We investigated nuclear factors that bind to δ1-crystallin enhancer core and regulate lens-specific transcription. A nuclear factor δEF1, which binds to the essential element of the δ1-crystallin enhancer core, was molecularly cloned from the chicken by a southwestern method. The protein organization of δEF1 deduced from the cDNA sequence indicated that it has heterogeneous domains for DNA-binding, two widely separated zinc fingers and a homeodomain, analogous to Drosophila ZFH-1 protein. The C-terminal zinc fingers were found to be responsible for binding to the δ1-crystallin enhancer core sequence. δEF1 had proline-rich and acidic domains common to various transcriptional activators. During embryogenesis, δEF1 expression was observed in the postgastrulation period in mesodermal tissues; initially, in the notochord, followed by somites, nephrotomes and other components. The expression level changed dynamically in a tissue, possibly reflecting the differentiation states of the constituent cells. Besides mesoderm, δEF1 was expressed in the nervous system and the lens, but other ectodermal tissues and endoderm remained very low in δEF1 expression. Cotransfection experiments indicated that this factor acts as a repressor of δ1-crystallin enhancer. Possession of heterogeneous DNA-binding domains and its dynamic change of expression in embryogenesis strongly suggest that EF1 acts in multiple ways depending on the cell type and the gene under its regulation.

Cell type-specific gene expression is mainly controlled in the process of transcriptional initiation by the differential activity of transcription factors. Cumulative evidence has suggested that cell-specific regulation is not necessarily elicited by the activity of a single kind of factor but generally by a combinatorial assortment of transcriptional regulators, each of which may have wider cell type distribution than the gene under its regulation. In addition, it has become increasingly clear that not only activators but also repressors play crucial roles in generating cell type-specific activation. Thus, to determine the mechanism of such regulation, it is essential to characterize each of the factors that participate and have a determinative role in eliciting the specificity.

We have been interested in the mechanism of lens-specific expression of the δ-crystallin gene, because this gene is the first to be turned on in lens cell differentiation during embryogenesis. In addition, since lens cells undergo terminal differentiation exceptionally early in embryogenesis through interaction of relatively simple cell lineages (Grainger, 1992), analysis of δ-crystallin gene regulation may also shed light on the mechanism of how transcription factors are regulated during embryonic development.

The lens specificity of the δ-crystallin expression is conferred by an enhancer located in the third intron of the gene (Hayashi et al., 1987; Goto et al., 1990), and a short stretch of the DNA sequence TTGCTCACCT in the enhancer core region was found to be essential for lens-specific activity of the enhancer (Funahashi et al., 1991). We have identified a nuclear factor, δEF1, which binds specifically to the DNA sequence and is present in lens cells and many other non-lens cells. To assess the significance of this factor in lens-specific regulation of the δ-crystallin gene, δEF1 cDNA was molecularly cloned and characterized.

It was found that δEF1 has interesting features as a DNA-binding protein and as an embryonic gene regulator. It has multiple DNA-binding domains, two separated zinc finger clusters and a homeodomain, and represses δ-crystallin enhancer element. It is turned on in the early process of organogenesis in mesodermal tissues, the nervous system and the lens, and its activity changes dynamically along with tissue maturation. These characteristics place δEF1 not only among the regulators of δ-crystallin enhancer but among the factors important for the regulation of postgastrulation embryogenesis.

cDNA libraries

Random-primed cDNAs of 13 day chicken embryonic lens and oligo(dT)-primed cDNAs of lens and brain of the same stage were prepared from poly(A)+ RNAs using a cDNA synthesis kit (Invitrogen). cDNAs were ligated with EcoRI-NotI linkers (Invitrogen) or EcoRI-BamHI-NotI linkers (Takara), inserted into the EcoRI sites of λgt10 or λgt11 and packaged using a packaging extract (Stratagene).

Southwestern screening (Vinson et al., 1988; Singh et al., 1988) of the cDNA libraries

The random-primed lens cDNA library made on λgt11 was screened for binding of HNW sequence (Funahashi et al., 1991). 2×105 phages were plated with Y1090 on ten 120 cm2 plates and incubated at 42°C for 3.5 hours. Then nylon filters (Nytran; Schleicher & Schuell) impregnated with 10 mM isopropylthio-β-D-galactopyranoside (IPTG) were overlaid, incubated at 37°C overnight, lifted, air-dried and processed for the guanidine-HCl denaturation-renaturation procedure (Vinson et al., 1988) in binding buffer [10 mM Hepes, 50 mM NaCl, 1 mM Na2HPO4, 0.1 mM ZnSO4, 1 mM dithiothreitol (DTT), 5% glycerol]. The filters were blocked with 5% skim milk (Difco) in binding buffer and washed twice in binding buffer containing 0.25% skim milk. A pair of filters were placed in a plastic bag containing binding buffer with 0.3 mg/ml poly(dA-dT)(Pharmacia), 0.25% skim milk and end-labeled, octamerized HNW probe (23 ng/ml, 3×105 disints/minute/ ml), and incubated at room temperature for 2 hours. The filters were washed by several changes of binding buffer and subjected to autoradiography.

Expression and purification of JF12 fusion proteins

JF12 cDNA insert of the original λgt11 clone was placed in pGEX-3X(NotI) for fusion with glutathione-S transferase (GST) or in pMAL-c2 vector (New England Biolabs) for fusion with a maltose-binding protein (MBP). To generate a GST fusion protein, the cDNA excised by NotI digestion was inserted into the NotI site of pGEX-3X(NotI) which was constructed by insertion of a NotI linker (New England Biolabs) at the SmaI site of pGEX-3X (Smith and Johnson, 1988) so that GST and JF12 coding sequences become in frame. To obtain MBP fusion protein, the cDNA was excised by EcoRI digestion and inserted in the EcoRI site of pMAL-c2. The fusion proteins were produced in E. coli and affinity-purified according to the published procedure (Smith and Johnson, 1988) or to the manufacturer’s instruction. In a rough estimate, 1 l of bacterial culture yielded 1 mg of GST fusion protein and 2 mg of MBP fusion protein. GST fusion protein was further purified using FPLC mono Q column (Pharmacia).

Generation of antiserum and affinity-purification of antibodies

Anti-GST:JF12 antiserum was raised in a rabbit by subcutaneous injection of the Mono Q-purified GST:JF12 fusion protein using Freund‘s adjuvant. The antiserum was loaded on a column of CNBr-activated Sepharose 4B (Pharmacia) coupled with MBP:JF12 fusion protein. Specific antibodies were eluted with 10 mM glycine (pH 2.5) and 100 mM triethylamine (pH 11.5). Eluates were combined and dialyzed against phosphate-buffered saline (PBS).

Gel mobility shift assay and western blotting of nuclear extracts

Nuclear extracts were prepared and subjected to gel mobility shift assay according to Funahashi et al. (1991). Immunoprecipitation and western blotting were done according to Kato et al. (1990) except that an ECL system (Amersham) was used in place of 125I-labeled protein A.

Transfection and luciferase assay

Construction of luciferase reporter gene and δEF1 expression vector pCMVX-δEF1 has been described by Kamachi and Kondoh (1993). The chicken embryonic cells in primary culture (Hayashi et al., 1987) were seeded at the density of 2×105 cells per 3.5 cm dish the day before, and transfected with 1 μg plasmid DNA according to Chen and Okayama (1987), washed after 6 hours, and harvested after 24 hours for luciferase assay (deWet et al., 1987). Luciferase activity was measured according to Kamachi and Kondoh (1993) at 25°C using a 1251 luminometer (Wallac).

Immunohistology

Embryos staged according to Hamburger and Hamilton (1951) were fixed with 3.5% paraformaldehyde in Hepes-buffered saline (HBS) at 4°C for 2-5 hours, washed, impregnated in a graded series of sucrose in HBS (10, 15 and 25%) at 4°C, embedded in OCT compound (Miles Scientific) and frozen on dry ice. Sections 4 μm thick were made in a cryostat (Bright), placed on glass slides coated with gelatin-chromealum and air-dried. Sections were incubated successively with purified anti-δEF1 antibodies, biotinylated anti-rabbit Ig (Amersham), fluorescein-conjugated streptoavidine (Amersham) and 4′,6′-diamidino-2-phenylindole dihydrochloride (DAPI; 20 μg/ml) in Tris-buffered saline (TBS) containing 0.1% Tween-20 and 10% skim milk (Difco), with washings with TBS between the steps. Finally, the sections were mounted in Gelvatol (PBS containing 20% polyvinyl alcohol, 20% glycerol, and 2.5% 1,4-diazabycyclo-[2,2,2]-octane) and examined under an Axioplan microscope (Zeiss).

Southwestern cloning of putative EF1 cDNA

For the purpose of cloning δEF1 cDNA, the southwestern approach (Vinson et al., 1988; Singh et al., 1988) was employed. A cDNA library was constructed from lens poly(A)+ RNA of 13 day chicken embryos in a λgt11 expression vector, and screened for binding of the cDNA-encoded protein fused with an N-terminal portion of β-galactosidase to the octameric HN fragment of the δ1-crystallin enhancer. The HN fragment was 55 bp long, contained the binding site of δEF1 near its 3′ end, and exhibited stringently lens-specific enhancer activity in its multimeric form (Funahashi et al., 1991). In the previous study (Funahashi et al., 1991), we noted that δEF1 binding was most clearly demonstrated in a gel mobility shift assay when poly(dA-dT) was used as competitor of non-specific DNA binding. Therefore, the same poly(dA-dT) was used in the screening.

From a total of 2×106 plaques screened, 5 positive clones were isolated. Analysis of binding specificity utilizing mutant probes (Funahashi et al., 1991) indicated that one of them encoded a protein that bound specifically to the δEF1-binding site (blocks 5 through 6; Fig. 1A) (data not shown). This clone, JF12, was further characterized. Two of the remaining four, JF9 and JF10, coded for proteins with HMG boxes (JF9-encoded protein was a chicken HMG1) and bound preferentially to AT-rich blocks 10 to 3, and the other two, JF11 and JF13, without known DNA-binding motifs bound specifically to block 10 (Fig. 1A). These nucleotide sequences will appear in EMBL/GenBank/DDBJ databases under the accession numbers D14314 (JF9), D14315 (JF10), D14316 (JF11), D14317/8 (JF13).

Fig. 1.

Gel mobility shift assay of δEF1 and GST:JF12 fusion protein, and the effect of anti-JF12 antibodies. (A) HN fragment sequence and its blocks. Mutant fragments had transversion type base alterations in the blocks indicated (Funahashi et al., 1991). DC5 is the 3′ half of the HN fragment used for probe of gel shifts. (B) Effect of mutations of HN fragments as competitor of DC5 gel shift. (a) Affinity-purified GST:JF12 (5 ng per lane), and (b) lens nuclear extract (10 μg per lane). −, no sequence-specific competitor; 3-7, HN fragments with mutations in the designated blocks. Arrowheads indicate the position of the complex with GST:JF12 (a) and δEF1 (b). Note that, for both GST:JF12 and δEF1, the mutation in block 5 reduced and that in block 6 totally abolished binding of DC5 sequence. (C) Effect of anti-GST:JF12 fusion protein serum to factor-probe complexes. (a) GST:JF12 (100 ng per lane), (b) lens nuclear extract (10 μg per lane) and (c) brain nuclear extract (5 μg per lane). −, no serum; P, preimmune serum; I, anti-fusion protein serum. Arrowheads indicate the complexes of the fusion protein or δEF1, and asterisk, the super-shifted complexes.

Fig. 1.

Gel mobility shift assay of δEF1 and GST:JF12 fusion protein, and the effect of anti-JF12 antibodies. (A) HN fragment sequence and its blocks. Mutant fragments had transversion type base alterations in the blocks indicated (Funahashi et al., 1991). DC5 is the 3′ half of the HN fragment used for probe of gel shifts. (B) Effect of mutations of HN fragments as competitor of DC5 gel shift. (a) Affinity-purified GST:JF12 (5 ng per lane), and (b) lens nuclear extract (10 μg per lane). −, no sequence-specific competitor; 3-7, HN fragments with mutations in the designated blocks. Arrowheads indicate the position of the complex with GST:JF12 (a) and δEF1 (b). Note that, for both GST:JF12 and δEF1, the mutation in block 5 reduced and that in block 6 totally abolished binding of DC5 sequence. (C) Effect of anti-GST:JF12 fusion protein serum to factor-probe complexes. (a) GST:JF12 (100 ng per lane), (b) lens nuclear extract (10 μg per lane) and (c) brain nuclear extract (5 μg per lane). −, no serum; P, preimmune serum; I, anti-fusion protein serum. Arrowheads indicate the complexes of the fusion protein or δEF1, and asterisk, the super-shifted complexes.

The cDNA insert of JF12 was 826 base pairs long and an open reading frame (ORF) spanned the entire insert length. The ORF encoded three consecutive zinc fingers as putative DNA-binding motifs in its deduced amino acid sequence, as described below, and these fingers must have served as the DNA-binding domain of the β-galactosidase:JF12 fusion protein.

Antibodies against JF12-encoded protein confirmed identity with βEF1

The cDNA sequence was excised from the phage vector and placed downstream of the GST sequence of a pGEX-based vector (Smith and Johnson, 1988). The GST:JF12 fusion protein was expressed in bacterial cells and purified using a glutathione column. This fusion protein had the same DNA-binding specificity as δEF1 in gel mobility shift assay (Fig. 1B). In addition, this binding was dependent upon the presence of zinc ions and inhibited by EDTA, consistent with zinc fingers being the DNA-binding domain (data not shown).

Antibodies were raised against the GST:JF12 fusion protein and examined for their effect on DNA-δEF1 complex. As shown in Fig. 1C, anti-GST:JF12 abolished the δEF1-DC5 complex in the same way as the GST:JF12-DC5 complex, and produced the supershifted bands. The effects of the antibodies on δEF1 were essentially the same regard-less of its source (Fig. 1C for lens and brain).

Thus, because of having the same sequence specificity of DNA binding and the same antigenicity, we concluded that the cDNA sequence of the clone JF12 represented a portion of δEF1 that included a DNA-binding domain.

Entire cDNA sequence and the encoded protein

The cDNA insert of JF12 was used to probe northern blots of poly(A)+RNAs prepared from various embryonic organs. A major RNA species of 5.5 kb was present in all organs examined (Fig. 2A), as expected from the apparently ubiquitous distribution of δEF1-binding activity (Funahashi et al., 1991). In addition, a second mRNA species of 4 kb was detected in non-lens tissues.

Fig. 2.

Northern blot analysis of δEF1 mRNA. 1.2 μg poly(A)+ RNAs from 13 day embryonic lens (1), brain (2) and heart (3) were electrophoresed and hybridized with JF12 probe (A), RS12-113 (5′) probe (B) and 3′ probe derived from 3′ half of RS12-32 (C) (see Fig. 3).

Fig. 2.

Northern blot analysis of δEF1 mRNA. 1.2 μg poly(A)+ RNAs from 13 day embryonic lens (1), brain (2) and heart (3) were electrophoresed and hybridized with JF12 probe (A), RS12-113 (5′) probe (B) and 3′ probe derived from 3′ half of RS12-32 (C) (see Fig. 3).

Starting from JF12, overlapping cDNA clones were isolated from λgt10 and λgt11 libraries of embryonic lens cDNAs, as shown in Fig. 3A(b). The cDNA sequence reconstructed from the overlapping clones was 5271 bases long without including poly(A), consistent with the 5.5 kb mRNA detected on northern blots. The longest ORF spanned 3342 bases (Fig. 3A), and coded for a protein of 124×103Mr. Western blotting of nuclear extracts of embryonic lens and brain with anti-JF12 (i.e. anti-δEF1) serum detected a band corresponding to 170×103Mr, apparently larger than the coding capacity of the cDNA. However, when the cDNA reconstructed from clones 12-118 and 12-119 was placed in an expression vector pCDM8 (Seed, 1987) and transfected to COS7 cells, the cDNA directed synthesis of the same 170×103Mr band (Fig. 4, lane 1). This indicated successful cloning of the entire δEF1 coding sequence.

Fig. 3.

cDNA clones of δEF1. (A) A scheme of the cDNA and the clones. (a) Reconstructed full-length cDNA and assignment of the coding region. The coding region is boxed, and the characteristic domains, zinc fingers (▩), the homeodomain (▦), the proline-rich domain (▧) and the acidic domain (■) are indicated. The wedges indicate locations of introns in the genomic sequence and the dots potential poly(A) addition sites. (b) Overlapping cDNA clones isolated from lens libraries. Clones RS12-32, RS12-118 and RS12-119 were from oligo(dT)-primed libraries and others from a random-primed library. (c) cDNA clones isolated from oligo(dT)-primed brain library screened with RS12-1 (5′ half) and RS12-113 as probes. (B) Nucleotide sequence and encoded amino acid sequence. The nucleotide sequence was generated from RS12-119, RS12-118 and RS12-32. The last nucleotide shown in the figure (position 5271) is the point to which a complete match was found with the genomic sequence, and was followed by a poly(A) tract in clone RS12-32. Potential poly(A) addition signals are boxed. In the amino acid sequence, zinc fingers are numbered and underlined, with landmark cystine and histidine residues shown stippled; the homeodomain are underlined by a broken line; potential glycosylation sites are circled; potential phosphorylation sites are boxed and stop codon are indicated by an asterisk. This nucleotide sequence will appear in EMBL/GenBank/DDBJ databases under the accession number D14313.

Fig. 3.

cDNA clones of δEF1. (A) A scheme of the cDNA and the clones. (a) Reconstructed full-length cDNA and assignment of the coding region. The coding region is boxed, and the characteristic domains, zinc fingers (▩), the homeodomain (▦), the proline-rich domain (▧) and the acidic domain (■) are indicated. The wedges indicate locations of introns in the genomic sequence and the dots potential poly(A) addition sites. (b) Overlapping cDNA clones isolated from lens libraries. Clones RS12-32, RS12-118 and RS12-119 were from oligo(dT)-primed libraries and others from a random-primed library. (c) cDNA clones isolated from oligo(dT)-primed brain library screened with RS12-1 (5′ half) and RS12-113 as probes. (B) Nucleotide sequence and encoded amino acid sequence. The nucleotide sequence was generated from RS12-119, RS12-118 and RS12-32. The last nucleotide shown in the figure (position 5271) is the point to which a complete match was found with the genomic sequence, and was followed by a poly(A) tract in clone RS12-32. Potential poly(A) addition signals are boxed. In the amino acid sequence, zinc fingers are numbered and underlined, with landmark cystine and histidine residues shown stippled; the homeodomain are underlined by a broken line; potential glycosylation sites are circled; potential phosphorylation sites are boxed and stop codon are indicated by an asterisk. This nucleotide sequence will appear in EMBL/GenBank/DDBJ databases under the accession number D14313.

Fig. 4.

Western blotting of δEF1 in embryonic tissues and expressed from cDNA using an anti-GST:JF12 antiserum. Lane 1, A nuclear extract of COS7 cells (6 μg) transfected with δEF1 cDNA (a composite of RS12-119 and RS12-118) on pCDM8 vector (Seed, 1987); lane 2, nuclear extract untransfected COS7 cells (6 μg); lane 3, chicken embryonic lens extract (6.5 μg); lane 4, chicken embryonic brain extract (5 μg); lane 5, immunoprecipitate of the same brain extract of a five times larger amount, showing that only the bands indicated by the arrowheads are the authentic antigens. The upper band indicated by the large arrowhead represents δEF1, and corresponds to the size of 170×103Mr. The lower band indicated by the small arrowhead (120×103Mr) and observed in the transfected COS7 cells and in brain extract represents proteolytic products of δEF1 generated during preparation, since inclusion of a protease inhibitor cocktail in the extraction medium significantly reduced the intensity of the bands.

Fig. 4.

Western blotting of δEF1 in embryonic tissues and expressed from cDNA using an anti-GST:JF12 antiserum. Lane 1, A nuclear extract of COS7 cells (6 μg) transfected with δEF1 cDNA (a composite of RS12-119 and RS12-118) on pCDM8 vector (Seed, 1987); lane 2, nuclear extract untransfected COS7 cells (6 μg); lane 3, chicken embryonic lens extract (6.5 μg); lane 4, chicken embryonic brain extract (5 μg); lane 5, immunoprecipitate of the same brain extract of a five times larger amount, showing that only the bands indicated by the arrowheads are the authentic antigens. The upper band indicated by the large arrowhead represents δEF1, and corresponds to the size of 170×103Mr. The lower band indicated by the small arrowhead (120×103Mr) and observed in the transfected COS7 cells and in brain extract represents proteolytic products of δEF1 generated during preparation, since inclusion of a protease inhibitor cocktail in the extraction medium significantly reduced the intensity of the bands.

In the reconstructed cDNA sequence, the ORF was preceded by an 11 base 5′ untranslated sequence, and followed by a 1918 base 3′ untranslated sequence. Examination of the 3′ untranslated region indicated a potential alternative poly(A) addition site immediately downstream of the 3′ end of the ORF. This alternative poly(A) addition site accounted for the 4 kb form of mRNA observed in non-lens tissues. In northern blots, the 5′ proximal cDNA probe detected both RNA species (Fig. 2B), but the probe downstream of the first poly(A) site detected only the 5.5 kb species (Fig. 2C).

When an oligo(dT)-primed cDNA library of 13 day embryonic brain was screened with clones RS12-113 and the 5′ half of RS12-1 [probes representing N-proximal and C-proximal regions, respectively, of the coding sequence; Fig. 3A(b)], the majority of the cloned cDNAs started near the initiator Met codon and terminated approximately 160 bases downstream of the termination codon [Fig. 3A(c)], consistent with the same mRNA start site between lens and brain, and also with the frequent use of the alternative poly(A) addition signal in the brain. cDNA sequences of the coding region were identical to those of the lens except for base changes to synonymous codons in a few places, which likely reflected genetic polymorphism in the chicken population. Thus, it was concluded that δEF1 proteins are identical in amino acid sequence in the lens and brain, and perhaps in other non-lens tissues.

Southern blot of chicken DNA using the δEF1 cDNA fragments as probe indicated that the δEF1 gene is likely present as a single copy per haploid genome (data not shown). Comparison of the nucleotide sequence of δEF1 cDNA with that of the cloned genomic gene (R. Sekido, M, Okanami, unpublished result) indicated that the cDNA sequence was divided into nine exons, as shown in Fig. 3A(a), and that the initial Met codon of the cDNA ORF was located in the proximity of the promoter region.

Multiple DNA-binding domains of δEF1: zinc finger clusters and a homeodomain

Amino acid sequence of δEF1 deduced from the cDNA sequence revealed an interesting feature of this DNA-binding protein (Fig. 3B). In addition to three zinc fingers near the C terminus (C fingers) initially identified in clone JF12, there were four additional zinc fingers of Cys2-His2 type near the N terminus (N fingers). The fourth N finger (finger 4) had a cystine in the place of a histidine in ordinary fingers, but other characteristics (location of phenylalanine, clustering of basic amino acid residues adjacent to a histidine, and spacings between typical amino acid residues) classified this finger as the Cys2-His2 type.

Very interestingly, a homeodomain was found in between the zinc finger clusters. The amino acid sequence of the homeodomain had the conserved landmark amino acid residues, and was similar to POU homeodomains [especially Drosophila I-POU (Treacy et al., 1992) and rat Brn-3 (He et al., 1989)] in the regions assigned to helices. An intriguing characteristic of the δEF1 homeodomain was that it lacks a basic amino acid cluster, which is usually found on its N-terminal side.

In the middle of the amino acid sequence, there was a domain rich in proline, and to the C-terminal side of the C fingers there was a domain highly rich in glutamic acid (acidic domain; Fig. 3A(a) and B), features common to transcriptional activators (Mitchell and Tjian, 1989; Seipel et al., 1992). There are several potential phosphorylation sites and a number of potential glycosylation sites (Fig. 3B).

A search for sequence similarity in DNA sequence data base revealed that the C-terminal half of the δEF1 was almost identical to human Nil-2-a (Williams et al., 1991). This will be discussed later.

A combination of multiply clustered zinc fingers and interposed homeodomain(s) has recently been reported for Drosophila ZFH proteins (ZFH-1 and 2) (Fortini et al., 1991; Lai et al., 1991) and for a rat α-fetoprotein enhancer binding protein (Morinaga et al., 1991). It is interesting to note that δEF1 and ZFH-1 not only have an organization of DNA-binding domains analogous to each other (Fig. 5A), but also have similar amino acid sequences in these DNA-binding domains (Fig. 5B). Especially fingers 3 to 4 of δEF1 were very similar to fingers 4 to 5 of ZFH-1 and also fingers 5 to 7 of δEF1 were similar to fingers 7 to 9 of ZFH-1 (Fig. 5B). It was also noted that fingers 3 to 4 and 6 to 7 of δEF1, i.e., the last two N fingers and C fingers are similar to each other (Fig. 5C). In regions other than the DNA-binding domains, no significant similarity of amino acid sequence was observed between δEF1 and ZFH-1.

Fig. 5.

Comparison of zinc fingers and homeodomains between δEF1 and ZFH-1. (A) Schematic alignment of zinc finger domains (cross-hatched) and homeodomains (hatched) on cDNA sequences where coding regions are boxed. (B) Comparison of amino acid sequences between DNA-binding domains. The cystine and histidine residues involved in zinc ion binding are indicated by asterisks. Identical amino acid residues are indicated by :, and the residues of similar hydrophobicity by . (C) Comparison of N fingers 3-4 and C fingers 6-7 of δEF1.

Fig. 5.

Comparison of zinc fingers and homeodomains between δEF1 and ZFH-1. (A) Schematic alignment of zinc finger domains (cross-hatched) and homeodomains (hatched) on cDNA sequences where coding regions are boxed. (B) Comparison of amino acid sequences between DNA-binding domains. The cystine and histidine residues involved in zinc ion binding are indicated by asterisks. Identical amino acid residues are indicated by :, and the residues of similar hydrophobicity by . (C) Comparison of N fingers 3-4 and C fingers 6-7 of δEF1.

Transrepression by δEF1

To address the question of what activity δEF1 has as a transcriptional regulator, we set up an experiment in which δEF1 expression vector was cotransfected with reporter luciferase gene carrying the HN fragment octamer (Fig. 6A). The HN fragment has strictly lens-specific enhancer activity (Funahashi et al., 1991). The luciferase gene was driven by the minimum δ-crystallin promoter that has no tissue preference (Hayashi et al., 1987; Kamachi and Kondoh, 1993). Octamerized HN enhancer fragment carrying the wild-type sequence (HNW) was placed upstream of the promoter to activate expression. To control the extent of activation, we used octamerized HN6 fragment carrying a mutation that totally inactivated the enhancer and δEF1 binding (Funahashi et al., 1991; Fig. 6B). In the absence of exogenous δEF1 expression, octameric HNW fragment elicited thirteen-fold activation over HN6 mutant in lens cells. By cotransfection of an increasing amount of δEF1 expression vector, there was a progressive decrease of the enhancer activity of the HN fragment (Fig. 6C). Under the condition that the molar input of δEF1 expression vector was nearly the same as that of the reporter luciferase gene, enhancer activity was decreased by fivefold. By contrast, the luciferase gene carrying the HN6 mutant fragments was not affected by exogenous δEF1 (Fig. 6E,F). The repressing effect of co-transfected δEF1 expression vector was observed only in lens cells (Fig. 6C,D). These results indicated that δEF1 functions as a repressor of the enhancer by binding to the HN fragment.

Fig. 6.

Transrepression of HN fragment-bearing luciferase gene by cotransfection of δEF1 expression vector. (A) Schemes of the δEF1 expression vector driven by cytomegalovirus promoter (top) and the luciferase-encoding reporter gene driven by the minimum δ-crystallin promoter and carrying octamerized HN fragment in the upstream (bottom). (B) The nucleotide sequences of wild-type HN fragment monomer (HNW) and of HN6 mutant. Linker sequences are in lower case letters and base sequence alteration is indicated by the stippled box. (C-F) The change of relative luciferase activity with increasing amount of δEF1 expression vector, taking the luciferase activity without exogenous δEF1 as 100%. Total copy number of expression vector plasmids and total amount of DNA (1 μg) were kept constant by inclusion of insert-less pCMVX vector and inert pUC plasmid in the transfection mixture, which minimized the effect of titration of transcription factors by exogenous DNA (Goto et al., 1990). (C) With wild-type HN fragments (HNW) in lens cells, (D) with HNW fragments in fibroblasts, (E) with HN6 mutant fragments in lens cells and (F) with HN6 mutant fragments in fibroblasts. The relative luciferase activity without exogenous δEF1 compared to HN fragment-less luciferase gene was 31, 0.5, 2.4 and 1.6 for C, D, E and F, respectively.

Fig. 6.

Transrepression of HN fragment-bearing luciferase gene by cotransfection of δEF1 expression vector. (A) Schemes of the δEF1 expression vector driven by cytomegalovirus promoter (top) and the luciferase-encoding reporter gene driven by the minimum δ-crystallin promoter and carrying octamerized HN fragment in the upstream (bottom). (B) The nucleotide sequences of wild-type HN fragment monomer (HNW) and of HN6 mutant. Linker sequences are in lower case letters and base sequence alteration is indicated by the stippled box. (C-F) The change of relative luciferase activity with increasing amount of δEF1 expression vector, taking the luciferase activity without exogenous δEF1 as 100%. Total copy number of expression vector plasmids and total amount of DNA (1 μg) were kept constant by inclusion of insert-less pCMVX vector and inert pUC plasmid in the transfection mixture, which minimized the effect of titration of transcription factors by exogenous DNA (Goto et al., 1990). (C) With wild-type HN fragments (HNW) in lens cells, (D) with HNW fragments in fibroblasts, (E) with HN6 mutant fragments in lens cells and (F) with HN6 mutant fragments in fibroblasts. The relative luciferase activity without exogenous δEF1 compared to HN fragment-less luciferase gene was 31, 0.5, 2.4 and 1.6 for C, D, E and F, respectively.

Expression of δEF1 during embryogenesis

Utilizing highly purified anti-δEF1 antibodies, tissue distribution of δEF1 was histologically examined focusing on early developmental stages of the chicken embryos. The results are summarized in Table 1. In brief, the major sites of δEF1 expression in the embryo were the mesodermal tissues, neuroectoderms, neural crest derivatives and the lens. In most of these tissues, δEF1 expression became detectable after outline of the tissue is established. Amnion expressed δEF1 from the beginning.

Table 1.

Summary of immunohistological analysis of δEF1 expression at various developmental stages

Summary of immunohistological analysis of δEF1 expression at various developmental stages
Summary of immunohistological analysis of δEF1 expression at various developmental stages

At stage 8 when three germ layers had just been established in the rostral half, no δEF1 expression was detectable (Fig. 7A,B). As the development proceeded past stage 10,δEF1 expression was initiated in the mesoderm, initially in the notochord and then in the somites, the nephrotomes and the lateral plates (Fig. 7C-F). This was followed by expression in the neural tube which began at stage 12. The beginning of δEF1 expression in these tissues was characterized by the scattered appearance of δEF1-positive nuclei in a tissue without remarkable polarity of distribution (Fig. 7C,E; Table 1, asterisks). With development, the fraction of δEF1-positive nuclei increased to nearly 100% in a tissue.

Fig. 7.

Immunohistology of δEF1 expression in the trunk of postgastrulation embryos. (A,B) Stage 8 cross section through neural fold (NF), somite (Sm) and notochord (Nc) stained with anti-δEF1 for FITC-fluorescence (A) and with DAPI (B). (C,D) Stage 14 cross section stained with anti-δEF1 (C) and DAPI (D). Neural tube (NT), somite (Sm) and notochord (Nc) are indicated. Note that most of the nuclei in the notochord are δEF1 positive, whereas the positive nuclei in the neural tube and the somite are scattered. (E,F) Sagittal (E) and parasagittal (F) sections of a stage 14 embryo through notochord and somites, respectively. Right is rostral. Note high δEF1 expression level throughout the notochord (E) and that, in the somites, δEF1-positive nuclei are present almost exclusively in the external epithelial layer (F). The bar indicates 50 μm for A to D and 100 μm for E and F.

Fig. 7.

Immunohistology of δEF1 expression in the trunk of postgastrulation embryos. (A,B) Stage 8 cross section through neural fold (NF), somite (Sm) and notochord (Nc) stained with anti-δEF1 for FITC-fluorescence (A) and with DAPI (B). (C,D) Stage 14 cross section stained with anti-δEF1 (C) and DAPI (D). Neural tube (NT), somite (Sm) and notochord (Nc) are indicated. Note that most of the nuclei in the notochord are δEF1 positive, whereas the positive nuclei in the neural tube and the somite are scattered. (E,F) Sagittal (E) and parasagittal (F) sections of a stage 14 embryo through notochord and somites, respectively. Right is rostral. Note high δEF1 expression level throughout the notochord (E) and that, in the somites, δEF1-positive nuclei are present almost exclusively in the external epithelial layer (F). The bar indicates 50 μm for A to D and 100 μm for E and F.

At stage 14 when somitic tissues differentiated, the level of δEF1 expression increased in the myotome and this became the most prominent site of expression (Fig. 8A,B for stage 18 embryo). In the same section, it is seen that expression of δEF1 in the neural tube was lower than in the somites. Migrating neural crest cells of the dorsal trunk did not express δEF1 (Fig. 8A, arrowheads), but those that colonized in the spinal ganglia (Fig. 8A,B) and in the mesenchymal compartment of the visceral arches (Fig. 8C,D) expressed δEF1. In the mesonephros, Wolffian duct initially expressed a modest level of δEF1 but soon ceased its expression (data not shown). Condensed mesenchymes and primitive mesonephric tubules strongly expressed δEF1, but once tubular structures were complete and connected to Wolffian duct, they stopped δEF1 expression (Fig. 8E,F).

Fig. 8.

Expression of δEF1 in mesoderm and neural crest derivatives. (A,B) A part of a cross section through a stage 18 embryo stained for δEF1 immunofluorescence (A) and for DAPI fluorescence (B). D, dermatome; M, myotome; NT, neural tube; Sc, sclerotome; SG, presumptive spinal ganglion. A cluster of migrating neural crest cells is indicated by the arrowheads. (C,D) A cross section through a visceral arch (hyoid arch) of a stage 18 embryo stained for δEF1 fluorescence (C) and for DAPI fluorescence (D). E, epithelial component derived from ecto- and endoderms; M, mesenchymal component mostly derived from the neural crest. (E,F). Mesonephros of a stage 23 embryo stained for δEF1 immunofluorescence (E) and for DAPI fluorescence (F). Note that nascent tubules (Tn) express high δEF1, while Wolffian duct (W) expresses little δEF1 at this stage and the tubules connected to the duct (T) express a very reduced level of δEF1. The bar indicates 50 μm.

Fig. 8.

Expression of δEF1 in mesoderm and neural crest derivatives. (A,B) A part of a cross section through a stage 18 embryo stained for δEF1 immunofluorescence (A) and for DAPI fluorescence (B). D, dermatome; M, myotome; NT, neural tube; Sc, sclerotome; SG, presumptive spinal ganglion. A cluster of migrating neural crest cells is indicated by the arrowheads. (C,D) A cross section through a visceral arch (hyoid arch) of a stage 18 embryo stained for δEF1 fluorescence (C) and for DAPI fluorescence (D). E, epithelial component derived from ecto- and endoderms; M, mesenchymal component mostly derived from the neural crest. (E,F). Mesonephros of a stage 23 embryo stained for δEF1 immunofluorescence (E) and for DAPI fluorescence (F). Note that nascent tubules (Tn) express high δEF1, while Wolffian duct (W) expresses little δEF1 at this stage and the tubules connected to the duct (T) express a very reduced level of δEF1. The bar indicates 50 μm.

When the lens placode was formed in the ectoderm and then began to invaginate (stage 14), δ-crystallin expression was clearly demonstrated but δEF1 was not detectable (data not shown). δEF1 became barely detectable only at the stage when the lens vesicle was formed (stage 18; Fig. 9A,B). This indicates that δEF1 itself is probably not required for initiation of δ-crystallin expression. However, as lens development proceeded and reached the stage of lens fiber formation, δEF1 became clearly detectable in all lens cell nuclei (Fig. 9C,D). Expression of δEF1 in the placodal components of the ectoderm was specific to the lens. Otic vesicles and nasal pit epithelia, for instance, remained δEF1 negative (data not shown). Later in lens development, e.g.,at stage 33 (day 8), δEF1 became expressed more strongly in the lens fiber cells than in the epithelial cells (Fig. 9E,F). Thus, it appeared that δEF1 was abundant in the nuclei of the cells that highly expressed the δ-crystallin gene.

Fig. 9.

δEF1 expression in the developing lens. (A,C,E) δEF1 immunofluorescence; (B,D,F) DAPI fluorescence. (A,B) Stage 18 embryo when lens vesicle is formed. A, amnion, L, lens vesicle and OC, optic cup. Note that δEF1 level is very low in the lens vesicle and in the continuing ectoderm, compared to the optic cup. (C,D) Stage 23 embryonic lens when lens fibers had elongated. Both external lens epithelium (LE) and internal lens fiber (LF) contained a high level of δEF1 in their nuclei, while the ectoderm remained very low in the δEF1 expression. R, retina; Cr, cornea. (E,F) Stage 33 (8 day) lens. δEF1 was positive in the annular pad (AP) and the peripheral fibers (PF) but negative in most of the epithelia area (E) and the cortical fiber cells (CF). The bar indicates 100 μm for A to D and 200 μm for E and F.

Fig. 9.

δEF1 expression in the developing lens. (A,C,E) δEF1 immunofluorescence; (B,D,F) DAPI fluorescence. (A,B) Stage 18 embryo when lens vesicle is formed. A, amnion, L, lens vesicle and OC, optic cup. Note that δEF1 level is very low in the lens vesicle and in the continuing ectoderm, compared to the optic cup. (C,D) Stage 23 embryonic lens when lens fibers had elongated. Both external lens epithelium (LE) and internal lens fiber (LF) contained a high level of δEF1 in their nuclei, while the ectoderm remained very low in the δEF1 expression. R, retina; Cr, cornea. (E,F) Stage 33 (8 day) lens. δEF1 was positive in the annular pad (AP) and the peripheral fibers (PF) but negative in most of the epithelia area (E) and the cortical fiber cells (CF). The bar indicates 100 μm for A to D and 200 μm for E and F.

Endodermal tissues and most other ectodermal tissues remained negative or very low in δEF1 expression.

The relationship with previously described Nil-2-a

The homology search for nucleotide/amino acid sequences in data bases identified human Nil-2-a (Williams et al., 1991) as having sequences very similar to δEF1. Nil-2-a was defined by a cDNA clone coding for a protein that binds to the −100 to −105 region of IL2 promoter, which has significant similarity to the block 6 of the HN fragment of the δ1-crystallin enhancer. Alignment of δEF1 and Nil-2-a amino acid sequence (Fig. 10) indicated that the latter matches very well to the C-terminal two-thirds of the δEF1 sequence. The sequence similarity extends to the region upstream of the first methionine residue of the Nil-2-a sequence, which is valine in the δEF1 sequence. This suggested that Nil-2-a-encoding cDNA is a partial cDNA of human δEF1 rather than an alternative form of δEF1. This has been supported by our comparative analysis of chicken and mouse δEF1 genes (R. Sekido, T. Takagi and Y. Higashi, unpublished result). Organization of the δEF1 gene is very well conserved between the chicken and the mouse, and hence likely in the human: the most N-proximal methionine codon of Nil-2-a is probably located in the middle of a large exon.

Fig. 10.

Alignment of amino acid sequences of δEF1 and Nil-2-a. Putative DNA-binding domains, zinc finger clusters and the homeobox are boxed, and identical amino acid residues on δEF1 and Nil-2-a sequences are stippled. The first methionine residue of Nil-2-a sequence is marked by an asterisk.

Fig. 10.

Alignment of amino acid sequences of δEF1 and Nil-2-a. Putative DNA-binding domains, zinc finger clusters and the homeobox are boxed, and identical amino acid residues on δEF1 and Nil-2-a sequences are stippled. The first methionine residue of Nil-2-a sequence is marked by an asterisk.

δEF1 as a ZFH family DNA-binding protein

An intriguing feature of δEF1 protein as a DNA-binding protein is that it has a multiplicity of potential DNA-binding domains as combinations of zinc-fingers and a homeo-domain in a way analogous to Drosophila zfh gene products (Fortini et al., 1991) and an α-fetoprotein enhancer binding factor (Morinaga et al., 1991). A striking similarity is found between δEF1 and ZFH-1 in the amino acid sequence of the DNA-binding domains and in their organization, which suggests that these two proteins have properties in common as DNA-binding proteins. It is interesting to note that δEF1 is expressed in most of the mesodermal lineages and, in Drosophila, zfh-1 expression is also primarily mesoderm-specific (Lai et al., 1991).

How the separate DNA-binding domains interact with genomic DNA sequences poses an interesting question concerning not only δEF1 itself but also proteins having multiple DNA-binding domains in general. The JF12 portion of δEF1 carrying only the C fingers had the same binding specificity as the native δEF1 when the DC5 fragment of the enhancer core was used as the probe of binding, indicating that C fingers alone account for the binding to the enhancer core region. In the case of PRDII-BF1, two zinc finger clusters were separately located close to the N and C termini, and these clusters had the same binding specificity (Fan and Maniatis, 1990). As shown in Fig. 5C, N fingers 3-4 had significant sequence similarities to C fingers 6-7, which may suggest that N fingers bind to nucleotide sequences having some similarity to C-finger-binding sites. It has been demonstrated that joining of zinc fingers of different binding specificity results in combined plural specificities (Keller and Maniatis, 1992), which may suggest that δEF1 has multiple sets of sequence specificities. It is not certain whether the homeodomain of δEF1 binds DNA in a sequence-specific manner, but the home-odomain of ZFH-1 does have such an activity (Fortini et al., 1991). Alternatively, the homeodomain of δEF1 may participate in protein dimer formation, as is suggested for POU homeodomains (Treacy et al., 1992) to which δEF1 homeodomain has an appreciable similarity.

δEF1 as a transcriptional regulator

An intriguing feature of δEF1 expression in the lens is that it begins after δ-crystallin expression is turned on and that it appears to increase along with δ-crystallin synthesis.

Cotransfection experiments using a δEF1 expression vector and HN enhancer fragment-carrying reporters demonstrated that δEF1 acts as a repressor of the δ-crystallin enhancer. By detailed mutational analysis of the HN fragment, we have shown that δEF1 competes with an activator in lens cells through overlapping binding sites (Kamachi and Kondoh, 1993). In embryonic lens cells where a high activator level is achieved, the repressor action of δEF1 is probably counteracted. However, in a variety of non-lens cells where the activator is not in effect, δEF1 likely represses δ-crystallin expression so as to make lens specificity very stringent. In fact, in a number of primary cultures and cell lines tested, δEF1 binding of HN fragment always resulted in repression of reporter genes (Kamachi and Kondoh, 1993).

How, then, can we reconcile the persistent and even augmented expression of δEF1 in developing lens with increase of δ-crystallin expression which seem to take place in parallel? It should be noted that there are a remarkable increase of crystallin synthesis and a substantial loss of non-crystallin expression, which occur simultaneously in maturating lens cells. Perhaps δEF1 acts to repress non-crystallin genes as long as they possess a δEF1-binding site but lack proper activating factors in lens cells.

However, considering the dynamic expression of δEF1 in early embryogenesis, it is tempting to speculate a positive function for δEF1 in a non-lens context. The domains rich in prolines in the middle and rich in glutamic acids near the C terminus resemble identified activation domains of various activators (Mitchell and Tjian, 1989; Seipel et al., 1992; and references therein). In a variety of cell-type-specific enhancers of non-crystallin origin, δEF1-binding sites have been identified. Thus, putative dual functions and heterogeneous DNA-binding domains associated with single δEF1 protein raise the possibility that this protein factor acts in a variety of ways depending on the cell type and the gene to be regulated.

Dynamic expression of δEF1 during embryogenesis

In the embryo, the primary site of δEF1 expression is the mesoderm. Expression of δEF1 begins when gastrulation is over and organogenesis has just begun. Most of the mesoderm derivatives express δEF1: the notochord first, and then somites, nephrotomes and lateral plates. It has been demonstrated that at stages 12 to 14 when all notochord nuclei contain a significant level of δEF1, the notochord is active in organizing dorsoventral polarity of the neural tube (Yamada et al., 1991). In the somite, δEF1 is turned on first in the external epithelial part leaving the cell mass in the core negative, and when three somitic compartments are differentiated, the myotome becomes and remains the site where δEF1 level is the highest. In the nephrotome, δEF1 is most conspicuous where tubular assembly is very actively going on. This dynamic change of expression strongly suggests that δEF1 has a significant role in the early histogenesis of mesodermal tissues.

In conjunction with mesodermal expression of δEF1, it is interesting to note that mesodermal expression of zfh-1 in Drosophila is dependent upon the activity of the snail and twist genes (Lai et al., 1991). Mouse homologues of the latter two, Sna (Smith et al., 1992) and M-twist (Wolf et al., 1991) have been cloned and their expression in the mid- to postgastrulation mouse embryos examined. Both Sna and M-twist are expressed in wide variety of mesodermal tissues and mesenchymal components of neural crest derivatives. Extrapolating the mouse data to chicken embryos, snail and twist homologues are expressed earlier than δEF1. It is possible that a hierarchy of mesodermal regulatory genes analogous to Drosophila exists in vertebrates.

Thus, expression pattern of δEF1 poses interesting problems of how this transcription factor gene is regulated in the spatiotemporal order of embryonic development, and how δEF1 exerts its effects on diverse cell types. Since genomic clones of the chicken and mouse δEF1 genes are already available (R. Sekido, M. Okanami, T. Takagi and Y. Higashi, unpublished), the problems can be solved if advantage is taken of transgenic and gene targeting technologies.

We thank S. Subramani for providing firefly luciferase gene, M. Okanami for assistance in cDNA analysis, and H. Nakamura, Y. Wakamatsu and Y. Higashi for discussion. This work was supported by grants from the Ministry of Education, Science and Culture of Japan to J. F., Y. K. and H. K., and by Special Coordinating Funds for Promoting Science and Technology from the Science and Technology Agency of Japan to H. K. In addition, J. F. and Y. K. are recipients of Fellowships from the Japan Society for the Promotion of Science for Japanese Junior Scientists.

Chen
,
C.
and
Okayama
,
H.
(
1987
).
High efficiency transformation of mammalian cells by plasmid DNA
.
Mol. Cell. Biol
.
7
,
2745
2752
.
deWet
,
J. R.
,
Wood
,
K. V.
,
Deluca
,
M.
,
Helinski
,
D. R.
and
Subramani
,
S.
(
1987
).
Firefly luciferase gene: structure and expression in mammalian cells
.
Mol. Cell. Biol
.
7
,
725
737
.
Fan
,
C-M.
and
Maniatis
,
T.
(
1990
).
A DNA-binding protein containing two widely separated zinc finger motifs that recognize the same DNA sequence
.
Genes Dev
.
4
,
29
42
.
Fortini
,
M. E.
,
Lai
,
Z.
and
Rubin
,
G. M.
(
1991
).
The Drosophila zfh-1 and zfh-2 genes encode novel proteins containing both zinc-finger and homeodomain motifs
.
Mech. Dev
.
34
,
113
122
.
Funahashi
,
J.
,
Kamachi
,
Y.
,
Goto
,
K.
and
Kondoh
,
H.
(
1991
).
Identification of nuclear factor δEF1 and its binding site essential for lens-specific activity of the δ1-crystallin enhancer
.
Nucl. Acids Res
.
19
,
3543
3547
.
Goto
,
K.
,
Okada
,
T. S.
and
Kondoh
,
H.
(
1990
).
Functional cooperation of lens-specific and nonspecific elements in the δ1-crystallin enhancer
.
Mol. Cell. Biol
.
10
,
958
964
.
Grainger
,
R. M.
(
1992
).
Embryonic lens induction shedding light on vertebrate tissue determination
.
Trends Genet
.
8
,
349
355
.
Hamburger
,
V.
and
Hamilton
,
H. L.
(
1951
).
A series of normal stages in the development of the chick embryo
.
J. Morphol
.
88
,
49
92
.
Hayashi
,
S.
,
Goto
,
K.
,
Okada
,
T. S.
and
Kondoh
,
H.
(
1987
).
Lens-specific enhancer in the third intron regulates expression of the chicken δ1-crystallin gene
.
Genes Dev
.
1
,
818
828
.
He
,
X.
,
Treacy
,
M. N.
,
Simmons
,
D. M.
,
Ingraham
,
H. A.
,
Swanson
,
L.W.
and
Rosenfeld
,
M. G.
(
1989
).
Expression of a large family of POU-domain regulatory genes in mammalian brain development
.
Nature
340
,
35
42
.
Kamachi
,
Y.
and
Kondoh
,
H.
(
1993
).
Overlapping positive and negative regulatory elements determine lens-specific activity of the δ1-crystallin enhancer
.
Mol. Cell. Biol., in press
.
Kato
,
K.
,
Kanamori
,
A.
and
Kondoh
,
H.
(
1990
).
Rapid and transient decrease of N-myc expression in retinoic acid-induced differentiation of OTF9 teratocarcinoma stem cells
.
Mol. Cell. Biol
.
10
,
486
491
.
Keller
,
A. D.
and
Maniatis
,
T.
(
1992
).
Only two of the five zinc fingers of the eukaryotic transcriptional repressor PRDI-BF1 are required for sequence-specific DNA binding
.
Mol. Cell. Biol
.
12
,
1940
1949
.
Lai
,
Z.
,
Fortini
,
M. E.
and
Rubin
,
G. M.
(
1991
).
The embryonic expression patterns of zfh-1 and zfh-2, two Drosophila genes encoding novel zinc-finger homeodomain proteins
.
Mech. Dev
.
34
,
123
134
.
Mitchell
,
P. J.
and
Tjian
,
R.
(
1989
).
Transcriptional regulation in mammalian cells by sequence-specific DNA binding proteins
.
Science
245
,
371
378
.
Morinaga
,
T.
,
Yasuda
,
H.
,
Hashimoto
,
T.
,
Higashio
,
K.
and
Tamaoki
,
T.
(
1991
).
A human α-fetoprotein enhancer-binding protein, ATBF1, contains four homeoproteins and seventeen zinc fingers
.
Mol. Cell. Biol
.
11
,
6041
6049
.
Seed
,
B.
(
1987
).
An LFA-3 cDNA encodes a phospholipid-linked membrane protein homologous to its receptor CD2
.
Nature
329
,
840
842
.
Seipel
,
K.
,
Georgiev
,
O.
and
Schaffner
,
W.
(
1992
).
Different activation domains stimulate transcription from remote (‘enhancer’) and proximal (‘promoter’) positions
.
EMBO J
.
11
,
4961
4968
.
Singh
,
H.
,
LeBowitz
,
J. H.
,
Baldwin
,
Jr. A. S.
and
Sharp
,
P. A.
(
1988
).
Molecular cloning of an enhancer binding protein, isolation by screening of an expression library with a recognition site DNA
.
Cell
52
,
415
423
.
Smith
,
D. B.
and
Johnson
,
K. S.
(
1988
).
Single-step purification of polypeptides expressed in Escherichia coli as fusions with glutathione S-transferase
.
Gene
67
,
31
40
.
Smith
,
D. E.
,
del Amo
,
F. F.
and
Gridley
,
T.
(
1992
).
Isolation of Sna, a mouse gene homologous to the Drosophila genes snail and escargot: its expression pattern suggests multiple roles during postimplantation development
.
Development
116
,
1033
1039
.
Treacy
,
M. N.
,
Neilson
,
L. I.
,
Turner
,
E. E.
,
He
,
X.
and
Rosenfeld
,
M. G.
(
1992
).
Twin of I-POU: a two amino acid difference in the I-POU homeodomain distinguishes an activator from an inhibitor of transcription
.
Cell
68
,
491
505
.
Vinson
,
C. R.
,
LaMarco
,
K. L.
,
Johnson
,
P. F.
,
Landschulz
,
W. H.
and
McKnight
,
S. L.
(
1988
).
In situ detection of sequence-specific DNA binding activity specified by a recombinant bacteriophage
.
Genes Dev
.
2
,
801
806
.
Williams
,
T. M.
,
Moolten
,
D.
,
Burlein
,
J.
,
Romano
,
J.
,
Bhaerman
,
R.
,
Godillot
,
A.
,
Mellon
,
M.
,
Rauscher
,
F. J.
and
Kant
,
J. A.
(
1991
).
Identification of a zinc finger protein that inhibits IL-2 gene expression
.
Science
254
,
1791
1794
.
Wolf
,
C.
,
Thisse
,
C.
,
Stoezel
,
C.
,
Thisse
,
B.
,
Gerlinger
,
P.
and
Perrin-Schmitt
,
F.
(
1991
).
The M-twist gene of Mus is expressed in subsets of mesodermal cells and is closely related to the Xenopus X-twi and the Drosophila twist genes
.
Dev. Biol
.
143
,
363
373
.
Yamada
,
T.
,
Placzek
,
M.
,
Tanaka
,
H.
,
Dodd
,
J.
and
Jessel
,
T. M.
(
1991
).
Control of cell pattern in the developing nervous system: polarizing activity of the floor plate and the notochord
.
Cell
64
,
635
647
.