A Composite Enhancer Element Directing Tissue-specific Expression of Mouse Mammary Tumor Virus Requires both Ubiquitous and Tissue-restricted Factors*

Mouse mammary tumor virus (MMTV) expression is restricted primarily to mammary epithelial cells. Sequences responsible for both the mammary-specific expression of MMTV and the activation of cellular oncogenes are located within two enhancer elements at the 5′-end of the long terminal repeat. Whereas the Ban2 enhancer (−1075 to −978) has been well characterized, the mammary-specific enhancer of MMTV from −956 to −862 has only recently been recognized as a key determinant of mammary-specific oncogene activation by MMTV. The present study identifies and characterizes three binding sites located within this element. Transient transfection of deletion and mutation constructs shows that all three sites contribute to the basal expression of MMTV in mammary cells. One of the binding activities (footprint I) is restricted to mammary cells, whereas the other two sites bind factors found in both mammary and nonmammary cells. The multimerized mammary-specific enhancer of MMTV on its own can enhance a minimal promoter in a mammary-specific fashion. However, the FpI binding site alone cannot mediate this effect. Thus, it is the binding of multiple factors in a combinatorial fashion that mediates the mammary-restricted expression of MMTV.

Infection by MMTV 1 leads to the development of mammary adenocarcinomas in laboratory mice (1,2). The retrovirus is passed horizontally by ingestion of milk-borne infectious viral particles or vertically, in certain strains, via the inheritance of an active proviral copy (3). Although MMTV is latently transforming, no oncogene is encoded in the viral genome (4). Integration of the provirus near the int genes results in the inappropriate expression of these cellular oncogenes and initiates events that lead to the formation of mammary tumors (5).
Entry into cells is achieved by the virion binding to a membrane receptor, which is expressed in many tissues (6). Low levels of MMTV RNA are detected in the epithelial cells from several tissues, such as salivary gland, lungs, kidney, and testes, as well as lymphoid tissue, but expression is approximately 500-fold higher in the lactating mammary gland (7). The MMTV LTR is widely used to direct the expression of transgenes to the mammary gland, but the basis for this specificity is not completely understood. The tissue specificity of MMTV expression is not due to its inducibility by steroid hormones. There are active and functional glucocorticoid receptors in most mouse tissues (8), yet MMTV expression is primarily detected in the mammary gland (7).
The 5Ј-end of the MMTV LTR contains enhancer elements that direct MMTV expression in a mammary-specific fashion (reviewed in Ref. 9). A complex array of protein binding sites in this region has been identified by gel shift and footprinting assays (10 -15). One region, termed the Ban2 enhancer, contains binding sites for four different proteins, including AP-2, an Ets-related factor, a member of the CTF/NF-I family, and an uncharacterized factor mp4 (11,14,15). Recently, we identified a second region in the 5Ј-end of the MMTV LTR that we termed the mammary-specific enhancer of MMTV (MEM element) (16). This element is important for both the mammary-specific expression of MMTV and the activation of cellular proto-oncogenes. Although the MEM element acts synergistically with the Ban2 enhancer in the MMTV LTR, multimerization of the MEM element itself can confer mammary cell-specific transcriptional activation (16).
In this study, we show that the MEM element is a composite element displaying at least three distinct footprinted domains that function synergistically. Deletion or mutation of individual binding sites decreases transcription from the MMTV promoter by up to 90% in mammary cells. Two of the domains are bound by ubiquitous factors, one of which is probably a member of the NF-1 family. At least one of the binding activities, corresponding to the protected region footprint I (FpI), appears to be restricted to mammary cells. However, multimerization of the FpI binding domain alone cannot activate transcription in mammary cells. Thus, it is the combination of multiple binding sites, one of which may be mammary-specific, that contributes to the mammary tropism of MMTV.

EXPERIMENTAL PROCEDURES
DNase I Footprinting-The DNase I footprinting assay was performed as described by Dynan (17). A 215-bp fragment of the MMTV promoter encompassing Ϫ969 to Ϫ754 of the MMTV LTR was uniquely labeled at the 5Ј-end of the noncoding strand. Approximately 80,000 cpm of probe were incubated with increasing amounts of crude nuclear extracts and 2 g of poly[d(I⅐C)] in a final reaction volume of 50 l for 15 min on ice. Fifty l of a solution containing 5 mM CaCl 2 and 10 mM MgCl 2 were added to each tube. A variable amount of DNase I was then added to each reaction and incubated for 1 min at room temperature. The reactions were terminated by the addition of 90 l of stop solution (200 mM NaCl, 30 mM EDTA, 1% SDS, and 100 g/ml yeast tRNA), extracted with phenol/chloroform, and ethanol-precipitated. DNA pellets were resuspended in 5 l of 0.1 N NaOH:formamide (1:2, v/v) containing xylene cyanol and bromphenol blue. The samples were boiled for 2 min, loaded onto a thin 8% urea polyacrylamide gel, and run in 1ϫ TBE for ϳ2 h at 1600 V. Following electrophoresis, the gel was dried and subjected to autoradiography.
Nuclear Extracts-Crude nuclear extracts were prepared from the cell lines listed in Table I according to the method of Dignam et al. (18). Ten to 20 plates of cells (T-175 flasks or 15-cm diameter dishes) were grown to confluency and harvested to make each extract. Protein concentrations were determined according to Bradford (19). Typical yields ranged from 5-18 mg/ml. Extracts were aliquoted, flash frozen in liquid nitrogen, and stored at Ϫ70°C.
Plasmids-The MMTV LTR in the MMTV-chloramphenicol acetyltransferase (CAT) vectors is a chimera derived from the C3H and GR strains of MMTV (GenBank TM accession numbers J02274 and V01175, respectively). The chimeric LTR has GR sequences from Ϫ291 to ϩ83 and C3H sequences from Ϫ1194 to Ϫ292 and from ϩ84 to ϩ99 Constructs with deletions in the MMTV LTR were made by cutting the vector with the indicated restriction enzymes, filling in the overhangs, and religating. A double-stranded 15-bp oligomer (TGAGGT-GAATTCTAG) was inserted at the Bsu36I site to alter the spacing between the footprinted regions. The FpIIm and FpIIIm constructs were made by inserting 41-bp oligomers with the indicated base pair changes between the Bsu36I and ClaI sites. The Amersham Pharmacia Biotech U.S.E. site-directed mutagenesis kit was used to generate the FpI mutant constructs indicated in Fig. 2. All base changes described were confirmed by sequencing.
To test enhancer activity, LTR sequences upstream of the basal MMTV promoter (Ϫ108 to ϩ99) in the MMTV-CAT reporter plasmid were replaced by three or four copies of a fragment (Ϫ969 to Ϫ862) encompassing the MEM element. Alternatively, a 33-bp oligonucleotide containing only the FpI site was multimerized and cloned upstream of the basal MMTV promoter. Constructs that contained four or six wildtype copies of the FpI site and a construct with six mutant copies were chosen and confirmed by sequencing.
Electrophoretic Mobility Shift Assay (EMSA)-EMSA probes were labeled by filling in the 5Ј-overhangs with Klenow, [␣-32 P]dCTP, and cold nucleotides. A 20-l reaction containing 15-25 g of crude nuclear extract and 1-2 g of poly[d(I⅐C)] was incubated for 15 min at 4°C in binding buffer (50 mM KCl, 10 mM Tris, pH 7.5, 5% glycerol, 1 mM EDTA, 2 mM MgCl 2 , 0.8 g of gelatin, 1.3 mM dithiothreitol). In some cases, cold competitor was then added and incubated for 10 min at 4°C. Finally, labeled probe (30,000 cpm) was added to each reaction and incubated 15 min at 4°C. The samples were loaded on a precooled 4% polyacrylamide gel and run in 0.25ϫ TBE at 250 volts for 2 h at 4°C. The gel was dried and subjected to autoradiography.

Multiple Binding Sites within a 107-bp Region of the MMTV
LTR-In our previous study (16), we identified a region in the 5Ј-end of the MMTV LTR that plays a role in the activation of cellular oncogenes, termed the MEM element. To begin to understand the mechanistic basis of the tissue specificity of MMTV, DNase I footprinting analysis was performed to identify protein binding sites in the MEM element (Fig. 1). Nuclear extracts from a human mammary carcinoma cell line (T47D) and a mouse fibroblast cell line (LtkϪ) were compared. FpIII was the 3Ј-most binding site and contained a consensus NF-I binding site of TTGGCN 5 GCCAA (25). FpIII binding activity was seen with both mammary and fibroblast extracts. The next binding activity, FpII, was also seen with both extracts. This small footprint corresponded to a previously identified binding site for an uncharacterized activity. Despite the fact that the binding activity is present in both mammary and nonmammary cells, it has been termed mammary cell-activating factor (10). A difference in binding activity between mammary and fibroblast extracts was observed in the third protected region, FpI. Although the protection was present most clearly with the T47D nuclear extracts, there was partial protection of this region using LtkϪ extracts. However, the footprint seen with the mammary extracts was accompanied by a DNase I hypersensitive site. Thus, the MEM element is composed of multiple binding sites, one of which appears to bind a tissue-restricted, if not mammary-specific, factor.
Synergistic Function of the Three Binding Sites-In order to determine the functional significance of the three binding sites delineated by footprinting analysis, a series of deletions and mutations were introduced into the MMTV LTR ( Fig. 2A). These MMTV-CAT constructs were transiently transfected into either T47D(A1-2) or LtkϪ cells to test their transcriptional activity. The largest deletion (⌬76), which removed 76 bp that included FpII, FpIII, and much of the FpI region, had the greatest effect, virtually ablating the basal expression of MMTV in mammary cells. This same deletion had no effect on transcription in fibroblast cells (compare Fig. 2B with Fig. 2C). Deletion of either much of FpI (⌬BB, Ϫ938 to Ϫ903) or of FpII/III (⌬BC, Ϫ903 to Ϫ862) resulted in a loss of promoter activity in mammary cells, suggesting that all three binding sites may be necessary for transcriptional activity.
This speculation was confirmed by making clustered point mutations in the three binding sites individually to assess the contribution of each subdomain. A 5-bp change in the FpIII binding site reduced activity of the MMTV promoter in mammary cells by 70%, and a 3-bp change in FpII/mammary cellactivating factor site reduced activity by 60% (Fig. 3B). A series of mutations along the FpI site (FpI mut 1-4) resulted in 50 -80% reductions of promoter activity. The spatial arrangement of the three binding sites of the MEM element is also critical for their functional synergy. A 15-bp oligonucleotide was inserted at the Bsu36I site separating FpI from FpII/III by 1.5 turns of the DNA helix while leaving the binding sites intact. This alteration in spacing resulted in a 70% drop in promoter activity. It appears that all three binding sites are needed, intact and in the correct context, in order to achieve mammary-specific transcription.
The MEM Element Can Function as an Independent Enhancer Unit-In the context of the MMTV LTR, the MEM element acts in synergy with the Ban2 element (16). To test whether the MEM element on its own is sufficient to enhance transcription, three or four copies of the enhancer were placed upstream of a minimal MMTV promoter (Fig. 3A). The minimal promoter (⌬Sst) had no activity in T47D(A1-2) cells. A single MEM element had little effect on this promoter. However, adding back three or four copies of the MEM element rendered the minimal promoter 5-10 times more active than the fulllength MMTV LTR in mammary cells (Fig. 3B). In fibroblast cells (Fig. 3C), there was no change in the transcriptional activity of any of the constructs. These data support the hypothesis that the MEM element is a key determinant in the mammary-specific expression of MMTV.
The MEM Element Is Not Involved in the Hormone Response-Studies on the hormonal regulation of MMTV have provided many of the insights that have shaped the current understanding of what is now termed "classical" hormone response elements. MMTV has a well characterized cluster of hormone response elements located between Ϫ80 and Ϫ200 in the MMTV LTR (26) that mediate the prodigious induction of promoter activity resulting from exposure to the appropriate steroid hormones. We wanted to test whether the MEM element was involved in the hormone responsiveness of the MMTV LTR. The constructs diagrammed in Fig. 2A were transiently transfected into T47D(A1-2) and LtkϪ cells and treated with 40 nM dexamethasone, a synthetic glucocorticoid, for 20 h before harvest. As shown in Fig. 4, none of the deletions or mutations in the MEM element significantly affected the hormone response of MMTV in either cell type. Thus, the MEM element is necessary for the basal transcription of MMTV but is not involved in the hormone response.
Characteristics of the FpI Binding Activity-To confirm the mammary-specific nature of the FpI-binding protein, a panel of nuclear extracts from 20 different cell lines (Table I) were tested for FpI binding activity. The extracts were bound to a probe containing the FpI binding site, and the resultant EMSA is shown in Fig. 5. Three DNA-protein complexes were formed with T47D extracts. Complex 1 was present in five of eight mammary extracts. In contrast, the complex was not found in any of 12 extracts from nonmammary cells both epithelial and nonepithelial in origin. Complex 2 was seen with all the extracts, and the mobility of a third, faster migrating complex (complex 3) varied among different extracts and was not always present. The presence of more than one EMSA complex using a FpI probe explains the partial footprinting seen using LtkϪ fibroblast extracts (Fig. 1). However, there is a mammaryrestricted binding activity associated with the FpI protected region, corresponding to complex 1.
Next, probes containing the four different FpI mutations (mut1-4) were used to test for FpI binding (Fig. 6A). Compared with a wild-type FpI sequence, none of the four mutants were able to bind complex 1 well. Additionally, mut2 did not form complex 2, and mut3/mut4 could not form complex 3. Because all four mutations lead to a reduction in the transcriptional activation by the MEM element (Fig. 2B), complex 1 exhibits the best correlation between FpI binding activity and transcriptional activity.
In order to determine the specificity of binding to the FpI probe, a competition EMSA was performed with a variety of competitor DNAs (Fig. 6B). A 10-fold molar excess of cold wild-type FpI oligonucleotide efficiently competed for binding of all the complexes. The FpI mutants could only compete for binding of the complexes they formed in Fig. 6A. Thus, mut2 was unable to compete well for complex 2 binding, and neither   5. The complex 1 FpI binding activity is found exclusively in mammary cell lines. EMSA experiments were performed using nuclear extracts made from the cell lines listed in Table I and a 33-bp probe containing the FpI binding site. Three specific complexes were identified and are indicated on the left. A fourth, faster migrating complex, is a nonspecific band that could be largely eliminated by the inclusion of poly[d(I⅐C)] during the incubation. Lanes 1-8 depict results with extracts from the indicated mammary cell lines, and lanes 9 -20 represent nonmammary cell lines. mut3 nor mut4 could compete for complex 3. The FpI mutants, with the exception of mut1, could not compete for complex 1 very efficiently. Two unrelated oligonucleotides could not compete for the binding of any of the complexes, demonstrating the specificity of the nuclear extracts for binding the FpI sequences.
The FpI-binding Protein Alone Cannot Activate Transcription-To test whether the FpI binding region was itself sufficient to activate transcription in mammary cells, reporter gene constructs containing either a wild-type or mutant 35-bp oligonucleotide spanning the FpI binding site were multimerized and cloned upstream of a minimal MMTV promoter (Fig. 7A). As before, the minimal MMTV promoter (⌬Sst) showed almost background levels of transcriptional activity in T47D(A1-2) cells (Fig. 7B). Adding back four or six copies of the wild-type FpI binding site failed to enhance the activity of the minimal promoter further. Therefore, the FpI domain requires the other domains of the MEM element to enhance transcription in mammary cells, even though the FpI domain alone appears to have mammary cell-specific binding activity.
Homology to a C/EBP Binding Site-The sequence of the FpI binding site shows similarity to the consensus binding sequence for the C/EBP family of transcription factors, including two CAAT-box motifs (Fig. 8A). A set of reciprocal competition gel shift assays revealed some similarities between FpI binding and C/EBP binding. EMSA experiments using T47D nuclear extracts and either a 33-bp FpI probe or a 20-bp C/EBP probe were performed and a 50-fold molar excess of competitor DNA was added. In Fig. 8B, both the 33-bp FpI oligonucleotide and a shorter 21-bp FpI oligonucleotide competed for binding to the FpI probe (lanes 2 and 3). Interestingly, a C/EBP binding site was also able to efficiently compete for the binding of complex 1 and partially for complex 2 (Fig. 8B, lane 4). The binding was specific, as a mutant C/EBP oligonucleotide could not compete for the complexes (lane 5). When a C/EBP binding site was used as a probe, both FpI oligonucleotides competed for binding almost as well as the C/EBP oligonucleotide (lanes 7-9). Again, binding to the C/EBP probe was specific because a mutant C/EBP binding site did not compete for the complexes (lane 10). These findings raise the possibility that a C/EBP family member contributes to the mammary-specific expression of MMTV and are discussed more fully below.

DISCUSSION
The 5Ј-end of the MMTV LTR is responsible for both the mammary-specific expression of the virus and the activation of cellular oncogenes (16). The fact that MMTV induces tumors primarily in the mammary gland appears to be linked with mammary-specific expression of the virus. A number of binding sites for nuclear proteins have been reported in this region of the LTR. These fall within two regions defined as having mammary-specific transcriptional activity: the Ban2 enhancer and the MEM element. When the MEM element is multimerized upstream of a minimal MMTV promoter, it can enhance transcription in a mammary-specific fashion. Similar experiments with the Ban2 enhancer show that it, too, can enhance transcription on its own (10, 15). However, we have shown that inactivation of either the Ban2 enhancer or the MEM element in the context of the LTR abrogates both the MMTV promoter and its ability to activate a nearby proto-oncogene promoter (16). Thus, the Ban2 and the MEM elements function synergistically when contained in a full-length LTR.
The Ban2 enhancer contains binding sites for four different proteins, including AP-2, an Ets-related factor, a member of the CTF/NF-I family, and an uncharacterized factor mp4 (11,14,15). The Ban2 fragment (Ϫ1075 to Ϫ978) upstream of the thymidine kinase promoter is active only in mammary cells and not in HepG2 liver cells (10,14). A 180-bp fragment (Ϫ1166 to Ϫ987) containing this enhancer can also target a transgene to the mammary gland of mice when linked to the SV40 promoter (27). Even when present as a single copy and independent of the MEM element, the Ban2 enhancer was able to direct mammary-specific transcription. This may be because the Ban2 enhancer is separated from the negative regulatory elements contained in the full-length MMTV LTR, or due to the use of a stronger heterologous promoter, rather than the very weak MMTV basal promoter. However, there were higher levels of transgene expression when the Ban2 and MEM elements were both present (10).
The MEM element has only recently been recognized as a distinct functional element (16). We have now characterized the functional interplay of the binding sites within the MEM element that constitute the active, tissue-specific enhancer. DNase I footprinting analysis revealed three protected regions within this element in extracts from T47D mammary carcinoma cells (Fig. 1). All three sites contribute to the mammaryspecific expression of the virus. Deletions that removed one, two, or all three of the binding sites reduced transcriptional activity by 85-96%, and base substitutions in individual sites reduced transcriptional activity by 50 -80%. These results with the MEM element contrast with the Ban2 enhancer, in which clustered point mutations in all four binding sites were required to reduce the activity by 80% (14). The spatial context of the MEM element binding sites with respect to one another is also important. When the three binding sites are left intact but a 15-bp insertion is introduced, the LTR can no longer support transcription in mammary cells. Thus, not only are all three binding activities of the MEM element important for the mammary-specific expression of MMTV, it may be that physical interaction between the factors or a common coactivator is required for MEM element activity.
In dissecting the MEM element, we have focused on the FpI protected region, because DNase I footprinting analyses revealed a difference in binding activity in extracts from T47D mammary carcinoma cells and fibroblasts. EMSA studies indicated three activities in T47D cell extracts that bound a FpI oligonucleotide. In order to be certain that there was a mammary-specific binding activity associated with the FpI factor, we assayed the binding activity of nuclear extracts from 20 cell lines representing mammary carcinomas, nontumor mammary epithelium, nonmammary epithelial cells, and nonepithelial cells (Table I). Complex 1 was found exclusively in mammary cells and was present in the majority of the mammary lines examined. A second complex, complex 2, was found in all cell types. Finally, a third, faster migrating complex was present in many cases, but its mobility varied depending on the cell line. Four different clustered point mutations within the FpI binding site decreased MEM element activity. Complex one correlated best with this functional data, as it was the only one of the three complexes of which the binding was decreased by all four mutations.
The FpI region cannot mediate mammary-specific transcription on its own, however, even in multiple copies. When the FpI binding site is multimerized upstream of a minimal MMTV promoter and separated from other two MEM element-binding proteins, there is no increase in transcriptional activity in a mammary cell line. Thus, the MEM element behaves as a composite enhancer, the function of which is dependent on functional synergy between tissue-specific (or at least tissuerestricted) and nonspecific factors.
The binding of multiple transcription factors to a composite enhancer element to allow cell-type specific transcription has been observed in a number of tissues, including the mammary gland, liver, lymphoid cells, heart, and muscle (28 -32). This combinatorial effect circumvents the requirement for tissuespecific transcription factors and instead mediates tissue-restricted gene expression by combining both ubiquitous and tissue-enriched transcription factors in a novel way. No factor specific to the mammary gland has been identified, although a couple of binding proteins were named for that putative property. Mammary gland factor is identical to Stat5 and is expressed in many tissues (33). Mammary gland factor/Stat5 does, however, mediate the prolactin signal and is important for the expression of milk proteins, such as ␤-casein and whey acidic protein (28). Recently, a functional Stat5 binding site has FIG. 9. Composite elements direct the mammary specific expression of MMTV. Shown is a schematic of the binding sites identified in the 5Ј-end of the MMTV LTR that participate in the mammary-specific expression of MMTV. The sites are clustered in two independently defined regions, the Ban2 enhancer and the MEM element. Boundaries of the binding sites are shown, as are alternate names or the putative factors that bind individual sites. been identified in the middle portion of the MMTV LTR and is important for the mammary-specific expression of MMTV (34). The mammary cell-activating factor was found to be a member of the Ets family of transcription factors and is not restricted to the mammary gland (35). Its activity has not been well characterized, but because of such a binding site in the Ban2 enhancer, it does play some role in the mammary-specific regulation of MMTV. Fig. 9 outlines the binding sites located within the two enhancer elements in the 5Ј-end of the MMTV LTR. Of the eight binding sites, six binding activities are found in both mammary and nonmammary cells. The uncharacterized mp4 factor was reported to show binding activity in mammary cell lines only (11), and the FpI complex 1 binding activity was shown to be restricted to mammary cell lines in this study. We found that the FpI region binds ubiquitous factors as well. An earlier study (10) that termed this region F4 observed nonspecific binding activities but not a mammary-specific complex as we have seen. Because we did not see the mammary-specific complex (complex 1) in all mammary cell lines, this may explain the discrepancy.
The competition EMSA data (Fig. 8) raise the interesting possibility that the FpI-binding protein may be a member of the C/EBP family of transcription factors (36). The 3Ј-half of the binding site is well conserved, and it has been documented that C/EBP binding sites may show significant homology to only one half of the palindrome (37). Three of the six isoforms, C/EBP ␣, ␤, and Ѩ, are expressed in the mouse mammary gland and are important for its development and function (38 -40). C/EBP ␤ is important for the mammary-specific expression of ␤-casein and whey acidic protein, in conjunction with other ubiquitous transcription factors, such as Stat5, YY1, NF-I, and the glucocorticoid receptor (28). Although the sequence of the FpI binding site, along with EMSA data, is consistent with a member of the C/EBP family, other data weigh against this conclusion. In contrast to the marked heat stability characteristic of C/EBP proteins, the FpI binding activity is heat labile at temperatures above 37°C. When nuclear extracts were heated to 60°C for 1, 2, 3, 4, or 5 min and used in EMSA analysis, the extracts retained the ability to form a complex with a C/EBP probe but not with a FpI probe (data not shown). Furthermore, the inclusion of antibodies against C/EBP ␣, C/EBP ␤, or C/EBP Ѩ neither upshifted any of the complexes binding an FpI oligonucleotide nor disrupted the complexes. Finally, none of these three forms of C/EBP exhibit a cell-type distribution as restricted as does the complex 1 binding activity. Initial attempts to enrich for complex 1 binding and study its biochemical properties have been hampered by its apparent lability to extensive manipulation.
Our previous results and the results from this study firmly establish the importance of the MEM element in the mammary-specific expression of MMTV. Further experiments to characterize the poorly defined binding activities of the mammaryspecific MEM and Ban2 elements need to be carried out. It will be critical to determine whether the FpI and mp4 factors binding these elements are truly restricted to the mammary gland and whether, at least in the case of FpI, it is related to the C/EBP family of transcription factors. It is apparent that a complex interaction of many transcription factors in a combinatorial fashion is required to allow the mammary-specific expression of MMTV. The mechanism by which combinatorial interactions are translated into tissue-specific, developmentally appropriate activity is a key question in the regulation of transcription.