Oncogene (1999) 18, 5923 ± 5935 ã 1999 Stockton Press All rights reserved 0950 ± 9232/99 $15.00 http://www.stockton-press.co.uk/onc The EGF/TGFa response element within the TGFa promoter consists of a multi-complex regulatory element

Rana Awwad1,2,4, Lisa E Humphrey1,4,5, Baskar Periyasamy1, William Scovell Jr1, Wenhui Li1,5, Kevin Coleman3, Mark Lynch2, Joan Carboni2, Michael G Brattain1,5 and Gillian M Howell*,1,4,5

1Department of Biochemistry and Molecular Biology, Medical College of Ohio, PO Box 10008, Toledo, Ohio, OH 43699-0008, USA; 2Department of Molecular Genetics, Oncology Drug Discovery, Bristol Myers Squibb, PO Box 4000, Princeton, New Jersey, NJ 08543-4000, USA; 3Box 123, P®zer Central Research, Eastern Point Road, Groton, Connecticut, CT 06340, USA

Autocrine TGFa is an important growth e€ector in the Introduction transformed phenotype. Growth stimulation of some colon cancer cells as well as other types of cancer cells TGFa (transforming growth factor a1) is a member of is e€ected by activation of the epidermal growth factor the family of mitogenic, autocrine growth factors receptor. Importantly, this receptor activation leads to which show homology to epidermal growth factor further stimulation of TGFa transcription and increased (EGF) and, like EGF, exert their biological e€ects via peptide synthesis. However, the molecular mechanism by the epidermal growth factor receptor (EGFr; Carpenter which TGFa transcription is activated is poorly under- and Cohen, 1990; Derynck, 1992; Todaro et al., 1990). stood. In this paper, we describe the localization of a cis- TGFa is secreted as a 50 amino acid peptide following sequence within the TGFa promoter which mediates this cleavage from a 160 (human) or 159 (rat) amino acid stimulation. This region contains parallel cis-acting precursor (Derynck, 1988) although secretion of larger elements which interact to regulate both basal and forms has been reported (Derynck, 1988; Watkins et EGF-induced TGFa expression. The well di€erentiated al., 1988). The precursor molecule can also exist as a colon carcinoma cell line designated FET was employed transmembrane glycoprotein which has biological in these studies. It produces autocrine TGFa but requires activity (Brachman et al., 1989; Anklesaria et al., exogenous EGF in the medium for optimal growth. 1990). Addition of EGF to FET cells maintained in the absence The use of neutralizing antibodies to both TGFa of EGF resulted in a 2 ± 3-fold increase of both TGF and the EGFr, and TGFa antisense approaches have promoter activity and endogenous TGFa mRNA at 4 h. all demonstrated that TGFa is an autocrine growth- This addition of EGF also stimulated protein synthesis. stimulatory factor in colon carcinoma (Markowitz et The use of deletion constructs of the TGFa promoter in al., 1990; Sizeland and Burgess, 1991, 1992; Ziober et chimeras with chloramphenicol acetyl transferase loca- al., 1993) as well as other tumors (Derynck et al., 1987; lized EGF-responsiveness to between 7247 and 7201 Ennis et al., 1989; Imanishi et al., 1989; Malden et al., within the TGFa promoter. A 25 bp sequence within this 1989; Lee et al., 1991). TGFa has also been shown to region conferred EGF-responsiveness to heterologous be an autocrine growth factor in normal cells promoter constructs. Further use of deletion/mutation (Markowitz et al., 1990; Malden et al., 1989; Co€ey chimeric constructs revealed the presence of at least two et al., 1987; Madtes et al., 1988; Bates et al., 1990; interacting cis-elements, one binding a repressor activity Kudlow and Bjorge, 1990; Lewis et al., 1990) and and the other, an activator. Gel shift studies indicate the during development (Wilcox and Derynck, 1988). presence of distinct complexes representing activator and However, solid tumors generally produce higher levels repressor binding, which are positively modulated by of this growth factor than their normal counterparts EGF. The type and amount of complexes formed by (Derynck et al., 1987). these proteins interact to regulate both the basal activity Interestingly, activation of the EGFr by either and EGF-responsiveness of the TGFa promoter. The TGFa or EGF stimulates TGFa production (Co€ey interaction of an activator protein with an EGF- et al., 1987, 1992; Bjorge et al., 1989). Addition of responsive repressor may serve to regulate the level of TGFa or EGF to cultured cells including keratino- this progression-associated, transforming protein within cytes, breast and colon carcinoma cells results in a tight limits. time- and dose-dependent increase in TGFa mRNA production. Increased TGFa protein secretion into Keywords: colon carcinoma; EGF-response element; culture media is also observed with similar kinetics. TGFa promotor; transcription; transforming growth Activation of the EGFr has been implicated in this factora autoinduction as neutralizing antibodies to EGFr inhibit both TGFa accumulation into the media and the growth of many types of tumor cells in vitro (Markowitz et al., 1990; Ennis et al., 1989; Imanishi et *Correspondence: GM Howell al., 1989). Furthermore, overexpression of TGFa in 4 These authors contributed equally to this work the weakly tumorigenic colon carcinoma cell line GEO 5 Current address: Department of Surgery, University of Texas by stable transfection of a TGFa expression vector Health Science Center at San Antonio, 7703 Floyd Curl Drive, San Antonio, Texas, TX 78284, USA resulted in the isolation of clones with increased Received 30 June 1998; revised 10 May 1999; accepted 12 May 1999 tumorigenicity both in vitro and in vivo (Ziober et al., Multi-complex regulation of the TGFa promoter by EGF/TGFa RAwwadet al 5924 1993). Signi®cantly, the increased TGFa expression workers (Raja et al., 1991), using TGFa promoter- from the vector resulted in induction of endogenous luciferase constructs, reported evidence indicating that TGFa mRNA levels. EGF-responsiveness was located in the general region Despite the importance of TGFa autocrine activity of 7373 to 759 bp of the TGFa promoter, but and its consequent autoinduction as a growth speci®c elements were not identi®ed, and no known stimulant in cancer cells, little is known of the consensus motifs are present in this region. Further- molecular mechanisms controlling TGFa expression, more, as this area of the TGFa promoter shows no particularly with regard to control of transcription. homology to other known EGF-responsive consensus In order to study TGFa transcriptional regulation, sequences, it seemed likely that the EGF/TGFa we cloned a 2813 bp fragment of the human TGFa response element in this gene would represent a promoter (Saeki et al., 1991). This fragment extended novel response element. Therefore, starting with a the TGFa promoter characterization by 1673 bp in series of deletion TGFa-CAT constructs ranging from the 5' direction over that reported by a previous 7370 to 7201, we undertook studies to identify this group (Jakobovits et al., 1988). Where sequence element. overlap occurs, the sequence is identical. The rat The well di€erentiated FET colon carcinoma cell TGFa promoter has also been cloned and is highly line was employed for these studies. The FET cell homologous to the human TGFa promoter (Blasband line was derived from a well di€erentiated primary et al., 1990). The notable features of the TGFa tumor and has retained di€erentiated characteristics promoter are that it is devoid of any CCAAT or in vitro (Brattain et al., 1984; Chantret et al., 1988; TATA box sequences. There are several Sp-1 sites Mulder and Brattain, 1989a,b; Wan et al., 1988). It located within the proximal 600 bp of DNA. There has been adapted to growth in a completely are several potential AP-2 sites located in the more chemically de®ned medium (Brattain et al., 1984). distal regions of the promoter between 7375 and This cell line expresses TGFa and EGFr and 7850 (Jakobovits et al., 1988). Since EGFr responds proliferatively to exogenous EGF (Mulder activation stimulates TGFa expression, the character- and Brattain, 1989a,b; Wan et al., 1988). While ization of the EGF/TGFa response element within exogenous EGF is not an absolute requirement for the human TGFa promoter is important in under- growth due to autocrine TGFa activity, optimal standing the control of this gene. Raja and co- growth is achieved only when EGF is present in the

Figure 1 Induction of TGFa transcriptional activity and mRNA by EGF. (a) FET clone 3, stably transfected with p2813-CAT, was grown to 60% con¯uency in serum-free medium minus EGF. Cells were then changed to serum-free medium containing 10 ng/ ml EGF and harvested at 1, 4 and 24 h for assay of CAT activity as described in Materials and methods. (b) Parental FET cells adapted to growth in serum-free medium without EGF were switched to medium containing 10 ng/ml EGF and harvested at 1, 4 and 24 h later for RNA quantitation by RNase protection analysis as described in Materials and methods. (c) Quantitation of the CAT assay Multi-complex regulation of the TGFa promoter by EGF/TGFa RAwwadet al 5925 media (Mulder and Brittain, 1989b). Therefore the Results FET cell line represents an excellent model to study regulation of the TGFa promoter by activated Kinetics of induction of TGFa transcription by EGF EGFr. In this report, we describe the use of deletion Initial studies were performed to con®rm the EGF- fragments of the TGFa promoter in chloramphenicol transcriptional responsiveness of the TGFa promoter acetyltransferase (CAT) reporter-gene chimeras to in the FET cell line. The largest TGFa promoter localize the EGF/TGFa response element. Subse- construct p2813-CAT (containing the 2813 bp of quently, we de®ned a 47 bp region which conferred TGFa promoter 5' to the translation start site) was EGF-responsiveness to heterologous promoter reporter stably transfected into FET cells. Following Neo constructs. Further analysis of this sequence identi®ed selection and subsequent testing for CAT activity, 25 bp of promoter which was responsive to EGF. clones were isolated and some of these cells were Moreover, mutation/deletion analysis of this sequence adapted to growth in serum-free medium lacking EGF. identi®ed both putative repressor and activator Experiments were performed with a typical clone, functions in this region. designated FET 3. Results obtained with this clone were similar to other clones stably transfected with p2813-CAT. The addition of EGF (10 ng/ml) to these cells resulted in very little change in TGFa promoter activity at 1 h but by 4 h CAT activity was increased 3 ± 4-fold (Figure 1a,c) and remained elevated at 24 h. In order to con®rm that these changes in promoter activity re¯ected changes in endogenous TGFa mRNA expression, we examined TGFa mRNA expression under similar conditions in the parental cell line. Following the addition of EGF to the parental cell line routinely grown in serum-free medium without EGF, the TGFa mRNA level showed little change at 1 h but was increased 2 ± 3.5-fold at 4 h (Figure 1b). The TGFa mRNA level returned to basal by 24 h. This apparent discrepancy in TGFa promoter activity versus TGFa mRNA expression may be explained by the extreme stability of the CAT mRNA. This responsive- ness to EGF is also re¯ected by protein levels in FET cells. FET cells maintained in serum-free medium minus EGF express only 73+10 pg/ml of TGFa per 106 cells in 48 h (6+s.e.mean; n=3). However, the addition of EGF results in increased TGFa secretion to 274+39 pg/ml per 106 cells over a 48 h period. Therefore, the changes in activity of the TGFa promoter-CAT chimeras re¯ect changes seen in both the endogenous TGFa mRNA and protein levels of FET cells di€erentially exposed to EGF and therefore, activation of the EGFr leads to transcriptional up- regulation of TGFa.

Localization of the EGF/TGFa response element As we had established that the TGFa promoter within FET cells was transcriptionally responsive to EGF, we then proceeded to de®ne the EGF-responsive region

Table 1 Fold-induction of TGF CAT construct activity by EGF Fold-di€erence in CAT activity in the presence CAT versus absence of EGF construct (Mean+s.e.mean; n=3) p370 2.75+0.87 p343 2.26+0.63 p247 2.44+0.82 p201 0.98+0.34 Figure 2 Localization of the EGF-response element. The The CAT assay data from cells plated and grown in serum-free deletion constructs p201 (a), p247 (b), p343 (c) and p370-CAT medium with (+) or without (7) EGF was quantitated on an (d) were transfected into FET cells adapted to continuous growth Ambis. The p370-, p343- and p247-CAT constructs showed in serum-free medium minus EGF. Following transfection, half signi®cant induction in the presence of EGF (P50.05; denoted by the cells were switched to serum-free medium containing EGF asterisk). The p201-CAT construct showed similar activity in cells and harvested for CAT assays at 24 h later grown + or 7 EGF Multi-complex regulation of the TGFa promoter by EGF/TGFa RAwwadet al 5926 within the TGFa promoter. In order to de®ne the In order to further de®ne the EGF-responsive region of the TGFa promoter involved in EGF/TGFa region of this promoter, two overlapping oligos responsiveness, a series of deletion constructs were spanning this region were synthesized. These se- transiently transfected into parental FET cells grown quences contained 7246 to 7220 designated (oligos continuously in the absence of EGF. Exogenous EGF 1/2) and 7225 to 7199 (designated oligos 3/4) of the was added for the indicated times. An RSV-b- TGFa sequence. As a control, oligonucleotides galactosidase construct was co-transfected to normal- corresponding to the region 7202 to 7176 ize for transfection eciency. A typical result of such (designated oligos 5/6) which lies downstream of the an experiment is shown in Figure 2. The TGFa EGF-responsive region were synthesized. These con- promoter deletion constructs p370-, p343- and p247- structs were also tested for their ability to confer CAT each show signi®cantly higher activity (P50.05, EGF-responsiveness to the heterologous tk promoter. n=3) in the presence of exogenous EGF than in its All three oligo sequences conferred some increase in absence (summarized in Table 1). In contrast, the p201- basal activity of the pBLCAT2 vector. However, only CAT, construct shows the same level of CAT activity the construct pBL-(3/4)-tk-CAT showed increased in the presence or absence of EGF. Consequently, the CAT activity in the presence of EGF (Figure 3). EGF/TGFa response element appeared to be localized The growth factor produced a 2 ± 2.5-fold increase in between 7201 and 7247 within the TGFa promoter, CAT activity. The activity of the pBL-(1/2)-tk-CAT placing it in the same region as described by Raja et al. and pBL-(5/6)-tk-CAT constructs showed no signifi- (1991). It is also evident from Figure 2 that the basal cant di€erence in activity in the presence or absence of level of transcription of the p201-CAT construct is EGF (Table 2). much reduced compared to the p247-, p343- and p370- In order to further investigate the properties of this CAT constructs. EGF-responsive region, plasmids were isolated in which the oligo pair 3/4 was inserted in reverse The EGF/TGFa cis-element confers EGF-responsiveness to a heterologous tk promoter

The region between 7201 and 7247 of the TGFa Table 2 Fold-induction of heterologous TK-CAT constructs promoter contains no suitable restriction enzyme sites Fold induction to allow further subcloning. Therefore, this 47 bp Constructs (mean+s.e.mean) sequence was synthesized and cloned just upstream of the tk promoter in the pBLCAT2 vector as described Control pBLCAT2 1.45+0.13 in Materials and methods. This 47 bp sequence PBL-3/4-CAT 2.40+0.27* PBL-5D-CAT 1.76+0.06 conferred EGF-responsiveness to the heterologous tk PBL-13R-CAT 0.96+0.04 promoter (data not shown). This sequence produced a PBL-1/2-CAT 1.23+0.15 2 ± 2.5-fold increase in CAT activity while the parental PBL-5/6-CAT 0.90+0.01 pBLCAT2 vector showed no response. Mean+s.e.mean n=3; *Denotes di€erences by t-test

Figure 3 Further localization and characterization of the EGF-response element. The heterologous construct p3/4-tk-CAT containing 7201 to 7225 of the TGFa promoter was transfected into FET cells maintained in serum-free medium minus EGF(7). Following transfection, half the cells were switched to medium containing EGF(+) and harvested for CAT assay 24 h later. The e€ect of EGF on the parental vector pBLCAT2 is shown as a control. The e€ect of EGF on two tandem sequences of oligonucleotide 3/4 inserted upstream of the promoter, p-5D-tk-CAT, is also shown Multi-complex regulation of the TGFa promoter by EGF/TGFa RAwwadet al 5927 orientation to the tk promoter (designated pBL-13R- correct orientation (designated pBL-5D-tk-CAT). tk-CAT) and in which two sets of the oligos were Notably, multiple insertions of the 25 bp EGF- inserted upstream of the tk promoter, both in the responsive region do not increase the responsiveness

Figure 4 E€ect of deletion of various sequence motifs within the EGF-response element on basal activity and EGF-responsiveness of the TGFa promoter construct p247-CAT. (a) CAT assay of the Null and GAGG deletion motifs. The CAT activity of the p247- CAT plasmid and the deletion vectors described (b) was measured in the presence (+) or absence (7) of EGF for 4 h following transient transfection into FET cells maintained in serum-free medium as described in Materials and methods. (b) Graphical representation of the basal activity and EGF-response of the deletion motif p247-CAT constructs. The sequence motifs deleted in each construct are shown on the left hand side of the panel. The boxed bp represent the deleted motif. The basal activity and EGF- response of each construct is normalized to the activity of the p247-CAT construct in the absence of EGF (activity=100%). Data represents 6+s.e.mean (n=3) Multi-complex regulation of the TGFa promoter by EGF/TGFa RAwwadet al 5928 to EGF (Figure 3 and Table 2). A double-insertion of E€ect of deletion of sequence motifs within the context the sequence results in only a 1.7 ± 2.0-fold increase in of the TGFa promoter CAT activity upon stimulation with EGF. This response is of the same order of magnitude but If the sequence contained within oligonucleotide 3/4 slightly decreased compared to a single sequence. represented the EGF-response element, then deletion of Furthermore, when inserted in reverse orientation, the this sequence from an EGF-responsive TGFa promoter basal activity of the sequence is reduced approxi- construct should abolish the EGF-response. Using the mately 50% compared to that of the correct Ex-Site protocol (Stratagene) a p247-CAT construct in orientation, and the reverse sequence does not which 18 bp of the sequence within oligonucleotide 3/4 respond to exogenous EGF (Table 2). was deleted (Figure 4b), designated the p247-(null)-

Figure 5 Graphical representation of the basal activity and EGF-responsiveness of the deleted/mutated EGF-response element- heterologous constructs. The deletion/mutation sequences within the 25 bp EGF-response element are shown on the left hand side of the ®gure. These deletion/mutation sequences were cloned upstream of the minimal AML65 promoter in the pAML65-CAT vector. These constructs were transiently transfected into FET cells as described in Materials and methods and the CAT activity was assayed following change of medium with (+) or without (7) EGF for 4 h. CAT activity was normalized to that of the wild-type EGF-response element (oligonucleotide 3/4) in the absence of EGF (activity=100%). The data represents mean+s.e.mean (n=3). The activity of the pAML65-CAT construct is shown to verify the non-EGF-responsiveness of this parental construct. The stars indicate signi®cant (P50.05) response to EGF Multi-complex regulation of the TGFa promoter by EGF/TGFa RAwwadet al 5929 CAT construct, was generated. This null construct generated (Figure 4b). Deletion of the sequence showed decreased basal activity compared to the p247- TGACGG (designated the p247-(TGAC)-CAT con- CAT vector in serum free medium minus EGF and struct) or the TAGC sequence (designated the p247- showed no responsiveness to exogenous EGF (Figure (TAGC)-CAT construct) resulted in constructs which 4a,b). showed very little e€ect on basal activity in the A series of sequential deletions throughout the absence of EGF (a non-signi®cant decrease) but which 25 bp sequence of oligonucleotide 3/4 were also no longer showed responsiveness to exogenous EGF

Figure 6 Gel shift analysis of nuclear constructs prepared from EGF-treated FET cells. (a) Nuclear extracts were prepared from FET cells treated with EGF for 1 or 4 h or non-treated controls (Oh) as described in Materials and methods. These extracts were run against labeled oligonucleotide 3/4 and the -GAGG, -TGAC and -TAGC deletion sequences described in Figure 5. There are ®ve major bands of interest, labeled 1 ± 5. The control lane of oligonucleotide run without protein is labeled OP. (b) Competition study. Nuclear extract from EGF-treated FET cells was run against labeled oligonucleotide 3/4 in the presence of increasing amounts of excess, unlabeled oligonucleotide 3/4. Bands 1 ± 4 are completely competed away while band 5 is not. The lanes in which no competitor cold oligonucleotide is added are labeled OC. The protein free probe lane is again labeled OP Multi-complex regulation of the TGFa promoter by EGF/TGFa RAwwadet al 5930 (Figure 4b). The most striking e€ect on TGFa further signi®cant increase in promoter activity promoter activity was seen when the sequence occurred. Therefore, the EGF-responsive region of GCGAGGAGG was deleted (the p247-(GAGG)- the TGFa promoter appeared to contain both an CAT construct). This construct showed a signi®cant EGF-responsive activator which might also be 2 ± 2.5-fold increase in basal activity (Figure 4a,b). involved in the maintenance of basal promoter Moreover, upon addition of exogenous EGF, a activity and a repressor of TGFa promoter activity.

Figure 7 Competition assay to assess binding speci®city putative activator and repressor binding. (a) Sequence of short oligonucleotides used in the competition assay. (b) Competition assay of putative activator binding. Activator binding to 32P-labeled deletion oligonucleotide (band e) was competed with a 10 ± 50-fold excess of the indicated unlabeled competitor sequences. Nuclear extracts were prepared from FET cells stimulated with EGF for 4 h prior to harvest. OC-no competitive oligonucleotide lane; OP- no protein lane. (c) Competition assay of putative repressor binding, Repressor binding to 32P-labeled TAGC-deletion oligonucleotide (bands 1 and 3) was competed with a 10 ± 50-fold excess of indicated competitor sequences Multi-complex regulation of the TGFa promoter by EGF/TGFa RAwwadet al 5931 band 2 and band 4 remained. Band 2 showed marked E€ect of sequential 4 bp deletion within oligonucleotide EGF-induction under these conditions. When the 3/4 on heterologous promoter activity TAGC deletion oligonucleotide was used as probe, If the TGFa promoter does consist of adjacent bands 1, 3 and 4 remained. In the case of the TGAC activator and repressor sequences, these functions deletion, again bands 1, 3 and 4 were seen, but bands 1 should also be apparent in the context of heterologous and 3 were markedly reduced in intensity. As band 4 is promoter constructs. For these experiments, a new present with all the deletion oligonucleotides whether heterologous promoter construct containing 65 bp of EGF-responsive or not, it probably represents a basal the AML65 promoter was made as described in transcription factor. To further test the binding Materials and methods due to the high basal activity speci®city of bands 1, 2 and 3, we ran competition of the pBLCAT2 vector. This new heterologous studies using the 32P-labeled GAGG and TAGC promoter contains only a TATA sequence. Therefore, deletion oligonucleotides against smaller promoter it has lower basal activity which facilitated not only fragments from the 25 bp EGF-responsive region detection of the EGF-response (which is generally only (Figure 7a). Band 2, which binds the GAGG deletion of approximately twofold magnitude) but also allowed oligonucleotide and represents the putative activator for ready identi®cation of loss of repressor sequences. binding activity is not competed by oligonucleotide D

On transfection into the FET cell line, the pAML65- (Figures 7b and 8) which contains the repetitive CAT construct showed no response to EGF when GAGG sequence (CGAGGAGGTG). Interestingly, assayed over a 24 h time course (data not shown). band 2 is only partially competed by the oligonucleo- The sequence of the 4 bp deletions is shown on the tides B and C, which separately contain the TGAC and left hand side of Figure 5. Oligonucleotides were also TAGC motifs (sequences GGAGTGACG and synthesized with the corresponding sequence mutated CGGTAGCCG, respectively). However, oligonucleo- to show that deletions were not spatially disrupting the tide AE (which is a slightly shorter version of the sequence and so causing the e€ects. The three 4 bp GAGG deletion oligonucleotide) and contains both the motifs deleted or mutated within oligonucleotide 3/4 TGAC and TAGC motifs in their correct spatial were based on the sequences of the p247-CAT relationship, completely inhibits binding of band 2 deletions, shown in Figure 4b and designated (Figure 7b). (-GAGG), (-TGAC) and (-TAGC), respectively. The In Figure 7c, competition of the 32P-labeled TAGC results were similar to the e€ect of deletions within the deletion oligonucleotide against oligonucleotide C and p247-CAT promoter construct and are presented D is presented. The TAGC oligonucleotide binds bands graphically on the righthand side of Figure 5. 1 and 3, which represent putative repressor binding. Speci®cally, deletion or mutation of the GAGG Oligonucleotide D (containing the putative repressor sequence again conferred an approximately twofold binding sequence CGAGGAGGTG) completely com- signi®cant increase on basal activity of the hetero- petes away band 1, while band 3 binding is partially logous construct in the absence of EGF. Addition of inhibited. In contrast, oligonucleotide C exogenous EGF resulted in a further signi®cant (CGGTAGCCG) which does not contain the GAGG increase in CAT activity. Deletion or mutation of the repressor motif, does not compete for either band 1 or TGAC or TAGC sequences had no signi®cant e€ect on 3. These studies suggest that the binding in band 1 basal activity of the construct, although TGAC clearly represents repressor binding while band 3 also deletion caused a slight increase and TAGC deletion play a role in repressor function. These gel shift studies a slight decrease. Signi®cantly, deletion/mutation of clearly indicate the presence of activator and repressor either the TGAC or TAGC motifs resulted in the loss binding which each show distinct binding speci®cities. of EGF-responsiveness.

Discussion Gel shift analysis In order to begin to study the trans-acting factors TGFa is an autocrine growth regulator in normal cells binding to this EGF-responsive element, and to and overexpression of this growth factor contributes to determine whether this binding is regulated by EGF, the loss of growth regulation and development of nuclear extracts were prepared from FET cells grown tumorigenicity in the transformed phenotype in colon in the absence of EGF or treated with EGF for 1 or cancer cells (Ziober et al., 1993). The importance of 4 h for use in gel shift studies (Figure 6a). When EGFr activation in the regulation of TGFa, either by oligonucleotide 3/4 is used as probe several bands are EGF or TGFa itself, has been established by the use of seen (labeled 1 to 5). The lowest mobility, highest EGFr antibodies (Markowitz et al., 1990; Ennis et al., molecular weight bands represented by bands 1, 2 and 1989; Imanishi et al., 1989). Although the TGFa 3 are particularly responsive to EGF, especially at 4 h promoter has been isolated, little is known about the following its addition. Band 4 shows a slight EGF- mechanism of response to EGFr activation. Therefore, response whereas band 5 does not. When completed studies were initiated to identify the DNA sequence with cold oligonucleotide 3/4, all but band 5 are involved. In this paper we report the localization of competed suggesting this band represents non-speci®c this EGF/TGFa response element to a 25 bp region binding (Figure 6b). between 7225 and 7201 of the TGFa promoter. This We also ran the nuclear extracts using the deletion ®nding is in agreement with previous ®ndings that series of oligonucleotides as probe to determine localized EGF-responsiveness to the ®rst 313 bp of the whether speci®c bands of protein binding were 5'-¯anking sequence of the TGFa promoter using a associated with speci®c motifs (Figure 6a). When the TGFa promoter-luciferase construct (Raja et al., 1991). GAGG deletion sequence was used as probe, only The ability of this 25 bp fragment to confer EGF- Multi-complex regulation of the TGFa promoter by EGF/TGFa RAwwadet al 5932

responsiveness to both the heterologous tk and AML65 EGF/TGFa stimulation of TGFa promoter activity. promoters indicated that it is an enhancer-type Gel shift analysis reveals that both motifs are required element. The evidence with the ®rst 65 bp of the for optimal binding of the activator which is positively AML promoter is particularly convincing since this regulated by EGF. Interestingly, the gel shift competi- region contains only a TATA box element. tion binding studies on the activator binding re¯ect the The level of induction of TGFa mRNA by EGF TGFa promoter activity exhibited by the deletion reported in this study is in agreement with previous constructs. Optimal TGFa promoter activity and results (Howell et al., 1995; Co€ey et al., 1987, 1992; optimal activator protein binding is observed when Bjorge et al., 1989). Similarly, the gastrin gene which is the GAGG sequence is deleted leaving both the TAGC transcriptionally responsive to EGF, shows only a and TGAC motifs (the GAGG deletion oligonucleo- maximal threefold increase in both mRNA levels and tide). This oligonucleotide has high basal activity, and promoter activity following stimulation by the growth as activator binding is responsive to EGF, activity is factor (Merchant et al., 1991). Although EGF e€ects a further stimulated by EGF. In contrast, neither the wide range of biological action, it produces only TGAC or TAGC notifs alone compete e€ectively for relatively small increases in individual gene activity. activator binding and so both the TAGC and TGAC This EGF response element, as well as regulating deletion oligonucleotides have very low basal activity responses to exogenous EGF, must be important in with very little or no e€ect of EGF. Thus, both the TGFa autoregulation, and so constitutes part of the TAGC and TGAC motifs are required for activator TGFa autocrine loop. The transcription factors binding and the presence of at least a part of the binding to these sites therefore represent downstream TGAC motif is also necessary for repressor binding. e€ector molecules of EGFr activation, both due to This activator must also act to o€set the e€ect of the exogenous EGF and endogenous TGFa. Whether these repressor and maintain some low level of TGFa transcription factors regulate other genes involved in promoter activity. the mitogenic response to EGFr activation remains to The multi-complex nature of this EGF-response be determined. element probably accounts for the fact that this cis- This study also suggests the importance of the sequence does not show all the properties of an TGFa-autoregulatory element in regulating basal enhancer. While this element can confer EGF- expression of the TGFa promoter. Deletion to 201 bp responsiveness to a heterologous promoter, it does of TGFa promoter which removes the adjacent EGF/ not function in reverse orientation. Moreover, two of TGFa response element, signi®cantly decreases the the EGF-response elements in tandem do not increase level of basal CAT activity. Other groups have also the level of response to exogenous EGF, probably reported that deletions which remove this element because as well as aligning two activator sites, two result in CAT constructs with reduced basal activity repressor sites are also brought into tandem. This (Jakobovits et al., 1988; Raja et al., 1991). Therefore, multi-complex regulation of a growth factor gene is not this EGF/TGFa response element may also be unique. Regulation of the c-fos gene by mitogens is important in maintaining a level of TGFa basal also e€ected by interactions of adjacent transcription transcription which facilitates the action of more factors (Dalton et al., 1993; Treisman, 1994; Price et distal (`upstream') elements. Deletion of the ®rst al., 1995). Transcription of the c-fos gene is stimulated 370 bp of DNA to generate a construct extending by the serum response factor (SRF) bound to the SRF from 71140 to 7370 yields very low CAT activity site and by ternary complex factor (TCF) bound at an (only 5% of the total activity) despite the presence of adjacent site which contains an Ets transcription factor several potential Sp-1 and AP-2 sites within this motif. TCF cannot bind independently of SRF. While fragment (Jakobovits et al., 1988). The remaining the SRF alone can mediate response to serum, the low-level basal transcription of the p201-CAT con- presence of TCF is required for activation of c-fos struct is probably due to Sp-1 binding at the two transcription in response to individual growth factor potential Sp-1 consensus sequences within this region. signaling pathways. Therefore, a complex array of cis- The presence of Sp-1 sites within this region of the rat and trans-acting factors presumably allows for the TGFa promoter which are required for basal in¯uence of many pathways to converge at a single transcription has been reported previously (Chen et regulatory region. These e€ects on the trans-acting al., 1992). factors could be both transcriptional and/or post- The results of this study indicate that this EGF- translational a€ecting, for example, phosphorylation response element comprises binding sites for both a state. In the case of the TGFa gene, interaction transcriptional activator and repressor. The GAGG between transcriptional activators and repressors in motif within the EGF-response element appears to this region might also be a mechanism for regulating bind a repressor activity which appears to control basal expression of this potentially transforming gene within expression of the TGFa promoter. Gel shift analysis tight limits. indicates that protein binding in the bands which The DNA sequence of the EGF/TGFa responsive appear to represent the repressor complexes is also region of the TGFa promoter does not show homology EGF-regulated. Therefore, as well as functioning to to other known EGF/TGFa response elements which maintain low basal levels of TGFa, the repressor may have been identi®ed. Other EGF/TGFa responsive also represent a feedback mechanism to regulate the elements include the AP-1 transcription factor which level of response to TGFa to exogenous growth factors has the minimal consensus sequence TGACTCA and as well as the positive stimulation of TGFa promoter binds jun family homodimers or jun ± fos heterodimers activity by the autocrine pathway. (Angel et al., 1987; Fisch et al., 1989; Curran and We have also identi®ed two motifs within the 25 bp Franza, 1988; McDonnell et al., 1990). Moreover, GC- sequence, TGAC and TAGC, which are necessary for rich motifs have also been reported to mediate EGF- Multi-complex regulation of the TGFa promoter by EGF/TGFa RAwwadet al 5933 responsiveness as in the Egr-1 (Cao et al., 1990; Cell transfection and isolation of stable transfectants Sukhatme, 1991) and gastrin (Merchant et al., 1991) Con¯uent FET cells were harvested by trypsinization and genes. These GC-rich motifs show weak homology to resuspended at 26107 cells per 800 ml of serum-free medium the consensus sequence for the transcription factor without EGF and transferred to a 0.4 cm electro gap cuvette. AP-2 which can mediate induction by both the cyclic Cells were electroporated at 250 volts and 960 mFarads in a AMP and protein kinase C pathways (Imagawa et al., BioRad gene pulser with 50 mg TGFa-promoter-CAT 1987). However, it is unlikely that they are true AP-2 construct. In transient transfection assays, an RSV-b- sites as they are only weakly induced by phorbol esters. galactosidase reporter gene (20 mg) was co-transfected to Another class of EGF-response element is that shared monitor transfection eciency). Transfected cells were plated by the prolactin and tyrosine hydroxylase genes in two dishes in serum-free medium without EGF and changed 12 h later. Cells were harvested at 36 to 48 h after (Elsholtz et al., 1986; Lewis and Chikaraiski, 1987). transfection following addition of EGF for the stated times. The two motifs in these genes are very similar but show Stable clones containing TGFa-promoter CAT constructs no homology to the other EGF-response elements or were transfected as described above with 30 mg vector and the EGF-response element reported here. 5 mg of an SV-40-Neomycin (Neo) plasmid and plated at The reason for the diversity of EGF-responsive 1 : 20 ± 1 : 30 dilution. Two days following plating, cells were motifs remains to be determined. The AP-1 and switched to medium containing 600 mg/ml Geneticin (G418; cAMP-responsive elements control a wide range of Gibco) to allow selection for Neo resistant clones. These gene responses by binding homodimers or heterodimers clones were expanded and tested for CAT expression. of transcription factor families (Curran and Franza, 1988; Zi€, 1990; Brindle and Montminy, 1992). CAT assays Activation of the EGFr stimulates a wide range of responses, encompassing proliferative, di€erentiation Following three washes in phosphate bu€ered saline, cells and developmental e€ects. Furthermore, the EGF were harvested in 1 ml of TEN bu€er (40 mM Tris, pH 7.4, family encompasses a growing number of members 1mM EDTA and 150 mM NaCl) and following centrifuga- tion, resuspended in 100 ml of 250 mM Tris-HCl, pH 7.8 including amphiregulin and cripto which also interact containing 15% glycerol (v/v) and lysed by three cycles of with the EGFr (Carpenter and Cohen, 1990; Derynck, freeze-thawing. Protein concentration in the lysate was 1992; Todaro et al., 1990). Diverse EGF/TGFa cis- determined by microassay using bicinchoninic acid reagent elements would be one way to encompass the range of (Pierce). The b-galactosidase activity was determined by EGF/TGFa (and indeed other EGF family members) macroassay as described by Promega. Cell lysates were added e€ects on gene transcription. This would certainly to the CAT assay corrected for b-galactosidase activity. The allow for diversity in temporal induction of e€ects, CAT assay was performed in a ®nal volume of 164 mlin especially those involved in precise cell-cycle regulation. 250 mM Tris-HCl, pH 7.8, containing 0.2 mCi [14C]- chloramphenicol (14C-D-threo[dichloroacetyl-1-14C] chloram- phenicol; 55 mCi/mmol; Amersham) and 3.6 mM acetyl CoA (Pharmacia) overnight at 378C. The acetylated products and Materials and methods remaining [14C]-chloramphenicol were extracted with ethyl acetate and analysed by thin layer chromatography using a Cell culture chloroform/methanol solvent system. The FET cell line is routinely maintained in a de®ned serum- free medium as previously described (Mulder and Britain, Design of the heterologous promoter-CAT constructs 1989b; Wan et al., 1988; Boyd et al., 1988). This consists of McCoy's 5A medium (Sigma) supplemented with amino The 47 bp of DNA sequence (both strands) corresponding to acids, pyruvate, antibiotics and the growth factors insulin 7201 to 7247 of the TGFa promoter were synthesized with (20 mg/ml; Sigma), transferrin (4 mg/ml; Sigma) and EGF BamHI `sticky ends' and were inserted in the BamHI site just (10 ng/ml; R&D Systems). However, in order to maximize upstream of the thymidine kinase tk promoter in the EGF-responsiveness, EGF was omitted from the medium in pBLCAT2 vector. Subsequently, three other sets of over- which cells were maintained and experiments were performed lapping oligos spanning 7246 to 7220 (sense and antisense after changing cultures to serum-free medium containing strand; oligo 1/2), 7225 to 7199 (oligo 3/4) and 7202 to EGF for the stated times. Working cultures were maintained 7176 (oligo 5/6) were synthesized according to the TGFa promoter sequence in reference 25 and also cloned into the at 378C in the presence of 5% CO2. Cells were subcultured using 0.25% trypsin in Joklik's medium (Gibco) containing pBLCAT2 vector. 0.1% EDTA and replated in serum-free medium. While the heterologous tk constructs based on the pBLCAT2 vector were used in initial characterizations of the EGF-response element, we switched to another construct Cloning of the TGFa promoter in the more detailed analyses. This was because of the high A normal human leukocyte genomic library (Clonetech) was background activity of the tk promoter. Therefore, we screend by plaque hybridization (Maniatis et al., 1982) with a synthesized the ®rst 65 bp of the adenonous major late (AML ) promoter (755 to +10) with BamHI and XhoI probe homologous to the TGFa coding sequence +104 to 65 +129 (based on the published TGFa cDNA sequence; ends. The promoter within the pBLCAT2 vector was excised with BamHI and XhoI and the AML sequence was inserted Derynck, 1988). A 2.8 kb PstI±SnaBI fragment containing 65 in this site, generating a vector designated pAML -CAT. the TGFa promoter sequence and the ®rst 125 bp of coding 65 sequence was isolated (Saeki et al., 1991). Following removal This 65 bp region contains only a TATA box and no other of the coding sequence with Ba131 exonuclease (to nucleotide known consensus regions. 74), the resulting fragment was cloned upstream of the bacterial CAT gene in the vector pGCAT, to give the largest RNA preparation construct designated p2813-CAT. By the use of appropriate restriction sites or the PCR technique, deletion fragments of Total RNA was isolated by the method of Chirgwin et al. the TGFa promoter were generated and inserted upstream of (1979) using a cesium tri¯uoroacetic acid gradient (Pharma- the bacterial CAT gene in the vector pGCAT-C. cia). Multi-complex regulation of the TGFa promoter by EGF/TGFa RAwwadet al 5934 RNase protection assay High speci®c activity double-stranded oligonucleotide labeled with 32P-dGTP by Klenow (Promega) was incubated To obtain a speci®c TGFa probe, a 306 bp EcoRI ± SphI with nuclear extracts in a ®nal volume of 20 ml as described fragment of TGFa cDNA was cloned into the pGEM 3Z(-) previously (Gorski et al., 1986). Following incubation on ice, vector (Promega). High speci®c activity riboprobe was the reaction mix was loaded onto a 4% polyacrylamide gel in synthesized in the presence of [32P]UTP (3000 Ci/Mmol; 25 m Tris-25 mM boric acid bu€er containing 0.5 mM NEN) by SP6-RNA polymerase. Following DNaseI diges- M EDTA and run at 150 ± 250 V at 48C (Scheidereit et al., tion of the DNA template, the riboprobe was isolated on a 1987). In order to test the binding speci®cities of the protein Nacs-52 column (Gibco BRL). complexes, competition experiments were performed. Increas- Total RNA (20 ± 40 mg) was hybridized to the riboprobe ing concentrations of unlabeled oligonucleotides (5 ± 100 ng) overnight in 5 M guanidine isothiocyanate (pH 7.0) contain- were added to incubations of 32P labeled probe (oligonucleo- ing 0.1 EDTA as described previously (Howell et al., M tide 3/4, deletion versions of 3/4) in the presence of nuclear 1995). Following RNase digestion of excess riboprobe, extract from EGF-treated cells. protected fragment was electrophoresed on a 6% urea- polyacrylamide sequencing gel. Equal loading of the gel was determined by simultaneous hybridization of the RNA with an actin probe transcribed from the BamHI ± HindIII fragment of a-actin (Enoch et al., 1986) inserted in the pGEM 3Z(-) vector and linearized with HinfI. This probe Abbreviations yields a 145 bp protected fragment when hybridized with TGFa, transforming growth factor a; EGF, epidermal human actin. growth factor; EGFr, epidermal growth factor receptor; bp, base pair(s); CAT, chloramphenicol acetyltransferase; Protein assay tk, thymidine kinase; cAMP, cyclic adenosine monopho- sphate;Neo,Neomycn;PCR,polymerasechainreaction. The TGFa protein was quantitated using the quantitative ELISA kit from Oncogene Science. Acknowledgements Gel retardation assay This work was supported by National Institutes of Health Nuclei were prepared from FET cells by hypotonic lysis as Grants CA34432 and CA54807. Work performed in partial described by Dignam et al. (1983). Nuclear proteins were salt- ful®llment of the requirements for the PhD degree for R extracted and concentrated as described by Gorski et al. Awwad. We thank Suzanne Payne, Jenny Zak and Lorraine Gilmore for preparation of the manuscript and (1986) and aliquots were snap-frozen in liquid N2 and stored at 7808C. Barbara Young for excellent technical assistance.

References

Angel P, Imagawa M, Chiu R, Stein B, Imbra RJ, Co€ey Jr RJ, Graves-Deal R, Dempsey PJ, Whitehead RH Rahmsdorf HJ, Jonat C, Herrlich P and Karin M. and Pittelkow MR. (1992). Cell Growth Di€er., 3, 347 ± (1987). Cell, 49, 729 ± 739. 354. Anklesaria P, Teizido J, Laiho M, Pierce JH, Greenberger JS Co€ey RJ, Derynck R, Wilcox JN, Bringman TS, Goustin and Massague J. (1990). Proc. Natl. Acad. Sci. USA, 87, AS, Moses HL and Pittelkow MR. (1987). Nature, 328, 3289 ± 3293. 817 ± 820. Bates SE, Valverius EM, Ennis BW, Bronzert DA, Sheridan Curran T and Franza Jr BR. (1988). Cell, 55, 395 ± 397. JP, Stampfer MR, Mendelsohn J, Lippman ME and Dalton S, Marais R, Wynne J and Treisman R. (1993). Dickson RB. (1990). Endocrinology, 126, 596 ± 607. Biological Sciences, 340, 325 ± 332. Bjorge JD, Paterson AJ and Kudlow JE. (1989). J. Biol. Derynck R. (1988). Cell, 54, 593 ± 595. Chem., 264, 4021 ± 4027. Derynck R. (1992). Adv. Cancer Res., 58, 27 ± 52. Blasband AJ, Rogers KT, Chen X, Azizkhan JC and Lee DC. Derynck R, Goeddel DV, Ullrich A, Gutterman JU, (1990). Mol. Cell. Biol., 10, 2111 ± 2121. Williams RD, Bringman TS and Berger WH. (1987). Boyd DD, Levine AE, Brattain DE, McKnight MK and Cancer Res., 47, 707 ± 712. Brattain MG. (1988). Cancer Res., 48, 2469 ± 2474. Dignam JD, Lebowitz RM and Roeder RG. (1983). Nucl. Brachman R, Lindquist RB, Nagashima M, Kohr W, Lipari Acids. Res., 11, 1475 ± 1489. T, Napier M and Derynck R. (1989). Cell, 56, 691 ± 700. Elsholtz HP, Mangalam HJ, Potter E, Albert VR, Supowit S, Brattain MG, Levine AE, Chakrabarty S, Yeoman LC, Evans RM and Rosenfeld MG. (1986). Science, 234, Willson JKV and Long BH. (1984). Cancer Metastasis 1552 ± 1557. Rev., 3, 177 ± 191. Ennis BW, Valverius EM, Bates SE, Lippman ME, Bellot F, Brindle PK and Montminy MR. (1992). Curr. Opin. Genet. Kris R, Schlessinger J, Masui H, Goldenberg A, Dev., 2, 199 ± 204. Mendelsohn J and Dickson RB. (1989). Mol. Endo., 3, Cao X, Koski RA, Gashler A, McKiernan M, Morris CF, 1830 ± 1838. Ga€ney R, Hay RV and Sukhatme VP. (1990). Mol. Cell. Enoch T, Zinn K and Maniatis T. (1986). Mol. Cell. Biol., 6, Biol., 10, 1931 ± 1939. 801 ± 810. Carpenter G and Cohen SJ. (1990). J. Biol. Chem., 265, Fisch TM, Prywes R and Roeder RG. (1989). Mol. Cell. 7709 ± 7712. Biol., 9, 1327 ± 1331. Chantret I, Barbat A, Dussaulx E, Brattain MG and Gorski K, Carneiro M and Schibler U. (1986). Cell, 47, 767 ± Zweibaum A. (1988). Cancer Res., 48, 1936 ± 1942. 776. Chen X, Azizkhan JC and Lee DC. (1992). Oncogene, 7, Howell GM, Ziober BL, Humphrey LE, Willson JKV, Sun 1805 ± 1815. L, Lynch M and Brattain MG. (1995). J. Cell. Physiol., Chirgwin JM, Przybyla AE, MacKondla RJ and Rutter WJ. 162, 256 ± 265. (1979). Biochemistry, 18, 5294 ± 5299. Imagawa M, Chiu R and Karin M. (1987). Cell, 51, 251 ± 260. Multi-complex regulation of the TGFa promoter by EGF/TGFa RAwwadet al 5935 Imanishi K-I, Yamaguchi K, Kuranami M, Kyo E, Hozumi Price MA, Rogers AE and Treisman R. (1995). EMBO J., 14, T and Abe K. (1989). J. Natl. Cancer Inst., 81, 220 ± 223. 2589 ± 2601. Jakobovits EB, Schlokat U, Vannice JL, Derynck R and Raja RH, Paterson AJ, Shin TH and Kudlow JE. (1991). Levinson AD. (1988). Mol. Cell. Biol., 8, 5549 ± 5554. Mol. Endocr., 5, 514 ± 520. Kudlow JE and Bjorge JD. (1990). Cancer Biology, 1, 293 ± Saeki T, Cristiano A, Lynch MJ, Brattain M, Kim N, 302. Normanno N, Kenney N, Ciardiello F and Salomon DS. Lee LW, Raymond VW, Tsao M-S, Lee DC, Earp HS and (1991). Mol. Endocr., 5, 1955 ± 1963. Grisham JW. (1991). Cancer Res., 51, 5238 ± 5244. Scheidereit C, Heguy A and Roeder RG. (1987). Cell, 51, Lewis EJ and Chikaraishi DM. (1987). Mol. Cell. Biol., 7, 783 ± 793. 3332 ± 3336. Sizeland AM and Burgess AW. (1991). Mol. Cell. Biol., 11, Lewis JJ, Goldenring JR, Modlin IM and Co€ey RJ. (1990). 4005 ± 4014. Surgery, 108, 220 ± 227. Sizeland AM and Burgess AW. (1992). Mol. Biol. Cell, 3, Madtes DK, Raines EW, Sakariassen KS, Assoian RK, 1235 ± 1243. Sporn MB, Bell GI and Ross R. (1988). Cell, 53, 285 ± 293. Sukhatme VP. (1991). Am. J. Kidney Dis., 17, 615 ± 618. Malden LT, Novak U and Burgess AW. (1989). Int. J. Todaro GJ, Rose TM, Spooner CE, Shoyab M and Plowman Cancer, 43, 380 ± 384. GD. (1990). Sem. Cancer Biol., 1, 257 ± 263. Maniatis T, Fritsch EF and Sanbrook J. (1982). Molecular Treisman R. (1994). Curr. Opin. Genetics Dev., 4, 96 ± 101. cloning: A Laboratory Manual. Cold Spring Harbor Wan C-W, McKnight KM, Brattain DE, Brattain MG and Laboratory, NY. pp. 202 ± 203. Yeoman LC. (1988). Cancer Letts., 43, 139 ± 145. Markowitz SD, Molkentin K, Gerbic C, Jackson J, Stellato Watkins LF, Brattain MG and Levine AE. (1988). Cancer T and Willson JKV. (1990). J. Clin. Invest., 86, 356 ± 362. Letts., 40, 59 ± 70. McDonnell SE, Kerr LD and Matrisian LM. (1990). Mol. Wilcox JN and Derynck R. (1988). Mol. Cell. Biol., 8, 3415 ± Cell. Biol., 10, 4284 ± 4293. 3422. Merchant JL, Demediuk B and Brand SJ. (1991). Mol. Cell. Zi€ EB. (1990). Trends in Genetics, 6, 69 ± 72. Biol., 11, 2686 ± 2696. Ziober BL, Willson JKV, Humphrey LE, Childress-Fields K Mulder KM and Brattain MG. (1989a). Mol. Endocr. 3, and Brattain MG. (1993). J. Biol. Chem., 268, 691 ± 698. 1215 ± 1222. Mulder KM and Brattain MG. (1989b). In: Augenlicht LH. (ed). The Cell and Molecular Biology of Colon Cancer. CRC Press: Boca Raton, Florida, pp. 45 ± 67.