The Basic Helix-Loop-Helix-Zipper Transcription Factor USF1 Regulates Expression of the Surfactant Protein-A Gene*

Expression of the rabbit pulmonary surfactant protein A (SP-A) gene is lung-specific, occurs primarily in type II cells, and is developmentally regulated. We previously identified two E-box-like enhancers, termed the distal binding element (DBE) and proximal binding element (PBE), in the 5′-flanking region of the rabbit SP-A gene. In the present study, the PBE was used to screen a rabbit fetal lung cDNA expression library; a cDNA insert was isolated which is highly similar in sequence to human upstream stimulatory factor 1 (hUSF1). By use of reverse transcription polymerase chain reaction, two isoforms of rabbit USF1 (rUSF1) mRNAs were identified in fetal rabbit lung and other tissues. The levels of rUSF1 mRNAs reach a peak in fetal rabbit lung at 23 days gestation, in concert with the time of initiation of SP-A gene transcription. Binding complexes of nuclear proteins obtained from fetal rabbit lung tissue and isolated type II cells with the DBE and PBE were supershifted by the addition of anti-rUSF1 IgG. Binding activity was enriched in type II cells compared with lung fibroblasts. Overexpression of rUSF1s in A549 adenocarcinoma cells positively regulated SP-A promoter activity of cotransfected reporter gene constructs. It is suggested that rUSF1s, which bind to two E-box elements in the SP-A gene 5′-flanking region, may serve a key role in the regulation of SP-A gene expression in pulmonary type II cells.

Surfactant, a developmentally regulated lipoprotein produced by pulmonary type II cells, acts to reduce alveolar surface tension and prevent atelectasis; surfactant production is initiated in fetal lung only after ϳ75% of gestation is completed. The adaptation of the fetus to extrauterine life is highly dependent upon the maturity of the lung at birth and its capacity to produce surfactant. Lung surfactant contains at least four associated proteins, surfactant protein (SP) 1 -A, SP-B, SP-C, and SP-D, which appear to serve important roles in surface activity, phospholipid reutilization, and immune function within the alveolus (1). The surfactant protein genes are developmentally regulated and expressed in a lung-specific manner. The gene encoding SP-A is expressed primarily in type II cells and to a lesser extent in bronchioalveolar epithelial (Clara) cells (2,3). SP-A gene expression in fetal lung is under multifactorial control; glucocorticoids and agents that increase cyclic AMP appear to play important roles in its regulation (4).
Transcription of the SP-A gene is initiated in fetal rabbit lung tissue on day 24 of a 31-day gestation period. Within the 5Ј-flanking region of the rabbit SP-A gene, we have identified two E-box-like motifs, termed distal binding element (DBE at Ϫ985 bp) and proximal binding element (PBE at Ϫ85 bp), which bind rabbit lung nuclear proteins in an apparently identical manner (5). Binding activity is enriched in type II pneumonocytes compared with whole lung tissue. The DBE and PBE share a similar sequence and compete for binding to the same size species of nuclear proteins of 69, 45, and 22 kDa (5). In type II cell transfection studies, we found that rabbit SP-A gene 5Ј-flanking sequences between Ϫ991 and Ϫ47 bp mediate maximal levels of basal and cyclic AMP-induced expression of SP-A promoter activity (6). The finding that mutagenesis of either of the E-boxes (5) or a cyclic AMP-response element (CRE)-like sequence (6,7) resulted in a marked decrease of basal and cyclic AMP-induced expression of SP-A:hGH fusion genes in transfected type II cells indicates that these elements act in a cooperative manner to regulate SP-A promoter activity.
In the present study, we screened a rabbit fetal lung cDNA expression library using the PBE sequence as probe. A cDNA insert was isolated encoding the rabbit homolog of human USF1. By use of reverse transcription-polymerase chain reaction (RT-PCR) we found that there are two alternatively spliced forms of USF1 mRNA in rabbit tissues. Rabbit USF1s (rUSF1s), which are enriched in type II cells, bind to the DBE and PBE and positively regulate SP-A promoter activity. The finding that the levels of rUSF1 mRNA reach a maximum in fetal rabbit lung just before the time of initiation of SP-A gene transcription suggests that this transcription factor may play a role in the developmental regulation of SP-A gene expression.

EXPERIMENTAL PROCEDURES
Cloning of a cDNA Insert Encoding a PBE-binding Protein-First strand cDNA was synthesized from poly(A) ϩ RNA isolated from 24-day fetal rabbit lung tissues using random hexanucleotides and was used to generate second strand cDNA using a cDNA synthesis kit (Pharmacia Biotech Inc., You-Prime cDNA synthesis kit). The double-stranded cDNAs with EcoRI/NotI linkers were inserted into a gt11 vector and packaged by use of Gigapack II gold packaging extract (Stratagene). A 32 P-labeled double-stranded oligonucleotide corresponding to the PBE was used to screen the gt11 cDNA expression library employing standard techniques (8). Upon screening of ϳ2 million recombinant phage clones, a specific cDNA insert encoding a protein that bound specifically to the radiolabeled DBE and PBE, but not to nonspecific DNA, was isolated. The cDNA insert was subcloned into the pGEM-7Z plasmid vector (Promega) and sequenced using Sequenase 2.0 (U. S. Biochemical Corp.). The nucleic acid sequence of the 950-bp cDNA insert, termed pG-U1b, was found to be highly similar to the sequence of human upstream stimulatory factor 1 (hUSF1) but lacked a 5Ј-untranslated region and 18 bp of sequence coding the amino terminus of the protein. This insert was labeled with 32 P and used to rescreen the cDNA library. Another cDNA clone of 1,566 bp in length, which contains 126 bp of 5Ј-untranslated region, the full-length coding region, and 594 bp of 3Ј-untranslated region, was isolated, subcloned into pGEM-7Z plasmid vector (termed pG-rUSF1b), and sequenced.
RT-PCR-Poly(A) ϩ RNA isolated from 28-day fetal rabbit lung tissue was annealed to primer 3 ( Fig. 1B and Table I) at 65°C for 5 min, and the first strand cDNAs were synthesized by RT using a first strand cDNA synthesis kit (Pharmacia). The synthesized first strand cDNAs were amplified using primers 1 and 2 ( Fig. 1B and Table I) by PCR, and plasmid pG-U1b was used as a template in a parallel control reaction. The PCR products were resolved on a 1.8% agarose gel. The DNAs in two bands on the gel were isolated; the DNA isolated from the upper band (termed fragment U1a) was amplified again by PCR. The two isolated RT-PCR products were sequenced by use of CircumVent thermal cycle sequencing kit (New England BioLabs).
For quantitative RT-PCR, a DNA fragment to be used as competitive internal standard (CIS) was generated by PCR amplification of the region from 275 to 545 bp of the plasmid pGEM-7Z (Promega) with a pair of primers that contain the sequences of primers 1 and 4 at their 5Ј-ends, which hybridize to the 303-324 and 932-949 nucleotide sequences in rUSF1a cDNA ( Fig. 1A and Table I), respectively. A first strand cDNA synthesis kit was used with 50 g of total RNA isolated from lung tissues of fetal rabbits of 21, 23, 25, and 28 days gestation as templates for reverse transcription of rUSF1 cDNAs using the primer 5 ( Fig. 1B and Table I), which hybridizes to the 967-987 nucleotide region of rUSF1b mRNA. The cDNAs were washed four times with TE buffer (pH 8.0) using Microcon-100 (Amicon) to remove free primers that were not incorporated into the cDNAs. The cDNAs (transcribed from 2.5 g of RNA) were then combined with 1 amol (10 Ϫ18 mol) of CIS and used as the templates for PCR, using the 32 P-labeled primers 1 and 4 ( Fig. 1A and Table I). After 25 cycles of amplification, aliquots (40% of products) were resolved on a 1.8% EtBr-agarose gel. The gel was dried, and an autoradiogram was generated and scanned using a 300-A computing laser densitometer. To monitor RT efficiency, 0.2 g of oligo(dT) primer (provided by Pharmacia) and 10 Ci of [␣-32 P]dCTP were used in the parallel reaction. The rate of [␣-32 P]dCTP incorporation was determined by measurement of radioactivity adsorbed to DE81 filters (9).
Construction of Plasmids (Table II)-To construct plasmid pG-rUSF1a, the RT-PCR product U1a containing the region of rUSF1a spliced out of rUSF1b was digested with HincII and Eco47III, gel purified, and used to replace the region between HincII and Eco47III of USF1b in pG-rUSF1b (Fig. 1). To construct the expression plasmid pQE-U1b (2-161), the region corresponding to amino acids 2-161 in pG-rUSF1b was amplified by PCR, and BamHI and SalI sites were introduced at the 5Ј-and 3Ј-ends, respectively. The PCR-amplified fragment cut with BamHI and SalI was subcloned into pQE-30 (Qiagen).
For in vivo binding studies using a yeast system, two cDNA fragments of rUSF1b were linked either to the Gal4 DNA binding domain (in yeast expression vector pGBT9 from CLONTECH), referred to as GalDBD or the Gal4 activation domain (in yeast expression vector pGAD424 from CLONTECH), referred to as GalAD. A cDNA fragment, termed U1b bHLH-LZ , which encodes the region from the bHLH-leucine zipper region to the stop codon (amino acid residues 127-282), was amplified from pG-U1b by PCR; EcoRI and BamHI sites were introduced at the 5Ј-and 3Ј-ends, respectively. The PCR-amplified fragment cut with EcoRI and BamHI was subcloned into the pGBT9 and pGAD424 to produce fusion constructs with sequences encoding either the Gal4 DNA binding or the transcription activation domains, pDBD-U1b bHLH-LZ and pAD-U1b bHLH-LZ , respectively. The other PCR-generated cDNA fragment, termed U1b full , which encodes the region from amino acid 7 to the stop codon (amino acid residues 7-282) (Fig. 1A), was used to construct pDBD-U1b full and pAD-U1b full . The yeast expression vector pGAL4, which encodes full-length GAL4 protein, was obtained from CLONTECH in the Matchmaker Two-hybrid System. To construct pTS381, an EcoRI/HindIII fragment of the rabbit SP-A 5Јflanking sequence from Ϫ381 to Ϫ49 bp (5) was ligated into the EcoRI FIG. 1. Rabbit lung contains two isoforms of USF1. Panel A, schematic diagrams of coding regions of USF1 cDNAs. Shown are hUSF1, U1b, isolated by screening the gt11 rabbit lung cDNA library with 32 P-labeled PBE, full-length rUSF1, both of which lack 28 amino acid residues corresponding to transactivation domain (TA) II of hUSF1 (indicated by the bent lines), as well as another isoform of USF1 in rabbit lung, rUSF1a, obtained by RT-PCR, which contains the 28-amino acid residue TA II present in hUSF1. rUSF1a and U1b cDNAs lack sequences encoding amino acid residues 1-6. The numbers above the lines indicate numbers of the deduced amino acid residues, and those below the lines indicate nucleotide numbers of the cDNA with the translation start site of rUSF1b numbered as 1. The regions to which RT-PCR primers 1, 4, and 5 anneal are indicated by arrows below the line representing rUSF1a cDNA. H, HincII site; E, Eco47III site; B, BglII site. Primers used in RT-PCR are indicated. Primer 3 was used to synthesize the first strand cDNA from day 28 fetal rabbit lung RNA by RT. The cDNA was amplified by PCR using two other oligonucleotides, primers 1 and 2. Panel B, RT-PCR products of two isoforms of rUSF1s resolved by agarose gel electrophoresis. Lane 1, rUSF1b cDNA (clone U1b) was used as template in PCR as a control; lane 2, products of RT-PCR of day 28 fetal rabbit lung RNA.
The E-box sequence, CACGTG, of the rabbit SP-A gene (DBE) and adenovirus major late (ADML) promoter is shown in bold letters.
b The E-box-like sequence, CTCGTG, of the rabbit SP-A gene (PBE) is shown in bold letters.
c The mutated nucleotides in the E-box are shown in italics and underlined.
d An 8-bp palindromic cyclic AMP-responsive element that binds to transcription factor CREB (13) is shown in bold letters.
site of pTH1 (a kind gift from Dr. S. Fields), filled in, and ligated.
For transfection studies, SP-A:hGH fusion genes SP-A Ϫ991 :hGH, SP-A Ϫ976 :hGH, SP-A Ϫ991PBE(Ϫ) :hGH, containing Ϫ991, Ϫ976 (lacking the DBE), and Ϫ991 (containing a mutation in the PBE) bp of 5Ј-flanking DNA from the rabbit SP-A gene linked to the human growth hormone (hGH) structural gene (5,6), and expression plasmids pACSKCMV2 and pCMV-nLac were utilized. An adaptor containing NsiI and EcoRV sites followed by the translation start codon ATG and an EcoRI site was inserted into the EcoRI site of pACSKCMV2 to generate vector pCMV. pCMV-USF1a and pCMV-USF1b were constructed by subcloning EcoRI fragments of pG-rUSF1a and pG-U1b into pCMV, respectively.
pGST-U1b, used for expression of the glutathione S-transferase (GST)-rUSF1b fusion protein in bacteria, was constructed by cloning the EcoRI fragment of pG-U1b into the EcoRI sites of pGEX-1T (Pharmacia).
The orientations and the sequences of linking sites were confirmed by DNA sequencing. In all cases where inserts were generated by PCR, the nucleotide sequences in the resulting plasmid inserts were confirmed by DNA sequencing.
Expression and Purification of GST and His-tag Fusion Proteins-GST and GST-fusion proteins were expressed and purified from Escherichia coli DH-5␣ (Life Technologies, Inc.) as described by Smith and Johnson (10). After binding to glutathione-Sepharose 4B (Pharmacia), the proteins were washed and eluted with reduced glutathione (Sigma). His-tag fusion proteins were expressed and purified from E. coli M15[pREP4] (Qiagen) according to the manufacturer's instructions. After binding to Ni-NTA-resin (Qiagen), the proteins were washed and eluted with imidazole (Sigma). The concentrations of the expressed proteins were determined by the method of Bradford (Bio-Rad). The purity and size of the eluted proteins were then evaluated by Coomassie staining of sodium dodecyl sulfate-polyacrylamide gels.
Preparation of USF1 Antibodies-Polyclonal antibodies to rUSF1 were generated by immunizing guinea pigs as described by Harlow and Lane (11) with His-tagged USF1b (2-161). The IgGs were purified by use of an ImmunoPure (A) IgG purification kit (Pierce) according to the manual. This IgG preparation recognized both expressed rUSF1a and rUSF1b by immunoblot analysis.
Electrophoretic Mobility Shift Assay (EMSA)-Rabbit lung nuclear proteins were prepared as described by Gorski et al. (12). Type II pneumonocytes and fibroblasts were prepared from cultured fetal rabbit lung explants (6). Binding reactions and gel electrophoresis all were performed as described previously (5). Double-stranded oligonucleotides corresponding to the DBE, PBE, and CRE (13) ( Table I) were end labeled using polynucleotide kinase and [␥-32 P]ATP. The nuclear extracts (10 g), bacterially expressed GST (240 ng) or GST-rUSF1b fusion proteins (75 ng) were incubated at room temperature for 20 min in binding buffer (20 mM HEPES (pH 7.6), 150 mM KCl, 0.2 mM EDTA, 20% glycerol) with radiolabeled DNA probe and poly(dI⅐dC)-poly(dI⅐dC) (Pharmacia) as nonspecific competitor. For antibody supershift analysis, this was followed by the addition of guinea pig preimmune IgG or the IgG against rUSF1s (13 ng in 1 l) and an additional 20-min incubation. The DNA-protein complexes were resolved on a 5% native polyacrylamide gel and visualized by autoradiography.
In Vivo Binding Analysis in Yeast-The reporter gene-containing plasmid pTS381 was linearized by digestion with PvuII in the LYS2 region and transformed into yeast w303-1a (a kind gift from Dr. S. Fields). The pTS381-integrated strain, termed S381, was selected by growing in medim lacking uracil and confirmed by Southern blotting. S381 was transformed with pDBD-U1b bHLH-LZ (expresses GalDBD-U1b bHLH-LZ ), pDBD-U1b full (expresses GalDBD-U1b full ), pAD-U1b bHLH-LZ (expresses GalAD-U1b bHLH-LZ ), and pAD-U1b full (expresses GalAD-U1b full ), as well as with control plasmids pGBT9 (expresses GalDBD), pGAD424 (expresses GalAD), and pGAL4 (expresses fulllength Gal4 protein). A single colony was grown at 30°C for 16 h in SD medium lacking tryptophan (ϪTrp) for the transformants of pGBT9derived plasmids, or in SD medium lacking leucine (ϪLeu) for the transformants of pGAD424-derived plasmids (14). The cultures were washed three times with water and diluted to 0.5 O.D. 600 /ml. Depending upon which of the transformed plasmids were being analyzed, 4 l of diluted yeast was applied either to plates lacking histidine (ϪHis) and tryptophan (ϪTrp) or ϪHis and ϪLeu SD/3-aminotriazole (5 mM) plates as well as to ϪTrp or ϪLeu SD plates. The plates were incubated Homodimerization of rUSF1 in a Yeast Two-hybrid System-To study the interaction of rUSF1 in vivo, the Gal4 DBD-USF1 fusion plasmids pDBD-U1b bHLH-LZ and pDBD-U1b full , the Gal4 activation domain-USF1 fusion plasmids pAD-U1b bHLH-LZ and pAD-U1b full , as well as control plasmids pGBT9, pGAD424, and pGAL4 were transformed into yeast HF7c (provided in the Matchmaker two-hybrid system kit, CLONTECH) individually or cotransformed in different combinations. The transformants were grown in appropriate medium and studied as described above.
Transient Transfection of USF1 Expression Vectors and Reporter Gene Constructs-A549 cells (ATCC CCL 185) (15) were plated at a density of 5-9 ϫ 10 6 cells/60-mm dish 1 day before transfection. The cells were maintained overnight in Waymouth's MB752/1 medium containing fetal calf serum (10%, v/v). The cells were then washed three times with Hanks' balance salt solution (Life Technologies, Inc.) and cotransfected with a number of SP-A:hGH reporter gene constructs, USF1 expression vectors, and control plasmids by incubation with 11 g of each DNA fragment and 44 g of DOTAP (Boehringer Mannheim) in Waymouth's MB752/1 for 18 h. The medium was then aspirated and replaced with fresh Waymouth's MB752/1 at 24-h intervals. Two days later, the media and cells were collected. The concentrations of hGH in the media were analyzed by radioimmunoassay using an Allegro hGH kit (Nichols Institute Diagnostics, San Juan Capistrano, CA). ␤-Galactosidase activity was determined as described by Miller (16).

RESULTS
Isolation of the Rabbit USF1 cDNA Clones-To isolate cDNA clones encoding proteins that bind to E-box motifs of the DBE and PBE in the 5Ј-flanking region of the rabbit SP-A gene, a gt11 cDNA expression library constructed from mRNA isolated from 24-day fetal rabbit lung tissue was screened using a 32 P-labeled double-stranded oligonucleotide corresponding to the PBE. From ϳ2 million recombinant phage clones, a specific cDNA clone encoding a protein that bound specifically to the radiolabeled DBE and PBE (Table I) but not to nonspecific DNA C2 (5) was isolated. The cDNA insert was subcloned into the pGEM-7Z plasmid vector and sequenced. The nucleic acid sequence of the 950-bp cDNA insert, termed pG-U1b, was found to be highly similar to the sequence of hUSF1. This clone contained an 828-bp open reading frame and 113 bp of 3Јuntranslated DNA but lacked 5Ј-untranslated region and 18 bp of coding sequence at the amino terminus of the protein. To isolate a full-length cDNA, the insert in pG-U1b was used to rescreen the cDNA library, and a 1,566-bp cDNA insert was isolated which contained a full-length cDNA for rUSF1b. This clone contained 126 bp of 5Ј-untranslated region, an 846-bp open reading frame, and 594 bp of 3Ј-untranslated region. A comparison of the sequences of hUSF1 and rUSF1b revealed that only two amino acid residues of rUSF1b differ from the corresponding sequence of hUSF1; however, rUSF1b lacks 28 amino acid residues in transactivation domain II of hUSF1 (17). This region is indicated by the angled line in Fig. 1A.
Rabbit Tissues Contain Two Isoforms of USF1 mRNA-To determine whether there is an isoform of USF1 in rabbit lung which contains the 28 amino acid residues that are lacking in rUSF1b, we designed an oligonucleotide primer (primer 3) which anneals to a region 100 bp downstream of the deleted region in the rUSF1b (Fig. 1A and Table I). This was used to synthesize, by reverse transcription, the first strand cDNA from 28-day fetal rabbit lung RNA. The cDNA was amplified by PCR using two other oligonucleotide primers, 1 and 2 ( Fig. 1A and Table I). As shown in lane 2 of Fig. 1B, two RT-PCR products were obtained. The major product corresponded in length to rUSF1b (lower band, 154 bp), whereas the other corresponded in length to hUSF1 (upper band, 238 bp, termed rUSF1a). Sequence analysis of the RT-PCR-amplified cDNA revealed that rUSF1a is 100% identical to rUSF1b except that rUSF1a contains the sequence encoding 28 amino acid residues absent in rUSF1b. This 28-amino acid sequence, deduced from the nucleotide sequence of the rUSF1a, is identical to the corresponding region in hUSF1. Interestingly, in RNA isolated from day 28 fetal rabbit lung, rUSF1a was found to be present at much lower levels than rUSF1b (Fig. 1B). The two isoforms of rUSF1 also were found in other tissues by RT-PCR (below). USF1 Comprises One Component of the Complex of Proteins Bound to the DBE and PBE-To obtain expressed rUSF1 proteins for analysis of their properties, the rUSF1b cDNA insert was subcloned into the GST fusion vector pGEX-1T (Pharmacia) for expression and purification of GST-USF1b fusion protein. The GST-rUSF1b fusion proteins expressed in E. coli were purified using glutathione-agarose beads and used to analyze binding activity for the DBE and PBE by EMSA. As shown in Fig. 2, 28-day fetal rabbit lung nuclear proteins (lanes 6, 9, and 12) and GST-rUSF1b (lanes 5, 8, and 11) bound to the radiolabeled DBE (upper panel) and PBE (lower panel). When in vitro translated rUSF1b, which does not contain GST, was used in EMSA, the same mobility shift complexes were observed (data not shown). Although GST-rUSF1b fusion protein is about 26 kDa larger than rUSF1b, its binding complexes with the DBE or PBE displayed an electrophoretic mobility similar to that of native rUSF1-containing complexes (lanes 5 versus 6), suggesting that folding and/or net charge of rUSF1 plays a more important role in mobility in native polyacrylamide gels than molecular weight. The binding complexes of the DBE and PBE with fetal rabbit lung nuclear proteins were partially supershifted by the addition of anti-rUSF1 IgG (lane 12), indicating that one of the nuclear proteins that bound to the DBE and PBE is USF1. The anti-rUSF1 IgG completely supershifted the complexes of expressed GST-rUSF1b bound to the DBE and PBE (lane 11). Although the sequence of the PBE (CTCGTG) differs by one nucleotide from the DBE E-box (CACGTG), which contains a sequence known to bind to hUSF1 (18), the PBE has binding activity similar to the DBE for expressed rUSF1b (Fig. 3) and rabbit lung nuclear proteins (5). The PBE and the DBE compete for binding to lung nuclear proteins (5). This may be explained by the finding that the nucleotide residues C and G are the most important nucleotides for binding (18). To analyze the importance of the nucleotide sequences of the DBE and PBE for rabbit lung nuclear protein binding, a series of oligonucleotides containing one (m1) or two (m2) point mutations at the known contact points (Table I) were generated and utilized in competition EMSA using the wild type DBE and PBE sequences as radiolabeled probes. As shown in Fig. 3, the binding of radiolabeled DBE either to bacterially expressed GST-rUSF1b or to lung nuclear proteins was effectively competed by a 500-fold molar excess of nonradiolabeled DBE or of a 24-bp oligonucleotide from the adenovirus major late promoter (ADML) (19) which contains the E-box core sequence, CACGTG, but a different flanking sequence (Table I). The ability to compete with DBE for binding to GST-rUSF1b and nuclear proteins was abolished when the 5Ј C in the E-box core sequence of the DBE was mutated to T (D-m1). A nonradiolabeled oligonucleotide in which the central C and G of the E-box core sequence of the DBE were mutated to T and A (D-m2, leaving the E-box consensus sequence, CANNTG, intact) also was an ineffective competitor of the radiolabeled DBE at a 500-fold molar excess (Fig. 3). When the DBE mutants were used in EMSA as radiolabeled probes, only D-m2, but not D-m1, bound GST-rUSF1b and lung nuclear proteins but with a very low activity (data not shown). The same results were obtained when the competition experiment was performed using radiolabeled PBE as probe and PBE mutants as competitors (Fig. 3).
The SP-A gene is expressed predominantly in type II pneumonocytes and to a lesser extent in Clara cells of rabbit lung tissue (2). Binding activity of rabbit fetal lung nuclear proteins for the DBE and PBE is enriched in type II cells compared with whole lung tissue (5). To evaluate the possible role of rUSF1 in the type II cell-specific regulation of SP-A gene expression, the binding activity of rUSF1 in nuclear proteins from an enriched population of the type II cells was compared with that of nuclear proteins from an enriched population of fibroblasts isolated from the same collagenase-digested cell suspension of cultured fetal rabbit lung tissue using an electrophoretic mobility supershift assay and radiolabeled DBE and PBE. As can be seen in Fig. 4, binding activity for radiolabeled DBE and PBE was greatly enriched in type II cells (lane T II ) compared with lung fibroblasts (lane Fb). The binding complexes of the DBE and PBE with fetal rabbit lung nuclear proteins were partially supershifted by the addition of anti-rUSF1 IgG, indicating that one of the type II cell nuclear proteins that bound to the DBE and PBE was USF1. To evaluate the integrity of the fibroblast and type II cell nuclear extracts, we compared their abilities to bind a radiolabeled double-stranded oligonucleotide containing the CRE sequence, TGACGTCA, derived from the rat somatostatin gene (Table I), which binds to the ubiquitous transcription factor CREB (13). Binding activities of nuclear proteins from fibroblast and type II cells for the CRE were similar (data not shown), indicating that these two preparations contained equivalent amounts of nuclear protein binding activity.
USF1 Homodimerizes in Vivo-It has been found that hUSF1 binds to the E-box motif as a dimer and that both helix-loop-helix and leucine zipper domains are required for dimerization (17,20,21). By studying the interaction of 35 Slabeled rUSF1b and bacterially expressed GST-USF1b fusion proteins in the presence or absence of the DBE and PBE in

FIG. 3. Analysis of the effects of mutation of the DBE and PBE on binding of GST-rUSF1b or 28-day fetal rabbit lung nuclear proteins using competitive EMSA. Radiolabeled DBE (left panels)
or PBE (right panels) was incubated for 30 min with bacterially expressed GST-rUSF1b fusion protein (upper panels) and with nuclear proteins isolated from lung tissues of 28-day fetal rabbits (lower panels) in the absence (Ϫ) or presence of double-stranded oligonucleotides containing DBE or PBE without or with one (m1) or two (m2) mutations in the E-box core sequence or ADML as competitors at 500-fold molar excess of the radiolabeled probe. The mutations in the E-box core sequence are shown in bold italics.

FIG. 4. rUSF1 binding activity for the DBE and PBE is enriched in lung type II cells. Radiolabeled DBE (upper panel) or PBE
(lower panel) was incubated with nuclear proteins isolated from fetal rabbit lung fibroblasts (Fb) or type II cells (T II ) for 20 min followed by incubation for 20 min in the absence or presence of anti-rUSF1 IgG, followed by electrophoresis. F, free probe; C, DNA-protein complexes; S, antibody supershift of DNA-protein complexes with anti-USF1 IgG.
vitro, we found that rabbit USF1 also can form homodimers in solution before binding to the specific E-box elements and that the binding of rUSF1 to the E-box does not influence dimerization of USF1 (data not shown). The yeast two-hybrid system (22) was used to determine whether homodimerization of rUSF1 occurs in vivo and whether the presence of the aminoterminal region compared with bHLH-LZ alone influences dimerization. As shown in Fig. 5, both the bHLH-LZ region of rUSF1 (U1b bHLH-LZ ) and full-length rUSF1b (U1b full ) homodimerized and transactivated expression of the His3 gene in yeast HF7c. The efficiency of the bHLH-LZ region for interaction was only one-third of that of the full-length rUSF1b. This could possibly be the result of a fusion protein artifact, since the bHLH-LZ region contains all essential domains for DNA binding and homodimerization (21).
The Levels of rUSF1a and rUSF1b mRNA in Rabbit Lung Tissue Change during Fetal and Postnatal Development-SP-A gene expression is developmentally regulated in fetal lung tissue (4). In rabbits, SP-A gene transcriptional activity is first detectable on day 24 and reaches maximal levels by day 28 of the 31-day gestation period (24). To analyze developmental changes in the levels of rUSF1 mRNAs, quantitative RT-PCR experiments were performed. First, we developed a CIS, a nonhomologous DNA fragment from the plasmid pGEM-7Z engineered to contain the rUSF1 primer templates (the regions corresponding to nucleotides 303-324 and 932-949 in rUSF1a cDNA, Fig. 1A) at 5Ј-and 3Ј-ends, respectively. The CIS was of a size to generate a CIS PCR product of 311 bp in length. 50 g of total RNA isolated from day 28 fetal rabbit lung was used as template for reverse transcription of rUSF mRNAs using the primer 5 that hybridizes to nucleotides 967-987 of rUSF1a mRNA (Fig. 1A). The cDNAs that were generated were then used as the templates for PCR, using the 32 P-labeled primers 1 and 4 that hybridize to nucleotides 304 -323 and 932-949 of rUSF1 cDNA (Fig. 1A). These primers flank the segment encoding the 28 amino acid residues that are deleted in rUSF1b; therefore, the cDNAs specific for both rUSF1a and rUSF1b can be amplified as 557-bp and 641-bp products, respectively. An aliquot of cDNAs corresponding to reverse transcripts from 2.5 g of RNA were used in each reaction in the presence of varying amounts of CIS. After 25 cycles of amplification, PCR product accumulation remained exponential so that product concentration was proportional to starting cDNAs. Aliquots of products (40% of total) were resolved on an agarose gel and visualized by autoradiography (upper panel of Fig. 6A). The majority of rUSF1 RT-PCR products was comprised of rUSF1b (557 bp), whereas rUSF1a (641 bp) constituted only a minor component. Under competitive conditions, the amount of rUSF cDNA synthesized is equivalent to the amount of CIS added to the reaction when the molar ratio of products ϭ 1. The point at which CIS and rUSF1b products were equal was reached at 2.5 ϫ 10 Ϫ19 mol of CIS (lower panel of Fig. 6B), suggesting that there are about 10 Ϫ19 mol of rUSF1b mRNA/g of total RNA in 28-day fetal rabbit lung.
To analyze developmental changes in the levels of rUSF1 mRNAs, aliquots of total RNA isolated from lung tissues of 21-28-day fetal rabbits and from neonates were used as templates to synthesize first strand cDNAs of rUSF1 by reverse transcription as described above. The efficiency of reverse transcription was monitored by use of oligo(dT) primer and [␣-32 P]dCTP in a parallel reaction. The percentage of [␣-32 P]dCTP incorporation was 1.22 Ϯ 0.17 (mean Ϯ S.D.). PCR was carried out using the cDNA transcribed from 2.5 g of RNA in the presence of 1 amol (10 Ϫ18 mole) of CIS. As shown in the autoradiogram (upper panel of Fig. 6B), the majority of rUSF1 RT-PCR products was comprised of rUSF1b (557 bp), whereas rUSF1a (641 bp) consisted of only a minor component at all developmental ages studied. The levels of rUSF1 mRNAs reached a peak at 23 days (lower panel of Fig. 6B), which is just prior to the time that initiation of SP-A gene transcription can be detected in nuclei from fetal rabbit lung (3). Interestingly, the ratio of rUSF1a to rUSF1b also reached a maximal level at this time.
mRNA Encoding rUSF1s Are Expressed Ubiquitously in Fetal Rabbit Tissues-USF1 is a ubiquitously expressed transcription factor (23). To determine whether USF1 exists in two isoforms in different tissues, aliquots of total RNA isolated from heart, skeletal muscle, liver, kidney, and brain of 23-day fetal rabbits were analyzed by RT-PCR using the same USF1 primers as in the quantitative RT-PCR experiments. As shown in Fig. 7, the same two RT-PCR products that were amplified from lung were isolated using RNA from all of these tissues; the major band corresponds in length to rUSF1b (557 bp), whereas the other corresponds in length to rUSF1a (641 bp). The levels of rUSF1a mRNA appeared to be highest in muscle and liver and lowest in kidney and brain. On the other hand, rUSF1b signals appeared similar in all tissues examined. This finding suggests that although expression of the USF1 gene is equivalent in these tissues, alternative splicing of the mRNA may be regulated by tissue-specific factors. Alternatively, the apparent equivalence of the USF1b signal could be the result of its  (Table II) were transformed alone or in the combinations indicated into yeast HF7c. The yeast transformants were grown on a ϪHis plate containing 5 mM 3-aminotriazole for 20 h. Panel B, densities at 600 nm were determined as described under "Experimental Procedures." Data are mean Ϯ S.E. of three determinations from one of two similar independent experiments. greater abundance, causing the number of PCR cycles used to exceed the linear phase of the reaction.

USF1 Interaction with an E-box Element within 5Ј-Flanking Region of the SP-A Gene in Yeast Results in Activation of Yeast
Gal1 Promoter Activity-To determine whether the binding of rUSF1 to the PBE E-box motif transactivates gene expression in vivo, we used a yeast one-hybrid system. Plasmids pAD-U1b full and pAD-U1b bHLH-LZ , which express GalAD-U1b full and GalAD-U1b bHLH-LZ , respectively, were transformed into the yeast S381, in which Ϫ381 to Ϫ49 bp of 5Ј-flanking region of the rabbit SP-A gene linked to a minimal Gal1 promoter-controlled His3 gene was incorporated into the yeast genome. The Ϫ381 to Ϫ49 bp 5Ј-flanking region contains the functional PBE, and this region has been shown to mediate basal and cAMPinducible activation of SP-A promoter activity in type II cells (5). Yeast expression plasmids pDBD-U1b full and pDBD-U1b bHLH-LZ , which express GalDBD-U1b full and GalDBD-U1 bHLH-LZ , respectively, as well as the control plasmids pGAD424, pGBT9, and pGAL4, which express GalAD, GalDBD, and full-length Gal4 protein, respectively, also were transformed into yeast S381. As shown in Fig. 8, both full-length USF1b and the bHLH-LZ region of USF1 linked to the Gal4 activation domain (GalAD-U1b full and GalAD-U1b bHLH-LZ , respectively) activated expression of the His3 gene in yeast S381, resulting in its ability to grow in medium lacking histidine (ϪHis). By contrast, neither GalDBD, GalAD, full-length Gal4 protein (GAL4), nor GalDBD linked either to full-length USF1b (GalDBD-U1b full ) or to the USF1b bHLH-LZ region (GalDBD-U1b bHLH-LZ ) activated expression of the His3 gene in yeast S381. These findings suggest that the rUSF1 bHLH domain functionally interacted with the E-box element within the 5Јflanking region of rabbit SP-A gene so that the fused Gal4 activation domain activates Gal1 promoter activity and expres-sion of the His3 gene. No apparent difference in efficiency for activation of His3 gene expression was observed between the full-length and bHLH-LZ region of rUSF1 (Fig. 8), although full-length USF1 exhibited a greater ability to form homodimers than the bHLH domain (Fig. 5).
USF1 Activates Expression of SP-A:hGH Fusion Genes in A549 Cells-In previous studies, we found that SP-A:hGH fusion genes containing Ϫ991 bp of SP-A 5Ј-flanking sequence (SP-A Ϫ991 :hGH) linked to the hGH structural gene were expressed in primary cultures of type II cells; expression of SP-FIG. 7. rUSF1a and rUSF1b mRNAs are expressed in different tissues of 23-day gestational fetal rabbits. Aliquots of total RNA isolated from different tissues of fetal rabbits of 23 days gestational age were used as templates for reverse transcription of rUSF cDNAs using the primer described under "Experimental Procedures." The cDNAs were amplified by PCR in the presence of CIS template at 10 Ϫ19 mol/g of RNA, a nonhomologous DNA fragment containing the rUSF1 primer templates which yields a PCR product of 311 bp in length. Aliquots of PCR products were resolved on an agarose gel. The gel was dried, and an autoradiogram was generated.
FIG. 6. The levels of rUSF1a and rUSF1b mRNAs in rabbit lung tissue change during fetal and postnatal development. Panel A, total RNA isolated from lung tissues of fetal rabbits of 28 days gestational age was used as template for reverse transcription of rUSF cDNAs using the primer described under "Experimental Procedures." The cDNAs were amplified by PCR in the presence of different amounts of CIS, a nonhomologous DNA fragment containing the rUSF1 primer templates which yields a PCR product of 311 bp in length. Aliquots of PCR products were resolved on an agarose gel. The gel was dried, and an autoradiogram was generated (upper panel) and scanned using a computing laser densitometer. Shown in the lower panel is the scan of rUSF1b and CIS arbitrary units. Panel B, total RNAs isolated from lung tissues of fetal rabbits of 21-28 days gestational age and from neonates (N) were used as templates for reverse transcription of rUSF1 cDNAs. The cDNAs were amplified by PCR in the presence of CIS template at 10 Ϫ19 mol/g of RNA. Aliquots of the PCR products were resolved on an agarose gel. The gel was dried, and an autoradiogram (upper panel) was generated and scanned using a computing laser densitometer (lower panel). C, control in which the RNA was subjected to PCR without prior incubation with reverse transcriptase. Data shown are from one of five similar experiments.
A Ϫ991 :hGH was stimulated upon incubation with Bt 2 cAMP. Mutagenesis of the DBE or PBE in the context of the 991 bp of 5Ј-flanking region resulted in a marked reduction of basal and cyclic AMP-induced fusion gene expression (5). To determine whether rUSF1a and rUSF1b regulate expression of the SP-A gene via the DBE and PBE, expression vector plasmids pCMV-USF1a and pCMV-USF1b were cotransfected into A549 cells with the SP-A Ϫ991 :hGH fusion gene, with SP-A Ϫ976 :hGH (which lacks the DBE) or with SP-A Ϫ991PBE(Ϫ) :hGH (containing a mutation in the PBE) (5). A549 is a lung adenocarcinomaderived cell line presumed to be of type II cell origin; however, this cell line does not produce SP-A mRNA and protein at detectable levels (15). As shown by the open bars in Fig. 9, when A549 cells were cotransfected with the intact SP-A Ϫ991 : hGH fusion gene and with pCMV-USF1a or pCMV-USF1b, a 3.1-and 2.4-fold induction, respectively, of hGH expression was found compared with that observed upon cotransfection of the SP-A Ϫ991 :hGH fusion gene with "empty" expression vector plasmid, pCMV, or calf thymus DNA shown as control. By contrast, when A549 cells were cotransfected with pCMV-USF1a or pCMV-USF1b and the SP-A Ϫ976 :hGH fusion gene, in which the DBE is deleted (hatched bars), no induction of hGH expression was found compared with that observed upon cotransfection with the empty vector plasmid pCMV or calf thymus DNA (Fig. 9). Similarly, no induction of hGH expression was detected when A549 cells were cotransfected with SP-A Ϫ991PBE(Ϫ) :hGH fusion genes and pCMV-USF1a or pCMV-USF1b. When A549 cells were cotransfected with SP-A Ϫ991PBE(Ϫ) :hGH fusion genes and calf thymus DNA or pCMV, hGH expression also was below detectable levels (Fig. 9). These findings suggest that rUSF1s bind to the E-box motifs in the 5Ј-flanking region of the rabbit SP-A gene and induce SP-A promoter activity. DISCUSSION Previously, we identified two structurally similar sequences within the 5Ј-flanking region of the rabbit SP-A gene, termed DBE and PBE (5). Deletion or mutation of these elements reduces both basal and cAMP-induced expression of SP-A:hGH fusion genes transfected into type II cells. The DBE contains the E-box core sequence CACGTG, which is known to interact with a number of gene regulatory proteins that contain a bHLH structure. The PBE contains the sequence CTCGTG, which differs by one nucleotide from the DBE (5). Because the PBE and the DBE compete for binding to lung nuclear proteins (5) we considered it likely that similar or identical proteins bind to these elements. In the present study, we screened a fetal rabbit lung cDNA expression library using the radiolabeled PBE as probe. A cDNA insert was isolated which encodes a protein that specifically binds to the DBE and PBE elements. Sequence analysis of the cDNA insert revealed that this protein is the rabbit homolog of hUSF1 (18,19,(25)(26)(27)(28)(29), a heat-stable protein, M r Х43,000, which binds to the E-box element CACGTG (26). Previously, we observed that heat-stable type II cell nuclear proteins of M r Ϸ 69,000, 45,000, and 22,000 bind both to the DBE and PBE; the 45-kDa protein(s) appears to be the predominant species and binds as a dimer (5). In the present study, we observed that anti-rUSF1 IgG supershifted a major portion of the binding complexes formed upon incubation of nuclear proteins from rabbit lung type II cells with the radiolabeled DBE and PBE. These findings suggest that USF1 comprises a principal component of the complex of lung type II cell nuclear proteins that bind to these elements.
The cDNA isolated in the screen of the rabbit lung cDNA library encodes rUSF1b, a human USF1 (hUSF1) homolog that lacks 28 amino acids in the second transactivation domain. By use of RT-PCR, we also identified in all tissues analyzed another mRNA species for USF1, which we termed rUSF1a, encoding the full-length 310-amino acid protein; however, we observed that rUSF1b is present at considerably higher levels than rUSF1a in all fetal rabbit tissues studied. Three forms of hUSF1 cDNA were characterized previously by Gregor et al. (26), termed z18, z32, and z6b. Only z6b encodes the full-length USF1 protein of 310 amino acids; z18 and z32 encode truncated 244-and 49-amino acid isoforms, respectively. Interestingly, none of these forms of hUSF1 cDNAs corresponds to rUSF1b,  Table II were transformed into yeast S381 in which the 5Ј-flanking region of the rabbit SP-A gene (Ϫ381 to Ϫ49 bp) linked to a minimal Gal1 promoter-controlled His3 gene had been incorporated into the yeast genome. The yeast transformants were grown on a ϪHis plate for 20 h. Growth was observed only in yeast transformed with GalAD-U1b bHLH-LZ and GalAD-U1b full . Densities at 600 nm were determined as described under "Experimental Procedures." Data are the mean Ϯ S.E. of three determinations from one of two similar independent experiments. and all contain the sequence encoding the 28 amino acids which is lacking in rUSF1b. In the mouse USF1 gene sequence, the 84 bp of DNA encoding these 28 amino acids are located at the 3Ј-end of exon 6 (30). An alternative splice site present within exon 6 can be used to splice out the 84 nucleotides and create a donor site that can link to the acceptor site at the 5Ј-end of exon 7 without causing a shift in the reading frame. The 84-nucleotide region in rUSF1a cDNA which is deleted in rUSF1b is bracketed by the splice junction consensus nucleotides AG:GT.
Although USF1 is a ubiquitously expressed transcription factor (23), it has been suggested to be involved in the regulation of tissue-specific expression of a growing number of genes (31)(32)(33)(34)(35)(36)(37)(38)(39)(40)(41). USF1 also is suggested to be involved in developmental regulation of Xenopus transcription factor IIIA gene expression (42). By use of RT-PCR, we found the same two species of rUSF1 mRNAs in all tissues, corresponding in size to rUSF1a and rUSF1b. In all tissues studied, rUSF1b was the predominant species. Interestingly, rUSF1 binding activity was found to be present at high levels in type II epithelial cells but was barely detectable in fibroblasts isolated from fetal lung, suggesting a lung cell-specific function of USF1. Considering that USF1s contain two protein interaction domains (helix-loophelix and leucine zipper), it is possible that USF1s might interact with different tissue-specific factors to activate expression of different genes in different tissues.
Transcriptional activity of the SP-A gene is developmentally regulated in fetal rabbit lung and is first detectable in lung nuclei on day 24, reaching maximal levels by day 28 of the 31-day gestational period (3). By use of quantitative RT-PCR, we observed that rUSF1a and rUSF1b mRNAs reach peak levels on day 23 of gestation just before the time of activation of SP-A gene transcription (3,23). Interestingly, the ratio of rUSF1a to rUSF1b also reaches maximal levels at this time. In previous studies, we observed that binding activity for the DBE and PBE of fetal rabbit lung nuclear proteins also reaches maximal levels on day 24 of gestation (5). These findings suggest that increased expression of rUSF1s and changes in alternative splicing which yield a relative increase in rUSF1a may serve an important role in the initiation of SP-A gene expression in fetal rabbit lung during development. Since the 28amino acid deletion in rUSF1b is located in the second transactivation domain, it is possible that rUSF1a has increased transactivation potential compared with rUSF1b. Whether a change in the ratio of rUSF1a to rUSF1b serves a role to regulate SP-A gene transcription in fetal lung during development is uncertain because we observed that cotransfected rUSF1a and rUSF1b have similar efficiencies in transactivating SP-A:hGH fusion genes in A549 lung adenocarcinoma cells (Fig. 9).
The transactivation domains in USF1 have been reported to be relatively weak (40); therefore, heterodimerization of USF1a and USF1b with other transcription factors may be critical in the regulation of SP-A gene expression. Preliminary findings indicate that rUSF1s interact with a related bHLH-LZ protein, USF2, upon binding to the DBE and PBE. 2 In previous studies, we characterized a CRE-like sequence (6, 7) and a GT-box (43) which are also required for basal and cyclic AMP regulation of SP-A promoter activity in transfected type II cells. Our findings suggest that a member of the nuclear receptor family interacts with the CRE-like sequence (7) and that Sp1 and related factors bind to the GT box (43). It, therefore, is likely that regulation of SP-A expression in lung type II cells is dependent upon the cooperative interaction of USF1s with the CRE-and GT box-binding proteins. Characterization of the USF1-interacting transcription factors and investigation of their regulation may provide further insight into the molecular mechanisms involved in the developmental, type II cell-specific, and cyclic AMP regulation of SP-A gene expression.