ABSTRACT
We performed in vivo dimethylsulfate footprinting of the 220 bp mouse proximal pro[alpha]1(I) collagen promoter and the 350 bp mouse proximal pro[alpha]2(I) collagen promoter in BALB/3T3 fibroblasts, primary mouse skin fibroblasts, S-194 B cells, NMuLi liver epithelial cells and RAG renal adenocarcinoma cells and in vitro DNase I footprinting of these promoters using nuclear extracts of these different cell types. Whereas pro[alpha]1(I) and pro[alpha]2(I) collagen RNAs were present in BALB/3T3 fibroblasts and primary fibroblasts, these RNAs could not be detected in the three other cell lines. Comparison of in vitro DNase Ifootprints for each of the two proximal collagen promoters indicated that the patterns of protection were very similar with the different nuclear extracts, suggesting that the DNA binding proteins binding to these promoters were present in all cell types tested. In contrast, in vivo footprints over these proximal promoters were cell-specific, occurring only in fibroblast cells and not in the other three cell types. The in vivo footprints were generally located within the in vitro footprinted regions. Our results suggest that although all cell types tested contained nuclear proteins that can bind to the proximal pro[alpha]1(I) and pro[alpha]2(I) collagen promoters in vitro, it is only in fibroblasts that these proteins bind to their cognate sites in vivo. We discuss possible regulatory mechanisms in type I collagen genes that can contribute to the cell-specific in vivo protein-DNA interactions at the proximal promoters.
Type I collagen, the most abundant among all collagen types, is expressed specifically in fibroblasts, osteoblasts, odontoblasts and some other mesenchyme-derived cells (1 ,2 ). It is a heterotrimer composed of two [alpha]1(I) chains and one [alpha]2(I) chain, each encoded by a single gene (3 ). The two genes are coordinately expressed in a 2:1 stoichiometry (4 ). Changes in the level of synthesis of type I collagen occur in a number of physiological, pathological and experimental conditions, including the appearance of type I collagen and modulation of its synthesis in specific cell types during embryonic development (1 ,5 ), its increased synthesis during wound healing and in fibrotic diseases (1 ,6 ,7 ), as well as the control of its synthesis by cytokines such as tumor growth factor-[beta] (TGF[beta]) and interleukin-1 (IL1) in tissue culture conditions (8 -10 ). The principal modulating mechanism of pro[alpha]1(I) and pro[alpha]2(I) collagen synthesis occurs at the level of transcription (11 ,12 ), but the precise mechanisms involved during development and in adult tissues are still poorly understood.
Previous studies have identified several functional cis-acting elements in the proximal promoter of type I collagen genes. In the pro[alpha]2(I) collagen gene, these include a binding site for the ubiquitous heterotrimeric CCAAT-binding factor (CBF), from nt -73 to -100 (13 ) and redundant GC-rich binding sites for several proteins from nt -65 to -105, -114 to -131 and -152 to -176 (14 ). Several classes of proteins that are mainly ubiquitous proteins bind to these redundant sites. These include SP1, proteins different from SP1 that bind to an SP1 consensus binding site, proteins that bind to a Krox consensus binding site and probably others. Transient expression and in vitro transcription experiments with wild-type and mutant templates showed that the segment from nt -40 to -170 was essential for promoter activation (14 ). Other studies identified three short cis-acting GC-rich elements in the human pro[alpha]2(I) collagen gene from nt -264 to -323 that were capable of binding SP1 (15 ). In the mouse promoter a binding site for CTF/NF1 from nt -290 to -305 is also present (16 ).
The sequence of the proximal promoter segment between nt -220 and the TATA box in the pro[alpha]1(I) gene presents homologies with the equivalent segment in the pro[alpha]2(I) gene. In the pro[alpha]1(I) promoter, the CBF binding site from nt -86 to -113 is surrounded by two GC-rich repeats that can bind SP1 and presumably other GC-rich binding proteins (17 ). The promoter also contains two apparently redundant sites from nt -190 to -170 and -160 to -133, which can bind a protein designated c-Krox that also binds to the pro[alpha]2(I) gene from nt -165 to -155 (18 ,19 ). Another DNA binding protein which binds to the pro[alpha]2(I) promoter from nt -180 to -152, also binds to the pro[alpha]1(I) promoter from nt -168 to -129 and nt -194 to -168 (20 ). Previous studies have also shown that a short promoter fragment containing sequences between nt -222 and +116 has strong transcriptional activity when tested either in an in vitro transcription assay (13 ) or in transient transfection experiments (17 ).
In addition to the proximal promoter elements, the type I collagen genes also contain upstream tissue-specific enhancers and enhancer elements in the first intron (21 -26 ). One can postulate that, to activate transcription in specific cell types, proteins binding to these enhancers cooperate with transcription factors that bind to the proximal promoters.
In this study, we asked whether the proximal promoters of the type I collagen genes would exhibit any cell type-specific protein-DNA interactions. We first performed in vitro DNase I footprinting over the 220 bp proximal pro[alpha]1(I) and the 350 bp proximal pro[alpha]2(I) promoters with nuclear extracts of fibroblasts, which express the type I collagen genes and with nuclear extracts of B cells, NMuLi liver cells and RAG renal adenocarcinoma cells, which do not express these genes. With all extracts tested, the patterns of these in vitro footprints were similar. We then asked whether in vivo occupancy of sites in the proximal promoters that are occupied in vitro occurs both in cells that express and that do not express the type I collagen genes. In contrast to the in vitro results, in vivo occupancy was seen only in cells in which the type I collagen genes are expressed.
BALB/3T3 murine fibroblasts, S-194 murine myeloma cells, NMuLi murine liver epithelial cells and RAG murine renal adenocarcinoma cells were obtained from American Type Culture Collection (ATCC) and cultured in Dulbecco's modified essential medium (DMEM) containing 10% fetal bovine serum and the antibiotics penicillin (50 U/ml) and streptomycin (50 [mu]g/ml) (all from Life Technologies, Inc.). Primary skin fibroblasts were isolated from newborn mice after sacrifice: the skin was digested with collagenase D (Boehringer, catalogue no. 1088874) in DMEM (3 mg/ml) for 3 h at 37oC, the cells dissociated from residual tissue pieces by pipetting and the dissociated cells collected. The collected cells were washed twice with phosphate-buffered saline (PBS) and then once with the culture medium (DMEM plus 10% fetal calf serum plus 50 U/ml penicillin plus 50 [mu]g/ml streptomycin) before plating. Cells at 80% confluence were used for all purposes, including preparation of total RNA, nuclear extracts, genomic DNA and in vivo methylation by dimethylsulfate (DMS) for in vivo DMS footprinting.
Total RNA was prepared from 80% confluent cells following standard guanidinium thiocyanate extraction procedures (27 ). The integrity of RNA was verified by ethidium bromide staining of a formaldehyde-agarose gel. RNase protection assay was done using the `Ambion Ribonuclease Protection Assay RPA II kit' (Ambion, catalogue no. 1410). Antisense RNA probes were transcribed and labeled with [32P]uridine triphosphate (UTP) and T7 RNA polymerase using the in vitro transcription kit MAXIscript (Ambion, catalogue no. 1310-1326). The pro[alpha]1(I) collagen RNA contained the sequence from nt -222 to +124, the pro[alpha]2(I) collagen RNA probe contained the sequence from nt -110 to +132. Total RNA (10 [mu]g) from different cells was hybridized with 8 * 104 c.p.m. of each of the probes separately or together with the control [beta]-actin probe, an RNase A/RNase T1 mixture was used to digest single-stranded RNA and the products were run on a 5% polyacrylamide, 8 M urea gel and subjected to autoradiography. The yeast RNA was used as negative control RNA sample.
For footprinting, nuclear extracts were prepared according to standard procedures (28 ) and in vitro DNase I footprinting was done as described (29 ). For the pro[alpha]1(I) collagen promoter footprinting, the EcoRI-HindIII fragment of the plasmid pG60 (17 ), containing nt -238 to +113 of the mouse pro[alpha]1(I) collagen gene, was labeled at the EcoRI end by polynucleotide kinase, for footprinting the upper DNA strand and by Klenow enzyme for footprinting the lower DNA strand. For the pro[alpha]2(I) collagen promoter footprinting, the NarI-BamHI fragment of the plasmid pLAG23, containing nt -350 to +7 of the mouse pro[alpha]2(I) collagen gene, was labeled at the NarI end by Klenow enzyme for footprinting the upper strand of DNA and by polynucleotide kinase for footprinting the lower strand of DNA. The binding reaction mixtures were made up in 50 [mu]l volumes using 15 000 c.p.m. of the end-labeled probes (~10 fmol) and 20 [mu]g of nuclear extracts from different cell types. The binding buffer was 15 mM Tris-HCl pH 7.5, 75 mM NaCl, 0.1 mM EDTA pH 8.0, 0.5 mM DDT, 5% glycerol and 2 [mu]g poly(dI.dC).poly(dI.dC). After the binding reaction mixture was incubated for 20 min at room temperature, 2 [mu]l DNase I (Worthington DPRF; stored as 1 mg/ml in 50% glycerol) diluted 85 times in cold DNase I dilution buffer (10 mM Tris-HCl pH 7.5, 50 mM NaCl, 10 mM CaCl2, 62.5 mM MgCl2, 1 mg/ml bovine serum albumin) was added to it and the reaction mixture was mixed and incubated for 1.5 min at room temperature. The reaction was then terminated by addition of 10 [mu]l of stop buffer (5% SDS, 125 mM EDTA). Then 20 [mu]g of yeast tRNA was added and phenol/chloroform extraction and ethanol precipitation performed. The DNA pellet was dried, resuspended in formamide loading buffer (80% formamide, 10 mM NaOH, 0.25 mg/ml bromophenol blue and 0.25 mg/ml xylene cyanol FF), heated to 90oC for 3 min and loaded on a denaturing 8% polyacrylamide, 8 M urea sequencing gel. One to two microliters of a 1:1500 dilution of a 1 mg/ml DNase I solution was used in controls lacking nuclear extracts. The end-labeled DNA fragments were also subjected to the G + A sequencing reaction of Maxam and Gilbert (30 ) and loaded on the gel for identification of protected sequences.
In vivo DMS footprinting of 80% confluent cultured cells was performed by ligation-mediated polymerase chain reaction (LM- PCR) (31 ), using Vent polymerase for both first-strand extension and amplification. Briefly, the intact cells were treated with 0.1% DMS (Aldrich Chemical Company, Inc., catalogue no. 18,630-9) for 1.5 min at room temperature. DNA was then isolated from these cells and treated with 1 M piperidine (Aldrich, catalogue no. 10,409-4) at 90oC for 30 min to cleave the DNA at the DMS-methylated G residues. Naked genomic DNA control was prepared from BALB/3T3 fibroblasts and then subjected to in vitro DMS treatment and subsequent piperidine cleavage of methylated G residues (30 ). Samples of in vivo or in vitro DMS-treated DNA (3 [mu]g) were hybridized with type I procollagen proximal promoter-specific primer sets (see primers) and then amplified by LM-PCR. The final products were analyzed on 6% polyacrylamide, 8 M urea gels. All in vivo footprints were highly reproducible: the complete experiments were repeated at least three times with similar results. Results of visual examination of the in vivo footprints were also confirmed quantitatively using Bio Image Intelligent Quantifier.
Primers were synthesized in an Applied Biosystem 391 DNA synthesizer and purified on a 10% polyacrylamide/8 M urea gel. The sequences of the different primers are presented in Table 1 .
We assayed five different cell types, BALB/3T3 mouse fibroblasts, primary mouse fibroblasts, S-194 mouse B cells, NMuLi mouse liver epithelial cells and RAG mouse renal adenocarcinoma cells, for type I collagen RNAs. Levels of pro[alpha]1(I) collagen mRNA and pro[alpha]2(I) collagen mRNA present in total RNA of these different cell lines were determined by RNase protection assay. As Figure 1 shows, 124 nt of the pro[alpha]1(I) collagen RNA probe (Fig. 1 A) and 132 nt of the pro[alpha]2(I) collagen RNA probe (Fig. 1 B), were protected by RNAs from BALB/3T3 fibroblasts and primary mouse fibroblasts (Fig. 1 A and B, lanes 6-8), but not by RNAs from S-194 B cells, NMuLi liver epithelial cells or RAG renal adenocarcinoma cells (Fig. 1 A and B, lanes 9-11). The lengths of the protected probes are consistent with the location of the transcription initiation sites in each of the two corresponding genes. In contrast, a control [beta]-actin probe was protected in all cell lines tested. These experiments indicated that the pro[alpha]1(I) collagen mRNA and the pro[alpha]2(I) collagen mRNA were present in BALB/3T3 fibroblasts and mouse primary fibroblasts, but not in S-194 B cells, NMuLi cells and RAG cells. Based on these experiments, we used all five cell types for in vitro DNase I and in vivo DMS footprinting experiments: BALB/3T3 and primary mouse fibroblasts as cells expressing both type I collagen genes, and S-194, NMuli and RAG cells as cells not expressing these genes.
In previous in vitro DNase I footprint experiments using nuclear extracts of BALB/3T3 fibroblasts, we identified several segments that were protected from DNase I digestion within the proximal 350 bp promoter of the mouse pro[alpha]2(I) collagen gene (14 ). For comparison, these footprints are shown in Figure 2 A and B, lane 3. The protected areas included a segment from nt -73 to -100 to which the heterotrimeric CCAAT binding factor CBF bound. In this region of the promoter, the area of DNA protected by purified CBF was identical to that seen with crude nuclear extracts (13 ). Two other protected segments extended from nt -114 to -131 and from nt -152 to -176. These two segments are GC-rich and each was shown to bind several classes of DNA binding proteins present in nuclear extracts of fibroblastic cells (14 ). There are two additional protected segments adjacent to each other and located further upstream, one from nt -258 to -285 and the other from nt -290 to -304, but protection in these segments was weaker than for the three proximal footprints. Using the proximal 350 bp promoter of the pro[alpha]2(I) collagen gene as template, we asked whether nuclear extracts of those cells in which no type I collagen RNA was present generated footprints different from footprints obtained with nuclear extracts of fibroblast cells. The results, presented in Figure 2 , show that, for both the upper strand of DNA (Fig. 2 A) and the lower strand of DNA (Fig. 2 B), the overall patterns of DNase I protection were similar with nuclear extracts from all cell types tested, BALB/3T3 fibroblasts, primary mouse fibroblasts, S-194 B cells, NMuli liver epithelial cells and RAG renal adenocarcinoma cells (Fig. 2 A and B, lanes 3-7). The segments that were protected with nuclear extracts of the fibroblast cells, nt -73 to -100, -114 to -131, -152 to -176, -258 to -285 and -290 to -304, were also protected with nuclear extracts of the other cells. The small differences in the protected areas between lanes in each panel of Figure 2 were not reproducible with different extract preparations of the same cells, presumably due to differences in individual preparations. These results suggested that nuclear proteins capable of binding in vitro to discrete sites in the 350 bp pro[alpha]2(I) collagen promoter were present in nuclei of cells that express type I collagen and those that do not.
The pro[alpha]1(I) collagen gene is coordinately expressed with the pro[alpha]2(I) collagen gene. Previous in vitro DNase I footprinting of the proximal 220 bp murine pro[alpha]1(I) collagen promoter showed protection of a large segment from nt -80 to -190 that could be subdivided into three subareas, from nt -80 to -130, from nt -133 to -160 and from nt -170 to -190 (17 ). The nt -80 to -130 segment contains the CBF binding site that is surrounded by two 12 bp identical GC-rich repeats. As for the proximal pro[alpha]2(I) collagen promoter, we asked whether footprints produced on the proximal 220 bp pro[alpha]1(I) promoter were similar using nuclear extracts from cells that express this gene and cells that do not. Figure 3 shows that, for both the upper strand (Fig. 3 A) and the lower strand (Fig. 3 B) of the proximal 220 bp pro[alpha]1(I) collagen promoter, the segments that are protected with nuclear extracts of BALB/3T3 fibroblasts and primary mouse fibroblasts (Fig. 3 A and B, lanes 3 and 4) were also protected with nuclear extracts of the other three cells, S-194 B cells, NMuLi liver epithelial cells and RAG renal adenocarcinoma cells (Fig. 3 A and B, lanes 5-7). The strongest protection occurred from nt -80 to -130 on both strands; less intense protection occurred from nt -133 to -155 and from nt -175 to -190. We conclude that the nuclear proteins capable of binding in vitro to discrete sites in the proximal 220 bp pro[alpha]1(I) collagen promoter are present in nuclei of all cell types tested.
To determine whether the DNA sequences that were protected in vitro were also occupied by DNA binding proteins in vivo, we performed in vivo DMS footprinting in various cells using LM-PCR. As shown in Figure 4 , the G residue at position -78 on the upper strand of DNA was hypersensitive to DMS methylation in primary fibroblasts and BALB/3T3 fibroblasts (lanes 2 and 3). This hypersensitivity to DMS methylation was not found in cells in which no pro[alpha]2(I) collagen RNA was detected: S-194 B cells, NMuli liver epithelial cells and RAG renal adenocarcinoma cells (lanes 4-6); in these cells the methylation pattern was the same as with naked DNA (lane 1). This DMS hypersensitive site which is immediately adjacent to the CCAAT motif (from nt -80 to -84) could result from protein-DNA interactions at and around this motif. In support of this interpretation, we observed a limited but reproducible protection at nt -81 on the upper strand (Fig. 4 ) and at nt -85 on the lower strand (data not shown). These results are consistent with the notion that among the cells tested, protein-DNA interactions occur at and immediately around the CCAAT motif of the promoter in fibroblasts but not in other cells.
Five different mouse cell types were used for in vivo footprints of the 350 bp proximal mouse pro[alpha]2(I) and the 220 bp proximal mouse pro[alpha]1(I) collagen promoters in intact cells in culture: primary fibroblasts, BALB/3T3 fibroblasts, S-194 B cells, NMuLi liver epithelial cells and RAG renal adenocarcinoma cells. Measurements of the pro[alpha]1(I) and the pro[alpha]2(I) collagen RNAs in these cells indicated that these RNAs are present in the two fibroblast cell types, but are absent in the three other cell lines. In both genes, discrete in vivo DMS footprints were detected in the proximal promoter segments only in cells containing type I collagen RNA. In the three other cell lines, the patterns of DMS methylation were identical to those obtained with naked DNA for each of the two type I collagen proximal promoters. These in vivo DMS footprints strongly suggested that DNA binding proteins occupy discrete segments of the two type I collagen proximal promoters in intact cells and that this occurs only in cells in which the type I collagen genes are expressed. In other genes, such as the major histocompatibility class II genes, binding sites for ubiquitous proteins have also been shown to be occupied only in cells that express the genes (33 ).
In contrast, in vitro DNase I footprinting, produced with nuclear extracts of cells that express and cells that do not express type I collagen genes, showed essentially identical protection patterns. Our interpretation of these results was that the DNA binding proteins that are responsible for these footprints are present in nuclear extracts of type I collagen-expressing and non-expressing cells and that they probably are mainly ubiquitous proteins. This conclusion was also supported for the proximal pro[alpha]2(I) collagen promoter by the demonstration of very similar patterns of DNA-protein complexes in gel shift assays with nuclear extracts of BALB/3T3 fibroblasts and S-194 B cells when oligonucleotide probes were used encompassing the in vitro DNase I footprints of this proximal promoter (14 and unpublished observation).
The in vivo DMS footprints in the pro[alpha]2(I) collagen proximal promoter are located within segments that are also protected in vitro, except for a protected residue at -144 and one that is hypersensitive to methylation at -149. Hypersensitivity to methylation might result from DNA bending or other structural changes in DNA caused by binding of proteins in adjacent segments. Transient expression and in vitro transcription experiments using 5' deletions and internal deletions have indicated that the segment between -40 and -170, in which the in vivo DMS footprints were detected, had a critical role in this promoter's activity (14 ). The in vivo footprints in the pro[alpha]1(I) collagen promoter found in the GC-rich repeats surrounding the CBF binding site indicate promoter occupancy in these segments and occur within a segment that is protected in vitro. The in vivo footprints at nt -20 in the pro[alpha]1(I) promoter just downstream of the TATA box could correspond to detectable occupancy by a component of the general transcription machinery. Although in vivo footprints suggest that specific areas of the promoter are occupied by DNA binding proteins, they do not reveal the nature of these proteins. It is, however, reasonable to postulate that both in vivo and in vitro, similar proteins bind to specific elements of the type I collagen promoters. We believe, therefore, that the proteins that bind to the DNA segments identified by in vivo footprints are mainly ubiquitous proteins.
Our recent experiments have identified strong upstream tissue-specific enhancers in the two type I collagen genes. In transgenic mouse experiments, separate cis-acting elements upstream of the proximal pro[alpha]1(I) promoter were observed to control expression of reporter genes in different type I collagen-producing cells; one such element located ~1600 bp upstream of the transcription start site was shown to control osteoblast expression, while another element located between -3.2 and -2.3 kb controlled expression in tendons and fascia (23 ,24 ). There is also evidence of a dermis-specific enhancer in the first intron of the human pro[alpha]1(I) gene (25 ). In the mouse pro[alpha]2(I) gene, a potent far-upstream enhancer element identified between 11.5 and 17.5 kb upstream of the start of transcription, was shown to control expression mainly in mesenchymal cells and fibroblasts (26 ). This element contains a cluster of three DNase I hypersensitive sites in chromatin that are cell specific.
If ubiquitous DNA binding proteins bind to discrete elements of the type I collagen proximal promoters in fibroblasts in vivo but not in cells that do not express the type I collagen genes, a mechanism must exist that prevents accessibility of these ubiquitous proteins to the proximal promoters in the latter cells. One possibility is that in the cells which do not express the type I collagen genes, the organisation of nucleosomes and eventually other chromatin components inhibit access of these ubiquitous DNA binding proteins to the promoter DNA. In contrast, in cells in which the type I collagen genes are expressed, these proteins would have access to their binding sites in the proximal promoters. We hypothesize that in cells expressing the type I collagen genes, proteins binding to the upstream cell-specific enhancers could recruit proteins that favor disruption of the nucleosomal structure in the proximal promoters allowing ubiquitous DNA binding proteins to gain access to the proximal promoters.
*To whom correspondence should be addressed. Tel: +1 713 792 2590; Fax: +1 713 794 4295; Email: benoit_decrombrugghe@molgen.mda.uth.tmc.edu
REFERENCES