Reduced Folate Carrier Gene Silencing in Multiple Antifolate-resistant Tumor Cell Lines Is Due to a Simultaneous Loss of Function of Multiple Transcription Factors but Not Promoter Methylation*

The human reduced folate carrier (hRFC) is the major uptake route for antifolates used in cancer chemotherapy. Here we explored the molecular basis for the decrease or loss of hRFC gene expression in seventeen tumor cell lines with resistance to multiple antifolates due to impaired antifolate transport. We studied the role of various cis-acting elements including CRE/AP-1-like element and GC-box in hRFC promoters A and B, respectively, as well as AP-2, Mzf-1 and E-box that are contained within or near four tandemly repeated sequences upstream of promoter A. Decreased or abolished binding either to [32P]GC-box, Mzf-1, AP-1, E-box, or CRE oligonucleotides was detected in ∼50-80% of antifolate-resistant cell lines. Strikingly, ∼80% of the cell lines displayed a simultaneously decreased binding to three or more of these hRFC promoter elements, whereas normal AP-2 binding was retained. The possible contribution of promoter methylation to hRFC gene silencing was also explored. None of the antifolate-resistant cell lines, except for MDA-MB-231 cells, showed hRFC promoter methylation; consistently, MDA-MB-231 was the only cell line that retained binding to all six cis-acting elements. Western blot analysis demonstrated decreased expression of transcriptional activators (pCREB-1, pATF-1, USF-1, c-Fos, c-Jun, Sp1, and Sp3) and/or increased expression of repressors (short Sp3 isoforms), whereas normal AP2α levels were retained. Transient expression of the relevant transcription factors restored, at least partially, both promoter binding and hRFC gene expression. This is the first report that transcriptional silencing of the hRFC gene in multiple tumor cell lines with resistance to various novel antifolates is a result of a simultaneous loss of function of multiple transcription factors but not promoter methylation.

The human reduced folate carrier (hRFC) is the major uptake route for antifolates used in cancer chemotherapy. Here we explored the molecular basis for the decrease or loss of hRFC gene expression in seventeen tumor cell lines with resistance to multiple antifolates due to impaired antifolate transport. We studied the role of various cis-acting elements including CRE/AP-1-like element and GC-box in hRFC promoters A and B, respectively, as well as AP-2, Mzf-1 and E-box that are contained within or near four tandemly repeated sequences upstream of promoter A. Decreased or abolished binding either to [ 32 P]GC-box, Mzf-1, AP-1, E-box, or CRE oligonucleotides was detected in ϳ50 -80% of antifolateresistant cell lines. Strikingly, ϳ80% of the cell lines displayed a simultaneously decreased binding to three or more of these hRFC promoter elements, whereas normal AP-2 binding was retained. The possible contribution of promoter methylation to hRFC gene silencing was also explored. None of the antifolate-resistant cell lines, except for MDA-MB-231 cells, showed hRFC promoter methylation; consistently, MDA-MB-231 was the only cell line that retained binding to all six cis-acting elements. Western blot analysis demonstrated decreased expression of transcriptional activators (pCREB-1, pATF-1, USF-1, c-Fos, c-Jun, Sp1, and Sp3) and/or increased expression of repressors (short Sp3 isoforms), whereas normal AP2␣ levels were retained. Transient expression of the relevant transcription factors restored, at least partially, both promoter binding and hRFC gene expression. This is the first report that transcriptional silencing of the hRFC gene in multiple tumor cell lines with resistance to various novel antifolates is a result of a simultaneous loss of function of multiple transcription factors but not promoter methylation.
Reduced folates are essential cofactors involved in various one-carbon transfer reactions resulting in the biosynthesis of nucleotides and amino acids essential for cell growth and DNA replication (1). Since mammalian cells lack the enzymatic capacity for the de novo biosynthesis of tetrahydrofolate cofac-tors, these vitamins must be obtained from the diet (1). Folate analogues (i.e. antifolates) serve as anticancer drugs as they potently inhibit various enzymes involved in purine and deoxythymidylate biosynthesis resulting in inhibition of DNA synthesis and consequent cell death.
Reduced folates and antifolates are divalent anions that cannot traverse membranes by passive diffusion and are therefore transported into mammalian cells primarily by the reduced folate carrier (RFC) 1 (2,3). Among the antifolates that are recognized as transport substrates by the RFC are the dihydrofolate reductase inhibitors MTX, PT523 (4), and edatrexate (5), the novel thymidylate synthase inhibitors raltitrexed (Tomudex) (6) and ZD9331 (7), the glycinamide ribonucleotide transformylase inhibitor AG2034 (8) as well as the multitargeted antifolate (MTA), pemetrexed (9). In order for antifolates to exert their cytotoxic effect on tumor cells, high levels of RFC-dependent transport activity must be maintained (2,3). Therefore, impaired drug transport results in decreased intracellular concentration of antifolates and consequent drug resistance; defective antifolate transport is a major mechanism of drug resistance in cultured tumor cells subjected to clinically relevant concentrations of various antifolates (10 -14). Consistently, decreased hRFC gene expression and/or impaired antifolate transport are documented mechanisms of drug resistance in acute lymphoblastic leukemia (ALL) (15)(16)(17) and osteosarcoma patients that undergo MTX-containing chemotherapy (18 -20). Using human osteosarcoma specimens at diagnosis we have shown recently that expression of low RFC protein levels at the time of initial biopsy correlates with a poor response to preoperative chemotherapy containing MTX (20).
Decreased RFC mRNA levels can result from transcriptional silencing; the lack of hRFC transcripts in MTX-resistant MDA-MB-231 breast cancer cells with impaired MTX transport was due to promoter methylation (21). Furthermore, recent studies suggest a complex regulation of hRFC gene expression in normal tissues and tumor cell lines (22)(23)(24). Therefore, we recently initiated studies that focus on the elucidation of the molecular basis of the loss of hRFC gene expression in antifolate-resistant tumor cells. To this end, we recently found that alterations in the expression and binding of transcription factors to cAMP response element (CRE)/AP-1-like element and GC-box in hRFC promoters A and B, respectively, underlie the loss of hRFC gene expression in two antifolate-resistant cell lines with impaired antifolate transport (25). Apart from the constitutive GC-box and inducible CRE/AP-1 element, additional promoter elements including AP-2, myeloid zinc finger 1 (Mzf-1), and E-box are contained within or near four tandemly repeated sequences upstream of promoter A (Fig. 1). Using electrophoretic mobility shift assay (EMSA) and antibody-mediated supershift analysis we screened 17 antifolate-resistant human leukemia and breast cancer cell lines displaying impaired antifolate transport associated with a marked decrease or complete loss of RFC mRNA expression. Strikingly, the vast majority of the cell lines displayed a simultaneous loss of binding to three or more hRFC promoter elements including CRE, E-box, AP-1, Mzf-1, and GC-box, whereas AP-2 binding was retained. The markedly decreased binding to these cis-acting elements was associated with decreased expression of various transactivators. This is the first demonstration that simultaneous alterations in the expression and function of multiple transcription factors that bind to various cis-acting elements are a major mechanism of loss of hRFC gene expression in tumor cell lines that display resistance to various novel antifolates due to impaired drug uptake. We further show that while hRFC promoter silencing via methylation is rare in antifolate-resistant cell lines, it appears to alleviate the stress that would otherwise lead to alterations in the expression and function of transcription factors. Cell Lines and Tissue Culture-CCRF-CEM, a human T-cell leukemia line, and its antifolate-resistant sublines (14,26) were maintained in RPMI 1640 medium containing 2.3 M folic acid (Biological Industries, Beth-Haemek, Israel) supplemented with 10% fetal calf serum (Invitrogen), 2 mM glutamine, 100 units/ml penicillin G (Sigma), and 100 g/ml streptomycin sulfate (Sigma). The cell lines were established by single step or stepwise antifolate selection of parental CCRF-CEM cells as previously described (14,26). Human breast cancer ZR-75 cells and the MTX-resistant cell lines ZR-75-MTX R and MDA-MB-231 were grown in Dulbecco's modified Eagle's medium containing 2.3 M folic acid (Biological Industries, Beth-Haemek, Israel) supplemented with 10% fetal calf serum (Invitrogen), 2 mM glutamine, 100 units/ml penicillin G (Sigma), and 100 g/ml streptomycin sulfate (Sigma).

Drugs-MTX
Electrophoretic Mobility Shift and Antibody-mediated Supershift Assays-Nuclear extracts were prepared from exponentially growing cells (2 ϫ 10 7 cells) as previously described (27). DNA-protein complexes were formed by incubating nuclear extract proteins (6 g) with [␣-32 P]dCTP or [␣-32 P]dATP end-labeled CRE, GC-box, AP-1, AP-2, Mzf1, and E-box double-stranded oligonucleotides (Table I) as detailed elsewhere (28). Oligonucleotide competition was performed with 10 -100-fold molar excess of non-radioactive consensus oligonucleotides as previously described (25). A 2-fold decrease (or more) in the binding intensity as revealed by scanning densitometry relative to parental cells was scored as altered binding. Furthermore, decreased binding of a single band upon electrophoretic mobility shift assay was sufficient for scoring it as altered binding to a given consensus binding site. For supershift analysis, nuclear proteins (6 g) were incubated overnight at 4°C with anti-CREB-1, -ATF-1, -Sp1, -Sp3, -c-Jun, Jun B, c-Fos, and -USF-1 antibodies (4 g; Santa Cruz Biotechnology) prior to the addition of radiolabeled oligonucleotides as previously described (25). Protein concentration was determined by the colorimetric method of Bradford (29).
Semi-quantitative RT-PCR Analysis of hRFC Gene Expression-Exponentially growing cells (1 ϫ 10 7 ) were harvested by centrifugation, washed with phosphate-buffered saline and total RNA was isolated using the Tri-Reagent kit according to the instructions of the manufacturer (Sigma). A portion of total RNA (20 g in a total volume of 20 l) was reverse-transcribed using M-MLV (180 units, Promega) in a reaction buffer containing random hexamer primers, dNTPs, and a ribonuclease inhibitor Rnasin (Promega). Portions of cDNA (ϳ50 ng) synthesized from parental CEM cells and their antifolate-resistant sublines were amplified using Expand polymerase (Roche Applied Science) in a reaction buffer (total volume 25 l) containing: 10 pmol of each primer in 2xReddyMix PCR master mix reaction buffer according to the instructions of the manufacturer (ABgene, Surrey, UK). The PCR reaction was performed as follows: initial melting at 95°C for 5 min, followed by 25-35 cycles each of 1 min at 95°C, annealing at 62°C for 45 s, elongation at 72°C for 1 min, followed by 10 min extension at 72°C. Then, the PCR products were resolved on 2% agarose gels containing ethidium bromide. The primers used for the semi-quantitative RT-PCR of hRFC and GAPDH were previously described (25).
Transient Transfections with Various Expression Constructs-Exponentially growing suspension cells (2 ϫ 10 7 ) were harvested by centrifugation and transfected by electroporation (1000 F, 234 V) with 10 g of the following expression plasmids: pCREB1-VP16 (kindly provided by Dr. M. E. Greenberg), pmyc-c-fos, pHA-c-Jun and pPacSp1 (a gift from Dr. G. Suske). Cells were then seeded at 2 ϫ 10 6 /ml in prewarmed growth medium. For transient RFC mRNA expression and EMSA, after 1. Schematic representation of the hRFC promoter region. Putative binding sites for various transcription factors are designated by geometrical shapes. The two dark gray boxes represent minimal promoters A and B (22), whereas the light gray boxes R1-R4 denote nearly identical tandem repeats upstream of promoter A. Nucleotide numbering was relative to translation initiation ATG which represents position ϩ1 using accession number 7717445.
24 h of incubation at 37°C, cells were harvested and total RNA and nuclear proteins were extracted.
Determination of hRFC Promoter Methylation-We have devised two independent assays in order to explore hRFC promoter methylation in the various antifolate-resistant cell lines.
A) Bisulfite DNA Sequencing Assay-The first assay is based on bisulfite modification of genomic DNA followed by genomic PCR and DNA sequencing. Bisulfite treatment of genomic DNA was carried out using CpGenome DNA modification kit according to the instructions of the manufacturer (Intergen) under conditions that allowed the complete conversion of all cytosine residues (but not 5-methylcytosine) to uracil. DNA treated in vitro with SssI methyltransferase (New England Biolabs) was used as a positive control for methylated hRFC alleles. Following bisulfite modification, DNA was PCR-amplified using Reddymix (ABgene) and 10 pmol of each primer (5Ј-ATCCCAAAAC-CCCAAAA-3Ј, and 5Ј-GGATTTTAGGGTTAGTTT-3Ј), corresponding to a promoter region of hRFC where methylation was previously found to result in transcriptional silencing (21) (see Fig. 9). All positions refer to ATG of the translation initiation as ϩ1 using accession number 7717445. Reaction mixtures were incubated at 95°c for 3 min, followed by 35 cycles of 1 min at 95°C, 2 min at 53°C, and 3 min extension at 72°C; a final extension at 72°C for 10 min was performed to complete the genomic amplification process. Products were then resolved by electrophoresis on a 1.5% agarose gel, purified (Qiagen), and sequenced.
B) Southern Blot-based Assay-Genomic DNA from the various cell lines was co-digested with NcoI (Fermentas) and MunI along with either SmaI (methylation-sensitive restriction enzyme) or XmaI (a methylation insensitive isozyme of SmaI) according to the instructions of the manufacturer (New England Biolabs). Following digestion, DNA was fractionated by electrophoresis (2.2 V/cm) on 0.8% agarose gels containing 0.2 g/ml ethidium bromide. DNA in the gel was then  P-labeled genomic hRFC probe (see Fig. 9). Following an overnight hybridization at 65°C, blots were washed under high stringency conditions with a final wash in a solution of 0.1ϫ SSC, 0.1% SDS at 65°C for 30 min, and visualized by a phosphorimaging device.

Loss of [ 3 H]MTX Transport and Antifolate
Resistance-In the present study we used a large panel of seventeen antifolateresistant leukemia and breast cancer cell lines, the features of which are summarized in Table II; these previously published characteristics also include initial rates of [ 3 H]MTX transport. The up to 99% loss of [ 3 H]MTX uptake in these antifolateresistant cell lines was associated with a marked decrease in RFC protein expression. Furthermore, in approximately two thirds of the cell lines, heterozygous nonsense mutations and/or missense mutations were present. Consequently, these cell lines displayed up to ϳ3,500-fold cross-resistance to various hydrophilic antifolates.
hRFC mRNA Expression-The decreased RFC protein levels suggested a consistent loss of hRFC gene expression. We therefore used semi-quantitative RT-PCR analysis in order to determine hRFC mRNA levels (normalized to GAPDH) in the various antifolate-resistant cell lines. Whereas readily detectable in parental CEM cells after 25 cycles of RT-PCR, hRFC cDNA was observed in neither of the antifolate transport-defective cell lines (Fig. 2A); consistently, after 30 cycles of RT-PCR, decreased RFC cDNA levels could be detected in eight of the seventeen antifolate-resistant cell lines, whereas no detectable expression could be observed in the remainder (Fig. 2B).
Decreased Binding of Transcription Factors to Various hRFC Promoter Elements in Antifolate-resistant Cell Lines-These RT-PCR data established that RFC gene expression was pro-foundly suppressed in various cell lines. To explore potential alterations in the binding of transcription factors to the various cis-acting elements in the hRFC promoter region, we used electrophoretic mobility shift assay (EMSA). Nuclear factor binding to consensus 32 P-labeled oligonucleotides including CRE, E-box, AP-1, Mzf-1, and GC-box was markedly decreased (Fig. 3, A-E, respectively), whereas normal AP-2 binding was retained in all antifolate-resistant cell lines (Fig. 3F). Fig. 4A summarizes the percentage of cell lines with decreased oligonucleotide binding; 47-76% of the cell lines had decreased transcription factor binding either to GC-box, Mzf-1, AP1, Ebox, or CRE. Strikingly, ϳ76% of the cell lines displayed a simultaneous decrease in the binding to three or more hRFC promoter elements, whereas, only ϳ18% of the cell lines had one or two hRFC elements altered in binding (Fig. 4B). In contrast, MDA-MB-231 was the only cell line that retained normal binding to all six cis-acting elements.
Alterations in the Expression of Various Transcription Factors-Recently we have shown that decreased CRE-and GCbox binding in the hRFC promoter is due to alterations in the expression of the relevant transcription factors (25). We therefore used Western blot analysis in order to survey the levels of pCREB-1, pATF-1, USF-1, c-Jun, c-Fos, Sp1, Sp3 isoforms, and AP-2␣ (Fig. 5).
pCREB and pATF-1-All cell lines with decreased CRE binding either had prominently diminished or undetectable levels of phosphorylated CREB-1 (pCREB), the transcriptionally active form of CREB (Fig. 5A). Consistently, the vast majority of cell lines with decreased pCREB levels retained near normal levels of pATF-1 (Fig. 5A), whereas only a few cell lines had decreased levels of both pCREB-1 and pATF-1 (Fig. 5A, compare lanes 3  and 11 with 1). A ϳ35-kDa pATF/pCREB family member that was barely detectable in parental cells was overexpressed in almost all cell lines with decreased CRE binding (Fig. 5A). USF-1-The majority of cell lines with decreased E-box binding had substantially diminished levels of USF-1 (Fig. 5B). Consistently, PT R0.5 N, CEM-T, and CEM-MTX R1 cells with a dramatic loss of E-box binding (Fig. 3B, see loss of the upper band in lanes 6, 15, and 16), displayed barely detectable levels of USF-1 (Fig. 5B). In contrast, CEM/MTX and EDX R0.03 cells with a markedly increased E-box binding (Fig. 3B, see the intense upper band in lanes 8 and 11) showed consistently elevated USF-1 levels (Fig. 5B, compare lanes 8 and 11 with 1).
c-Jun and c-Fos-Whereas parental CEM cells expressed comparable levels of c-Jun and c-Fos, various cell lines with decreased AP-1 binding had a substantial decrease in c-Jun (Fig. 5C) and/or c-Fos levels (Fig. 5D).
Sp1 and Sp3 Isoforms-With the exception of CEM-MTX R1 cells in which Sp1 levels were markedly diminished, decreased GC-box binding was not associated with decreased Sp1 expression in the various cell lines (Fig. 5E). However, consistent with our previous studies (25), several cell lines with deceased GCbox binding had elevated levels of the short Sp3 isoform (Fig.  5F), an established repressor of RFC transcription (22). However, decreased GC-box binding in several cell lines could not be attributed to alterations in Sp1 or Sp3 levels.
AP-2␣-Consistent with the intact AP2 binding in all cell lines was the maintenance of parental AP2␣ levels (Fig. 5G).
Antibody-mediated Supershift Analysis-To identify the individual transcription factors that participate in the binding to the various hRFC promoter elements, we used antibody-mediated supershift analysis (Fig. 6).
CRE-Preincubation of nuclear proteins with antibodies to CREB-1 and ATF-1 or a combination of the two, followed by [ 32 P]CRE protein-antibody complex formation, revealed supershifts in parental CEM cells (Fig. 6A, left panel, complexes A  and B in lanes 2 and 3, 4, respectively; note the elimination of band 1 with anti-CREB-1 and band 2 with anti-ATF-1 antibodies, respectively). In contrast, CEM/MTX R1 cells with a markedly decreased CRE binding had only residual ATF-1 complex formation (Fig. 6A, right panel). These results are consistent with the retention of pATF-1 expression (Fig. 5A).
E-box-Antibodies against USF-1 supershifted band 1 and formed a high molecular weight complex A (Fig. 6B, compare lane 2 with 1). In contrast, CEM/MTX R1 cells that lost band 1 but not band 2 of E-box binding (Fig. 3B) did not display any supershift (Fig. 6B, right panel).
AP-1-We used antibodies against AP-1 activators (c-Jun and c-Fos) and repressors (Jun B) (Fig. 6C). Anti-c-Jun antibodies revealed a supershifted complex B (Fig. 6C, left panel). Whereas anti-c-Fos antibodies eliminated bands 2 and 3, anti-Jun B antibodies eliminated band 2 and formed complex A (Fig.  6C, left panel). Consistently, co-treatment with both anti-c-Jun, c-Fos, and Jun B antibodies eliminated bands 2 and 3, reduced the intensity of band 1 and formed complexes A and B (Fig. 6C,  left panel). In contrast, CEM/MTX R1 with an abolished AP-1 binding (Fig. 3C) failed to show any detectable supershift (Fig.  6C, right panel).
GC-box-Anti-Sp1 antibody eliminated band 2 and formed complex A, whereas anti-Sp3 antibody eliminated bands 1 and 3, and formed complexes B and C in nuclear extracts from parental CEM cells (Fig. 6D, left panel). Consistently, treatment with both antibodies eliminated bands 1, 2, and 3 and formed complexes A, B, and C (Fig. 6D, left panel). In contrast, CEM/MTX R1 cells (Fig. 3E), which had a marked loss of GC-box binding, showed only residual supershifts with both anti-Sp1 and -Sp3 antibodies (Fig. 6D, right panel).
Restoration cell lines prompted us to explore the impact of their transient expression on restoration of: (a) transcription factor binding to the various hRFC promoter elements, and (b) RFC mRNA expression. Transfection of representative cell lines with expression constructs harboring CREB-1, c-Fos, and c-Jun, or Sp1 resulted in a partial or complete restoration of binding to CRE (Fig. 7A), AP-1 (Fig. 7B), and GC-box (Fig. 7C), respectively. This resulted in at least a partial restoration of hRFC mRNA expression relative to parental cells (Fig. 8).
Assessment of hRFC Promoter Methylation in Antifolate-Resistant Cell Lines-Recently it was shown that methylation of hRFC promoter A occurs in MTX-resistant MDA-MB-231 cells resulting in transcriptional silencing and impaired drug transport (21). Thus, to explore the frequency of hRFC promoter methylation in the various antifolate-resistant cell lines, we devised two hRFC promoter methylation assays. The first involves the bisulfite DNA sequencing technique that is based on the treatment of genomic DNA with bisulfite thereby convert-ing only unmethylated cytosines to uracils. Examination of a region upstream to promoter A that contains 13 CpGs (Fig. 9A, see dotted line) that was shown (21) to be fully methylated in MDA-MB-231 cells (Fig. 9B), revealed that none of the parental and antifolate-resistant cell lines showed CpG island methylation (Fig. 9B). In the second approach we devised a Southern blot assay that is based on genomic DNA digestion with methylation-sensitive (SmaI) and -insensitive restriction isoenzymes (XmaI); this assay was designed to analyze the methylation status of a more downstream region of the hRFC promoter A. Blots were then hybridized with a 32 P-labeled 537 bp genomic hRFC probe (Fig. 9A, see light gray box). The 600-bp band obtained after SmaI digestion was diagnostic for unmethylated DNA and was observed in all cell lines except for MDA-MB-231 cells that contained methylated CpG islands and therefore could not undergo digestion with SmaI (Fig. 9C). onstrate that regions upstream and downstream of hRFC promoter A are not methylated in any of the antifolate-resistant cell lines examined here except for MDA-MB-231 cells. Strikingly, the only MTX-transport defective cell line that retained intact binding to all hRFC promoter elements was MDA-MB-231 (Fig. 3, A-F). In contrast, no promoter methylation was found in ZR-75-MTX R , another MTX-resistant breast cancer cell line with impaired antifolate transport, decreased Mzf-1 binding (Fig. 3D) and a heterozygous premature translation termination mutation in the hRFC (Table II). Thus, the loss of hRFC mRNA expression in MDA-MB-231 cells that did not harbor any RFC mutation (Table II) could be solely attributed to hRFC promoter methylation. These results suggest that alterations in the binding of transacting factors and promoter methylation appear to be mutually exclusive mechanisms of transcriptional silencing of the hRFC gene. DISCUSSION We have recently shown that loss of hRFC gene expression in two drug-resistant cell lines with impaired antifolate transport is associated with alterations in the expression and binding of transcription factors to the inducible CRE/AP-1 and constitutive GC-box in hRFC promoters A and B, respectively (25). The aim of the current study was to explore whether multiple antifolate-resistant cell lines with decreased (or abolished) hRFC mRNA expression display altered binding of transcription factors to additional cis-acting elements that are contained within or near four tandemly repeated sequences upstream of promoter A including E-box, AP-2 and Mzf-1. Extending on our recent findings (25), we also examined the status of transcription factor binding to CRE/AP-1 and GC-box in this large panel of cell lines that are resistant to various antifolates. We strikingly found that the majority of these cell lines displayed loss of binding to E-box, AP-1, and Mzf-1, cis-acting elements that were not previously associated with decreased hRFC gene expression. Furthermore, as much as ϳ80% of the cell lines displayed a simultaneous loss of binding to three or more hRFC promoter elements; this was consistently associated with decreased expression of the relevant transcription factors that act as transcriptional activators including USF-1, c-Fos, c-Jun, pCREB-1, and pATF-1. In contrast, although binding to AP-2 was previously shown to promote hRFC gene expression (23) Drug-resistant cell lines with decreased (or loss) of binding to cis-acting elements due to specific alterations in transcription factor expression, were transiently transfected with expression vectors harboring specific transcription factors. Then, the binding to 32 P-labeled CRE, AP-1, and GC-box (panels A, B, and C, respectively) was determined.
part, both binding to the cis-acting elements and hRFC gene expression. Using two experimental approaches, we found that none of the antifolate-resistant cell lines contained hRFC promoter methylation, except for MDA-MB-231. Consistently, the latter cells retained normal binding to all hRFC promoter elements. We conclude that the simultaneous decrease in the expression and/or function of multiple transcription factors involved in activation of hRFC gene expression is a frequent mechanism of resistance to multiple novel antifolates in malignant tumor cell lines with impaired antifolate transport. Our results further suggest that decreased binding of transactivators to sequences upstream to promoter A including four nearly identical tandem repeats containing Mzf-1 and GC-box as well as a 5Ј-neighbor E-box appear to play an important role in the silencing of the hRFC gene in antifolate-resistant cell lines.
Of the seventeen antifolate-resistant cell lines, only MDA-MB-231 cells retained normal binding to all six hRFC promoter elements that were examined here. This inherently MTX-resistant breast cancer cell line was recently found to completely lack hRFC transcripts (30) due to silencing of the hRFC gene via promoter methylation (21). In contrast, the MTX-resistant breast cancer cell line ZR-75-MTX R with barely detectable hRFC mRNA levels displayed a markedly decreased Mzf-1 binding. Furthermore, neither ZR-75-MTX R cells nor any of the other antifolate-resistant cell lines had hRFC promoter methylation. These results strongly suggest that hRFC promoter methylation and alterations in the binding of various transcription factors to different cis-acting elements in the hRFC promoter region are presumably mutually exclusive mechanisms of gene silencing. The following evidences support the conclusion that the presence of CpG island methylation in the hRFC promoter may alleviate the biochemical stress imposed by antifolates that would otherwise lead to decreased binding of transcription factors to various hRFC promoter elements. (a) Methylated CpG islands can directly interfere with the binding of transcription factors to their cognate recognition sites (31). Most mammalian transcription factors have GC-rich binding sites and contain many CpGs in their DNA recognition elements. Binding by several of these factors is impeded or abolished by CpG methylation. (b) DNA methylation can directly influence the translational positioning of a nucleosome at specific DNA sequences in vitro and could lead to chromatin remodeling and consequent masking of essential regulatory elements by nucleosomes (32). (c) In addition to these direct modalities of physical masking of cis-acting elements, evidence exists for mechanisms of indirect repression mediated by proteins that specifically bind to methylated DNA. These include methyl-CpG-binding protein 1 (MeCP1) (33) and MeCP2 (34), as well as a family of novel mammalian proteins known as methyl-CpG binding domain 1, 2, and 4 (MBD1, 2 and 4) (35)(36)(37). Strikingly, MeCP2 and MBD2 that harbor a transcriptional repression domain (TRD) are thought to repress transcription by recruiting a histone deacetylase complex that modifies chromatin structure (37,38). Indeed, Worm et al., (21) have shown histone deacetylase complex recruitment in the hRFC gene in MDA-MB-231 cells with promoter methylation. Hence, these evidences emphasize the high efficiency of the direct and indirect mechanisms of transcriptional repression mediated by promoter methylation thereby eliminating the biochemical stress that would otherwise lead to loss of function of various transactivators.
Several lines of evidence indicate that loss of RFC function typically associated with a ϳ95% decrease in MTX transport activity in multiple antifolate-resistant cell lines is based upon the coexistence of several modalities of drug resistance: (a) Two-thirds of the seventeen antifolate-resistant cell lines studied here harbored antifolate transport inactivating mutations (missense and/or nonsense). The large proportion of premature translation termination mutations in these antifolate-resistant cell lines could clearly contribute to the loss of hRFC mRNA expression via nonsense-mediated mRNA decay (NMD) (39,40). Recent studies have indicated that an mRNA species harboring a premature translation termination mutation (i.e. stop codon) may be rapidly degraded via a quality control mechanism known as NMD. However, since these nonsense mutations were all heterozygous and as human leukemia cells frequently contain three RFC alleles, one or more normal RFC allele(s) were retained, thereby subjecting them for additional modalities of loss of function. (b) Indeed, all the antifolateresistant cell lines examined here either had a markedly decreased or a complete loss of hRFC gene expression. This was associated with the simultaneous loss of binding of multiple transcription factors to their hRFC promoter elements. (c) Using Southern blot analysis we recently found that the majority of these antifolate-resistant cell lines displayed hRFC allele loss. 2 This finding is consistent with previous studies that also identified hRFC deletions in antifolate-resistant human erythroleukemia cells exhibiting a 90% loss of hRFC mRNA expression (41).
Loss of binding to CRE, E-box, and AP-1 in the multiple antifolate-cell lines could be explained by decreased levels of the relevant activating transcription factors including pCREB-1 and pATF-1, USF-1, as well as c-Jun and c-Fos, respectively. However, the levels of Mzf-1 could not be exam-2 Y. Kaufman, I. Ifergan, L. Rothem, G. Jansen, and Y. G. Assaraf, submitted for publication. ined due to the lack of commercially available antibodies (23). In contrast, the loss of binding to GC-box could not be attributed to decreased Sp1 or Sp3 levels in various antifolate-transport-defective cell lines, except for CEM-MTX R1 and PT R0.5 Ϫ N cells, which displayed markedly decreased Sp1 and Sp3 levels, respectively. One viable possibility that may be consistent with these results is that Sp1 and Sp3 undergo post-translational modifications that compromise their transactivator capabilities. Members of the Sp/Krü ppel-like family of transcription factors undergo several post-translational modifications that modulate their transactivating potential including O-linked glycosylation, serine and/or threonine phosphorylation and acetylation (42). It was specifically shown that the introduction of an O-linkage of N-acetylglucosamine to the activation domain of Sp1 inhibits its transcriptional capability in vitro (43) and in vivo (44). Furthermore, the corepressor mSin3A was found to recruit O-linked N-acetylglucosamine transferase to promoters resulting in O-linked glycosylation of transcription factors like Sp1 yielding transcriptional repression (42). Interestingly, O-linked glycosylated proteins are also phosphoproteins and these two types of modifications may be reciprocally regulated in Sp1 (45). Threonine phosphorylation of the zincfinger domain of Sp1 by casein kinase II results in a reduced affinity of Sp1 for its consensus binding site (46,47). In the present study we found that two cell lines (AG2034 R2 and EDX R0.03 ) with loss of GC-box binding had normal Sp1 levels but markedly increased levels of both the long and short forms of Sp3. These results may be consistent with the current literature that the short forms of Sp3 act as transcriptional repressors. Furthermore, the long form of Sp3 that, as opposed to all members of the Sp family, contains an inhibitory domain that may inhibit transcription; this activity is apparently regulated FIG. 9. Analysis of CpG island methylation in hRFC promoter A region (A) in parental cells and their antifolate-resistant cell lines by bisulfite sequencing (B) and Southern blots analysis after DNA digestion with methylation-sensitive restriction endonuclease (C). Panel A, schematic representation of hRFC promoters A and B regions with emphasis on the sites analyzed for DNA methylation; the dotted line spans a region upstream of promoter A that was examined for CpG island methylation using the bisulfite sequencing method. The light gray box represents the genomic probe from a downstream region to promoter A that was used for the assessment of DNA methylation by Southern blot analysis. Panel B, genomic DNA was treated with sodium bisulfite, PCR-amplified and sequenced (see "Materials and Methods"). Methylated cytosine residues in CpG dinucleotides are represented by asterisks. The top and bottom colored tracings of DNA sequencing represent a tract of nucleotides from the upstream region of promoter A in MDA-MB-231 and ZR-75-MTX R cells, respectively (see dotted line in panel A). Note that all cytosines are methylated in MDA-MB-231 but not in ZR-75-MTX R cells. Panel C, the methylation status of the hRFC promoter A region was also determined by co-digesting genomic DNA from the various cell lines with NcoI and MunI along with either SmaI (methylation-sensitive) or XmaI (methylation-insensitive isozyme of SmaI). Digested DNA was resolved by agarose gel electrophoresis and transferred to a nylon membrane. Blots were then hybridized with a 537-bp genomic probe (see light gray box in panel A). When digesting with SmaI, the 600-bp band obtained represents the unmethylated RFC allele while its absence denotes a methylated allele. Note that the control XmaI band is obtained whether or not DNA is methylated, as this enzyme is methylation-insensitive. by acetylation of a critical lysine residue in the inhibitory domain (48). Further studies are underway to explore whether post-translational modifications of Sp1 and Sp3 including Olinked glycosylation, phosphorylation, and acetylation occur in antifolate-resistant cells displaying loss of GC-box binding while retaining normal Sp1 levels.