Tumor Cell-selective Cytotoxicity of Matrix ... - Cancer Research

[CANCER RESEARCH 60, 6061– 6067, November 1, 2000]

Tumor Cell-selective Cytotoxicity of Matrix Metalloproteinase-activated Anthrax Toxin Shihui Liu, Sarah Netzel-Arnett, Henning Birkedal-Hansen, and Stephen H. Leppla1 Oral Infection and Immunity Branch [S. L., S. H. L.] and Matrix Metalloproteinase Unit [S. N-A., H. B-H.], National Institute of Dental and Craniofacial Research, NIH, Bethesda, Maryland 20892

ABSTRACT Matrix metalloproteinases (MMPs) are overexpressed in a variety of tumor tissues and cell lines, and their expression is highly correlated to tumor invasion and metastasis. To exploit these characteristics in the design of tumor cell-selective cytotoxins, we constructed two mutated anthrax toxin protective antigen (PA) proteins in which the furin protease cleavage site is replaced by sequences selectively cleaved by MMPs. These MMP-targeted PA proteins were activated rapidly and selectively on the surface of MMP-overexpressing tumor cells. The activated PA proteins caused internalization of a recombinant cytotoxin, FP59, consisting of anthrax toxin lethal factor residues 1–254 fused to the ADP-ribosylation domain of Pseudomonas exotoxin A. The toxicity of the mutated PA proteins for MMP-overexpressing cells was blocked by hydroxamate inhibitors of MMPs, including BB94, and by a tissue inhibitor of matrix metalloproteinases (TIMP-2). The mutated PA proteins killed MMPoverexpressing tumor cells while sparing nontumorigenic normal cells when these were grown together in a coculture model, indicating that PA activation occurred on the tumor cell surface and not in the supernatant. This method of achieving cell-type specificity is conceptually distinct from, and potentially synergistic with, the more common strategy of retargeting a protein toxin by fusion to a growth factor, cytokine, or antibody.

INTRODUCTION Tumor cell-selective cytotoxins have been created by replacing the receptor-recognition domains of bacterial and plant protein toxins with cytokines, growth factors, and antibodies (1). The protein toxins used contain an enzymatic domain that acts in the cytosol to inhibit protein synthesis and a domain that achieves translocation of this catalyst from a vesicular compartment to the cytosol, as well as the cell-targeting domain that is replaced or altered so as to achieve tumor cell specificity. Certain of these “immunotoxins” derived from diphtheria toxin, Pseudomonas exotoxin A, and ricin have shown efficacy and have been approved for clinical use. However, a recurrent problem with these materials is that therapeutic doses typically damage other tissues and cells (2). This is not surprising because very few of the tumor cell surface receptors or antigens that are targeted are totally absent from normal tissue. Therefore, even in the best cases, some toxin uptake will occur in normal bystander cells. Because these toxins act catalytically, even a small amount of internalized toxin can seriously damage normal tissue. Even a single molecule delivered to the cytosol can kill a cell (3). Several strategies have been used to improve the therapeutic indices of immunotoxins. However, attempts have rarely been made to exploit characteristics unique to the particular toxins. Many protein toxins require proteolytic cleavage to separate the catalytic domain from the receptor-binding and translocation domains (4). Anthrax toxin is unique in that proteolytic cleavage must occur on the cell surface to achieve binding and internalization of a catalytic polypeptide. Anthrax Received 2/18/00; accepted 8/23/00. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1 To whom requests for reprints should be addressed, at Oral Infection and Immunity Branch, National Institute of Dental and Craniofacial Research, NIH, 30 Convent Drive MSC 4350, Building 30, Room 316, Bethesda, MD 20892-4350. Phone: (301) 594-2865; Fax: (301) 402-0396; E-mail: [email protected].

toxin is a three-part toxin secreted by Bacillus anthracis consisting of PA2 (Mr 83,000), LF (Mr 90,000), and EF (Mr 89,000; Refs. 5–7). These three proteins are individually nontoxic. PA binds to an unidentified receptor (8) and is cleaved at the sequence RKKR167 by cell-surface furin or furin-like proteases (9, 10). The COOH-terminal Mr 63,000 fragment (PA63) remains bound to receptor, associates to form a heptamer, and binds LF and EF (6, 11–13). The resulting oligomeric complex is internalized by endocytosis (14), produces a channel in the endosomal membrane, and translocates LF and EF to the cytosol. LF and EF induce cytotoxic events in the cytosol through their respective enzymatic actions. The combination of PA ⫹ LF, named anthrax lethal toxin, lyses mouse macrophages (15, 16), and kills animals (17, 18). These effects result from the proteolytic action of LF on mitogen-activated protein kinase kinases and possibly other cytosolic proteins (19, 20). The combination of PA⫹ EF, named edema toxin, damages cells because of the intracellular adenylate cyclase activity of EF (21). LF and EF have substantial sequence homology at aa 1–250 (7). This region constitutes the PA-binding domain (22). LF aa 1–254 (LFn) are sufficient to achieve translocation of “passenger” polypeptides to the cytosol of cells in a PA-dependent process (23, 24). Thus, an LFn fusion to the ADP-ribosylation domain of Pseudomonas exotoxin A, named FP59, kills any cell type possessing PA receptors (24, 25). FP59 may be used as a potent cytotoxic agent when delivered to the target cells with a target-specific PA. The unique requirement that PA be activated on the target cell surface provides an opportunity to re-engineer this protein to make its activation dependent on eukaryotic cell surface proteases. In particular, it should be possible to exploit the fact that many tumor cells overexpress certain cell surface-associated proteases. MMPs constitute a family of zinc-dependent, multidomain, neutral endopeptidases that play a leading role in both the normal tissue remodeling and pathological destruction of the extracellular matrix (26). Family members include secreted and membrane bound collagenases, stromelysins, gelatinases, and membrane-type metalloproteinases (26). MMPs are overexpressed in a variety of tumor tissues and tumor cell lines and are highly correlated to tumor invasion and metastasis (27). Of the MMPs, MMP-2 (gelatinase A), MMP-9 (gelatinase B), and membrane-type 1 MMP (MT1-MMP) are reported to be most related to invasion and metastasis in various human cancers (27, 28). Current models view MMPs as acting during tumor invasion and metastasis by breaking down tissue extracellular matrix and dissolving epithelial and endothelial basement membranes, thereby enabling tumor cells to invade through stroma and blood vessels. Certain MMPs may also participate in tumor neoangiogenesis because they are selectively up-regulated in proliferating endothelial cells in tumor tissues (29). Furthermore, a group of incompletely characterized MMPs can contribute to the sustained growth of established tumor foci by the ectodomain cleavage of membrane-bound pro-forms of growth fac2 The abbreviations used are: PA, protective antigen; EF, edema factor; LF, lethal factor; aa, amino acid(s); FP59, fusion protein of LF aa 1–254 and Pseudomonas exotoxin A domain III; MMP, matrix metalloproteinase; MTT, 3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide; PA20, NH2-terminal Mr 20,000 fragment of PA; PA63, COOH-terminal Mr 63,000 fragment of PA; TIMP, tissue inhibitor of matrix metalloproteinase.

6061

MMP-ACTIVATED ANTHRAX CYTOTOXINS

tors, releasing peptides that are mitogens for tumor cells and/or tumor vascular endothelial cells (30, 31). The catalytic activities of MMPs are highly regulated. The MMPs are expressed as zymogens, which are activated by proteolysis of an NH2-terminal propeptide that blocks the active site cleft (32, 33). Additionally, a family of proteins, the TIMPs, function as highly effective MMP inhibitors (Ki ⬃ 10⫺10 M; Ref. 34). Through these control mechanisms, the activities of MMPs appear to be restricted to the surface of tumor cells, where they may play a key role in the degradative events associated with invasion and metastasis. Recognition of the role of MMPs in tumor development and metastasis has encouraged the development of synthetic inhibitors of MMPs as potential therapeutic agents (35, 36). Some of these agents are already used in cancer treatment (36). The efficacy of these MMP-directed agents argues that other strategies targeting MMPs may provide selective antitumor agents. In the work described here, we constructed two mutated anthrax PA proteins, PA-L1 and PA-L2, in which the furin recognition site is replaced by sequences susceptible to cleavage by MMP-2 and MMP-9. When combined with FP59, these PA proteins showed selective killing of MMP-overexpressing human tumor cell lines including fibrosarcoma HT1080 cells, breast cancer MDA-MB-231 cells, and melanoma A2058 cells.

In Vitro Cleavage of PA, PA-L1, and PA-L2 by MMP-2, MMP-9, and Furin. Reaction mixtures of 50 ␮l containing 5 ␮g of the PA proteins were incubated at 37°C with 5 ␮l of soluble furin or 0.2 ␮g of active MMP. Furin cleavage was done in 25 mM HEPES (pH 7.4), 150 mM NaCl, 0.2 mM EDTA, 0.2 mM EGTA, 100 ␮g/ml ovalbumin, 1.0 mM CaCl2, and 1.0 mM MgCl2. Cleavage with MMPs was done in 50 mM HEPES (pH 7.5), 10 mM CaCl2, 200 mM NaCl, 0.05% (v/v) Brij-35, and 50 ␮M ZnSO4. Aliquots (5 ␮l) withdrawn at intervals were separated by PAGE using 10 –20% gradient Tris-glycine gel (Novex, San Diego, CA) and electroblotted to a nitrocellulose membrane (Novex). Cleavage was assessed by Western blotting with a rabbit anti-PA antibody. Membranes were blocked with 5% (w/v) nonfat milk, incubated sequentially with rabbit anti-PA polyclonal antibody (#5308) and horseradish peroxidase-conjugated goat antirabbit antibody (sc-2004; Santa Cruz Biotechnology, Inc., Santa Cruz, CA), and developed with TMB Stabilized Substrate (Promega Corp., Madison, WI). Cells and Culture Medium. Vero cells, COS-7 cells, human fibrosarcoma HT1080 cells, human melanoma A2058 cells, and human breast cancer MDAMB-231 cells were obtained from American Type Culture Collection (Manassas, Virginia). All cells were grown in DMEM with 0.45% glucose, 10% FCS, 2 mM glutamine, and 50 ␮g/ml gentamicin. Cells were maintained at 37°C in a 5% CO2 environment. Preparation of Cell Extracts and Conditioned Media for Gelatin Zymography. Cells were grown to 80 –100% confluence in 75-cm2 flasks, washed twice with serum-free DMEM, and lysed for 10 min on ice with 1 ml/flask of 0.5% (v/v) Triton X-100 in 0.1 M Tris-HCl (pH 8.0). Lysates MATERIALS AND METHODS collected by scraping with a rubber policeman were centrifuged at 10,000 rpm for 10 min at 4°C, and the protein concentrations of the supernatants were Reagents. Enzymes for DNA manipulation and modification were pur- determined by the BCA procedure (Pierce, Rockford, IL) and adjusted to 1 chased from New England Biolabs (Beverly, MA). FP59 and a soluble form of mg/ml with lysis buffer. For collection of conditioned media, the cells were furin were prepared in our laboratory as described (37). Activated MMP-2 and incubated for 24 h with 4 ml/flask of serum-free DMEM. The culture superTIMP-2 were gifts from Dr. William Stetler-Stevenson (NIH, Bethesda, MD) natants were harvested, and cellular debris was removed by centrifugation at and active MMP-9 was purchased from Calbiochem (San Diego, CA). MMP 10,000 rpm for 10 min at 4°C. inhibitors BB94 (Batimastat) and BB2516 (Marimastat) were gifts from British Gelatin Zymography. To enrich and assay gelatinases, cell extracts (1 ml) Biotechnology Ltd. (Oxford, United Kingdom), and GM6001 was a gift and amounts of conditioned media derived from the same number of cells were from Dr. Richard E. Galardy (Glycomed, Inc., Alameda, CA) (38). Rabbit incubated at 4°C for 1 h in an end-over-end mixer with 50 ␮l of gelatinanti-PA polyclonal antibody (#5308) was made in our laboratory. Rabbit Sepharose 4B (Amersham Pharmacia Biotech, Piscataway, NJ) equilibrated anti-MT-MMP1 (AB815) was purchased from Chemicon International, Inc. with 50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 5 mM CaCl2, 0.02% (v/v) Tween (Temecula, CA). 20, and 10 mM EDTA. After four washes with 1 ml of equilibration buffer Construction of PA MMP Substrate Proteins. A modified overlap PCR containing 200 mM NaCl, the beads were eluted with 30 ␮l of 4⫻ nonreducing method was used to construct the mutated PA proteins PA-L1 and PA-L2 in SDS sample buffer. The supernatants were loaded onto 10% gelatin zymogram which the furin site is replaced by the gelatinase substrate sequences GPLgels (Novex). After electrophoresis, gels were developed using the buffers and GMLSQ and GPLGLWAQ, respectively. The PA expression plasmid pYS5 procedures specified by Novex. (39) was used as template. We used 5⬘ primer F, AAAGGAGAACGCytotoxicity Assay with MTT. Cells were seeded into 96-well plates at TATATGA (Shine-Dalgarno and start codons are underlined), and the phos⬃25% confluence. The next day, cells were washed twice with serum-free phorylated primer R1, pTGAGTTCGAAGATTTTTGTTTTAATTCTGG, anDMEM to remove residual serum. Serial dilutions of PA, PA-L1, or PA-L2 nealing to the sequence corresponding to P154-S163, to amplify a fragment (0 –1000 ng/ml) combined with FP59 (50 ng/ml) in serum-free DMEM were designated “N.” We used the mutagenic phosphorylated primer H1, pGGACCATTAGGAATGTGGAGTCAAAGTACAAGTGCTGGACCTAC- added to the cells to give a total volume of 200 ␮l/well. In some experiments, GGTTCCAG, encoding the gelatinase substrate GPLGMLSQ and S168-P176, MMP inhibitors were added 30 or 60 min prior to toxin addition. In some and reverse primer R2, ACGTTTATCTCTTATTAAAAT, annealing to the experiments, cells was incubated with the toxins for 6 h, after which the sequence encoding I589-R595, to amplify a mutagenic fragment “M1.” We used medium was replaced with fresh DMEM supplemented with 10% FCS. Other a phosphorylated mutagenic primer H2, pGGACCATTAGGATTATGGGCA- cells were incubated in the serum-free DMEM for the full extent of the toxin treatment, usually 48 h. Cell viability was then assayed by adding 50 ␮l of 2.5 CAAAGTACAAGTGCTGGACCTACGGTTCCAG, encoding gelatinase mg/ml MTT. The cells were incubated with MTT for 45 min at 37°C, the substrate GPLGLWAQ and S168-P176, and reverse primer R2, to amplify a mutagenic fragment “M2.” Primers F and R2 were used to amplify the ligated medium was removed, and the blue pigment produced by viable cells was products of N ⫹ M1, and N ⫹ M2, respectively, resulting in the mutagenized solubilized with 100 ␮l/well of 0.5% (w/v) SDS, 25 mM HCl, in 90% (v/v) isopropanol. The plates were vortexed, and the oxidized MTT was measured as fragments L1 and L2, in which the coding sequence for the furin site (RKKR167) is replaced by gelatinase substrate sequences GPLGMLSQ and A570 using a microplate reader. Cytotoxicity Assay in a Coculture System. A coculture model was deGPLGLWAQ, respectively. The HindIII/PstI digests of L1 and L2 were cloned between the HindIII and PstI sites of pYS5. The resulting expression plasmids signed to mimic the in vivo condition to verify whether PA-L1 and PA-L2 kill were named pYS-PA-L1 and pYS-PA-L2, and their expression products, the MMP-overexpressing tumor cells while not affecting MMP nonexpressing cells. Vero, HT1080, A2058, and MDA-MB-231 cells were cultured in sepaPA mutated proteins, were accordingly named PA-L1 and PA-L2. Expression and Purification of PA, PA-L1, and PA-L2. To express PA, rate chambers of eight-chamber slides (Nalge Nunc International, Naperville, PA-L1, and PA-L2, the expression plasmids pYS5, pYS-PA-L1, and pYS- IL) to 80 –100% confluence. The cells were washed twice with serum-free PA-L2 were transformed into nonvirulent strain B. anthracis UM23C1–1 and DMEM, the chamber partition was removed, and the slide was put into a grown in FA medium (39) with 20 ␮g/ml of kanamycin for 16 h at 37°C. PA culture dish with serum-free medium, so that all of the cells were bathed in the proteins were purified by ammonium sulfate precipitation, followed by chro- same medium. PA, PA-L1, or PA-L2 (300 ng/ml) and FP59 (50 ng/ml) were matography on a MonoQ column (Amersham Pharmacia Biotech, Piscataway, added individually or in combination and cells were exposed for 48 h. Then MTT (0.5 mg/ml) was added for 45 min at 37°C, the partitions were reNJ), as described previously (40). 6062


grown in medium containing 500 ␮g/ml G418 (Life Technologies, Rockville, MD). Cells expressing the MT1-MMP-EGFP fusion protein, named COSgMT1, were sorted from nonexpressing cells by flow cytometry with a FACStar Plus (Becton Dickinson), using excitation at 488 nm. Expression of MT1-MMP-EGFP was assessed by Western blotting using a specific antiserum (AB815, Chemicon International, Temecula, CA).

RESULTS

Fig. 1. Production of mutated PA proteins that can be specifically cleaved by gelatinases. A, schematic representation of MMP substrate PA proteins. The furin cleavage site RKKR was replaced with gelatinase substrate sequences GPLGMLSQ in PA-L1 and GPLGLWAQ in PA-L2. Arrows, sites cleaved by furin and the gelatinases. B–D, Western blot analyses to show cleavage of PA (B), PA-L1 (C), and PA-L2 (D) by MMP-2, MMP-9, and furin. Proteins were incubated with MMP-2, MMP-9, and furin for the times indicated, and samples were analyzed by Western blotting with rabbit polyclonal antibody against PA.

mounted, and the oxidized MTT in each chamber was dissolved to determine the viability of each cell type. Binding and Processing of PA, PA-L1, and PA-L2 by Cultured Cells. Cells were grown in 24-well plates to 80 –100% of confluence and washed twice with serum-free DMEM to remove residual serum. Then the cells were incubated with 1000 ng/ml of PA, PA-L1, or PA-L2 at 37°C in serum-free DMEM for different lengths of time. When TIMP-2 was tested, it was incubated with cells for 1 h before the PA proteins were added. The cells were washed three times to remove unbound PA proteins. Cells were lysed in 100 ␮l/well modified RIPA lysis buffer [50 mM Tris-HCl (pH 7.4), 1% NP40, 0.25% sodium deoxycholate, 150 mM NaCl, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, and 1 ␮g/ml each of aprotinin, leupeptin, and pepstatin] on ice for 10 min. Equal amounts of protein from cell lysates were separated by PAGE using 10 –20% gradient Tris-glycine gels (Novex). Western blotting to detect PA and its cleavage products was performed as described above. Construction and Transfection of MT1-MMP cDNA into COS-7 Cells. A cDNA encoding human MT1-MMP was a generous gift of Dr. Motoharu Seiki (University of Tokyo, Tokyo, Japan). The MT1-MMP coding sequence was isolated by digestion with TthIII, the ends were filled in with Pfu, and the fragment was inserted into the SmaI site of the mammalian expression vector pEGFPN1 (Clontech Laboratories, Palo Alto, CA). The protein expressed from this plasmid contains MT1-MMP at the NH2 terminus of EGFP (red shifted variant of green fluorescent protein). COS-7 cells (2 ⫻ 105 per dish) were transfected with the expression vector (2 ␮g) using 10 ␮g of SuperFect (Qiagen, Valencia, CA). Cells were incubated for 3 h with the DNA-SuperFect complex in the presence of serum and antibiotic-containing medium, washed, and grown in fresh serum-containing medium for 48 h. Thereafter, cells were

Generation of Mutated PA Proteins That Can Be Cleaved by MMPs. The crystal structure of PA shows that the furin cleavage site, RKKR167, is in a surface-exposed, flexible loop composed of aa 162–175 (41). Cleavage in this loop by furin-like proteases is essential to toxicity. We constructed mutated PA proteins in which the furinsensitive sequence is replaced by the MMP-sensitive sequences GPLGMLSQ and GPLGLWAQ, designated L1 and L2, respectively (Fig. 1A). These two sequences were shown to be favored gelatinase substrates by examining the rates of hydrolysis of ⬎50 synthetic oligopeptides (42, 43). The octapeptide sequences selected are excellent substrates for MMP-2 and MMP-9 but are also susceptible to other MMP species (42, 43), including MT1-MMP (44). These two sequences were modeled after the MMP cleavage site in the ␣1 chain of type I collagen and contain substitutions that enhanced cleavage by MMP-2 and MMP-9 (L1) or by multiple members of the MMP family (L2). Plasmids encoding the mutated proteins PA-L1 and PA-L2 were constructed by a modified overlap PCR method, cloned into the Escherichia coli-Bacillus shuttle vector pYS5, and efficiently expressed in B. anthracis UM23C1-1. The expression products were secreted into the culture supernatants at 20 –50 mg/l. The mutated PA proteins were purified by ammonium sulfate precipitation, followed by MonoQ chromatography. The purified, mutated PA proteins PA-L1 and PA-L2 comigrated with PA in SDS-PAGE but migrated faster than PA during nondenaturing PAGE (data not shown), because the four positively charged residues RKKR of the furin site were replaced by noncharged amino acids. To verify that the mutated proteins had the expected susceptibility to proteases, they were subjected to cleavage with a soluble form of furin and active forms of MMP-2 and MMP-9. As expected, PA was very sensitive to furin but completely resistant to MMP-2 and MMP-9 (Fig. 1B). In contrast, PA-L1 and PA-L2 were completely resistant to furin but were efficiently processed into the two expected fragments, PA63 and PA20, by MMP-2 and MMP-9 (Fig. 1, C and D). Thus, the mutated PA proteins had the desired susceptibilities to the gelatinases. PA-L1 and PA-L2 Selectively Kill MMP-overexpressing Tumor Cells. To test the hypothesis that PA-L1 and PA-L2 would preferentially kill MMP-overexpressing cells, cytotoxicity assays were done with three human tumor cell lines: fibrosarcoma HT1080, melanoma

Fig. 2. Zymographic analysis of gelatinase production. Triton X-100 cell extracts (left panel) and serum-free conditioned medium (right panel) of Vero, HT1080, A2058, and MDA-MB-231 cultures were analyzed. Cell extract protein (1 mg) or volumes of conditioned media (3– 4 ml) corresponding to 1 mg of cell extract were analyzed by gelatin zymography as described in “Materials and Methods.” Parallel gels (not shown) containing molecular weight markers identified the two bands as MMP-9 and MMP-2, as indicated.

6063


Table 1 Toxicities of wild-type and mutated PA proteins for cultured cells Values in the table are EC50 in ng/ml, where EC50 is the concentration of toxin required to kill half of the cells. EC50s are interpolated from Figs. 3, 4, and 8. Values are for 48-h toxin treatments, except that values in parentheses are for 6-h toxin treatments. All incubations included 50 ng/ml FP59. Nicking of PA-L1 and L-2 (last two lines) was by MMP-2.

PA PA-L1 PA-L2 Nicked PA-L1 Nicked PA-L2

Vero

HT1080

A2058

MDA-MB-231

COS-7

COSgMT1

5 (6) ⬎1000 (⬎1000) ⬎1000 (⬎1000) (20) (20)

2.5 (5.5) 2 (10) 2 (10)

2 (6) 4 (20) 7 (25)

1 (2) 3 (15) 4 (30)

6 (15) ⬎1000 (⬎1000) ⬎1000 (⬎1000)

20 (30) 20 (40) 20 (20)

A2058, and breast cancer MDA-MB-231. A nontumor monkey cell line, Vero, was used as control. Gelatin zymography of cell lysates showed that HT1080 expressed both gelatinases, A2058 expressed MMP-2 but not MMP-9, and MDA-MB-231 expressed MMP-9 but not MMP-2 (Fig. 2). Conditioned serum-free medium gave qualitatively similar results, indicating that the gelatinases expressed by these three tumor cell lines were associated with the cell surface and partially secreted into the medium (Fig. 2). In contrast, Vero cells had very low expression of gelatinases (Fig. 2). Cytotoxicity of PA and the mutated PA proteins to these cells was measured in 96-well plates. PA, PA-L1, and PA-L2 combined with FP59 were incubated with cells for 6 or 48 h, and viability in both cases was measured after 48 h. The EC50s (concentrations needed to kill half of the cells) for PA and the mutated PA proteins are summarized in Table 1. The gelatinase-nonexpressing Vero cells were resistant to PA-L1 and PA-L2 but sensitive to PA in a dose-dependent manner (Fig. 3A). However, PA-L1 and PA-L2 that were first nicked by MMP-2 in vitro efficiently killed Vero cells, even with the 6-h toxin challenge (Fig. 3B). This demonstrates that the resistance of Vero cells to PA-L1 and PA-L2 was attributable to the inability of the cells to proteolytically activate the mutated PA proteins. In contrast to Vero cells, the three gelatinase-expressing tumor cells, HT1080, A2058, and MDA-MB-231, were quite susceptible to PA as well as to PA-L1 and PA-L2 (Fig. 4). The selective cytotoxicity of PA-L1 and PA-L2 for the tumor cells was retained when the experiments were repeated in medium containing FCS (data not shown). This indicates that serum proteases do not activate the PA proteins, nor do serum protease inhibitors block proteolytic cleavage by the cell surface proteases. To simplify further analyses, all subsequent experiments were performed in serum-free medium. Binding and processing of PA, PA-L1, and PA-L2 on the surface of cells were also assessed. Vero and HT1080 cells were incubated with PA, PA-L1, and PA-L2 for various times, and cell lysates were examined by Western blotting to detect processing of the PA proteins

Fig. 3. Cytotoxicity of mutated PA proteins for Vero cells. Vero cells grown in serum-free DMEM medium were treated with intact or MMP-2-treated PA, PA-L1, and PA-L2, together with 50 ng/ml FP59. The toxins were incubated with cells for 48 h or were removed after 6 h and replaced with fresh, serum-containing DMEM. MTT was added to determine cell viability at 48 h. Nicked PA-L1 and PA-L2 were prepared by cleavage of PA-L1 and PA-L2 by active MMP-2 at 37°C for 3 h. A, effect of PA proteins after 6 or 48 h exposure. B, effect of intact and MMP-nicked toxins after 6 h of exposure. The analyses were performed two additional times with results similar to those presented. Bars, SE.

to the active PA63 species. PA, PA-L1, and PA-L2 could be detected in the Vero and HT1080 cell lysates as soon as 10 min after incubation, showing rapid binding to the cell surface (Fig. 5, A and B). PA was processed by both cell types. In contrast, PA-L1 and PA-L2 were processed by the MMP-overexpressing HT1080 cells but not by the Vero cells (Fig. 5, A and B). This result was consistent with the previous evidence that PA-L1 and PA-L2 were processed only by MMPs (Fig. 1, B and C) and selectively killed MMP-expressing tumor cells (Figs. 3 and 4). Although HT1080 cells processed PA-L1 and PA-L2 along with PA, the latter was cleaved somewhat more rapidly (Fig. 5B). To verify that the cleavage observed on the surface of HT1080 cells was attributable to MMPs, we added the specific inhibitor, TIMP-2 (Fig. 5C). The inhibitor had no effect on cleavage of PA (by furin) but strongly decreased cleavage of both PA-L1 and PA-L2. We also analyzed the processing of PA-L1 and PA-L2 in culture supernatants of HT1080 cells, and could not detect the active PA63 species in overnight culture supernatants (data not shown). To further demonstrate that the cytotoxicity of the mutated PA proteins for tumor cells was dependent on MMP activity, we characterized the effects of several well-characterized MMP inhibitors, BB94 (Batimastat), BB2516 (Marimastat), and GM6001. All three MMP inhibitors, and especially GM6001, conferred strong protection to HT1080 cells against challenge with PA-L1 and PA-L2 plus FP59 but did not protect the cells against PA plus FP59 (Fig. 6). In addition, the highly specific physiological MMP inhibitor, TIMP-2, also protected cells. Thus, the killing the tumor cells by PA-L1 and PA-L2 is highly dependent on the MMP activity expressed by the tumor cells. PA-L1 and PA-L2 Retain Selectivity for MMP-overexpressing Cells in a Coculture Model. We designed a coculture model to mimic in vivo conditions to test whether PA-L1 and PA-L2 can selectively kill gelatinase-expressing tumor cells but not bystander cells. Vero, HT1080, MDA-MB-231, and A2058 cells were cultured in separate compartments of eight-chamber slides. When the cells reached confluence, the chamber partitions were removed, and the slides were put into culture dishes with serum-free medium so that all cells on the slide were bathed in the same medium. PA, PA-L1, or PA-L2 (each at 300 ng/ml) plus FP59 (50 ng/ml) or FP59 alone were added to the culture dishes and incubated for 48 h before measuring viability. PA killed all cells, whereas PA-L1 and PA-L2 killed only the HT1080, MDA-MB-231, and A2058 cells while not affecting the gelatinase nonexpressing Vero cells (Fig. 7). This result shows that PA-L1 and PA-L2 are not activated in the tissue culture medium by secreted proteases, nor do PA proteins proteolytically activated on the surface of one cell dissociate and rebind on other cells. Activated MMPs in the culture supernatant would have led to killing of the Vero cells, because Fig. 3 shows that PA-L1 and PA-L2 cleaved in solution become cytotoxic. MT1-MMP Plays a Role in Activation of PA-L1 and PA-L2. Northern blot analyses have shown that the three tumor cells used in these studies express MT1-MMP in addition to gelatinases (data not shown), suggesting that MT1-MMP may contribute to cleavage of the mutated PA proteins. To examine the potential role of MT1-MMP, we used COS-7 cells, after showing by zymography that they express

6064


quences susceptible to cleavage by MMPs. When combined with FP59, derived from Pseudomonas exotoxin A, these two mutated PA proteins selectively killed MMP-overexpressing human tumor cells. The potency of the mutated PA proteins for the tumor cells closely approached that of native PA, consistent with the demonstration that cleavage of PA-L1 and PA-L2 on cell surfaces was almost as rapid as cleavage of native PA. This result is somewhat surprising, because it might have been expected that steric constraints or localization to differing membrane domains would limit contact between receptorbound PA and cell-surface MMPs molecules. Cytotoxicity assays in a coculture model, where cells were equally accessible to the toxins in the supernatant, showed that PA-L1 and PA-L2 killed only the three MMP-overexpressing tumor cells and not bystander Vero cells. This result demonstrated that activation of PA-L1 and PA-L2 occurs principally on cell surfaces. Thus, the

Fig. 4. Cytotoxicity of PA proteins for HT1080 cells (A), A2058 cells (B), and MDA-MB-231 cells (C). Cells were cultured and treated with toxin for 6 or 48 h as in Fig. 3. The analyses were performed two additional times with results similar to those presented. Bars, SE.

negligible amounts of gelatinases (Fig. 8A, inset). Consistent with the lack of gelatinases, COS-7 cells were resistant to PA-L1 and PA-L2 plus FP59 but susceptible to PA plus FP59 (Fig. 8A). A plasmid expressing human MT1-MMP-EGFP was transfected into COS-7 cells, yielding a stable transfectant, COSgMT1, in which expression of MT1-MMP-EGFP was verified by Western blotting (Fig. 8B, inset). In contrast to COS-7 cells, COSgMT1 were very sensitive to PA-L1 and PA-L2 (Fig. 8B), indicating that MT1-MMP can cause activation of these mutated PA proteins, either by directly cleaving the cell-bound PA proteins or in an indirect way by activation of proMMP-2 or other MMP zymogens, which in turn cleaved the PA proteins. In support of the former explanation, preliminary experiments have shown that purified MT1-MMP can cleave the mutated PA proteins (data not shown).

Fig. 5. Binding and processing of PA proteins on the surface of Vero and HT1080 cells. Vero cells (A) and HT1080 cells (B and C) were cultured in 24-well plates to 80 –100% of confluence, washed, and incubated in serum-free medium. In C, TIMP-2 was added at 5 ␮g/ml for 1 h prior to toxin addition. The PA proteins were added to the cells at a final concentration of 1000 ng/ml and incubated for the indicated times, and lysates were prepared for Western blotting analysis with rabbit anti-PA polyclonal antibody.

DISCUSSION The close association between MMP overexpression and tumor metastasis has been known for several decades. Recognition that MMPs contribute to tumor development and metastasis has led to the development of novel therapies using synthetic inhibitors of MMPs (35, 36). However, it is expected that these inhibitors will only slow the growth of tumors, without having a direct cytotoxic action that could eradicate the malignant cells. The present study is the first attempt to exploit the localization of MMPs to achieve cell-type selective targeting of cytotoxic bacterial toxin fusion proteins. In this study, we constructed two mutated anthrax toxin PA proteins, PA-L1 and PA-L2, in which the furin recognition site is replaced by se-

Fig. 6. MMP inhibitors protect HT1080 cells from MMP-targeted PA proteins. HT1080 cells were cultured to 80% of confluence in a 96-well plate and washed twice with serum-free DMEM. MMP inhibitors GM6001, BB94, and BB2516 were added to the cells at a final concentration of 10 ␮M in serum-free DMEM. TIMP-2 was used at a final concentration of 10 ␮g/ml. After 30 min preincubation with the MMP inhibitors, 300 ng/ml of PA, PA-L1, or PA-L2 were added combined with 50 ng/ml FP59. After 6 h, the medium containing the toxins and MMP inhibitors was removed, and fresh serumcontaining medium was added, and incubation continued to 48 h, at which time MTT was added to determine cell viability. The analysis was performed one additional time with results similar to those presented.

6065


Fig. 7. PA-L1 and PA-L2 selectively kill MMP-overexpressing tumor cells in a coculture model. Vero, HT1080, A2058, and MDA-MB-231 cells were cultured in the separate chambers of eight-chamber slides to 80 –100% confluence. Then the partitions were removed, and the slides were placed in 100-mm culture dishes with serum-free medium, so that the different cells were in the same culture environment. PA, PA-L1, or PA-L2 (300 ng/ml) each combined with FP59 (50 ng/ml) were separately added to the cells and incubated 48 h. MTT was added to determine cell viability. The slides were photographed (inset), the partitions were replaced, and the insoluble blue dye in each chamber was dissolved and measured in a microplate reader (bar graph). The analysis was performed one additional time, with results similar to those presented.

and/or modulated on cell surfaces, MMPs constitute specific markers for tumor tissues and provide a means for targeting these cells. It was originally thought that the only role of MMPs in cancer was to break down tissue barriers to promote tumor invasion and metastasis. It is now understood that MMPs also participate in tumor neoangiogenesis and are selectively up-regulated in proliferating endothelial cells (29). Therefore, the modified bacterial toxins described here may have the additional therapeutic effect of targeting the dividing vascular endothelial cells that are essential to neoangiogenesis in tumor tissues. Therefore, the MMP-targeted toxins may also kill tumor cells by starving the cells of necessary nutrients and oxygen. Previous efforts to develop anthrax toxin fusion proteins as therapeutic agents have focused on modification of domain 4, the receptorbinding domain of PA. Work is ongoing to create cell type-specific cytotoxic agents by modifying or replacing domain 4 to direct PA to alternate receptors (40, 49). This work follows the example of the development of immunotoxins from other protein toxins, as cited earlier (1). We suggest that combining two conceptually distinct targeting strategies in a single PA protein will yield agents having higher therapeutic indices. A protein that is both retargeted to a tumor cell surface protein and dependent on MMPs for activation may achieve therapeutic effects while being free of the side effects observed with many of the existing immunotoxins.

cytotoxicity of PA-L1 and PA-L2 is directed selectively to the MMPoverexpressing tumor cells, suggesting that these cytotoxins may retain cell-type specificity if used in vivo. These results encourage the further testing of these materials in animal models. Several factors may explain the cell-type selectivity obtained with the mutated PA proteins in the coculture system described above. PA proteins bind to cells rapidly and with high affinity (Kd, ⬃1 nM). Therefore, even at low PA concentrations, PA receptors will be highly occupied. As a result, any PA that becomes activated in the supernatant or dissociates from a cell after cleavage would be unable to locate a free receptor by which to bind to cells and internalize FP59. The bystander normal cells would have their receptors occupied by the uncleavable mutated PA proteins. These nonproductively occupied receptors may be cleared from the cell surface of the normal cells, further reducing the possibility that the cell could be targeted by a PA protein that becomes activated in either the surrounding milieu or on the surface of an MMP-expressing cell. Another factor that may explain the selectivity observed in the coculture model is that the enzymatic activities of MMPs appear to be enhanced through their localization on cell membranes. This focuses their activities on extracellular matrix substrates and renders them more resistant to the proteinase inhibitors present in the extracellular milieu. Recently, it has been shown that physiological concentrations of plasmin can activate both MMP-2 and MMP-9 on the surface of HT1080 cells by a mechanism independent of MMP or acid proteinase activities (45). In contrast, in the soluble phase, plasmin degrades both MMP-2 and MMP-9 (45). Thus, plasmin may provide a mechanism for restricting gelatinase activities to the cell surface. The recently identified, membrane-anchored MMP family member, MT1-MMP, also serves to localize MMP activity to the cell surface. It functions both as an MMP activator and as a receptor for MMP-2 but has no effect on MMP-9 (46). A MMP-2/TIMP-2 complex binds to MT1-MMP on the cell surface and then is proteolytically activated by an adjacent MT1-MMP. Recent work has shown that adhesion receptors, such as ␣v␤3 integrin (47) and the cell surface hyaluronan receptor CD44 (48), may act to retain soluble active MMP-2 or MMP-9 on invasive tumor cell surfaces, where their proteolytic activities are more likely to promote cell invasion. Because MMP activities involved in tumor invasion and metastasis are localized

Fig. 8. Expression of MT1-MMP makes COS-7 cells sensitive to MMP-targeted PA proteins. A, cytotoxicity of PA-L1 and PA-L2 to COS-7 cells. COS-7 cells were cultured to 80 –100% of confluence, washed, and incubated in serum-free DMEM medium. PA proteins combined with 50 ng/ml FP59 were added to the cells and incubated for 6 or 48 h. MTT was added to determine cell viability at 48 h. Inset, zymographic analysis of cell extracts and culture supernatants of COS-7 cells, compared with a lysate of HT1080 as control. B, cytotoxicity of PA-L1 and PA-L2 to COSgMT1. COSgMT1 cells were treated as in A. Inset, expression of MT1-MMP-EGFP from COSgMT1 cells as measured by Western blotting using a rabbit anti-MT1-MMP antibody (AB815; Chemicon International, Inc.). The analyses were performed two additional times with results similar to those presented.

6066


ACKNOWLEDGMENTS We thank Guangqing Tang for helpful discussions and Dana Hsu for assistance in protein purification.

REFERENCES 1. Kreitman, R. J. Immunotoxins in cancer therapy. Curr. Opin. Immunol., 11: 570 –578, 1999. 2. Frankel, A. E., Tagge, E. P., and Willingham, M. C. Clinical trials of targeted toxins. Semin. Cancer Biol., 6: 307–317, 1995. 3. Yamaizumi, M., Mekada, E., Uchida, T., and Okada, Y. One molecule of diphtheria toxin fragment A introduced into a cell can kill the cell. Cell, 15: 245–250, 1978. 4. Gordon, V. M., Klimpel, K. R., Arora, N., Henderson, M. A., and Leppla, S. H. Proteolytic activation of bacterial toxins by eukaryotic cells is performed by furin and by additional cellular proteases. Infect. Immun., 63: 82– 87, 1995. 5. Smith, H., and Stanley, J. L. Purification of the third factor of anthrax toxin. J. Gen. Microbiol., 29: 517–521, 1962. 6. Leppla, S. H. The bifactorial Bacillus anthracis lethal and oedema toxins. In: J. E. Alouf and J. H. Freer (eds.), Comprehensive Sourcebook of Bacterial Protein Toxins, pp. 243–263. London: Academic Press, 1999. 7. Leppla, S. H. Anthrax Toxins. In: J. Moss, B. Iglewski, M. Vaughan, and A. Tu (eds.), Bacterial Toxins and Virulence Factors in Disease. Handbook of Natural Toxins, Vol. 8, Ed. 1, pp. 543–572. New York: Marcel Dekker, Inc., 1995. 8. Escuyer, V., and Collier, R. J. Anthrax protective antigen interacts with a specific receptor on the surface of CHO-K1 cells. Infect. Immun., 59: 3381–3386, 1991. 9. Klimpel, K. R., Molloy, S. S., Thomas, G., and Leppla, S. H. Anthrax toxin protective antigen is activated by a cell-surface protease with the sequence specificity and catalytic properties of furin. Proc. Natl. Acad. Sci. USA, 89: 10277–10281, 1992. 10. Molloy, S. S., Bresnahan, P. A., Leppla, S. H., Klimpel, K. R., and Thomas, G. Human furin is a calcium-dependent serine endoprotease that recognizes the sequence Arg-X-X-Arg and efficiently cleaves anthrax toxin protective antigen. J. Biol. Chem., 267: 16396 –16402, 1992. 11. Leppla, S. H., Friedlander, A. M., and Cora, E. Proteolytic activation of anthrax toxin bound to cellular receptors. In: F. Fehrenbach, J. E. Alouf, P. Falmagne, W. Goebel, J. Jeljaszewicz, D. Jurgen, and R. Rappuoli (eds.), Bacterial Protein Toxins, Ed. 1, pp. 111–112. New York: Gustav Fischer, 1988. 12. Milne, J. C., Furlong, D., Hanna, P. C., Wall, J. S., and Collier, R. J. Anthrax protective antigen forms oligomers during intoxication of mammalian cells. J. Biol. Chem., 269: 20607–20612, 1994. 13. Benson, E. L., Huynh, P. D., Finkelstein, A., and Collier, R. J. Identification of residues lining the anthrax protective antigen channel. Biochemistry, 37: 3941–3948, 1998. 14. Gordon, V. M., Leppla, S. H., and Hewlett, E. L. Inhibitors of receptor-mediated endocytosis block the entry of Bacillus anthracis adenylate cyclase toxin but not that of Bordetella pertussis adenylate cyclase toxin. Infect. Immun., 56: 1066 –1069, 1988. 15. Friedlander, A. M. Macrophages are sensitive to anthrax lethal toxin through an acid-dependent process. J. Biol. Chem., 261: 7123–7126, 1986. 16. Hanna, P. C., Kochi, S., and Collier, R. J. Biochemical and physiological changes induced by anthrax lethal toxin in J774 macrophage-like cells. Mol. Biol. Cell, 3: 1269 –1277, 1992. 17. Beall, F. A., Taylor, M. J., and Thorne, C. B. Rapid lethal effect in rats of a third component found upon fractionating the toxin of Bacillus anthracis. J. Bacteriol., 83: 1274 –1280, 1962. 18. Ezzell, J. W., Ivins, B. E., and Leppla, S. H. Immunoelectrophoretic analysis, toxicity, and kinetics of in vitro production of the protective antigen and lethal factor components of Bacillus anthracis toxin. Infect. Immun., 45: 761–767, 1984. 19. Duesbery, N. S., Webb, C. P., Leppla, S. H., Gordon, V. M., Klimpel, K. R., Copeland, T. D., Ahn, N. G., Oskarsson, M. K., Fukasawa, K., Paull, K. D., and Vande Woude, G. F. Proteolytic inactivation of MAP-kinase-kinase by anthrax lethal factor [see comments]. Science (Washington DC), 280: 734 –737, 1998. 20. Vitale, G., Pellizzari, R., Recchi, C., Napolitani, G., Mock, M., and Montecucco, C. Anthrax lethal factor cleaves the N-terminus of MAPKKs and induces tyrosine/ threonine phosphorylation of MAPKs in cultured macrophages. Biochem. Biophys. Res. Commun., 248: 706 –711, 1998. 21. Leppla, S. H. Anthrax toxin edema factor: a bacterial adenylate cyclase that increases cyclic AMP concentrations of eukaryotic cells. Proc. Natl. Acad. Sci. USA, 79: 3162–3166, 1982. 22. Quinn, C. P., Singh, Y., Klimpel, K. R., and Leppla, S. H. Functional mapping of anthrax toxin lethal factor by in-frame insertion mutagenesis. J. Biol. Chem., 266: 20124 –20130, 1991. 23. Arora, N., Klimpel, K. R., Singh, Y., and Leppla, S. H. Fusions of anthrax toxin lethal factor to the ADP-ribosylation domain of Pseudomonas exotoxin A are potent cytotoxins which are translocated to the cytosol of mammalian cells. J. Biol. Chem., 267: 15542–15548, 1992. 24. Arora, N., and Leppla, S. H. Residues 1–254 of anthrax toxin lethal factor are sufficient to cause cellular uptake of fused polypeptides. J. Biol. Chem., 268: 3334 –3341, 1993.

25. Milne, J. C., Blanke, S. R., Hanna, P. C., and Collier, R. J. Protective antigen-binding domain of anthrax lethal factor mediates translocation of a heterologous protein fused to its amino- or carboxy-terminus. Mol. Microbiol., 15: 661– 666, 1995. 26. Birkedal-Hansen, H. Proteolytic remodeling of extracellular matrix. Curr. Opin. Cell Biol., 7: 728 –735, 1995. 27. Stetler-Stevenson, W. G., Aznavoorian, S., and Liotta, L. A. Tumor cell interactions with the extracellular matrix during invasion and metastasis. Annu. Rev. Cell Biol., 9: 541–573, 1993. 28. Jones, L., Ghaneh, P., Humphreys, M., and Neoptolemos, J. P. The matrix metalloproteinases and their inhibitors in the treatment of pancreatic cancer. Ann. NY Acad. Sci., 880: 288 –307, 1999. 29. Chambers, A. F., and Matrisian, L. M. Changing views of the role of matrix metalloproteinases in metastasis. J. Natl. Cancer Inst., 89: 1260 –1270, 1997. 30. Arribas, J., Coodly, L., Vollmer, P., Kishimoto, T. K., Rose-John, S., and Massague´, J. Diverse cell surface protein ectodomains are shed by a system sensitive to metalloprotease inhibitors. J. Biol. Chem., 271: 11376 –11382, 1996. 31. Merlos-Suarez, A., Fernandez-Larrea, J., Reddy, P., Baselga, J., and Arribas, J. Pro-tumor necrosis factor-␣ processing activity is tightly controlled by a component that does not affect notch processing. J. Biol. Chem., 273: 24955–24962, 1998. 32. Springman, E. B., Angleton, E. L., Birkedal-Hansen, H., and Van Wart, H. E. Multiple modes of activation of latent human fibroblast collagenase: evidence for the role of a Cys73 active-site zinc complex in latency and a “cysteine switch” mechanism for activation. Proc. Natl. Acad. Sci. USA, 87: 364 –368, 1990. 33. Murphy, G., Stanton, H., Cowell, S., Butler, G., Knauper, V., Atkinson, S., and Gavrilovic, J. Mechanisms for pro matrix metalloproteinase activation. Acta Pathol. Microbiol. Immunol. Scand., 107: 38 – 44, 1999. 34. Birkedal-Hansen, H., Moore, W. G., Bodden, M. K., Windsor, L. J., Birkedal-Hansen, B., DeCarlo, A., and Engler, J. A. Matrix metalloproteinases: a review. Crit Rev. Oral Biol. Med., 4: 197–250, 1993. 35. Brown, P. D. Matrix metalloproteinase inhibitors: a novel class of anticancer agents. Adv. Enzyme Regul., 35: 293–301, 1995. 36. Drummond, A. H., Beckett, P., Brown, P. D., Bone, E. A., Davidson, A. H., Galloway, W. A., Gearing, A. J., Huxley, P., Laber, D., McCourt, M., Whittaker, M., Wood, L. M., and Wright, A. Preclinical and clinical studies of MMP inhibitors in cancer. Ann. NY Acad. Sci., 878: 228 –235, 1999. 37. Gordon, V. M., Benz, R., Fujii, K., Leppla, S. H., and Tweten, R. K. Clostridium septicum ␣-toxin is proteolytically activated by furin. Infect. Immun., 65: 4130 – 4134, 1997. 38. Grobelny, D., Poncz, L., and Galardy, R. E. Inhibition of human skin fibroblast collagenase, thermolysin, and Pseudomonas aeruginosa elastase by peptide hydroxamic acids. Biochemistry, 31: 7152–7154, 1992. 39. Singh, Y., Chaudhary, V. K., and Leppla, S. H. A deleted variant of Bacillus anthracis protective antigen is non-toxic and blocks anthrax toxin action in vivo. J. Biol. Chem., 264: 19103–19107, 1989. 40. Varughese, M., Chi, A., Teixeira, A. V., Nicholls, P. J., Keith, J. M., and Leppla, S. H. Internalization of a Bacillus anthracis protective antigen-c-Myc fusion protein mediated by cell surface anti-c-Myc antibodies. Mol. Med., 4: 87–95, 1998. 41. Petosa, C., Collier, R. J., Klimpel, K. R., Leppla, S. H., and Liddington, R. C. Crystal structure of the anthrax toxin protective antigen. Nature (Lond.), 385: 833– 838, 1997. 42. Netzel-Arnett, S., Fields, G. B., Birkedal-Hansen, H., Van Wart, H. E., and Fields, G. Sequence specificities of human fibroblast and neutrophil collagenases. J. Biol. Chem., 266: 6747– 6755, 1991. 43. Netzel-Arnett, S., Sang, Q. X., Moore, W. G., Navre, M., Birkedal-Hansen, H., and Van Wart, H. E. Comparative sequence specificities of human 72- and 92-kDa gelatinases (type IV collagenases) and PUMP (matrilysin). Biochemistry, 32: 6427– 6432, 1993. 44. Will, H., Atkinson, S. J., Butler, G. S., Smith, B., and Murphy, G. The soluble catalytic domain of membrane type 1 matrix metalloproteinase cleaves the propeptide of progelatinase A and initiates autoproteolytic activation. Regulation by TIMP-2 and TIMP-3. J. Biol. Chem., 271: 17119 –17123, 1996. 45. Mazzieri, R., Masiero, L., Zanetta, L., Monea, S., Onisto, M., Garbisa, S., and Mignatti, P. Control of type IV collagenase activity by components of the urokinaseplasmin system: a regulatory mechanism with cell-bound reactants. EMBO J., 16: 2319 –2332, 1997. 46. Polette, M., and Birembaut, P. Membrane-type metalloproteinases in tumor invasion. Int. J. Biochem. Cell Biol., 30: 1195–1202, 1998. 47. Brooks, P. C., Stromblad, S., Sanders, L. C., von Schalscha, T. L., Aimes, R. T., Stetler-Stevenson, W. G., Quigley, J. P., and Cheresh, D. A. Localization of matrix metalloproteinase MMP-2 to the surface of invasive cells by interaction with integrin ␣v␤3. Cell, 85: 683– 693, 1996. 48. Yu, Q., and Stamenkovic, I. Localization of matrix metalloproteinase 9 to the cell surface provides a mechanism for CD44-mediated tumor invasion. Genes Dev., 13: 35– 48, 1999. 49. Varughese, M., Teixeira, A. V., Liu, S., and Leppla, S. H. Identification of a receptor-binding region within domain 4 of the protective antigen component of anthrax toxin. Infect. Immun., 67: 1860 –1865, 1999.

6067