Contributory Effects of de N o v o Transcription and Premature Transcript Termination in the Regulation of Human Epidermal Growth Factor Receptor Proto-oncogene RNA Synthesis*

Overexpression of the epidermal growth factor (EGF) receptor (c-erbB) proto-oncogene is a frequent occurrence in human carcinoma and appears to accom- pany autocrine or paracrine transforming growth fac-tor-a expression, which in model systems can result in activation of EGF receptor tyrosine kinase activity and phenotypic transformation. Here we have investigated the transcriptional regulation of the EGF receptor gene, by run-on transcription in isolated nuclei derived from epithelioid tumor lines. The level of transcription was measured at various points on the 100-kilobase pair EGF receptor gene locus, on either sense or anti- sense DNA strands. We find the level of sense strand transcription along exon 1 is 8-fold higher than tran- scription in exons 2-26. Primary EGF receptor transcripts appear to pause or terminate prematurely be- tween exons l and 2. Termination was mapped to a sequenced region approximately 2 kilobase pairs 3‘ of exon 1, proximal to a previously reported DNase I hypersensitive site and an enhancer-like activity. Transcription in the CpG-rich region surrounding exon 1 is bidirectional, with antisense transcripts ini- tiating in intron 1 and extending through the coding first exon. Activation of protein kinase C results in a &fold induction of EGF receptor transcription, accom-panied by a slow release in the block RNA elongation between exon 2 and exon 26, showing that EGF receptor results shown were typical of three independent experiments. Elevated transcription of exon 1 templates was observed on both sense and antisense strands, indicating bidirectional transcription in the exon 1 region. Significantly reduced transcription on the exon 3-16 sense strand template (with a 4-fold greater length and T content) could be detected. No antisense transcription of exons 2-26 was observed. R, increasing hybridization wash temperature to control for potential nonspecific hybridization to CC-rich sequences in the exon 1 region. Run-on hybridizations to single-stranded templates were washed and RNase treated as de- scribed in Fig. l, and then finally washed twice at 70 "C (lane l ), 80 "C (lane 2 ) or 60 "C (lane 3 ) in 0.1 X SSC, 1% SDS for 30 min. In lane 4, run-on transcription was aholished by addition of 2 pg/ml of the RNA polymerase I1 inhibitor n-amanitin. Opposite strand templates which detect sense transcripts are designated +; conversely, templates which hybridize to antisense transcripts are designated In two separate experiments, hybridization to exon 1 templates was preserved even at 80 "C, while control signals were reduced, indicating the EGFR exon 1 hybridization signal was specific.

Overexpression of the epidermal growth factor (EGF) receptor (c-erbB) proto-oncogene is a frequent occurrence in human carcinoma and appears to accompany autocrine or paracrine transforming growth factor-a expression, which in model systems can result in activation of EGF receptor tyrosine kinase activity and phenotypic transformation. Here we have investigated the transcriptional regulation of the EGF receptor gene, by run-on transcription in isolated nuclei derived from epithelioid tumor lines. The level of transcription was measured at various points on the 100-kilobase pair EGF receptor gene locus, on either sense or antisense DNA strands. We find the level of sense strand transcription along exon 1 is 8-fold higher than transcription in exons 2-26. Primary EGF receptor transcripts appear to pause or terminate prematurely between exons l and 2. Termination was mapped to a sequenced region approximately 2 kilobase pairs 3' of exon 1, proximal to a previously reported DNase I hypersensitive site and an enhancer-like activity. Transcription in the CpG-rich region surrounding exon 1 is bidirectional, with antisense transcripts initiating in intron 1 and extending through the coding first exon. Activation of protein kinase C results in a &fold induction of EGF receptor transcription, accompanied by a slow release in the block RNA elongation between exon 2 and exon 26, showing that EGF receptor RNA synthesis may be altered by changes in de novo transcription and by a block to RNA elongation.
The epidermal growth factor receptor proto-oncogene transduces a mitogenic signal produced by the binding of ligands comprising the epidermal growth factor family (EGF,' transforming growth factor-a, vaccinia virus growth factor, and amphiregulin). Ligand binding results in an activation of the intrinsic receptor tyrosine kinase activity and an increase in the steady-state RNA concentration of a variety of cellular genes, for example c-myc, c-fos, and the EGF receptor (EGFR) * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
itself. Receptor overexpression is frequently observed in human carcinoma, often in conjunction with autocrine or paracrine transforming growth factor-a expression (Ozanne et al., 1986), and similarly, constitutive activation of the EGF receptor tyrosine kinase has been shown to result in phenotypic transformation of immortalized mammalian fibroblasts in culture (Stern et al., 1987;Velu et al., 1987Velu et al., , 1989; DeFiore et Wells and Bishop, 1988;Haley et al., 1989).
The EGF receptor mRNA is encoded by 26 exons spanning 100 kb (Haley et al., 1987a) of chromosome 7 q12-14 (Shimizu and Kondo, 1982) and is transcribed in tissues from all three germ cell layers; however, expression is not observed in cells of the hematopoietic lineage (reviewed in Carpenter, 1987). The EGFR promoter and adjacent sequences in intron 1 are enriched in the relatively rare dinucleotide CpG. In a variety of cellular genes with constitutive "housekeeping" functions, rare CpG-rich sequences at the 5' end of such genes are present in an "open" or exposed chromatin conformation and represent domains associated with transcription initiation (Bird, 1986). Consistent with an open chromatin conformation, two DNase I nuclease-hypersensitive sites are observed in the 5' end of the EGFR gene: one located in the repeating TCC sequences in the promoter region and another in the first intron approximately 2 kb 3' of exon I Johnson et al., 1988b).
Transcription of the EGFR gene has been previously shown to initiate at multiple sites -400 to -150 from the ATG start codon, without a common RNA polymerase I1 TATA motif Haley et al., 1987a). A requirement for transcription factor Spl (Dynan and Tjian, 1983) in maximal EGF receptor promoter activity has been shown (Haley et al., 1987a;Kageyama et al., 1988a) and five Spl binding sites have been identified by DNase protection (Johnson et al., 1988a). Two additional transcription factors, TCF and ETF, which contribute to maximal EGF receptor promoter activity have been identified and partially characterized (Johnson et al., 198813;Kageyama et al., 1988b).
Upon transient transfection, the EGF receptor promoter shows a high activity relative to "strong" promoters such as the SV40 early promoter or the Rous sarcoma virus LTR (Haley et al., 1987a;Johnson et al., 1988a) even when transfected into cell lines expressing low or undetectable numbers of EGF receptors, for example in Chinese hamster ovary cells. Additionally, the CpG-rich character and multiple Spl binding sites, commonly found in genes with metabolic or housekeeping functions, further suggest some constitutive function of the EGFR promoter. In contrast both EGF and phorbol ester positively regulate EGF receptor RNA levels in a variety of cell lines, even in the absence of protein synthesis. The positive effect of EGF on steady-state levels of EGF receptor RNA is attributable to both RNA stabilization  and to a lesser extent to de novo transcription (Fernandez-Pol et al., 1989;Hudson et al., 1989) in actively growing epithelioid cell lines. This may constitute a pathway by which EGF receptors are replenished at the cell surface, following receptor down-regulation and degradation. However, between tissues or cell lines expressing the EGF receptor, large variations in receptor number (Adamson and Rees, 1981) and steady-state EGFR RNA levels are observed. This suggests that additional mechanisms may be required to account for the large differential in EGF receptor expression between expressing cell types.
Here we have investigated the control of EGF receptor expression at the level of de nouo transcription and RNA elongation. We have measured the distribution of RNA polymerase I1 actively engaged in transcription at defined exons of the EGF receptor gene in the human carcinoma cell lines, by run-on transcription (Groudine and Weintraub, 1982) in isolated nuclei. Radiolabeled RNA transcribed by this procedure was hybridized to an excess of single-or double-stranded EGFR or control DNAs immobilized on nylon membranes. The extent of hybridization reflects the average distribution of RNA polymerase I1 at various points on the EGFR gene.
We describe pausing and/or premature termination of the EGF receptor primary transcript between exons 1 and 2, which was mapped to a sequenced region approximately 2 kb 3' of exon 1. The partial block to transcription is not attributable to gene rearrangement. Transcription initiation from the CpG-rich sequences surrounding exon 1 was observed on both DNA strands. Stimulation of transcription was observed 60 min following addition of phorbol ester or fetal calf serum to serum-deprived cells. The relatively slow increase in transcript elongation in exons 2-26 between 120 and 240 min and the magnitude of increased readthrough transcription suggest the block to RNA elongation may be a point at which EGF receptor RNA synthesis is controlled.

MATERIALS AND METHODS
Cell Lines-The vulval carcinoma cell line A431-7 was maintained in Dulbecco's modified Eagle's media (DMEM) containing 10% fetal calf serum (Gibco) and 1 mM sodium pyruvate. In serum deprivation experiments, cells were washed three times in DMEM and cultured without serum for 24-48 h prior to stimulation with phorbol 12myristate 13-acetate or 10% fetal calf serum. The MDA468 and MDA231 breast carcinoma lines and the HN-5 squamous carcinoma line were cultured under similar conditions. Nuclear Run-on Transcription-Nuclei from A431, MDA468, MDA231, or HN5 cell lines were prepared (Prywes and Roeder, 1986) from two 150-mm culture dishes, where approximately 2 X 10' cells were washed in cold calcium and magnesium free phosphate-buffered saline, scraped into 5 ml of cold phosphate-buffered saline and centrifuged a t 500 X g for 5 min a t 4 "C. The pellet was resuspended in 1.8 ml of ice-cold 10 mM Tris-C1, p H 7.4, 10 mM NaCI, 3 mM MgC12, and Nonidet P-40 added to 0.5%. After 5 min on ice the nuclei were pelleted (500 X g, 5 min, 4 "C), washed in 4 ml of ice-cold Nonidet P-40 lysis buffer, pelleted, and resuspended in 0.2 ml of 50 mM Tris-C1, pH 8.3, 5 mM MgC12, 0.1 mM EDTA, 40% glycerol. Samples (100 pl) were flash-frozen in liquid nitrogen and stored at -70 "C.
were carried a t 30 "C by addition of 100 JLI of 10 mM Tris-CI, p H 8.0, Nuclear run-on transcription assays (Eick and Bornkamm, 1986) 6 mM MgC12, 300 mM KCI, 0.25 mM ATP, GTP, CTP, and 100 pCi of [a-:"P]UTP (800 Ci/mmol, SP6 grade; Amersham Corp.) for 25 rnin, followed by treatment with RNase-free DNase (50 units for 5 min; Boehringer Mannheim) and addition of 0.2 ml of 0.5% SDS, 100 mM Tris-C1, pH 7.4, 50 mM EDTA, and 30 p1 of 3 mg/ml proteinase K (Merck; preactivated a t 37 "C for 30 min in 10% SDS). After 60 min a t 37 "C the digestion was terminated by phenol/chloroform extraction and the reaction desalted using a 2-ml Sephadex G-50 spun column equilibrated in 10 mM Tris-C1, p H 7.4, 1 mM EDTA, 1% SDS. Samples were boiled for 5 min and hybridized to the immobilized template DNAs at 44 "C in 50% deionized formamide, 0.75 M NaC1, 0.075 M sodium citrate (5 X SSC), 5 x Denhardt's solution (Denhardt, 1966), 50 pg/ml sonicated salmon sperm DNA, 50 mM sodium phosphate, p H 6.8, 1 mM sodium pyrophosphate, 100 JLM ATP, and 1% SDS for 24 h. Filters were washed for 30 min each, twice in 2 X SSC, 1% SDS at 60 "C, rinsed twice in 2 X SSC, twice in 2 X SSC with 10 yg/ml RNase A, twice with 0.2 X SSC, 1% SDS at 60 "C (except Fig. 2B, where the final wash was performed in 0.1 X SSC, 1% SDS at 70 and 80 "C), and the filters air-dried and autoradiographed. Fig. l B was quantitated using a scanning laser densitometer and software developed by Protein Databases Inc. (Huntington Station, NY). Fig. 6A was quantitated using a LKB Ultrascan XL laser densitometer. DNA Probes and Templates-The hybridization templates for nuclear run-on transcription assays comprised both double-stranded and single-stranded templates. The double-stranded EGFR cDNAs and genomic DNAs shown in Fig. 1A were previously described (Ullrich et aL, 1984;Haley et al., 1987b, respectively). One pg of template DNA was immobilized on Genescreen Plus membrane (Dupont) according to the manufacturer's protocol, using a slot blot apparatus. pUC18 vector DNA was used as a negative control. A yprotein kinase C subclone containing 5"flanking sequence was used as a positive control template. RNA polymerase I1 transcription from this template was relatively constant between experiments and showed only slight induction during the phorbol ester and fetal calf serum stimulation experiments.
The structure of the 5' end of the EGFR gene from A431 was analyzed by Southern hybridization using the 3889-bp EGFR genomic probe LIII encompassing 850 bp of 5"flanking sequence, 351 bp of exon 1 and 2688 bp of intron 1 or the EGFR cDNA probes shown in Fig. 1A. Hybridization probes were purified, radiolabeled using random primers (Taylor et al., 1976) and DNA polymerase Klenow fragment, desalted on a 2-ml Sephadex G-50 column, and hybridized to Southern-blotted A431 DNA restricted with EcoRI or Hind111 (Maniatis et al., 1982).

Transcription of EGFR RNA in Epithelioid Cell Lines Is Elevated in Exon 1 Relative to Exons 2-26-
The increase in steady-state levels of EGFR RNA following addition of EGF, phorbol ester, cycloheximide, or fetal calf serum to both serum-supplemented or serum-deprived epithelioid cell lines has been well documented Kudlow et al., 1986;Earp et al., 1986;Bjorge and Kudlow, 1987;Haley et al., 1987a). To examine the potential role of de nouo transcription in the regulation of EGF receptor expression in epithelioid cell lines, the distribution of RNA polymerase I1 along the EGFR gene was measured by nuclear run-on transcription. Nuclei were prepared from A431 vulval carcinoma as well as control carcinoma cell lines, radiolabeled, and run-on RNA isolated and hybridized to templates representing various regions of the EGFR gene. Seven double-stranded EGF receptor cDNA probes and one double-stranded genomic probe spanning the 100-kb gene were initially used as hybridization templates, as illustrated in Fig. 1A.
In run-on transcription assays of nuclei isolated from the vulval carcinoma cell line A431, the distribution of RNA polymerase I1 was unequal along the length of the EGF receptor gene. Elevated levels of run-on transcription were observed upon hybridization to templates containing exon 1 ; lune 9, pUC18 negative control DNA; lune 10, y-protein kinase C positive control DNA ( g R 6 ) . Hybridized membranes were washed twice in 2 X SSC (1 X SSC, 0.15 M NaCI, 0.015 M trisodium citrate), 1% SDS at 37 "C for 30 min, twice in 2 X SSC for 5 min, twice in 2 X SSC, 10 pg/ml RNase A, 37 "C for 30 min and twice in 0.1 X SSC, 1% SDS a t 60 "C for 30 min. Hybridized filters were dried and subjected to autoradiography. Similar results were obtained in four independent experiments. Equal loading of EGFR and pUC18 template DNAs was verified by control hybridization of the template filter to vector DNA. C, quantitation of nuclear run-on hyhridization hy scanning laser densitometry and normalization of the derived values of the T content of the template. Template HS18 ( R , lune I ) is omitted due to the heterogeneity of transcription start sites within the EGFR promoter and suhsequent imprecision in the assignment of T content. relative to templates containing any of the exons from 2 to 26 (Fig. 1B). Quantitation of run-on hybridization showed an average &fold higher rate of transcription of exon 1 relative t o exons 2-26 (Fig. 1C). Similar results were obtained in four independent experiments using the EGFR templates illustrated in Fig. L4, with the consistent finding that the level of transcription in exon 1 was significantly higher than in the remainder of the 100-kb locus. Elevated transcription in exon 1 was also observed in the carcinoma lines MDA468, MDA231, and HN-5 and is not restricted to the A431 cell line (not shown). In contrast, no EGFR run-on hybridization was observed to exon 1 templates in nuclei from the B-cell line IM9, a cell line which does not express EGF receptor (not shown). These observations suggest that a partial block to RNA elongation occurs between exons 1 and 2.
In order to verify the observation that transcription of exons 2-26 was reduced with respect to exon 1, we addressed the following questions: 1) Is transcription from the CpG-rich region surrounding exon 1 bidirectional and can the combined transcription of sense and antisense DNA strands account for the elevated transcription of exon 1 observed? 2) Is the elevated level of transcription around exon 1 sensitive to (1amanitin, a potent inhibitor of RNA polymerase I1 transcription? 3) Could these results be attributed to artifact and derive: (i) from either the increasing length of the primary transcript 5' "tail" during transcription of the 100-kb gene (addressed by partial alkaline hydrolysis of the run-on RNAs); (ii) by the possibility of RNA polymerase reinitiating at the promoter (addressed by addition of n-laurylsarcosine to the reaction to block reinitiation); (iii) by variation in the concentration of template DNA immobilized on the filter membrane (addressed by control hybridization with vector sequence); or (iv) by nonspecific GC base pairing during hybridization (addressed by increasing the stringency of the hybridization wash conditions).
Elevated Transcription in Exon I Is Bidirectional, Sensitive to a-Amanitin, and Nonartifactual-To determine whether the observed block to transcription elongation derived from the coding or "sense" DNA strand, RNA radiolabeled by nuclear run-on transcription in A431 nuclei was hybridized to single-stranded EGFR templates ( Fig. 2A). Analysis of this experiment allows several points to be made: 1) Elevated transcription of exon 1 was observed using sense strandspecific probes. 2) EGFR transcription along exon 1 was also bidirectional, with specific hybridization to templates containing both sense and antisense exon 1 sequences. No antisense transcription could be detected in exons 2-26.3) Transcription was sensitive to a-amanitin (0.5-2 pglml), characteristic of transcription by RNA polymerase 11. 4) In control experiments, partial alkaline hydrolysis of the run-on primary, unspliced transcript RNAs (0.1 M NaOH, 10 min, 0 "C) had no effect on the pattern of run-on hybridization to various templates along the EGFR gene. Blocking transcriptional reinitiation by addition of 0.5% n-laurylsarcosine to the nuclear run-on reaction also had little observable effect to the hybridization pattern. Finally, the hybridization pattern did not derive from differences in the amount of DNA template immobilized on the nylon filter as shown by hybridization of the "processing control" filter with vector M13 DNA.
In a separate control experiment, it was shown that hybridization of nuclear run-on RNA to exon 1 templates resulted from specific base pairing and did not derive from nonspecific interaction with GC-rich sequences. Stringent washing of the run-on RNA hybridizations in 0.1 x SSC, 1% SDS at 70 and 80 "C did not abolish hybridization to a single-stranded exon 1 template with a 62% GC content (Fig. 2B) or independently to a variety of double-stranded exon 1 templates of varying GC content (not shown).
Localization of a Block to RNA Elongation to a Gene Region 2 kb 3' of Exon I-To map regions of possible transcription termination, single-stranded M13 templates containing sequences within the EGFR promoter, exon 1, and intron 1 were prepared (Fig. 3A). Nuclei were isolated from subconfluent A431 cells, run-on transcripts labeled with [a-" In two separate experiments, hybridization to exon 1 templates was preserved even a t 80 "C, while control signals were reduced, indicating the EGFR exon 1 hybridization signal was specific. and hybridized to immobilized M13 single-stranded template DNA. The result of one of several consistent experiments is shown in Fig. 3B, where templates which detect sense strand transcripts are designated plus (+; a-g) and antisense strand transcripts minus (-; h-o). Several points can be made: 1) Run-on transcription initiated from the EGFR promoter region as expected (Fig. 3B, 6 and c )  ( Fig. 3B, a ) . 2) Transcription terminated prematurely approximately 2 kb 3' of exon I. Identical results were obtained with the XbaI-Hind111 template (Fig. 3R, R ) and a similar StuI-Hind111 template (not shown).
3) Antisense transcription initiated in the 800-bp CpG-rich region immediately 3' of exon 1 (Fig. 3B, k), proceeding through exon I and then showed a reduction in antisense transcription 5' of the promoter (Fig. 323, h). 4) In control experiments similar to those shown in Fig. 2, no run-on hybridization was observed when nuclei were pretreated with the RNA polymerase I1 inhibitor Premature Termination of EGFR Transcription a-amanitin (2 pg/ml), equal template immobilization was confirmed by hybridization with labeled M13 DNA, and no effect was observed when run-on transcriptions were performed in the presence of 0.5% n-laurylsarcosine to block reinitiation of transcription or when transcripts were subjected to partial alkaline hydrolysis as described previously (not shown).
A 2.7-kb region comprising the 5' end of the first intron of the EGFR gene was sequenced by dideoxy chain termination (Fig. 4). The CpG-rich base composition previously observed in the 5"untranslated sequences extended approximately 800 bp into the first intron. Seven potential transcription factor S p l binding sites were observed in the 800 bp 3' of exon 1 which may promote antisense transcripts which initiate within this region. A long tract of alternating purine-pyrimi-

T G A U~V \~C C C~T A C A G C C T C C C C T C G G A C C C C G C~~C A~~T T T C T G A G A~~~~C T C C C C G C C~C C 640
CGCTCCGCGCAGGTCTCIUULCTGhRGCCGGCG===G=CAGCCT~CCC~=====TCCAGGTCCCCGCGATCCTCCTT 120

T G C T G G C U C G G T T A G T T T = = I C G T T~~~A~~~T -T C C~~T T G G~G C A G~C C C C U C C G C T C aeo
GCCTCGCCCGGTGCGCCCTCGTCTTGCCTATC~GAGTGCC~C~CTC~ACCCCAGCTCCCTCCGC~CCGffiC 960

G G G~U~T A G I I C T G T~T~T A A A~~~A~T T = T =~T~~W~U G G~T C I T C T C * C
I 920 SBDI PILI

A G A T -T T T T M C A G I U U L~T G T T T~~G~T C T A T C~T T C A G T U A T T T T A T T C M G A T G C A C T T T G T T
0 " l sac1 2320 2400 XbaI IgO

S t " 1
MGTGACTTTCCULCAGCTGGTUGAG~AGACGGG~ITCACACCAGGGGCTTCC~CTCCAGI\TCCCTCTCTCMCT 2640 termination. The deduced amino acid sequence of exon 1 is indicated, potential transcription factor S p l binding sites are indicated by small boxes, a potential Z-DNA structure is marked by a wavy underline and two potential T-rich termination signals indicated by the double underline. A potential stem-loop structure immediately upstream of the poly(T) tract is shown (arrow). The SrnaI-XbaI (1958-2441) transcription termination region is boxed. The positions of nuclear factor binding sites (Maekawa et aL, 1989) proximal to the termination region are marked by brackets. Restriction cleavage sites are marked and the sequence length, in base pairs, is shown to the right. The location and orientation of run-on hybridization templates containing intron 1 (Fig. 3) are marked by the corresponding letter. dine nucleotides was found at position 1266, similar to those sequences associated with allelic conversion events (Shen et al., 1981). Several poly(T)-rich sequences were noted, particularly TTTTGTTTGGTT at position 2490 and TTCTCTTTATTTT at position 2040 (which was preceeded by a potential stem-loop structure GTCTCTgccgaAGAGAC), within the proposed region of premature transcription termination, homologous with previously characterized RNA polymerase I1 termination signals . No truncated RNAs were observed by Northern analysis using exon 1 probes (not shown).

The Block to EGFR RNA Elongation Does Not Derive from
Gene Rearrangement-The A431 cell line contains an approximate 30-fold amplification in EGF receptor gene copy number and overexpresses EGF receptor protein (-3 X IO6 receptors/cell; Ullrich et al., 1984;Merlino et al., 1984;Lin et al., 1984). It should be noted that approximately one-half the EGFR genes in the A431 cell line contain a single detectable intrachromosomal rearrangement near exon 13 Haley et al., 1987b), resulting in the expression of a secreted, EGF receptor extracellular domain.
In order to address the possibility that the block to RNA elongation might derive from gene rearrangement proximal to exon 1, the genomic structure of the EGFR gene was investigated by Southern hybridization analysis.
No rearrangement in the exon 1 region of the EGFR gene was observed. Hybridization of the 3.9-kb EGFR genomic DNA probe (LIII) to EcoRI or Hind111 restricted A431, MDA468, and placental DNAs revealed no obvious sequence rearrangement in the region surrounding exon 1 (Fig. 5A). Furthermore, Southern hybridization with the LIII, A21,64.1, 64.3, and 62.1 probes to A431 DNA (Fig. 5B) show the expected restriction pattern for the EGFR locus (Haley et al., 1987a). These data confirm the specificity of the template DNA used in run-on hybridization and demonstrate the absence of repetitive sequences in the probes used. Thus hybridization specifically reflects transcription of the EGF receptor gene and not transcription of genes whose RNAs contain highly repetitive sequence motifs.
EGFR Transcription and RNA Elongation in Serum-deprived A431 Cells May Be Elevated by Activation of Protein Kinase C or Serum-dependent Signaling Pathways-Activation of growth factor signaling pathways has been shown to regulate the progression of RNA transcripts through the termination site(s) at the exon l/intron 1 boundary of the Cmyc gene (Nepveu et al., 1987). In order to evaluate whether the transcriptional block observed between exons 1 and 2 in the EGFR gene might be a point of regulation, nuclear runon transcription assays were performed with nuclei derived from serum-deprived A431 cells restimulated with either phorbol ester or serum. Subconfluent A431 cells were grown in serum-free DMEM for 36 h, then stimulated with either phorbol12-myristate 13-acetate (10 ng/ml) or fetal calf serum (10%). Cells were washed with ice-cold phosphate-buffered saline at time points over a 4-h interval and nuclei prepared as described under "Materials and Methods." The 30-and 60min time points gave similar results to the 15-min time point, with a small increase in exon 1 transcription by 60 min observable and were omitted from Fig. 6A for brevity. Serum deprivation of subconfluent A431-7 cells lead to a decrease in the rate of transcription from exon 1 relative to cells maintained in 10% fetal calf serum. When serum starved cells were restimulated with either phorbol ester (Fig. 6A) or with serum (not shown), an increase in EGFR transcription over the 240min time course was observed. Quantitation of duplicate runon hybridization experiments by scanning laser densitometry  (lanes 1-3) or HindIII (lanes 4-6) restriction endonuclease digested genomic DNA probed with random primed, 32P-labeled 3.9-kb LIII DNA containing exon 1 of the EGF receptor gene. Lanes 1 and 4 were loaded with 4 pg of digested A431 DNA; lanes 2 and 5 were loaded with 8 pg of digested MDA468 DNA; lanes 3 and 6 were loaded with 25 pg of digested human placental DNA. No rearrangement of exon 1 or flanking regions was observed. Positions of the 8-kb EcoRI (lanes 2-31 and 3.9-kb HindIII (lanes 4-6) hybridizing bands are shown (arrowheads). B, no repetitive DNA elements were observed in A431 DNA digested with EcoRI (kft panel) and HindIII (right panel) by Southern hybridization using the run-on hybridization probes (LIII (exon l), A21 (exons 1-3), 64.1 (exons 3-16), 64.3 (exons 16-22), 62.1 (exons 22-26)) under standard washing conditions (0.1 X SSC, 60 "C, 60 min). Thus nuclear run-on hybridization to these EGFR probe sequences will not detect repetitive sequences found in cellular RNAs under the conditions described and run-on hybridization experiments specifically reflect EGF receptor transcript levels.
was performed, in which filter background was subtracted and values for each filter normalized to the positive control (Fig.  6B). The initial increase in de novo transcription following phorbol ester addition was specific to exon 1, with little readthrough to exons 2-26 within the 60-min time period. The increased RNA elongation through exons 2-26 in the 120and 240-min time period roughly paralleled the increased transcription from exon 1, although at a reduced level. The temporal lag in phorbol ester stimulation of transcript levels in exons 2-26 in the 30-to 120-min time period relative to that of exon 1, suggests some functional dissociation of transcription initiation and subsequent RNA elongation in the EGFR gene.

DISCUSSION
Deregulation of proto-oncogene transcription resulting in overexpression of the gene product has been implicated in oncogenesis (Bishop, 1987). Specifically, EGF receptor overexpression has been associated with the development and/or progression of epithelioid neoplasia in vitro and in vivo. Here, on transcript hybridization to serum-deprived A431 cells stimulated with phorbol ester over a 240-min time period. Each point reflects the average of two separate experiments in which two templates for exon 1 (HS18 and A21) and exons 2-22 (64.4 and 64.3) are also averaged. Background was subtracted for each filter and values for each filter normalized to the positive control template. Note that values were not further normalized to T content of the templates, due to the heterogeneous transcription start sites within the HS18 template, which understates the differential in hybridization intensity of exon 1 versus exons 2-26. Quantitation was performed by scanning laser densitometry.
we have examined EGFR transcription at the level of transcription initiation and RNA elongation along the 100-kb EGFR gene in response to agents which activate protein kinase C-or serum-dependent signal transduction pathways.
Induction of EGFR Transcription Initiation through Protein Kinase C Activation-Transcription of the EGF receptor gene may be induced by protein kinase C-dependent and seruminduced pathways with an intermediate time course of induction. EGFR transcription during in vitro nuclear run-on transcription assays showed a significant induction of de m u 0 RNA synthesis approximately 60-120 min following phorbol ester or serum stimulation of serum-deprived A431 cells (Fig.  6). Serum deprivation of A431 cells resulted in a decrease in EGFR transcription, which could be reversed by addition of phorbol ester or by readdition of fetal calf serum, leading to a significant stimulation of EGFR transcription above the level observed in cells maintained in serum. The reduction in EGFR RNA synthesis by serum deprivation may account for the reduction in EGF receptor number observed in rat liver during fasting (Freidenberg et al., 1986).
Activation of protein kinase C by phorbol ester and subsequent phosphorylation of the EGF receptor leads to a reduction in high affinity binding of EGF to its receptor (Hunter et al., 1984;Downward et al., 1985) and a decrease in receptor tyrosine kinase activity (Davis, 1988). However, phorbol ester addition to several epithelioid cell lines increases steady-state EGFR RNA levels, reaching a maximum over 4-6 h Bjorge and Kudlow, 1987;Haley et al., 1987a;Bjorge et al., 1989). These results, together with our finding that phorbol ester treatment increases de m u 0 transcription of the EGFR gene, support a positive feedback model in which EGF receptor biosynthesis may be stimulated in response to receptor down-regulation by phorbol ester.
It has been proposed that EGF and protein kinase C share a common pathway by which EGF receptor synthesis may be increased (Bjorge and Kudlow, 1987). Recent data on the induction of EGFR transcription by EGF (Fernandez-Pol et al., 1989) over a similar time period to that reported here for activation of protein kinase C are further evidence for a common pathway. In transient transfection experiments with EGFR promoter-reported constructs, maximal stimulation by EGF and phorbol ester was not additive (Hudson et al., 1989), suggesting convergence of the two signaling pathways on EGFR transcription element(s) within a "core" sequence. This CpG-rich core DNA region is important in maximal activity of the EGF receptor promoter (Haley et al., 1987a) and contains multiple potential binding sites for transcription factor AP2, a factor known to transduce signals generated by protein kinase C (Imagawa et al., 1987), and which may account for the induction of de novo transcription by phorbol ester or serum treatment observed here.
Bidirectional Transcription-By use of strand-specific runon transcription templates we have demonstrated that both sense and antisense transcripts initiate within an 800-bp CpG island surrounding exon 1. EGFR antisense RNA polymerase I1 transcripts initiate within a 400-bp region immediately 3' of exon 1, with transcripts extending through exon 1 and decreasing in number in the 5"flanking region. Some variation in the comparative levels of sense and antisense transcription is observed between individual experiments, suggestive of independent regulation, but the precise mechanism for this remains obscure. Previous transient transfection experiments with EGFR promoter-reporter constructs suggested a bidirectional pattern of transcription from the promoter core region (Haley et al., 1987a) which, in the antisense orientation, might be terminated upstream of the MstII site, -485 from the ATG start codon (Hudson et al., 1989). These data are consistent with the nuclear run-on experiments here which show decreased antisense transcription 5' of the MstII site (Fig. 3, h). Bidirectional transcription from CpG-rich "islands" (Bird, 1986) has similarly been described for the dihydrofolate reductase (Linton et al., 1989) and c-Ha-ras (Lowndes et al., 1989). Multiple sequence motifs potentially binding to the transcription factors Spl and AP2 are located within the EGFR intron and promoter regions (Fig. 4) may be in part responsible for antisense transcription. Whether antisense transcripts could potentially hybridize with exon 1 coding sequence of the mature EGFR mRNA and alter the efficiency of its translation is untested and remains a formal possibility.
A Block to EGFR RNA Elongation in the First Intron- The epidermal growth factor receptor contains a partial block to RNA elongation between exons 1 and 2, approximately 2 kb downstream of the first exon within a SmaI-XbaI restriction fragment . Termination occurs close to a previously reported DNase I-sensitive site (Ishii et ai., 1985) similar to transcription of the c-myb gene in which premature termination occurs within a region containing multiple DNase I-hypersensitive sites (Bender et al., 1987), a topological feature associated with transcriptionally active chromatin (Weintraub et al., 1981). This region of the first intron contains an enhancer-like activity which has been shown to cooperate with a second enhancer upstream of the EGFR promoter (Maekawa et al., 1989), and DNase footprint mapping in this region shows nuclear factor binding a t multiple sequences on both DNA strands. Premature termination and subsequent alteration in steady-state RNA concentration has been described for the c-myc, c-myb, c-fos genes as well as a number of viral genes. Changes in c-myc RNA elongation between exons 1 and 2 have been well described during cellular differentiation (Bentley and Groudine, 1986;Nepveu and Marcu, 1986;Eick and Bornkamm, 1986) or in response to growth factor stimulation (Nepveu et al., 1987;reviewed Spencer and Groudine, 1990).
Sequence analysis of the termination region revealed several poly(T) tracts on the coding strand which show homology with transcription termination signals identified in viral genes  where stem-loop structures, similar to the structural requirements observed for abortive transcription in procaryotes, have been observed (see Resnekov et al., 1989). Short transcripts prematurely terminating in intron 1 were not detectable by Northern analysis using poly(A)-selected RNA (not shown) and are likey to have a short halflife. Thus the exact localization of sequence motifs responsible for the premature termination in the EGFR gene await precise characterization of in vitro run-off transcripts.
The partial block to RNA elongation in itself may be construed as a form of regulation of gene expression. One possible function for the block to elongation could be to prevent overloading of RNA polymerase I1 complex along the 100-kb locus, initiated by a highly active promoter. A high level of EGFR transcription, comparable to the SV40 early and Rous sarcoma virus long terminal repeat promoters, was observed in previous transient transfection experiments and derived in part from relatively constitutive transcription factor Spl. In the A431 vulval carcinoma cell line used in this study, overexpression has been shown to correlate with the level of gene amplification (Lin et al., 1984) and not apparently through release in the block to elongation between exons 1 and 2. This appears to be true of the MDA468, MDA231, and HN-5 carcinoma lines as well, suggesting the block to elongation may not be readily circumvented, although small changes in readthrough transcription to exons 2-26 would escape detection in the nuclear run-on assays described.
A perhaps more interesting question is whether the partial block to EGFR RNA elongation may be altered in response to hormonal stimuli. First, the block to elongation does not remain constant. Increasing transcriptional initiation, by protein kinase C activation results in an increased readthrough to exons 2-26 and elevated mRNA synthesis. Second, a protein kinase C-mediated increase in transcription initiation was observable within 60 min, while elevated RNA elongation to downstream exons was detected only between 120 and 240 min. In vivo transcription of the 18kb first intron should occur in approximately 10 min (assuming a rate of 30 nucleotides/s), extending to exon 16 in approximately 30 min. The time differential between increased initiation and RNA elongation through to exon 2 would not simply derive from the normal rate of RNA polymerase 11, suggesting some independent regulation of the two events.
Cellular regulation of RNA elongation has been described for both the c-myc and hsp7O genes. For c-myc, stimulation of serum-arrested A31 cells with EGF or with agents which elevate cellular CAMP levels leads to a release of the transcriptional block between exons 1 and 2, thereby increasing steady-state c-myc RNA levels (Nepveu et al., 1987;Miller et al., 1989). In the Drosophila hsp7O gene, heat shock results in the release of a block to RNA elongation and an increase in the concentration of hsp70 RNA (Rouguie and Lis, 1988). In vitro transcription data using Ad2 and HIV templates suggest antitermination may be mediated by specific factor(s), for example the tat gene in the case of HIV (Kao et al., 1987). The close proximity of protein binding sites (Maekawa et al., 1989) immediately upstream of the EGFR intron region associated with premature termination of transcription, raises the possibility that the formation of protein-DNA complexes may similarly influence RNA elongation through this DNA region. Given that the number of cell surface EGF receptors is known to vary over a three log range in the different cell lines and tissues known to express the receptor, premature termination might serve, in part, to define tissue-specific or differentiation state-specific transcription of EGFR RNA, as observed for the c-myc gene during HL-60 differentiation (Bentley and Groudine, 1986). In this regard, the decreased level of EGF binding observed following differentiation of embryonal carcinoma cell lines (Carlin and Andrews, 1985) or following differentiation of primary keratinocytes (Boonstra et al., 1985), may indicate cellular systems in which this hypothesis can be tested.