Transcription of the Bone Sialoprotein Gene Is Stimulated by v-Src Acting through an Inverted CCAAT Box¹

Morbidity associated with cancer often occurs as a consequence of the metastasis of tumor cells to secondary sites. The formation of skeletal metastases occurs in the majority of patients with advanced breast, prostate, and thyroid cancers (1), reflecting a selective affinity for mineralized tissues. However, the basis of osteotropism of metastatic cells from these tumors is unknown. Recent studies have demonstrated that BSP,³ a glycoprotein that is essentially unique to mineralizing tissues, is expressed in the majority of human breast (2, 3) and lung (4) cancer lesions. Moreover, in breast carcinomas, BSP expression has been linked to poor prognosis, independent of lymph node status (5), and is expressed on the cell surface of estrogen-receptor-positive cells that metastasize to bone with high frequency (6).

BSP is a phosphorylated (ser/thr) and sulfated (tyr) glycoprotein that is produced almost exclusively in mineralizing tissues (reviewed in Ref. 7), where it is believed to nucleate and bind to hydroxyapatite crystals through two to three stretches of polyglutamic acid (8). BSP can also mediate the attachment of bone cells to mineralized tissue surfaces through an RGD motif that recognizes the α_vβ₃ integrin (9). Expression of BSP at ectopic sites has been associated with mineral crystal formation, including the formation of microcrystalline deposits of hydroxyapatite in breast and lung tumors (2, 4). Notably, the expression of OPN, another bone matrix protein with similar properties to BSP, including the presence of hydroxyapatite and cell binding motifs, has also been associated with cellular transformation (10, 11). Although the expression of OPN has been shown to be regulated by several tumor promoters including v-ras (12) and v-src (13), mechanisms of induction of BSP in transformed cells have not been reported.

Over-expression of EGF family members (14) or the erB-2 (neu) receptor (15, 16, 17) induce neoplastic growth in mammary glands of transgenic mice through a mechanism requiring the physical association of the activated neu receptor with the proto-oncogene c-src (15). Although mutations of the src proto-oncogene alone is not sufficient to induce mammary tumors (17), over-expression of c-src is, nevertheless, a key step in oncogenesis (18). Moreover, the v-src gene is responsible for the potent transforming activity of the Rous sarcoma virus. Cellular transformation caused by expression of the constitutively active src oncogene induces transcription of genes coding for 9E3/CEF-4, collagenase, TGF-β, c-Fos, JunB, and matrix metalloproteinase-9, which have been implicated in tumor growth and metastasis (19). In normal cells, Src is involved in the regulation of a number of signal transduction pathways, including those that target cell division and/or cell activation, recovery from oxidative stress, and cytoskeletal arrangement.

To determine whether oncogenes associated with the formation of tumors that frequently metastasize to bone influence the expression of BSP, we have studied the transcriptional regulation of BSP by v-Src. Here we show that v-Src can markedly stimulate BSP gene transcription by targeting an inverted CCAAT box in the proximal promoter of the BSP gene.

Primary fibroblastic cells, derived from the 129/sv mouse embryos homozygous for c-src disruption (src−/−; Ref. 20) and used for transient transfection assays were provided by Dr. B. B. Mukherjee (McGill University, Montreal, Canada). FBS was purchased from Sigma Chemical Co. (St. Louis, MO). The FBS was heat-inactivation by heating at 56°C for 30 min. Cells were grown in DMEM with high glucose, supplemented with 5% FBS, and subcultured before reaching confluence.

Plasmid pEcoRIB, containing the RSV src region, was obtained from American Type Culture Collection. An EcoRI fragment containing the v-src-coding region was excised from pEcoRIB and cloned into the pCMV5 vector under the control of the cytomegalovirus promoter. Transcription analyses were performed on BSP promoter constructs ligated to a luciferase reporter gene. Constructs pRBSP-43, pRBSP-116, pRBSP-424, and pRBSP-801 were made by excising the rat BSP promoter inserts from pLUC 2, 3, 4, and 5 (21), respectively, with HindIII and SalI and religating the inserts into the pGL-3-Basic vector (Promega, Madison, WI) at the SmaI restriction site. Other luciferase constructs were made by ligating the rat BSP promoter segment-2992 (XbaI) to+60 (SalI), including the 5′ Phage arm, into the pGL-3-Basic vector (Promega) at the SmaI restriction site. This ∼3-kb insert was then deleted unidirectionally with Exo III (Amersham Pharmacia Biotech, Oakville, Ontario, Canada), and the 5′-ends of each deletion determined by nucleotide sequencing. Two chimeric constructs encompassing nts −323 to +79 (pHBSP-323 and nts −1038 to +79 (pHBSP-1038) of the human BSP gene promoter were prepared by unidirectional deletions of the human BSP promoter (22) ligated into the pGL-3-Basic vector. The expression vector for pSV-β-Gal was purchased from Amersham Pharmacia Biotech; the dominant-negative expression vector for NF-Y (NFY DN) was from R. Mantovani (Universita degli Studi di Milano, Milan, Italy), and a dominant-negative expression vector for Ras (Ras DN) was from Dr. A. Guha (Mount Sinai Hospital, Toronto, Canada).

Site-directed point mutations (5′-ATTGG-3′ to 5′-ATTTT-3′) in the CCAAT box were generated by PCR, as described by Ausubel et al. (23). The designated mutations and fidelity of PCR were confirmed by nucleotide sequencing.

All transfection assays were performed on exponentially growing src−/− cells. Cells were plated onto 12-well multiwell dishes 24 h before transfection at a density of 5 × 10⁴ cells/well. Cells at 30–50% confluence were transfected with 0.5 μg of a luciferase construct, 0.5 μg of pBluescript (Stratagene, La Jolla, CA), as a nonspecific DNA, and 0.5 μg of pSV-β-Gal as an internal control, with or without the v-src expression or empty vector using the calcium phosphate-DNA coprecipitation method (23). After 3 h, glycerol shock was done by incubating the cells for 1 min in 10% glycerol, and the cells given fresh 5% FBS media. After an additional 3 h of incubation, the cells were washed three times with DMEM with high glucose and given fresh 0.5% FBS media, and harvested 48 h after the transfection.

The luciferase assay was carried out using the Luciferase Assay System (Promega). Briefly, cells were washed twice in ice-cold PBS and thoroughly drained before 400 μl of 1X Reporter Lysis Buffer (Promega) was added to lyse the cells, which were then incubated for 15 min at RT before being scraped into a 1.5-ml microfuge tube. The samples were kept on ice until vortexed for 20 s and then centrifuged at 12,000 × g in a Microfuge. The luciferase assay was carried out with 20 μl of the supernatant (cell extract) and 50 μl of Luciferase Assay Reagent (Promega). The light emitted was measured for 20 s using a Berthold Lumat LB-9501 luminometer. To normalize for the transfections, a β-Gal assay was carried out using the GalactoLight Plus luminescent reporter assay kit (Tropix, Boston, MA). To inactivate endogenous β-Gal activity, 20 μl of the cell extract was heated to 48°C for 50 min and, after returning to RT, 200 μl of Reaction Buffer were added and incubated for 1 h. After the addition of 300 μl of Light Emission Accelerator, the emitted light was measured for 10 s.

Oligonucleotides used for GMS assays were purchased from Life Technologies, Inc., and the consensus oligonucleotides for CTF/NF1 and CREB were purchased from Promega. Anti-NF-Y antibodies (24) were a generous gift from Dr. R. Mantovani. The respective double-stranded oligonucleotides (3.5 pmol) were end-labeled with [γ-³²P]ATP (New England Nuclear), using T₄ polynucleotide kinase (Pharmacia Biotech). Nuclear extracts (1–6 μg) were incubated with 0.035 pmol of the labeled probe at 21°C for 30 min using the binding conditions described by Dorn et al. (25). For the antibody and competition experiments, antibodies or competing oligonucleotides were preincubated with the binding reaction mixture for 10 min at RT before the probes were added and incubated for an additional 20 min. The components in the reaction mixture were resolved by electrophoresis on a 5% nondenaturing acrylamide gel (80:1, acrylamide:bis acrylamide), run at 150 V at RT. After electrophoresis, the gel was dried and exposed to Kodak X-OMAT AR film for 3–24 h at-70°C.

Nuclear extracts used for GMS assays were prepared from NIH3T3, NIHv-src, murine c-src, src−/−, and src+/− cells according to a modified Digman’s method, as described in Ausubel et al. (23). Protein concentration was determined using the Bio-Rad Protein Assay (Bio-Rad, Hercules, CA). The concentration of the nuclear extracts from the different cells was approximately 3–4 mg/ml.

Transient transfection analysis using various length constructs of the rat BSP promoter linked to a luciferase reporter demonstrated that cotransfection of the expression vector containing the v-Src gene is able to up-regulate luciferase expression by <5-fold relative to controls in which the empty expression vector was cotransfected (Fig. 1). The up-regulated expression was observed in the shortest construct that included nts −116 to +60 of the promoter. Similarly, transient transfection assays using human BSP promoter constructs showed a similar v-Src-mediated increase in transcription in the highly conserved (22) immediate promoter region (data not shown). To delineate further the region through which v-Src was acting to increase BSP transcription, shorter constructs of the rat promoter were prepared. An increase in transcription was still present as the promoter was deleted from −116 to −60. However, with deletions beyond nt −43 the increase in transcription was no longer evident (Fig. 2). The only recognizable consensus sequence in the region encompassing nucleotides −43 to −60 in the promoter was an inverted CCAAT box (Fig. 3).

To determine whether the CCAAT box is involved in the up-regulation of the rat BSP promoter by v-Src, two point mutations were introduced (5′-ATTGG-3′ to 5′-ATTTT-3′) in the CCAAT box sequence within the construct pRBSP-116. Although these mutations markedly decreased the luciferase activity of the construct, the up-regulation of the rat BSP promoter by v-Src was completely eliminated (Fig. 4). To confirm that v-Src was acting through the CCAAT box, double-stranded oligonucleotides containing the CCAAT box were ligated upstream of a thymidine kinase promoter cloned into the pGL-3-Basic luciferase reporter vector and analyzed by transient transfection assays. The pTk-S-CCAAT construct, which contained the complete CCAAT box, was up-regulated about 50% by v-src cotransfection. However, constructs pTk-5′-CAAT and pTk-CCAA-3′, in which the CCAAT box was truncated were not significantly affected by v-src expression. Similarly, the construct that contained the mutated CCAAT box oligonucleotide (pTk-M-AAAAT), with the same mutation present in the mutated pRBSP-116 construct, showed no v-Src effect on transcriptional activity (Fig. 5). These results demonstrate that an intact CCAAT box is required for regulation of BSP promoter activity by v-Src.

To identify nuclear components that might mediate v-Src effects on BSP transcription, a number of double-stranded oligonucleotide probes were constructed for GMS assays. The sequence of these probes as well as the regions in the rat BSP promoter that they span are depicted in Fig. 3. A 45-mer oligonucleotide probe (l-CCAAT) was used initially to identify nuclear components that bind in the region around the CCAAT box. GMS analysis with increasing amounts of nuclear extract showed a dose-dependent increase in the formation of three complexes, identified as α, β and γ, that seem to bind the L-CCAAT probe specifically (Fig. 6), because competition GMS assays resulted in the selective disappearance of these three bands with the corresponding unlabeled double-stranded oligo (data not shown). To identify in more detail the regions where these three components interact, GMS assays were carried out with shorter constructs encompassed by the L-CCAAT probe. Two of the three bands, α and β, bound to a double-stranded oligonucleotide that included the complete CCAAT box (S-CCAAT) as well as the NL-CCAAT probe, which extends further 3′ from S-CCAAT (Fig. 6). Notably, much stronger binding of the β component to these oligonucleotides was observed compared with the L-CCAAT probe. Using a 5′-CAAT probe that extends from the 5′-end of the L-CCAAT (nt −69) sequence to the 3′-end of the CCAAT box, but omitting the last nucleotide (nt −47) to prevent the association of the factor(s) that bind to the CCAAT box, a new complex distinct from α and β which binds 5′ to the CCAAT box was revealed (Fig. 6). Therefore, the nuclear components that form complexes designated δ and γ seem to bind at each side of the CCAAT box, independently from α and β.

Because the intact CCAAT box seemed to be necessary for mediating the up-regulation of the rat BSP transcription by v-Src, a GMS assay was carried out to determine whether the formation of the complexes α and β was specific. The 5′-CAAT and CCAA-3′ probes that flank the CCAAT box, but engineered to have an incomplete CCAAT box sequence, thereby preventing the association of CCAAT binding factors, were unable to compete for the binding of the bands α and β (Fig. 7). Moreover, a probe (M-AAAAT) identical in sequence and length to the S-CCAAT probe, but including two point mutations in the CCAAT box (5′-ATTGG-3′ to 5′-ATTTT-3′) was also unable to compete for the binding, demonstrating that the intact CCAAT box is required for binding α and β (Fig. 7). That the binding was specific was indicated by the loss of bands when the binding is competed with unlabeled S-CCAAT probe (Fig. 7).

Although there are a number of CCAAT box binding transcription factors, the CCAAT box found in the rat BSP gene has a consensus sequence recognized by the transcription factor NF-Y. Therefore, a polyclonal antibody, α-NF-YB, raised against the NF-YB subunit (24), was used in the GMS assay to identify the CCAAT box-binding components. The antibody was found to selectively supershift the α complex dose-dependently (Fig. 8), confirming that the α band was the transcription factor NF-Y. Notably, NF-Y is a heterotrimeric complex made up of three subunits (A, B, and C) that are collectively required for DNA binding (26). Consequently, the α band likely respresents the trimeric protein bound to the double-stranded oligonucleotide.

To ascertain that NF-Y was indeed necessary for the transcriptional regulation of the rat BSP promoter by v-Src, transient transfection assays were carried out using a dominant-negative form of NF-Y (NF-Y DN). Cotransfection of NF-Y DN expression vector with v-src expression vector abolished the increase in transcription observed when v-src expression vector is used alone (Fig. 9). This effect was only seen when the CCAAT box was present in the promoter construct, the effect being absent when the CCAAT box was removed (Fig. 9).

To determine whether v-Src expression affected the levels of NF-Y, CCAAT binding proteins in nuclear extracts prepared from NIH3T3, NIHv-src, murine c-src, as well as from src−/− and src+/− cells were analyzed by GMS using the double-stranded S-CCAAT oligonucleotide. Although slightly higher binding of the NF-Y and β components was seen in the two cell lines overexpressing v-Src and c-Src when compared with the src−/− and src+/− cells, even higher binding was apparent in the control NIH3T3 cells (Fig. 10). Also, because the binding was comparable in the src−/− and src+/− cells, the amount of NF-Y capable of binding seemed to be independent of the level of expression and activity of Src.

To determine whether the v-Src effects on NF-Y were mediated through the Ras pathway transient transfection assays with the v-src expression vector were performed in the presence and absence of a dominant negative ras (Ras DN) expression vector. Although the Ras DN effectively blocks the Ras pathway, it failed to have any significant effect on the v-Src-induced stimulation of expression by the BSP-Luc constructs (results not shown).

To determine whether the endogenous BSP gene is induced by Src expression, mRNA from NIH3T3, NIHv-src, murine c-src, and src−/− cells transfected with the v-src expression vector were probed with a mouse BSP cDNA by Northern hybridization. BSP mRNA could not be detected in any of these cell lines, but was observed in control mouse bone MC3T3-E1 cells (results not shown).

Studies of breast tumor tissues have shown that BSP, which is normally expressed only in mineralizing tissues, is synthesized by malignant mammary epithelial cells (3). Moreover, higher expression of BSP in mammary tumors is associated with a more aggressive potential, suggesting that BSP expression could contribute to the invasive phenotype of metastasizing breast cancer cells (6). A retrospective analysis of breast cancer patients has revealed that the frequency of bone metastases is higher in patients with BSP-positive tumors (22%) than in patients with BSP-negative tumors (7%; 5). Furthermore, the majority of patients with advanced breast cancer have evidence of skeletal metastases by the time of death (27), indicating that bone is the preferred target for metastatic breast cancer cells. Bone is also a target of prostate, multiple myeloma, lung, thyroid, and kidney metastatic cancer cells that express BSP (5, 6). Thus, the ectopic expression of BSP, which can mediate the attachment of cells to bone mineral through RGD and polyglutamate motifs (7), could be a critical component in the metastasis of cancer cells to bone. In this regard, BSP is present on the cell surface of estrogen-receptor-positive breast cancer cell lines, such as MCF-7 and T47-D (3), and osteotropism of breast adenocarcinoma cells is closely associated with its estrogen receptor status (28).

In our studies, transient transfection analysis revealed that v-Src up-regulates transcription of both rat and human BSP promoter constructs <5-fold, but did not induce expression of the endogenous BSP gene. Because BSP expression is induced in various cancer cells, studies of cell transformation can provide insights into mechanisms of BSP gene induction. In models of mammary carcinoma development, aberrations in EGF signaling have been shown to invoke tumorigenesis (14, 15) through a mechanism requiring the physical association of the activated neu receptor with the proto-oncogene c-src (16). Also, Src kinase is rapidly stimulated by estrogen in MCF-7 mammary tumor cells (29). Notably, overexpression of c-src alone does not induce mammary tumor formation (17), although it is a key step in oncogenesis (18). v-Src is reported to modulate gene expression through serum-responsive elements of the Egr1/TIS8 gene (29), a dyad symmetry element, and the Sis-inducible factor-responsive element of the c-fos gene (30), a Src-responsive site of the 9E3/CEF-4 gene (31), an ATF/CRE element of the TIS10/PGS2 gene (32), and a CCAAT element in the proximal promoter of the OPN gene (13). It has also been reported that v-Src regulates the junB gene through the CCAAT or TATA elements (33), although which of these elements was responsible for this regulation was not demonstrated. Previous studies have shown that OPN expression, which is also related to cellular transformation (10, 11) is up-regulated by v-Src, which stimulates OPN gene transcription through an inverted CCAAT element immediately upstream from the TATA box (13). Although OPN expression is also stimulated by EGF, no response to EGF was evident in BSP transcription (results not shown). In this regard, the EGF and Src effects do not seem to be linked because OPN expression can be up-regulated by EGF in the absence of Src expression (34).

The variation observed in the degree of stimulated transcription by v-Src in different experiments could be attributed to a response to serum, which also seems to act through the CCAAT box (35). That the CCAAT box was the target of the v-Src response is evident from the following three observations. First, the v-Src response was abolished by the introduction of a two-point mutation in the CCAAT box. Second, a heterologous reporter construct, pTK-S-CCAAT, in which the CCAAT sequence was placed upstream of a thymidine kinase minimal promoter, was responsive to v-Src, whereas the same construct with the two point mutations in the CCAAT box (pTK-M-CCAAT) was not. Third, two other heterologous constructs that contain the 5′ and 3′-flanking regions of the CCAAT box, but with an incomplete CCAAT box, were not up-regulated by v-Src. Although the increase in transcription with pTK-S-CCAAT was less (50%) than observed with BSP-Luc constructs, and could suggest the involvement of flanking or upstream sequences, more modest responses are frequently observed from heterologous contructs, even when multiple copies of the element are incorported, as observed in studies of the OPN CCAAT box (13).

Of the four components identified by GMS assays to bind to the CCAAT box region of the rat BSP gene promoter, only two components (α and β) bound to the CCAAT box (Fig. 6). Because the δ and γ complexes do not require an intact CCAAT box for binding, they do not seem to be involved in Src-mediated transcription of BSP. The complex α appears to represent an association of the heterotrimeric complex of the A, B, and C subunits of the transcription factor NF-Y (36) with the CCAAT element because it could be supershifted with an antibody to NF-Y. The faster migrating β complex seems to represent a novel component binding in the CCAAT box region. Although the β complex was competed effectively by the consensus oligonucleotide for CREB, this is possibly due to the presence of a CCAAT-like flanking sequence because the migration of the β complex does not correspond to the mobility of CREB protein complexes. Moreover, the involvement of the β component in Src-mediated transcription does not seem likely because the formation of the β complex was markedly compromised in the extended CCAAT oligonucleotide, L-CCAAT (Fig. 6), indicating that its binding in the context of the complete promoter may not be significant. Consequently, NF-Y seems to be the most likely target of Src, as observed in the OPN gene, in which two components binding in the region of the CCAAT box were also identified (13). The second complex binding 5′ of the CCAAT box in the OPN promoter appears to be equivalent to component δ, which does not seem to be involved in the regulation of OPN or BSP genes by v-Src, as demonstrated by transient transfection assays performed using constructs that include the binding element for δ (Fig. 5).

Although NF-Y seems to mediate the v-Src stimulation of BSP and OPN transcription, the mechanism of NF-Y transcriptional activation remains elusive. Similar to previous studies (13, 33), we were unable to demonstrate any significant differences in the amounts of NF-Y that bound to the CCAAT box in nuclear extracts of src−/− and src+/− cells, nor were the levels significantly elevated or suppressed in cells transfected with v-Src and c-Src expression vectors (Fig. 10). Although it is possible that NF-Y may be activated by chemical modification, such as phosphorylation, presently there is no evidence for this. Indeed, in most instances, NF-Y functions with more specific transcriptional activators. For example, recent studies have shown that the NF-YB and NF-YC can interact with the TATA-binding protein through conserved regions adjacent to putative histone fold motifs (37). In view of the close proximity of the inverted CCAAT and TATA boxes in the BSP promoter (Fig. 5) and the marked increase in basal transcription of promoter constructs that include the CCAAT box (22, 38), it is conceivable that the NF-Y complex may regulate BSP gene transcription by direct association with the preinitiation complex.

The signaling pathway that links v-Src activity with NF-Y is not known. Because estrogens activate the signal-transducing Src/p21ras/ERK pathway in breast cancer cells through an interaction of Src with the estrogen receptor (39) and inhibiting Ras activity can block the ability of v-Src to transform fibroblasts (40), we studied the effects of a dominant-negative ras expression vector on v-Src activity in transient transfection assays. However, no significant effect was observed indicating that the Ras pathway was not involved in the transcriptional increase of the rat BSP promoter by v-Src. Another pathway that could be involved in the transcriptional increase by Src is through STAT proteins, which are a group of latent cytoplasmic transcription factors that function as signal transducers and activators of transcription (41). Among the members of this family, Stat3, which is activated by various cytokines, is directly activated by v-Src (42). v-Src associates with Stat3 and phosphorylates in vitro, at which time Stat3 is able to migrate to the nucleus and activate transcription through the sis-inducible element (42). However, in our studies the target for v-Src is the CCAAT box. Consequently, further experimentation will be necessary to identify the relevant pathway that links v-Src and BSP gene transcription.

We thank Drs. Mantovani and Gupta for generosity in providing constructs and antibodies used in these studies. We also thank Kam-Ling Yao for expert technical assistance and Drs. B. Ganss and S. Cheifetz for advice with experiments.