The enhancer of the human transferrin gene is organized in two structural and functional domains.

We previously identified a 300-base pair long enhancer, located 3.6 kilobases upstream of the cap site of the human transferrin gene. A 5' deletion up to position 86 of the enhancer resulted in complete loss of the enhancer activity. Here we show by competition footprint analysis, gel retardation assays, and transient expression studies in hepatoma and HeLa cells that the enhancer is composed of two distinct structural and functional domains, A (nucleotides 1-86) and B (nucleotides 87-291). Each domain is a proto-enhancer of a different type. Domain A is a proto-enhancer that, when multimerized, is able by itself to stimulate transcription from the heterologous SV40 promoter, both in Hep3B and HeLa cells. It contains the octanucleotide TGTTTGCT sequence and is the binding site of two liver-specific nuclear factors and of a different HeLa nuclear factor. Domain B contains four binding sites interacting with several liver nuclear proteins. In order to bind, any of these proteins requires the presence of all the others. This domain is able to block the activity of a downstream negative element, but it has no enhancer activity by itself. In the presence of the transferrin promoter, full enhancer activity requires the association of the two domains A and B.


The Enhancer of the Human Transferrin Gene Is Organized in Two
Structural and Functional Domains* (Received for publication, September 24,1990) Franpois Boissier, Corinne Auge-Gouillou, Evelyne Schaeffer, and Mario M. ZakinS From the Laboratoire d%xpression des Ggnes Eucaryotes Znstitut Pasteur, 75724 Paris Cedex 15, France We previously identified a 300-base pair long enhancer, located 3.6 kilobases upstream of the cap site of the human transferrin gene. A 5' deletion up to position 86 of the enhancer resulted in complete loss of the enhancer activity. Here we show by competition footprint analysis, gel retardation assays, and transient expression studies in hepatoma and HeLa cells that the enhancer is composed of two distinct structural and functional domains, A (nucleotides 1-86) and B (nucleotides 87-291). Each domain is a proto-enhancer of a different type.
Domain A is a proto-enhancer that, when multimerized, is able by itself to stimulate transcription from the heterologous SV40 promoter, both in Hep3B and HeLa cells. It contains the octanucleotide TGTTTGCT sequence and is the binding site of two liver-specific nuclear factors and of a different HeLa nuclear factor. Domain B contains four binding sites interacting with several liver nuclear proteins. In order to bind, any of these proteins requires the presence of all the others. This domain is able to block the activity of a downstream negative element, but it has no enhancer activity by itself. In the presence of the transferrin promoter, full enhancer activity requires the association of the two domains A and B.
The molecular basis of the human transferrin (Tf)' gene expression was first analyzed in liver (Brunel et al., 1988;Schaeffer et al., 1989) and more recently in Sertoli cells of rat testis (Guillou et al., 1991). We previously showed that, in liver, the regulatory elements are distributed in four functionally different regions: a minimal cell type-specific promoter, a distal promoter, a negative-acting region, and an upstreamlocated enhancer. The Tf enhancer, situated between -3.6 and -3.3 kb relative to the transcriptional start site of the gene, fulfills the criteria established for enhancers. It stimulates transcription of an homologous and heterologous promoter in an orientation-and position-independent manner. Transient expression studies revealed that the Tf enhancer is * This work was supported by the Centre National de la Recherche Scientifique (Unite de Recherche Associee 1129) and the Institut National de la Sante et de la Recherche Medicale Grant 87 1018. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBankTM/EMBL Data Bank with accession number(s) 62823.
$ To whom correspondence should be addressed. not tissue-specific since it is able to stimulate transcription from the SV40 promoter both in hepatoma and in epithelial carcinoma HeLa cells. Thus it appears that the liver-specific expression of the Tf gene is only mediated by the nuclear protein transferrin-liver factor 1 and the CCAAT/enhancer binding protein-related factor, binding to adjacent sites of the proximal Tf promoter (Schaeffer et al., 1989;Ochoa et al., 1989). Recent studies of cellular enhancers have started to reveal the complexity of their &-acting elements; some of them were shown to present a modular organization resembling that of the viral SV40 enhancer (Herbst et al., 1989;Godbout et al., 1988;Grayson et al., 1988;Kruse et al., 1988;Knepel et al., 1990). An extensive analysis of the SV40 enhancer defined distinct levels of organization of an enhancer (Herr and Clarke, 1986;Fromental et al., 1988;Ondek et al., 1988). The basic structural unit was called an enhanson and was defined as the binding site for a trans-acting factor. The combination of two or more copies of the same enhanson or of different enhansons generates an enhancer module (Schaffner et al., 1988;Dynan, 1989), also called a proto-enhancer. In the SV40 enhancer, each enhanson displays a characteristic cell typespecific activity (Kanno et al., 1989).
The aim of this work was to determine the structural and functional organization of the Tf enhancer. The interactions of nuclear factors with different regions of the enhancer were analyzed by DNase I footprint and competition experiments and by gel retardation assays. They defined two distinct structural domains: the domain A composed of one enhanson and the domain B composed of four enhansons.
To establish the function of each domain, single and multiple copies of each domain were cloned upstream of the Tf and the SV40 promoter in expression vectors that were introduced into Hep3B and HeLa cells.
The data demonstrate the modular organization of the Tf enhancer and clarify the role of each domain in transcriptional stimulation. They stress the importance of the octanucleotide TGTTTGCT sequence, which is also present in several other liver-expressed genes. This sequence was detected in domain A, which is the binding site of distinct proteins in different cell types.

MATERIALS AND METHODS
Materials-T4 DNA ligase, DNA polymerase I (Klenow fragment), T4 polynucleotide kinase, and restriction endonucleases were purchased from New England Biolabs; alkaline phosphatase and F-12 and Dulbecco's modified Eagle's media from Boehringer Mannheim; poly(d1-dC) from Pharmacia LKB Biotechnology Inc.; DNase I from Worthington; 14C-labeled chloramphenicol and 32P-labeled dNTPs from Amersham Corp.; and acetyl coenzyme A from Sigma.
The first nucleotide after the EcoRI site in the polylinker of (-3600, +39) Tf-CAT was defined as position 1 so that the HindIII site at -3300 bp was position 291. Oligonucleotide I1 (positions 58-86) was filled in and multimerized with T4 DNA ligase. The oligomers containing one, two, and four copies were inserted in the SmaI site of pUC19-CAT2 and in the NdeI blunt ended site of pTf-CAT to obtain, respectively, IlTf"CAT2, 4(Il)Tf-CAT2, IlpTf-CAT, 2(Il)pTf-CAT, and 4(Il)pTf-CAT. The orientation of the inserts was determined by restriction enzyme mapping or DNA sequencing. All the inserts were in sense orientation.
DNase I Footprinting-DNase I footprints were performed with nuclear extracts of rat liver and HeLa cells, as described by Galas and Schmitz (1981) with some modifications (Brunel et al., 1988). In each sample, 70 pg of crude nuclear extracts were added.
DNA-Protein Mobility Shift Assay-The standard assay was performed in a final volume of 20 pl containing 1 ng of 5' 32P-labeled double strand oligonucleotide (about 40,000 cpm), 1 pg of poly(d1-dC), 50 ng of sonicated salmon sperm DNA, 30 ng of sonicated Escherichia coli DNA, 100 mM KCl, 0.2 mM EDTA, 0.5 mM dithiothreitol, 20 mM HEPES, pH 7.9, 20% glycerol, 0.1% Nonidet P-40, and 6 pg of rat liver nuclear proteins. After 15 min at 4 "C, the mixture was loaded onto an 8% polyacrylamide gel and electrophoresed as indicated in Ochoa et al. (1989) for 120 min at 15 V/cm. The gel was dried and autoradiographed. For the competition experiments, various amounts of unlabeled competitor oligonucleotides were added in the binding reaction, as indicated in each assay.
Cell Culture, Transfections, and CAT Assays-Human hepatoma Hep3B cells and HeLa cells were cultured and transfected and the cell extracts were used for CAT assays as described by Schaeffer et al. (1989).

Domains, A and B
We established previously that the fragment located between -4.0 and -3.3 kb relative to the cap site exhibits typical characteristics of an enhancer (Schaeffer et al., 1989). To localize more precisely the limits of the enhancer, we measured the transient expression of 5'-deleted Tf-CAT constructs, transfected in the hepatoma Hep3B cells. As shown in Fig. 1, vectors containing inserts ending at -4000 and -3600 bp gave the highest CAT level, chosen as 100%. When the 5' deletion reached position -3500, the CAT level dropped to about one-third, identical to the level reached by the -620 to +39 promoter region. The 5' deletion ending at position -3200 bp resulted in a dramatic decrease in CAT expression. This minimal expression resulted both from the absence of the enhancer and from the presence of the -1000 to -620 negative-acting region, which is active only in absence of the enhancer, as shown previously (Schaeffer et al., 1989).
The enhancer was numbered position 1 after the EcoRI site in the polylinker of (-3600, +39) Tf-CAT so that the HindIII site at -3300 bp was position 291. It appears that the fragment located between -3600 and -3500 bp relative to the cap site, which corresponds, respectively, to positions 1 and 86 of the enhancer, is crucial for enhancer activity. The truncated enhancer fragment, located between -3500 and -3300 relative to the cap site and corresponding to positions 87 and 291, was unable to stimulate transcription; however, the presence of this fragment was sufficient by itself to block the negative action of the -1000 to -620 region. It is therefore tempting to suggest the presence of two functional domains, A and B, respectively located upstream and downstream of position 86.
Footprint Analysis of the Whole and Truncated Tf Enhancer-To identify the nuclear factors interacting with the whole (1-291) and truncated (87-291) enhancer regions, we performed a DNase I footprinting analysis, using rat liver nuclear extracts. Fig. 2, A and B, shows the protection pattern of each strand of the whole enhancer. The three protected regions numbered I, 111, and IV have been described previously (Schaeffer et al., 1989). A careful analysis of the footprint pattern using several nuclear extract preparations indicated the presence of the short slightly protected region numbered 11. As expected, regions 11, 111, and IV were protected in the truncated enhancer. Interestingly, motif I was still protected in the truncated enhancer between positions 87 and 121 ( Fig.  2C), compared with positions 58 and 121 in the whole enhancer; moreover, the DNase I hypersensitive site located at position 99 in the whole enhancer disappeared. These results suggest that at least two proteins are able to interact with region I of the whole enhancer, one binding to region Ia (58-86) and the other binding to region Ib (87-121). Fig. 3 presents the DNA sequence of the enhancer, as well as the localization of the four protected regions as determined in Fig. 1. Nuclear proteins present in HeLa cell extracts generated an identical footprinting pattern.
Competition Footprint Analysis of the Tf Enhancer-To further explore the interactions of nuclear proteins with the putative domains A (1-86) and B (87-291), footprinting experiments were performed in the presence of different competitor oligonucleotides. The synthetic oligonucleotides I1 (58-86), I2 (81-107), and I3 (100-128) correspond to overlapping sequences of region I (Fig. 4). As shown in Fig. 5a, when unlabeled oligonucleotide I1 was used as a competitor in 50and 100-fold excess, the protection of motif Ia disappeared as expected, as well as the hypersensitive site located at position 64; the footprint patterns of regions Ib, 11, 111, and IV were unchanged. The exact delimitation of the 3' border of the protection of motif Ia is difficult to assess by the competition footprinting experiments; however, the results shown in Fig.  5a indicated that the presence of the hypersensitive site at position 64 on the lower strand correlates with the binding of a nuclear protein to this motif. Moreover, methylation interference data performed with oligonucleotide I1 showed that the purines interfering with the interaction of the protein are located at position 64 on the upper strand and at positions 67 and 69-72 on the lower strand ( Fig. 3 and results not shown). This confirms the importance of the hypersensitive site at position 64 and the correlation with the binding of a nuclear factor to the region around nucleotides 64-72. As shown in Fig. 5b, addition of increasing amounts of unlabeled oligonucleotide I2 (nucleotides 81-107) inhibited not only the protection of motif Ib, but also the formation of footprints I11 and IV, whereas motif Ia was not affected the hypersensitive site located at position 99 disappeared, whereas the site located in motif Ia at position 64 was still present. The protection of motif I1 was in general only slightly visible; therefore, it is difficult to detect a clear modification of its pattern. Finally, oligonucleotide I3 (nucleotides 100-128) was unable to compete for the binding of any nuclear factor; the footprinting pattern obtained with a 200 molar excess of unlabeled oligonucleotide I3 was identical with the pattern observed without any competitor (Fig. 5, lunes 0).
In addition, competition experiments were performed with synthetic oligonucleotides corresponding to motif I1 (oligo- EcoRI-Hind111 290-bp fragment from -3.6 to -3.3 kb was 3' end-labeled a t Hind111 (upper strand) ( A ) and at EcoRI (lower strand) ( B ) . C, the EcoRI-HindIII205-bp fragment from -3.5 to -3.3 kb was 3' end-labeled a t EcoRI (lower strand). S, the G+A sequence ladder; F, free DNA without nuclear extracts; L, experiments with rat liver nuclear extracts. Brackets delineate sequences I, 11,111, and IV protected against DNase I digestion. Numbers along the autoradiographs indicate the position relative to position 1 defined as in the legend to Fig. 3. Horizontal arrows indicate the sites of enhanced DNase I cleavage.
nucleotide 111, nucleotides 81-107) and to overlapping sequences of motifs I11 and IV (Fig. 4, upper panel). Oligonucleotides 111, 1112, and IV2 at a 100-fold excess were able not only to displace the protection of their homologous sequence, but also to inhibit the protection of motifs Ib, 11, 111, and IV without affecting the protection of motif Ia. In each case, the resulting pattern was identical with the one observed in the competition assay performed with an excess of oligonucleotide I2 as shown in Fig. 5b. Oligonucleotides 1111 (173-203) and IV1 (229-260), like oligonucleotide 13, were also unable to affect the footprinting pattern. These results help to establish an interaction profile of the trans-acting factors with their preferential binding sites on each enhancer region (Fig. 4, lower punel). Analysis of the trans-Acting Factors by Gel Retardation Assays-To determine the composition of the factor(s) bound to each domain, we performed gel retardation experiments. Each oligonucleotide 11, 12, 111, 112, and IV2 that was previously shown in the competition footprint assay to interact with a nuclear protein was 5' end-labeled and used in gel mobility experiments in the presence of liver nuclear extracts (Fig. 6A, rightpanel). The electrophoretic profile of the shifted DNA-protein bands was clearly different with oligonucleotide I1 compared with the others. Moreover, when oligonucleotide I1 was used as a probe, cross-competition experiments performed with 10, 50, and 100 molar excess of unlabeled oligonucleotides 11, 12, 111, 1112, and IV2 showed that only the homologous oligonucleotide was able to compete for the protein binding (Fig. 6B).
Oligonucleotides I11 and I112 appear to interact with different liver protein(s) according to the different retarded bands (Fig. 6A, rightpanel), and the absence of cross-reactivity with all heterologous nucleotides (Fig. 6C). Interestingly, when oligonucleotide I112 was used as an unlabeled competitor, it was able to partially displace the complex formed by liver protein(s) with oligonucleotides I2 and IV2 (Fig. 6C). This suggests that oligonucleotide I112 contains sequences that may be binding sites for the protein(s) interacting with oligonucleotides I2 and IV2.
Precisely, oligonucleotides I2 and IV2 gave rise to an almost similar retarded complex, different from all the others (Fig.  6A, right panel). Moreover, in cross-competition experiments, the complex formed with oligonucleotide I2 could be partially competed for by an excess of unlabeled oligonucleotide IV2 and vice versa (Fig. 6C). The same is true when oligonucleotide I112 was used as a competitor as indicated above, whereas no competition could be observed with oligonucleotides I1 and I11 (Fig. 6C).

Analysis of the Enhancer and Cell Type-specific Activity of Domains A and B Transient Expression Analysis in Hep3B and HeLa Cells-
As presented previously, the Tf enhancer is not liver-specific since it activates the transcription of the SV40 early promoter not only in Hep3B cells but also to a lower extent in the epithelial carcinoma HeLa cells. When the enhancer is directly linked to the homologous promoter, the expression is restricted to the Hep3B cells since the Tf promoter is not functional in HeLa cells (Schaeffer et al., 1989).
The cell type-specific activity of each enhancer domain A and B was tested by transient expression experiments in Hep3B and HeLa cells. These cells were transfected with a series of Tf-CAT constructs, containing the domain A (oli- gonucleotide 11) or the 0.2-kb long domain B in single and multimerized copies linked either to the homologous Tf promoter (Fig. 7 ) or to the SV40 early promoter (TATA box and 21-bp repeats) (Fig. 8). The levels of CAT expression were expressed relative to the level of pSV2-CAT containing the SV40 promoter and enhancer. As shown in Figs. 7 and 8, domain B, either in single or multiple copies, was unable to significantly activate either the Tf or the SV40 promoters.
On the contrary, four copies of oligonucleotide 11, corresponding to domain A, were able to strongly activate the transcription of the SV40 promoter 6-fold in Hep3B cells and &fold in HeLa cells (Fig. 8). This activation can be compared to the 3-and 5-fold stimulation of the whole 0.3-kb enhancer, respectively, in HeLa and in Hep3B cells. In contrast, multimers of the oligonucleotide I1 were hardly able to stimulate transcription from the Tf promoter. Their enhancing ability was far below that of the whole 0.7-or 0.3-kb long enhancer region, which activated transcription of the Tf promoter about &fold (Fig. 7).

Mobility Shift Experiments with Oligonucleotide I1 in the Presence of Rat Liver and HeLa
Cell Extracts-Since the multimerized oligonucleotide I1 was able to function as an enhancer both in Hep3B and HeLa cells in the presence of the SV40 promoter, it was interesting to determine whether or not the same factor binding to this site was present in the two cell types. We already determined that the footprint protection of domain A was identical with rat liver and HeLa extracts. As shown in Fig. 6A (left panel), we performed gel mobility shift experiments, using oligonucleotide I1 in the presence of, respectively, four and three different preparations of liver and HeLa nuclear extracts. The liver proteins gave rise in all experiments to two retarded DNA-protein complexes called C1 and C2, whereas the HeLa extracts always gave rise to a unique retarded complex of a different electrophoretic mobility. With liver extracts, these retarded bands could be competitively inhibited by unlabeled oligonucleotide 11, as shown in Fig. 6B. A similar result b a s obtained using the HeLa extracts. These data suggest that the nuclear factors interacting with motif Ia are different in liver and in HeLa cells.

DISCUSSION
This report presents the structural and functional characterization of the enhancer of the human transferrin gene. The region situated between -3.6 and -3.3 kb upstream of the transcriptional start site of the Tf gene was shown in our previous paper to present the characteristics of a typical enhancer (Schaeffer et al., 1989). As shown before, this enhancer is not liver-specific since it was able to stimulate the heterologous SV40 promoter in hepatoma and, to a lower extent, in HeLa cells. When the Tf enhancer or the 72-bp sequence of the SV40 enhancer was linked to the Tf promoter, the expression was strictly hepatoma-specific. In transient expression studies, the liver-specific expression of the Tf gene is solely conferred by the combined action of the Tf-LF1 nuclear protein and the C/EBP-related factor binding to adjacent proximal promoter sites. In contrast to most of the cellular enhancers, the Tf enhancer is not responsible for the cell type-specific expression of the human Tf gene.
Existence of Two Structural Domains-The interactions of trans-acting factors with the Tf enhancer were first examined by in uitro techniques. DNase I footprinting assays allowed the detection of four protected regions (I (58-125), I1 (145-161), I11 (189-222), and IV (232-281)), with nuclear extracts of rat liver (Figs. 2, A and B, and 3) and of HeLa cells.
The first evidence of the existence of two distinct structural domains was given by the DNase I footprint analysis of the (87-291) truncated enhancer region. As shown in Fig. 2C, even in the absence of the 58-86 sequence, the remaining region I was still protected by liver nuclear extracts. This suggests that region I can be divided at position 86 into two structural motifs, Ia and Ib, and that each motif corresponds to the binding site of different proteins.
Competition footprint experiments provided further evidence about the two structural domains. These assays were performed with unlabeled synthetic oligonucleotides corresponding to overlapping sequences of the enhancer (Fig. 4).
The results provided a precise mapping of the preferential sites of DNA-protein interactions (Fig. 4, lower panel). They indicated that only sequences of oligonucleotide I1 and part of oligonucleotides 12, 111, 1112, and IV2 correspond to the binding sites of nuclear factors. Second, they clearly showed the absence of cross-competition between oligonucleotide 11, corresponding to region Ia, and all other oligonucleotides, corresponding to regions Ib, 11, 111, and IV. When the competition assay was performed with oligonucleotide I1 (58-86), only the corresponding motif Ia was modified (Fig. 5a). To the contrary, competition with oligonucleotide I2 (81-107), corresponding to motif Ib (Fig. 56) or with oligonucleotides 111, 1112, and IV2, modified the footprinting pattern and DNase 1-hypersensitive sites of regions Ib, 11, 111, and IV, without affecting the protection of motif Ia. Thus, it appears that the four protected regions can be divided into domain A, containing region Ia, and domain B, including regions Ib, 11, 111, and IV (Fig. 4, lower panel). Moreover, the fact that the elimination of any one of the proteins interacting with any one of regions Ib, 11,111, or IV strongly modified the protection pattern of the three other regions suggests a cooperation for binding between the trans-acting factors of the B domain. It is likely that protein-protein interactions play an essential role in the stability of the final B domain DNA-nuclear protein complex. Gel retardation studies performed with liver nuclear extracts (Fig. 6A, rightpanel) and cross-competition assays (Fig.  6, B and C) indicated that the proteins binding to oligonucleotide I1 are different from the proteins interacting with the oligonucleotides corresponding to domain B. Oligonucleotides I11 and 1112 seem to bind distinct proteins, different from the others; oligonucleotides I2 and IV2 may bind similar proteins, but different from the others. These data suggest that the four protected motifs of domain B are the binding sites of at least three different liver nuclear factors.
Liver nuclear extracts gave rise to two specific retarded DNA-protein complexes (C1 and C2) with oligonucleotide 11, whereas HeLa nuclear extracts gave a unique band of a higher electrophoretic mobility (Fig. 6A, left panel). This result strongly suggests the existence in liver and HeLa cells of different proteins able to interact with domain A.
Existence of Two Functional Domains-Transient expression studies demonstrated that the two structural domains A (1-86) and B (87-291) correspond to two functional active domains. Compared to the fully active (-3600, +39) Tf-CAT vector, the 5"deleted (-3500, +39) Tf-CAT construct had lost all enhancer activity; in the hepatoma Hep3B cells, its level of CAT expression was identical with that reached by the (-620, +39) Tf-CAT vector containing only the Tf promoter (Fig. 1). This showed that the deleted sequence, which corresponds to position 1-86 of the enhancer, is crucial for the enhancer activity. It also indicated that the 0.2-kb long domain B had no enhancer activity by itself. However, the presence of this domain B was essential to maintain the activity of the promoter and to block the action of the (-1000, -620) negative-acting element described in our previous paper (Schaeffer et al., 1989).
The enhancer activity of each isolated domain was further tested in transient expression experiments by inserting each domain either in single copy or in multimers in front of different promoters. The resulting data clearly indicate that domain B is unable to enhance the activity of either the Tf promoter (Fig. 7) or the heterologous SV40 early promoter, either in Hep3B or in HeLa cells (Fig. 8). In contrast, multi-mers of the oligonucleotide 11, corresponding to domain A, were able to strongly stimulate the expression of the SV40 promoter 6-fold in Hep3B cells and &fold in HeLa cells (Fig.  8). Thus the active liver and HeLa nuclear factors binding to domain A (Fig. 6, left panel) are able to form functional interactions with the factor Spl (Dynan and Tjian, 1983) binding to the SV40 promoter.
This contrasts with the results obtained with the homologous Tf promoter. Two or four copies of the oligonucleotide I1 were unable to significantly enhance the activity of the Tf promoter in Hep3B cells (Fig. 7). It appears that the interactions of the cis-and trans-acting elements of the Tf enhancer and promoter are quite stringent. Only the combination of the two domains A and B results in full enhancer activity in the presence of the Tf promoter. This emphasizes the importance of protein-protein interactions in bringing together the two enhancer domains and the Tf promoter, either directly or via some component(s) of the transcriptional machinery, as proposed by Ptashne's group Carey et al., 1990).

Each Domain A and B Is a Proto-enhancer of a Different
Type-A fine dissection of the SV40 enhancer revealed that it is composed of multiple motifs, binding nuclear proteins which work synergistically to generate enhancer function. Distinct types of motifs have been identified and three levels of enhancer organization have been defined (Herr and Clarke, 1986;Fromental et al., 1988;Ondek et al., 1988;Dynan, 1989).
The first level is the enhanson, which corresponds to a protein binding site; the second level corresponds to combinations of identical or different enhansons to form a proto-enhancer, the enhancer minimal element. Where a proto-enhancer is not active by itself, a third level of organization exists, corresponding to the combination of at least two proto-enhancers (Fromental et al., 1988).
It is tempting to define the Tf enhancer organization following the SV40 enhancer structure. Consequently, the Tf enhancer is composed of five enhansons: Ia, Ib, 11, 111, and IV. Enhanson Ia corresponds to the structural and functional domain A. According to the definition proposed by Fromental et al. (1988), it is an enhanson of class C, since it exhibits enhancer activity when a single copy of the motif is oligomerized. Thus, it is also a proto-enhancer of type 111. Enhansons Ib, 11,111, and IV of domain B correspond to a proto-enhancer that has no activity by itself; only the association with the enhanson Ia generates enhancer activity. In conclusion, the Tf enhancer is made up of two proto-enhancers of different types.
The cooperation between the proteins interacting with B domain and the synergistic activation observed when the two proto-enhancers A and B are present may have important functional consequences. A variation in the amount of a single factor should result in a modification of the Tf enhancer activity. This emphasizes the importance of each single factor in the overall modulation of the Tf enhancer function. The Proto-enhancer A Interacts with Cell Type-specific Proteins-The 30-bp long proto-enhancer A, corresponding to the oligonucleotide 11, contains what appears more and more like a liver-specific sequence TGTTTGCT, since the identical motif has been described in control regions of other liverspecific genes. The same octanucleotide is also found in the promoter of the a-fetoprotein and a,-antitrypsin genes (Scott and Tilghman, 1983;Ciliberto et al., 1985), in the human a-Transferrin Gene Enhancer fetoprotein enhancer (Watanabe et al., 1987), and in the E site of the hepatitis B virus enhancer (Shaul and Ben Levy, 1987). The heptanucleotide sequence TGTTTGC is present in the mouse a-fetoprotein enhancer (Godbout et al., 1988) and in the mouse albumin enhancer (Herbst et al., 1989;Zaret et al., 1990).
The absence of liver specificity of the Tf enhancer previously observed (Schaeffer et al., 1989) could have resulted from a different cell type specificity of each proto-enhancer. However, our mobility shift data (Fig. 6A, left panel) strongly suggest that the proto-enhancer A interacts in liver and in HeLa cells with different nuclear factors. Like the octamer sequence of immunoglobulin genes (Schaffner, 1989), the sequence of the Tf proto-enhancer A appears to stimulate cell-specific transcription by binding distinct factors in different cell types. In this case, the Tf enhancer behaves in hepatoma and in HeLa cells as a tissue-specific enhancer, each proto-enhancer interacting in a particular cell type with its cognate cell-specific proteins. Purification and characterization of the different nuclear proteins are obviously necessary to confirm this attractive model.