Localization of DNA Protein-binding Sites in the Proximal and Distal Promoter Regions of the Mouse cu-Fetoprotein Gene*

DNase I footprinting assays were performed to iden- tify the binding sites for putative trans-acting factors involved in the control of a-fetoprotein (AFP) gene expression using mouse AFP promoter fragments (-839 to +56) and nuclear protein extracts from fetal, newborn, and adult livers and from brain and kidney. Our studies have shown that with nuclear protein from adult mouse liver, there are 14 protected regions in the AFP promoter up to -839 base pairs (bp). Region I (-82 to -43) was protected by at least three different factors, one of which is CCAAT-bindinglenhancer-binding protein. This region is highly conserved in the mouse, rat, and human AFP genes and has been shown previously to be essential for the regulation of tissue-specific expression in mouse. Differences in DNase I protection with fetal, newborn, and adult nuclear pro- teins have been observed in the proximal promoter region (up to -202 bp) and in regions further upstream (up to -839 bp). Significant differences among liver, kidney, and brain nuclear protein-binding sites have also been observed. In these studies, we have mapped the fetal and adult nuclear protein-binding

DNase I footprinting assays were performed to identify the binding sites for putative trans-acting factors involved in the control of a-fetoprotein (AFP) gene expression using mouse AFP promoter fragments (-839 to +56) and nuclear protein extracts from fetal, newborn, and adult livers and from brain and kidney. Our studies have shown that with nuclear protein from adult mouse liver, there are 14 protected regions in the AFP promoter up to -839 base pairs (bp). Region I (-82 to -43) was protected by at least three different factors, one of which is CCAAT-bindinglenhancerbinding protein.
This region is highly conserved in the mouse, rat, and human AFP genes and has been shown previously to be essential for the regulation of tissuespecific expression in mouse. Differences in DNase I protection with fetal, newborn, and adult nuclear proteins have been observed in the proximal promoter region (up to -202 bp) and in regions further upstream (up to -839 bp). Significant differences among liver, kidney, and brain nuclear protein-binding sites have also been observed.
In these studies, we have mapped the fetal and adult nuclear protein-binding sites of the cis-acting DNA sequences of the mouse AFP proximal promoter (up to -200) and have identified specific protein-binding sites in the distal promoter (-200 to -839).
We have also identified the sites of the AFP promoter which bind nuclear proteins from highly differentiated tissues in which AFP is not expressed.
a-Fetoprotein is an oncofetal protein whose regulatory characteristics encompass activation in the early stages of fetal liver development (l), repression in the newborn (2-5), and reactivation in the adult during liver regeneration (2-4) and chemical hepatocarcinogenesis (6)(7)(8). The gene is also expressed in yolk sac, fetal gut, and fetal kidney cells (9-16) during specific stages of development, but the major site of synthesis is the fetal liver, in which the gene is transcribed throughout the period of fetal liver development. These regulations occur primarily at the level of transcription (5) and provide an excellent model for studying the molecular mechanisms of both tissue-specific and developmental regulation of eukaryotic gene expression.
In recent studies, the cis-acting elements of the rodent and human AFP' genes have been identified by transient expression assay (17)(18)(19)(20)(21)(22) and by insertion of various expression vectors into mice as stable transgenes (23,24). These studies have demonstrated that liver-specific expression as well as repression can be conferred by cis-acting DNA sequences localized up to 1 kilobase upstream of the transcription initiation site and/or within the coding region of the gene. Similar studies with other liver-specific serum protein genes such as albumin (25)(26)(27)(28)(29)(30), cut-antitrypsin (31-34), transthyretin (35), and fibrinogen (36) genes have demonstrated that sequences in their proximal promoter regions are highly conserved and that these sequences are the potential binding sites for transacting factors that are both liver specific, such as HNF-1 and C/EBP (30,34,(36)(37)(38)(39)(40)(41), and ubiquitous, such as NF-1 (42-45) and 47). For example, purified HNF-1 has been shown to interact with promoter regions from albumin, AFP, (Y-and P-fibrinogen, cY1-antitrypsin, and transthyretin (36,37, 41, 48). The promoter of each of these genes contains a common DNA motif that is recognized by HNF-1. Another liver-specific nuclear protein, C/EBP, which was initially identified as a protein that binds to the viral enhancer core element (49,50) and to certain CCAAT-containing sequences (51), recognizes several different DNA sequence motifs (48). This protein also binds to several sites in the albumin promoter (30,38,39) and to the promoter of other liver-specific genes such as transthyretin (42) and cY1-antitrypsin (35,36,41). These nuclear proteins have been shown to function as positive transcription factors with the albumin and al-antitrypsin promoters in in vitro transcription assays. These studies suggest that certain trans-acting factors should bind to conserved sequences in the AFP promoter and function in its liver-specific regulation. The AFP gene is unique, however, because its regulation requires positive trans-acting factors in the fetal liver and negative trans-acting factors in the adult liver. We have conducted experiments designed to identify potential trans-acting protein factors of the mouse AFP pro-

RESULTS
Localization of Sites in the Mouse AFP Proximal Promoter Which Bind Adult Liver Nuclear Proteins (the Repressed AFP Gene)-cis-Acting DNA sequences of the mouse AFP promoter which are essential for tissue-specific and developmental regulation have been mapped to within 1 kilobase of the AFP promoter region (17)(18)(19)(20)23,24). Godbout et al. (18) have shown by deletion analysis that the region between -85 and -52 is essential for tissue-specific AFP expression. Deletion of this essential region results in virtually complete extinction of the transcriptional activity in human hepatoma (Hep G2) cells (18). Studies with transgenic animals have shown that the signals that direct postnatal repression of the AFP gene are included in 1 kilobase of 5'-flanking DNA (24) and in a portion of the structural gene (61). Since the interaction of cis-acting sequences with transacting factors is the proposed mechanism for these regulatory processes, we have conducted DNase I protection assays to map the regions of the mouse AFP promoter which bind nuclear proteins from fetal, newborn, and adult mouse liver and to determine whether the protein-binding sites correlate with the regions identified by deletion analysis to be of functional importance.
A restriction map of the mouse AFP promoter, up to -839, the TATA box, and transcription initiation site are shown in DNase I protection assays is also shown. To identify nuclear protein-binding sites of the AFP proximal and distal promoter regions and to establish a standard map as a basis for comparison between fetal liver and nonliver cell proteins, we performed DNase I protection analyses using adult liver nuclear proteins (Fig. 2) quence is shown in Fig. 3, and the protected regions are protein concentrations, nucleotides -63 to -82 within region indicated by solid lines either above the sequence (coding Ib were protected. The sequence of this region, TATstrand data) or under the sequence (noncoding strand data).
GTTTGCTCA, is similar to sites D (TATGATT) and F DNase I protection analysis indicates that region I (-82 to (TATGTTT) in the mouse albumin promoter (30, 38). These -43) binds more than one protein (Fig. 2, A  to CCAAT-containing sequences (51) and the core enhancer sequence (49). To verify whether region Ib is protected by C/ EBP, an oligonucleotide containing the consensus C/EBPbinding site, and purified C/EBP (both gifts of Dr. Steven L. McKnight, Dept. of Embryology, Carnegie Institute of Washington, Baltimore, MD) were used in the protection assay.
The results are shown in Fig. 4. In panel A, when increasing amounts of C/EBP oligonucleotide competitor are used (lanes l-4), protection of region Ib is significantly but not completely abolished. In panel B, it is shown clearly that region Ib is fully protected by purified C/EBP (lanes 1-3). We interpret these experiments to indicate that although region Ib is a binding site for C/EBP, there are also other factor(s) present in adult mouse liver nuclei which bind to this region. When the protein concentration is increased to 60 pg, region -= -- Lane C is the control of the DNase I protection with the 50 pg of adult mouse liver nuclear extract.
All the other labels are the same as in Fig. 2. Ia, from -43 to -62, is protected (Fig. 2, A and B). This regioncontainsthesequence(-62)GTTACTAGTTAAC(-50), which is homologous to the hepatocyte nuclear factor 1 (HNF-1)-binding consensus sequence GTTAATNATTAAC (36). Region II (-126 to -98) contains a sequence GTTAAT-TATTGGC(-128 to -116), which is homologous to the HNF-1 consensus site and is linked at its 3' end with a sequence (-llS)TGGCAAATTGCCT(-107), which is homologous to the NF-1 consensus site (TGGA/,N5 GCCA). These sequence data suggest that region II could bind either HNF-1 or NF-1. Our experiments also show that C/EBP is another factor that binds to region II (Fig. 4). The protection of the 3' end of region II with adult liver nuclear extract was partially competed out by the C/EBP consensus sequence (Fig. 4A, lanes  2-4). The fact that purified C/EBP also exhibits protection of region II (Fig. 4B) is very strong evidence that C/EBP is one of several factors present in adult liver nuclei which bind to this region.
Sequence analysis of region III (-146 to -131) and region IV (-169 to -152) indicates that they contain sequences (C)CTG(C)TCT (-136 to -129)   DNA probes used in panels A and C are coding strand-labeled AFP-1 and AFP-3, respectively. The DNA probe used in panel B is the anticoding strand-labeled AFP-2. DNA fragments were incubated with increasing amounts of nuclear proteins from fetal, newborn, and adult mouse liver, as indicated at the top of each lane, in the presence of 4000 ng of ds-poly(d1. dC). All the other labels are the same as in Fig. 2. reported by Guertin et al. (22) as a binding site for the glucocorticoid receptor (GR) protein in the rat AFP promoter. To verify the identity of the proteins that bind to regions III and IV, an oligonucleotide containing the GRE consensus sequence, and purified glucocorticoid receptor protein (gifts from Dr. Keith Yamomoto) were used. The results showed that the purified GR exhibits weak binding to region IV and does not bind at all to region III (data not shown). Furthermore, the GRE oligonucleotide consensus sequence does not compete for either of these sites. We conclude that the proteins binding to regions III and IV, which we call nuclear protein III (NP-III) and nuclear protein IV (NP-IV), respectively, are not GR.
Region V (-177 to -202) consists of a sequence GTGGAAA, -179 to -185, which is homologous to the virus enhancer core sequence GTGGT/AT/~T/,+ A similar sequence in SV40, Moloney sarcoma virus, and polyoma virus enhancers binds C/EBP (49,59,60). To test whether protection of region V is due to the binding of C/EBP, the oligonucleotide containing the consensus C/EBP-binding site and purified C/ EBP protein were used to perform DNase I footprinting experiments. As shown in Fig. 4A, the C/EBP-binding oligo-mer prevents binding of the factor from adult mouse liver nuclei to the upstream part of region V. In addition, purified C/EBP also shows protection of the same upstream sequence of region V (Fig. 4B). These results suggested that C/EBP is involved in the protection of region V, but only from -183 to -202 (Fig. 3). This observation was unexpected because the sequence homologous to the virus enhancer core sequence was not fully protected by C/EBP. These data also indicate that the sequence adjacent to the enhancer core sequence is the site for C/EBP binding in region V. Since the region protected by nuclear extract is significantly larger, our results indicate that the downstream part of region V may be protected by another as yet unidentified factor.
Localization of Sites in the Mouse a-Fetoprotein Distal Promoter Which Bind Adult Liver Nuclear Proteins-We classify the sequences upstream of region V (from -200 to -839) as the distal promoter region. The protected sequences in this domain, with the possible exception of those in region XI, have not been identified as protein-binding sites in any liverspecific genes prior to this report. Camper and Tilghman (24) DNA probes used in pat& A, C, and D are anticoding strand-labeled AFP-1, AFP-3, and AFP-4, respectively.
The DNA probe used in panel B is the coding strand-labeled AFP-2.
DNA fragments were incubated with increasing amounts of nuclear proteins from fetal and newborn mouse liver as indicated at the top of each lane, in the presence of 4000 ng of ds-poly(dI.dC).
All the other labels are the same as in Fig. 2. expression. There is also evidence that this region exhibits enhancer activity in hepatoma (61) and fetal kidney cells.3 Regions VI (-230 to -247), VII (-260 to -291), VIII (-330 to -310), IX (-438 to -402), X (-543 to -514), XII (-742 to -730), XIII (-777 to -760), and XIV (-796 to -789) are all binding sites of the distal AFP promoter which exhibit interesting potential structural characteristics. For example, regions VI and VII are separated by an inverted repeated sequence (-266 to -249) TGAA(T)GAAT(T)ATTCTTCA, which forms a hypersensitive site when bound by proteins of the adult nuclear extract. Region VIII, whose sequence is GA(G)TTACATAGTAA(G)TC (-326 to -310), also exhibits an inverted repeat. Region IX contains at its 3' end a direct repeat motif from -412 to -398 (TGAATAGCCTGAACT), which is found as an inverted repeat between regions VI and VII. Regions XIII and XIV exhibit direct repeats with the sequence motif AGTTC/T. One of these is found in region XIV, and two are found in region XIII. 'J. P. Rabek, D. Hsie, and J. Papaconstantinou, manuscript in preparation.
In Region XI (-579 to -550), the protected sequence of the coding strand consists of a partial HNF-1 sequence linked to a half-NF-1 site. The ability of these proteins to bind to this site is being tested. Another interesting characteristic of region XI is that the hypersensitive site formed between -562 and -556 with protein(s) of the adult nuclear extract does not form with proteins of fetal nuclear extract (Fig. 5). This will be discussed in detail below.

Localization of Sites in the Mouse a-Fetoprotein Promoter
Which Bind Fetal and Newborn Liver Nuclear Proteins (the Transcriptionally Active AFP Gene)-Footprint analyses were done to identify differences in binding patterns between the transcriptionally active and repressed AFP genes (Figs. 5 and 6). One of the most significant differences occurs between the protein(s) binding in regions Ia and Ib (Figs. 5 and 6). The data indicate that there is very weak or no protection of region Ia by fetal and newborn nuclear proteins, whereas the region Ib binding site shows strong protection. In the adult, however, both regions Ia and Ib are protected. The protection of region Ib with fetal liver nuclear protein can be fully competed out by the oligonucleotide that contains the C/EBP consensus sequence (data not shown). This indicates that C/EBP is the only factor in fetal nuclear protein extract which binds to region Ib; whereas in the adult, the nuclear extracts appear to contain more than one factor that binds to this region (Fig.  4). Regions III, IV, and V also exhibited developmental differences in binding patterns. The sequences of regions III and IV are not protected strongly by fetal or newborn nuclear proteins (Figs. 5 and 6), but are clearly protected by proteins in the adult nuclear extract (Figs. 2 and 5). Thus, the absence of NP-III and NP-IV binding in fetal and newborn liver is associated with the transcribing AFP gene, and their binding in the adult is associated with the repressed gene. Region V, on the other hand, exhibits strong binding by the fetal, newborn, and adult nuclear extracts.
The DNase I protection patterns for region IX show a significant developmental difference among fetal, newborn, and adult nuclear proteins. The protection pattern indicates that all three nuclear extracts protect the middle sequence of this region (-439 to -416) and that the 5' end of the region is protected with fetal and newborn nuclear proteins (up to -464), whereas the 3' end is strongly protected by adult nuclear proteins (down to -402). The staggered nature of this protection suggests that these interactions may involve several proteins or a single protein whose binding properties may change through modification. Another significant developmental difference is seen in region XI where it appears that there may be fetal-and adultspecific proteins that bind to this region. This is based on the fact that binding by adult nuclear protein to this site generates a hypersensitive site between -562 and -556 (Figs. 2 and 5), whereas this does not occur with the proteins from fetal and newborn liver nuclear extract (Figs. 5 and 6).
Regions VI, VII, VIII, and XII show similar binding characteristics with fetal, newborn, and adult nuclear proteins. Therefore, these regions may not be involved in the developmental regulation of the AFP gene but may be involved in its tissue-specific functions as indicated by data with proteins from kidney and brain. This is discussed in detail below.
Regions XIII and XIV are detected as discrete binding sites with the nuclear proteins from adult mouse liver (Fig. 2, panels G and H). There are 9 base pairs that separate regions XIII and XIV in the presence of adult protein (Fig. 3). In the presence of fetal nuclear proteins, regions XIII and XIV are  (Fig. 6D) and are detected as a single protected region.
Liver Specificity of Nuclear Proteins-Nuclear extracts from liver, kidney, and brain of adult C3H/He mice were used in DNase I footprinting assays to detect tissue-specific nuclear protein binding and resulting protection patterns (Fig. 7). Because of difficulties in preparing large quantities of nuclear proteins from kidney and brain, the amounts of protein used for these experiments were only increased up to 30 pg. Liverspecific binding sites were identified as regions Ib, V, VIII, and X. As mentioned above, the liver nuclear protein that binds to regions Ib and V is C/EBP (49-51). Region VIII was also only protected by nuclear protein from the liver, indicating that this may be a unique AFP-binding site. Region X is a binding site that shows liver specificity. Interestingly, this site shows its strongest binding properties with the adult nuclear proteins, indicating that it may represent an adult liver-specific nuclear protein-binding site (Figs. 5 and 6).
Sites that bind nuclear proteins from two or all three of the tissues tested but also exhibit differences in binding properties are regions Ia, II, VI, VII, IX, and XI. Kidney nuclear protein, for example, only protected region Ia (-73 to -43). There is no protection in region I when nuclear protein from brain was used. The region II binding site is protected by kidney nuclear extract from -128 to -110 and liver nuclear protein from -126 to -98, clearly demonstrating differences between the nuclear proteins of these two differentiated cell types. Only half of the NF-1 site is protected by the kidney nuclear extract, and the protection appears to extend into region III. Protection by brain nuclear proteins, on the other hand, is similar to that of the liver nuclear proteins.
Regions VI and VII were protected by nuclear extracts from all three cell types, but there were differences in the pattern of protection. In this case, the patterns of protection for the kidney and brain proteins were similar, whereas that for the liver was unique.
In region IX, basically the same protection pattern occurs with liver and kidney nuclear protein, but the data also indicate that the kidney may have higher levels of this factor.
Although region XI was protected by nuclear proteins from all three tissues, a DNase I-hypersensitive site was formed by liver nuclear protein and to a lesser extent by kidney nuclear extract. However, no hypersensitive site could be detected in the presence of brain nuclear proteins. These data indicate that there are differences in the liver, kidney, and brain nuclear proteins that bind to the sequences in region XI. DISCUSSION We have identified the regions in the mouse AFP promoter (from -1 to -839) which bind nuclear proteins from fetal, newborn, and adult mouse (C3H/He) liver. Significant similarities and differences between the protection maps for the transcriptionally active (fetal) and repressed (adult) gene have been demonstrated. We also detected differences among liver, kidney, and brain.
Our studies indicate that the DNA sequences of regions Ia and Ib are the binding sites for two liver-specific trans-acting factors, which we believe are HNF-1 (37) and C/EBP (49-51) (Fig. 8). These proteins are liver-specific transcription factors for the mouse albumin (38, 39, 56), cul-antitrypsin (34-37), transthyretin (36, 37, 40), and fibrinogen genes (36, 37). The localization of region Ib as the binding site for C/EBP is based on competition studies (Fig. 4) with the C/EBP oligonucleotide and with pure C/EBP protein. Localization of region Ia as the binding site for HNF-1 is based on experiments by Jose-Estanyol and Danan (57), who showed that this region of this rat AFP promoter, which is homologous to the region Ia sequence of the mouse AFP promoter, is protected by protein from a rat liver nuclear extract, and this protection can be competed out by an oligonucleotide containing the HNF-1 consensus sequence (58). Furthermore, using purified rat HNF-1 for DNase I protection analysis, Courtois et al. (37) have shown binding to region Ia of the mouse AFP promoter. These studies support our conclusion that HNF-1 is the protein in adult mouse liver nuclear extract which results in the protection of region Ia (Fig. 2). Studies from our laboratory and others indicate that region II is a binding site for NF-1, HNF-1, and C/EBP. Sequence data suggest that NF-1 and HNF-1 could bind to this region. Jose-Estanyol and Danan (57) have shown that NF-1 oligonucleotide sequences can compete for a protein that binds to a region in the rat AFP promoter which is homologous to region II of the mouse. Our DNase I protection analyses with nuclear proteins fractionated by heparin-agarose chromatography (data not shown) have revealed that HNF-1 binds to this region at -133 to -111, but only in the absence of other region II binding factors. Since region II protection with nuclear extracts extends from -126 to -98 and since this region does not fully include the HNF-1 sequence, we believe that the protection exhibited by nuclear extract is due to the binding of NF-1. Our experiments also show that C/EBP is a third factor that binds to region II. This diversity of binding by C/EBP, which is a proven liver-specific transcription factor, poses interesting questions concerning its role in the regulation of AFP. One of the most important of these is whether the binding properties of C/EBP are altered in the presence of other factors. Since it has been proposed that the alignment of these proteins may be essential for gene regulation (39), any alteration in binding due to competition by other nuclear proteins may be an important factor in this regulation.
The competition studies with C/EBP oligonucleotide and pure C/EBP protein indicate that C/EBP is the only protein in fetal liver extract which binds to region Ib, whereas similar experiments with adult mouse liver nuclear protein indicate the presence of other factors besides C/EBP which bind to region Ib (data not shown). Maire et al. (39) have proposed that site D of the mouse albumin promoter binds an adultspecific protein, DBP, as well as C/EBP. Since region Ib exhibits strong homology to albumin site D and since Ib is located in the AFP promoter at relatively the same position as site D in the albumin promoter (see Fig. 8), it is possible that DBP is one of the proteins in adult nuclear extract which binds to region Ib. On the other hand, these same experiments show that C/EBP binds very strongly to region V although the protected region does not coincide with the enhancer core consensus sequence. The enhancer core consensus sequence in region V was protected by liver nuclear protein extracts. These studies also indicate that region V may bind another protein.
Interestingly, binding of HNF-1 and C/EBP shows overlapping protection of regions Ia and Ib when fractionated proteins are used in DNase I protection assays (data not shown). Our studies have shown that binding in both regions occurs with adult nuclear extract, whereas with fetal nuclear extracts we only detect strong binding at the C/EBP site. These data indicate that HNF-1 is at a low concentration or absent in the fetal liver and that it may be an adult-specific trans-acting factor. Alternatively, HNF-1 may be present in the fetal liver in a form that cannot bind to its site in the presence of C/ EBP, thus making its detection difficult by protection assay. In this mechanism, we postulate that C/EBP binding may be replaced by other region Ib binding factors such as DBP, which is also an adult-specific protein, and that this in turn facilitates the binding of HNF-1. Functional analysis by Godbout et al. (18) has shown that sequences in region I are essential for the tissue-specific expression of AFP gene. On the basis of these studies plus our footprinting experiments, we propose that the liver-specific factor C/EBP plays a major role in the transcription of the AFP gene in the fetus. This is supported by our observation that there is no protection at the HNF-1 site in the AFP promoter when fetal or newborn mouse liver nuclear extracts are used in the protection assays.
In vitro transcription analyses have demonstrated that HNF-1, C/EBP, and DBP are positive transcription factors for the mouse albumin gene (39,54) and that HNF-1 is the dominant positive transcription factor in directing liver-specific albumin transcription in the adult (39). Since these factors also bind to the promoter of the repressed AFP gene in the adult, the underlying basis for these differences may be in the presence of unique binding sites of the AFP promoter and/or the alignment of trans-acting factors relative to the transcription start site. A major difference between fetal and adult AFP promoter binding patterns occurs in regions III and IV. The proteins (NP-III and NP-IV) that bind to both of these regions are either not present in the fetal nuclear extracts or are in a form that cannot bind. Furthermore, our analyses detect a gradual increase in protection in the newborn, and the highest level is detected in the adult, suggesting that they are specific factors of the postnatal liver. Guertin et al. (22) have reported that this site in the rat AFP promoter binds the glucocorticoid receptor complex. However, our data indicate that NP-III and NP-IV are not GR and that the proteins bound to regions III and IV may be involved in the repression of the AFP gene. Although each of these regions contains consensus sequences for glucocorticoid receptor-binding sites (GRE) and although glucocorticoid enhances the repression of the AFP gene in the newborn (64-66), our experiments indicate that the glucocorticoid receptor protein complex is not the protein that binds to these sites. Furthermore, our studies indicate that NP-III is not related to CR and that NP-IV may belong to a family of GR-related proteins because purified GR exhibits weak binding to region IV.
The protection map of the promoter of the repressed AFP gene indicates that HNF-l-, C/EBP (or DBP)-, NF-l-, regions III (NP-III)-, IV (NP-IV)-, and V (C/EBP)-binding sites are occupied (Fig. 8). One major difference between the albumin (active) and AFP (repressed) proximal promoters is the absence of NP-III-and NP-IV-binding sites in the albumin promoter. This is further indication that regions III and IV may be the binding sites for proteins that function in the repression of the AFP gene in the adult. However, what the relationship of these proteins is to products of the raf locus is unknown.
Several regions in the distal promoter (from -200 to -839) exhibit significant differences in binding patterns by adult compared with fetal nuclear extracts. For example, both nuclear extracts protect the middle sequence of region IX, whereas protection is extended upstream with fetal nuclear extract and downstream with adult. Regions XIII and XIV also show a change in their protection patterns associated with development. In the adult, these regions are clearly separated by an unprotected region of approximately 9 nucleotides. In the fetus, however, the regions are not separated, and protection extends through XIII and XIV. Recently, the distal promoter region has been shown to have enhancer activity by chloramphenicol acetyltransferase assay in both hepatoma (61) and kidney cells.3 In addition, this region of the promoter has been reported by Camper and Tilghman (24) to function in the repression of the AFP gene in the adult hepatocyte. We believe that transacting factors binding to these regions may be involved in these regulatory activities.
Analysis of protection patterns of kidney and brain nuclear extracts provides information on tissue-specific nuclear factors. The C/EBP-binding sites (regions Ib and V) are not protected by kidney or brain nuclear proteins, which is further evidence that C/EBP is a liver-specific protein. On the other hand, the kidney protein does produce protection at the HNF-1 site. This kidney protein may be a variable HNF-1 such as that detected in dedifferentiated hepatoma cells and in lung cells (62). Also, protection of the NF-l-binding site varies between liver and kidney proteins, indicating that the kidney nuclear extract contains a different NF-1 member of the proposed NF-1 family (63). Both kidney and brain nuclear extracts contain proteins that bind to regions III and IV. Thus, in all three tissues (adult liver, kidney, and brain) in which the AFP gene is repressed, regions III and IV are strongly protected.
It has been proposed that in early liver development, the AFP and albumin genes are activated by the AFP enhancers located within the intergenic region (24) and that subsequent regulation occurs via interactions of regulatory proteins with the promoter sequences for each gene. Thus, the repression of the AFP gene at birth and continued expression of albumin are attributed to the regulatory sequences of each promoter. On the basis of our studies, we propose the following model for the independent regulation of the AFP gene. (a) The high level of expression of the AFP gene in the fetal liver is due to the combinatorial activity of C/EBP, NF-1, C/EBP, in the proximal promoter. (b) In the adult, the AFP gene is repressed due to the binding of NP-III and NP-IV between NF-1 and region V (C/EBP). We propose that these factors may function to prevent the formation of an active transcription complex. However, this does not exclude the possible involvement of the distal promoter transacting factors in this regulatory process. This would explain the positive transcription activity of HNF-1, C/EBP, and DBP with the albumin gene (39) and their association with repressed activity in the adult AFP gene.
Our studies demonstrate that the distal promoter region shows differences in protein-binding patterns between fetus and adult. Since there are indications that this region may also be associated with the postnatal repression of the gene (24), their role in developmental regulation and/or tissuespecific regulation must await further study.