A dual specificity promoter system combining cell cycle ... - Nature

Gene Therapy (1999) 6, 1276–1281  1999 Stockton Press All rights reserved 0969-7128/99 $12.00 http://www.stockton-press.co.uk/gt

A dual specificity promoter system combining cell cycle-regulated and tissue-specific transcriptional control ´ ˆ ¨ DM Nettelbeck, V Jerome and R Muller Institute of Molecular Biology and Tumor Research (IMT), Philipps-University Marburg, Emil-Mannkopff-Strasse 2, D-35033 Marburg, Germany

The expression of both proliferation-associated and cell type-specific genes is a hallmark of both cancer cells and tumor endothelial cells. The possibility to combine both features in a single transcriptional control unit would greatly increase the selectivity of vectors used for cancer gene therapy. Previous studies by our laboratory have shown that the transcription of several cell cycle genes is regulated by a novel cell cycle-regulated repressor, termed CDF-1. This repressor functions by blocking in resting cells the transcriptional activation by specific factors binding to the upstream activating sequence (UAS), most notably the CCAAT-box binding factor NF-Y/CBF. Based on this work we have developed a dual specificity promoter system that combines cell type specificity with cell cycle regulation. A

chimeric transcription factor (Gal4/NF-Y) consisting of the transactivation domain of NF-Y and the DNA-binding domain of Gal4 is expressed from a tissue-specific promoter. Gal4/NF-Y can bind to a second promoter consisting of a minimal cyclin A promoter with multiple Gal4 binding sites replacing the normal UAS. This leads to the tissue-specific expression of Gal4/NF-Y whose stimulatory activity on the promoter is restrained in resting cells by the recruitment of the CDF-1 repressor to the promoter. The functionality of this system is demonstrated for the specific transcriptional targeting of proliferating melanoma cells, where cell cycle regulation was ⬎20-fold and cell type specificity was ⬎50-fold.

Keywords: transcriptional targeting; cell cycle regulation; tissue-specific transcription; melanoma; cyclin A promoter; tyrosinase promoter

Introduction Cell proliferation is one of the hallmarks of both cancer cells and tumor endothelial cells that can be exploited for the targeting of malignant tumors by gene therapy.1 A frequently employed strategy makes use of tissue-specific promoters2 in conjunction with effector systems whose efficacy is dependent on cell division, such as the thymidine kinase/ganciclovir system.1,3 A limiting factor of this approach is the fact that tumor cells that are not proliferating at the time of treatment will not be eliminated. It would therefore be advantageous to achieve cell cycle dependence through the expression vector rather than the therapeutic gene. One way to accomplish this goal would be to combine cell type-specific2 and cell cycle regulated4 gene expression in the same promoter. We have previously shown that the cell cycle regulated transcription of a number of genes, including cdc25C and cyclin A, is determined by a common sequence motif comprised of two contiguous repressor elements, termed CDE and CHR.5,6 These elements are specifically occupied in G1-arrested cells as shown by in vivo footprinting, and interact with a novel repressor recognizing the CDE in the major groove and the CHR in the minor groove, ¨ Correspondence: R Muller Received 18 November 1998; accepted 8 March 1999

termed CDF-1.6,7 In both promoters, repression is dependent on the upstream activating sequence (UAS), indicating that the repressor complex functions by blocking transcriptional activators.8 Further studies employing defined activation domains tethered to a minimal cdc25C promoter showed that CDF-1 mediated repression is specific for a subset of activators and that the serine-threonine-glutamine-rich activation domain of the CCAATbox binding factor NF-Y/CBF9–20 is repressed most efficiently.21 In S/G2 cells, the CDE/CHR is not occupied so that activation via the UAS is permissible.8 The goal of the present study was to develop of a new strategy for the transcriptional targeting of proliferating tumor cells based on CDF-1 mediated repression, which we refer to as dual specificity promoter (DSP) system. In this approach, a combination of cell type-specific and cell cycle-regulated transcription is achieved by expressing from a tissue-specific promoter a chimeric transcription factor consisting of the serine-threonine-glutamine-rich transactivation domain of NF-Y19 and the DNA-binding domain of Gal4,22,23 which can bind to a minimal cyclin A promoter containing multiple Gal4 binding sites replacing the normal upstream activators. This leads to the tissue-specific expression of the NF-Y/Gal4 fusion protein whose stimulatory activity on the cyclin A promoter is under negative control of the cell cycle-regulated CDF-1 repressor in resting cells, as shown in a model of human melanoma cells.

Combination of cell cycle-regulated and tissue-specific transcriptional control DM Nettelbeck et al

1268

Results Description of the system The strategy established in the present study is based on the idea to regulate tissue-specific transcription in a cell cycle-dependent fashion by making use of the transcriptional repressor module CDE/CHR which interacts with the G0/G1-specific CDF-1. We have previously shown that CDF-1 mediated repression is specific for a subset of transcriptional activators and has the most dramatic effect on NF-Y.21 This activator specificity precludes the possibility of simply to insert a CDE/CHR module into a tissue-specific promoter. In fact, a series of such constructs was generated with a number of cell type-specific promoters/enhancers, including tyrosinase, tyrosinaserelated protein (TRP), troponin C, myogenin and von Willebrand factor, but in no case was it possible to achi¨ eve cell cycle regulated transcription (S Brusselbach, VJ, DMN and RM, unpublished observations). It was therefore necessary to design a strategy that takes into account the activator specificity of CDF-1-mediated repression. Toward this end we designed a system (DSP system) consisting of two components: (1) a chimeric transcriptional activator composed of the DNA-binding domain of Gal422 and the repressible activation domain of the A subunit of NF-Y,14 whose expression is driven by a tissuespecific promoter (activator construct); and (2) a minimal cyclin A promoter (harboring a CDE/CHR)8 with eight Gal4-binding sites24 driving the reporter or therapeutic gene (Figure 1). The constructs used for testing the DSP system are shown in Figure 2. As a model we used proliferating melanoma cells as the target and the human tyrosinase promoter25 for cell type-specific transcription. To minimize the size of the constructs we did not use the entire tyrosinase promoter, but a fusion of two copies of the major upstream enhancer (TDE) and the basal promoter region (TPE).25 This artificial promoter was able to drive transcription in melanoma cells efficiently and specifically: transcriptional activity was 40–70% of pGL3control (containing SV40 promoter and enhancer) in melanoma cells (B16, MeWo), and 10- to ⬎100-fold higher than in NIH3T3 fibroblasts, PC3 prostate carcinoma cells, HepG2 hepatoma cells, Saos-2 osteosarcoma cells and HeLa cells (data not shown). This promoter was used to drive expression of the Gal4/NF-Y activator. This construct will be referred to as Tyr-GN, the activator-responsive reporter plasmid as GalCycA. Cell cycle regulation Firstly, we investigated whether the DSP system allows for proliferation-dependent gene expression. MeWo cells synchronized in G1 were obtained by methionine deprivation, which decreases the fraction of S-phase cells to ⭐1 (BrdU incorporation) without any signs of apoptosis (Hoechst 33258 staining). MeWo cells were cotransfected with Tyr-GN and the reporter construct GalCycA and luciferase activities were determined in both normally growing and G1-arrested cells and standardized using the SV40 promoter construct SV40p. Figure 3 shows that the level of cell cycle regulation (ratio of activities in growing and arrested cells) was 23-fold which is comparable to the extent of regulation seen with the natural cyclin A promoter (26-fold). Importantly, the activity in G1 cells was in the range of the promoterless vector. As expected,

Figure 1 Schematic outline of the dual specificity promoter (DSP) system used to restrict the expression of transgenes to proliferating cells of a specific cell type. (a) A tissue-specific promoter directs the expression of a recombinant transcriptional activator (RTA), which is a fusion protein of the Gal4 DNA-binding domain and the NF-YA transcription activation domain (activator construct). The RTA binds to Gal4-binding sites upstream of a minimal cyclin A promoter harboring the CDE/CHR and drives the transcription of the reporter or therapeutic gene. The minimal cyclin A promoter is inactive on its own, because it lacks all upstream activator binding sites. Activation of transcription in the DSP system will therefore occur only in proliferating cells (CDE/CHR not occupied) of the target tissue (RTA synthesized). (b) In quiescent cells of the target tissue, the CDE/CHR element is occupied by a cell-cycle regulated repressor CDF-1 (CDE/CHR-binding factor-1) which inhibits transactivation by NF-YA. CDF-1 binds to the CDE/CHR element in quiescent and G1-cells only. (c) In cells of non-target tissues, the tissue-specific promoter is silent and the RTA is not expressed. CDE, cell cycle dependent element; CHR, cell cycle homology region; Gal4 bs, Gal4 binding sites.

mutation of the CDE (Tyr-GN + GalmCycA), which abrogates binding of the cell cycle-regulated repressor CDF1,8 greatly diminished the dependence on cell proliferation. This indicates the CDF-1 mediated repression is largely responsible for cell cycle regulation in the DSP system. The residual cell cycle regulation of five-fold seen with Tyr-GN + GalmCycA appears to be due to the intrinsic cell cycle dependence of the tyrosinase promoter alone (Tyr in Figure 3; 4.2-fold). The nature of this marginal cell cycle regulation of the tyrosinase promoter remains obscure, but this issue is not relevant for the purpose of the present study. The increased activity of TyrGN + GalmCycA in proliferating cells compared to TyrGN + GalCycA is likely to be a consequence of the loss of CDF-1 mediated repression in the G1 fraction of the normally cycling cell population. These results clearly


show the efficiency of the DSP system for controlling transcription in a proliferation-dependent manner.

Cell type-specific transcription Next, we tested the DSP system for cell type-specific transcription. MeWo cells, NIH3T3 fibroblasts and PC-3 prostate carcinoma cells were co-transfected with Tyr-GN and GalCycA, and luciferase activities were determined in normally growing cells and standardized as above. Figure 4 shows a 73-fold specificity for MeWo cells relative to NIH3T3 cells and a 58-fold specificity compared with PC-3 cells, with basically undetectable levels in the nontarget cells (similar to the promoterless vector). As expected, no significant activity and/or specificity were observed with the GalCycA reporter construct alone or with a cotransfected ‘activator’ plasmid lacking the NFY activation domain (Tyr-G). Likewise, cotransfection of an activator (CMV-GN) driven by the cytomegalovirus (CMV) early promoter/enhancer rather than the tyrosinase promoter led to a dramatically reduced specificity (approximately two- to eight-fold). These observations show that a clear tissue-specific expression can be achieved by the DSP system.

Figure 2 Establishment of the DSP system for proliferating melanoma cells: structure of the constructs. CMVp/e, CMV promoter and enhancer; Tyrp, tyrosinase promoter; TDE, tyrosinase promoter distal element; TPE, tyrosinase promoter proximal element; CycAp, cyclin A promoter; Gal4 bs, Gal4 binding sites.

Figure 3 Proliferation-dependent transcription achieved by the DSP system after transient cotransfection of MeWo cells with activator (Tyr-GN) or pUC19 (controls) and luciferase reporter constructs. For technical details see Materials and methods. Values were standardized individually for proliferating and methionine-deprived G1-arrested cells using an SV40 promoter (SV40p) construct. The numbers to the right of the bars represent the ratio of luciferase activities in proliferating cells and methionine-deprived cells. The data shown are averages of triplicates ± standard deviation.

Induction of a biological effect by the DSP system Although the activity in MeWo cells was lower than with the tyrosinase promoter construct on its own, it was approximately three-fold higher than that seen with the SV40 promoter, indicating that the DSP system allows for an efficient transcriptional activity. In order to investigate whether the level of transcription achieved by the DSP system is sufficient to obtain a biological effect we performed an in vitro TNF-␣ cytolysis assay.26 This assay, which measures cytotoxic effects on the TNF-␣ sensitive L929 cell line, was performed with the medium of MeWo cells cotransfected with activator (Tyr-GN) and effector (GalCycATNF) constructs. Details of the assay can be found in the Materials and methods. Forty-eight hours

Figure 4 Cell type-specific transcription achieved by the DSP system. Melanoma (MeWo) and non-melanoma (NIH3T3, and PC-3) cell lines were transiently transfected with activator (CMV-GN, Tyr-GN or TyrG) or pUC19 (controls) and luciferase reporter constructs. Values were standardized individually for each cell line using an SV40 promoter (SV40p) construct. The numbers to the right of the bars represent the ratio of luciferase activities in melanoma cells and control cells. The data shown are averages of triplicates ± standard deviation.

1269


1270

after transfection cytolysis was in the range of 60% (corresponding to approximately 1.0 ng/ml TNF␣ in the supernatant; Figure 5a, b). In contrast, the supernatant of cells transfected with the GalCycATNF effector construct alone showed a negligible extent of cell death (⬍2%). These results show that the level of transcription achievable by the DSP system is sufficient to obtain a clear biological effect.

Discussion In the present study, we have introduced a new strategy for the transcriptional targeting of proliferative disorders, such as malignancies, by gene therapy. This DSP system eliminates a common disadvantage associated with the use of tissue-specific promoters, which is the fact that expression of the transgene occurs also in the normal tissue from which the tumor originated. The principle of the DSP system is to couple tissue-specific transcription with cell cycle regulation. This was achieved by controlling in a cell cycle-dependent fashion (through the G0/G1-specific repressor CDF-1) the activator function of

Figure 5 TNF␣ cytolysis assay showing that the DSP system gives rise to a level of gene expression that is sufficient to obtain a biological effect. (a) Standard curve showing the survival of L929 cells exposed to recombinant TNF␣ (in conditioned medium form untransfected MeWo cells) giving an IC50 value of 0.7 ng/ml. For technical details see Materials and methods. (b) Assay of supernatants from untransfected MeWo cells ± TNF␣ and transfected MeWo cells expressing TNF␣ directed by the DSP system (as in Figures 3 and 4). Cotransfection of activator (Tyr-GN) and effector (GalCycATNF) constructs lead to approximately 60% cytolysis of L929 cells (corresponding to approximately 1.0 ng/ml TNF␣ in the supernatant). The supernatant of cells transfected with the reporter construct alone showed a marginal effect only (⬍2% cytolysis). The percentages of cytolysis were calculated relative to the −TNF and +TNF readings (0% and 100% cytolysis, respectively). Values are the mean of triplicate transfections; the error bars indicate the standard deviations.

an artificial transcription factor (Gal4/NF-YA), which is expressed from a tissue-specific promoter. In order to prove the principle of the DSP system we have used a human melanoma model. We show that expression of a luciferase reporter gene occurs preferentially in the proliferating melanoma cells using a combination of a Gal4/NF-YA-activated cyclin A promoter and elements of the tyrosinase promoter driving the Gal4/NF-YA activator. Cell type specificity was in the range of 58- to 73fold when compared with cell lines of other origins, and the extent of cell cycle regulation was 23-fold, which is similar to the wild-type cyclin A promoter. Importantly, transcription directed by the DSP system was undetectable in both non-proliferating cells and non-target cells, ie comparable with the promoterless reporter plasmid. In contrast, in proliferating target cells, efficient transcription was observed. Thus, the level of expression that can be obtained with the DSP system was sufficient to induce a clear biological effect in a TNF-␣ cytolysis assay. Several important questions need to be addressed now. Firstly, it has to be analyzed whether the system with respect to its transcriptional activity can be further improved, eg by incorporating all components into a single plasmid vector. This should not prove difficult in view of the relatively small size of all components (approximately 2.5 kb in total excluding the reporter/therapeutic gene for the melanoma system described in the present study). Another possibility would be to implement a positive feedback loop which amplifies the signal once the promoter becomes active.27 Secondly, we have shown that the DSP system is functional using a non-viral vector, ie a plasmid–cationic lipid complex. At present viral vectors are the preferred means for transduction in vivo. It will therefore be important to address the critical question as to whether viral sequences will interfere with the DSP system. We are currently in the process of generating replication-deficient adenoviruses harboring the DSP system to investigate potential transcriptional interference and to study the functionality of the DSP system in vivo. Thirdly, the DSP system has to be tested with other tissue-specific promoters. In this context, endothelial cell-specific promoters will be of great interest, since the DSP system might allow for the specific transcriptional targeting of proliferating endothelial cells found in the angiogenic areas of tumors. The availability of the DSP system should facilitate the development of new vectors for the targeting of proliferating diseases, since the only determinants of selective gene expression are cell type and cell proliferation. Such diseases include not only malignant tumors, but also conditions such as arthritis, vascular diseases and certain immunological disorders.28 We also believe that the development of the DSP system represents a good example of how the understanding of fundamental transcriptional control mechanisms, such as the CDF-1 mediated cell cycle regulation of gene expression, can be translated into applied science and provide new tools for the development of novel therapeutic strategies.

Materials and methods Cell culture The murine cell lines NIH3T3 (mouse fibroblast; obtained from Dr R Treisman, ICRF, London, UK), and L929


(transformed fibroblast, provided by Dr M Clauss, M91 Bad Nauheim, Germany), the human cell lines PC-3 (prostate carcinoma; ATCC CRL-1435; obtained from Prof ¨ G Aumuller, University of Marburg, Germany), and MeWo (melanoma; provided by Prof I Hart, St Thomas Hospital, London, UK)29 were maintained at 37°C in a humidified atmosphere of 5% CO2. PC-3 and L929 cells were cultured in RPMI 1640 (Boehringer Ingelheim BioProducts, Heidelberg, Germany) and the other cell lines in Dulbecco-Vogt modified Eagle’s medium (DMEM, Boehringer Ingelheim BioProducts). RPMI 1640 and DMEM were supplemented with 5% (L929 cells) or 10% (all others) fetal bovine serum (FBS, Life Technologies, Eggenstein, Germany) and 2 mm l-glutamine (Boehringer Ingelheim BioProducts). Cell proliferation was measured by incorporation of 5-bromodeoxyuridine (5-BrdU) followed by immunostaining as described.30 Apoptosis was determined by staining with Hoechst 33258.31

Plasmid constructs pGL3basic (basic), pGL3promoter (SV40p) and pGL3control (containing both SV40 promoter and enhancer) were purchased from Promega, Mannheim, Germany. The basic construct lacking eukaryotic promoter and enhancer sequences was used for determining the vectorrelated background activity. The SV40p construct was used for standardizing transfection efficiencies. pUC19 was obtained from New England Biolabs (Schwalbach, Germany). pGL3CMV and CycA were cloned by inserting a BamHI/HindIII fragment of pcDNA3 (Invitrogen, Groningen, The Netherlands) and a human cyclin A promoter fragment (−214 to +100),8 respectively, into the BglII/HindIII sites of pGL3basic. The Tyr construct contains twice the human tyrosinase enhancer element TDE (nucleotides −2014 to −1811)25 cloned as 5⬘KpnI/3⬘-NheI and 5⬘-NheI/3⬘-XhoI PCR-amplified fragments and the tyrosinase promoter element TPE (−209/+51)25 cloned as 5⬘-XhoI/3⬘-BglII PCR-amplified fragment into the corresponding restriction sites of pGL3. The reporter constructs GalCycA (see Figure 2), and GalmCycA contain eight 17-mer GAL4 binding sites (5⬘CGGAGTACTGTCCTCCG-3⬘)24 inserted upstream of the cyclin A promoter (−40/+94)32 harboring either the wildtype CDE or a mutant CDE (TCGCGGG → TCGCTGG)8 in pGL3. For constructing GalCycATNF, the luciferase cDNA of GalCycA was replaced by the TNF␣ cDNA of pAS3 (obtained from M Clauss, Bad Nauheim, Germany). pAS3 contains the mouse TNF␣ cDNA cloned into the PstI/EcoRI sites of pBluescriptIISK. The activator constructs (see Figure 2) encode fusion proteins containing the sequences of the Gal4 DNA binding domain (aa 1– 147)22 and the murine NF-YA S-, T- and Q-rich transcriptional activation domain (aa 1–261).14 For cloning of CMV-GN and Tyr-GN, the luciferase gene of the pGL3CMV and Tyr plasmids were replaced by an EcoRI/BglII fragment of the GAYA11 plasmid obtained from R Mantovani, University of Milan, Italy (including the sequences coding for the Gal4/NF-YA fusion (see Figure 2).14 The NF-YA deleted control Tyr-G was obtained by replacing the mNF-YA transcriptional activation domain of Tyr-GN by a stop codon inserted immediately downstream of codon 147 of Gal4. All plasmids were amplified in E. coli and purified according to a standard protocol (Qiagen, Hilden, Germany). Plasmids

obtained by PCR-amplification were verified by DNAsequencing.

Transfections and luciferase assays NIH3T3, PC-3 and MeWo cells were transfected at 70% confluence in six-well plates using the cationic liposome DOTAP as described by the manufacturer (Boehringer Mannheim). One microgram of reporter plasmid and 2 ␮g of activator plasmid or pUC19 were used per well and mixed with 6 ␮l of DOTAP in transfection buffer. The transfection mixture was mixed with OptiMEM (Life Technologies, Eggenstein, Germany) and added to the cells. After 6 h of incubation, cells were washed once with PBS and fresh medium for growing cells or methioninefree medium containing 1/100 ITS-A supplement (both Life Technologies) for synchronization in G1 was added. Luciferase activities were determined after 25 h (growing cells) or 60 h (methionine deprived cells) as described.33 Transfections were performed in triplicates and repeated at least once with an independent plasmid preparation. To achieve higher transfection efficiencies for the tumor necrosis factor-␣ (TNF␣) bioassay, MeWo cells were transfected using Lipofectin (Life Technologies) according to the manufacturer’s protocol. One microgram of GalCycATNF and 2 ␮g pUC19 or Tyr-GN were mixed with 10 ␮l lipofectin in OptiMEM and incubated with the cells for 6 h. TNF␣ bioassay using L929 cells MeWo cells were co-transfected with GalCycATNF/ pUC19 or GalCycATNF/Tyr-GN in triplicates. Twentyfour hours after transfection the medium was replaced and collected after another 24 h. The culture supernatants were assayed for TNF␣ bioactivity by determining the cytotoxicity on the transformed mouse fibroblast line L929 as previously described.26 Briefly, L929 cells were seeded in microtiter plates at 4 × 104 cells per well. Serial dilutions of mouse TNF␣ in conditioned medium of nontransfected MeWo cells and the supernatants of the transfected cells were added in triplicates 16 h later, together with actinomycin D at a final concentration of 1 ␮g/ml. After 24 h, the remaining L929 cells were fixed and stained with crystal violet and the retained dye was measured with an ELISA-reader at ␭540 nm.

Acknowledgements ¨ We are grateful to Dr S Brusselbach for useful dis¨ cussions, to Professors G Aumuller, I Hart and HH Sedla¨ cek for tumor cell lines, to Drs M Clauss and D Mannel for help with the TNF␣ assay and reagents and to Dr M Krause for synthesis of oligonucleotides.

References 1 Sikora K. Gene therapy for cancer. Trends Biotech 1993; 11: 197–201. 2 Miller N, Whelan J. Progress in transcriptionally targeted and regulatable vectors for genetic therapy. Hum Gene Ther 1997; 8: 803–815. 3 Rigg A, Sikora K. Genetic prodrug activation therapy. Mol Med Today 1997; 3: 359–366. ¨ 4 Muller R. Transcriptional regulation during the mammalian cell cycle. Trends Genet 1995; 11: 173–178. 5 Lucibello FC et al. Periodic cdc25C transcription is mediated by a novel cell cycle-regulated repressor element (CDE). EMBO J 1995; 14: 132–142.

1271


1272

6 Zwicker J et al. Cell cycle regulation of the cyclin A, cdc25C and cdc2 genes is based on a common mechanism of transcriptional repression. EMBO J 1995; 14: 4514–4522. 7 Liu N et al. CDF-1, a novel E2F-unrelated factor, interacts with cell cycle-regulated repressor elements in multiple promoters. Nucleic Acids Res 1997; 25: 4915–4920. 8 Zwicker J et al. Cell cycle regulation of cdc25C tanscription is mediated by the periodic repression of the glutamine-rich activators NF-Y and Sp1. Nucleic Acids Res 1995; 23: 3822–3820. 9 Dorn A et al. A multiplicity of CCAAT box-binding proteins. Cell 1987; 50: 863–872. 10 Maity SM, Golumbek PT, Karsenty G, de Crombrugghe B. Selective activation of transcription by a novel CCAAT binding factor. Science 1988; 241: 582–585. 11 van Hudisduijnen HR et al. Co-evolution from yeast to mouse: cDNA cloning of the two NF-Y (CP-1/CBF) subunits. EMBO J 1990; 9: 3119–3127. 12 Maity SN, Vuorio T, Crombrugghe B. The B subunit of a rat heteromeric CCAAT-binding transcription factor shows a striking sequence identity with the yeast Hap2 transcription factor. Proc Natl Acad Sci USA 1990; 87: 5378–5382. 13 Maity SN, Sinha S, Ruteshouse EC, de Crombrugghe B. Three different polypeptides are necessary for DNA binding of the mammalian heteromeric CCAAT binding factor. J Biol Chem 1992; 267: 16574–16580. 14 Li X-Y et al. Intron–exon organization of the NF-Y genes. Tissuespecific splicing modifies an activation domain. J Biol Chem 1992; 267: 8984–8990. 15 Mantovani R et al. Monoclonal antibodies to NF-Y define its function in MHC class II and albumin gene transcription. EMBO J 1992; 11: 3315–3322. 16 Mantovani R et al. Dominant negative analogs of NF-YA. J Biol Chem 1994; 269: 20340–20346. 17 Sinha S, Maity SN, Lu J, de Crombrugghe B. Recombinant rat CGF-C, the third subunit of CGF/NFY, allows formation of a protein–DNA complex with CBF-A and CBF-B and yeast HAP2 and HAP3. Proc Natl Acad Sci USA 1995; 92: 1624–1628. 18 Sinha S et al. Three classes of mutations in the A subunit of the CCAAT-binding factor CBF delineate functional domains involved in the three-step assembly of the CBF-DNA complex. Mol Cell Biol 1996; 16: 328–337.

19 Coustry F, Maity SN, de Crombrugghe B. Studies on transcription activation by the multimeric CCAAT-binding factor CBF. J Biol Chem 1995; 270: 468–475. 20 Currie RA. NF-Y is associated with the histone acetyltransferases GCN5 and P/CAF. J Biol Chem 1998; 273: 1430–1434. 21 Zwicker J et al. CDF-1 mediated repression of cell cycle genes targets a specific subset of transactivators. Nucleic Acids Res 1997; 25: 4926–4932. 22 Lech K, Anderson K, Brent R. DNA-bound fos proteins activate transcription in yeast. Cell 1988; 52: 179–184. 23 Chasman DI, Kornberg RD. GAL4 protein: purification, association with GAL80 protein and conserved domain structure. Mol Cell Biol 1990; 10: 2916–2923. 24 Webster N et al. The yeast UASG is a transcriptional enhancer in HeLa cells in the presence of the GAL4 trans-activator. Cell 1988; 52: 169–178. 25 Siders WM, Halloran PJ, Fenton RG. Transcriptional targeting of recombinant adenoviruses to human and murine melanoma cells. Cancer Res 1996; 56: 5638–5646. 26 Hwu P et al. Use of soluble recombinant TNF receptor to improve detection of TNF secretion in cultures of tumor infiltrating lymphocytes. J Immunol Meth 1992; 151: 139–147. ´ ˆ ¨ 27 Nettelbeck DM, Jerome V, Muller R. A new strategy for the enhancement of weak cell type-specifc promoters. Gene Therapy 1998; 5: 1656–1664. ¨ ¨ 28 Muller R, Nettelbeck D, Sedlacek HH. Chimare Promotormodule mit rekombinanten Transkriptionsfaktoren. Patent Application 1998; DE 198 314 20.5. 29 Bean MA et al. Cell-mediated cytotoxicity for bladder carcinoma: evaluation of a workshop. Cancer Res 1975; 35: 2902–2913. 30 Sewing A et al. Human cyclin D1 encodes a labile nuclear protein whose synthesis is directly induced by growth factors and suppressed by cyclic AMP. J Cell Sci 1993; 104: 545–555. ¨ ¨ 31 Lau QC, Brusselbach S, Muller R. Abrogation of c-Raf expression induces apoptotis in tumor cells. Oncogene 1998; 16: 1899–1902. 32 Henglein B et al. Structure and cell cycle-regulated transcription of the human cyclin A gene. Proc Natl Acad Sci USA 1994; 91: 5490–5494. 33 Ausubel I, Frederick M. Current Protocols in Molecular Biology. John Wiley: New York, 1991.