THE JOURNAL OF BIOL~CICAL CHEMISTRY 0 1994 by The American Society for Biochemistry and Molecular Biology, Inc.

Val. 269, No. 12, Issue of March 25, pp. 9298-9303, 1994 Printed in U.S.A.

Expression of the Dibasic Proprotein Processing Enzyme Furin Is Directed by Multiple Promoters*

(Received for publication, November 5 , 1994)

Torik A. Y. Ayoubi+, John W. M. Creemers, Anton J. M. Roebroek, and Wim J. M. Van de Ven From the Laboratorv for Molecular Oncolom. Center for Human Genetics, University of Leuven, Herestraat 49, B-3000 Leuven, Belgium

The prototype mammalian proprotein processing en- zyme furin is shown to be encoded by three distinct FUR mRNA isoforms which differ only in their 5“untrans- lated regions. By primer extension analysis, the tran- scription start sites of the three mRNA isoforms were defined. The genomic regions located immediately u p stream of the three alternative transcriptional start sites were shown to possess promoter activity in trans- fection experiments using the luciferase encoding gene as reporter. In a liver cell line, the P1 promoter appeared to be the strongest; in a lung cell line, the P1A promoter. Human FUR promoter P1 but not P1A or P1B was trans- activated by transcription factor CIEBPP. Other mem- bers of this family of bZIP transcription factors, CIEBPa and CIEBPS, were not able to transactivate the P1 pro- moter. Promoter P1A and P1B have characteristics of promoters of housekeeping genes. They lack TATA or CAAT boxes upstream of the transcription start site but are very GC-rich and contain several SP1 sites. Pro- moter P1, on the other hand, has a TATA box in the proximal promoter region. In electromobility shift as- says and DNase I footprinting analysis, transcription factor SP1 was found to bind to the proximal region of the P1 promoter. Altogether, our results indicate that expression of the human FUR gene is directed by alter- native promoters, housekeeping (GC-rich) as well as regulated (TATA-containing) promoters, suggesting that their differential use may be a mechanism to modulate levels of the furin enzyme.

The proprotein processing enzyme furin is the mammalian prototype of a novel family of subtilisin-like serine endopro- teases which possess cleavage specificity for sites involving multiple basic amino acid residues and are involved in the processing of precursor proteins of a variety of regulatory pep- tides and proteins (for recent reviews, see Refs. 1-3). Furin has been shown to be able to process a variety of proproteins into their mature products; among others those for von Willebrand factor (4,5), p-nerve growth factor (61, albumin (71, complement C3 (7), factor M (8), influenza virus hemagglutinin (9), HIV-1 glycoprotein gp120 (lo), and anthrax toxin protective antigen (11). The furin enzyme, which is encoded by the FUR gene (121,

* This work was supported by the “Geconcerteerde Onderzoekacties 1992-1996,” EC Contract BIOT-CT91-0302, and “Stichting Technische Wetenschappen” Grant STW-22.2726. This text presents results of the Belgian Programme on Interuniversity Poles of Attraction initiated by the Belgian State, Prime Minister’s Ofice, Science Policy Program- ming. The scientific responsibility is assumed by its authors. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertise- ment” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Biology Organization (EMBO). $Holder of a long-term fellowship from the European Molecular

has been reported to be ubiquitously expressed (13, 14). This is in contrast to other members of this family which display a tissue-specific expression pattern; e.g. the prohormone conver- tases (PCs)’ PClPC3 and PC2, which are neuroendocrine-spe- cific (15-171, and PC4, which is expressed only in spermato- genic cells (18, 19). In this context, it should be noted, furthermore, that levels of FUR mRNA were found to vary in a tissue-dependent manner (13). The mechanisms by which dif- ferential expression of these novel enzymes is controlled are not yet known. In the present report, we have studied regulation of expression of the human FUR gene.

In previous studies, we have cloned and characterized the genomic DNA encompassing the entire human FUR gene (20) as well as corresponding cDNAs (12, 21). Upon nucleotide se- quence analysis of these DNAs, the genetic organization of human FUR could be determined. The gene spans about 20 kbp, with the coding sequences dispersed over 15 exons, start- ing in exon 2 and ending in exon 16 (4). Interestingly, a human FUR cDNA with an alternative 5‘-noncoding exon has been described (14). Alignment of the alternative 5’-noncoding se- quence and the human FUR genomic sequences revealed that it corresponded to a region within the first intron of the FUR gene and that it was flanked at the 3‘-side by a splice donor site. We will refer to this alternative exon as exon 1A. The alternative exon 1A was also found in a mouse FUR cDNA (22). Further- more, the 5’-noncoding region of a rat FUR cDNA (23) appeared to align with yet another region within the first intron of hu- man FUR, and it was also flanked by a splice donor site. We have designated this putative third alternative first exon as exon 1B. It remains to be established whether the three alter- native 5’-noncoding exons in FUR mRNAs arise by transcrip- tion from different promoters or by differential splicing from a single primary transcript.

Here we describe experiments aimed to elucidate the mecha- nism by which expression of human FUR is controlled. The transcription start sites of alternative FUR mRNA isoforms were determined using primer extension analysis. Promoter activity of the sequences upstream of the transcription initia- tion sites was examined in transient transfection assays. The results indicate that at least three promoters can drive the expression of the FUR gene. Furthermore, characteristics of the various promoter regions were established. Two of them appeared to have architectural features of housekeeping pro- moters. The third one turned out to be of a different category as it appeared to be trans-activated by transcription factor C/EBPp.

MATERIALS AND METHODS Plasmids-Molecular cloning was performed according to standard

techniques (24). Clone pAR5, which contains a human genomic DNA

The abbreviations used are: PC, prohormone convertase; C/EBP, CCAAT/enhancer-binding protein; kbp, kilobase pair(s); bp, base paifis).

9298

Regulation of FUR Gene Expression 9299

insert of about 20 kbp including the complete coding region of FUR, was used for subcloning (21). DNAfragments starting at the PstI site in exon 1 and extending to a PstI site located 3.7 kbp upstream, a Sac1 site located 2.6 kbp upstream, a NheI site located 1.3 kbp upstream, or a KpnI site located 500 bp upstream were inserted into the multicloning site of pGLa-Basic, a promoter- and enhancerless vector containing the firefly luciferase encoding sequence. The resulting clones were desig- nated pGL2-P1, pGL2-Pl-Sac1, pGLa-Pl-NheI, and pGLa-Pl-KpnI, re- spectively. A DNA fragment starting at the XmnI site in exon 1A and extending to a NheI site located 2.2 kbp upstream was inserted into pGLa-Basic, resulting in clone pGL2-P1A. Similarly, a DNA fragment starting at the PtlMI site in exon 1B and extending to a SacI site located 1.2 kbp upstream was inserted into pGLa-Basic, resulting in clone pGL2-P1B. The ClEBP expression vectors pMSV-ClEBPa, pMSV-C/ EBPp, and pMSV-ClEBPS were kindly provided by Drs. S. L McKnight and W.-C. Yeh.

Primer Extension Analysis-Primer extension was performed essen- tially by established methods (24) using oligonucleotides specific for each of the alternative first FUR exons. The following primers were used in this study: OL-I: 5'-ITCACCATCCAGCTCCAGAG-3' corre- sponding to nucleotides 1-20 relative to the 3'-end of FUR exon 1; OL-1A 5'-AGCCTGGCGGGGAAACl"-3' corresponding to nucleotides 8-27 relative to the 3'-end of FUR exon lA, OL-1B: 5"GACTGCTGm- TCCCAGCAC-3' corresponding to nucleotides 10-28 relative to the 3'- end of FUR exon 1B; OL-ml: 5'-CGTCACCATCTAGCTCCAGA-3' cor- responding to nucleotides 1-20 relative to the 3'-end of mouse exon 1. Primers were 5'-end-labeled using Ly-32PlATP (6000 Cilmmol, DuPont NEN) and T4 polynucleotide kinase (Boehringer Mannheim). Each ra- diolabeled primer (5 x los dpm) was added to 10 pg of total RNA in 25 pl of hybridization buffer (80% deionized formamide, 40 m Pipes, pH 6.4.400 mM NaCl, 1 m EDTA). The mixture was heated at 85 "C for 5 min, and DNA was then allowed to hybridize at 42 "C for 16 h. Subse- quently, the primer was extended in a reverse transcriptase reaction at 42 "C for 60 min using reverse transcription buffer and 100 units of reverse transcriptase (Life Sciences) in a 2 0 4 reaction volume. RNase A (2 pl of a 0.5 pg/pl solution in 0.5 M EDTA) was then added, and incubation was continued at 37 "C for 15 min. After ethanol precipita- tion, the extended products were fractionated on a 6% polyacrylamide sequencing gel containing 8 M urea.

Cell Lines, DNA Dansfer, and Dansient Gene Expression-The HepG2 (human hepatocellular carcinoma), COS-1 (African green mon- key kidney), RIN 1027 B2 (somatostatin secreting rat cell line), and NCI-H82 (human small cell carcinoma of the lung) cell lines were ob- tained from the American Type Culture Collection (ATCC) and grown according to the ATCC recommendations. Promoter activity of DNA fragments was assayed using transient gene expression. DNAs were purified via Qiagen columns. Transfections were carried out using lipo- some-mediated DNA transfer (DOTAB, Boehringer) according to the guidelines of the manufacturer. For each experiment, luciferase activity was determined in duplicate or triplicate wells, and each DNA construct was tested in two to five separate experiments. Therefore, results are expressed as the mean of 4-15 values.

DNase Footprint Assay-The noncoding strand of a 170-bp BamHI- PstI DNAfragment of PGL2-P1 was first end-labeled using [a-32P]dCTP and Klenow polymerase and, subsequently, purified by gel electrophore- sis. DNase1 footprinting was performed using affinity-purified human SP1 protein and a commercial footprinting system (Promega) according to the protocol provided by the manufacturer. Partially digested DNA was recovered and, after phenol extraction and ethanol precipitation, analyzed on a 6% polyacrylamide gel containing 8 M urea.

Electromobility S'hifl Assay--Two complementary oligonucleotides

and its antisense strand) that spanned the putative low aflhity SPI

tides (OL-SP1: 5'-AlTCGATCGGGGCGGGGCGAGC-3' and its anti- binding site in the P1 promoter and two complementary oligonucleo-

sense strand) that contained a high affinity SP1 binding site were used in mobility shift assays. The complementary strands were first an- nealed and then end-labeled using T4 polynucleotide kinase and [Y-~~PIATP; 10,000 dpm were used for each binding reaction. Rat liver nuclear extracts were prepared by the method of Gorski et al. (26). For binding reactions, 15 ng of purified SP1 or 5 pg of rat liver nuclear extract were first preincubated with or without 100-fold molar exce: of unlabeled oligonucleotide or with 20 m ZnC1, in 25 nm Tris-HC1 (pH 8.0), 50 m KCI, 6.0 m MgCl,, 0.5 mM EDTA, 1 m dithiothreitol, 1 pg of poly(dI-dC), and 10% glycerol on ice for 20 min. Subsequently, oligo- nucleotide DNAs were added, and the incubation continued for 20 min on ice. Finally, the reaction mixtures were loaded directly onto a non- denaturing 690 polyacrylamide gel (aerylamidehisacrylamide, 19:l)

(OL-hFURPlSP1: 5"'IITTAGCTCCTCCCCCCACCCACAGACACC-3'

1 1A 18 2 . . . . . . . .

I I I 1

P S N K P N X S Pf AT0

8 m m

OL-1 OL-1A OL-1B

n 1 kb

FIG. 1. Schematic representation of the genetic organization of the 5"region of the human FUR gene. The relative positions of exon 1, exon lA, and exon 1B are depicted (boxes). K, KpnI; N, NheI; P , PstI; Pf, PfiMI; S , SacI; X, XmnI.

prepared in Tris boratelEDTA buffer (0.5 x TBE: 45 m Tris base, 45 m boric acid and 1 m EDTA). Electrophoresis was performed at 4 "C at a current of 35 mA using 0.5 x TBE as the running buffer.

RESULTS

Detection of FUR mRNA Isoforrns and Analysis of the Genomic DNA Regions Immediately Upstream of FUR Noncod- ing Exons I, IA, and IB-In a previous Northern blot study of FUR expression in mouse (221, evidence was obtained that furin can be encoded by at least three distinct mRNA isoforms. In order to determine whether all three first exons were ex- pressed in human tissues, primer extension analysis was per- formed. Primers were designed complementary to sequences of exons 1, IA, and 1B (Fig. 1). Hybridization of each of these primers to RNA of human liver and HepG2 cells showed that both in human liver and HepG2 cells the three alternative 5"noncoding exons were expressed (Fig. 2A 1. In all three cases, heterogeneous start sites were identified; one major start site and several minor ones. From the size of the extension products obtained with each primer (only those corresponding to the transcripts originating from the major start sites are consid- ered here), exon 1 was estimated to be about 158 nucleotides, exon 1A about 370 nucleotides, and exon 1B about 180 nucle- otides. Results obtained in S1 nuclease analysis seemed to con- firm the pattern of transcription initiation as observed in the primer extension studies (not shown). Examination of the 5'- flanking region of exon 1 revealed a putative CCAAT, an SP1, and a TATA element in close proximity to the transcription start site. Interestingly, the TATA element turned out to be downstream of the major transcription start site as determined by primer extension analysis. There are only a few reports in the literature of TATA elements downstream of the major tran- scription start site (27-30). It may thus be the case that this particular FUR promoter belongs to a class of RNA polymerase 11-dependent promoters with a functional TATA element within the transcribed sequence. However, although the sequences of the putative promoter regions immediately upstream of exon 1 in mouse and human FUR are more similar (80% sequence identity) than those of the exons themselves (67% sequence identity) (21,22), the major start site in mouse appeared to be about 25 nucleotides downstream of the TATA element (Fig. 2 0 , as expected. One interesting difference between the mouse and human sequences is the presence of an SP1 site within human but not mouse exon 1. The relevance of this SP1 site within human exon 1 for start site selection remains to be established. The genomic DNA regions immediately upstream of exon 1A as well as 1B have all the architectural features of promoters of so-called housekeeping genes. Unlike the 5'-up- stream region of exon 1, the 5"upstream regions of exons 1A and 1B lack obvious TATAor CCAAT sequences; they are, how- ever, very G€-rich (the GC content in the 100 most proximal nucleotides in the upstream region of exon 1A is 88%) and contain several SP1 elements. Transcription initiation ap-

9300 Regulation of FUR Gene Expression

FIG. 2. Primer extension analysis. A, a "2P-labeled oligonucleotide complemen- tary to exon 1 (OL-1, lanes 1 and 2 ) , exon 1A (OL-lA, lanes 3 and 4) , or exon 1B (OLIB, lanes 5 and 6 ) was annealed to 10 pg of total RNA from adult human liver (lanes 1,3, and 5) or HepG2 cells (lanes 2, 4, and 6). B, a "2P-labeled oligonucleotide complementary to exon 1 (OL1, lanes 1 and 4) , exon 1A (OLlA, lanes 2 and 51, or exon 1B (OL-lB, lanes 3 and 6 ) was an- nealed to 10 pg of total RNA from fetal human liver (lanes 1, 2, and 3 ) or NCI- H82 cells (lanes 4, 5, and 6). C, a 32P- labeled oligonucleotide complementary to human exon 1 (OL1, lane h ) or to the corresponding position in mouse exon 1 (OLml , lane rn) was annealed to 10 pg of total RNA from HepG2 cells (lane h ) or mouse liver (lane rn). Lane M, molecular weight markers.

A

M I .. 404

309

242

21 7

20 1 1 9 0

1 8 0

1 6 0

147

122 - Y

110 - I,

peared to occur heterogeneously, especially the transcripts en- coding sequences of exon 1A.

Relative Expression Levels of the Various FUR Dan- scripts-To obtain insight in the natural occurrence of the vari- ous FUR transcripts, relative expression levels of the three alternative noncoding exons of the human FUR gene were de- termined in various human cell lines and tissues by primer extension analysis. Selected human cell lines included HepG2 and the human lung carcinoma cell line NCI-H82; as human tissues, adult and fetal liver were studied. From these studies, it became clear that the human cell lines tested expressed all three alternative FUR mRNAs and that their relative amounts were comparable to those in human liver RNA (results not shown). In the small cell lung carcinoma cell line NCI-H82, however, the relative levels of the transcripts encoding exon 1 or 1A were lower than in liver cells (Fig. B). This latter ob- servation suggests that variations in the relative expression levels of the alternative transcripts may exist. From Fig. 1, it is also clear that the results obtained with RNA from human liver and HepG2 cells are almost identical, making the HepG2 cell line an appropriate model system for studying human FUR gene expression.

Promoter Activity of the FUR Genomic DNA Regions Imme- diately Upstream of the Alternative 5'-Noncoding Exons-The human genomic DNA regions upstream of exons 1, lA, and 1B were analyzed for functional promoter activity via transient expression into the HepG2 cells. As controls, each experiment included pGL2-Control (luciferase expression driven by SV40 early promoter and enhancer) and pGL2-Basic (promoterless construct). Transfection of HepG2 cells with pGLPP1, pGL2- PlA, or pGL2-P1B DNA showed a consistently higher activity of the P1 promoter region as compared to those of promoter regions P1A and PlB, respectively. Transfection with the vari- ous FUR-luciferase fusion constructs resulted in at least a 10- fold higher luciferase activity, as compared to the correspond-

"

6 C

2 3 4 5 6 1 2 3 4 5 6 M h M m " . .

404

404- -

309 309. "

242 242 - : =

217

20 1

190

217- - - 201 . - "

1 9 0 - . .- , 180" 1 8 0 - "

1 6 0 1 6 0 - - 147 147- c

122 - u 122 * .I

110. 110-1)

ing promoterless DNA constructs (Fig. 3A) . Similar results were obtained in transfection experiments with COS-1 instead of HepG2 cells (data not shown). These results provide func- tional evidence for the existence of three distinct promoter re- gions in the human FUR gene and they are likely to drive expression of the three alternative FUR mRNA isoforms.

To define the DNA sequences required for activity of the P1 promoter, various deletion mutants of the pGL2-PI DNA insert were analyzed. Deletion mutants were obtained as described under "Materials and Methods," and their promoter activity was assayed in HepG2 cells. Setting the luciferase activity obtained after transfection of pGL2-P1 DNA a t loo%, pGL2- P1-KpnI DNA, which contained only 400 bp upstream of the TATA element, appeared to express still 75% of the activity of pGL2-P1 (Fig. 3B). We conclude from this result that elements for core P1 promotor activity are contained within these 400 bp.

Differential Regulation of the FUR Promoters-The presence of CCAAT sites in the 5"flanking region of exon 1 and the observation that, of the three FUR promoters, the activity of the P1 promoter was the highest in HepG2 cells, prompted us to investigate the possible involvement of CIEBP proteins in the regulation of FUR gene expression. Functional interaction of CIEBPP with the FUR P1 promoter could be demonstrated in co-transfection experiments. Transfection of HepG2 cells with pGL2-P1 DNA, which contained the luciferase gene under the control of the 3.6-kbp human FUR P1 promoter region, to- gether with an expression vector for C/EBPa, CIEBPP, or CIEBPG resulted in a 2.5-fold stimulation of luciferase expres- sion by CIEBPP as compared to C/EBPa (Fig. 3 0 . This effect was specific for the P1 promoter sequences, since it was not observed with pGL2-P1A or pGL2-P1B DNA, in which expres- sion of the luciferase reporter gene was driven by promoter P1A or P1B sequences, respectively.

The primer extension studies described above showed that the sequences corresponding to the three alternative leader

Regulation of FUR Gene Expression 9301

A B 1400 1200

1200 1000 '

1000 800 -

800 800 .

600

400 400

200 - 200

0 0 I P I P I A PlI3 Control Bsaic PI Sac1 Nhel Kpnl Control Baslc

C D 2000 1000

1500 800

600 -

1000

400 -

500 200 -

0 0 -

P1 PIA P I 0 P1 P I A COnlrOl Baeic

immediately upstream of exon 1 (PI ), exon 1A (PIA), or exon 1B (PIB) or the SV40 early promoter and enhancer (Control) were compared to the FIG. 3. Transient gene expression assays.A, transfection of HepG2 cells with reporter constructs in which the promoter activity of sequences

empty vector (Basic). B, transfection of HepG2 cells with reporter constructs containing P1 DNA fragments of various lengths. P1, SacI, NheI, and KpnI indicate results obtained with 5"flanking DNA inserts of 3.6, 2.5, 1.2, or 0.4 kbp, respectively. C, trans-activation of the P1 promoter by CIEBPP. Effector plasmids CIEBPa (black bars), CIEBPP (hatched bars), or CIEBPG (shaded bars) were co-transfected with P1, PlA, or P1B promoter constructs. D, transfection of NCI-H82 cells with reporter constructs in which the promoter activity of sequences immediately upstream of exon 1 (PI) or exon 1A ( P I A ) or the SV40 early promoter and enhancer (Control) were compared to the empty vector (Basic).

exons were expressed, but that their relative levels varied somewhat in HepG2 cells as compared to neuroendocrine cells of the small cell lung carcinoma cell line NCI-H82. Therefore, the activities of the P1 and P1A promoters were compared in HepG2 and NCI-H82 cells. In HepG2 cells, the P1 promoter appeared to be consistently stronger compared to the P1A pro- moter. In NCI-H82 cells, however, the P1 promoter appeared to be considerably weaker than the P1A promoter (Fig. 30 1. Simi- lar results were obtained with endocrine cell line RIN 1027 B2 (data not shown). The results of these studies point toward differential regulation of the P1 and P1A promoters in hepa- toma and endocrine cell lines.

Binding of SPl to FUR DNA Encompassing the P1 Pro- moter-The presence of an SP1 element just upstream of the TATA element in the P1 promoter and several SP1 sites up- stream of the transcription start sites of mRNAs containing the sequences of exons 1A or 1B suggested that SP1 might be involved in the regulation of transcription of the human FUR gene. In view of the housekeeping character of the P1A and P1B promoters, we tested whether purified SP1 transcription factor could bind to the putative SP1 site in the TATA-containing P1 promoter region. The proximal promoter P1 fragment of the human FUR gene (extending from the PstI site within exon 1 to the BamHI site 170 bp upstream) was subjected to DNase1 footprint analysis in the presence or absence of purified human SP1 (Fig. 4). Purified SP1 produced a footprint in this fragment covering the sequence indicated in Fig. 4. The SP1 element in the proximal region of promoter P1 appeared to have a se- quence that resembles SP1 low affinity binding sites (25,311. In order to show that this SP1 site was illdeed a low affinity binding site for SP1, an oligonucleotide (OLhFUR-PlSP1) with the putative low affinity binding site was used in electro- mobility shift assays; in these assays, purified SP1 transcrip- tion factor and rat liver nuclear extracts were used. OL-hFUR- PlSPl was found to bind purified human SP1, forming a major shifted band (Fig. 5). This binding was specific, as it was com- peted by excess unlabeled oligonucleotide. A 100-fold molar

M 1 2 3

4 B 76 -

B 4 5 % 34 - a 6 V W v w

FIG. 4. DNase I footprinting of the human P1 promoter region with affinity-purified SPl. The "P-labeled DNA used for footprint- ing analysis was a 170-bp BamHI-PstI DNA fragment of PGL2-P1 that was end-labeled on the noncoding strand. DNase cleavage pattern in the absence (lane I ) or in the presence of 1 (lane 2 ) or 2 (lane 3) footprint units of SP1 protein. The sequence of the protected nucleotides is outlined. The bored nucleotide sequence corresponds to the SP1 rec- ognition sequence.

excess of unlabeled oligo OGhFUR-PlSP1 already competed completely for SP1 binding. In contrast, when oligo OLSP1, which contained a high-affinity SP1 binding site, was used, a 100-fold molar excess of unlabeled oligo OLSPl still resulted in some binding. OLhFUR-PlSP1 was also bound by proteins of nuclear extracts of rat liver. In this case, double bands were

9302 1 2 3 4 5 6 7

Regulation of FUR Gene Expression 8 9 10 11 12

OL-SPI + OL-hfur SPI - SP1 protein + extract competitor - 20 mM ZnCI, -

-

+ + + +

+ - "

- + + + + -

+ - - + "

- " - - + + + + + + + - - - - - + + + + - - + - - + " +

FIG. 5. Electromobility shift assay. A double-stranded 32P-labeled oligonucleotide corresponding to a high affinity SP1-binding site (lanes 1-61 or a double-stranded :l"P-labeled oligonucleotide containing the

human FUR gene (lanes 7-12) was incubated with affinity-purified SP1 SP1 recognition sequence in the proximal P1 promoter region of the

protein (lanes 1 3 and 7-9) or rat liver nuclear extract (lanes 4-6 and 10-12). Binding reactions were performed in the presence of a 100-fold molar excess of the corresponding unlabeled double-stranded competi- tor oligonucleotide (lanes 2, 5, 8, and 11 1 or in the presence of 20 m~ ZnC12 (lanes 3, 6, 9, and 12).

observed. The observation of such double bands has been de- scribed in the literature, and several studies show that most probably only one of these shifted bands results from SP1 bind- ing whereas the other is the result of binding of SP1-like pro- teins (32,33). The same double bands were also obtained with OL-SP1, which contained a high-affinity SP1 site, except that in this case the intensity of the double bands were stronger. I t should be noted that the same rat liver nuclear extract was used as in the experiments with OL-hFUR-PlSP1 described above. The observed difference in intensities is probably due to the fact that OL-SP1 has a higher affinity for the SP1 protein than OL-hFUR-PlSP1. An additional indication that the shifted bands were due to SP1 binding was obtained by show- ing that addition of 20 mM ZnClz to the binding reaction com- pletely abolished binding. SP1 is a zinc finger protein known to need zinc for its activity, but high levels of zinc are known to disrupt its binding to DNA (34).

DISCUSSION In the present study, the major transcription start sites for

three alternative noncoding exons of the human FUR gene have been identified. Furthermore, the locations of the corre- sponding promoter regions were determined, and their func- tioning was analyzed in transient expression assays. The re- sults of the transfection experiments in the HepG2 and NCI- H82 cell lines are in reasonable agreement with the relative pattern of promoter usage seen by primer extension analysis of endogenous RNA from these cells. This strengthens the conclu- sion that transcription differences observed with both tech- niques are real. In HepG2 cells, the P1 promoter was consis- tently stronger than the P1A promoter, whereas the opposite was the case in neuroendocrine cells such as NCI-H82 and RIN 1027 B2. This suggests that the P1 promoter could be more specific for liver than for endocrine cells in general. In this regard, it is of interest to note that several studies have shown a lower level of FUR mRNA in endocrine as compared to non- endocrine cell lines (6, 14, 22).

There appeared to be a considerable heterogeneity in the length of the 5"untranslated region of exon 1A. The human P1A promoter is assumed to initiate transcription from mul- tiple start sites. Interestingly, examination of the proximal pro- moter sequence did not reveal either a TATA or a CCAAT ele-

ment, features usually found in the proximal promoter regions of tissue-specific promoters. The absence of the TATA element may explain the wide range of transcription initiation from the P1A promoter since the TATA element is thought to be impor- tant for the precise positioning of transcription initiation sites (35). The promoters P1A and P1B have a very high GC content and several SP1 elements, typical features of .promoters of housekeeping genes. Promoter P1 has both a TATA and CCAAT element, but contains also a low affinity SP1 binding site. The CCAAT element is favorably positioned to regulate the P1 pro- moter, whose use might be more predominant in the liver due to the high content of CEBP transcription factors in hepatic cells (36, 37). In agreement with this is the observed higher activity of the P1 promoter in HepG2 cells as compared to NCI-H82 cells. In addition, the P1 promoter was truns-acti- vated by CIEBPP and not CIEBPu or CIEBPG, which is in agree- ment with the finding that CIEBPP is the predominant CIEBP isoform expressed in hepatoma cells (38). The fact that the P1 promoter was trans-activated only 2.5-fold could be due to the fact CIEBPP is constitutively present but undergoes an in- crease in activity and nuclear localization via a post-transla- tional mechanism when the cells are exposed to cytokines or elevated levels of CAMP (39-41). The presence of a proper stimulus during certain physiological conditions may enhance the observed trans-activation of the P1 promoter by CIEBPP. Altogether, the human FUR gene can be included into the group of genes containing a combination of both regulated TATA and ubiquitously active GC-rich promoters (42).

We have demonstrated that purified SP1 binds to the P1 promoter region of the FUR gene and that rat liver nuclear extracts contain an SP1 activity that binds to this element with low affinity and produces an electromobility shift similar to that produced by purified SP1. Our data suggest that the SP1 protein may be important for the expression of the FUR gene. Under normal physiological conditions, the FUR gene is ex- pressed a t low levels; expression of furin a t relatively high levels using expression vectors containing strong promoters seems to be toxic to cells.2 In this context, the lower affinity of the SP1 binding site in the P1 promoter region for the SP1 transcriptional activator may be functional in that it may ex- plain the low levels of expression of the FUR gene. The FUR mRNA isoforms are predicted to encode identical proteins since there is an in-frame stop codon just upstream of the start ATG in exon 2. For this reason, it is remarkable that the nucleotide sequences of the 5"untranslated exons of mouse and human are highly similar (90% sequence identity between mouse and human exon 1A). This suggests that the 5"untranslated exons might have some important and specific function(s) which have been conserved during evolution.

The observation that the FUR gene apparently possesses three distinct promoters, P1, PlA, and PlB, directing the syn- thesis of the mRNA-1, mRNA-lA, and mRNA-1B isoforms, re- spectively, provides an explanation for the synthesis of these different FUR mRNA species from a single gene. At the mo- ment, it remains to be determined what the functional signifi- cance of this is for control of expression. Alternative promoters have been identified in several eukaryotic genes and their dif- ferential use can provide an extra level of flexibility in the control of the expression of these genes (42). One of the pro- moters can be functional only at a particular developmental stage or in a given tissue while another can drive transcription in a different tissue. Here, we provide evidence for different relative levels of expression of the three mRNA isoforms of human FUR in the small cell lung carcinoma cell line NCI-H82

T. A. Y. Ayoubi, J. W. M. Creemers, A. J. M. Roebroek, and W. J. M. Van de Ven, unpublished observations.

Regulation of FUR Gene Expression 9303

as compared to other cells, pointing toward the possibility of J., Barr, P. J.. and Thomas, G. (1990) J. Cell Bid. 111,2851-2859 cell type-dependent variation in differential promoter usage of 7. Misumi, Y., Oda, K., Fujiwara, T., Takami, N., Tashiro, K., and Ikehara, y.

the FUR gene. Expression of the FUR gene has been found to 8. WasIey, L. c., Rehemtulla, A., Bristol, J. A,, and Kaufman, R. J. (1993) J . Biol. (1991) J. Bid. Chem. 266, 16954-16959

be ubiquitous, b u t mRNA levels vary among tissues, with the Chem. 268,845W465 highest expression in liver and kidney (13, 14,22,43). Expres- 9. Stieneke-Grober, A,, Vey, M., Angliker, H., Shaw, E., Thomas, G . , Roberts, C.,

sion of FUR mRNA in the rat central nervous system is ubiq- 10. Hallenberger, S., Angliker, H., Shaw, E., Klenk, H. D., and Garten, W. (1992) Klenk, H. D., and Garten, W. (1992) EMBO J. 11,2407-2414

uitous, as was determined by in situ hybridization analysis Nature 360,358-361 (44). All brain areas exhibited low to moderate levels in both

11. Klimpel, K. R., Molloy, S. S., Thomas, G., and Leppla, S. H. (1992) Proc. Natl. Acad. Sci. U. S. A. 89, 10277-10281

neuronal and non-neuronal tissues. However, some brain re- 12. Roebroek, A. J. M., Schalken, J . A,, Leunissen, J. A. M., Onnekink, C., Bloem-

exhibited high levels Of FUR mRNA’ The high- 13. Schalken, J. A,, Roebroek, A. J. M., Oomen P. P. C. A., Wagenaar, Sj. Sc., ers, H. P. J., and Van de Ven, W. J. M. (1986) EMBO J. 5,2197-2202

est level of FUR mRNA was observed in the granular cells of the islands of Calleja. The observed higher levels of FUR

Debruyne, F. M. J., Bloemers, H. P. J., and Van de Ven, W. J. M. (1987) J. Clin. Inuest. 80, 1545-1549

m~~~ in the brain regions were not simply due to a greater 14. Ban, P. J., Mason, 0. B., Landsberg, K. E., Wow, P. A., Kiefer, M. C., and

“packing density” of the cells in these areas, but appeared to be 15. Seidah, N. G., Gaspar, L., Mion, P., Marcinkiewicz, M., Mbikay, M., and Chre- Brake, A. J . (1991) DNA Cell Bid. 10,319-329

due to higher levels of FUR mRNA in individual cells. Alto- tien, M. (1990) DNA Cell Biol. 9, 415424

be considered to constitute a housekeeping gene, its expression Lett. 300,82-88 gether, these suggest that, the FUR gene may 17. Creemers, J. W. M., Roebroek, A. J. M., and Van de Ven, W. J. M. (1992)FEBS

level may differ depending on cell type or physiolo~cal 18. Nakayama, K., Kim, W. S., Toni, S., Hosaka, M., Nakagawa. T., Ikemizu, J.,

tion. I t is not unreasonab1e to assume that the strong 19. Seidah, N. G., Day, R., Hamelin, J., Gaspar, A. M., Collard, M. W., and Chre- of FUR gene expression as found in primary human non-small tien, M. (1992) Mol. Endocrinol. 6, 1559-1570 cell lung carcinomas, but not in small cell lung carcinomas (13), 20. Van den Ouweland, A. W. M., Van Groningen, J. J. M., Roebroek, A. J. M.,

can also be explained in such terms. It would be of interest to Onnekink, C., and Van de Ven, W. J. M. (1989) Nucleic Acids Res. 17, 7101-7102

16. Smeekens, S. P., and Steiner, D. F. (1990) J. Biol. Chem. 265, 29973000

Baba, T., and Murakami, K. (1992) J. Biol. Chem. 267,5897-5900

evaluate this in the context of promoter usage. Gene transcription from multiple promoters also allows post-

transcriptional control. By generating transcripts with differ- ent untranslated sequences, the stability or the translation efficiency of the mRNA can be affected (45). Promoters P1A and P1B of the human FUR gene are GC-rich and contain multiple SP1 sites that might be involved in maintaining the housekeep- ing levels of the enzyme. Transcription driven by these promot- ers are predicted to result in long GC-rich and highly struc- tured 5’-untranslated exons 1A and lB, respectively. mRNAs containing GC-rich leader sequences are usually not efficiently translated (451, so it is possible that the FUR mRNA isoform containing sequences for exon 1 is the major functional mRNA. Finally, the fact that the FUR gene possesses multiple promot- ers might ensure its ubiquitous expression pattern and reflect its relevance for cell growth and homeostasis in a large variety of eukaryotic cell types, if not all.

21. Van den Ouweland, A. M. W., Van Duijnhoven, J. L. P., Keizer, G. D., Dorssers,

22. Creemers, J. W. M., Roebroek, A. J . M., Van den Ouweland. A. W. M., Van L. C. J., and Van de Ven, W. J. M. (1990) Nucleic Acids Res. 18, 664

Duijnhoven, J. L. P., and Van de Ven, W. J . M. (1992) Mol. Bid. Life Sci. Adu. 11, 127-138

””

23. Misumi, Y., Sohda, M., and Ikehara, Y. (1990) Nucleic Acids Res. 18. 6719 24. Samhrook, J., Fritsch, E. F., and Maniatis, T. (1989) Molecular Cloning: a

Laboratory Manual, Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY

25. Kadonaga, J. T., Jones, K. A., and Tjian, R. (1986) Trends Biochem. Sci. 11, 2&23

26. Gorski, K., Carneiro, M., and Schibler, U. (1986) Cell 47,767-776 27. Cam, F. E., Need, L. R., and Chin, W. W. (1987) J. Biol. Chem. 262,981-987 28. Wolf, O., Kourides, I. A., and Gurr, J. A. (1987) J. Biol. Chem. 262, 16596

29. Carcamo, J., Maldonado, E., Cartes, P., Ahn, M.-H., Kasai, Y., Flint, J., and

30. Kasai, Y., Chen, H., and Flint, S. J. (1992) Mol. Cell. Biol. 12, 2884-2897 31. Robidoux, S., Gosselin, P., Harvey, M., Leclerc, S., and Guerin, S. L. (1992) Mol.

32. Kingsley, C., and Winoto, A. (1992) Mol. Cell. Biol. 12, 4251-4261 33. Udvadia, A. J., Rogers, K. T., Higgins, P. D. R, Murata, Y., Martin, K. H.,

Humphrey, P. A,, and Horowitz, J. M. (1993) Proc. Natl. Acad. Scr. U. S. A. 90.3265-3269

16603

Reinberg, D. (1990) Genes & Deu. 4, 1611-1622

Cell. Biol. 12, 37963806

REFERENCES 1. Seidah, N. G., and Chretien, M. (1992) Trends Endocrinol. Metab. 3, 133-140 2. Steiner, D. F., Smeekens, S. P., Ohagi, S., and Chan, S. J. (1992) J. Biol. Chem.

267,23435-23438 3. Van de Ven, W. J. M., Van Duijnhoven, J. L. P., and Roebroek, A. J. M. (1993)

Crit. Rev. Oncogen. 4, 115-136 4. Van de Ven, W. J. M., Voorberg, J., Fontijn, R., Pannekoek, H., Van den

Ouweland, A. M. W., Van Duijnhoven, J. L. P., Roebroek, A. J. M., and Siezen

5. Wise, R. J., Barr, P. J. , Wong, P. A., Kiefer, M. C., Brake, A. J., and Kaufman, R. J . (1990) Mol. Biol. Rep. 14, 265-275

6. Bresnahan, P. A., Leduc, R., Thomas, L., Thorner, J . , Gibson, H. L., Brake, A. R. J. (1990) Proc. Natl. Acad. Sci. U. S. A. 87, 9378-9382

38. Baumann, H., Morella, K. K., Campos, S. P., Cao, Z., and Jahreis, G . P. (1992)

39. Metz, R., and Ziff, E. (1991) Genes & Deu. 6, 1754-1766 40. Ramji, D. P., Vitelli, A., Tronche, F., Cortese, R., and Ciliherto, C. (1993)

41. Trautwein, C., Caelles, C., van der Geer, P., Hunter, T., Karin, M., and

42. Schibler, U., and Sierra, F. (1987) Annu. Reu. Genet. 21, 237-257 43. Hatsuzawa, K., Hosaka, M., Nakagawa, T., Nagase, M., Shoda, A,, Murakami,

44. Schiifer, M. K.-H., Day, R., Cullinan, W. E., Chretien, M., Seidah, N. G., and

45. Kozak, M. (1991) J. Cell Bid. 115,887-903

J. Bid. Chem. 267, 19744-19751

Nucleic Acids Res. 21, 289-294

Chojkier, M. (1993) Nature 364,544-547

K., and Nakayama, K. (1990) J. Biol. Chem. 265, 22075-22078

Watson, S. J. (1993) J. Neurosci. 3, 125S1279