Aberrant Promoter Methylation of the ABCG2 Gene in Renal Carcinoma

Abstract

ABCG2 is a ubiquitous ATP-binding cassette transmembrane protein that is important in clinical drug resistance. Little is known about the mechanism(s) regulating the expression of ABCG2. We hypothesized that DNA methylation could play a role in the epigenetic regulation of ABCG2 gene expression. The promoter methylation status of three renal carcinoma cell lines was assessed with restriction enzyme digestion-coupled PCR and bisulfite genomic sequencing. Both UOK121 and UOK143, with known methylation of the VHL promoter, showed induction of ABCG2 expression after 5-aza-2′-deoxycytidine (5-aza-dC) treatment, suggesting that aberrant methylation of the ABCG2 gene was associated with gene silencing. In vitro methylation of the ABCG2 promoter-driven luciferase reporter vector resulted in a significant inhibition of transcription. Our data suggested that the ABCG2 gene is regulated coordinately at both histone and DNA levels. A chromatin immunoprecipitation assay demonstrated that the methylated promoter in UOK121 and UOK143, but not the unmethylated one in UOK181, is associated with the methyl CpG binding domain proteins (MBDs), MBD2 and MeCP2. Histone deacetylase 1 and a corepressor, mSin3A, were identified binding to the promoter region containing the CpG island, thereby suppressing ABCG2 transcription. Reactivation of ABCG2 was achieved by treatment with 5-aza-dC, a demethylating agent, concomitant with the release of MBDs from the promoter. Furthermore, the association of methylated lysine 9 on histone H3, a hallmark of promoter methylation, with the promoter was reduced following 5-aza-dC treatment. These data suggest that DNA methylation-dependent formation of a repressor complex in the CpG island contributes to inactivation of ABCG2.

ABCG2 (previously named BCRP, MXR, or ABCP) is an ATP-binding cassette half-transporter, originally identified as a multidrug resistance transporter (2, 15, 39). It is highly expressed in many normal tissues, including the epithelium of the small intestine and the liver canalicular membrane (36). In addition to conferring resistance in cancer cells to chemotherapeutic agents such as mitoxantrone, topotecan, and methotrexate (14, 56), ABCG2 has been shown to mediate apically directed drug transport and to play a significant role in absorption, distribution, and elimination of its substrates (8, 55).

To date, little is known about the molecular mechanisms controlling ABCG2 expression. Characterization of the ABCG2 gene promoter revealed that it is a TATA-less promoter with several Sp1, AP1, and AP2 sites and a CCAAT box downstream from a putative CpG island. Both positive and negative cis-regulatory elements have been suggested in the ABCG2 promoter (4). Although the role of these cis-elements in ABCG2 transcription has not been assessed, recent studies have demonstrated functional hormone (16) and hypoxia (32) response elements in the ABCG2 promoter. Early observations in our laboratory revealed that ABCG2 expression in some renal clear cell carcinoma (RCC) cell lines could be upregulated by treatment with a DNA-demethylating agent, 5-aza-2′-deoxycytidine (5-aza-dC). This observation led to the hypothesis that DNA methylation might play a role in the epigenetic regulation of the ABCG2 gene.

DNA methylation at the 5-position carbon of cytosine within 5′-CpG-3′ dinucleotides is the predominant epigenetic modification in the regulation of gene expression. While DNA methylation is essential for mammalian development (34), distortion of the genomewide DNA methylation profile in cancer cells results in aberrant methylation patterns and silencing of tumor suppressor genes (5, 17, 26). Methylation is initiated by the de novo DNA methyltransferases DNMT3a and DNMT3b, which transfer methyl groups from S-adenosylmethionine to C-5 of CpG base pairs. During DNA replication, the maintenance methyltransferase DNMT1 then methylates the newly synthesized strand complementary to the hemimethylated DNA.

Hypermethylation of CpG islands in gene promoters leads to suppression of transcriptional activity (6, 28). It has been reported that methylation levels of gene promoters are inversely proportional to transcriptional activity (30). The epigenetic silencing of genes is mediated through recruitment of a group of proteins, called methyl CpG binding proteins (MBDs), that act as docking sites for corepressor proteins such as Sin3a, histone deacetylases (HDACs), histone methyltransferases, and heterochromatin protein 1a (HP1a) (21, 22, 57). As a consequence, the chromatin becomes more condensed and not favorable for transcription (22, 27). Mammalian cells contain five MBDs with highly homologous methyl CpG-binding domains, MBD1, MBD2, MeCP2, MBD3, and MBD4. The mechanism of MBDs targeting to different regions of the genome is not known. Recent studies from lower eukaryotes and mammals have shown that DNA methylation and histone modifications, specifically hypoacetylation and H3K9 methylation, are interrelated (51, 52).

The present study was undertaken to explore the molecular mechanisms underlying the epigenetic regulation of ABCG2 gene expression. Since the expression of the VHL gene is known to be regulated by methylation of its 5′ regulatory regions (23) and the gene methylation status was well characterized in the renal carcinoma cell lines (23, 58), analysis of VHL gene expression was performed together with ABCG2 gene expression to serve as a validation of the assays. Our results indicate that aberrant promoter methylation represses ABCG2 expression in RCC cells, associated with coordinated changes in histone methylation and deacetylation. Inhibition of methyltransferase can reactivate the expression. MBDs and the HDACs are recruited to the methylated promoter and appear to contribute to the repression of ABCG2.

MATERIALS AND METHODS

Tissue culture.

Human sporadic clear cell renal carcinoma cell lines UOK121, UOK143, and UOK181 were kindly provided by M. W. Linehan (NCI, Bethesda, MD). The methylation status of UOK121 and UOK143 has been described previously (1, 23). UOK121 and UOK143 had lost one VHL allele and retained a single heavily methylated allele. In contrast, the promoter/exon 1 region of the VHL gene is not methylated in UOK181. The cell lines were maintained in Iscove's modified Eagle's medium containing 10% fetal bovine serum and incubated at 37°C in 5% CO₂.

5-Aza-2′-deoxycytidine (5-Aza-dC) and depsipeptide treatment.

Cell lines were treated for 4 days with 5-aza-2′-deoxycytidine (5-aza-dC) (Sigma, St. Louis, MO) at a concentration of 1 μM. Culture media (with 5-aza-dC) were replaced every 24 h. In parallel experiments, cells were treated with depsipeptide at 2 ng/ml plus verapamil at 5 μg/ml for 24 h. Stock solutions of 5-aza-dC, depsipeptide, and verapamil were dissolved, respectively, in 50% acetic acid, dimethyl sulfoxide, and water.

Semiquantitative RT-PCR.

Total RNA was isolated using the Trizol reagent (Invitrogen, Carlsbad, CA). RNA (1 μg) was reverse transcribed using PowerScript reverse transcriptase (Clontech, Mountain View, CA). Amplification of cDNA was done using primers specific for ABCG2, VHL, and β-actin. Primers specific for ABCG2 cDNA amplification were 5′-CAATGGGATCATGAAACCTG-3′ (forward) and 5′-GAGGCTGATGAATGGAGAA-3′ (reverse). The primer pair specific for VHL cDNA was 5′-TCATCTTCTGCAATCGCAGTCC-3′ (forward) and 5′-AATCTCCCATCCGTTGATGTGC-3′ (reverse). The primers specific for β-actin amplification were 5′-GCTCGTCGTCGACAACGGCTC-3′ (forward) and 5′-CAAACATGATCTGGGTCATCTTCTC-3′ (reverse). Amplification of β-actin cDNA served as an internal control. PCR amplification for the ABCG2, VHL, and β-actin mRNA was performed at an annealing temperature of 55°C to yield 584-, 420-, and 330-bp products, respectively. The PCR products were resolved on 2% agarose gel, stained with ethidium bromide, and quantitated.

Restriction enzyme digestion coupled with PCR to assess promoter methylation.

A methylation-sensitive PCR assay was used to analyze the methylation status of the promoter region of ABCG2 and VHL (23). DNA was treated with methylation-sensitive restriction endonuclease followed by PCR amplification. Total genomic DNA (5 μg) was isolated using the genomic DNA isolation kit (Promega, Madison, WI) and digested with 100 U of either HpaII or MspI (New England Biolabs, Ipswich, MA) at 37°C overnight. HpaII is a methylation-sensitive restriction endonuclease that cleaves DNA at CCGG sequences when the internal cytosine residue is unmethylated on both strands. MspI cleaves DNA regardless of methylation status and serves as a control for digestion efficiency. Since the VHL gene is known to be unmethylated in placental DNA (Sigma), it was used as a control to confirm that the HpaII/MspI digestions were complete. Digested DNA was then amplified by PCR using primers designed for the evaluation of DNA methylation status of the ABCG2 and VHL genes (Table 1). PCRs contained 1.5 mM MgCl₂, 200 μM deoxynucleoside triphosphates (dNTPs), 0.4 μM of each primer, 100 ng of digested DNA, and 2.5 U of Taq polymerase (Bioline, Randolph, MA). The PCR protocol was 1 cycle of 94°C for 3 min; 30 cycles of 94°C for 1 min, 55°C for 1 min, and 72°C for 1 min; and finally 1 cycle of 10 min at 72°C.

TABLE 1.

Primers used for the restriction enzyme-coupled PCR assay and ChIP assay

Primer set	Oligonucleotide sequence	Amplicon size (bp)
Restriction enzyme-coupled PCR
VHL
VHLPF1	5′-CAGTAACGAGTTGGCCTAGC-3′	440
VHLE1R	5′-GTCGAAGTTGAGCCATACGG-3′
ABCG2
Distal sense	5′-CTCCTCCTGTAGTGCCTTCAGATCTTGCT-3′	259 (nt −1527 to −1268)
Distal antisense	5′-TTGCAAATGACCCGAGATCCCACCA-3′
Proximal sense	5′-GATGCAGCAGGTAGATGTTGGGA-3′	707 (nt −687 to +20)
Proximal antisense	5′-GCCAGAGCTGAACGCAGTGG-3′
ChIP assays
ABCG2
Distal sense	5′-CTCCTCCTGTAGTGCCTTCAGATCTTGCT-3′	259 (nt −1527 to −1268)
Distal antisense	5′-TTGCAAATGACCCGAGATCCCACCA-3′
Proximal
P1 sense	5′-GATGCAGCAGGTAGATGTTGGGA-3′	108 (nt −687 to −579)
P1 antisense	5′-TGTGCAATATTCCGATGGTGTGGA-3′
P2 sense	5′-CCATTCACCAGAAACCACCCATTT-3′	157 (nt −428 to −271)
P2 antisense	5′-GCTCATTGGGCTGATCAGTACCT-3′
P3 sense	5′-AGGTACTGATCAGCC CAATGAGC-3'	154 (nt −293 to −139)
P3 antisense	5′-TGAGCCGCCAGCAGGACT-3′
P4 sense	5′-AGTCCTGCTGGCGGCTCA-3′	166 (nt −146 to +20)
P4 antisense	5′-GCCAGAGCTGAACGCAGTGG-3′
GAPDH
GAPDH sense	5′-TACTAGCGGTTTTACGGGCG-3′	166
GAPDH antisense	5′-TCGAACAGGAGGAGCAGAGAGCGA-3′

Bisulfite genomic sequencing.

Genomic DNAs were isolated from cell lines with the aforementioned genomic DNA purification kit (Promega), and they were treated with a sodium bislfite modification kit (Chemicon, Temecula, CA) according to the manufacturer's recommendation. The CpG island region of the ABCG2 gene was then amplified with three pairs of gene-specific primers (designed by a computer program called MethPrimer [35]) (Table 2) in a mixture containing 100 ng bisulfite-modified DNA, 1× PCR buffer (20 mM Tris-HCl, pH 8.4, 50 mM KCl), 200 μM each dNTP, 1.5 mM MgCl₂, and 0.5 μM each primer in a final volume of 50 μl. Reactions were hot started at 94°C for 3 min before the addition of 2.5 U of Taq polymerase (Bioline). In order to avoid the preferential amplification of either methylated or unmethylated sequences, no CpG dinucleotide was included in the bisulfite primer sequences. PCR amplification products were purified from a 2% agarose gel with the QIAquick gel extraction kit (QIAGEN, Valencia, CA). The gel-purified PCR fragments were TA cloned into the pCR2.1-TOPO vector (Invitrogen) and transformed with the JM109-competent cells (Promega). Individual clones were sequenced with M13 forward primer. Sequencing was performed by the DNA Sequencing Minicore Facility at the Center for Cancer Research, National Cancer Institute, NIH.

TABLE 2.

PCR primers for bisulfite genomic sequencing

Primer set	Oligonucleotide sequencea	Size (bp)	Anneal temp (oC)	Genomic positionb
A
Sense	5′ TTTTTTTAGAATTTTATTTATTAGAAATTA 3′	248	52	−440
Antisense	5′ CAAAAACAAACCAAACACTCC 3′			−193
B
Sense	5′ GGAGTGTTTGGTTTGTTTTTG 3′	223	52	−217
Antisense	5′ CAATAACCCCTCCCCAA 3′			+6
C
Sense	5′ TTGGGGAGGGGTTATTG 3′	225	52	−11
Antisense	5′ ACACAACCCCTTCCCCCAA 3′			+214

^aCytosine residues modified by bisulfite treatment are underlined.

^bGenomic position is the location of the 5′ nucleotide of the primer in relation to the major transcriptional start site (GenBank accession no. AF151530) (4).

Construction of reporter gene plasmids and in vitro methylation.

Fragments of ABCG2 promoter were amplified by PCR and fused upstream of the firefly luciferase gene in the pGL3-Basic vector (Promega). A series of constructs were made with the 3′ end terminating at +396 relative to the designated transcriptional start site (GenBank accession no. AF151530) (4). The 5′ ends began at bases −1662, −1289, −628, −407, −312, −295, −243, −115, and −31. Another construct (−1662 to −629) was prepared, missing the CpG island as predicted by the CpG Island Searcher on the web (www.cpgislands.com) (54). Reporter constructs were confirmed by restriction mapping and DNA sequencing.

Whole ABCG2 reporter plasmids were methylated using SssI methylase (M.SssI; New England Biolabs), which methylates all cytosine residues within the double-stranded dinucleotide recognition sequence 5′-CG-3′. Ten micrograms of each plasmid was incubated with 30 U of M.SssI, and in a parallel control reaction, 10 μg of the same plasmid was mock methylated in the absence of S-adenosylmethionine (13). Methylated and mock-methylated plasmids were digested with SacI/HindIII and religated back to SacI/HindIII-restricted pGL3-Basic vector (Promega). The extent of methylation was determined by digestion with a mixture of methylation-sensitive restriction enzymes HpaII and HhaI.

Transient transfection and luciferase reporter assays.

The unmethylated (mock methylated) or methylated ABCG2 promoter/firefly luciferase fusion genes (200 ng DNA) were transfected in UOK121 cells using Fugene 6 (Roche, Indianapolis, IN). In each experiment, the phRG-Basic plasmid (50 ng), encoding Renilla luciferase (Promega), was cotransfected for normalization proposes. Luminescence was measured 48 h after transfection using the dual-luciferase reporter assay system (Promega). The pGL3-Basic (promoterless) plasmid was used in each experiment to determine the basal levels. Reporter activity was normalized by calculating the ratio of firefly to Renilla values. Each construct was tested in three independent transfections.

Cytotoxicity assay.

The cytotoxicity assays were performed using the sulforhodamine method described previously (47). Briefly, cells were plated in flat-bottom 96-well plates at a density of 2,000 cells per well and allowed to attach for 24 h at 37°C in 5% CO₂. The cells were then treated with one of three ABCG2-substrate chemotherapeutic drugs (mitoxantrone, topotecan, or SN-38) at a range of concentrations with or without 5 μM fumitremorgin C (FTC) and allowed to incubate at 37°C in 5% CO₂ for 96 h. Fumitremorgin C was synthesized by Thomas McCloud, Developmental Therapeutics Program, Natural Products Extraction Laboratory, National Institutes of Health (Bethesda, MD). Mitoxantrone and topotecan were purchased from Sigma. SN-38 was obtained from LKT Laboratory (St. Paul, MN). After incubation, the cells were fixed in 50% trichloroacetic acid and stained with sulforhodamine B solution (0.4% sulforhodamine B [wt/vol] in 1% acetic acid). Optical densities were read on a Bio-Rad plate reader at an absorbance of 570 nm. Each concentration was tested in quadruplicate, and controls were done in replicates of eight.

ChIP.

Cells growing on 10-cm culture dishes were incubated in culture medium containing 1% formaldehyde for 10 min at 37°C to cross-link DNA to chromatin-associated proteins. The cross-linking was quenched by adding 0.125 M glycine and incubated at room temperature for 5 min. The cells were then rinsed in ice-cold phosphate-buffered saline containing 5 mM sodium butyrate (Sigma) and 1× protease inhibitor cocktail (Calbiochem, La Jolla, CA), scraped, and collected by centrifugation at 4°C. Chromatin immunoprecipitation (ChIP) was carried out according to the manufacturer's recommendations with the following antibodies: anti-acetyl-histone H3 (Lys 9), anti-trimethyl-histone H3 (Lys9), anti-HDAC1, and anti-MBD2 (Upstate); anti-MeCP2 (Abcam); and anti-DNMT1, anti-DNMT3a, anti-DNMT3b, or anti-mSin3A (Santa Cruz Biotechnology). The DNA-protein complexes were sonicated using a Misonix sonicator MX2020 under conditions that gave a range in DNA fragments from 200 to 600 bp, as determined by agarose gel electrophoresis.

Amplification of the immunoprecipitated DNA was achieved using Taq DNA polymerase (Bioline) and 1 μl of either immunoprecipitated DNA, a normal immunoglobulin G (IgG) control or a 1:10 dilution of input chromatin. Experimental reactions were performed to determine optimal PCR conditions so that the yield of PCR products was dependent on the amount of input DNA (data not shown). The conditions for all reactions were as follows: 94°C for 3 min, followed by 30 cycles at 94°C for 1 min, 55°C for 1 min, 72°C for 1 min, and 72°C for 7 min. Primers were designed to evaluate the distal region (distal) and proximal regions (P1, P2, P3, and P4) spanning the putative CpG island of the ABCG2 promoter (Table 1). PCR products were electrophoresed on 2% agarose gels and stained with ethidium bromide. Enrichment (fold) in each immunoprecipitation was determined by quantifying the intensities of the PCR products in immunoprecipitated DNA versus input DNA (total chromatin).

RESULTS

Effect of 5-aza-dC on ABCG2 and VHL expression in RCC cell lines.

The expression level of ABCG2 in three renal clear cell carcinoma cell lines was determined by semiquantitative RT-PCR. Two of the cell lines, UOK121 and UOK143, carry a methylated VHL gene, while the third line, UOK181, does not. Interestingly, the ABCG2 expression level in UOK121 and UOK143 was about half of that in the UOK181 cell line (Fig. 1A and C). Treatment with 1 μM, 1.5 μM, or 2 μM 5-aza-dC significantly increased ABCG2 expression in the two hypermethylated VHL RCC cell lines, UOK121 and UOK143, but not in the unmethylated VHL line UOK181, as determined by RT-PCR (Fig. 1A) and immunoblot analysis (Fig. 1B). A minimum nontoxic dose of 1 μM 5-aza-dC was chosen for all subsequent studies because there was cell death at higher doses. VHL was also reexpressed by 5-aza-dC treatment in UOK121 and UOK143, which is in line with previous findings (Fig. 1A) (1). VHL expression was constitutively higher in UOK181 and did not respond to 5-aza-dC treatment.

FIG. 1.

Semiquantitative RT-PCR analysis (A, C, and D) and immunoblot analysis (B and C) of ABCG2 and VHL expression in UOK121, UOK143, and UOK181 cells. (A) Dose-dependent upregulation of ABCG2 after treatment with 5-aza-dC (1, 1.5, or 2 μM) for 4 days. The numbers denote the upregulation of ABCG2 and VHL relative to the untreated sample for each cell line after normalization to β-actin. (B) Cells were treated as in panel A and were harvested for immunoblot analysis of ABCG2 protein expression. (C) Cells were untreated or treated with either 5-aza-dC (1 μM) for 4 days or depsipeptide (2 ng/ml) plus verapamil (5 μg/ml) for 1 day. Representative results from three independent experiments are shown. The upper panel shows the RT-PCR analysis of ABCG2 and VHL. The numbers indicate the upregulation of ABCG2 and VHL relative to the untreated UOK121 cells after normalization to β-actin. The agarose gel image for ABCG2 was taken from a 33-cycle PCR to enable the visualization of the band corresponding to ABCG2 in the methylated cells. The actual quantitation was done with a lower-cycle-number PCR (25 cycles), previously found to be linear. The lower panel represents the immunoblot analysis of ABCG2 protein expression. β-Actin was used as a loading control. (D) Dose-dependent upregulation of ABCG2 after treatment with depsipeptide (1, 2, 5, or 10 ng/ml) for 24 h. The numbers denote the upregulation of ABCG2 relative to the untreated sample for each cell line after normalization to β-actin.

Since transcriptional regulation by DNA methylation has been suggested to involve histone deacetylation (20), the ABCG2 expression was also examined after treatment with depsipeptide, a specific histone deacetylase inhibitor. Upon depsipeptide treatment (2 ng/ml), the ABCG2 expression was only slightly upregulated in UOK121 and UOK143 but more remarkably in UOK181 at the mRNA and protein level (Fig. 1C). Although higher doses of depsipeptide could lead to a more remarkable upregulation of ABCG2 in UOK121 and UOK143 (Fig. 1D), these doses were associated with lower cell viability (reflected in cell growth rate and propidium iodide staining analysis; data not shown). A nontoxic dose of 2 ng/ml of depsipeptide (24 h) was thus chosen for all subsequent studies. Verapamil (5 μg/ml) was always added together with depsipeptide to prevent Pgp-mediated efflux of depsipeptide from the cells.

5-Aza-dC has been shown to be capable of transcriptionally activating genes with unmethylated promoters (50), inducing histone acetylation and H3 lysine 4 methylation (44), suggesting this drug can induce chromatin remodeling independently of its effects on cytosine methylation. Therefore, further experiments were set out to determine whether the upregulation of ABCG2 following 5-aza-dC treatment was mediated by demethylation.

Computational analysis of the CpG island in the ABCG2 gene.

A simplified schematic overview of the region showing the location of CpG sites in the ABCG2 promoter is provided in Fig. 2A. A CpG island is located at bases −599 to +329 according to Gardiner-Garden/Frommer criteria as listed in the CpG Island Searcher on the Internet (54). As the ABCG2 promoter contains a region meeting the typical CpG island criteria of >60% C+G content and an observed CpG frequency of >0.6, we decided to analyze its methylation status.

FIG. 2.

Methylation pattern of the ABCG2 gene. (A) Schematic representation of the putative CpG island of the ABCG2 gene. The distribution of the CpG dinucleotides as indicated by the vertical solid lines reveals the dense clustering of these sites spanning the 5′ flanking region and constituting a CpG island. HpaII/MspI recognition sites within the CpG island are shown by ovals. Double-arrowed lines indicate regions (distal and proximal) that were PCR amplified after restriction enzyme digestion. (B) Genomic DNA from UOK121, UOK143, and UOK181 digested by HpaII and MspI was amplified by PCR, respectively, using primers shown in panel A for the ABCG2 gene. UN, uncut DNA; P, placental DNA. (C) PCR results as in panel B using primers that flanked the promoter/exon 1 region of the VHL gene.

Methylation pattern of the ABCG2 gene promoter.

The proximal region of the ABCG2 promoter (−599 to +329) harbors 11 HpaII/MspI sites (Fig. 2A). To examine the methylation status of the ABCG2 promoter region in RCC lines, a pair of primers was used to amplify across these MspI/HpaII sites (Fig. 2A). Genomic DNA was treated with a methylation-sensitive restriction endonuclease followed by PCR amplification. Methylated cytosine residues at HpaII recognition sequences prevent enzyme cleavage and can be detected by PCR amplification product recovery. Conversely, restriction enzyme cleavage at unmethylated sites induces DNA strand breaks and abrogates PCR amplification. As a comparator, DNA was also treated with MspI endonuclease. MspI is an isoschizomer of HpaII that is not sensitive to methylation and cleaves DNA regardless of methylation status. The lack of amplified product in both HpaII- and MspI-digested DNA confirmed that the promoter is unmethylated in UOK181 cells (Fig. 2B). Comparable levels of PCR product in undigested and HpaII-digested DNA from UOK121 and UOK143 cells indicate that the ABCG2 promoter in these cells is methylated in the proximal promoter region found within the identified CpG island (Fig. 2B). Of note, amplification products were obtained for all three cell lines when using PCR primers designed to anneal to the distal region of the ABCG2 gene, where there is no MspI/HpaII recognition site (Fig. 2B). Thus, this result confirmed the correlation between methylation status and ABCG2 gene expression levels in the RCC lines, suggesting that changes in the methylation status could be a mechanism for regulating ABCG2 expression.

Again, since the methylation pattern of VHL is well characterized in the RCC cell lines, primers VHLPF1 and VHLE1R were synthesized to amplify the promoter/exon 1 region of the VHL gene to validate the method (Table 1) (58). In this region, there is a large CpG island of 10 CCGG sequences (33). No PCR product was obtained with HpaII or MspI digestion prior to PCR amplification in UOK181 (Fig. 2C), which agrees with the fact that the region is undermethylated in this cell line (58). Since the VHL gene would be expected to be unmethylated in placental DNA (58), it was employed as a control to confirm that HpaII/MspI digestion was complete. As expected for the known methylated VHL promoter/exon 1 in the UOK121 and UOK143 lines, PCR products were obtained from both HpaII-digested and control uncut DNA (Fig. 2C).

Methylation status of individual CpG dinucleotides in the ABCG2 promoter as determined by bisulfite genomic sequencing.

Bisulfite sequencing was employed to investigate the methylation profile of the ABCG2 CpG island in the three renal carcinoma cell lines. Genomic DNA was modified with sodium bisulfite, converting all unmethylated cytosines to uracils. Methylated cytosines were not converted to uracils. The core region of the ABCG2 CpG island (nucleotides [nt] −440 to +154) was then amplified with primer sets A, B, and C (Fig. 3 and Table 2). The specificity of the primers for bisulfite-converted templates was tested by their failure to PCR amplify non-bisulfite-treated human genomic DNA. PCR products were cloned into the pCR2.1-TOPO vector. The resulting plasmids were transformed using competent cells and plated to obtain colonies. Individual clones were then selected for sequencing. Each clone thus represents an individual methylation profile of the cell population. By examining the non-CpG cytosines, bisulfite conversion of unmethylated cytosines to uracils was virtually complete. In agreement with the methylation-sensitive restriction enzyme-coupled PCR above, the ABCG2 promoters of UOK121 and UOK143 were extensively methylated, whereas that of UOK181 was mostly unmethylated (Fig. 3). Interestingly, methyl groups were not distributed uniformly across the 66 CpG sites studied but instead were concentrated in the distal part of the CpG island studied (CpGs 1 to 52; nucleotides −388 to +114).

FIG. 3.

Methylation profile of individual CpG dinucleotides in ABCG2. (Upper panel) Genomic map of the CpG island in ABCG2. CpG sites and their genomic positions within the ABCG2 CpG island (−599 to +329) are represented by vertical lines at the top. Nucleotide positions are numbered relative to the major transcriptional start site (+1) defined in the published DNA sequence (GenBank accession no. AF151530) (4). (Lower panel) Methylation status of individual CpG dinucleotides. The numbering bar at the top of each cell line panel represents the 594-bp region from nucleotides −440 to +154 analyzed in this study. CpG sites on the numbering bar are designated from 1 to 66 according to their 5′-to-3′ order in the ABCG2 genomic sequence, and they are spaced out evenly for simplicity. PCR products A (−440 to −193), B (−213 to +6), and C (−11 to +154), amplified from bisulfite-treated DNA for sequencing, are shown. Eight clones were analyzed for each PCR fragment in each cell line. Each square denotes a CpG site across PCR fragments amplified from UOK121, UOK143, and UOK181. Filled squares, methylated; open squares, unmethylated.

Inhibition of transcriptional activity by methylation of the ABCG2 promoter.

A series of 5′ deletion reporter gene constructs were generated from the ABCG2 promoter region with the 3′ end terminating at +396 bp (Fig. 4A). Construct −312/+396 contains 287 of the 928 bases of the CpG island, the CCAAT box, and the predicted promoter. Construct −1662/−629 is missing the entire CpG island and the predicted promoter region. To study the possible cause and effect relationship between CpG methylation and transcriptional activity, we methylated a series of ABCG2 promoter fragments, cloned them into reporter constructs, and then compared the abilities of in vitro-methylated and mock-methylated ABCG2 promoters to direct the reporter gene expression in transiently transfected UOK121 cells. To confirm that in vitro methylation of the promoter constructs had been successful, we carried out double-restriction digestion of the promoter constructs. Complete methylation was demonstrated by the blockade of HpaII and HhaI digestion and migration of the vector as a single band as shown by ethidium bromide staining (Fig. 4B).

FIG. 4.

Methylation silences the activity of the ABCG2-promoter reporter constructs in a transient transfection assay. (A) Schematic representations of the 5′ deletion ABCG2-promoter constructs. The 5′ end of each of the constructs relative to the transcription start site (arrows) is indicated. Alignment with the putative CpG island is indicated above the deletion constructs. (B) Complete methylation of the ABCG2 promoter-pGL3 reporter constructs as revealed by restriction analysis. The mock-methylated (M.SssI −) and methylated (M.SssI +) constructs were doubly digested with a mixture of two methylation-sensitive restriction enzymes that fail to cut their recognition sequence when the cytosine is methylated (HpaII and HhaI). While the fully methylated constructs were undigested, the mock-methylated constructs were digested extensively, as shown by the production of several small fragments compared with the undigested constructs. (C) ABCG2 transcriptional activity in UOK121 cells transiently transfected with the mock-methylated or M.SssI-methylated ABCG2 promoter constructs depicted. The mean reporter activity ± standard deviation (firefly/Renilla luciferase units [RLU]) from three independent experiments is presented. pGL3-Basic represents the vector backbone without the ABCG2 promoter insert. The −1662 to −629 ABCG2 promoter construct does not contain the putative CpG island.

Using a transient transfection assay system in UOK121 cells, all of the ABCG2 promoter-luciferase constructs, except −1662/−629, demonstrated good activities above the pGL3-Basic background (Fig. 4C). In agreement with the findings by Bailey-Dell et al. (4), the ABCG2 proximal promoter (−312 to the transcriptional start site) was demonstrated to have necessary elements for basal expression. However, none of our constructs provided luciferase activity at levels significantly higher than the rest.

Comparison of results between the methylated and mock-methylated reporter plasmids indicated that methylation resulted in only residual reporter gene activity, in contrast to the cells transfected with mock-methylated constructs (Fig. 4C). However, for the reporter construct −1662/−629 devoid of the putative CpG island, methylated and mock-methylated samples gave similar reporter activities, with an absolute signal just three- to fourfold above the pGL3-Basic background.

Methylated cell lines were more sensitive to ABCG2 substrate drugs.

To evaluate the therapeutic implication of the lower ABCG2 expression due to promoter methylation, we performed 4-day cytotoxicity assays on the three renal cancer cell lines in the presence of one of three ABCG2 substrate anticancer drugs, mitoxantrone, topotecan, or SN-38. A summary of results is shown in Table 3. For all three drugs, the methylated cell lines UOK121 and UOK143 were found to be more sensitive than the unmethylated cell line UOK181. The cytotoxicity assays were repeated with or without 5 μM FTC to assess the impact of ABCG2 inhibition on the cytotoxic effect of the drugs on the cells. FTC is a specific inhibitor of ABCG2 (46), and at 5 μM, FTC did not affect cell viability. In UOK121 and UOK143, 5 μM FTC had little effect on the cytotoxicity of the three ABCG2 substrates tested. In contrast, UOK181 cells were partially sensitized to the drugs in the presence of 5 μM FTC. The IC₅₀s for all the three drugs were reduced to approximately one-third of that without FTC (Table 3).

TABLE 3.

Partial sensitization of the unmethylated cell line UOK181 to ABCG2 substrate drugs by FTC

Cell line	IC50 (nM) ofa:
	Mitoxantrone		Topotecan		SN-38
	−FTC	+FTC	−FTC	+FTC	−FTC	+FTC
UOK121	36.4 ± 4.0	34.7 ± 6.8	39.0 ± 3.6	40.2 ± 1.2	47.9 ± 0.8	39.3 ± 0.8
UOK143	47.1 ± 10.3	44.2 ± 10.8	52.9 ± 7.5	49.1 ± 7.7	38.5 ± 5.6	29.9 ± 5.5
UOK181	304.2 ± 84.4	99.8 ± 31.8	1,075.0 ± 196.1	326.0 ± 30.2	3,918.0 ± 373.1	1,347.5 ± 278.7

^aIC₅₀ is the drug concentration effective in inhibiting 50% of cell growth as measured by the sulforhodamine B assay after a 96-h drug exposure. Results are shown as the mean ± standard deviation. The assays were repeated three times.

ABCG2 promoter methylation is associated with histone deacetylation and methylation at lysine 9 on histone H3 (H3K9).

Since transcriptional regulation by DNA methylation has been suggested to involve both deacetylation and methylation of histone H3 (20), the differences in these posttranslational modifications of histone, if any, between the ABGC2 alleles of UOK121, UOK143, and UOK181 cells were explored using a chromatin immunoprecipitation assay with antibodies against acetylated and trimethylated histone H3. Five sets of primers spanning both the distal and proximal (P1 to P4) ABCG2 promoter regions were designed for the ChIP assay (Table 1). Using an anti-acetyl-H3 antibody recognizing acetyl groups on lysine 9 of histone H3, analysis of immunoprecipitated DNA by PCR indicated that ABCG2 DNA levels (P2 to P4 regions) relative to GAPDH DNA levels were lower in immunoprecipitates from UOK121 and UOK143 compared with UOK181. (Figure 5 shows the representative P3 results.) In contrast, ChIP assays demonstrated that methylated H3K9 was present prominently at the methylated and silenced ABCG2 promoter in UOK121 and UOK143 cells, but was less abundant at the unmethylated and unrepressed ABCG2 promoter in UOK181 cells (Fig. 5). The result demonstrates a direct correlation between ABCG2 promoter methylation, histone methylation, and deacetylation (Lys 9) and loss of ABCG2 expression.

FIG. 5.

ABCG2 chromatin in the methylated cell lines is associated with less acetylated but more methylated histone H3 (K9) than the unmethylated cell line. (Upper panel) ChIPs were performed with UOK121, UOK143, and UOK181 cells. Acetylated and methylated H3 histones (K9) associated with the P3 region (−293 to −139) in the ABCG2 promoter were analyzed by PCR. Soluble chromatin used in immunoprecipitations had a typical size of <0.5 kb, as visualized by gel electrophoresis. Input, DNA isolated from the lysate before immunoprecipitation; IgG, ChIP using normal IgG for immunoprecipitation; 1, 2, and 3, UOK121, UOK143, and UOK181, respectively. (Lower panel) Quantitative analyses of the occupancy of acetylated or methylated H3 to the ABCG2 promoter (P3 region) in the UOK cells. The results are expressed as the percentage of immunoprecipitate (IP) over total input DNA utilized. Error bars show the standard deviations of three different experiments with independent chromatin preparations.

Treatment of cells with a demethylating agent has been demonstrated to restore histone acetylation on a methylation-mediated transcriptionally silent promoter (11), so histone acetylation on the ABCG2 promoter was examined. To test this proposition, ChIPs with antibodies against acetyl-H3 (Lys 9) were performed in UOK cells treated with 5-aza-dC for 4 days (Fig. 6A). 5-Aza-dC treatment induced hyperacetylation of histone H3 in the P2 to P4 regions in UOK121 and UOK143, but not in UOK181, compared to their respective untreated controls (Fig. 6A). The proportion of acetylated histone H3 (Lys 9) bound to the promoter region (P3) was enhanced approximately eight- and sevenfold in UOK121 and UOK143 cells, respectively. Interestingly, the enrichment of acetylated H3 in the distal region remained essentially unchanged after 5-aza-dC treatment (Fig. 6A). As a control, depsipeptide, a specific histone deacetylase inhibitor, also led to histone H3 hyperacetylation (P2 to P4 regions) in all the three RCC lines. Taken together, these results show that even in the absence of HDAC inhibition, gene demethylation can independently induce local promoter-associated histone hyperacetylation in the cell lines bearing the methylated ABCG2 alleles. Cellular levels of acetylated histone H3 and HDAC1 were also examined in whole-cell lysates to demonstrate that the findings on the promoter were specific but not due to global nonspecific events. In contrast to its promoter-specific effect, 5-aza-dC did not alter the levels of global histone acetylation, as seen in immunoblot analysis. Depsipeptide treatment, however, induced both global and gene-specific histone hyperacetylation (Fig. 6B). Combining 5-aza-dC and depsipeptide did not lead to a further increase in histone acetylation (data not shown).

FIG. 6.

ABCG2 chromatin is enriched with acetylated histone following gene demethylation. (A) Histone H3 (K9) acetylation and trimethylation enrichment on ABCG2 chromatin. ChIPs were performed with UOK121, UOK143, and UOK181 cells treated with 5-aza-dC (1 μM) for 4 days or depsipeptide (2 ng/ml) plus verapamil (5 μg/ml) for 24 h. Acetylated H3 or methylated H3 associated with the distal and P3 regions in the ABCG2 promoter were analyzed by PCR. a, b, and c represent no treatment, 5-aza-dC treatment, and depsipeptide-plus-verapamil treatment, respectively. The bottom panel shows the quantitative analyses of the occupancy of acetylated H3 or methylated H3 to the ABCG2 promoter (P3 and distal region) in the UOK cells. The results are expressed as the percentage of immunoprecipitate (IP) over total input DNA utilized. Error bars show the standard deviations of three independent experiments. (B) Immunoblot analysis of acetyl-H3 and HDAC-1. 5-Aza-dC did not alter the levels of global histone acetylation, whereas HDAC inhibition by depsipeptide resulted in hyperacetylation of H3 in whole-cell lysates. HDAC1 expression was not significantly affected by either treatment. Whole-cell lysates were prepared from UOK121, UOK143, or UOK181 cells for acetylated-histone H3 (17 kDa) and HDAC-1 (65 kDa) detection, respectively.

The other histone code, H3K9 methylation, was also examined in 5-aza-dC- or depsipeptide-treated cells. Interestingly, the association of methylated H3K9 with the proximal ABCG2 promoter (P2 to P4) was remarkably decreased in 5-aza-dC-treated UOK121 and UOK143 (Fig. 6A shows the representative P3 results). On the contrary, the association of methylated H3K9 with the proximal ABCG2 promoter in the unmethylated UOK181 cell line was lower and unaffected by 5-aza-dC treatment (Fig. 6A). Upon treatment with depsipeptide, the strong association of methylated H3K9 with the proximal ABCG2 promoter in the methylated cell lines was not affected, whereas the much weaker association in the unmethylated cell line was reduced (Fig. 6A).

Binding of methyl-CpG binding proteins to the ABCG2 promoter.

Previous work demonstrated that members of the methyl-CpG binding (MBD) protein family, including MBD1, MBD2 (MeCP1), MBD3, and MeCP2, mediate methylation-dependent transcriptional repression and are present in distinct HDAC-containing complexes (28, 37, 41-43). Among these methyl-CpG binding proteins, MBD2 and MeCP2 are better understood and provide a link between the two global mechanisms of gene regulation, DNA methylation and histone acetylation. A number of publications demonstrate conclusively that they are involved in the suppression of aberrantly methylated tumor suppressor genes by binding to methylated regulatory regions (3, 19). To assess whether the MBD proteins were involved in ABCG2 silencing, the protein expression of MBD2 and MeCP2 was determined by immunoblot analysis and their association with the ABCG2 promoter was examined using the ChIP assay.

MBD2 was strongly expressed in all of the three UOK cell lines (Fig. 7A). MeCP2 was highly expressed in UOK121 and UOK181, but much lower in UOK143 (Fig. 7A). Treatment of the three cell lines with either 5-aza-dC or depsipeptide did not alter MBD2 and MeCP2 protein expression levels in the whole-cell lysates.

FIG. 7.

The methylated ABCG2 promoter is associated with MBD proteins. (A) Immunoblot analysis of MBD proteins (MBD2 and MeCP2), HDAC1, and mSin3A in UOK cell lines. β-Actin was used as a loading control. Levels of protein expression of MBD2, MeCP2, HDAC1, and mSin3A were not changed in cells treated with 5-aza-dC or depsipeptide. (B) ChIP analyses targeting the P3 and the distal region of the ABCG2 promoter in UOK121, UOK143, and UOK181 cells, using antibodies against MBD2 and MeCP2. 1, 2, and 3 represent UOK121, UOK143, and UOK181, respectively. (C) A ChIP assay was performed targeting the GAPDH promoter in UOK cells using antibodies against MBD2 or MeCP2. GAPDH has no MBD2 or MeCP2 associated with it.

ChIP assay demonstrated that MBD2 was associated with both UOK121 and UOK143 promoters (P2 to P4 regions). Only results from the P3 region are shown for clarity. The association of MBD2 with the methylated (UOK121 and UOK143) promoter was much higher than their association with the unmethylated (UOK181) promoter, which was barely detectable (Fig. 7B). This is consistent with the hypomethylation status of the ABCG2 alleles in UOK181. Similarly, MeCP2 was shown to bind to the methylated ABCG2 promoter (P2 to P4 regions) in UOK121 cells (Fig. 7B). The weak association of MeCP2 with the other methylated promoter (UOK143) probably reflects the low cellular expression level of MeCP2 (Fig. 7A). Importantly, the weak binding of MeCP2 and MBD2 with the promoter in UOK181 cells again strongly suggested that the ABCG2 promoter is not methylated in these cells. As a control, the binding of MBD2 and MeCP2 to the promoter of a constitutively expressed GAPDH gene was also evaluated. As expected, no binding of any of these proteins to the GAPDH promoter could be detected (Fig. 7C), demonstrating the specificity of MBD protein binding to the ABCG2 promoter. There was also no binding of these MBDs to the ABCG2 distal region (Fig. 7B). Comparison of the quantitated levels of MBDs and acetylated H3 at each of the three CpG islands (UOK121, UOK143, and UOK181) demonstrated an inverse relationship between the two proteins. Islands occupied by MBDs were significantly deacetylated, whereas islands unbound by MBDs demonstrated higher levels of histone H3 acetylation (Fig. 5 and 7B).

The association of MBD2 and MeCP2 with the ABCG2 promoter was also analyzed by the ChIP assay in the UOK cells after treatment with 5-aza-dC. As shown in Fig. 8, 5-aza-dC significantly decreased the association of both MBD2 and MeCP2 with the ABCG2 promoter in UOK121 and UOK143 cells (region P3), whereas it did not have any impact on the already low binding of MBDs to the ABCG2 promoter in UOK181 (region P3). Interestingly, depsipeptide did not appreciably affect the binding of MBDs with the ABCG2 promoter in any cell line (Fig. 8).

FIG. 8.

5-Aza-dC treatment led to a release of MBDs, HDAC1, and mSin3A from the methylated ABCG2 promoter. ChIP analyses were carried out targeting the P3 and distal regions of the ABCG2 promoter in UOK121, UOK143, and UOK181 cells after 4 days of treatment with 5-aza-dC (1 μM) or 1 day of treatment with depsipeptide (2 ng/ml) plus verapamil (5 μg/ml), using antibodies against MBD2, MeCP2, HDAC1, or mSin3A. a, b, and c represent no treatment (No Tx), 5-aza-dC treatment, and depsipeptide treatment, respectively. The bottom panel shows the quantitative analyses of the occupancy of MBD2, MeCP2, HDAC1, and mSin3A to the ABCG2 promoter (P3 and distal region) in the UOK cells. The results are expressed as the percentage of immunoprecipitate (IP) over total input DNA utilized. Error bars show the standard deviations of three independent experiments.

MeCP2 has been proposed to interact with the transcriptional corepressors mSin3A and HDAC1 to direct a large component of gene repression in model systems (28, 41). Therefore, the association of HDAC1 and mSin3A with the ABCG2 promoter was assessed after DNA demethylation. UOK121 and UOK143 cells treated with 5-aza-dC consistently show reduced HDAC1 and mSin3A binding on the promoter, concomitant with the release of MBD2 and MeCP2 from the ABCG2 promoter (Fig. 8). These results are consistent with the conclusion that HDAC1 and mSin3A are brought to the promoter through association with MBD2 or MeCP2 on hypermethylated chromatin (28, 41).

Decreased association of DNMT1 and DNMT3a with the ABCG2 promoter upon 5-aza-dC treatment.

5-Aza-dC leads to DNA methylation by inhibiting DNMTs. The renal carcinoma cell lines were treated with either 5-aza-dC or depsipeptide, and the whole-cell lysates were subjected to immunoblot analysis with specific antibodies against the DNMTs. After a 4-day treatment of 5-aza-dC, the levels of all three DNMTs decreased significantly (Fig. 9A). However, depsipeptide did not alter expression of the DNMTs.

FIG. 9.

5-Aza-dC treatment inhibited the three known methyltransferases (DNMT1, DNMT3a, and DNMT3b), resulting in less association of them with the ABCG2 promoter. (A) Immunoblot analysis of DNMT1, DNMT3a, and DNMT3b in the UOK cell lines. β-Actin was used as a loading control. Levels of protein expression of the DNMTs were reduced significantly by 5-aza-dC treatment (4 days), but were not affected by treatment with depsipeptide. (B) ChIP analyses targeting the proximal P3 region of the ABCG2 promoter in UOK121, UOK143, and UOK181 cells using antibodies against DNMT1, DNMT3a, and DNMT3b. 1, 2, and 3 represent no treatment (No Tx), 5-aza-dC treatment, and depsipeptide-plus-verapamil treatment, respectively. (C) Quantitative analyses of the occupancy of DNMT1, DNMT3a, and DNMT3b to the proximal ABCG2 promoter (P3) in the UOK cells. The results are expressed as the percentage of immunoprecipitate (IP) over total input DNA. Error bars show the standard deviations of three independent experiments.

The association of the DNMTs with the ABCG2 promoter was assessed with the ChIP assay using the respective antibodies. The results demonstrate that among the three DNMT isoforms, DNMT1 and DNMT3a were associated with the ABCG2 promoter in all the three renal cancer cell lines, with a stronger association (about three times stronger) in the methylated UOK121 and UOK143 cell lines than the unmethylated UOK181 (Fig. 9B and C). Upon treatment with 5-aza-dC, but not depsipeptide, the association of both DNMT1 and DNMT3a was reduced to barely detectable levels. These results correlate with the decrease in their total cellular levels. Anti-DNMT3b antibodies did not result in immunoprecipitation of the ABCG2 promoter from the control or treated cells. The data suggest that activation of ABCG2 by 5-aza-dC in the methylated cell lines is correlated with the inhibition and the resulting relieved association of DNMT1 and DNMT3a with the promoter.

DISCUSSION

Aberrant DNA methylation patterns have been linked to altered gene expression in certain genetic diseases and tumors (5, 18). These observations are based on the fact that genome methylation regulates gene transcription and higher-order chromatin structure (30). Methylation of transcriptional regulatory sequences has been proposed to repress transcription of several mammalian genes by a variety of different molecular mechanisms, such as by direct interference with transcription factor binding, by altering the structure of chromatin, and/or by attracting MBD proteins (7, 38).

In this study, the methylation-mediated transcriptional repression of ABCG2 was examined in RCC cell lines. Determination of the methylation pattern in the ABCG2 gene was made using simple HpaII digestion followed by PCR amplification of the promoter region flanking the putative CpG island. Methylation at specific CpG dinucleotides was confirmed by bisulfite genomic sequencing. The ABCG2 promoter in the UOK121 and UOK143 cells was found to be methylated, but UOK181 was not methylated, the pattern of which paralleled the well-characterized methylation status in the VHL gene (23, 58). This can be of therapeutic importance. UOK121 and UOK143, having a methylated ABCG2 promoter and expressing a lower level of ABCG2, were found to be more sensitive to ABCG2 substrate drugs than the unmethylated UOK181. Moreover, partial sensitization of cells to ABCG2 substrate drugs by FTC, an ABCG2-specific inhibitor, was only appreciable in the unmethylated UOK181.

Consistent with the role of DNA methylation in ABCG2 silencing, incubation of methylated cell lines UOK121 and UOK143 with 5-aza-dC, a specific inhibitor of DNA methyltransferase, resulted in upregulation of ABCG2 expression in a concentration-dependent manner. No effect was observed in the unmethylated UOK181 cells. Finally, in vitro methylation of a luciferase reporter gene construct confirmed the silencing effect of methylation on the ABCG2 promoter.

It has been suggested that DNA methyltransferases may act only on chromatin that is methylated at lysine 9 on histone H3 (H3K9) (31, 51). Indeed, H3K9 methylation is sufficient for initiating a gene repression pathway in vivo (49). Our results suggest that modifications (methylation and deacetylation) of histone H3 assembled at the ABCG2 promoter and DNA methylation of the CpG island coordinately cause silencing of the ABCG2 gene. The ABCG2 expression in the methylated cell lines (UOK121 and UOK143) can be restored by 5-aza-dC treatment, and the reactivation is associated with hyperacetylation of H3 at lysine 9. Thus, apart from demethylating the ABCG2 promoter, 5-aza-dC also resets the histone code, switching it from methylation to acetylation at H3K9.

MBD proteins mediate silencing of genes by facilitating the establishment of a repressive chromatin environment (7, 57). To date, five methyl-CpG binding proteins have been identified: MeCP2, MBD1, MBD2, MBD3, and MBD4 (41-43, 48). These proteins recruit chromatin remodeling enzymes such as histone deacetylases and mSin3 to the DNA with their transcriptional repression domains, creating an inactive chromatin configuration (28, 38, 41). We evaluated the binding of these various proteins to the ABCG2 promoter. ChIP assays demonstrated binding of MeCP2, MBD2, HDAC1, and mSin3A to the CpG island region in UOK121 and UOK143, but only minimally in UOK181. ChIP analyses also revealed that 5-aza-dC treatment in UOK121 and UOK143 cells facilitated an enrichment of acetyl-H3, a release of MBDs (i.e., MeCP2 and MBD2), and a decreased occupancy of HDAC1 and mSin3A on the ABCG2 promoter, consistent with a more open chromatin conformation that would allow transcriptional activation. However, no appreciable change in the binding of transcription factors, Sp1 and c-JUN, was detected after 5-aza-dC treatment in the renal cell lines (data not shown). Taken together, the data support the notion that DNA methylation-dependent formation of a repressor complex in the CpG island contributes to inactivation of the ABCG2 gene.

The binding of the MBDs to methylated DNA results in recruitment of HDACs to support transcriptional repression. In our experiments, the HDAC inhibitor depsipeptide could upregulate ABCG2 expression most effectively in the unmethylated UOK181, but to a lesser extent in the methylated UOK121 and UOK143. It has been proposed that DNA methylation is dominant over other epigenetic factors in the regulation of gene transcription (10). At a dose of depsipeptide (2 ng/ml) that induced significant acetylation of histone H3 in the whole-cell lysate of the methylated UOK121 and UOK143 cells, limited ABCG2 upregulation was observed despite the fact that increased accumulation of acetylated histone on the ABCG2 promoter was noted. This suggests that DNA demethylation is more critical than histone acetylation for the ABCG2 gene chromatin to switch from a transcriptionally nonpermissive to a permissive configuration. In this regard, the repressive histone code, MeH3K9, was found binding strongly to the proximal ABCG2 promoter in the methylated cell lines, which was not affected by depsipeptide treatment. In contrast, depsipeptide did reduce the association of MeH3K9 with the ABCG2 promoter in the unmethylated cell line. This may explain why depsipeptide has less effect on upregulation of ABCG2 in the methylated cell lines than in the unmethylated cell line. A similar finding has been reported for FMR1 transcriptional silencing in fragile X cells (11). Interestingly, inhibition of histone deacetylation by trichostatin A following partial demethylation by 5-aza-dC does lead to a robust activation of a variety of methylated genes, such as MLH1, TIMP3, INKN2A, INKN2B, and MDR1 (10, 53). However, there was no synergistic effect of depsipeptide and 5-aza-dC on ABCG2 expression, in contrast to what has been reported for those aforementioned genes silenced by promoter methylation (10, 11). It is likely that methylation-dependent transcriptional repression of ABCG2 involves corepressors other than depsipeptide-sensitive histone deacetylase(s).

Although methylation-mediated silencing of ABCG2 is proposed in this study, it is interesting to note that ABCG2 expression level in the unmethylated cell line UOK181 is only about twofold higher that that in the methylated cell line. Notably, ABCG2 can be upregulated by fivefold in UOK181 by treatment with depsipeptide. Thus, a DNA methylation-independent pathway may exist for the transcriptional repression of ABCG2 in UOK181. Investigation into this proposed regulatory mechanism is now under way.

DNA methylation can confer a selective growth advantage to cells when it occurs in the promoter regions of genes repressing the expression of tumor suppressor genes, resulting in the development of cancer (26, 40). Since ABCG2 normally functions as an efflux transporter, the physiological significance of this methylation-dependent repression in cancer is not clear. Whether the repression of ABCG2 would provide an advantage to the cell, or merely be an epiphenomenon, is open to speculation. Folate deprivation has been reported to decrease ABCG2 expression (24), though the underlying mechanism is not known. Interestingly, down-regulation through hypermethylation of H-cadherin (25) and p16INK4A (45) has been reported in response to folate depletion. Since folates have been reported to be substrates for ABCG2, demethylation of its promoter could provide protection for the cell from folate deficiency. Regardless of whether ABCG2 repression could be of benefit to cancer cells, the identification of patients whose tumors have repressed ABCG2 could be important. Drugs that are substrates for ABCG2 would be expected to be more effective in such a patient population.

A more important role of ABCG2 methylation could be that in normal physiology. Gene methylation is believed to be the basic mechanism for the establishment and maintenance of genomic imprinting (12). Imprinted genes are marked in the male and female germ lines and retain the molecular memory of their parental origin, resulting in allelic expression differences. However, gene regulation at the promoter level during normal growth and development is not well understood. Abcg2 expression is increased in the murine lactating breast epithelium (29). Presumably, the levels decline during involution of the mammary gland by an unknown mechanism, and the involvement of methylation has not been excluded. Furthermore, ABCG2 is normally expressed in the placenta and in stem cells (9, 59). The possible role of promoter methylation in regulating ABCG2 expression in stem cells and their progeny has not been evaluated. Since it would require isolation of different populations of cells, including stem cells, progenitor cells, and terminally differentiated cells, this pursuit must await progress in the identification of specific markers for cells at different stages of development.