Department of Molecular Genetics, University of Texas M. D. Anderson Cancer Center, Houston, Texas, TX 77030, USA

Correspondence to: Michèle Sawadogo, Department of Molecular Genetics, University of Texas M. D. Anderson Cancer Center, Houston, Texas, TX 77030, USA

USF is a family of transcription factors that are structurally related to the Myc oncoproteins and also share with Myc a common DNA-binding specificity. USF overexpression can prevent c-Myc-dependent cellular transformation and also inhibit the proliferation of certain transformed cells. These antiproliferative activities suggest that USF inactivation could be implicated in carcinogenesis. To explore this possibility, we compared the activities of the ubiquitous USF1 and USF2 proteins in several cell lines derived from either normal breast epithelium or breast tumors. The DNA-binding activities of USF1 and USF2 were present at similar levels in all cell lines. In the non-tumorigenic MCF-10A cells, USF in general, and USF2 in particular, exhibited strong transcriptional activities. In contrast, USF1 and USF2 were completely inactive in three out of six transformed breast cell lines investigated, while the other three transformed cell lines exhibited loss of USF2 activity. Analyses in cells cultured from healthy tissue confirmed the transcriptional activity of USF in normal human mammary epithelial cells. These results demonstrate that a partial or complete loss of USF function is a common event in breast cancer cell lines, perhaps because, like Myc overexpression, it favors rapid proliferation.

USF; breast cancer cell lines; c-Myc; transcription activation

Introduction

USF is a family of helix - loop - helix transcription factors that display strong similarities with the Myc oncoproteins both in their overall protein structure (Murre et al., 1989; Gregor et al., 1990; Sirito et al., 1994) and DNA-binding specificity (Blackwell et al., 1990; Kerkhoff et al., 1991; Bendall and Molloy, 1994). In mammals, the two ubiquitously expressed genes, Usf1 and Usf2, play overlapping, essential roles in embryonic development (Sirito et al., 1998). All USF proteins interact with DNA as dimers and recognize E-box elements characterized by a central CACGTG or CACATG sequence. The major USF species present in most tissues and cell types is the USF1·USF2 heterodimer. USF1 homodimers are less abundant and USF2 homodimers usually quite scarce (Sirito et al., 1994, 1998; Viollet et al., 1996).

Strong activation of transcription was observed both in vitro and in vivo following binding of USF to specific sites in gene promoters (Sawadogo and Roeder, 1985; Luo and Sawadogo, 1996b; Roy et al., 1997). This transcriptional activity requires, in addition to the C-terminal DNA-binding domain, activation domains present in the N-terminal region of the USF proteins. In both USF1 and USF2, a specialized activation domain, located within the highly conserved USF-specific region (USR), is necessary and sufficient for transcriptional activation of promoters containing an initiator element in addition to a TATA box (Luo and Sawadogo, 1996b). USF2 also contains a second, more classical transcriptional activation domain encoded by the fifth exon of the gene. Together with the USR, this exon 5 domain is required for transcriptional activation by USF2 of promoters lacking an initiator element (Luo and Sawadogo, 1996b; Qyang et al., 1999). Other USF2 domains, encoded by the first four exons, regulate the activity of the exon 5 activation domain (Luo and Sawadogo, 1996b).

It is noteworthy that, although the USF proteins are ubiquitously expressed, they may not be transcriptionally active in all cell types. We recently reported a comparison between the activities of USF1 and USF2 in the HeLa and Saos-2 cell lines (Qyang et al., 1999). In HeLa cells, the USF proteins were transcriptionally active and, when overexpressed, caused marked growth inhibition. In contrast, USF1 and USF2 were both transcriptionally inactive in Saos-2 cells and, consequently, their overexpression had no effect on proliferation. In mutational analysis, the inability of USF to activate transcription in Saos-2 cells correlated with a complete inactivity of the USR domain. These results suggest the existence of a USF-specific coactivator that interacts with the USR domain to mediate the transcriptional activity of USF1 and USF2 in cells such as HeLa and is either absent or somehow inactivated in Saos-2 cells (Qyang et al., 1999).

Like USF, c-Myc is a ubiquitously expressed transcription factor. c-Myc is known to play a key role in the control of cellular proliferation and its overexpression, whether due to gene amplification or translocation, is a frequent event in tumor progression (reviewed in Marcu et al., 1992; Garte, 1993). In breast cancer in particular, there is a strong correlation between c-Myc overexpression and an elevated risk of relapse, especially in the case of node-negative patients (Varely et al., 1987; Roux-Dosseto et al., 1993; Escot et al., 1993). There is also a clear correlation between the c-myc expression level and the proliferative capacity of breast cancers and breast cancer cell lines. For instance, amplification of the c-myc gene, which is observed in 20 - 30% of all breast cancers, occurred in more than 50% of mammary carcinomas selected for their high proliferation rates (Kreipe et al., 1993).

The transforming ability of c-Myc is best exemplified by its ability to elicit the complete transformation of primary cells when coexpressed with a second oncoprotein such as activated Ras (Land et al., 1983). The effect of the USF proteins on cellular transformation was also investigated by focus formation assay in primary embryonic fibroblasts (Luo and Sawadogo, 1996a). In that assay, cotransfection of either USF1 or USF2 with activated Ras did not result in the appearance of foci of morphologically transformed cells, demonstrating that the function of USF in transformation is different from that of c-Myc. Instead, cotransfection of USF was found to abolish cellular transformation mediated by c-Myc and activated Ras. This inhibition of cellular transformation by USF required not only its DNA-binding domain but also domains involved in transcriptional activation, indicating that the effect was not a simple DNA-binding competition with Myc. Rather, it seems that the activity of USF can antagonize the transforming ability of Myc. The inhibitory activity of USF1 in the focus formation assay was specific to the Myc pathway since USF1 overexpression had no effect on the cellular transformation of embryonic fibroblasts mediated by E1A and Ras. In contrast, USF2 overexpression inhibited focus formation mediated by a variety of oncogenes, probably as a consequence of its ability to strongly inhibit the growth of certain transformed cells (Luo and Sawadogo, 1996a).

Considering the key role of Myc in breast cancer together with the antagonism between the cellular functions of USF and Myc, it is clearly important to determine whether USF is also implicated in breast carcinogenesis. Here, we report an analysis of the transcriptional activities of USF1 and USF2 in different normal and breast cancer cell lines. The results indicate that, like Myc overexpression, loss of USF function is common in breast cancer cells.

Results

Expression and DNA-binding activity of USF in human breast cell lines

To investigate the expression of endogenous USF in breast cells of various origins, we assembled a panel of cell lines including two originally derived from normal breast epithelium (MCF-10A and HBL-100) and six derived from breast tumors: Hs578T and T47D (ductal carcinomas), and BT-20, MCF-7, MDA-MB-231 and MDA-MB-468 (adenocarcinomas). Among these cancer cell lines, two were strongly estrogen responsive (MCF-7 and T47D) while others were estrogen receptor negative (BT-20, MDA-MB-231, HBL-100, Hs578T and MDA-MB-468). These different cell lines were also characterized by different degrees of tumorigenicity and metastatic potential, ranging from non-tumorigenic (MCF-10A) to highly metastatic (MDA-MB-231). Note finally that only MCF-10A cells were reported to contain a normal or nearly normal complement of chromosomes (Soule et al., 1990). While derived from epithelial cells in the milk of a nursing mother and originally non-tumorigenic, HBL-100 cells are known to contain two integrated repeats of the SV40 genome and become increasingly tumorigenic at passages above 35 (Krief et al., 1989). These cells were used in our experiments at passages 35 to 41. For comparison, two other human cell lines, HeLa (cervical carcinoma) and Saos-2 (osteosarcoma), were added to the panel because they had been previously well characterized for USF expression and activity (Qyang et al., 1999).

For each cell line, nuclear extracts were prepared using identical volumes of buffers and identical cell numbers. The levels of endogenous USF in these nuclear extracts were determined by Western blotting (Figure 1). This analysis revealed that the USF1 and USF2 polypeptides were present and displayed the same electrophoretic mobilities in all cell lines (Figure 1A). There was also no significant variation in the relative levels of USF1 and USF2, although the absolute levels of the two USF polypeptides varied from cell line to cell line. These variations may be related to the fact that the expression of USF1 and USF2 changes considerably during the cell cycle (T Lu and M Sawadogo, unpublished result) and that cell lines growing at different rates probably contain different distributions of cells in the various phases of the cell cycle. Nevertheless, it was clear from this analysis that the USF1 and USF2 proteins were ubiquitously expressed in breast cell lines.

Next, an electrophoretic mobility shift assay (EMSA) was used to monitor the DNA-binding activity of USF in the same nuclear extracts (Figure 1B). In this experiment, a 150-bp DNA fragment was used as the radiolabeled probe under conditions that allow separation of the complexes containing different USF dimers (Sirito et al., 1994, 1998). Total USF DNA-binding activity in the different extracts varied in proportion to the concentration of USF polypeptides detected by Western blotting, indicating that there was no difference between cell lines in the ability of the USF proteins to bind DNA (Figure 1). Just as in HeLa and Saos-2, complexes corresponding to the USF1·USF2 heterodimers were the predominant species in all breast cell lines tested, followed by the USF1 homodimer-containing complexes. With the notable exception of MCF-10A, complexes containing USF2 homodimers were barely detectable in most cell lines. From this analysis, we concluded that USF proteins active in DNA binding were ubiquitously expressed in both normal and cancerous breast cells. Therefore, if USF was implicated in breast carcinogenesis, changes in expression or DNA-binding activity were probably not involved.

Optimization of transfection efficiencies in different breast cell lines

In order to analyse the transcriptional activity of USF in the different cell lines, we needed efficient transfection capability. We therefore explored for each cell line the use of different transfecting agents, incubation times, and plated cell numbers. In these optimization experiments, transfection efficiencies were monitored using the RSV-Luc reporter construct in which transcription of the luciferase gene is driven by the strong Rous Sarcoma virus promoter. Satisfactory conditions were obtained for seven cell lines (see Materials and methods and summary in Table 1). Transfections in BT-20 cells remained poor under all conditions tested, so the transcriptional activity of USF in this cell line could not be determined.

As illustrated in Figure 2, transcription from the RSV promoter was, under optimum transfection conditions, comparable in the seven breast cell lines, with less than fourfold difference between the largest and smallest values of luciferase activities. Next, we used EMSA to determine whether ectopic USF could be efficiently expressed in the same cell lines. In all cases, high overexpression was observed after transient transfection of expression vectors encoding either USF1, USF2, or the mutant USF2N (Figure 3). USF2N lacks all USF2 domains implicated in transcriptional activation, but localizes properly to the nucleus and binds DNA as efficiently as wild-type USF2 (Luo and Sawadogo, 1996b). Therefore, this mutant provides a useful control for monitoring the expression of reporter genes in the absence of USF transcriptional activity.

Transcriptional activity of USF in a non-tumorigenic breast epithelial cell line

The reporter plasmids used to investigate the transcriptional activity of USF in different breast cell lines by transient cotransfection assays are schematically represented in Figure 4a. pMLLuc and pU3MLLuc both contain the adenovirus major late minimum promoter driving transcription of the luciferase gene. The two reporters differ only by the presence or absence of three USF-specific binding sites inserted in pU3MLLuc upstream of the TATA box. Our earlier studies had shown that transcription from the pU3MLLuc reporter can be strongly activated in HeLa cells by cotransfection with either USF1 or USF2 (Luo and Sawadogo, 1996b).

MCF-10A, a spontaneously immortalized cell line, was used to evaluate the transcriptional activity of USF in normal breast epithelial cells (Figure 4). In these cells, transfection of the pU3MLLuc reporter yielded a luciferase activity 13-fold higher than that observed for pMLLuc, indicating that endogenous proteins activated transcription by binding to the USF sites (Figure 4b). To determine whether these endogenous proteins were USF itself, we cotransfected expression vectors encoding mutants with dominant negative effect on the activity of the USF or, as a control, the Myc family of transcription factors. A-USF and A-Max are constructs in which the basic region of USF1 or Max, respectively, was replaced by an acidic sequence (Qyang et al., 1999; Krylov et al., 1997). This substitution greatly stabilizes heterodimer formation between the wild-type and mutant proteins. For example, at a 3 : 1 ratio, A-USF essentially abolishes DNA binding by USF in vitro (Qyang et al., 1999). Cotransfection of A-USF significantly reduced the activity of the pU3MLLuc reporter in MCF-10A cells, while cotransfection of A-Max was much less inhibitory. Cotransfection of either A-USF or A-Max had essentially no effect on the activity of the pMLLuc reporter (Figure 4b). From these results, we concluded that endogenous USF, rather than Myc-related proteins, was primarily responsible for the strong activity of the pU3MLLuc reporter in MCF-10A cells.

The ability of exogenous USF proteins to also activate transcription in MCF-10A cells was monitored by cotransfecting pU3MLLuc with expression vectors encoding either USF1, USF2, or, as a control, the USF2N mutant lacking all transcriptional activation domains. As shown in Figure 4c, transcription from the pU3MLLuc reporter was further increased 1.8-fold by cotransfection of USF1 and 3.3-fold by cotransfection of USF2, demonstrating that exogenous USF proteins were transcriptionally active in MCF-10A cells. In contrast, cotransfection of USF2N repressed the activity of the pU3MLLuc reporter, probably by competing for the binding of endogenous USF to the promoter DNA (Figure 4c). From these experiments, we concluded that both endogenous and exogenous USF proteins were active transcription factors in normal breast epithelial cells.

Transcriptional activity of USF in transformed breast cell lines

In contrast to the pRSV-Luc reporter, which yielded comparable luciferase activities in all cell lines (Figure 2), the USF-dependent pU3MLLuc reporter was very poorly transcribed in all six tumorigenic breast cell lines tested as compared to MCF-10A (Figure 5). This was striking since USF was present at similar levels in all cell lines examined (Figure 1). Given the strong transcriptional activity of endogenous USF in MCF-10A (Figure 4b), this result suggested that endogenous USF exhibited decreased transcriptional activity in cancer cells as compared to normal cells.

The ability of exogenous USF1 and USF2 to activate transcription in different breast cancer cell lines was next analysed by cotransfecting the pU3MLLuc reporter and USF1, USF2, USF2N, or the corresponding empty vector (Figure 6). In three cell lines, Hs578T, MDA-MB-468 and T47D, strong transcriptional activity was detected for USF1 (13 - 22-fold activation), while USF2 stimulated transcription only 2 - 3-fold over the low endogenous level. Given that USF2 was actually more active than USF1 in MCF-10A, this result suggested a deficiency affecting specifically the transcriptional activity of USF2 in these three cell lines.

Exogenous USF1 and USF2 were both completely inactive at the pU3MLLuc reporter in the other three cell lines: MCF-7, MDA-MB-231 and HBL100 (Figure 6). This result indicated a total loss of USF function in these cells, similar to that previously characterized for the Saos-2 osteosarcoma cell line (Qyang et al., 1999). Also as in Saos-2, transcription of the pU3MLLuc reporter could be activated in breast cancer cells by USF2-VP16, a fusion protein containing the ubiquitous activation domain of VP16 fused at the C-terminus of USF2 (data not shown). This control demonstrated that the inactivity of the natural USF proteins was not due to an inability to localize to the nucleus or bind to the promoter DNA, but to subsequently activate transcription in these cells. Taken together, these experiments suggested that events affecting to various extents the transcriptional activity of USF were extremely common in breast cancer cell lines.

Transcriptional activity of USF in human mammary epithelial cells (HMECs)

The different activities exhibited by USF in breast cancer cell lines as compared to the non-tumorigenic MCF-10A cell line were suggestive of a relationship between the loss of USF function and the transformed phenotype of these cells. However, it remained possible that MCF-10A cells could be, for some reason, unusual with regards to the activity of USF. To eliminate this possibility and confirm the suspected link between USF inactivation and carcinogenesis, we carried out additional analyses using HMECs. First, early-passage HMECs were transfected with the pRSV-Luc reporter to evaluate their transfectability. This control yielded luciferase activities similar to those demonstrated in established cell lines, indicating that reliable transfection results could be obtained in these cells (data not shown). Next, HMECs were transfected with either the pMLLuc or pU3MLLuc reporter to determine whether these cells contained endogenous factors that could mediate transcriptional activation through USF-specific binding sites. In three separate experiments, the reporter containing USF sites exhibited luciferase activities that were 30 - 70-fold higher than those of the matching reporter lacking these sites (Figure 7a). This result indicated that, like MCF-10A, but unlike the various breast cancer cell lines, HMECs contained endogenous USF-like transcription factors that could activate promoters by interacting with upstream USF sites.

To directly examine the transcriptional activities of USF1 and USF2 in HMECs, cotransfection experiments were carried out with the pU3MLLuc reporter and different USF expression vectors (Figure 7c). In one of these experiments, the overexpression of USF in these cells was analysed by EMSA, which confirmed the efficiency of transfection (Figure 7b). Just as in MCF-10A, the strong endogenous activity of the pU3MLLuc reporter in HMECs was further enhanced by a factor of 2 - 3-fold by cotransfection of either USF1 or USF2, while expression of the transcription activation-deficient USF2N mutant repressed promoter activity between 2.7 - 4.0-fold (Figure 7c). In three out of four experiments, exogenous USF2 exhibited a higher transcriptional activity than exogenous USF1, which was again reminiscent of the results obtained in MCF-10A (Figure 4c), but clearly different from those obtained in other cell types (Luo and Sawadogo, 1996b).

From this analysis, we concluded that USF was transcriptionally active in normal mammary epithelial cells, and that this normal USF activity was preserved in the immortalized MCF-10A cell line. Taken together with the depressed or absent activity of USF in the various breast cancer cell lines, these results strongly suggested a correlation between the loss of USF function and tumorigenesis.

c-Myc expression levels in breast cancer cell lines

To compare the frequency of USF inactivation in breast cancer cells to that of c-Myc overexpression, we examined by Western blotting the levels of the c-Myc polypeptide in the same panel of cell lines. As shown in Figure 8, c-Myc expression was lowest in the normal cell line, MCF-10A. Overexpression was observed in all the other cell lines, ranging from 2.3 - 6.2-fold in the different breast cancer cell lines and reaching almost tenfold in HeLa cells. This result demonstrated that c-Myc overexpression was a general characteristic of transformed cell lines. There was no clear correlation between the extent of Myc overexpression and the partial or complete loss of USF activity in the different cancer cell lines.

Discussion

To probe the involvement of transcription factor USF in carcinogenesis, we analysed the expression and activity of the ubiquitous USF1 and USF2 proteins in a representative panel of normal and breast cancer cell lines. These experiments revealed similar expression levels and DNA-binding activities of USF1 and USF2 in all cell lines. However, there was a striking difference between normal and transformed breast epithelial cells in the transcriptional activity of the USF proteins. Endogenous as well as exogenous USF proteins demonstrated transcriptional activities in the non-tumorigenic MCF-10A cells, with USF2 being more active than USF1 (Figure 4). The behavior of MCF-10A cells was representative of normal mammary epithelial cells, since similar results were obtained in early passage HMECs (Figure 7). In contrast, USF transcriptional activity was either decreased or absent in all transformed breast cell lines tested. Three of these cancer cell lines exhibited strong exogenous USF1 activity, but comparatively very low USF2 activity (Figure 6). In these cells, stimulation by overexpressed USF1 was more pronounced than in MCF-10A cells, reflecting the fact that the basal level due to endogenous transcription factors was considerably lower. However, the enhanced reporter activities in the presence of overexpressed USF1 were similar in these particular cell lines and in MCF-10A. Three other tumorigenic cell lines exhibited, in addition to a very low endogenous activity, a complete loss of transcriptional activity for exogenous USF1 as well as USF2 (Figure 6). Together, these results indicate that a partial or complete loss of USF activity is a very common event in breast cancer cell lines. Therefore, it seems likely that this event confers a selective advantage to these cells, at least during in vitro propagation.

It is noteworthy that all of the breast cancer cell lines studied here also demonstrated increased expression of the c-Myc oncoprotein as compared to the normal MCF-10A cells (Figure 7). The fact that Myc overexpression and loss of USF function were present in the same cells indicates either that the selective advantages conferred by these two events are not identical or that, if the same pathway is affected, the advantages are additive.

What are the molecular mechanisms involved in the loss of USF function in breast cancer cells? Given the similarity of the observed phenotypes, the gene responsible for the simultaneous inactivity of USF1 and USF2 in certain breast cancer cell lines is likely to be the same as the gene responsible for the inactivity of USF in Saos-2 cells (Qyang et al., 1999). In Saos-2, mechanistic studies showed that the loss of USF function was linked to the inactivity of the USR, an essential domain that is highly conserved between USF1 and USF2. The USR acts as an autonomous activation domain at promoters containing an initiator element, probably by recruiting a specialized transcriptional coactivator. In the case of USF2, interaction of this putative coactivator is also thought to induce a conformational change that unmasks the exon 5 activation domain necessary for activation of promoters lacking an initiator element (Luo and Sawadogo, 1996b; Qyang et al., 1999). Therefore, the simplest explanation for our results is that the coactivator that interacts with the USR to mediate the transcriptional activity of the USF proteins is either absent or inactivated in Saos-2 cells as well as in MCF-7, MDA-MB-231 and HBL-100 cells. Clearly, once the corresponding gene is identified, it will be very important to determine how it is altered in these cells and whether similar mutations also occur in tumors.

A different phenotype was observed in the other three breast cancer cell lines. In Hs578T, MDA-MB-468 and T47D, endogenous factors as well as exogenous USF2 were inactive, but exogenous USF1 was apparently active. The underlying mechanisms for this specific inactivation of USF2 remain to be determined. Possibilities include overexpression of a USF2-specific repressor, inactivation of a USF2-specific coactivator, or differences in posttranslational modifications affecting the interaction of USF2 with a regulatory cofactor. In any of these cases, it is important to note that specific inactivation of USF2 alone may be sufficient to confer a selective advantage to breast cancer cells since, in other cell types, USF2 displayed a much stronger antiproliferative activity than USF1 (Luo and Sawadogo, 1996a). Alternatively, it is also conceivable that the genetic alteration present in these three breast cancer cell lines actually causes the simultaneous inactivation of both USF proteins. Such a mechanism would be more consistent with the lack of endogenous activity, and the different effects of cotransfected USF1 versus USF2 could be explained by a concentration-dependent repression mechanism that is relieved more efficiently by overexpression of USF1 than by overexpression of USF2.

Whether endogenous USFs are the only transcription factors responsible for the strong endogenous activity of the pU3MLLuc reporter in MCF-10A and HME cells remains to be firmly established. Correlation between the loss of exogenous USF activity and the loss of endogenous E-box activity in the transformed cell lines strongly suggests that the two phenomena are related. Yet, cotransfection of A-USF or USF2N, which would be expected to greatly reduce or even abolish endogenous USF activity, failed to repress the activity of the pU3MLLuc reporter down to the basal level observed with the pMLLuc reporter lacking USF sites. This limited efficiency of the dominant negative mutants may reflect an intrinsic kinetic problem in this type of analysis. High expression of the dominant negative mutant may occur too late to prevent assembly of endogenous USF into stable transcription complexes at the promoter of the cotransfected reporter gene. Alternatively, it could be that additional E-box-binding transcription factors contribute to the strong activity of pU3MLLuc in normal breast epithelial cells. Possible candidates in this case include the different members of TFE3 family of transcription factors, which can all bind the adenovirus major late USF site (Beckmann et al., 1990; Fisher et al., 1991). However, TFE3 does not efficiently activate the pU3MLLuc reporter (Qyang et al., 1999). A role of c-Myc seems even more unlikely considering the lack of inhibition by A-Max (Figure 4B) and the fact that c-Myc expression is considerably lower in MCF-10A than in the cancer cell lines (Figure 8). There remains the possibility of an unknown transcription factor, although EMSA analysis did not reveal the presence in MCF-10A of novel proteins interacting with the USF site (Figure 1B). Furthermore, this unknown factor would also have to be commonly inactivated in breast cancer cells.

Considering the strong inhibitory effect of USF overexpression on transformation and proliferation in other cell types, the very common loss of USF transcriptional activity in cancer cell lines strongly suggests participation of these transcription factors in a general tumor suppression pathway. In this regard, it is striking that the transcriptional activity of USF is always altered indirectly, rather than directly by a change in the expression of the USF proteins. Together with the early embryonic lethality of the USF-null mice (Sirito et al., 1998; Vallet et al., 1998), this observation strongly suggests that, in addition to their important role in the transcriptional regulation of growth-related genes, the USF proteins may serve another function that is essential for cellular viability.

Materials and methods

Cell culture

The MCF-10A, MDA-MB 231 and T47D cell lines were generous gifts from M-C Hung, D-H Yu (University of Texas MD Anderson Cancer Center, Houston, TX, USA), and K Galaktionov (Baylor College of Medicine, Houston, TX, USA), respectively. All other breast cancer cell lines were obtained from the American Tissue Type Collection and propagated as monolayers in a humidified incubator at 37°C with 5% CO₂. Most cell lines were propagated in Dulbecco modified Eagle medium (DMEM) containing 60,000 U/ml penicillin and 200 U/ml streptomycin (GIBCO - BRL) and supplemented with 10% fetal bovine serum (Hyclone). For T47D cells, this medium was supplemented with 10 g/ml insulin (Sigma). MCF-10A cells were grown in 50% DMEM 50% F-12 medium (GIBCO - BRL) supplemented with 5% fetal bovine serum, 10 g/ml insulin (Sigma), 100 ng/ml cholera toxin (Calbiochem), 0.5 g/ml hydrocortisone (Sigma), 150 ng/ml epidermal growth factor (GIBCO - BRL), 1.05 mM calcium chloride, 30 000 U/ml penicillin and 100 U/ml streptomycin. Human breast epithelial cells (HMECs) were obtained for Clonetics (San Diego, CA, USA) and propagated according to the supplier's instructions.

Plasmids

The reporter plasmids pA₃RSVLUC (`pRSV-Luc'), pMLLuc, and pU3MLLuc were as described (Wood et al., 1989; Luo and Sawadogo, 1996b). Derivation of the pSG5-derived psvUSF1, psvUSF2 and pUSF2N (pU2N) expression vectors was previously reported (Meier et al., 1994; Luo and Sawadogo, 1996a,b). Vector alone controls for these expression vectors were provided by transfecting the pSG424 vector containing the same SV40 promoter (Sadowski et al., 1988) instead of the parental pSG5 vector, which, for unknown reasons, strongly inhibited the transcription of cotransfected reporters in several of the cell lines. Expression vectors driven by the cytomegalovirus promoter, pCMV-USF1 and pCMV-USF2, were constructed by subcloning the USF1 and USF2 cDNA inserts from psvUSF1 and psvUSF2, respectively, into the BamHI site of the pCMV₀ vector (identical to pCMV-NEO-BAM, Kern et al., 1992).

Transient transfection assays

About 16 h prior to transfection, cells were seeded in 6-well plates at the density appropriate to obtain 60 - 75% confluency (Table 1). For each cell line, DNA amounts and incubation times were titrated using pRSV-Luc with different transfecting agents to achieve maximal transfection efficiency. These optimized conditions, summarized in Table 1, were used in all subsequent experiments. For cotransfection assays, half of the total DNA amount was the reporter plasmid and the other half the USF expression vector or corresponding empty vector. Separate cotransfections were carried out in all cell lines with the SV40- and CMV-driven USF expression vectors. Stronger USF expression was obtained in T47D and MDA-MB-468 cells with the CMV-driven vectors and only these results were used. In the other cell lines, the two expression systems yielded identical results and the two sets of data were combined. Transfections using lipofectamine or Fugene 6 (Boehringer-Mannheim) were carried out according to the manufacturer's instructions. After 48 h, the cells were scraped in 100 l of lysis buffer and luciferase assays were performed as described previously (Luo and Sawadogo, 1996b). For comparison between cell lines, the arbitrary units of luciferase activities, determined using a scintillation counter, were all calculated for 10 l of lysate.

Transfection experiments were all repeated multiple times, and the results from a minimum of three independent experiments were used to calculate average luciferase activities and standard deviations.

Nuclear extracts

For all cell lines, mini nuclear extracts were prepared starting with 1.5´10⁶ cells according to the procedure of Schreiber et al. (1989).

EMSA and Western blotting

Standard DNA-binding reactions (10 l) contained 10 mM Tris-HCl (pH 7.9), 60 mM KCl, 1 mM dithiothreitol, 0.1% Triton-X100, 1 g of poly(dI-dC), and either 0.5 l of nuclear extract or 1.5 l of whole cell extract as prepared for the luciferase assay. The radiolabeled probe (0.1 ng) was either a 150-bp DNA fragment or a 33-bp oligonucleotide, as indicated in the Figure legends. The DNA-binding reactions also contained as competitor DNA 10 ng of a 30-bp oligonucleotide containing or lacking the USF consensus binding site. After 20 min incubation at 30°C, 2 l of 15% Ficoll were added and the reactions were analysed by electrophoresis on 4% acrylamide/0.2% bisacrylamide/22 mM Tris-borate (pH 8.3)/0.5 mM EDTA gels.

For Western blot analysis, proteins from 10 l of nuclear extracts, corresponding to 300 000 cells, were resolved by SDS - PAGE on 12% acrylamide gels. The proteins were then transferred to nitrocellulose membranes and probed with USF1-, USF2-, or c-Myc-specific peptide antibodies (Santa Cruz Biotechnology). Secondary antibody binding was detected using the ECL method (Amersham). Blots were stripped in a solution containing 62.5 mM Tris-HCl pH 6.8, 100 mM -mercaptoethanol and 2% SDS at 55°C for 30 min prior to re-use.

Acknowledgements

We thank M-C Hung, D-H Yu and K Galaktionov for cell lines, X Luo, S Maity and G Lozano for plasmids, H-X Yang for technical assistance, and MN Szentirmay for critical reading of the manuscript. This work was supported by Grants DAMD17-96-1-6221 from the Department of the Army and CA79578 from the National Cancer Institute (MS), by institutional funds, and by a postdoctoral fellowship from the National Cancer Institute Training Grant CA09299 (T Lu).

Beckmann H, Su LK and Kadesch T. (1990). Genes Dev. 4, 167-179. MEDLINE

Bendall AS and Molloy PL. (1994). Nucleic Acids Res. 22, 2801-2810. MEDLINE

Blackwell TK, Kretzner L, Blackwood EM, Weintraub H and Eisenman RN. (1990). Science 250, 1149-1151. MEDLINE

Escot C, Simony-Lafontaine J, Maudelonde T, Puech C, Pujol H and Rochefort H. (1993). Oncogene 8, 969-974. MEDLINE

Fischer DE, Carr CS, Parent LA and Sharp PA. (1991). Genes Dev. 5, 2342-2352. MEDLINE

Garte SJ. (1993). Crit. Rev. Oncogen. 4, 435-448.

Gregor PD, Sawadogo M and Roeder RG. (1990). Genes Dev. 4, 1730-1740. MEDLINE

Kerkhoff E, Bister K and Klempnauer KH. (1991). Proc. Natl. Acad. Sci. USA 88, 4323-4327. MEDLINE

Kern SE, Pietenpol JA, Thiagalingam S, Seymour A, Kinzler KW and Vogelstein B. (1992). Science 256, 827-830. MEDLINE

Kreipe H, Feist H, Fisher L, Felgner J, Heidorn K, Mettler L and Parwaresch R. (1993). Cancer Res. 53, 1956-1961. MEDLINE

Krief P, Saint-Ruf C, Bracke M, Boucheix C, Billard M, Cassingena R, Jasmin C, Mareel M and Assorone B. (1989). Int. J. Cancer 43, 658-664. MEDLINE

Krylov D, Kasai K, Echlin DR, Taparowsky EJ, Arnheiter H and Vinson C. (1997). Proc. Natl. Acad. Sci. USA 94, 12274-12279. MEDLINE

Land H, Parada LF and Weinberg RA. (1983). Nature 304, 596-602. MEDLINE

Luo X and Sawadogo M. (1996a). Proc. Natl. Acad. Sci. USA 93, 1308-1313. MEDLINE

Luo X and Sawadogo M. (1996b). Mol. Cell. Biol. 16, 1367-1375. MEDLINE

Marcu KB, Bossone SA and Patel AJ. (1992). Annu. Rev. Biochem. 61, 809-860. MEDLINE

Meier JL, Luo X, Sawadogo M and Straus SE. (1994). Mol. Cell. Biol. 14, 6896-6906. MEDLINE

Murre C, McCaw PS and Baltimore D. (1989). Cell 56, 777-783. MEDLINE

Qyang Y, Luo X, Lu T, Ismail P, Krylov D, Vinson C and Sawadogo M. (1999). Mol. Cell. Biol. 19, 1508-1517. MEDLINE

Roux-Dosseto M, Romain S, Dussault N, Desideri C, Piana L, Bonnier P, Tubiana N and Martin PM. (1993). Eur. J. Cancer 28A, 1600-1604.

Roy AL, Du H, Gregor PD, Novina CD, Martinez E and Roeder RG. (1997). EMBO J. 16, 7091-7104. MEDLINE

Sadowski I, Ma J, Triezenberg S and Ptashne M. (1988). Nature 335, 563-564. MEDLINE

Sawadogo M and Roeder RG. (1985). Cell 43, 165-175. MEDLINE

Schreiber E, Matthias P, Muhler MM and Schaffner W. (1989). Nucleic Acids Res. 17, 6419. MEDLINE

Sirito M, Lin Q, Deng JM, Behringer RR and Sawadogo M. (1998). Proc. Natl. Acad. Sci. USA 95, 3758-3763. MEDLINE

Sirito M, Lin Q, Maity T and Sawadogo M. (1994). Nucleic Acids Res. 22, 427-433. MEDLINE

Soule HD, Maloney TM, Wolman SR, Peterson Jr WD, Brenz R, McGrath CM, Russo J, Pauley RJ, Jones RF and Brooks SC. (1990). Cancer Res. 50, 6075-6086. MEDLINE

Vallet VS, Casado M, Henrion AA, Bucchini D, Raymondjean M, Kahn A and Vaulont S. (1998). J. Biol. Chem. 273, 20175-20179. MEDLINE

Varley MJ, Swallow JE, Brammar WJ, Whittaker JL and Walker RA. (1987). Oncogene 1, 423-430. MEDLINE

Viollet B, Lefrancois-Martinez AM, Henrion A, Kahn A, Raymondjean M and Martinez A. (1996). J. Biol. Chem. 271, 1405-1415. MEDLINE

Wood WM, Kao MY, Gordon DF and Ridgway EC. (1989). J. Biol. Chem. 264, 14840-14847. MEDLINE

Figure 1 Expression and DNA-binding activity of endogenous USF in breast epithelial cell lines. The endogenous levels of USF expression in different epithelial cell lines were examined by using mini nuclear extracts prepared in case under identical conditions. (a) Western blot analysis. For each cell line, the proteins present in 10 l of nuclear extract (3´10⁵ cell equivalents) were resolved by SDS - PAGE and transferred to nitrocellulose. The same blot was probed successively with USF2- and USF1-specific antibodies. Migrations of the 43-kDa USF1 and 44-kDa USF2 polypeptides are indicated at left. (b) EMSA. DNA-binding reactions were assembled using 0.1 ng of a 150-bp radiolabeled DNA fragment containing a USF-specific binding site and 0.5 l of mini nuclear extract. Competitor DNA (10 ng) under the form of a 30-bp oligonucleotide containing the USF consensus binding site was added as indicated. Migrations of the free probe and the major USF-DNA complexes are indicated at left

Figure 2 Activity of the RSV-Luc reporter in different cell lines. Transfections were carried out using the optimized conditions described in Table 1 and the resulting luciferase activities were determined by scintillation counting. For each cell line, a minimum of three independent transfection experiments were performed. The results were used to calculate averages, indicated by the height of each bar, and standard deviations, indicated in each case by the error bars

Figure 3 Expression of exogenous USF in different breast cell lines. Cells were transfected as described in Table 1 with expression vectors encoding either USF1, USF2, of USF2N, or with the corresponding empty vector, as indicated above each lane. USF overexpression was monitored by EMSA using in each case a 1.5-l aliquot of the whole cell extract prepared for luciferase assay. The radiolabeled probe (0.1 ng) was a 33-bp oligonucleotide containing the USF-consensus binding site

Figure 4 Transfectional activity of USF in normal breast epithelial cells. (a) Schematic representation of the pMLLuc and pU3MLLuc reporter plasmids containing the adenovirus major late minimum promoter with or without three upstream USF-binding sites. (b) The transcriptional activity of endogenous USF in MCF-10A cells was monitored by transfecting each reporter in the presence or absence of dominant negative mutants of the USF (A-USF) or Myc (A-Max) transcription factors, as indicated. The results from three independent experiments were analysed and plotted as described for Figure 2. (c) The transcriptional activity of exogenous USF in MCF-10A was monitored by cotransfecting the pU3MLLuc reporter with expression vectors for either USF1, USF2 or USF2N, or with the corresponding empty vector as a control. Shown are the averaged results from a minimum of three independent experiments

Figure 5 Activity of the pU3MLLuc reporter in different cell lines. Transfections were carried out exactly as described for Figure 2, except that the pU3MLLuc reporter was used instead of pRSV-Luc. Shown are the averaged results from three independent experiments

Figure 6 Transcriptional activity of exogenous USF in transformed breast cell lines. The transcriptional activity of exogenous USF1 and USF2 in the different cell lines was determined by cotransfecting pU3MLLuc with the indicated USF expression vectors or corresponding empty vector, as indicated above each bar. For comparison, the results obtained with the normal MCF-10A cells (Figure 4c) are also shown

Figure 7 Transcriptional activity of USF in HMECs (a) Cells at passage 8, 9 or 10, were seeded in 6-well plates at an initial density of about 1´10⁵ per well. Transfections were carried out the following day using either 6 l of Fugene 6 and 3 g of plasmid DNA (exp. 1) or 3 l of Fugene 6 and 2 g of plasmid DNA (exp. 2 and 3). In each case, half of the DNA was the pMLLuc or pU3MLLuc reporter and the other half was pSG5. Whole cell extracts were prepared 48 h after transfection for luciferase assay. Exp., experiment. (b) Overexpression of USF by transient transfection in HMECs was monitored by EMSA as described for Figure 3. (c) Activity of exogenous USF in HMECs. Cells at passage 10 or 11 were seeded at initial densities of 1 - 2´10⁵ well and transfected the following day using Fugene 6 (3 l). Cotransfected plasmids included 1 g of the pU3MLLuc reporter and 1 g of the SV40 expression vector for USF1, USF2, USF2N, or the corresponding empty pSG5 vector, as indicated above each bar

Figure 8 Expression of c-Myc in the different cell lines. The same blot used to determine the endogenous levels of USF in nuclear extracts from different cell lines (Figure 1a) was stripped and reprobed with Myc-specific antibodies. Migration of the 62-kDa c-Myc polypeptide is indicated at left. Quantitation of this band is shown below each lane

 Optimized conditions for transfection in various breast cell lines