An Alternative, Human SRC Promoter and Its Regulation by Hepatic Nuclear Factor-1 a *

The SRC gene encodes the proto-oncogene pp60 c- src , a tyrosine kinase implicated in numerous signal trans-duction pathways. In addition, the SRC gene is differentially expressed, developmentally regulated, and frequently overexpressed in human neoplasia. However, the mechanisms regulating its expression have not been completely explored. Here we describe the isolation of a new distal SRC promoter and associated exon, designated 1 a , which we mapped to a position 1.0 kilobase upstream of the previously described SRC 1A housekeeping promoter. Differential use of these promoters and their associated exons coupled with subsequent splicing to a common downstream exon results in c-Src transcripts with different 5 * ends but identical coding regions. Promoter analysis following transient transfections into HepG2 cells mapped the minimal 1 a promoter to a region 145 bp upstream of the major transcription start site. This region contained a consensus binding site for hepatic nuclear factor-1 (HNF-1), a liver-en-riched transcription factor implicated in the regulation of a number of genes in liver, kidney, stomach, intestine, and pancreas. Subsequent mobility shift assays confirmed that HNF-1 a isoform was the predominant factor interacting with this region of the promoter. Mutation of the HNF-1 site resulted in a dramatic reduction in SRC promoter activity. Cotransfection studies

SRC encodes pp60 c-src , a nonreceptor tyrosine kinase impli-cated in numerous growth receptor-mediated signaling pathways leading to cellular proliferation, motility, and adhesion as well as differentiation and transformation (1,2). More recently SRC and its close family members have been linked to both cell survival and vascular endothelial growth factor-mediated angiogenesis (3). Since its discovery as the homologue of the v-src transforming gene of Rous sarcoma virus, it has long been speculated that c-Src could play an important role in human cancer. Over the years activation of this kinase has been reported in many human neoplasms, especially those of the colon (4 -8) but also breast (9,10), lung (11,12) pancreas (13,14), and liver (15). Most recently, activating mutations in the SRC gene have been identified in a small number of advanced colon cancers (16). However, in many tumors and cell lines, a significant portion of this activation process appears to result from c-Src overexpression (7,8). Interestingly, overexpression of normal c-Src is sufficient to transform mouse fibroblasts (17), especially in combination with receptor tyrosine kinases such as epidermal growth factor receptor (18,19). In addition, antisense-mediated down-regulation of c-Src in the human colon cancer cell line (HCCL) 1 HT29 and the SKOv-3 ovarian cancer cell line, both of which constitutively overexpress c-Src, results in decreased proliferation and tumor forming ability (20,21). These data suggest that overexpression of c-Src can contribute to both transformation and tumorigenicity. Our own laboratory has recently shown that this common overexpression results from transcriptional up-regulation of the SRC gene in many HCCLs. 2 Although the SRC gene product has been the subject of intense scrutiny, SRC transcriptional regulation in both normal and malignant cells has received relatively little attention. Whereas c-Src expression in normal tissues and cells is fairly ubiquitous, expression levels vary tremendously, and the gene is developmentally regulated (1,2). In addition, SRC is induced at the transcriptional level during differentiation of several cell types (22)(23)(24)(25) and during both maturation and activation of osteoclasts, where SRC function is essential (26,27).
The SRC gene is composed of 14 exons; the first three, designated 1A, 1B, and 1C, encode the 5Ј-untranslated region of the mRNA. Exons 2-12 encode the protein and 3Ј noncoding region (28,29). A promoter, with all the hallmarks of a housekeeping gene (including high GC content and multiple start sites), is associated with exon 1A (29). We have shown that this promoter is regulated by members of the Sp1 family as well as a novel factor, SPy, which interacts with several unusual polypurine:polypyrimidine tracts that are essential for full SRC promoter activity (30). However, during the course of our studies with HCCLs, we noted that many c-Src transcripts did not originate from this promoter. We report here the isolation and characterization of an alternative SRC promoter that is active in a number of tumor cell lines, and we demonstrate that the homeodomain-containing transcription factor HNF-1 is responsible for regulating the expression of c-Src in a tissue-restricted fashion.
RNA Isolation and Cloning of 5Ј RACE Products-Total cellular RNA was isolated from semi-confluent tissue culture plates as described (31) and quantitated by A 260 . Concentration and integrity were confirmed by formaldehyde agarose gel electrophoresis. For 5Ј RACE analysis, poly(A) ϩ mRNA was isolated from HT29 cells, using a mRNA purification kit (Amersham Pharmacia Biotech, as described by the manufacturer. A 5Ј RACE kit (Life Technologies, Inc.) was used, essentially as described by the manufacturer, with some modifications. Briefly, 0.5 g of poly(A) ϩ mRNA was reverse transcribed using a SRC exon 2-specific primer (5Ј RACE primer 1; 5Ј-GGTCTGCGAGGGCGGGGAAAGCG) complementary to sequence located 111 bp downstream of the initiating AUG codon. Following RNase treatment and purification, one-fifth of the single-stranded cDNA was dC tailed with terminal deoxynucleotidyl transferase. A fifth of the this reaction was then subjected to amplification using a second SRC exon 2-specific primer (5Ј RACE primer 2; 5Ј-TTGGGCTTGCTCTTGTTGCTAC) complementary to sequence located 26 bp downstream of the AUG codon and the 5Ј RACE abridged anchor primer supplied with the kit (35 cycles of denaturation at 94°C for 45 s, annealing at 58°C for 60 s, and extension at 72°C for 90 s). At this point a clear PCR product was observed following agarose gel electrophoresis. This PCR product (one-tenth volume) was then subjected to a second round of amplification using the 5Ј RACE primer 2, the abridged universal amplification primer supplied with the kit, and 2.5 units of Pfu (Stratagene) (30 cycles of denaturation at 94°C, annealing at 53°C for 60 s, and elongation at 72°C for 90 s). Following phosphorylation with polynucleotide kinase and ATP, the PCR product was gel purified and cloned into the EcoRV site of pBluescript (Stratagene). A number of individual PCR derived inserts were then sequenced using the T7 sequencing kit (Amersham Pharmacia Biotech).
The plasmid pSRC1␣ chimera was prepared by restricting the longest 5Ј RACE product, cloned in pBluescript, with HincII (which cuts within exon 1␣) and EcoRV (located in the multiple cloning site), thus removing the first 46 bp of exon 1␣ sequence. In parallel a EcoRV/ HincII fragment encompassing this 46 bp of exon1␣ and an additional 532 bp of upstream genomic sequence were isolated from pBam 4.8. This fragment was then cloned into the EcoRV/HincII-digested 5Ј RACE plasmid. The resulting pSRC1␣ chimeric construct contained 532 bp of presumptive genomic promoter sequence spliced directly to the three noncoding exons (1␣, 1B, and 1C). The Ϫ2852, Ϫ777, and Ϫ145SRC1␣-CAT3-Basic clones were derived from the genomic pBam 4.8 clone and contained a common 3Ј end defined by an MscI site located in exon 1␣ at ϩ153. The individual fragments were produced by complete digestion of pBam 4.8 with BamHI followed by a partial digestion with MscI. Fragments of the predicted size (3005, 929, and 297 bp, respectively) were gel purified and blunt end ligated into the SmaI site of pCAT3-Basic. The Ϫy375SRC1␣-CAT3-Basic vector was derived from a RsaI/XhoI digest of Ϫ777SRC1␣, and the resulting 533-bp fragment was cloned into the SmaI/XhoI site of pCAT3-Basic. The ϩ19 SRC1␣-CAT3-Basic vector was prepared by isolating the 133-bp HincII/ XhoI fragment from Ϫ145SRC1␣-CAT3-Basic followed by cloning into the SmaI/XhoI site of pCAT3-Basic. The vector, Ϫ777⌬-136SRC1␣-CAT3-Basic was derived from Ϫ777SRC1␣-CAT3-Basic by digesting with StuI and XhoI. Following an end fill reaction with T4 DNA polymerase, the resulting plasmid fragment was gel purified and religated. In all cases the identity of each construct was confirmed by DNA sequence analysis.
CAT reporter constructs containing a mutated version of the HNF-1 site (Ϫ145mutSRC1␣ and Ϫ777mutSRC1␣) were generated from the wild type versions of the vectors using the QuikChange site directed mutagenesis protocol (Stratagene). Sense 5Ј-TGGGAAGCTGCGGTTC-CTCTTGGAGCCAGCTTGCAAAC-3Ј and antisense 5Ј-GTTTGCAAG-GCTGGCTCCAAGAGGAACCGCAGCTTCCCA-3Ј mutagenic oligonucleotides (Life Technologies, Inc.) spanning the region Ϫ62 to Ϫ24 were used for this process. Mutated sequences are underlined. Following confirmation of mutation by sequence analysis the promoter fragments were isolated and recloned into pCAT3-Basic.
Isolation of the 1␣ Genomic Region and Mapping of Transcription Start Sites-The genomic exon 1␣ region was mapped to a position approximately 1.0 kilobase upstream of the previously identified exon 1A and associated promoter. This was confirmed by both our own sequencing of the genomic pBam 4.8 clone (29) and the extensive sequence analysis available for this region of Chromosome 20 performed at the Sanger Institute. Mapping of transcription start sites was carried out using an S1 Nuclease Protection Assay Kit (Ambion) essentially as described by the manufacturer. Briefly, a single-stranded 32 Plabeled probe was prepared by linearizing pSRC1␣ chimera with StuI (which cuts 107 bp upstream of the longest 5Ј RACE clone analyzed) and carrying out a primer extension reaction in the presence of [␣-32 P]dCTP, unlabeled dNTPs, a primer complementary to sequence in exon 1B (5Ј-CTGGCTCTGTCTCTCATAGCTGG) and Taq polymerase. Following isolation of the full-length S1 probe on an acrylamide gel, 50,000 cpm were hybridized to 25 g of HT29 total RNA at 55°C overnight. Following S1 digestion, the products were resolved on a standard DNA sequencing gel. Start sites were mapped by a direct comparison with a sequencing ladder generated with pSRC1␣ chimera and the exon 1B-specific primer. S1 Analysis of Human Tumor Cell Lines-Two S1 probes specific for either 1␣ or 1A containing c-Src transcripts were generated from the plasmids pSRC1␣ chimera and pSRC1A chimera linearized with StuI or SacI respectively. Analysis was performed exactly as described for mapping of transcription start sites (see above) except that an exon 1C-specific primer (5Ј-GAGTCAGGGGTCTCGAAATAGAG) was used to generate probes. Hybridization was carried out at 48 and 60°C overnight for the 1␣ and 1A probes, respectively.
Multi-tissue Blots-A multi-tissue RNA blot was purchased from CLONTECH containing standardized levels of poly(A) ϩ mRNA from various normal human tissues. The blot was first hybridized with a 32 P-labeled SRC cDNA fragment (a NcoI/KpnI fragment encompassing the first 875 bp of the SRC coding region) exactly as described by the manufacturer. The blot was then stripped according to manufacturers instructions and reprobed with a 32 P-labeled 1␣-specific probe (a 200-bp StuI/HpaII genomic fragment encompassing Ϫ135 to ϩ165 of the promoter region).
Electrophoretic Mobility Shift Assays-HepG2 nuclear extracts were prepared according to the method of Andrews and Faller (33). A 298-bp MscI fragment covering the Ϫ145 to ϩ152 region of the SRC1␣ promoter was blunt end ligated into the SmaI site of pBluescript. A 32 Plabeled probe encompassing the region Ϫ145 to ϩ19 was then prepared by digesting this clone with ClaI and HincII followed by an in-fill reaction with Klenow fragment and [ 32 P]dCTP. EMSAs were carried out as described previously (30). Competitor double-stranded oligonucleotide sequences (Life Technologies, Inc.) used in EMSAs were SrcHNF (Ϫ56 to Ϫ35, 5Ј-GCTGCGGTTAATCTTTAAGCA-3Ј), mHNF (5Ј-GCTGCGGTTCCTCTTGGAAGCCA-3Ј) and a consensus HNF-1 sequence (34), cHNF (5Ј-GTGGTTAATNATTAAC-3Ј). For competition experiments, oligonucleotides (25 M excess) were incubated with nuclear extract for 15 min prior to the addition of probe. For antibody experiments 1 or 2 g of Supershift anti-HNF-1␣ or HNF-1␤ (Santa Cruz) were used.
Immunoblot Analysis-Cells from semi-confluent 10-cm plates were lysed directly in a loading buffer containing 65 mM Tris-HCl (pH 7.0), 2% (w/v) SDS, 5% ␤Ϫmercaptoethanol, 10% glycerol, and 0.05% (w/v) bromphenol blue. Following protein concentration determination using a Lowry kit (Sigma), 30 g of total cellular protein was resolved on a 10% SDS-polyacrylamide gel. Gel transfer to nitrocellulose and membrane blocking were performed using standard procedures (7). Blots were first incubated with antibodies specific for HNF-1␣, Sp1, or Sp3 (Santa Cruz) at 2 g/ml, washed, and then subsequently probed with the appropriate secondary antibody conjugated to horseradish peroxidase (Santa Cruz) diluted 1:2000 as described (7). Membranes were incubated in chemiluminescence reagents (PerkinElmer Life Sciences) and exposed to Kodak X-OMAT Blue XB-1 film for detection.
Transfections and CAT Assays-All plasmids used in transfections were isolated using an EndoFree Plasmid Maxi Kit (Qiagen). In a typical transfection 0.5 g of CAT vector and 0.5 g of pCMV␤-gal were mixed with 6.0 l of PLUS reagent and 4 l of LipofectAMINE each in 100 l of OptiMEM (all reagents from Life Technologies, Inc.). For coexpression studies an extra 0.5 g of either pBJ5-HNF-1␣, pBJ5-HNF-1␤, or an empty control vector were included. Following a 30-min incubation at room temperature, the DNA complexes were added directly to cells in serum-free medium (4 ϫ 10 5 cells/well). After 3 h, serum was added to 10% in a final volume of 2 ml. Cells were harvested 48 h later with provided lysis buffer, and protein levels were determined using Bio-Rad Bradford reagent. ␤-Galactosidase activity was determined as described previously (30), and CAT levels were quantified using a CAT ELISA kit (Roche Molecular Biochemicals). CAT levels were corrected for transfection efficiency and were repeated at least three times in duplicate.

Isolation of c-Src 5Ј RACE Clones and Mapping of the SRC1␣
Promoter and Its Associated Exon-As part of our efforts to understand the regulation of SRC in HCCLs, we carried out a series of S1 and RNase protection experiments using singlestranded 32 P-labeled probes complementary to the 5Ј noncoding region of c-Src mRNA (encoded by exons 1A, 1B, and 1C). Surprisingly, the protected fragments observed were frequently much shorter than expected and were consistent with mRNA species, which lacked sequence encoded by exon 1A (results not shown). To test this hypothesis, we carried out 5Ј RACE using mRNA isolated from the HCCL, HT29, as described under "Experimental Procedures." A major species of approximately 500 bp was isolated and cloned, and a number of individual constructs were sequenced (Fig. 1A). All of these clones contained sequence encoded by exons 1B, 1C, and 2 (the first coding exon). However, novel sequence data were obtained 5Ј of exon 1B, indicating the presence of a new exon. None of the clones analyzed contained sequence encoded by exon 1A. Using traditional mapping methods, as well as the extensive sequence data available for this region of the genome (Chromosome 20 project, Sanger Institute, Cambridge, UK), we mapped this new exon to a position approximately 1.0 kilobase upstream of exon 1A and named it exon 1␣. (Figs. 1B and 2A) To map the transcription start site(s) for this new proposed SRC promoter, we constructed a chimeric clone consisting of genomic sequence derived from exon 1␣ spliced in frame to exons 1B, 1C, and 2 derived from the 5Ј RACE product, as described under "Experimental Procedures." Single-stranded 32 P-labeled probes were then synthesized from this construct and used in S1 mapping experiments. A single major transcription start site, as well as several minor sites, were identified (Fig. 2B). The start sites mapped were very similar or identical to those predicted by sequence analysis of the 5Ј RACE clones and were also confirmed by RNase protection (results not shown).
Relative SRC Promoter Usage in Human Tumor Cell Lines-To examine the relative abundance of transcripts arising from promoters 1␣ and 1A, S1 probes were used in protection experiments with RNA isolated from various human cell lines (Fig. 3). Using similar S1 probes to those described above, we found both full-length protection, demonstrating SRC1␣ promoter usage, and a shorter protected species consistent with transcripts arising from 1A (Fig. 3A). These data demonstrate that both promoters are used in HCCLs such as HT29, WiDr, COLO201, COLO205, previously shown to constitutively overexpress c-Src (7). Interestingly, we also found the human hepatocarcinoma cell line HepG2, which also expresses similarly high constitutive levels of c-Src, 3 also utilizes both promoters, although with a strong preference for 1␣. HCCLs, which we have previously shown to express low levels of c-Src (SW620 and SW480), reflected this low expression in protection experiments, although species consistent with usage of both promoters were observed with prolonged exposures. In contrast, protection of RNA derived from T47D (breast cancer), HTB 10 (neuroblastoma), and HTB 18 (retinoblastoma) was consistent with predominant utilization of the 1A promoter. This was confirmed by repeating these experiments with S1 probes synthesized from vector pSRC1A chimera, where the 1␣ portion was replaced with 1A sequence (Fig. 3B). In this case fulllength protection was seen with HTB 18 and HTB 10, confirming that these cell lines preferentially use the 1A promoter.
Expression of 1␣-derived c-Src Transcripts in Normal Human Tissues-To determine whether transcripts originating from this new promoter were expressed in normal tissues, we first examined the expression pattern of c-Src using a multi- FIG. 1. Analysis of an alternative c-Src transcript using 5 RACE and genomic organization of the SRC promoter locus. A, a 500-bp 5Ј RACE fragment (shown with an arrow in the right panel) amplified from HT29 mRNA was cloned, and several individual clones were sequenced. The sequence of the clone with the most extensive unique 5Ј region is shown (left panel). The arrows mark the position of the various splice sites as determined by comparison with genomic sequence. The complementary sequence of the two c-Src-specific primers used in the generation of the 5Ј RACE product are underlined. B, the position of the new 1␣ exon was mapped relative to the original 1A exon as described in the text. Several important restriction sites for subsequent cloning of CAT expression vectors are shown, including the SalI/ NaeI sites, which were used in the construction of the previously described 0.54SRC1A-CAT construct (29). Exon 1B is located some 20 kilobases downstream of the promoter locus. The entire sequence of this promoter region has been deposited in GenBank TM (accession number AF272982). tissue mRNA dot blot (CLONTECH) with a cDNA probe derived from the SRC coding region. As shown in Fig. 4A, c-Src mRNA was ubiquitously expressed, although levels were highly variable. Expression was absent in adult tissues such as skeletal muscle, low in heart, liver, colon, and thymus, and highest in tissues such as testis, stomach, and adrenal gland. In contrast, when a probe specific for transcripts arising from exon 1␣ was used, a very different pattern was observed (Fig.  4B). Here expression was highest in pancreas, stomach, kidney, and fetal lung with lower levels detectable in colon, liver, prostate, and some other fetal tissues such as kidney and liver. Expression of SRC1␣ transcripts was undetectable in most other tissues, including brain, testis, and adrenal gland. These data suggest that although overall c-Src mRNA expression is widespread, transcripts originating from the SRC1␣ promoter display a much more restricted pattern of expression. Unfortunately, because of an extremely high GC content, we have had difficulty constructing a reliable 1A-specific probe to complement these data.
Analysis of SRC 1␣ Promoter Activity Following Transient Transfection-To begin characterization of this new promoter, we constructed a series of 5Ј and 3Ј SRC1␣ promoter deletion CAT expression vectors (using pCAT-3basic, Promega). Be-cause the SRC1␣ promoter appeared to be particularly well utilized in the HepG2 cells, we carried out our initial analysis in this cell line. As shown in Fig. 5, deletion from the 5Ј end resulted in a series of stepwise increases in CAT levels, suggesting the presence of several upstream negative regulatory elements. Highest CAT levels were achieved with the Ϫ145SRC1␣ construct, which was approximately four times more active than the Ϫ2852SRC1␣ construct and produced CAT levels similar to that of the SRC1A promoter construct, 0.54SRC1A-CAT3. However, deletion of sequences between Ϫ145 and ϩ19 completely abolished this activity. Similarly, a 3Ј deletion of the sequence up to Ϫ136 (Ϫ777⌬Ϫ136SRC1␣-CAT3) resulted in the complete loss of promoter activity. Thus, sequences in the Ϫ136 to ϩ19 region were identified as critical for SRC 1␣ promoter function. Examination of this region ( Fig.  2A) revealed several consensus transcription factor binding sites including a weak Sp1 site (GGCGGG, positions Ϫ81 to Ϫ69), NF-IL6 (TKNNGNAAK, positions Ϫ13 to Ϫ5), and a site very similar to HNF-1 (GGTTAATNTTTAAG, positions Ϫ51 to Ϫ38). The presence of a HNF-1 site was particularly intriguing because factors binding to this site are known to regulate a variety of genes in tissues such as liver, kidney, stomach, intestine, and pancreas (35). Our multi-tissue RNA blots revealed specific SRC1␣ expression in all of these tissues (albeit at highly variable levels). In addition, it was clear from our S1 analysis (Fig. 3) that SRC1␣ transcripts were present in several colon cancer cell lines and were especially abundant in the hepatocarcinoma cell line, HepG2.
HNF-1␣ Binds to and Regulates the Minimal SRC1␣ Promoter-To identify factors interacting with the minimal SRC1␣ promoter we employed EMSAs using a 32 P-labeled 163-bp fragment encompassing the Ϫ145 to ϩ18 region of the promoter, as detailed under "Experimental Procedures." As shown in Fig.  6A, one major diffuse complex was formed between this probe and a HepG2 nuclear extract. This complex formation was abolished by a 25 molar excess of a synthetic oligonucleotide encompassing the SRC1␣ HNF-1 like sequence (Src-HNF, Ϫ56 to Ϫ35) or by a consensus HNF-1 oligonucleotide (cHNF-1). In contrast a mutated version of the SRC1␣ HNF-l oligonucleotide (mHNF) failed to compete. Because HNF-1 consists of two isoforms (HNF-1␣ and HNF-1␤) that bind the same sequence as either homo-or heterodimers (36,37), we carried out additional EMSAs in the presence of antibodies specific to either HNF-1␣ or HNF-1␤. We found that the entire complex was supershifted in the presence of anti-HNF-1␣. In contrast, anti-HNF-1␤ had little or no effect on the complex (Fig. 6A).
To further explore the role this sequence plays in regulating SRC1␣ in vivo, we introduced mutations into the Ϫ56 to Ϫ35 HNF-1 site in both Ϫ777SRC1␣-CAT3 and Ϫ145SRC1␣-CAT3. As shown in Fig. 6B this mutation resulted in the destruction of promoter activity following transfection into HepG2 cells. Because our S1 protection experiments demonstrated that the HTB 10 cell line did not express c-Src transcripts from the SRC1␣ promoter, we repeated these transfection experiments in these cells. None of the SRC1␣ constructs demonstrated measurable activity in HTB 10 cells, nor did nuclear extracts form a complex in EMSAs (results not shown). Western blot analysis demonstrated a complete absence of HNF-1␣ in the HTB 10 cell line, compared with intermediate levels in HT-29 cells and a strong signal from HepG2 cells (Fig. 6C). In contrast to this tissue-specific pattern of expression, Sp1 and Sp3 (which regulate SRC1A expression) appeared to be present at comparable levels.
Finally, we carried out cotransfection experiments using SRC1␣-CAT3 constructs and the expression plasmids pBJ5-HNF-1␣ and pBJ5-HNF-1␤. Because of the high endogenous FIG. 2. Transcriptional start site mapping of the SRC1␣ promoter. A, exon 1␣ sequence (based on the longest 5Ј RACE clone) is shown boxed and shaded (Ϫ30 to ϩ 175) and is bordered at its 3Ј end by a consensus splice site. The major start site is designated as ϩ1 (see below). The HNF-1 consensus sequence (Ϫ56 to Ϫ35) is boxed. R (G/A) at Ϫ13 refers to a single base pair polymorphism identified at this site. B, a S1 probe was synthesized from pSRC1␣ chimera (C) and used to map transcription start sites as described under "Experimental Procedures." Lane 1, yeast RNA (25 g); lane 2, HT29 total RNA (25 g). The same primer and plasmid were used to generate a sequencing ladder. Numbering is based on the designation of ϩ1 for the major start site.
HNF-1␣ levels in the HepG2 cells, we chose to carry out these experiments in HT29 cells. Interestingly, all the SRC1␣-CAT3 constructs had much lower activity in this cell line relative to both pSV2CAT and 0.54SRC1A-CAT3. Nevertheless, as shown in Fig. 7, the Ϫ145SRC1␣ϪCAT3 vector was strongly activated by HNF-1␣ but not by the related factor HNF-1␤. In addition, this transactivation was shown to be absolutely dependent on the presence of the wild type HNF-1 binding site because the mutated versions of Ϫ145SRC1␣-CAT3 were not transactivated in the presence of the HNF-1 expression vectors. Taken together, these data demonstrate that the only major nuclear species interacting with the SRC1␣ promoter is the HNF-1␣ transcription factor and that this interaction is absolutely required for promoter activity. DISCUSSION Although the SRC gene product has been the target of intensive study for many years, we still know surprisingly little about the transcriptional regulation of this important gene. We have isolated a new SRC promoter located just upstream of the previously described SRC1A housekeeping promoter (29,30). Differential use of these two promoters and their associated exons, coupled with subsequent splicing to common downstream exons, results in c-Src transcripts with different 5Јuntranslated regions that still encode identical proteins. The SRC1␣ promoter lies outside the CpG island, which encompasses the housekeeping promoter, and, in contrast to the multiple start sites mapped in the 1A promoter, transcription initiates from a single major site. Several lines of evidence lead us to conclude that this new SRC promoter is regulated by HNF-1␣. First, HNF-1␣ regulates expression of genes mainly in the liver, kidney, stomach, pancreas, and intestinal tract (reviewed in Refs. 35 and 38). Multi-tissue mRNA blots and S1 analysis of various human cell lines confirmed SRC1␣ expression was restricted to many of these same tissues. Second, FIG. 3. Differential use of the SRC1␣ and 1A promoters in human tumor cell lines. A, S1 protection using a 1␣-specific probe. A single-stranded 32 P-labeled probe was generated from pSRC1␣ chimera and used in S1 analysis of various human cell lines (colon cancer unless specified otherwise) as described under "Experimental Procedures." Lanes 1-11, yeast RNA, HT29, WiDr, COLO201, COLO205, HepG2 (hepatocarcinoma) SW480, SW620, T47D (breast carcinoma), HTB 10 (neuroblastoma), and HTB 18 (retinoblastoma), respectively. The prominent protected species at 345 bp represents full-length protection, whereas the 172-bp species is consistent with a c-Src mRNA species lacking exon 1␣ sequence. B, S1 protection using a 1A-specific probe. A S1 probe was synthesized from the pSRC1A chimera vector in a fashion similar to that described above and used to analyze RNA isolated from various cell lines. Lanes 1-7, undigested probe, yeast RNA, HT29, HepG2, SW480, HTB 10, and HTB 18. Full-length protection is seen at 274 bp, whereas the species at 172 bp is consistent with a c-Src mRNA species lacking exon 1A. Lane M, size markers were generated from a 32 P-labeled HpaII digest of pBluescript.

FIG. 4. Comparison of overall c-Src and 1␣-specific expression levels.
A, a multi-tissue RNA blot was purchased from CLONTECH containing standardized levels of poly(A) ϩ mRNA from various normal human tissues (see C for key). The blot was hybridized with a 32 P-labeled c-Src-specific probe derived from the coding region of a SRC cDNA. B, the same filter was stripped according to manufacturer's instruction and then rehybridized with a 32 P-labeled probe specific for exon 1␣. The autoradiographs for c-Src and exon 1␣ were generated by exposing filters to x-ray film at 80°C in the presence of an intensifying screen for 18 and 36 h, respectively. regulation of this promoter in HepG2 and HT29 cells was absolutely dependent on the presence of a cis-acting element that we demonstrated was predominantly bound by HNF-1␣. Third, SRC1␣ transcripts were absent in cell lines such as HTB 10, which do not express HNF-1. Lastly, coexpression of SRC1␣ϪCAT3 vectors with a HNF-1␣ (but not the related HNF-1␤) expression vector in HT29 cells resulted in transactivation of the promoter. The observation that HNF-1␤ is unable to transactivate the SRC promoter is consistent with a number of reports that describe HNF-1␤ as a weak transactivator of other HNF-1-dependent promoters (39 -42). The dual promoter system described here is therefore likely to be involved in both basal transcription in the majority of cells and tissues (most likely from the 1A promoter) as well as during transcriptional up-regulation of SRC in specific tissues during development, differentiation, or for other specific physiological reasons. A system of dual, or even multiple promoters, is found in a number of other genes, including other SRC family members such as LCK (43), and adds considerable potential flexibility to the transcriptional regulation of a gene (44). However, specific examples of how such promoters are differentially regulated are rare. The S1 protection experiments and transcription factor expression studies described here prompts us to suggest the following working model for SRC promoter regulation: in tissues or cells where HNF-1␣ is present at high levels, transcription would take place predominantly from the 1␣ promoter, even in the presence of Sp1, because the frequent transcriptional elongation from 1␣ would inhibit or disrupt formation of preinitiation complexes at 1A. An example of this scenario is seen in the HepG2 cell line where there is high HNF-1␣ expression and predominant expression of SRC1␣, despite the presence of Sp1 at levels similar to those of other cell lines. Interestingly, we found that the SRC1␣ and SRC1A-CAT3 expression vectors independently possess similar activities in transfection experiments in HepG2. However, S1 data strongly indicate expression from SRC1␣ predominates in vivo. Various restriction fragments from the 1␣ promoter region were cloned into the pCAT3-Basic vector using a common 3Ј MscI restriction site, as described under "Experimental Procedures." Numerical designation of each construct is based on assigning the primary transcription start site as ϩ1. CAT levels were determined by CAT ELISA corrected for transfection efficiency with a ␤-galactosidase expressing vector and are presented as percentages of the activity of the vector pSV2CAT. CAT values represent means Ϯ S.D. of at least three experiments performed in duplicate.
FIG. 6. Interaction of HNF-1␣ with the SRC1␣ construct. A, a 32 P-labeled fragment from the SRC1␣ promoter region (Ϫ145 to ϩ19) was incubated with HepG2 nuclear extract (5 g) in the absence or presence of specific oligonucleotide competitors or antibodies. Complexes were then resolved on polyacrylamide gels as described under "Experimental Procedures." When included, competitor oligonucleotides were present at a 25 molar excess relative to probe. C1 represents the most abundant complex formed with this probe, and Free refers to the mobility of uncomplexed probe. Complexes in lanes 3 and 5 contain 1 g of antibody, whereas lanes 4, 6, and 7 contain 2 g. All lanes contain HepG2 nuclear extract with the exception of lane 1, which contains probe alone. B, specific mutations at the HNF-1 binding site (Ϫ52 to Ϫ35) were introduced into the CAT expression vectors Ϫ777 and Ϫ145SRC1␣-CAT3 (Ϫ777wt and Ϫ145wt) to produce the corresponding mutant versions (Ϫ777mut and Ϫ145mut) as described under "Experimental Procedures." Following transient transfection into HepG2 cells, CAT levels were assayed by ELISA as described under "Experimental Procedures." CAT levels are expressed relative to pSV2CAT and represent the averages and S.D. of at least three experiments performed in duplicate. C, whole cell extracts from the cell lines HT-29, HepG2, and HTB 10 were prepared and separated by SDS-polyacrylamide gel electrophoresis. Following transfer to nitrocellulose, the blot was incubated sequentially with antibodies specific for HNF-1␣, Sp1, and Sp3 as described under "Experimental Procedures." In cells where there is less HNF-1␣ but comparable levels of Sp1 (e.g. HT-29 cells), we would expect to see relatively more expression originating from SRC1A because the preinitiation complex has an increased likelihood of forming as a result of decreased expression originating from 1␣. Indeed, based on our transfection data, SRC1A is a much stronger promoter than SRC1␣ in HT-29 cells, but S1 data suggest that the two promoters are used at similar levels in vivo. Lastly, in cells that do not express HNF-1␣, there would be exclusive expression from 1A, as is the case with the HTB 10 cell line. Clearly, future experiments using a single expression vector that contains both promoters will be a prerequisite to testing these models. Furthermore, we cannot rule out the possibility that additional factors are required for full expression of the SRC1␣ promoter, especially in HT29 cells. Indeed, the close physical proximity of these two SRC promoters suggests that cross-talk between factors associated with each may be possible or even likely.
Our observation that SRC expression can be regulated by a tissue-specific transcription factor such as HNF-1␣ raises obvious questions as to what possible roles SRC might play in these tissues? Assigning specific functions to c-Src is always difficult because of the functional redundancy associated with the large c-Src family (Ref. 45 and references therein). However, an examination of the cell types in which HNF-1 is expressed presents some possible clues; HNF-1 was originally described as a liver-specific factor involved in the regulation of genes such as albumin, ␣-fetoprotein, and ␤-fibrinogen, among others (35). HNF-1␤ is expressed earlier in development, has a broader range of expression (46 -48) and appears to be essential for visceral endoderm differentiation (47). HNF-1␣, on the other hand, is expressed at a later stage and has been implicated in the maintenance or establishment of a correct differentiation state (49,50). In contrast to the embryonic lethal phenotype of HNF-1␤ knockouts (47), HNF-1␣ gene disruption in mice does not seriously affect correct embryonic development. Offspring, however, exhibit liver enlargement, a wasting syndrome, and severe kidney dysfunction resulting from renal proximal tubular failure (50). Both HNF-1 isoforms are expressed in specialized polarized epithelial cells (50). For example, HNF-1 expression is particularly high in the crypts of mouse small and large intestine (51). Interestingly, high pp60 c-src activity has also been reported in such crypt cells, suggesting c-Src is involved in either the proliferation or differentiation of these cells, possibly under the control of HNF-1 (52). In addition, HNF-1 expression in kidney is restricted to the proximal tubules (49); c-Src plays a key role in the regulation of the NHE3 Na/H antiporter in these proximal tubules (53) suggesting again that c-Src, under the control of HNF-1, may play a role in ion transport in the kidney and perhaps in the polarized epithelia of other tissues.
Lastly, we are particularly interested in the regulation of the SRC gene in cancer cells. It is interesting to note that many of the cancers most frequently linked to c-Src overexpression (colon, liver, and pancreas) arise from tissues in which HNF-1 activity is present. A recent report has linked the ratio of HNF-1␣ and HNF-1␤ to the histological differentiation status of hepatocellular carcinoma (54), and c-Src expression levels have in the past been linked to the differentiation status of colon cancer cell lines (55). Thus, transcriptional deregulation of the SRC gene in certain cancers and its subsequent contribution to tumorigenesis may ultimately result from changes in HNF-1 expression that in turn may result from a corruption of normal differentiation pathways during neoplasia. Results described here provide an important first step in elucidating the role HNF-1 plays in c-Src overexpression in cancer, as well as expanding our understanding of the regulation of SRC transcription under normal physiological conditions.