D2001 Nature Publishing Group 0929-1903/01/$17.00/+0 www.nature.com/cgt

Adenoviral delivery of TIMP1 or TIMP2 can modify the invasive behavior of pancreatic cancer and can have a significant antitumor effect in vivo Anne S. Rigg, and Nicholas R. Lemoine Imperial Cancer Research Fund Molecular Oncology Unit, Imperial College School of Medicine, London W12 0NN, UK.

Pancreatic carcinomas overexpress several matrix metalloproteinases (MMPs), in particular MMP2 and MMP9. These enzymes are involved in the degradation of the to aid tumor invasion. The aim of this study was to investigate the effect of TIMP therapy on pancreatic cancer. Human TIMP1 or TIMP2 has been introduced in pancreatic tumor cells under the control of a constitutive promoter using adenoviral vectors, and the effect on tumor invasion observed. It has been demonstrated in vitro that the TIMP-expressing pancreatic tumor cells were significantly less invasive than those cells transfected with a control vector. In vivo, adenoviral delivery of TIMP1 or TIMP2 to nude mice harboring intraperitoneal human pancreatic cancers resulted in prolonged survival compared with control mice if the gene therapy was given early (P<.009 and P<.0293, respectively). The in vivo experiments demonstrated evidence of gene transfer by adenoviral vectors to tumor cells and murine mesenteric cells. There was no evidence of transgene expression in distant organs. These experiments have proved the hypothesis that TIMP overexpression in pancreatic cancer cells can modify the invasive phenotype. Also, TIMP gene transfer to human tumor cells is possible both in vitro and in vivo. Cancer Gene Therapy (2001) 8, 869–878

Key words: Tissue inhibitor of metalloproteinases; pancreatic cancer; adenovirus; gene therapy.

denocarcinoma of the exocrine pancreas is an invasive Pancreatic tumor samples probed for MMP2 mRNA Aneoplasm that is usually lethal. Due to the deep, intra- demonstrate MMP2 in both tumor epithelial cells and abdominal location of the pancreas and a phenotype stromal cells. TIMP1 and TIMP2 are expressed in the characterised by early invasion, the cancer tends to present majority of pancreatic tumor epithelia, but not stromal late at an advanced inoperable stage. Despite surgery, elements.12 Overall, there is a relative paucity of TIMPs radiotherapy, and chemotherapy, the 5-year survival for a favoring the activities of the MMPs. It is the net balance patient with adenocarcinoma of the pancreas is virtually between the proteases and their inhibitors that is critical to zero.1 Given the typically late presentation of the disease, it their effect on the environment. is appropriate to concentrate new treatment strategies on Given that disturbances of the levels of the MMPs and limiting pancreatic carcinoma cell invasion. TIMPs are implicated in tumor growth and metastasis, it is It is known that the enzymes logical to attempt to alter the balance by reducing the level of (MMPs) are involved in proteolysis of the extracellular MMP and/or increasing the level of TIMP. The majority of matrix, aiding tumor cell invasion and the establishment of attention have been focused on the inhibition of the MMP metastatic deposits.2,3 In addition, MMP expression is enzyme by synthetic inhibitors or TIMP gene therapy. associated with the neovascularization required for a tumor Several groups have opted to use the TIMPs themselves as blood supply to develop.4 A group of four molecules has anticancer agents. The first reports of the use of TIMPs as been identified, which can specifically inhibit both anticancer agents centered on recombinant TIMP1 and proenzyme MMPs and active MMPs.5–9 These tissue TIMP2. Both have been shown to reduce the inhibitors of metalloproteinases (TIMPs) occur naturally invasion of malignant cells through membranes in vitro, and within the extracellular matrix. Pancreatic carcinomas limit the development of tumors in vivo.13 – 16 express several MMPs, in particular MMP2 and MMP9, A variety of vectors have been used to transfer TIMP and the level of MMP2 has been directly correlated with the in target cells including plasmids, retroviruses, and degree of disruption of the basement membrane.10,11 adenoviruses. Several groups have demonstrated that malignant cells genetically modified to overexpress TIMP1 or TIMP2 form fewer, smaller, and less invasive tumors than Received August 2, 2001. 17 – 21 Address correspondence and reprint requests to Nicholas R. Lemoine, control cells in vivo. There have been recent reports of MD, PhD, Imperial Cancer Research Fund Molecular Oncology Unit, TIMP3 being used for gene therapy and also inhibiting 22,23 Imperial College School of Medicine, Hammersmith Campus, Du Cane invasion in vitro when delivered by an adenoviral vector. Road, London W12 0NN, UK. E-mail address: [email protected] The advantage of TIMP gene therapy is definitive MMP

Cancer Gene Therapy, Vol 8, No 11, 2001: pp 869–878 869 870 RIGG AND LEMOINE: TIMP GENE THERAPY inhibition by the naturally occurring substance that can be EcoRI fragment corresponding to bases 23–653 of the produced in excess by cells at the desired location. Genetic published cDNA. The TIMP2 cDNA consisted of a 660-bp manipulation of TIMP levels locally may also avoid the HindIII/EcoRI fragment corresponding to bases 255–915 of toxicity associated with synthetic MMP inhibitors. the published cDNA (sequencing of the transgene revealed a Another area of interest has been that TIMPs may not be neutral change from C to T at position 36 compared to the inducing an antitumor effect by pure MMP inhibition. published sequence). TIMP1, TIMP2, and TIMP3 have been demonstrated to The Adenovirus serotype 2 vector system was obtained reduce angiogenesis in vitro and in vivo,24 – 28 and inhibit the from Genzyme (Framingham, USA). The shuttle vector release of latent growth factors in the extracellular matrix. consists of the 50 region of the adenovirus serotype 2 genome TIMP3 has been shown to be proapoptotic in , (bases 1–356), a CMV promoter followed by a small cervical cancer, and vascular smooth muscles.22,23 multiple cloning site for insertion of the desired transgene, The primary aim of this research was to examine the the SV-40 polyadenylation signal, and a further region of effect of constitutive expression of TIMP gene products on adenovirus 2 genome (4020-10685). human pancreatic cancer cells. Replication-defective adeno- The second component, Ad gal, consists of a serotype 2 viral vectors were used to achieve gene transfer. Four adenovirus with an E1 deletion extending through the pancreatic carcinoma cell lines were genetically modified to IX sequence, with relocation of protein IX to the E4 overexpress the TIMP proteins and used to assess the effect locus. The E4 region has been modified to include only on invasive behavior in vitro. A second aspect of the project ORF6. The Escherichia coli -galactosidase gene (LacZ) has been the usage of a mouse model that closely resembles has been inserted in the empty E1 region under the control of advanced pancreatic cancer with peritoneal metastases to the E1 promoter. As there is a region of examine the effects of adenoviral delivery of TIMP1 or between the Ad gal and the Ad2 shuttle plasmid in the TIMP2 in vivo. Previous work has exclusively focused on presence of 293 cells, homologous recombination can occur the use of recombinant TIMP proteins or TIMP gene transfer as the E1 functions are provided in trans. for tumors established as subcutaneous xenografts or The TIMP cDNA were amplified by polymerase chain pulmonary metastases, but this is not representative of the reaction with primers to add SpeI and MluI restriction sites at human pattern of pancreatic cancer. Also, generally, the the 50 and 30 ends, respectively, to allow simple cloning in the tumor cells were modified ex vivo prior to injection in a Ad2 shuttle vector. murine host. In this study we have attempted to genetically TIMP1 primers 50 ACTAGTACTAGTGAATTCGAGCT- modify the behavior of human pancreatic cancers estab- CGGTACC 30 Upstream 50 ACGCGTACGCGTGAATTCT- lished in the peritoneal cavities of mice as disseminated CAGGCTATCTG 30 Downstream. metastases. TIMP2 primers 50 ACTAGTACTAGTAAGCTTATGGG- CGCCGCG 30 Upstream 50 ACGCGTACGCGTGAATTCT- TATGGGTCCTC 30 Downstream. METHODS The shuttle plasmids containing TIMP1 or TIMP2 were Cell lines and culture conditions linearized by restriction with BstBI and the Ad gal genome was digested with the restriction endonuclease PshA1 to The human pancreatic carcinoma cell lines used in this study remove the 50 end of the viral DNA including the LacZ gene. were obtained from the following sources: Panc1 obtained The resulting DNA from the two digests was then transfected from the American Type Culture Collection (Rockwell, in 293 cells for homologous recombination and packaging as MD); PSN-1 obtained from Dr. T. Sugimura (National described.29 Cancer Center Research Institute, Tokyo, Japan); and T3M4 After ultracentrifugation on a cesium chloride gradient obtained from Dr. T. Okabe (University of Tokyo, Tokyo, and desalting, the total number of adenoviral particles Japan). Suit2/MMP2 is derived from the Suit2 cell line present in the eluate was calculated by measuring the optical following transfection with the expression plasmid density at 260 nm.30 pGW1HG containing the human gene for MMP2 under The titer of the recombinant adenoviral vectors was control of the CMV promoter and was kindly provided by established by logarithmic limiting dilution on 293 cells. The Dr. S. Bramhall (Department of Surgery, University of lowest dilution at which CPE occurred was taken to be the Birmingham, Birmingham, UK). Human kidney embryonic titer of the virus. Adenoviral stocks were screened for cell 293 cells constitutively expressing Adenovirus serotype replication-competent adenovirus (RCA) by titration on 5 early 1 proteins were obtained from Microbix (Toronto, nonpermissive Panc1 cells. Canada). T3M4 cells were cultured in RPMI 1640 medium and all other cell lines in Dulbecco’s modified Eagle’s medium. For all cell lines, the medium was supplemented Adenoviral infection of pancreatic carcinoma cell lines with 10% heat-inactivated fetal calf serum (FCS), penicillin 5 1Â10 human pancreatic carcinoma cells were infected with (100 U/mL), and streptomycin (100 g/mL). Ad gal virus to give multiplicities of infection (MOI) of 10, 50, 100, 500, and 1000 in the presence of 2% FCS. The Generation of recombinant adenoviruses infected cells were incubated for 18 hours. The medium was The human TIMP1 and TIMP2 cDNA were kindly provided removed and replaced with fresh medium plus 10% FCS. in the plasmid vector pGW1HG by British Biotech (Oxford, Forty-eight hours after infection, the cells were washed UK). The TIMP1 cDNA consisted of a 630-bp EcoRI/ with phosphate-buffered saline (PBS), fixed, and stained

Cancer Gene Therapy, Vol 8, No 11, 2001 RIGG AND LEMOINE: TIMP GENE THERAPY 871 for -galactosidase expression [incubated in 1 mg/mL Assessment of apoptosis D X-gal (5-bromo-4-chloro-3-indoyl- - -galactopyrano- Analysis to determine if apoptosis was occurring in sidase), 10 mM sodium phosphate buffer (pH 7.3), genetically modified cells was performed using the 150 mM NaCl, 1 mM MgCl2, 3.3 mM K3Fe(CN)6,3.3mM Apoptosis Detection System, Fluorescein (Promega, South- K4Fe(CN)6]. ampton, UK). The human pancreatic cancer cell lines were transduced with adenoviral vectors as described above. Assay of TIMP1 and TIMP2 production After 48 hours, the cells were detached with trypsin and seeded on sterile polylysine-coated slides. The cells were 1 105 human pancreatic carcinoma cells were infected with  allowed to grow on the slides for 24 hours. The cells were Ad gal, AdTIMP1, or AdTIMP2 at an MOI of 100 for then fixed in 4% paraformaldehyde (in PBS, pH 7.4) for Panc1, T3M4, and Suit2 and an MOI of 500 for PSN-1. The 25 minutes at 48C. The slides were washed twice in PBS. cells were then incubated with the adenovirus for 18 hours. The cells were permeabilized in 0.2% Triton X-100 in PBS Culture medium was then changed and the conditioned for 5 minutes on ice and washed twice with PBS. Cells were medium collected at 24 hours. Cells were detached with covered with 100 mL of equilibration buffer (200 mM trypsin, resuspended, and counted. Supernatants were potassium cacodylate, pH 6.6, 25 mM Tris–HCl, pH 6.6, centrifuged to remove any cellular debris and transferred to sterile microfuge tubes. To assess the time course of TIMP1 or TIMP2 expression by Panc1 cells after infection with the adenoviral vectors, 1Â105 cells were seeded in each well of six-well plates. The cells were incubated overnight at 378C in E4+10% FCS. Cells were infected at an MOI of 100 with AdTIMP1 or AdTIMP2. The following day, cells were detached with trypsin and seeded on six-well plates at a density of 1Â105 cells/well. Supernatants were harvested every 12 hours for a total of 96 hours. TIMP1 and TIMP2 protein concentrations were deter- mined using the Biotrak2 TIMP1 ELISA System or the Biotrak2 TIMP2 ELISA System according to the manu- facturer’s instructions.

Assessment of invasion by cell lines 3Â106 human pancreatic carcinoma cells were infected with Ad gal, AdTIMP1, or AdTIMP2 at the appropriate MOI for the cell line. Twenty-four hours after infection, the cells were detached with trypsin, resuspended in medium, and counted. 1Â105 cells were seeded in the upper chambers of 12-well Transwell plates (Costar, Corning, NY). These consist of an upper chamber whose base is a porous membrane (8 m pore size) and has been coated with reconstituted basement membrane (Matrigel; Becton Dickinson Labware, Franklin Lakes, NJ, USA). One milliliter of medium plus 10% FCS was added to the lower chamber. The Transwell plates were then incubated at 378C for 72 hours. Cells that had invaded the membranes were then subjected to an MTT assay. Twenty microliters of 5 mg/mL MTT [3-(4,5-dimethylthiazole-2-yl)-2,5- diphenyltetrazolium bromide] was added to 1 mL of medium in the lower chamber and incubated at 378C for 2–4 hours. The upper chambers were then transferred to another 12-well plate containing 200 L of DMSO in each well. After incubation for 5 minutes at room temperature, the formazan from the cells on the underside of the membrane (i.e., the cells that have invaded the membrane) had dissolved in the DMSO. Each sample was transferred Figure 1. Transduction of 1Â105 Panc1 cells with AdTIMP1 or to a 96-well plate allowing absorbance to be measured at AdTIMP2 at an MOI of 100 was performed. Serum-free medium 560 nm. Absorbance could be plotted on standard curves was placed on the cells for a period of 12 hours and then harvested. for each cell line of optical density against cell number. This was repeated over a 96-hour period. The samples were Each experiment was performed four times with nine subjected to ELISA to estimate the amount of TIMP protein present replicates tested for each sample. at each time point.

Cancer Gene Therapy, Vol 8, No 11, 2001 872 RIGG AND LEMOINE: TIMP GENE THERAPY

0.2 mM DTT, 0.25 mg/mL BSA, 2.5 mM cobalt chloride) 10% glycerol. The adenoviral aliquots were stored at À708C and left at room temperature for 5 minutes. The equilibration and thawed on ice prior to administration. Mice received an buffer was then tapped off and replaced with incubation intraperitoneal injection of 1Â106 Panc1 cells on day 0. They buffer (90 L of equilibration buffer, 10 L of nucleotide then received an intraperitoneal injection of 1Â109 pfu mix, 2 L of TdT enzyme). The nucleotide mix consists of adenovirus on day 1, 8, or 15. The primary endpoint was 50 M fluorescein-12-dUTP, 100 M dATP, 10 mM Tris– survival. HCl, pH 7.6, and 1 mM EDTA. Plastic coverslips were All mice underwent a postmortem at the time of death. applied and the cells allowed to incubate at 378C for Any ascites present was aspirated and measured. Mouse 60 minutes inside a humidified chamber. The slides were organs (spleen, liver, kidney, heart, lungs, and brain) were protected from direct light. Positive controls consisted of weighed, as was any tumor that was present. All organs, cells treated with DNase I prior to incubation and negative tumor, and ascitic fluid were snap-frozen in liquid nitrogen controls consisted of cells incubated in buffer lacking TdT and stored at À708C. enzyme. The reaction was stopped by immersing the slides in 2Â SSC for 15 minutes. The slides were washed in PBS Assessment of in vivo transgene expression and then stained with propidium iodide (1 g/mL in PBS) Mice received 1Â106 Panc1 cells as an intraperitoneal for 15 minutes in the dark. The slides were washed with injection on day 0. They then received an intraperitoneal deionized water three times. One drop of Anti-Fade solution injection of 1Â109 pfu Ad gal on day 1, 8, or 15. The mice (Molecular Probes, Cambridge, UK) was added to the cells were sacrificed 2 or 7 days after the adenoviral admin- and a glass coverslip applied. The edges were sealed. The istration. Tumor and organs were harvested and were snap- cells were analyzed using a fluorescent microscope set to frozen, and composite blocks were mounted in OCT resin. 520 nm to view fluorescein and >620 nm to view propidium Frozen sections (8–10 m) of murine tissue to be stained iodide. were thawed and immediately fixed in 0.25% glutaraldehyde and 2% formaldehyde (in PBS) at 48C for 30 minutes. The In vivo experiments sections were then washed in PBS three times for 10 minutes All animals used for the in vivo experiments were female each. Complete X-gal solution was then placed over each nu/nu mice (ICRF BSU, Clare Hall, South Mimms). All section of tissue and the slides incubated overnight in a procedures were carried out according to Home Office humidified chamber at 378C. The slides were washed in PBS regulations and were covered by Home Office License and postfixed in the same fixation solution. The tissue was 90/717. counterstained with eosin. The tissue sections were then Human pancreatic carcinoma cell lines were cultured and observed under the light microscope for blue cells. On each washed with PBS before being detached from the plates by occasion that X-gal staining was performed, a positive trypsin, counted, and resuspended. A total of 1Â106 cells in control of 293 cells infected with an adenovirus expressing 0.1 mL of sterile saline were injected in the peritoneal cavity -galactosidase was prepared and tested. of each mouse. Another cohort of mice received an intraperitoneal Replication-defective adenoviruses Ad gal, AdTIMP1, injection of 5Â106 Suit2/MMP2 cells on day 0. They then and AdTIMP2 were used for in vivo work. All adenoviruses received an intraperitoneal injection of 1Â109 pfu Ad gal on were prepared as aliquots of 1Â109 pfu in 0.5 mL of PBS and day 19 and were sacrificed on day 21. Tumor and organs

Figure 2. Matrigel invasion assays of human pancreatic carcinoma cells trans- duced with adenoviral vectors expressing gal, TIMP1, or TIMP2. Each experiment was performed three times and with six replicates. Statistically significant reduc- tionsininvasionwerecalculatedby Student’s t test.

Cancer Gene Therapy, Vol 8, No 11, 2001 RIGG AND LEMOINE: TIMP GENE THERAPY 873 were harvested and frozen sections prepared. The frozen generated and purified by three cycles of limiting dilution sections were subjected to staining for -galactosidase as described. Restriction endonuclease digestion and expression as described above. Southern blot analysis using 32P-labelled TIMP1 or TIMP2 probes confirmed the presence of TIMP1 and TIMP2 Statistical analysis transgenes. Large-scale production of the recombinant adenoviruses ELISA and invasion data were analysed using an unpaired 12 Student’s t test and are shown as the mean±SEM. Statistical gave yields of approximately 2Â10 total particles per milliliter for each virus. AdTIMP2 was found to have a titer significance was defined as P<.05. Analysis of the survival 10 11 data for the in vivo experiments was performed by the log- of 3Â10 pfu/mL, whereas AdTIMP1 had a titer of 1Â10 rank test. These results are expressed as a P value. Kaplan- pfu/mL. Therefore, the ratio of infectious particles to total Meier curves were also derived. Hazard ratios could be particles was less than 1:100. calculated to estimate the risk of death at any time in a particular group. A value of less than 1 suggests a lower risk Adenoviral infection efficiency of death than the control group; likewise, a value greater than The optimal MOI for subsequent in vitro work was defined 1 suggests a higher risk of death. as the number of plaque-forming units per cell required to achieve 100% transduction. Mock-infected Panc1, T3M4, Suit2/MMP2, and PSN-1 cells were not positive with RESULTS staining for -galactosidase expression. When infected with Ad gal, the optimal MOI was 100 for Panc1, Suit2/MMP2, Generation of AdTIMP vectors and T3M4 cells, and for PSN-1 cells it was 500. These MOIs Replication-incompetent adenoviruses containing cDNA were used to infect cells in all subsequent in vitro experi- encoding gal, TIMP1, or TIMP2 were successfully ments. Toxicity was not observed with MOI <1000.

Figure 3. Statistical analysis of the re- sults from the administration of AdTIMP1 to mice bearing Panc1 human pancreatic tumours. The Kaplan-Meier survival curve is shown (a). Hazard ratios were calculated to show the hazard of death in a particular treatment group as com- pared to the untreated group (b) or the hazard ratios for comparisons between different treatment groups (c).

Cancer Gene Therapy, Vol 8, No 11, 2001 874 RIGG AND LEMOINE: TIMP GENE THERAPY

Analysis of transgene expression totic or antiapoptotic effect in difference circumstances. To To investigate the level of TIMP protein production in the investigate this possibility, pancreatic carcinoma cells were days after adenoviral transduction of tumor cells, super- infected with Ad gal, AdTIMP1, or AdTIMP2. The cells natants were collected at 12-hour intervals over 96 hours were fixed and subjected to fluorescein labelling of nicked from Panc1 cells infected with AdTIMP1 or AdTIMP2. ends of DNA. The cells were counterstained with propidium TIMP1 expression was low at time 0 and had increased iodide. The positive control (cells treated with DNase I prior within 12 hours and remained elevated until the final to labelling) demonstrated a number of cells with fragmented measurement at 96 hours. TIMP2 expression also rose within and condensed nuclear material at 520 nm. No apoptotic 12 hours of transduction with AdTIMP2 and was still cells were seen in the fluorescein-negative control. The four elevated at 96 hours (Fig 1). cell lines tested (Panc1, T3M4, Suit2/MMP2, PSN-1) demonstrated only occasional apoptotic cells whichever Effect of TIMP on invasion adenovirus they had been infected with, suggesting that the adenoviral vectors delivering -galactosidase, TIMP1, or Matrigel invasion assays were performed to assess the TIMP2 had no effect on the level of apoptosis within the cell invasive ability of human pancreatic carcinoma cells infected populations. with Ad gal, AdTIMP1, or AdTIMP2. The number of cells invading the lower surface of the membrane was assessed by In vivo experiments MTT assay, and from the standard curve for each cell line, Survival. A series of in vivo experiments was performed to the number could be obtained. The results for the four cell investigate whether adenoviral delivery of TIMP1 or TIMP2 lines are shown in Figure 2. In all cases, there was a would influence the survival of female nu/nu mice harboring significant reduction of invasion by cells infected with human intraperitoneal pancreatic carcinomas. The model AdTIMP1 or AdTIMP2 as compared with Ad gal. used 1Â106 Panc1 human pancreatic carcinoma cells injected in the intraperitoneal cavity of a nude mouse on Effect of adenoviral TIMP on cellular apoptosis day 0. The adenoviral vectors Ad gal, AdTIMP1, or There are several descriptions in the literature of gene AdTIMP2 were administered as a single intraperitoneal therapy, with TIMP1, TIMP2, or TIMP3 having a proapop- injection of 1Â109 pfu per mouse on day 1, 8, or 15. For all

Figure 4. Statistical analysis of the results from the administration of Ad- TIMP2 to mice bearing Panc1 human pancreatic tumours. The Kaplan-Meier survival curve is shown (a).Hazard ratios were calculated to show the hazard of death in a particular treatment group as compared to the untreated group (b) or the hazard ratios for comparisons between different treat- ment groups (c).

Cancer Gene Therapy, Vol 8, No 11, 2001 RIGG AND LEMOINE: TIMP GENE THERAPY 875

advantage compared to untreated control mice (P=.0293) (Fig 4). No survival benefit was seen for the AdTIMP2 treatment on days 8 and 15. The estimated hazard ratio for mice treated with AdTIMP2 on day 1 suggested a lower risk of death than the control mice (.265). A log-rank test of the three AdTIMP2 treatment groups showed that there was a significant difference in survival between the groups (P=.046). The administration of Ad gal did not show a survival benefit over control for any of the treatment time points at the end of 26 weeks. Tumor burden and volume of ascites. At the time of death, the volume of ascites and tumor weight were measured. Any mice alive and well at 26 weeks were defined as having no tumor. The results were subjected to statistical analysis by a Student’s t test. For the mice receiving AdTIMP1 or AdTIMP2 early, there was a significant reduction in tumor weight compared to the untreated control mice (P=.042 and P=.011, respectively). The intermediate and late adminis- tration of AdTIMP1 or AdTIMP2 did not significantly alter tumor weight (Fig 5).

Figure 5. Graphs demonstrating the mean tumor weight (a) and mean volume of ascites (b) for the groups of mice that received a single intraperitoneal injection of AdTIMP1. Each treatment group is compared with the control group by a two-sample t test and statistical significance is defined as P<.05 (denoted as *). the mice, the major endpoint was death. The experiment was continued for 26 weeks. Statistical analysis of the survival results for the mice receiving AdTIMP1 at the three time points was performed using the log-rank test (Fig 3). There was significantly longer survival for the mice receiving AdTIMP1 on day 1 compared to the control mice (P<.009) and the estimated hazard ratio suggested a reduced likelihood of death than the control mice (.173). Mice treated with AdTIMP1 on day 8 did not have a significant survival advantage over the control mice (P=.145), but the estimated hazard ratio suggested a lower risk of death. AdTIMP1 administered on day 15 did not significantly alter survival. When compar- ing the survival times of the three AdTIMP1 groups, there Figure 6. Graphs demonstrating the mean tumor weight (a) and is strong evidence that the three groups differ significantly mean volume of ascites (b) for the groups of mice that received a (P<.001). single intraperitoneal injection of AdTIMP2. Each treatment group is The mice that received AdTIMP2 on day 1 of the compared with the control group by a two-sample t test and statistical experiment did have a statistically significant survival significance is defined as P<.05 (denoted as *).

Cancer Gene Therapy, Vol 8, No 11, 2001 876 RIGG AND LEMOINE: TIMP GENE THERAPY

Similar analyses were performed for the volume of ascites This work has demonstrated that the modification of found in each mouse at the time of death. None of the human pancreatic carcinoma cells to overexpress TIMP1 AdTIMP1- or AdTIMP2-treated groups showed a signifi- or TIMP2 using adenoviral vectors can significantly alter cant difference in the volume of ascites from the control invasion in vitro and prolong survival in a pancreatic group (Fig 6). Ad gal had no effect on tumor weight or cancer animal model. Adenoviral vectors were chosen as volume of ascites at any timepoint. they can infect a broad host range, high titers are Assessment of gene transfer. It was important to establish the achievable, and they do not have an envelope to induce extent of gene delivery by adenoviral vectors in this model complement activation. Replication-incompetent adenovi- of intraperitoneal human pancreatic carcinoma in a nude ral vectors encoding the human TIMP1 or TIMP2 cDNA mouse. Mice were seeded with Suit2/MMP2 tumors on were successfully generated. These adenoviral vectors day 0 received 1Â109 pfu Ad gal on day 19 and were could be produced in large quantities at satisfactory titer sacrificed days after the adenoviral injection. Of note, no and be shown to be free of replication-competent tumor was seen in the mice bearing Panc1 cells on viruses. sacrifice so that only normal organs could be examined. For each cell line, a variety of different MOIs were tested. Tissues and tumors were harvested for transgene analysis. It was found that for Panc1, T3M4, and Suit2/MMP2 cells, The majority of tumors were found around the mesentery an MOI of 100 was required to achieve 100% transduction. and over the pancreas. Therefore, frozen sections were PSN-1 required an MOI of 500 to achieve 100% trans- made of these tissues. All six mice showed blue cells duction. These levels reflect favorably with reports in the literature of between 100 and 1000 adenoviral MOI being around the periphery of the tissues, but also regions of 23 tumor that had infiltrated the pancreas expressing - necessary to transduce tumor cells. galactosidase (data not shown). Sections from composite Adenoviral delivery of TIMP1 and TIMP2 produces dramatic reductions in invasion for the four cell lines in vitro. organ blocks demonstrated that there was no -galactosi- 22 dase expression in the brain, heart, liver, kidneys, lung, or These findings are compatible with those of Ahonen et al spleen after the intraperitoneal administration of Ad gal to who demonstrated a reduction in invasion of melanoma cells these mice. in vitro after infection with TIMP-expressing adenoviral vectors. There are several papers that describe the ability of TIMPs DISCUSSION to influence apoptosis.22,34 This was relevant to the reduction in cellular invasion, as it might have been due to increased One of the criticisms of gene therapy is that for many of the apoptosis of modified cells such that there were fewer approaches to achieve a cure, 100% of the tumor cells would available to migrate through the matrigel and porous need to be modified. Therefore, strategies that affect membrane. None of the adenoviruses altered the background nontransduced as well as transduced cells would be of use. level of occasional apoptotic cells. The ‘‘bystander effect’’ is described for genetic prodrug It was important to try and develop an in vivo model that activation therapy,31 but the concept also applies to stromal represented the scenario of a human pancreatic cancer. targeting approaches. The stromal and endothelial cells There have been experiments with orthotropic injection of within a tumor play a crucial role in the structure of the human tumor cells in the pancreas of an immunosuppressed extracellular matrix and hence the invasion and angiogenesis mouse. While this provides a tumor in the correct that take place. A gene therapy strategy that disrupted the anatomical location, it requires laparotomy under general extracellular environment to limit invasion would influence anaesthetic.35 A less invasive method is the injection of nontransduced and transduced cells alike. This ‘‘field effect’’ human pancreatic cancer cells into the peritoneal cavity. has been demonstrated by xenograft tumors in nude mice The tumors form as numerous intraperitoneal nodules and becoming encapsulated in a fibrous layer with restricted mimic the widespread metastases often seen with intra- invasion after treatment with TIMPs.17,20 abdominal cancers. Several groups have demonstrated that Other researchers have investigated the behavior of this model produces disseminated tumors that can be high TIMP-expressing tumor cells.17,19,21 The tumor targeted by gene therapy.36 –38 cells were transfected ex vivo, so while proving the Using this model of intraperitoneal pancreatic cancer, we principle of tumor growth limitation by TIMPs, this does investigated survival in tumor-bearing nude mice following not reflect the clinical scenario of using TIMP gene a single injection of an adenoviral vector. AdTIMP1 and therapy to treat a tumor that is already established. To AdTIMP2 showed a statistically significant increase in date, very little work has been published on the in vivo survival for mice that received therapy on the day after transduction of cells with TIMPs. One group injected tumor seeding. Administration of AdTIMP1 or AdTIMP2 1 retroviral vector producer cells in subcutaneous tumors. or 2 weeks after tumor seeding did not improve survival. The retroviruses that were released encoded TIMP2 The prolonged survival with early administration of cDNA. The growth and invasion of the injected tumors AdTIMP1 and AdTIMP2 may reflect the fact that tumor were reduced compared to control tumors.20 TIMP gene cells were not able to establish themselves in the peritoneal therapy has also been attempted for coronary artery cavity by invasion and promoting angiogenesis. It is disease. The gene transfer of TIMP2 or TIMP3 to the possible that administration after tumor cell implantation intimal cells of human saphenous veins limits neointimal has occurred is too late to inhibit tumor as it is already thickening in vitro.32,33 established.

Cancer Gene Therapy, Vol 8, No 11, 2001 RIGG AND LEMOINE: TIMP GENE THERAPY 877

In the in vivo experiments, the weight of tumor and the 6. Docherty AJP, Lyons A, Smith BJ, et al. Sequence of human volume of ascites were measured at the time of death. For the tissue inhibitor of metalloproteinases and its identity to mice receiving early AdTIMP1 or AdTIMP2, there was a erythroid-potentiating activity. Nature. 1985;318:66–69. reduction in tumor burden at 26 weeks compared to the 7. Stetler-Stevenson WG, Krutzsch H, Liotta LA. Tissue inhibitor of metalloproteinase (TIMP-2): a new member of the untreated control mice. The intermediate and late adminis- metalloproteinase inhibitor family. J Biol Chem. 1989;264: tration of AdTIMP1 or AdTIMP2 did not cause a reduction 17374–17378. in measurable disease. There was no statistically significant 8. Uria JA, Ferrando AA, Velasco G, Freije JMP, Lopez-Otin C. effect on the volume of ascites. Structure and expression in breast tumors of human TIMP-3, a These experiments have shown that TIMP1 or TIMP2 new member of the metalloproteinase inhibitor family. Cancer delivered by an adenoviral vector can improve survival in Res. 1994;54:2091–2094. tumor-bearing mice provided that the treatment is given 9. Leco KJ, Apta SS, Taniguchi GT, et al. Murine tissue inhibitor of early. This raises difficulties for the clinical setting as the metalloproteinases 4 (TIMP-4): cDNA isolation and express- vast majority of pancreatic tumors present late. Therefore, ion in adult mouse tissues. FEBS Lett. 1997;401:213–217. it is possible that for such patients, TIMP administration 10. Gress TM, Muller-Pillasch F, Lerch MM, Friess H, Buchler M, Adler G. Expression and in situ localization of genes encoding would be best suited as a combination therapy with for extracellular matrix degrading proteases in pancreatic cytotoxic agents. Certainly, for a mouse model of lung cancer. Int J Cancer. 1995;62:407–413. cancer, it has been shown that a synthetic MMP inhibitor 11. Satoh K, Ohtani H, Shimosegawa T, Koizumi M, Sawai T, had a synergistic effect when given with cyclophospha- Toyota T. Infrequent stromal expression of gelatinase A and mide over either treatment alone.39 There are other clinical intact basement membrane in intraductal neoplasms of the settings in which TIMP gene therapy might be appropriate. pancreas. Gastroenterology. 1994;107:1488–1495. As the antitumor effect is most pronounced if the treatment 12. Bramhall SR, Neoptolomemos J, Stamp GWH, Lemoine is given early, it might be of use in the perioperative NR. Imbalance of expression of matrix metalloproteinases period for patients who have had tumor resections. This (MMPs) and tissue inhibitors of the matrix metallopro- teinases (TIMPs) in human pancreatic carcinoma. J Pathol. could be helpful for patients with intra-abdominal 1997;182:347–355. malignancies who are at high risk of tumor recurrence. 13. Schultz RM, Silberman S, Persky B, Bajkowski AS, The occurrence of port site metastases has been described Carmichael DF. Inhibition by human recombinant tissue for patients undergoing minimal access surgery who are inhibitor of metalloproteinases of human amnion invasion found to have occult tumors. The use of TIMP gene and lung colonization by murine B16-F10 melanoma cells. therapy in this setting might be advantageous. Also, TIMP Cancer Res. 1988;48:5539–5545. gene therapy might be considered for palliation of 14. Alvarez OA, Carmichael DF, DeClerck YA. Inhibition of symptoms, e.g., the control of pleural effusions or ascites collagenolytic activity and metastasis of tumor cells by a following fluid drainage. recombinant human tissue inhibitor of metalloproteinases. J Natl Cancer Inst. 1990;82:589–595. 15. DeClerck YA, Yean T-D, Chan D, Shimada H, Langley KE. Inhibition of tumor invasion of smooth muscle cell layers by recombinant human metalloproteinase inhibitor. Cancer Res. ACKNOWLEDGMENTS 1991;51:2151–2157. 16. Albini A, Melchiori A, Santi L, Liotta LA, Brown PD, Anne Rigg was supported by the Mike Stone Cancer Stetler-Stevenson WG. Tumor cell invasion inhibited by Research Fund. The authors thank Gillian Murphy for her TIMP-2. J Natl Cancer Inst. 1991;83:775–779. guidance, and Bruce Roberts and Rhonda Doll at Genzyme 17. DeClerck YA, Perez N, Shimada H, Boone TC, Langley KE, for donation of the adenoviral vector system and advice on Taylor SM. Inhibition of invasion and metastasis in cells its usage. transfected with an inhibitor of metalloproteinases. Cancer Res. 1992;52:701–708. 18. Khokha R. Suppression of the tumorigenic and metastatic abilities of murine B16-F10 melanoma cells in vivo by the overexpression of the tissue inhibitor of the metalloproteinase-1. REFERENCES J Natl Cancer Inst. 1994;86:299–304. 19. Montgomery AMP, Mueller BM, Riesfeld RA, Taylor SM, 1. Monitor Population and Health, Office for National Statistics, DeClerck YA. Effect of tissue inhibitor of the matrix UK, 1998. metalloproteinases-2 expression on the growth and sponta- 2. Stetler-Stevenson WG, Liotta LA, Kleiner DE Jr. Extracellular neous metastasis of a human melanoma cell line. Cancer Res. matrix 6: role of matrix metalloproteinases in tumor invasion 1994;54:5467–5473. and metastasis. FASEB J. 1993;7:1434–1441. 20. Imren S, Kohn DB, Shimada H, Blavier L, DeClerck YA. 3. Matrisian L. The matrix-degrading metalloproteinases. Overexpression of tissue inhibitor of metalloproteinases-2 by Bioessays. 1992;14:455–463. retroviral-mediated gene transfer in vivo inhibits tumor growth 4. Senger D. Molecular framework for angiogenesis: a complex and invasion. Cancer Res. 1996;56:2891–2895. web of interactions between extravasated plasma proteins and 21. Koop S, Khokha R, Schmidt EE, et al. Overexpression of endothelial cell proteins induced by angiogenic cytokines. Am J metalloproteinase inhibitor in B16F10 cels does not affect Pathol. 1996;149:1–7. extravasation but reduces tumor growth. Cancer Res. 5. Murphy G, Docherty AJP. The matrix metalloproteinases 1994;54:4791–4797. and their inhibitors. Am J Respir Cell Mol Biol. 1992;7: 22. Ahonen M, Baker AH, Kahari V-M. Adenovirus-mediated 120–125. gene delivery of tissue inhibitor of metalloproteinases-3

Cancer Gene Therapy, Vol 8, No 11, 2001 878 RIGG AND LEMOINE: TIMP GENE THERAPY

inhibits invasion and induces apoptosis in melanoma cells. 32. George SJ, Baker AH, Angelini GD, Newby AC. Gene transfer Cancer Res. 1998;58:2310–2315. of tissue inhibitor of metalloproteinase-2 inhibits metallopro- 23. Baker AH, George SJ, Zaltsman AB, Murphy G, Newby AC. teinase activity and neointimal formation in human saphenous Inhibition of invasion and induction of apoptotic cell death of veins. Gene Ther. 1998;5:1552–1560. cancer cell lines by overexpression of TIMP-3. Br J Cancer. 33. George SJ, Lloyd CT, Angelini GD, Newby AC, Baker AH. 1999;79:1347–1355. Inhibition of late vein graft neointimal formation in human and 24. Albini A, Fontanini G, Masiello L, et al. Angiogenic potential in porcine models by adenovirus-mediated overexpression of vivo by Kaposi’s sarcoma cell-free supernatants and HIV-1 tat tissue inhibitor of metalloproteinase-3. Circulation. 2000;101: product: inhibition of KS-like lesions by tissue inhibitor of 296–304. metalloproteinase-2. AIDS. 1994;8:1237–1244. 34. Guedez L, Stetler-Stevenson WG, Wolff L, et al. In vitro 25. Anand-Apte B, Pepper MS, Voest E, et al. Inhibition of suppression of programmed cell death of B cells by tissue angiogenesis by tissue inhibitor of metalloproteinase-3. Invest inhibitor of metalloproteinases-1. J Clin Invest. 1998;102: Ophthalmol Vis Sci. 1997;38:817–823. 2002–2010. 26. Johnson MD, Kim HR, Chesler L, Tsao Wu G, Bouck N, 35. Yang L, Hwang R, Pandit L, Gordon EM, French Anderson W, Polverini PJ. Inhibition of angiogenesis by tissue inhibitor of Parekh D. Gene therapy of metastatic pancreas cancer with metalloproteinase. J Cell Physiol. 1994;160:194–202. intraperitoneal injections of concentrated retroviral herpes 27. Murphy AN, Unsworth EJ, Stetler-Stevenson WG. Tissue simplex thymidine kinase vector supernatant and ganciclovir. inhibitor of metalloproteinases-2 inhibits bFGF-induced Ann Surg. 1996;224:405–417. human microvascular endothelial cell proliferation. J Cell 36. Aoki K, Yoshida T, Sugimura T, Terada M. Liposome- Physiol. 1993;157:351–358. mediated in vivo gene transfer of antisense K-ras construct 28. Takigawa M, Nishida Y, Suzuki F, Kishi J, Yamashita K, inhibits pancreatic tumor dissemination in the murine perito- Hayakawa T. Induction of angiogenesis in chick yolk sac neal cavity. Cancer Res. 1995;55:3810–3816. membrane by polyamines and its inhibition by tissue inhibitors 37. Aoki K, Yoshida T, Matsumoto N, et al. Gene therapy for of metalloproteinases (TIMP and TIMP-2). Biochem Biophys peritoneal dissemination of pancreatic cancer by liposome- Res Commun. 1990;171:1264–1271. mediated transfer of herpes simplex thymidine kinase gene. 29. Gerald R, Meidel RS. Adenoviral vectors. In: Glover DM, Hum Gene Ther. 1997;8:1105–1113. Hames BD, eds. DNA Cloning 4. A Practical Approach. 38. Rosenfeld ME, Vickers SM, Raben D, et al. Pancreatic Oxford: Oxford Univ. Press; 1996:285–305. carcinoma cell killing via adenoviral mediated delivery of the 30. Mittereder N, March KL, Trapnell BC. Evaluation of the herpes simplex virus thymidine kinase gene. Ann Surg. concentration and bioactivity of adenovirus vectors for gene 1997;225:609–620. therapy. JVirol. 1996;70;7498–7509. 39. Anderson IC, Shipp MA, Docherty AJP, Teicher BA. 31. Freeman SM, Abboud CN, Whartenby KA, et al. The Combination therapy including a gelatinase inhibitor and ‘‘bystander effect’’: tumor regression when a fraction of the cytotoxic agent reduces local invasion and metastasis of tumor mass is genetically modified. Cancer Res. 1993;53: murine lewis lung carcinoma. Cancer Res. 1996;56:715– 5274–5283. 718.

Cancer Gene Therapy, Vol 8, No 11, 2001