|Home | About | Journals | Submit | Contact Us | Français|
In a previous genomic analysis, using somatic methyltransferase (DNMT) knockout cells, we showed that hypomethylation decreased the expression of as many genes as were observed to increase, suggesting a previously unknown mechanism for epigenetic regulation. To address this idea, the expression of the BAG family genes was used as a model. These genes were used because their expression was decreased in DNMT1−/−, DNMT3B−/−, and double knockout cells and increased in DNMT1-overexpressing and DNMT3B-overexpressing cells. Chromatin immunoprecipitation analysis of the BAG-1 promoter in DNMT1-overexpressing or DNMT3B-overexpressing cells showed a permissive dimethyl-H3-K4/dimethyl-H3-K9 chromatin status associated with DNA-binding of CTCFL/BORIS, as well as increased BAG-1 expression. In contrast, a nonpermissive dimethyl-H3-K4/dimethyl-H3-K9 chromatin status was associated with CTCF DNA-binding and decreased BAG-1 expression in the single and double DNMT knockout cells. BORIS short hairpin RNA knockdown decreased both promoter DNA-binding, as well as BAG-1 expression, and changed the dimethyl-H3-K4/dimethyl-H3-K9 ratio to that characteristic of a nonpermissive chromatin state. These results suggest that DNMT1 and DNMT3B regulate BAG-1 expression via insulator protein DNA-binding and chromatin dynamics by regulating histone dimethylation.
One mechanism regulating the epigenome involves specific chromosomal methylation patterns controlled via a complex interplay of at least three independently encoded DNA methyltransferases (DNMT): DNMT1, DNMT3A, and DNMT3B (1, 2). It has been shown previously that DNMT1 and DNMT3B seem to cooperatively maintain DNA methylation and gene silencing, and genetic disruption of both significantly inhibited DNMT activity and reduced genomic DNA methylation by roughly 95% (3). Using these somatic knockout (DNMT1−/−, DNMT3B−/−) and double knockout (DKO) cell lines, our group has reported a comprehensive study comparing the effects of altered DNMT activity on gene expression in the cancer epigenome (4).
One of the most interesting findings in this study involved the observation that a substantial fraction of genes was down-regulated, rather than up-regulated, after genetic DNMT knockout (4). These results suggest that a hypomethylated chromosomal status inhibits a subset of gene expression, whereas methylation may activate some of these same genes. Since this publication, data from another group (5) have also shown that overexpression of DNMT1 increased the expression of a specific subset of genes. In this work, genomic analysis of Rat-1 cells overexpressing DNMT1 showed an increased expression of slightly more genes than were observed to decrease, suggesting that regulation of the epigenome may be more complex than previously thought. However, neither of these studies suggests a mechanism to account for these surprising observations.
Chromatin insulators are DNA boundary elements that inhibit gene expression by partitioning chromosomal domains, and they function by blocking the effects of nearby enhancers in a position-dependent manner (6, 7). Chromatin insulators binding proteins, such as CTCF and CTCFL/BORIS, bind to specific DNA boundary elements that isolate chromosomal domains from cis-acting transcriptional regions (6, 7). CTCF is thought to be a ubiquitous, highly conserved, zinc-finger (ZF) DNA-binding protein that has multiple roles in gene regulation (6), as well as chromatin and X chromosome inactivation (8). A paralogue of CTCF, brother of the regulator of imprinted sites (BORIS), exhibits extensive homology to CTCF in the 11 ZF regions, suggesting a similar DNA-binding spectrum (9–11). However, there are significant differences in their NH2 and COOH terminal domains (9), suggesting that these regions interact with different binding proteins to alter gene expression. BORIS is thought to be primarily a testis-specific protein (12); however, BORIS has also been shown to be increased in a variety of tumors (10). CTCF seems to repress gene expression of a specific subset of genes (13, 14), whereas BORIS has been shown to activate gene expression (10).
In the current study, we show that regulation of BAG-1 expression by CTCF and BORIS through CTCF-binding sites located in its upstream promoter region. Chromatin immunoprecipitation (ChIP) analysis of the BAG-1 upstream regulatory region in DNMT somatic knockout cells showed a nonpermissive chromatin status, indicated by a lower ratio of dimethyl-H3-K4/dimethyl-H3-K9. In addition, we showed preferential CTCF binding to the promoter region, as well as decreased BAG-1 expression in DNMT somatic knockout cells. In contrast, ChIP analysis of the BAG-1 promoter in cell lines overexpressing either DNMT1 or DNMT3B showed a permissive chromatin status, indicated by a higher ratio of dimethyl-H3-K4/dimethyl-H3-K9. Furthermore, preferential BORIS binding to the promoter and increased BAG-1 expression were observed in these cells. BORIS short hairpin RNA (shRNA) knockdown decreased BORIS binding to the BAG-1 promoter, decreased BAG-1 expression, and changed the ratio of dimethyl-H3-K4/dimethyl-H3-K9 to a nonpermissive chromatin state. Finally, transient cotransfection experiments also showed that BORIS activated and CTCF inhibited BAG-1 promoter activity. The results of these experiments suggest that (a) cellular DNMT expression seems to influence histone methylation, as reflected by changes in the dimethyl-H3-K4/dimethyl-H3-K9 ratio, (b) there is a reciprocity of BORIS and CTCF function to coordinately dictate gene expression via histone methylation, and (c) DNA methylation may activate specific genes or gene families via a mechanism involving BORIS binding to the promoter.
DNMT1 knockout (DNMT1−/−) and DNMT3B knockout (DNMT3B−/−) cell lines and DNMT1 and DNMT3B DKO cell lines, as well as a parental colon cancer cell line (HCT116), were described (3) and grown in McCoy’s 5A medium (Invitrogen). HCT116 cells overexpressing DNMT1 (DNMT1+), DNMT3B (DNMT3B+), DNMT1, and DNMT3B double knock-in (DKI) were cultured in McCoy’s 5A medium supplemented with 10% fetal bovine serum (FBS), 1% penicillin/streptomycin, and 750 μg/mL G418. Rat-1 and Rat-1 CMV-DNMT1+ cell lines (a kind gift from Dr. Tom Curran) were cultured in MEM (α modification), containing 5% FBS, penicillin (100 units/mL), streptomycin (100 μg/mL), and G418 (300 μg/mL) in a humidified 37°C incubator with 5% CO2.
Total cellular RNA was isolated as described previously (4). Reverse transcription–PCR (RT-PCR) and quantitative real-time RT-PCR were carried out as previously described (15). RT-PCR reactions were done thrice for each sample, as previously described (16). Gene-specific primers were purchased from Applied Biosystems as premade TaqMan gene expression assays: BAG-1-Hs00185390_m1, BAG-3-Hs00188713_m1, BAG-4-Hs00362193_m1, and glyc-eraldehyde-3-phosphate dehydrogenase (GAPDH)–HS99999905_m1. The relative amounts of transcript of the tested genes were normalized by GAPDH. The average ratios of three replicate runs were calculated.
Nuclear extracts were prepared using a modification of the previously described method (17). No reducing agents were added to the buffers used to make extracts. Protein concentrations were determined using the Bradford assay (Bio-Rad Labs.) and a Beckman DU-640 spectrophotometer (18). After preparation and protein analysis, samples were stored at −80°C and thawed on ice immediately before use.
After preparation of nuclear extracts, equal protein amounts (10–20 μg/sample) were mixed with Laemmli lysis buffer, boiled for 5 min, and loaded into denaturing SDS-polyacrylamide gels for electrophoresis, as previously described (19). Membranes were probed with antibodies against BAG-1 (Santa Cruz Biotechnology) and BAG3 [a kind gift from Elise Kohn, National Cancer Institute (NCI)] and incubated with the appropriate horseradish peroxidase–conjugated secondary antibody (Santa Cruz Biotechnology). Bands were analyzed using an enhanced chemiluminescence protocol (Amersham Biosciences) and visualized on radiographic film (Eastman-Kodak). Figures are representative of the outcome in at least two independent experiments.
The 827-bp fragment encompassing −854 to −28 of the hBAG-1 promoter was PCR amplified from an Invitrogen BAC clone (RP11-366F19). The fragment was subcloned into pGL3-Basic (Promega) to construct the pBAG-1-luc plasmid. All internal deletions and truncations of the hBAG-1 promoter were done by subcloning various PCR fragments into either pBAG-1-FL-luc or pGL3-Basic. Two complementary oligos (ggtaccAGGCTGGGCGGCGGgagct-cacgcgtCCGCGAAGAAACgctagcctcgagCCTCCTGGCGTTTagatct) carrying all three putative CTCF-binding sites that reside in the hBAG-1 promoter were synthesized, annealed, and subcloned into pTAL-Luc (Clontech) to construct p3x-CTCF-tk-LUC.
BORIS shRNA was obtained from OriGene and transfected into HCT116 cells using FuGENE (Roche) according to the manufacturer’s recommendation. At 48 h after transfection, cells were put in selective media containing 1 μg/mL puromycin. Colonies were isolated after 3 wk of antibiotic selection, and Western blotting was performed to screen for positive colonies with decreased BORIS expression. HCT116-based and Rat-1–based cells were plated at 2 × 105 cells per 60-mm plate and transfected using FuGENE according to manufacturer’s protocol. For each transfection, 1 μg of either pBAG-1-Luc or p3x-CTCF-tk-LUC was used. As an internal control, 1 μg of the β-galactosidase expression plasmid (pCMV-β-gal) was used, and carrier DNA (pUC) was added to each sample to maintain a total of 6 μg per transfection. For all transfections, the total amount of CMV promoter-containing plasmids was adjusted so that the same amount was present in each reaction. Thirty-six hours after transfection, cells were harvested and prepared for luciferase and β-galactosidase assay (Promega). The total amount of luciferase and β-galactosidase activities were measured independently using a Wallac plate reader (Perkin-Elmer) equipped with the appropriate filters. The relative fold induction of luciferase activity was normalized to β-galactosidase activity. The BORIS expression vector, a kind gift from Dr. David Schrump (NCI), and the CTCF expression vector, a kind gift from Dr. Carl Barrett (NCI), were subcloned into pCIneo.
Cells were cultured for 24 h and prepared using a ChIP assay kit from Upstate Biotechnology, Inc., according to the manufacturer’s recommendations. Briefly, cells were fixed in 1% formaldehyde at 37°C for 10 min before being washed, scraped, and pelleted in ice-cold PBS. Cells were lysed in SDS lysis buffer [1% SDS, 10 mmol/L EDTA, 50 mmol/L Tris-HCl (pH 8.1)] for 10 min on ice. Cell lysates were sonicated using a Misonix 3000 sonicator equipped with a cup horn for 2 × 2 min, 20 s pulse, and 20 s rest interval at setting 4. Sonicated cell lysates were centrifuged at 13,000 rpm for 10 min at 4°C. An aliquot of the supernatant was subjected to reverse cross-linking with 5 mol/L NaCl to check for sonication efficiency. Samples with DNA sheared between 1000 and 200 bp were used for subsequent ChIP reactions.
Immunoprecipitations were performed using CTCF (C-20) antibody from Santa Cruz Biotechnology, Inc., or BORIS antibody (Supplementary Figs. S2 and S3). ChIP-enriched DNA along with input DNA was purified by phenol/chloroform extraction and ethanol precipitation. Purified DNA samples were analyzed by quantitative real-time PCR using SYBR Green PCR Master Mix (Applied Biosystems) with the ABI Prism 7500 Detection System (Applied Biosystems) with the following primer pair: 5′-GTCTTCCGGG-GATGGAGAGCA-3′and 5′-TTGACCGCCCAGCGATGG-3′. The conditions for ChIP-PCR were 40 two-step cycles of 95°C for 15 sec and 60°C for 60 sec. Data were collected at 60°C and analyzed by comparative C t methods (20). All results for ChIP experiments (including those described below) are presented as fold enrichment relative units that is calculated as the mean ± SD fold enrichment of immunoprecipitated DNA relative to no antibody (BORIS or CTCF) IgG controls, which represents ChIP specificity, done in duplicate for each experiment.
Immunoprecipitations were performed using either an antibody to histone H3 dimethyl K4 or histone H3 dimethyl K9 (Abcam). ChIP was performed as described previously (21). Each experiment was repeated at least twice independently. Pulled down DNA was analyzed using real-time PCR normalized by input DNA. The sequences of primers at BAG-1/CTCF-binding sites 1 and 2 are forward, 5′-GTGCCAATGCAGTCAGTCA and reverse, 5′-TGGTTGCTCTAGTGT-TTGGG; the sequences of primers at BAG 1/CTCF-binding site 3 are forward, 5′-ACCAGAAACGGA-AGCAGAGT and reverse, 5′-CCTCCCACTCCTTCTACCAA.
Protein samples were prepared, stored, and quantified by methods described earlier. No reducing agents were added to the electrophoretic mobility shift assays (EMSA) or the buffers used to make nuclear extracts. EMSAs were performed as previously described (22) using a 32P-radiolabeled oligonucleotide corresponding to the second CTCF DNA-binding site in the BAG-1 promoter. Nuclear extracts (10 μg) were incubated with poly dI-dC for 10 min on ice, followed by the addition of radiolabeled oligonucleotide (100,000 cpm of radiolabeled probe per reaction) and incubation at 25°C for 20 min. Samples were electrophoresed on a 6% nondenaturing PAGE gel, dried, exposed to a phosphoimager screen, and analyzed using a Typhoon 8600 Phosphor-imager (Molecular Dynamics).
We have previously shown that roughly equal numbers of genes were observed to increase as those shown to decrease after genetic knockout of the DNMT genes (4). Interestingly, an additional genomic study of DNMT1 overexpression in Rat-1 cells showed an increase in expression of roughly 380 genes while ~300 genes were decreased (5). Taken together, these results implied that DNA methylation may increase, as well as, silence specific genes in a coordinate manner. To address this issue, a more complete analysis of our microarray data was done to examine genes that exhibit decreased gene expression in DNMT somatic knockout cells. This analysis identified a family of prosurvival genes (BAG gene family) that was decreased in DNMT1−/− and DNMT3B−/−cell lines; this decrease was even greater in the DKO cells (Table 1). In contrast, an increase in BAG-1, BAG-3, and BAG-4 gene expression was observed in HCT116 cells overexpressing either DNMT1 (DNMT1+) or DNMT3B (DNMT3B+). Because the expression of all three of the BAG genes was altered as a function of DNMT status, it seemed logical to use the BAG family genes as a model.
To confirm the microarray results, RNA was isolated from the HCT116, DNMT1−/−, DNMT3B−/−, and DKO cell lines and was used for quantitative real-time RT-PCR analysis for BAG-1, BAG-3, and BAG-4 expression. These experiments authenticated the microarray data and showed that gene expression for the BAG genes is intermediately down-regulated in the DNMT1−/− and DNMT3B−/− cells and significantly decreased in the DKO cells (Fig. 1A). Western blot analysis with antibodies against either BAG-1 (Santa Cruz Biotechnology) or BAG-3 (a kind gift from Dr. Elise Kohn) showed that the BAG-1 and BAG-3 protein levels were decreased in the somatic knockout cell lines and further decreased in the DKO cells (Fig. 1B).
The upstream regulatory regions of BAG-1, BAG-3, and BAG-4 were analyzed, and interestingly, the 5′upstream regulatory regions of all three BAG genes revealed several common sequences, including CpG islands (Fig. 1C). To investigate a mechanism for BAG gene regulation, the BAG-1 promoter region was cloned into a luciferase reporter vector (pBAG-1-Luc) and transient transfection assays were done in the DNMT1−/−, DNMT3B−/−, DKO, and DNMT1+ cells. Consistent with the RNA and protein data, cells lacking DNMT genes transfected with pBAG-1-Luc showed decreased BAG-1 promoter activity (Fig. 1D, vertical columns 1–4). In contrast, HCT116 cells overexpressing DNMT1 (DNMT1+ cells) displayed a roughly 2-fold increase in luciferase activity (column 5). In addition, a pair of matched Rat-1 fibroblast cell lines overexpressing DNMT1 (23) showed a similar increase in the BAG-1 promoter activity (Fig. 1D, vertical columns 6 versus 7). An increase in BAG-1 promoter activity was also seen in HCT116 cell lines overexpressing DNMT3B or both DNMT1 and DMNT3B (DKI; Supplementary Fig. S1), suggesting that overexpression of DNMT genes also alters BAG-1 gene expression.
To define the critical DNA cis-acting sequence(s) in the BAG-1 promoter responsible for regulating gene expression, a series of progressive BAG-1 promoter 5′ deletion mutants were constructed. HCT116 cells transiently transfected with these plasmids (Fig. 2A) clearly showed that the transcriptional regulatory region of interest is located within the BAG-1 CpG island. Similar results were observed in HeLa cells (data not shown). A series of internal deletion mutants were constructed that systematically remove various parts of the CpG island, including the three CTCF DNA-binding sites. Transient transfection experiments showed that deletion of the two upstream CTCF-binding sites (Fig. 2B) increased expression by roughly 2-fold (third column from the top), and the removal of all three CTCF-binding sites ( fifth column from the top) resulted in an additional, but very modest, increase in the luciferase activity. These data suggest two points: (a) the CTCF DNA-binding sites in the CpG island are potentially the cis-acting DNA sequence inhibiting BAG-1 gene expression and (b) the first and second CTCF-binding sites have a more significant effect on gene expression than the third.
To determine a potential mechanism, we examined the methylation pattern of the BAG-1 promoter in both the somatic DNMT knockout and the DNMT-overexpressing cells. Multiple regions of the BAG-1 promoter, including the CpG island and CTCF-binding sites, were analyzed by bisulfite sequencing and failed to show any significant differences in methylation (Fig. 3A). We also looked at histone methylation to determine if changes in chromatin compaction might account for the changes in BAG-1 gene expression. Histone H3 dimethylation at residue K4 (H3-K4) or histone H3 dimethylation at residue K9 (H3-K9) have been shown to define distinct chromatin regions either permissive or nonpermissive for gene expression, respectively. ChIP analysis of the BAG-1 first, second (Fig. 3B), and third (Fig. 3C) CTCF-binding sites in DNMT somatic knockout cells showed a nonpermissive chromatin status indicated by a low dimethyl-H3-K4/dimethyl-H3-K9 ratio compared with that of control HCT116 cells (Fig. 3B and C). In contrast, a permissive chromatin status with a high dimethyl-H3-K4/dimethyl-H3-K9 ratio was observed in DNMT1+ and DNMT3B+ cells (Fig. 3B and C). Interestingly, no change in the dimethyl-H3-K4/dimethyl-H3-K9 ratio was observed for the third CTCF DNA-binding site in the DNMT3B−/− cells (Fig. 3B), whereas the first and second region did change (Fig. 3C). This result matches with the RT-PCR data where only a very modest decrease in BAG-1 expression was observed in the DNMT3B−/− cells (Fig. 1A). The results of these experiments suggest that DNMT1 and DNMT3B may selectively influence H3-K4 and H3-K9 dimethylation, and this may be one mechanism for the regulation of BAG-1 gene expression.
It has been reported that CTCF can block the transcriptional activity of enhancer elements (13, 24). Thus, it seemed logical to determine if CTCF associates with the BAG-1 promoter and if this might play a role in the regulation of gene expression. CTCF ChIP analysis of the BAG-1 promoter in HCT116, DNMT1−/−, DNMT3B−/−, and DKO cells showed an increase in CTCF binding to the BAG-1 promoter in the DNMT1−/− and DNMT3B−/− cells and a further increase in the DKO cells (Fig. 4A). In contrast, CTCF binding to the BAG-1 promoter was markedly decreased in DNMT1+, DNMT3B+, and DKI cells (Fig. 4B). Overall, these results showed an inverse relationship between CTCF binding and the expression of BAG-1, suggesting a potential mechanism whereby CTCF inhibits transcription and lack of binding prevents repression and increases the expression of BAG-1 promoter.
The above experiments used primers that flank the BAG-1 promoter CTCF-binding sites, but these results do not a priori show that the CTCF-binding site is the target sequence. To address this, electrophoretic mobility shift assays (EMSAs) were done using the CTCF-binding site oligos. These experiments showed a significant increase in CTCF-binding activity with protein extracts harvested from HCT116, DNMT1−/−, and DNMT3B−/− cells (Fig. 4C, right, lane 1 versus lanes 2 and 3). Surprisingly, instead of an expected decrease in DNA binding, a significant increase in binding was observed in the DNMT1+, DNMT3B+, and DKI cells (Fig. 4C, lanes 4–7) compared with that observed in the HCT116 cells (lane 1) or the somatic cell knockouts (lanes 2 and 3). In addition, not only was there an increase in DNA-binding in the DNMT-overexpressing cells, but the shifted band ran slightly lower than that seen in the somatic knockout cells. These experiments suggest that another protein and/or protein complex is binding to the CTCF sites in the BAG-1 promoter in cells overexpressing DNMT.
CTCFL/BORIS, a testis-specific protein that is structurally similar to CTCF, shares most of its homology in the ZF DNA-binding domain (9–11). This similarity suggested that the change in the shifted protein-DNA band in Fig. 4C (lanes 4–7 versus lanes 1–3) might be due to BORIS binding to the CTCF-binding sequence. To address this idea, ChIP analysis of the BAG-1 promoter with anti-BORIS antibody (Supplementary Figs. S2 and S3) was performed and showed an increase in BORIS binding to the BAG-1 promoter region in the DNMT1+, DMNT3B+, and DKI cells (Fig. 4D) compared with HCT116 cells. This observation matches the increase in BAG-1 gene expression (Fig. 1D and Supplementary Fig. S1). Very little, if any, BORIS binding to the BAG-1 promoter was observed in the analysis using the HCT116 somatic DNMT knockout cells (data not shown).
The above experiments suggest a role for BORIS in the regulation of BAG-1 expression; however, no cause and effect was established. To address this idea more critically, tissue culture cells were treated with shRNA to knockdown BORIS expression. Treatment of HCT116 cells with BORIS shRNA resulted in a decrease in BORIS protein levels (Fig. 5A), and ChIP analysis with an anti-BORIS antibody showed a decrease in BORIS binding to the BAG-1 promoter (Fig. 5B). Finally, BORIS shRNA knockdown changed the dimethyl-H3-K4/dimethyl-H3-K9 ratio at the BAG-1 promoter (Fig. 5C) to a nonpermissive chromatin status and decreased BAG-1 expression (Fig. 5D). The result of these experiments suggests a connection between BORIS binding to the BAG-1 promoter, histone methylation status, and BAG-1 expression.
To determine if BORIS and CTCF can alter BAG-1 expression, a series of transient cotransfection experiments were done, using either pBAG-1-Luc or a reporter plasmid containing three CTCF DNA-binding sites in tandem (p3x-CTCF-tk-Luc). Transient cotransfection using a CTCF expression vector (pCMV-CTCF) with pBAG-1-Luc (Fig. 6A) or p3x-CTCF-tk-LUC (Fig. 6B) showed decreased luciferase activity. In contrast, cotransfection using a BORIS expression vector (pCMV-BORIS) with pBAG-1-Luc (Fig. 6C) or p3x-CTCF-tk-Luc (Fig. 6D) showed increased luciferase activity. The results of these experiments extend the results above, suggesting that BORIS and CTCF use the same cis-acting DNA site to either repress (through binding by CTCF) or induce (through binding by BORIS) BAG-1 gene expression.
Epigenetic inheritance involves three primary mechanisms, all of which seem to play a role in carcinogenesis: DNA methylation; posttranslational modification of histones, including methylation, phosphorylation, or acetylation; and modifications by specific DNA-binding proteins, such as insulators binding to boundary elements (6, 7). One of the mechanisms that regulate chromatin compaction involves chromatin insulators, which demarcate specific semiautonomous genomic regions into units (25). The boundaries created by the insulators seem to block or alter the interaction between cis-acting enhancers or silencers in a position-dependent manner (26, 27). We have previously shown that genetic inhibition of DNMT down-regulates, as well as up-regulates, an equal number of genes (4). DNMT1 overexpression also increases the expression of as many genes as are silenced (data not shown), which others have also shown (5). However, the explanation for this observation is likely complex, and given the intricate nature of the epigenome, this may be due to indirect, as well as direct, effects. To address this, we used the BAG gene family, more specifically BAG-1, as a model to determine a potential mechanism to explain this gene expression pattern.
The results presented here suggest that BORIS binding to the BAG-1 promoter coincides with changes in the dimethyl-H3-K4/dimethyl-H3-K9 ratio, which indicates a permissive chromatin status for the expression of BAG-1. This was further suggested by the BORIS shRNA knockdown experiments that showed decreased BORIS binding to the BAG-1 promoter and decreased BAG-1 expression, as well as a nonpermissive chromatin status indicated by the change in the dimethyl-H3-K4/dimethyl-H3-K9 ratio. Furthermore, CTCF binding to the BAG-1 promoter corresponds with a decrease in gene expression, as well as a nonpermissive chromatin status reflected by a decrease in the dimethyl-H3-K4/dimethyl-H3-K9 ratio. A 5′and internal deletion analysis of the BAG-1 promoter clearly showed that the CTCF-binding sites in the BAG-1 CpG island are the target sequences for this regulatory process. As such, these results suggest that insulator proteins, such as BORIS and CTCF, may be critical to the regulation of BAG-1 and very likely to the other BAG genes as well.
These data showed that CTCF binding to the BAG-1 promoter increases as the DNMT gene is deleted. In addition, BORIS binding increases as a function of DNMT overexpression. However, the mechanism underlying these changes as a function of DNMT status is unclear. One possibility is that BORIS and CTCF compete for binding to CTCF-binding sites, and binding is a function of overall intracellular protein levels. However, the protein levels of BORIS and CTCF are roughly equal in both the DNMT somatic knockout and the DNMT-overexpressing cells (Supplementary Fig. S4). While this question, in the context of this manuscript, remains unanswered, it does seem that histone methylation, and likely chromatin remodeling, is associated with the specific insulator protein bound to the BAG-1 promoter.
It has been suggested that hypermethylation, presumably via the activation of DNMT genes, may play a role in carcinogenesis via the silencing of genes, such as p16 and RB (28, 29). In this model, the loss of gene function results in a genomic environment that is permissive for cellular damage, such as genomic instability, and this is an early step in carcinogenesis. However, cellular transformation also requires the activation of prosurvival and proproliferative genes. It is well established that in vitro DNMT1 overexpression can induce immortalized cells to transform; however, the mechanisms are complex (23). The results presented above suggest that two contrasting DNA methylation phenomena may coexist, causing epigenetic silencing of tumor suppressor genes, such as p16 and RB, and the activation of oncogenes, such as BAG family genes; both of which may be required for transformation. This idea is being pursued.
This observation is interesting because BAG genes seem to be evolutionarily conserved gene family with homologues found in yeast, amphibians, plants, and mammals. The human genome contains four BAG genes: BAG-1, BAG-3 (CAIR-1), BAG-4 (SOD), and BAG-6 (Scythe, BAT3; refs. 30–32). BAG proteins contain at least one copy of a conserved domain, allowing for an interaction with HSP70 (33, 34), and operate as cochaperones that ultimately affect cellular processes, including division, migration, differentiation, and pro-survival pathways (35–37). The BAG family proteins, thus, might be viewed as molecular bridges, with the NH2 terminal domains serving as scaffolds for interactions with specific proteins and the COOH terminal BAG domain serving as a recruit for HSP70 (38).
Our data raise an intriguing question: can changes in the regulation of the epigenome silence tumor suppressor genes and simultaneously activate proproliferative/survival pathways, creating a permissive environment for genomic instability and conferring a cell growth advantage? Although BORIS is initially thought to be primarily a testis-specific protein (11), it is also expressed in a wide variety of tumors (9), suggesting a potential role in carcinogenesis. Our results suggest that at least one potential mechanism for cellular transformation may involve BORIS-induced activation of a series of preprogrammed prosurvival/proproliferative genes or gene families that favor cell division, creating a permissive environment for division. These results also suggest that DNMT1 and DNMT3B expression seems to influence the reciprocity of BORIS and CTCF function to coordinately dictate gene expression via histone methylation as reflected by changes in the dimethyl-H3-K4/dimethyl-H3-K9 ratio.
Grant support: NIH grants CA65145 (A.P. Feinberg), CA72602 (D. Gius), and CA75556 (D. Gius). This research was supported in part by Intramural Research Program of the NIH National Cancer Institute Center for Cancer Research.
We thank Ina Rhee and Bert Vogelstein for the HCT116 knockout cell lines.