DCIS is suspected to be a direct, although not obligate, precursor of invasive breast cancer, and aberrant DNA methylation is believed to play a crucial role in breast tumorigenesis. Considering that epigenetic changes often become apparent in early phases of the disease, we speculated that the identification of DCIS-specific methylated biomarkers might be crucial to elucidate the molecular mechanisms underlying the initiation and development of breast cancer and to conceive effective strategies for early diagnosis.
To acquire valuable information into the epigenetic switches that may promote and/or contribute to the initial neoplastic events, we have analyzed the DNA methylation profile of DCIS, on a MIRA-based CpG island microarray platform. This novel and sensitive genome-wide screening approach has led to the identification of 108 CpG islands that display aberrant levels of DNA methylation in early breast lesions. Of the 81 CpG islands associated with known genes, only 37 map to promoter regions or further upstream (46%). In agreement with recently published data [40
], more than one-half of the methylated CpG islands in normal genomes fall within the body of the gene or in downstream regions. The functional role of these intergenic and intragenic CpG-rich elements remains obscure, but it has been suggested that they may constitute short independent transcriptional units [40
Several targets have been examined by conventional bisulfite methods and found to be differentially methylated in multiple breast tumors, in complete agreement with the MIRA results. Most importantly, these gene candidates display methylation frequencies ranging from 50% to 83% in DCIS and up to 93% in stage I breast cancer, depending on the target gene, and these candidates hold great promise, alone or in combination, for future diagnostic applications. We were also able to identify several genes, such as CDKN2A
, and WT1
(Table ) well known to be methylated and transcriptionally silent in breast cancer [10
]. Many other 'conventional' methylation markers, however, were not represented in our data set. This apparent discrepancy with previous reports can be ascribed, in part, to the different conditions utilized in microarray data analysis to define thresholds; consequently, some genes may be classified as false negative simply because they fall below the statistical cutoff points. Digestion of genomic DNA, prior to the MIRA pull down, with the methylation-sensitive endonuclease Hha
I can also be crucial in excluding those CpG islands that are not methylated at Hha
I restriction sites (5'-GCGC), although they are methylated at surrounding CpGs within the target sequence. In addition, genes previously described as methylated in advanced-stage breast cancer may not be methylated in DCIS.
Surprisingly, most of the targets identified in the present study have never been linked to epigenetic errors during breast carcinogenesis and may shed new light into the molecular mechanisms underlying the insurgence of breast cancer. The employment of undissected breast tissue that fails to discern the epigenetic contribution of the single cell subtypes and the restricted number of DCIS used in the microarray analysis, however, may represent important limitations to this study and need to be kept in consideration.
Apart from the potential discovery of novel tumor suppressor genes and/or methylation biomarkers, relevant for a better comprehension and management of the disease, the present study has uncovered a broad epigenetic phenomenon that occurs at the onset of breast cancer development. Interestingly, we found that 32% of the total hypermethylated CpG islands (26 out of the annotated 81 hits) are associated with members of multiple homeobox gene subfamilies – a surprisingly high percentage considering that, so far, only ~300 homeobox genes have been identified in the human genome (roughly 1% of the presumed battery of protein-coding genes) [41
]. CpG methylation of homeobox genes has been sporadically observed during breast tumorigenesis; that is, methylation of the HOXB13
] and members of the HOXA cluster [42
]. The extent and the recurrence of this epigenetic event in mammary carcinoma, however, have never been emphasized until now. Robust and frequent methylation of homeobox genes is not restricted to breast cancer, and occurs at significant frequencies (~10% to 20% of all methylated genes) in early-stage lung carcinoma [28
] – suggesting a common epigenetic pathway involving the homeobox gene network. Yet, the diverse and nonidentical methylation spectra exhibited by DCIS and stage I lung cancer at homeobox gene-associated CpG islands cautions against the existence of a common epigenetic phenotype among different tumor types. This is not surprising since the pattern and function of the homeobox gene networks are exclusive for a particular tissue [44
] and no specific expression and/or CpG-island methylation signatures across tumors have so far been reported.
We cannot deduce why homeobox genes become preferential targets of aberrant CpG methylation during breast tumorigenesis and whether this extensive methylation can shift their finely tuned homeostasis, thus triggering tumorigenesis, or is merely associated with the neoplastic event. The widespread and recurrent nature of this phenomenon, however, seems to suggest that a common mechanistic pathway may exist in cancer cells, which promotes de novo methylation of these targets at the onset of tumor development.
Recent data have unraveled the role of Polycomb repressor complexes in targeting and modulating homeobox genes. At least six independent genome-wide studies have identified several common Polycomb targets in vertebrates and flies, most of which are represented by homeobox genes and other developmental transcription factors [45
]. Interestingly, 43 out of the 81 annotated genes identified in the present study (~53%) and found to be hypermethylated in early-stage breast cancer overlap with known Polycomb targets, strongly supporting the PcG link [46
]. Moreover, most of these DCIS-specific methylated CpG islands are embedded in regions other than promoters, consistent with the finding that the Polycomb repressive complex 2 subunit SUZ12 is distributed across large domains of developmental genes spanning from the promoter up to 2 to 35 kb into the gene [46
]. SUZ12 is required for the histone methyltransferase activity and silencing function of the EED–EZH2 complex and is upregulated in different tumors, including breast tumors [47
]. EZH2, another key Polycomb repressor complex 2 component, undergoes gene amplification in several tumor types [48
] and is overexpressed in prostate cancer and breast cancer [49
]. EZH2 physically interacts with all three DNA methyltransferases in mammalian cells, and has been suggested to play a crucial role in regulating de novo
DNA methylation and its maintenance at target sequences [51
Further support of this mechanistic connection between Polycomb silencing and tumor-associated DNA methylation comes from recent studies linking Polycomb occupancy of genes in noncancerous cells and tissues (including embryonic stem cells) with cancer-associated hypermethylation events [28
]. Paradoxically, however, several homeobox genes are upregulated rather than downregulated in breast cancer and other tumor types, suggesting that several tiers of regulation, other than DNA methylation, may concur in determining homeobox misregulation. Several genome-wide PcG profiling studies have reported that 10% to 20% of the identified PcG targets are transcriptionally active [46
]. Bracken and colleagues have suggested that, in undifferentiated cells, PcG complexes have the potential to target genes poised for silencing as well as target genes predisposed to activation [57
]. The transition between alternative modes of PcG regulation may require additional signals upon differentiation (and likewise during tumorigenesis), which may include recruitment of additional transcriptional activators and/or competition with PcG antagonists, the tritorax group (trxG) proteins. These signals may all have a counteracting effect to the PcG-mediated gene repression [57
In a similar scenario, it is conceivable that many of the homeobox gene-associated CpG islands that become methylated in DCIS might have already switched off their active transcriptional state in the normal breast epithelium or its progenitor cells. If that were the case, a hypothesis linking DNA methylation of homeobox gene CpG islands mechanistically to tumorigenesis would not be sustainable. Unfortunately, no RNA samples from DCIS were available to test the relationship between DNA methylation and gene expression.