Epigenetic Reprogramming of Pericentromeric Satellite DNA in Premalignant and Malignant Lesions

Pericentromeric chromosomal domains, composed of satellite 2 and 3 DNA repeats, exist adjacent to the centromere on multiple human chromosomes. They are gene-poor regions that present as chromatin dense heterochromatic structures in cells and are important for maintaining proper segregation of sister chromatids during mitosis (1, 2). Under normal conditions, pericentromeric satellite DNA, as well as other types of repeat DNA, is retained as condensed, transcriptionally inert, heterochromatin to maintain genomic stability. However, in various types of cancer cells, the epigenetic control and genomic stability of these domains are perturbed. Accordingly, these domains, including the megabase satellite DNA domain at 1q12, are common sites of chromosome rearrangements in various types of cancer (3–6). For instance, 1q12 translocations or deletions are among the most frequent karyotypic abnormalities in breast cancer (5). Also in immunodeficiency, centromeric instability, and facial anomalies (ICF), genomic instability involves pericentromeric satellites (7, 8). In both cancer and ICF, the instability of pericentromeric regions is associated with DNA demethylation and chromatin decondensation (9–11). Furthermore, recent studies have demonstrated that pericentromeric satellite DNA is expressed in various types of cancer and provided a link between this expression and genomic instability (12–16). Thus, epigenetic dysregulation of the 1q12 domain and other pericentromeric satellite DNA domains may play an important role in tumorigenesis.

Constitutive heterochromatin, such as pericentromeric satellite domains, is typically marked with H3K9me3, which is established by the lysine methyl-transferases SUV39H1/2, and recruits heterochromatin protein 1 (HP1). HP1 interacts with other epigenetic factors to implement a repressive state that involves repressive marks such as DNA methylation and H4K20me2/3 (1, 17). The polycomb-group (PcG) proteins traditionally represent another type of chromatin repression normally enriched on facultative heterochromatin together with H3K27me3 and H2AK119ub, and are traditionally not considered associating with pericentromeric heterochromatin (18, 19). However, several studies have demonstrated that under some circumstances, PcG proteins can be found on pericentromeric satellite DNA (20–24). Furthermore, the frequent co-occurrence of H3K27me3 and H3K9me3 marks (25–27) and possible cooperation between HP1 and Polycomb Repressive Complex 2 (PRC2) suggest that the HP1 and PcG repressive systems are not mutually exclusive (27). Thus, there seems an intimate and dynamic exchange between HP1- and Pc-mediated repression of satellite DNA.

In some cell types, PcG proteins are found in relatively large nuclear aggregates, referred to as PcG bodies (28), of which the structural composition and function has remained elusive. We show herein that these structures are, in fact, Polycomb Repressive Complex 1 (PRC1) deposited on the 1q12, and possibly other, pericentromeric satellite DNA domains. The data presented reveal epigenetic reprogramming of satellite DNA as a premalignant event and may help elucidate the role of satellite DNA in cancer development.

Cell lines were cultured in RPMI (Thermo Fisher Scientific; FM6, FM79, and A375) or DMEM (Sigma-Aldrich; MEL-ST) supplemented with 10% fetal bovine serum (Thermo Fisher Scientific), penicillin (100 U/mL), and streptomycin (100 mg/mL). Primary neonatal melanocytes obtained from human foreskin and pooled from five donors (Yale Dermatology Cell Culture Facility, Yale University School of Medicine, New Haven, CT) were cultured in RPMI supplemented with 10% FBS, 200 nmol/L TPA (Sigma-Aldrich), 200 pmol/L cholera toxin (Sigma-Aldrich), 10 nmol/L endothelin (Bachem), and 10 ng/mL human stem cell factor (Thermo Fisher Scientific). Cells lines were maintained at 5% CO₂ and primary melanocytes at 10% CO₂. Where indicated, cells were treated with 2 to 10 μmol/L of GSK126 (Selleckchem) or 2 to 10 μmol/L of SGI-110 (ApexBio) for 72 hours. Melanoma cell lines were: FM2, FM3, FM6, FM28, FM45, FM55-M1, FM55-P, FM57, FM72, FM79, FM81, FM82, FM86, FM88, A375, MZ2-MEL, SK-MEL-28, SK-MEL-37b, and SK-MEL-44. Breast cancer cell lines were: MDA-MB-435s, T-47-D, Hs578T, MCF7, ZR-75-1, BT-474, SK-BR-3, M-4A4, NM-2C5, CAL-51, MDA-MB-231, MDA-MB-157, BrCa-MZ01, BT-20, and MDA-MB-468. Cell lines were kept at low passage numbers (<20) and, when appropriate, their identity was verified using DNA fingerprinting by short tandem repeat analysis (Cell IDTM system, Promega). Cell lines were frequently tested for Mycoplasma (MycoAlert, Mycoplasma testing kit, Lonza).

A375 cells were transfected with pLX304 plasmids encoding PCGF2, CBX6, or CBX6 with a C-terminal V5-tag (Harvard Medical School PlasmID Repository, Boston, MA) or with the pPM plasmid encoding PHC1 with a N-terminal HA-tag (Abmgood).

Tissue specimens were collected from patients treated at Odense University Hospital or obtained from US Biomax. The experiments were conducted in compliance with the Helsinki declaration and approved by the ethical committee of Funen and Vejle County (VF20050069). Informed consent from participants was not needed. Normal tissues included in this study were skin, tonsil, esophagus, parotis, lung, thyroid, spleen, thymus, liver, gallbladder, colon, duodenum, ventricle, muscle, testis, prostate, bladder, kidney, pancreas, cerebellum, uterus, and placenta.

Tissue sections were deparaffinized, treated with 1.5% H₂0₂ in Tris-buffered saline (pH 7.5) for 10 minutes to block endogenous peroxidase activity, rinsed in distilled H₂O, subjected to antigen retrieval by microwave boiling for 15 minutes in 10 mmol/L Tris, 0.5 mmol/L EGTA, pH 9.0 and then stained using one of the following two procedures: (i) Sections were washed in TNT buffer (0.1 M Tris, 0.15 M NaCl, 0.05% Tween-20, pH 7.5) and incubated with monoclonal rabbit anti-BMI1 (Cell Signaling Technology) diluted in antibody diluent (S2022, DakoCytomation) for 1 hour at room temperature. Sections were washed with TNT and incubated with horseradish peroxidase-conjugated “Ready-to-use” EnVision+ polymer K4001 (DakoCytomation) for 30 minutes, followed by another wash with TNT. The final reaction product was visualized by incubating with 3,3′-diaminobenzidine (DAB)+ substrate-chromogen for 10 minutes, followed by washing with H₂O and counterstaining of sections with Azure B before mounting in AquaTex (Merck Inc.). (ii) Sections were blocked with 5% normal goat serum for 60 minutes, incubated overnight with monoclonal rabbit anti-BMI1 (Cell Signaling Technology) diluted 1:100 in 0.25% BSA/0.3% Triton X100/PBS, and washed in dilution buffer. The sections were then incubated with goat anti-rabbit Alexa Fluor 488 (1:500; Invitrogen) for 90 minutes in dilution buffer, washed in dilution buffer, and mounted with Prolong gold antifade solution with DAPI (Thermo Fisher Scientific). Cells grown on coverslips were fixed in 4% formaldehyde and permeabilized in 0.2% Triton X100, PBS. For staining, cells were blocked in 3% BSA, PBS and immunostained with indicated antibodies.

Tissues stained by method 1 (see above) were scored using light microscopy as either BMI1-positive or -negative. Cells were considered positive if staining was convincingly observed in the nuclei regardless of intensity. Parallel sections were stained as above and scored using fluorescence microscopy in four categories representing the frequency of cells with BMI1 bodies: 0 (≤1%), 1 (>1%–<10%), 2 (10%–50%), and 3 (>50%).

Cells were attached to microscope slides using cytospin, fixed and stained as described above. After another fixation, cells were incubated in 2xSSC, 0.05% Tween-20 (pH 7) for 2 minutes and dehydrated for 2 minutes per step in a series of 70%, 85%, and 100% ethanol. Air-dried cells were heated with 1q12 satellite III FISH probe (Cytocell) in hybridization buffer under a coverslip, sealed with rubber cement, for 2 minutes at 75° and incubated for 37° for 16 hours in a humidified atmosphere. Then cells were washed for 2 minutes in 0.25xSSC (pH 7) and 2 × 2 minutes with 2xSSC, 0.05% Tween-20 (pH 7). Mounting on glass slides was done with Prolong gold antifade mountant with DAPI (Thermo Fisher Scientific).

Rabbit anti-BMI1 (Cell Signaling Technology; used 1:200 for immunofluorescence (IF) and 1:100 for fluorescence IHC), mouse anti-BMI1 (Ab14389, Abcam; used 1/1,500 for IF), rabbit anti-histone H3K27me3 (Clone C36B11, Cell Signaling Technology; used 1:200 for IF), rabbit anti-RING1A (Clone D2P4D, Cell Signaling Technology; used 1:800 for IF); rabbit anti-RING1B (clone D22F2, Cell Signaling Technology; used 1/800 for IF), rabbit anti-H2K119Ub (clone D27C4, Cell Signaling Technology; used 1:200 for IF), and rabbit anti-H3 (pAb FL136, Santa Cruz Biotechnology; used 1:1,000 for Western blotting).

Genomic DNA was purified with the DNeasy Blood & Tissue kit (Qiagen). Quantification of overall DNA methylation levels was done with the Methylated DNA Quantification kit. For bisulfite sequencing, 500 ng genomic DNA was bisulfite-converted using the EZ DNA Methylation Gold kit (Zymo Research). PCR amplification of 1q12 satellite II DNA was carried out essentially as previously published (29). Products were cloned into a sequencing plasmid and a representative number of clones were sequenced.

For quantitative measurement of satellite expression, DNAse-treated total RNA was used for cDNA synthesis primed with 5′-AAT CAT CAT CCA ACG GAA GCT AAT G-3′ (AS1) for SATIII, a mix of sense and antisense primers for SATa (ab85782, Abcam), or SATII (ab85781, Abcam). cDNA was used for PCR analysis using SYBR green mastermix (Qiagen) with conditions recommended by the manufacturer. For SATIII, PCR primers were 5′- AGT CCA TTC AAT GAT TCC ATT CCA GT -3′ and AS1 and for SATa and SATII primers were as above. PCR cycles for the SATII and SATa assays were 40 cycles of 15 seconds at 94°, 30 seconds at 60°, and 30 seconds at 72°. The SATIII assay was adapted from Enukashvily and colleagues (12). Samples were run on a StepOne real-time PCR system (Thermo Fisher Scientific).

Sporadic reports of PcG bodies in cancer cells encouraged us to investigate whether this phenomenon is associated with malignant transformation. Because PcG protein BMI1 has been reported to be a canonical subunit of PRC1, we used immunohistochemical staining of BMI1 to investigate the organization of PcG structures in normal, premalignant and malignant tissues. As expected, BMI1 was widely expressed in normal adult tissues. In 22 of the 24 tissues examined, we found cellular subtypes that expressed BMI1, consistent with results reported in the Human Protein Atlas (https://www.proteinatlas.org). In normal cells, BMI1 generally exhibited an inhomogeneous nuclear distribution resembling that of chromatin (Fig. 1A; Supplementary Fig. S1). BMI1 was also frequently detected in cancer cells, with 85% of melanoma cell lines and 71% of breast cancer cell lines being positive. In contrast to the nuclear pattern of BMI1 in normal cells, large BMI1 PcG bodies were observed in 76% of melanoma cell lines and 50% of breast cancer cell lines (Fig. 1B). PcG bodies were also seen in patient tumors of different cancer types, but were most prevalent in melanomas (Fig. 1B). Eight of 10 BMI1⁺ melanoma tumors contained PcG bodies, and 7 had PcG bodies in >80% of tumor cells. PcG bodies were less frequently observed in other cancer types and these generally exhibited a much lower frequency of tumor cells containing PcG bodies compared with melanoma (10%–50%). In both cancer cell lines and patient tumors, the number of PcG bodies per cell was highly variable (i.e., between 1 and 7 PcG bodies per cell).

These results clearly suggested that PcG bodies are a phenomenon specific to malignant cells and predominantly present in melanoma. To validate this hypothesis, we examined BMI1 in tumor development using melanoma as a model. In melanoma tumor development, melanocytes in the epidermis can acquire genetic and epigenetic changes that support increased proliferation and the formation of benign lesions called nevi (30). Although having a benign phenotype, nevi cells exhibit characteristics of melanoma cells (e.g., activation of oncogenes) and may develop further into melanoma cells. In cultured neonatal primary melanocytes, BMI1 demonstrated a genome-wide, chromatin-associated distribution similar to that of normal cell types (Fig. 2A), but very different from the large PcG bodies observed in cancer cells. The high-magnification images of primary melanocytes suggested a somewhat punctuate distribution on chromatin. We also investigated the BMI1 organization in melanocytes immortalized by transduction with SV40 and hTERT (31). These cells proliferate indefinitely in culture, but do not form tumors in immunodeficient mice. The immortalized melanocytes also contained BMI1 PcG bodies (0–4 per cell; Fig. 2A), providing an interesting link between extensive proliferation and formation of PcG bodies. In compound and dermal nevi, BMI1 was expressed in 32 of 33 specimens. Interestingly, the BMI1 expression pattern was highly diverse between and within nevi (Fig. 2A and B; Supplementary Table S1). BMI1 PcG bodies were observed in all, or a subset, of cells in 29 of 32 BMI1⁺ nevi. There were also variations in number (approximately 1–3 PcG bodies per cell) and size of BMI1 PcG bodies. Staining of a larger cohort of melanomas demonstrated that 34% of BMI1⁺ tumors contained BMI1 PcG bodies (24/70; Fig. 2A and B; Supplementary Table S1). There was no significant difference between primary melanomas versus metastases. Interestingly, more melanomas exhibited a relatively homogenous pattern of BMI1 nuclear distribution compared with nevi, with most cells in individual tumors either PcG body-positive or -negative (Fig. 2B). Melanoma generally also had more PcG bodies (approximately 1–8 PcG bodies per cell) per cell than nevi.

These results demonstrate that PcG bodies are a general phenomenon in cancer, but more prevalent in melanoma than other cancers. We also demonstrate that, at least in melanoma, PcG body formation is a premalignant event.

Multiple forms of mammalian PRC1 complexes containing the subunits BMI1, PH1, CBX4, and RING1B or their orthologous and, in some cases, non-PcG partners, have been described (32). Apart from the existence of BMI1 in PcG bodies, the composition of these structures has remained largely elusive (28). We investigated the presence of different variants of canonical PRC1 subunits in two melanoma cell lines, A375 and FM79, and found that PcG bodies in melanoma cells of both cell lines comprised different variants of all four types of canonical PRC1 subunits (Fig. 3A and B; Supplementary Fig. S2). In addition, melanoma PcG bodies were positive for the two PcG-associated histone marks H2AK119ub and H3K27me3. The latter, catalyzed by EZH2, has been shown to be important (but not required in all cases) for PRC1 deposition on chromatin. Treatment of melanoma cells with a small molecule inhibitor of EZH2 (i.e., GSK126; ref. 33) reduced H3K27me3 (Fig. 3C–D). Interestingly, reduced levels of H3K27me3 were associated with increased levels of histone H3 for unknown reasons (Fig. 3C). EZH2 inhibition also induced loss or deformation of PcG bodies in a number of cells (Fig. 3E and F). Although H3K27me3 levels seemed highly reduced in most cells (>90%) treated with 10 μmol/L GSK126 (Fig. 3D), only 18% and 12% exhibited loss or deformation of PcG bodies, respectively (Fig. 3E and F), suggesting that H3K27me3 is not essential for maintenance of PcG bodies, but is required for de novo formation of PcG bodies in daughter cells upon cell division. These results suggest that PcG bodies contain PRC1 complexes with a canonical composition, and that their chromatin recruitment complies with known mechanisms (Fig. 3G).

PcG proteins are believed to catalyze the formation of facultative heterochromatin, which maintains developmental genes at a repressed state (34). Drosophila studies have provided evidence of the association of PcG-repressed genes with PcG bodies, but in general little is known about the genomic loci contained in PcG bodies. In human cancer cells, PcG bodies have been shown to associate with the 1q12 pericentromere composed of satellite 2 and 3 repeats (35). We used a combined IF-in situ hybridization technique that allows simultaneous detection of PcG bodies and a 1q12 satellite III probe to confirm this association in multiple cancer cells lines and in melanoma tumors (Fig. 4A and B). In every cell, the probe stained 1 to 8 foci that, indeed, corresponded to PcG bodies (Fig. 4A and B). In most cases, there was complete agreement between the number of PcG bodies and 1q12 foci. In other cases (e.g., U2OS cells), there were more PcG bodies than 1q12 foci. The differences in number of satellite III/PcG bodies between cells lines may reflect chromosome 1 copy number variations or spreading of PcG deposition to other satellite structures. PcG bodies in cells of melanocytic nevi and melanoma tumors were also found to be equivalent to the 1q12 pericentromeric domain (Fig. 4B). In primary melanocytes and normal somatic tissues (e.g., colon and thyroid cells), there were no PcG bodies, and therefore no association between PcG bodies and the 1q12 domain (Fig. 4A and B). These results clearly show that PcG bodies are nuclear subdomains in which PRC1 accumulates on the 1q12 pericentromeric satellite DNA and possibly on other satellite DNA domains. Moreover, our results strongly suggest that epigenetic conversion of pericentromeric satellites DNA into a polycomb state is a feature of premalignant and malignant cells.

Although previously believed to be maintained in a constitutive highly repressed state, pericentromeric satellite DNA is now known to be expressed under specific conditions. For instance, satellite transcripts were detected spatially and temporarily during mammalian development and cellular differentiation (36, 37). Satellite II and III expression has also been reported in different types of cancer cells (12, 13) and in cells responding to heat shock, osmotic stress and radiation (38, 39). Moreover, unfolding and expression of satellite DNA was described as an early event in multiple types of senescence (12, 40). The epigenetic mechanisms underlying this depression of transcription has not been fully described. We investigated whether epigenetic switching to a PcG state is associated with increased transcription from the affected loci, and quantification of expression from satellite II and III repeats present at 1q12 clearly suggested that polycomb deposition on these DNA structures permits increased transcription (Fig. 5A and B). A very low basal level of satellite II and III transcription was detected in primary neonatal melanocytes (from 25 ng of reversely transcribed RNA), which are devoid of PcG bodies. The level was slightly enhanced (5.5 times for satellite II) in immortalized melanocytes, where all cells have PcG bodies. This suggests that the conversion of these loci to a PcG state moderately enhances transcription. The transcriptional level of 1q12 satellite II and III was further enhanced in melanoma cells compared with immortalized melanocytes (i.e., up to 76.5 times relative to primary melanocytes for satellite II), which may reflect the presence of more PcG bodies or additional epigenetic changes that promote satellite expression in melanoma cells. Expression from centromeric alpha-satellite DNA was not detected in primary melanocytes, immortalized melanocytes, or melanoma cells (Fig. 5A and B). Noncoding RNAs are important for PcG recruitment (41) and it can be speculated that enhanced transcription from the 1q12 pericentromeric satellites is functionally implicated in its conversion to a PRC1 domain. Similarly, noncoding RNAs are central in the formation of HP1 heterochromatic domains (42).

Because PcG complexes repress chromatin, we tested the possibility that disruption of PcG bodies would result in increased transcription of satellite DNA. A375 cells were treated with an EZH2 inhibitor (GSK126) to disrupt PcG bodies (Fig. 3E and F), and quantification of satellite 2 expression indeed showed increased transcription from satellite 2 DNA (Fig. 5C). This suggests that although conversion of 1q12 pericentromeric DNA into PcG bodies is associated with increased transcription, PcG bodies still represent a transcriptionally repressive milieu.

Recent reports suggest that lack of DNA methylation facilitates association of PcG proteins with pericentromeric heterochromatin. For instance, absence of DNA methylation and H3K9me3 was demonstrated to recruit PcG and H3K27me3 to pericentromeric heterochromatin in mouse embryonic stem cells (24). Similarly, Dnmt knockout recruited PRC1 to pericentromeric heterochromatin and increased H3K27me3 and H2AK119ub at these regions (21). This is in agreement with the ability of unmethylated CpGs to act as nucleation sites for PcG recruitment through binding of KDM2B and possibly other ZF-CxxC binding proteins (43–45). Pericentromeric satellites, including the 1q12 domain, are generally highly methylated in normal tissues, but often demethylated in cancer (5, 46, 47). Thus, demethylation could provide the basis for formation of PcG bodies. To test this, we compared the overall and 1q12 satellite II methylation status in primary melanocytes, immortalized melanocytes, and melanoma cells and found that the formation of PcG bodies in immortalized melanocytes and melanoma cells indeed correlated with both an overall reduction in DNA methylation levels and a specific demethylation of satellite repeats of the 1q12 domain (Fig. 6A).

To further establish the importance of DNA demethylation in recruitment of PcG complexes to 1q12 satellite DNA, we treated primary melanocytes with the DNA methyltransferase inhibitor SGI-110, which induced formation of PcG bodies in a subset of cells, suggesting that demethylation alone may be sufficient to induce PcG body formation (Fig. 6B and C). In most cells, two PcG bodies were formed corresponding to the number of 1q12 domains per cell. Combined IF and in situ hybridization analysis confirmed that these PcG bodies indeed formed on 1q12 pericentromeric satellite DNA (Fig. 6D). In a small number of cells, there were more than two PcG bodies, suggesting that they also formed on additional pericentromeric satellite DNA domains. The epigenetic conversion induced by SGI-110 resulted in only a modest increase in transcription from 1q12 satellite repeats (2.9-fold for satellite II; Fig. 5).

These results suggest that demethylation is the primer for PcG body formation. Future studies should address whether loss of activity of specific DNMTs are involved.

The biology of DNA satellites has remained largely unexplored due to their highly repetitive nature, which complicates their investigation. Furthermore, until recently repetitive sequences were considered junk DNA with no functional properties. Conversely, various types of satellite DNA have long been implicated in cancer development as sites of chromosome rearrangements (3, 48). A major site of chromosome rearrangement in various types of malignancies, such as breast cancer and myeloma, is the megabase sized 1q12 pericentromeric satellite DNA domain (3, 4, 6, 9), which is also implicated in the ICF syndrome as a site of genomic instability (7). The involvement of the 1q12 and other pericentromeric satellite DNA domains in chromosome breakage is associated with their unfolding, which is likely promoted by epigenetic changes. Indeed, the instability of pericentromeric satellite DNA is connected with the demethylation of these domains (9, 10). Still, understanding the epigenetic control of satellite DNA and its role in cancer development remains elusive. In this study, we present novel insight into the epigenetic control of the 1q12 pericentromeric satellite DNA in premalignant and malignant cells.

We show that the 1q12 satellite DNA domain is reprogrammed into a PcG state in melanocytic cells of nevi, which can be considered premalignant lesions, and that this chromatin domain corresponds to what is known as PcG bodies. We further find that this phenotype is conserved in primary and malignant melanoma tumors and also evident in other types of cancer. We show that this epigenetic conversion coincides with global and satellite DNA demethylation and can be induced by inhibition of DNMTs, suggesting that PcG bodies form on pericentromeric satellites in response to DNA demethylation. DNA methylation is functionally linked with H3K9me2/3, HP1 proteins, and associated factors normally important for maintaining repression of pericentromeric heterochromatin, suggesting that PcG bodies form on pericentromeric satellites DNA domains as a compensatory repressive mechanism. Despite this installation of repressive PcG complexes on 1q12 satellite DNA, it becomes permissive for transcription. This may be surprising because expression of satellites in cancer cells has been linked to genomic instability (15, 16). On the other hand, this seems to be context-specific because expression of satellite sequences has also been detected in a number of normal tissues (36, 37). Our results are backed up by findings by Hall and colleagues who demonstrated hypomethylation-dependent BMI1 deposition on 1q12 satellite DNA domain in cancer cells (49). However, in contrast to our data, this study did not find increased expression from the 1q12 loci upon PcG deposition, perhaps due to differences in the investigated cell types (primary melanocytes vs. U2OS cells) or assay sensitivity (PCR vs. RNA hybridization).

The question of the functional purpose of this epigenetic conversion of pericentromeric satellite DNA in premalignant cells is clarified by our finding that PcG body formation precedes malignant transformation (e.g., melanocytic nevi), and it can be speculated that it primes cells for this process. However, only about half of primary melanoma tumors have PcG bodies corresponding to frequencies observed among nevi cells. Thus, there does not seem to be a selection of cells with PcG bodies from the premalignant to the malignant stage, indicating that PcG bodies have no direct role in malignant transformation. Instead, we speculate that PcG bodies may be indirectly important for cancer development. As mentioned above, PcG body formation could be an essential mechanism for maintaining stability of demethylated satellites in cells that experience global demethylation, which may be beneficial for other reasons. Indeed, our results suggest that PcG body formation is associated with global demethylation. Although genomic instability, such as that caused by, for instance, unfolding of pericentromeric satellite DNA domains, may be beneficial to premalignant and malignant cells, it likely needs to be balanced to maintain mitotic fidelity, and PcG body formation may represent such a mechanism. This is supported by our finding that disruption of PcG bodies by inhibition of EZH2-mediated H3K27me3 derepresses satellite 2 DNA, as indicated by increased expression (Fig. 5).

PcG bodies may also be important for preventing cells from undergoing senescence in response to oncogene expression and increased proliferation. Typical nevi contain one million or more cells and require at least 20 cell doublings to form. This suggests that expression of oncogenic BRAF or NRAS, as occurs in melanocytic cells of many nevi, does not directly induce senescence, but that senescence rather occurs as a response to telomere erosion after oncogene-induced proliferation (30). Our observation that PcG bodies are present in melanocytes with increased proliferation and lifespan due to expression of SV40ER and hTERT, as well as in nevi cells, indicates that this phenomenon is linked to enhanced proliferative capacity. Interestingly, unfolding and depression of pericentromeric satellite DNA has been observed in several types of senescence and may be mechanistically implicated in the senescence response (40). Thus, epigenetic reprogramming of satellite DNA into PcG bodies may serve as a means to prevent senescence development. This fits well with results demonstrating that PcG proteins such as BMI1 and EZH2 can delay or prevent senescence (50–52). Although this involves silencing of the INK4A-ARF locus and reduced function of p16 (53–55), repression of pericentromeric satellite DNA may represent an additional means to repress senescence.

We have described a novel type of premalignant epigenetic perturbation that is conserved in many melanomas and other types of cancer. Because the reprogramming of pericentromeric satellite DNA domains into PcG bodies seems to be specific to premalignant and malignant cells, it may be useful as a diagnostic marker for precancerous changes and represent a novel therapeutic entry point for treatment of cancer.

No potential conflicts of interest were disclosed.

Conception and design: N.H. Brückmann, C.B. Pedersen, M.F. Gjerstorff

Development of methodology: N.H. Brückmann, C.B. Pedersen, M.F. Gjerstorff

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): N.H. Brückmann, C.B. Pedersen, M.F. Gjerstorff

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): N.H. Brückmann, C.B. Pedersen, M.F. Gjerstorff

Writing, review, and/or revision of the manuscript: N.H. Brückmann, H.J. Ditzel, M.F. Gjerstorff

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): N.H. Brückmann, M.F. Gjerstorff

Study supervision: H.J. Ditzel, M.F. Gjerstorff

This work was supported by the Velux Foundation, the Danish Cancer Society, Academy of Geriatric Cancer Research (AgeCare), and the Danish Research Council for Independent Research. We thank Ole Nielsen and Lisbet Mortensen (Department for Pathology, Odense University Hospital) for technical assistance with immunohistochemical staining, M. K. Occhipinti for editorial assistance and Dr. Robert A. Weinberg (Whitehead Institute for Biomedical Research) for providing MEL-ST cells.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.