Cloning and characterization of an ATBF1 isoform that expresses in a neuronal differentiation-dependent manner.

The human ATBF1 cDNA reported previously, now termed ATBF1-B, encodes a 306-kDa protein containing 4 homeodomains and 18 zinc fingers including one pseudo zinc finger motif. Here, we report the isolation of a second ATBF1 cDNA, 12 kilobase pairs long, termed ATBF1-A. The deduced ATBF1-A protein is 404 kDa in size and differs from ATBF1-B by a 920-amino acid extention at the N terminus. Analysis of 5′-genomic sequences showed that the 5′-noncoding sequences specific to ATBF1-A and ATBF1-B transcripts were contained in distinct exons that could splice to a downstream exon common to the ATBF1-A and ATBF1-B mRNAs. The expression of ATBF1-A transcripts increased to high levels when P19 and NT2/D1 cells were treated with retinoic acid to induce neuronal differentiation. Preferential expression of ATBF1-A transcripts was also observed in developing mouse brain. Transient transfection assays showed that the 5.5-kilobase pair sequence upstream of the ATBF1-A-specific exon (exon 2) supported expression of the linked chloramphenicol acetyltransferase gene in neuronal cells derived from P19 cells but not in undifferentiated P19 or in F9 cells, which do not differentiate into neurons. These results showed that ATBF1-A and ATBF1-B transcripts are generated by alternative promoter usage combined with alternative splicing and that the ATBF1-A-specific promoter is activated during neuronal differentiation.

The ATBF1 (AT motif binding factor 1) cDNA was first isolated from HuH-7 human hepatoma cells based on the ability of its product to bind to an AT-rich enhancer element of the human ␣-fetoprotein gene (AFP) (1). This protein is characterized by a large size (306 kDa) and the presence of four homeodomains (I-IV) and 18 zinc fingers including 1 pseudo zinc finger motif. Transient transfection assays showed that ATBF1 suppressed the activity of the AT-rich element of the enhancer and promoter of the AFP gene but not those of the albumin gene (2). This effect is thought to be mediated by specific interaction between homeodomain IV of the ATBF1 molecule and the AT-rich element of the AFP gene (1,2).
Besides ATBF1, there are three transcription factors that contain both homeodomain and zinc finger motifs. Chicken ␦EF1 has one homeodomain and nine zinc finger motifs. This protein has been shown to repress activities of the DC segment of the ␦1-crystallin enhancer (3,4) and the E2 element of the immunoglobulin chain and muscle creatine kinase enhancer (5). Its expression pattern in chicken embryos suggests that ␦EF1 plays a role in mesoderm development and embryonic myogenesis (3)(4)(5). Drosophila ZFH-1 contains one homeodomain and nine zinc finger motifs (6). It is expressed in the embryonic mesoderm and nervous system (7), and phenotypic analysis of loss-of-function mutant embryos has shown that ZFH-1 determines cell fate or positioning (8). Drosophila ZFH-2 contains 3 homeodomains and 16 zinc finger motifs (6) and is expressed almost exclusively in the central nervous system of Drosophila embryos (7). It binds to the RCS element of the opsin gene through homeodomain III (6,9) and activates the SER L element of the DOPA decarboxylase gene through homeodomain II (6,9).
Sequence comparison shows that homeodomains I, II, and III of ATBF1 are 77, 69, and 61% identical with the corresponding homeodomains of ZFH-2 (9,10). Homeodomain IV of ATBF1 is 46% identical with homeodomain III of ZFH-2. 13 zinc fingers of ATBF1 and ZFH-2 show identities ranging from 22 to 89%. All of these homologous domains are colinearly arranged in the ATBF1 and ZFH-2 molecules. These observations suggest that ATBF1 and ZFH-2 may have similar functions, raising the possibility that ATBF1 plays a role in mammalian central nervous system development. In fact, the level of ATBF1 transcripts is highly elevated in embryonic and neonatal mouse brains (11). In addition, ATBF1 expression is activated during retinoic acid-induced neuronal differentiation of P19 embryonal carcinoma cells (11).
In this paper, we report the isolation of a second ATBF1 cDNA, termed ATBF1-A, which is 3.3 kb 1 longer than the previously reported clone, now termed ATBF1-B. We show that ATBF1-A transcripts are generated by alternative splicing and alternative usage of a promoter region that is activated in neuronally differentiating cells. ␣MEM with 10% fetal bovine serum. To induce neuronal differentiation, P19 cells were allowed to form aggregates in bacteriological grade Petri dishes for 4 days in the presence of 0.5 M retinoic acid and transferred to tissue culture grade Petri dishes without retinoic acid (12). The day of adding retinoic acid to the medium was assigned day 0. F9 mouse embryonal carcinoma cells were grown in Dulbecco's modified Eagle's medium with 10% fetal bovine serum on gelatin-coated dishes. Visceral endoderm-like cells were induced from these cells by treatment with 0.05 M retinoic acid (13). NT2/D1(NTERA-2.cl.D1) human embryonal carcinoma cells were grown in Dulbecco's modified Eagle's medium with 10% fetal bovine serum. Neuronal differentiation was induced from NT2/D1 cells by treatment with 10 M retinoic acid for 3 weeks (14). HuH-7 human hepatoma cells were maintained in a chemically defined medium, ISE-RPMI 1640 (15). huH-1 human hepatoma cells were grown in RPMI 1640 with 1% fetal bovine serum.

Cells
Isolation of cDNA Clones-An M426 human fibroblast cDNA library, pCEV15 (16), was screened using a human ATBF1 cDNA, 488 (1), labeled with [␣-32 P]dCTP as a probe. This resulted in the isolation of B and E clones. These clones were then used as probes to screen a second M426 fibroblast cDNA library, pCEV-lacZ (17), to isolate I and ME clones. A short sequence between ME and I was isolated by RT-PCR using primers corresponding to ME and I sequences.
Isolation of Genomic DNA Clones-A WI38 human lung fibroblast genomic DNA library in FIX II (Stratagene) was screened using three cDNA fragments, E, ME, and MEM, as probes ( Fig. 1). E probe (0.5 kb) was prepared from the 5Ј portion of E clone corresponding to a 5Јnoncoding region of ATBF1-B. ME probe (0.7 kb) was prepared from the 5Ј portion of ME clone. It consisted of 0.4 kb of the 5Ј-noncoding sequence and 0.3 kb of the 5Ј-coding sequence of ATBF1-A. MEM probe (0.7 kb) was prepared from a coding region common to ATBF1-A and ATBF1-B. These DNA fragments were labeled with [␣-32 P]dCTP.
DNA Sequencing-Nucleotide sequences were determined from both strands by the dideoxy chain termination procedure with standard sequencing primers and custom-made primers using Sequenase Kit (Stratagene) with [␣-35 S]dATP or using Taq DyeDeoxy Terminator Cycle Sequencing Kit (Applied Biosystems).
Northern Blot Analysis-Total RNA was isolated by the guanidium isothiocyanate-cesium chloride method (18). ATBF1 transcripts were detected by hybridization with I probe (0.9 kb) prepared from I clone ( Fig. 1) for human samples or the mouse ATBF1 cDNA clone (11) for mouse samples. Glyceraldehyde-3-phosphate dehydrogenase mRNA was detected using a 0.78-kb glyceraldehyde-3-phosphate dehydrogenase cDNA probe as described previously (11). The probes were labeled with [␣-32 P]dCTP.
RNase Protection Assays-To prepare MD14 probe for RNase protection assays for human ATBF1-A and ATBF1-B transcripts, 256 bp of human ATBF1-A cDNA sequence from nucleotides 3353-3608 was inserted into pBluescript II KS(ϩ) (Stratagene) (Fig. 1). This was digested with BamHI and transcribed by T7 RNA polymerase (Boehringer Mannheim) in the presence of [␣-32 P]CTP to produce a radioactive 287-bp fragment consisting of 256 bp of the ATBF1-A sequence and 31 bp of the vector sequence. This probe yielded a 256-bp fragment with ATBF1-A mRNA and a 216-bp fragment with ATBF1-B mRNA (Fig. 4).
To prepare BX151 probe for mouse ATBF1 mRNA, 151 bp of mouse ATBF1 cDNA corresponding to nucleotides 3353-3503 of human ATBF1-A cDNA was inserted into pBluescript II KS(ϩ) (Fig. 1). This was digested with BamHI and transcribed by T3 RNA polymerase (Boehringer Mannheim) in the presence of [␣-32 P]CTP to produce a radioactive 228-bp fragment consisting of 151 bp of the mouse ATBF1 sequence and 77 bp of the vector sequence. This probe yielded a 151-bp fragment with ATBF1-A mRNA and a 111-bp fragment with ATBF1-B mRNA (Fig. 4).
Total RNA (5 g) was hybridized with 5 ϫ 10 5 cpm of probe in 40 mM PIPES (pH 6.4), 80% formamide, 0.4 M NaCl, and 1 mM EDTA at 45°C overnight. The reaction mixture was digested with 40 g/ml RNase A and 2 g/ml RNase T1 at 30°C for 30 min. The sample was then treated with 130 g/ml proteinase K, electrophoresed on a 5% polyacrylamide/ urea gel, and autoradiographed.
Mapping of the 5Ј-End of mRNA-For the determination of the 5Ј-end of ATBF1-B mRNA, the 603-bp StyI-DraI fragment consisting of a 5Ј-portion of exon 1 and its 5Ј-flanking sequence was cloned into pBluescript II KS(ϩ), digested with StyI, and transcribed by T3 RNA polymerase. The resultant 680-bp fragment consisting of 603 bp of the genomic sequence and 77 bp of the vector sequence was used as a probe (1B probe, Fig. 6). For the determination of the 5Ј-end of ATBF1-A mRNA, the 428-bp StuI-BamHI fragment consisting of a 5Ј portion of exon 2 and its 5Ј-flanking sequence was cloned into pBluescript II KS(ϩ), digested with StuI, and transcribed by T7 RNA polymerase. The resultant 474-bp fragment consisting of 428 bp of the genomic sequence and 46 bp of the vector sequence was used as a probe (1A probe, Fig. 6). Both probes were labeled with [␣-32 P]CTP. Total RNA (10 g) from M426 cells was hybridized with 5 ϫ 10 5 cpm of RNA probes, digested with RNase A and RNase T1 followed by proteinase K, and analyzed by gel electrophoresis as described for RNase protection assays above.
CAT Plasmids-To construct pUMS-CAT, the CAT gene and the poly(A) signal were removed from pSV2-CAT (19) and inserted into pBluescript II KS(ϩ). The upstream mouse sequence (UMS) of the c-mos gene was removed from pUBT-luc (20) and inserted upstream of the CAT gene. To construct pA5.5-CAT, the 5.5-kb (SalI-BamHI) 5Јflanking sequence of exon 2 (the first exon specific to ATBF1-A mRNA) was inserted between UMS and the CAT gene of pUMS-CAT. To construct pB3.5-CAT, the 3.5-kb (NotI-DraI) 5Ј-flanking sequence of exon 1 (the first exon specific to ATBF1-B mRNA) was inserted between UMS and the CAT gene of pUMS-CAT.
CAT Assays-Cells were transfected with 10 g of chimeric CAT plasmids by the calcium phosphate precipitation method. pCH110 ␤galactosidase expression plasmid (10 g) (Pharmacia Biotech Inc.) was cotransfected to monitor for transfection efficiency to normalize CAT activity. Cells were harvested 48 h later and lysed by five cycles of freezing and thawing. Part of the lysate was used to assay ␤-galactosidase activity (21). The remaining lysate was heated at 65°C for 10 min and centrifuged at 12,000 ϫ g for 5 min, and the supernatant was analyzed for CAT activity (19).

RESULTS
Isolation of ATBF1-A cDNA-Screening of pCEV15 cDNA library with the 488 human ATBF1 cDNA (1) as a probe resulted in the isolation of B and E clones (Fig. 1). These clones were then used as probes to screen pCEV-lacZ cDNA library to isolate I and ME. Sequence analysis showed that I and B were identical, representing the 3Ј-half of the ATBF1 cDNA previously reported (1). E and ME represented the sequence upstream of I, but 527 nucleotides at the 5Ј-end of E were entirely different from the corresponding region of ME. This segment of E contained the 5Ј-noncoding sequence followed by the coding sequence of the ATBF1 cDNA reported previously. In ME, we found a reading frame extending upstream of the translation initiation codon in E. Thus, ME represented a new cDNA sequence with a longer coding sequence. To complete sequence analysis of the new ATBF1 cDNA, we isolated a short region between ME and I by RT-PCR. The sequence of the RT-PCR product was identical with the corresponding region of ATBF1 previously reported. Combining ME, I, and the RT-PCR product, the sequence of the new ATBF1 cDNA was constructed (Fig. 2). This cDNA is referred to as ATBF1-A and the ATBF1 cDNA reported previously as ATBF1-B.
The ATBF1-A cDNA encodes a protein of 3703 amino acids with a molecular mass of 404 kDa. This protein contains 4 homeodomains and 23 zinc finger motifs including 1 pseudo zinc finger motif. ATBF1-A is longer than ATBF1-B by 920 amino acids added at the N terminus. This extended region contains five zinc fingers, two acidic domains (amino acids 110 -145 and 432-510), and one region rich in both serine and threonine (39%, amino acids 396 -431) (Fig. 3). In addition, computer analysis (22) (23,24), and an RNA-binding motif, QphpRlaR (amino acids 708 -715) (23,25). An ATP-binding motif, ASGSAGKS (7), at amino acids 2930 -2937 and two lysine residues at 2959 and 2962 (26) may also be involved in ATP binding functions (in the amino acid sequences described above, the residues indicated in upper case are those found in the consensus sequences).

Expression of ATBF1-A and ATBF1-B Transcripts in Cell
Lines and Developing Mouse Brain-Northern blot analysis using the I clone as a probe detected faint bands in the region of 10 -13 kb in M426, HuH-7, and huH-1 human cell lines (Fig.  4A, lanes 1-3). These bands corresponded to ATBF1-A and ATBF1-B transcripts, but the levels of their expression were very low in all these cell lines. Similarly, P19 embryonal carcinoma cells contained small amounts of ATBF1-A and ATBF1-B transcripts (Fig. 4A, lane 4). However, when P19 cells were treated with retinoic acid to induce neuronal differentiation, there were large increases in the levels of these transcripts (Fig. 4A, lane 5) (11).
To analyze more precisely the relative amount of ATBF1-A and ATBF1-B transcripts, we conducted RNase protection assays using MD14 probe capable of distinguishing these two types of transcripts. The results showed that the three human cell lines described above expressed both types of transcripts (Fig. 4B, lanes 2-4). The amounts of ATBF1-A transcripts were 5-10-fold higher than those of ATBF1-B transcripts, although the absolute levels of these transcripts were low. Similarly, small amounts of ATBF1-A and ATBF1-B transcripts were detected in undifferentiated NT2/D1 human embryonal carcinoma cells (Fig. 4B, lane 5). The band representing ATBF1-B transcripts was very weak but visible after long exposure. Upon induction of neuronal differentiation from these cells by treatment with retinoic acid, the level of ATBF1-A transcripts increased about 50-fold (Fig. 4B, lane 6). The level of ATBF1-B transcripts also increased but to a much lesser extent. Thus the ratio of ATBF1-A and ATBF1-B transcripts in differentiated NT2/D1 cells was 50:1 as compared to 5:1 in undifferentiated cells.
Similar assays of P19 cells using BX151 probe showed that, as in the case of NT2/D1 cells, the level of ATBF1-A transcripts was 50-fold higher in retinoic acid-treated cells than in untreated cells (Fig. 4B, lanes 8 and 9). Also, as in the case of NT2/D1 cells, the increase in the level of ATBF1-B transcripts was much less, giving rise to a 50:1 ratio of ATBF1-A and ATBF1-B transcripts in retinoic acid-treated cells. These results showed that neuronal differentiation of P19 and NT2/D1 cells is accompanied by preferential synthesis of ATBF1-A transcripts.
RNase protection assays of ATBF1 transcripts were also conducted on embryonic and neonatal mouse brains. The results showed high levels of ATBF1 expression, predominantly in the form of ATBF1-A, in brains of 15-day-old embryos and 1-day-old neonates (Fig. 5). Expression of ATBF1 transcripts was decreased in brains of 5-and 7-day-old neonates.
Identification of Alternatively Spliced Exons-To explore the mechanism by which ATBF1-A and ATBF1-B mRNAs are generated, we analyzed 5Ј-genomic sequences encompassing the 5Ј-ends of these transcripts. Two genomic clones were isolated. One clone, FIX-17 (Fig. 6), could hybridize to ME and E probes representing the 5Ј-noncoding sequences of ATBF1-A and ATBF1-B mRNAs, respectively (Fig. 1). The other clone, FIX-13 (Fig. 6), could hybridize to ME probe and also to MEM probe, which represents a coding region common to ATBF1-A and ATBF1-B (Fig. 1).
Southern blot and restriction analysis of these clones showed that FIX-17 contained exons 1 and 2, and FIX-13 carried exons 3 and 4 (Fig. 6). Exons 1 and 2 in FIX-17 were further characterized by sequence analysis and RNase protection assays to define the 5Ј boundaries (Fig. 7). The results showed that exon 1 carried most of the 5Ј-noncoding sequence of ATBF1-B mRNA and exon 2 that of ATBF1-A mRNA (Fig. 2). Exon 3 in FIX-13 contained the remaining ATBF1-A noncoding sequence (49 bp) followed by the ATBF1-A protein coding sequence. Exon 4 contained the remaining ATBF1-B noncoding sequence (23 bp) followed by the ATBF1-B protein coding sequence. Analysis of exon-intron junctions confirmed the presence of the consensus splice donor (GT) sites and acceptor (AG) sites (results not shown). In addition, it was found that exon 4 could splice to either exon 1 or exon 3 (Fig. 6). In the former case, the ATBF1-B mRNA sequence is generated, and in the latter case, the ATBF1-A mRNA sequence is generated with the ATBF1-B noncoding sequence in exon 4 becoming a part of the ATBF1-A coding sequence (Fig. 2). These results are consistent with the mechanism that ATBF1-A and ATBF1-B mRNAs are generated by alternative splicing.
Determination of Promoter Activity-The presence of distinct 5Ј exons specific to ATBF1-A and ATBF1-B sequences suggested that two different promoters may be used to initiate transcription of the two types of ATBF1 mRNAs. To examine whether the 5Ј-flanking sequences of exons 1 and 2 have transcriptional activity, they were linked to the CAT reporter gene (Fig. 8A) and transfected into undifferentiated and neuronally differentiated P19 cells. The 5.5-kb 5Ј-flanking sequence of ATBF1-A-specific exon 2 (pA5.5-CAT) supported CAT expression in differentiated P19 cells (Fig. 8B, lane 7) but not in undifferentiated P19 cells (Fig. 8B, lane 3). No CAT activity was expressed in F9 embryonal carcinoma cells with or without retinoic acid treatment (Fig. 8C). Treatment of F9 cells with retinoic acid resulted in induction of endodermal cells but not neuronal cells. pA5.5-CAT expressed CAT activity in other neuronal cells, such as Neuro-2a and retinoic acid-treated NT2/D1 cells, but not in non-neuronal cells, such as HepG2 (results not shown). These results showed that the 5Ј-flanking sequence of exon 2 has neuronal cell-specific promoter activity.
The 3.5-kb 5Ј-flanking sequence of ATBF1-B-specific exon 1 (pB3.5-CAT), on the other hand, did not support CAT expression in undifferentiated or differentiated P19 cells or any other cell lines described above. DISCUSSION We report here the isolation of a second human ATBF1 cDNA, termed ATBF1-A. This cDNA differs from the previously reported ATBF1 cDNA, termed ATBF1-B, by an extra 3.3-kb sequence at the 5Ј-end. Since the extended region can  encode five additional zinc fingers, helicase-related sequences, and domains rich in acidic amino acids and serine and threonine, it is possible that ATBF1-A may have functions not associated with the ATBF1-B isoform.
RNase protection assays detected two sizes of mRNAs corresponding to ATBF1-A and ATBF1-B in various cell lines and mouse brain. In all cases, ATBF1-A transcripts were present in larger amounts than ATBF1-B transcripts, but the absolute levels of these transcripts were low in various cell lines and adult mouse brain. Similarly, the amounts of these ATBF1 mRNAs were low in undifferentiated P19 and NT2/D1 embryonal carcinoma cells. However, neuronal cells derived from these cells by treatment with retinoic acid contained much higher levels of these transcripts. The increased expression was particularly pronounced with ATBF1-A transcripts, relative to ATBF1-B transcripts. Preferential expression of the ATBF1-A form was also observed in developing mouse brain.
The isolation and analysis of 5Ј-genomic sequences defined the basis for the generation of the two species of ATBF1 mRNAs. We found that 5Ј-noncoding sequences specific to ATBF1-A and ATBF1-B mRNAs were contained in distinct exons. We also found a downstream exon that can splice to either the ATBF1-A-or ATBF1-B-specific exon. These results showed that alternative splicing is involved in the generation of the two ATBF1 isoforms. Transient transfection experiments showed that the 5Ј-flanking region of exon 2, the first exon specific to ATBF1-A, exhibited promoter activity in neuronal cells derived from P19 cells but not in undifferentiated P19 cells. No promoter activity was expressed in F9 embryonal carcinoma cells or other non-neuronal cells. These results indicate that the 5Ј-flanking sequence of the ATBF1-A-specific exon functions as a neuronal cell-specific promoter. It is likely that this promoter is responsible for the observed increase in ATBF1-A transcripts in association with neuronal differentia-tion of P19 or NT2/D1 cells. Recent CAT assays showed that the 5Ј-flanking sequence of exon 2 could be shortened to 300 bp without losing promoter activity in neuronal cells. This indicates that the relatively short promoter region is sufficient to confer neuronal cell specificity. Computer search has revealed that the 300-bp region contains putative binding sites for several transcription factors, including c-fos, AP-2, SP-1, and zif268 (also known as egr-1, krox24, or NGF-A). Whether these factors are in fact involved in promoter activation in neuronally differentiating P19 cells is being investigated in our laboratory.
Although it is possible that ATBF1-B transcripts are produced by a differentially regulated promoter, we have not yet detected transcriptional activity associated with the 5Ј-flanking region of exon 1 (ATBF1-B-specific exon). Our failure to detect promoter activity in this region is likely due to very low levels of ATBF1-B expression in the cell lines used for CAT transfection assays. Obviously, these cells are deficient of certain transcription factors important for the synthesis of the ATBF1-B isoform. We are currently searching for cell lines expressing higher levels of ATBF1-B transcripts to be used for determination of ATBF1-B-specific promoter activity.