Stromelysin-3 Is Induced in Tumor/Stroma Cocultures and Inactivated via a Tumor-specific and Basic Fibroblast Growth Factor-dependent Mechanism*

Stromelysin-3 (STR-3) is a recently characterized matrix metalloproteinase (MMP) with a unique pattern of expression and substrate specificity. Unlike other MMPs, STR-3 is consistently and dramatically overexpressed by multiple epithelial malignancies, including carcinomas of the breast, lung, colon, head and neck, and skin. Recent studies suggest that STR-3 promotes the local establishment of epithelial malignancies, contributing to tumor cell survival and implantation in host tissues; however, STR-3’s mechanism of action remains undefined. STR-3 is a stromal cell product, prompting speculation that infiltrating stromal cells secrete STR-3 in response to tumor-derived factors. To explore this possibility, we developed a tumor/“stroma” coculture assay in which non-small cell lung cancer (NSCLC) cell lines were grown on confluent monolayers of normal pulmonary fibroblasts. In these tumor/stroma cocultures, NSCLCs stimulate normal pulmonary fibroblasts to secrete STR-3 and release extracellular basic fibroblast growth factor. Thereafter, STR-3 is processed at a unique internal sequence via a basic fibroblast growth factor- and MMP-dependent mechanism to a previously unidentified 35-kDa protein that lacks enzymatic activity. 35-kDa STR-3 is the most abundant STR-3 protein in tumor/stroma cocultures and is only detected when normal pulmonary fibroblasts are cultured with malignant bronchial epithelial cells. Therefore, the tumor-specific processing of STR-3 to the 35-kDa protein is likely to be an important regulatory mechanism.

Matrix metalloproteinases (MMP) 1 are zinc-dependent endopeptidases that promote the local invasion and distant metastasis of epithelial malignancies and the neovascularization of tumor cell deposits (1)(2)(3)(4)(5) and participate in normal tissue remodeling (6). Currently recognized families of MMPs include collagenases, gelatinases, stromelysins, membrane-type MMPs, and additional single enzymes such as matrilysin and metalloelastase; these enzymes differ in substrate specificity, regulation, tissue-specific expression, and potential interactions with additional MMP family members (2,3).
Stromelysin-3 (STR-3) (MMP-11) is a recently characterized MMP with a unique pattern of expression and substrate specificity (7,8). The enzyme was originally isolated on the basis of its overexpression in primary breast cancers and identified as a MMP family member because of its predicted amino acid (aa) sequence (7). Like other MMPs, STR-3 has a highly conserved "pro" domain which is cleaved when the enzyme is converted to its active form. STR-3 also contains a characteristic catalytic domain with a zinc-binding consensus sequence and a "hemopexin" domain with sequence similarity to the heme-binding proteins. However, STR-3 differs from other previously characterized MMPs that are secreted as inactive zymogens. The STR-3 pro domain contains an additional recognition site for the Golgi-associated pro-protein convertase, furin (9). Consequently, the ϳ60-kDa STR-3 proenzyme is processed within the constitutive secretory pathway and released as a ϳ45-kDa active enzyme (9,10).
The 45-kDa STR-3 protein is currently thought to be the major active form of the enzyme (9). However, significant questions remain regarding the biological activity of 45-kDa STR-3. Although STR-3 has the characteristic structure of a MMP, its substrate specifically differs markedly from that of other MMP family members (11). A fragment of recombinant murine STR-3 which lacks the C-terminal hemopexin domain displays the properties of a weak metalloproteinase (11). However, the human 45-kDa STR-3 does not degrade classic MMP substrates such as gelatin, casein, and elastin (8). Moreover, the human STR-3 protein contains an amino acid substitution in the highly conserved MMP "met turn" which may alter the activity of the enzyme (12).
To date, the only known substrates for STR-3 are the serine protease inhibitors (serpins), ␣1-proteinase inhibitor (␣1-PI, ␣1-antitrypsin), and ␣2-antiplasmin (8). Because ␣1-PI is the major circulating inhibitor of elastase, the degradation of ␣1-PI by STR-3 may increase elastase-mediated tissue damage. However, additional MMPs including MMP-1 (tissue collagenase) and MMP-3 (stromelysin-1) also hydrolyze ␣1-PI (13). Because STR-3 also degrades ␣2-antiplasmin, the enzyme could also indirectly increase local plasmin levels and promote plasminmediated conversion of additional pro-MMPs to their active forms (8). However, the unique aspects of STR-3 regulation, * 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. structure, and function (8,9,11,12,14) suggest that the protein has additional as yet undefined biological activities in normal tissues and epithelial malignancies.
STR-3 is expressed during normal embryogenesis and the remodeling of certain adult tissues. In human embryos, STR-3 is expressed in developing digits (15); in murine embryos, the enzyme is found during limb, tail, and snout morphogenesis (16). The enzyme is expressed by stromal elements in contact with epithelial cells in normal embryonic and adult tissues. In certain settings, STR-3-positive stromal elements are in contact with epithelial cells undergoing regional apoptosis and selected cell survival. For example, during frog morphogenesis, STR-3 is specifically expressed in small intestine mesenchyme during a time in which primary intestinal epithelial cells undergo apoptosis and replacement by secondary epithelial cells (17). In humans and rodents, STR-3 is expressed in tissues that undergo extensive remodeling such as placenta, uterus, and post-lactation mammary glands (18,19). For example, female mice who have completed weaning express STR-3 in involuting mammary glands (18). Taken together, these data suggest that specific changes in the viability of normal epithelial cells affect the expression of STR-3 in adjacent stroma.
The settings in which STR-3 is normally expressed provide insights regarding the role of the enzyme in primary tumors. Unlike other MMP family members, STR-3 is consistently and dramatically overexpressed by a variety of primary epithelial malignancies, including carcinomas of the breast, lung, colon, head and neck, and skin (20 -25). In our own recent studies, virtually all newly diagnosed primary non-small cell lung cancers (NSCLC) expressed significantly higher levels of STR-3 than adjacent normal lung specimens (22). Although STR-3 is overexpressed in primary and metastatic carcinomas, the enzyme is synthesized by interdigitating stromal cells (20 -25). In primary and metastatic tumors, STR-3 levels decline as a consequence of the distance between malignant epithelial and normal stromal cells with the highest levels of the enzyme at the tumor/stroma interface (7).
STR-3 has also been identified in stromal elements of in situ carcinomas and precursor lesions and linked with the grade and local invasiveness of early stage tumors (21,25). In a series of precancerous lesions of the respiratory tract, STR-3 was absent in hyperplastic and metaplastic lesions; in contrast, the enzyme was frequently expressed in preinvasive (dysplastic and in situ) lesions and uniformly identified in invasive carcinomas (26).
The fact that STR-3 is uniformly expressed by early stage tumors suggests that the enzyme may participate in the initial development of these malignancies. Consistent with this hypothesis, STR-3 was recently shown to promote the establishment of local tumors in nude mice by contributing to tumor cell implantation and survival in host tissues (27). Although these recent studies (27) provide the first evidence that STR-3 promotes local tumor development, the molecular mechanism by which STR-3 exerts its effects is not yet known. STR-3 appears to be expressed as a consequence of a specific interaction between malignant epithelial cells and surrounding stromal elements, suggesting that the enzyme may participate in the earliest stages of local tumor development in which malignant epithelial cells traverse the basement membrane, invade the surrounding stroma, and directly contact normal stromal elements.
To explore these possibilities in a controlled and easily accessible system, we developed a tumor/"stroma" coculture assay in which NSCLC cells are grown on confluent monolayers of normal pulmonary fibroblasts. In these tumor/stroma cocultures, NSCLC cells stimulate pulmonary fibroblasts to secrete STR-3 and release bFGF. Following the release of STR-3 and bFGF, the active 45-kDa STR-3 enzyme is processed at a unique internal sequence via a bFGF-and MMP-dependent mechanism to a major 35-kDa protein that lacks enzymatic activity. Because 35-kDa STR-3 is the most abundant STR-3 protein in tumor/stroma cocultures and is only detected when normal pulmonary fibroblasts are cultured with malignant bronchial epithelial cells, these findings provide additional insights into the regulation and role of STR-3 in epithelial carcinomas.
In selected experiments, cells from the tumor/stroma cocultures were washed twice in PBS and lysed at 4°C in PBS/0.5% Triton X-100. Cell lysates were subsequently incubated for 30 min on a 4°C rocking platform, centrifuged for 15 min at 4°C and at 10,000 ϫ g to remove insoluble material, and assayed for protein content.
Transwell Cocultures-Confluent monolayers of pulmonary fibroblasts were plated in the lower chambers and NSCLC cells in the upper chambers of transwell apparatuses (0.4-m pore size, Costar, Cambridge, MA). After an initial 18-h incubation in MEM, 10% FCS, the transwell tumor/stroma cocultures were washed, and serum-free media were added as described previously. At designated 24 -72-h intervals, conditioned media were harvested, and cell lysates were prepared as indicated.
Centrifugal Elutriation-After A549 NSCLC cells and CCL-153 fibroblasts were directly cocultured as described above, the fibroblasts and tumor cells were separated by centrifugal elutriation (J.E-5.0 elutriation system, Beckman Instruments, Inc., Fullerton, CA). Cells from the cocultures were trypsinized and washed three times in RPMI 1640 medium containing 1% FCS, 10 mM HEPES, 0.3 mM EDTA, 50 units/ml penicillin, and 50 g/ml streptomycin at 4°C. Thereafter, cells were loaded into the elutriation chamber at a pump speed of 8 ml/min (rotor speed 2000 rpm/min) in the same medium. The pump speed was increased slowly to 15 ml/min to achieve an equal distribution of cells in the chamber. After equilibrium was achieved, the pump speed was increased from 20 to 120 ml/min in 5 ml/min increments and 20 sequen-tial 100-ml cell fractions were collected. Collection was completed at a pump speed of 120 ml/min.
To assess the percentage of tumor cells and fibroblasts in each cell fraction, an aliquot of each fraction was added to an individual well of a multichamber glass slide (Lab-Tek chamber slides, Nunc, Naperville, IL); slides were air-dried and immunostained for keratin expression as described previously (22). Thereafter, the 20 individual 100-ml cell fractions were separately centrifuged and lysed for RNA extraction.
RNA Preparation and Analysis-Total RNAs from specific cell fractions were prepared and analyzed as described previously (35). In brief, total RNAs were isolated by acid guanidinium thiocyanate/phenol chloroform extraction (RNA STAT-60 kit, Tel-Test Inc., Friendswood, TX), size-fractionated on a 1% agarose gel under denaturing conditions, and transferred to a nylon membrane (Hybond Nϩ, Amersham Corp.). The blot was then hybridized with a 32 P-labeled STR-3 cDNA probe as described previously (22).
␣2-Macroglobulin Entrapment Assay-␣2-Macroglobulin entrapment was performed as described previously (8). In brief, aliquots of STR-3-containing conditioned media were incubated with 10 g of purified human ␣2-macroglobulin (Sigma) for 18 h at room temperature in the absence or presence of 5 m of the broad spectrum MMP inhibitor, BB-94. Thereafter, samples were size-fractionated on 10% SDS-PAGE gels under non-reducing conditions, blotted, and probed with the anti-STR-3 mAb 5ST4A9 as described above.
STR-3 Immunoprecipitation-Antisera directed against the STR-3 C terminus (RAST Ig) was generated by immunizing two New Zealand rabbits with an ovalbumin-coupled peptide containing the 25 C-terminal STR-3 aa (aa 464 -488) (7). Affinity-purified RAST Ig was used to immunoprecipitate STR-3 from the conditioned media of NSCLC (A549)/pulmonary fibroblast (CCL-153) cocultures. In brief, 100 l of 15 ϫ conditioned media was incubated for 2 h at 4°C with or without 2 g of affinity-purified RAST Ig; protein A-Sepharose (25 g/ml) was added for an additional 30 min at 4°C. Thereafter, samples were centrifuged at 10,000 ϫ g for 2 min and corresponding immunoprecipitates and immunodepleted conditioned media samples were collected. Immunoprecipitates and aliquots of immunodepleted conditioned media were size-fractionated on 10% SDS-PAGE gels under reducing conditions, blotted, and analyzed with the 5ST4A9 STR-3 mAb as described above.
The 35-kDa STR-3-enriched fraction (fraction 1) was subsequently loaded on an immunoabsorbent column made by covalently coupling 1 mg of immunoaffinity purified RAST IgG to 1 ml of wet protein A beads (Pharmacia Biotech Inc.). After extensive washing in 20 mM Tris-HCl, pH 7.4, 1 M NaCl, 1 mM CaCl 2 , the RAST-Ig column was eluted with 0.1 M glycine HCl, pH 2.5. Eluted fractions were immediately adjusted to pH 7.4 and assayed for 35-kDa STR-3 by immunoblotting. 35-kDa STR-3-containing fractions were pooled and concentrated 20 ϫ by ultrafiltration (Centricon-10, Amicon, Beverly, MA). The concentrated 35-kDa STR-3 sample was size-fractionated by SDS-PAGE, transferred to an Immobilon P membrane, and stained with Amido Black (Sigma). Thereafter, the ϳ35-kDa STR-3 band was excised and subjected to N-terminal sequence analysis (ABI model 492A, Worcester Foundation for Biomedical Research, Shrewsbury, MA).

STR-3 Is Induced in Tumor/Stromal Cell Cocultures-In
previous studies, STR-3 was overexpressed by stromal cells in primary NSCLC of all stages and pathologic subtypes (22). To determine whether STR-3 was secreted by stromal cells in response to tumor-derived factors, an in vitro assay was developed in which NSCLC cell lines were cocultured with normal pulmonary fibroblasts for 1-3 days in serum-free media. Thereafter, conditioned media from the tumor/stroma cocultures were size-fractionated, immunoblotted, and analyzed for STR-3 ( Fig. 1).
When cultured alone, NSCLC cell lines such as A549 (tumor, T) secrete virtually no STR-3 ( Fig. 1, lanes 4 and 8). Normal fetal (CCL153) and adult (CCL-210) pulmonary fibroblasts that are cultured alone for 1-3 days also secrete little 45-or 60-kDa STR-3 ( Fig. 1, lanes 1 and 5). In marked contrast, tumor/stroma cocultures secrete increased quantities of the 60-kDa proform and the 45-kDa active enzyme (CCL-153, 5.5 ϫ base line and CCL-210, 24.5 ϫ base line, Fig. 1, lanes 3 and 7). Similar results were obtained when pulmonary fibroblasts were cocultured with additional NSCLC cell lines (data not shown). For these reasons, RNAs from fibroblast-and tumor cellenriched fractions (fractions 1 and 20, respectively) were prepared and analyzed by Northern blot for STR-3 (Fig. 2, lanes 4  and 5); RNAs from CCL-153 fibroblasts and A549 NSCLC cells cultured separately (Fig. 2, lanes 1 and 3) or directly cocultured and harvested together (Fig. 2, lane 2) were also analyzed. As expected, STR-3 transcripts were significantly more abundant in tumor/stroma cocultures than in fibroblasts or tumor cells cultured separately (Fig. 2, compare lane 2 with lanes 1 and 3). STR-3 transcripts were also significantly more abundant in the fibroblast-enriched fraction than in the tumor cell-enriched fraction of the separated coculture (Fig. 2, compare lanes 4 and  5). Taken together, these data identify the fibroblasts as the primary source of STR-3 in the tumor/stroma cocultures.

A Previously Unidentified 35-kDa STR-3 Protein Is the Major Form of STR-3 in Tumor/Stroma
Cocultures-Although the fibroblast-derived mature active 45-kDa STR-3 enzyme was present in conditioned media from tumor/stroma cocultures, the major form of STR-3 was a previously unidentified 35-kDa protein (Fig. 1, lanes 3 and 7). This 35-kDa STR-3 protein was of particular interest because it constituted ϳ70% of all STR-3 in conditioned media from prolonged (day 3) cocultures (Fig. 1,  lane 7).
The 35-kDa STR-3 Protein Is Processed at the N Terminus and Lacks a Portion of the Catalytic Domain-To determine whether the newly identified 35-kDa STR-3 protein was a processed form of the larger active enzyme, conditioned media from tumor/stroma cocultures were initially immunoprecipitated with an antiserum directed against the C-terminal STR-3 peptide (RAST). RAST-STR-3 immunoprecipitates were subsequently immunoblotted and analyzed with an antibody directed against the STR-3 hemopexin domain (5ST4A9). Fig. 3A includes RAST-STR-3 (lane 2) and control immunoprecipitates (lane 1) and aliquots of conditioned media following immunodepletion with RAST (lane 4) or protein A alone (lane 3). The RAST antiserum removes the majority of 35-kDa STR-3 from tumor/stroma cell-conditioned media, indicating that 35-kDa STR-3 contains the full C terminus in addition to the 5ST4A9 epitope from the hemopexin domain (Fig. 3A, lanes  2 and 4). To determine whether 35-kDa STR-3 undergoes additional processing at the N terminus, a duplicate blot of the indicated samples was analyzed with an antibody directed against the STR-3 catalytic domain (5ST4C10). Although the 45-kDa active enzyme reacted with 5ST4C10, 35-kDa STR-3 was not identified by this antibody (data not shown). Taken together, these data indicate that 35-kDa STR-3 is processed at the N terminus and that this smaller STR-3 protein lacks the 5ST4C10 epitope from the catalytic domain.

35-kDa STR-3 Does Not Entrap ␣2
Macroglobulin-Because 35-kDa STR-3 contains the C-terminal hemopexin domain but lacks a portion of the N-terminal catalytic domain, we compared the biological activity of 35-kDa STR-3 and the mature 45-kDa active enzyme using an ␣2-macroglobulin entrapment assay (8). ␣2-Macroglobulin is a broad range protease inhibitor which complexes with all of the previously characterized metalloproteinases, including 45-kDa STR-3 (8).
The Generation of 35-kDa STR-3 Requires Tumor-specific Interactions with Surrounding Stromal Cells-To determine whether malignant epithelial cells are required for the generation of 35-kDa STR-3, we cocultured pulmonary fibroblasts with either NSCLC (Fig. 4, A and C, lanes 4 -6) or SV40 immortalized non-tumorigenic tracheal epithelial cells (Fig. 4,  A and C, lanes 1-3) and assayed STR-3 in the resulting conditioned media. As previously demonstrated (Fig. 1), CCL-153 fibroblasts constitutively secrete higher levels of STR-3 than the CCL-210 fibroblasts (compare Fig. 4, A, lanes 1 and 4 with  C, lanes 1 and 4).
The induction of STR-3 was over 5-fold greater in cocultures of pulmonary fibroblasts and NSCLC cells than in cocultures of fibroblasts and SV40-immortalized non-tumorigenic tracheal epithelial cells (compare Fig. 4, A, lane 5 versus 2, and C, lane 5 versus 2). In addition, although 35-kDa STR-3 was the most abundant STR-3 protein in tumor/stroma cocultures (Fig. 4, A and C, lane 5), 35-kDa STR-3 was undetectable in cocultures of SV40-immortalized tracheal epithelial cells and pulmonary fibroblasts (Fig. 4, A and C, lane 2). These data indicate that the induction of 45-kDa STR-3 is more efficient in tumor/stroma cocultures and that the generation of 35-kDa STR-3 is tumorspecific and dependent upon the interaction between malignant bronchial epithelial cells and pulmonary fibroblasts. For these reasons, we attempted to identify the factors responsible for the generation of 35-kDa STR-3.
The Generation of 35-kDa STR-3 in Tumor/Stroma Cocultures Requires bFGF-Because factors including PDGF, EGF, and bFGF increase STR-3 transcript abundance in normal pulmonary fibroblasts in vitro (7,22), we evaluated the potential role of these factors in STR-3 induction and processing in the tumor/stroma cocultures. As indicated in Fig. 4 (A and C), tumor/stroma cocultures were performed in the presence or absence of neutralizing antibodies directed against PDGF, the EGF receptor, or bFGF. None of these neutralizing antibodies reduced the quantities of 45-kDa STR-3 in conditioned media from tumor/stroma cocultures (Fig. 4, A and C, compare lanes  5, 8, 11, and 14); PDGF and EGF receptor neutralizing antibodies also had no effect on the generation of 35-kDa STR-3
Because extracellular bFGF was implicated in the generation of 35-kDa STR-3, additional aliquots of conditioned media from the day 3 tumor/stroma cocultures ( Fig. 4A and C, NA) were also assayed for bFGF content (Fig. 4, B and D). Although neither fibroblasts nor NSCLC cells that were cultured alone released detectable quantities of bFGF (Fig. 4, B and D, lanes  1 and 3), tumor/stromal cocultures released readily detectable bFGF (Fig. 4, B and D, lane 2).

The Generation of 35-kDa STR-3 and the Release of bFGF Do Not Require Direct Tumor/Stromal Cell Contact: Transwell
Assays-To determine whether bFGF induction/release and bFGF-mediated processing of STR-3 require physical contact between NSCLC cells and pulmonary fibroblasts, the two cell types were cocultured in a transwell apparatus which permits only the diffusion of soluble factors. As indicated in Fig. 5A  (lanes 2 and 5), 35-kDa STR-3 and bFGF were equally abundant in conditioned media from transwell and direct tumor/ stroma cocultures.
Because bFGF release was not dependent upon tumor/stromal cell contact, the transwell cocultures were also used to identify the cellular source of bFGF. Fibroblasts and tumor cells from transwell cocultures were separately lysed and analyzed for bFGF (Fig. 5B, lanes 5 and 6); direct fibroblast/tumor cell cocultures were similarly evaluated (Fig. 5B, lane 2). It is readily apparent that the fibroblasts are the primary source of bFGF in tumor/stroma cocultures (Fig. 5B, compare lanes 5 and  6) and that the fibroblasts produce increased bFGF following their exposure to NSCLC cells (Fig. 5B, compare lanes 4 and 5,  lanes 1 and 2).
The Generation of the 35-kDa STR-3 Is Inhibited by BB-94 -bFGF increases the expression of multiple proteinases including MMPs and serine proteases of the plasmin activator/plasmin system (2, 36), prompting speculation that bFGF-mediated processing of STR-3 might require additional MMPs or plasmin. To assess this possibility, normal pulmonary fibroblasts were cocultured with NSCLC cells in the presence or absence of a broad spectrum MMP inhibitor (BB-94) (30 -32) or the serine protease inhibitor aprotinin (33) (Fig. 6). As indicated, BB-94 completely inhibited the generation of 35-kDa STR-3 in tumor/ stroma cocultures, whereas aprotinin had no detectable effect (Fig. 6, A and B, compare lanes 2, 5 and 8). In additional experiments, neither the more specific plasmin inhibitor, ␣2AP, nor the specific inhibitor of urokinase-type plasmin activator, PAI-1, reduced the generation of 35-kDa STR-3 (data not shown). Therefore, bFGF-mediated processing of STR-3 to the major 35-kDa protein is not dependent upon the plasmin activator/plasmin system but does require MMP activity.
Generation and Processing of STR-3 in Tumor/Stroma Cocultures: Proposed Model-Taken together, the data from the NSCLC/pulmonary fibroblast cocultures indicate that NSCLCs stimulate normal pulmonary fibroblasts to secrete STR-3 and release bFGF (Fig. 7). Thereafter, STR-3 is processed at the N terminus via a tumor-specific and bFGF-and MMP-dependent mechanism to a major previously undescribed 35-kDa protein which differs in biological activity from 45-kDa STR-3 (Fig. 7).

35-kDa STR-3 Purification and Analysis of N-terminal
Sequence-Because the 35-kDa STR-3 lacks the 5ST4C10 epitope from STR-3 catalytic domain and fails to entrap ␣2-macroglobulin (Fig. 3B), this major STR-3 protein is unlikely to have the enzymatic activity of its 45-kDa precursor. To specifically compare mature active 45-kDa STR-3 with the 35-kDa processed protein, we purified 35-kDa STR-3 and identified its N terminus.
35-kDa STR-3 Purification-An unusual NSCLC cell line, SL-6, was used in the large scale purification of 35-kDa STR-3. Unlike the majority of NSCLC cell lines that do not secrete STR-3, SL-6 cells secrete STR-3 and process the mature active enzyme to 35-kDa protein in absence of normal pulmonary fibroblasts (Fig. 8A, lane 1). In SL-6 cells, the processing of STR-3 to a 35-kDa protein is also bFGF-and MMP-dependent; 35-kDa STR-3 is less abundant when SL-6 is cultured in presence of a neutralizing bFGF mAb (Fig. 8A, lane 2) or concentrations of Ն1 M BB-94 (Fig. 8, A and B). In SL-6 cells cultured with increasing amounts of the MMP inhibitor (BB-94), the concentrations of both 60-kDa pro-STR-3 and mature 45-kDa STR-3 increase as that of the 35-kDa STR-3 decreases, confirming the precursor/product relationship between the larger and smaller STR-3 proteins (Fig. 8B).
N-terminal Sequence of 35-kDa STR-3-To determine the N-terminal sequence of 35-kDa STR-3, the protein was purified from SL-6 conditioned media using a three-step procedure (see "Materials and Methods"). The highly purified STR-3 sample contained a single major ϳ35-kDa band after SDS-PAGE, transfer to an Immobilon membrane, and Amido Black staining (data not shown). This 35-kDa STR-3 band was excised and subjected to N-terminal sequencing. Three N-terminal sequences were identified in the broad ϳ35-kDa STR-3 band (Fig. 9). These N-terminal sequences, aa 190 -195, 192-194, and 194 -199 from full-length STR-3, yield a protein with the calculated mass of ϳ35-kDa.
The identification of three related N-terminal 35-kDa STR-3 sequences suggests that the protein may be cleaved at a single site and subsequently degraded by an autocatalytic mechanism (37,38) or by additional enzymes in SL-6 conditioned media. In this regard, it is of interest that the most N-terminal 35-kDa STR-3 sequence (Glu 190 -Tyr 195 ) is preceded by a non-polar glycine residue (aa 189) and a unique highly polar 5-aa sequence (KTHRE, aa 184 -188) (Fig. 9). Although the KTHRE sequence is completely conserved in human, murine, and amphibian STR-3, the sequence differs in its entirety from the analogous non-polar aa sequence (GPGIG) in other MMP family members (Fig. 9, inset) (39). These data suggest that the tumor-specific and bFGF-mediated processing of STR-3 occurs via a mechanism unique to STR-3.

35-kDa STR-3 Lacks Specific N-terminal Residues
Required for MMP Enzymatic Activity-Of note, 35-kDa STR-3 lacks specific N-terminal residues that have been implicated in MMP-mediated cleavage. These missing aa include a highly conserved Ala (Ala 178 in STR-3) that plays an important role in the reaction mechanism and the secondary zinc ligands Asp 164 , Asp 166 , His 179 , and a calcium ligand Asp 171 that contribute to the structural integrity of MMPs (40,41). Taken together, the epitope mapping (Fig. 3A) and preliminary structural and functional analyses (Figs. 3B and 9) suggest that 35-kDa STR-3 lacks components of the catalytic domain necessary for enzymatic activity.  (aa 190 -195, 192-194, and 194 -199) were identified in the broad ϳ35-kDa STR-3 band. The most N-terminal 35-kDa STR-3 sequence (Glu 190 -Tyr 195 ) is preceded by a non-polar glycine residue (aa 189) and a highly polar 5aa sequence (KTHRE) which is completely conserved in human, murine, and amphibian STR-3; this 5-aa sequence differs from the analogous non-polar conserved aa sequence (GPIGG) in other MMP family members (inset) (39).
To assess directly the proteolytic activity of 35-kDa STR-3, recombinant 35-and 45-kDa STR-3 proteins were synthesized and compared in ␣1-PI degradation assays. As described previously (8,12), 200 ng of r45-kDa STR-3 cleaved ␣1-PI (2 g), and this activity was inhibited by the addition of BB-94 (Fig.  10, compare lanes 1, 3, and 4). In marked contrast, 200 ng of r35-kDa STR-3 did not cleave ␣1-PI (Fig. 10, lane 6); increased amounts of r35-kDa STR-3 (up to 2 g) also had no effect (data not shown). These data provide further direct evidence that the major 35-kDa STR-3 protein lacks the proteolytic activity of the mature active 45-kDa enzyme. DISCUSSION We have developed a tumor/stroma coculture assay in which NSCLC cells stimulate normal pulmonary fibroblasts to release STR-3 and the potent angiogenic peptide, bFGF (Fig. 7). In these cocultures, STR-3 is processed via a tumor-specific and bFGF-dependent mechanism to yield a major 35-kDa protein that lacks enzymatic activity (Fig. 7). Because 35-kDa STR-3 is the most abundant STR-3 protein in tumor/stroma cocultures and is only detected when normal pulmonary fibroblasts are cultured with malignant bronchial epithelial cells, these findings provide additional insights into the regulation and potential role of STR-3 in epithelial carcinomas.
STR-3 differs from other MMP family members in several key ways as follows: 1) the mature enzyme does not degrade common MMP substrates (8,11); 2) the human STR-3 protein contains an amino acid substitution in the highly conserved MMP met turn which may alter its enzymatic activity (12); and 3) STR-3 undergoes tumor-specific processing at a highly conserved sequence that is specific to STR-3 (Fig. 9). In addition, large quantities of recombinant STR-3 are required to degrade the only described native substrate, ␣1-PI (Ref. 8 and Fig. 10). Taken together, these data suggest that STR-3 may not function in a manner analogous to other previously characterized MMPs and that the protein may have an as yet unidentified function.
The currently described coculture assays provide important additional clues regarding the unique nature of STR-3. In the tumor/stroma cocultures, tumor cells stimulate normal stromal cells to secrete STR-3 and mediate the processing of 35-kDa STR-3. Although 35-kDa STR-3 is the major form of the enzyme in conditioned media from tumor/stroma cocultures (Figs. 1 and 4), 35-kDa STR-3 was not previously detected in conditioned media from normal pulmonary fibroblasts or STR-3 Cos cell transfectants (8,27). This is not surprising because the generation of 35-kDa STR-3 requires an interaction between tumor and stromal cells (Fig. 4). These observations underscore the utility of an in vitro assay which mimics the in vivo setting in which tumor cells invade the basement membrane and come into direct contact with normal stromal elements.
We performed the coculture assays with two types of normal pulmonary fibroblasts and either malignant bronchial epithelial cells or non-tumorigenic tracheal epithelial cells (Fig. 4). When normal pulmonary fibroblasts are cocultured with malignant bronchial epithelial cell lines, 45-kDa STR-3 is secreted and processed to the major 35-kDa protein. When fibroblasts are cocultured with non-tumorigenic tracheal epithelial cells, there is reduced but detectable secretion of 45-kDa STR-3; however, there is no detectable processing to 35-kDa STR-3 protein (Fig. 4, A and C, lane 2). The demonstrated enzymatic activity of STR-3 is so different from that of other MMPs that it remains to be determined whether 45-kDa STR-3 is functioning as a classic MMP in vivo. Furthermore, the unique nature of STR-3 processing and tumor-specific and bFGF-dependent generation of 35-kDa STR-3 prompt speculation regarding additional functions for the processed protein.
Our analyses of STR-3 induction and processing in tumor/ stroma cocultures indicated that bFGF increased the processing but not the induction of STR-3. Although recombinant bFGF increased STR-3 transcript abundance in normal pulmonary fibroblasts in previous in vitro studies (7,22), neutralizing bFGF antibodies did not inhibit the induction and secretion of STR-3 in tumor/stroma cocultures (Fig. 4, A and C, lane 14). However, neutralizing bFGF antibodies dramatically decreased the processing of STR-3 to the major 35-kDa protein in these assays (Fig. 4, A and C, lane 14). Because bFGF mediates the processing of STR-3 to the enzymatically inactive 35-kDa protein in tumor/stroma cocultures, we analyzed bFGF release in these conditions. As indicated in Figs. 4 and 5, tumor/stroma coculture increased the production and release of fibroblastderived bFGF.
bFGF stimulates the growth and metastasis of many types of tumors (42)(43)(44)(45)(46)(47)(48)(49) and promotes tumor angiogenesis (42,50,51). There are several alternatively spliced 18 -23-kDa bFGF isoforms that are thought to have different mechanisms of action; the low molecular mass 18-kDa bFGF isoform is most likely to be released and to interact with high affinity cell-surface receptors (52,53). Consistent with these observations, 18-kDa bFGF is the primary isoform detected in conditioned media from tumor/stroma cocultures (Fig. 5A). However, bFGF lacks a classical leader sequence and appears to be released by novel secretory mechanisms (50,54). For these reasons, the tumor/ stroma cocultures may represent a useful model system in which to analyze potential mechanisms of bFGF release.
Because bFGF stimulates the production of multiple proteinases including MMPs and serine proteases of the plasmin activator/plasmin system (2,3,36), we explored the possibility that bFGF-mediated STR-3 processing occurred via an additional MMP or plasmin. The broad spectrum MMP inhibitor (BB-94) inhibited the generation of 35-kDa STR-3 in tumor/ stroma cocultures indicating that bFGF-mediated processing of 35-kDa requires additional MMP activity. In tumor/stroma cocultures, the major processed form of STR-3 occasionally appears as a ϳ37/35-kDa STR-3 doublet (Figs. 4 and 5), prompting speculation that STR-3 may undergo initial MMPmediated cleavage and subsequent additional processing. The identification of three related 35-kDa STR-3 N-terminal amino acid sequences (Fig. 9) is also consistent with this observation. Recently, other MMP family members have also been reported to undergo initial MMP-mediated cleavage and subsequent autocatalytic processing (37,38).
In summary, the data derived from the tumor/stroma coculture assay provide additional evidence regarding the unique nature of STR-3. STR-3 differs from other MMP family members in its initial activation (9), substrate specificity, and biological activity (9, 10) and near-uniform overexpression in epithelial malignancies (20 -25). The current studies demonstrate  3 and 4) or the truncated recombinant 35-kDa STR-3 (lanes 6 and 7) in the absence (lanes 3 and 6) or the presence (lanes 4 and 7) of 5 M BB-94. After incubation, reaction products were analyzed by SDS-PAGE and stained with Coomassie Blue. I and C indicate intact and cleaved ␣1-PI, respectively. that in tumor/stroma cocultures, STR-3 is also processed at a unique internal sequence via a tumor-specific and bFGF-and MMP-dependent mechanism to a major 35-kDa protein that lacks enzymatic activity. The regulatory nature of the major 35-kDa STR-3 protein and the relationship between bFGF and STR-3 in tumor/stroma cocultures and primary epithelial malignancies will be of further interest.