Gene Therapy (2001) 8, 725–729  2001 Nature Publishing Group All rights reserved 0969-7128/01 $15.00 www.nature.com/gt BRIEF COMMUNICATION A synthetic zipper-based dimerization system for combining multiple specificities

VJe´roˆme and R Mu¨ ller Institute of Molecular Biology and Tumor Research (IMT), Philipps University, Emil-Mannkopff-Strasse 2, 35033 Marburg, Germany

One of the biggest challenges facing gene therapy is the with different selectivity. A crucial determinant of the DCTF development of vectors that direct the activity of therapeutic system is the heterodimerization interface that should pro- genes specifically to the sites of disease. To achieve this vide for a high affinity interaction without interference by goal, the restriction of transgene via synthetic endogenous proteins. Here, we describe such a dimerization promoters that are endowed with multiple specificities rep- system based on engineered Fos and Jun leucine zippers. resents a particularly promising strategy. Towards this end, We show the usefulness of this system for the combination we have developed a generally applicable strategy (DCTF of cell type-specific and -regulated transcription system) where a synthetic promoter is driven by an artificial and demonstrate its functionality in an in vivo setting. Gene heterodimeric whose DNA-binding and Therapy (2001) 8, 725–729. transactivating subunits are expressed from two promoters

Keywords: cell cycle regulation; DCTF system; ; tissue-specific promoter; transcriptional targeting

The dual specificity chimeric transcription factor (DCTF) acids in the Jun LZ and the negatively charged residues system has been established for the targeting of prolifer- in the Fos counterpart leading to strong self-repulsion of ating melanoma cells.1 Two subunits harboring the DNA- both proteins.11,12 Both, Jun and Fos family members are binding domain of the yeast transcription factor GAL4 or expressed to variable extents in all cells. To make the the transactivation domain of VP16 Jun/Fos LZs suitable for the DCTF system it was there- were expressed from the tyrosinase and cyclin A pro- fore necessary to prevent interactions between the chim- moter, respectively. Heterodimerization was achieved eric transcription factors of the DCTF system and the through the interaction surfaces of the CD4 and the Lck endogenous Jun and Fos proteins. We hypothesized that protein kinase. Although proof of principle could be this could be possible by swapping between the two LZs demonstrated by this system, it was clear that the overall several of the oppositely charged amino acids involved transcriptional activity had to be improved for a potential in heterodimerization. application in gene therapy.1 Since the CD4–Lck Based on the resolved crystal structure12 we introduced interaction is weak and potential interactions with three acidic amino acids from c-Fos into the c-Jun LZ endogenous molecules might result in the sequestration (mJun) and three basic amino acids from c-Jun into the c- of the chimeric transcription factor used in the DCTF sys- Fos LZ (mFos), as depicted in Figure 1a, thereby creating tem, we reasoned that the replacement of the hetero- mixed-charged structures. To analyze the oligomeriz- dimerization interface would lead to a substantial ation properties of transcription factors carrying such improvement. engineered LZs in the context of the herpes simplex virus The leucine zippers (LZs) represent particularly strong VP16 transactivation domain13 (VP16-mJun) or the yeast interaction domains commonly frequent in transcription GAL4 DNA-binding domain14 (GAL4-mFos) we investi- factors2 and have been shown to be functional in a heter- gated the association of in vitro translated proteins by ologous context.3–5 Dimerization through LZs is mediated radioimmunoprecipitation (RIP). 35S-methionine-labeled by regularly spaced (heptad repeats) in parallel VP16-Jun proteins (VP16-mJun, VP16-wtJun with a wild- ␣-helices through hydrophobic interactions, while the type Jun LZ, and VP16-⌬Jun lacking a dimerization choice of the dimerization partner is determined by other domain; Figure 1b) were mixed with unlabeled GAL4- amino acids, mainly charged residues forming salt mFos proteins and immunoprecipitated with an bridges.6–11 This is exemplified by the Fos and Jun pro- directed against GAL4. Figure 1c shows that the proteins teins which preferentially form heterodimers and only containing engineered LZs formed heterodimers, but that have a very limited ability to homodimerize. This speci- complex formation between VP16-wtJun and GAL4-mFos ficity is brought about by the positively charged amino or VP16-⌬Jun and GAL4-wtFos was prevented. Likewise, VP16-mJun did not interact with GAL4-wtFos in an analogous experiment (not shown). Correspondence: R Mu¨ller These observations suggest that the mJun and mFos LZ Received 8 December 2000; accepted 30 January 2001 domains could be suitable for mediating specific intra- In vivo transcriptional targeting VJe´roˆme and R Mu¨ller 726

Figure 1 (a) Structure of the chimeric transcription factors. Schematic representation of the GAL4-fos and VP16-jun fusion proteins. Electrostatic interactions between Fos and Jun LZ are shown by arrows.12 The amino acids swapped between the leucine zippers are indicated by black arrows. (b) SDS-PAGE of 35S-methionine-labeled in vitro translated Fos and Jun LZ fusion proteins. (c) Complex formation between VP16-mJun and GAL4-mFos LZ fusion proteins. Wild-type VP16-wtJun and mutated VP16-mJun were translated in the presence of 35S-methionine, complexed with unlabeled mutated GAL4-mFos and immunoprecipitated with a GAL4-specific antibody. The precipitated proteins were analyzed by SDS-PAGE followed by autoradiography. As a control an empty pGEM4 vector was translated in the presence of 35S-methionine and complexed with unlabeled GAL4-mFos; as a second control the labeled VP16-⌬Jun was incubated with unlabelled GAL4-mFos. For in vitro transcription/translation assay, the chimeric transcrip- tion factors were placed under the control of the T7 promoter. They were excised from the pGAL-wtFos, pGAL-mFos, pVP⌬Jun, pVP-wtJun, and pVP- mJun constructs (see legend to Figure 2) as HindIII/XbaI fragments and cloned in the same sites into the pGEM4 vector (Promega, Mannheim, Germany) to yield GAL4-wtFos, GAL4-mFos, VP16-⌬Jun, VP16-wtJun andVP16-mJun. In vitro translation was performed with the TNT T7 quick (Promega) according to the manufacturer protocol. Briefly, 1 ␮g of DNA template was mixed with 40 ␮l of TNT T7 master mix and 2 ␮l 35S-methionine (1000 Ci/mmol, 10 mCi/ml; Amersham, Freiburg, Germany) and incubated at 30°C for 90 min. To assess the quality of the translated products, 5 ␮l of the translation reaction was analyzed on SDS-PAGE 12.5%. To avoid nonspecific interactions of the rabbit reticulocyte lysate with the rabbit anti- serum used later on to perform the immunoprecipitation, the in vitro translated products were pre-cleared. Twenty ␮l of proteins were incubated with 5 ␮l of normal rabbit serum (Vector Laboratories, Gru¨nberg, Germany) at 4°C for 1 h before adding 20 ␮l of protein-A sepharose (Pharmacia, Freiburg, Germany) and incubating for 30 min. After centrifugation at 4°C for 5 min at 13 000 g the pre-cleared supernatant was ready for the in vitro association assay. For in vitro protein-binding assays, equal volumes of pre-cleared supernatant (5 ␮l each) containing 35S-methionine-labeled VP16-⌬Jun, VP16- wtJun or VP16-mJun proteins in 50 mm Tris pH 7.2 were combined and incubated at 30°C for 90 min as described.20 Protein complexes were mixed with 2 ␮l rabbit polyclonal IgG GAL4 (DBD) (Santa Cruz, Heidelberg, Germany). The immune complexes were collected by the addition of sepharose- A, washed in RIPA buffer (10 mm Tris-Cl (pH 7.5), 150 mm NaCl, 1% NP40, 1% deoxycholate, 0.25% SDS 1 mm DTT) and prepared for SDS- PAGE. The immunoprecipitated proteins were separated on 12% gels followed by autoradiography.

cellular protein–protein interactions. We therefore ana- wild-type LZs, nor the combination of one wild-type and lyzed in the next step the functionality of the VP16-mJun one mutant LZ, was able to enable heterodimerization. and GAL4-mFos proteins in a cellular environment using The former is presumably due to the sequestration of the the DCTF system. For this purpose, we designed recombinant factors by endogenous Fos and Jun proteins, expression constructs where VP16-wtJun, VP16-mJun the latter due to structural reasons, as supported by the and VP16-⌬Jun were under the control of the cyclin A in vitro association data shown in Figure 1b. promoter, and GAL4-wtFos and GAL4-mFos were driven We next attempted to optimize the transcriptional by a tyrosinase promoter.1 An SV40 promoter-driven activity achievable by the Jun/Fos DCTF system by clon- luciferase gene with 10 GAL4 binding sites (10xgal4RE- ing both expression cassettes into one plasmid in a head- SVpGL3) served as reporter construct. Cotransfections of to-tail configuration, rationing that this would lead to melanoma cells (MeWo) in various combinations clearly equimolar concentrations of both subunits in the trans- showed that only the coexpression of the VP16-mJun fected cell. As shown in Figure 2b, this strategy lead to (pVP-mJun) and GAL4-mFos (pGAL-mFos) proteins an ෂ24-fold increase (36:1.5) in luciferase activity in mela- gave rise to any significant activity (ෂ1.5-fold SV40 pro- noma cells (36-fold SV40 promoter). We also compared moter; Figure 2a). Cotransfection of GAL4-mFos with the activity of the Jun/Fos DCTF system with that of the either VP16-wtJun (pVP-wtJun) or VP16-⌬Jun (pVP- previously described CD4-LCK dimerization system.1 ⌬Jun) gave only marginal signals (in the range of the pro- The transcriptional activity of the Jun/Fos ‘one-plasmid’ moterless luciferase plasmid pGL3), as did the coex- system was found to be ෂ7-fold greater relative to the pression of GAL4-wtFos (pGAL-wtFos) with any of the corresponding CD4-LCK system (36:5.5). Overall, the VP16-Jun proteins. Thus, neither the combination of two DCTF system described in the present study gives rise to

Gene Therapy In vivo transcriptional targeting VJe´roˆme and R Mu¨ller 727

Figure 2 (a) Intracellular formation of specific functional heterodimers. The human melanoma (MeWo) cell line was transiently cotransfected with 2 ␮g pGAL-wtFos or pGAL-mFos, 1 ␮g pVP-wtJun or pVP-mJun or pVP-⌬Jun and 0.5 ␮g reporter construct (10× gal4RE-SVpGL3). As positive or negative controls, 1 ␮g SV40pGL3 or pGL3 were transfected. (b) Comparative analysis of ‘one-plasmid’ and ‘two-plasmids’ DCTF using either mJun/mFos or CD4/LCK dimerization interfaces. In the case of the ‘one-plasmid’ system, MeWo cells were cotransfected with 1 ␮g of the indicated expression plasmid and 1 ␮g10× gal4RE-SVpGL3. For the ‘two-plasmids’ system, cells were cotransfected with 1 ␮g of each expression plasmid and 1 ␮g10× gal4RE- SVpGL3. The ⌬Jun and ⌬CD4 plasmids lacking a dimerization domain were included as negative controls. The numbers to the right of the bars represent the activity relative to SV40pGL3. Data represent the average of four separate experiments, each performed in duplicate, ± standard deviation. (c) Selective transgene expression in proliferating melanoma cells by the mJun/mFos DCTF system. Cells were cotransfected with 1 ␮g pGALmFos-VP⌬Jun or pGALmFos-VPmJun and 0.5 ␮g of the reporter construct (10× gal4RE-SVpGL3). The cells were then cultured either in normal or in methionine- free medium and assayed for luciferase activity 54 h after-transfection. The graph shows the standardized luciferase activities (ratio of pGALmFos- ⌬ VPmJun to pGALmFos-VP Jun) in G1-arrested (methionine-deprived) and growing cells. Data represent the average from four separate experiments, each performed in duplicate, ± standard deviation. pGL3-basic (pGL3) and pGL3promoter (SV40pGL3) were purchased from Promega. The pGL3promoter construct was used as a cloning vector and for normalizing transfection efficiencies. The 10× gal4RE-SVpGL3 and pVP-⌬Jun (CycA-AD in the previous publication) constructs have been described previously.1 To generate GAL4-wtFos, the c-Fos leucine zipper (amino acids 160–196)12 was fused to the 147 amino-terminal residues of GAL414 with the nuclear localization signal of the simian virus 40 large T antigen19 (pGAL-wtFos). The pGAL-mFos construct was generated as described above and carried the following : E-167→K; E-172→K; E-181→R. The expression of these fusions is driven by the chimeric human tyrosinase promoter containing two tyrosinase enhancers upstream from the basal promoter.1 The VP16-wtJun was constructed by fusing the c-Jun leucine zipper (amino acids 276–312)12 to the nuclear localization signal of SV40 large T antigen and the transcrip- tional activation domain of HSV VP16 (amino acids 411–455)13 (pVP-wtJun). The pVP-mJun construct was generated as described above and carried the following mutations in the leucine zipper (K-283→E; K-288→E; R-302→E). The expression of this fusion is driven by the human cyclin A promoter (−214 to +100; CycA).1 To generate the pGALmFos-VP⌬Jun and pGALmFos-VPmJun constructs, the cassettes containing the CycA-driven chimeric transcription factors were cloned in a tail-to-tail orientation into the pGAL-mFos plasmid. The constructs pGALLck-VP⌬CD4 and pGALLck-VPCD4 were obtained by cloning the CycA-driven chimeric transcription factors in a tail-to-tail orientation into the Tyr-LCK plasmid.1 The human cell lines MeWo21 (melanoma), PC3 (prostate carcinoma; ATCC CRL-1435) and NCI-H322 (bronchioloalveolar carcinoma; ATCC CRL-5806) were maintained ° at 37 Cin5%CO2 in DMEM or RPMI 1640 supplemented with 10% fetal bovine serum. The MeWo cell line was obtained from Prof I Hart (ICRF, London, UK) and the NCI-H322 cell line from Dr M Favrot (Institut Albert Bonniot, Grenoble, France). Cells were plated on 35 mm (diameter) tissue culture plates at a density yielding 60 to 80% confluency at the time of the transfection with DOTAP as described by the manufacturer (Boehringer

Mannheim, Mannheim, Germany). For synchronization in G1, cells were cultivated for 48 h in methionine-free medium (GIBCO-BRL, Karlsruhe, Germany) with 1% ITS-A supplement (GIBCO-BRL) after the removal of DNA–DOTAP complexes. Fifty-four hours after transfection, luciferase activities were determined as described.22

Ͼ a 100-fold higher transcriptional activity compared carcinoma) and in cycling versus G1-arrested cells. The with the originally published system1 (36:0.2). It is poss- latter cells were obtained by methionine deprivation as ible that the performance of the DCTF system can be described previously.1 As shown in Figure 2c, the activity further improved by incorporating all components into a in both H322 and PC3 cells was marginal (р3.5% of single plasmid vector, which is particularly relevant if the MeWo). In addition, transcriptional activity was ෂ9-fold system is to be used in the context of a viral vector. This greater in normally proliferating relative to G1-arrested should not be difficult in view of the relatively small size MeWo cells. Thus, only the cycling target cells provide of all components (Ͻ3 kb in total excluding the an environment that allows for the Jun/Fos DCTF system reporter/therapeutic gene). to be active. To assess the cell-type specificity and cell cycle regu- Finally, we addressed the functionality of the Jun/Fos lation of the ‘one-plasmid’ Jun/Fos DCTF system we ana- DCTF system in vivo. We established melanoma (MeWo) lyzed its transcriptional activity in different cell types and lung carcinoma (H322) xenografts in nude mice and (MeWo melanoma, H322 lung cancer and PC3 prostate the pGL3promoter vector or the ‘one-plasmid’ Jun/Fos

Gene Therapy In vivo transcriptional targeting VJe´roˆme and R Mu¨ller 728 DCTF system was injected intratumorally as described in tors it was possible to enhance promoter activity more the legend to Figure 3. To be able to standardize the data than 100-fold to a level that makes this system now appli- and to correct the results for variations in transduction cable to gene therapy. In addition to this specific appli- efficiency we co-injected a SV40 promoter Renilla lucifer- cation, the mJun/mFos LZ domains should be useful in ase construct, whose activity can be determined indepen- addressing many other problems, for example, directing dently from the firefly luciferase expressed by the DCTF partner proteins to a specific subcellular location,15 to system. As shown in Figure 3, transcriptional activity was increase the activity of multi-enzyme systems by virtue readily detectable with the SV40 promoter construct in of a physical link between the enzymes16,17 or to improve both tumor types demonstrating that the transfer of the function of heterologous proteins in a gene thera- naked DNA in these xenografts was successful. More- peutic setting by enabling oligomerization.18 Taken over, no significant activities were measured with the together, our data suggest that the dimerization system control plasmids lacking a functional interaction domain described in the present study, and its particular appli- (pGALmFos-VP⌬Jun). Most importantly, the Jun/Fos cation to the DCTF system, will provide invaluable tools DCTF system was 48-fold more active in the MeWo xeno- for gene therapy and protein engineering. grafts compared with the H322 tumors. These results clearly demonstrate that the Jun/Fos DCTF system is functional in an in vivo setting of experimental Acknowledgements melanoma. We are grateful to I Hart for the MeWo cell line, to N In summary, our observations show that LZs can be Favrot for the H322 cell line and to M Krause for oligonu- engineered in a way that they enable a specific and strong cleotide synthesis. We thank Simone Schmitt for per- interaction of heterologous proteins in an intracellular forming the in vitro translation association assay. This environment, and that such chimeric proteins also func- work was supported by the Dr Mildred Scheel Stiftung. tion in vivo. This is exemplified in the present study by the application of this technology to the DCTF system which suffered from a very low transcriptional activity. References By incorporating the mJun/mFos dimerization system 1Je´roˆme V, Mu¨ ller R. Tissue-specific, cell cycle-regulated chimeric and expressing the resulting subunits from the same vec- transcription factors for the targeting of gene expression to tumor cells. Hum Gene Ther 1998; 9: 2653–2659. 2 Busch SJ, Sassone-Corsi P. Dimers, leucine zippers and DNA- binding domains. Trends Genet 1990; 6: 36–40. 3 Neuberg M, Adamkiewicz J, Hunter JB, Mu¨ller R. A Fos protein containing the Jun leucine zipper forms a homodimer which binds to the AP1 binding site. Nature 1989; 341: 243–245. 4 Kouzarides T, Ziff E. Leucine zippers of fos, jun and dic- tate dimerization specificity and thereby control DNA binding. Nature 1989; 340: 568–571. 5 Schmidt-Do¨rr T et al. Construction, purification and characteriz- ation of a hybrid protein comprising the DNA binding domain of the LexA repressor and the Jun leucine zipper: a circular dichroism and mutagenesis study. Biochemistry 1991; 30: 9657– 9664. 6 Kouzarides T, Ziff E. The role of the leucine zipper in the fos– jun interaction. Nature 1988; 336: 646–651. 7 O’Shea EK, Rutkowski R, Stafford WFD, Kim PS. Preferential heterodimer formation by isolated leucine zippers from fos and jun. Science 1989; 245: 646–648. 8 Schuermann M et al. The leucine repeat motif in Fos protein Figure 3 Functionality of the DCTF system in vivo. pGALmFos-VP⌬Jun mediates complex formation with Jun/AP-1 and is required for or pGALmFos-VpmJun, 10× gal4RE-SVGL3 and pRL-SV40 were injected transformation. Cell 1989; 56: 507–516. intratumorally into subdermally implanted MeWo and H322 tumors in 9 Oas TG et al. Secondary structure of a leucine zipper determined nude mice. As a positive control pGL3-promoter (SV40pGL3; encoding by nuclear magnetic resonance spectroscopy. Biochemistry 1990; firefly luciferase) and pRL-SV40 (encoding Renilla luciferase) for stan- 29: 2891–2894. dardization were injected. After 24 h, luciferase activities were measured 10 Schuermann M, Hunter JB, Hennig G, Mu¨ller R. Non-leucine in tumor extracts. The results are presented as firefly luciferase activity residues in the leucine repeats of Fos and Jun contribute to the (FF) relative to Renilla luciferase activity (RL) per mg protein. Values stability and determine the specifity of dimerization. Nucleic ± represent the mean of five different experiments, standard deviation. Acids Res 1991; 19: 739–746. × 5 Nude mice were injected intradermally with 5 10 MeWo cells as pre- 11 O’Shea EK, Rutkowski R, Kim PS. Mechanism of specificity in 23 × 7 24 viously described or subcutaneously with 2 10 H322 cells. Three to the Fos-Jun oncoprotein heterodimer. Cell 1992; 68: 699–708. 4 weeks later tumors had reached a size of 4–5 mm and were injected with 12 Glover JN, Harrison SC. Crystal structure of the heterodimeric plasmid DNA. Thirty micrograms of pGL3promoter or 15 ␮g pGALmFos- bZIP transcription factor c-Fos-c-Jun bound to DNA. Nature VP⌬Jun or pGALmFos-VPmJun plus 15 ␮g10× gal4RE-SVpGL3 were 373 diluted in 50 ␮l glucose–0.01% Triton ×100 and injected intratumorally 1995; : 257–261. with insulin syringes (30 gauge, 0.5-inch needles; Becton Dickinson, Hei- 13 Triezenberg SJ, Kingsbury RC, McKnight SL. Functional dissec- delberg, Germany). For standardization, 6 ng of pRL-SV40 were co- tion of VP16, the trans- of herpes simplex virus injected. The mice were killed 24 h later, the tumors were lysed in 500 ␮l expression. Genes Dev 1988; 2: 718–729. Passive Lysis Buffer (Promega) and the Firefly and Renilla luciferase 14 Chasman DI, Kornberg RD. GAL4 protein: purification, asssoci- activities were measured with the dual-luciferase reporter assay system as ation with GAL80 protein, and conserved domain structure. Mol recommended by the manufacturer (Promega, Munich, Germany). The Cell Biol 1990; 10: 2916–2923. protein concentration was measured using the detergent-compatible Bio- 15 Dang CV et al. Intracellular leucine zipper interactions suggest Rad protein assay (Bio-Rad, Munich, Germany). c- hetero-oligomerization. Mol Cell Biol 1991; 11: 954–962.

Gene Therapy In vivo transcriptional targeting VJe´roˆme and R Mu¨ller 729 16 Davidson JN et al. The evolutionary history of the first three 21 Bean MA et al. Cell-mediated cytotoxicity for bladder carcinoma: enzymes in pyrimidine biosynthesis. BioEssays 1993; 15: 157–164. evaluation of a workshop. Cancer Res 1975; 35: 2902–2913. 17 Gontero B et al. Structural and functional properties of a multi- 22 Ausubel I, Frederick M. Current Protocols in Molecular Biology enzyme complex from spinach chloroplasts. 2. Modulation of John Wiley & Sons: New York, 1991. the kinetic properties of enzymes in the aggregated state. Eur J 23 Cornil I, Man S, Fernandez B, Kerbel RS. Enhanced tumorigenic- Biochem 1993; 217: 1075–1082. ity, melanogenesis, and metastases of a human malignant mela- 18 Walczak H et al. Tumoricidal activity of tumor necrosis factor- noma after subdermal implantation in nude mice. J Natl Cancer related -inducing in vivo. Nature Med 1999; 5: Inst 1989; 81: 938–944. 157–163. 24 Coll JL et al. Antitumor activity of bax and naked gene trans- 19 Kalderon D, Roberts BL, Richardson WD, Smith AE. A short fer in lung cancer: in vitro and in vivo analysis. Hum Gene Ther amino acid sequence able to specify nuclear location. Cell 1984; 1998; 9: 2063–2074. 39: 499–509. 20 Ransone LJ, Verma IM. Nuclear proto- fos and jun. Annu Rev Cell Biol 1990; 6: 539–557.

Gene Therapy