A Methylation-responsive MDBP/RFX Site Is in the First Exon of the Collagen α2(I) Promoter*

DNA methylation inhibits transcription driven by the collagen α2(I) promoter and the 5′ end of the gene in transient transfection and in vitro transcription assays. DNA-binding proteins in a unique family of ubiquitously expressed proteins, methylated DNA-binding protein (MDBP)/regulatory factor for X box (RFX), form specific complexes with a sequence overlapping the transcription start site of the collagen α2(I) gene. Complex formation increased when the CpG site at +7 base pairs from the transcription start site was methylated. The identity of the protein was demonstrated by co-migration and cross-competition for a characteristic slowly migrating doublet complex formed on MDBP/RFX recognition sequences and the collagen sequences by band shift assays. A RFX1-specific antibody supershifted the collagen DNA-protein complexes. Furthermore, in vitro translated RFX1 protein formed a specific complex with the collagen sequence that was also supershifted with the RFX1 antibody. MDBP/RFX displayed a higher affinity binding to the collagen sequence if the CpG at +7 was mutated in a manner similar to TpG. This same mutation within reporter constructs inhibited transcription in transfection and in vitro transcription assay. These results support the hypothesis that DNA methylation-induced inactivation of collagen α2(I) gene transcription is mediated, in part, by increased binding of MDBP/RFX to the first exon in response to methylation in this region.

DNA methylation inhibits transcription driven by the collagen ␣2(I) promoter and the 5 end of the gene in transient transfection and in vitro transcription assays. DNA-binding proteins in a unique family of ubiquitously expressed proteins, methylated DNA-binding protein (MDBP)/regulatory factor for X box (RFX), form specific complexes with a sequence overlapping the transcription start site of the collagen ␣2(I) gene. Complex formation increased when the CpG site at ؉7 base pairs from the transcription start site was methylated. The identity of the protein was demonstrated by co-migration and cross-competition for a characteristic slowly migrating doublet complex formed on MDBP/RFX recognition sequences and the collagen sequences by band shift assays. A RFX1-specific antibody supershifted the collagen DNA-protein complexes. Furthermore, in vitro translated RFX1 protein formed a specific complex with the collagen sequence that was also supershifted with the RFX1 antibody. MDBP/RFX displayed a higher affinity binding to the collagen sequence if the CpG at ؉7 was mutated in a manner similar to TpG. This same mutation within reporter constructs inhibited transcription in transfection and in vitro transcription assay. These results support the hypothesis that DNA methylationinduced inactivation of collagen ␣2(I) gene transcription is mediated, in part, by increased binding of MDBP/ RFX to the first exon in response to methylation in this region.
Type I collagen, the most abundant collagen molecule within the collagen family, normally consists of a heterotrimer of two ␣1(I) chains and one ␣2(I) chain. Synthesis of these genes is often down-regulated upon oncogenic transformation of cells in culture (1)(2)(3)(4). We have previously demonstrated down-regulation of the collagen ␣2(I) gene, encoding one of the subunits of Type I collagen in an epithelial-like cell line from rat liver, K16 cells upon their conversion to a tumorigenic line, W8, after treatment with the carcinogen 2-N-(acetoxyacetyl)-aminofluorine (3). The promoter-5Ј region of the ␣2 gene was methylated in the nonexpressing W8 cells and not in the expressing K16 cells (5). Furthermore, reporter gene expression downstream of the 218-bp 1 promoter and the 54-bp 5Ј region of the rat and human collagen ␣2(I) genes was inactivated by in vitro DNA methylation in transient transfection experiments, whereas an analogous expression plasmid with the SV40 early promoter/ enhancer driving expression was not (6). We also demonstrated that a minimal collagen ␣2(I) promoter containing the preinitiation region (Ϫ41 to ϩ54) driving expression of the luciferase reporter gene was inactivated by DNA methylation (7). The inhibition of reporter gene expression was attributable to CpG methylation, specifically of collagen ␣2(I) sequences. However, all the methylation sites were located in the first exon, not in the promoter.
DNA methylation in the promoter and 5Ј region of genes often correlates with decreased transcription of vertebrate genes, and many studies indicate that this methylation is often causally involved in down-regulation of gene expression (8 -14). Different mechanisms have been hypothesized to explain the inactivation due to methylation. In certain cases, methylation inhibits binding of positive transactivating factors. Alternatively, DNA methylation can induce the binding of the nonspecific methyl-sensitive proteins, such as MeCP1 or MeCP2, that bind to methylated cytosine regardless of surrounding bases and act as global repressors by condensing chromatin. In addition, a family of closely related proteins called methylated DNA-binding protein (MDBP) or regulatory factor for X box (RFX) 1-4 (15)(16)(17)(18) can bind methylated DNA sequences with higher affinity within a sequence-specific 14-bp consensus sequence, 5Ј-RT(m 5 C/T)RYYA(m 5 C/T)RG(m 5 C/T)RAY-3Ј (where (m 5 C/T) indicates 5-methylcytosine or T, R indicates G or A, and Y indicates C or T). Methylation-dependent binding sites were located for this protein at the beginning of the human genes for hypoxanthine phosphoribosyl transferase; ␣-galactosidase A; human leukocyte antigens (HLA)-A2, -A3, and -A25 antigens; and the apoferritin H gene (19). For the first two X-linked genes, DNA methylation may help down-regulate gene expression on the inactive X chromosome by increasing binding of MDBP at the three MDBP sites in the hypoxanthine phosphoribosyl transferase promoter/5Ј region (20). The MDBP sites at the beginning of the first exon of the ␣-galactosidase A gene are at least partially methylated on the inactive X chromosome but completely unmethylated on the active X chromosome (21).
Cytosine methylation-independent sites have been identified that contain T residues replacing 5-methylcysteine residues (22) in hepatitis B virus, polyoma virus enhancers, cytomegalovirus (CMV) enhancers and c-Myc intron (19,(23)(24)(25)(26)(27). Therefore, MDBP family proteins, consisting of homo-or het-erodimers of RFX1-4 subunits (17), can bind in a cytosine methylation-independent or -dependent fashion to their cognate sites, depending on the sequence of these sites. These constitutively expressed proteins can act as repressors or activators in a context-dependent fashion (28,29). An activation domain containing a glutamine-rich region is found in the N-terminal half of RFX1, whereas a region with repressor activity overlaps the C-terminal dimerization domain.
In our previous studies (7), we demonstrated that methylation sites within the first exon, which inactivated the ␣2(I) collagen promoter, bind to a sequence-specific methylationresponsive protein. Also, there was decreased formation of a TATA binding complex on methylated DNA ligands in gel shift experiments. This report demonstrates that MDBP/RFX1 binds to the first 20 bases of the first exon. When this sequence, which matches the consensus sequence for MDBP at 10 out of 14 positions in the center of the oligonucleotide, is methylated at its one CpG dinucleotide pair or is mutated to TpG at this site, there is increased binding of MDBP/RFX and inhibition of collagen gene transcription. Therefore, MDBP/RFX protein is likely to contribute to down-regulation of collagen gene repression and might do so in response to increased methylation associated with oncogenic transformation.
Electrophoretic Mobility Gel Shift Assay-Nuclear extracts were prepared essentially according to Dignam et al. (30), with some modifications. Extractions of protein from isolated nuclei were performed at higher salt conditions than normal using 500 mM NaCl or 420 mM NaCl rather than 350 mM NaCl in Buffer C. All buffers contained the protease inhibitors leupeptin (40 g/ml), aprotinin (200 g/ml), pepstatin A (40 g/ml), and phenylmethylsulfonyl fluoride (0.5 mM) as well as the phosphatase inhibitor orthovanadate (1 mM). Protein concentration of the extracts was determined by the Bradford reagent using bovine serum albumin as a standard. Collagen sequences (Table I) or MDBP/ RFX consensus sequences (Table II) with HindIII overhangs were synthesized (Oligo Etc. and Integrated DNA Technology) as complementary strands, annealed to make double stranded oligonucleotides and radiolabeled using the [␣-32 P]dATP and the Klenow fragment to fill in the HindIII overhang. For the DNA mobility shift assay, the binding reaction was performed for 30 min at room temperature in 20 l of binding buffer containing 90,000 -100,000 cpm/200 fmol of labeled probe, 1 g of poly(dI-dC)⅐poly(dI-dC), and nuclear extract containing 4.5-5.0 g of protein. Double-stranded annealed complementary oligonucleotides (Oligo Etc. and Integrated DNA Technology) were used as competitors (Tables I and II). Separation of free radiolabeled DNA from DNA-protein complexes was carried out on a 4 -5% nondenaturing polyacrylamide gel with a standard Tris-borate electrophoresis buffer at 300 V in the cold (4°C). Autoradiography was performed by overnight exposure to Kodak Biomax film (Eastman Kodak Co.). The intensities of the bands were quantified using Instant Imager (Packard Instrument Co.). In the antibody experiment, the nuclear extract and antibodies were preincubated for 20 min at room temperature before the radiolabeled probe was added, followed by another 20 min incubation with the probe. The antibody (kindly supplied by Dr. W. Reith to Dr. M. Ehrlich) is a polyclonal antibody to recombinant RFX1 (31), and its specificity for other family members has been described (16).
In Vitro Transcription and Translation-The RFX1 cDNA in the sense orientation in the pBK-RSV vector (Stratagene) was transcribed and translated in vitro using a rabbit reticulocyte lysate (Promega; TNT translation kit) following the manufacturer's protocol. The in vitro translated proteins were used in electrophoretic mobility gel shift assay.
In Vitro Mutagenesis-Mutation at the ϩ7 and at ϩ23 sites (C to T) in the collagen ␣2(I) gene (all positions given relative to the transcription start site) were performed by site-directed mutagenesis based upon Kunkel's method (Muta-Gene phagemid mutagenesis kit, Bio-Rad) following the manufacturer's protocol. The mutated constructs were then cloned into pH 20 (Ϫ220 to ϩ54 of the collagen ␣2(I) fused to the luciferase coding sequence) at SmaI-HindIII sites. The DNA sequence of the mutated constructs were confirmed by DNA sequencing (U. S. Biochemical Corp.) prior to their use in transfection and in vitro transcription assays.
Transient Transfection and Luciferase Assays-Plasmid DNA was transfected by lipofection (LipofectAMINE, Life Technologies, Inc.) into rat fibroblasts 24 h after plating cells. Plasmids containing the wild type or mutated bp in Ϫ220 to ϩ54 of collagen ␣2(I) promoter/5Ј region driving expression of the luciferase coding sequence were co-transfected with a reference plasmid, pCMV-green fluorescent protein (CLON-TECH) containing the CMV immediate early promoter driving expression of the gene encoding the green fluorescent protein. CMV-green fluorescent protein was used to normalize the transfection efficiency.
Luciferase assays were performed under standard conditions (Luciferase kit; Promega Corp.). Briefly, the cells were washed twice with phosphate-buffered saline buffer and scraped with lysis reagent. The cells and solution were centrifuged at 12,000 ϫ g to pellet the debris. The cell extract was mixed with the luciferase assay reagent, and light emission was measured in a scintillation counter. The luciferase activity was assayed in duplicate within the linear range of the instrument. Ten readings at 60-s intervals were averaged in each assay. Values were normalized to fluorescence of the green fluorescent protein.
In Vitro Transcription Assay-The reaction mixture for in vitro transcription contained 50 -90 g of nuclear extract, 1 g of super-coiled template DNA (purified on a CsCl gradient), 20 mM HEPES, pH 7.9, 4 mM MgCl 2 , 60 mM KCl, 2 mM EDTA, 0.5 mM dithiothreitol, 12% glycerol, 600 M of each of rNTP in a final volume of 25 l. Reaction was carried out at 30°C for 1 h and terminated by the addition of 175 l of stop solution, which contained 0.3 M sodium acetate, 0.5% SDS, 3 g/ml tRNA, pH 5.2. After extraction of protein with phenol/chloroform, the RNA was precipitated by ethanol. To detect the newly synthesized transcript, antisense oligonucleotide primer corresponding to a sequence in the luciferase gene was generated ( Table I). The primer was end labeled with polynucleotide kinase and [␥-32 P]ATP, hybridized to in vitro transcription products and extended using Moloney murine leukemia virus reverse transcriptase. The primer-extended products were analyzed by 5% polyacrylamide gel electrophoresis containing 7 M urea. Transcription reaction and primer extension reactions always included an RNase inhibitor (RNase inhibitor protein, cloned human pancreatic RNase A lytic enzyme inhibitor, Ambion, Inc.). Gels were dried and autoradiographed at Ϫ80°C with an intensifying screen.

RESULTS
In our earlier study (7), we demonstrated that the Ϫ25 to ϩ30 sequence of ␣2(I) promoter could bind sequence-specific nuclear proteins preferentially when CpGs were methylated. This sequence contains two CpG sites, at ϩ7 and ϩ23, respectively, relative to the transcription start site. In order to investigate which sites are important for the methylation responsiveness, a gel shift experiment was performed using the wild type and mutated probes (both unmethylated and methylated) with rat fibroblast nuclear extracts (Fig. 1). The DNA fragment with a CpG to TpG transition at position ϩ23 specifically complexed with proteins in the extract in a similar manner as the wild type DNA fragment. There was a 3-fold increase in FIG. 1. Methylation and mutation at the ؉7 CpG site increases the protein-DNA complex formation on the ␣2(I) initiator probe as judged by electrophoretic mobility shift assay. Duplex oligonucleotide sequences corresponding to positions Ϫ25 to ϩ30 of the ␣2(I) gene containing C to T mutations at ϩ7, ϩ23, or both were incubated with Sss I methylase with (M) and without (U) S-adenosylmethionine as described under "Materials and Methods." Nuclear extracts (5 g of protein) were incubated with wild type or mutated radiolabeled probes (specific activity of all probes, 90,000 cpm/200 fmol) and subjected to electromobility shift assay. This x-ray is representative of five experiments with three different extracts.
binding when the CpG at position ϩ7 was methylated whether or not the ϩ23 site was mutated. This result suggests that the ϩ23 CpG site is not important for the increased binding upon cytosine methylation. On the other hand, there was no difference in binding between unmethylated and methylated DNA fragments when the ligand had a CpG to TpG mutation at the ϩ7 position, and only the ϩ23 CpG was differentially methyl-ated. This indicates that only the ϩ7 site is important for increasing protein binding on methylated constructs. In addition, the amount of complex formation with the ϩ7 mutated DNA fragment was approximately 3 times more than with the analogous wild type unmethylated fragment.
Ten base pairs out of 14 in the CpG methylated sequence from position Ϫ1 to ϩ13 of the collagen ␣2(I) 5Ј region match the consensus sequence for the transcription regulatory protein MDBP/RFX ( Fig. 2A). Furthermore, the CpG of this sequence is in the same position as the CpG of several previously described methylation-dependent MDBP/RFX sites (19). Lastly, CpG to TpG transitions in the methylation-dependent sites have been shown to increase binding by MDBP/RFX in a similar fashion to cytosine methylation at these sites. Therefore, we suspected that the collagen ␣2(I) sequence in this region is an MDBP/RFX site.
Four short oligonucleotides from different parts of the 55-bp ␣2(I) initiator region DNA fragment (position Ϫ25 to ϩ30) initially were used as ligands in electrophoretic mobility shift assays to determine whether there was methylation sensitive binding of a nuclear protein to these shorter probes. Nuclear extracts from rat skin fibroblasts were used in these assays with oligonucleotide duplexes containing ␣2(I) sequences Ϫ20 to ϩ1, Ϫ14 to ϩ6, ϩ10 to ϩ30, or Ϫ1 to ϩ20 bordered by a HindIII site (AAGCTT) added to ends of both strands (see Table I). The sequence in the beginning of the first exon of the ␣2(I) gene containing the region homologous to MDBP/RFX was the only sequence that resulted in a methylation responsive formation of a specific complex when used as a radiolabeled ligand (data not shown). Two specific, slowly migrating DNA-protein complexes formed with the Ϫ1 to ϩ20 oligonucleotide that migrated only slightly faster (commensurate with the small size of the ligand) than the methylation-responsive complexes formed from the longer Ϫ25 to ϩ30 oligonucleotide. Furthermore, just as methylation increased the amount of this specific complex formation from the Ϫ25 to ϩ30 oligonucleotide, methylation increased binding by the smaller Ϫ1 to ϩ20 oligonucleotide 2.5-3-fold (Fig. 2B, lanes 1 and 2). This complex formation was specific, as it is competed by an excess of the identical unlabeled sequence (Fig. 2, lanes 3 and 7) but not a similar excess of other tested ␣2(I) sequences (Ϫ14 to ϩ6, ϩ10 to ϩ30, and Ϫ20 to ϩ1) (Fig. 2B, lanes 4 -6 and 8). Competition for binding to methylated sequence (Ϫ1 to ϩ20) probe increased when the specific competitor was methylated (Fig. 2B,  lanes 3 and 7), whereas methylation of the ϩ10 to 30 competitor did not visibly change the amount of labeled complex formation (Fig. 2B, lanes 6 and 8).
The ability of various other oligonucleotide duplexes to com- Methylation-dependent sequence-specific binding activity is located at the transcription start site of ␣2(I) collagen gene (؊1 to ؉13). A, sequence homology between consensus sequence for MDBP/RFX and ␣2(I) collagen at the transcription start site (Ϫ1 to ϩ13). R, purine; m, methylated C or T; Y, pyrimidine. B, methylationsensitive and sequence-specific binding to the ␣2(I) collagen transcription start site in electrophoretic mobility shift assay. Unmethylated (U) and methylated (M) probes (␣2(I) Ϫ1 to ϩ20 with HindIII ends; see Table I) are compared in lanes 1 and 2. Methylation status of the probe and the competitor is indicated at the top. Competitors at 50-fold molar excess over the labeled ligand (200 fmol of labeled ligand) were incubated with nuclear proteins, and then radioactive probes were added. Different regions of the ␣2(I) collagen gene were used as competitors as follows: lane 3 and 7, Ϫ1 to ϩ20; lane 4, Ϫ14 to ϩ6; lane 5, Ϫ20 to ϩ1; lanes 6 and 8, ϩ10 to ϩ30 (see Table I for sequence information). Competitor sequences are methylated in lanes 7 and 8. The assay conditions were the same as in Fig. 1. The arrows indicate the protein-DNA complexes generated by MDBP/RFX and ␣2(I) sequence. TTCATAGCCTTATGCAGTTGCTCTCCAGCG pete for complex formation with the methylated ␣2(I) probe was also tested (Fig. 3A). Sequences unrelated to MDBP/RFX sites, namely mTAE (Fig. 3A, lane 9; Table I) and a sequence in the ␣1(I) gene (Fig. 3A, lane 8; Table I) did not compete for complex formation. In contrast, the known binding sites for MDBP/RFX sequences (Table II) competed at 50-fold molar excess over the labeled ligand (Fig. 3A, lanes 2-7). These include EP (a high affinity, cytosine methylation-independent MDBP/RFX binding site present in hepatitis B viral enhancer (29)) Py1 (a similar binding site in polyomavirus enhancer B), and pB1 (an in vitro methylated cytosine methylation-depend-entsiteintheplasmidpBR322 (19)).Inaddition,twomethylationdependent human MDBP/RFX sites, the human ␣-galactosidase A site, close to the transcription start site (␣-GalA, ϩ49), and the human apoferritin H (hFer ϩ202) site, as well as a cytosine methylation-independent site known as X-box in the MHC class II gene promoters, competed effectively for binding to the ␣2(I) site (29). The complexes that formed on all of these MDBP/RFX sites co-migrated with the ␣2(I) promoter-protein complexes with the same rat fibroblast nuclear extract (Fig. 3B,  lanes 3-7). In addition, the collagen sequence was as methylation sensitive under our conditions as pB1 and was more sensitive than ␣-GalA or hFer sequences (not shown). Use of antiserum specific for the RFX1 polypeptide of MDBP/ RFX complex also indicated that the methylation-responsive protein recognizing the Ϫ1 to ϩ20 sequence of the ␣2(I) gene is MDBP/RFX. Preincubating the reaction mixture for the electrophoretic mobility shift assay with this antibody resulted in much slower migration of the protein-DNA complex formed between the nuclear protein and the ␣2(I) sequence (Fig. 4A,  lanes 1-3). The same supershift was obtained when the known MDBP/RFX-specific EP DNA sequence was used as the ligand (Fig. 4A, lanes 4 -6). Presumably, the large size of the MDBP/ RFX dimer complexed with the antibody, as well as with the DNA ligand, was responsible for these supershifted ␣2(I) or EP ligands not entering the gel.
Further evidence for the identity of the protein binding in a sequence-specific and methylation-specific manner to the methylated Ϫ1 to ϩ20 sequence from the ␣2(I) gene was provided in electrophoretic mobility shift assays with in vitro translated MDBP/RFX1 protein. This protein product formed a complex with the methylated Ϫ1 to ϩ20 sequence that comigrated with a complex formed between the same ligand and rat fibroblast nuclear extracts (Fig. 4B, lanes 2 and 8). Because MDBP/RFX from nuclear extracts contains different related homodimers or heterodimers of related polypeptide chains and the RFX1 homodimer is the largest of these, it is not surprising that the main complex seen from the in vitro translation product corresponds to the slower moving of the two bands observed in binding reactions with the nuclear extract. We also tested the ability of RFX1 antiserum to supershift complex formed using the in vitro transcribed/translated MDBP/RFX1protein (Fig. 4B, lane 3). The same supershift was seen as for the nuclear protein Fig. 4A, lanes 3 and 6). Rabbit control antiserum did not bind to the protein-DNA complex (Fig. 4B,  lane 4).
To test for control of gene expression at this sequence in the beginning of the ␣2(I) gene, C to T mutations at ϩ7 and/or ϩ23 sites were introduced by site-directed mutagenesis into the ␣2(I) promoter-luciferase construct, pH 20, containing the Ϫ220 to ϩ54 region of the promoter. These expression plasmids were transfected into rat fibroblast cells. The ϩ7 mutation inhibited transcription to a greater extent (80 Ϯ 2.6% (S.D.)) than did the ϩ23 mutation (47 Ϯ 3.8%) (Fig. 5A). The same constructs were used in an in vitro transcription assay. The ϩ7 mutation decreased the transcription by 56% but the ϩ23 mutation decreased only 23% (Fig. 5B) in a representative experiment repeated three times. Therefore, a CpG-to-TpG mutation at the ϩ7 site, which increases the formation of specific protein complexes in vitro, also decreases transcription in vivo and in reaction mixtures containing a nuclear extract. This mutation had more of an effect on transcription than did an analogous mutation at position ϩ23, which does not match the MDBP/ RFX consensus sequences.  (Table II). The methylation-sensitive competitor sites (␣GalA, hFer, and pB1) were methylated (lanes 5-7). Competitor sequences unrelated to MDBP/RFX binding sites are ␣1(I) (lane 8) and mTAE (lane 9) (Table I). Lane 1 contains no competitors. The arrows indicate the protein-DNA complexes generated by MDBP/RFX and the ␣2(I) methylated sequence. B, nuclear extracts were incubated with the same five known MDBP/RFX ligands in addition to the ␣2(I) sequence from position Ϫ1 to ϩ20. All probes were labeled at the same specific activity (100,000 cpm/200 fmol). MDBP/RFX sites used as probes are shown in lanes 3-7 (lane 3, EP from hepatitis B virus enhancer 1; lane 4, X box from MHC type II; lane 5, ␣-GalA from the ␣-galactosidase A gene; lane 6, pBsite1 from pBR322; lane 7, hFer from the apoferritin gene) (Table II). Methylation status of probes is indicated at the top.
The arrows indicate the protein-DNA complexes generated by MDBP/ RFX and the ␣2(I) methylated sequence.

DISCUSSION
One of the unusual characteristics of MDBP/RFX proteins is their behavior with respect to CpG-containing recognition sites. These sequence-specific DNA-binding proteins show increased binding to these sites in their CpG-methylated form. The sequence specificity of the binding site distinguishes these proteins from methylated DNA-binding proteins, such as MeCP1, MeCP2, DBP-m, and MDBP-2-H1 (22,(32)(33)(34)(35). We demonstrated that the MDBP/RFX could bind to a sequence at the very beginning of the ␣2(I) gene first exon in the rodent and human genomes in a methylation-dependent manner. Similar methylation-responsive MDBP sites are present at the beginning of the human genes for hypoxanthine phosphoribosyl transferase; HLA-A2, -A3, -A25 antigens; and ␣-galactosidase A (19).
MDBP/RFX also binds to many of its sites in a cytosine methylation-independent manner if 1-3 of the CpGs in the consensus sequence are replaced by TpG or TpA. There is a CpG site 7 base pairs downstream from the transcription start site of the ␣2(I) gene that is within the MDBP/RFX site. This CpG matches the central CpG of the MDBP/RFX consensus sequence ( Fig. 2A), which confers methylation-dependent binding on other methylation-responsive binding sites for this family of closely related proteins (19). The location of this MDBP/ RFX site in the ␣2(I) gene was defined by homology to many known MDBP/RFX sites, electrophoretic mobility shift assays with rat fibroblast nuclear extracts and with the in vitro transcription/translation product from an MDBP/RFX1 template, and supershift assays using specific antibody to MDBP/RFX1 (Figs. 2-4). When we replaced this unmethylated CpG within the MDBP/RFX site with a TpG, there was increased specific complex formation comparable to the increase seen upon methylation of this CpG dinucleotide pair (Fig. 1), just as has been observed with other methylation-dependent MDBP/RFX sites (19). When the next downstream CpG in this region of the ␣2(I) gene, namely, the CpG at position ϩ23, was converted to a TpG, there was no effect on complex formation with MDBP/ RFX proteins in a nuclear extract, as expected, because this dinucleotide is outside the MDBP/RFX site.
A reporter containing the ␣2(I) promoter/5Ј region gene driving expression of a luciferase gene was equivalently mutated at the CpG at position ϩ7. The observed decrease in promoter activity in the TpG mutant in both transfection experiments in cultured rat fibroblast and in vitro transcription with fibroblast extracts may be due to increased binding of MDBP/RFX to the mutated sequence (Fig. 5). Although this mutation may also affect formation of the preinitiation complex for transcription in a manner independent of MDBP/RFX binding, a similar mutation in a more downstream methylation-dependent MDBP/RFX site (48 bp downstream of the major transcription start site) in the human ␣-galactosidase A also decreases gene expression in transient transfection assays. Furthermore, when that MDBP/RFX site was mutated to a yeast GAL4 binding site and yeast GAL4 DNA-binding domain chimeras with mammalian transcription factors were present, a hybrid GAL4 DNA-binding domain-MDBP/RFX down-regulated reporter gene expression, whereas the intact GAL4 transcription factor and a hybrid GAL4 DNA-binding domain-VP16 activation domain up-regulated expression of the reporter gene. 2 All of the homo-and heterodimeric members of the MDBP/ RFX1-3 family of DNA-binding protein exhibit methylation-dependent binding to certain of their cognate sites (16,17). In contrast, RFX5 is a more distant member of this family, which is involved in positive regulation of major histocompatibility type II genes and does not display methylation-dependent binding (17). Yeast proteins involved in cell cycle control and DNA damage are also present in this family (36,37). MDBP/ RFX family members have very similar DNA binding domains but different N-terminal regions. RFX2-4 show appreciable tissue-specificity in their distribution, whereas RFX1 is present at similar levels in a variety of examined tissues (16). RFX1 homodimer, a large protein with 979-amino acids per subunit, contains both transcription repression and activation domains, the ability of which to positively or negatively modulate transcription may vary depending upon the location of its cognate sites in a given gene region and the other proteins with which it interacts (28,29). Positive regulation of transcription by MDBP/RFX has been demonstrated for the methylationindependent binding site in the EP sequence of the hepatitis B virus enhancer (38). In contrast, one of us previously demonstrated that a low affinity methylation-independent binding site beginning at position ϩ5 of the CMV IE transcription unit can down-regulate transcription when its binding by MDBP/ RFX is increased by mutation (39).
Dual function transcription factors have been described that switch their function by interaction with different co-activators or co-repressors (40 -42), different neighboring transacting factors (43)(44)(45), or interaction with specific DNA sequences in different locations relative to the transcription start site (46). MDBP/RFX interacts with c-Abl, greatly stimulating its autophosphorylation (47). We have preliminary data using antibody supershifting experiment suggesting that c-Abl protein is present in the complex formed with the collagen sequence. 3 MDBP/ RFX can also interact with at least one TAFII factor. 4 Furthermore, MDBP/RFX binding sites implicated in transcription control are sometimes located in the proximal promoter region (48) or intron sequences (49), as well as in DNA sequences immediately downstream of or spanning the transcription start site, as in the ␣2(I) sequence described in this report. In the proximal promoter of proliferating cell nuclear antigen (PCNA), there is an MBDP/RFX binding site that overlaps with a cAMP response element-binding site (48,50,51). Using mutation analysis in and surrounding the RFX binding site, it was demonstrated that binding of RFX protein inversely correlates with transcription activation of the promoter. Therefore, MDBP/RFX plays an inhibitory role along with the retinoblastoma-related tumor suppressor protein, p107, in inhibiting activation of the PCNA promoter activity. RFX also represses gene expression of c-Myc at a site in the first intron (26,52). The data presented here indicate that MDBP/RFX behaves as transcriptional repressor in the context of methylated collagen ␣2(I) sequence. We have previously demonstrated that in a rat cell line that has lost expression of this gene, there is methylation in the promoter region and that treatment of this cell line with the DNA demethylating agent 5-azacytidine results in the gain of expression of this gene (43). Because abnormal hypermethylation of the 5Ј region of tumor suppressor genes is so frequent in cancer and down-regulation of collagen gene expression has been proposed to contribute to carcinogenesis, abnormal methylation of the 5Ј region of the ␣2(I) gene, including the MDBP/RFX site, might occur in cancers and play a role in down-regulation of the expression of this gene. Future genomic sequencing experiments will allow testing of this hypothesis and of the possibility that the RFX/MDBP is a transcriptional repressor involved in tissue-specific regulation of the collagen ␣2(I) gene transcription in response to naturally occurring differential methylation.  (lanes 1-3) or EP (lanes 4 -6) radiolabeled probes. Specific MDBP/RFX1 antibody was added at a 1:100 dilution (lanes 3 and 6), and nonimmune (NI) serum was added at a 1:100 dilution (lanes 2 and 5). The upper arrow indicates the supershifted protein-DNA complexes, and the lower arrow indicates the protein-DNA complexes generated by MDBP/RFX and the methylated ␣2(I) sequence. B, RFX1 cDNA was in vitro transcribed and translated and the product used for mobility shift assays. Specific RFX1 antibody (lane 3 and 6) or nonimmune (NI) serum (lanes 4 and 7) at a 1:100 dilution was added to the reaction mixture containing the in vitro translated protein (lanes 2-4) or, as a control, just the reticulocyte lysate (lanes 5-7) for 20 min before the addition of radiolabeled ␣2(I) ligand. Lane 8 contained nuclear extracts instead of the in vitro translation product incubated with the ligand. Lane 1 contained probe without protein.
The arrows indicate the protein-DNA complexes generated by MDBP/RFX and the ␣2(I) methylated sequence. NS, nonspecific binding from the reticulocyte lysate.  in vitro transcription assays (B). A, rat skin fibroblasts were transfected by the LipofectAMINE method with collagen promoter-luciferase constructs (2 g) containing C to T mutations at ϩ7 or ϩ23. A CMV promoter driving the green fluorescent protein gene (1 g) was co-transfected to normalize the transfection efficiency. The graph represents the average luciferase activity/g of protein with standard error from four different experiments performed in duplicate. B, the control and mutated ␣2(I) collagen promoter-luciferase constructs (1 g) containing bases Ϫ220 to ϩ54 of the collagen gene were transcribed in vitro with rat fibroblast nuclear extract. The RNA transcript was detected by primer extension as described under "Materials and Methods." This is a representative of an experiment performed three times.