Because the somatic changes in DNA methylation that drive epigenetic gene silencing in cancer cells, although functionally equivalent to gene deletion, do not involve irretrievable loss of DNA sequence information, such changes present an attractive "rational" therapeutic target for cancer treatment and prevention. As such, several approaches to reversal of CpG island hypermethylation-associated gene silencing have emerged. One strategy, now in late-stage clinical development, features the use of inhibitors of DNMTs, such as 5-aza-cytidine (5-aza-C) (recently approved by the US Food and Drug Administration [FDA] for the treatment of myelodysplastic disorders ), 5-aza-deoxycytidine (5-aza-dC), zebularine, procainamide, or hydralazine, to reduce 5-mCpG density at the CpG island sequences in dividing cancer cells (66-69) (Fig. 6). Another approach, also under active clinical scrutiny, has been the use of inhibitors of histone deacetylases (HDACs), such as sodium phenylbutyrate, valproic acid, suberoylanilide hydroxamic acid, and many others, to limit the formation of repressive chromatin conformation near the genes carrying abnormally methylated CpG islands (70-73). Finally, the targeting of other CpG island methylation-associated chromatin regulators, such as the 5-meC-binding domain (MBD) family proteins, histone methyltransferases, and ATP-dependent remodeling enzymes, is in its earliest stages.
Nucleoside analog inhibitors of DNMTs, such as 5-aza-C and 5-aza-dC, have been widely used in attempts to reverse abnormal DNA methylation changes in cancer cells in vitro for the purpose of restoring "silenced" gene expression (74). Unfortunately, despite some apparent successes using preclinical models and some promising results in early clinical trials, the clinical usefulness of these compounds for cancer has not yet been fully realized, even though 5-aza-C (Vidaza®, Pharmion Corporation) has gained an FDA-approved indication for myelodysplastic disorders (65). In a phase II study (n = 14 men) of 5-aza-dC (decitabine) for androgen-independent, progressive, metastatic
prostate, with the drug given at a dose of 75 mg/m2 intravenously every 8 hours for three doses, and treatment cycles repeated every 5 to 8 weeks to allow for resolution of toxicity, 2 of 12 men evaluable for response had stable disease with a time to progression of more than 10 weeks (75). However, this may not be the best dose and dose-schedule for reversal of epigenetic gene silencing (74). One of the most significant limitations of the nucleoside analog DNMT inhibitors in clinical trials, especially at the doses and dose-schedules used thus far, has been treatment-associated side effects, such as myelo-toxicity with resultant neutropenia and thrombocytopenia, which are characteristic of other nucleoside analogs in general, including nucleoside analogs that are not DNMT inhibitors (64). Another concern regarding the use of nucleoside analogs as DNMT inhibitors has been that incorporation of the nucleoside analogs into genomic DNA might lead to mutations and/or cancer development (76).
Procainamide, a drug approved by the FDA for the treatment of cardiac arrhythmias, and hydralazine, a drug approved for the treatment of hypertension, are non-nucleoside analogs that both also seem to inhibit DNMT activity (77-79). Procainamide may be selective for DNMT1, the enzyme most consistently associated with tumorigenesis in genetic studies in mouse models; nucleoside analogs seem to inhibit each of the DNMTs (79). However, long-term use of procainamide (or of hydralazine) carries a risk of drug-induced lupus, more common in women than in men. In animal models, both 5-aza-C and procainamide seem to trigger autoimmunity, although whether or not autoimmunity is an unavoidable side effect of DNMT inhibition is not known (78,80). Finally, as mentioned earlier, mice carrying one disrupted Dnmtl allele and one hypomorphic Dnmtl allele, resulting in 10% of normal DNMT activity, have been reported to exhibit genomic instability and to develop T-cell lymphomas, hinting that therapeutic reductions in 5-mCpG dinucleotides might promote the appearance of certain cancers (e.g., lymphomas) while attenuating the appearance of others (e.g., epithelial tumors; see refs. 5 and 6). Thus, the clinical use of DNMT inhibitors is likely to be limited by both mechanism-based and mechanism-independent side effects.
Similar to DNMT inhibitors, HDAC inhibitors have also exhibited promising preclinical activity in cancer models. HDAC inhibitors under clinical development include sodium phenylbutyrate (and other butyrates), valproic acid, suberoylanilide hydroxamic acid, pyroxamide, N-acetyl dinaline (CI-994), and depsipeptide. However, the early clinical experience with these agents suggests that side
effects, such as nausea, vomiting, diarrhea, fatigue, edema, and so on, can occur, although severe adverse events seem rare (70,71,73). In addition to DNMT inhibitors and HDAC inhibitors administered as single agents, combinations of DNMT inhibitors and HDAC inhibitors also seem to have intriguing activity in preclinical models (24,81). Whether combinations of the currently available collection of DNMT inhibitors and HDAC inhibitors can reactivate silenced cancer genes, without unacceptable toxicity, in human clinical trials against prostate cancer, has not yet been determined.
Two MBD family proteins have been implicated in the silencing of critical genes in cancer cells carrying abnormally hypermethylated CpG island sequences, and may define pathways amenable to targeting for reversal of epigenetic gene silencing (Fig. 7). One of the MBD family proteins, MeCP2, contains an approx 70-amino acid minimal region that mediates selective binding to DNA containing 5-mCpG (an MBD motif), and a transcriptional repression domain that permits interaction with the transcriptional repressor Sin3 and associated HDACs (82). MeCP2 can, thus, act as a CpG island hypermethylation-dependent transcriptional repressor by binding transcriptional regulatory sequences carrying 5-mCpG and recruiting HDACs. For this reason, MeCP2-mediated inhibition of 5-mCpG-containing promoter activity can usually be alleviated by treatment with trichostatin A, an inhibitor of HDACs (82). Another MBD family protein, MBD2, which can also bind selectively to DNA containing 5-mCpG, has been found to be a component of a mega dalton (MD) transcription repression complex, MeCP1, that also contains the Mi-2/NuRD chromatin remodeling complex subunits MBD3, HDAC1, and HDAC2, the histone-binding proteins RbAp46 and RbAp48, the SWI/SNF helicase/ ATPase domain-containing protein Mi2, MTA2, and two uncharacterized polypeptides of 66- and 68-kDa (83). Although present in the Mi-2/NuRD complex, the MBD family protein MBD3 does not seem to recognize 5-mCpG-containing DNA. As a result, in the absence of MBD2, Mi-2/NuRD complexes, capable of catalyzing ATP-dependent chromatin remodeling, are incapable of selectively binding hypermethylated transcriptional regulatory sequences. In the MeCP1 complex, MBD2 acts to recruit the Mi-2/NuRD chromatin remodeling complex to 5-mCpG-containing DNA (83). Of interest, although MeCP2-mediated transcriptional repression can typically be alleviated by treatment with HDAC inhibitors, MeCP1-mediated inhibition of 5-mCpG-containing promoter activity is often not affected by HDAC inhibitor exposure.
MBD family proteins have not yet been systematically targeted for the reactivation of critical genes in cancer cells that have been silenced by CpG island hypermethylation. However, MBD family proteins, particularly MBD2, make attractive therapeutic targets for two key reasons. First, the anticipated efficacy of drugs that interfere with MBD family protein activity is provocative, and second, the likely side effects may be minimal. As for efficacy, MBD2 is a validated target for new drugs. MBD2 selectively binds the GSTP1 CpG island when it is methylated, as occurs ubiquitously and early during prostatic carcinogenesis, and siRNA-mediated reduction in MBD2 levels activates GSTP1 expression despite CpG island hypermethylation (31,84,85). Similarly, cells from Mbd2-/-mice are unable to repress transcription from exogenously hypermethylated promoters in transient transfection assays (86). Also, ApcMinl+Mbd2-/- mice develop far fewer intestinal adenomas, and survive longer, than do ApcMin/+Mbd2+/- or ApcMin/+Mbd2+/+ mice (87). As for toxicity, other than a maternal behavior defect, the Mbd2-/- mice seem fairly unremarkable, and have maintained normal gene imprinting, repression of endogenous retroviral sequences, and no obvious ectopic gene expression (86). In contrast, Dnmt1-/-, Dnmt3a-/-, and Dnmt3b-- mice are not viable. Furthermore, unlike mice with defective Dnmt1 genes, Mbd2-/- mice do not seem prone to increased lymphomagenesis when crossed to mice carrying disrupted p53 genes (88). These observations hint that drugs targeted at MBD2 might be able to reactivate epigenetically silenced genes in cancer cells, or in precancerous lesions, with a significant margin of safety. For this reason, chemical biology approaches, featuring high-throughput screening of chemical libraries, have been applied to the search for such compounds (X. Lin, Z. Reichardt, and W. G. Nelson, Personel Communication). Provocatively, antisense inhibitors directed at MBD2 mRNA have been reported to attenuate human cancer cell growth in vitro and in vivo (89,90).
To fully exploit the therapeutic opportunities presented by targeting epigenetic gene silencing in prostate cancer, both new drugs and new clinical trial strategies for developing and evaluating such treatments will be needed. Similar to other rational "targeted" treatments for human cancers, the expectation is that epigenetic modulation might be accomplished with chronic or prolonged treatments that have minimal side effects. If used to prevent cancers, the safety demands of such treatments will be even greater. For these reasons, early clinical studies will seek to ascertain the "optimal biologic dose" presumably a dose that restores the expression and function of epigenetically silenced genes, rather than the "maximally tolerated dose." Also, because combinations of inhibitors (e.g., DNMT inhibitor plus HDAC inhibitor) may be more effective at reactivating silenced genes than individual inhibitors, the design of clinical trials needed to establish dose and dose-schedule may prove very complex. Almost certainly, the new technologies permitting sensitive detection of epige-netically silenced genes, such as MS-PCR, will be required as pharmacodynamic end points for such clinical development strategies.
Was this article helpful?