Abstract
ARHI is a maternally imprinted tumor suppressor gene that maps to a site on chromosome 1p31 where loss of heterozygosity has been observed in 40% of human breast and ovarian cancers. ARHI is expressed in normal ovarian and breast epithelial cells, but ARHI expression is lost in a majority of ovarian and breast cancers. Expression of ARHI from the paternal allele can be down-regulated by multiple mechanisms in addition to loss of heterozygosity. This article explores the role of DNA methylation in silencing ARHI expression. There are three CpG islands in the ARHI gene. CpG islands I and II are located in the promoter region, whereas CpG island III is located in the coding region. Consistent with imprinting, we have found that all three CpG islands were partially methylated in normal human breast epithelial cells. Additional confirmation of imprinting has been obtained by studying DNA methylation and ARHI expression in murine A9 cells that carry either the maternal or the paternal copy of human chromosome 1. All three CpG islands were methylated, and ARHI was not expressed in A9 cells that contained the maternal allele. Conversely, CpG islands were not methylated and ARHI was expressed in A9 cells that contained the paternal allele of human chromosome 1. Aberrant methylation was found in several breast cancer cell lines that exhibited decreased ARHI expression. Hypermethylation was detected in 67% (6 of 9) of breast cancer cell lines at CpG island I, 33% (3 of 9) at CpG island II, and 56% (5 of 9) at CpG island III. Hypomethylation was observed in 44% (4 of 9) of breast cancer cell lines at CpG island II. When methylation of CpG islands was studied in 20 surgical specimens, hypermethylation was not observed in CpG island I, but 3 of 20 cases exhibited hypermethylation in CpG island II (15%), and 4 of 20 cases had hypermethylation in CpG island III (20%). Treatment with 5-aza-2′-deoxycytidine, a methyltransferase inhibitor, could reverse aberrant hypermethylation of CpG island I, II and III and partially restore ARHI expression in some, but not all of the cell lines. Treatment with 5-aza-2′-deoxycytidine partially reactivated ARHI expression in cell lines with hypermethylation of CpG islands I and II but not in cell lines with partial methylation or hypomethylation of these CpG islands. To test the impact of CpG island methylation on ARHI promoter activity more directly, constructs were prepared with the ARHI promoter linked to a luciferase reporter and transfected into SKBr3 and human embryo kidney 293 cells. Methylation of the entire construct destroyed promoter activity. Selective methylation of CpG island II alone or in combination with CpG island I also abolished ARHI promoter activity. Methylation of CpG I alone partially inhibited promoter activity of ARHI. Thus, hypermethylation of CpG island II in the promoter region of ARHI is associated with the complete loss of ARHI expression in breast cancer cells. Other epigenetic modifications such as hypermethylation in CpG island III may also contribute to the loss of ARHI expression.
INTRODUCTION
Alterations in DNA methylation can contribute to human oncogenesis. An estimated 50% of human genes have clusters of CpG dinucleotides (CpG islands) in their 5′-regulatory sequences. In the vast majority of these genes, CpG islands are not methylated. In human cancers, aberrant methylation has been observed, including global hypomethylation, increased DNA methyltransferase activity, and local DNA hypermethylation of CpG islands (1 , 2) . Aberrant methylation of promoter-associated CpG islands has emerged as a distinct molecular pathway leading to malignant transformation (3 , 4) . By silencing key regulatory genes such as tumor suppressor genes, DNA methylation can provide the epigenetic equivalent of mutation/deletion during oncogenesis. Several tumor suppressor genes are transcriptionally silenced or down-regulated by methylation of CpG dinucleotides in the promoter region. Decreased gene transcription has been correlated with CpG methylation in the promoter region in putative tumor suppressor genes, including RB (5) and P16 (6) . Hypermethylation of the BRCA1promoter in breast and ovarian cancers has been shown to reduce BRCA1 mRNA expression in several studies (7 , 8) .
ARHI 3 is a novel imprinted tumor suppressor gene that encodes a small GTPase with 60% homology to Ras and Rap (9) . In contrast to Ras and Rap, re-expression of ARHI suppresses clonogenic growth of breast cancer cells, reduces their invasiveness, and induces apoptosis (10) . ARHI expression is down-regulated in 40% of DCIS and 70% of invasive breast cancers, judged by in situ hybridization and immunohistochemistry. When DCIS and invasive cancer were present in the same sample, much lower expression of ARHI was detected in invasive tumor when compared with DCIS (11) .
ARHI is one of some 40 genes across the entire human genome that are known to be imprinted. Imprinting results in the differential expression of the maternal and paternal alleles in zygotes and in normal adult cells. Genetic imprinting is thought to play an important role in mammalian development and is associated with some inherited diseases, as well as certain cancers. In normal adult cells, ARHI is expressed only from the paternal allele (9) , and the maternal copy is silenced. Mutations of ARHIhave not been found in the coding and promoter regions in all cancers tested to date, but LOH was found in 40% of breast and ovarian cancers (12) . As ARHI expression is lost in a greater fraction of breast cancers, we have sought other mechanisms by which ARHI could be down-regulated.
To investigate possible epigenetic dysregulation of the ARHI gene, we evaluated the methylation status of ARHIin normal and malignant mammary cells. Aberrant methylation of CpG islands was found in many breast cancer cell lines. Hypermethylation of certain CpG islands correlated with decreased expression of ARHI and demethylation of these same islands was associated with re-expression of ARHI.
MATERIALS AND METHODS
Cell Lines and Breast Cancers.
Mouse A9 cells and mouse A9 hybrids that contain a single human chromosome 1 of known parental origin (paternal: A9-1P, maternal: A9-1M) were developed at the Tottori University (Yonago, Tottori, Japan) and maintained in DMEM supplemented with 10% FBS and 3 μg/ml blasticidin S hydrochloride (Sigma, St. Louis, MO) for A9-1P and 800 μg/ml G418 (Invitrogen, Carlsbad, CA) for A9-1M cells. Six human primary mammary epithelial cells were cultured in 1:1 MCDB 105 and medium 199 with 10% FBS and 10 ng/ml epidermal growth factor (Sigma). Nine breast cancer cell lines (MDA-MB-231, MDA-MB-435, MDA-MB-453, MDA-MB-468, BT20, BT474, CAMA-1, MCF-7, and SKBr3) were purchased from the American Type Culture Collection (Manassas, VA) and maintained in RPMI 1640 with 10% FBS. Human embryo kidney cell line 293 was purchased from Microbix Biosystems, Inc. (Toronto, Ontario, Canada). Human breast cancer tissues were obtained using protocols approved by The University of Texas M. D. Anderson Institution Review Board. Twenty cases of primary invasive ductal carcinoma were fixed in formalin and embedded in paraffin.
Immunohistochemistry.
Sections (3 μm) were cut from formalin-fixed, paraffin-embedded tissues. Antigen retrieval was performed by heating 3 min under pressure in a pressure cooker in 10 mm sodium citrate (pH 6.0). Endogenous peroxidase activity was quenched by incubating sections in 3% hydrogen peroxide. The ARHI monoclonal antibody (IgG1) was applied overnight at 1.16 μg/ml in 1% BSA-PBS at 4°C overnight. The UltraVision Large Detection System AntiPolyvalent, horseradish peroxidase (Lab Vision Corporation), was used for detecting the primary antibody. A diaminobenzidine tetrachloride-supersensitive substrate kit (Biogenex, San Ramon, CA) was used to visualize the antibody-antigen complex. Counterstaining was performed with hematoxylin. Sections of normal breast epithelia served as a positive control. Two negative controls were carried out in all samples: ARHI antibody was replaced by PBS and leukocyte common antigen monoclonal antibody (IgG1).
DNA/RNA Extraction.
Genomic DNA was extracted from cell lines using QIAamp DNA Mini Kit (Qiagen). RNA was extracted using RNeasy Mini Kit (Qiagen) based on the manufacturer’s protocol. Paraffin slides were dewaxed by xylene, then stained by hematoxylin and microdissected. After microdissection, tumor tissues were digested with 1 × TK buffer [0.25% Tween 20, 2 mg/ml proteinase K, and 1 × 10 mm Tris (pH 8)-1 mm EDTA] and incubated at 56°C overnight. The lysates were heated at 100°C in a block for 10 min, then treated with bisulfite.
5-Aza-dC Treatment.
Breast cancer cell lines MDA-MB-231, MDA-MB-435, MDA-MB-453, MDA-MB-468, BT20, MCF-7, and SKBr3 were seeded at a density of 1 × 106 cells/100-mm dish in RPMI 1640 with 10% FBS and allowed to attach over a 24-h period. 5-Aza-dC (Sigma) was then added to a final concentration of 0.2–1 μm, and the cells were allowed to grow for 5 days. The medium with or without 5-aza-dC was changed each day. At the end of the treatment, the medium was removed, and the RNA and DNA were extracted for real-time quantitative RT-PCR and bisulfite analysis.
COBRA.
Bisulfite treatment of DNA and COBRA were performed as described on line (13) . 4 After treatment, 2-μl aliquots were amplified in 50-μl reaction mixtures containing 5 μl of 10× PCR buffer II, 14 μl of 25 mm MgCl2, 2.5 μl of 25 mm deoxynucleotide triphosphate, 1 μl of each primer (300 ng/μl), and 0.5 unit of AmpliTaq Gold DNA polymerase (Roche). Primers were designed based on the nucleotide sequence of the ARHIgene submitted to GenBank (AF202543). Primers used for COBRA were as follows: (a) CpG I-F/R, 5′-GTAAGGGAGAAAGAAGTTAGA-3′/5′-TACTATCCTAACAAAACCCTC-3′; (b) CpG II-F/R, 5′-GTTGGGTTAGTTTTTTATAGTTGGTT-3′/5′-AACCAAACAACCTAAAAAACAAATAC-3′; and (c) CpG III-F/R, 5′-GTTTTTAAGTTTTATAGGAAGATT-3′/5′-ATAATATACAAAAAAAACACACC-3′. After amplification, 20–80% of PCR products were digested with the restriction enzyme TaqI (New England Biolabs) for CpG I and III or BstUI (New England Biolabs) for CpG II, both of which recognize sequences unique to the methylated (bisulfite-unconverted) alleles but cannot recognize unmethylated (bisulfite-converted) alleles. DNA was then precipitated and electrophoresed in 6% polyacrylamide gels. The gels were stained with ethidium bromide, and the intensity of methylated alleles was calculated by densitometry using ImageQuant software (Molecular Dynamics). A methylation density cutoff point of 15% was considered significant.
Real-Time Quantitative-RT-PCR.
Total RNA from different cell lines was purified as above. cDNA was synthesized using 2 μg of total RNA. Oligo(dT)16 and SuperScript II reverse transcriptase (Invitrogen) were used for the reverse transcriptase reaction according to the manufacturer’s instructions. Real-time quantitative PCR was performed in a reaction mixture with two gene-specific primers (NY2P1, 5′-TCTCTCCGAGCAGCGCA-3′; and NY2P2, 5′-TGGCAGCAGGAGACCCC-3′), a labeled probe 5′-TGTCTTCTAGGCTGCTTGGTTCGTGCC-3′ (5′-fluorescent label, 6-carboxyfluorescein; 3′-label, 6-carboxytetramethylrhodamine), 2 μl of reverse transcriptase reaction mixture, and 12.5 μl of Master Mix on an ABI Prism 7700 Sequence Detection System (Perkin-Elmer), following the protocol of the manufacturer. The amount of ARHI expression, normalized to a human glyceraldehyde-3-phosphate dehydrogenase endogenous reference and relative to a calibrator, is given by: 2 −ΔΔCT (derivation of the formula is from user bulletin no. 2 of the manufacturer, page 11–13; Perkin-Elmer). 5
Promoter Constructs.
To investigate the function of the ARHI promoter, we prepared two constructs P2 and P9 using a TA Cloning Kit (Invitrogen). Construct P2 (−466/+13) includes CpG II alone, and P9 (−1622/+13) includes both CpG I and CpG II (Fig. 1A) ⇓ . P2 and P9 have been separately linked to a luciferase reporter (Promega). P2 and P9 plasmids have been transformed to Escherichia coli and purified by Qiagen Plasmid Maxi Kit (Qiagen).
A, ARHIgenomic structure and the position of promoter constructs P2 and P9. The ARHIgene contains two exons interrupted by a large intron. ▪ and □ represent the coding and noncoding regions, respectively. are CpG island regions. The locations of P2 and P9 are indicated. B, schematic diagram outlining construction of individually methylated CpG islands within promoter constructs linked to a luciferase promoter. Fragments containing CpG islands I or II were excised by appropriate restriction enzymes. The cut fragments and remaining constructs were treated with or without SssI methylase, respectively. The products were re-ligated as indicated in the diagram. Unmethylated cut fragments and remaining constructs were re-ligated as an experimental control of the process of the enzymatic cutting and re-ligation procedure. I: CpG island I; II: CpG island II; L: luciferase reporter; and M and ▪: methylated.
Transient Transfection and Luciferase Assays.
SKBr3 and 293 cells were seeded at a density of 3 × 105 /well in a 6-well dish and grown to 60–70% confluence in complete growth media containing RPMI 1640 (SKBr3) or DMEM (293) supplemented with 10% FBS. For each well, 2 μg of plasmid DNA and 0.01 μg of pRL-TK control vector (triplicate for each group) were cotransfected into cells with FuGene 6 (Roche) according to the manufacturer’s instructions. After incubation for 42 h, cells were harvested according to the manufacturer’s instructions (Promega). Luciferase activities were measured by using a Dual-Luciferase Report Assay System in an Analytical Luminometer Monolight 2010 (PharMingen) according to the manufacturer’s instructions. A pGL2-basic control vector without an insert was used as a negative control in the transfection experiments. Relative luciferase activities were calculated as the ratio of firefly/renilla luciferase.
In Vitro Methylation.
In vitro methylation was carried out according to the methods described by Song (14) . Briefly, ARHIpromoter construct P2 or P9 was incubated overnight with three units of SssI (CpG) methylase (New England Biolabs) in the presence (methylated) or absence (mock-methylated) of 1 mm S-adenosylmethionine, as recommended by the manufacturer. After purification with a QIAquick Gel Extraction Kit (Qiagen) and quantitation of DNA at A260 nm, equal amounts (2 μg) of methylated and mock-methylated reporter constructs were transiently transfected into SKBr3 or human embryo kidney 293 cell lines, and luciferase activities were measured as described above. Individual reactions were monitored by digestion with BstUI restriction enzyme.
Selective in Vitro Methylation of CpG Island I or II.
To examine the selective effect of methylation on individual CpG islands I or II, ARHI promoter construct P9 was cut with appropriate restriction enzymes to excise CpG I (MluI + NdeI) and CpG II (NdeI + SphI) from the original construct. The experiment is shown schematically in Fig. 1B ⇓ . After cutting, the fragment of CpG island I or II and the remaining portion of the construct containing the other CpG island and the luciferase reporter vector were purified from agarose gel by QIAquick Gel Extraction Kit (Qiagen) and DNA quantitated at A260 nm. One-third of the fragments or the remaining portion of the construct was treated with SssI methylase, and two-thirds were untreated (unmethylated). Methylation-specific restriction enzyme BstUI was used to monitor the methylation efficiency to make sure that the fragments were completely methylated. The methylated or unmethylated CpG island fragments and the corresponding remaining portions of the constructs were re-ligated after methylation. The ligation reaction was performed at 16°C overnight using the same amount of DNA at an insert/vector molar ratio of 2:1. The ligated products were purified by gel purification (Qiagen), and the ligation efficiency was monitored by gel electrophoresis.
The luciferase activities of the constructs (Fig 1B) ⇓ containing selectively methylated CpG island fragments were measured. The constructs containing a methylated luciferase reporter and re-ligated constructs without methylation treatment were used as controls. The luciferase activities of various amounts of original P9 construct (0.325–1.95 μg), corresponding to the 5–30% of the original luciferase activity of the P9 construct, were also measured to permit quantitation.
RESULTS
Methylation of CpG Islands and Silencing of the Maternal Allele of ARHI in Normal Cells.
Oshimura et al. (15, 16, 17) have prepared mouse A9 hybrids containing single alleles of human chromosome 1 of either maternal (A9-1M) or paternal (A9-1P) origin, introduced via microcell-mediated chromosome transfer. Using RT-PCR and quantitative real-time PCR, we confirmed that only A9-1P cells that bear the paternal allele express ARHI expression (103–104 copies/100 ng of cDNA), whereas no ARHI expression was detectable in A9-1M cells that bear the maternal allele. When the parental cell lines were tested, ARHI was expressed in the diploid human MK fibroblasts (103–104 copies/100 ng cDNA) but not detectable in the murine A9 cells, consistent with the fact that an ARHI homologue has not been observed, to date, in the mouse. COBRA was used to analyze the methylation status of CpG islands associated with DNA of the ARHIgene in these cells. DNA from the hybrid and parental cell lines was treated with bisulfite, amplified, and digested with restriction enzymes that can distinguish methylated and unmethylated fragments. We found that all three CpG islands were methylated in A9-1M-1 (A9-1M-1 and A9-1M-2) cells but not methylated in A9-1P (A9-1P-1 and A9-1P-2) cells (Fig. 2) ⇓ . Consistent with single allele methylation, all three CpG islands in MK cells were partially (single allele) methylated. The results additionally confirmed that ARHIis a maternally imprinted gene in which the maternal allele methylated and silenced.
Methylation status of A9-chromosome 1 hybrids. A9 hybrids containing a single allele of human chromosome 1, either maternal (A9-1M) or paternal (A9-1P), were treated with bisulfite and digested with restriction enzymes. All maternal A9 hybrids cells (1M1, 1M2) contained methylated CpG islands and all paternal A9 hybrids (1P1, 1P2) contained unmethylated CpG islands. Unmethylated (gray arrows) and methylated (dark arrows) bands were quantitated by densitometry. Methylation density (percentage of restricted versus unrestricted fragment) is shown below each lane.
Consistent with imprinting of ARHIin all normal human cells, CpG islands I, II, and III were all partially methylated (one allele) in each of six normal mammary epithelial cell cultures (Fig. 3) ⇓ .
Methylation status of normal breast epithelial cells. Methylation of ARHIDNA from six cultures of normal breast epithelial cells was characterized as described in Fig. 2 ⇓ . All normal epithelial cells showed methylation of only a single allele of ARHI. C: control, using genomic DNA without bisulfite treatment.
Aberrant Methylation of ARHICpG Islands in Breast Cancer Cell Lines and Tumors.
In contrast to the consistent partial methylation of ARHIin normal breast epithelial cells, aberrant methylation of ARHIwas found in many breast cancer cell lines that had decreased ARHI expression. Hypermethylation was found in 6 of 9 breast cancer cell lines (67%) at CpG island I, in 3 of 9 (33%) at CpG island II, and in 5 of 9 (56%) at CpG island III (Fig. 4) ⇓ . Moreover, hypomethylation was observed in 44% of breast cancer cell lines (4 of 9) at CpG island II (Fig. 4) ⇓ .
Aberrant methylation of ARHICpG islands in breast cancer cell lines. Methylation of ARHIDNA was studied in nine breast cancer cell lines and characterized as described in Fig. 2 ⇓ .
Although all of the cancer cell lines lacked ARHI expression, differences were observed among different cells lines in methylation of CpG islands. Three types of methylation have been observed. In type 1, CpG island II is hypermethylated (methylation in both alleles). In type 2, CpG island II is hypomethylated (neither is methylated), but CpG islands I and III are hypermethylated. In type 3, all CpG islands are partially methylated (methylation in only one allele; Table 1 ⇓ ). Hypomethylation of CpG island II in type II cells suggested that the maternal-specific methylation was lost in these tumor cells, a phenomenon existing in some cancers termed the LOI.
Methylation status and expression of ARHI after treatment with 5-aza-dC
We have analyzed the methylation status of 20 cases of breast cancer by microdissection of tumor cells from paraffin slides. No case had hypermethylation in CpG island I; 3 of 20 cases had hypermethylation in CpG island II (15%); and 4 cases had hypermethylation in CpG island III (20%). Hypomethylation was found in 3 cases at CpG island I (15%), 1 case at CpG island II (5%), and 2 cases at CpG island III (10%). Representative cases with different methylation patterns are shown in Fig. 5 ⇓ . By histochemical staining, we found ARHI expression was down-regulated in 70% of these 20 cases, including the 3 cases with hypermethylation of CpG island II.
Aberrant methylation of ARHICpG islands in human breast cancers. Methylation of ARHIDNA was studied in breast cancer tissues and characterized as described in Fig. 2 ⇓ . T5, T7, and T14 were exhibited hypermethylation of CpG island II. T13 had hypomethylation of CpG island II but hypermethylation of CpG island III. T1 had partial methylation of all three CpG islands. T16 had hypomethylation of CpG island III.
ARHI Expression after Treatment with Demethylating Agents.
To explore the regulatory function of methylation in breast cancer cells, a DNA demethylating agent, 5-aza-dC, was used to treat breast cancer cell lines that exhibited different abnormalities of ARHICpG methylation. For those types 1 and 2 breast cancer cell lines that exhibited hypermethylation (both alleles) of CpG islands I, II, or III, treatment with the demethylating agent 5-aza-dC restored a normal state of partial methylation (one allele; Fig. 6A ⇓ ). Successful demethylation could partially restore ARHI expression in type 1 (shown as MDA-MB-435) and type 2 cells (shown as BT20) that exhibited hypermethylation of either CpG island I, II, or III but not in type 3 cells (shown as MCF-7), which did not exhibit hypermethylation at either CpG island (Fig. 6B ⇓ , Table 1 ⇓ ). Thus, ARHI expression could be partially reactivated only in cell lines in which the promoter region CpG islands I and II were hypermethylated but not in cancer cells with partial or hypomethylation of these CpG islands. These data suggest that hypermethylation of the promoter region of ARHIis one of the important mechanisms by which ARHIgene expression can be down-regulated, although this is not the only mechanism of ARHI regulation.
Effect of 5-aza-dC on methylation and expression of ARHI in breast cancer cells. A, treatment with 5-aza-dC decreased hypermethylation (both alleles) to partial methylation (only a single allele) in MDA-MB-435 (CpG island II) and BT20 (CpG islands I and III). Methylation was not altered in other CpG islands. B, ARHI expression as defined in the “Materials and Methods” was restored only in cell lines with hypermethylation of either CpG island II or I (MDA-MB-435 and BT20 cells) but not in cell lines with partial methylation of CpG islands (MCF-7 cells). ARHI expression was compared in treated and untreated sample.
ARHIPromoter Activity after DNA Methylation.
To define more critically the role of CpG island methylation in silencing ARHI expression, the ARHIpromoter segment P2, which includes CpG island II, and the ARHIpromoter segment P9, which includes both CpG islands I and II, have been linked to a luciferase reporter. By using CpG methyltransferase, the promoter segments were methylated and transfected into SKBr3 cells. Promoter activity was completely eliminated by methylation (Fig. 7) ⇓ , consistent with the possibility that methylation of the ARHIpromoter plays an important role in regulating its activity, provided that methylation does not affect the inherent activity of the luciferase construct. Experiments shown below suggest that methylation of the luciferase reporter might reduce but not eliminate the luciferase activity of the entire construct.
Methylation of ARHIpromoter (P2 and P9) completely eliminated transcriptional activity. ARHIpromoter fragments P2 and P9 were linked to a luciferase reporter and transfected into SKBr3 cells using a dual-luciferase report assay system. A pGL2-basic control vector without insert was used as a negative control. Relative luciferase activities were calculated as the ratio of firefly/renilla luciferase. ARHIpromoter construct P2 or P9 was incubated with SssI (CpG) methylase in the presence (methylated) or absence (mock-methylated) of 1 mm S-adenosylmethionine.
ARHI Activity after Methylation of CpG Islands I and II.
To understand the role of two different CpG islands in the regulation of ARHI expression, we have selectively methylated each island to determine the impact on ARHIpromoter activity. The fragment containing CpG island I or II was excised by appropriate restriction enzymes and treated with SssI methylase. The products were reinserted into a nonmethylated promoter reporter construct. Conversely, as a control, nonmethylated fragments were re-ligated with a methylated promoter reporter construct (which included the other methylated CpG island and the luciferase reporter) to compare the effect of methylation on the remaining constructs. Nonmethylated cut fragments and remaining constructs were re-ligated back as an experimental control for the process of the enzymatic cutting and re-ligation procedure.
To determine the reporter activity lost during experimental manipulation of the constructs, we compared the luciferase activities of re-ligated constructs to that of the original P9 construct, using comparable amounts of DNA. Enzymatic cutting and re-ligation could restore ∼10–15% of the original luciferase activity (data not shown), which compares favorably to the efficiency of DNA ligation. Accordingly, relatively larger amounts of DNA were used to generate enough luciferase activity.
As shown in Fig. 8B ⇓ , methylation of both CpG island I and luciferase reduced transcription by only 20%, whereas methylation of CpG island II alone reduced transcription to a background level. In Fig. 8A ⇓ , methylation of CpG island I did reduce transcription significantly but did not to background level, whereas methylation of CpG island II completely abolished transcription. These data suggest that although CpG island I methylation could contribute to reduced expression but that methylation of CpG island II is more critical for complete silencing. Methylation of the luciferase reporter could not demolish the activity of the constructs because at least some reporter activity was retained in the luciferase constructs that had been methylated and re-ligated (Fig. 8) ⇓ .
Selective methylation of ARHICpG islands down-regulates transcriptional activity of the ARHI promoter. Methylation of CpG island I partially reduced and methylation of CpG island II completely eliminated transcriptional activity. I: CpG island I; II: CpG island II; L: luciferase reporter; and M: methylated. A, selective methylation of CpG island I assay; B, selective methylation of CpG island II assay. The related constructs and procedures were described in “Materials and Methods.”
DISCUSSION
Epigenetic modification has been identified as a crucial event in carcinogenesis. Aberrant cytosine methylation of promoter regions represents one possible mechanism for gene silencing in cancer. Cytosine methylation has been found inversely correlated with transcription of many genes. The high content of CG nucleotides, known as CpG islands, which were found mainly in the promoter regions of housekeeping genes and some tissue-specific genes, are protected from methylation in normal mammalian cells. The methylation pattern, however, is altered in cancer cells. Widespread hypomethylation accompanied by region-specific CpG island hypermethylation is frequently observed in many tumor suppressor genes and can result in transcriptional down-regulation or silencing. Recent studies have identified a link between two important epigenetic modifications, DNA methylation and histone acetylation, in regulation of gene expression and silencing. Although the precise mechanism of transcriptional repression by DNA methylation is still not clear, it is known that DNA methylation can induce repressive chromatin remodeling by causing massive histone deacetylation at the methylated sites.
In this study, we have confirmed and extended earlier observations that ARHI, as an imprinted gene, is expressed only from the paternal allele on chromosome 1 and that the maternal allele is silenced and methylated (9) . Complete methylation of all three CpG islands of ARHIwas observed in the A9-1M cells that bear maternal human chromosome 1, whereas no methylation of CpG islands was observed in the ARHIgene from A9-1P cells that bear human paternal chromosome 1. Consistent with this observation, partial methylation (methylation of one allele) was observed for each of the CpG islands from normal breast epithelial cells. Although methylation plays an important role in ARHIimprinting, it may not be the only mechanism. Other mechanisms such as histone deacetylation and methylation have been suggested to gene imprinting.
As a putative tumor suppressor gene, the function of both alleles of ARHI must be lost to facilitate tumor progression. Through imprinting, expression of the maternal allele of ARHI has been lost in all normal cells. Thus, as an imprinted gene, ARHIhas already sustained one of two hits required for carcinogenesis in Knudson’s model of tumor suppressor gene inactivation (18) . Any genetic or epigenetic alteration in the remaining paternal allele of ARHIcould completely eliminate function of the ARHIgene. Silencing of ARHI is associated with LOH in 40% of breast and ovarian cancers. We have examined 9 informative cases in which paired DNA from tumor and normal tissue could be analyzed for both LOH and methylation (9 , 12) . The retained allele was methylated in 7 of 9 patients with LOH (78%), suggesting that deletion of the functional allele of this imprinted gene provided a second hit for silencing ARHI.
Loss of ARHI expression may also occur through epigenetic mechanisms. In this study, we explored the potential role of aberrant methylation in silencing of ARHI. Three types of aberrant methylation have been observed in breast cancer. In type 1, CpG island II is hypermethylated with variable methylation status in other CpG islands. In type 2, CpG island II is hypomethylated, and CpG islands I and III are hypermethylated. In type 3, no hypermethylation is found, and all three CpG islands are partially methylated or hypomethylated. Treatment with the demethylating agent, 5-aza-dC, led to partial restoration of ARHI expression in type 1 and type 2 cells that exhibited hypermethylation of either CpG island I or II but not in type 3 cells, which did not exhibit hypermethylation at either CpG island. To establish a causal link between methylation of DNA and gene expression, we have established an in vitro methylation assay. Using this assay, we demonstrated that methylation of the promoter region of ARHIcompletely eliminated its transcriptional activity, suggesting that CpG island hypermethylation in the promoter region could account for silencing of ARHI. This is, of course, a model system that addresses only the issue of hypermethylation and may or may not take into account the association of methylated DNA with chromatin. Histone deacetylation has, however, been implicated in similar studies of methylation associated gene silencing in Xenopus (19) .
There are two CpG islands located in the promoter region of ARHI. Methylation of CpG island I partially blocked ARHI expression, whereas methylation of CpG island II completely eliminated the transcriptional activity of the ARHI promoter. These data suggest that CpG island II plays the more critical role in regulating ARHI expression. CpG island II is located immediately upstream of exon I of the ARHIgene near the TATA box. CpG island II spans the region covering several active transcription factors elements, including multiple SP1 sites (20) . The SP1 transcription factor may compete with DNA methylases in activation of specific genes (21) . The mechanism by which CpG methylation blocks ARHI transcription remains to be elucidated. Methylation of CpG islands in the promoter might prevent transcriptional initiation, inhibit the binding of transcription factors, or alter chromatin structure.
Interestingly, hypermethylation of CpG island III in the coding region of exon 2 has been observed frequently in breast cancer cell lines (56%). The impact of the downstream CpG island III methylation on ARHI expression is unclear. Most earlier studies of other genes suggest that only methylation of CpG islands in the promoter region plays a role in repressing gene expression (22) . Recent studies, however, suggest that methylation of exonic CpG islands in coding regions might also affect gene expression (23 , 24) . Nguyen et al. (25) have suggested that exonic CpG islands are more susceptible to de novo methylation than promoter islands and may serve as an early event in aberrant methylation. Subsequent spreading of aberrant methylation from exons into the promoter region with consequent inactivation of promoter segments might contribute to the genesis of specific cancers.
Among the patterns of aberrant methylation, type 2 methylation is of particular interest because these breast cancer cell lines lose their maternal-specific methylation of CpG island II and appear to exhibit LOI. LOI was first discovered in Wilms’ tumor; Steenman et al. (26) reported early in 1994 that LOI of insulin-like growth factor-2 is linked to reduced expression and abnormal methylation of H19 in Wilms’ tumor. Frequent LOI of PEG1/MEST had been found in invasive breast cancer (27) . Scelfo et al. (28) showed that loss of the maternal KvDMR1 methylation was common, ranging from 30 to 50%, to a variety of adult neoplasms, including liver, breast, cervical, and gastric carcinomas. They found that 11p15.5 loci were concomitantly hypomethylated, indicating that loss of KvDMR1 methylation occurred in the context of a common mechanism affecting the methylation of a large 11p15 subchromosomal domain. Therefore, it seems possible that, contrary to the repression of promoter activity caused by hypermethylation, loss of gene expression at 11p15.5 may result from the activation, by hypomethylation, of one or more negative regulatory elements (28) . The mechanism and effect of LOI in ARHI CpG island II are still unknown.
In this study, we demonstrated that the transcriptional silencing of ARHI is correlated with the hypermethylation of CpG islands in the ARHI promoter. Consistent with imprinting, CpG islands in the maternal allele of normal cells are consistently methylated and silenced, whereas paternal alleles are not. In cancers with decreased ARHI expression, aberrant patterns of CpG methylation were observed in some but not all cell lines. Hypermethylation of both alleles of CpG island II completely eliminated ARHIpromoter activity. Thus, imprinting and methylation of CpG islands in ARHIprovides a new mechanism for the down-regulation of this growth regulatory gene in breast cancers.
Footnotes
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.
↵1 This work was supported by NIH Grants CA 64602 (to R. C. B.) and CA 80957 (to Y. Y.) and a Susan G. Komen Breast Cancer Foundation Grant (to Y. Y.).
↵2 To whom requests for reprints should be addressed, at The University of Texas, M. D. Anderson Cancer Center, 1515 Holcombe Boulevard, Box 354, Houston, TX 77030. Phone: (713) 792-3790; Fax: (713) 745-2107; E-mail: yyu{at}mdanderson.org
↵3 The abbreviations used are: ARHI, ras homologue member I; FBS, fetal bovine serum; 5-aza-dC, 5-aza-2′deoxycytidine; LOH, loss of heterozygosity; LOI, loss of imprinting; COBRA, combined bisulfite restriction analysis; RT-PCR, reverse transcription-PCR.
↵4 Internet address: http://www3.mdanderson.org/leukemia/methylation.
↵5 Internet address: http://docs.appliedbiosystems.com/pebiodocs/04303859.pdf.
- Received November 7, 2002.
- Accepted May 9, 2003.
- ©2003 American Association for Cancer Research.
References
Volume 63, Issue 14
