Expression of integrin β6 enhances invasive behavior in oral squamous cell carcinoma

Matrix Biology 21 (2002) 297–307

0945-053X/02/$ - see front matter� 2002 Elsevier Science B.V. and International Society of Matrix Biology. All rights reserved.PII: S0945-053XŽ02.00002-1

Expression of integrinb6 enhances invasive behavior in oral squamouscell carcinoma

Daniel M. Ramos *, Maria But , Joseph Regezi , Brian L. Schmidt , Amha Atakilit ,a, a a b c

Dongmin Dang , Duncan Ellis , Richard Jordan , Xiaowu Lia a a a

Department of Stomatology, University of California at San Francisco, San Francisco, CA 94143-0512, USAa

Department of Oral and Maxillofacial Surgery, University of California at San Francisco, San Francisco, CA 94143, USAb

The Lung Biology Center, University of California at San Francisco, San Francisco, CA 94143, USAc

Received 19 July 2001; received in revised form 20 November 2001; accepted 19 December 2001

Abstract

Oral squamous cell carcinoma(SCC) is characterized by invasive growth and the propensity for distant metastasis. Theexpression of specific adhesion receptors promotes defined interactions with the specific components found within the extracellularmatrix (ECM). We previously showed that theavb6 fibronectin receptor is highly expressed in oral SCC. Here we forcedexpression of theb6 subunit into poorly invasive SCC9 cells to establish the SCC9b6 cell line and compared these two cell linesin several independent assays. Whereas adhesion to fibronectin was unaffected by the expression ofb6, migration on fibronectinand invasion through a reconstituted basement membrane(RBM) were both increased. Function-blocking antibodies toavb6(10D5) reduced both migration on fibronectin and invasion through an RBM, whereas anti-a5 antibodies were effective only insuppressing migration on fibronectin, not invasion. Expression ofb6 also promoted tumor growth and invasion in vivo andmodulated fibronectin matrix deposition. When grown as a co-culture with SCC9 cells, peritumor fibroblasts(PTF) organized adense fibronectin matrix. However, fibronectin matrix assembly was decreased in co-cultures of SCC9b6 cells and PTF and thisdecrease was reversed by the addition of function-blocking anti-avb6 antibodies. The expression ofb6 also resulted in increasedlevels of matrix metalloproteinase 3. Addition of the general MMP inhibitor GM6001 to SCC9b6yPTF co-cultures dramaticallyincreased fibronectin matrix assembly in a similar fashion as incubation with anti-avb6 antibodies. These results demonstrate thatexpression ofb6 (1) increases oral SCC cell motility and growth in vitro and in vivo;(2) negatively affects fibronectin matrixassembly; and(3) stimulates the expression and activation of MMP3. We suggest that the integrinavb6 is a key component oforal SCC invasion and metastasis through modulation of MMP-3 activity.� 2002 Elsevier Science B.V. and International Societyof Matrix Biology. All rights reserved.

Keywords: b6 integrin; Fibronectin; Squamous cell carcinoma; Invasion

1. Introduction

Squamous cell carcinoma(SCC) of the oral cavity isthe sixth most common cancer worldwide(Silverman,1998). Every year, approximately 400 000 new cases oforal and oro-pharyngeal tumors are diagnosed(Silver-man, 1998). In the United States alone, over 30 000new cases of oral cancer are diagnosed annually. Despitetreatment advances that have resulted in reduced patientmorbidity, the overall survival rates for the disease have

*Corresponding author. Tel.:q1-415-502-4905; fax:q1-415-476-4204.

E-mail address: [email protected](D.M. Ramos).

remained consistently poor(;40%) over the past 50years(Silverman, 1998).Invasion and metastasis are the most challenging and

important aspects of tumor control and the most preva-lent cause of death in patients with oral SCC. Invasionoccurs through degradation of the basement membraneand the interstitial extracellular matrix(ECM), whichresults in cellular infiltration into the adjacent tissue.During tumor invasion, the ECM is continually remod-eled, exposing tumor cells to an ever-changing array ofmatrix molecules, including fibronectin, vitronectin, andtenascin-C(Thomas et al., 2001). Fibronectin has pre-viously been identified as a major component of the

298 D.M. Ramos et al. / Matrix Biology 21 (2002) 297–307

interstitial stroma in invasive oral SCC(Liu et al.,1998). Fibronectin is expressed in a temporally andspatially regulated fashion, and is composed of multiplesubunits. Additionally, fibronectin is capable of beingalternatively spliced, generating an even greater numberof putative adhesion molecules(Liu et al., 1998; Ramoset al., 1998). Alternative splicing also gives rise to thepossibility of generating several distinct sets of signals,due to the expression of discrete domains within eachspliced variant(Hynes, 1992). Fibronectin possessesadhesive properties that are important during develop-ment, wound healing and tumorigenesis(Morla et al.,1994). It is chemotactic for a variety of cell types,including oral SCC cells(Hynes, 1990; Ramos et al.,1998), and is produced by many cell types, includingfibroblasts, epithelial cells, endothelial cells, myoblastsand glial cells(Mosher, 1984).One way tumor cells interact with the ECM is through

the integrin family of cell surface adhesion receptors.Integrins are a family of transmembrane adhesion mol-ecules that mediate cell–cell and cell–matrix interac-tions (Hynes, 1992). Each integrin heterodimer consistsof a non-covalently associateda and b subunit. Theyprovide a direct link integrating the ECM with thecytoskeleton, integrating the outside environment withthe intracellular one, and functioning as bidirectionaltransducers of extra- and intracellular signals, potentiallymodulating tumor cell invasion(Akiyama et al., 1995).Integrins recognize the structure of the surroundingmatrix and transmit messages to the cells, leading toguided degradation of the molecular barriers to invasion.Malignant transformation is characterized by disrupt-

ed cytoskeletal organization, decreased adhesion, andaltered adhesion-dependent responses. Studies of inte-grin expression in transformed cells suggest that variousintegrin subunits may contribute either positively ornegatively to the transformed cell phenotype(Ruoslahti,1991). Changes in the expression of fibronectin-bindingintegrins have been observed in some types of trans-formed fibroblasts, whereas changes in the expressionof other integrins have been observed in different malig-nant cell types(Plantefaber and Hynes, 1989). Altera-tions in cell adhesion receptors is a general theme ofthe transformed cell type(Akiyama et al., 1990). Forexample, high levels of thea5b1 fibronectin receptorseem to correlate with a low level of transformation incertain tumors, whereas increased expression ofavb3appears to correlate positively with malignant transfor-mation in melanomas(Albelda et al., 1990). One con-sistent finding is that the spatial arrangement of integrinsin carcinomas becomes disordered. Changes in integrincell surface distribution can affect ligand-binding affinityand correlate with concomitant disorganization of thestructure of the basement membrane itself. Reducedexpression levels of thea5, a3 anda2 integrins havebeen reported in carcinomas whereas increased levels of

a6b4 appear in head, neck and skin tumors(Julianoand Varner, 1993; Giancotti and Mainiero, 1994). Itseems that several combinations of changes in theintegrin expression pattern may precede tumor formationin a given tissue. Although variable, shifts in integrinexpression during the progression of head and neckcarcinoma have been reported(Jones et al., 1997). Forexample, expression of thea6b4 complex in oral SCChas been suggested to be associated with a poor prog-nosis of this disease(Wolf et al., 1990). Others havereported that poorly differentiated oral SCC tumors losea6b4, a2b1 and a3b1 (Jones et al., 1993). Morerecently, we and others have shown that theavb6integrin is neo-expressed in oral cancer(Breuss et al.,1995; Ramos et al., 1997).Both quantitative and qualitative alterations in integrin

expression have been documented both in vitro and invivo. Some integrins are either overexpressed or nolonger expressed whereas others become phosphorylat-ed, thereby affecting their cytoskeletal and extracellularligand-binding properties(Fawcett and Harris, 1992).Shifts in specific integrin composition result in cytoske-letal rearrangements affecting cell adhesion and migra-tion as well as growth. In addition to modulating celladhesion, the integrin family of cell adhesion receptorshas been implicated in inducing MMP secretion in othercell types (Werb et al., 1989). It has recently beendemonstrated that the expression of theb6 integrinresults in MMP9 secretion, demonstrating a delicatebalance between expression of an integrin and a matrixdegrading enzyme(Niu et al., 1998). A better under-standing of the early interactions between carcinoma,stromal cells and the ECM could lead to the identifica-tion of prognostic indicators and the development ofnovel therapeutic strategies. In this study, we sought todetermine how the forced expression of theavb6integrin would affect the invasive behavior of oral SCCcells both in vitro and in vivo.

2. Results

2.1. Stable expression of the integrin b6 in oral SCCcells by retroviral transduction

SCC9 cells were retrovirally transduced with the full-length b6 cDNA and designated as the SCC9b6 cellline. FACS analysis forb6 expression was performedon the SCC9, SCC9pLXSN(blank retroviral vector)and the SCC9b6 cells. Expression ofb6 increased byapproximately 15-fold over the original SCC9 cells orthe cells expressing the blank vector(SCC9pLXSN)(Fig. 1a).We also wished to determine if the forced expression

of the b6 subunit altered the expression of other FN-binding integrins. FACS analysis using anti-av, anti-a5,anti-b1, anti-b3, and anti-b5 antibodies was next per-

299D.M. Ramos et al. / Matrix Biology 21 (2002) 297–307

Fig. 1. The effect ofb6 transduction on integrin expression in oralSCC cells. SCC9pLXSN(vector only) and SCC9b6 cells were incu-bated with antibodies toavb6 (RG69) (a). SCC9pLXSN cells arelabeled as histogram 1; SCC9b6 cells as histogram 2. Note an approx-imately 15-fold increase in reactivity of the SCC9b6 cells with RG69compared with the SCC9pLXSN cells, representing an increase inb6expression. SCC9pLXSN and SCC9b6 cells were also analyzed byFACS after incubation with antibodies toav (L230) (b); a5 (BIIG2)(c); anti-b1 (P5D2) (d); anti-b3 (LM609) (e); and anti-b5 (P1F6)(f) and analyzed by flow cytometry. Expression ofav was increasedin the b6, expressing cells(b). In contrast, expression levels ofa5(c); b1 (d); b3 (e); andb5 (f) were unchanged by the expression oftheb6 subunit.

Fig. 2. Expression ofavb6 affects oral SCC cell motility, not adhe-sion. (a) Expression ofavb6 does not affect adhesion to fibronectin.SCC9pLXSN and SCC9b6 cells were plated on fibronectin(10mgyml) for 1 h in the absence or presence of function-blocking mAbsto a5 (B2GII) and to avb6 (10D5). Loosely adherent cells wereremoved, and the number of cells firmly adhering to the substrate wasdetermined by a colorimetric assay at 595 nm. Adhesion to fibronectinwas significantly reduced in both cell lines by the addition of anti-a5antibodies but was unaffected by the addition of anti-avb6 antibodies.(b) Expression ofb6 promotes SCC cell migration on fibronectin.The 2=10 SCC9pLXSN and SCC9b6 cells were analyzed for migra-4

tion on fibronectin(10 mgyml) in the absence and presence of mAbsBIIG2 and 10D5 in a 2-h assay. Migration by SCC9b6 cells wassuppressed by both BIIG2 and 10D5 function z blocking antibodies.(c) Expression ofb6 promotes oral SCC cell invasion through areconstituted basement membrane. The 2=10 SCC9pLXSN and4

SCC9b6 cells were seeded onto RBM-coated Transwell filters in thepresence or absence of BIIG2 or 10D5. SCC9b6 cells were more thanfivefold as invasive as SCC9pLXSN cells, and this invasion was sup-pressed by the addition of mAb 10D5. Addition of BIIG2 did notaffect SCC cell invasion. The number of cells invading through theRBM was determined by counting five random fields in triplicate.Values in a, b and c represent the mean of three separateexperiments"S.E.M.

formed (Fig. 1b–f). Expression ofav increased in theSCC9b6 cells, whereas no change in the expression ofthe a5, b1, b3, or b5 subunits could be detected. Wesuggest that the increase inav expression is due to anexcess pool ofav that becomes available for associationafter retroviral transduction ofb6. Our experimentssuggest that no other integrin complexes are affected byexpression of theb6 subunit.

2.2. Expression of b6 does not affect oral SCC celladhesion to fibronectin

Short-term adhesion assays were performed by plating2=10 cells (SCC9pLXSN and SCC9b6) onto fibro-4

nectin and allowing them to attach for 1 h. After 1 h,loosely adherent cells were removed and the number ofcells remaining attached was determined by colorimetricassay at 595 nm. We found that initial attachment tofibronectin was not increased by overexpressing theb6integrin subunit in SCC9 cells(Fig. 2a).


Fig. 3. Expression ofb6 promotes SCC cell proliferation on fibro-nectin. The 1=10 SCC9pLXSN and SCC9b6 cells were seeded3

serum free onto fibronectin-coated 96-well plates. Every 24 h tripli-cate wells were analyzed. After 4 days the assay was terminated, andthe relative number of cells was determined using the MTT methodand analyzed by colorimetric assay at 595 nm. By day 4, the SCC9b6cells significantly out-proliferated the SCC9pLXSN cells. Values rep-resent means"S.E.M.

Adhesion to fibronectin was analyzed in the presenceof anti-a5 (BIIG2) and anti-avb6 (10D5) function-blocking antibodies. Adhesion of both cell lines tofibronectin was almost completely suppressed by theaddition of BIIG2 (Fig. 2a), whereas 10D5 had littleeffect on the adhesion of either cell line. These resultssuggest that initial SCC cell adhesion to fibronectin isprimarily a function ofa5b1, notavb6.

2.3. Expression of b6 enhances oral SCC cell migrationon fibronectin

To determine if expression ofb6 influences SCC cellmotility, migration assays were performed(Li et al.,1998). SCC9pLXSN and SCC9b6 cells were seededonto Transwell filters(8-mm pore size) precoated onthe undersurface with 10mgyml of fibronectin. Thelower compartment contained peritumor fibroblast con-ditioned medium as a chemoattractant. Migrationincreased more than 50% in cells transduced with theb6 cDNA (Fig. 2b). SCC9b6pLXSN cell migration wasnot affected by expression of the blank vector.We used blocking antibodies toa5 and toavb6 to

determine if they mediate SCC cell migration. Antibod-ies toa5 (BIIG2) reduced migration on fibronectin bymore than 50% in both the SCC9pLXSN and theSCC9b6 cells (Fig. 2b). Anti-avb6 (10D5) antibodieshad little effect on SCC9pLXSN cell migration onfibronectin, but reduced the migration of SCC9b6 cellsby approximately 50%(Fig. 2b). These results indicatethat botha5b1 andavb6 contribute to oral SCC cellmigration on fibronectin.

2.4. Expression of b6 enhances oral SCC cell invasionthrough a reconstituted basement membrane

We next examined tumor cell invasion of an RBM invitro. Briefly, 2=10 SCC9pLXSN and SCC9b6 cells4

were seeded onto RBM-coated filters(in serum-freemedium) and incubated at 378C. After 16 h the assayswere terminated and differences in invasion were deter-mined. The SCC9b6 cells were more than fivefold asinvasive as the mock transfected SCC9pLXSN cells(Fig. 2c).Suppression ofb6 and a5 function was analyzed

using anti-avb6 and anti-a5 function-blocking antibod-ies(10D5 and BIIG2, respectively). Invasion was almostcompletely suppressed by the addition of the 10D5antibodies, whereas BIIG2 had virtually no effect(Fig.2c). These results indicate that the invasion of the oralSCC cells through an RBM is mediated throughavb6,not a5b1.

2.5. Expression of b6 results in increased proliferationby oral SCC cells

We performed proliferation assays on fibronectin todetermine whetherb6 expression alters oral SCC cell

growth in vitro. The 1=10 SCC9pLXSN and SCC9b63

cells were plated, serum free, into individual wells of a96-well plate previously coated with 10mgyml offibronectin(Fig. 3). By 24 h the SCC9b6 cells had out-proliferated the SCC9pLXSN cells, and continued to doso throughout the 4-day assay. By day 4, the SCC9b6cell proliferation was approximately 30% greater thanthat of the mock transfected SCC9pLXSN cells, indi-cating that expression ofb6 enhances the growth of theoral SCC cells on fibronectin.

2.6. Expression of b6 enhances oral SCC tumor growthand invasion in nude mice

To evaluate the effect ofb6 on oral tumor growth invivo, we injected 2=10 SCC9pLXSN or SCC9b6 cells5

into the floor of the mouth of 12 athymic mice per cellline as previously described(Ramos et al., 1997). Tumorgrowth was then followed for 6 weeks. Tumors formedby the SCC9b6 cells (Fig. 4a) grew five-fold largerthan those formed by the SCC9pLXSN cells(Fig. 4b).The SCC9b6 tumors weighed on average fivefold more(;0.5 g) than the SCC9 tumors(;0.1 g) (Fig. 4c).These results indicate that a fundamental differenceexists between tumor growth ofb6-positive andb6-negative oral SCC cells.The harvested tumors were snap-frozen in liquid

nitrogen, cryostat sectioned(4-mm sections) and proc-essed for routine histochemistry. A significant differencein invasive behavior between the two tumor types wasnoted (Fig. 5). Specifically, the tumors formed by theSCC9pLXSN cells grew adjacent to, but did not invade,the underlying muscle(Fig. 5a). A basement membraneencapsulated the SCC9pLXSN tumor, whereas thetumors formed by the SCC9b6 cells invaded extensively,


Fig. 4. Expression ofb6 enhances oral SCC cell growth in vivo. The2=10 SCC9b6 (a) or SCC9pLXSN(b) cells were injected transor-5

ally into the floor of the mouth of 12 athymic mice per cell line. After6 weeks the animals were killed, and the tumors were excised andevaluated for growth in vivo. Visual examination of the animals dem-onstrated that the SCC9b6 cells produced significantly larger tumorsthan those formed by SCC9 cells. To quantitate differences in tumordevelopment, the tumors were weighed(c). The tumors formed bythe SCC9b6 cells weighed on average fivefold greater(mean 0.5g"S.E.M.) than those formed by theb6-negative SCC9pLXSN cells(mean 0.1 g"S.E.M.).

Fig. 5. Expression ofb6 increases oral SCC cell invasion. Tumorsformed by injection of SCC cells were harvested and processed forimmunohistochemistry by staining with hematoxylin and eosin. Thefrozen sections were then evaluated microscopically for the extent oflocal invasion. The tumors formed by the SCC9b6 cells invadedextensively into the underlying muscle(a) whereas those producedby the SCC9pLXSN cells did not invade but became encapsulatedand retained within a basement membrane(b).

infiltrating into the underlying muscle(Fig. 5b). Theseresults indicate that expression ofb6 both enhancestumor growth in vivo and promotes oral SCC invasion.

2.7. Expression of b6 suppresses organization of fibro-nectin matrices

Interaction with the ECM is an important property oftumor cell invasion. Therefore, to evaluate the effect ofSCC expression ofb6 on the organization of a fibro-nectin matrix, we grew co-cultures of SCC9pLXSN cellswith peritumor fibroblasts(PTF) or SCC9b6 cells withPTF on glass coverslips for 16 h at 378C, and thenexamined the cultures for fibronectin matrix assemblyusing immunofluorescence microscopy(Ramos et al.,1998). A highly organized FN matrix was deposited bythe SCC9pLXSNyPTF co-cultures when analyzed usinganti-FN antibodies(Fig. 6a), whereas the fibronectinmatrix organized by the SCC9b6yPTF co-cultures wasreduced and less organized(Fig. 6b).To determine whetherb6 activity was responsible for

the decreased fibronectin assembly, we examined the

co-cultures in the presence of theavb6 function-block-ing mAb, 10D5. When SCC9b6yPTF co-cultures weregrown in the presence of 10D5, fibronectin matrixassembly was almost completely restored to that seen inSCC9pLXSNyPTF co-cultures(Fig. 6c). These resultsclearly show that the expression ofb6 negatively influ-ences the organization of the fibronectin extracellularmatrix, and that suppressing its function can restorefibronectin matrix assembly.

2.8. Expression of b6 induces the expression of activeMMP-3

MMP expression(specifically MMP-9) as a result ofb6 expression has been described previously by others(Agrez et al., 1999). Gelatin zymography was performedon conditioned medium from SCC9pLXSN(mock trans-fected), SCC9 parental, and the SCC9b6 cells(Fig. 7a,upper panel). Zones of clearing represent relativeenzyme activity. All three cell lines expressed roughlyequivalent amounts of cleared zones with the apparentmolecular mass equivalent to MMP-9 and MMP-2. Todetect putative MMP-3 activity, we next performedcasein zymography on the conditioned medium fromthe same cell lines and determined that the SCC9b6cells expressed more than 10-fold greater amounts ofcleared zones corresponding to MMP-3 compared to theSCC9pLXSN(mock transfected) and the SCC9 parentalcell lines(Fig. 7a, lower panel).To verify the identity of the cleared zones identified

by casein zymography, we performed Western blottingusing anti-MMP-3 antibodies. Conditioned medium fromthe SCC9b6 cell line expressed higher levels of the 59-kDa proenzyme(Fig. 7b). Additionally, only the con-ditioned medium from the SCC9b6 cells expressed the


Fig. 6. Expression ofavb6 suppresses fibronectin matrix assemblyby SCCyPTF co-cultures. SCC9pLXSN(a) or SCC9b6 (b,c) cellswere co-cultured with peritumor fibroblasts(PTF) on glass coverslipsat 37 8C for 24 h. In some experiments, SCC9b6yPTF co-cultureswere grown in the presence of the mAb 10D5 toavb6 (c). Thecultures were fixed, stained with mAbs to fibronectin, and examinedby immunofluorescence microscopy. Fibronectin matrix assembly bythe SCC9yPTF co-cultures(a) was substantially decreased in the co-cultures of SCC9b6yPTF cells(b). Note that the addition of 10D5almost completely restored fibronectin matrix assembly in theSCC9b6yPTF co-cultures(c).

Fig. 7. Expression ofb6 increases MMP-3, but not MMP-2 or MMP-9 expression in oral SCC cells.(a) Conditioned medium fromSCC9pLXSN(lane 1), SCC9 parental(lane 2), and SCC9b6 cells(lane 3) was analyzed by gelatin zymography(upper panel).Although all three cell lines express MMP-9 and MMP-2, there is noapparent change in the expression as a result ofb6 expression. Incontrast, SCC9b6 conditioned medium analyzed by casein zymogra-phy demonstrated a dramatic increase in zones cleared with the appar-ent molecular mass of MMP3 compared with SCC9 and SCC9pLXSN(lane 3, lower panel). (b) To confirm the identity of MMP3, Westernblotting was performed on conditioned medium from SCC9pLXSNand SCC9b6 cells. There was a significant increase in the productionand activation of MMP3 in SCC9b6 cell conditioned medium asdetected with anti-MMP3 antibodies(AB-1) when compared withSCC9pLXSN(b). This activation was suppressed by growing the cellsin the presence of GM6001, an inhibitor of MMPs. These resultssuggest that expression ofb6 specifically results in an increase inMMP3 expression and activation.

activated 45- and 29-kDa forms. Conditioned mediumfrom cultures grown in the presence of the MMPinhibitor GM6001 was also analyzed by Western blot-ting. Expression of the 45- and 29-kDa activated formswas inhibited, demonstrating suppression of MMP-3activation. These results indicate that expression ofb6promotes both the production and activation of MMP-3and this activation may be partly responsible for the

increased motility and invasion shown by the SCC9b6cell line.

2.9. Incubation with GM6001 restores fibronectin matrixassembly

Our experiments thus far showed that expression ofb6 by SCC cells negatively affects fibronectin matrixassembly in SCC9b6yPTF co-cultures. Our data alsodemonstrate that expression of theb6 integrin stimulatesactivation of MMP-3. To examine whether suppressionof MMP-3 might affect the organization of fibronectinmatrices, we grew SCC9b6yPTF co-cultures in theabsence(Fig. 8a) or presence(Fig. 8b) of GM6001(10mM) for 16 h. After 16 h, fibronectin matrix organiza-tion was evaluated by immunofluorescence microscopy


Fig. 8. Incubation of SCC9b6yPTF co-cultures with GM6001 increases fibronectin matrix assembly. SCC9b6yPTF co-cultures were grown for24 h in the absence(a) or presence(b) of 10 mM GM6001 to determine if MMP production was responsible for degradation of fibronectinmatrix assembly. SCC9b6yPTF co-cultures grown in the presence of GM6001 for 24 h(b) showed an increase in fibronectin matrix assembly,suggesting that assembly is modulated by MMP production, presumably MMP3.

using anti-fibronectin antibodies. We found that incu-bation with GM6001 resulted in increased fibronectinmatrix assembly(Fig. 8b). These results indicate thatMMP activation(MMP-3) decreases fibronectin matrixassembly by the SCC9b6yPTF co-cultures.

3. Discussion

In this study, we focused on how the forced expressionof the b6 subunit in a poorly invasive oral SCC cellline (SCC9) influences cell growth and invasion in vitroand in vivo.Our findings substantiate the work of others indicating

that b6 enhances the invasiveness of epithelial-derivedtumor cells and suggest that it may increase theirmetastatic potential(Thomas et al., 2001). Studies byother laboratories have shown that forced expression ofavb6 in malignant keratinocytes promotes invasion andleads to increased capacity for migration towards FN(Thomas et al., 2001).Tumor cells must adhere to the ECM before they can

modify and migrate upon it. Therefore, it was surprisingto find that adhesion to fibronectin was not increasedby expression of theavb6 FN receptor. This suggeststhat the initial adhesion to FN is not the primary functionof avb6 in oral SCC. This agrees with the work ofother investigators who showed thatavb6 weakly pro-motes adhesion to fibronectin, functioning in more of asecondary fashion when other, stronger fibronectinreceptors, such asa5b1, are present(Koivisto et al.,2000).Huang et al.(1998) also demonstrated that theavb6

complex promotes keratinocyte cell migration. Whenthe motility of the SCC9 and the SCC9b6 cell lines wasevaluated, we found that SCCb6 cell migration onfibronectin as well as invasion through a reconstitutedbasement membrane were both significantly increased.On fibronectin, migration required both theavb6 andthe a5b1 fibronectin receptors. We suggest that theb6integrin may be more important in mediating post-

ligand-binding events such as cell growth, motility andpossible invasion than in mediating initial adhesion. Ourprevious work(Ramos et al., 1998) does not confirmthat by Hamidi et al.(2000), who show thatavb6 isup-regulated in a portion of oral leukoplakias. We couldnot detect expression at this stage. Perhaps the differencein observations can be explained by variations in samplesize or antibody–antigen access due to differences inimmunohistochemical protocols.Our evaluation of several parameters in vitro(prolif-

eration, migration on fibronectin, invasion of an RBM)suggested that expression ofb6 may be important forinvasion by oral SCC cells. A primary characteristic ofneoplastic disease is unregulated cell growth. Othershave shown that integrins transmit signals controllingcell proliferation (Giancotti and Ruoslahti, 1990). Wealso found that the expression of theb6 integrin enhanc-es cell proliferation by 30% in oral SCC cells. In vivo,a variety of factors and events affect the ability of tumorcells to grow and expand into adjacent tissue. Toassociate proliferation in vitro with growth in vivo, weperformed tumorigenesis assays. Both the SCC9b6 andthe mock transfected SCC9pLXSN cells were injectedtransorally into athymic mice and evaluated for tumordevelopment. Although athymic animals do not com-pletely reproduce the events encountered by tumor cellsin the human(e.g. immune surveillance), other requi-rements such as the necessity for angiogenesis, degra-dation of complex matrices, and local invasion intomuscle and bone can be recapitulated in this system.Our experiments showed that expression of theb6

integrin enhanced tumor growth fivefold when comparedwith growth of the original cell line, providing strongevidence for the role ofb6 as a possible tumor promoter.More importantly, the SCC9b6 cells invaded deep intothe underlying muscle, whereas the SCC9 cells grewadjacent to the muscle but did not invade. Using gelzymography, we found that the SCC9, SCC9pLXSNand the SCCb6 cells expressed equivalent amounts ofMMP-2 and MMP-9. In contrast, when conditioned


medium from the same three cell lines was evaluatedby casein zymography, only the SCC9b6 cells expressedactivated MMP-3. The identity of MMP-3 in the caseingel was verified using Western blotting with anti-MMP-3 antibodies(AB-1). To attempt to understand howinvasion was increased in theb6-expressing cells, wecompared the expression of MMP-3 in both cell types.Our experiments indicate that the expression ofb6results in the activation of MMP-3. Others have shownthat ligation of thea5b1 complex results in the up-regulation of MMP-3 and MMP-1(Werb et al., 1989).More recently, it was demonstrated that expression ofavb6 resulted in increased expression of MMP-9 incolon carcinoma cells(Agrez et al., 1999). Expressionof these matrix-degrading enzymes would then facilitateinvasion of the ECM through extensive remodelingfollowed by enhanced cell motility. It is possible thatwhen b6 binds to its ligand, there is a subsequentincrease in one or more of these MMPs locally, directlydegrading the surrounding ECM. We suggest that theexpression of theavb6 complex by SCC cells may becritical to the local and distant invasion that is charac-teristic of this type of tumor. These results also suggestthat whenb6 is expressed, a series of events occurs,resulting in a quicker-growing and more invasive formof this disease.The microenvironment, an important part of the inva-

sive process, provides a substrate for cell adhesion, avehicle by which to transmit signals, and a reservoir fora wealth of growth factors with which tumor cells mayinteract. Stromal cell–tumor cell interactions are impor-tant for maintaining the existing ECM architecture aswell as reorganizing the environment to facilitate theprogression of oral SCC. During the invasive process,the ECM is continually remodeled through degradationof the existing matrix and replacement with novelmolecules. Specific molecules deposited within theestablished tumor environment can potentially modifyexisting signals, resulting in altered cell behavior. Ourprevious work showed that suppression of specific inte-grin function affected both tenascin-C and fibronectinmatrix organization(Ramos et al., 1998). In this study,we used b6-negative andb6-positive cell lines todetermine if theb6 integrin affects the deposition of afibronectin matrix. Co-cultures of SCC9pLXSN cellsand PTF grown under serum-free conditions for 24 horganized a dense, fibrillar fibronectin matrix. In con-trast, co-cultures of SCC9b6 cells and PTF deposited aless extensive fibronectin matrix in which the fibrilswere truncated and less organized. The correlationbetween fibronectin matrix deposition andb6 expressionwas further supported when fibronectin matrix assemblywas restored by the addition ofavb6 (10D5) function-blocking antibodies to the co-cultures. We suggest, asothers have, that integrin–ligand interactions result inalterations to gene expression of MMPs, thereby altering

the immediate environment through degradation andpossible deposition of novel molecules.We speculate that the modulation of extracellular

matrix ligands secondary to the expression of specificadhesion receptors may promote oral SCC progressionby reducing the presence of adhesion-promoting mole-cules such as fibronectin. A reduction in deposition offibronectin could theoretically promote tumor cell motil-ity by reducing the availability of an adhesive ligand,replacing it with an anti-adhesive molecule, such astenascin-C (unpublished results). That fibronectinassembly was restored by the addition of function-blocking antibodies toavb6 further implicates thisintegrin complex as a potential negative regulator offibronectin matrix assembly. Our hypothesis that fibro-nectin matrix assembly could be modulated by MMP-3was supported by the increased expression of the acti-vated form of MMP-3 in theb6-expressing cells. Theclose association betweenb6 and MMP-3 expressionwas further demonstrated using the general MMP inhib-itor GM6001: when SCC9b6yPTF co-cultures weregrown in its presence, fibronectin matrix organizationwas substantially increased, indicating that MMP expres-sion (presumably MMP-3) is involved in the observedreduction of fibronectin matrix assembly seen in theSCC9b6yPTF co-cultures. Together, these data suggestthatb6 promotes oral SCC growth and invasion throughup-regulation of MMP-3, which is partially responsiblefor reducing fibronectin matrix assembly.Although it is probable that ultimately several cell-

surface receptors contribute to the cancerous phenotype,our study provides evidence that theavb6 integrin isup-regulated in oral cancer and is a predominant playerin the migration of these cells as well as in their reduceddeposition and assembly of fibronectin matrices. Wesuggest that these events contribute to the poor outcomeassociated with this disease. The increased growth andinvasion associated withavb6 expression may be pre-dictive of metastatic progression in oral SCC.

4. Experimental procedures

4.1. Cell culture

The poorly invasive human oral SCC9 cell line,derived from a tongue lesion(Rheinwald and Beckett,1981), was a generous gift of Dr Randall Kramer,University of California at San Francisco(UCSF). TheSCC9 cell line is anchorage-dependent and tumorigenicin nude mice, and produces epidermal keratins and lowlevels of involucrin. Peri-tumor fibroblasts(PTF) wereisolated from excess material taken during head andneck surgery at UCSF. Cells were cultivated in Dulbec-co’s modified Eagle’s medium(DMEM) (45%) andHam’s F12 medium(45%) with 10% fetal calf serum.For co-culture experiments, the PTF and SCC cells were


grown in a 2:1 ratio and plated at 30 000 cellsycm2

onto glass coverslips. For production of PTF conditionedmedium, confluent fibroblasts were washed three timeswith phosphate-buffered saline(PBS) and incubated inDMEM without serum for 72 h. The conditioned medi-um was collected and filtered through 0.22-mm-porefilters (Millipore Corp.) and stored aty20 8C. In someexperiments the hydroxamic acid matrix metalloprotei-nase (MMP) inhibitor 3-(N-hydroxycarbamoyl)-2(R)-isobutylpropionyl-L-tryptophan methylamide(GM6001)was dissolved at a concentration of 100 mM in DMSOand added to the culture medium at a final concentrationof 10 mM. GM6001 is a general inhibitor of all MMPswith dissociation constant of enzyme-inhibitor complex-es(K ) values of 100 nM(Gijbels et al., 1994).i

4.2. Reagents

Mouse monoclonal antibodies(mAbs) to avb6(RG69 and 10D5) were a generous gift of Dr DeanSheppard(UCSF). Rat mAbs to thea5 subunit(BIIG2)were provided by Dr Caroline Damsky(UCSF). Rabbitpolyclonal antibody to fibronectin was purchased fromSigma Chemical(cat.�7338). GM6001 was kindly pro-vided by Dr Zena Werb(UCSF).

4.3. Retroviral transduction experiments

The b6 sense retroviral expression vector was gener-ated using the Retro-X system(Clontech, Palo Alto,CA) (Li et al., 1998). The b6 cDNA was a generousgift of Dr Dean Sheppard(UCSF). The cDNA wasinserted in the sense orientation into the multiple cloningsite (MCS) of the retroviral vector pLNCX. The vectorwas transfected into the PT67 packaging cell line usingLipofectAMINE (Life Technologies, Inc.). Conditionedgrowth medium containing secreted retrovirus was col-lected(Li et al., 1998) and the retroviral endpoint titerswere determined on NIH3T3 cells. After G418 selection,the supernatant was supplemented with 8mgyml ofpolybrene and sterile filtered. The SCC9 cells wereincubated with PT67 cell conditioned medium contain-ing the b6 sense retroviral construct for 24 h. Trans-fected cells expressing the retrovirus-encoded cDNAwere selected with G418 growth medium. Pooled cul-tures of G418-resistant cells were isolated by fluores-cence-activated cell sorting(FACS) using mAb toavb6(RG69). Cells overexpressing theb6 subunit weredesignated the SCC9b6 cell line.

4.4. Gelatin and casein zymography

MMP-2, MMP-9, and MMP-3 were analyzed using10% SDS–substrate gels. To detect MMP-2 and MMP-9 activity, 1 mgyml gelatin was added to the gel.Similarly, to detect MMP-3 activity, 0.5 mgyml of casein

was added. The conditioned medium was used withoutconcentration for gelatin zymography, whereas the medi-um was concentrated 10-fold for casein zymography.Tumor cell conditioned medium was collected underserum-free conditions and mixed with substrate gelbuffer (10% SDS, 50% glycerol, 25 mM Tris–HCl, pH6.8 and 0.1% bromophenol blue) and 20ml was addedonto the gel without prior boiling. Following electropho-resis, the gels were washed twice in 2.5% Triton-X-100for 30 min at room temperature to remove the SDS.The gels were then incubated at 378C overnight insubstrate buffer containing 50 mM Tris–HCl and 5 mMCaCl at pH 8.0. The gels were stained with 0.15%2

Coomassie blue R250(BioRad, CA) in 50% methanol,10% glacial acetic acid for 20 min at room temperatureand de-stained in the same solution without Coomassieblue. The gelatin-degrading enzymes were identified asclear bands against the blue background of the stainedgel.

4.5. Western blotting

Cells were seeded at 70% confluence on fibronectinwithout serum for 24 h in the presence or absence of10mM GM6001. The conditioned medium was collectedand concentrated 50-fold using a 10-kDa cutoff Amiconfilter (Millipore, Beverly, MA). Protein concentrationswere determined by BCA protein assay kit(Pierce,Rockford, IL). 35mg was loaded per well. The sampleswere then separated by SDS–polyacrylamide gel elec-trophoresis(PAGE) followed by Western blotting usingcommercially available mouse monoclonal anti-MMP-3antibodies (AB-1, Oncogene). AB-1 recognizes theproenzyme(61 kDa) as well as the activated forms(48and 29 kDa) of the enzyme.

4.6. Flow cytometry

Cultured SCC9 and SCC9b6 cells were detached with2 mM EDTA:0.05% bovine serum albumin(BSA) inPBS and washed with PBS containing 1% normal goatserum. The cell suspensions were then incubated withanti-avb6 (RG69) (1:10) or anti-a5 (BIIG2) (1:100)mAbs in PBS for 45 min at 48C. After washing, thecells were incubated with secondary fluorescein isothio-cyanate(FITC)-labeled goat anti-mouse IgG or goatanti-rat IgG (Jackson ImmunoResearch) for 20 min.After a final wash, the cells were fixed in 0.1% formal-dehyde for 10 min at 48C and then analyzed by flowcytometry (FACScan, UCSF). A negative control wasprocessed as described above, using an irrelevant mousemAb as the primary antibody. Cell analysis was gatedon forward and size scatter intensities. The results werepresented in single-parameter flow cytometrichistograms.


4.7. Cell adhesion assay

Cell adhesion to immobilized fibronectin was meas-ured as previously described(Ramos et al., 1991, 1997).Briefly, individual wells of polystyrene 96-well flat-bottom microtiter plates were coated with 10mgyml ofpurified fibronectin overnight at 48C. The wells werewashed, and non-specific binding was then blocked with0.1% BSA in PBS for 30 min. Cells were detachedfrom culture dishes by treatment with 2 mMEDTA:0.05% BSA in PBS for 10 min at 378C, washed,and resuspended in serum-free DMEM(containing 200mgyml CaCl , 200mgyml MgCl ) with 0.1% BSA. The2 2

cells were plated at 2=10 cellsywell in 50 ml of4

medium(with or without antibodies) and then incubatedat 378C in a humidified 5% CO incubator for 1 h. The2

assay was terminated by washing, and the number ofadherent cells was then determined using a microcolor-imetric assay for crystal violet at 595 nm. The data wereexpressed as the mean"S.E.M. of three separate exper-iments, with each performed in triplicate.

4.8. Migration and invasion assay

Migration assays were performed as previouslydescribed(Li et al., 1998). Cell migration was assayedusing 24-well, 6.6-mm-i.d. Transwell cluster plates(8.0-mm pore size; Costar, Cambridge, MA) as described(Li et al., 1998). Briefly, the lower surfaces of the filterswere coated with fibronectin(10 mgyml) overnight at4 8C. Filters were then washed and blocked using 0.1%BSA. Cells were removed from tissue culture disheswith 2 mM EDTA as described above, then pelleted andresuspended in serum-free medium. Next, 300ml offibroblast conditioned medium was placed in the lowerchamber as a chemoattractant(Ramos et al., 1998).Then, 2=10 tumor cells were placed onto the upper4

surface of the filter, and the chambers were incubatedat 378C for 2 h. In some experiments, function-blockingmAbs toavb6 (10D5) or toa5 (BIIG2) were incubatedwith the cells. When the experiment was terminated, thefilters were fixed, the upper surface was wiped cleanand stained, and the number of cells crossing the filterwas determined. Five random fields of each Transwellfilter were counted at=200 power, and the cell numberswere calculated as total migrated cells per filter.For invasion of a reconstituted basement membrane

(RBM), the upper surface of the filter was coated witha thin layer of RBM for 1 h at 378C to allow forpolymerization. Fibroblast conditioned medium wasplaced in the lower chamber as a chemoattractant(300ml) as described for migration. Then, 2=10 cells were4

seeded onto the RBM-coated filters, and the chamberswere incubated at 378C for 16 h. Upon termination ofthe experiment, the filters were processed as describedfor migration. The number of cells invading to the

undersurface was determined as described above. Insome assays, function-blocking mAbs toavb6 (10D5)or a5 (BIIG2) were added with the cells.

4.9. Cell proliferation assay

Cell proliferation was assayed using the MTT(3-w4,5-dimethylthiazol-2ylx-2,5-diphenyltetrazolium bro-mide) assay. The 1=10 SCC9 and SCC9b6 cells were3

seeded in individual wells of a 96-well plate coatedwith 10 mgyml of fibronectin. At predetermined timepoints (time zero, 24 h, 48 h, 72 h, 96 h), 100 ml ofMTT (1 mgyml) was added to each well, and the plateswere incubated in a 5% CO incubator at 378C for 42

h. Next, 200ml of DMSO was added to each well, andthe plates were incubated at room temperature for anadditional 30 min. The samples were analyzed bycolorimetric assay at 595 nm. The experiment wasrepeated in triplicate.

4.10. In vivo tumor growth assay

Tumor cells were injected into the floor of the mouthof 12 athymic mice per cell line to compare tumorgrowth generated byb6-positive andb6-negative oralSCC cells. Athymic mice were anesthetized peritoneallywith Avertin (0.015 ml of a 2.5% solutionyg body�

wt.).The 2=10 tumor cells in 50ml of DMEM were5

deposited into the submental space using a 30-gauge,0.5-inch long needle(Becton Dickinson & Co.). Theneedle was carefully inserted into the subcutaneoustissues in the midline. The point of injection wasimmediately posterior to the mandibular symphysis andbetween the bellies of the digastric muscle. The cellswere deposited over 30 s.When the mice regained consciousness, they were

placed back in cages and evaluated twice weekly fortumor formation. After 4 weeks, when the tumorsbecame palpable, the animals were monitored threetimes weekly to assure their ability to take food. Theanimals were then sacrificed and the lesions weredissected, weighed, embedded in Tissue-Tek optimalcutting temperature(OCT) compound, and quick-frozeninto liquid nitrogen.

4.11. Immunohistochemistry

Briefly, 4-mm cryostat sections of tumor tissue sam-ples were air dried for 30 min, fixed iny18 8C acetonefor 5 min, and air dried again. Endogenous peroxidaseactivity was blocked with Peroxoblock Solution(ZymedLaboratories) for 45 s at room temperature. After rinsingin PBS for 15 min to remove the OCT, sections wereincubated with 0.5% caseiny0.05% thimerosalyPBS for30 min at room temperature and stained with hematox-


ylin and eosin. Sections were subsequently dehydratedthrough graded alcohol solutions, transferred to xylene,and mounted with Permount(Fisher Scientific). Thetissue sections were evaluated microscopically for inva-sion into the underlying tissue.

5. Nomenclature

ECM: extracellular matrixFACS: fluorescence-activated cell sortingFN:: fibronectinmAb: monoclonal antibodyMMP: matrix metalloproteinasePTF: peritumor fibroblastsRBM: reconstituted basement membraneSCC: squamous cell carcinomaSCC9b6: SCC9 cells retrovirally transduced to express

b6 integrin

Acknowledgments

Grant Support: National Institute for Craniofacial andDental Research(R01 DE12856, P01 DE13904, R29DE01193) and the University of California CancerResearch Coordinating Committee(to D.M.R.). We wishto thank Ms. Evangeline Leash for editing themanuscript.

References

Agrez, M., Gu, X., Turton, J., et al., 1999. Theavb6 integrin inducesgelatinase B secretion in colon cancer cells. Int. J. Cancer 81,90–97.

Akiyama, S.K., Larjava, H., Yamada, K.M., 1990. Differences in thebiosynthesis and localization of the fibronectin receptor in normaland transformed cultured human cells. Cancer Res. 50, 1601–1607.

Akiyama, S.K., Olden, K., Yamada, K.M., 1995. Fibronectin andintegrins in invasion and metastasis. Cancer Metastasis Rev. 14,173–189.

Albelda, S.M., Mette, S.A., Elder, D.E., et al., 1990. Integrin distri-bution in malignant melanoma: association of the beta 3 subunitwith tumor progression. Cancer Res. 50, 6757–6764.

Breuss, J.M., Gallo, J., DeLisser, H.M., et al., 1995. Expression ofthe beta 6 integrin subunit in development, neoplasia and tissuerepair suggests a role in epithelial remodeling. J. Cell Sci. 108,2241–2251.

Fawcett, J., Harris, A.L., 1992. Cell adhesion molecules and cancer.Curr. Opin. Oncol. 4, 142–148.

Giancotti, F.G., Mainiero, F., 1994. Integrin-mediated adhesion andsignaling in tumorigenesis. Biochim. Biophys. Acta 1198, 47–64.

Giancotti, F.G., Ruoslahti, E., 1990. Elevated levels of thea5b1fibronectin receptor suppress the transformed phenotype of Chinesehamster ovary cells. Cell 60, 849–859.

Gijbels, K., Galardy, R.E., Steinman, L., 1994. Reversal of experi-mental autoimmune encephalomyelitis with a hydroxamate inhibi-tor of matrix metalloproteases. J. Clin. Invest. 94, 2177–2182.

Hamidi, S., Sato, T., Kainulainen, T., Epstein, J., Lerner, K., Larjava,H., 2000. Expression ofavb6 integrin in oral leukoplakia. Br. J.Cancer 82, 1433–1440.

Huang, X., Wu, J., Spong, S., Sheppard, D., 1998. The integrinalphavbeta6 is critical for keratinocyte migration on both its knownligand, fibronectin, and on vitronectin. J. Cell Sci. 111, 2189–2195.

Hynes, R.O., 1990. In: Rich, A.(Ed.), Fibronectins. Springer Seriesin Molecular Biology. Springer-Verlag, New York, pp. 393–538.

Hynes, R.O., 1992. Integrins: versatility, modulation, and signalingin cell adhesion. Cell 69, 11–25.

Jones, J., Sugiyama, M., Watt, F.M., Speight, P.M., 1993. Integrinexpression in normal, hyperplastic, dysplastic, and malignant oralepithelium. J. Pathol. 169, 235–243.

Jones, J., Watt, F.M., Speight, P.M., 1997. Changes in the expressionof alpha v integrins in oral squamous cell carcinomas. J. OralPathol. 26, 63–68.

Juliano, R.L., Varner, J.A., 1993. Adhesion molecules in cancer: therole of integrins. Curr. Opin. Cell Biol. 5, 812–818.

Koivisto, L., Grenman, R., Heino, J., Larjava, H., 2000. Integrinsa5b1, avb1, andavb6 collaborate in squamous carcinoma cellspreading and migration on fibronectin. Exp. Cell Res. 255, 10–17.

Li, X., Chen, B.L., Blystone, S.D., McHugh, K.P., Ross, F.P., Ramos,D.M., 1998. Differential expression ofav integrins in K1735melanoma cells. Invasion Metastasis 18, 1–14.

Liu, H., Chen, B., Zardi, L., Ramos, D.M., 1998. Soluble fibronectinpromotes migration or oral squamous cell carinoma cell. Int. J.Cancer 78, 261–267.

Morla, A., Zhang, Z., Ruoslahti, E., 1994. Superfibronectin is afunctionally distinct form of fibronectin. Nature 367, 193–196.

Mosher, D.F., 1984. Physiology of fibronectin. Ann. Rev. Med. 35,561–575.

Niu, J., Gu, X., Turton, J., Meldrum, C., Howard, E.W., Agrez, M.,1998. Integrin mediated signaling of gelatinase B secretion in coloncarcinoma cells. Biochem. Biophys. Res. Commun. 249, 287–291.

Plantefaber, L.C., Hynes, R.O., 1989. Changes in integrin receptorson oncogenically transformed cells. Cell 56, 281–290.

Ramos, D.M., Cheng, Y.F., Kramer, R.H., 1991. Role of laminin-binding integrin in the invasion of basement membrane matricesby fibrosarcoma cells. Invas. Metastas. 11, 125–138.

Ramos, D.M., Chen, B.L., Boylen, K., et al., 1997. Stromal fibroblastsinfluence oral squamous-cell–carcinoma cell interactions with ten-ascin-C. Int. J. Cancer 72, 369–376.

Ramos, D.M., Chen, B.L., Regezi, J., Zardi, L., Pytela, R., 1998. TN-C matrix assembly in oral squamous cell carcinoma. Int. J. Cancer75, 680–687.

Rheinwald, J.G., Beckett, M.A., 1981. Tumorigenic keratinocyte linesrequiring anchorage and fibroblast support cultures from humansquamous cell carcinomas. Cancer Res. 41, 1657–1663.

Ruoslahti, E., 1991. Integrins J. Clin. Invest. 87, 1–5.Silverman, S., Jr., 1998. Oral Cancer, 4th ed. B.C. Decker, Ontario.Thomas, G., Lewis, M.P., Whawell, S.A., et al., 2001. Expression ofthe avb6 integrin promotes migration and invasion in squamouscell carcinoma cells. J. Invest. Dermatol. 117, 67–73.

Werb, Z., Tremble, P.M., Behrendtsen, O., Crowley, E., Damsky,C.H., 1989. Signal transduction through the fibronectin receptorinduces collagenase and stromelysin gene expression. J. Cell Biol.109, 877–889.

Wolf, G.T., Carey, T.E., Schmaltz, S.P., et al., 1990. Altered antigenexpression predicts outcome in squamous cell carcinoma of thehead and neck. J. Natl. Cancer Inst. 82, 1566–1572.

Expression of integrin β6 enhances invasive behavior in oral squamous cell carcinoma

Documents

Transcript of Expression of integrin β6 enhances invasive behavior in oral squamous cell carcinoma