|Home | About | Journals | Submit | Contact Us | Français|
Gene silencing by aberrant epigenetic chromatin alteration is a well-recognized event contributing to tumorigenesis. While genetically engineered tumor-prone mouse models have proven a powerful tool in understanding many aspects of carcinogenesis, to date few studies have focused on epigenetic alterations in mouse tumors. To uncover epigenetically silenced tumor suppressor genes (TSGs) in mouse mammary tumor cells, we conducted initial genome-wide screening by combining the treatment of cultured cells with the DNA demethylating drug 5-aza-2′-deoxycytidine (5-azadC) and the histone deacetylase inhibitor trichostatin A (TSA) with expression microarray. By conducting this initial screen on EMT6 cells and applying protein function and genomic structure criteria to genes identified as upregulated in response to 5-azadC/TSA, we were able to identify 2 characterized breast cancer TSGs (Timp3 and Rprm) and 4 putative TSGs (Atp1B2, Dusp2, FoxJ1 and Smpd3) silenced in this line. By testing a panel of ten mouse mammary tumor lines, we determined that each of these genes is commonly hypermethylated, albeit with varying frequency. Furthermore, by examining a panel of human breast tumor lines and primary tumors we observed that the human orthologs of ATP1B2, FOXJ1 and SMPD3 are aberrantly hypermethylated in the human disease while DUSP2 was not hypermethylated in primary breast tumors. Finally, we examined hypermethylation of several genes targeted for epigenetic silencing in human breast tumors in our panel of ten mouse mammary tumor lines. We observed that the orthologs of Cdh1, RarB, Gstp1, RassF1 genes were hypermethylated, while neither Dapk1 nor Wif1 were aberrantly methylated in this panel of mouse tumor lines. From this study, we conclude that there is significant, but not absolute, overlap in the epigenome of human and mouse mammary tumors.
Tumorigenesis is a multi-step process stemming from increased oncogene activity in concert with constrained tumor suppressor gene (TSG) activity (Weinberg 1996). Such disease-producing changes in cellular phenotype result from dysregulation of gene expression and/or protein function attributable to genetic and/or epigenetic changes within the genome.
Several well-studied TSGs and other growth regulatory genes undergo de novo hypermethylation and consequential transcriptional silencing in human cancer cells. Aberrant cytosine methylation leading to gene silencing occurs at discreet, clustered 5′-CG-3′ dinucleotides (referred to as CpG islands) within the genome (Bird 1992). This cancer-associated CpG methylation is part of a complex set of epigenetic events that transform chromatin from a transcriptionally-active to an inactive state (Wolffe 2001) and epigenetic gene silencing is now widely recognized as either a causative or correlative event in tumor development (Jones and Baylin 2002; Robertson 2005) and is now widely regarded as one of the ‘hits’ in the Knudsen hypothesis leading to TSG inactivation (Jones and Laird 1999).
In recent years, numerous laboratories have developed tumor-prone mouse models owing to engineered over-expression of oncogenes or lost TSG expression. Clearly, these models have provided a powerful tool in studying tumorigenesis and unique insight into the molecular mechanisms that give rise to cancer in humans. For example, BRCA1, a TSG associated with breast and ovarian tumors in humans (Bishop 1999; Struewing et al., 1997), when conditionally deleted within the mammary epithelium of p53 heterozygous mice greatly increases mammary tumor incidence (Xu et al., 1999). This observation lends strong evidence as to the role BRCA1 plays in breast tumor suppression.
What remains less explored is the role that epigenetics plays during tumorigenesis in mice. While it seems straightforward to presume that gene silencing occurs in a parallel fashion in both human and mouse tumors, mice have not been widely utilized as a platform for either the discovery of epigenetically silenced genes or to study the role that gene silencing plays in the process of tumorigenesis. In one of the few studies to utilize a tumor-prone model system to dissect epigenetic events during disease progression, Yu et al (2005a) used Restriction Landmark Genomic Scanning (RLGS) to investigate aberrant DNA methylation events in a mouse model of T/natural killer acute lymphoblastic leukemia. These investigators observed nonrandom, aberrant DNA methylation occurring during tumorgenesis but was not present in benign preleukemic cells. Further, the inhibitor of DNA binding 4 (Idb4) gene was identified as a putative TSG silenced via promoter hypermethylation in mouse, and this gene was found epigenetically silenced in human leukemia as well. Similarly, Costello and colleagues found that SLC5A8, which encodes a sodium monocarboxylate cotransporter, was epigenetically silenced in both mouse and human brain tumors (Hong et al., 2005). These studies underscore the high potential of mice as a model system to understand the epigenetic basis of cancer and, potentially, other diseases as well.
A list of well-characterized TSG including BRCA1, CDKN2A, SFN, CDH1, and CST6 are targets for epigenetic silencing in breast cancer (Dobrovic and Simpfendorfer 1997; Esteller et al., 2000; Ferguson et al., 2000; Nass et al., 2000; Holst et al., 2003; Ai et al., 2006). However, at present, we are unaware of any genes that have been identified as targets for epigenetic silencing in mouse mammary tumors. To initially address this question, we conducted a screen for epigenetically silenced genes in a mouse mammary tumor line using an approach termed “pharmacological unmasking” (Esteller 2007). This often used approach examines changes in gene expression in response to the global DNA demethylating drug 5-aza-2′-deoxycytidine (5-azadC) and the histone deacetylase (HDAC) inhibitor trichostatin A (TSA) via microarray. Our screen was conducted using a mouse mammary cell line and identified ~1,000 genes that were upregulated ≥ 5-fold in response to 5-azadC/TSA in this line. Based on gene function and genomic architecture in mouse and man, several candidates identified in the initial screen were selected for further study.
Mouse mammary tumor lines 410.4, 410 and 66.1 were obtained from Dr. A. Fulton (University of Maryland, Baltimore, MD), D2F2 and 4T1 from Dr. F. Miller (Karamanos Cancer Foundation, Detroit, MI), LM2 and LM3 from Dr. E. Bal de Kier Joffé (Universidad de Buenos Aires, Argentina) and D1K2 and NeuT from Dr. B. Law (University of Florida, Gainesville, FL). The mouse mammary tumor line EMT6 was purchased from American Type Culture Collection (ATCC, Manassas, VA).
LM2 and LM3 were grown in Eagle’s MEM with 2 mM L-glutamine, 1.5g/L sodium bicarbonate, 0.1 mmol/L nonessential amino acids, and 1.0 mmol/L sodium pyruvate. 4T1 were cultured in RPMI 1640 with 2 mmol/L L-glutamine, 1.5 g/L sodium bicarbonate, 4.5 g/L glucose, 10 mmol/L HEPES, and 1.0 mmol/L sodium pyruvate. 410, 410.4, 66.1, NeuT, D2F2, and D1K2 were maintained in DMEM with 2mM L-glutamine, 0.1 mmol/L nonessential amino acids. EMT6 was cultured in Waymouth’s MB 752/1 medium with 2mM L-glutamine. All cell lines were maintained in medium supplemented with 10% (v/v) heat-inactivated fetal bovine serum (Invitrogen, Carlsbad, CA) at 37°C in 5% CO2 humidified atmosphere. Human breast tumor cell lines were cultured as previously outlined (Ai et al., 2006).
Where indicated, cell lines were treated with 5 μM 5-azadC for 5 days (fresh drug was added every 24 hours) followed by a 24-hour treatment with 100 nM TSA. All drugs were purchased from Sigma.
Fifteen fresh-frozen archival breast tumors were obtained from the University of Florida Shands Cancer Center Molecular Tissue Bank. All specimens and pertinent patient information were treated in accordance with policies of the institutional review board (IRB) of the University of Florida Health Sciences Center. Tumors analyzed in this study were examined by a surgical pathologist and identified as invasive breast adenocarcinoma (stage II or III). Where indicated, matched normal breast tissue was obtained from disease-free surgical margin.
Total RNA was extracted from EMT6 cells either untreated or treated with the global DNA demethylating agent 5-aza-2′-deoxycytidine (5-azadC) and the histone deacetylase inhibitor (HDAC) TSA. The quantity and purity of the extracted RNA was evaluated using a Nanodrop ND-100 spectrophometer (Nanodrop Technologies, Wilmington, DE) and its integrity measured using an Agilent Bioanalyzer. Five hundred nanograms (500 ng) of total RNA was amplified and labeled with a fluorescent dye (Cy3) using a Low RNA Input Linear Amplification Labeling kit (Agilent Technologies, Palo Alto, CA). Abundance and quality fluorescently labeled cRNA was also assessed using a Nanodrop ND-1000 spectrophotometer and an Agilent Bioanalyzer. The labeled cRNA (1.65 μg) was hybridized to an Agilent Mouse Whole Genome Oligo Microarray (Agilent Technologies, Inc., Palo Alto, CA) for 17 hrs prior to washing and scanning. Data was extracted from scanned images using Agilent’s Feature Extraction Software.
RT-PCR was carried out according to established protocols (Ai et al., 2006). Briefly, untreated and 5-azadC/TSA treated mouse mammary tumor cells were homogenized in Trizol (Invitrogen), and total RNA purified according to the manufacturer’s instructions. First-strand cDNA synthesis was carried out using SuperScript III reverse transcriptase (Invitrogen). Subsequently, this cDNA was used in PCR reactions using Timp3, Rprm, Smpd3, Atp1B2, Dusp2 and FoxJ1-specific oligonucleotide primers as listed in Supplemental Table I. Mouse glyceraldehyde-3-phosphate dehydrogenase (Gapdh) served as a control to assure equivalent RNA in each reaction. Following PCR, reactions were analyzed by electrophoresis on 10% polyacrylamide gels followed by staining with ethidium bromide.
Genomic DNA (gDNA) was isolated from human breast tumor tissue samples and cultured human and mouse cell lines using a Qiagen Blood and Cell Culture DNA kit (Qiagen, Inc., Valencia, VA) and stored at −20°C before use. DNA was modified with EZ DNA Methylation kit (ZYMO Research Co., Orange, CA) according to the manufacturer’s instructions. Briefly, ~1.0 μg denatured gDNA was treated with sodium bisulfite at 50°C for ~16 hours in the dark. Following this, samples were applied to supplied columns, desulfonation conducted, and DNA was subsequently eluted in a volume of 20 μL with supplied elution buffer. One μL of bisulfite-converted gDNA solution was generally used in subsequent PCR reactions.
Primers used for Methylation Specific-PCR (MSP) analyses are listed in Supplemental Table II. MSP reactions were conducted using Qiagen HotStarTaq DNA polymerase and supplied 1X PCR buffer supplemented with 0.1 mmol/L dNTPs, 2.5 mmol/L MgCl2, and 0.5 μmol/L each of forward and reverse primers and bisulfite-converted gDNA. Following thermocycling, PCR products were loaded 10% polyacrylamide gels and visualized using ethidium bromide.
Human placental gDNA (Biochain Institute, Hayward, CA) and mouse gDNA isolated from tail snips of wildtype C57BL/6 mice were used in primer validation experiments of human and mouse MSP, respectively and where indicated DNA was methylated in vitro using SssI methylase (NEB, Ipswich, MA) as outlined previously (Vo et al., 2004).
PCR reactions were conducted on bisulfite-converted gDNA as described above. Supplemental Table III summarizes pyrosequencing primers and sequence analyzed in each gene. Following PCR amplification using a biotinylated reverse primer, 5–20 μL of PCR product was mixed with 2 μL of streptavidin-coated sepharose beads (GE Heathcare Bio-Sciences AB, Uppsala, Sweden), 40ul of binding buffer (10 mM Tris-HCl, 2 M NaCl, 1mM EDTA, 0.1% Tween 20, pH 7.6), and brought up to 80 μL total volume with dH2O. This mixture was immobilized on streptavidin beads by shaking constantly at 1400 rpm for 10 minutes at RT. The beads were subsequently harvested using a vacuum preparation workstation and appropriate sequencing primer added to a final concentration of 330 nM. The sequencing primers were subsequently annealed to the template by incubating the samples at 80°C for 2 min and cooled at RT for 10 min prior to analysis. Pyrosequencing was conducted using a PyroMark MD system (Biotage AP, Uppsala, Sweden), and sequence analysis conducted using supplied software.
The focus of this investigation was to compare mouse and human mammary cancer for potential overlap in tumor epigenotype across these two species. Our screening strategy entailed utilizing the mouse mammary tumor line EMT6 in a pharmacological unmasking screen (Esteller 2007). Briefly, EMT6 cells were treated with the global DNA demethylating agent 5-aza-2′-deoxycytidine (5-azadC) and the histone deacetylase (HDAC) inhibitor trichostatin A (TSA) to de-repress epigenetically silenced genes. After drug treatment, which was found to have minimal toxicity in EMT6 cells by MTT assay (data not shown), total RNA was harvested from two replicate cultures of drug treated and untreated EMT6 cells and used to probe Agilent Mouse Whole Genome microarrays.
Of the ~44,000 transcripts measured on these microarrays, an average of 54% were detected as significantly above background levels in each of the hybridizations (range: 51 to 57%). The intensity values from the replicate samples for both the treated and untreated cultures had Pearson correlation values greater than 0.97, suggesting that the reproducibility of the results was very high. Reproducibility of the within-group samples, along with the differences between the two groups, were also apparent when the z-score transformed intensity values for the 2,252 transcripts found to be differentially expressed (984 upregulated, 1,268 downregulated) using an absolute fold change criterion of 5-fold and a log ratio p-value of 0.001 were compared (Fig 1).
Following the initial screen by expression microarray, we established a series of bioinformatic criteria to narrow our search for epigenetically silenced tumor suppressor genes (TSGs) in the EMT6 cell line. From the entire list of genes measured by the microarray, we first excluded genes that showed a moderate change in expression in response to 5-azadC/TSA. Therefore, we required that a candidate gene demonstrate at least a 5-fold increase with a log ratio p-value of 0.001 to be considered further. This narrowed our focus to 984 genes of interest (Fig 2). (A listing of all 984 genes identified is given in Supplemental Table IV.)
Next, we used protein function and genomic structure criteria to further define potential TSGs that are targets for epigenetic silencing in EMT6 cells. We were principally interested in studying genes that encode proteins with an established tumor suppressor activity(ies), or a function consistent with tumor suppressive activity. Specifically, we included in our study gene products where previously published studies support a gene function that either antagonizes cell growth, activates apoptosis or suppresses anti-apoptotic processes, restrains cellular invasiveness, or inhibits angiogenic processes.
We next excluded from the list of putative/characterized TSGs those genes that do not possess a canonical CpG island within 5′ flanking region in both the human and mouse orthologs using the UC Santa Cruz genome browser (http://genome.ucsc.edu) to investigate genomic architecture of various candidate genes. We used the algorithm within the UCSC genome browser that scores the presence of CpG islands in genomic DNA using criteria originally established by Gardiner-Garden and Frommer (Gardiner-Garden and Frommer 1987). Our rationale for establishing these criteria is that genes which possess both features (i.e., putative TSG activity and the presence of a CpG island that overlaps the translational start site of the gene) are likely to identify high-value candidate targets for epigenetic silencing. Of the 984 genes initially identified as being upregulated ≥5-fold in our expression microarray screen, this bioinformatic investigation identified 37 genes meeting the outlined criteria (see Supplemental Table V). Of this, 26 genes were further investigated in this study.
Our next test was to confirm 5-azadC/TSA-induced upregulation of gene expression by RT-PCR. This revealed 12 genes where the gene upregulation detected by microarray was confirmed by RT-PCR and 14 genes whose upregulation could not be independently confirmed using an independent set of RNA samples isolated from drug treated/untreated EMT6 cells. Methylation-Specific PCR (MSP), a common methodology to examine DNA methylation (Herman et al., 1996), was subsequently used to judge methylation within the CpG islands associated with remaining 12 candidate genes. This analysis revealed that six genes exhibited no methylation as judged by MSP. This may indicate that the 5-azadC/TSA-induced increase in expression of these genes is indirect in nature and/or that silencing of these genes is controlled by chromatin modifications other than CpG methylation (e.g., transcriptionally-repressive histone modifications). The later of these possibilities is quite possible since our drug treatment regimen included the histone deacetylase inhibitor TSA. As a result of these efforts our focus narrowed to six genes: Tissue Inhibitor of Metalloproteinase 3 (Timp3), TP53-dependent G2 Arrest Mediator Candidate also termed Reprimo (Rprm), Sphingomyelin Phosphodiesterase 3 Neutral (Smpd3), ATPase, Na+/K+ Transporting, Beta 2 Polypeptide (Atp1B2), Dual-Specificity Phosphatase 2 (Dusp2), and ForkheadboxJ1 (FoxJ1).
Timp3 and Rprm are well-characterized TSGs silenced in human breast cancer (Bachman et al., 1999; Takahashi et al., 2005). To test how commonly these genes are epigenetically silenced in mouse mammary tumor lines, we acquired 9 distinct cultured mouse mammary tumor lines (designated 410.4, 410, LM3, LM2, D1K2, NeuT, D2F2, 66.1 and 4T1), in addition to the EMT6 cell line. As outlined above, RT-PCR was used to confirm increased expression in response to 5-azadC/TSA; however, due to high-levels of toxicity to this drug cocktail, we found that only four of these cell lines (EMT6, 410.4, 410, and LM3) showed sufficiently low toxicity (>75% viability 48 hrs after 5 μM 5-azadC treatment) to be appropriate subjects for this experiment. These lines were cultured in the presence or absence of 5-azadC and TSA, total RNA harvested and subsequently cDNA synthesized. Results of this RT-PCR experiment clearly indicated that the abundance of Timp3 and Rprm transcripts increased substantially following treatment of each line with 5-azadC/TSA (Fig 3A). Parallel reactions using primers specific for the mouse form of Gapdh confirmed equivalent RNA abundance in each PCR reaction.
We next investigated the methylation status of Timp3 and Rprm in our full panel of mouse mammary tumor line by designing MSP primers using a publically available on-line primer design program (www.mspprimer.org/cgi-mspprimer/design.cgi). All primer sets designed for the analysis for DNA methylation status (both MSP and pyrosequencing analysis) were designed to assay DNA methylation at, or in close proximity to, the transcriptional start site (TSS) of candidate genes since this region is most likely to have undergone dense methylation during the process of epigenetic silencing (Bird 2002; Jones and Baylin 2002).
Primers were designed specific for both the mouse Timp3 and Rprm genes and validated using mouse genomic DNA extracted from tail clippings of C57BL/6 mice. This mouse genomic DNA was either methylated in vitro with SssI methylase or not methylated prior to bisulfite conversion. MSP reactions conducted using this control DNA indicated that our designed primers discriminate between the Timp3 and Rprm genes in either a methylated or unmethylated state (Fig 3B). Subsequently, these MSP primer sets were used to probe the methylation status of the Timp3 and Rprm gene in our panel of 10 mouse mammary tumor cell lines (Fig 3C). With the exception of D2F2, a methylated-specific (M) amplicon was detected when the Timp3 gene was analyzed using bisulfite-converted genomic DNA isolated from each cell line. While amplification of both unmethylated and methylated-specific MSP products is a common finding in MSP analysis (Herman et al., 1996), potentially attributable to heterogeneous promoter methylation within the tumor cell population, this result indicates the presence of methylated CpG within the Timp3 gene in at least a subset of cells of each line except D2F2 where the Timp3 promoter appears to be unmethylated.
As an independent means of quantitatively analyzing DNA methylation, we conducted pyrosequencing on mouse Timp3 and Rprm genes in the EMT6 and LM3 cell lines. Pyrosequencing, like MSP, analyzes bisulfite-converted genomic DNA as a template for PCR; however, pyrosequencing is conducted on PCR products amplified using primers that do not contain internal CpG dinucleotides. Thus, amplification occurs in a manner that is unbiased by DNA methylation status. Subsequently, methylation at distinct CpGs can be determined quantitatively in a pool of genomic DNA (Colella et al., 2003; Tost et al., 2003).
The primer design developed for the mouse Timp3 gene allowed for analysis of 9 CpG dinucleotides within a ~45 bp portion of the CpG island associated with this gene. Pyrosequencing analysis, consistent with MSP results, indicated that Timp3 was methylated in both EMT6 and LM3 cell lines. Moreover, all the CpG dinucleotides investigated using this approach displayed methylation suggesting a high level of methylation density, although CpG methylation was not uniform over the region interrogated. The average CpG methylation measured in this region was 79% in EMT6 and 76% in LM3 (Fig 3D). The pyrosequencing protocol developed for Rprm analyzed 6 CpG dinucleotides over a ~35 bp region of the CpG island within this gene. Similar to Timp3 analysis, all CpG within this region of the Rprm gene exhibit significant methylation with average CpG methylation of 84% and 86% in EMT6 and LM3 cells, respectively (Fig 3E).
Pyrosequencing analysis was also conducted on control (mouse tail) DNA samples (Fig 3D,E). For both Timp3 and Rprm genes, average CpG methylation measured was 7%, indicating very low levels of DNA methylation in this control DNA sample. This result indicates that methylation of the CpG islands associated with the Timp3 and Rprm genes in the mouse genome is a tumor-associated event. Moreover, when our DNA methylation analysis is combined with the results of our RT-PCR experiments using the DNA demethylating agent 5-azadC, we conclude that the Timp3 and Rprm genes are targeted for epigenetic silencing in mouse mammary tumor lines through aberrant gene methylation. Considering most of the 10 distinct cell lines display this feature, silencing of these genes is apparently a common event in cultured mouse mammary tumor cells. Since both Timp3 and Rprm are also targets for epigenetic silencing in human breast tumor line and primary tumors, these results reinforce the validity of our expression microarray screen for epigenetically silenced genes in mouse mammary tumor lines and establish a point of epigenetic overlap between mouse and human mammary gland tumors.
In addition to Timp3 and Rprm, our screen identified four genes (Atp1B2, Smpd3, Dusp2 and FoxJ1) that were strong candidates for epigenetic silencing in the EMT6 line. RT-PCR was conducted on EMT6, 410.4, 410, and LM3 cells cultured in the presence or absence of 5-azadC/TSA using primers designed specifically for the mouse orthologs of these genes. RT-PCR analysis of the Atp1B2 transcript in EMT6 cells showed a clear upregulation of this transcript in response to 5-azadC/TSA (Fig 4A). Conversely, 410.4 and LM3 cells displayed no detectable upregulation of Atp1B2 expression and analysis of 410 cells showed a slight (~2-fold) increase in Atp1B2 expression following 5-azadC/TSA treatment. RT-PCR results showed strong upregulation of Dusp2 expression in 5-azadC/TSA treated EMT6, 410 and 410.4 cells while LM3 cells showed no increased expression of this gene in drug treated LM3 cells. Smpd3 showed marked increases in gene expression in each cell line in response to 5-azadC/TSA. Similarly, FoxJ1 showed clear upregulation in drug treated EMT6 and LM3 cells while expression was more modestly increased in drug treated 410 and 410.4 cells.
Based on the positive RT-PCR results outlined above, we designed MSP primers for the mouse forms of the Atp1B2, Smpd3, Dusp2 and FoxJ1 genes. As shown (Fig 4B), validation experiments using in vitro methylated/unmethylated mouse genomic DNA confirmed that MSP primer sets were capable of determining the methylation state of each gene. In EMT6 and 410 cells, MSP analysis of the Atp1B2 gene indicated abundant M-specific amplicon indicating that this gene is hypermethylated in these cell lines. This result is consistent with the increased expression of this gene observed in RT-PCR experiments. In 410.4 cells, slight amplification of the M-specific band was noted and no M-specific amplicon was detected when genomic DNA isolated from LM3 cells was assayed, indicating that methylation of this gene is not a prominent feature in this cell line. MSP analysis of the other lines in our panel of mouse mammary tumor cells indicated that only 66.1 cells display prominent amplification of the M-specific product from the Atp1B2 gene. Reduced amplification of the M-specific product was observed in 4T1 and D2F2 cells while no M-specific product was obtained from the other 3 lines (LM2, D1K2 and NeuT) in our panel.
MSP analysis of the Dusp2 gene in EMT6, 410.4 and 410 cells indicated strong amplification of the M-specific product (Fig 4C). When combined with the clear upregulation of the Dusp2 transcript when each of these lines were treated with 5-azadC/TSA, this result indicates that this gene is epigenetically silenced in these lines. Conversely, no aberrant methylation of the Dusp2 gene in LM3 cells was observed. In the remaining six lines in our panel, methylation of the Dusp2 gene was observed in D1K2, NeuT and 66.1 cells and no M-specific amplicon was observed when LM2, D2F2 and 4T1 cells were assayed by MSP.
When the Smpd3 gene was analyzed by MSP, methylation of this gene was observed in all of the cell lines in our panel with the exception of NeuT and D2F2 (Fig 4C). Since we observed upregulation of the Smpd3 in each of the four cell lines treated with 5-azadC/TSA, MSP results are consistent with epigenetic silencing of the Smpd3 gene in cultured mammary tumor lines. Similarly, all ten cultured mammary tumor lines displayed methylation of the FoxJ1 gene as judged by MSP. Again, this is consistent with our observation of increased FoxJ1 expression in 5-azadC/TSA treated cells and indicates that the FoxJ1 is commonly targeted for epigenetic silencing in cultured mouse mammary tumor lines. In sum, these lines of experimentation indicate that the Atp1B2, Dusp2, Smpd3 and FoxJ1 genes are epigenetically silenced in mouse mammary tumor lines; however, the occurrence of these aberrant events is variable within this panel of cultured lines.
The ATP1B2, DUSP2, SMPD3 and FOXJ1 genes have not previously been characterized as targets for epigenetic silencing in human breast tumors. Thus, we examined the methylation status of each of these genes in a panel of eight cultured human breast cancer cell lines by designing MSP primer sets for the human ortholog of each gene. Validation experiments using in vitro methylated/unmethylated human placental genomic DNA indicated that each of these primer sets can discriminate between methylated and unmethylated forms of these four genes (Fig 5A). MSP analysis of the ATP1B2 gene in our panel of human breast tumor lines indicated robust amplification of the methylation-specific product in MCF-7, MDA-MB-435 and BT-474 cells while MDA-MB-468 cells show moderate accumulation of the M amplicon and low level amplification of the M product was observed in MDA-MB-231. Other lines in our panel (T-47D, BT-549 and SKBR3) displayed no evidence of ATP1B2 methylation. MSP analysis of the SMPD3 gene determined that MDA-MB-231, MDA-MB-435 and BT-474 cells display methylation of this gene, while the other cell lines show no evidence of SMPD3 methylation. Similarly, DUSP2 was found to be methylated in MCF7, MDA-MB-231 and BT-549 cells but not in T-47D, MDA-MB-468, SKBR3, MDA-MB-435 and BT-474 cells. Finally, MSP analysis of the FOXJ1 gene indicates strong amplification of the M-specific amplicon in MDA-MB-231, MDA-MB-435 and BT-474 cells. Weak amplification of the M-specific FOXJ1 amplicon was observed in MCF-7, MDA-MB-468 and BT-549. No amplification of the FOXJ1 M-specific product was observed in either T-47D or SKBR3 cells.
To confirm the MSP results obtained when the methylation status of the ATP1B2, DUSP2, SMPD3 and FOXJ1 genes was scored by MSP, we conducted pyrosequencing. The pyrosequencing assay developed for the human ATP1B2 gene scores methylation of seven CpG dinucleotides within a ~30 nt region of the CpG island within the 5′ end of the human ATP1B2 gene. When genomic DNA harvested from MCF-7 cells was analyzed by pyrosequencing, we observed low (3%) to moderate (29%) levels of CpG methylation within this cell line in the ATP1B2 gene (Fig 5C). Overall, average CpG methylation measured within the ATP1B2 gene was 17% in this cell line. In BT-474 cells, we observed markedly higher methylation within the ATP1B2 gene with average CpG methylation of 34%. Pyrosequencing of control (human placental) DNA showed a low level of CpG methylation (4.6%) in this gene.
Pyrosequencing analysis of the SMPD3 gene showed high average levels of CpG methylation (56% and 71%) within the SMPD3 gene in MDA-MB-231 and BT-474 cells, respectively (Fig 5D). Conversely, low levels of CpG methylation (0.8%) of the SMPD3 gene was measured in control DNA. Consistent with MSP results, we observed methylation of the CpG island associated with the human DUSP2 gene in MDA-MB-231 (88%) and BT-549 (44%) cells with low (5.5%) CpG methylation measured in control DNA (Fig 5E). We also observed CpG methylation within the human FOXJ1 gene in MDA-MB-231 (40%) and BT-474 (63%) cells and control DNA displayed low (5.5%) CpG methylation within FOXJ1 (Fig 5F). Taken together, both MSP and pyrosequencing analyses indicate that the ATP1B2, DUSP2, SMPD3 and FOXJ1 are aberrantly methylated, albeit at differing frequencies, in cultured human breast tumor lines.
Next, we conducted MSP analysis to examine ATP1B2, DUSP2, SMPD3 and FOXJ1 gene methylation in primary breast tumor/normal mammary tissue samples. Fifteen frozen human breast tumors were obtained and patient-matched disease-free breast tissue were obtained with two of these tumor samples (Tumor72 and Tumor75). Genomic DNA was isolated, bisulfite modified, and used in MSP reactions with primers specific for the ATP1B2, DUSP2, SMPD3 and FOXJ1 genes. In all normal breast tissue samples, we observed exclusive amplification of the unmethylated-specific (U) amplicon, indicating that these genes are not methylated in normal mammary tissue (Fig 6A). When patient-matched breast tumor samples were analyzed by MSP, we observed methylation of the SMPD3 gene in both Tumor72 and Tumor75. Further, methylation of the FOXJ1 gene was observed in Tumor72. No methylation of either DUSP2 or ATP1B2 was observed in either Tumor72 or Tumor75.
Figure 6B shows a representative subset of additional breast tumors analyzed by MSP. We observed methylation of the SMPD3 gene in Tumor 62, 66 and 70. Overall, we observed SMPD3 methylation in 60% of breast tumors investigated (9 out of 15). We observed methylation of the ATP1B2 gene in Tumor 61, 68 and 70 and, overall, detected a 20% rate (3/15) of ATP1B2 methylation in breast tumors. In contrast, FOXJ1 was found to be commonly methylated in human breast tumors; specifically, we observed that 11 out of 15 tumors surveyed (73%) displayed aberrant FOXJ1 methylation as scored by MSP. Taken together, MSP analysis suggests that promoter hypermethylation the ATP1B2, SMPD3 and FOXJ1 occur in primary human breast tumors with varying occurrence.
Despite the fact that we documented hypermethylation of DUSP2 in cultured human breast cancer lines, this gene was found to be unmethylated in all human breast tumors examined (Fig 6B). DUSP2 encodes a phosphatase (also known as PAC1) that dephosphorylates and inactivates mitogen-activated protein (MAP) kinases (Chu et al., 1996) and is a transcriptional target of p53 in signalling apoptosis (Yin et al., 2003). These properties make this gene an excellent candidate TSG and DUSP2 hypermethylation has been reported in prostate cancer (Yu et al., 2005b). However, we did not observe methylation of this gene in primary tumors leading us to conclude that epigenetic silencing of DUSP2 in human breast cancer is likely a rare event.
Finally, we determined the methylation status of well-defined TSGs known to be targets for epigenetic silencing in human breast cancer in our panel of cultured mouse mammary tumor lines. We designed MSP primer sets to assay methylation within the CpG islands associated with the mouse orthologs of the Cdh1, RarB, Gstp1, Dapk1, RassF1 and Wif1 genes. Validation experiments using in vitro methylated/unmethylated mouse genomic DNA indicate that each of these primer sets can delineate gene methylation status (Fig 7A).
Our MSP analyses indicate that RassF1 and RarB were commonly methylated in our panel of mouse mammary tumor lines with 90% and 80% showing significant methylation, respectively (Fig 7B). In contrast, methylation of Cdh1 and Gstp1 were far less commonly encountered in this panel of cell lines with 30% and 10%, respectively, displaying methylation of these genes. In all mouse mammary cell lines, we observed exclusive amplification of the unmethylated-specific products for Dapk1 and Wif1 genes indicating that neither gene is methylated in this panel of mouse cell lines.
By coupling genome-wide screening with bioinformatics analyses and inspection of publically available mouse and human genome sequence, we were able to characterize 6 candidate genes as targets for epigenetic silencing in the EMT6 line. These genes are: Timp3, Rprm, Smpd3, Dusp2, Atp1B2, and FoxJ1. Of these, Timp3 and Rprm have already been characterized as TSGs and targets for epigenetic silencing in human breast and other tumor types (Esteller et al., 2001; Takahashi et al., 2005; Kim et al., 2006;). The fact that our initial screening identified these TSGs as silencing candidates in EMT6 underscores the successful nature of our initial screen. Moreover, since analysis of 10 distinct mouse mammary tumor lines shows that both Timp3 and Rprm are commonly silenced in cell lines of this type, we conclude that both human and mouse mammary tumors share silencing of these tumor suppressor genes in common.
A combination of RT-PCR and MSP confirmed that Smpd3, Dusp2, Atp1B2, and FoxJ1 are silenced in mouse mammary tumor lines. Our findings indicate that Atp1B2 is a target, although likely an uncommon one, for silencing in mouse mammary cell lines. The other 3 genes identified in our screen appear to be more commonly hypermethylated in mouse mammary tumor lines. We observed significant amplification of the M-specific amplicon specific for Dusp2 in 60% of the cell lines tested, Smpd3 in 80%, and FoxJ1 in 100%. When a panel of eight human breast tumor cell lines were assayed for methylation by pyrosequencing and/or MSP, we observed substantial methylation of the ATP1B2 gene in 50% (4/8) cell lines assayed. Also, SMPD3, DUSP2, and FOXJ1 methylation was detected in 38% (3/8) of the cell lines surveyed. Of note, methylation of each of these genes, with the exception of DUSP2, was observed in human primary breast tumors and MSP analysis of normal breast tissue showed methylation of these genes to be a disease-associated event. The fact that DUSP2 methylation was observed in cultured human cell lines but not primary tumors strongly suggests that silencing of this gene is likely a rare event in human breast tumors.
As part of our study we also examined the methylation status of 6 genes known to be targets of epigenetic silencing in human breast tumors in our panel of cultured mouse mammary tumor lines. In this investigation the mouse ortholog of the Cdh1, Gstp1, RarB, Wif1, RassF1, and Dapk1 genes were assayed by MSP. We observed, to various extents, that Cdh1, Gstp1, RarB, and RassF1 were hypermethylated in our panel of mouse tumor lines. However, we obtained no evidence that either the Wif1 or Dapk1 genes are aberrantly methylated in any of these lines.
When the results of our studies on various epiegentically silenced genes in human and mouse cultured mammary tumor cell lines and human primary breast tumors are considered collectively, we conclude that there is significant overlap in the mammary tumor cell epigenome between these two species. Of the twelve genes studied, only Wif1 and Dapk1 were found not be silenced in both species. While Dusp2 was found to be commonly hypermethylated in mouse cell lines, methylation of this gene was less commonly observed in cultured human lines and no evidence for hypermethylation was observed in primary human breast tumors. These results indicate that some genes targeted for epigenetic silencing in human breast tumors are also aberrantly methylated in mouse mammary tumor lines demonstrating considerable but not absolute overlap in the mammary tumor epigenome between these two species.
Sphingomyelinases (SMases) belong to a multigene family of phosphodiesterases. Each of the three well-characterized SMases (SMPD1, SMPD2, and SMPD3) hydrolyze sphingomyelin to ceramide and phosphocholine. The acid sphingomyelinase SMPD1 is located in the endolysosomal compartment, while the neutral SMases (nSMases) SMPD2 and SMPD3 are located within the endoplasmic reticulum membranes and in the Golgi apparatus, respectively (Stoffel et al., 2007). nSMases are a major intracellular regulator of ceramide, a bioactive lipid.
Several lines of evidence support the view that the ability of nSMases contribute to proliferation of cancer cells and/or to their invasiveness. Of note, the SMPD3 locus is present, as well as six members of the cadherin family, with chromosome 16q22.1, and genetic alterations in this region have been implicated in the progression of various cancers, notably breast cancer (Chalmers et al., 2001; Driouch et al., 1997). Moreover, it has recently been shown that reduction of ceramide levels due to inhibition of nSMase in Hs578T breast adenocarcinoma cells overexpressing caveolin-1 results in a significant enhancement in Akt activation a well-characterized pro-growth, pro-survival mechanism (Wu et al., 2007). Ample evidence also supports a role for ceramide in the activation of apoptosis. For example, Luberto et al (2002) showed inhibition of TNF-α-induced cell death in MCF7 cells using a novel inhibitor of nSMase activity. Data obtained in this study indicates that ceramide induces conformational changes in Bax and activates mitochondrial phosphatase 2A, which, in turn, dephosphorylates Bcl-2. These two events clearly link ceramide and nSMases to the activation of cell death. Similarly, studies link ceramide to the regulation of cell cycle progression by inducing growth arrest by promoting dephosphorylation of the Rb tumor suppressor (Alesse et al., 1998).
In this report, we show that the SMPD3 gene in both mouse and man is a target for epigenetic silencing in mammary tumors. To our knowledge, this is the first report of the SMPD3 gene being targeted for silencing during tumorigenesis. Functionally, this aberrant event will result in decreased ceramide production in response to normal physiological cues promoting dysregulation of both cellular death and growth programs. Thus, this event has clear implications in the process of breast tumorigenesis.
The protein encoded by ATP1B2 is a member of the Na+/K+ ATPase beta chain protein family. Na+/K+ ATPases, which are composed of two subunits (a large catalytic subunit (alpha) and a smaller glycoprotein subunit (beta)), are integral membrane proteins responsible for establishing and maintaining the electrochemical gradients of Na+ and K+ ions across the plasma membrane. This process is essential for osmoregulation, for sodium-coupled transport of a variety of organic and inorganic molecules, and for electrical excitability of nerve and muscle (Dobretsov and Stimers 2005).
The adhesion molecule on glia (AMOG) protein was originally identified as a neural cell adhesion molecule and controls neuron–astrocyte adhesion as well as neural cell migration in vitro (Antonicek et al., 1987), was later found to be encoded by the ATP1B2 gene (Gloor et al., 1990). Others (van den Boom et al., 2006) recently showed that the ATP1B2 gene was targeted for epigenetic silencing via aberrant gene methylation in malignant glioma cells. This finding, supported by the observation that reduced ATP1B2 expression is associated with increased tumor grade (Senner et al., 2003), led these investigators to propose that ATP1B2 silencing could be a significant event in glioma development. Our results indicate that Atp1B2 is epigenetically silenced in a subset of cell lines in our panel of mouse mammary tumor lines and its human ortholog is subject to gene hypermethylation in both human breast tumor cell lines and primary tumors. Given the role that this gene product plays in controlling glial migration, these findings suggest that silencing of ATP1B2 may influence breast tumor cell migration as well. This compelling notion requires further investigation.
FOXJ1 is a member of the fork head gene family of transcription factors originally identified in Drosophila, and a large family of FOX genes are currently characterized in humans (Katoh 2004). In mammalian cells, FOXJ1 has been shown to suppress NFκB transcriptional activity in vitro and suppress inflammation in vivo (Lin et al., 2004). While NFκB is a key regulator of immune response, dysregulation of NFκB often drives deregulated growth, resistance to apoptosis, and increased metastatic potential (Lu and Stark 2004). Since FOXJ1 functions as an antagonist to NFκB activity it is reasonable to suggest that this molecule functions as a TSG, and others have commented that FOXJ1 is an excellent candidate for anti-cancer therapy (Chen 2004). While the role of FOXJ1 in cancer suppression remains to be explored, our finding that this gene is a common target for aberrant hypermethylation in both cultured cells and primary tumors strongly suggests that FOXJ1 silencing is potentially an important event in breast tumorigenesis.
In conclusion, our study has documented considerable, but not absolute, overlap exists between the epigenome of mouse and human mammary tumors. This finding supports the possible utility of using mouse as a model system to understand early epigenetic events during tumorigenesis as well as the general impact of aberrant chromatin remodeling events in cancer. Moreover, our study revealed several putative TSG that are novel targets for epigenetic silencing in human breast cancer.
The authors would like to thank Drs. Lingbao Ai and Wan-Ju Kim for technical assistance and advise during the completion of this study and Dr. Martha Campbell-Thompson for supplying breast tumors.
Supported, in part, by a grant from the NIH (R21 CA102220) to KDB.