Antitumor Activity and Bystander Effect of Adenovirally ... - Cell Press

doi:10.1006/mthe.2002.0606, available online at http://www.idealibrary.com on IDEAL

ARTICLE

Antitumor Activity and Bystander Effect of Adenovirally Delivered Tissue Inhibitor of Metalloproteinases-3 Matti Ahonen,1,2 Risto Ala-Aho,1,2 Andrew H. Baker,3 Sarah J. George,4 Reidar Grénman,5 Ulpu Saarialho-Kere,6 and Veli-Matti Kähäri1,2,* 1 Centre for Biotechnology, University of Turku and Åbo Akademi University, Turku, Finland Departments of Medical Biochemistry and of Dermatology, University of Turku, Turku, Finland 3 Department of Medicine and Therapeutics, University of Glasgow, Glasgow, UK 4 Bristol Heart Institute, Bristol Royal Infirmary, Bristol, UK 5 Department of Otorhinolaryngology–Head and Neck Surgery, Turku University Central Hospital, Turku, Finland 6 Department of Dermatology, University of Helsinki, Helsinki, Finland 2

*To whom correspondence and reprint requests should be addressed. Fax: (+358) 2-3338000. E-mail: [email protected].

We have studied the effect of a newly identified tumor suppressor tissue inhibitor of metalloproteinases-3 (TIMP-3) on the growth of human melanoma and squamous-cell carcinoma (SCC). Adenoviral delivery of the TIMP-3 gene to human melanoma (A2058) and SCC (UT-SCC-7) cells ex vivo inhibited tumorigenesis after subcutaneous (s.c.) injection of the infected cells into SCID/SCID mice. Three daily consecutive intratumoral injections of 1.4 ⫻ 109 plaque-forming units (pfu) of TIMP-3 adenovirus (RAdTIMP-3) inhibited the growth of preestablished melanoma and SCC xenografts in SCID/SCID mice, whereas growth of control virus-injected tumors was not affected. The antitumor effect of RAdTIMP-3 was obtained with in vivo adenoviral transduction efficiency of 8–10%, and it was more potent than that of adenovirally delivered p53. Adenovirusmediated expression of TIMP-3 potently reduced gelatinolytic activity, increased the number of apoptotic cells, and inhibited vascularization of melanomas. Escalation of the adenoviral dose to three rounds of three daily consecutive injections with 1.4 ⫻ 109 pfu of RAdTIMP-3 every 6 days entirely inhibited growth of injected melanomas for 32 days. Mixing RAdTIMP-3-infected A2058 cells with uninfected cells in 1:1 ratio in culture resulted in death of all cells in 96 hours. Adenovirally delivered TIMP-3 was also expressed by A2058 cells in soluble form into the culture medium, where it exerted a cytotoxic effect on uninfected A2058 cell cultures after relocating to the cell layer. These results identify TIMP-3 as a novel type of secreted tumor suppressor, which has antiinvasive, antiangiogenic, and proapoptotic effects in vivo, and which displays a potent bystander effect validating further exploration of its applicability in human cancer gene therapy. Key Words: melanoma, squamous-cell carcinoma, tissue inhibitor of metalloproteinase, matrix metalloproteinase, adenovirus

INTRODUCTION Growth, invasion, and metastasis of malignant tumors are dependent on their ability to degrade basement membranes and extracellular matrix (ECM) in an effective and controlled manner [1]. Matrix metalloproteinases (MMPs) are a family of at least 21 zinc-dependent endopeptidases collectively capable of degrading essentially all ECM components, and they play an important role in tumor invasion and metastasis [1]. Malignant melanomas are cutaneous neoplastic tumors characterized by high invasion and metastasis capacity resulting in poor prognosis [2]. Upregulation of the expression of several MMPs, including collagenase-1 (MMP-1), gelatinase-A (MMP-2), gelatinase-B (MMP-9), collagenase-3 (MMP-13), and membrane MOLECULAR THERAPY Vol. 5, No. 6, June 2002 Copyright © The American Society of Gene Therapy 1525-0016/02 $35.00

type-1 MMP (MT1-MMP; MMP-14), has been detected in invasive primary and metastatic melanomas in vivo, suggesting a role for them in melanoma growth and invasion [3–8]. The activity of MMPs is specifically inhibited by tissue inhibitors of metalloproteinases (TIMPs), which bind to active MMPs in 1:1 molar stoichiometry in the extracellular space [1,2]. TIMP-1, TIMP-2, and TIMP-4 are secreted in soluble form, whereas TIMP-3 is associated with ECM of cells in culture and in vivo [9–11]. TIMP-1 inhibits the activity of most MMPs, with the exception of MT1-MMP and MMP-2 [1,2]. TIMP-3 inhibits the activity of MMP-1, MMP-2, stromelysin-1 (MMP-3), MMP-9, and MMP-13 [12]. In addition, TIMP-3 inhibits the activity of

705

ARTICLE


TABLE 1: TIMP-3 inhibits melanoma and squamous cell carcinoma tumor growth in vivo A2058 Treatment

PBS

RAdLacZ

Tumor take

5/5

5/5

Mean tumor volume (mm3)

1453 ± 165 1019 ± 321

UT-SCC-7 RAdTIMP-1

RAdTIMP-3

PBS

RAdpCA3

RAdTIMP-3

3/5

0/5*

5/5

4/4

0/4**

869 ± 751

0

1044 ± 268

889 ± 268

0

Human melanoma (A2058) and squamous cell carcinoma (UT-SCC-7) cells were transduced in culture with recombinant replication deficient adenoviruses harboring E. coli ␤-galactosidase (RadLacZ), TIMP-1 (RAdTIMP-1), and TIMP-3 (RAdTIMP-3), or with empty control virus (RAdpCA3), as indicated at MOI 100 (A2058) or 500 (UT-SCC-7), which gives 100% infection efficiency in these cells. Control cultures were mock-transduced with phosphate buffered saline (PBS). Cells were detached 16 hours after infection and implanted subcutaneously in the back of SCID/SCID mice. For A2058 cells tumor take as number of mice with tumors 4 weeks (PBS, RAdLacZ, and RAdTIMP-1) and 12 weeks (RAdTIMP-3) after injection, and mean volume ± SD of tumors are shown. For UT-SCC-7 cells tumor take as number of mice with tumors 9 weeks (PBS, RAdpCA3, and RAdTIMP-3) after injection, and mean volume ± SD of tumors are shown. Statistical significance of tumor take determined with Fisher’s exact t-test: for A2058 cells, *P = 0.0079 (RAdTIMP-3 versus RadLacZ); RAdTIMP-1 versus PBS and RadLacZ, not significant; for UT-SCC-7 cells **P = 0.029 (RAdpCA3 versus RAdTIMP-3).

membrane-anchored proteinases containing disintegrin and metalloproteinase (ADAM) and thrombospondin-like (TS) domains: tumor necrosis factor-␣ (TNF-␣)–converting enzyme (TACE; ADAM-17), ADAM-10, aggrecanase-1 (ADAM-TS4), and aggrecanase-2 (ADAM-TS5) [13–15]. TIMP-3 inhibits shedding of ectodomains of several cellsurface proteins, such as TNF-␣, TNF-␣ receptor-I (TNF-RI), syndecan-1 and -4, interleukin-6 receptor, and L-selectin [13,15–19]. Recent studies have also identified TIMP-3 as a putative tumor suppressor, the expression of which is specifically repressed during oncogenic transformation and inactivated by methylation of the gene in several types of malignant tumors, such as kidney, lung, colon, breast, brain, and pancreatic cancers [20–24]. Accordingly, TIMP3 has been shown to promote apoptosis in several types of normal and malignant human cells in culture [25–28]. The ability of TIMPs to inhibit MMP activity has raised interest in their applicability to combat growth and invasion of malignant tumors. Accordingly, overexpression of TIMP-1, TIMP-2, and TIMP-4 has been shown to inhibit malignant tumor growth in vivo [29–31]. We have shown that adenovirally mediated expression of TIMP-3 inhibits invasion and adhesion of melanoma cells in culture and promotes apoptosis in these cells [27]. In the present study we have examined the effect of adenovirally delivered TIMP-3, which potently inhibits MMP activity and has a direct tumor-suppressing effect. Here we show for the first time that adenovirus-mediated gene delivery of TIMP-3 to melanoma and squamous-cell carcinoma (SCC) cells ex vivo inhibits their implantation and growth in immunocompromised mice in vivo. We also demonstrate that intratumoral injection of preestablished melanomas and SCC cells in vivo with recombinant adenovirus harboring TIMP3 inhibits tumor growth more potently than tumor suppressor gene p53 and results in reduced gelatinolytic activity, induction of apoptosis, and inhibition of tumor angiogenesis. In addition, we show that adenovirally delivered TIMP-3 is also produced by melanoma cells in soluble form into the culture medium, where it exerts a cytotoxic effect on uninfected melanoma cell cultures after homing to the cell layer. These data demonstrate for the first time that adenoviral TIMP-3 expression potently

706

suppresses tumor growth in vivo and exerts a cytotoxic bystander effect on uninfected tumor cells.

RESULTS TIMP-3 Inhibits Melanoma and Squamous-Cell Carcinoma Tumor Growth To examine the effect of TIMP-3 on tumor growth in vivo, we first infected human A2058 melanoma cells ex vivo with RAdTIMP-3 (a replication-deficient adenovirus harboring TIMP-3), in parallel with RAdTIMP-1 (a TIMP-1coding adenovirus) and RAdLacZ (an adenovirus harboring the Escherichia coli ␤-galactosidase gene), at a multiplicity of infection (MOI) of 100, which gives 100% transduction efficiency in these cells [27]. Control cultures were mock-transduced in parallel with PBS. The cells were trypsinized 16 hours after infection and cells (5 ⫻ 106) injected s.c. into the backs of SCID/SCID mice. By 4 weeks all mice injected with mock-transduced and RAdLacZinfected melanoma cells had prominent tumors (Table 1). None of the mice injected with RAdTIMP-3-infected melanoma cells had tumors by 4 weeks after injection, and no tumor growth was observed during an additional 8-week observation period. The difference in tumor formation with RAdTIMP-3-infected melanoma cells compared with control virus-infected cells was significant at P = 0.0079 (Table 1). In comparison, RAdTIMP-1-infected melanoma cells formed tumors in three of five mice, and the tumors were somewhat smaller than those formed by RAdLacZ- and PBS-infected cells (Table 1). However, the difference in tumor occurrence and size between RAdTIMP-1-infected cells and PBS- or RAdLacZ-transduced melanoma cells was not significant (Table 1). No histologic differences were detected between tumors formed by RAdLacZ- or RAdTIMP-1-infected melanoma cells, as compared with uninfected melanoma cells (data not shown). We also studied the effect of TIMP-3 on tumor cell implantation with UT-SCC-7 cells, established from metastasis of cutaneous SCC [32]. We transduced UT-SCC-7 cells with RAdTIMP-3 or empty adenovirus RAdpCA3 at a MOI of 500 (which gives 100% transduction efficiency in these cells [32]), trypsinized the cells 16 hours later, and injected

MOLECULAR THERAPY Vol. 5, No. 6, June 2002 Copyright © The American Society of Gene Therapy


A

ARTICLE

B

FIG. 1. Adenovirus-mediated gene delivery of TIMP-3 inhibits melanoma growth in vivo. A2058 melanoma cells (1 ⫻ 106) were implanted s.c. into the backs of SCID/SCID mice, and tumors were allowed to grow to 50–100 mm3 in size. (A) Growth of tumors treated every 24 hours for 3 days with injections of PBS (100 ␮l), or with adenoviruses harboring Escherichia coli ␤-galactosidase (RAdLacZ), TIMP-1 (RAdTIMP-1), and TIMP-3 (RAdTIMP-3; 1.4 ⫻ 109 pfu each). (B) Growth of tumors injected once every 24 hours for 3 days, and then repeating this series of injections altogether three times with 3 days in between series, with PBS (100 ␮l), empty control adenovirus (RAd66), or RAdTIMP-3 (1.4 ⫻ 109 pfu each). Statistical significance determined by Student’s t-test: (A) at day 28, RAdTIMP3 versus RAdLacZ, P < 0.00002; RAdTIMP-3 versus PBS, P < 0.0004; RAdTIMP-1 versus RAdLacZ, not significant; RAdTIMP-1 versus PBS, not significant; RAdLacZ versus PBS, not significant. RAdTIMP-3 (n = 5), RAdTIMP-1 (n = 5), RAdLacZ (n = 5), PBS (n = 6). (B) At day 34, RAdTIMP-3 versus RAd66, P < 0.007; RAdTIMP3 versus PBS, P < 0.0002; RAd66 versus PBS, not significant. RAdTIMP-3 (n = 5), RAdTIMP-1 (n = 4), RAd66 (n = 5), PBS (n = 4).

them (5 ⫻ 106 cells) s.c. into SCID/SCID mice. All mice injected with mock-infected and control virus-infected cells had detectable tumors 9 weeks after injection, whereas we detected no tumors in mice injected with RAdTIMP-3-transduced UT-SCC-7 cells (Table 1). The difference in tumor occurrence between the RAdTIMP-3 group and the control virus group was significant at P = 0.024 (Table 1). TIMP-3 Inhibits Growth of Melanoma Xenografts To examine the effect of TIMP-3 on the growth of preestablished tumors in vivo, we implanted A2058 melanoma cells (1 ⫻ 106) s.c. into the backs of SCID/SCID mice, which developed palpable tumors (50–100 mm3) in 14 days. We then injected the tumors once every 24 hours for 3 days with RAdTIMP-3, RAdTIMP-1, and RAdLacZ (1.4 ⫻ 109 pfu each) or with PBS (100 ␮l), and measured the size of tumors twice a week. The growth of RAdTIMP-3-injected tumors was clearly inhibited, whereas RAdTIMP-1- and RAdLacZ-injected tumors continued to grow after the injections (Fig. 1A). The tumors in the RAdTIMP-3 group were 84% smaller than those in the PBS group and 75% smaller than in the RAdLacZ group, as measured 14 days after the first injection with recombinant adenoviruses or PBS (Fig. 1A). RAdTIMP-1 had no marked effect on tumor growth, as compared with PBS- or RAdLacZ-injected tumors (Fig. 1A). Next, we escalated the adenoviral dose by injecting the melanomas with 1.4 ⫻ 109 pfu of RAdTIMP-3, empty control virus RAd66, or PBS (100 ␮l), once every 24 hours for 3 days and then repeating this series of injections


altogether three times with 3 days in between series (total of 21 days; Fig. 1B). The growth of RAdTIMP-3-injected tumors was entirely inhibited after starting the injections, whereas PBS- and RAd66-injected tumors continued to grow rapidly. The mice in these groups had to be killed 18 days after the first injection. At this time the tumors in the RAdTIMP-3 group were 85% smaller than those in the PBS group and 81% smaller than in the RAd66-injected group. No significant growth in RAdTIMP-3-injected tumors was detected during an additional 14-day followup, extending up to 32 days after the first injection (Fig. 1B). Transduction of Melanoma Cells in Vivo by Recombinant Adenoviruses We verified the expression of TIMP-1 and TIMP-3 in melanomas following intratumoral injection of RAdTIMP1 and RAdTIMP-3 by in situ hybridization. Expression of TIMP-3 and TIMP-1 mRNA was detected in ~ 5–10% of tumor cells at day 1 after three daily injections of 1.4 ⫻ 109 pfu of the corresponding virus (Fig. 2, top). The level of TIMP-3 and TIMP-1 mRNA expression was similar at day 7, and the expression gradually declined by day 14 after infection, although the expression was still detectable (data not shown). However, we did not detect expression of TIMP-3 and TIMP-1 mRNA in PBS- and RAd66-injected melanomas (Fig. 2, top, and data not shown). To estimate the efficacy of adenovirus-mediated gene delivery in vivo, we injected the tumors once every 24 hours for 3 days with RAdLacZ (1.4 ⫻ 109 pfu) and stained them for ␤-galactosidase. Cells positive for ␤-galactosidase were detected mainly adjacent to the injection site but

707

ARTICLE


FIG. 2. TIMP-3 inhibits gelatinolytic activity and promotes apoptosis in melanomas. A2058 melanoma xenografts grown in SCID/SCID mice were injected every 24 hours for 3 days with adenoviruses harboring TIMP-3 (RAdTIMP-3), TIMP-1 (RAdTIMP-1), and with empty control adenovirus (RAd66; 1.4 ⫻ 109 pfu each), or with PBS (100 ␮l). Tumors were harvested and analyzed on day 1 after the injections. (Top) Expression of TIMP-1 and TIMP-3 mRNA was determined by in situ hybridization. Presence of autoradiographic grains seen in white in dark-field exposure indicating expression of TIMP-1 and TIMP-3 mRNAs in melanoma cells in tumors injected with RAdTIMP-1 and RAdTIMP-3, respectively. No expression of TIMP-3 mRNA is detected in PBS- and RAd66-injected melanomas. (Middle) Gelatinolytic activity in tumors determined with in situ gelatinase zymography. Marked gelatinase activity was noted as white areas of gelatin degradation in PBS- and RAd66-injected tumors. Reduction in gelatinase activity was detected in RAdTIMP-1- and RAdTIMP-3-injected tumors. (Bottom) The presence of apoptotic cells was determined by TUNEL staining of tumor sections. Increased number of TUNEL-positive apoptotic cells (brown staining, arrows) is seen in RAdTIMP3-injected tumors. Scale bars, 100 ␮m.

also at the edges of tumor below the fibrous capsule surrounding the tumor (data not shown). Using this injection protocol, 8% of melanoma cells stained positive for ␤galactosidase 24 hours after the third injection (data not shown). No ␤-galactosidase-positive cells were detected in PBS-injected tumors (data not shown). Following intratumoral injection of RAdLacZ, we also detected single ␤galactosidase-positive cells in the liver, whereas lungs and kidney were negative for ␤-galactosidase activity (data not shown). TIMP-3 and TIMP-1 Inhibit Gelatinolytic Activity in Melanomas in Vivo To verify the inhibitory effect of adenovirally delivered TIMP-3 and TIMP-1 on MMP activity in vivo, we studied tumor sections by in situ gelatin zymography 1, 7, and 14 days after three consecutive daily injections with RAdTIMP-3, RAdTIMP-1, RAd66, or PBS. Marked gelatinase activity was observed at all time points in RAd66and PBS-injected tumors (Fig. 2, middle, and data not shown). However, potent reduction in gelatinolytic activity was evident in RAdTIMP-3- and RAdTIMP-1-injected tumors, as compared with RAd66- and PBS-injected tumors at day 1 (Fig. 2, middle). At day 7, reduction in gelatinase

708

activity in RAdTIMP-3- and RAdTIMP-1-injected tumors was even more potent than on day 1, and a gradual increase in gelatinase activity was noted at day 14 in RAdTIMP-3- and RAdTIMP-1-injected tumors (data not shown). Although we detected the expression of TIMP-3 and TIMP-1 mRNA in a limited number of tumor cells (Fig. 2, top), inhibition of gelatinolytic activity was noted throughout the tumor, indicating secretion of functional TIMP-3 and TIMP-1 to the extracellular space and distribution of these MMP inhibitors within the tumor tissue from the site of production. TIMP-3 Promotes Apoptosis in Melanoma Cells in Vivo We have previously noted that adenovirally mediated TIMP-3 expression promotes apoptosis in melanoma cells and other types of malignant cells in culture, as well as in smooth-muscle cells in culture and in vein grafts in vivo [26–28,33]. In this instance, we observed apoptotic melanoma cells in RAdTIMP-3-injected tumors in the vicinity of the injection site and outer borders of tumor (Fig. 2, bottom). In contrast, only single apoptotic cells were detectable in RAdTIMP-1-, RAd66-, and PBS-injected tumors. Quantitation of the percentage of terminal



A

ARTICLE

B

FIG. 3. TIMP-3 promotes apoptosis and inhibits angiogenesis in melanomas. A2058 melanoma xenografts grown in SCID/SCID mice were injected every 24 hours for 3 days with adenovirus harboring TIMP-3 (RAdTIMP-3), TIMP-1 (RAdTIMP-1), and with empty control adenovirus (RAd66; 1.4 ⫻ 109 pfu each), or with PBS (100 ␮l). (A) Relative area occupied by TUNEL-positive cells in tumor sections was determined with digital-image analysis on day 1, 7, or 14 after the injections. Relative number of apoptotic cells in RAdTIMP-3-injected tumors is increased on days 1 and 7, as compared with tumors injected with PBS, control virus RAd66, and RAdTIMP-1. (B) Blood vessels were immunostained with anti-human von Willebrand factor in tumors at day 14 after three consecutive injections with PBS, RAd66, RAdTIMP-1, and RAdTIMP-3. The number of von Willebrand-positive blood vessels containing erythrocytes was counted in five adjacent microscopic fields with ⫻10 magnification in the margin of two tumors. The number of blood vessels is reduced in RAdTIMP-3- and RAdTIMP-1-injected tumors, as compared with PBS- and RAd66-injected tumors.

deoxynucleotide transferase nick-end labeling (TUNEL)–positive cells revealed a marked increase (15-fold) in the number of apoptotic cells in RAdTIMP-3-injected tumors at day 1 after the three injections, as compared with RAd66-injected tumors (Fig. 3A). Increased number of apoptotic cells in RAdTIMP-3-injected tumors was also evident at day 7 after the third injection, and decreased by day 14 (31 days after implantation), apparently as a result of reduced TIMP-3 expression. TIMP-3 and TIMP-1 Inhibit Vascularization of Melanoma Xenografts Both TIMP-1 and TIMP-3 have been shown to inhibit angiogenesis in vivo [34,35]. Therefore, we determined vascularization of the melanoma xenografts by counting blood vessels with endothelial cells positive for von Willebrand factor and containing intraluminal erythrocytes. At day 14 after the three injections, we observed significant reduction (64%) in vasculature of RAdTIMP-3injected melanomas, as compared with PBS-injected tumors (Fig. 3B). Injection of tumors with RAdTIMP-1 also resulted in 43% reduction in blood vessel number, as compared with RAd66-injected tumors (Fig. 3B). No marked difference in vascularization was detected between RAd66and PBS-injected tumors (Fig. 3B). Characterization of Bystander Effect of Adenovirally Delivered TIMP-3 on Melanoma Cells The observation that adenoviral TIMP-3 gene expression by tumor cells in vivo resulted in potent inhibition of


gelatinase activity and tumor growth of melanoma xenografts regardless of the limited transduction efficiency suggests that TIMP-3 has a potent bystander effect in this model. To examine whether adenovirally delivered TIMP3 exerts a cytotoxic bystander effect on adjacent uninfected cells, we infected A2058 melanoma cells with RAdTIMP-3 at a MOI of 100, then detached and mixed them in different ratios with uninfected cells. Determination of the number of viable cells revealed that in cultures containing 25% RAdTIMP-3 infected A2058 cells, 74% of the cells were killed within 96 hours, and that all melanoma cells were dead in cultures containing 50%, 75%, and 100% RAdTIMP-3-transduced cells (Fig. 4A). Infection of cells with RAdTIMP-1 or RAd66 had no effect on cell viability. To examine whether the cytotoxic bystander effect of adenovirally expressed TIMP-3 requires cell–cell contact, we infected A2058 melanoma cells in parallel with RAdTIMP-3, RAdTIMP-1, and RAdLacZ at a MOI of 100, incubated them for 24, 48, or 72 hours, and determined the levels of TIMP-3 and TIMP-1 in culture medium and cell layers by western blot analysis. Transduction of melanoma cells with RAdTIMP-3 resulted in accumulation of TIMP-3 in the cell layer within 24 hours of infection, and the levels of TIMP-3 in the cell layer remained unaltered until 72 hours (Fig. 4B). Glycosylated and unglycosylated TIMP-3 were also detected in the medium of RAdTIMP-3-infected cells at 24 hours after infection, and further accumulation was detected by 72 hours, apparently because TIMP-3 binding sites in the matrix were

709

ARTICLE


A

C

B

D

FIG. 4. Bystander effect of adenovirally expressed TIMP-3. (A) A2058 Melanoma cells were infected at a MOI of 100 with adenoviruses harboring E. Coli ␤-galactosidase (RAdLacZ), TIMP-1 (RAdTIMP-1), or TIMP-3 (RAdTIMP-3), detached after 10 hours, and mixed with uninfected A2058 cells in different ratios, as indicated, and plated. Cultures were incubated for 96 hours, and cell viability was assessed with MTT assay. (B) A2058 cells were transduced with RAdLacZ, RAdTIMP1, or RAdTIMP-3 at a MOI of 100 and incubated for 24, 48, and 72 hours. The levels of TIMP-3 and TIMP-1 in medium and cell layer were determined by western blot analysis. Glycosylated and unglycosylated forms of TIMP-3 were detected in the medium and cell layer of RAdTIMP-3-infected cells at all time points. (C) A2058 Melanoma cells were infected with RAdTIMP-3- and RAdLacZ (MOI 100)-conditioned medium (CM), harvested at 72 hours, and then transferred to cultures of uninfected A2058 cells. The levels of TIMP-3 in conditioned medium and cell layer were determined by western blot analysis at 12 and 24 hours of incubation. TIMP-3 has entirely relocated in the cell layer in 12 hours, whereas TIMP-1 remains in the medium. (D) A2058 cells were treated for 96 hours in serum-free medium containing 15% and 35% of conditioned medium of uninfected, and RAd66-, RAdTIMP-1-, and RAdTIMP-3-transduced A2058 cell cultures incubated for 72 hours. Cell viability was determined at the 96-hour time point by MTT assay.

saturated (Fig. 4B). No TIMP-3 was detectable in the cell layer of RAdLacZ-infected cells. In comparison, we noted accumulation of TIMP-1 in conditioned medium of cells infected with RAdTIMP-1, whereas the levels of TIMP-1 in the cell layer remained constant until 72 hours (Fig. 4B). The levels of TIMP-1 were not altered in the medium or cell layer of A2058 cultures infected with RAdLacZ. To determine whether soluble TIMP-3 can induce apoptosis in melanoma cells, we treated uninfected A2058 cells with conditioned medium from RAdTIMP-3- and

710

RAdTIMP-1-infected cells. TIMP-3 disappeared entirely from the conditioned medium in 12 hours and accumulated in the cell layer, where it remained at least until 24 hours of incubation (Fig. 4C). No TIMP-3 was detectable in the medium or cell layer of A2058 cultures infected with RAdLacZ. In contrast, TIMP-1 remained in the conditioned medium of RAdTIMP-1-infected cells at least for 24 hours, and we did not detect it in the cell layer (Fig. 4C). Incubation of melanoma cells in culture medium containing 15% of conditioned medium of RAdTIMP-3-infected cells resulted



A

ARTICLE

B

C

FIG. 5. Comparison of antitumor effect of TIMP-3 and p53 in melanoma and squamous-cell carcinoma. A2058 Melanoma cells (1 ⫻ 106) and UT-SCC-7 squamous-cell carcinoma cells (5 ⫻ 106) were implanted s.c. into the backs of SCID/SCID mice, and tumors were allowed to grow to 50–100 mm3 (A2058) and 25–50 mm3 (UT-SCC-7) in size. (A) Growth of A2058 melanoma cell tumors injected once every 24 hours for 3 days with adenoviruses harboring E. coli ␤-galactosidase (RAdLAcZ), p53 (RAdp53), and TIMP-3 (RAdTIMP-3; 1.4 ⫻ 109 pfu each). (B) Growth of UT-SCC-7 xenograft treated as in (A). (C) MTT assay of A2058 cells transduced at a MOI of 100, and UT-SCC-7 cells at a MOI of 500 with RAdLacZ, RAdp53, or RAdTIMP-3, and control (no virus) incubated for 24, 48, 72, and 96 hours. Number of viable cells at each time point was normalized to uninfected control cultures. (D) Hoechst staining of A2058 (left) and UT-SCC-7 (right) 72 hours after infection with RAdLacZ, RAdp53, or RAdTIMP-3 as in (C). Statistical significance determined by Student’s t-test (A) at day 28, RAdTIMP-3 (n = 5) versus RAdLacZ (n = 4), P = 0.026; RAdTIMP-3 (n = 5) versus RAdp53 (n = 5), P = 0.002; RAdLacZ (n = 4) versus RAdp53 (n = 5), not significant. (B) At day 54, RAdTIMP-3 (n = 6) versus RAdLacZ (n = 5), P = 0.004; RAdp53 (n = 6) versus RAdLacZ (n = 5), P = 0.021; RAdp53 (n = 6) versus RAdTIMP-3 (n = 6), P = 0.044.

D

in death of 47% of A2058 cells within 96 hours, and inclusion of 35% of the same medium killed all A2058 cells (Fig. 4D). Incubation of A2058 cells with conditioned medium of uninfected A2058 cells, or with A2058 cell cultures infected with RAd66 or RAdTIMP-1, had no effect on

cell viability. Together these results show that adenovirally expressed TIMP-3 exerts a cytotoxic bystander effect that does not require cell–cell contact but is mediated by soluble TIMP-3, which homes to the cell layer of A2058 cultures.


711

ARTICLE


Comparison of Antitumor Effect of TIMP-3 and p53 in Melanoma and Squamous-Cell Carcinoma Next, we examined the antitumor effect of TIMP-3 in another tumor model, SCC, and in comparison to a well-characterized tumor suppressor gene, p53. There is evidence for the antitumor efficacy of adenovirally delivered p53, which also involves a bystander effect [36,37]. In this context, we compared the tumor suppressor effect of RAdTIMP-3 with RAdp53, an adenovirus harboring wild-type p53, in melanoma and SCC xenografts in SCID/SCID mice. We injected mice s.c. with 1 ⫻ 106 A2058 cells or with 5 ⫻ 106 UT-SCC-7 cells, and tumors were allowed to grow to the size of 50–100 mm3 and 25–50 mm3, respectively. We then injected both types of tumors once every 24 hours for 3 days with 1.4 ⫻ 109 pfu of RAdLacZ, RAdp53, or RAdTIMP-3, and followed the growth of tumors. RAdp53 did not significantly inhibit melanoma growth, whereas RAdTIMP-3 inhibited A2058 xenograft growth significantly (Fig. 5A). In contrast, we observed significant inhibition (71%, P = 0.021) of SCC xenograft growth with RAdp53 19 days after the first injection, as compared with RAdLacZ-injected tumors (Fig. 5B). However, the inhibition of SCC xenograft growth obtained with RAdTIMP-3 was clearly more potent than with RAdp53, because the tumor growth was inhibited by 93% (P = 0.004) compared with RAdLacZ-injected tumors and by 79% (P = 0.044) compared with RAdp53-injected tumors at day 19 after the first injection (Fig. 5B). The estimated transduction efficiency of UT-SCC-7 tumor after a single injection with RadLacZ was 3% (data not shown). To compare the effect of TIMP-3 and p53 on tumor cell viability, we transduced A2058 cells and UT-SCC-7 cells with RAdLacZ, RAdp53, and RAdTIMP-3 at MOIs of 100 and 500, which give 100% transduction efficiency in these cells, respectively. Adenoviral delivery of p53 had no effect on the viability of A2058 cells, determined by MTT assay, whereas adenoviral expression of TIMP-3 killed all A2058 cells in 96 hours after infection (Fig. 5C), in accordance with our previous observations [27]. In addition, using Hoechst staining we detected cells with typical apoptotic nuclei in RAdTIMP-3transduced A2058 cultures 72 hours after transduction, whereas no apoptotic cells were present in RAdp53- and RAdLacZ-transduced A2058 cultures (Fig. 5D, left). In parallel cultures, we noted that significant reduction in UT-SCC7 cell viability had already occurred at 48 hours after transduction with both RAdp53 and RAdTIMP-3 (28%, P = 0.0031 and 34%, P = 0.014, respectively); moreover, at 72 hours nearly all cells in RAdTIMP-3-infected UT-SCC-7 cultures were killed, and in RAdp53-infected cultures 85% of cells were dead (P = 0.01). Hoechst staining detected the presence of apoptotic cells in both RAdTIMP-3- and RAdp53-transduced UT-SCC-7 cultures 72 hours after transduction (Fig. 5D, right).

DISCUSSION In the present study, we have examined the effect of adenovirus-mediated delivery of two natural MMP inhibitors,

712

TIMP-3 and TIMP-1, on the growth of human tumor cells in SCID/SCID mice. Our results show for the first time that adenoviral delivery of the TIMP-3 gene to melanoma and SCC cells ex vivo inhibits completely the implantation and growth of these cells after s.c. injection in SCID mice. In addition, our results show that intratumoral injections of preestablished subcutaneous melanoma and SCC xenografts in SCID mice with TIMP-3 adenovirus result in significant suppression in the growth of these tumors. In general, TIMP-3 possesses several potential antitumor properties, which may all contribute to inhibition of melanoma tumorigenesis and growth in a synergistic way, that is, inhibition of MMP activity, inhibition of cell adhesion to the ECM, inhibition of angiogenesis, and promotion of apoptosis. There is considerable evidence that MMP activity is instrumental for growth and invasion of malignant tumors, including melanoma. Overexpression of MMP-1, MMP-2, MMP-9, MMP-13, and MT1-MMP (MMP-14) has been associated with the invasion capacity of primary melanomas in vivo [3–7]. In the current work, the estimated adenoviral transduction efficiency in vivo after three adenovirus injections was 8–10%, whereas gelatinolytic activity in RAdTIMP-3- and RAdTIMP1-injected tumors was inhibited throughout the tumor, indicating distribution of both TIMPs from the site of production within the tumor tissue. However, adenovirus-mediated expression of TIMP-1 had no marked effect on the growth of preestablished melanomas in vivo, although potent inhibition of gelatinolytic activity was noted. This may be due to its different MMP-inhibitory profile and its inability to induce apoptosis. TIMP-1 does not inhibit the activity of MMP-2 and MT1-MMP, both associated with invasion of melanomas, whereas the activity of both these MMPs is potently inhibited by TIMP-3 [12]. In addition, TIMP-1 has been shown to protect mammary tumor and Burkitt lymphoma cells from apoptosis in vivo [38,39]. Based on these results, it appears that inhibition of MMP activity alone is not sufficient to suppress growth of malignant tumors. Our results also show that adenovirus-mediated expression of TIMP-3 inhibits vascularization of melanomas in vivo. The ability of TIMP-3 to induce apoptosis in vascular smoothmuscle cells may potentiate its ability to inhibit angiogenesis through destabilization of the vascular bed [26,33]. Previous work with stable cell lines expressing TIMP-1, TIMP2, or TIMP-4 have shown significant regression of tumor growth in in vivo animal models [29–31,39]. It is likely that suppression of tumor growth in these studies was caused by inhibition of tumor invasion and angiogenesis, because TIMP1, TIMP-2, and TIMP-4 overexpression had no effect on viability of these cells in culture. These observations are in accordance with the results of our ex vivo experiments, showing that TIMP-1 expression by melanoma cells at the time of injection in subcutaneous tissue can inhibit their implantation and suppress subsequent growth of tumors to some extent. However, we did not obtain a significant suppression of tumor growth of preestablished melanoma xenografts by adenovirus-mediated gene delivery of TIMP-1, although it



also inhibited vascularization of melanoma xenografts. This result may have occurred because the level of TIMP-1 production obtained in vivo by adenoviral gene transfer was not as high as in studies using stably transfected cell lines, in which all cells produce high levels of TIMP-1, and therefore the level of angiogenesis inhibition obtained by adenoviral TIMP-1 delivery may not be sufficient to suppress tumor growth. We have previously noted that adenovirus-mediated gene delivery of TIMP-3 promotes apoptosis in melanoma cells in culture, whereas TIMP-1 and TIMP-2 had no effect on viability of melanoma cells, although all three TIMPs mentioned potently inhibited the invasion of melanoma cells [27]. Stable transfection of colon carcinoma cells with TIMP-3 inhibits their implantation and growth in vivo in nude mice [25]. This experimental model is analogous with our ex vivo approach, which also demonstrated the potent inhibitory effect of TIMP-3 expression on melanoma and SCC cell implantation and growth in vivo (Table 1). However, because high expression of TIMP-3 would be expected to induce apoptosis, it is likely that these stably transfected cell lines either produce limited levels of TIMP-3 or have acquired resistance to TIMP3-induced apoptosis. The results presented here show, for the first time, that adenoviral delivery of TIMP-3 into preestablished melanomas induces apoptosis in these tumor cells in vivo. It has been suggested that TIMP-3-induced apoptosis involves stabilization of the TNF-RI ectodomain [16]. This notion is supported by the recent finding that TIMP-3 inhibits the activity of TACE (ADAM-17), which can shed TNF-RI as well as TNF-␣ from the cell surface [13,15]. In addition, TIMP3 inhibits shedding of ectodomains of receptors and other membrane proteins, which may sensitize cells to apoptotic signals [16–18]. The apoptosis-inducing domain has been recently mapped to the N terminus of TIMP-3, which is essential for inhibition of MMP activity. It was also shown that mutation of a conserved cysteine, which is essential for inhibition of MMP activity, abolished both MMPinhibitory activity and apoptosis-promoting activity of TIMP-3 [40]. We have also previously shown that infection of melanoma cells with RAdTIMP-3 reduces their adhesion to the ECM, which could result in loss of survival signals from ECM and in subsequent apoptosis [27]. The inhibition of cell adhesion by TIMP-3 may also play an important role in inhibiting melanoma tumorigenesis, growth, and invasion in vivo. The observation that adenovirally mediated TIMP-3 gene delivery into melanoma cells in vivo resulted in potent inhibition of gelatinase activity and tumor growth of melanoma xenografts despite limited transduction efficiency suggests a potent bystander effect of TIMP-3 in this model. This notion was further supported by the observation that melanoma cells infected with TIMP-3 adenovirus exerted a cytotoxic effect on uninfected A2058 cells in culture. This cytotoxic effect was not dependent on cell–cell contact but could be transferred to uninfected cultures by conditioned medium containing soluble


ARTICLE

glycosylated and unglycosylated TIMP-3. Soluble TIMP-3 rapidly relocated from medium to cell layer, followed by the death of all melanoma cells. These results show that high expression of TIMP-3 by adenovirally infected cells rapidly saturates their matrix and results in distribution of soluble TIMP-3 to the surrounding tissue. However, it appears that soluble TIMP-3 quickly homes to the A2058 cell layer, formerly devoid of TIMP-3. These observations provide an explanation for the widespread MMP inhibition and apoptosis noted in RAdTIMP-3-injected melanoma xenografts, despite limited transduction efficiency in vivo. Furthermore, these observations suggest that the majority of TIMP-3 produced locally after adenoviral infection remains within the ECM of tumor tissue, maximizing its cytotoxic effect on tumor cells and minimizing its entry to the systemic circulation. Comparison of adenovirally delivered TIMP-3 and p53 showed that p53 overexpression did not inhibit melanoma growth and was not capable of inducing apoptosis in A2058 melanoma cells. This is in accordance with the findings that melanomas rarely contain mutation in the p53 gene and that adenovirally delivered wild-type p53 rarely induces apoptosis in melanoma cells [41]. However, adenoviral expression of p53 potently inhibited the growth of SCC tumors, and this result may also involve a bystander effect [37]. Our results show that adenoviral TIMP-3 expression inhibits SCC tumor growth more potently than p53 adenovirus, most likely because of a more widespread bystander effect of secreted TIMP-3, as compared with the intracellular tumor suppressor p53. Furthermore, our observations show that the tumor-suppressing effect of TIMP-3 is not dependent on the p53 status of the tumor cells. Treatment of primary malignant melanoma involves excision of the tumor, which is curative in most cases. However, the high metastatic tendency of melanoma poses a serious clinical problem due to the limited response of metastatic melanoma to current chemotherapy regimens [42]. Alone or in combination with other therapeutic modalities, cytotoxic gene therapy for malignant melanoma should be based at least in part on systemic delivery of gene therapy vectors so as to reach the majority of the tumor cells and achieve maximal cytoreductive effect. To achieve this goal without severe adverse effects, targeting of viral vectors is obviously essential. This can be achieved by modification of adenoviral vector tropism to infect melanoma cells preferentially and by transcriptional targeting of the expression of the tumor-suppressing transgene to melanoma cells. However, our results suggest that topical adenoviral delivery of TIMP-3 may be useful in gene therapy of local malignant tumors, such as treatment-resistant primary and metastatic head and neck SCCs. Altogether, our results warrant further studies to elucidate the role of TIMP-3, a secreted, antiinvasive, and antiangiogenic tumor suppressor gene product, in therapy of malignant tumors.

713

ARTICLE

MATERIALS


AND

METHODS

Melanoma and squamous-cell carcinoma cultures. Melanoma cell line A2058, established from metastasis of human malignant melanoma, was obtained from the American Type Culture Collection (ATCC, Manassas, VA) and cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% FCS, 2 mM glutamine, 100 IU/ml penicillin G, and 100 ␮g/ml streptomycin. Human SCC cell line UT-SCC-7, which has one mutant p53 allele, was established from metastasis of cutaneous SCC [32] and was cultured in DMEM supplemented with 10% FCS, 2 mM glutamine, 100 IU/ml penicillin G, and 100 ␮g/ml streptomycin and nonessential amino acids. Adenoviral vectors. Recombinant replication-deficient adenovirus RAdLacZ (RAd35) [43], which contains the Escherichia coli ␤-galactosidase (LacZ) gene under the control of the cytomegalovirus immediate-early (CMV IE) promoter, and corresponding empty adenovirus (RAd66) [43] were provided by Gavin W. G. Wilkinson (University of Cardiff, Wales). Construction and characterization of replication-deficient adenoviruses containing the coding region of human TIMP-1 (RAdTIMP-1) or TIMP-3 (RAdTIMP-3) genes driven by the CMV IE promoter have been described [26,44]. Wild-type p53-carrying adenovirus (RAdp53) was provided by Prem Seth (Des Moines University, Des Moines, IA) [45]. The empty adenovirus (RAdpCA3) contains the CMV promoter and polyadenylation signal, but no transgene (S. Leivonen et al., manuscript in preparation]. Virus propagation and titer determination of recombinant adenoviruses were carried out as described [27,44]. Growth of melanoma xenografts in SCID/SCID mice. We carried out all experiments with mice according to institutional animal care guidelines and with permission of the animal test review board of the University of Turku, Finland. We used 4- to 8-week-old severe-combined-immunodeficiency (SCID/SCID) mice in all experiments. Each experimental group contained an equal number of male and female mice. In ex vivo experiments we infected A2058 melanoma cells at a MOI of 100, and UT-SCC-7 cells were infected at a MOI of 500, which gives 100% transduction efficiency in these cells. We then incubated the infected cells for 16 hours, washed them with PBS, and detached them with trypsin [27,32]. Trypsin was neutralized with 10% FCS in DMEM, and cells (5 ⫻ 106/mouse) in 100 ␮l of PBS were injected s.c. into the backs of mice. For in vivo experiments, we established tumors by injecting 1 ⫻ 106 A2058 and 5 ⫻ 106 UT-SCC-7 s.c. into the backs of mice and allowed tumors grow for 14 and 35 days, respectively. We then injected tumors with adenoviruses (1.4 ⫻ 109 pfu) in 100 ␮l PBS and control tumors with 100 ␮l PBS. We measured tumor diameter twice a week by calipers. For X-gal staining, we first fixed tumor pieces (5 mm ⫻ 5 mm) in 4% paraformaldehyde for 30 minutes, washed them three times in PBS, and stained them overnight with 1 mg/ml of 5-bromo-4chloro-3-indoyl-␤-galactopyranoside (X-gal) at room temperature [46]. In vivo transduction efficiency was determined from X-gal-stained sections from two tumors (two sections per tumor) by quantitating the ratio of blue area to total tumor area with Microcomputer Image Device system (MCID; Imaging Research Inc.). Liver, lungs, and kidney of RAdLacZ-injected mice were similarly stained by X-gal. For TUNEL staining, immunohistochemistry, and in situ hybridization, we excised tumors from euthanized animals and immediately fixed them in 4% formaldehyde overnight, incubated them in 40% ethanol for 24 hours and in 70% ethanol for 24 hours, and embedded them in paraffin. For in situ zymography, tumors were mounted into Tissue-Tek and flash-frozen in liquid isopentane. In situ hybridization. We conducted in situ hybridization with the 35Slabeled human TIMP-1 and TIMP-3 antisense RNA probes, as described [47]. After hybridization at 55⬚C, the slides were washed under stringent conditions, including treatment with RNase A. Following 14–40 days of autoradiographic exposure, the photographic emulsion was developed, and slides were stained with hematoxylin and eosin. In situ TUNEL staining. We used ApopTag kit (Oncor Inc., Gaithersburg, MD) to detect fragmented DNA of apoptotic cells in vivo. Briefly, after deparaffinization the tissues were digested with proteinase K (20 ␮g/ml) and endogenous peroxidase was inactivated with 2% hydrogen peroxide. Subsequently, a mixture of digoxigenin-dUTP and terminal deoxynucleotidyl transferase enzyme was added, and the samples were incubated

714

for 1 hour at 37⬚C. This was followed by treatment with anti-digoxigeninperoxidase, staining with diaminobenzidine, and counterstaining with hematoxylin. The number of TUNEL-positive cells was quantitated by measuring the area of brown color/total tumor area of two tumors (one section each) by MCID image analysis, necrotic areas excluded. In situ gelatin zymography. We determined in vivo gelatinolytic activity by gelatin in situ zymography, as described [33]. Briefly, 7-␮m frozen sections (four sections per sample, n = 2 per group) were applied to glass slides and coated with LM-1 photographic emulsion (Amersham International, UK) diluted 1:2 with incubation medium (50 mM Tris-HCl, 50 mM NaCl, 10 mM CaCl2, 0.05% (wt/vol) Brij 35, pH 7.6). After overnight incubation at 37⬚C, slides were developed in the light with Kodak D-19 developer (Kodak, Bridgend, Wales, UK) and fixed using Kodak Unifix solution (Kodak, Bridgend, Wales, UK). Gelatinolytic activity was identified as white areas of lysis on the black background. As controls, sections incubated with incubation buffer supplemented with 20 mmol/liter EDTA or 200 nmol/liter of the MMP inhibitor Ro 31-9790 (Roche Diagnostics Ltd., Welwyn Garden City, UK). Immunohistochemical staining. We immunostained blood vessels using rabbit anti-human von Willebrand factor antibody (A0082; Dako, Carpinteria, CA). The peroxidase/anti-peroxidase technique was applied using diaminobenzidine or aminoethylcarbazole as chromogenic substrates and hematoxylin as counterstain. The samples were pretreated with pepsin and incubated overnight with antibody at 1:300 dilution. The number of von Willebrand-positive blood vessels containing erythrocytes was counted in five adjacent microscopic fields with ⫻10 magnification in the margin of two tumors. Assessment of bystander effect. We transduced A2058 melanoma cells at a MOI of 100 with RAd66, RAdTIMP-1, or RAdTIMP-3, and incubated them in serum-free culture medium for 10 hours. Cells were thereafter detached, washed four times with PBS, suspended in serum-free culture medium, mixed in different ratios with uninfected A2058 melanoma cells, and plated onto 96-well plates. The number of viable cells was determined at different time points by MTT assay, using CellTiter 96 AQueous nonradioactive cell proliferation assay (Promega, Madison, WI) as described [27]. To obtain conditioned medium of melanoma cells, we infected 1.5 ⫻ 106 A2058 cells at a MOI of 100 with RAd66, RAdTIMP-1, and RAdTIMP-3 for 12 hours, washed the infected cells four times with PBS, and incubated them for 72 hours in serum-free culture medium. Conditioned medium was added on confluent A2058 cells grown on 96-well plates mixed with fresh serum-free culture medium in different ratios. Viability of cells was determined after 72 hours and 96 hours incubation with MTT assay. Western blot analysis. We determined the levels of TIMP-3 and TIMP-1 in conditioned medium and cell layer by western blot analysis, as described earlier [27]. Briefly, 1.5 ⫻ 106 melanoma cells were transduced in serumfree conditions with RAd66, RAdTIMP-1, or RAdTIMP-3 at a MOI of 100. Cell layers were washed with PBS and collected in Laemmli sample buffer. Samples of cell layer and conditioned medium were reduced in 5% ␤-mercaptoethanol in Laemmli sample buffer. Aliquots of conditioned medium and cell layer were fractionated electrophoretically on SDS–polyacrylamide (10%) gels, transferred to nitrocellulose filter (Amersham, England), and incubated at 20⬚C with polyclonal antibodies against TIMP-1 or TIMP-3 (Chemicon International Inc., CA) for 1 hour at a dilution of 1:1000. The filters were then incubated with secondary anti-rabbit horseradish peroxidase–conjugated antibodies for 1 hour at room temperature and subjected to ECL reaction (Amersham Corp., UK); positive labeling was then detected with autoradiography. Determination of apoptotic cells. For determination of the effect of RAdp53 and RAdTIMP-3 on cell viability, we seeded 1 ⫻ 104 cells on 96well plates and incubated them for different periods of time after the adenovirus infection. The number of viable cells was determined by MTT assay, as above. Apoptotic cells were visualized by staining with Hoechst-33342 (10 ␮g/ml) and fluorescent microscopy.

ACKNOWLEDGMENTS We thank Hanna Haavisto, Sari Pitkänen, and Marjo Hakkarainen (all from University of Turku) for technical assistance. This study was supported by grants



from the Academy of Finland (grant 45996), Sigrid Jusélius Foundation, the Cancer Foundation of Finland, Turku University Central Hospital (EVO grant 13336), and by research contract with Finnish Life and Pension Insurance Companies. M.A. and R.A. are students in Turku Graduate School in Biomedical Sciences. RECEIVED FOR PUBLICATION SEPTEMBER 5, 2001; ACCEPTED MARCH 27, 2002.

REFERENCES 1. Westermarck, J., and Kähäri, V. M. (1999). Regulation of matrix metalloproteinase expression in tumor invasion. FASEB J. 13: 781–792. 2. Hofmann, U. B., Westphal, J. R., Van Muijen, G. N., and Ruiter, D. J. (2000). Matrix metalloproteinases in human melanoma. J. Invest. Dermatol. 115: 337–344. 3. Woolley, D. E., and Grafton, C. A. (1980). Collagenase immunolocalization studies of cutaneous secondary melanomas. Br. J. Cancer 42: 260–265. 4. Väisänen, A., Tuominen, H., Kallioinen, M., and Turpeenniemi-Hujanen, T. (1996). Matrix metalloproteinase-2 (72 kD type IV collagenase) expression occurs in the early stage of human melanocytic tumour progression and may have prognostic value. J. Pathol. 180: 283–289. 5. Van den Oord, J. J., Paemen, L., Opdenakker, G., and de Wolf-Peeters, C. (1998). Expression of gelatinase B and the extracellular matrix metalloproteinase inducer EMMPRIN in benign and malignant pigment cell lesions of the skin. Am. J. Pathol. 151: 665–670. 6. Airola, K., et al. (1999). Expression of collagenases-1 and -3 and their inhibitors TIMP-1 and -3 correlates with the level of invasion in malignant melanomas. Br. J. Cancer 80: 733–743. 7. Hofmann, U. B., et al. (2000). Expression and activation of matrix metalloproteinase-2 (MMP-2) and its co-localization with membrane-type 1 matrix metalloproteinase (MT1MMP) correlate with melanoma progression. J. Pathol. 191: 245–256. 8. Nikkola, J., et al. (2001). High collagenase-1 expression correlates to favourable chemoimmunotherapy response in human metastatic melanoma. Melanoma Res. 11: 157–166. 9. Pavloff, N., Staskus, P. W., Kishnani, N. S., and Hawkes, S. P. (1992). A new inhibitor of metalloproteinases from chicken: ChIMP-3. A third member of the TIMP family. J. Biol. Chem. 267: 17321–17326. 10. Leco, K. J., Khokha, R., Pavloff, N., Hawkes, S. P., and Edwards, D. R. (1994). Tissue inhibitor of metalloproteinases-3 (TIMP-3) is an extracellular matrix-associated protein with a distinctive pattern of expression in mouse cells and tissues. J. Biol. Chem. 269: 9352–9360. 11. Yu, W. H., Yu, S. S., Meng, Q., Brew, K., and Woessner, J. F., Jr. (2000). TIMP-3 binds to sulfated glycosaminoglycans of the extracellular matrix. J. Biol. Chem. 275: 31226–31232. 12. Apte, S. S., Olsen, B. R., and Murphy, G. (1995). The gene structure of tissue inhibitor of metalloproteinases (TIMP)-3 and its inhibitory activities define the distinct TIMP gene family. J. Biol. Chem. 270: 14313–14318. 13. Amour, A. S., et al. (1998). TNF-␣ converting enzyme (TACE) is inhibited by TIMP-3. FEBS Lett. 435: 39–44. 14. Amour, A., et al. (2000). The in vitro activity of ADAM-10 is inhibited by TIMP-1 and TIMP-3. FEBS Lett. 473: 275–279. 15. Kashiwagi, M., Tortorella, M., Nagase, H., and Brew, K. (2001). TIMP-3 is a potent inhibitor of aggrecanase 1 (ADAM-TS4) and aggrecanase 2 (ADAM-TS5). J. Biol. Chem. 276: 12501–12504. 16. Smith, M. R., Kung, H., Durum, S. K., Colburn, N. H., and Sun, Y. (1997). TIMP-3 induces cell death by stabilizing TNF-␣ receptors on the surface of human colon carcinoma cells. Cytokine 9: 770–780. 17. Fitzgerald, M. L., Wang, Z., Park, P. W., Murphy, G., and Bernfield, M. (2000). Shedding of syndecan-1 and -4 ectodomains is regulated by multiple signaling pathways and mediated by a TIMP-3-sensitive metalloproteinase. J. Cell. Biol. 148: 811–824. 18. Hargreaves, P. G., et al. (1998). Human myeloma cells shed the interleukin-6 receptor: inhibition by tissue inhibitor of metalloproteinase-3 and a hydroxamate-based metalloproteinase inhibitor. Br. J. Haematol. 101: 694–702. 19. Borland, G., Murphy, G., and Ager, A. (1999). Tissue inhibitor of metalloproteinases-3 inhibits shedding of L-selectin from leukocytes. J. Biol. Chem. 274: 2810–2815. 20. Andreu, T., Beckers, T., Thoenes, E., Hilgard, P., and von Melchner, H, (1998). Gene trapping identifies inhibitors of oncogenic transformation. The tissue inhibitor of metalloproteinases-3 (TIMP3) and collagen type I ␣2 (COL1A2) are epidermal growth factorregulated growth repressors. J. Biol. Chem. 273: 13848–13854. 21. Pennie, W. D., Hegamyer, G. A., Young, M. R., and Colburn, N. H. (1999). Specific methylation events contribute to the transcriptional repression of the mouse tissue inhibitor of metalloproteinases-3 gene in neoplastic cells. Cell Growth Differ. 10: 279–286. 22. Loging, W. T., and Reisman, D. (1999). Inhibition of the putative tumor suppressor gene TIMP-3 by tumor-derived p53 mutants and wild type p53. Oncogene 18:


ARTICLE

7608–7615. 23. Bachman, K. E., et al. (1999). Methylation-associated silencing of the tissue inhibitor of metalloproteinase-3 gene suggests a suppressor role in kidney, brain, and other human cancers. Cancer Res. 59: 798–802. 24. Ueki, T., et al. (2000). Hypermethylation of multiple genes in pancreatic adenocarcinoma. Cancer Res. 60: 1835–1839. 25. Bian, J., et al. (1996). Suppression of in vivo tumor growth and induction of suspension cell death by tissue inhibitor of metalloproteinases (TIMP)-3. Carcinogenesis 17: 1805–1811. 26. Baker, A. H., Zaltsman, A. B., George, S. J., and Newby, A. C. (1998). Divergent effects of tissue inhibitor of metalloproteinase-1, -2, or -3 overexpression on rat vascular smooth muscle cell invasion, proliferation, and death in vitro. TIMP-3 promotes apoptosis. J. Clin. Invest. 101: 1478–1487. 27. Ahonen, M., Baker, A. H., and Kähäri, V.-M. (1998). Adenovirus-mediated gene delivery of tissue inhibitor of metalloproteinases-3 inhibits invasion and induces apoptosis in melanoma cells. Cancer Res. 58: 2310–2315. 28. Baker, A. H., George, S. J., Zaltsman, A. B., Murphy, G., and Newby, A. C. (1999). Inhibition of invasion and induction of apoptotic cell death of cancer cell lines by overexpression of TIMP-3. Br. J. Cancer 79: 1347–1355. 29. Khokha, R., Zimmer, M. J., Wilson, S. M., and Chambers, A. F. (1992). Up-regulation of TIMP-1 expression in B16-F10 melanoma cells suppresses their metastatic ability in chick embryo. Clin. Exp. Metastasis 10: 365–370. 30. Imren, S., Kohn, D. B., Shimada, H., Blavier, L., and DeClerck, Y. A. (1996). Overexpression of tissue inhibitor of metalloproteinases-2 retroviral-mediated gene transfer in vivo inhibits tumor growth and invasion. Cancer Res. 56: 2891–2895 31. Wang, M., et al. (1997). Inhibition of tumor growth and metastasis of human breast cancer cells transfected with tissue inhibitor of metalloproteinase 4. Oncogene 14: 2767–2774. 32. Ala-Aho, R., Grénman, R., Seth, P., and Kähäri, V.-M. (2002). Adenoviral delivery of p53 gene suppresses expression of collagenase-3 (MMP-13) in squamous carcinoma cells. Oncogene 21: 1187–1195. 33. George, S. J., Lloyd, C. T., Angelini, G. D., Newby, A. C., and Baker, A. H. (2000). Inhibition of late vein graft neointima formation in human and porcine models by adenovirus-mediated overexpression of tissue inhibitor of metalloproteinase-3. Circulation 101: 296–304. 34. Johnson, M. D., et al. (1994). Inhibition of angiogenesis by tissue inhibitor of metalloproteinase. J. Cell. Physiol. 160: 194–202. 35. Anand-Apte, B., et al. (1997). Inhibition of angiogenesis by tissue inhibitor of metalloproteinase-3. Invest. Ophthalmol. Vis. Sci. 38: 817–823. 36. Clayman, G. L., Frank, D. K., Bruso, P. A., and Goepfert, H. (1999). Adenovirus-mediated wild-type p53 gene transfer as a surgical adjuvant in advanced head and neck cancers. Clin. Cancer Res. 5: 1715–1722. 37. Frank, D. K., Frederick, M. J., Liu, T. J., and Clayman, G. L. (1998). Bystander effect in the adenovirus-mediated wild-type p53 gene therapy model of human squamous cell carcinoma of the head and neck. Clin. Cancer Res. 4: 2521–2528. 38. Li, G., Fridman, R., and Kim, H. R. (1999). Tissue inhibitor of metalloproteinase-1 inhibits apoptosis of human breast epithelial cells. Cancer Res. 59: 6267–6275. 39. Guedez, L., et al. (2001). Tissue inhibitor of metalloproteinase-1 alters the tumorigenicity of Burkitt’s lymphoma via divergent effects on tumor growth and angiogenesis. Am. J. Pathol. 158: 1207–1215. 40. Bond, M., et al. (2000). Localisation of the death domain of TIMP-3 to the N–terminus: metalloproteinase inhibition is associated with pro-apoptotic activity. J. Biol. Chem. 275: 41358–41363. 41. Yamashita, et al. (2001). Induction of apoptosis in melanoma cell lines by p53 and its related proteins. J. Invest. Dermatol. 117: 914–919. 42. McMasters, K. M., Sondak, V. K., Lotze, M. T., and Ross, M. I. (1999). Recent advances in melanoma staging and therapy. Ann. Surg. Oncol. 6: 467–475. 43. Wilkinson, G. W., and Akrigg, A. (1992). Constitutive and enhanced expression from the CMV major IE promoter in a defective adenovirus vector. Nucleic Acids Res. 20: 2233–2239. 44. Baker, A. H., Wilkinson, G. W., Hembry, R. M., Murphy, G., and Newby, A. C. (1996). Development of recombinant adenoviruses that drive high-level expression of the human metalloproteinase-9 and tissue inhibitor of metalloproteinase-1 and -2 genes: characterization of their infection into rabbit smooth muscle cells and human MCF-7 adenocarcinoma cells. Matrix Biol. 15: 383–395. 45. Katayose, D., Wersto, R., Cowan, K. H., and Seth, P. (1995). Effects of a recombinant adenovirus expressing WAF1/Cip1 on cell growth, cell cycle, and apoptosis. Cell Growth Differ. 6: 587–597. 46. Jaakkola, P., Ahonen, M., Kähäri, V.-M., and Jalkanen, M. (2000). Transcriptional targeting of adenoviral gene delivery into migrating wound keratinocytes using FiRE, a growth factor-inducible element. Gene Ther. 7: 1640–1647. 47. Airola, K., et al. (1998). Human TIMP-3 is expressed during fetal development, hair growth cycle, and cancer progression. J. Histochem. Cytochem. 46: 437–448.

715