Combination Gene Transfer to Potentiate Tumor Regression

T Ohno1,4, Z Yang1,4, X Ling2, M Jaffe1,4, EG Nabel2, D Normolle4 and GJ Nabel1,4 1Howard Hughes Medical Institute, University of Michigan Medical Center, Departments of 2Internal Medicine and 3Biological Chemistry, Ann Arbor, MI; and 4Comprehensive Cancer Center, University of Michigan Medical Center, Ann Arbor, MI, USA

Recent efforts to treat malignancy using gene transfer have shown, a significant reduction of tumor size was observed met with varying degrees of success. In this paper, we with each vector. Combination treatment, in which both report the results of studies using two recombinant adeno- vectors were administered, resulted in a trend toward a viral vectors to examine the efficacy of combination gene reduced tumor growth greater than with either vector alone. transfer to cause tumor regression in vivo. One of these In order to characterize the mechanism of tumor vectors encodes the murine MHC class I gene, H-2K b regression, cytolytic T lymphocyte (CTL) assays against (ADV-Kb), which induces an immune response that stimu- the allogeneic molecule, H-2K b, were performed. Mice lates tumor regression. The second vector encodes the treated with ADV-K b showed specific CTL activity against human p21 cyclin dependent kinase inhibitor (ADV-p21). the H-2Kb molecule, demonstrating that the immune This gene product arrests cell cycle progression and pre- response against the H-2Kb gene product involved in vents proliferation of tumor cells. Both vectors were tested tumor regression was potentiated by expression of the p21 in a murine model in vivo for antitumor effect. As previously gene which affects cell cycle progression.

Keywords: gene therapy; cancer; p21; MHC

Introduction through the administration of either a retroviral vector or DNA–liposome complexes in vivo. This strategy is It has been suggested that most malignancies have designed to induce an immune response to syngeneic evolved mechanisms to evade immune detection, but the tumor cells, initially directed to the allogeneic MHC class ability to induce an immune response against tumors has I gene (HLA-B7 for humans or H-2Ks for mice), conse- been demonstrated after immunologic activation, and the quently generating tumor-specific immune responses.20 stimulation of specific antitumor immune recognition Alternative genetic targets may also be candidates to offers the potential to provide therapy for malignancy. A limit tumor growth after gene transfer in vivo. Among variety of approaches for inducing immune recognition these approaches, genes which affect cell cycle pro- of tumors or increasing the strength of the immune gression, specific cyclin/cyclin-dependent kinase inhibi- response have been studied. Among the methods tors (CKIs), have shown antitumor effects in animal mod- developed to elicit immune rejection are the stimulation els.21 Expression of the p21 CKI using gene transfer of major histocompatibility-restricted lysis through a techniques suppressed the growth of malignant cells in 1 cytotoxic T cell (CTL) response, tumor infiltrating lym- vitro and in vivo.21 In the previous study, the arrest of cell 2–4 5 phocytes, immunization with tumor cells, adminis- cycle progression was due to growth arrest associated 6 tration of cytokines or tumor-associated antigen sensi- with terminal differentiation. In this report, we have 6–8 9,10 tized lymphoid cells, natural killer cells, and tested the efficacy of gene transfer of the murine MHC macrophages. Others have attempted to introduce new gene, H-2Kb, in combination with the gene for p21, using antigens into tumor cells by treatment with drugs or adenoviral vectors in vivo. We sought to determine 11–13 virus infection. Furthermore, the expression of cyto- whether combination gene transfer with an immunomod- kine genes by genetic modification of tumor cells has ulatory and antiproliferative gene could potentiate the been shown to induce an antitumor response, mediated antitumor immune response. by lymphocytes or other inflammatory cells.14–18 We have previously developed a gene transfer protocol which utilizes the human MHC class I molecule, HLA- Results B7, for human trials,19 or a murine gene product, H-2Ks, in mouse studies,20 for the modification of tumor cells Construction of adenoviral vectors Recombinant adenoviral vectors which encode the H-2Kb gene (ADV-Kb) and the human p21 gene (ADV-p21) were Correspondence: GJ Nabel, Howard Hughes Medical Institute, University constructed using homologous recombination techniques of Michigan Medical Center, Departments of Internal Medicine and Bio- logical Chemistry, 1150 W. Medical Center Drive, 4520 MSRB I, Ann (Figure 1a and Refs 21 and 22). These vectors were exam- Arbor, MI 48109-0650, USA ined for their ability to express the allogeneic MHC class Received 27 August 1996; accepted 23 December 1996 I molecule, H2Kb (ADV-Kb) or to inhibit the growth of Combination immune and antiproliferative gene transfer T Ohno et al 362

Figure 1 Construction of recombinant adenoviruses. (a) Strategy for the construction of adenoviral vectors. The recombinant adenoviruses were constructed by homologous recombination between sub360 genomic DNA, an Ad5 derivative with a deletion in the E3 region, and H-2Kb expression plasmid, pAd-RSV-Kb or p21 expression plasmid described previously.22 This Figure shows the construct of ADV-Kb. These recombinant adenoviral vectors have the deleted sequences in the E1A and E1B region, impairing the ability of these viruses to replicate and transform nonpermissive cells. (b) Flow cytometric analysis of the expression of H-2Kb molecules on Renca cells infected with ADV-Kb. Cells were infected with either ADV-Kb or control vector, ADV-⌬E1 at an MOI of 200. Twenty-four hours after infection, cells were stained with a murine anti-H-2Kb antibody (solid line) or an isotype control antibody (dashed line) as a ﬁrst antibody and FITC-conjugated anti-mouse IgG antibody as a second antibody. Cells infected with ADV-Kb (left panel), ADV-⌬E1 (middle panel) and Renca-Kb which were stable transfectants of Renca cells with the H-2Kb gene (right panel) were analyzed by ﬂow cytometry (FACScan).

Renca cells (p21).15 When Renca cells were infected with Treatment of established tumors with adenoviral vectors ADV-Kb at a multiplicity of infection (MOI) of 200, cells in vivo were able to express the H-2Kb protein on the cell surface, To examine the efficacy of ADV-Kb and ADV-p21 for the in contrast to cells transduced with the control adeno- treatment of tumors in vivo, BALB/c mice were pre- virus, ADV-⌬E1 (Figure 1b). Similar to previous studies, immunized twice with 1 × 106 irradiated splenic lympho- Renca cells transduced with ADV-p21 reduced prolifer- cytes of C57BL6 mice, then injected subcutaneously on ation, in contrast to the control vector ADV-⌬E1 (Figure the right flank with Renca cells (106) on day 0. Six days 2). No expression of H-2Kb was observed in these cells after inoculation, tumors of approximately 2 × 2 mm were transduced with ADV-p21 (data not shown). Thus, each established in each mouse. Adenoviral infection was vector was able specifically to modify the phenotype of initiated by direct injection into tumor tissues once each Renca cells with respect to cell surface antigen expression day for 5 days during the first week and twice at 2 day (ADV-Kb) or cell cycle progression (ADV-p21) in vitro. intervals during the second week (a total of seven Combination immune and antiproliferative gene transfer T Ohno et al 363 The growth rates in groups treated with ADV-Kb, ADV-p21 or both were significantly less than the growth rates in the PBS-treated group (Figure 3, P Ͻ 0.01). The groups treated with ADV-⌬E1/ADV-Kb or ADV- ⌬E1/ADV-p21 showed a lower rate of tumor growth when compared to the group treated with ADV-⌬E1 alone (Figure 3, P = 0.23 and P = 0.07, respectively); however, only the mice treated with ADV-p21/ADV-Kb showed a trend toward lower growth rates than those treated with ADV-⌬E1 alone (Figure 3, P = 0.02). Tumor growth was almost completely arrested by day 15 in seven of nine mice (tumors were less than 3 × 3 mm) in the group treated with the combination of ADV-Kb and ADV-p21. Figure 2 Effect of ADV-p21 on proliferation of Renca cells in vitro. 104 We also examined the rate of generation of tumor-free ⌬ Renca cells were infected with ADV-p21 and ADV- E1 at an MOI of mice in each group (Table 1). The group treated with PBS 200. Cell numbers were counted using trypan blue exclusion on days 2, had no tumor-free mice. The group treated with ADV- 4, 6 and 7 to examine the effect of p21 on cell growth. ⌬E1 or ADV-Kb alone showed one tumor-free mouse out of eight treated mice evaluated (12.5%). In the group injections). Mice received 2 × 108 p.f.u. (50 ␮l per mouse) treated with ADV-p21 alone, three out of seven (42.9%) of adenovirus in each injection. Treatment with either p21 mice were tumor-free, and in the combination group of b or Kb was conducted by injecting 1 × 108 p.f.u. each of ADV-K and p21, five out of eight (62.5%) mice were ADV-⌬E1 mixed with an equal amount of either ADV- tumor-free on day 44. Thus, the frequency of tumor Kb or ADV-p21 to adjust the total viral dose among the regression correlated with the ability to inhibit local groups. In the combination experiment, 1 × 108 p.f.u. each tumor growth in this model. b of ADV-K and ADV-p21 were mixed together and b administered. PBS was injected as an additional control. Generation of CTL response to H-2K Tumor size in individual mice was measured following Induction of CTL activity against the allogeneic protein, b inoculation of the tumors. In all mice treated with PBS, H-2K , was measured in each treatment group. Splenic tumors grew aggressively by day 15. Although the mean lymphocytes prepared from each mouse group were b tumor growth rate in mice treated with the control sensitized in vitro with Renca-K cells, a stable transfect- adenoviral vector, ADV-⌬E1, demonstrated slower tumor ant of Renca cells containing the expression vector enco- b growth than the PBS-treated group, the difference ding the murine H-2K gene and with 20 U/ml human between the two groups was not statistically significant recombinant IL-2 for 5 days. CTL activity against Renca- b (Figure 3, P = 0.16). Exposure of Renca cells to adeno- K or parental Renca cells respectively, was measured virus, regardless of the insert of the vector, caused a (Figure 4). Splenic lymphocytes prepared from the mice treated with PBS (data not shown), ADV-⌬E1 or ADV- slight cytopathic effect in vitro which could explain the b slower tumor growth in mice treated with ADV-⌬E1 p21 alone showed no CTL activity against Renca-K cells, (data not shown). demonstrating the failure of the induction of CTL activity against the H-2Kb molecule in these groups (Figure 4a and b). Splenic cells obtained from the mice treated with ADV-Kb or with the combination of ADV-Kb and ADV- p21 showed a significant CTL response against Renca-Kb cells, depending on the effector-to-target ratio (E:T), but not against parental Renca cells (Figure 4c and d). No response against parental Renca cells was observed in any of the treatment groups, suggesting that the tumor rejection in the groups treated with ADV-Kb were par- tially or completely dependent upon the specific immune response against the allogeneic molecule, H-2Kb.

Discussion In this study, we have examined the efficacy of combination gene therapy for the treatment of tumors established in mice in vivo. We used two genes which Figure 3 Tumor growth rates after various viral treatments in vivo. Mice exert effects through distinct mechanisms to prevent were treated with PBS, ADV-⌬E1, ADV-Kb, ADV-p21, or ADV-Kb and tumor growth. One gene encodes an allogeneic MHC ADV-p21, as described in the text. Each animal’s tumor growth was mod- class I molecule, H-2Kb, intended to induce the host eled by estimating parameters for initial tumor size and growth rate in immune response against tumors which were modified an exponential growth model. Each symbol represents the estimated with this product, facilitating rejection of tumor tissues.20 growth rate parameter for a single animal. The means are depicted by the The other gene is a cyclin-dependent kinase inhibitor, line. The groups treated with ADV-Kb, ADV-p21, and ADV-Kb and ADV-p21 were all significantly different from PBS-treated animals (P Ͻ p21, which arrests cell cycle progression, preventing the 0.01), and the animals treated with ADV-Kb and ADV-p21 were signifi- growth of tumors, and allowing cells to commit to ter- cantly different from animals treated with ADV-⌬E1 (P = 0.02). minal differentiation.21 Both genes were incorporated Combination immune and antiproliferative gene transfer T Ohno et al 364 Table 1 Complete regression of tumors in response to single or combination gene transfer

Treatment PBS ADV-⌬E1 ADV-Kb ADV-p21 ADV-Kb + ADV-p21

Percentage tumor-free 0 12.5 12.5 42.95 62.5

The percentage of mice which remained tumor-free 44 days after inoculation with 106 Renca cells and treatment with the indicated vectors as described in Materials and methods is shown. Each group included eight mice, except for ADV-p21 (n = 7).

Figure 4 Measurement of CTL activity against H-2Kb molecules of tumor-bearing mice with various treatments. Splenic lymphocytes prepared from each group (3 × 107 cells) were stimulated with 6 × 105 irradiated C57BL6 splenic lymphocytes and 20 U/ml recombinant human IL-2 for 5 days in vitro. Stimulated lymphocytes derived from mice whose tumors were injected with ADV-⌬E1 (a), ADV-p21(b), ADV-Kb (c), or ADV-Kb and ADV- p21 (d) were used for CTL assays in each E:T ratio. Cr51 labeled Renca cells () and Renca-Kb cells (í) were used as target cells. Speciﬁc lysis was calculated using the following formula: speciﬁc lysis (%) = (experimental release (c.p.m.) − spontaneous release (c.p.m.))/(maximum release (c.p.m.) − spontaneous release (c.p.m.)).

into adenoviral vectors and transferred alone, or in com- mice were found in the groups infused with ADV-p21 or bination, directly into tumor tissues. As reported pre- with the combination of ADV-Kb and ADV-p21 adeno- viously, the introduction of the p21 gene alone into tumor virus. The proportion of mice available for evaluation in tissues showed a reduction in tumor size in vivo.21 Similar each group which became tumor-free was 42.9% and to results using other foreign class I MHC genes with 62.5%, respectively after 44 days. The existence of tumor- different vectors, H-2Kb gene transfer alone also inhibited free mice in the ADV-⌬E1 treated group (12.5%) suggests tumor growth. We demonstrated here that the combi- a background effect of the adenoviral vector for the nation therapy using recombinant adenoviruses encoding rejection of tumor. In fact, we observed some toxic effects the H-2Kb and p21 gene respectively can reduce the with the infection of adenovirus on Renca cells in vivo growth of tumor cells and eliminate tumor tissues more (Figure 2) and in vitro (data not shown). However, treat- effectively than the introduction of either gene alone ment with ADV-Kb and ADV-p21 together resulted in a (Figure 3). Importantly, increased numbers of tumor-free trend towards an increased rate of rejection of tumor Combination immune and antiproliferative gene transfer T Ohno et al 365 cells, and the combination therapy provided the highest (DMEM; Gibco, Grand Island, NY, USA), supplemented frequency for generating tumor-free mice. with 10% fetal calf serum (FCS; Gibco). A stable transfect- Cytotoxic T cell assays provided evidence that the ant of the Renca cells which expresses the H-2Kb mol- induction of an immune response against H-2Kb occurred ecule was established by co-transfection with RSV-neo in ADV-Kb and ADV-Kb plus ADV-p21 treated mice in and RSV-Kb expression vectors. The expression of the H- vivo, supporting the observation that tumor rejection in 2Kb gene on RSV-Kb was driven by the Rous sarcoma the groups treated with ADV-Kb is mainly dependent on virus enhancer and the promoter, and the transcription the host defense mechanism which recognizes and was terminated by the SV40 large T antigen polyadenyl- removes tumors modified with an allogeneic MHC class ation signal sequence. I gene. When the parental Renca cell was used as a target cell line in this study, no CTL response against parental Preparation of recombinant adenovirus cells was detected. This finding suggests that a specific The recombinant adenoviral vectors, ADV-Kb and ADV- anti-H-2Kb response was induced in cytotoxic T lympho- p21, were constructed by homologous recombination cytes. In our previous studies, the generation of an anti- between sub360 genomic DNA, an Ad5 derivative with parental response using the murine MHC class I mol- a deletion in the E3 region, and a mouse MHC class I ecule, H2-Ks was observed in vivo.20 Possibly, the use of molecule expression plasmid, pAd-RSV-Kb, or a human a different gene transfer vector (adenovirus) or a different p21 expression plasmid, pAd-p21, respectively.21,22 The MHC haplotype was responsible for this effect. Alterna- sequence in the E1A and E1B region is deleted in these tively, the different tumor cell line, Renca, in the context recombinant adenoviral vectors, impairing their ability to of a different mouse genetic background, may not have replicate and transform nonpermitting cells. The pAd- been sufficiently immunogenic to allow induction of the RSV-Kb plasmid was prepared by introducing the NdeI– antitumor immune response in vivo. The CTL analysis BamHI fragment of an H-2Kb expression plasmid, pRSV- demonstrated no activity against parental Renca cells or Kb into the BglII site of pAd-BglII.22 The pAd-p21 plasmid Renca-Kb cells in the group treated with ADV-p21 alone. was constructed using the NruI–DraIII fragment of a p21 This result is consistent with our previous findings that expression plasmid, pRc/CMV/p21 (25), as well. The expression of the p21 gene product does not lead to the recombinant adenoviruses were purified as described recognition of tumor-specific antigens on Renca cells by previously.22 the host, but promotes the arrest of cell cycle progression and the induction of terminal differentiation of tumors Flow cytometry analysis by the inhibition of cyclin-dependent kinase activity.21 Renca cells were infected with either ADV-⌬E1 or ADV- Recently, combination gene transfer has been exam- Kb at an MOI of 200, as described previously.22 One day ined using other genes for its ability to enhance tumor after infection, cells were trypsinized and suspended in regression.23,24 In one study, a combination of a prodrug DMEM supplemented with 10% FCS. Cells were incu- ° converting enzyme, herpes simplex virus thymidine kin- bated at 37 C with 5% CO2 for 1 h to allow the re- ase, and a cytokine gene, IL-223,24 was examined. In expression of the H-2Kb molecule on the cell surface. another case, a combination of cytokine genes, tumor Cells were then incubated with murine anti-H-Kb anti- necrosis factor-alpha (TNF␣), and interferon-gamma body (Pharmingen, San Diego, CA, USA) as a first anti- (IFN␥), or IL-2 and IFN␥ has been explored.24 In both body at 4°C for 30 min followed by incubation with FITC- cases, combination therapies conferred a longer survival conjugated anti-mouse IgG (Sigma, St Louis, MO, USA) rate and a lower rate of tumor growth than either treat- using the same conditions. Cell surface expression of the ment alone. In this report, we used the combination of H-2Kb molecule was analyzed by flow cytometry H-2Kb and p21 genes. Although the combination therapy (FACScan; Becton Dickinson, San Jose, CA, USA). of both genes can augment the rate of tumor rejection when compared with either treatment alone, our findings In vivo treatment of tumors in in vivo studies suggest that the set of H-2Kb and p21 Five-week-old BALB/c mice (Charles River Lab, Wil- genes renders an additive effect rather than a synergistic mington, MA, USA) were pre-immunized twice by intra- effect for tumor rejection. One explanation for this obser- peritoneal (i.p.) injection with irradiated (25 Gy) spleen vation is that both genes were subcloned into inde- cells from C57BL6 mice. Following the second pre- pendent adenoviruses. Although expression of both gene immunization, mice were inoculated subcutaneously products within the same vector may enhance this effect with 1 × 106 Renca cells on the right flank. After the further, the present approach facilitates the analysis of tumor was established, recombinant adenoviral vectors different genes for combination treatment. Such protocols (2 × 108 p.f.u. in 50 ␮l of PBS) were injected directly into for combination gene transfer may advance efforts to tumor tissues.20,21 Infusions of the adenoviral vectors develop improved treatments for malignancy. In were performed once per day for 5 days during the first addition, the ability to use such gene transfer approaches week and twice at 2-day intervals during the second together with conventional pharmacologic or immuno- week. therapeutic approaches may potentiate the effects of each treatment modality and increase the possibility of erad- Cytolytic T lymphocyte (CTL) assay icating minimal residual disease in malignancy. Splenic lymphocytes were prepared and two mice containing an average-sized tumor were killed from each Materials and methods group. Cells were sensitized in vitro with Renca-Kb cells, a stable transfectant of Renca cells containing the Cell culture expression vector encoding the murine H-2Kb gene. Sens- A mouse renal carcinoma cell line, Renca (H-2Kd,15), was itization was continued for 5 days with 20 U/ml human maintained in Dulbecco’s modified Eagle’s medium recombinant IL-2. CTL assays were performed using Combination immune and antiproliferative gene transfer T Ohno et al 366 radiolabeled Renca cells or Renca-Kb cells as target cells 6 Lafreniere R, Rosenberg SA. Successful immunotherapy of at an appropriate E:T ratio as described previously.20 murine experimental hepatic metastases with lymphokine- activated killer cells and recombinant interleukin 2. Cancer Res Statistical analysis 1985; 45: 3735–3741. 7 Rosenberg SA et al. Use of tumor-infiltrating lymphocytes and All statistical modeling was done with PROC MIXED interleukin-2 in the immunotherapy of patients with metastatic (SAS Institute, Cary, NC, USA) on a Sun Microsystems melanoma. New Engl J Med 1988; 319: 1676–1680. SPARCStation 4 (Mountain View, CA, USA). The tumor 8 Yoshizawa H, Sakai K, Chang AE, Shu SY. Activation by anti- sizes were logarithmically transformed (tumor sizes of 0 CD3 of tumor-draining lymph node cells for specific adoptive were changed to 1 before transformation), and the trans- immunotherapy. Cell Immunol 1991; 134: 473–479. formed sizes were fit to a mixed effects linear model to 9 Oldham RK. Natural killer cells: artifact to reality: an odyssey produce an exponential growth model. In this model, an in biology. Cancer Metast Rev 1983; 2: 323–336. intercept and slope were fit to each animal’s growth 10 Herberman RB. Multiple functions of natural killer cells, includ- curve, where the intercepts (logarithm of initial size) ing immunoregulation as well as resistance to tumor growth. Concepts Immunopathol 1985; 1: 96–132. were assumed to be random effects with a mean common 11 Itaya T et al. Xenogenization of a mouse lung carcinoma (3LL) by to all treatment groups, and the animals’ growth rates transfection with an allogeneic class I major histocompatibility were assumed to be random effects with means that complex gene (H-2Ld). Cancer Res 1987; 47: 3136–3140. varied between treatment groups. The modeled values 12 Pucetti P, Romani L, Fioretti MC. Chemical xenogenization of were graphically compared to the tumor profiles to experimental tumors. Cancer Metast Rev 1987; 6: 93–111. ensure that the model was appropriate. Differences 13 Sugiura C et al. Xenogenization of tumor cells by transfection between the mean growth rates of the treatment groups with plasmid containing env gene of friend leukemia virus. Jpn and positive and negative control groups were evaluated J Cancer Res 1988; 79: 1259–1263. by means of Dunnett’s test. 14 Tepper RI, Pattengale PK, Leder P. Murine interleukin-4 dis- plays potent antitumor activity in vivo. Cell 1989; 57: 503–512. 15 Fearon ER et al. Interleukin-2 production by tumor cells Acknowledgements bypasses T helper function in the generation of an antitumor response. Cell 1990; 60: 397–403. We gratefully acknowledge Donna Gschwend for manu- 16 Gansbacher B et al. Interleukin 2 gene transfer into tumor cells script preparation, Judith Stein for help in editing, and abrogates tumorigenicity and induces protective immunity. Nancy Barrett for graphic arts. This work was supported J Exp Med 1990; 172: 1217–1224. in part by grants from the Public Health Service (P01 17 Gansbacher B et al. Retroviral vector-mediated gamma- CA59327, GJN). interferon gene transfer into tumor cells generates potent and long lasting antitumor immunity. Cancer Res 1990; 50: 7820–7825. 18 Esumi N, Hunt B, Itaya T, Frost P. Reduced tumorigenicity of References murine tumor cells secreting gamma-interferon is due to non- specific host responses and is unrelated to class I major histo- 1 Kern O, Klarnet J, Jensen M, Greenberg PJ. Requirement for rec- compatibility complex expression. Cancer Res 1991; 51: 1185– ognition of class II MHC molecules and processed antigen for 1189. optimal generation of syngeneic tumor specific class I MHC 19 Nabel GJ et al. Direct gene transfer with DNA–liposome com- restricted cytotoxic T lymphocytes. J Immunol 1986; 136: 4303– plexes in melanoma: expression, biologic activity, and lack of 4310. toxicity in humans. Proc Natl Acad Sci USA 1993; 90: 11307– 2 Rosenberg SA, Spiess P, Lafreniere R. A new approach to the 11311. adoptive immunotherapy of cancer with tumor-infiltrating lym- 20 Plautz GE et al. Immunotherapy of malignancy by in vivo gene phocytes. Science 1986; 223: 1318–1321. transfer into tumors. Proc Natl Acad Sci USA 1993; 90: 4645–4649. 3 Muul LM, Spiess PJ, Director EP, Rosenberg SA. Identification 21 Yang Z et al. The p21 cyclin-dependent kinase inhibitor sup- of specific cytolytic immune responses against autologous presses tumorigenicity in vivo. Nature Med 1995; 1: 1052–1056. tumor in humans bearing malignant melanoma. J Immunol 1987; 22 Ohno T et al. Gene therapy for vascular smooth muscle cell pro- 138: 989–995. liferation following arterial injury. Science 1994; 265: 781–784. 4 Topalian SL, Solomon D, Rosenberg SA. Tumor-specific cytoly- 23 Chen SH et al. Combination gene therapy for liver metastasis of sis by lymphocytes infiltrating human melanomas. J Immunol colon carcinoma in vivo. Proc Natl Acad Sci USA 1995; 92: 1989; 142: 3714–3725. 2577–2581. 5 Shu S, Rosenberg SA. Adoptive immunotherapy of a newly 24 Sun WH et al. In vivo cytokine gene transfer by gene gun reduces induced sarcoma: immunologic characteristics of effector cells. tumor growth in mice. Proc Natl Acad Sci USA 1995; 92: 2889– J Immunol 1985; 135: 2895–2903. 2893.