Rb, accompanied by phosphorylation [56]. Rb is involved in transcriptional control of E2F-regulated cell cycle genes. Rb binds to SUV39h1 and HP1 to mediate gene silencing of target genes [58]. Importantly, Rb mutants that disrupt interactions with or cause mislocalization of SUV39h1 give rise to cancer [59].

Rb is regulated by oncogenic Ras [60]. Suv39h1 knock-out mice over-expressing a constitutively active oncogenic form of Ras in blood progenitor cells developed lymphoma [59]. Analysis of cells from these mice showed increased levels of p16INK4a, consistent with activation of the Rb pathway [60]. Interestingly, loss of SUV39h1 appeared to bypass the requirement for the loss of p53 in the formation of lymphoma, without evidence of aneuploidy due to chromosomal segregation defects. These results broadly suggest an anti-oncogenic role for SUV39h1 as an early barrier in the prevention of cancer in cases of unregulated Ras signaling [59]. Simultaneous treatment of cells expressing oncogenic Ras with the histone deacetylase inhibitor TSA and the DNA demethylation inhibitor 5'-aza-deoxycytodine resulted in increased cancer and early death, similar to the effects caused by the loss of SUV39h1 [59]. These findings suggest a connection between histone modifications and DNA methylation (Figure 1b), and suggest the possibility of adverse outcomes upon treatment with such inhibitors.

3. HP1 and disease

One function of K9H3 methylation is to serve as a binding site for HP1 (Figure 1b). HP1 is conserved among species, with mice and humans each possessing three genes encoding HP1-like proteins [61]. In humans these are referred to as HP1Hsα, HP1Hsβ and HP1Hsγ. These proteins share significant amino acid sequence identity, yet have distinct chromosomal localization patterns [61-63]. HP1 proteins contain a chromo domain that binds methylated K9H3, and a chromo shadow domain that dimerizes [64]. Dimerization establishes a platform in which nuclear proteins containing the PxVxL pentapeptide motif interact [65]. Association of HP1 with a target gene causes alterations in chromatin structure and gene silencing by mechanisms that are not completely understood [66,67].

Understanding the function of HP1 in chromosome dynamics and gene regulation has implications for determining disease mechanisms. A reduction in HP1 levels causes kinetochore defects, loss of chromosome cohesion and condensation, as well as aberrant chromosome segregation [68-73]. In addition to these defects associated with centromere function, telomere function is also impaired. In Drosophila, telomeric fusions are observed in mutants lacking HP1 [74]. In mammals, over-expression of HP1Hsα or HP1Hsβ, but not HP1Hsγ, results in telomere fusions and telomere shortening, due to reduced interactions with hTERT, the catalytic component of telomerase [75-77]. Thus, modulation of HP1 levels could contribute to aneuploidy and telomere fusion events that occur during cancer progression.

Taking into account the multiple roles HP1 proteins play in many aspects of nuclear function, it is easy to understand how a reduction in HP1 could impact processes related to cell growth and cancer progression. Connections between HP1 and breast cancer were derived from studies comparing breast cancer cell lines with different invasive/metastatic potential [78]. Invasion properties are based on an in vitro assay [79], whereas the ability to metastasize is based on behavior of cells following injection into an immune compromised mouse [80]. HP1Hsα was found to be down-regulated in highly invasive/metastatic cells compared with poorly invasive/non-metastatic breast cancer cells and metastasis tissue from breast cancer patients [81] (Table 1). A causal role in invasion has been demonstrated by modulation of HP1Hsα levels [82]. Increased levels of HP1Hsα in highly invasive/metastatic cells decreased invasion. In contrast, knock-down of HP1Hsα with RNAi in poorly invasive/non-metastatic cells increased invasion. Alterations in invasion were not accompanied by changes in cell growth rate [82]. However changes in gene expression were observed, suggesting HP1Hsα regulates genes required for an invasive phenotype (T.J.M and L.L.W, unpublished data). Collectively, these findings suggest that HP1Hsα is a member of a small class of proteins known as metastasis suppressors. Unlike tumor suppressors that alter growth of a primary tumor, metastasis suppressors regulate metastasis without affecting cancer cell growth [83,84]. Studies in which HP1Hsα levels are altered in human breast cancer cells and assayed for metastasis in a mouse model are needed to determine whether HP1Hsα is a metastasis suppressor. In addition to changes in the levels of HP1Hsα in breast cancer cells, decreased HP1Hsα correlates with tumor progression in multiple human cancers. HP1 _Hs_α mRNA levels are reduced in advanced forms of papillary thyroid carcinoma [85]. HP1Hsα mRNA is also down-regulated in medulloblastoma; this down-regulation correlates with treatment failure [86]. Thus, HP1Hsα might play a role in the progression of many types of cancer.

In addition to the connections between HP1 and cancer, viruses have usurped HP1 for functions related to replication within host cells (Table 1). In virally permissive cells, human cytomegalovirus lytic genes are expressed; in non-permissive cells HP1Hsβ association with viral promoter sequences and correlates with viral latency [87]. Upon differentiation into macrophages, HP1Hsβ association is lost, and gene activation occurs [87-89]. Similarly, in the case of Kaposi's sarcoma-associated herpes virus, an interaction of the viral latency-associated nuclear antigen with exogenously supplied mouse HP1α and HP1β caused transcriptional inhibition [90]. For HIV-1, transcriptional inhibition occurs by a different mechanism in which a complex consisting of CTIP2 and HP1 sequesters the viral _trans_-activator Tat [91]. In addition to these examples in which HP1 governs viral transcriptional control, the polyoma JC virus has co-opted HP1 by an independent mechanism. The viral encoded protein Agno subverts the association of HP1 with the nuclear envelope protein lamin B receptor (LBR) causing increased nuclear envelope flexibility and more efficient escape of viral particles [92]. Thus, viruses utilize HP1 for both transcriptional control and regulation of nuclear architecture.

4. Connections among epigenetic gene silencing systems

Through the course of investigating the role of epigenetic modifications involved in disease, connections among DNA methylation, histone acetylation and histone methylation have become apparent (Figure 1). Studies in model organisms have specifically revealed a connection between DNA methylation and histone methylation. In Neurospora, a screen for mutants that lacked CpG DNA methylation identified a K9H3 histone methyltransferase [93]. In Arabidopsis, mutants lacking a histone H3 methyltransferase show altered CpNpG methylation and reactivation of endogenous retrotransposons [94,95]. The reverse situation can also exist in which methylated CpG sites in Arabidopsis direct sites of K9H3 methylation [96-98]. Similarly, in mammals, methylation of CpG islands within pericentric repetitive DNA elements is directed by histone methylation [99-102]. Conversely, in some instances DNA methylation precedes histone modifications, as methyl-CpG binding proteins such as MeCP2 recruit histone deacetylases and histone methyltransferases [103-105]. In fact, HDACs interact with many factors involved in epigenetic silencing [106] (Figure 1). Thus, the interplay between DNA methylation and histone modifications appears to serve in reinforcing the transcriptionally silent state.

5. Dynamics of epigenetic modifications and therapy

The dynamic nature of epigenetic gene regulation is important to consider in the context of disease. Both the loss and gain of gene silencing at target genes can potentially be reversed through drug treatment [106]. Drugs that inhibit DNA methylation can reactivate silenced genes in cancer cells, possibly re-establishing cell cycle control [107]. Drugs that inhibit histone deacetylation block cell cycle progression and cause apoptosis by unknown mechanisms [107]. While HDAC inhibitors alter the abundance of acetylated histones, other possible targets that could account for the observed affects include acetylated chromatin proteins such as E2F and p53 that regulate cell cycle. This network of epigenetic modifications might favor transient drug treatment, as initial reversal of the transcriptional status of a gene could be propagated by other epigenetic modifiers in the absence of prolonged drug treatment. Alternatively, the connections among epigenetic modifications might hinder successful drug treatment. Evidence for this comes from a recent study in which drug treatment resulted in gene activation, yet a subset of the epigenetic gene silencing marks remained [108]. Such marks could serve as memory to seed subsequent silencing events in the absence of continual drug treatment. The recent discovery of histone demethylases provides additional drug targets for reversing the transcriptionally silent state [109]. A combination of drugs that alter DNA methylation, histone acetylation, and histone methylation might ultimately be the most effective means of combating disease.

Acknowledgements

We apologize to the many investigators whose research could not be cited due to space limitations. We would like to thank Al Klingelhutz and members of the Wallrath lab for comments on the manuscript ,and Judith Kassis for discussions. Research is supported by an NIH grant (GM61513) to L.L.W., a grant from the Department of Defense Breast Cancer Research Foundation (DAMD17-02-1-0424) to L.L.W. and a Susan G. Komen Dissertation Research Award (DISS0403121) to T.J.M.

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References