Growth-inhibitory Activity and Downregulation of the Class II Tumor-suppressor Gene H-rev107 in Tumor Cell Lines and Experimental Tumors

The EJ-ras transformed rat hepatoma cell line ANR4, the rat colon carcinoma cell lines DHD and WB2054, the rat pancreatic tumor cell lines CRI-D2 and CRI-D11, and the Ki-RAS transformed rat kidney cell line KNRK were obtained from the European Collection of Animal Cell Cultures (Porton Down, UK). The immortal rat pancreatic cell line AR4-IP and the transformed rat pancreatic cell line AR4-2J were obtained from Dr. O. Hagenbüchle (Lausanne, Switzerland). The immortal rat kidney cell lines, NRK 49F and NRK 52E, were obtained from the American Type Culture Collection (Rockville, MD). 208F (Quade, 1979) and REF52 (Franza et al., 1986) are immortalized, nontumorigenic rat fibroblast cell lines. The FE-8 cell line is an HRAS-transformed derivative of 208F (Griegel et al., 1986). Clone F9 is a phenotypic revertant cell line derived from FE-8 (Schäfer et al., 1988). Cells were grown in DME medium supplemented with 10% fetal calf serum, penicillin (100 U/ml), and streptomycin (100 μg/ml), standard medium. ANR4 cells were grown in Williams E medium supplemented with 10% fetal calf serum, penicillin (100 U/ml), and streptomycin (100 μg/ml). DHD cells were grown in DME/Hams F10 medium (1:1); AR4-2J, AR4-IP, and WB2054 cells were grown in RPMI medium supplemented with the same additives as above.

Total cellular RNA was isolated as described (Chomczynski and Sacchi, 1987). 10 μg of RNA was denatured using glyoxal and separated in 1% agarose gels (McMaster and Carmichael, 1977), and transferred to nylon membranes (Genscreen plus, New England Nuclear, Boston, MA). Northern blots were hybridized with a 523-bp H-rev107 cDNA fragment encoding the 5′ untranslated sequence and part of the coding sequence as a probe. Fragments were ³²P labeled by random priming (Feinberg and Vogelstein, 1983), and hybridization was carried out either in 50% formamide, 10% dextran sulfate, 1 M sodium chloride, 5X Denhardt's solution, and 100 μg/ml denatured salmon sperm DNA at 42°C, or in the Quickhyb solution from Stratagene (La Jolla, CA) at 65°C.

H-rev107 cDNA was cloned into expression vectors in which transcription of the inserted DNA is driven by the cytomegalovirus promoter. Plasmids pCMVTK107+ and pCMVTK107− encode H-rev107 mRNA in sense and in antisense orientation, respectively (Hajnal et al., 1994). Plasmid pCDNA-107+ was constructed by cloning the 900-bp BamHI insert, obtained from the cDNA clone pBS107C, into the BamHI site of pCDNA3 (Invitrogen, San Diego, CA). This fragment contains the complete coding region of H-rev107, 33 bp of 5′ untranslated region, and 386 bp of 3′ untranslated region. Plasmid pCDNA-107− drives transcription of antisense H-rev107 mRNA. The pCDNA107 ΔC vector was constructed by PCR amplification of a 416-bp fragment obtained from pBS107C using the H-rev107 specific primers 5′ TTGAATTCAT GCCCATACCAGAACCCAAG 3′ and 5′ GGGAGCTCTCGCCACCTTGACGATGTCT 3′, followed by insertion of the PCR product into the EcoRI and XhoI sites of pCDNA3. The termination codon is just downstream of the XhoI site in the vector. This cDNA clone expresses only the first 135 amino acids of the H-rev107 protein. For the production of a tetracycline-inducible H-rev107 (pUHD107), a 714-bp BamHI fragment from cDNA clone pBS107B (Hajnal et al., 1994) was cloned into the BamHI site of the plasmid pUHD 10-3 (Gossen & Bujard, 1992).

Plasmid DNA was transfected into 208F, FE-8, and ANR4 cells by calcium– phosphate precipitation, as described (Wigler et al., 1978). 5 μg of linearized recombinant vector DNA, 1 μg of linearized plasmid pY3 carrying the hygromycin B resistance gene, and 20 μg genomic carrier DNA were used for transfection of 10⁶ recipient cells. 48 h after transfection, cells were subjected to selection in standard medium containing 400 μg/ml hygromycin B. When colonies had reached a size of 500–1,000 cells, the cells were trypsinized and pooled. H-rev107 protein expression was analyzed by immunofluorescence using the anti-H-rev107 antibody. To obtain FE-8 cells harboring a tetracyclin-inducible H-rev107, the cells were first transfected with the plasmid pUHD172-1neo and stable clones selected as described above. This plasmid expresses the modified tetracycline repressor which was turned into an activator (Gossen et al., 1995). These clones were analyzed by transient transfection using the plasmid pUHC13-3 containing the luciferase-reporter gene under the control of a tet operator (Gossen and Bujard, 1992). Clones showing a preferable ratio of induction to background luciferase actvity were chosen for further transfection with the pUHD-107 plasmid and clones selected by cotransfection with the pY3 plasmid as described above. 48 h after induction of H-rev107 expression using 2 μg/ml doxycyclin, protein expression was monitored either by Western blot analysis or by immunofluorescence. For control cells (10-3-3), the pUDH10-3 vector alone was used instead of pUHD-107. For tumorigenicity assays, 10⁶ cells were injected subcutanously into either side of ICR nu/nu mice. Tumor growth was controlled daily. After the indicated time the animals were killed, tumors removed, and either frozen in liquid nitrogen or explanted in tissue culture medium after digestion with trypsin/EDTA. Tumor cells were incubated in standard medium containing 50 μg/ml gentamicin.

For subcellular fractionation, adherent cells were washed twice in PBS, incubated in hypotonic lysis buffer (10 mM Tris, 0.1 mM dithiothretiol, 1 mM PMSF, 5 μg/ml aprotinin, 5 μg/ml leupeptin, 50 μg/ml TLCK, pH 8) for 5–10 min on ice, and lysed using a Dounce homogenizer. Nuclei were obtained after centrifugation at 500 g at 4°C for 10 min, washed twice in PBS, and purified via centrifugation through a sucrose gradient (Blobel and Potter, 1966). The 500 g supernatant was centrifuged at 100,000 g at 4°C for 30 min to obtain the membrane fraction. Cytoplasmic proteins were precipitated from the 100,000 g supernatant using methanol/chloroform (Wessel and Függe, 1984). All pellets were resuspended in SDS sample buffer, boiled for 5 min, and stored at −20°C. Protein concentration was determined using amido black staining of protein extracts on nitrocellulose filters (Schaffner and Weissmann, 1973). Equal amounts of protein were separated on 16% polyacrylamide gels by SDS-gel electrophoresis (Laemmli, 1970) and transferred to nitrocellulose membranes (Immobilon P; Millipore Corp., Milford, MA). Membranes were blocked overnight in TBST (10 mM Tris, pH 8, 150 mM NaCl, 0.05% Tween 20) containing 1% BSA, incubated with affinity purified anti-H-rev107 antibody (Hajnal et al., 1994) diluted 1:1,000 in TBST for 4 h. Membranes were washed three times for 5 min in TBST and incubated for 45 min with an anti–rabbit horse-radish peroxidase coupled antibody and processed according to the manufacturer's instructions (ECL-Kit; Amersham Intl., Buckinghamshire, England).

Indirect immunoperoxidase staining on frozen tissue sections was performed using the Vectastain-kit (Vector Laboratories, Burlingame, CA) according to the protocol of the manufacturer. In blocking experiments, the H-rev107 antiserum was preabsorbed with 8 μg/ml of bacterially expressed H-rev107 protein prior to incubation with the sections. For immunofluorescence analysis, cells were grown on glass coverslips coated with polyornithine (100 μg/ml). After fixation in 3% paraformaldehyde in phosphate-buffered saline for 20 min, cells were permeabilized using 0.5% Triton X-100 in PBS for 2 min. Coverslips were incubated with the H-rev 107 antiserum, diluted 1:100 in PBS/1% BSA, washed in PBS, incubated with a DTAF-conjugated goat anti–rabbit antibody (Dianova GmbH, Hamburg, Germany), washed in PBS, mounted in 5% Mowiol and analyzed with a fluorescence microscope (Zeiss Inc., Oberkochen, Germany). For nuclear staining, cells were incubated with 1 μg/ml Hoechst 33258 fluorescent dye together with the second antibody.

REF52 cells which overexpress H-rev107 (Hajnal et al., 1994), are resistant to RAS transformation, and forced overexpression of RAS blocks their growth and eventually causes cell death (Hirakawa and Ruley, 1988). Interestingly, the adenovirus E1A nuclear protein, which represses the H-rev107 gene, abolishes the resistance of REF52 cells to RAS (Hajnal et al., 1994). These observations suggested that H-rev107 is responsible for RAS- induced growth arrest or cell death. To test the hypothesis that overexpression of H-rev107 is growth inhibitory for RAS-transformed cells but not for normal cells, we have stably transfected different H-rev107 cDNA expression constructs into immortal, nontumorigenic 208F rat fibroblasts, into the HRAS transformed rat fibroblasts FE-8, or into the EJ-RAS transformed rat hepatoma cell line ANR4. FE-8 cells do not express endogenous H-rev107, while ANR4 cells exhibit very low levels of endogenous H-rev107 mRNA (Fig. 1,A). The plasmid pCDNA107+ directs moderate expression of the full-length H-rev107 cDNA. An antisense expression vector pCDNA107−, the plasmid pCDNA107 ΔC expressing the truncated H-rev107 cDNA, and the plasmid pCDNA3 were used as controls. The cloning efficiencies of FE-8 and ANR4 cells, following the introduction of the plasmid pCDNA107+, were reduced by 62 and 75%, respectively, as compared to control transfections (Table I).

After stable transfection of FE-8 cells with pCDNA107+, only 5 of 13 isolated FE-8 clones showed significantly elevated levels of the H-rev 107 protein, as determined by Western blots. In contrast, eight out of nine clones transfected with pCDNA107 ΔC expressed the truncated form of the H-rev107 protein (Fig. 1,B). Immunofluorescence analysis of the five independent pCDNA107+ transfectant clones using the H-rev107 antiserum, revealed that only 20–50% of the cells were positive for the protein. Within a few weeks of culture, the fraction of H-rev107 positive cells decreased to <10% (data not shown). A significant reduction of colony formation was not observed in transfections of nontumorigenic 208F cells excluding an unspecific toxicity of H-rev107 expression plasmids (Table I). These experiments showed that the level of H-rev107 expression inversely correlated with the clonogenicity of HRAS-transformed cells transfected with the full-length H-rev107 cDNA. Moreover, there was a strong selection against H-rev107 expressing cells during continued culture of transfectants.

To confirm this, we also transfected the pCMVTK107+ vector, which yields very high H-rev107 protein levels in the transfected cells due to the presence of the thymidine kinase leader sequence cloned in front of the H-rev107 cDNA. Transfection of this H-rev107 expression vector did not yield individual stable ANR4 transfectant clones, while only few colonies were formed after selection of FE-8 cells. None of 11 clones tested expressed H-rev107, whereas HRAS protein expression was maintained, indicating that a very high H-rev107 expression in the presence of HRAS does not permit permanent proliferation of transfectants in vitro. In addition to the stably transfected cells, FE-8 fibroblasts harboring the H-rev107 cDNA under the control of a tetracycline-inducible promoter (pUHD107), were tested for their cloning efficiencies with or without induction of the H-rev107 protein (Fig. 2). A reduction in cloning efficiency similar to that observed with constitutively expressed H-rev107 was obvious in cells with pUHD-107 (Fig. 2). Upon prolonged culture of these cells in the presence of doxycycline, there was again a selection against H-rev107 expressing cells (data not shown).

To analyze the effects of H-rev107 overexpression on proliferation in vivo, we generated FE-8 and ANR4 transfectants with the same expression vectors as above. To minimize loss of H-rev107 expressing cells due to prolonged culture in vitro, colonies resistant to the selective agent were trypsinized, after they had reached a size of more than 200 cells, and pooled. 8 d after trypsinization and replating we performed immunofluorescence analysis to determine the fraction of H-rev107 positive cells. A representative culture of FE-8 transfectants is shown in Fig. 3 (A and C). To quantitate the H-rev107 positive fraction of cells transfected with pCMVTK107+, we counted the cells present in several independent microscopic fields of FE-8 and ANR4 transfectants. The proportion of H-rev107 positive cells was 56% in a total of 614 cells counted. FE-8 and ANR4 cells transfected with the antisense expression vector, or transfected with a pCDNA3 plasmid directing the expression of an unrelated gene, formed progressively growing tumors after a latency period of 4–6 d. In transfected FE-8 populations, the latency period increased only marginally, while it increased to 9–13 d in ANR4 populations (Fig. 4). The tumors formed after injection of FE-8 and ANR4 cell pools transfected with the high expressing plasmid pCMVTK107+ enlarged more slowly than control transfectants until about day 11. Afterwards, the differences in tumor sizes originating from cells transfected either with full length H-rev107 cDNA or H-rev107 antisense constructs gradually disappeared. No significant differences in the size of tumors could be observed at 20 d after injection. Cells transfected with the truncated H-rev107 cDNA (pCDNA107 ΔC) did not suppress tumor growth in vivo (Fig. 4). After 14 or 20 d the mice were killed, and cells from the tumors were explanted, grown in tissue culture, and analyzed for H-rev107 protein by immunofluorescence. The proportion of transfected tumor cells positive for H-rev107 protein was reduced as compared to the original cells. A representative culture of tumor explants is shown in Fig. 3 (B and D). Counting of independent microscopic fields of FE-8 transfectants (866 cells counted) and ANR4 transfectants (784 cells counted) revealed an average of 15% of cells staining positive with the H-rev107 antibody. These results suggest that either H-rev107 expression was downregulated in most of the cells during in vivo passaging or that the H-rev107 expressing cells are efficiently overgrown by the nonexpressing cells.

The product of the H-rev107 gene is an intracellular protein of 160 amino acids found either in the cytoplasmic or in the membrane fractions of cell extracts. Density-dependent growth arrest of 208F fibroblasts leads to an accumulation of the protein in the membrane fraction (Hajnal et al., 1994). To analyze the subcellular localization of this in vitro, weakly expressed protein in more detail, H-rev107 was transiently overexpressed in COS cells or in 208F immortal rat fibroblasts. Expression of the full-length protein using the strongly expressing vector pCMVTK-107+ was analyzed by Western blot technique, after biochemical fractionation of cellular extracts, and by immunofluorescence staining and subsequent inspection by confocal laser microscopy. Western blotting of nuclear, membrane, and cytoplasmic extracts of cells overexpressing H-rev107 revealed presence of the protein in all three fractions (Fig. 5 A). The cytoplasmic fraction shows a single protein band, while an additional band with slightly lower electrophoretic mobility is detected in the nuclear and the membrane fractions, suggesting posttranslational modification of the protein.

A plasmid expressing a truncated H-rev107 cDNA (pCDNA-107 ΔC), lacking the last 25 codons, was also overexpressed in COS cells. The missing region encodes predominantly hydrophobic amino acids and is probably responsible for membrane association. Deletion of this hydrophobic region leads to a loss of the H-rev107 protein in the nuclear and in the crude membrane fractions. The shortened H-rev107 ΔC protein is entirely found in the cytoplasmic fraction (Fig. 5,A), suggesting that the COOHterminal hydrophobic domain mediates the binding of H-rev107 to an intracellular membrane. In cells overexpressing full-length H-rev107, immunofluorescence analysis revealed no staining within the nucleus and at the plasma membrane. Rather, the H-rev107 protein appears to be linked to the nuclear membrane and to membranes in the perinuclear space (Fig. 5, B and C). As a considerable amount of the protein is present in the cytoplasm, staining of the free protein presumably masks the exact localization of the membrane-bound polypeptide.

High levels of H-rev107 mRNA expression have been observed in the RAS-transformation-resistant fibroblasts REF 52 and EK3 and in the phenotypic revertant cell line F9 obtained from HRAS-transformed cells. To further investigate the role of H-rev107 in neoplasia, we compared the expression of the gene in various normal rat tissues and tumor cell lines. Northern blot analysis of mRNA obtained from adult rat tissue showed ubiquitous expression of the H-rev107 gene (Fig. 6). Strong expression was observed in brain, kidney, liver, stomach, large intestine, spleen, muscle, and heart. A weaker expression was detected in breast and small intestine, and a very weak signal was observed in placenta. The H-rev107 protein is detectable in protein extracts from epithelial tissues of adult rats (Hajnal et al., 1994). Immunohistochemistry on cryosections of normal rat tissues using the H-rev107 antibody was used to localize H-rev107 expression in the tissues. In the colon, strong H-rev107-specific staining was limited to the differentiated epithelial cells. Similarly, epithelial specific expression was found in stomach, small intestine, and esophagus, in tracheal and bronchial epithelium, in testis, and in transitional epithelium of the bladder (Fig. 7). Weaker expression was also observed in the cardiac muscle and in spleen (data not shown).

In contrast to the normal tissues the rat liver carcinoma cell line ANR4 expressed only low levels of H-rev107 mRNA. No mRNA was detectable in two immortal kidney cell lines, NRK 49F and NRK 52E, or in the Ki-RAS transformed kidney cell line KNRK. The highest expression levels were detected in the colon carcinoma cell lines WB2054 and DHD (Fig. 1,A). The H-rev107 sequences transcribed in these tumorigenic cell lines are presumably wildtype, since analysis of RNAs prepared from WB 2054 and DHD cells by RT-PCR, SSCP, and sequencing of PCR fragments did not provide evidence for mutations in the coding sequences of H-rev107. However, analysis of the H-rev107 protein expression using the H-rev107 antibody revealed very low levels in WB2054 and no detectable staining in DHD (data not shown). H-rev107 mRNA was also expressed in the immortal, nontumorigenic pancreatic cell line AR4 IP and in normal human pancreas (Husmann, K., C. Sers, and R. Schäfer, manuscript in preparation). Very low levels were present in the two pancreatic carcinoma lines Cri-D11 and Cri-D2, and no H-rev107 mRNA could be detected in the tumorigenic pancreas line AR4 2J (Fig. 1 A).

To find out if H-rev107 expression is altered in malignant tissues, we analyzed tumors derived from transgenic mice carrying an activated HRAS oncogene controlled by the whey acidic protein promoter. The whey acidic protein promoter is activated upon onset of lactation in the mammary epithelial cells, leading to the development of mammary tumors in certain transgenic lines (Andres et al., 1987). H-rev107 mRNA expression was absent or very low in tumors L 25, 25.33G, and 25.72 (Fig. 8). Compared to the two control samples (Fig. 8; normal-A, normal-B), also the tumors 25.81, 25III13B, and 25.89 show a clear reduction in H-rev107 mRNA levels. Immunohistochemical analysis of the same malignant tissues failed to detect expression of the H-rev107 protein in spite of detectable mRNA. Western blot analysis of preneoplastic 208F rat fibroblasts and mouse NIH3T3 cells had shown a weak signal derived from the endogenous protein (Hajnal et al., 1994 and Sers, C., unpublished observations), but detection of the protein in individual cells was not possible. This is most likely due to a very low protein level, a phenomenon also observed in the rat colon carcinoma line WB2054, substantiating the notion that H-rev107 downregulation can occur at the RNA and the protein level.

Several lines of evidence suggest that H-rev107 is a transformation-suppressing gene. We have previously reported that H-rev107 mRNA is below the limit of detection in HRAS transformed cells as shown by Northern blot analysis, and that the gene was upregulated in the phenotypic revertant cell line F9 and in two different HRAS-transformation-resistant fibroblast lines (Hajnal et al., 1994). In this paper, we have shown that H-rev107 downregulation is frequent in tumorigenic rodent cell lines and in murine tumors, while the gene is ubiquitously expressed in the corresponding normal tissues. Most importantly, transfection of H-rev107 cDNA expression vectors into two different HRAS-oncogene expressing cell lines resulted in the suppression of colony formation. Clones permanently overexpressing the H-rev107 protein could not be obtained. In vivo, ectopic expression of H-rev107 leads to the attenuation of tumor growth in nude mice. Populations of explanted tumor cells did not express significant amounts of the H-rev107 protein.

What is the mechanism of growth suppression due to the overexpression of the H-rev107 gene? Only a minor reduction of colony formation was observed in cells transfected with the truncated H-rev107 gene (Table I). Expression of the truncated gene can be maintained stably in culture and did not result in a reduction of tumor growth (Fig. 1,B and Fig. 4). This suggests that the membrane attachment of the protein plays an important role in the proliferation–suppressive function in vivo. The H-rev107 protein is clearly not associated with the plasma membrane in 208F fibroblasts but rather with yet undefined intracellular membranes and presumably with the nuclear membrane. An attractive speculation is that the H-rev107 protein may serve as a negative regulator of an oncogenic HRAS- mediated signal by sequestration of kinases or of other growth-regulatory proteins. Members of the signal transduction kinase cascade have been shown to bind to anchoring proteins. These proteins target their catalytically active partners to specific subcellular sites and thereby control activation and phosphorylation of specific substrates (Mochly-Rosen et al., 1991; Coghlan et al., 1995). To understand the role of the H-rev107 protein in growth suppression in molecular detail, it is important to identify the target membrane and interacting proteins.

With constitutive expression of oncogenic RAS, slow growing cell populations overexpressing the H-rev107 gene are easily taken over by the fast-growing, nonexpressing cell populations. Thus, a detailed analysis of growth arrest in specific stages of the cell cycle was hampered by such an instability of H-rev107 high-expressers among RAS-transformants and by the low concentrations of the endogenous H-rev107 protein in cultured cells. The endogenous H-rev107 protein was detected exclusively in differentiated epithelial cells of several tissues, suggesting that it plays a role in differentiation-related growth arrest. We did not observe a significant fraction of apoptotic cells in transfected cell cultures (data not shown). Thus, the reduction of colony formation of transfected HRAS-transformed cells and the replacement of H-rev107-expressing cells by negative cells in vivo and in vitro may be explained by induced senescence, which is regarded as an important mechanism of tumor suppression (O'Brien et al., 1986; Shay and Wright, 1991).

The H-rev107 gene can be categorized as a class II tumor suppressor gene which has lost a growth-constraining function in tumor cells due to an expression block (Lee et al., 1991). In class I tumor suppressor genes such as TP53 or RB (for review see Vogelstein and Kinzler, 1992; Riley et al., 1994), loss of function results from mutation or deletion of DNA. Furthermore, the concept of two classes of tumor suppressor genes suggests that some class II genes are the targets of class I genes encoding transcriptional regulators (Lee et al., 1991). An increasing number of class II tumor suppressor genes have been identified on the basis of their downregulation in tumorigenic cells and owing to their transformation-inhibiting activity. The product of the K-rev gene, a Ras-like GTP-binding protein, was shown to interfere with oncogenic signaling (Kitayama et al., 1989; Frech et al., 1990; Culine et al., 1992). Other class II genes encode the cytoskeletal proteins tropomyosin (Fernandez et al., 1992) and vinculin (Prasad et al., 1993), the matrixmodifying enzyme lysyl oxidase (Contente et al., 1990; Krzyzoziak et al., 1992; Hajnal et al., 1993), gap junction proteins (Lee et al., 1992), and the invasion modulating serine protease maspin (Zou et al., 1994).

We have observed a concerted expression block to several H-rev genes in HRAS-transformed cells. Downregulation affected the H-rev107 gene (Hajnal et al., 1994), the H-rev142 gene encoding lysyl oxidase (Hajnal et al., 1993; Oberhuber et al., 1995), the H-rev18 gene encoding Ril, a novel member of LIM double zinc finger proteins involved in differentiation (Kiess et al., 1995), and Tsp-1 encoding the angiogenesis inhibitor thrombospondin (Shäfer, R., unpublished observations). These findings support the notion that continuous mitogenic signaling by mutated RAS specifically results in the coordinated downregulation of critical genes.

The repression of class II suppressor gene activity mediated by onco-proteins provides an important novel link between the activity of oncogenes and anti-oncogenes. The most intriguing feature of class II tumor suppressor inactivation is perhaps its reversibility triggered by various agents (Contente et al., 1990; Krzyzosiak et al., 1992; Oberhuber et al., 1995). Upregulation of H-rev gene expression in phenotypic revertants may occur in a coordinated fashion (Hajnal et al., 1993; Oberhuber et al., 1995). Several drugs have been identified that are able to reestablish the expression of downregulated genes and revert the transformed phenotype (Krzyzosiak et al., 1992; Kumar et al., 1995). The further development of those drugs may provide new therapeutic options for interfering with the activity of oncogenes commonly activated in human neoplasia.

We are grateful to Jacqueline Gheyselinck for technical assistance, Hannes Nef, Norbert Wey, and Christa Scholz for photographic work, and Marie-Theres Dours-Zimmermann and Dieter Zimmermann for helpful discussion. We also thank Otto Hagenbüchle, Lausanne, for AR4-IP and AR4-2J cells and H. Bujard, Heidelberg, for kindly providing the tetracycline plasmids.

This work was supported by the Schweizerischer Nationalfonds (grant 31.37525.93) and by the Krebsliga des Kantons Zürich.