INTRODUCTION

The transcriptional regulatory elements of eukaryotic genes are located both upstream and downstream of the transcription initiation site. Downstream regulatory elements for genes transcribed by RNA polymerase II are found within both coding and noncoding proximal exons and further downstream within introns or within the region 3 of the last exon. However, examples of regulatory elements found within distal coding exons of mammalian genes are much less common. Here, we investigate a potential regulatory element located within exon 6 of the keratin 18 (K18)¹ gene.

The K18 gene contains seven exons that code for a type I intermediate filament protein that is normally coexpressed with its heteropolymeric partner, keratin 8 (K8), in a variety of internal, single-layered epithelial tissues such as liver, kidney, and intestine. K18 and K8 expression generally persists in cancers that arise from these tissues (1). A 10-kilobase genomic fragment containing the K18 gene has all of the genetic information necessary for position-independent, copy number-dependent, and tissue-specific expression in transgenic mice (2). K18 expression is regulated in part by proximal promoter elements, including multiple SP-1 sites and a TATA box, three negative regulatory elements within the first intron, and a complex 100-bp enhancer element also within the first intron (3, 4). Furthermore, an analysis of DNase I-hypersensitive regions of the gene in transgenic mouse tissues indicates that a potential regulatory element appears to reside within exon 6, the penultimate exon of the gene (5). Deletion of this region of the gene decreased transient expression of transfected constructions, and reintroduction of exon 6 into the gene or the cDNA restored expression levels.

We have now characterized this element further and have evaluated its importance in vivo. An element within exon 6 binds AP-1 components (c-Jun and c-Fos) and interacts with the first intron enhancer to promote AP-1 responsiveness. Transgenic mice generated from a mutant K18 gene containing a 33-bp substitution of exon 6 with the corresponding region of keratin 19 (K18/19), without an AP-1 site, express significantly less RNA in adult liver and intestine. In addition, we have compared this chimeric K18/19 gene to a related construction that alters the reading frame of K18 in exon 6 to generate a premature termination codon but does not change the AP-1 site. The alterations within the K18/19 gene and the exon 6 frameshift result in decreased expression in transgenic mice by different mechanisms. The frameshift mutation leads to an increased rate of RNA turnover and little RNA accumulation, whereas the alteration of exon 6, which eliminates the AP-1 site, does not change the stability of the mRNA. In addition, the frameshift mutation interferes with the activity of the intron enhancer. These results show that regulatory elements within a distal coding exon of K18 cooperate with the first intron enhancer to facilitate optimal expression of the gene but are not essential for either integration site-independent or copy number-dependent expression.

EXPERIMENTAL PROCEDURES

In the K18/19 gene, 33 bp of exon 6 were replaced with the corresponding coding sequence of the K19 cDNA by introducing a synthetic double-stranded oligonucleotide between the BamHI and SmaI sites of exon 6 in a fragment of the whole K18 gene. Subsequently, the wild type K18 gene fragment was replaced by a fragment defined by a unique BstEII site in exon 5 and a KpnI site in exon 7. To generate the K18-Bam gene, the BamHI site in exon 6 was digested and filled with the use of the Klenow fragment of DNA polymerase. This 4-bp insertion altered the reading frame. The resulting plasmid contained the ClaI site at the place of BamHI site and the termination codon TGA 18 bp downstream of the ClaI site in exon 6 (Fig. 1).

[View Larger Version of this Image (19K GIF file)]

The XKCAT, XKCATIS, and XKCATE100 constructions for CAT assay have been described previously (4). Exon 6 and its K18/19 and BamHI modifications were made by polymerase chain reaction amplifications from the corresponding plasmids with primers containing SalI sites. The polymerase chain reaction fragments were cloned into the XhoI sites of XKCAT, XKCATIS, and XKCATE100 plasmids upstream of the K18 gene promoter.

Transgenic mice were prepared by standard procedures as described previously (2, 6-8). The HindIII fragments of K18/19 and K18-Bam constructions were used for microinjections. The transgenic animals were identified by dot-blot hybridization of tail DNA with a K18 cDNA hybridization probe. K18-Bam founder mice were generated in C57Bl6 × SJL F2 hybrid mice by the DNX corporation (Princeton, NJ) under a contract from the National Institutes of Child Health and Human Development Mouse Development Facility. Subsequent breeding of the founder mice was performed with C57Bl6 mice. K18/19 mice were generated by The Burnham Institute Transgenic Mouse Service in the FVB/N strain of mice.

F9 and HR9 cells were transfected as described (3, 4). CAT constructs were cotransfected with LK444-Lac vector representing the human -actin promoter driving the -galactosidase gene for normalization of transfection efficiency. Because the LK444-Lac vector is responsive to AP-1 activity but not to Ets2 activity (data not shown), normalization was performed only within the set of transfected cells receiving the same transactivators. CAT and -galactosidase activities were measured as described previously (3, 4, 9).

Transgene copy number was determined by dot-blot or by Southern blot hybridization with a K18 probe followed by quantitation in an Ambis radioactivity image analyzer. Signals were normalized for each tissue by hybridizing the stripped filter with a mouse L32 ribosomal protein cDNA hybridization probe. The estimation of copy number of the transgene was done as described previously (7). RNA was purified from dissected mouse tissues using guanidine isothiocyanate and ultracentrifugation in cesium chloride (10). The levels of K18 RNA were determined by RNase protection assay using an SP6 RNA polymerase [³²P]UTP-labeled antisense transcript of the 431-bp fragment of the K18 gene, overlapping the RNA start site and part of the first exon (5). The neo RNA was detected with an SP6 RNA polymerase antisense transcript of the 245-bp fragment of the neo gene (5). L32 ribosomal gene RNA was measured simultaneously. A 147-bp fragment of the L32-4A gene (nucleotides 103-250) (11) was amplified by polymerase chain reaction with primers containing EcoRI and BamHI sites and cloned into pGEM1. Digestion with EcoRI and transcription by SP6 polymerase yielded a 187-nucleotide probe. The protected K18 signal was normalized to the signal obtained from the endogenous mouse L32 RNA or the cotransfected neo RNA. The specific radioactivity of the L32 riboprobe was adjusted to 10 times less than that of the K18 riboprobe.

Footprint analysis with purified c-Fos and c-Jun was performed as described previously for the K18 intron enhancer (4). Exon 6 was amplified using oligonucleotides with external XbaI sites, matching the K18 gene at K18 nucleotide 5525 and ending with nucleotide 5748. The amplified XbaI fragment was cloned into pUC19. The fragment was end labeled and excised using the EcoRI and HindIII sites of the polylinker. The gel-purified probe was incubated with bovine serum albumin, purified c-Jun and c-Fos, or the Ets2 DNA binding domain in 50 µl containing 10% glycerol, 2% polyvinyl alcohol, 25 mM Tris-HCl, pH 7.9, 6.25 mM MgCl₂, 50 mM KCl, 1 mM EDTA, 0.5 mM dithiothreitol, and 1 µg of poly(dI·dC) (Sigma). After digestion with the indicated concentrations of DNase I, the products were displayed on a sequencing gel and were detected by autoradiography. Recombinant c-Jun and c-Fos were gifts from Dr. T. Deng (University of Florida, Department of Biochemistry and Molecular Biology). The DNA binding domain of Ets2 was used as described previously (1, 9). This recombinant protein binds consensus Ets binding sites and functions as a dominant negative inhibitor of Ets2 transactivation (12).

HR9 parietal endodermal cells were transfected by the calcium phosphate method with 5 µg of the pMC1NeoA plasmid and 15 µg of the K18 (pGC1853) (13, 14), K18/19, or the K18-Bam plasmids. After overnight incubation with the plasmid precipitates, the cells were washed and incubated in complete culture medium (Dulbecco's modified Eagle's medium containing 10% fetal bovine serum) and 5 µg/ml actinomycin D. RNA was extracted at 0, 0.5, 1, 2, and 4 h after exposure to the drug using the acidic phenol method (15). Total RNA was digested with RNase-free DNase to remove residual plasmid DNAs (16) and was analyzed by RNase protection assay for the neo and K18 RNAs simultaneously.

RESULTS

Exon 6 of the K18 gene contains potential AP-1 and Ets binding sites (Fig. 1). To assess whether these sites are capable of binding AP-1 or Ets proteins, the region was subjected to DNase I footprinting in the presence of either purified c-Fos and c-Jun or recombinant Ets2 DNA binding domain. Purified AP-1 components recognized the putative AP-1 site and resulted in a clear footprint over the suspected element (Fig. 2, lanes 4 and 5). No evidence for Ets2 binding to the region was evident, even though a potential Ets binding site is found just upstream of the AP-1 site (Fig. 2, lanes 2 and 3). The same conditions resulted in a clear footprint of the Ets site in the first intron enhancer (9).

[View Larger Version of this Image (82K GIF file)]

Three lines of evidence have previously implicated the sixth exon of the K18 gene in transcriptional regulation. First, a DNase-hypersensitive site is found associated with the exon in transgenic tissues that express K18 (5). Second, an HpaII restriction enzyme site within the exon is less methylated in tissues that express the gene (5). Third, deletion of portions of exon 6 and 7 from the K18 gene reduces the level of K18 RNA expressed in transfected cells, and reintroduction of exon 6 rescues the deficiency (5). To investigate the mechanism and biological role of exon 6 in further detail, we have assayed the regulatory activity of the exon removed from the context of the K18 gene. Fragments corresponding to exon 6 sequences of the K18 gene, a 33-bp substitution of K18 exon 6 with the homologous coding sequence of K19 (K18/19) which lacks the AP-1 site or a 4-bp frameshifting insertion in exon 6 (K18-Bam) (Fig. 1), were inserted into a CAT vector driven by the 250-bp proximal promoter of the K18 gene (XKCAT). In addition, the fragments were tested in the presence of either the K18 first intron (XKCATIS) or with a 100-bp fragment of the first intron, corresponding to the approximate minimal enhancer (XKCATE100) (Fig. 3A). The vectors were transfected into either F9 embryonal carcinoma cells that have low AP-1 activity or into HR9 parietal endodermal cells that express mouse K18 and have substantial AP-1 activity. In addition, an expression vector for c-Fos was cotransfected with the vectors in F9 cells to test the response of the constructs to elevated AP-1 activity. c-Fos has been shown previously to be the most potent AP-1 activator of the K18 gene in F9 cells (3, 17, 18). The results are shown in Fig. 3, B, C, and D. The expression of the XKCAT proximal promoter constructs in F9 cells was weak, but the presence of exon 6 or the K18/19 modification of this exon increased expression 3-5-fold (Fig. 3B, lanes 2 and 3). The coexpression of c-Fos did not stimulate the expression of any of these constructs. The addition of the first intron or the Ets/AP-1 oncogene response element resulted in increased activity and significant induction when c-Fos was coexpressed as shown previously (3) (Fig. 3C, lanes 2, 3, 6, and 7). The smaller 100-bp enhancer element (XKCATE100) increased activity further in F9 cells, as reported previously. The differences in activity of XKCATIS (Fig. 3C, lane 1) and XKCATE100 (Fig. 3C, lane 5) are caused by the presence of three negative regulatory elements in the larger intron sequence which are active in F9 cells but not HR9 cells (9). The K18/19 modification of exon 6 did not stimulate either the basal activity or the Fos-induced activity as effectively as the wild type exon. This suggests that the AP-1 site found in the wild type K18 exon 6, but not in the K18/19 exon 6, contributes modestly to the transcriptional stimulatory activity of exon 6 but is not the only positive regulatory element. We have shown previously that mutation of the AP-1 binding site within the intron enhancer of the K18 gene abolishes the response to coexpressed AP-1 in transfection experiments (9). Exon 6 stimulates the K18 proximal promoter activity independent of AP-1 but also modulates the AP-1 responsiveness of the intron enhancer. In contrast to the wild type exon 6, the K18-Bam alteration, which inserts 4 bp into the exon, abolished the stimulatory activity of exon 6 both in the absence (Fig. 3B, lane 4) and presence of the larger intron fragment (Fig. 3C, lane 4) and the smaller 100-bp intron enhancer (Fig. 3C, lane 8).

[View Larger Version of this Image (31K GIF file)]

In contrast to F9 cells, the addition of the exon 6 fragments did not augment activity of either construct in differentiated HR9 cells (Fig. 3D, lanes 2, 3, 6, and 7). This result is consistent with previous results that showed that the AP-1-dependent enhancer coded for by the 100-bp intron 1 fragment was fully active in HR9 cells, presumably because AP-1 components were not limiting these reporter constructions. No difference is observed between XKCATIS and XKCATE100 in HR9 cells because the three silencer elements found in the larger intron fragment are not active in differentiated cells (9). However, in HR9 parietal endodermal cells, the construct containing exon 6 from the K18-Bam gene, in combination with the longer intron fragment, repeatedly resulted in much less activity (Fig. 3D, lane 4). This negative regulatory activity was dependent on the presence of the intron sequences containing the three negative regulatory elements because exon 6 of K18-Bam did not influence the activity of the XKCATE100 vector in HR9 cells (Fig. 3, panel D, lane 8, and panel C, lane 8). This negative activity of exon 6 containing the BamHI mutation is unusual because it requires an intron fragment that contains elements that were thought to be active only in undifferentiated F9 cells but not in HR9 cells. The combination of the longer intron fragment and the mutant exon 6 now appears inhibitory in differentiated HR9 cells.

Because the AP-1 site within exon 6 appeared to interact with other regulatory elements of the K18 gene, the importance of exon 6 was evaluated within the context of the whole gene. Immunofluorescent staining of transiently transfected HR9 cells indicated that the chimeric K18/19 gene was capable of generating K18/19 RNA and protein that were incorporated into the endogenous keratin intermediate filament network (data not shown). The transcriptional response of this construct was assessed by transfection into F9 cells with or without coexpressed c-Jun, c-Fos, and Ets2. To standardize for differences in transfection efficiency, the pMC1NeoA plasmid coding for the neo gene was cotransfected with each plasmid mixture, and simultaneous assessment of the neo RNA was measured in the RNase protection analysis. The results are shown in Fig. 4A, and normalized data are shown in Fig. 4B. In F9 cells, the K18 gene is expressed at a low level (Fig. 4, panel A, lane 1; panel B, lane A). Coexpression of c-Jun stimulates expression of the gene only modestly (Fig. 4B, lane B). However, coexpression of c-Fos is much more potent (Fig. 4, panel A, lane 3; panel B, lane C). Similar to c-Jun, Ets2 activates the K18 gene only modestly by itself. However, coexpression of Ets2 with either c-Jun or c-Fos increases expression (Fig. 4, panel A, lanes 5 and 6; panel B, lanes E and F). Maximal stimulation is seen when all three factors are expressed simultaneously (Fig. 4, panel A, lane 7; panel B, lane G).

[View Larger Version of this Image (28K GIF file)]

In contrast to the wild type K18 gene, the K18/19 construct was poorly activated by c-Fos (Fig. 4, panel A, lane 10; panel B, lane C), consistent with the results of the CAT constructions. In addition, the response to Ets2 and the combination of Ets2 and AP-1 components was greatly reduced for the K18/19 gene. When supplemented with all three transcription factors, the activity of the K18/19 gene was moderately reduced (Fig. 4, panel A, lane 14; panel B, lane G). Because the primary Ets-responsive element of the K18 gene is within the first intron enhancer element and Ets2 does not bind directly to exon 6 sequences, these results indicate an interaction between the regulatory elements of exon 6 and the first intron, resulting in effects that are dependent upon the particular AP-1 and Ets component expressed. The maximum level of expression of the K18/19 gene in F9 cells when supplemented with Ets2, c-Fos, and c-Jun is about 58% of the wild type gene. Thus it appears that regulatory elements found within exon 6, including the AP-1 site, act to modulate transcriptional activity of the K18 gene in transfected cultured cells.

To evaluate the importance of exon 6 elements in vivo, transgenic mice were generated with the K18/19 and K18-Bam genes. Standard pronuclear injection of fertilized mouse embryos resulted in four lines of each type of mouse (Table I). Southern blot analysis was performed to ensure that all of the transgenic lines contained the mutant K18 genes and to confirm the normal head-to-tail tandem organization predominantly found in transgenic mice. All lines contained the K18/19 gene in tandem array as assessed by Southern blot patterns using restriction enzyme, which cut the transgene only once with no apparent rearrangements (data not shown). Quantitative dot-blot analysis was performed as described previously to estimate the total number of integrated genes (Table I).

Table I. RNA expression by transgenic mice K18 RNA was measured by RNA protection analysis and was normalized to the amount of L32 RNA in the same probe. RNA values are represented as pg of K18 RNA/10 µg of total RNA. Line Copy no. Organ K18 RNA pg RNA/gene Avg. RNA/gene S.D. Avg. % K18 K18 avg. Liver 11.9 4.0 K18 avg. Intestine 8.3 2.8 K18 avg. Kidney 6.5 2.3 K18/19-11 5.7 Liver 15 2.6 K18/19-16 5.3 Liver 29 5.5 K18/19-18 18.0 Liver 26 1.4 K18/19-20 14.0 Liver 40 2.9 3.1 1.7 26 K18/19-11 5.7 Intestine 14 2.5 K18/19-16 5.3 Intestine 18 3.3 K18/19-18 18.0 Intestine 21 1.2 K18/19-20 14.0 Intestine 18 1.3 2.1 1.0 25 K18/19-11 5.7 Kidney 34 6.0 K18/19-16 5.3 Kidney 84 15.8 K18/19-18 18.0 Kidney 383 21.8 K18/19-20 14.0 Kidney 88 6.3 12.4 7.7 191 K18-Bam-1 16.0 Liver 5 0.3 K18-Bam-2 40.0 Liver 5 0.1 K18-Bam-3 12.6 Liver 3 0.3 K18-Bam-4 19.8 Liver 2 0.1 0.2 0.1 2 K18-Bam-1 16.0 Intestine 6 0.3 K18-Bam-2 40.0 Intestine 19 0.5 K18-Bam-3 12.6 Intestine 8 0.7 K18-Bam-4 19.8 Intestine 11 0.5 0.5 0.1 6 K18-Bam-1 16.0 Kidney 9 0.6 K18-Bam-2 40.0 Kidney 48 1.2 K18-Bam-3 12.6 Kidney 37 3.0 K18-Bam-4 19.8 Kidney 5 0.2 1.2 1.2 19

Four transgenic mouse lines were also established from the K18-Bam construction (Fig. 1), which contains a 4-bp insertion in exon 6 and subsequently generates a frameshift terminating five codons downstream of the original K18 reading frame. This mutation was created in an attempt to generate mice expressing a dominant acting mutant form of K18, capable of disrupting K8/K18 intermediate filament structure. A very similar truncated K18 protein was characterized previously after expression of the truncated form of the cDNA from a human -actin promoter (20). Fig. 5 shows the results of Southern blot analysis for the K18-Bam gene of three lines, in which the normal BamHI site is destroyed, and a diagnostic ClaI site results from the manipulations of the position of the BamHI site. Separate analysis of the fourth line confirmed the same pattern (data not shown). Dot-blot analysis of tail DNAs and Southern blot analysis of the genomic DNA from these animals confirmed a normal head-to-tail integration pattern and 13-40 copies of the K18-Bam gene inserted (Table I).

[View Larger Version of this Image (56K GIF file)]

K18 RNA expression in several organs of the K18/19 and K18-Bam mice was measured by RNase protection assay. Fig. 6A shows one result, comparing the levels of expression of the different lines in intestine, kidney, and liver. The levels of K18 RNA were compared with synthetic standard K18 mRNA, in this and other assays, to estimate the level of expression. RNAs from K18-Bam mice were analyzed similarly (Fig. 6B). Table I shows the values for representative mice of all lines, and a summary of data for each line is shown in Fig. 7.

[View Larger Version of this Image (65K GIF file)]

[View Larger Version of this Image (29K GIF file)]

K18 RNA was found in epithelial organs of all transgenic animals, which suggests that expression of both mutant genes is independent of the particular integration site of each mouse. However, the K18/19 mice expressed approximately 26% of the level of wild type K18 mice in liver and 25% of the level in intestine. These decreased levels of expression are consistent with the lower levels of K18 RNA found in transfection experiments with the entire gene (Fig. 4B, lane G), although the degree of inhibition was greater in vivo. Surprisingly, expression in kidney was elevated over values found for K18 mice. However, the overlap of the deviations of the wild type K18 and K18/19 kidney samples precludes a firm conclusion. Additional analysis of other organs in which K18 RNA is usually very low (muscle and spleen) confirmed that the tissue specificity of the gene was not altered dramatically (data not shown). The levels of RNA for all four lines of mice correlated fairly well with the number of integrated genes, showing that the K18/19 construct was likely still capable of copy number-dependent expression. These results indicate that the 33 bp of exon 6 containing the AP-1 site are important for maximal expression in liver and intestine. Furthermore, the greater effect of the mutation in vivo compared with transfection analysis probably reflects a more realistic assessment of the importance of the 33-bp region of exon 6. It is likely that the combination of specific members and levels of transcription factors like AP-1 combine with the interactions among the different K18 regulatory elements to result in greater effects in vivo.

K18-Bam mice also expressed K18 RNA in all three tissues of all four lines of mice. However, the levels of expression per gene copy were dramatically lower than wild type or K18/19 mice (note the strong signal for the L32 standard (III) in Fig. 6B relative to Fig. 6A). This result mirrors the inhibitory activity of this mutation observed in transfection experiments (Fig. 3, panel C, lanes 4 and 8; panel D, lanes 4 and 8). Clearly, in vivo, this 4-bp insertion has dramatic effects upon the efficiency of expression of K18 RNA.

As intragenic elements that affect the rate of RNA degradation are well known, the stability of K18, K18/19, and K18-Bam RNAs was measured in cultured cells. HR9 cells were cotransfected with one of the three genes and additionally the pMC1NeoA plasmid for normalization of the transfection efficiency. The transfected cells were then treated with actinomycin D, and the amount of K18 and neo RNAs was measured by RNase protection assay after increasing drug exposure times. Fig. 8 shows the results of one such experiment. The neo RNA behaved similarly in each of the three sets of transfections, with an apparent half-life of 3.3 h (Fig. 8A). The initial levels of K18 RNA from the three constructions reflected the regulatory activity of exon 6. K18/19 RNA was lower than wild type K18, and K18-Bam RNA was much less than K18/19 RNA. However, after normalization, the rate of loss of the K18 signal relative to the neo signal still permitted an evaluation of the stability of the RNAs. Both the K18 and K18/19 mRNAs were more stable than the neo RNA, thus resulting in an apparent increase in K18 RNA relative to neo. This result shows that the K18/19 RNA has a stability similar to that of the wild type K18 RNA. However, the K18-Bam RNA was degraded at a rate similar to the neo RNA, which was faster than either K18 or K18/19. Thus, the mutation introduced into the K18-Bam construct results in increased turnover, which likely contributes to the low level of K18-Bam RNA in transgenic mice tissues.

[View Larger Version of this Image (14K GIF file)]

DISCUSSION

The results of K18/19 expression in transgenic mice indicate that sequences within exon 6 modulate expression of the K18 gene. These sequences stimulate the proximal promoter and interact with the first intron enhancer to modulate the response of the gene to AP-1. However, this interaction appears complex. First, exon 6 sequences containing the AP-1 site do not confer AP-1 responsiveness in the absence of the intron enhancer. Second, the K18/19 gene acts like the wild type gene in responding modestly to coexpressed c-Jun. However, activation of expression of K18/19 by c-Fos in F9 cells is altered dramatically (Fig. 4). This particular sensitivity to c-Fos is further reinforced by the refractory response to the combination of coexpressed c-Fos and Ets2. The details of the differential responsiveness of the two regulatory elements to specific components of AP-1 in F9 cells remain unclear because c-Fos normally interacts with Jun family members to mediate AP-1 binding, and exon 6 can clearly bind c-Fos/c-Jun well. F9 cells contain little c-Jun, JunB, or c-Fos, but do express JunD (21). Furthermore, F9 cells express two AP-1 repressors that may, in part, be responsible for the low AP-1 activity in these cells (22). Thus, the effects of expressing particular AP-1 members in F9 cells may involve the titration of inhibitory activities as well as direct activation. Furthermore, the expression of c-Jun or c-Fos may also mediate the activation of other members of the AP-1 or CREB families and thus result in an indirect activation that still utilizes one or both K18 AP-1 sites. Although the AP-1 site found within exon 6 is the most obvious transcriptional regulatory element, other neighboring potential factor binding sites are likely important for the activity of exon 6. Deletion of the AP-1 site from exon 6 (by substitution with K19 sequences) decreases but does not destroy the modulatory activity of exon 6 in F9 cells (Fig. 3). We have no evidence of the direct interaction of Ets2 with sequences within exon 6. Nevertheless, the alteration of exon 6 alters the responsiveness of the gene to coexpressed Ets2. The cooperation between the first intron Ets and AP-1 sites was revealed by mutations of either site alone (4, 9). The cooperation of intron 1 with exon 6 with respect to activation by AP-1 would suggest that the primary collaboration between the intron AP-1 and ETS sites is modulated further by factors binding to exon 6.

Regulatory elements located within the transcribed portions of genes are quite common, although examples of such elements in coding exons are much less frequent. Some examples include c-myc (23), rabbit -globin (24), the glial fibrillary acidic protein gene (25), and the transcriptional silencer in the blood clotting factor VIII cDNA (26). Histone genes are well known for containing transcriptional regulatory elements within the coding regions (27-31). One practical consequence of the transcriptional activity of exon 6 is that the K18 cDNA, and perhaps other cDNAs, are not neutral with respect to transcriptional regulation. Expression of cDNAs by heterologous promoters may be altered significantly by the interaction of regulatory elements within coding exons or with the regulatory elements of the particularly expression vector.

K18 is expressed primarily in epithelial tissues, and transgenic expression reflects that specificity. However, direct transfection studies in cultured cells result in promiscuous K18 expression (5, 13). The analysis of K18/19 transgenic mouse tissues revealed that both liver and intestinal expression were decreased in K18/19 mice, consistent with a modulatory role on the level of expression of the gene revealed by transfection analysis. The 75% decrease in K18/19 RNA per gene observed in adult liver and intestine contrasts with the 40% decrease observed in cultured cells; it likely reflects the different milieu of transcription factors available to the gene in animals. Exon 6 appears not to be as important for expression in kidney or HR9 parietal endodermal cells as for liver and intestine. The analysis of selected other organs did not reveal changes in tissue specificity. For example, K18 RNA levels were very low in spleen and muscle in K18/19 mice (data not shown). Thus, exon 6 appears to modulate the level of expression but does not alter the basic tissue specificity of the gene.

The characteristics of position-independent and copy number-dependent expression of some genes are dependent upon multicomponent regulatory elements known as the locus control region (32-36). The internal regulatory elements of the K18 gene might also contribute to such activity for the K18 gene. However, the integration site and the largely copy number-dependent behavior of the K18/19 gene indicate that 33 bp of exon 6 are not necessary for these characteristics. This reinforces previous demonstrations that position-independent and copy number-dependent expression in transgenic mice can be conferred on both the herpes simplex thymidine kinase gene (7) and on the metallothionein-human growth hormone fusion genes (37) by the distal 5- and 3-flanking regions of the K18 gene. These flanking elements are sufficient for position-independent and copy number-dependent expression of at least two different heterologous genes. Exon 6 is probably not necessary for such behavior of the K18 gene.

The unexpected dramatic effect on expression in transgenic mice of the 4-bp insertion in the K18-Bam construct is likely the result of both a decrease in RNA stability (Fig. 8) and the creation of a negative regulatory element. The mutation in the BamHI site of exon 6 abolishes the stimulatory activity of the exon (Fig. 3, panel B, lane 4; panel C, lane 8). Furthermore, it inhibits the action of the intron enhancer, but only in the context of the larger intron fragment. The most likely interacting elements may be the three negative regulatory elements identified previously within intron 1 (9). However, these silencer elements were active only in stem cells such as F9 and not in differentiated cells, like HR9. Exon 6 may contain an anti-silencer type activity, as is found in the vimentin gene (37), which neutralizes the effect of the silencer elements. The K18-Bam mutation may have disrupted this anti-silencer activity, thus rendering the silencers now active in differentiated tissues. The presence of this suppressor activity, combined with the decrease in RNA stability, results in the very weak expression of the K18-Bam transgene in all tissues.

The regulation of the K18 gene is complex because multiple regulatory elements are found both flanking and within the gene, and these dispersed elements interact to modulate activation by transcription factors. Exon 6 is an example of a coding region that also functions as a transcriptional regulatory element. Our studies have emphasized the importance of evaluating particular elements in vivo and in the proper context. Transfection studies and reporter genes do not provide a complete picture of the transcriptional regulation of the K18 gene, which is expressed in a variety of different epithelial cell types.