Identification of a ras-activated enhancer in the mouse osteopontin ...

MOLECULAR AND CELLULAR BIOLOGY, Jan. 1995, p. 476–487 0270-7306/95/$04.0010 Copyright q 1995, American Society for Microbiology

Vol. 15, No. 1

Identification of a ras-Activated Enhancer in the Mouse Osteopontin Promoter and Its Interaction with a Putative ETS-Related Transcription Factor Whose Activity Correlates with the Metastatic Potential of the Cell XIAOJIA GUO,1† Y. PAUL ZHANG,1 DUANE A. MITCHELL,1 DAVID T. DENHARDT,1* 2 AND ANN F. CHAMBERS Department of Biological Sciences, Rutgers University, Piscataway, New Jersey 08855-1059,1 and London Regional Cancer Centre and Department of Oncology, University of Western Ontario, London, Ontario N6A 4L6, Canada2 Received 22 November 1993/Returned for modification 7 January 1994/Accepted 18 October 1994

The role of RAS in transducing signals from an activated receptor into altered gene expression is becoming clear, though some links in the chain are still missing. Cells possessing activated RAS express higher levels of osteopontin (OPN), an avb3 integrin-binding secreted phosphoprotein implicated in a number of developmental, physiological, and pathological processes. We report that in T24 H-ras-transformed NIH 3T3 cells enhanced transcription contributes to the increased expression of OPN. Transient transfection studies, DNAprotein binding assays, and methylation protection experiments have identified a novel ras-activated enhancer, distinct from known ras response elements, that appears responsible for part of the increase in OPN transcription in cells with an activated RAS. In electrophoretic mobility shift assays, the protein-binding motif GGAGGCAGG was found to be essential for the formation of several complexes, one of which (complex A) was generated at elevated levels by cell lines that are metastatic. Southwestern blotting and UV light cross-linking studies indicated the presence of several proteins able to interact with this sequence. The proteins that form these complexes have molecular masses estimated at approximately 16, 28, 32, 45, 80, and 100 kDa. Because the ;16-kDa protein was responsible for complex A formation, we have designated it MATF for metastasisassociated transcription factor. The GGANNNAGG motif is also found in some other promoters, suggesting that they may be similarly controlled by MATF. plasma membrane where it triggers a cascade of protein phosphorylation events mediated by kinases such as MAPKK, MAPK(ERK), and RSK (6, 19, 25). The final targets of such phosphorylation include transcription factors, like those acting on AP-1 and ets elements in the polyomavirus enhancer, the VL30 transcriptional element, and the c-fos and collagenase promoters, that cause changes in gene expression—the metallo- and cysteine-proteinases for example (9, 10, 14, 15). We have sought to elucidate the RAS-stimulated signal transduction pathway by studying the promoter of osteopontin (OPN), a gene whose expression is enhanced by activation of RAS (12), in order to identify an effector molecule that could then be traced back upstream. OPN is a secreted RGD-containing phosphoprotein that interacts with the avb3 integrin and is capable among other things of promoting cell attachment, regulating intracellular Ca21 levels, and influencing gene expression (21). OPN has long been recognized as a protein whose expression is increased when cells are transformed (47). OPN is responsive to various cell growth signals, including estrogen and progesterone, calcitriol (vitamin D3), and 12-O-tetradecanoylphorbol-13-acetate (TPA) (17, 18, 39). OPN (also called Eta-1) expression is enhanced upon activation of T cells and macrophages (43). OPN is induced during multistage carcinogenesis in mouse skin, and its expression correlates both with the metastatic potential of mouse fibroblasts and with the extent of RAS activation (12, 16). In this work, we present evidence that the increase in OPN expression in H-ras-transformed cells is driven at least in part at the transcriptional level. The sequence GGAGGCAGG, which is distinct from any sequence reported in the eukaryotic tran-

The ras gene family is part of a superfamily, which includes the rho and rab families, that encodes small (;20-kDa) proteins able to bind guanosine nucleotides and to participate in a variety of biochemical pathways in the eukaryotic cell (7). RAS transmits its signal via protein-protein interactions when complexed with GTP. Signaling ceases upon hydrolysis of the bound GTP to GDP by a GTPase, intrinsic to RAS, whose activity is accelerated by a GTPase-activating protein, or GAP. Signaling commences again upon replacement of the bound GDP by GTP, a process facilitated by a guanosine diphosphate dissociation stimulator, or GDS. Mutations in specific ras codons enhance the lifetime of the active, GTP-bound state and convert wild-type RAS into an unregulated active oncoprotein. RAS can be activated by an upstream sequence of events initiated by the interaction of the phosphotyrosine residues of a stimulated cell surface receptor with SH2 domains in Grb2; the SH3 domains of Grb2 in turn interact with proline-rich segments of a GDS protein like Sos to promote dissociation of GDP-RAS complexes (35). These interactions involving SH2 and SH3 domains are important because they influence both the conformation and the localization of the protein in the cell. RAS activation facilitates the attachment of c-Raf to the

* Corresponding author. Mailing address: Nelson Biological Laboratories, P.O. Box 1059, Piscataway, NJ 08855. Phone: (908) 445-4569. Fax: (908) 445-0104. Electronic mail address: denhardt@biology. rutgers.edu. † Present address: Department of Pharmacology, Cornell University Medical College, New York, NY 10028. 476

A ras-ACTIVATED ENHANCER IN THE opn PROMOTER

VOL. 15, 1995

scription factor recognition site data base (release 7.3, September, 1993), is a nuclear target for RAS action on the opn promoter, possibly via the action of a member of the ETS family of transcription factors. (Much of the research reported here was conducted by Xiaojia Guo in partial fulfillment of the requirements for a Ph.D. from Rutgers University [24b]). MATERIALS AND METHODS Cell lines and cell culture. NIH 3T3 cells (3T3 hereafter) and a series of T24 H-ras-transformed 3T3 cell lines (PAP0, PAP2, C2P0, C2P2, C5P0, and C5P2) were maintained as described previously (13, 26). In some experiments, 3T3 cells were treated with 0.1 mM sodium orthovanadate. In serum starvation experiments, the medium (Dulbecco modified Eagle medium [DMEM] plus 10% calf serum) of the subconfluent cell culture was changed to DMEM lacking serum 18 h prior to preparation of the nuclear extract. Cells were serum stimulated by the addition of fresh medium containing 10% calf serum 1 or 8 h prior to nuclear extraction. 2H1 cells (C3H 10T1/2 cells transfected with a mutated c-H-ras gene under the transcriptional control of the metallothionein I promoter [51]) were provided by T. Haliotis (Queen’s University, Kingston, Ontario, Canada) and cultured in DMEM supplemented with 10% fetal calf serum. To induce rasexpression in 2H1, subconfluent cultures were pretreated with 50 mM ZnSO4 for 24 h. Human melanoma cell lines IGR39 and IGR37 (24) were maintained in a-MEM supplemented with 10% fetal calf serum. Mouse mammary carcinoma cell lines D2.A1, D2.OR, and D2.1 were maintained in DMEM plus 10% fetal calf serum as described previously (37). Nuclear run-on assay. Gel-purified cDNA inserts of mouse b-actin, glyceraldehyde-3-phosphate dehydrogenase (GAPD), tissue inhibitor of metalloproteinase 1 (TIMP-1), calcyclin, OPN, and mitogen-regulated protein (proliferin) (MRP) were slot blotted onto a nitrocellulose membrane (2 mg/slot) in duplicate. Nuclei isolated from subconfluent cultures of 3T3 and PAP2 cells were employed in run-on reactions in the presence of [a-32P]UTP. Equal amounts (counts per minute) of radioactively labeled transcripts were incubated with the slot blots and then subjected to RNase digestion and autoradiography as described previously (2). Autoradiographic intensities were quantified by densitometric scanning and normalized to the actin signal for each preparation. Construction of recombinant DNA. Standard methods were used to subclone DNA sequences. The 59-deletion fragments of the opn gene were generated from the genomic clone of the mouse OPN gene (17) either by cleavage with restriction enzymes or by PCR amplification with appropriate primers. They were subcloned into pXP2 encoding luciferase (from the American Type Culture Collection, ATCC 37577) or pSDKlacZpA encoding b-galactosidase (provided by J. Rossant, University of Toronto). A synthetic oligonucleotide identical to the opn sequence from 2740 to 2713 or a mutant version was inserted upstream of pOpn(288/179)Luc, resulting in pRAEOpnLuc or pMutOpnLuc (see Fig. 6). Constructs containing PCR-amplified fragments or synthetic oligonucleotides were verified by sequencing using standard dideoxynucleotide techniques. Synthetic oligonucleotides were purchased from the Rutgers University-Robert Wood Johnson Medical School core facility or generously made available by G. Rodan (Merck Sharp & Dohme Research Laboratories, West Point, Pa.). Transient transfection and reporter gene product assays. 3T3 and PAP2 cells were transfected with the recombinant constructs by calcium phosphate coprecipitation (2). The basal luciferase construct pOpn(288/179)Luc was cotransfected with each b-galactosidase expression construct in order to normalize for transfection efficiency. Cells were harvested and lysed by scraping in a solution consisting of 25 mM Tris-phosphate (pH 7.8), 2 mM dithiothreitol, 2 mM 1,2diaminocyclohexane-N,N,N9,N9-tetraacetic acid (CDTA), 10% glycerol, and 1% Triton X-100. The b-galactosidase activity was measured as previously described (38); the amount of luciferase produced by the transfected cells was determined by using the Promega (Madison, Wis.) luciferase assay system and a Berthold Lumat LB9501. Measured light units were converted to femtogram of luciferase per transfection plate by using a standard curve obtained with purified luciferase (Sigma, St. Louis, Mo.). The plasmid pSV-H-ras, which expresses the H-rasVal-12 oncoprotein under the control of the simian virus T-antigen promoter, was provided by S. Powers, University of Medicine and Dentistry of New Jersey, and was cotransfected with pRAEOpnLuc or pMutOpnLuc into 3T3 cells. Formation of nuclear extracts and EMSA. Nuclear extracts were prepared (33) and electrophoretic mobility shift assays (EMSAs) were performed (45) as previously described. The DNA sequences to be tested were end labeled with 32P by using the Klenow fragment of DNA polymerase I. A typical DNA-binding reaction contained 10 mg of nuclear proteins, 104 cpm of DNA probe, 4 mg of poly(dI-dC) (Pharmacia, Piscataway, N.J.), and 10 mg of bovine serum albumin. In some EMSAs, purified proteins, such as PEA3 and GABP, were used instead of nuclear extracts. The recombinant cDNA clone encoding a glutathione-Stransferase (GST)PEA3 fusion protein was obtained from J. Hassell (McMaster University, Hamilton, Ontario, Canada), and the GST-PEA3 fusion protein was expressed in Escherichia coli and purified (53). In some EMSAs, a 100-fold molar excess of an unlabeled DNA fragment containing the transcription recognition site was included in the binding reaction.

477

Methylation interference. Methylation interference was carried out as described previously (2) with a fragment containing 12 bp of a HindIII linker sequence joined 59 of the 35-bp opn sequence (2742 to 2708), which was labeled at its 39 EcoRI site with [a-32P]dATP. The probe was methylated for 5 min with dimethyl sulfate and incubated with nuclear extracts. DNA-binding reactions were scaled up fivefold and applied onto a nondenaturing gel. Unbound probe and complex A in PAP2 cells were purified from the gel and incubated with piperidine prior to loading on a DNA sequencing gel (equivalent amounts of radioactivity were loaded on an 8% polyacrylamide–7 M urea gel). Southwestern blot analysis. Nuclear extracts were fractionated by electrophoresis in sodium dodecyl sulfate (SDS)–12% polyacrylamide gels and transferred onto polyvinylidene fluoride membranes (Immobilon-P; Millipore Corp., Bedford, Mass.). After transfer, the membranes were denatured for 10 min in a solution consisting of 6 M guanidine-HCl, 25 mM N-2-hydroxyethylpiperazineN9-2-ethanesulfonic acid (HEPES), 4 mM KCl, and 3 mM MgCl2 (pH 7.9), renatured in DNA-binding buffer (25 mM HEPES [pH 7.9], 4 mM KCl, 3 mM MgCl2), blocked with 5% nonfat milk plus 10 mg of sonicated salmon sperm DNA per ml in DNA-binding buffer for 60 min at 48C, and then washed with 0.25% nonfat milk in DNA-binding buffer for 5 min at 48C. Membranes were then incubated for 4 h at 48C in DNA-binding buffer containing 1 mM dithiothreitol, 10 mg of sonicated salmon sperm DNA per ml and the 32P-labeled opn fragment from 2740 to 2713 (106 cpm/ml). The membranes were washed four times (15 min for each wash) with DNA-binding buffer at 48C to decrease nonspecific binding. The proteins present on the membranes and able to bind the labeled DNA were visualized by autoradiography. UV cross-linking. Synthetic 6-mers complementary to the two 39 ends of the two strands making up the 2740 to 2713 duplex were used to prime synthesis by the Klenow polymerase on the separately synthesized strands of the duplex in the presence of the four deoxynucleoside triphosphates, including [a-32P]dATP and [a-32P]dGTP. The two internally labeled probe preparations were mixed together and used for protein binding and UV cross-linking as previously described (2) except that the irradiation was with two germicidal lamps at an intensity of 10 mW/cm2 for the time indicated. Nucleotide sequence accession number. The opn promoter sequence from 21361 to 2852 has been deposited in GenBank and given accession number M38399.

RESULTS OPN transcription is enhanced upon RAS activation. PAP2 cells, derived from pooled clones of 3T3 cells transformed with the human bladder cancer T24 H-ras oncogene (26), secrete more OPN than the parent 3T3 cells (12, 16). To determine whether enhanced OPN mRNA and protein levels were due to increased transcription of the gene, nuclear run-on transcription assays were performed with preparations from 3T3 and PAP2 nuclei. Radioactively labeled run-on transcripts were hybridized to filter-bound denatured cDNA clones of OPN and, for comparative purposes, actin, GAPD, calcyclin, TIMP-1, and MRP. Actin was chosen as the benchmark, since actin mRNA levels in the two cell lines were the same as judged by Northern (RNA) blot analysis (16). As shown in Fig. 1, the apparent rate of transcription of the opn (and also calcyclin) gene was increased some sixfold in the ras-transformed cells; TIMP-1 transcription was changed little, while GAPD and MRP transcription was increased to an intermediate extent relative to actin transcription. Scan of the opn promoter for a sequence responsive to ras activation. The entire mouse OPN gene and almost 1 kb of 59-flanking sequence have been characterized (5, 17, 36). Here we report additional sequence, 21361 to 2852, upstream of the opn gene. A SphI-SacI fragment (21361 to 2852) was inserted into M13 mp18 and mp19 and sequenced in both directions. This fragment overlaps with the sequence (2910 to 179) reported earlier (17). The 1.44-kb sequence (21361 to 179) was then analyzed for transcription factor-binding sites using the Genetics Computer Group program (release 6.2, August 1993) and found to possess many potential binding sites for transcription factors; some of them, including AP-1, AP-2, AP-3 (PEA3), AP-4, and AP-5, are underlined in Fig. 2. The multiplicity of potential regulatory elements suggests that expression of this gene may be controlled in complex ways in different types of cells.

478

GUO ET AL.

MOL. CELL. BIOL.

FIG. 1. Nuclear run-on assay. Nuclei isolated from subconfluent cultures of 3T3 and PAP2 cells were employed in run-on reactions in the presence of [a-32P]UTP as described previously (2). Equal amounts (counts per minute) of radioactively labeled RNA synthesized in either PAP2 or 3T3 nuclear extracts were used to probe duplicate slot blots containing 2 mg of gel-purified cDNA inserts. The intensity of each band in the autoradiograph was quantitated by densitometry and normalized to the actin signal.

To search for a sequence in the opn promoter conferring responsiveness to ras transformation in PAP2 cells, reporter plasmids containing the nested portions of the opn promoter linked to the b-galactosidase gene shown in Fig. 3A were constructed, giving rise to the pOpnLacZ series or, when linked to the firefly luciferase gene, the pOpnLuc series. Subsequent to the transfection of recombinant reporter plasmids into PAP2 and control 3T3 cells, extracts were made and assayed for reporter gene activity. Figure 3B shows the b-galactosidase activity induced by increasingly 59-extended segments of the opn promoter in both cell lines. The smallest region of the opn promoter tested, 288 to 179, was capable of stimulating expression of the reporter gene in response to RAS activation. There was an additional positive element between 2670 and 2740 of the promoter. Negative regulatory elements particularly effective in the ras-transformed cells may reside between residues 288 and 2258 and between residues 2777 and 2882. Given the relatively constant level of expression observed in 3T3 cells, we do not think the diminished expression exhibited by the longer constructs in PAP2 cells was due simply to the size of the DNA. Luciferase assays and in vitro transcription assays (data not shown) yielded very similar results and expression patterns, indicating that the differences observed were not the consequences of particular plasmid DNA preparations or reporter gene idiosyncrasies. The sequence GGAGGCAGG at 2725 to 2717 functions as a ras-activated enhancer (RAE). We dissected the regions 2740 to 2670 and 288 to 179 into smaller pieces and examined the ability of each to bind proteins in nuclear extracts made from 3T3 and PAP2 cells. The purpose was to search for a protein able to bind to a segment that was differentially expressed in PAP2 and 3T3 cells. EMSAs with PAP2 nuclear extracts and fragments from 2712 to 2670, 288 to 254, 253 to 147, and 148 to 179 produced the same gel shift patterns as did 3T3 nuclear extracts (data not shown). The 28-bp fragment from 2740 to 2713, on the other hand, formed seven reasonably discrete DNA-protein complexes, one of which, complex A, was found only in PAP2 cells (Fig. 4A); complex F was usually the most abundant and sometimes slightly more prominent in 3T3 extracts than in PAP2 extracts (see Fig. 8 also). The binding was sequence specific because an excess of unlabeled DNA (‘‘cold probe’’) of identical sequence blocked complex formation (see also Fig. 10). Methylation interference assays were undertaken in order to

define the sites of protein-DNA contact in complex A. Scaled-up binding reactions of PAP2 nuclear extracts with methylated probe (2740 to 2713) were electrophoresed on a nondenaturing gel to separate free and bound DNA. The free and retarded (in complex A) DNA molecules were purified from the gel, reacted with piperidine, and electrophoresed on a sequencing gel to fractionate the cleavage products. Missing bands reveal the position of purine residues whose methylation interfered with protein binding (Fig. 4B). Methylation of any one of the six guanine residues (also the two adenines) in the sequence GGAGGCAGG (2725 to 2717, Fig. 4C) resulted in reduced complex A formation. To confirm the significance of this protein-binding site, we used synthetic oligonucleotides to make three mutant versions of this 9-bp sequence in the 2740 to 2713 oligonucleotide: mut1, mut2, and mut3 are TTCGGCAGG, GGATTAAGG, and GGAGGCCAA, respectively (Fig. 5A). The mutated sequences were tested for protein-binding activity. As shown in Fig. 5B, mutations in the 59 GGA or the 39 AGG of the sequence from 2725 to 2717 (mut1 or mut3) abolished the formation of complexes A, B, and C; mutation of the central GGC drastically impaired or abolished formation of complexes B, E, and F, while enhancing complex G formation. The formation of complex D (always a minor species, perhaps the result of a complex with PEA3 [see Fig. 8 and 10]) did not seem to be sensitive to these 3-nucleotide changes. The biological relevance of these DNA-protein interactions was investigated with various chimeric reporter plasmids constructed by inserting the 28-bp enhancer sequence from 2740 to 2713 or a mutant version thereof (mut1, mut2, or mut3) into the HindIII site of pOpn(288/179)Luc in plasmid pXP2. The 288 to 179 portion of the opn promoter was included to provide basal promoter elements such as the CAAT and TATA boxes. DNA sequencing (not shown) confirmed the addition of the intended element and further revealed the number of inserts and their orientations. The results of transient cotransfection experiments with the RAS expression vector pSV-H-ras using these constructs in 3T3 cells are shown in Fig. 6; the number and orientation of the inserts in each construct are indicated by the arrows on the left. There was no significant difference among the constructs in the assays in the absence of activated RAS (data not shown). However, in the presence of active RAS (cotransfection with a ras expression vector), the pRAE-OpnLuc construct showed higher activity

VOL. 15, 1995


479

FIG. 2. Sequence of the opn promoter from 21361 to 179. The SphI-SacI fragment (21361 to 2852) was subcloned into M13 mp18 and mp19 and sequenced by standard dideoxynucleotide sequencing. The sequence from 2910 to 179 was reported previously (17). Upstream promoter elements and potential enhancers were identified by the GCG program in the eukaryotic transcription factor recognition site data base (release 6.2, August 1993); only sequences with no mismatch are underlined. The RAE identified here is underlined twice.

than pOpnLuc did, confirming that this enhancer sequence was sensitive to RAS activation and capable of mediating augmented transcription in cells expressing activated RAS. Therefore, we named this sequence ras-activated enhancer (RAE). Results with pRAE3- and pRAE4-OpnLuc indicated that the element could function in either orientation. In these experiments compared with the wild-type RAE, mut1 and mut3 appeared to have slightly lower enhancer activities and mut2 appeared to have slightly higher enhancer activity. These results indicate that in the context of this particular basal pro-

moter element (288 to 179) the protein complexes that can be formed (complex F by mut1 and mut3 and complexes A, C, and G by mut2 [Fig. 5]) are functional. Nuclear proteins with molecular masses of about 16, 28, 32, 45, 80 and 100 kDa can bind the RAE. To determine the protein composition of the complexes detected in the EMSAs (Fig. 4), we carried out Southwestern blotting and UV crosslinking studies. The Southwestern blot analysis shown in Fig. 7A examined individual protein-DNA interactions by probing with a 32P-labeled RAE a polyvinylidene fluoride membrane

480

GUO ET AL.

MOL. CELL. BIOL.

FIG. 3. Scan of the opn promoter for ras response elements. (A) The regions of the opn promoter used to drive expression of the reporter gene. The fragments of the opn promoter were generated from the l9090 genomic clone (17) either by cleavage with restriction enzymes or by PCR amplification with appropriate primers; they were subcloned either into pSDKlacZpA, which encodes b-galactosidase (4, 31), or into pXP1, which encodes luciferase (22). (B) b-Galactosidase assay. Cells were transfected by calcium phosphate coprecipitation and processed after 48 h for b-galactosidase activity as described in Materials and Methods. Four independent experiments were performed and found to yield the data shown; bars indicate the standard errors.

on which nuclear proteins from 3T3 and PAP2 cells, fractionated by SDS-polyacrylamide gel electrophoresis, had been blotted. As evidence for the specificity of the interaction of the RAE with the individual proteins, we demonstrated that the unlabeled RAE but not the unlabeled sequence from 288 to 254 was able to compete (data not shown). This experiment reveals that the RAE sequence is capable of binding specifically to six different proteins with apparent molecular masses of about 16, 28, 32, 45, 80 and 100 kDa in nuclear extracts of both 3T3 and PAP2 cells. On many occasions, extracts were made in the presence of various cocktails of proteinase inhibitors (some or all of the following: phenylmethylsulfonyl fluoride, leupeptin, aprotinin, pepstatin A, antipain, and L-1-tosylamide-2-phenyl-ethyl-chloromethyl ketone). In no case did we observe any difference that might be indicative of proteolysis affecting either the EMSA complexes formed or the proteins detected in Southwestern blots. UV cross-linking was performed on the complexes present in the nuclear extracts of PAP2 and 3T3 cells respectively. Figure 7B shows a time course of the cross-linking of proteins to the 32P-labeled RAE; within the resolution of this experiment (the complexes had been UV cross-linked and digested with DNase prior to electrophoresis), no prominent differences were detected in the major proteins from 3T3 and PAP2 extracts that could be cross-linked to the RAE. Proteins with apparent molecular masses of somewhat less than 20, 30 and 40 kDa were detected by virtue of the residual labeled deoxyribonucleotide still attached to each. In a different approach the individual DNA-protein complexes formed in an EMSA experiment with a PAP2 extract were separately eluted and analyzed on an SDS-polyacrylamide gel as shown in Fig. 7C; here the proteins were cross-linked to the DNA while in the gel, and the complexes were not digested with DNase. This experiment reveals the apparent sizes of the covalent DNAprotein complexes present in each of the major complexes detected by EMSA; we were unable to detect labeled proteins

FIG. 4. Protein interactions with the 2740 to 2713 element of the opn promoter. (A) EMSA of the 32P-labeled opn sequence from 2740 to 2713 with no nuclear extract (2) or with nuclear extracts (NE) made from either 3T3 (N) or PAP2 cells (P) in the absence or presence of a 100-fold molar excess of unlabeled probe sequence. Arrows marked A to G indicate reproducibly retarded complexes. (B) Complex A, which was unique to PAP2 cells, was subjected to a methylation interference study. EMSA of methylated probe (2742 to 2708) with PAP2 nuclear extract was performed. The free and bound fragments in complex A were purified from the gel, reacted with piperidine to cleave next to purines, and electrophoresed on an 8% sequencing gel. Missing bands (asterisks) reveal the positions of purine residues whose methylation interfered with protein binding. (C) Sequence of the probe used in panels A and B. The six guanine residues marked with an asterisk in the sequence from 2725 to 2717 GGAGGCAGG were protected from methylation by the formation of complex A.

VOL. 15, 1995

FIG. 5. EMSA studies with mutant RAE sequences. (A) Bases identified on the basis of methylation interference were altered in groups of three as indicated. (B) Each mutant element was tested in nuclear extracts (NE) of both 3T3 (N) and PAP2 (P) cells for its ability to form the various complexes A to G. Wt, wild type.


481

from complex E or G, perhaps because they were not present in sufficient abundance or were inefficiently cross-linked. In a parallel study of RAE-binding proteins performed with 3T3 extracts, proteins similar in size to those detected in PAP2 extracts were detected in complexes B, C, and F. Since the proteins are seen by virtue of the entire (we assume) 18-kDa 32 P-labeled DNA bound to them, the actual molecular weights of the individual proteins must be somewhat less than would otherwise be deduced from the markers. (Because it is not known whether one or both strands of the DNA remain associated with the protein after UV irradiation and boiling in SDS, there is an element of imprecision in our estimation of the molecular weights of the associated proteins.) From these experiments we conclude that at least complexes A and F are defined by interactions of the RAE with unique proteins of about 16 and 100 kDa, respectively. The protein composition of the complexes of intermediate mobility (B to D) are harder to define unambiguously but appear to represent interactions with one or more proteins in the range of 25 to 100 kDa. Detection of protein factors interacting with RAE in many cell lines and correlation of the ability to form complex A with metastatic potential. To determine whether the RAE-binding proteins detected in 3T3 and PAP2 cells (Fig. 4) are expressed in other cell types, we examined nuclear extracts from more than 30 cell lines by EMSAs. Nuclear extracts were prepared from these cell lines and screened for the presence of complexes A to G by incubation with a labeled RAE sequence. Figure 8 shows the results obtained from some of these cell lines. PAP0 is a pooled population of ras-transformed 3T3 cells that is metastatic both in the chicken embryo and in nude mice. PAP2 cells are a more highly metastatic population selected by two passages through the chicken embryo for the ability to metastasize to the liver (26). C2P0 and C5P0 are two clonal lines that are poorly metastatic; C2P2 and C5P2 were derived from C2P0 and C5P0, respectively, by two rounds of selection

FIG. 6. Abilities of mutant RAEs and multiple RAEs to activate a ‘‘basal’’ promoter. The pXP2 luciferase expression plasmid was used together with the 288 to 179 ‘‘basal’’ opn promoter to assess the potential of the indicated elements, RAE, or its mutated sequences, to drive transcription and expression of luciferase. Arrows indicate the orientation of the element; the number represents the copy number of inserted elements. Luciferase constructs were cotransfected with a ras expression vector, pSV-H-ras, into 3T3 cells by calcium phosphate coprecipitation. After 40 h of transfection, cells were lysed and the luciferase activity of each lysate was measured. Quadruplicate transfections were averaged to generate the values and standard errors shown. A second experiment with different plasmid preparations yielded a similar pattern.

482

GUO ET AL.

FIG. 7. Analysis of RAE-binding proteins. (A) Southwestern blot analysis of proteins interacting with RAE. Nuclear extracts of 3T3 (N) and PAP2 (P), along with 14C-labeled protein markers were electrophoresed and blotted as described in Materials and Methods. Proteins on the membrane were denatured, renatured, and incubated with 32P-labeled RAE probe. Molecular mass (M) markers are indicated on the left; arrows on the right point to the locations of proteins able to bind RAE. (B and C) Proteins bound to RAE as detected by cross-linking with UV radiation. In panel B, the duplex 28-mer RAE (Fig. 4C) uniformly labeled with 32P was incubated with nuclear extracts from 3T3 (N) and PAP2 (P) cells and then subjected to UV radiation for 0, 5, and 30 min. After DNase I and micrococcal nuclease digestion, the proteins were fractionated on a SDS–10% polyacrylamide gel, which was then dried and autoradiographed. The markers are 43 kDa (ovalbumin), 29 kDa (carbonic anhydrase), 18.4 kDa (b-lactoblobulin), and 14.3 kDa (lysozyme). In panel C, the complexes formed between the 32 P-labeled RAE and proteins in a PAP2 nuclear extract were electrophoresed in an EMSA gel, which was then subjected to UV radiation for 30 min. Gel strips containing the complexes were excised, and the individual complexes were eluted and fractionated (without nuclease digestion) on a SDS–12% polyacrylamide gel, which was then dried and autoradiographed. Molecular mass (M) markers are indicated on the left, and the approximate apparent molecular masses of the DNA-protein complexes are indicated on the right.

for metastatic capability as described above. The relative levels of RAS expression in these various cell populations, determined by immunoblotting, and their metastatic potentials are PAP2 . C2P2 . C5P2 . PAP0 . C2P0 . C5P0 . 3T3 (13,

MOL. CELL. BIOL.

26). Except for PAP0, the abundance of complex A increased and that of complex F decreased, roughly in proportion to the RAS level and the metastatic potential. The wider generality of the result that complex A was formed most abundantly in highly metastatic cells was established with the ras-transformed derivative of C3H 10T1/2 cells (2H1), human melanoma tumor cell lines IGR37 and IGR39, and mouse mammary carcinoma cell lines D2.A1, D2.1, and D2.OR (Fig. 8, lanes 9 to 15). 2H1 has the ras oncogene under the control of the metallothionein promoter and its ras expression is inducible (32). In these cells there was an increase in the abundance of the protein giving rise to complex A when ras expression was induced by activating the metallothionein promoter with Zn21. IGR37 and IGR39 were isolated from a metastatic melanoma and the primary melanoma, respectively, from the same patient. They possess very different properties in vitro and in vivo; IGR37 is considerably more metastatic than IGR39, which is relatively normal (13a). D2.A1, D2.1, and D2.OR are three related murine mammary carcinoma cell lines of recent origin (37). They have spontaneously arrived at different stages of tumor progression. It is noteworthy that D2.A1, which is the most metastatic and most invasive, formed only complex A. D2.1 and D2.OR are similar to each other, with D2.1 in some cases being somewhat more malignant than D2.OR, determined by experimental and spontaneous metastasis assays (37). Nuclear extracts made from cells ranging from nontumorigenic primary mouse embryo fibroblasts and proximal tubule epithelial cells to various 3T3 lines and nonmetastatic human melanoma cells yielded band shift patterns with RAE similar to that obtained with the 3T3 cells (data not shown). Complex A was not detected in LTA and LTA-based mouse cell lines (Fig. 8, lanes 16 to 18). LTA cells (mouse thymidine kinase-negative and adenine phosphoribosyl transferase-negative L cells) are tumorigenic but not metastatic, and remain so when transfected with ras (52). Neo and P1rasP2 are LTA cells transfected with pSV2neo and pSV2neo-T24-H-ras, respectively. Both are nonmetastatic and unable to form complex A. Because of the good correlation between the capacity to form complex A and the metastatic potential of the cell, we have assigned the acronym MATF (metastasis-associated transcription factor) to the protein responsible for the formation of complex A. The ;16-kDa protein seen in the Southwestern blot of PAP2 and 3T3 nuclear proteins appears likely to be the MATF on the basis of the cross-linking study shown in Fig. 7C. Induction of complex A formation by transient RAS transfection and serum stimulation. We studied how RAS might determine MATF-directed activation of the RAE by examining the EMSA pattern produced by extracts of 3T3 cells transfected with a RAS expression vector (Fig. 9, lane 4). Neither a calcium phosphate precipitate (lane 2) nor the pSV2CAT vector (lane 3) caused complex A formation in 3T3 cells. However, expression of activated RAS specifically altered the EMSA pattern: complex A appeared, the abundance of complex C was increased, and complexes B and D to G disappeared. The origin of the new complex between D and E (lanes 4 and 7) is unknown. There was a transient appearance of complex A after serum stimulation of the 3T3 cells (lanes 6 to 8). The EMSA pattern of 3T3 cells transfected with the ras expression vector is different from that of ras-transformed cell lines such as PAP2 and C2P2 (compare Fig. 9, lane 4 with Fig. 8). We speculate that this is because activated RAS is expressed to an unusually high level in the transiently transfected cells compared with permanently transfected lines. Relation of MATF to other transcription factors. The sequence GGAGGCAGG is consistent with the hypothesis that

VOL. 15, 1995


483

FIG. 8. EMSA patterns of the RAE observed with extracts of various metastatic and nonmetastatic lines. Nuclear extracts prepared from subconfluent cultured cells (see Materials and Methods) were incubated with a 32P-labeled RAE sequence at room temperature for 15 min and then subjected to electrophoresis as described in the legend to Fig. 4. Nonmetastatic lines are marked with an asterisk. NO NE, no nuclear extract.

the MATF is a member of the ETS superfamily of GA-binding transcription factors (30); this family includes AP-2 (also known as PEBP2 or PEA2), AP-3 (PEA3), GABP, and ETS1-3. To determine if the MATF was the same as any one of these, a series of competition experiments was set up as described in the legend to Fig. 10. The results revealed that sequences capable of binding the GABP, PEBP2, and ETS1-3 proteins were unable to compete effectively for the proteins that interacted with RAE. Confirmation of the conclusion that these proteins were not responsible for the formation of complex A in particular came from DNA-protein EMSAs using (i) the RAE with purified PEA3 and GABP proteins (no binding) and (ii) nuclear extracts from PAP2 and 3T3 cells with PEBP2, GABP, and AP-1 probes (the complexes formed differed from the RAE complexes). We also tested whether antibodies to GABP or ETS1-3 could interact with complex A; neither of these antibodies caused any change in the mobility (‘‘supershift’’) of the complexes we have characterized here (data not shown). A minor species, complex D, may contain PEA3, because purified PEA-3 protein can interact with RAE and form a complex that migrates in the same general region in a native gel (Fig. 10, PEA3 lane). Seven of nine bases in the opn sequence from 2728 to 2720 (AGTGGAGGC) match the ets1-binding site AGCGGA AGCG in the Moloney murine sarcoma virus long terminal repeat (40). The ETS1 oncoprotein, which exists in several different isoforms and may be regulated both by serine phosphorylation and by the redox state of the cell, binds as a monomer to oligonucleotides containing the GGAA sequence that is found in the PEA3 motif (23, 48). Satake et al. (45) reported a 24-bp sequence in polyomavirus DNA that interacted with polyomavirus enhancer-binding protein 2 (PEBP2), which is a target of H-ras. PEBP2 is present in 3T3 cells and converts into a faster-migrating factor, termed PEBP3, in 3T3 cells transformed by the activated H-ras oncogene (45). Purified PEBP3 is a heterodimer composed of a (30- to 35-kDa)

FIG. 9. EMSA of the RAE with nuclear extracts made from 3T3 cells under various conditions. Free probe (lane 1), mock-transfected cells (lane 2), cells transfected with the control vector pSV2CAT (lane 3) or with pSV-H-ras (lane 4) for 40 h prior to nuclear extract preparation, and cells serum starved for 18 h (lane 5) and then stimulated with serum for 8 (lane 6) or 1 h (lane 7) prior to preparation of nuclear extracts are shown. In lane 8, cells were treated with sodium orthovanadate at final concentration of 0.1 mM for 1 h prior to extraction.

484

GUO ET AL.

MOL. CELL. BIOL.

FIG. 10. Inability of selected response elements to compete with RAE. Nuclear extracts were made from PAP2 cells, combined first with a 1003 excess of unlabeled competitor and then with 32P-labeled RAE. After a 15 min of incubation, the mixture was electrophoresed. The PEA3 lane shows a gel shift with the purified PEA3 protein and RAE. The PEA3 protein was a fusion containing a portion of the GST (53). The ETS 1-3 (59-GATCTCGAGGCGGAAGTTCG A-39) sequence was obtained from R. Fisher; the GABP (59-CAACGGAGCG GAAACCGCCG-39) sequence was obtained from T. Brown; PEBP2 (59-GTAA CTGACCGCAGAGGGC-39) and its mutant (59-GTAACTCACGGCAGAG GGC-39) were obtained from Y. Ito; the AP-1 (59-CGTGACTCAGCGCGC-39) was obtained from S. Rittling; the opn sequence from 288 to 254 (59-GCAAA AACCTCATGACACATCACTCCACCTCCTG-39) was synthesized in the Rutgers University-Robert Wood Johnson Medical School core facility and does not contain the RAE sequence. In all cases, duplex oligonucleotides were actually tested even though only the sequence of one strand is shown. 2, no cold competitor (control).

and b (20- to 25-kDa) subunits (28). The central CCPuC of the PEBP2-binding sequence (Pu/T)ACCPuCA (Pu stands for purine) is particularly important for PEBP2 binding (27a). Southwestern blotting with a PEBP2 probe (not shown) revealed five proteins, three of which had mobilities similar to those of proteins able to bind RAE. Because of the different binding specificities and lack of competition with the factor we have identified here, we conclude that MATF is distinct from ETS1-3, GABP, PEBP2, and AP-1. DISCUSSION Osteopontin is expressed in several different types of cells in response to specific signals and likely is important in more than one context in the development and functioning of the mammalian organism (21). Thus, it is likely that transcription of the gene is subject to a number of different control elements, reflecting the diversity of cell types in which expression must be controlled. Consistent with this thought is the fact that a large number of potential recognition motifs exist in the promoterenhancer region. In this work we have identified one of the links that connect RAS to a change in gene expression regulated at the transcriptional level. The good agreement between the increase in opn transcription measured in the nuclear run-on (6.4-fold) and the increase in mRNA abundance determined from Northern

blots (8.5-fold) suggests that increased transcription largely accounts for the increased expression. Since a portion of the increased transcription is determined by regions of the promoter not able to bind to the transcription factor detected here, we conclude that this binding protein is one of several contributors to the overall increase in the rate of transcription of opn. Transient transfection with pOpn(288/179)LacZ suggests that the region from 288 to 179 of the opn promoter, which contains potential AP-1, -2, and -5 sites, contributes to increased expression in ras-transformed cells. Although c-jun mRNA levels are reduced in the ras-transformed PAP2 cells (52), an increased rate of opn transcription could still result from an increase in the activity of FOS-JUN dimers. There was little difference in the EMSA patterns formed by incubation of PAP2 and 3T3 nuclear extracts with this region or its subfragments from 288 to 254, 253 to 147, and 148 to 179 (data not shown). Transient transfection assays indicated that there might also be several negative transcriptional regulators binding to elements in the promoter; we have not studied further this negative regulation. The sequence GGAGGCAGG in the opn promoter is a novel ras-activated enhancer. The sequence from 2740 to 2713 in the opn promoter conferred increased transcriptional activity in ras-transformed 3T3 cells and when cotransfected with a ras expression vector into 3T3 cells. It was able to form several distinct protein complexes (A to G), one of which (A) was dependent upon activation of RAS for its formation. Methylation interference studies revealed the importance of the core binding site was GGAGGCAGG at 2725 to 2717, and mutation analysis of this sequence confirmed that formation of complexes A to C, E, and F was dependent upon it. UV cross-linking experiments suggested that the formation of complexes A and F was dependent upon interactions with proteins of about 16 and 100 kDa, respectively. By Southwestern blotting we were able to detect six RAE-binding proteins, two of which had apparent molecular weights consistent with those that we had determined for isolated complexes A and F. The expression cloning of the mRNAs encoding these proteins that is currently under way should help us understand the complex regulation that is taking place in this portion of the OPN promoter. The G-rich sequence GGAGGCAGG is similar (6 of 9 bp) to the GABP-binding site GGAAGCGGAAG in the enhancer of the herpes simplex virus ICP4 gene (49). Unlike many DNA-protein interactions in which the DNA sequence possesses dyad symmetry and interacts with protein dimers, GABP, an a2b2 tetramer, binds to a purine-rich motif in the ICP4 enhancer that is present as a direct repeat. Since the MATF forms a complex with RAE different from the ETS family members tested (Fig. 10), we suggest that it is a novel member of the ETS family. Indicative of a potentially wide significance, we note in Fig. 11 that the recognition motif GGANNNAGG or its complement occurs in other enhancers, for example, the Pit-1-pGHF1 factor that is involved in mediating basal and regulated expression of the rat growth hormone gene by thyroid hormone (54). Several sequences in different genes have been reported to be ras response elements. Imler et al. (27) showed that the polyomavirus PEA1-binding site CTGACTCA, which is equivalent to an AP-1 site, is ras responsive. This sequence also mediated activation by TPA and serum. These results suggested that the RAS signal was transmitted from outside the cell to a transcription factor in the nucleus by a protein kinase C-dependent signal transduction pathway. The sequence TGA CTCT in VL30, also present in transforming growth factor b1 and NVL-30, and the sequence GTGACGTCAC in the human

VOL. 15, 1995


485

FIG. 11. Enhancers containing the GGANNNAGG motif or its complementary CCTNNNTCC. The sequences are taken from the following sources: osteopontin (this work and reference 25a), deltaE1 (44), T-Ag-SV40-III (20), ovotransferrin S element 1 (34), Pit-1-pGHF1 (24a), IgPE-1-IgH (46), CuE2.2 (1), GR-HMTIIa-II (3), and AP1-JE (50). Two and four motifs were found in AP1-JE and ovotransferrin S element 1, respectively.

DNA b-polymerase promoter were demonstrated to be ras response elements (29, 41, 42). However, deletion of the ras response element in the VL30 gene did not affect stimulation by TPA, and the binding of a protein in nuclear extracts to the ras response element in the human b-polymerase promoter was not inhibited by an oligonucleotide corresponding to an AP-1-binding element. All three of the above sequences have in common a TGAC sequence that may indicate a common site of interaction. Taken together, these results suggest that the RAS and TPA signal transduction mechanisms for transcriptional activation are distinct but possibly overlapping in some cell types (8, 11, 14). Relation of complex A formation to RAS activation and metastasis. We were disappointed that the Southwestern blots did not reveal a protein absent in the 3T3 cells but present in the ras-transformed derivative, PAP2. This result would have been consistent with the idea that in the ras-transformed cells a protein that was responsible for complex A formation had been induced. Given that this idea does not seem to be true, we have developed the hypothesis diagrammed in Fig. 12 that the complex-forming and transcription-regulating ability of the 16kDa protein responsible for complex A formation is controlled by an inhibitor protein (X) that is regulated via the signaling pathway under the control of RAS. We suggest that under one set of conditions, in 3T3 cells for example, the 16-kDa transcriptional activator is prevented from interacting with the RAE sequence by the inhibitor protein. We further speculate that under other conditions (after stimulation with serum or activation of RAS or in the presence of orthovanadate) protein X is modified by phosphorylation so that it can no longer prevent the 16-kDa protein from forming complex A with RAE. As a consequence of this interaction, transcription of the OPN gene, and perhaps other genes, is enhanced. Another possibility is that the action of RAS is to increase the activity of the RAE-binding protein, for example by phosphorylation. However, various experiments designed to detect phosphorylation, tyrosine phosphorylation in particular, of MATF itself were negative. Of complexes A to G that were detected in many cell lines, complex A appeared the most reliable indicator of the metastatic phenotype. It was found in cell lines independently selected for enhanced metastatic capability, such as mouse PAP2, C2P2, C5P2, D2.A1, D2.1, D2.OR, and

FIG. 12. Simplified model for the regulation of complex A formation by RAS. The Southwestern blot analysis suggests that the set of proteins capable in principle of interacting with RAE are the same in 3T3 cells and PAP2 cells. The protein responsible for complex F formation, some 100 kDa in size, interacts particularly with the central GGC of the RAE site in the opn promoter; our results suggest that it may exert negative control. The protein responsible for complex A formation, some 16 kDa in size, interacts with the GGA and AGG sequences bracketing the central GGC. Because complex A was detected only in cells containing activated RAS, or when the cells were stimulated with serum or treated with orthovanadate, we suggest that in the absence of any of these perturbations the ;16-kDa protein is complexed with a protein (X) that blocks it from forming complex A. That orthovanadate, a phosphatase inhibitor, can alleviate the block suggests that phosphorylation may be the mechanism regulating the activity of X, although other possibilities cannot be excluded. The strong correlation between the presence of complex A and the metastatic phenotype lead us to propose the name metastasis-associated transcription factor (MATF) for this RAE-binding protein. Since the binding sites for the ;16- and ;100-kDa proteins overlap, there must clearly be some interaction or competition between the two proteins. The other proteins able to bind to the RAE and to participate in the formation of the other complexes have been omitted from the diagram for simplicity.

human IGR37. Consistently, we did not detect complex A in the ras-transformed LTA lines, including NEO and P1rasP2 cells. LTA is a tumorigenic, but nonmetastatic, cell line whose degree of malignancy is not altered by RAS expression; furthermore, opn expression is repressed by RAS in these cells (52). The good correlation between the presence of complex A and the metastatic phenotype suggests that this putative metastasis-associated transcription factor MATF could control (directly or indirectly) one or more genes important to the metastatic process (12). The ability of a tumor extract to form complex A may be useful clinically as a prognostic indicator and could serve as a target for developing chemotherapeutic agents specific for preventing RAS from participating in tumor progression and metastasis. ACKNOWLEDGMENTS This research was supported by grants from the National Cancer Institute of Canada (to A.F.C.) and by funds from Rutgers University, the Charles and Johanna Busch Endowment, and the U.S. Public Health Service (to D.T.D.). A.F.C. is a Career Scientist of the Ontario Cancer Treatment and Research Foundation. X.G. was supported by a Medical Research Council of Canada Studentship, a Rutgers University Graduate Assistantship, and the Anne B. and James H. Leathem Scholarship Fund. Y.P.Z. was supported by a Rutgers-

486

GUO ET AL.

MOL. CELL. BIOL.

UMDNJ Core Curriculum Graduate Fellowship. D.A.M. was supported by a Henry Rutgers Undergraduate Fellowship. We thank J. Rossant for making pSDKlacZpA available; T. Brown for gifts of DNA oligomers containing the GABP site, purified proteins GABPa and GABPb, and antibodies against GABP; S. Powers for the gift of pSV-H-ras; T. Haliotis for providing the 2H1 cells; J. Hassell for providing PEA3; Y. Ito for providing PEBP2; R. J. Fisher for providing ETS1-3; S. Rittling for providing AP-1; S. Walther and K. Curtis for help in preparing the manuscript; M. Noda and G. Rodan for gifts of oligonucleotides; and S.-M. Hwang for assistance. REFERENCES 1. Augereau, P., and P. Chambon. 1986. The mouse immunoglobulin heavychain enhancer: effect on transcription in vitro and binding of proteins present in HeLa and lymphoid B cell extracts. EMBO J. 5:1791–1797. 2. Ausubel, F. M., R. Brent, R. E. Kingston, D. D. Moore, J. G. Seidman, J. A. Smith, and K. Struhl. 1989. Current protocols in molecular biology. Greene Publishing Associates and Wiley-Interscience, New York. 3. Beato, M., J. Arnemann, G. Chalepakis, E. Slater, and T. Willmann. 1987. Gene regulation by steroid hormones. J. Steroid Biochem. 27:9–14. 4. Beckwith, J. R. 1978. lac: the genetic system, p. 11–30. In J. H. Miller and W. S. Reznikoff (ed.), The operon. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, N.Y. 5. Behrend, E. I., A. F. Chambers, S. M. Wilson, and D. T. Denhardt. 1993. Comparative analysis of two alternative first exons reported for the mouse osteopontin gene. J. Biol. Chem. 268:11172–11175. 6. Blenis, J. 1993. Signal transduction via the MAP kinases: proceed at your own RSK. Proc. Natl. Acad. Sci. USA 90:5889–5892. 7. Bokoch, G. M., and C. J. Der. 1993. Emerging concepts in the ras superfamily of GTP-binding proteins. FASEB J. 7:750–759. 8. Bollag, G., and F. McCormick. 1991. Regulators and effectors of ras proteins. Annu. Rev. Cell Biol. 7:601–632. 9. Bortner, D. M., S. J. Langer, and M. C. Ostrowski. 1993. Non-nuclear oncogenes and the regulation of gene expression in transformed cells. Crit. Rev. Oncog. 4:137–160. 10. Bruder, J. T., G. Heidecker, and U. Rapp. 1992. Serum-, TPA-, and RASinduced expression from Ap-1/ETS-driven promoters requires Raf-1 kinase. Genes Dev. 6:545–556. 11. Cantley, L. C., K. R. Auger, C. Carpenter, B. Duckworth, A. Graziani, R. Kapeller, and S. Soltoff. 1991. Oncogenes and signal transduction. Cell 64: 281–302. 12. Chambers, A. F., E. I. Behrend, S. M. Wilson, and D. T. Denhardt. 1992. Induction of expression of osteopontin (OPN; secreted phosphoprotein) in metastatic, ras-transformed NIH3T3 cells. Anticancer Res. 12:43– 48. 13. Chambers, A. F., G. H. Denhardt, and S. M. Wilson. 1990. RAS-transformed NIH3T3 cell lines, selected for metastatic ability in chick embryos, have increased proportions of p21-expressing cells and are metastatic in nude mice. Invasion Metastasis 10:225–240. 13a.Chambers, A. F., and J. F. Harris. Unpublished results. 14. Chambers, A. F., and A. B. Tuck. 1993. Ras-responsive genes and tumor metastasis. Crit. Rev. Oncog. 4:95–114. 15. Coffer, P., M. de Jonge, A. Mettouchi, B. Binetruy, J. Ghysdael, and W. Kruijer. 1994. junB promoter regulation: Ras mediated transactivation by cEts-1 and cEts-2. Oncogene 9:911–922. 16. Craig, A. M., G. T. Bowden, A. F. Chambers, M. A. Spearman, A. H. Greenberg, J. A. Wright, M. McLeod, and D. T. Denhardt. 1990. Secreted phosphoprotein mRNA is induced during multi-stage carcinogenesis in mouse skin and correlates with the metastatic potential of murine fibroblasts. Int. J. Cancer 46:133–137. 17. Craig, A. M., and D. T. Denhardt. 1991. The murine gene encoding secreted phosphoprotein 1 (osteopontin): promoter structure, activity, and induction in vivo by estrogen and progesterone. Gene 100:163–171. 18. Craig, A. M., J. H. Smith, and D. T. Denhardt. 1989. Osteopontin, a transformation-associated cell adhesion phosphoprotein, is induced by 12-O-tetradecanoylphorbol 13-acetate in mouse epidermis. J. Biol. Chem. 264:9682– 9689. 19. Crews, C. M., and R. L. Erikson. 1993. Extracellular signals and reversible protein phosphorylation: what to Mek of it all. Cell 74:215–217. 20. DeLucia, A. L., B. A. Lewton, R. Tjian, and P. Tegtmeyer. 1983. Topography of simian virus 40 A protein-DNA complexes: arrangement of pentanucleotide interaction sites at the origin of replication. J. Virol. 46:143– 150. 21. Denhardt, D. T., and X. Guo. 1993. Osteopontin—a protein of many functions. FASEB J. 7:1475–1482. 22. DeWet, J. R., K. V. Wood, M. DeLuca, D. R. Helinski, and S. Subramani. 1987. Firefly luciferase gene: structure and expression in mammalian cells. Mol. Cell. Biol. 7:725–737. 23. Fisher, R. J., S. Koizumi, A. Kondoh, J. M. Mariano, G. Mavrothalassitis, N. K. Bhat, and T. S. Papas. 1992. Human ETS1 oncoprotein: purification,

isoforms, -SH modification, and DNA sequence specific binding. J. Biol. Chem. 267:17957–17965. 24. Foa, C., and C. Aubert. 1977. Ultrastructural comparison between cultured and tumor cells of human malignant melanoma. Cancer Res. 37:3957–3963. 24a.Genetics Computer Group. 1993. Genetics Computer Group data reference set, vol. 3. Data file 1993, p. 77. Genetics Computer Group, Madison, Wis. 24b.Guo, X. 1994. Ph.D. dissertation. Rutgers University, Piscataway, N.J. 25. Hall, A. 1994. A biochemical function for Ras—at last. Science 264:1413– 1414. 25a.Hijiya, N., M. Setoguchi, K. Matsuura, Y. Higuchi, A. Akizuki, and S. Yamamoto. 1994. Cloning and characterization of the human osteopontin gene and its promoter. Biochem. J. 303:255–262. 26. Hill, S. A., S. M. Wilson, and A. F. Chambers. 1988. Clonal heterogeneity, experimental metastatic ability, and p21 expression in H-ras-transformed NIH3T3 cells. J. Natl. Cancer Inst. 80:484–490. 27. Imler, J. L., C. Schatz, C. Wasylyk, B. Chatton, and B. Wasylyk. 1988. A Harvey-ras responsive transcription element is also responsive to a tumourpromoter and to serum. Nature (London) 332:275–278. 27a.Ito, Y. Personal communication. 28. Kamachi, Y., E. Ogawa, M. Asano, S. Ishida, Y. Murakami, M. Satake, Y. Ito, and K. Shegesada. 1990. Purification of a mouse nuclear factor that binds to both the A and B cores of the polyomavirus enhancer. J. Virol. 64: 4808–4819. 29. Kedar, P. S., D. R. Lowy, S. G. Widen, and S. H. Wilson. 1990. Transfected human b-polymerase promoter contains a ras-responsive element. Mol. Cell. Biol. 10:3825–3856. 30. Klemsz, M. J., S. R. McKercher, A. Celada, C. Van Beveren, and R. A. Maki. 1990. The macrophage and B cell-specific transcription factor PU.1 is related to the ets oncogene. Cell 61:113–124. 31. Kothary, R., S. Clapoff, S. Darling, M. D. Perry, L. A. Moran, and J. Rossant. 1989. Inducible expression of an hsp68-lacZ hybrid gene in transgenic mice. Development 105:707–714. 32. Krook, A., M. J. Rapoport, S. Anderson, H. Pross, Y. C. Zhou, D. T. Denhardt, T. L. Delovitch, and T. Haliotis. 1992. p21ras and protein kinase C function in distinct and interdependent signaling pathways in C3H 10T1/2 fibroblasts. Mol. Cell. Biol. 13:1471–1479. 33. Lee, K. A. W., A. Bindereif, and M. G. Green. 1988. A small-scale procedure for preparation of nuclear extracts that support efficient transcription and pre-mRNA splicing. Gene Anal. Tech. 5:22–31. 34. Maroteaux, L., R. Heilig, D. Dupret, and J. L. Mandel. 1983. Repetitive satellite-like sequences are present within or upstream from 3 avian proteincoding genes. Nucleic Acids Res. 11:1227–1243. 35. McCormick, F. 1993. How receptors turn RAS on. Nature (London) 363: 15–16. 36. Miyazaki, Y., M. Setoguchi, S. Yoshida, Y. Higuchi, S. Akizuki, and S. Yamamoto. 1990. The mouse osteopontin gene: expression in monocytic lineages and complete nucleotide sequence. J. Biol. Chem. 265:14432– 14438. 37. Morris, V. L., A. B. Tuck, S. M. Wilson, D. Percy, and A. F. Chambers. 1993. Tumor progression and metastasis in murine hyperplastic alveolar nodule mammary tumor cell lines. Clin. Exp. Metastasis 11:103–112. 38. Neilson, D. A., J. Chou, A. J. MacKrell, M. J. Casadaban, and D. F. Steiner. 1983. Expression of a preproinsulin-b-galactosidase gene fusion in mammalian cells. Proc. Natl. Acad. Sci. USA 80:5198–5202. 39. Noda, M., R. L. Vogel, A. M. Craig, J. Prahl, H. F. DeLuca, and D. T. Denhardt. 1990. Identification of a DNA sequence responsible for binding of the 1,25-dihydroxyvitamin D3 enhancement of mouse secreted phosphoprotein 1 (Spp-1 or osteopontin) gene expression. Proc. Natl. Acad. Sci. USA 87: 9995–9999. 40. Nye, J. A., J. M. Petersen, C. V. Gunther, M. D. Jonsen, and B. J. Graves. 1992. Interaction of murine ETS-1 with GGA-binding sites establishes the ETS domain as a new DNA-binding motif. Genes Dev. 6:975–990. 41. Owen, R. D., D. M. Bortner, and M. C. Ostrowski. 1990. ras oncogene activation of a VL30 transcriptional element is linked to transformation. Mol. Cell. Biol. 10:1–9. 42. Owen, R. D., and M. C. Ostrowski. 1990. A nuclear factor that binds to ras-responsive enhancer elements is present in human cells. Cell Growth Differ. 1:601–606. 43. Patarca, R., R. A. Saavedra, and H. Cantor. 1993. Molecular and cellular basis of genetic resistance to bacterial infection: the role of early T-lymphocyte activation-1/osteopontin gene. Crit. Rev. Immunol. 13:225–246. 44. Redondo, J. M., S. Hata, C. Brocklehurst, and M. S. Krangel. 1990. A T cell-specific transcriptional enhancer within the human T cell receptor d locus. Science 247:1225–1229. 45. Satake, M., T. Ibaraki, and Y. Ito. 1988. Modulation of polyomavirus enhancer binding proteins by Ha-ras oncogene. Oncogene 3:69–78. 46. Schlokat, U., D. Bohmann, H. Scholer, and P. Gruss. 1986. Nuclear factors binding specific sequences within the immunoglobulin enhancer interact differentially with other enhancer elements. EMBO J. 5:3251–3258. 47. Senger, D. R., C. A. Perruzzi, and A. Papadopoulos. 1989. Elevated expression of secreted phosphoprotein 1 (osteopontin, 2ar) as a consequence of neoplastic transformation. Anticancer Res. 9:1291–1300.

VOL. 15, 1995 48. Seth, A., R. Ascione, R. J. Fisher, G. J. Mavrothalassitis, N. K. Bhat, and T. S. Papas. 1992. The ets gene family. Cell Growth Differ. 3:327–334. 49. Thompson, C. C., T. A. Brown, and S. L. McKnight. 1991. Convergence of ets- and notch-related structural motifs in a heteromeric DNA binding complex. Science 253:762–768. 50. Timmers, H. T. M., G. J. Pronk, J. L. Bos, and A. J. van der Eb. 1990. Analysis of the rat JE gene promoter identifies an AP-1 binding site essential for basal expression but not for TPA induction. Nucleic Acids Res. 18:23–36. 51. Trimble, W. S., P. W. Johnson, N. Hozumi, and J. C. Roder. 1986. Inducible cellular transformation by a metallothionein-RAS hybrid oncogene leads to


487

natural killer cell susceptibility. Nature (London) 321:782–784. 52. Tuck, A. B., S. M. Wilson, R. Khokha, and A. F. Chambers. 1991. Different patterns of gene expression in RAS-resistant and RAS-sensitive cells. J. Nat. Cancer Inst. 83:485–491. 53. Xin, J.-H., A. Cowie, P. Lachance, and J. A. Hassell. 1992. Molecular cloning and characterization of PEA3, a new member of the ets oncogene family that is differentially expressed in mouse embryonic cells. Genes Dev. 6:481–496. 54. Ye, Z.-S., and H. H. Samuels. 1987. Cell- and sequence-specific binding of nuclear proteins to 59-flanking DNA of the rat growth hormone gene. J. Biol. Chem. 262:6313–6317.