INTRODUCTION

S100A4 (p9Ka) is a member of the S100 family of calcium-binding proteins first identified in a cultured myoepithelial-like cell line, Rama 29 (1). The S100A4 (p9Ka) gene has been isolated and its nucleotide sequence determined (2). Elevation of the level of S100A4 (p9Ka) in benign, non-metastatic rat (3) and mouse (4) cells induces metastatic capability in some of the cells. Although mice containing multiple copies of a rat (5) or mouse (6) S100A4 transgene show no observable phenotype, mice which contain cells that express both the rat S100A4 transgenes and the activated neu oncogene, or mouse S100A4 transgenes in a genetic background that yields primary mammary tumors, succumb to metastatic mammary tumors (6, 7), strongly suggesting that overexpression of S100A4 in mammary tumors can yield the metastatic phenotype. Some rat and mouse cultured mammary epithelial cell lines of high metastatic potential contain elevated levels of S100A4, when compared with their non-metastatic counterparts (8, 9).

Mechanisms for the regulation of transcription of the murine S100A4 (mts1) gene have been investigated. It has been reported that cis-acting elements 5 of the region immediately upstream of the TATA box play little or no part in transcription of the mouse S100A4 (mts1) gene (10), but the loss of methylation of DNA may be associated with transcriptional activation of this gene in mice and human lymphoma cells (10, 11). Further analysis of the natural mouse gene revealed two elements within the first intron that may be involved in regulating the expression of the S100A4 (mts1) gene in this species (12). The nucleotide sequence of this reported 16-bp¹ protected element is identical in both the mouse and the rat genomes (2, 12). However, the expression of the rat S100A4 (p9Ka) transgene under the control of its own promoter in transgenic mice suggests a different pattern of expression between these two species (5). Thus, additional regulatory elements in the rat S100A4 (p9Ka) gene have been sought which might affect rat S100A4 (p9Ka) levels in rat mammary epithelial cells.

EXPERIMENTAL PROCEDURES

E. coli strains DH5 and XL-1 blue (Stratagene) were made competent for DNA uptake using the procedure of Hanahan (13). Competent cells were stored at 70 °C. DNA manipulations were carried out using pBluescript II (Stratagene).

Oligodeoxyribonucleotides were synthesized using -cyanoethyl phosphoramidite chemistry on a Perceptives Biosystems Expedite 8909 DNA synthesizer. A cDNA to human GC-factor mRNA was produced by reverse transcript-polymerase chain reaction (PCR). Initially, 3 µg of RNA isolated from A431 cells (kindly provided by T. Gorman, Department of Medicine, University of Liverpool) was heated at 70 °C for 10 min with 20 µg of human placental ribonuclease inhibitor and 1 µg of oligo(dT) in a total volume of 22 µl, and the mixture was then cooled on ice. Reverse transcription was carried out for 1 h at 37 °C with 0.5 mM dNTPs, 10 mM dithiothreitol, and 2 µl of Superscript (Life Technologies, Inc.) in the buffer system supplied by the manufacturer. A 5-µl sample of the reaction mixture was used for amplification of GC-factor cDNA by PCR using a pair of specific primers for 35 cycles.

A 3,198-bp cloned fragment of the normal rat S100A4 (p9Ka) gene (2) consisting of 2,696 bp of the upstream region of the S100A4 (p9Ka) gene, the entire first exon, first intron, and 15 bp of the second exon of the S100A4 (p9Ka) gene was cloned into the multiple cloning site of pBluescript II to form recombinant plasmid pS100A4/4 (see Fig. 1A). This recombinant plasmid was partially digested with restriction enzyme EcoRI, and a 4.6-kbp EcoRI-digested DNA cassette consisting of a secreted placental alkaline phosphate reporter gene (14) linked to an SV-40 poly(A)-addition signal (SPAP/SV) was inserted 16 nucleotides from the 5 end of the second exon of the S100A4 (p9Ka) gene (see Fig. 1) to form the expression plasmid pS100A4/SPAP (see Fig. 1B).

Deletions in the S100A4 (p9Ka) promoter were produced using the plasmid pS100A4/4. Following sequencing of the 2,696 bp of the upstream region, restriction sites were identified. Four deletion mutants of the upstream region of the S100A4 (p9Ka) gene were constructed by digesting pS100A4/4 with restriction enzyme NotI (site located in the pBluescript II vector) and with one of the following restriction enzymes, SpeI, NcoI, NheI, or HindIII, all of which have sites uniquely located in the upstream region of the rat S100A4 (p9Ka) gene (see Fig. 2A). Following digestion, recircularization of the plasmids, and tranformation, colonies of bacteria bearing plasmids were propagated, and the absence of the deleted fragments was checked by restriction enzyme digestion and sequencing. Following the selection of suitable clones and isolation of plasmid DNA, the SPAP/SV cassette was inserted as described above. The resulting plasmids bearing deletions were designated p2122, p1818, p582, and p303. The correct site of insertion and orientation of the SPAP gene/SV-40 poly(A) addition cassette was confirmed by the correct pattern of bands arising from digestion of the plasmids with restriction enzyme XbaI.

pS100A4/4 was cleaved at the single NotI site, the sticky ends were in-filled with Klenow polymerase using -thiophosphate dNTPs (15), and the plasmid was subsequently digested with restriction enzyme NcoI (see Fig. 1). Unidirectional nested deletions from the 5 end of the p9Ka gene were produced using a time course of digestion with exonuclease III as described previously (2), removing the resulting single-stranded overhangs with mung bean nuclease (16). Following circularization of the plasmid DNA by ligation and transformation into E. coli cells, recombinant plasmids from E. coli colonies representing different time points were sequenced to precisely identify the extent of deletion of the upstream region. Clones with deletions between 1,818 and +17 were selected. The SPAP/SV cassette was inserted as described above. The resulting selected nested deletion mutants containing the downstream SPAP/SV cassette were termed p1670, p1400, p1225, p993, p762, p115, and p+17.

Rat mammary epithelial cells expressing low levels of S100A4 (p9Ka) (17) were the normal rat mammary-derived Rama 704 (18) and the benign tumor-derived Rama 25 (19) and Rama 37 (20) cell lines. S100A4 (p9Ka)-expressing (17) derivative cell lines of these epithelial cells were respectively, Rama 711 (18), Rama 29 (19), and Rama 37-E8 (20). Metastatic epithelial cells expressing S100A4 (p9Ka) were the Rama 600 (21) and Rama 800 (22) cell lines, the metastatic cells, KP1-R37, arising from the transfection of the benign epithelial Rama 37 cell line with additional copies of the S100A4 (p9Ka) gene, and a cell line derived from a resulting lymph node metastasis, KP1-LNT1 (3). The cells were grown in a 10% CO₂, 90% air mixture at 37 °C in Dulbecco's modified Eagle's medium supplemented with 5% fetal bovine serum (Life Technologies, Inc.), 50 ng/ml hydrocortisone, 50 ng/ml insulin, as described previously (19). Cells were transiently transfected using calcium phosphate-mediated DNA transfection (23), as described previously (24). The maximum level of expression of the SPAP reporter gene occurred 72 h after transfection into the Rama 800 cells (data not shown). To correct for dish-to-dish variation in transfection efficiencies, the cells were co-transfected with vector pGL2 (Promega), which contains a luciferase gene under the control of an SV40 enhancer and promoter. Assays for SPAP activity were carried out on culture supernatants, and luciferase assays were performed on cell extracts according to the recommendations of the manufacturers (Promega and Tropix Inc.).

Extracts of nuclei were prepared from the three cell lines, Rama 37, Rama 37-E8, and Rama 800 as described previously (25). Briefly, 10⁶ cells were washed twice with PBS, and the harvested cells were pelleted by brief centrifugation at 1,000 × g. The resulting cell pellet was resuspended in 500 µl of ice-cold extraction buffer (20 mM HEPES, pH 7.8, 10 mM KCl, 0.1 mM EDTA, 1 mM dithiothreitol, and 0.5 mM phenylmethylsulfonyl fluoride) and allowed to stand on ice for 15 min, after which time, 30 µl of 10% (v/v) Nonidet P-40 was added and the mixture was vortexed for 10 s. The cells were repelleted by centrifugation, the supernatant was discarded, the pellet was resuspended in 50 µl of extraction buffer containing 400 mM NaCl, and the mixture was agitated vigorously for 15 min at 4 °C. After centrifuging the mixture at 12,000 × g for 5 min at 4 °C, the supernatant contained the nuclear extract, and its protein concentration was determined using the method of Bradford (26). Appropriate volumes of extracts prepared from the three cell lines containing equal amounts of protein were used in the relevant experiments.

The assay was performed as described previously (27). DNA was amplified by PCR using the following primers: upper strand (5-3) GGTGAGGGGTGGATAGTCT, lower strand (5-3) GACCCAGTGTGTCCCACC, which correspond to nucleotide positions 1,403 to 1,385 and 1,224 to 1,241 upstream of the start site of transcription. The primer for amplifying the upper strand was labeled at its 5 end using T4 polynucleotide kinase with ³²P from [-³²P]ATP (28). The amplified DNA resulting from the PCR reactions was purified by electrophoresis through low melting temperature agarose gels. DNA/protein binding reactions consisted of the volume of nuclear extract containing 5 µg of protein, 1 µg of poly(dI-dC), 20 mM Tris-HCl, pH 7.9, 2 mM MgCl₂, 50 mM NaCl, 1 mM EDTA, 10% (v/v) glycerol, 0.1% (v/v) Nonidet P-40, 1 mM dithiothreitol, and 50 µg/ml BSA in a total volume of 19 µl. The tube was incubated at room temperature for 5 min followed by the addition of 1 µl of ³²P-labeled DNA (about 12,000 cpm), and incubation was continued for a further 20 min at room temperature. Samples of 10 µl were analyzed by polyacrylamide gel electrophoresis carried out as described previously (27).

Competitive DNA mobility shift assays were set up essentially as described above, except that two additional reactions were carried out for each cell line. Prior to adding the radioactively-labeled, PCR-amplified DNA fragment, varying amounts of the same DNA fragment or a similarly-sized unrelated DNA (a 128-bp PvuII-KpnI fragment of pBluescript II vector) in unlabeled form were added to the binding reaction mixtures.

The assay was performed as described previously (29). The procedures for the preparation of nuclear extracts, for the end-labeling of DNA with ³²P, and procedures for the DNA-protein binding reactions were as described above. After completion of the DNA-protein binding reactions, the 20-µl reaction mixtures were diluted to 100 µl by the addition of distilled water, followed by the addition of an equal volume of a salt mixture consisting of 10 mM MgCl₂, 2 mM CaCl₂, and 1.25 × 10² units of DNase I (Boehringer Mannheim). The reaction mixture was incubated at room temperature for 1 min, and the reaction was stopped by adding an equal volume of a buffer consisting of 0.2 M NaCl, 40 mM EDTA, 1% (w/v) SDS, 125 µg/ml tRNA (Pharmacia), and 100 µg/ml proteinase K (Boehringer Mannheim) and incubating at 37 °C for 15 min. The mixture was extracted with phenol/chloroform, and the DNA was precipitated with ethanol. Denaturing polyacrylamide gel electrophoresis and autoradiography were performed as described previously (29). Maxam and Gilbert DNA sequencing was performed as described previously (28).

10⁶ cultured cells were collected by mild digestion with trypsin and washed twice with PBS. The cells were resuspended with Dulbecco's modified Eagles' medium containing 0.05% (v/v) dimethyl sulfoxide (Me₂SO) for 5 min. Genomic DNA was isolated and treated with 1 M piperidine as described previously (30). Ligation-mediated PCR (LMPCR) procedures were adapted from those of Garrity et al. (31). Specific primers for LMPCR were as follows (positions in the S100A4 (p9Ka) gene sequence shown in parentheses and all sequences shown 5-3): for the upper strand, primer 1, AGCTACCCACAAGCTTTAGAG (1,436 to 1,416); primer 2, AGAGGCTTGTCTAGCTGGTG (1419 to 1400); primer 3, GGTGAGGGGTGGATAGTCT (1,403 to 1,385); and for the lower strand, primer 1, CCAGTGGTAAGGAGCATAAG (1,184 to 1,203); primer 2, GGAGCATAAGGAACAAGGAC (1,194 to 1,213); primer 3, CAAGGACTCAGTCCCAAGACC (1,207 to 1,227). The unidirectional linker was formed by annealing two primers (5-3): primer 1, GCGGTGACCCGGGAGATCTGAATTC; primer 2, GAATTCAGAT.

Total RNA (10 µg) was isolated from cell lines Rama 37, Rama 37-E8, and Rama 800 using a previously published procedure (32, 33). RNA preparations were subjected to 0.8% (w/v) agarose gel electrophoresis in the presence of formaldehyde, and the fractionated RNA was blotted onto Hybond nylon membranes (Amersham Intl.). The nylon membranes were incubated with a PCR-amplified cDNA corresponding to GC-factor mRNA that had been radioactively labeled using the method of random-primed synthesis (34) to a specific activity of between 5 × 10⁷ and 1 × 10⁸ cpm/µg. After hybridization, the filter was washed twice for 20 min each wash with a solution of 0.1 × SSC (15 mM sodium chloride, 1.5 mM sodium citrate) 0.1% (w/v) SDS at 65 °C, and subjected to autoradiography. Equivalent loading of RNA samples onto each lane was ensured by hybridizing the filter to a probe for the constitutively expressed rat non-muscle actin mRNA (35).

RESULTS

When the expression construct pS100A4/SPAP consisting of the SPAP gene inserted downstream of 2,696 bp of the upstream region of the S100A4 gene (Fig. 1) was transfected into the rat mammary cell lines, Rama 37, Rama 37-E8, and the metastatic epithelial cell line, Rama 800, the level of activity of SPAP arising from transient expression of the SPAP reporter gene in Rama 37-E8 and Rama 800 cells was 7-fold higher relative to that in the Rama 37 cells (not shown). The level of reporter gene activity reflects approximately the levels of S100A4 (p9Ka) expression in these cell lines (1, 17).

To find out whether the region upstream of the rat S100A4 gene contains cis-acting elements that regulate the rate of transcription of the S100A4 (p9Ka) mRNA, an initial series of constructs containing deletions were produced using restriction enzymes (see "Experimental Procedures"). When constructs containing these deletion mutants were transfected into the S100A4 (p9Ka)-expressing Rama 37-E8 and Rama 800 cell lines, removal of DNA between 2,696 and 303 bp upstream of the start site of transcription of the S100A4 (p9Ka) gene made no significant difference to the level of SPAP expression in either of these cell lines (Fig. 2A). When the same constructs were transfected into the low S100A4 (p9Ka)-expressing benign epithelial cell line, Rama 37, a similar result was obtained when DNA between 2,696 and 1,818 bp upstream was deleted. In contrast to the results with the S100A4-expressing Rama 37-E8 and Rama 800 cells, further deletion of DNA between 1,818 and 582 resulted in an approximately 6-fold increase in the level of expression of the reporter gene (Fig. 2A) relative to that achieved in the originally transfected Rama 37-E8 and Rama 800 cells. This result suggests that there is a region of DNA between 1,818 and 582 bp upstream of the start site of transcription of the rat S100A4 (p9Ka) gene that inhibits expression of the reporter gene in Rama 37 cells, but not in Rama 37-E8 and Rama 800 cells.

Fine mapping of the region between 1,818 and 582 bp, using nested deletion mutants, located this inhibitory region to between nucleotides 1,400 and 1,225 bp (Fig. 2B). Deletion of all the DNA between 762 and 115 bp upstream of the transcription start site of the rat S100A4 (p9Ka) gene slightly reduced the level of transient expression of the SPAP reporter gene but only by a further 13%. In contrast, removal of all upstream region DNA to a position of +17 bp relative to the start site of transcription, which includes removal of the TATA box, resulted in a virtually complete abolition of the transient expression of the reporter gene in these cells (Fig. 2B).

DNA mobility shift assays carried out using ³²P-labeled inhibitory region containing DNA 179 bp between 1,402 and 1,223, which had been amplified by PCR, revealed that nuclear extracts from all three cell lines tested formed retarded DNA-protein complexes. However, the signal from the complex formed with the nuclear extract from Rama 37 cells was much more intense than that formed with nuclear extracts from either S100A4-expressing cell line Rama 37-E8 or Rama 800 when the results were normalized for equal amounts (5 µg) of total nuclear protein from the three cell lines and for equal amounts of DNA probe (Fig. 3A). Three additional retardation products of much lower intensity were formed, which were more abundant in the lanes containing extracts from S100A4-expressing Rama 37-E8 and Rama 800 cell lines (Fig. 3A). The nature of these retardation products is not presently known. Increasing concentrations of unlabeled inhibitory region DNA competed effectively with the labeled inhibitory region DNA (Fig. 3B), whereas there was no reduction in the signal when up to at least a 40-fold weight excess of unrelated DNA was included in the assay (Fig. 3B).

DNase I footprinting analysis in vitro of the upstream region of the S100A4 (p9Ka) gene showed that 20 bp of DNA was completely protected in both the upper and lower strands when bound with a nuclear extract from Rama 37 cells. However, the 20-bp sequence was not exactly complementary in the upper and lower strands. In the upper strand, the protected 20-bp sequence corresponds to nucleotides 1,302 to 1,283 bp (5 to 3) upstream of the start site of transcription, and in the lower strand, the protected 20-bp sequence corresponds to 1,289 bp at the 5 end to 1,308 bp at the 3 end from the start site of transcription. Only 14 bp (nucleotides 1,302 to 1,289) out of the protected 20 bp are protected on both strands (Fig. 4, A and B). In contrast to the results with nuclear extracts from the Rama 37 cells, nuclear extracts of Rama 37-E8 or Rama 800 cells did not completely protect the sequence from digestion by DNase I (Fig. 4, A and B) when used at the same concentration as the Rama 37 extract, a result that is consistent with the gel mobility shift assays.

Footprinting in vivo carried out using LMPCR revealed that in the 14-bp inhibitory element described above, the G residues were completely protected from Me₂SO-mediated methylation and subsequent cleavage with piperidine in Rama 37 cells, but were only partially protected in Rama 37-E8 cells and were virtually unprotected in Rama 800 cells (Fig. 5, A and B). The results of DNA mobility shift and DNA footprinting in vitro and in vivo taken together strongly suggest that a transcriptional factor, which binds to the 14-bp inhibitory element in the upstream region of the S100A4 (p9Ka) gene, might be responsible for the repression of the level of S100A4 (p9Ka) mRNA and that this factor might itself be at higher concentration in the Rama 37 cell line than in Rama 37-E8 and Rama 800 cell lines.

The nucleotide sequence of the 14-bp inhibitory element is GC-rich (TGGCAGGGGCCTCC) and contains a sequence, GGCAGGGGCC, that bears significant homology to the binding site of a GC transcriptional repressor factor (36). A ³²P-labeled oligonucleotide containing the consensus GC-factor binding sequence (NNGCGGGGCN) (36) formed a very strong complex with a component of the nuclear extracts of the Rama 37 cells (Fig. 6A). In contrast, complexes formed between the GC-factor consensus sequence and nuclear extracts from the Rama 37-E8 and Rama 800 cells appear weaker although the amount of total nuclear protein and oligonucleotide used in binding reactions was the same for all three cell lines (Fig. 6A). Competitive oligonucleotide gel mobility shift assays, as described above, showed that the resulting complex was specific (Fig. 6B). Furthermore, the oligonucleotide containing the consensus GC-factor binding site almost completely abolished the complex formed between the inhibitory region DNA of the S100A4 (p9Ka) gene and the component of the nuclear extract of the Rama 37 cells at an equal molar ratio (Fig. 7). These results suggest that the trans-acting factor bound to the inhibitory element of the rat S100A4 (p9Ka) gene is closely-related to GC-factor (36). However, the binding between the inhibitory element of the S100A4 (p9Ka) gene and GC-factor appears to be somewhat weaker than that between GC-factor and its consensus sequence binding site.

The results obtained so far suggest that there might be quantitatively different levels of GC-factor-like activity in the three cell lines studied. The relative levels of mRNA for GC-factor in the three cell lines were measured by Northern blotting using a GC-factor-specific cDNA generated by reverse transcript-PCR as a probe (see "Experimental Procedures"). The probe hybridized to an mRNA of 2.9 kb in all three cell lines. The level of this mRNA was significantly higher in Rama 37 cells than in cell lines Rama 37-E8 and Rama 800 (Fig. 8, Table I), and the level of mRNA in the three cell lines corresponded with the degree of protection seen in DNase I footprinting experiments (Figs. 4 and 5). In general, there is an inverse correlation between the level of GC-factor mRNA and S100A4 (p9Ka) mRNA in these cell lines (Fig. 8, Table I). Furthermore, in five additional mammary cell lines containing various levels of S100A4 (p9Ka) mRNA, the level of S100A4 (p9Ka) mRNA also has an inverse relationship with the level of GC-factor mRNA (Table I). There was one exception to this observation. The cell strain KP1-R37, which was produced by the transfection of Rama 37 cells with additional copies of the cloned rat S100A4 (p9Ka) gene, exhibited elevated levels of S100A4 (p9Ka) mRNA, but did not show a lower level of GC-factor mRNA than that in the Rama 37 cells.

Table I. Relative levels of GC-factor and S100A4 mRNAs in various rat mammary cell lines Total RNA was isolated from rat mammary cell lines and subjected to agarose gel electrophoresis. Northern blotting and sequential hybridization with probes to GC-factor, S100A4 (p9Ka), and rat non-muscle actin mRNAs were carried out as described under "Experimental Procedures." mRNA levels were determined by densitometric scanning of the resulting autoradiographs (Fig. 8). The ratio of mRNA content is the peak area for GC-factor or S100A4 mRNA in each cell line relative to GC-factor or S100A4 mRNA in cell line Rama 37, and each value is the mean of two separate blot scans ± S.D. Cell line Ratio of GC-factor mRNA content Ratio of S100A4 mRNA content Rama 37 1.0 1.0 Rama 37-E8 0.22  ± 0.03 6.83  ± 0.39 Rama 800 0.12  ± 0.02 8.65  ± 0.13 Rama 600 0.26  ± 0.05 6.63  ± 0.83 Rama 25 0.96  ± 0.07 1.65  ± 0.32 Rama 29 0.31  ± 0.04 8.44  ± 0.94 Rama 704 1.18  ± 0.05 1.64  ± 0.11 Rama 711 0.22  ± 0.03 7.15  ± 1.17 KP1-R37 1.02  ± 0.11 10.27  ± 1.03 KP1-LNT1 0.94  ± 0.04 1.34  ± 0.15

DISCUSSION

A 14-nucleotide pair cis-acting negative control element has been located approximately 1,300 bp upstream of the start site of transcription of the rat S100A4 (p9Ka) gene. Deletion of this region results in a 6-fold increase in the level of activity of a SPAP reporter gene when transiently transfected into a benign epithelial rat mammary cell line, Rama 37, which naturally displays only low levels of p9Ka mRNA and protein (17). No such increase is observed when constructs bearing the same deletion are transfected into cell lines, Rama 37-E8 and Rama 800, which naturally express high levels of S100A4 (p9Ka). Gel mobility shift assays, DNase 1 footprinting in vitro, and footprinting experiments in vivo strongly suggest that there is a factor present in nuclear extracts of the Rama 37, Rama 37-E8, and Rama 800 cells that binds to this cis-acting element. The degree of protection of nuclear extracts of the various cell lines corresponds inversely with the level of S100A4 (p9Ka) in the cell lines, suggesting a quantitative variation of a factor binding to the inhibitory region.

The nucleotide sequence of the protected inhibitory region suggests that it contains a sequence which closely resembles the consensus binding sequence of GC factor (36). This factor was originally identified by its binding to a regulatory region in the promoters of the genes for the human epidermal growth factor receptor,  actin, and the calcium-dependent protease, calpain (36). The present paper is the first report of a GC factor binding site in a rat gene, and the first report of such a functional site in the gene for a S100 protein. It is not known whether the variation of the recognition sequence from the human consensus sequence reflects a species difference. However, the fact that an oligonucleotide corresponding to the consensus human GC-factor binding region effectively competes with the rat S100A4 (p9Ka) inhibitory sequence for binding of a factor in the rat cell nuclei suggests that the binding specificity of the rat factor is similar to that of the human GC-factor.

Several lines of evidence point to the factor that resembles GC-factor being a regulator of the transcription of the S100A4 (p9Ka) gene in the rat mammary cells in vivo. The degree of protection of DNA in footprinting experiments in vivo and in vitro is in inverse relationship to the levels of S100A4 (p9Ka) mRNA in the corresponding cells, raising the possibility that the GC-factor-like activity might regulate S100A4 expression in vivo. This interpretation is confirmed by the observation that a GC-factor-specific probe hybridizes to an mRNA from the rat cells that is the same size (3 kb) as that already described previously in human cells (36). The level of this mRNA, as measured by a quantitative Northern blotting procedure using a GC-factor-specific probe, exhibits an inverse relationship with the level of S100A4 (p9Ka) mRNA in nine out of ten rat mammary cell lines that exhibit differing levels of S100A4 (p9Ka) mRNA. The one exception is a cell line made metastatic by the transfection of additional copies of the S100A4 (p9Ka) gene (3). This cell line expresses S100A4 (p9Ka) mRNA at a high level but shows the same level of GC-factor mRNA as the low expressing Rama 37 cells. The explanation of this observation is likely to be that the additional 10-20 transfected copies of the S100A4 (p9Ka) gene lead to elevated levels of S100A4 (p9Ka) mRNA and protein by titrating out molecules of the GC-factor in the recipient Rama 37 cells. However, the alternative explanation that levels of GC-factor protein but not mRNA levels are reduced by some other mechanism by the transfection procedure cannot be entirely ruled out at this stage. The observation that a cell line, KP1-LNT1, from a lymph node metastasis arising from the KP1-R37 cells exhibits an S100A4 (p9Ka) level consistent with its level of GC factor mRNA suggests that other mechanisms in the development of a metastasizing tumor might regulate S100A4 (p9Ka) mRNA levels in these cells. Such a mechanism, in this case, might be methylation of the S100A4 (p9Ka) gene, as reported previously for the mouse S100A4 (mts1) gene (11).

The results of the present experiments contrast with those previously obtained using metastatic and non-metastatic murine mammary cells in which no evidence of important cis-acting regulatory sequences were found in the 5 region of the mouse S100A4 (mts1) gene in the region 41 to 1,897 upstream of the transcriptional start site (10). One possible explanation for the difference between the results with the rat and mouse S100A4 genes is that in the mouse S100A4 (mts1) gene, any GC-factor-binding sequence might lie further upstream than the 0 to 1,897 region examined (10). This explanation is unlikely since an exactly conserved GC-factor binding site is located at 1,472 bp upstream of the start site of transcription of the S100A4 gene in the human genome (37) (Fig. 9A). An alternative explanation is that in the murine cells, the predominant mechanism regulating the S100A4 (mts1) gene is the 16-base pair protein binding region, 290 bp downstream of an enhancer element, in the first intron of the S100A4 (mts) gene, which resembles the CD3 enhancer of T lymphocytes (12). Differential patterns of in vivo footprinting between metastatic and non-metastatic variants of mouse mammary epithelial cells arise from differential methylation of this region of the genome in the murine cell lines studied (12), and possibly from the resultant creation by methylation of a novel inhibitory AP-1 site, 40 bp downstream of the 16-bp binding region (38). The sequence that constitutes this potential AP-1 site specifically appears to be absent by deletion in the first intron of the rat S100A4 (p9Ka) gene, despite generally good conservation of the sequence of the flanking regions (Fig. 9B) and the first intron generally between rat and mouse. This observation suggests a possible difference between inhibitory mechanisms in mouse and rat S100A4 genes. The findings might serve to explain differences between the distribution of S100A4 in the tissues of rats and mice reported previously (5) from an examination of the tissue distribution of the endogenous S100A4 gene in rats (39) and mice (40) and by the pattern of expression of rat S100A4 transgenes under the control of their own rat promoter in the tissues of transgenic mice (5). In these mice, the tissue distribution of the rat S100A4 (p9Ka) mRNA in the mouse cells was identical to that found in the rat and did not correspond to the distribution of the endogenous mouse S100A4 (mts1) mRNA when the respective mRNAs were analyzed by Northern blotting procedures on the same tissue specimens (5).

These results suggest that there is a different emphasis in the regulation of the S100A4 gene between rats and mice. However, the consequence of overexpression of the S100A4 gene and its protein products in rat and mouse epithelial cells is the same, i.e. the induction of the metastatic phenotype in non-metastatic benign rodent mammary epithelial cells.

As noted above, the sequence that is recognized by GC-factor in the rat epithelial cells is conserved exactly between rat and human DNA although its position relative to the start site of transcription is slightly different in the two species, 1,300 in the rat and 1,427 in the human (Fig. 9A). In the human, the location is only 91 base pairs downstream of the poly(A)-addition site of the S100A3 gene (37). Since in the rat cells, the level of GC-factor mRNA is inversely related to the level of S100A4 (p9Ka) mRNA, suggesting a correlation between GC-factor mRNA and S100A4 (p9Ka) mRNA levels, it is possible that changes in GC-factor levels might also be associated with the acquisition of metastatic capability in human breast cancer cells.