Journal of Cell Science 113, 2909-2921 (2000) 2909 Printed in Great Britain © The Company of Biologists Limited 2000 JCS1511

Mouse keratinocytes immortalized with large T antigen acquire α3β1 -dependent secretion of MMP-9/gelatinase B

C. Michael DiPersio1,2,*, Michael Shao1, Lara Di Costanzo2, Jordan A. Kreidberg3 and Richard O. Hynes1 1Howard Hughes Medical Institute, Center for Cancer Research and Department of Biology, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, USA 2Center for Cell Biology and Cancer Research, Albany Medical College, Albany, New York 12208, USA 3Department of Medicine, Children’s Hospital and Department of Pediatrics, Harvard Medical School, Boston, Massachusetts 02115, USA *Author for correspondence at address 2 (e-mail: [email protected])

Accepted 12 June; published on WWW 20 July 2000

SUMMARY

Remodeling of the during tissue cells derived from wild-type mice, but not in MK cells development, wound repair and tumor cell invasion derived from α3-null mice. Reconstitution of α3β1 depends on the coordinated regulation of cell adhesion expression in α3-null MK cells through transfection with receptors, matrix and enzymes that proteolyse the the α3 subunit restored MMP-9 secretion, indicating an extracellular matrix. Integrin α3β1 is a major receptor on α3β1-dependent pathway for MMP-9 production. α3β1- epidermal keratinocytes for laminin-5 in the cutaneous dependent expression of MMP-9 was associated with the basement membrane and is required for normal basement immortalized phenotype, since nonimmortalized, primary membrane organization during skin development. α3β1 is keratinocytes required soluble growth factors, but not also expressed at high levels in the majority of adherent α3β1, for efficient expression of MMP-9. Our results transformed cells and in most tumors, and it could have suggest that an α3β1-independent pathway(s) for MMP-9 similar roles in extracellular matrix remodeling during production is suppressed in keratinocytes immortalized tumorigenesis and cell invasion. In the present study, we with large T antigen, and that an α3β1-dependent pathway show that α3β1 expression is required in immortalized is required for sustained production of MMP-9 in the mouse keratinocytes (MK) for the production of the matrix absence of other pathways. metalloproteinase MMP-9/gelatinase B, an MMP that is coexpressed with α3β1 in epithelial cell carcinomas and during wound healing, and contributes to the invasive Key words: Integrin, Matrix metalloproteinase, Extracellular matrix, potential of some tumor cells. MMP-9 was expressed in MK Keratinocyte

INTRODUCTION the repertoire of expressed by a cell determines its ability to invade an ECM that it encounters. Upon binding to Cell-mediated remodeling of the extracellular matrix (ECM) extracellular ligands, integrins can initiate distinct ‘outside-in’ during tissue development, wound healing, tumor cell invasion signaling events that affect downstream functions (Clark and and other tissue remodeling events requires extracellular Brugge, 1995; Schwartz et al., 1995), including cell migration proteinases that process ECM proteins during matrix assembly and invasion. Signal transduction functions of pre-existing or degrade existing ECM at the invading/migrating cell front integrins can be modulated during cell differentiation (Adams (Werb, 1997). The matrix metalloproteinases (MMPs) are and Watt, 1990) and transformation (Tapley et al., 1989; secreted as inactive proenzymes that are activated in the Schwartz et al., 1996; Rosales and Juliano, 1996; Sanders et pericellular environment through their own proteolytic al., 1998; Ruoslahti, 1999). cleavage by other MMPs or proteinases (Werb, 1997). Epidermal keratinocytes produce many of the proteinases Coupling the expression and activation of MMPs with the cell- involved in ECM remodeling during wound healing, squamous adhesion functions of ECM receptors is a potential mechanism cell carcinoma and pathological conditions of the skin (Kähäri for temporal and spatial regulation of ECM remodeling and Saarialho-kere, 1997; Johnsen et al., 1998; Westermarck (Huhtala et al., 1995; Brooks et al., 1996; Pilcher et al., 1997; and Kähäri, 1999). MMP-9/gelatinase B appears to be Lochter et al., 1999; Pozzi et al., 2000). Integrins are αβ particularly important for keratinocyte migration during re- heterodimers that function as the major receptors for cell epithelialization of cutaneous wounds (Salo et al., 1994; Okada adhesion to the ECM (Hynes, 1992). Most cell types express et al., 1997; Madlener et al., 1998) and invasion of tumor cells several integrins with distinct ligand-binding specificities, and (Westermarck and Kähäri, 1999; Ramos-DeSimone et al., 2910 C. M. DiPersio and others

1999). A potential mechanism for keratinocytes to acquire mutation as described previously (DiPersio et al., 1997). The presence and/or maintain proteolytic potential is through integrin- of at least one copy of the H-2Kb-tsA58 transgene was confirmed by dependent signaling pathways; however, interactions between PCR using a forward primer (AGCGCTTGTGTCGCCATTGTATTC) integrins and MMP-9 are poorly understood. and a reverse primer (GTAACACCACAGAAGTAAGGTTCC) that Integrin α3β1 is expressed at high levels by basal produced a PCR product of approximately 1000 base pairs in the keratinocytes of the epidermis, where it is a receptor for following PCR reaction: 95°C, 1 minute; 58°C, 2 minutes; 70°C, 3 minutes; 38 cycles. laminin-5 in the basement membrane that separates the Primary keratinocytes were prepared from neonatal mice as epidermis from the dermis (Carter et al., 1991; Delwel et al., described previously (DiPersio et al., 1997; Hodivala-Dilke et al., 1994). α3β1-deficient mice display basement membrane 1998). Keratinocyte growth medium consisted of Eagle’s Minimum disorganization at the dermal-epidermal junction in skin Essential Medium (EMEM; BioWhittaker, Walkersville, MD, USA) (DiPersio et al., 1997; Hodivala-Dilke et al., 1998), as well as supplemented with 4% fetal bovine serum (FBS) (Intergen, Purchase, in kidney and lung (Kreidberg et al., 1996), suggesting a NY, USA) from which Ca2+ had been chelated. Growth medium was µ generally important role for this integrin in ECM assembly supplemented with 0.05 mM CaCl2, 0.4 g/ml hydrocortisone α3β1 (Calbiochem, La Jolla, CA, USA), 5 µg/ml insulin (Sigma, St Louis, during development. is also highly expressed during −10 cutaneous wound healing (Larjava et al., 1993b; Watt and MO, USA), 10 M cholera toxin (ICN Biomedicals, Inc., Costa Mesa, CA, USA), 10 ng/ml epidermal growth factor (EGF; Gibco BRL), Hertle, 1994; Salo et al., 1994;) and in most metastatic 2×10−9 M T3 (Sigma), 100 i.u./ml penicillin and 100 µg/ml carcinomas (Natali et al., 1993; Bartolazzi et al., 1994; streptomycin (Gibco BRL). Cell cultures from one wild-type mouse and Patriarca et al., 1998), suggesting that it may also have a role one α3-null mouse were expanded for approximately 15 passages at the in ECM remodeling during these processes. In the current permissive temperature for large T antigen function (33°C) in study, we established conditionally immortalized mouse keratinocyte growth medium supplemented with 10 units/ml IFNγ keratinocyte (MK) cell lines derived from wild-type or α3β1- (Genzyme, Cambridge, MA, USA, or Gibco BRL). Individual wild- deficient mice. We show that secretion of MMP-9/gelatinase B type or α3-null cell lines were then cloned from each culture by limiting by the MK cell lines was dependent on the presence of dilution and were expanded under permissive conditions (33°C, 10 γ endogenous or transfected α3. α3β1-dependent secretion of units/ml IFN ) for approximately five more passages. MK cell lines MMP-9 was associated with the immortalized phenotype, since were maintained at 33°C, 8% CO2, in keratinocyte growth medium (without cholera toxin) supplemented with 10-20 units/ml IFNγ. freshly isolated primary keratinocytes secreted MMP-9 independently of α3β1 expression. We present evidence that MK cell growth assays immortalization of mouse keratinocytes with large T antigen 14 distinct MK lines (seven wild-type and seven α3-null) were tested results in suppression of certain signaling pathway(s) for for temperature-sensitive growth. MK cultures were trypsinized and MMP-9 production, and that integrin α3β1 is required for counted using Trypan Blue exclusion, then seeded onto culture plates sustained production of MMP-9 in the absence of these other at a density of 50 cells/cm2. Cells were cultured overnight under pathways. permissive conditions (33°C, 10 units/ml IFNγ), then either kept under permissive conditions or shifted to nonpermissive conditions (39°C, no IFNγ) and cultured for 8 days to allow colony formation. Colonies were fixed with 4% formaldehyde and stained with 0.1% Crystal Violet. MATERIALS AND METHODS Preparation of laminin-5-rich ECM from SCC-25 cells Antibodies Laminin-5-rich ECM was prepared from the human squamous cell Rabbit antisera against the cytoplasmic domains of the β1 and α3 carcinoma line SCC-25 (Rheinwald and Beckett, 1981), as described integrin subunits were described previously (Marcantonio and Hynes, previously (Xia et al., 1996). Briefly, SCC-25 cells were grown on 1988; DiPersio et al., 1995). Rabbit antisera against the α5 and αv tissue culture plates in DMEM:HAM’S F-12 (1:1 mix; BioWhittaker), subunits were purchased from Chemicon International (Temecula, 10% FBS, 0.4 µg/ml hydrocortisone for several days until confluent, CA, USA). Monoclonal antibody P1B5 against the human α3 subunit then removed with 0.05% trypsin, 1 mM EDTA in PBS. Plates coated was purchased from Gibco BRL (Gaithersburg, MD, USA). with the laminin-5-rich ECM were treated with 0.5 mg/ml soybean Monoclonal antibody GoH3 against the murine α6 subunit was trypsin inhibitor (Sigma) in PBS, then blocked with 1 mg/ml BSA in purchased from Immunotech (Westbrook, ME, USA). Goat antiserum PBS. to MMP-9 (M-17) was purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA). Rabbit antiserum against keratin 14 was a Iodination and immunoprecipitation of integrins kind gift from Dr E. Fuchs (University of Chicago, Chicago, IL, MK cells were seeded onto 10 cm plates coated with denatured USA), and rabbit antisera against the α3 and γ2 subunits of laminin- collagen or laminin-5-rich ECM and grown for 2 days under 5 were a generous gift from Dr G. Meneguzzi (Faculte de Medecine, permissive conditions (33°C, 10 units/ml IFNγ). Cells were then either Nice, France). kept under permissive conditions or shifted to nonpermissive conditions (39°C, no IFNγ) and grown for an additional 24 hours. Establishment of conditionally immortalized MK cell lines Monolayers were surface-labeled with 0.5 mCi [125I]NaI (New To generate conditionally immortalized cell lines from wild-type and England Nuclear, Boston, MA, USA) using the lactoperoxidase- α3β1-deficient keratinocytes, we exploited the ImmortoMouse glucose oxidase method, and cell lysates were prepared in octyl-β-D- (Charles River Laboratories, Wilmington, MA, USA), which glucopyranoside detergent buffer as described previously (DiPersio et expresses a temperature-sensitive mutant of SV40 large T antigen al., 1997). 150 µg of were immunoprecipitated with anti- (tsA58) under the control of the interferon γ (IFNγ)-inducible H-2Kb integrin antibodies, as described (DiPersio et al., 1997; Hodivala- promoter (Jat et al., 1991). Mice heterozygous for the α3-null Dilke et al., 1998). Rabbit polyclonal antisera were mutation (Kreidberg et al., 1996) were bred with the Immortomouse immunoprecipitated using protein A-sepharose (Pharmacia LKB, to generate α3+/− mice that contained the H-2Kb-tsA58 transgene. Piscataway, NJ, USA); rat mAb GoH3 was immunoprecipitated using These mice were then mated and primary keratinocytes were isolated goat anti-rat IgG-agarose (Sigma). Nonreducing SDS-PAGE was from the offspring, which were genotyped by PCR for the α3-null performed on 5% polyacrylamide gels. Integrin α3β1 regulates MMP-9 expression 2911

Immunofluorescence gels were stained with Coomassie Blue and destained in 10% MK cells were seeded onto glass coverslips coated with laminin-5- methanol, 5% acetic acid. Proteins with gelatinolytic activity were rich ECM and cultured for 2 days under permissive conditions. Cells revealed as clear bands on a blue background. Purified murine were fixed in 4% paraformaldehyde for 10 minutes, permeabilized in MMP-9, or a mixture of human MMP-9 and MMP-2 (Chemicon 0.5% IGEPAL/PBS for 10 minutes, and blocked in 10% normal goat International), served as a positive control. serum (NGS) in PBS. For immunofluorescence of α6 integrin, cells were stained with rat mAb GoH3 (1:100 dilution in 10% NGS), Zymography of full thickness skin cultures and primary followed by fluorescein-conjugated goat anti-rat IgG (TAGO keratinocytes BioSource, Camarillo, CA, USA). Filamentous actin was stained Cultures of primary keratinocytes and full thickness skin were prepared − using fluorescein-conjugated phalloidin (Sigma) at 1:5000 dilution. from a single neonatal litter consisting of five α3+/+, five α3+/ and five − − Representative fields were photographed on a Zeiss Axiophot α3 / mice (lacking the H-2Kb-tsA58 transgene). For each skin, a portion microscope (Thornwood, NY, USA). of approx. 1.5×1.5 cm was minced into smaller pieces for skin culture, and the remaining skin was used to prepare keratinocyte cultures. Skin Transfection of MK cells with recombinant α3 integrin slices were cultured overnight on 24-well plates in keratinocyte growth subunit medium, rinsed several times with PBS, then cultured for an additional Human α3 cDNA was a gift from Dr M. Hemler (Dana-Farber Cancer 48 hours in 0.3 ml serum-free medium. Primary cells were seeded onto Institute, Boston, MA, USA). The α3 cDNA used in these studies SCC-25 ECM at 33°C, then grown for an additional 48 hours in serum- consisted of the human α3 extracellular domain fused to the chicken free medium. Culture media from skin and primary cells were assayed α3A cytoplasmic domain at a conserved BclI restriction site near the for MMPs by zymography or western blot. carboxy-terminal portion of the transmembrane domain; the α3A For growth factor induction of MMP-9, primary cells were starved cytoplasmic domain is highly conserved (Takada et al., 1991). An XbaI for 1 day in serum-free EMEM/0.05 mM CaCl2 without growth factor restriction fragment encompassing the α3A cDNA was subcloned into or hormonal supplements, then cultured for an additional 2 days in the the XbaI polylinker site in pcDNΑ3.1/Zeo(+) (Invitrogen, San Diego, same medium that was either fully supplemented, supplemented only CA, USA) downstream of the cytomegalovirus promoter, and junctions with 10 ng/ml EGF, or unsupplemented. MMPs were concentrated were confirmed by DNA sequencing. The α3 expression plasmid, or from culture media on gelatin-agarose beads (see below). pcDNΑ3.1/Zeo as a control, was introduced into the α3-deficient MK- 5.4.6 cell line by calcium phosphate-mediated transfection. Western blotting Transfection mix contained 25 µg plasmid DNA/ml in 125 mM CaCl2, Cell lysates were prepared from primary keratinocytes or MK cells in Hepes-buffered saline (140 mM NaCl, 5 mM KCl, 0.75 mM Na2HPO4, RIPA buffer (1% Triton-X-100, 1% sodium deoxycholate, 0.1% SDS, 6 mM dextrose, 25 mM Hepes). MK-5.4.6 cells transfected with full- 158 mM NaCl, 10 mM Tris, pH 7.3, 1 mM EGTA, 2 mM PMSF, 12.5 length human α3 using lipofectamine (Gibco BRL) produced similar µg/ml leupeptin, 15 µg/ml aprotinin) and quantitated using a BCA results in zymography experiments (V. Iyer and C. M. DiPersio, Protein Assay Kit (Pierce, Rockford, IL, USA). Following cell unpublished). Stable transfectants were selected with 200 µg/ml extractions, ECM fractions were scraped from plates into reducing Zeocin (Invitrogen), and approximately 50 Zeocin-resistant colonies SDS-sample buffer (2% SDS, 80 mM Tris, pH 6.8, 2 mM EDTA, 100 were pooled. 10,000 cells expressing high levels of α3 were sorted by mM dithiothreitol, 2 mM PMSF, 12.5 µg/ml leupeptin, 15 µg/ml FACS with the mAb P1B5 against human α3 as primary antibody aprotinin). For skin lysates, whole skin was minced and lysed in (1:500 dilution) and fluorescein-conjugated goat anti-mouse as reducing SDS-sample buffer, and cornified layers were removed by secondary antibody (TAGO BioSource; 1:1000 dilution), then centrifugation. Protein preparations were subject to nonreducing 10% expanded under permissive culture conditions. Greater than 90% of the SDS/PAGE for α3 integrin and keratin 14 blots (5 µg/lane) or presorted pool expressed transfected α3. reducing 7% SDS/PAGE for laminin-5 blots (25 µg/lane), then transferred to 0.2 µm PVDF membranes (Bio-Rad, Hercules, CA, Gelatin zymography of MK cell culture media USA). Membranes were blotted with primary antisera at the following Equal numbers of MK cells were attached to laminin-5-rich ECM dilutions: anti-α3A integrin, 1:500; anti-keratin 14, 1:1000; anti- overnight, then rinsed and fed with serum-free medium that was laminin-5 (α3 or γ2 subunit), 1:500. Peroxidase (HRP)-conjugated otherwise fully supplemented, as described above. For the goat anti-rabbit IgG (Pierce) was used as secondary antibody at experiments in Fig. 5B, 4.3×104 cells/cm2 were plated; for all other 1:15,000 dilution. Chemiluminescence was performed with the experiments, 2.8×104 cells/cm2 were plated. Cells were cultured for SuperSignal Kit (Pierce). an additional 48 hours under permissive conditions (33°C, 10 units/ml For western blotting of MMP-9, gelatinases were concentrated IFNγ) or nonpermissive conditions (39°C, no IFNγ). Culture media from culture media by binding to gelatin-agarose beads (Sigma). were collected for zymography, and lysates were prepared from the Medium from MK cell cultures (2 ml), primary keratinocytes (1 ml) cell monolayer for western blots (see below). For some experiments or skin cultures (150 µl) was incubated overnight at 4°C with 100 µl MMPs were concentrated from culture medium by binding to gelatin- of gelatin-agarose. Control incubations included purified murine agarose beads (see below). Equal volumes of culture medium were MMP-9 or human MMP-2/MMP-9 mix in EMEM. Agarose beads mixed with zymography sample buffer (final: 2.25% SDS, 9% were recovered by centrifugation and eluted in electrophoresis sample glycerol, 45 mM Tris, pH 6.8, Bromophenol Blue) and resolved by buffer. Samples were immunoblotted with goat polyclonal antiserum nonreducing SDS/PAGE on 10% polyacrylamide gels impregnated against MMP-9 (1:200 dilution) followed by HRP-conjugated anti- with 1 mg/ml gelatin (Sigma). Following electrophoresis, gels were goat IgG (Santa Cruz Biotechnology) at 0.4 µg/ml. soaked in 2.5% Triton-X-100 to replace SDS, washed twice with water, then incubated at 37°C for 10-24 hours in MMP activation buffer (50 mM Tris, pH 8.0, 5 mM CaCl2). In control experiments, RESULTS calcium was replaced with 10 mM EDTA in the activation buffer. For inhibitor studies, 5 mM PMSF (Sigma; isopropanol as solvent) or 10 Establishment of conditionally immortalized mouse mM 1,10-phenanthroline (Sigma; methanol as solvent), or solvent keratinocyte (MK) cell lines from wild-type or α3β1- only, was added to the Triton and activation buffers. For activation studies, samples were preincubated with 1.5 mM p- deficient skin aminophenylmercuric acetate (APMA; Sigma), or solvent only We established clonal, mouse keratinocyte (MK) cell lines (DMSO), at 37°C for 6 hours prior to electrophoresis. Zymography that were either wild-type (MK+/+) or homozygous for the 2912 C. M. DiPersio and others

Fig. 1. The conditionally immortalized MK cell lines show temperature-sensitive growth. MK cells were plated at low density and cultured for 7 days under either permissive (33°C) or Fig. 2. Integrin expression on MK+/+ and MK−/− cells. The MK-1.16 nonpermissive (39°C) conditions for large T antigen function, then (MK+/+) cell line or the MK-5.4.6 (MK−/−) cell line were cultured at fixed and stained with Crystal Violet to assay colony growth. the permissive (33°C) or nonpermissive (39°C) temperature for large Representative results are shown for one cell line derived from wild- T antigen then surface-labeled with 125I, and cell lysates were type keratinocytes (MK+/+) and for one cell line derived from α3β1- immunoprecipitated with antibodies against specific integrin − − deficient keratinocytes (MK / ). Similar results were obtained for subunits. Immunoprecipitations were performed with preimmune seven distinct wild-type lines and seven distinct α3β1-deficient lines. serum (preim.), with antisera specific for β1, α3A, α3B, αv or α5, or with the monoclonal antibody GoH3 against α6, as indicated. Molecular mass markers are indicated.

α3-null mutation (MK−/−) by conditional immortalization of primary keratinocytes with the temperature-sensitive tsA58 Integrin expression in the MK cell lines is similar to mutant of SV40 large T antigen (Jat et al., 1991; see Materials that in primary keratinocytes and Methods for details). Cell lines conditionally Throughout this report, we will refer to wild-type MK cell lines immortalized with thermolabile large T antigen mutants grow as MK+/+ and to α3-null MK cell lines as MK−/−. Cell surface- robustly when cultured at the permissive temperature for iodination and immunoprecipitation of integrins in an MK+/+ large T antigen (33°C or 34°C) but show reduced growth at line (MK-1.16) and an MK−/− line (MK-5.4.6) showed that the non-permissive temperature (39°C) (Banks-Schlegel and expression of other integrins was not altered dramatically in Howley, 1983; Zaret et al., 1988; Jat and Sharp, 1989; the α3β1-deficient MK−/− cells. Identical results were obtained Morgan et al., 1994; Lefebvre et al., 1995). We compared with cells cultured on laminin-5-rich ECM (Fig. 2) or collagen- seven independent MK+/+ lines and seven independent coated plates (not shown). α3Aβ1 integrin was abundant in MK−/− lines for temperature-sensitive colony growth (see MK+/+ cells (Fig. 2, upper gel), since both subunits were Materials and Methods). All 14 lines formed colonies under coimmunoprecipitated with antiserum against either the β1 permissive conditions (33°C, plus 10 units/ml IFNγ), but subunit (approx. 115 kDa) or the α3A subunit (approx. 150 failed to form colonies under nonpermissive conditions kDa); we did not detect the α3B cytoplasmic domain variant (39°C, no IFNγ), indicating that cell growth was dependent (de Melker et al., 1997). As expected, α3A was completely on functional large T antigen. Representative examples are absent from MK−/− cells, and total β1 integrin was greatly shown in Fig. 1. Primary mouse keratinocytes failed to form reduced (Fig. 2, lower gel), indicating that α3Aβ1 is the major colonies under these conditions (not shown). Neither the β1 integrin in MK+/+ cells. We also determined whether either omission of IFNγ at 33°C, nor the addition of IFNγ at 39°C, the lack of α3β1 or the immortalization process altered the affected colony formation (data not shown), showing that expression of other integrins found on the MK cells. Very low IFNγ did not directly affect cell growth. Although the H-2Kb levels of α5β1 and moderate levels of an αv integrin, probably promoter that drives expression of the tsA58 is induced αvβ5 (Adams and Watt, 1991; Watt and Hertle, 1994), were by IFNγ, it is still active at lower levels in various cell types detected in both MK+/+ and MK−/− cells. The anti-α6 antibody in the absence of IFNγ (Jat et al., 1991). Therefore, it was not GoH3 immunoprecipitated from both MK+/+ and MK−/− cells surprising that IFNγ was dispensable for the expression of a series of bands expected for the α6 subunit (approx. 140 kDa) functional large T antigen in MK cells grown at 33°C. and proteolytic fragments of the associated β4 subunit (see Nevertheless, IFNγ was included at 10-20 units/ml during DiPersio et al., 1997, and references therein). Although we did routine culture of MK cell lines to maximize expression of not directly assay α2β1 expression, the low level of β1 integrin the H-2Kb-tsA58 transgene. remaining in MK−/− cells suggests that α2β1 is considerably Integrin α3β1 regulates MMP-9 expression 2913 less abundant than α3β1 in MK+/+ cells, consistent with our subunits of laminin-5 (Fig. 4). Both MK+/+ and MK−/− cells previous findings in primary mouse keratinocytes (Hodivala- expressed the unprocessed 190-200 kDa α3 subunit, together Dilke et al., 1998). For both cell lines, integrin expression was with previously described 165 kDa and 145 kDa processed similar under permissive and nonpermissive conditions for forms (Marinkovich et al., 1992) (Fig. 4A, LM-5: α3). In some large T antigen (Fig. 2, compare 33°C and 39°C). Furthermore, experiments, subtle differences in abundance of the fully the integrin expression patterns in the MK+/+ and MK−/− cell processed α3 chain were evident between MK+/+ and MK−/− lines were similar to those for primary keratinocytes isolated cells. Both cell lines also expressed the 155 kDa γ2 precursor; from wild-type and α3-null mice, respectively (Hodivala-Dilke the 105 kDa processed form of γ2 was not detected (Fig. 4A, et al., 1998); however, we cannot rule out functional changes LM-5: γ2). Immunoblot patterns for the α3 and γ2 subunits in existing integrins induced by immortalization or by absence were similar in MK+/+ and MK−/− cell cultures, whether we of α3β1. compared laminin-5 from total cell lysates or from the ECM fraction (Fig. 4B); for both cell lines, processed forms of α3 − − MK+/+ and MK / cell lines display differences in were slightly more abundant in the ECM fraction. MK+/+ and actin cytoskeletal structure when cultured on MK−/− cells produced comparable amounts of either α3 or γ2 laminin-5-rich ECM subunit, indicating equivalent laminin-5 expression levels. In addition to retaining an integrin repertoire typical of Since epithelial cell transformation is often associated with keratinocytes, the MK cell lines retained other characteristics changes in expression or processing of laminin-5 (Ryan et al., of keratinocytes, including expression of keratin 14 (see Fig. 1994; Gianelli et al., 1997), we compared laminin-5 expression 6) and involucrin (not shown). When MK+/+ or MK−/− cell lines between the MK cell lines and freshly isolated primary were grown to confluence on laminin-5-rich ECM, each line keratinocytes to identify possible changes associated with displayed a ‘cobblestone’ morphology, typical of epithelial cell immortalization. Laminin-5 expression levels varied only cultures (Fig. 3A,B). However, at lower cell densities MK−/− slightly among nine independent primary cultures; blots for a cultures contained a higher proportion of poorly spread cells representative culture are shown in Fig. 4A (wt). Laminin-5 than did MK+/+ cultures. MK−/− cells transfected with α3 expression and processing were similar in the MK cell lines showed restored cell spreading (not shown). Phalloidin- and normal keratinocytes, although in the MK cells expression staining of MK−/− cells on laminin-5-rich ECM revealed reduced actin stress fibers compared to MK+/+ cells (Fig. 3C,D). Similarly, primary mouse keratinocytes (DiPersio et al., 1997) and human keratinocytes (Xia et al., 1996) require α3β1 for efficient cell spreading on purified laminin-5 or laminin-5-rich ECM. In contrast with α3β1, integrin α6β4 mediates stable adhesion of keratinocytes to laminin-5 through hemidesmosmes (Borradori and Sonnenberg, 1999). The absence of α3β1 from MK cells did not affect the localization of α6β4 to the basal cell surface in a punctate pattern (Fig. 3E,F), similar to that observed previously in primary mouse keratinocytes (Hodivala-Dilke et al., 1998). MK+/+ cells, MK−/− cells and primary mouse keratinocytes display similar expression and processing of laminin-5 Laminin-5 is secreted as a 460 kDa heterotrimer of three subunits, α3 (190-200 kDa), β3 (140 kDa) and γ2 (155 kDa), and it is converted to smaller forms through proteolytic processing of the α3 and γ2 subunits (Marinkovich et al., 1992). Differential processing of the α3 or γ2 subunit has been correlated with changes in stable adhesion and migration of epithelial cells (Gianelli et al., 1997; Goldfinger et al., Fig. 3. Morphologies of MK+/+ and MK−/− cells. MK-1.16 cells (MK α3+/+) or MK-5.4.6 1998). To compare laminin-5 expression in cells (MK α3−/−) were grown under permissive conditions on laminin-5-rich ECM. +/+ −/− the MK and MK cell lines, we (A,B) Phase contrast of confluent MK cultures. (C,D) Phalloidin staining of actin stress performed immunoblot analysis of MK cell fibers in MK+/+ cells and MK−/− cells, respectively. (E,F) Immunofluorescence staining lysates with antisera specific for the α3 or γ2 for integrin α6β4 in MK+/+ cells and MK−/− cells, respectively. Bars, 50 µm. 2914 C. M. DiPersio and others

Fig. 4. Laminin-5 synthesis in MK+/+ and MK−/− cells. (A) Wild-type primary keratinocytes (wt), MK-1.16 cells (MK+/+) or MK-5.4.6 cells (MK−/−) were cultured on plastic tissue culture plates at 33°C for 3 days. Total cell lysates were prepared and assayed by immunoblot for expression of the laminin-5 α3 chain (LM-5: α3) or γ2 chain (LM-5: γ2). Blots were stripped and reprobed for keratin 14 (K14) or α3 integrin, as indicated. (B) Total cell lysates (cell) or extracellular matrix fractions (ECM) were prepared from cultures of MK-1.16 cells (+/+) or MK-5.4.6 cells (−/−) and assayed for laminin-5 expression as described in A. Molecular mass markers are indicated.

Fig. 5. MK cell lines produce an α3β1- dependent gelatinase. (A) MK-1.16 cells (+/+), MK-5.4.6 cells (−/−), or MK-5.4.6 cells transfected with α3A (−/−, α3) or vector alone (−/−, V), were cultured on laminin-5-rich ECM under either permissive (33°C) or nonpermissive (39°C) conditions for 2 days, and secreted gelatinases were assayed by gelatin zymography. Molecular mass markers are indicated. (B) Five other MK cell lines, derived independently from wild-type (1.14, 1.19) or α3-null keratinocytes (5.3.3, 5.5, 5.12), were assayed for gelatinase expression. (C) FACS analysis with the monoclonal antibody P1B5 shows high levels of α3 expression on the surface of α3-transfected MK-5.4.6 cells (+α3); MK−/−, MK-5.4.6 cells transfected with vector alone. levels were at the lower end of the normal range. These (Fig. 5A,B). A gelatinolytic protein of approximately 105 kDa results show that neither the lack of α3β1 integrin, nor was detected in culture medium from all three MK+/+ lines (+/+ immortalization by large T antigen, grossly affected expression lanes) but was barely detectable or absent in culture medium or processing of laminin-5 in mouse keratinocytes. Therefore, from all four MK−/− lines (−/− lanes), suggesting a secreted it is unlikely that α3β1-dependent alterations in laminin-5 proteinase that was α3β1-dependent. To verify that the absence contribute to differences in cell spreading or other cell of the gelatinase from MK−/− cells was due to absence of functions (discussed below) that we observed between MK+/+ endogenous α3β1, rather than to clonal variation between the and MK−/− cells. MK+/+ and MK−/− cell lines, we stably transfected the MK- 5.4.6 (α3−/−) cell line with a plasmid expressing α3A cDNA. Gelatinase secretion by cells is α3β1-dependent A pool of MK−/− cells that expressed high levels of transfected MMP-9/gelatinase B and MMP-2/gelatinase A have each been α3 was sorted by FACS with the monoclonal antibody P1B5, implicated in matrix remodeling during wound healing and and subsequent FACS analysis of the α3-enriched pool squamous cell carcinoma (Kähäri and Saarialho-kere, 1997; confirmed stable α3β1 expression on the cell surface (Fig. 5C). Johnsen et al., 1998; Westermarck and Kähäri, 1999). We Secretion of the 105 kDa gelatinase was restored in MK−/− cells compared three distinct MK+/+ lines and four distinct α3−/− transfected with α3A (Fig. 5A; −/−, α3), while MK−/− cells lines for gelatinase production. MK+/+ or MK−/− cells were transfected with vector alone (Fig. 5A; −/−, V) showed a cultured for 2 days on laminin-5-rich ECM in serum-free background level of activity comparable to that seen in medium, supplemented with growth factors and hormones (see untransfected MK−/− cells. Gelatinase secretion was also Materials amd Methods), and culture medium were collected dependent on α3β1 under nonpermissive culture conditions and assayed for gelatinase activity using gelatin zymography (Fig. 5A, 39°C), even when the MK cells were preincubated at Integrin α3β1 regulates MMP-9 expression 2915

Fig. 6. The α3β1-dependent gelatinase in MK cells shows properties of a matrix metalloproteinase. (A) Culture medium from MK+/+ cells (+/+) or α3-transfected MK−/− cells (α3) were assayed by gelatin zymography, as in Fig. 5. Gels were processed normally (untr.), incubated with 5 mM EDTA instead of calcium (EDTA), or incubated in the presence of 5 mM PMSF (PMSF), 10 mM 1,10-phenanthroline in methanol (phenan.), or methanol only (MetOH). For MMP activation, 1.5 mM aminophenylmercuric acetate (APMA) in DMSO, or DMSO only, was added to samples for 6 hours prior to zymography. M9, purified murine MMP-9; proenzyme and activated forms of MMP-9 and molecular mass markers are indicated.

39°C for 3 days prior to the assay (not shown), indicating that α3β1-dependent secretion did not require sustained activity of large T antigen. In both MK+/+ cells and α3-transfected MK−/− cells, gelatinase levels appeared slightly lower under permissive conditions (Fig. 5A, 33°C) than under nonpermissive conditions (Fig. 5A, 39°C). Reduced gelatinase production under permissive conditions may be due to the presence of IFNγ, which has been shown previously to inhibit expression of MMP-9 in transformed keratinocytes (Ala-aho et al., 2000). The α3β1-dependent gelatinase secreted by MK cells is MMP-9/gelatinase B α β The 3 1-dependent gelatinase in MK cells comigrated with Fig. 7. Identication of the α3β1-dependent gelatinase in MK cells as the 105 kDa proenzyme form of murine MMP-9/gelatinase B MMP-9. Culture medium from MK+/+ cells (+/+), MK−/− cells (−/−), (Fig. 6, lane M9), which is slightly larger than the 92 kDa or MK−/− cells transfected with α3 (−/−, α3) or vector alone (−/−,V) human MMP-9 homolog. As expected for a metalloproteinase, were assayed by immunoblot for the presence of MMP-9, following the activity was inhibited in both MK+/+ and α3-transfected isolation of MMPs on gelatin agarose beads as described in Materials MK−/− cells by EDTA or the zinc chelator 1,10-phenanthroline, and Methods. Purified MMP-9 from human (h) and mouse (m) were but not by the serine proteinase inhibitor PMSF (Fig. 6). The included as controls. Lysates from the corresponding cell monolayers proenzyme forms of gelatinases display proteolytic activity on were immunoblotted for the α3 integrin subunit or keratin 14 (K14), zymography gels and can be distinguished from the activated as indicated. forms by their migration properties. Treatment with APMA, an activator of metalloproteinases, resulted in a shift of the 105 confirmed the presence of intact protein in lysates (Fig. 7, K14 kDa gelatinolytic band to a lower band that comigrated with blot). the processed, activated form of murine MMP-9 (Fig. 6). To confirm that the α3β1-dependent gelatinase was MMP- Immortalization with large T antigen suppresses 9, we performed immunoblots of MK cell culture medium (Fig. MMP-9 production in α3β1-deficient keratinocytes 7). Since the gelatinase concentration was too low for detection but not in wild type keratinocytes by direct blotting, gelatinases were concentrated from culture To determine whether MMP-9 secretion is dependent on α3β1 medium prior to blotting by binding to gelatin-agarose beads in skin, we assayed gelatinase production in cultures of full- (see Materials and Methods). In control samples containing thickness skin. Skin slices were prepared from five wild-type purified MMPs, an antiserum specific for MMP-9 detected (α3+/+), five heterozygous (α3+/−), and five α3-null (α3−/−) human and murine MMP-9 (Fig. 7, h and m, respectively), but neonatal mice from a single litter and cultured in serum-free not human MMP-2 (not shown). This antiserum also detected medium for 2 days prior to gelatin zymography (Fig. 8A, MMP-9 in culture medium from MK+/+ cells (Fig. 7, +/+ lanes) representative samples shown). Secreted proteinases ranged in but not from MK−/− cells (Fig. 7, −/− lanes). MMP-9 secretion size from about 45 kDa to 105 kDa, and included proteins was restored in MK−/− cells transfected with α3 (Fig. 7, −/−, comigrating with pro-MMP-9, activated MMP-9 and pro- α3) but not MK−/− cells transfected with vector only (Fig. 7, MMP-2 (as indicated). Smaller bands probably represent −/−, V). Immunoblots of the corresponding cell lysates with intermediate and processed forms of MMP-2 (Okada et al., anti-α3 confirmed the absence of α3β1 from cells that failed 1997) and other proteinases. It is not known whether these to secrete MMP-9 (Fig. 7, α3 blot); control blots for keratin 14 proteinases are expressed in neonatal skin or were induced in 2916 C. M. DiPersio and others a wound healing response to local disruption of cell-ECM factors and cytokines, including EGF (Westermarck and and/or cell-cell contacts during preparation of skin slices. In Kähäri, 1999). To determine whether α3β1-mediated pathways either case, gelatinolytic patterns were generally similar in of MMP-9 induction were masked by other pathways of skins from wild-type mice (Fig. 8A, +/+) and from mice that induction in nonimmortalized keratinocytes, we cultured were homozygous (Fig. 8A, −/−) or heterozygous (not shown) primary keratinocytes in the presence or absence of soluble for the α3-null mutation. factors that are normally present in our standard growth Since several cell types found in skin can secrete medium (see Materials and Methods). Primary cultures were proteinases, we isolated primary keratinocytes from portions starved for 1 day in serum-free medium devoid of growth factor of the same skins used in Fig. 8A and assayed for secreted and hormonal supplements, then grown for 2 days in serum- gelatinases (Fig. 8B). Freshly isolated keratinocytes were free medium that was fully supplemented, unsupplemented or seeded directly onto laminin-5-rich ECM and cultured under supplemented with EGF only. Zymography of culture media the same conditions that were used for the MK cell lines (see showed that MMP-2 was expressed constitutively by primary Fig. 5). Control blots with anti-α3 and anti-keratin 14 keratinocytes under each of these conditions (Fig. 9). In confirmed the genotypes of corresponding skin and contrast, MMP-9 expression was regulated in response to keratinocyte cultures (Fig. 8C). Surprisingly, a gelatinase soluble factors present in the medium. Specifically, MMP-9 activity that comigrated with MMP-9 was detected consistently expression was dramatically reduced during the starvation in all medium samples from wild-type (Fig. 8B, +/+), period for both wild-type and α3β1-deficient cells (data not heterozygous (not shown) or α3-null keratinocytes (Fig. 8B, shown), but was restored to high levels by culture in fully −/−), showing that MMP-9 was secreted by nonimmortalized, primary cells independently of α3β1. MMP-9 expression was confirmed in α3-null primary cells isolated from eight different mice distributed over four litters. The presence of 20 units/ml IFNγ for up to 6 days had no effect on the constitutive secretion of MMP-9 by wild-type or α3β1-deficient primary keratinocytes (data not shown). A low level of activity comigrating with MMP-2 was also detected frequently in primary cultures, as has been reported for human keratinocytes (Salo et al., 1994). Immunoblotting confirmed the presence of MMP- 9 in both wild-type and α3-null skin cultures or primary cultures (Fig. 8D); however, only the 105 kDa proenzyme form of MMP-9 was secreted by primary keratinocytes (Fig. 8B). There was no consistent relationship between α3 genotypes and the level of MMP-9 secretion, which appeared to vary with cell density. It was difficult to plate primary cells quantitatively, however, since epidermal cell preparations include a large number of nonadherent suprabasal keratinocytes upon initial seeding. Therefore, our assays with primary cells were not sensitive enough to detect possible small effects of α3β1 on MMP-9 secretion. MMP-9 expression can be induced by certain growth

Fig. 8. Secretion of MMP-9 from skin and freshly isolated, primary keratinocytes is independent of α3β1. (A) Slices of full thickness skin were prepared from wild-type (+/+) or α3-null (−/−) neonatal mice, cultured for 2 days in serum-free medium, then assayed for secreted gelatinases by gelatin zymography, as in Fig. 5. (B) Primary keratinocyte cultures were prepared from portions of the same skins used in A and seeded at similar densities onto laminin-5-rich ECM overnight, then grown for 2 days in serum-free medium and assayed by gelatin zymography. For A and B, a mixture of purified human pro-MMP-2 (68 kDa) and human pro-MMP-9 (92 kDa) was used as standard (black arrows). Expected positions of the pro- and activated forms of murine MMP-9 are indicated by open arrows. (C) Lysates from the corresponding primary cell monolayers were immunoblotted for the α3 integrin subunit or keratin 14 (K14), as indicated. Lanes are aligned for A, B and C so that skin, primary cells and cell lysates are from the same animal. (D) Media from skin or primary keratinocyte cultures were assayed by immunoblot for MMP-9, as in Fig. 6. +/+, wild-type; −/−, α3-null. Purified MMP-9 standards were from mouse (m) or human (h). Integrin α3β1 regulates MMP-9 expression 2917

previously that primary α3-null keratinocytes on denatured collagen formed increased actin stress fibers compared to wild- type cells, suggesting that α3β1 expression inhibited cell spreading on non-laminin-5 substrates such as collagen (Hodivala-Dilke et al., 1998). Importantly, this inhibitory effect was not observed in keratinocytes cultured on substrates rich in laminin-5, for which α3β1 is a strong receptor. In the current study, we plated cells on an exogenous laminin-5-rich ECM in order to minimize effects of any potential differences in endogenous ECM secreted by MK+/+ or MK−/− cells. Indeed, alternative processing of the α3 or the γ2 subunit of laminin-5 can alter epithelial cell motility (Goldfinger et al., 1998; Gianelli et al., 1999). We were, however, able to rule out α3β1- dependent differences in laminin-5 production by MK cells, since the absence of α3β1 from MK−/− cells did not alter the Fig. 9. Induction of MMP-9 by EGF in nonimmortalized primary production or processing of laminin-5 in a detectable way. In keratinocytes. Primary keratinocytes that were either wild type (+/+) α3 − − addition, the MK cell lines showed only slightly reduced or -null ( / ) were cultured on laminin-5-rich ECM in the α3 presence or absence of growth factor and hormonal supplements, expression of laminin-5 and similar processing of the and then assayed for secreted gelatinases as described in Fig. 5. γ2 subunits when compared with primary keratinocytes. Following a 1 day starvation period in serum-free, unsupplemented Nevertheless, our results do not rule out α3β1-dependent medium, cells were cultured for an additional 2 days in the presence differences in other ECM proteins secreted by MK cells, or a of all medium supplements as described in the Materials and potential role for α3β1 in regulating laminin-5 processing Methods (+suppl.), in the absence of supplements (−suppl.) or in the during certain tissue remodeling events in vivo. presence of 10 ng/ml EGF only (+EGF). Molecular mass markers and positions of MMP-2 and MMP-9 are indicated. Immortalization of keratinocytes with large T antigen reveals an α3β1-dependent pathway for MMP-9 supplemented medium (Fig. 9, + suppl.). In contrast, MMP-9 production expression was not detected in either wild-type or α3β1- During wound healing, tumor invasion and other tissue deficient keratinocytes cultured in unsupplemented medium remodeling events, individual integrins regulate specific (Fig. 9, + suppl.). Addition of EGF alone restored MMP-9 proteinases in response to distinct ligands present in a complex secretion in primary keratinocytes independently of α3β1 extracellular matrix (Pilcher et al., 1997; Lochter et al., 1999). expression (Fig. 9, + EGF). In contrast, the presence of EGF MMP-9/gelatinase B is known to be expressed by keratinocytes did not induce MMP-9 secretion by immortalized keratinocytes during wound healing and in carcinomas of the skin (Pyke et deficient for α3β1 (Fig. 5, and data not shown). al., 1992; Juarez et al., 1993; Salo et al., 1994; Okada et al., 1997; Madlener et al., 1998; Westermarck and Kähäri, 1999). We showed that MK cells cultured on laminin-5-rich ECM DISCUSSION required expression of α3β1 for MMP-9 secretion. Three distinct MK+/+ (α3-positive) cell lines secreted MMP-9 into the Previous studies that utilized ‘function-perturbing’ antibodies culture medium, while MMP-9 was barely detectable in, or directed against α3β1 have provided valuable insights into the absent from, culture media of four distinct MK−/− (α3- roles of this integrin in regulating adhesion, migration and deficient) cell lines. MMP-9 secretion was completely restored signal transduction of epithelial cells (Kim et al., 1992; Xia et in MK−/− cells transfected with the α3 integrin subunit, al., 1996; Gonzales et al., 1999). Nevertheless, a disadvantage indicating a pathway for MMP-9 production that requires of this approach is that anti-α3β1 antibodies may fail to block α3β1. Nonimmortalized primary keratinocytes secreted MMP- all functions of the integrin, and they may even induce some 9 in a manner that was independent of α3β1, suggesting that functions (Symington and Carter, 1995). The α3β1-deficient α3β1-dependent MMP-9 secretion was activated de novo in MK−/− cell lines provide a valuable system for cell biological the MK cell lines upon immortalization. Alternatively, studies of integrin function that avoid these limitations. Indeed, immortalization may have occurred selectively in a the MK cell lines retained many characteristics of primary subpopulation of keratinocytes that utilize α3β1-dependent keratinocytes, including an identical repertoire of integrins and pathways of MMP-9 production. MMP-9 secretion was similar expression and processing of laminin-5. The current induced by EGF in nonimmortalized primary cells study, however, also revealed a novel MMP regulatory function independently of α3β1 (Fig. 9). In contrast, EGF and other for α3β1 integrin that was associated with the immortalized soluble factors were not sufficient to maintain MMP-9 phenotype, illustrating the need to exercise caution when expression in α3β1-deficient MK−/− cells (Fig. 5). EGF did extrapolating from data obtained with cell lines to the biology moderately enhance MMP-9 expression in MK+/+ cells (data of primary cells from which they were derived. not shown), suggesting that integrin α3β1 and the EGF α3β1-deficient MK−/− cells displayed reduced actin stress receptor (or possibly other receptors) may collaborate for fiber formation compared to MK+/+ cells when cultured on optimal expression of MMP-9 in the immortalized laminin-5-rich ECM, consistent with previous reports that keratinocytes. α3β1 is required for efficient keratinocyte spreading on The fact that α3-null keratinocytes lost MMP-9 expression laminin-5 (Xia et al., 1996; DiPersio et al., 1997). We showed after immortalization with tsA58 large T antigen suggests that 2918 C. M. DiPersio and others immortalization suppresses MMP-9 expression in these cells. antigen-immortalized keratinocytes. Loss of this pathway may Similarly, kidney collecting duct epithelial cells that were occur through loss of the receptor, as shown in the figure, or immortalized with tsA58 secreted abundant MMP-9 when through suppression of downstream signaling events. The grown at the nonpermissive temperature for large T antigen, activation of an alternative, α3β1-dependent pathway in the but failed to produce MMP-9 when grown at the permissive immortalized MK+/+ cells may ‘rescue’ MMP-9 production in temperature (Piedagnel et al., 1999). Expression of large T these cells (Fig. 10A). Since MK−/− cells lack α3β1, this rescue antigen also suppressed MMP-9 production in fibroblasts does not occur and suppression of the putative growth factor- (Logan et al., 1996). Large T antigen may inhibit MMP gene mediated pathway is revealed (Fig. 10B). expression by binding to and modulating the functions of Maintenance of MMP-9 secretion in immortalized MK+/+ cellular regulatory proteins. For example, large T antigen can cells could be due to activation of signal transduction pathways bind to the transcription factor AP-2 and inhibit its DNA that are initiated specifically by α3β1 binding to laminin-5, or binding activity (Mitchell et al., 1987). AP-2 appears essential possibly to other ligands present in the crude ECM preparation for MMP-2 gene transcription (Somasundaram et al., 1996) (Xia et al., 1996; Gonzales et al., 1999). In addition, it is and may activate MMP-9 transcription in epithelial cells (Fini possible that associations between α3β1 and certain cell surface et al., 1994; Munaut et al., 1999). Large T antigen can also bind adapter proteins are upregulated in immortalized/transformed to and inhibit the functions of the retinoblasoma (Rb) and p53 keratinocytes, thereby coupling MMP-9 production to α3β1 tumor suppressor proteins (for a review, see Levine and expression. For example, integrin α3β1 can associate with Momand, 1990), each of which has been implicated in MMP certain members of the TM4 (transmembrane-4) superfamily of gene activation. Indeed, p53 can bind and activate the MMP-2 cell surface proteins (Hemler, 1998), some of which have been promoter (Bian and Sun, 1997), and quenching of p53 function implicated in regulation of MMP-2 expression (Sugiura and by large T antigen may contribute to suppression of MMP gene Berditchevski, 1999). α3β1 can also associate with expression (Piedagnel et al., 1999). Similarly, the MMP-9 CD147/EMMPRIN, an inducer of MMP synthesis that is promoter contains an Rb binding element that appears to be implicated in tumor cell invasion (Berditchevski et al., 1997; involved in transcriptional activation (Himelstein et al., 1997), Hemler, 1998). Alternatively, MMP-9 expression may be suggesting that large T antigen may suppress MMP-9 regulated more generally through changes in cell spreading, expression by inhibiting Rb function. since the absence of MMP-9 secretion from MK−/− cells was The molecular mechanisms whereby MMP-9 expression is correlated with reduced spreading and actin stress fiber suppressed in large T antigen-immortalized MK−/− cells are not formation on laminin-5-rich ECM. Indeed, collagenase- yet known. Preliminary northern blots showed that MMP-9 1/MMP-1 gene expression was induced in fibroblasts by mRNA levels were reduced in MK−/− cells (V. Iyer and C. M. integrin-dependent rearrangement of the actin cytoskeleton and DiPersio, unpublished data), but further experiments are subsequent activation of the GTP-binding protein Rac 1 necessary to determine whether suppression occurs at (Kheradmand et al., 1998). the transcriptional or post- transcriptional level. The fact that MMP-9 suppression did not require wild type A primary keratinocyte MK +/+ sustained function of large T cell line growth growth antigen (Fig. 5) suggests that it may factor MMP-9 L-TAg MMP-9 factor occur primarily through an indirect, perhaps irreversible mechanism, MMP-9 MMP-9 rather than through inhibitory binding of large T antigen to

transcription factors that directly α3β1 LM-5 α3β1 LM-5 activate the MMP-9 gene. For example, immortalization of α epithelial cells by large T antigen 3-null MK -/- can result in loss of certain B primary keratinocyte cell line growth growth hormone receptors (Piedagnel et factor MMP-9 L-TAg factor al., 1999), suggesting that loss of essential receptors from the MMP-9 MMP-9 immortalized MK cells may have eliminated signaling pathways that maintain MMP-9 production in LM-5 LM-5 keratinocytes. We propose a model, illustrated Fig. 10. A model for acquisition of α3β1 integrin-dependent MMP-9 expression by large T antigen- in Fig. 10, in which MMP-9 immortalized keratinocytes. Integrin-independent pathways for induction of MMP-9 exist in secretion by normal keratinocytes nonimmortalized keratinocytes. For illustrative purposes, a growth factor-mediated pathway is shown. (A) The growth factor-mediated pathway is suppressed in large T antigen-immortalized is regulated primarily through an +/+ α3β1 α3β1 MK cells, and an -mediated pathway is activated in order to maintain MMP-9 expression. -independent pathway (a (B) α3β1-deficient primary cells express MMP-9 via the growth factor-mediated pathway, but the growth factor-mediated pathway is pathway is suppressed in the immortalized MK−/− cells. The absence of α3β1-mediated pathways, indicated as an example), and that due to the lack of α3β1, results in loss of MMP-9 expression from the MK−/− cells. LM-5, laminin- this pathway is lost from large T 5; L-TAg, large T antigen. Integrin α3β1 regulates MMP-9 expression 2919

A potential role for α3β1 in maintaining MMP-9 for critical comments on the manuscript. We also thank Susan production by transformed epithelial cells LaFlamme, Jane Sottile, Anthony Mastrangelo, and Vandana Iyer for α3β1 valuable discussions and critical reading of the manuscript. We thank -deficient mice display perinatal lethality and α3 disorganized basement membranes in developing skin, kidney Martin Hemler for the human cDNA, Elaine Fuchs for antiserum against keratin 14, and Guerrino Meneguzzi for antisera against the and lung (Kreidberg et al., 1996; DiPersio et al., 1997). In α3 and γ2 subunits of laminin-5. This research was supported by contrast, MMP-9-deficient mice are viable and show normal grants from the National Institutes of Health to R.O.H. organization of epithelial basement membranes, at least in (R01CA17007) and to C.M.D. (R01CA84238). R.O.H. is an kidney (Miner et al., 1997). In addition, we report here that full Investigator of the Howard Hughes Medical Institute. thickness skin explants from α3β1-deficient neonatal mice secreted MMP-9. Therefore, it seems unlikely that loss of MMP-9 function is involved in the basement membrane disorganization that occurs in α3β1-deficient mice. This ECM REFERENCES disorganization may instead result from alterations in other Adams, J. C. and Watt, F. M. (1990). Changes in keratinocyte-extracellular proteinases. Possible candidates include bone morphogenetic matrix interactions during terminal differentiation: reduction in fibronectin protein-1 (BMP-1) and related splice variants, since these binding precedes loss of α5β1 integrin from the cell surface. Cell 63, 425- proteinases are produced by keratinocytes (Lee et al., 1997), 435. and BMP-1-deficient mice display developmental defects in Adams, J. C. and Watt, F. M. (1991). Expression of β1, β3, β4, and β5 integrins by human epidermal keratinocytes and non-differentiating basement membrane organization (Suzuki et al., 1996; keratinocytes. J. Cell Biol. 115, 829-841. Burgeson and Christiano, 1997). We suggest that α3β1- Ala-aho, R., Johansson, N., Grénman, R., Fusenig, N. E., López-Otín, C. dependent regulation of MMP-9 may instead have important and Kähäri, V.-M. (2000). Inhibition of collagenase-3 (MMP-13) roles in postdevelopmental ECM remodeling, as suggested expression in transformed human keratinocytes by interferon-γ is associated previously (Shapiro, 1998). with activation of extracellular signal-regulated kinase-1,2 and STAT1. Oncogene 19, 248-257. Carcinogenesis is generally accompanied by dramatic Banks-Schlegel, S. P. and Howley, P. M. (1983). Differentiation of human changes in MMP expression, ECM synthesis and integrin epidermal cells transformed by SV40. J. Cell Biol. 96, 330-337. function that facilitate tumor growth and cell invasion (Werb, Bartolazzi, A., Cerboni, C., Nicotra, M. R., Mottolese, M., Bigotti, A. and 1997; Johnsen et al., 1998; Westermarck and Kähäri, 1999). Natali, P. G. (1994). Transformation and tumor progression are frequently associated with expression of the α3β1 heterodimer in solid tumors. Int. J. MMP expression during cancer cell invasion is regulated by Cancer 58, 488-491. multiple environmental factors, including growth factors, Berditchevski, F., Chang, S., Bodovora, J. and Hemler, M. E. (1997). cytokines, and integrin-mediated cell contacts with extracellular Generation of monoclonal antibodies to integrin-associated proteins. matrix (Westermarck and Kähäri, 1999; Lochter et al., 1999). Evidence that α3β1 complexes with EMMPRIN//OX47/M6. J. Biol. Certain integrin-mediated pathways of MMP production may be Chem. 272, 29174-29180. Bian, J. and Sun, Y. (1997). Transcriptional activation by p53 of the human activated during tumor development to facilitate cell invasion. type IV collagenase (gelatinase A or matrix metalloproteinase 2) promoter. α3β1 is a good candidate for activating such pathways, since it Mol. Cell. Biol. 17, 6330-6338. is expressed at high levels in the majority of transformed Borradori, L. and Sonnenberg, A. (1999). Structure and function of adherent cells (Plantefaber and Hynes, 1989; Tsuji et al., 1990; hemidesmosomes: more than simple adhesion structures. J. Invest. Dermatol. 112, 411-418. Dedhar and Saulnier, 1990; Melchiori et al., 1995) and in most Brooks, P. C., Stromblad, S., Sanders, L. C., von Schalscha, T. L., Aimes, primary and metastatic tumors (Natali et al., 1993; Bartolazzi R. T., Stetler-Stevenson, W. G., Quigley, J. P. and Cheresh, D. A. (1996). et al., 1994; Patriarca et al., 1998). MMP-9 is activated and Localization of matrix metalloproteinase MMP-2 to the surface of invasive coexpressed with α3β1 in epithelial squamous cell carcinomas cells by interaction with integrin αvβ3. Cell 85, 683-693. (Pyke et al., 1992; Juarez et al., 1993; Westermarck and Kähäri, Burgeson, R. E. and Christiano, A. M. (1997). The dermal-epidermal junction. Curr. Opin. Cell Biol. 9, 651-658. 1999) and other cancers, and its proteolytic activity contributes Carter, W. G., Ryan, M. C. and Gahr, P. J. (1991). Epiligrin, a new cell to the invasive potential of tumor cells (Ramos-DeSimone et al., adhesion ligand for integrin α3β1 in epithelial basement membranes. J. Cell 1999). The fact that suppression of MMP-9 in the α3β1-deficient Biol. 65, 599-610. MK−/− cell lines was overcome by expression of exogenous Clark, E. A. and Brugge, J. S. (1995). Integrins and signal transduction α3β1 pathways: the road taken. Science 268, 233-239. suggests an interesting role for this integrin in maintaining de Melker, A. A., Sterk, L. M. T., Delwel, G. O., Fles, D. L. A., Daams, H., MMP-9 expression in transformed epithelial cells. Such a role Weening, J. J. and Sonnenberg, A. (1997). The A and B variants of the is supported in a recent study by A. Albini and coworkers, which α3 integrin subunit: tissue distribution and functional characterization. Lab. showed that function-blocking antibodies against α3β1 caused Invest. 76, 547-563. a decrease in MMP-9 production and inhibited cell invasion in Dedhar, S. and Saulnier, R. (1990). Alterations in integrin receptor expression on chemically transformed human cells: specific enhancement of mammary carcinoma cells (Morini et al., 2000). In contrast, laminin and collagen receptors. J. Cell Biol. 110, 481-489. function-blocking antibodies against α3β1 failed to inhibit Delwel, G. O., de Melker, A. A., Hogervorst, F., Jaspars, L. H., Fles, D. L. MMP-9 production in primary mucosal keratinocytes (Larjava A., Kuikman, I., Lindblom, A., Paulsson, M., Timpl, R. and et al., 1993a), consistent with our results showing high MMP-9 Sonnenberg, A. (1994). Distinct and overlapping ligand specificities of the α3β1 α3Aβ1 and α6Aβ1 integrins: recognition of laminin isoforms. Mol. Biol. expression in nonimmortalized, -deficient primary Cell 5, 203-215. keratinocytes. Collectively, our study and the latter studies DiPersio, C. M., Shah, S. and Hynes, R. O. (1995). α3Aβ1 integrin localizes support the notion that α3β1-dependent regulation of MMP-9 to focal contacts in response to diverse extracellular matrix proteins. J. Cell expression is activated in some immortalized/transformed cells Sci. 108, 2321-2336. to facilitate cell invasion. DiPersio, C. M., Hodivala-Dilke, K. M., Jaenisch, R., Kreidberg, J. A. and Hynes, R. O. (1997). α3β1 integrin is required for normal development of the epidermal basement membrane. J. Cell Biol. 137, 729-742. We are grateful to Kairbaan Hodivala-Dilke for her advice and help Fini, M. E., Bartlett, J. D., Matsubara, M., Rinehart, W. B., Mody, M. K., during establishment and characterization of the MK cell lines, and Girard, M. T. and Rainville, M. (1994). The rabbit gene for 92-kDa matrix 2920 C. M. DiPersio and others

metalloproteinase: role of AP1 and AP2 in cell type-specific transcription. integrins mediate invasive activity of mouse mammary carcinoma cells J. Biol. Chem. 269, 28620-28628. through regulation of stromelysin-1 expression. Mol. Biol. Cell 10, 271-282. Gianelli, G., Falk-Marzillier, J., Schiraldi, O., Stetler-Stevenson, W. G. and Logan, S. K., Hansell, E. J., Damsky, C. H. and Werb, Z. (1996). T-antigen Quaranta, V. (1997). Induction of cell migration by matrix metalloprotease- inhibits metalloproteinase expression and invasion in human placental cells 2 cleavage of laminin-5. Science 277, 225-228. transformed with temperature-sensitive simian virus 40. Matrix Biol. 15, 81- Goldfinger, L. E., Stack, M. S. and Jones, J. C. R. (1998). Processing of 89. laminin-5 and its functional consequences: role of plasmin and tissue-type Madlener, M., Parks, W. C. and Werner, S. (1998). Matrix plasminogen activator. J. Cell Biol. 141, 255-265. metalloproteinases (MMPs) and their physiological inhibitors (TIMPs) are Gonzales, M., Haan, K., Baker, S. E., Fitchmun, M., Todorov, I., differentially expressed during excisional skin wound repair. Exp. Cell Res. Weitzman, S. and Jones, J. C. R. (1999). A cell signal pathway involving 242, 201-210. laminin-5, α3β1 integrin, and mitogen-activated protein kinase can regulate Marcantonio, E. E. and Hynes, R. O. (1988). Antibodies to the conserved epithelial cell proliferation. Mol. Biol. Cell 10, 259-270. cytoplasmic domain of the integrin β1 subunit react with proteins in Hemler, M. E. (1998). Integrin associated proteins. Curr. Opin. Cell Biol. 10, vertebrates, invertebrates, and fungi. J. Cell Biol. 106, 1765-1772. 578-585. Marinkovich, P. M., Lunstrum, G. P. and Burgeson, R. E. (1992). The Himelstein, B. P., Lee, E. J., Sato, H., Seiki, M. and Muschel, R. J. (1997). anchoring filament protein kalinin is synthesized and secreted as a high Transcriptional activation of the matrix metalloproteinase-9 gene in an molecular weight precursor. J. Biol. Chem. 267, 17900-17906. H-ras and v-myc transformed rat embryo cell line. Oncogene 14, 1995- Melchiori, A., Mortarini, R., Carlone, S., Marchisio, P. C., Anichini, A., 1998. Noonan, D. M. and Albini, A. (1995). The α3β1 integrin is involved in Hodivala-Dilke, K. M., DiPersio, C. M., Kreidberg, J. A. and Hynes, R. melanoma cell migration and invasion. Exp. Cell Res. 219, 233-242. O. (1998). Novel roles for α3β1 integrin as a regulator of cytoskeletal Miner, J. H., Betsuyaku, T., Shipley, J. M. and Senior, R. M. (1997). Renal assembly and as a transdominant inhibitor of integrin receptor function in function is normal in gelatinase B deficient mice [abstract]. Mol. Biol. Cell keratinocytes. J. Cell Biol. 142, 1357-1369. 8, 403. (Abstract) Huhtala, P., Humphries, M. J., McCarthy, J. B., Tremble, P., Werb, Z. and Mitchell, P. J., Wang, C. and Tijan, R. (1987). Positive and negative Damsky, C. H. (1995). Cooperative signalling by α5β1 and α4β1 integrins regulation of transcription in vitro: enhancer-binding protein AP-2 is regulates metalloproteinase gene expression in fibroblasts adhering to inhibited by SV40 T antigen. Cell 50, 847-861. fibronectin. J. Cell Biol. 129, 867-879. Morgan, J. E., Beauchamp, J. R., Pagel, C. N., Peckham, M., Ataliotis, P., Hynes, R. O. (1992). Integrins: versatility, modulation, and signaling in cell Jat, P. S., Noble, M. D., Farmer, K. and Partridge, T. A. (1994). Myogenic adhesion. Cell 69, 11-25. cell lines derived from transgenic mice carrying a thermolabile T antigen: Jat, P. S. and Sharp, P. A. (1989). Cell lines established by a temperature- a model system for the derivation of tissue-specific and mutation-specific sensitive simian virus 40 large-T-antigen gene are growth restricted at the cell lines. Dev. Biol. 162, 486-498. nonpermissive temperature. Mol. Cell. Biol. 9, 1672-1681. Morini, M., Mottolese, M., Ferrari, N., Ghiorzo, F., Buglioni, S., Jat, P. S., Noble, M. D., Ataliotis, P., Tanaka, Y., Yannoutsos, N., Larsen, Mortarnini, R., Noonan, D. M., Natali, P. and Albini, A. (2000). The L. and Kioussis, D. (1991). Direct derivation of conditionally immortal cell alpha3beta1 integrin is associated with mammary carcinoma cell metastasis, lines from an H-2Kb-tsA58 transgenic mouse. Proc. Natl. Acad. Sci. USA invasion and gelatinase B (MMP-9) activity. Int. J. Cancer, in press. 88, 5096-5100. Munaut, C., Salonurmi, T., Kontusaari, S., Reponen, P., Morita, T., Johnsen, M., Lund, L. R., Romer, J., Almholt, K. and Dano, K. (1998). Foidart, J.-M. and Tryggvason, K. (1999). Murine matrix Cancer invasion and tissue remodeling: common themes in proteolytic metalloproteinase 9 gene: 5′-upstream region contains cis-acting elements matrix degradation. Curr. Opin. Cell Biol. 10, 667-671. for expression in osteoclasts and migrating keratinocytes in transgenic mice. Juarez, J., Clayman, G., Nakajima, M., Tanabe, K. K., Saya, H., Nicolson, J. Biol. Chem. 274, 5588-5596. G. L. and Boyd, D. (1993). Role and regulation of expression of 92-kDa Natali, P. G., Nicotra, M. R., Bartolazzi, A., Cavaliere, R. and Bigotti, A. type-IV collagenase (MMP-9) in 2 invasive squamous-cell-carcinoma cell (1993). Integrin expression in cutaneous malignant melanoma: association lines of the oral cavity. Int. J. Cancer 55, 10-18. of the alpha 3/beta 1 heterodimer with tumor progression. Int. J. Cancer 54, Kähäri, V.-M. and Saarialho-Kere, U. (1997). Matrix metalloproteinases in 68-72. skin. Exp. Dermatol. 6, 199-213. Okada, A., Tomasetto, C., Lutz, Y., Bellocq, J.-P., Rio, M.-C. and Basset, Kheradmand, F., Werner, E., Tremble, P., Symons, M. and Werb, Z. P. (1997). Expression of matrix metalloproteinases during rat skin wound (1998). Role of Rac1 and oxygen radicals in collagenase-1 expression healing: evidence that membrane type-1 matrix metalloproteinase is a induced by cell shape change. Science 280, 898-902. stromal activator of pro-gelatinase A. J. Cell Biol. 137, 67-77. Kim, J. P., Zhang, K., Kramer, R. H., Schall, T. J. and Woodley, D. T. Patriarca, C., Alfano, R. M., Sonnenberg, A., Graziani, D., Cassani, B., de (1992). Integrin receptors and RGD sequences in human keratinocyte Melker, A., Colombo, P., Languino, L. R., Fornaro, M., Warren, W. H., migration: unique anti-migratory function of α3β1 epiligrin receptor. J. Coggi, G. and Gould, V. E. (1998). Integrin laminin receptor profile of Invest. Dermatol. 98, 764-770. pulmonary squamous cell and adenocarcinomas. Hum. Pathol. 29, 1208- Kreidberg, J. A, Donovan, M. J., Goldstein, S. L., Rennke, H., Shepherd, 1215. K., Jones, R. C. and Jaenisch, R. (1996). Alpha 3 beta 1 integrin has a Piedagnel, R., Murphy, G., Ronco, P. M. and Lelong, B. (1999). Matrix crucial role in kidney and lung organogenesis. Development 122, 3537- metalloproteinase 2 (MMP2) and MMP9 are produced by kidney collecting 3547. duct principal cells but are differentially regulated by SV40 large-T, arginine Larjava, H., Lyons, I. G., Salo, T., Mäkelä, M., Koivisto, L., Birkedal- vasopressin, and epidermal growth factor. J. Biol. Chem. 274, 1614-1620. Hansen, H., Akiyama, S. K., Yamada, K. M. and Heino, J. (1993a). Anti- Pilcher, B. K., Dumin, J. A., Sudbeck, B. D., Krane, S. M., Welgus, H. G. integrin antibodies induce type IV collagenase expression in keratinocytes. and Parks, W. C. (1997). The activity of collagenase-1 is required for J. Cell. Physiol. 157, 190-200. keratinocyte migration on a type I collagen matrix. J. Cell Biol. 137, 1445- Larjava, H., Salo, T., Haapasalmi, K., Kramer, R. H. and Heino, J. 1457. (1993b). Expression of integrins and basement membrane components by Plantefaber, L. C. and Hynes, R. O. (1989). Changes in integrin receptors wound keratinocytes. J. Clin. Invest. 92, 1425-1435. on oncogenically transformed cells. Cell 56, 281-290. Lee, S., Solow-Cordero, D. E., Kessler, E., Takahara, K. and Greenspan, Pozzi, A., Moberg, P. E., Miles, L. A., Wagner, S., Soloway, P. and Gardner, D. S. (1997). Transforming growth factor-β regulation of bone H. A. (2000). Elevated matrix metalloprotease and angiostatin levels in morphogenetic protein-1/procollagen C-proteinase and related proteins in integrin α1 knock-out mice cause reduced tumor vascularization. Proc. Natl. fibrogenic cells and keratinocytes. J. Biol. Chem 272, 19059-19066. Acad. Sci. USA 97, 2202-2207. Lefebvre, V., Garofalo, S. and de Crombrugghe, B. (1995). Type X Pyke, C., Ralfkiaer, E., Huhtala, P., Hurskainen, T., Dano, K. and collagen gene expression in mouse chondrocytes immortalized by a Tryggvason, K. (1992). Localization of messenger RNA for Mr 72,000 and temperature-sensitive Simain Virus 40 large tumor antigen. J. Cell Biol. 92,000 type IV collagenases in human skin cancers by in situ hybridization. 128, 239-245. Cancer Res. 52, 1336-1341. Levine, A. J. and Momand, J. (1990). Tumor suppressor : the p53 and Ramos-DeSimone, N., Hahn-Dantona, E., Sipley, J., Nagase, H., French, retinoblastoma sensitivity genes and gene products. Biochim. Biophys.Acta D. L. and Quigley, J. P. (1999). Activation of matrix metalloproteinase-9 1032, 119-136. (MMP-9) via a converging plasmin/stromelysin-1 cascade enhances tumor Lochter, A., Navre, M., Werb, Z. and Bissell, M. J. (1999). α1 and α2 cell invasion. J. Biol. Chem. 274, 13066-13076. Integrin α3β1 regulates MMP-9 expression 2921

Rheinwald, J. G. and Beckett, M. A. (1981). Tumorigenic keratinocyte lines (1996). Failure of ventral body wall closure in mouse embryos lacking a requiring anchorage and fibroblast support cultures from human squamous procollagen C-proteinase encoded by Bmp1, a mammalian gene related to cell carcinomas. Cancer Res. 41, 1657-1663. Drosophila tolloid. Development 122, 3587-3595. Rosales C. and Juliano, R. (1996). Integrin signaling to NF-κB in monocytic Symington, B. E. and Carter, W. E. (1995). Modulation of epidermal leukemia cells is blocked by activated oncogenes. Cancer Res. 56, 2302- differentiation by epiligrin and integrin α3β1. J. Cell Sci. 108, 831-838. 2305. Takada, Y., Murphy, E., Pil, P., Chen, C., Ginsberg, M. H. and Hemler, Ruoslahti, E. (1999). Fibronectin and its integrin receptors in cancer. Adv. M. E. (1991). Molecular cloning and expression of the cDNA for α3 subunit Cancer Res. 76, 1-20. of human α3β1 (VLA-3), an integrin receptor for fibronectin, laminin, and Ryan, M. C., Tizard, R., VanDevanter, D. R. and Carter, W. G. (1994). collagen. J. Cell Biol. 115, 257-266. Cloning of the LamA3 gene encoding the α3 chain of the adhesive ligand Tapley, P., Horwitz, A. F., Buck, C. A., Burridge, K., Duggan, K., Hirst, epiligrin. J. Biol. Chem. 269, 22779-22787. R. and Rohrschneider, L. (1989). Analysis of the avian fibronectin receptor Salo, T., Makela, M., Kylmaniemi, M., Autio-Harmainen, H. and Larjava, (integrin) as direct substrate for pp60v-src. Oncogene 4, 325-333. H. (1994). Expression of matrix metalloproteinase-2 and −9 during early Tsuji, T., Yamamoto, F.-i., Miura, Y., Takio, K., Titani, K., Pawar, S., human wound healing. Lab. Invest. 70, 176-182. Osawa, T. and Hakomori, S.-i. (1990). Characterization through cDNA Sanders, R. J., Mainiero, F. and Giancotti, F. G. (1998). The role of integrins cloning of galactoprotein b3 (Gap b3), a cell surface membrane glycoprotein in tumorigenesis and metastasis. Cancer Invest. 16, 329-344. showing enhanced expression on oncogenic transformation. Identification Schwartz, M. A., Schaller, M. D. and Ginsberg, M. H. (1995). Integrins: of Gap b3 as a member of the integrin superfamily. J. Biol. Chem. 265, 7016- emerging paradigms of signal transduction. Annu. Rev. Cell Dev. Biol. 11, 7021. 549-599. Watt, F. M. and Hertle, M. D. (1994). Keratinocyte integrins. In The Schwartz, M. A., Toksoz, D. and Khosravi-Far, R. (1996). Transformation Keratinocyte Handbook (ed. I. M. Leigh, E. B. Lane and F. M. Watt), pp. by Rho exchange factor oncogenes is mediated by activation of an integrin- 153-164. Cambridge: Cambridge University Press. dependent pathway. EMBO J. 15, 6525-6530. Wayner, E. A. and Carter, W. G. (1987). Identification of multiple cell Shapiro, S. D. (1998). Matrix metalloproteinase degradation of the adhesion receptors for collagen and fibronectin in human fibrosarcoma extracellular matrix: biological consequences. Curr. Opin. Cell Biol. 10, cells possessing unique α and common β subunits. J. Cell Biol. 105, 1873- 602-608. 1884. Somasundaram, K., Jayaraman, G., Williams, T., Moran, E., Frisch, S. Werb, Z. (1997). ECM and cell surface proteolysis: regulating cellular and Thimmapaya, B. (1996). Repression of a matrix metalloprotease gene ecology. Cell 91, 439-442. by E1A correlates with its ability to bind to cell type-specific transcription Westermarck, J. and Kähäri, V.-M. (1999). Regulation of matrix factor AP-2. Proc. Natl. Acad. Sci. USA 93, 3088-3093. metalloproteinase expression in tumor invasion. FASEB J. 13, 781-792. Sugiura, T. and Berditchevski, F. (1999). Function of α3β1-tetraspanin Xia, Y., Gil, S. G. and Carter, W. G. (1996). Anchorage mediated by integrin protein complexes in tumor cell invasion. Evidence for the role of the α6β4 to laminin 5 (epiligrin) regulates tyrosine phosphorylation of a complexes in production of matrix metalloproteinase 2 (MMP-2). J. Cell membrane-associated 80-kDa protein. J. Cell Biol. 132, 727-740. Biol. 146, 1375-1389. Zaret, K. S., DiPersio, C. M., Jackson, D. A., Montigny, W. J. and Suzuki, N., Labosky, P. A., Furuta, Y., Hargett, L., Dunn, R., Fogo, A. B., Weinstat, D. L. (1988). Conditional enhancement of liver-specific gene Takahara, K., Peters, D. M. P., Greenspan, D. S. and Hogan, B. L. M. transcription. Proc. Natl. Acad. Sci. USA 85, 9076-9080.