84 LI ET AL. MOLECULAR CARCINOGENESIS 22:84–94 (1998)

© 1998 WILEY-LISS, INC.

Immunological Characterization of Cell-Surface andSoluble Forms of Membrane Type 1 MatrixMetalloproteinase in Human Breast Cancer Cellsand in FibroblastsHui Li,1 Delbert E. Bauzon,1 Xinyun Xu,1 Harald Tschesche,2 Jian Cao,3 and QingXiang Amy Sang1*1Department of Chemistry, Biochemistry Division, Florida State University, Tallahassee, Florida2Department of Biochemistry, Faculty of Chemistry, University of Bielefeld, Bielefeld, Germany3Department of Medicine, State University of New York, Stony Brook, New York

Membrane type (MT) 1 matrix metalloproteinase (MMP) activates progelatinase A (pro–MMP-2), a type IVcollagenase, on the cell surface of tumors; however, its function in breast cancer progression and metastasis isnot fully understood. To examine the expression of MT1-MMP in breast cancer cells and fibroblasts, a specificrabbit antibody (Ab) directed against a unique synthetic peptide derived from the human MT1-MMP catalyticdomain was produced, purified, and characterized. This Ab is not likely to cross-react with MT2-, MT3-, or MT4-MMP or any other MMPs. MT1-MMP expression and pro–MMP-2 activation were stimulated by concanavalin Ain two human breast carcinoma cell lines (BT549 and MDA-MB-231) and in normal human fetal-lung fibro-blasts (HFL-1) and were slightly upregulated by breast cancer cell–fibroblast interactions. Both pro-MT1-MMPin plasma membrane (63.4 kDa) and the soluble forms of the enzyme in culture medium (57.6 and 25–30 kDa)were detected by immunoblot analysis, suggesting that cell-surface MT1-MMP exhibits an active conformationwithout the removal of its propeptide domain and that the mature enzyme is shed into the medium. In breastcancer cells, MT1-MMP and a recombinant catalytic domain of MT1-MMP were unable to activate pro-matrilysin,indicating that MT1-MMP is not a universal activator of all MMPs. MT1-MMP may play an important role in theinvasive growth and spread of breast cancer cells by specifically activating pro–MMP-2 to cleave the connec-tive-tissue barrier. Furthermore, use of the specific Ab may aid in the investigation of the role of MT1-MMP inhuman tumors. Mol. Carcinog. 22:84–94, 1998. © 1998 Wiley-Liss, Inc.

Key words: antibody production; gelatinase A; matrilysin; proteinase processing; invasion and metastasis

INTRODUCTION

Matrix metalloproteinases (MMPs) are a family ofclosely related hydrolases that require zinc for ca-talysis and calcium for structural integrity [1–3]. Be-cause of their ability to dissolve connective-tissuecomponents such as collagens, laminins, fibronectin,and proteoglycans, MMPs may be one of the mostimportant classes of molecules used by invading cellsto facilitate invasive growth and spread. All of thesecreted MMPs are produced as latent, inactive proen-zymes (pro-MMPs) that may be activated by cellularactivators. The activated MMPs can be inhibitedby tissue inhibitors of metalloproteinase (TIMPs)[1,2,4]. The activation and inhibition of MMPs iswell regulated under normal physiological condi-tions. In a number of pathological situations,however, the strict regulatory mechanisms arelost, leading to tumor metastasis and connectivetissue diseases such as arthritis.

To spread to other parts of the body, breast cancercells must break down the basement membranes un-derneath the epithelial breast cancer cell layer andunderneath the endothelial cell layer of blood ves-

sels. The cancer cells must also cleave interstitial bar-riers, such as type I collagen. The proteolytic activi-ties of MMPs, in particular those of type IVcollagenases, enable the destruction of the connec-tive-tissue barrier and allow the cancer cells to es-cape from confined environments, enter bloodvessels to reach a distant site, and form a secondarytumor. Both breast epithelial cancer cells and stro-mal cells produce the 72-kDa gelatinase A/type IVcollagenase (pro-MMP-2) in vitro and in vivo [5,6].This proteinase is partially responsible for digesting

*Correspondence to: Department of Chemistry, 203 Dittmer Labo-ratory of Chemistry Building, Florida State University, Tallahasee, FL32306-4390.

Received 6 June 1997; Revised 18 November 1997; Accepted 19December 1997

Abbreviations: MMP, matrix metalloproteinase; TIMP, tissue inhibi-tor of metalloproteinase; MT, membrane type; ECM, extracellularmatrix; Ab, antibody; cdMT1-MMP, catalytic domain of MT1-MMP;DMEM, Dulbecco’s modified Eagle’s medium; CM, conditioned cell-culture medium; ConA, concanavalin A; TBS, Tris-buffered saline;DT, detergent; AQ, aqueous; mAb, monoclonal antibody; IgG, im-munoglobulin G; ELISA, enzyme-linked immunosorbent assay; PL,pellet; SDS, sodium dodecyl sulfate.

MT1-MMP IN HUMAN BREAST CANCER 85

interstitial collagens and basement membrane typeIV collagen, laminin, and some glycoproteins.

The production or secretion of inactive pro–MMP-2 alone is not sufficient to induce tumor-cell inva-sion because the zymogen cannot hydrolyze theconnective-tissue barriers. The significance of pro–MMP-2 activation during tumor invasion has beenreviewed [7]. Davies et al. [8] reported that when thetotal protein level of MMP-2 (both latent and active)is considered, there is no difference in biopsy samplesof benign and malignant human breast diseases.However, the amount of active enzyme is signifi-cantly higher in malignant disease. It is only whenthe proenzyme is activated and the levels of activatedenzyme exceed the local concentrations of enzymeinhibitors that type IV collagenase activity is ex-pressed [5,6]. The production of the cellular activa-tor for pro-MMP is most critical in determining theinvasive potential of breast cancer cells.

Membrane type (MT) 1–MMP, one of the cellularactivators of pro–MMP-2, has been identified as be-ing on the surface of invasive tumor cells [9,10]. Themolecular mass calculated from the primary sequenceof MT1-MMP is 62.7 kDa for the zymogen (550 resi-dues) and 53.8 kDa for its active form (471 residues)[3]. Although this novel MMP has many structuralcharacteristics of the MMP family, such as thePRCGVPD consensus sequence in the propeptidedomain and similar catalytic and hemopexin-likedomains, it is distinct in containing a furin recogni-tion sequence (Arg-Arg-Lys-Arg) in the propeptidedomain, a unique insertion in the catalytic domain,and a transmembrane domain at the C terminus [9].Thus, MT1-MMP may also have unique functions.MT1-MMP is expressed by stromal cells of humanbreast carcinoma specimens [11,12] and by highlymetastatic human breast cancer cells [13]. Furtherevidence has shown that this cell-surface enzyme notonly activates pro–MMP-2 but also cleaves fibro-nectin, laminin, vitronectin, dermatan sulfateproteoglycan, and gelatin [14] as well as interstitialcollagens [15]. It also activates human breast cancerpro–collagenase 3 [16]. Therefore, this enzyme mayplay an important role in breast cancer invasion andmetastasis [17].

The fundamental regulation of MMP function isat the level of the net enzyme activity, i.e., theamount of active proteinase in excess of its inhibi-tors. This fact is often overlooked in studies thatmonitor the mRNA and protein levels of a singlecomponent of this system. We have been investigat-ing how pro-MMPs are activated and how MMP ac-tivities are regulated. We previously demonstratedthat matrilysin (MMP-7) and human fibroblast-typeinterstitial collagenase (MMP-1) can activate pro–MMP-2 and 92–98-kDa gelatinase B (MMP-9) a typeIV collagenase [18,19]. MMP-7 can also activate pro–MMP-1 [19]. The natural extension of our biochemi-cal studies has led us to investigate the role of

MT1-MMP in the activation of pro–MMP-2 and pro–MMP-7 in breast cancer cell/fibroblast culture systems.

Breast malignancies are almost exclusively epithe-lial cancers; however, the growth and disseminationof breast cancer requires complex interactions be-tween epithelial cancer cells and stromal components[20]. The growth and progression of the cancer isintimately related to the microenvironment, i.e., theextracellular matrix (ECM) and the stromal cells [21].Breast cancer cells constantly interact with the ECMand communicate with surrounding cells. The inter-actions of tumor and stroma in regulation of MMPexpression and in tumor progression have been sum-marized [20]. The expression of most MMP familymembers in breast cancer represents a tumor-inducedhost response. Most MMPs are expressed in the tu-mor stromal cells immediately adjacent to the can-cer cells [12]. The complex relationship betweenstromal and epithelial cells is an important factor inthe control of MMPs extremely relevant to breastcancer. Thus, the potential influence on MT1-MMPexpression by epithelial breast cancer cell–stromalcell interaction was also examined in this study.

A rabbit polyclonal antibody (Ab) against a uniqueinsertion sequence in the catalytic domain of hu-man MT1-MMP (cdMT1-MMP) was produced andcharacterized to aid in the study of MT1-MMP ex-pression in breast cancer cells and fibroblasts and instudies of the activation mechanism of pro–MMP-2.This Ab is highly specific; it does not cross-react withMMP1, MMP2, or MMP9, and it is not likely to cross-react with MT2-, MT3-, or MT4-MMPs or any otherMMPs. Both the plasma-membrane–associated 63.4-kDa form and the lower-molecular-mass solubleforms of MT1-MMP were detected with the Ab, sug-gesting that the cell-surface MT1-MMP still containsthe propeptide domain. Our data indicate that pro–MMP-2 is activated on the cell surface and releasedinto the extracellular medium. Pro–MMP-7 was notactivated by membrane-associated MT1-MMP or bythe soluble cdMT1-MMP; thus, pro–MT1-MMP mayexhibit an active conformation on the cell surfaceand specifically activate pro–MMP-2. MT1-MMP isnot a universal activator for all pro-MMPs. This re-search may lead to the identification of new and spe-cific treatment strategies based on the inhibitorstargeted to cell-surface MMP to disrupt the invasivespread of breast cancer.

MATERIALS AND METHODS

Cell Lines and Cell Culture

Two highly invasive human breast carcinoma celllines, BT549 and MDA-MB-231, were purchased fromthe American Type Culture Collection (Rockville,MD). BT549 has been shown to produce high levelsof both MMP-2 and MMP-9 [22]. MDA-MB-231 hasbeen shown to express MT1-MMP mRNA [11] andprotein [13]. The normal human fetal-lung fibroblas-

86 LI ET AL.

tic cell line HFL-1 was also purchased from Ameri-can Type Culture Collection. The cells were culturedin Dulbecco’s modified Eagle’s medium (DMEM;Mediatech, Washington, DC) with 4.5 g/L glucoseon plastic dishes in the presence of 10% fetal bovineserum (Hyclone Laboratories, Inc., Logan, UT). After3–8 d in culture, the cells were washed with and sub-cultured overnight in serum-free conditioned cell-culture medium (CM). Concanavalin A (ConA; SigmaChemical Co., St. Louis, MO) was shown to stimu-late MT1-MMP production [13]; therefore, the cellswere cultured in the presence and absence of 20 µg/mL ConA in serum-free CM. The CM was collectedand centrifuged, and the supernatant was concen-trated by using Centriprep-10 ultrafiltration concen-trators (Amicon, Inc., Beverly, MA). Proteinconcentrations were determined by using the MicroBCA Protein assay with bicinchoninic acid (PierceChemical Company, Rockford, IL).

Plasma-Membrane Preparation

The human breast cancer cell lines MDA-MB-231and BT549 and the normal human fetal-lung fibro-blastic cell line HFL-1 were cultured in the presenceor absence of 20 µg/mL ConA. The cell membranefraction was prepared as previously described [23,24].Briefly, cells from two dishes (100 mm × 20 mm) werecollected, centrifuged, and resuspended in 1 mL ofTris-buffered saline (TBS; 50 mM Tris and 150 mMNaCl, pH 7.4) containing 1.5% (v/v) Triton X-114. Thecells were homogenized with a Dounce homogenizer(Wheaton, Millville, NJ), and the whole-cell homoge-nate was centrifuged at 10 000 ×g for 2 min at 4°C.The pellet was dissolved in 1 mL of TBS. The superna-tant fraction was collected and phase-partitioned intodetergent (DT) and aqueous (AQ) phases at 37°C for 3min and then centrifuged at 5000 ×g for 1 min. TheDT phase containing the plasma membrane–enrichedfraction was diluted with cold TBS to 1 mL.

Antibody Production

A specific peptide corresponding to a unique in-sertion sequence in the cdMT1-MMP (Arg160-Glu-Val-P ro -Tyr -A la -Tyr - I l e -Arg -Glu -Gly -Hi s -Glu -Lys-Gln174-NH2) was synthesized and purified [9]. Thispeptide has very little or no homology to other mem-bers of the MMP family [3,25]. The molecular massof this peptide is 1.87 kDa, its isoelectric point is 7.5,and its molar extinction coefficient at 280 nm is 2560/M/cm based on calculations previously described[3,4]. The MT1-MMP peptide was synthesized by theBiochemical Analysis, Synthesis, and Sequencing Ser-vice Laboratory in the Department of Chemistry atFlorida State University (Tallahasee, FL). The peptidewas purified by Sephadex-G25 size-exclusion chro-matography, and the purity was verified by reverse-phase high-performance liquid chromatography. Thepurified MT1-MMP peptide (160–174) was coupledto keyhole limpet hemocyanin (Sigma Chemical Co.)

with 2% glutaraldehyde in 0.2 M phosphate buffer,pH 7.0. One milligram of the conjugated peptide wasemulsified in Freund’s complete adjuvant (SigmaChemical Co.) and injected into each of four NewZealand white rabbits. Subsequently, 1 mg of theconjugated peptide was emulsified in Freund’s incom-plete adjuvant (Sigma Chemical Co.) and used toboost the rabbits at 6-wk intervals [26]. The rabbitswere bled 1 wk after each boost. Rabbit sera weretaken before and after the immunization. Mono-clonal antibodies (mAbs) against MMP-2 and MMP-9 were purchased from Oncogene Research Product(CalBiochem, Cambridge, MA).

Ab Characterization

Peptide antigen immobilized on nitrocellulosemembrane was used for affinity purification of spe-cific Ab according to a published method [27] withsome modifications. Briefly, the nitrocellulose mem-brane was incubated with 1 mg/mL peptide antigenat 4°C overnight. The membrane was dried and thenincubated with rabbit antiserum at 4°C overnight.The membrane was rinsed in TBS and the Ab waseluted with 0.1 M glycine, pH 2.8. Cold 1 M Trisbuffer, pH 8.5, was immediately added to the elutedAb solution to bring the pH to 7.4. A two-step affin-ity chromatography method was also used to purifythe specific Ab. The peptide was coupled to Affi-Gel10 gel (Bio-Rad Laboratories, Hercules, CA) accord-ing to manufacturer’s instructions. The immunoglob-ulin G (IgG) fraction of the antipeptide antiserumwas first purified by protein A affinity column (SigmaChemical Co.) according to manufacturer’s direc-tions. The specific IgG molecules were then furtherpurified using the antigen-affinity Affi-Gel 10 gelcolumn [28]. The purified Ab was concentrated andstored at –80°C in 0.1 M glycine, 60 mM Tris, and0.02% sodium azide, pH 7.4. The Abs purified by theone-step nitrocellulose membrane method and bythe two-step affinity chromatography method wereboth specific. Preimmune IgG molecules were puri-fied from preimmune serum by protein A affinitychromatography for negative control experiments.

An enzyme-linked immunosorbent assay (ELISA)was used to assess the strength and specificity of therabbit polyclonal Ab [28]. Polyvinyl micro ELISAplates (96 well; Costar Corp., Cambridge, MA) werecoated with 50 µL/well of different concentrationsof either the synthetic peptide antigen or the recom-binant human cdMT1-MMP [29] in borate-bufferedsaline. Alkaline phosphatase–conjugated affinity-purified goat anti–rabbit IgG (Sigma Immuno Chemi-cals, St. Louis, MO) was used as the secondary Ab,and 1 mg/mL p-nitrophenyl phosphate in 1.0 Mdiethanolamine buffer, and 0.5 M MgCl2, pH 9.8 wasused as a substrate. After incubation, the plates wereread at 405 nm with a Titertek Multiscan MC-340automatic microplate reader (Flow Laboratories,McLean, Virginia). The propeptide, human fibroblast

MT1-MMP IN HUMAN BREAST CANCER 87

collagenase (MMP-1), gelatinase A (MMP-2), and neu-trophil gelatinase B (MMP-9) were used to test thespecificity and cross-reactivity of the Ab. The sourcesof these human MMPs and of recombinant humanMMP-7 are described by Sang et al. [18,19].Preimmune sera and IgG were also tested for theircross-reactivity with the peptide antigen.

Vector Construction, Transfection of COS-1 Cells,and Preparation of Plasma-Membrane–EnrichedCell Fraction

The pcDNA3 plasmids for expression of wild-typePro-MT1-MMP and MTDpro, propeptide-deletedMT1-MMP lacking the entire propeptide domain ofMT1-MMP were constructed by using a two-steppolymerase chain reaction procedure [30,31]. COS-1 cells are a monkey kidney cell line transformedby the T-antigen of simian virus 40. These cells donot express MT1-MMP and thus serve as a negative-control cell line for MT1-MMP [9,31]. COS-1 cellswere transfected with pcDNA3 vector alone (nega-tive control), MTDpro, propeptide-deleted MT1-MMP, MTDpro cDNA, or wild-type MT1-MMP cDNAby using the calcium phosphate method [32]. Trans-fected cells were cultured for 48 h in DMEM con-taining 10% fetal calf serum. The cells wereharvested by scraping, and cell membranes wereprepared from transfected cells after nitrogen cavi-tation as previously described [33]. The post-nuclearsupernatant (770 ×g for 10 min) was collected, andheavy organelles were removed by centrifugationat 6000 ×g for 15 min. The supernatant was centri-fuged at 100 000 ×g for 1 h at 4°C to recover theplasma membrane–enriched cell fraction in the pel-let. Cell membranes were characterized by electronmicroscopy as previously described [33].

Immunoblot Analysis (Western Blotting)

Immunoblot analysis (western blotting) was per-formed according to standard procedures [34,35]. Aspecific polyclonal rabbit Ab directed against a hu-man MT1-MMP peptide was generated as describedabove. Electrophoretic transfer of proteins from 10%polyacrylamide gels to nitrocellulose sheets (Bio-RadLaboratories) was performed in 25 mM Tris, 192 mMglycine, and 20% (v/v) methanol, pH 8.3, at 80–100V in a cold room for 2–4 h. After blocking of non-specific binding sites with 3% skim milk (blotting-grade blocker, nonfat dry milk; Bio-Rad Laboratories),the nitrocellulose blot was incubated with the pri-mary Ab diluted in 10 mM Tris, 150 mM NaCl, 0.05%Tween-20, and 0.5% skim milk, pH 8.0. The second-ary Ab was alkaline phosphatase–conjugated goatanti–rabbit IgG (Sigma Chemical Co.). The second-ary Ab for mAbs was the alkaline phosphatase–con-jugated, affinity-purified goat anti–mouse IgG (H +L) (Bio-Rad Laboratories). A deep blue-purple color,which indicates a positive result, developed whenthe blot was incubated with 0.15 mg/mL 5-bromo-

4-chloro-3-indolyl phosphate and 0.30 mg/mLnitroblue tetrazolium both from Sigma Chemical Co.in alkaline phosphatase color-development buffer(100 mM Tris and 5 mM MgCl2, pH 9.5).

Zymography Analysis

Zymography analysis was performed as previouslydescribed [34–36]. Heat-denatured collagen (porcinegelatin, EIA-grade reagent; Bio-Rad Laboratories) wascopolymerized in 10–12% polyacrylamide gels at afinal concentration of 1 mg/mL. Protein samples ofthe plasma membrane preparation, CM, DT phase,AQ phase, and the resolubilized portion of the pellet(PL) fraction were treated with standard sample buffer[18,19] without reducing agents and heating. Thetreated samples (normalized to the same number ofcells) were loaded onto substrate-copolymerized so-dium dodecyl sulfate (SDS)-polyacrylamide gels. Af-ter electrophoresis, the gels were soaked in 2.5% TritonX-100 to displace SDS and then incubated in 50 mMTricine buffer containing 0.2 M NaCl and 10 mMCaCl2, pH 7.5, at 37°C for 4 d. The gels were thenstained with Coomassie Blue and destained. Becausethe protein substrate (gelatin) is copolymerized in thegel, the whole gel stains blue, except for the areaswhere the proteinases are located. The gelatinase ac-tivities are visualized as transparent (white) bands onthe blue background because the proteinases digestthe substrate in situ.

RESULTSSpecificity of the Antibody Against MT1-MMPActive Peptide

The Ab against the MT1-MMP peptide (160–174)was characterized by ELISA (Figure 1). ELISA plateswere coated with 50 µL of cdMT1-MMP in concen-trations ranging from 0 to 270 nM (Figure 1A) orthe MT1-MMP peptide (160–174) in concentrationsranging from 0 to 110 nM (Figure 1B). The primaryAb was purified by using the nitrocellulose-mem-brane method; its optimal dilution range was from1:200 to 1:1600. The specificity of the Ab was deter-mined by testing its cross-reactivity against humanfibroblast collagenase (MMP-1), gelatinase A (MMP-2), and human neutrophil gelatinase B (MMP-9) pro-teins. At protein concentrations up to 320 nM (50µL/well) and primary Ab dilutions of 1:200 and1:400, no cross reactivity was detected (data notshown). The preimmune serum and IgG did not re-act with the peptide antigen. Propeptide (21–40)concentrations up to 250 nM did not have a posi-tive reaction with the Ab. Therefore, the Ab was veryspecific to MT1-MMP. Attempts to use the MT1-MMP peptide (160–174) [9] as a competitor in theELISA assays were unsuccessful.

The Ab directed against the MT1-MMP peptide(160–174) was further characterized by immunoblotanalysis (Figure 2). The plasma membrane prepara-tions of the parental COS-1 cells and the COS-1 cells

88 LI ET AL.

transfected with the plasmid not containing MT1-MMP cDNA served as negative controls for MT1-MMP. The plasma membrane–enriched fractionsfrom the COS-1 cells transfected with the full-lengthwild-type pro-MT1-MMP and MT1-MMP without thepropeptide domain were used as positive controls.The plasma-membrane fraction (DT phase) and theCM from the two breast cancer cell lines, MDA-MB-231 and BT549, were also examined by immunoblot

analysis. The molecular masses of pro–MT1-MMP andMT1-MMP were calculated to be 63.4 kDa and 57.6kDa, respectively, according to the standard proteinmarkers shown in Figure 2. The MT1-MMP speciesdetected in the DT phase had the same molecularmass as the pro–MT1-MMP band, and species de-tected in the CM phase had the same molecular massas the MT1-MMP band (Figure 2). This result indi-cates that plasma membrane–associated MT1-MMP

Figure 1. Characterization of the Ab against MT1-MMP pep-tide (160–174) by ELISA. ELISA plates were coated with variousconcentrations of cdMT1-MMP (A) and MT1-MMP peptide (160–174) (B). The primary Ab was the rabbit antihuman MT1-MMP

peptide antigen affinity-purified by using the nitrocellulose-membrane method. The Ab dilutions are indicated on eachdiagram.

MT1-MMP IN HUMAN BREAST CANCER 89

was the full-length pro–MT1-MMP, and the solubleform in CM was the mature MT1-MMP. Attempts toimmunize rabbits with an MT1-MMP propeptide-spe-cific sequence (residues 21–40, Ala-Leu-Ala-Ser-Gly-Ser-Ala-Gln-Ser-Ser-Ser-Phe-Ser-Pro-Glu-Ala-Trp-Leu-Gln-NH2) [9] have not been successful.Therefore, we cannot confirm the activation statusof the MT1-MMP species.

The peptide antigen sequence Arg160-Glu-Val-Pro-Tyr-Ala-Tyr-Ile-Arg-Glu-Gly-His-Glu-Lys-Gln174-NH2

shares less than 50% identity with the correspond-ing sequences of the other three membrane-typeMMPs. This peptide sequence was utilized in asearch-and-sequence alignment analysis performedagainst all the most current protein data banks byusing basic local alignment search tool (BLAST) pro-grams (National Center for Biotechnology Informa-tion (Bethesda, MD), http://www.ncbi.nlm.nih.gov),which are used to perform database searches andrigorous statistical analyses for evaluating the sig-nificance of matches [37,38]. The only sequencesidentified were those of human, mouse, and rabbitMT1-MMP. This Ab may cross-react with mouse and

rabbit MT1-MMP, however, it probably does notcross-react with any other types of MT-MMPs orMMPs.

Both Membrane-Associated and Soluble Forms ofMT1-MMP in Breast Cancer Cells and Fibroblasts

The plasma membrane–associated MT1-MMPwas detected by immunoblot analysis (westernblotting, Figure 3). MT1-MMP samples were pre-pared from the plasma-membrane fraction (DTphase) of cells in breast cancer/fibroblast co-cul-ture systems in the absence and presence of ConA.ConA was added to the serum-free CM to stimu-late MT1-MMP expression [13]. The breast cancercells were co-cultured with fibroblasts to deter-mine whether epithelial-stromal cell interactionsupregulate MT1-MMP expression. Figure 3A showsthe detection of MT1-MMP in BT549/HFL-1 co-culture at various BT549:HFL-1 ratios. Figure 3Bshows MT1-MMP detected in MDA-MB-231/HFL-1 co-culture at various MDA-MB-231:HFL-1 ratios.The estimated molecular mass of MT1-MMP de-tected in the DT phase was 63.4 kDa (Figures 2and 3). The results show that ConA stimulatedMT1-MMP expression and that the MT1-MMP pro-tein level was slightly upregulated in co-culture.

Figure 2. Characterization of the Ab directed against MT1-MMP peptide (160–174) by immunoblot analysis. The plasma-membrane preparations of the parental COS-1 cells and theCOS-1 cells transfected with the plasmid without MT1-MMPcDNA are negative controls for MT1-MMP. The plasma mem-brane–enriched fractions from the COS-1 cells transfected withthe full-length wild-type pro-MT1-MMP and MT1-MMP with-out the propeptide domain are the two positive controls. Theplasma-membrane fraction (DT phase) and the CM from thetwo breast cancer cell lines, MDA-MB-231 and BT549, werealso examined by immunoblot analysis. The samples werederived from approximately 104–105 cells per lane on theimmunoblot. The primary Ab was the Ab against MT1-MMPpeptide (160–174) purified by the two-step affinity-chroma-tography method. The Ab dilution was 1:200 to a final con-centration of 5 µg/mL. The secondary Ab dilution was 1:20 000.The molecular masses of pro-MT1-MMP and MT1-MMP werecalculated to be 63.4 kDa and 57.6 kDa, respectively, accord-ing to the standard protein markers.

Figure 3. Detection of the plasma membrane-associatedMT1-MMP by immunoblot analysis (western blotting). MT1-MMP samples were prepared from the plasma-membrane frac-tions (DT phase) of cells in breast cancer/fibroblast co-culturesystems in the absence and presence of ConA. The primary Abwas rabbit anti-human MT1-MMP peptide (160–174) Ab anti-gen-affinity–purified by using the nitrocellulose-membranemethod. The Ab dilution was 1:100 to a final Ab concentrationof 10 µg/mL. The secondary Ab dilution was 1:20 000. (A) Theresults of MT1-MMP in BT549/HFL-1 co-culture at variousBT549:HFL-1 ratios. Each lane on the immunoblot containsplasma membrane from 5500 cells. (B) The results of MT1-MMPMDA-MB-231/HFL-1 co-culture at various MDA-MB-231:HFL-1ratios. Each lane on the immunoblot contains plasma mem-brane from 6300 cells.

90 LI ET AL.

In addition, ConA also induced a very low levelof a 30–35 kDa species that may be a fragment ofcdMT1-MMP.

Low levels of soluble forms of MT1-MMP were de-tected in CM by immunoblot analysis (Figure 4).MT1-MMP samples were collected from CM in breastcancer cell/fibroblast co-culture systems in the ab-sence and presence of ConA. Figure 4A shows MT1-MMP in BT549/HFL-1 co-culture at various BT549:HFL-1 ratios. Figure 4B shows MT1-MMP in MDA-MB-231/HFL-1 co-culture at various MDA-MB-231:HFL-1 ratios. The estimated molecular mass ofthe MT1-MMP detected in the CM was 57.6 kDa.ConA stimulated MT1-MMP expression, and the co-culture system slightly increased the MT1-MMP pro-tein level. ConA also induced a very low level of a25–30 kDa species.

Activation of Pro–MMP-2 by MT1-MMP onthe Cell Surface

Pro–MMP-2, active MMP-2, and pro–MMP-9 indifferent cell fractions in the BT549/HFL-1 co-cul-ture system were visualized by gelatin zymography(Figure 5). Pro-MMP is activated by SDS-induced con-formational change without a decrease in its molecu-

lar mass; therefore, both pro-MMP and active MMPcan be detected by zymography. A small amount ofpro–MMP-2 and a large amount of pro–MMP-9 weresecreted by BT549 cells in the CM fraction. A highlevel of pro–MMP-2 and a low level of pro–MMP-9were secreted by HFL-1 fibroblasts. The identities ofMMP-2 and MMP-9 were confirmed by immunoblotanalysis with mAbs against these two enzymes (datanot shown). Pro–MMP-2 was not seen in the DT, AQ,or PL fractions in the absence of ConA. In the pres-ence of ConA, both pro–MMP-2 and active MMP-2were seen in the DT fraction, and interestingly, activeMMP-2 was seen in the CM, PL, DT, and AQ frac-tions, indicating that pro–MMP-2 was activated onthe cell surface and then released into other fractions.ConA decreased the amount of pro–MMP-9 in the CM.A low level of pro–MMP-9 appeared in the DT phaseupon treatment with ConA, suggesting that pro–MMP-9 may also be associated with the plasma membrane.

Pro–MMP-2, active MMP-2, and pro–MMP-9 in dif-ferent cell fractions in the MDA-MB-231/HFL-1 co-culture system were also visualized by gelatinzymography. As shown in Figure 6 (CM) MDA-MB-231 cells only secreted a small amount of pro–MMP-9. In the absence of ConA, pro–MMP-2 was not seenin the DT or PL fractions, and it was seen only slightlyin the AQ fraction. The conversion of pro–MMP-2 toactive MMP-2 was seen in the presence of ConA, andactive MMP-2 was seen in the CM, PL, DT, and AQfractions. However, ConA decreased the secretion ofpro–MMP-9, and very little or no pro–MMP-9 wasseen in any of the four fractions. In the PL fraction, atrace amount of gelatinolytic activity was seen atabout 50–55 kDa in MDA-MB-231 alone and in MDA-MB-231/HFL-1 co-culture at a ratio of 3:1. The iden-tity of this gelatinase activity remains to bedetermined.

No Pro–MMP-7 Activation by MT1-MMP

To identify the role of MT1-MMP in pro–MMP-7processing, experiments were conducted to deter-mine whether MT1-MMP activates pro-MMP-7 inthe MDA-MB-231 cell culture system and in the pu-rified enzyme system. MDA-MB-231 cells were cho-sen because they produce MT1-MMP, very littlepro–MMP-9, no MMP-2, and no MMP-7. In the ab-sence or presence of ConA, pro–MMP-7 was addedto the cell culture medium at a final concentrationof 10 nM. After 5 h and overnight incubation, thecell medium was collected and concentrated. Acasein zymogram (12% polyacrylamide gel co-po-lymerized with 1 mg/mL bovine casein) was usedto detect the activation of pro–MMP-7 by cellularMT1-MMP. CM, DT, AQ, and PL fractions were ex-amined. Pro–MMP-7 was not activated under theconditions tested (data not shown).

cdMT1-MMP was used to activate pro–MMP-7. Theactivity of cdMT1-MMP was verified by its case-inolytic activity and its ability to cleave a fluorogenic

Figure 4. Detection of the soluble form of MT1-MMP in CMby immunoblot analysis (western blotting). MT1-MMP sampleswere collected from CM in breast cancer cell/fibroblast co-cul-ture systems in the absence and presence of ConA. The pri-mary Ab is the rabbit Ab directed against the human MT1-MMPpeptide (160–174) purified using the two-step affinity method.The Ab dilution is 1:200 to a final Ab concentration of 5 µg/mL.The secondary Ab dilution is 1:20 000. (A) shows the results ofMT1-MMP in BT549/HFL-1 co-culture at various BT549:HFL-1ratios. Each lane on the immunoblot represents highly concen-trated medium derived from a culture of 55 000 cells. (B) showsthe results of MT1-MMP in MDA-MB-231/HFL-1 co-culture atvarious MDA-MB-231:HFL-1 ratios. Each lane on the immuno-blot represents highly concentrated medium derived from aculture of 63 000 cells.

MT1-MMP IN HUMAN BREAST CANCER 91

Figure 5. Detection of pro–MMP-2 and active MMP-2 in vari-ous cell fractions in the BT549/HFL-1 co-culture by gelatinzymography. BT549 and HFL-1 were co-cultured at varoiusBT549:HFL-1 ratios in the absence and presence of ConA. Each

lane on the zymograms contains a sample derived from a cul-ture of 5500 cells. Pro–MMP-2, active MMP-2, and pro–MMP-9were examined.

Figure 6. Detection of pro–MMP-2 and active MMP-2 in dif-ferent cell fractions in the MDA-MB-231/HFL-1 co-culture sys-tem by gelatin zymography. MDA-MB-231 and HFL-1 wereco-cultured at various MBA-MB-231:HFL-1 ratios in the absence

and presence of ConA. Each lane on the zymogram representsa sample derived from a culture of 6300 cells. Pro–MMP-2, ac-tive MMP-2, and pro–MMP-9 are visualized.

92 LI ET AL.

peptide substrate, Mca-Pro-Leu-Gly-Leu-Dpa-Ala-Arg-NH2 (BACHEM, King of Prussia, PA) [39]. Pro–MMP-7 samples (0.5 µM) were incubated with 0.056,0.188, 0.563, 1.875, and 5.625 µM cdMT1-MMP in40 µL of 50 mM Tricine, 0.2 M NaCl, 10 mM CaCl2,and 50 µM ZnSO4, pH 7.5, at 37°C for 5 h. Aliquotsof the reaction mixtures were examined by SDS–poly-acrylamide gel electrophoresis and silver-staininganalysis and by casein zymography. No activationof pro–MMP-7 was observed (data not shown). A posi-tive control experiment demonstrated that pro–MMP-7 was converted to the active form byautocatalysis upon treatment of 0.1M HgCl2 (datanot shown). We concluded that MT1-MMP is not anactivator of pro–MMP-7.

DISCUSSION

We report here that a specific rabbit Ab against aunique human MT1-MMP peptide sequence was pro-duced, purified, and characterized. This peptide se-quence corresponded to residues 160–174 ofMT1-MMP (Arg160-Glu-Val-Pro-Tyr-Ala-Tyr-Ile-Arg-Glu-Gly-His-Glu-Lys-Gln174-NH2), which are locatedin a unique insertion sequence in the catalytic do-main [9]. This Ab was highly specific to MT1-MMPand did not cross-react with other members of theMMP family. It may cross-react with MT1-MMP fromother species such as mouse and rabbit, according toits sequence homology. Using the Ab as a tool, wedemonstrated that MT1-MMP was produced by breastcancer cells and fibroblasts both as an integral mem-brane protein and in soluble forms. The productionof MT1-MMP was stimulated by ConA, and the in-creased levels of MT1-MMP may, in turn, be respon-sible for the ConA-induced activation of pro–MMP-2on the cell surface. Our data are consistent with thereport by Yu et al. [13] that ConA induced the MT1-MMP level necessary to activate pro–MT1-MMP. Therole of MT1-MMP in activating another importantmember of the MMP family, pro–MMP-7, has beentested. Pro–MMP-7 was not activated by breast can-cer cell–associated MMP. Although mercury chlorideinduced autoproteolytic activation of pro–MMP-7,this zymogen was not activated by a tenfold higherconcentration of cdMT1-MMP. Moreover, consistentwith a previous report, we found that MT1-MMP didnot activate pro–MMP-9 [9], demonstrating that MT1-MMP is not a universal activator for all pro-MMPs.

To investigate the expression of MT1-MMP byhuman breast cancer cells and the potential effect ofepithelial-stromal cell interactions on MT1-MMPexpression, breast cancer cell lines BT549 and MDA-MB-231 were each co-cultured with normal humanfetal-lung fibroblasts (HFL-1). The cells were solubi-lized with a nonionic detergent (Triton X-114), andthe soluble material was submitted to phase separa-tion as described previously [23]. Hydrophilic pro-teins were found exclusively in the AQ phase, andintegral membrane proteins with an amphiphilic

nature were recovered in the DT phase. A 63.4-kDaprotein was detected in the DT phase by using Ab tothe MT1-MMP peptide (Figures 2 and 3). This spe-cies may be the pro-MT1-MMP compared with thestandard recombinant pro-MT1-MMP band on Fig-ure 2. A faint band of 30–35 kDa was also detectedin the presence of ConA (Figure 3). This may be anautocleavage product of MT1-MMP.

The major MT1-MMP species detected in the CMby immunoblot analysis had a molecular mass of 57.6kDa, which was the same as the recombinant ma-ture MT1-MMP (Figures 2 and 4). In addition, a smallamount of a 25- to 30-kDa band was also observedin culture media and may be the cdMT1-MMP. Inthe AQ phase, several weak bands were detected atthe molecular masses ranging from 40 to 80 kDa. Atrace amount of MT1-MMP was also detected in theresolubilized fraction of the PL (Li and Sang, unpub-lished results). These observations indicate that MT1-MMP may exist in shed soluble forms. That therewere soluble forms of MT1-MMP may have impor-tant functional implications. Transmembrane do-main–deleted mutants of MT1-MMP activatepro–MMP-2 and degrade extracellular componentssuch as fibronectin, laminin, vitronectin, andproteoglycan [14] as well as interstitial collagens [15].MT1-MMP is a gelatinolytic enzyme and is secretedin a complex with TIMP-2 [40]. Although MT1-MMPhas some gelatinolytic activity, the concentration wastoo low to be detected in our cell-culture system (Fig-ures 5 and 6). In HT-1080 fibrosarcoma cells, ConAenhanced a 43-kDa form of MT1-MMP, and its ap-pearance correlated with pro–MMP-2 processing ac-tivity [41]. Furthermore, the soluble recombinantcdMT1-MMP can activate pro–MMP-2 and it is regu-lated by TIMP-2 and TIMP-3 [29,42].

Generally, pro–MMP-2 was only detected in theCM fraction. Upon treatment with ConA, pro–MMP-2 and active MMP-2 were detected in the DT phase,and active MMP-2 was also detected in all other frac-tions (Figures 5 and 6), indicating that pro–MMP-2is activated on the cell surface and then released intocell medium. The activation of pro–MMP-2 involvestwo steps. MT1-MMP initially cleaves the Asn66-Leu67

peptide bond of pro–MMP-2, and then MMP-2autoproteolytically activates itself [10,42]. Theplasma membrane–dependent pro–MMP-2 activationis initiated by the cleavage of the Asn66-Leu67 pep-tide bond by MT1-MMP to generate an intermediateform [43]. Cell-surface binding concentrates theMMP-2 intermediate form locally to allow auto-proteolytic processing to the fully active form moreeffectively [44]. Generally, pro–MMP-9 activity wasnot detected in the DT phase. Although ConA re-duced the protein level of pro–MMP-9 in the me-dium, it promoted the association of this enzyme toplasma membrane (Figure 5). This result may sug-gest that pro–MMP-9 also forms a complex with MT1-MMP and MMP-2. Pro–MMP-9 can be activated by

MT1-MMP IN HUMAN BREAST CANCER 93

MMP-2 [45]; therefore, it may be recruited by acti-vated MMP-2 to the cell surface.

There is strong evidence that tumor-stroma inter-actions are involved in breast cancer invasion andprogression. In in situ hybridization studies of mRNAlevels of MT1-MMP and other members of the MMPfamily in human breast cancer specimens, MT1-MMPand most other MMPs were localized to the fibro-blasts of tumor stroma of invasive cancers [12]. There-fore, both breast cancer cells and stromal fibroblastscan contribute to the invasive growth and spread ofbreast cancer cells. We have studied the influence ofbreast cancer cell/fibroblast interaction on the levelof MT1-MMP expression. The human fetal-lung fi-broblast cell line HFL-1 was selected for this studybecause the lung is one of the preferred sites for sec-ondary breast-tumor growth. Host (lung) microen-vironmental factors, such as fibroblasts, may interactwith tumor cells reciprocally to produce type IV col-lagenases, their activators, or both to facilitate theextravasation, invasion, and growth of breast cancercells in lungs. Both breast cancer cells and fibroblastsproduce low levels of MT1-MMP protein in cell-cul-ture systems. The expression of this enzyme wasstimulated by ConA and was slightly upregulated byepithelial cancer cell and stromal fibroblast interac-tions, as demonstrated in Figures 3–6. These resultsseem to confirm a report by Polette et al. [46] thatbreast adenocarcinoma cells can induce the expres-sion of MT1-MMP in human fibroblasts, althoughour study did not show high-level MT1-MMPupregulation, as the Polette et al. report did.

Full-length and some truncated MT1-MMP speciescan catalyze the hydrolysis of substrates. The recom-binant deletion mutant of MT1-MMP lacking thetransmembrane domain and the native MT1-MMPform secreted from MDA-MB-231 cells digest ECMsubstrates [15]. Pro–MT1-MMP may be activated byproteolytic removal of the propeptide or by a con-formational switch induced by microenvironmen-tal factors. MT1-MMP contains the RRKR111 sequence,which is a potential recognition sequence for intra-cellular furin-like proteases. Furin is a Golgi-associ-ated proprotein convertase, i.e., a paired basicamino-acid cleaving enzyme [31]. Pro–MT1-MMP canbe activated by furin-induced intracellular cleavageat the Arg111-Tyr bond [14,47]. A form of MT1-MMPisolated from tumor cell lysates has Tyr112 at the Nterminus [48], suggesting that furin may be respon-sible for the activation of MT1-MMP in these cells.An alternative pathway is autocatalytic activation togenerate active MT1-MMP species with N-terminalIle114 [29]. Interestingly, MT1-MMP species contain-ing the propeptide domain also activate pro–MMP-2; therefore, furin-induced cleavage of the propeptidedomain of pro–MT1-MMP is not a prerequisite forpro–MMP-2 activation [31]. Our results also suggestthat, in the plasma-membrane fraction, the MT1-MMP species still has its propeptide domain. Those

results suggest that pro–MT1-MMP in the plasmamembrane may exist in an active conformation thatcan hydrolyze its substrates on the cell surface. Themembrane microenvironment may allow the acti-vation of pro–MT1-MMP by changing its conforma-tion without removing its propeptide domain.

ACKNOWLEDGMENTS

This research was supported in part by grants fromthe Elsa U. Pardee Foundation, the Gustavus andLouise Pfeiffer Research Foundation, and the RGKFoundation to QXAS and from the Deutsche Forsch-ungsgemeinschaft to HT.

REFERENCES

1. Matrisian LM. The matrix-degrading metalloproteinases. Bio Es-says 14:455–463, 1992.

2. Birkedal-Hansen H, Moore WGI, Bodden MK, et al. Matrixmetalloproteinases: A review. Crit Rev Oral Biol Med 4:197–250,1993.

3. Sang QA, Douglas DA. Computational sequence analysis of matrixmetalloproteinases. J Protein Chem 15:137–160, 1996.

4. Douglas DA, Shi YE, Sang QA. Computational sequence analysisof the tissue inhibitor of metalloproteinase family. Minireview. JProtein Chem 16:237–255, 1997.

5. Stetler-Stevenson WG, Aznavoorian S, Liotta LA. Tumor cell in-teractions with the extracellular matrix during invasion and me-tastasis. Ann Rev Cell Biol 9:541–573, 1993.

6. Monteagudo C, Merino MJ, San-Juan J, Liotta LA, Stetler-Stevenson WG. Immunohistochemical distribution of type IVcollagenase in normal, benign, and malignant breast tissue. AmJ Pathol 136:585–592, 1990.

7. Stetler-Stevenson WG. Progelatinase A activation during tumorcell invasion. Invasion Metastasis 14:259–268, 1994–1995.

8. Davies B, Miles DW, Happerfield LC, et al. Activity of type IVcollagenases in benign and malignant breast disease. Br J Can-cer 67:1126–1131, 1993.

9. Sato H, Takino T, Okada Y, et al. A matrix metalloproteinaseexpressed on the surface of invasive tumor cells. Nature 370:61–65, 1994.

10. Kinoshita T, Sato H, Takino T, Itoh M, Akizawa T, Seiki M. Pro-cessing of a precursor of 72-kilodalton type IV collagenase/gelatinase A by a recombinant membrane-type 1 matrixmetalloproteinase. Cancer Res 56:2535–2538, 1996.

11. Okada A, Bellocq J-P, Rouyer N, et al. Membrane-type matrixmetalloproteinase (MT-MMP) gene is expressed in stromal cellsof human colon, breast, and head and neck carcinomas. ProcNatl Acad Sci USA 92:2730–2734, 1995.

12. Heppner KJ, Matrisian LM, Jensen RA, Rodgers WH. Expressionof most matrix metalloproteinase family members in breast can-cer represents a tumor-induced host response. Am J Pathol149:273–282, 1996.

13. Yu M, Sato H, Seiki M, Thompson EW. Complex regulationof membrane-type matrix metalloproteinase expression andmatrix metalloproteinase-2 activation by concanavalin A inMDA-MB-231 human breast cancer cells. Cancer Res55:3272–3277, 1995.

14. Pei D, Weiss SJ. Transmembrane-deletion mutants of the mem-brane-type matrix metalloproteinase-1 process progelatinase Aand express intrinsic matrix-degrading activity. J Biol Chem271:9135–9140, 1996.

15. Ohuchi E, Imai K, Fujii Y, Sato H, Seiki M, Okada Y. Membranetype 1 matrix metalloproteinase digests interstitial collagens andother extracellular matrix macromolecules. J Biol Chem272:2446–2451, 1997.

16. Knauper V, Will H, Lopez-Otin C, et al. Cellular mechanisms forhuman procollagenase-3 (MMP-13) activation—Evidence thatMT1-MMP (MMP-14) and gelatinase A (MMP-2) are able to gen-erate active enzyme. J Biol Chem 271:17124–17131, 1996.

17. Sato H, Seki M. Membrane-type matrix metalloproteinases (MT-MMPs) in tumor metastasis. J Biochem (Tokyo) 119:209–215,1996.

18. Sang Q-X, Birkedal-Hansen H, Van Wart HE. Proteolytic and non-proteolytic activation of human neutrophil progelatinase. BiochimBiophys Acta 1251:99–108, 1995.

94 LI ET AL.

19. Sang QA, Bodden MK, Windsor LJ. Activation of humanprogelatinase A by collagenase and matrilysin and activation ofprocollagenase by matrilysin. J Protein Chem 15:243–253, 1996.

20. Crawford HC, Matrisian LM. Tumor and stromal expression ofmatrix metalloproteinases and their role in tumor progression.Invasion Metastasis 14:234–245, 1994–1995.

21. Nicolson GL. Cancer progression and growth: Relationship ofparacrine and autocrine growth mechanisms to organ prefer-ence of metastasis. Exp Cell Res 204:171–180, 1993.

22. Azzam HS, Arand G, Lippman ME, Thompson EW. Associationof MMP-2 activation potential with metastatic progression inhuman breast cancer cell lines independent of MMP-2 produc-tion. J Natl Cancer Inst 85:1758–1764, 1993.

23. Bordier C. Phase separation of integral membrane proteins inTriton X-114 solution. J Biol Chem 256:1604–1607, 1981.

24. Monsky WL, Lin C-Y, Aoyama A, et al. A potential markerprotease of invasiveness, seprase, is localized on invadopodiaof human malignant melanoma cells. Cancer Res 54:5702–5710, 1994.

25. Puente XS, Pendas AM, Llano E, Velasco G, Lopez-Otin C. Mo-lecular cloning of a novel membrane-type matrix metal-loproteinase from a human breast carcinoma. Cancer Res56:944–949, 1996.

26. Harlow E, Lane D. Antibodies—A Laboratory Manual. Cold SpringHarbor Laboratory Press, Cold Spring Harbor, New York, 1988.

27. Smith DE, Fisher PA. Identification, developmental regulation,and response to heat shock of two antigenically related forms ofa major nuclear envelope protein in Drosophila embryos: Appli-cation of an improved method for affinity purification of anti-bodies using polypeptides immobilized on nitrocellulose blots. JCell Biol 99:20–28, 1984.

28. Stetler-Stevenson WG, Krutzsch HC, Wacher MP, Margulies IMK,Liotta LA. The activation of human type IV collagenase proen-zyme—Sequence identification of the major conversion productfollowing organomercurial activation. J Biol Chem 264:1353–1356, 1989.

29. Lichte A, Kolkenbrock H, Tschesche H. The recombinant cata-lytic domain of membrane-type matrix metalloproteinase-1 (MT1-MMP) induces activation of progelatinase A and progelatinase Acomplexed with TIMP-2. FEBS Lett 397:277–282, 1996.

30. Zucker S, Conner C, DiMassmo BI, et al. Thrombin induces theactivation of progelatinase A in vascular endothelial cells. Physi-ologic regulation of angiogenesis. J Biol Chem 270:23730–23738, 1995.

31. Cao J, Rehemtulla A, Bohou W, Zucker S. Membrane type ma-trix metalloproteinase 1 activates pro-gelatinase A without furincleavage of the N-terminal domain. J Biol Chem 271:30174–30180, 1996.

32. Cao J, Sato H, Takino T, Seiki M. The C-terminal region of mem-brane type matrix metalloproteinase is a functional transmem-brane domain required for pro-gelatinase A activation. J BiolChem 270:801–805, 1995.

33. Zucker S, Wieman JM, Lysik RM, et al. Enrichment of collagenand gelatin degrading activities in the plasma membranes ofhuman cancer cells. Cancer Res 47:1608–1614, 1987.

34. Sang Q-X, Stetler-Stevenson WG, Liotta LA, Byers SW. Identi-fication of type IV collagenase in testicular cell culture: Influ-ence of peritubular-Sertoli cell interactions. Biol Reprod43:956–964, 1990.

35. Sang Q-X, Thompson EW, Grant D, Stetler-Stevenson WG, ByersSW. Soluble laminin and arginine-glycine-aspartic acid contain-ing peptides differentially regulate type IV collagenase messen-ger RNA, activation, and localization in testicular cell culture.Biol Reprod 45:387–394, 1991.

36. Sang Q-X, Dym M, Byers SW. Secreted metalloproteinases intesticular cell culture. Biol Reprod 43:946–955, 1990.

37. Altschul SF, Gish W, Miller W, Myers EW, Lipman DJ. Basic localalignment search tool. J Mol Biol 215:403–410, 1990.

38. Altschul SF, Madden TL, Schaffer AA, et al. Gapped BLAST andPSI-BLAST: A new generation of protein database search pro-grams. Nucleic Acids Res 25:3389–3402, 1997.

39. Knight CG, Willenbrock F, Murphy G. A novel coumarin-labeledpeptide for sensitive continuous assays of the matrixmetalloproteinases. FEBS Lett 296:263–266, 1992.

40. Imai K, Ohuchi E, Aoki T, et al. Membrane-type matrixmetalloproteinase 1 is a gelatinolytic enzyme and is secreted in acomplex with tissue inhibitor of metalloproteinases 2. CancerRes 56:2707–2710, 1996.

41. Lohi J, Lehti K, Westermarck J, Kahari V-M, Keski-Oja J. Regula-tion of membrane-type matrix metalloproteinase-1 expressionby growth factors and phorbol 12-myristate 13-acetate. Eur JBiochem 239:239–247, 1996.

42. Will H, Atkinson SJ, Butler GS, Smith B, Murphy G. The solublecatalytic domain of membrane type 1 matrix metalloproteinasecleaves the propeptide of progelatinase A and initiatesautoproteolytic activation—Regulation by TIMP-2 and TIMP-3. JBiol Chem 271:17119–17123, 1996.

43. Strongin AY, Marmer BL, Grant GA, Goldberg GI. Plasma mem-brane-dependent activation of the 72-kDa type IV collagenase isprevented by complex formation with TIMP-2. J Biol Chem268:14033–14039, 1993.

44. Sato H, Takino T, Kinoshita T, et al. Cell surface binding andactivation of gelatinase A induced by expression of membrane-type-1-matrix metalloproteinase (MT1-MMP). FEBS Lett 385:238–240, 1996.

45. Fridman R, Toth M, Pena D, Mobashery S. Activation ofprogelatinase B (MMP-9) by gelatinase A (MMP-2). Cancer Res55:2548–2555, 1995.

46. Polette M, Gilles C, Marchand V, Seiki M, Tournier JM, BirembautP. Induction of membrane-type matrix metalloproteinase 1 (MT1-MMP) expression in human fibroblasts by breast adenocarcinomacells. Clin Exp Metastasis 15:157–163, 1997.

47. Sato H, Kinoshita T, Takino T, Nokayama K, Seiki M. Activationof a recombinant membrane type 1-matrix metalloproteinase(MT1-MMP) by furin and its interaction with tissue inhibitor ofmetalloproteinases (TIMP)-2. FEBS Lett 393:101–104, 1996.

48. Strongin AY, Collier I, Bannikov G, Marmer BL, Grant GA,Goldberg GI. Mechanism of cell surface activation of 72-kDatype IV collagenase: Isolation of the activated form of the mem-brane metalloprotease. J Biol Chem 270:5331–5338, 1995.