Nuclear factor 1-C2 contributes to the tissue-specific activation of a milk protein gene in the differentiating mammary gland.

Members of the nuclear factor 1 (NF1) transcription factor family have been postulated to be involved in the regulation of milk genes. In this work we have been able to identify the splice variant NF1-C2 as an important member of a tissue-specific activating complex that regulates the milk gene encoding carboxyl ester lipase (CEL). Mutation of the NF1-binding site in the CEL gene promoter results in a drastic reduction of the gene expression to about 15% in mammary epithelial cells. Furthermore, we demonstrate that the NF1-C2 protein interacts with a higher affinity to the NF1-binding site in the CEL gene promoter than other NF1 family members do and that NF1-C2 in the mouse mammary gland is a phosphorylated protein. During development of the mouse mammary gland, binding of NF1-C2 to the CEL gene promoter is induced at midpregnancy, in correlation with the induction of CEL gene expression. The fact that the NF1-C2 involving complex remains throughout the lactation period and decreases during the weaning period, when the CEL gene is down-regulated, supports its importance in the regulation of CEL gene expression. To our knowledge, this is the first report identifying a specific, endogenously expressed NF1 isoform to be involved in the tissue-specific activation of a gene.

Mammary epithelial cell differentiation is a complex process in which quiescent ductular cells proliferate and form alveolar structures that express their specialized products, the milk proteins. The differentiation process is driven by the cooperative action of multiple steroid and peptide hormones (1). As milk protein genes are expressed only in differentiating epithelial cells, they act as markers for the differentiation of these cells. Studies of the regulation of milk protein genes are therefore of great interest for the understanding of mammary gland development and function. Extensive studies have defined multiple cis-acting elements and transcription factors involved in the regulation of milk protein production. These include binding sites for the glucocorticoid receptor (2), signal transducers and activators of transcription (3,4), CAAT/enhancer-binding protein (5), and nuclear factor 1 (NF1) 1 (4).
In a previous paper (6), we demonstrated that a member(s) of the NF1 family plays an important role in the expression of the milk protein gene carboxyl ester lipase (CEL). The CEL gene is highly expressed in the mouse mammary gland during pregnancy and lactation and in the mouse exocrine pancreas (7). However, we showed that the involvement of NF1 was mammary gland-specific. Initially, NF1 was identified as a factor required for the replication of adenovirus DNA (reviewed in Ref. 8) but has since been recognized as a potent transcriptional regulator of many viral and cellular genes (9,10). The NF1 family in vertebrates is composed of four members, NF1-A, NF1-B, NF1-C, and NF1-X, that are all differentially spliced and expressed in unique but overlapping patterns (11)(12)(13). NF1 proteins bind to DNA as homo-or heterodimers to the consensus binding site, TTGG(C/A)(N 5 )(G/T)CCAA. Functional NF1-binding sites have been characterized in genes expressed in almost every tissue. They have been shown to regulate both ubiquitous and tissuespecific genes (reviewed in Ref. 14). With such a diverse set of tissue-specific and developmentally regulated genes under the control of NF1 proteins, it appears likely that NF1 proteins play a major role in development.
Streuli et al. (15) have shown that there is a connection between NF1 binding and the differentiated stage of the mammary gland epithelium. This is further supported by Furlong et al. (16) who described a switch in expression and binding of different NF1 proteins as mammary epithelial cells move from the fully differentiated stage to the involution stage. These findings suggest that different NF1 family members might be important for both the development and the regression of the mammary gland epithelial cells. However, the particular forms of NF1 that are active in these processes have not been characterized.
Here we report that the particular isoform NF1-C2 binds to the NF1-binding site in the mouse CEL gene promoter. The DNA binding activity of NF1-C2, an ϳ50-kDa phosphoprotein, is increased at day 13 of pregnancy and decreased at involution in mice, which is in concordance with the expression of the CEL gene. We also show that binding of NF1-C2 increases the expression of the CEL gene in the mouse mammary epithelial cell line HC11 and that this activation is mammary gland-specific. Mutation of the NF1-binding site in the CEL gene promoter reduced the CEL gene expression to about 15%. Furthermore, by showing that NF1-C2 binds to the NF1-binding site with higher affinity than NF1-A1, we provide evidence that different NF1 family members can bind with different affinity to the same site.

EXPERIMENTAL PROCEDURES
Nuclear Protein Preparation-The inguinal mammary glands from different stages of development were dissected from F1:C57Bl6 ϫ CBA * This work was supported by grants from the Swedish Medical Research Council, Assar Gabrielsson Foundation, and Fredrik and Ingrid Thuring Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
EMSA-The complementary oligonucleotides (5Ј-GTTCTGCTTG-GCGTGTTATCAAG-3Ј and 5Ј-AGCAACCTTGATAACACGCCAAG-3Ј) were annealed and radiolabeled by filling in with Klenow polymerase in the presence of [␣-32 P]dCTP, creating a probe referred as the NF1 oligonucleotide, representing Ϫ1792 to Ϫ1764 of the mouse CEL promoter (6). A GC box oligonucleotide was created by annealing and end labeling the complementary oligonucleotides 5Ј-CTGAGGGGGTA-GAGGGGAGGGAGTGC-3Ј and 5Ј-TCAGGCACTCCCTCCCCTCTAC-CCCC-3Ј and a USF oligonucleotide by annealing and end labeling the complementary oligonucleotides 5Ј-TCTGTCCCAGAAGTCACGTG-3Ј and 5Ј-CCGAGCACGTGACTTCTGGGA-3Ј. The WAP oligonucleotide was the same as described previously (20). Nuclear extracts (4 -8 g) were incubated with 1.5 g of poly(dI-dC) and 25,000 cpm of labeled probe in EMSA binding buffer (20 mM Hepes, pH 7.9, 50 mM KCl, 10% glycerol, 2 mM MgCl 2 , 0.5 mM EDTA, 0.1 mg/ml bovine serum albumin, 0.5 mM dithiothreitol) in a 20-l reaction volume for 15 min at room temperature. For supershift experiments, 2 l of anti-NF1 antibody (rabbit polyclonal antiserum, 8199, reacting with the C-terminal half of NF1-C, kindly provided by Dr. N. Tanese, New York University Medical Center, New York) or 2 l of anti-Stat 5a antibody (Santa Cruz Biotechnology) were included in the binding reaction and preincubated for 15 min before addition of probe. DNA-protein complexes were resolved on a 6% polyacrylamide gel (Tris glycine, 5% glycerol).
Western Blot Analysis-Nuclear extracts (20 g) were electrophoresed through a 10% SDS-polyacrylamide gel followed by electroblotting onto Hybond-P filter (Amersham Biosciences). For the dephosphorylation experiments, nuclear extracts (20 g) were treated with 1.5 units of potato acid phosphatase (PAP) (Sigma) in 0.1 M BisTris buffer (pH 6.0) in a 60-l reaction volume at 30°C for 1 h. To detect endogenous NF1-C, filters were incubated with a 1/1000 dilution of anti-NF1 antibody (8199), and the primary antibody was detected with peroxidaseconjugated anti-rabbit IgG using the BM Chemiluminescence Blotting Substrate peroxidase (Roche Molecular Biochemicals) and ECL films (Amersham Biosciences). In the overexpression experiments the overexpressed NF1 proteins were detected by incubating the filters with anti-HA antibody (Roche Molecular Biochemicals) or anti-NF1-C antibody (8199). The primary antibodies were detected with peroxidaseconjugated anti-mouse or anti-rabbit IgG (Roche Molecular Biochemicals), respectively.
UV Cross-linking-Nuclear extracts (20 g) were incubated with 4.5 g of poly(dI-dC) and 50,000 cpm of labeled probe in EMSA binding buffer in a 50-l reaction volume for 15 min at room temperature. The reactions were UV cross-linked in a UV Stratalinker 2400 (Stratagene) at 254 nm for 15 min and separated on a 10% SDS-polyacrylamide gel.
The gel was dried and exposed to an x-ray film at Ϫ70°C. RNA Analysis-The inguinal mammary glands from different stages of development were dissected from F1:C57Bl6 ϫ CBA mice. Total RNA was extracted from these glands, HC11 cells, and AR4 -2J cells by Trizol (Invitrogen) according to the manufacturer's instructions. Poly(A) ϩ RNA was purified using Oligotex mRNA kit (Qiagen). RT-PCR experiments were carried out using the Titan TM One Tube RT-PCR System (Roche Molecular Biochemicals). The primers used for NF1-C amplification were "FwdC," 5Ј-GCCGGCATGAGAAGGACTCTAC-CCA-3Ј (bp 1164 -1188, GenBank TM accession number Y07693), and "RevC," 5Ј-AGGAGGGATGGGAAGGCAACCTCGG-3Ј (bp 1736 -1760, GenBank TM accession number Y07693), yielding a 597-bp fragment of the mouse NF1-C2 and a 520-bp fragment of the mouse NF1-C5. For mouse GAPDH amplification, the primers used were 5Ј-CACCACCAT-GGAGAAGGCCGGGGCC-3Ј and 5Ј-TTGAAGTCGCAGGAGACAAC-CTGGT-3Ј, yielding a 554-bp fragment. For each reaction, 10 ng of poly(A) ϩ RNA was used, and the reactions were incubated at 50°C for 30 min, 97°C for 2 min, followed by 10 cycles of 1 min at 97°C, 1 min at 55°C, and 4 min at 68°C, and 30 cycles (or 20 cycles for GAPDH) of 30 s at 97°C, 30 s at 55°C, and 1 min at 68°C. From cycle 11 the 68°C step was extended by 5 s every cycle. Finally the reactions were incubated at 68°C for 7 min.
For Northern blotting, poly(A) ϩ RNA was separated on a 1% agarose/ formaldehyde gel and transferred to a GeneScreen Plus TM (PerkinElmer Life Sciences) nylon filter. The filter was hybridized with probes detecting NF1-C (a 550 bp-fragment excised by NaeI/BglII digestion from pCHNF1-C2 kindly provided by Dr. R. M. Gronostajski, Lerner Institute (19)) or human ␤-actin (CLONTECH).
Cell Extracts and Reporter Gene Assays-HC11 cells, stably transfected with CEL promoter/luciferase constructs with an intact (mCEL-1831Luc) or mutated (mCEL-1831NF1:1mutLuc) NF1-binding site (previously described (6)) were grown to different degrees of confluence and then harvested as described previously (21). Luciferase assays were performed using the Promega kit with 50 l of cell lysate and assayed in a luminometer (Berthold). The luciferase activity was normalized to the protein concentration of each extract, determined by the method of Bradford (18).

NF1-C Binding to the Mouse CEL Gene Promoter Is Affected by the Cellular Differentiation
Stage-Our earlier promoter studies in mouse mammary gland-derived cells revealed that a major positive element in the CEL gene promoter interacts with a member(s) of the NF1 family (6). The CEL gene is activated between day 11 and 14 of pregnancy in mice, and we wanted to analyze if this activation correlates with binding of NF1 at this specific stage of differentiation. EMSA analysis with an oligonucleotide including the NF1-binding site and nuclear extract prepared from mouse mammary gland tissue at different stages of development revealed that there is an increased intensity of the NF1 complex at day 13 of pregnancy (P13) (Fig. 1A). The intensity is maintained at day 16 of pregnancy (P16) and at day 1 of lactation (L1) but is reduced 2 days after weaning (W2). An EMSA with the same extracts but an unrelated probe containing a GC box confirmed that the difference in NF1 binding to the NF1-binding site was not due to unequal quantification of the extracts (Fig. 1A).
Epithelial cells have been shown previously (12) to express factors of the NF1-C family. We preincubated the EMSA binding reaction with an anti-NF1 antibody that specifically recognizes the C-terminal domain of NF1-C proteins (Fig. 1B). Because the NF1 complex was supershifted, we could conclude that it contains NF1-C. No supershifted band was observed with a control anti-Stat5a antibody. The specificity of the antibody is shown in Fig. 1C. Together these results indicate that NF1-C binding could be coupled to the developmental regulation of the CEL gene. Furthermore, by EMSA and supershift analysis we have demonstrated that NF1-C also interacts with the promoter of the rat whey acidic protein (WAP) gene (data not shown), another milk protein gene induced simultaneously with the CEL gene. This suggests that NF1-C is important for the expression of different milk genes induced at midpregnancy.
An NF1-C Protein of ϳ50 kDa Binds to the CEL Gene Promoter-Earlier reports (22,23) have shown that NF1 proteins can range in sizes from 30 to 100 kDa, which is due to the many isoforms and different kinds of post-translational modifications such as phosphorylation and glycosylation (14). To examine the sizes of the NF1-C proteins in the extracts used, a Western blot was performed. This analysis revealed that the expression patterns of the NF1-C proteins vary during mammary gland development (Fig. 2). The NF1-C protein of ϳ50 kDa increases and decreases with the degree of differentiation in a pattern similar to the binding pattern observed in the EMSA experiment. Hence, we conclude that this protein is most likely responsible for the interaction with the NF1-binding site in the CEL gene promoter. The ϳ74-kDa NF1 protein appearing in W2 is presumably the same as that described earlier (16) as an NF1 protein triggered in early involution of the mouse mammary gland. Our data demonstrate that even this factor is an NF1-C protein since we used an NF1-C-specific antibody. This shows that different NF1-C proteins are present in the mammary gland during development and regression.
One way to determine the size of DNA-binding proteins is by covalently cross-linking the proteins to its regulatory sequence using UV light. Because the efficiency of UV cross-linking is usually low, it is exceedingly rare for more than one crosslinking event to occur in a complex. Consequently, the observed molecular masses, when using the NF1-binding site, are likely to be those of monomers rather than dimers. Hence, this gives us the opportunity to investigate if the ϳ50-kDa protein is responsible for the interaction with the NF1-binding site. As can be seen in Fig. 3, the indicated species with high intensity generated in the extract from L1 were reduced in the extract from W2. The band corresponding to free, unbound probe was estimated to ϳ30 kDa, which was subtracted from the size of the bound species. The resulting size of about 50 kDa is in agreement with that of the NF1-C protein observed in the Western blot analysis.
The ϳ50-kDa NF1-C Protein Has Activation Potential in Mammary Gland Epithelial Cells-To analyze the activation potential of NF1-C we used the mammary gland epithelial cell line HC11. The cells were grown to different degrees of confluence in media containing epidermal growth factor and insulin, because increased confluence has been shown previously to affect the differentiation ability. EMSA analysis of nuclear extracts from the different stages of confluence showed a change in NF1-C binding activity (Fig. 4A). Surprisingly, there was a higher intensity of the NF1-C complex, in the extract from cells grown to the lowest degree of confluence (stage 1), although in extract from cells grown to the highest degree of confluence (stage 3) there was no binding of this complex at all. We observed no difference in DNA binding activity for these extracts to the control GC box oligonucleotide (data not shown). Western blot analysis also confirmed that the ϳ50-kDa proteins were most abundant at stage 1, and minimal levels were found in the extract from stage 3 (Fig. 4B). Hence, HC11 cells provided us with a suitable system to investigate the activation potential of the ϳ50-kDa protein in a mammary epithelial context, because we now had a stage where the factor was present and one in which it was not.
HC11 cells stably transfected with the mCEL-1831Luc and the mCEL-1831NF1:1mutLuc constructs (6) were grown to different degrees of confluence and then analyzed with a luciferase reporter gene assay and EMSA. The overall activity of the two constructs increased with increasing degrees of confluence (Fig. 4C). However, the proportional difference in luciferase activity between the two constructs decreased as the cells became more confluent (Fig. 4D). At stage 1 the activation potential was 7-fold higher for the wild type construct compared with that of the mutant construct. At the highest stage of confluence there was almost no difference at all. By comparing the results from the reporter gene analysis and the EMSA we suggest that the presence and binding of the ϳ50-kDa NF1-C protein to the NF1-binding site in the CEL gene promoter activates the CEL gene. Mutation of the NF1binding site precludes binding of NF1-C, and accordingly the expression was reduced to ϳ15%.
NF1-C2 Is the Specific Isoform Responsible for CEL Promoter Activation-Because the ϳ50-kDa protein is the dominant NF1-C protein in HC11 cells as well as in the mammary gland at P13 to L1, we wanted to investigate which NF1-C isoform it represents. Based on the sequence of the mouse NF1-C gene, we devised an RT-PCR strategy that allowed us to detect and distinguish the different NF1-C isoforms that are known to exist in mouse (Fig. 5A). Poly(A ϩ ) RNA from HC11 cells at stages 1 and 2 of confluence was isolated and subjected to RT-PCR (Fig. 5B). Two NF1-C products were detectable corresponding to the isoforms NF1-C2 (exon 9 spliced) and NF1-C5 (exons 9 and 10 spliced). However, the expression level of NF1-C5 was barely detectable which implies that the ϳ50-kDa protein is NF1-C2. Subcloning and sequencing verified the identity of the NF1-C2 transcript.
The Increased Binding of NF1-C2 during Pregnancy Is Not Regulated at the mRNA Level-As shown in Fig. 1A and Fig. 2, the binding of NF1-C2 to the NF1-binding site in the CEL gene promoter increased and decreased congruently with the expression of the NF1-C2 protein. To investigate if the amount of NF1-C was regulated at the transcriptional level, a Northern blot was performed with mRNA prepared from different stages of the mammary gland (Fig. 6). Two major transcripts of ϳ4 and 6.5 kb were detected throughout mammary gland development, from day P10 to L1, whereas the mRNA level was drastically reduced in the involution stage (W2). The relationship between these two transcripts is not clear, although it is believed that the larger transcript is a precursor of the smaller mRNA (24). NF1-C transcripts were also detected in the virgin stage (V) at levels comparable with those at day P10 to L1 (data not shown). Our data are in contrast with Mukhopadhyay et al. (20) who could not detect any NF1-C transcripts in the lactating mouse mammary gland. The reason for this discrepancy is unclear, but the fact that we could detect not only NF1-C transcripts but also NF1-C proteins in the lactating mammary gland, as well as similar findings by Kane et al. (25), demon-strates that NF1-C proteins are indeed present. The presence of NF1-C transcripts in early stages of the mammary gland development (V and P10) indicates that the expression of

FIG. 4. Evaluation of the effect of NF1-C binding.
A, EMSA was performed with the NF1 oligonucleotide and extracts from HC11 cells grown to different degrees of confluence as indicated. The arrow indicates the NF1-C complex. B, Western blot with nuclear extracts from HC11 cells grown to different degrees of confluence. The extracts were run on a 10% SDS-polyacrylamide electrophoresis gel and blotted onto a Hybond-P filter. The filter was incubated with the anti-NF1-C antibody (8199). The arrow indicates the ϳ50-kDa NF1-C protein. C, luciferase activity was measured in HC11 cells stably transfected with the mCEL-1831Luc and mCEL-1831NF1:1mutLuc constructs, grown to different degrees of confluence as indicated. The relative luciferase activity was normalized to the total protein concentration of each extract. Data represent mean and S.E. of at least three independent experiments. D, to illustrate the proportional difference between the two promoter constructs, the average luciferase activity of the wild type construct from each stage was divided with the average activity of the mutant construct from each stage.

NF1-C2 Contributes to CEL Gene Activation in Mammary Gland
NF1-C proteins at day P13 is not regulated at the transcriptional level.
The NF1-C2 Protein in the Mouse Mammary Gland Is a Phosphoprotein-It is known that NF1 proteins can be phosphorylated in vivo (26 -28). We therefore wanted to investigate if the NF1-C2 protein in the mammary gland is a phosphorylated protein. Nuclear extracts from HC11 cells at stage 1 of confluence were treated with potato acid phosphatase (PAP) and analyzed by Western blot using the anti-NF1-C antibody (Fig. 7). This analysis revealed that PAP converted the ϳ50-kDa NF1-C2 protein to a faster migrating species with the size of about ϳ35 kDa. The data indicate that the NF1-C2 protein involved in binding to the NF1-binding site in the CEL gene promoter is a phosphoprotein. Analysis of the same PAPtreated extract by EMSA and supershift experiment revealed that PAP converted the NF1-C2 complex to the faster migrating complex that can be seen in Fig. 1A in the extracts P13-L1 (data not shown). The existence of these less phosphorylated NF1-C2 proteins in extracts not treated with PAP is probably an artifact that results from endogenous phosphatase activity in the nuclear extracts because the intensity of this complex is increased with longer incubation times at room temperature. Taken together, these data suggest that the NF1-C2 protein responsible for the interaction with the NF1-binding site in the CEL gene promoter is a phosphorylated protein. However, a phosphorylation/dephosphorylation event cannot be responsible for regulating the binding of this factor because NF1-C2 proteins of only one degree of phosphorylation, the ϳ50-kDa proteins, is detected during mammary gland development.

NF1-C2 Has Higher Affinity Than NF1-A1 to the NF1-binding Site in the CEL Promoter-Northern blot analysis revealed
that not only the NF1-C gene, but also the NF1-A, -B, and -X genes are expressed in the mammary gland during pregnancy and lactation as well as in HC11 cells (data not shown). However, the facts that the NF1 complex binding to the NF1binding site in the CEL gene promoter can be supershifted with the NF1-C-specific antibody and that only a few strong bands appear in the UV cross-linking experiment suggest that this site might be specific to NF1-C2 or that NF1-C2 binds to this site with higher affinity than other NF1 family members. To investigate this we overexpressed the NF1-C2 or NF1-A1 proteins in HC11 cells, prepared nuclear extracts, and investigated DNA binding with EMSA (Fig. 8A). When overexpressing NF1-A1, a new, weak, and slower migrating band appeared, but the endogenous NF1-C2 complex remained at the same intensity. In contrast, when overexpressing NF1-C2 the new band that appeared had strong intensity, whereas the endogenous band was almost outcompeted. EMSA with an oligonucleotide containing a binding site for upstream stimulating factor (USF oligonucleotide) confirmed that the different extracts were equally quantified. A Western blot analysis revealed that the recombinant NF1-A1 and NF1-C2 proteins were equally expressed (Fig. 8B). These data demonstrate that NF1-C2 binds to the NF1-binding site in the CEL promoter with higher affinity than NF1-A1.
If an NF1-binding site is found to which NF1-A1 has higher affinity than NF1-C2, i.e. the opposite binding preferences, these NF1 isoforms would have different DNA binding specificities as well. The rat NF1-A1 has been reported previously (29) to bind to the adenovirus replication origin with higher affinity than rat NF1-C2, which thus implies that the DNA binding specificities might differ between these two isoforms. We investigated this by EMSA using the same extracts and the adenovirus replication origin oligonucleotide. In contrast to the results with rat NF1s, we found that mouse NF1-C2 had higher affinity to the adenovirus replication origin than that of mouse NF1-A1 (data not shown). Accordingly, these isoforms seem to have the same DNA-binding specificity, which is in agreement with all earlier reports so far. In conclusion, our data suggest Extract from HC11 cells grown to the lowest degree of confluence (stage 1) was treated or not treated with 1.5 units of PAP and then run on a 10% SDS-polyacrylamide electrophoresis gel. The gel was blotted onto a Hybond-P filter, which was incubated with the anti-NF1-C antibody (8199) as described under "Experimental Procedures." The arrows indicate the NF1-C proteins before and after PAP treatment. that NF1-C2 binds to the CEL gene promoter with a comparatively higher affinity than that of NF1-A1.
NF1-C2 Contributes to the Tissue-specific Expression of the CEL Gene in the Mammary Gland-We have shown previously (6) that the importance of NF1 in CEL gene regulation is mammary gland-specific because no effect of NF1 binding could be detected in pancreatic cells. These studies were performed in the rat pancreatoma cell line AR4 -2J, which like pancreas expresses CEL at high levels (30). We then speculated that the different NF1 complexes detected were composed of different NF1 family members or reflect tissue-specific combinations. However, EMSA and supershift experiments in the present study show that the same NF1-C2 complex that is involved in activation of the CEL gene in mammary epithelial cells also binds to the NF1-binding site in AR4 -2J cells (Fig. 9A). Furthermore, RT-PCR with mRNA isolated from these cells also confirmed the presence of NF1-C2 (Fig. 9B). This suggests that NF1-C2 is not a tissue-specific activator itself but interacts with other factor(s) in the mammary gland to mediate tissue specificity.

DISCUSSION
Transcriptional control plays a central role in determining the level of gene expression in various tissues during development and differentiation. Differential gene expression is controlled by a complex regulatory network in which special-ized transcription factors such as NF1 relay signals to specific target genes. Because of the rather ubiquitous expression of the NF1 genes, NF1-A, -B, -C, and -X, they were first believed to be basal transcription factors that contribute to the expression of "housekeeping" genes. However, the NF1 family has since been shown to be involved in the transcriptional activation and repression of both ubiquitous and tissuespecific genes. The existence of a great number of structurally different NF1 splice variants, their differential expression, as well as the involvement of NF1-binding sites in tissue-specific gene expression suggest that individual isoforms may have distinct functions (14).
The observation of differentiation-related changes in NF1 protein binding during preadipocyte differentiation (31) as well as differentiation of hematopoietic cell lines (32) suggests a potential role for NF1 proteins in developmental processes. Our study provides further support for this suggestion because alterations in expression and binding of NF1 accompany cellular differentiation of the mouse mammary gland. It is also apparent that the differentiation-related change in NF1 binding is important for the expression of milk protein genes. Binding of NF1 has been reported to play a critical role in determining the overall activity of the rat whey acidic protein (WAP) gene, another milk protein gene induced at about the same time point during mammary gland development as the CEL gene (4). In transgenic mice the transgene expression was totally abrogated when mutating the NF1-binding sites in the WAP promoter. In this study we report that binding of NF1-C2 to the CEL gene promoter is increased at day 13 of pregnancy. We have earlier shown that the CEL gene is activated in mouse between day 11 and 14 of pregnancy. Furthermore, by EMSA and supershift analysis we have demonstrated that the NF1 protein interacting with the WAP promoter NF1-binding site is NF1-C2 as well. These findings suggest that there is a connection between induction of milk genes and the differentiation-related binding of NF1-C2 at midpregnancy. It is also obvious from this study that different NF1-C proteins are present in the mammary gland during development and regression.
Northern blot analysis showed the presence of NF1-C transcripts throughout all observed stages of the mammary gland development. This is in contrast to Mukhopadhyay et al. (20) who recently demonstrated that NF1-C transcript is barely detectable during lactation. This contradictory observation is hard to explain, but because the antibody that we used is NF1-C-specific it is obvious that there are NF1-C proteins in the mammary gland during lactation. Our Northern blot analyses revealed that not only the NF1-C gene but also the NF1-A, -B, and -X genes are expressed in the mammary gland during pregnancy and lactation (data not shown). This is in accordance with the report by Mukhopadhyay et al. (20) which showed that the major NF1 transcripts present during mammary gland development are NF1-A4, NF1-B2, and NF1-X1. However, even though there are NF1 transcripts, one cannot be sure that the corresponding NF1 protein products exist. This is illustrated by the fact that we found NF1-C transcripts in cells from virgin as well as p10, where there apparently exist few or no NF1-C proteins. Accordingly, the amount of NF1-C proteins in mammary epithelial cells is not mainly regulated at the transcriptional level.
Earlier it was believed that all NF1 proteins bind to a given DNA-binding site with apparently similar affinities (33,34), but the emerging view is that the affinity can differ (29). In this report we clearly demonstrate support for this fact as we show that NF1-C2 binds with higher affinity than NF1-A1, another isoform expressed in the mammary gland (25), to the NF1binding site in the CEL promoter. Furthermore, in accordance with earlier reports (29,34), we found no difference in DNA binding specificity between the different isoforms.
Differently spliced forms of NF1-C have been observed to have strikingly different activation properties in different tissues or even on different promoters in the same cell type. Thus, different NF1 isoforms modulate transcription in a cell typespecific as well as promoter-specific manner, and the same isoform can even be an activator in one cell type and a repressor in another (11,(35)(36)(37)(38)(39)(40). Overexpression experiments of human NF1-C isoforms in yeast showed that NF1-C5 (lacks exon 9 ϩ 10) was the strongest activator, whereas NF1-C2 (lacks exon 9) had no activating ability at all (37,38). In contrast, we show in this study that NF1-C2 has activation capability in the mouse mammary gland. Mutation of the NF1-binding site in the CEL gene promoter prevented binding of NF1-C2, which resulted in a decreased CEL gene expression to about 15%. NF1-C2 can also bind to the CEL gene promoter in pancreatic cells without eliciting activation, demonstrating that the activation is tissuespecific. Furthermore, NF1-C2 had no activation potential when binding to the CEL gene promoter NF1-binding site in cotransfection experiments in JEG-3 choriocarcinoma cells shown earlier (41) to be NF1-C-deficient (data not shown). These observations indicate that NF1-C2 alone cannot mediate activation but needs to cooperate with other factors present in the mammary gland. Gao and Kunos (42) showed that one of the mechanisms whereby NF1 can affect the expression of genes is by recruiting cell type-specific cofactors, and the role of NF1-C2 in the CEL promoter could be to recruit a mammary gland-specific coactivator.
The transcriptional activity of NF1 has been shown to be modulated by phosphorylation (26 -28). However, the activation ability of NF1-C2 in the mammary gland is not regulated by a phosphorylation/dephosphorylation event because the level of phosphorylation of NF1-C2 does not change during mammary gland development. The precise mechanism of NF1mediated activation is unknown. Possibly, the alternative splicing of NF1-C gene generates proteins with different binding specificities for potential coactivators or corepressors. Direct interactions have been observed between NF1 and components of the basal transcriptional machinery as well as a number of other transcription factors and coactivators (14).
Other mechanisms, such as modulated affinity for certain promoter contexts or influence on chromatin structure, are also possible explanations. Interactions have indeed been demonstrated with histones H1 (43) and H3 (44). Further studies are needed to determine the mechanism by which NF1-C2 and other isoforms modulate transcription and to identify potential isoform-specific coactivator proteins.