The cystic fibrosis gene has a "housekeeping"-type promoter and is expressed at low levels in cells of epithelial origin.

Evaluation of the expression of the cystic fibrosis (CF) gene in human epithelial cell lines demonstrated active, but low level, transcription of the gene. Analysis of 3.8 kilobases of genomic sequences 5' to exon 1 of the CF gene demonstrated no TATA promoter element, but a high G + C content, multiple transcription start sites, and several potential Sp1 binding sites. Fragments of 5'-flanking sequences from 2.2 kilobases to as small as 102 base pairs 5' to the major transcription start site supported constitutive reporter gene expression in epithelial cells, but at low levels, and independent of the length of the 5' fragment. CF gene transcription was down-regulated by phorbol myristate acetate. Importantly, evaluation of freshly isolated normal human bronchial cells also demonstrated CF gene transcription at a relatively low rate. Together, these observations suggest that although the normal CF gene promoter has characteristics of a "housekeeping"-type gene, and the gene is expressed at low levels in cells of organs that manifest the clinical disorder "cystic fibrosis," its expression can be modulated transcriptionally, implying a possible therapeutic approach for the disease.

Cystic fibrosis (CF),' a fatal recessive hereditary disorder characterized by abnormalities of electrolyte transport in organs with exocrine glands, results from mutations of a 27exon 250-kb gene on chromosome 7 at q31 (1-7). The major clinical manifestations of CF are in the lung, with mucus impaction, bacterial colonization, chronic infection, and consequent derangements of airways and lung parenchyma. In the gastrointestinal tract, pancreatic insufficiency and large bowel water and electrolyte abnormalities can cause diarrhea or intestinal obstruction (1). In the skin, sweat gland secretions have a high content of C1-, a feature that forms the basis of the classic laboratory diagnosis of the disease (1).

* This work was supported in part by The American Cystic Fibrosis
Foundation and The French Cystic Fibrosis Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in thispaper has been submitted to the GenBank?"/EMBL Data Bank with accession number(s) M58478.
The predicted primary translation product of the CF gene is a 1480-amino acid protein termed the "cystic fibrosis transmembrane conductance regulator (CFTR)" (3). Although the function of this putative protein is unknown, evidence from a variety of sources suggests it modulates C1-, and possibly Na+, transport across the apical surface of epithelial cells (8)(9)(10)(11)(12)(13)(14). In this regard, compared with normal cells, cultured airway and pancreatic epithelial cells derived from individuals with CF secrete less C1-in response to secretagogues that increase intracellular CAMP or activate protein kinase C (10)(11)(12)(13)(14). Importantly, transfer of a normal CF cDNA reverses these abnormalities in C1-secretion (15)(16)(17). In addition to modulating the electrolyte milieu on epithelial apical surfaces, it has been hypothesized that CFTR may function to regulate electrolyte transport of intracellular vesicles (18).
The purpose of the present study is to evaluate the level of CF gene transcription in epithelial cells and to characterize the promoter of the gene. Interestingly, the structure of the promoter and its function in cells of epithelial origin show that it has the characteristics of a promoter of a "housekeeping"-type gene, but is capable of being modulated at the transcriptional level, an observation with important implications for strategies designed to reverse the CF phenotype in uiuo.

MATERIALS AND METHODS
Cell Cultures-Evaluation of CF gene expression was carried out in the colon adenocarcinoma cell lines HT-29 (American Type Culture Collection (ATCC) H T B 38) and T84 (ATCC CCL 248). HT-29 cells were cultured in Dulbecco's modified Eagle's medium (Whittaker Bioproducts) supplemented with 10% fetal bovine serum (Biofluids) and T84 cells in Dulbecco's modified Eagle's medium with 5% fetal bovine serum using standard techniques.
C F mRNA Transcripts-CF mRNA transcripts were evaluated by Northern analysis (19). Total RNA was isolated from HT-29 or T84 cells by the guanidinium thiocyanate-CsC1 gradient method. RNA (10 pg) was subjected to formaldehyde-agarose gel electrophoresis, transferred to a nylon membrane (Nytran, Schleicher and Schuell), hybridized with a "P-labeled 1.6-kb CF cDNA probe (PuuII-AccI fragment spanning exons 2-12 (ATCC 61136)) generated by random priming method, and evaluated by autoradiography.
C F Gene Transcription Rate-CF gene transcription rate was examined by nuclear transcription run-on analysis (20). Nuclei isolated from 5 X lo7 cells of each cell line were incubated with 5 mM ATP, 2 mM CTP, 2 mM UTP, 250 pCi of [a-"P]GTP (Amersham Corp., >400 Ci/mmol), and RNase inhibitor (RNasin, Promega) to label actively transcribed RNA. The RNA was recovered by the acid guanidinium thiocyanate-phenol-chloroform method, purified by Sephadex G-50 column chromatography, and hybridized to excess amounts (5 pg) of DNA targets (see below) immobilized on Nytran. The membranes were then washed, exposed to RNase A and RNase T1, followed by proteinase K (all from Boehringer Mannheim), and evaluated by autoradiography. The DNA targets included a human CF cDNA (pTG4964; a 4.5-kb cDNA encompassing the entire coding sequence, constructed by standard methods from polymerase chain Cystic Fibrosis Gene Promoter 9141 reaction (PCR) amplified fragments of cDNA derived from human lung poly(A)' RNA), genomic clones for c-fos and c-myc (both from Lofstrand Labs), human @-actin cDNA (pHF@A-1) (21), and, as a negative control, the plasmid pUC19. T o determine the relative transcription rate of the CF gene compared with the @-actin gene (defined as lOO%), the autoradiograms were quantified by laser densitometric scanning and the values normalized to the relative length of the mRNA coding sequences within DNA targets (CF, 4.5 kb; 8actin, 2.0 kb, respectively). T o evaluate the effect of phorbol 12myristate 13-acetate (PMA) on CF gene transcription, HT-29 and T84 cells were incubated in the absence or presence of PMA (HT-29, 100 nM; T84, 40 nM) for 24 h. Nuclear run-on analyses were carried out as described above, and the changes in the transcription level of the CF gene as well as the control @-actin gene were assessed by the laser densitometry.
Sequence of the 5"Flanking Region of the C F Gene-A 4.3-kb EcoRI DNA fragment containing CF gene exon 1 and 5"flanking region was isolated from a human chromosome 7-specific library (ATCC 57722), subcloned into pUC19 (pPB236), and sequenced by the dideoxy chain termination method (22). An homology search between the CF gene 5"flanking region sequence and sequences in GenBank was performed using DNASIS (Hitachi America) and FASTA (23).
Identification of Transcription Start Sites-Three methods were used to identify the CF gene transcription start sites in HT-29 and T84 cells: primer extension analysis, S1 nuclease mapping, and RNase protection analysis. Primer extension analysis was performed using two different antisense primers located in exon 1 (CFEXI-2 (5'-GTTTGGAGACAACGCTGGCCTTTTC-3') and HCF-13 (5"CTG-AAAAAAAGTTTGGAGACAACGC-3')) 5' end-labeled with [y'"P] ATP (Amersham Corp., >5000 Ci/mmol). Briefly, 100 pg of total RNA was hybridized with 5 X lo5 cpm of CFEXI-2 or HCF-13, incubated with Moloney murine leukemia virus reverse transcriptase (GIBCO/Bethesda Research Laboratories) and unlabeled nucleotide trisphosphates. Primer extension products were then fractionated on denaturing polyacrylamide gels and evaluated by autoradiography. S1 nuclease mapping was carried out using standard techniques with a '"P-end-labeled 800-bp HindIII-AuaI fragment of pPB236. RNase protection analysis was done using a 1.3-kb HindIII-EcoRI fragment of pPB236 after subcloning into the transcription vector pBluescript I1 SK+ (Stratagene). A '"P-labeled antisense cRNA was synthesized by T 3 RNA polymerase, hybridized with cellular RNA, treated with RNase A and RNase T1, and evaluated by denaturing polyacrylamide gel electrophoresis.
Evaluation of Promoter Activity of C F Gene B'-Flanking Sequences-Transfection vectors containing fusion genes of CF gene 5"flanking sequences and a luciferase reporter gene were constructed from a pUC8-derived vector (pCMV-luciferase). Sequentially deleted fragments of the CF gene 5"flanking region were prepared by polymerase chain reaction amplification with pPB236 as a template and substituted for the cytomegalovirus (CMV) promoter in the pCMVluciferase expression vector.
The Rous sarcoma virus (RSV)-long terminal repeat (LTR) promoter-luciferase construct (pRSVL) (24) was used as the positive control, and a promoterless luciferase plasmid (pLuc0) as the negative control. HT-29 cells were cotransfected using calcium phosphate precipitation with 15 pg of each luciferase expression plasmid vector and 5 pg of a CMV promoter-chloramphenicol acetyltransferase (CAT) expression plasmid (pCMV-cat) as an internal control. Cells were evaluated for luciferase activity after 48 h (24). CAT activity was assayed by standard methods. Levels of luciferase expression by the various vectors were normalized by CAT activity and shown relative to the expression of pRSVL (defined as 100%).
C F Gene Expression in Normal Human Bronchial Cells-To evaluate expression of the CF gene at the mRNA transcript level, normal bronchial epithelium was obtained from nonsmoking normal individuals by fiberoptic bronchoscopy using a standard cytology brush. Total RNA was isolated and converted to cDNA using standard techniques. cDNA was then subjected to limited polymerase chain reaction amplification (25 cycles) using Taq DNA polymerase (Perkin-Elmer Cetus Instruments) and CF gene-specific primers in exon 21 (HCF-12: 5'-AGTGGAGTGATCAAGAAATATGG-3') and exon 24 (HCF-6 5'-TCCACGAGCTCCAATTCCATGAGG-3') (25). The Southern analysis using a :'YP-labeled nested probe prepared by the random priming method. T o evaluate the transcription rate of the CF gene in bronchial epithelial cells, normal human bronchi were obtained from lung tissue of nonsmokers resected for localized neoplasms. Bronchi distant from the neoplasm were dissected free of parenchyma, and nuclei from bronchial epithelial cells were isolated nmnlifiprl rnNA Wac o v a l l d n r l h y rrflnvncn p 1 alnot*nphn*oa;s nnsl and evaluated by nuclear transcription run-on analysis as described above.

RESULTS AND DISCUSSION
Consistent with the concept that the disease "cystic fibrosis" is manifest in epithelial tissues, and the observation of CF mRNA transcripts in the pancreatic adenocarcinoma cell line CFPAC-1 (26) and cells derived from skin exocrine sweat gland ducts (3), analysis of total cellular RNA by Northern hydridization demonstrated 6.5-kb CF mRNA transcripts in the human colon epithelial cell lines HT-29 and T84 (Fig.   lA). Furthermore, consistent with the presence of CF mRNA transcripts, HT-29 and T84 cells actively transcribed the CF gene (Fig. 1B). However, the rate of transcription of the CF gene was relatively low, representing approximately 5% of that of a control 0-actin gene in both HT-29 and T84 cells (data based on densitometric scanning of the autoradiograms in Fig. 1B).
Surprisingly, but consistent with the relatively low rate of transcription of the CF gene, sequence analysis of 3.8 kb 5' to exon 1 of the CF gene demonstrated structural features typical of a housekeeping gene (27) (Fig. 2). First, there was no TATA promoter element within 500 bp upstream of the major transcription start site defined as +1 by primer extension analysis (Fig. 2B). Two sequences with homology to a CCAAT box and one inverted CCAAT box motif flanking exon 1 were observed, but they did not occur a t positions typical for functional CCAAT sites in most regulated genes (28). Second, the G + C content was high; for example, it averaged 65% in the 500 bp immediately 5' to exon 1, compared with 40% for the entire human genome. Third, multiple transcription start sites were observed in cells expressing the CF gene. This was observed among mRNA transcripts in both  (20). In comparison to the data for the CF gene, data for the genes cfos, c-myc, and p-actin are shown. The plasmid pUC19 is a negative .control.

1.
FIG. 2. Structural organization of the 5"flanking region of the C F gene. A, sequence of 5"flanking region of exon 1. Of the 3.8 kb sequenced 5' to exon 1, the nucleotide sequence from -1000 to +99 is shown; the sequence from -3825 to +574 has been submitted to GenBank. Numbers are relative to the major transcription start site (indicated by the arrow) as defined below in R. The putative start codon in exon 1 is located a t position +72, and the deduced amino acid sequence is shown below (single-letter code). There are two CCAAT-like sequences (-168 and -52, underline) and one inverted CCAAT motif (-60, ooerline). Three putative binding sites for the nuclear transcription factor Spl are indicated by the boxed GGGCGG core sequences a t -950, -334, and -255. A 38-bp segment from -168 to -141 has an 89% homology with the promoter sequence of human nl(1) collagen gene (dashed ooerline). R, determination of the transcription start sites. Shown is the primer extension analysis of CF mRNA from HT-29 and T84 cell lines utilizing the antisense primer CFEXI-2. The results are shown diagramatically at the top with representative autoradiograms at the bottom. Multiple transcription start sites were observed ranging over 110 bp in both HT-29 and T84 cells, with one prominent band (defined as +1) with microheterogeneity. Minor bands were observed a t -61, -60, -1, +2, +4, +35, +44, +52, and +53.
HT-29 and T84 cells, where multiple minor transcription start sites were observed in addition to the major start site 72 bp 5' to the putative translation start codon in exon 1. The same observations were made using three separate analytical methods, primer extension analysis (Fig. 2B), S1 nuclease mapping, and RNase protection analysis (data not shown). Finally, there were three potential binding sites for the transcription factor Spl (GC boxes a t -950, -334, and -255 relative to the major transcription start site), sequences commonly observed in promoters of housekeeping genes (27,29).
Although the CF gene 5"flanking sequence showed many features typical of a housekeeping gene, it also contained sequences with possible relevance to the modulation of the CF gene transcription. Among these were multiple potential AP-1 binding sites (at positions -1058, -976, -745, -283), elements that for some genes respond to cell activation with phorbol esters (29,30). In this regard, exposure of T84 cells to PMA resulted in a decrease in CF mRNA transcript levels and a parallel adoption of the cystic fibrosis C1-secretory phenotype, i.e. an inability to increase C1-secretion in response to agents such as forskolin.2 Consistent with these observations, the transcription rate of the CF gene decreased to 35-60% of the basal level by PMA induction in both cell lines, whereas the 8-actin transcription rate increased, suggesting that CF gene expression can be modulated in spite of the characteristics of the CF gene promoter as a housekeeping gene (Table I) Fig. 1A. The data represent densitometric measurement (in arbitrary units) of the individual signals after the negative control (pUC19) has been subtracted from CF or @actin data, respectively. In parentheses below the densitometric data is the calculated relative transcription rate relative to resting cells.
Epithelial cells Gene Experi- intracellular CAMP is one mechanism by which the CF gene product may modulate C1-secretion (8-14, 18, 26). There were also a number of potential glucocorticoid response elements (positions -1466, -1226, -1140, -941) as well as AP-2 binding sites (-1107, -344) and one C/EBP binding site 1 of the CF gene to that of other promoters demonstrated no striking similarities except for a 28-bp sequence at -168 that had an 89% homology with a segment of the promoter of the human al(1) collagen gene, a sequence important in the regulation of that gene (32). In addition, CF gene sequences Transfection of fusion genes composed of CF gene 5'flanking sequences and a luciferase reporter gene into HT-29 cells proved that these CF gene sequences had promoter function in cells that normally transcribed the endogeneous CF gene (Fig. 3). Interestingly, the region including only 102 bp immediately 5' of the major transcription start was capable of promoting luciferase expression. Consistent with the housekeeping gene concept (27), the relative amount of reporter gene expression did not change appreciably by progressively including additional 5' sequences up to -2244 from the transcription start. Importantly, and consistent with the relatively low level CF gene transcription observed in epithelial cell lines (Fig. lB), the relative level of luciferase expression supported by the CF gene 5"flanking sequence in transfection experiments with HT-29 cells was quite low, invariably <5% of that of the RSV-LTR promoter control.

HT-29
Although the endogenous CF promoter and transfected CF promoter functioned similarly in regards to low level expres- This evaluation is particularly relevant to CF, as the bronchial epithelium is the major site of the clinical manifestations of the disease. Not only were CF mRNA transcripts observed in freshly isolated bronchial cells (Fig. 4A), but nuclei isolated from normal bronchial epithelium demonstrated active transcription of the CF gene (Fig. 4B). Importantly, as observed in HT-29 and T84 cells, the level of transcription was quite low, representing only approximately 6% of that of the pactin gene.
Together, the structural features of the CF promoter and the low level expression of the CF gene in epithelial cells suggest that it should be classified as a housekeeping gene, together with other genes such as dihydrofolate reductase, hypoxanthine phosphoribosyl transferase, adenosine deaminase, and phosphoglycerate kinase genes (27). However, whereas most housekeeping genes code for proteins that are required for general cellular metabolic functions, the CF gene product probably is not absolutely required for function of the cell itself, i.e. the housekeeping function of the CF gene product may be more relevant to the overall function of the organ, rather than for the well-being of the cell. In this context, although mutations of the CF gene have profound consequences to organs such as the lung or pancreas, expression of the CF gene in epithelial cells is not necessary to sustain life or even to maintain the viability of the cells that express the gene (2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)26). Consistent with this concept, there is evidence that epithelial cells expressing the CF gene have multiple mechanisms controlling C1-secretion (9)(10)(11)(12)(13)(14)34), i.e. the CF gene product probably is not critical for maintaining the intracellular electrolyte milieu, but rather the extracellular milieu and/or within specialized vesicles such as those of bronchial mucus-producing cells (18,35,36). Alternatively, the electrolyte-related functions of CFTR may be only one part of its function, with other specialized functions linked to the disease process of cystic fibrosis. Thus, the CF gene may be in a subgroup of genes with structural chmactprinticn nf hnnnPkwping genec, hilt that have ticc11especific functions such as the human nerve growth factor receptor gene (37) and the Pim-1 proto-oncogene (33) and/or that exhibit regulation in gene expression by specific agents.
Irrespective of the classification of the promoter of the CF gene, the observation that the CF gene is transcribed in normal bronchial tissues at a low rate, together with the knowledge that individuals homozygous for the APhe508 mutation have CF mRNA transcript levels in respiratory epithelium comparable with normals (31), has important implications for the feasibility of directly correcting CF by in vivo gene transfer. In this regard, if the normal CF gene can be safely transferred to the respiratory epithelium, it should not be difficult to match the low level expression of the two endogenous CF genes sufficiently to convert the respiratory epithelium to the status of at least that of a CF heterozygote. Furthermore, the observation of the down-regulation of CF gene transcription by PMA leads to the possibility that the disease might be treatable by developing agents that would mimic this effect.