human fibronectin and mmp-2 collagen binding domains compete

Matrix Biology 21 (2002) 399–414

0945-053X/02/$ - see front matter� 2002 Elsevier Science B.V. and International Society of Matrix Biology. All rights reserved.PII: S0945-053XŽ02.00032-X

Human fibronectin and MMP-2 collagen binding domains compete forcollagen binding sites and modify cellular activation of MMP-2

Bjorn Steffensen *, Xiaoping Xu , Pamela A. Martin , Gustavo Zardenetaa, a a b

Department of Periodontics MC 7894, University of Texas Health Science Center at San Antonio, 7703 Floyd Curl Drive, San Antonio,a

TX 78229-3900, USADepartment of Oral and Maxillofacial Surgery, University of Texas Health Science Center at San Antonio, San Antonio, TX 78229-3900, USAb

Received 18 December 2001; accepted 23 April 2002

Abstract

The region of fibronectin(FN) surrounding the two type II modules of FN binds type I collagen. However, little is knownabout interactions of this collagen binding domain with other collagen types or extracellular matrix molecules. Among severalexpressed recombinant(r) human FN fragments from the collagen binding region of FN, only rI6–I7, which included the twotype II modules and both flanking type I modules, bound any of several tested collagens. The rI6–I7 interacted specifically withboth native and denatured forms of types I and III collagen as well as denatured types II, IV, V and X collagen with apparentK values of 0.2–3.7=10 M. Reduction with DTT disrupted the binding to gelatin verifying the functional requirement fory7d

intact disulfide bonds. The FN fragments showed a weak, but not physiologically important, binding to heparin, and did not bindelastin or laminin. The broad, but selective range of ligand interactions by rI6–I7 mirrored our prior observations for the collagenbinding domain(rCBD) from matrix metalloproteinase-2(MMP-2) wJ. Biol. Chem. 270(1995) 11555x. Subsequent experimentsshowed competition between rI6–I7 and rCBD for binding to gelatin indicating that their binding sites on this extracellular matrixmolecule are identical or closely positioned. Two collagen binding domain fragments supported cell attachment by ab1-integrin-dependent mechanism although neither protein contains an Arg–Gly–Asp recognition sequence. Furthermore, activation of MMP-2 and MMP-9 was greatly reduced for HT1080 fibrosarcoma cells cultured on either of the fibronectin fragments compared tofull-length FN. These observations imply that the biological activities of FN in the extracellular matrix may involve interactionswith a broad range of collagen types, and that exposure to pathologically-generated FN fragments may substantially alter cellbehavior and regulation.� 2002 Elsevier Science B.V. and International Society of Matrix Biology. All rights reserved.

Keywords: Fibronectin; Collagens; Matrix metalloproteinase; MMP; MMP-2; Gelatinase A; Enzyme activation; Cell attachment

Abbreviations: BSA, bovine serum albumin; DMSO, dimethyl sulfoxide; DTT, dithiothreitol; ECL, enhanced chemiluminescence; ECM,extracellular matrix; ELISA, enzyme-linked immunoabsorbance assay; FN, fibronectin; Gelatin, denatured collagen;K , dissociation constant;a-d

MEM, a-minimal essential medium; MMP, matrix metalloproteinase; PAGE, polyacrylamide gel electrophoresis; PBS, phosphate-buffered saline;rII1–II2, r14AA, rI6–I7, recombinant(r) domains from FN containing type II modules; rCBD, recombinant collagen-binding domain in MMP-2; SDS-PAGE, sodium dodecyl sulfate-polyacrylamide gel electrophoresis.

*Corresponding author. Tel.:q1-210-567-3564; fax:q1-210-567-6858.E-mail address: [email protected](B. Steffensen).

400 B. Steffensen et al. / Matrix Biology 21 (2002) 399–414

1. Introduction

Fibronectin(FN) is a high molecular weight dimericglycoprotein of the extracellular matrix(ECM) thatcontains a number of discrete binding sites for extracel-lular matrix molecules including collagen, heparin,fibrin, integrin receptors, DNA and bacteria(Hynes,1990). By virtue of these interactions, FN plays a keyrole in many biological functions including maintainingstructural integrity of the ECM, cell attachment, spread-ing, migration, and control of cell morphology anddifferentiation.The FN structure is characterized by types I, II and

III homology repeats that are;40, 60 and 90 aminoacids long, respectively(Skorstengaard et al., 1986). Inthe following, as an example, we will use the term ‘II1’to designate the type II module number one. Analysisof proteolytically-generated fragments of FN first local-ized the type I collagen binding properties to a 42-kDa

terminal fragment, which contained the only twoNH -2type II modules of the molecule, II1 and II2(Hahn andYamada, 1979; Zardi et al., 1985). Because only frag-ments of FN with the type II modules bound collagen,considerable effort has been directed towards elucidatingthe precise contribution of these two modules to theinteractions with collagen(Owens and Baralle, 1986;Ingham et al., 1988; Isaacs et al., 1989; Skorstengaardet al., 1994). The FN modules II1 and II2 are positionedbetween the modules I6 and I7(I6-II1-II2-I7). Proteo-lytic subfragments containing the regions II2–I7, I8–I9and I6–II1 possessed only weak gelatin binding prop-erties (Ingham et al., 1989; Litvinovich et al., 1991)and Owens and Baralle(Owens and Baralle, 1986)proposed that a 14 amino acid residue flanking sequencefrom the COOH-terminal I7 was essential for type IImodule interaction with gelatin. However, when Skor-stengaard et al.(1994) functionally mapped the collagenbinding region of FN using a series of precisely definedrecombinant constructs the smallest segment of FN toretain type I gelatin binding properties consisted of I6-II1–II2-I7 (Skorstengaard et al., 1994). Together, theevidence favors a model where the two FN type IImodules require at least part of, and most likely thecomplete flanking I6 and I7 modules for specific inter-action with type I collagen(Owens and Baralle, 1986;Banyai et al., 1990; Skorstengaard et al., 1994).Although several studies have investigated the type I

collagen binding properties of the FN type II modules,no reports have yet characterized the interactions withother collagen types or molecules of the ECM. In thiscontext, we previously demonstrated that the three tan-dem FN type II-like modules in matrix metalloprotei-nase-2 (MMP-2) form a collagen binding domain(CBD) which displayed interactions with a broaderrange of molecules than previously reported for eitherMMP-2 or FN (Steffensen et al., 1995). In addition to

native and denatured type I collagen, rCBD boundelastin, denatured types IV and V collagen, and heparin(Steffensen et al., 1995). Because FN type II-likemodules are present in several other molecules, includ-ing blood clotting factor XII(MacMullen and Fujikawa,1985), bovine seminal fluid proteins PDC-109 and BSP-A3 (Esch et al., 1983; Seidah et al., 1987), the insulin-like growth factor II receptorymannose 6-phosphatereceptor(Lobel et al., 1987; Morgan et al., 1987), themannose receptor(Taylor et al., 1990), and the 72-kDaand 92-kDa gelatinases(MMP-2 and -9) (Collier et al.,1988; Wilhelm et al., 1989), it is essential to understandwhether the duplicated modules have undergone func-tional specialization during evolution. Therefore, wehave characterized ligand interactions for three recom-binant proteins from human FN that all included bothtype II modules but had varying extension into theflanking type I6 and I7 modules.Exposure to FN fragments alters cellular regulation

of MMP expression. For example, the 120-kDa centralFN fragment, which contains the cell binding region,elevated collagenase and stromelysin expression in rab-bit synovial fibroblasts(Werb et al., 1989; Huhtala etal., 1995) and up-regulated the urokinase-type plasmin-ogen activator in periodontal ligament cells(Kapila etal., 1996). Although a 45-kDa gelatin binding fragmentof FN did not elicit similar proteinase responses inperiodontal ligament cells(Kapila et al., 1996), thisfragment and an additional 29-kDa NH -terminal gela-2

tin-binding FN fragments increased the gelatinolytic andcollagenolytic activities as well as the proteoglycanrelease in bovine articular cartilage explant cultures(Homandberg et al., 1992). Cellular interactions withFN also enhanced activation of proMMP-2 by a mech-anism involving processing of the cellular membrane-type metalloproteinase-1(Stanton et al., 1998) and wehave shown that exposure of fibroblasts to soluble rCBDpromotes MMP-2 activation. Recently, fibroblast inter-action with the high-affinity heparin-binding domain ofFN (modules III10–III15) was found to be associatedwith induction of apoptotic events(Kapila et al., 1999).These in vitro effects may translate to tissue remodelingevents in vivo where elevated FN fragmentation hasbeen demonstrated in wound exudates from chronicinflammatory conditions such as arthritis(Carsons etal., 1985; Xie et al., 1992), chronic ulcers(Wysockiand Grinnell, 1990; Grinnell et al., 1992), symptomatictemporo-mandibular joints(Zardeneta et al., 2000), andperiodontal disease(Talonpoika et al., 1993). Thus,understanding the molecular mechanisms of ligand inter-actions and cellular regulation in response to FN frag-ments is of considerable biological interest.We report here that isolated collagen binding domain

from FN interacts not only with type I collagen, butwith a series of collagen types. All collagen interactionsrequired the presence of the two type II modules as well

401B. Steffensen et al. / Matrix Biology 21 (2002) 399–414

as their two flanking type I modules as represented inrI6–I7 (I6-II1–II2-I7). Of potentially important biolog-ical significance, rI6–I7 competed for gelatin bindingwith the corresponding collagen binding domain fromMMP-2, supported cell attachment, and reduced cellularactivation of pro-MMP-2.

2. Experimental procedures

2.1. Extracellular matrix molecules, chromatographymedia, and antibodies

Acid-soluble native type I collagen was prepared fromrat tail tendons without pepsin treatment according tothe methods of Piez(1967) with modifications(Steffen-sen et al., 1995). Additional non-pepsin-treated collagentypes studied included bovine type II and human typeIII and VI collagens from Biødesign, and human typesIII and V collagens and murine type IV collagen fromBecton-Dickinson. Dr. Kirsch, University of Pennsylva-nia School of Dental Medicine, kindly provided type Xcollagen isolated from hypertrophic chicken growth platecartilage. To prepare gelatin(denatured collagen), nativeforms of the collagens were heat-denatured for 30 minat 56 8C.

Plasma FN was purified from bovine platelet-poorserum by gelatin-Sepharose affinity chromatography aspreviously described(Steffensen et al., 1995) and storedat –80 8C until use. Additional bovine plasma FN waspurchased from Life Technologies. Bovine elastin wasfrom Sigma and Matrigel was from Becton Dickinson.�

Gelatin-Sepharose 4B, heparin-Sepharose CL-6B,CM-Sepharose Fast Flow, Sephadex G-10, and chelatingSepharose 6B were from Amersham-Pharmacia. AffiGel10 and 15 and goat anti-rabbit polyclonal IgG(HqL)antibody conjugated with alkaline phosphatase werefrom BioRad. Affinity-purified, polyclonal antibody tothe His peptide included in the NH -terminal flanking6 2

sequence of the expressed proteins was kindly providedby Dr Overall(University of British Columbia, Vancou-ver, Canada), polyclonal antibody to human FN was thegenerous gift by Dr Grinnell(Southwestern MedicalCenter, Dallas, Texas), andb -integrin subunit blocking1

monoclonal antibody Mab13 was graciously providedby Dr Yamada(National Institute of Dental and Crani-ofacial Research, NIH).

2.2. Expression constructs for recombinant fibronectindomains

Three different segments of the collagen bindingregion in human FN were expressed as recombinantproteins in E. coli based on the sequence of humanplasma FN presented by Kornblihtt et al.(1985)(GenBank X02761). In the following, we will use theprefix r for recombinant, i.e. rII1–II2, whereas the

¯ ¯ ¯

modules in the native protein will be referred to as II1–II2. The II1–II2 expression construct was designed toencode the two FN type II modules only(Thr314–Met432). The second construct, 14AA, contained thecoding sequence for II1–II2 and also the 14 amino acidsof the flanking COOH-terminal I7(Thr314–Met446)that Owens and Baralle(Owens and Baralle, 1986)proposed to be important residues for FN binding oftype I collagen. The third construct, I6–I7, coded formodules II1-II2 and all of the two flanking modules I6and I7(Gly275–Ala479), which were required for typeI collagen binding in studies by Skorstengaard et al.(1994). Coding cDNA fragments were amplified byreverse transcription-polymerase chain reactions(RT-PCR) (25 cycles; 948C 30 s, 628C 1 min, 72 8C 1min 50 s) from total RNA isolated from human gingivalfibroblasts as previously described(Chomczynski andSacchi, 1987). The oligonucleotide pairs(Table 1)added 59-NheI and 39-PstI restriction sites for directionalligation into the pGYMX expression vector, whichexpresses recombinant proteins with a short NH -termi-2

nal fusion tag comprising an initiating methionine, aHis tag, and a factor X cleavage site(Guillemette et6 a

al., 1991; Eltis et al., 1994). To verify the fidelity ofthe reactions and the reading frame, the expressionconstructs were verified by double-stranded chain ter-mination DNA sequencing(Sanger et al., 1978) at theUniversity of Texas Health Science Center, Facility forAdvanced DNA Technologies.

2.3. Corrections to the published FN DNA sequence

DNA sequencing of the expression constructs revealedfour single base deviations from the published GenBanksequence(X02761) for FN that is based on Kornblihttet al.(1985) resulting in two amino acid changes(Table2). We confirmed that these sequence deviations incDNA from human gingival fibroblasts were true andnot the result of mispriming or random PCR errors bysequencing cDNA from four additional cell lines afteramplification by RT-PCR using high-fidelity Taq poly-merase(Platinum Taq, Life Technologies). The cellswere two fibroblast cell lines, IMR90 and WI38, andtwo mammary breast cancer cells, MRC5 and Hs578T,the cell line from which the original DNA sequence wasderived (Hackett et al., 1977; Kornblihtt et al., 1985).The corrected sequence has been submitted to GenBankand is accessible under the accession� AF313299.

2.4. Expression and purification of recombinant proteins

Expression and purification of the recombinant FNproteins followed procedures modified slightly fromprior reports(Steffensen et al., 1995, 1998). In brief,freshly transformed colonies ofE. coli grown in superbroth under selective pressure by ampicillin showed


Table1

Olig

onuc

leotides

used

forge

neratin

greco

mbina

ntFN

cons

truc

tsa

Con

structs(am

inoac

ids)b

Seq

uenc

es

rII1–II2

(Thr31

4–Met43

2)Sen

sean

tisen

se5

9-TCGCTA

GCACAGCTGTA

ACCCAGACTTA

CG-3959-TAAGCTTCTGCAGTCATTA

CATGGGGCAGAACCCAAA-39

r14A

A(T

hr31

4–Met44

6)Sen

sean

tisen

se5

9-TCGCTA

GCACAGCTGTA

ACCCAGACTTA

CG395

9-TA

AGCTTCTGCAGTCATTA

CATGACCCCTTCATTGGTT-39

rI6–I7(G

ly27

5–Ala47

9)Sen

sean

tisen

se5

9-ATCGCTA

GCGGCCACTGTGTCACAGACAG-3959-TAAGCTTCTGCAGTCATTA

GGCAATGCATGTCCATTCC-39

Exp

erim

entald

etails

ofRT-PCR

arepres

entedin

Sec

tion2.

a

The

aminoac

idse

quen

cein

FN

enco

dedby

theex

pres

sion

cons

truc

t.b

good yields of recombinant proteins. Because proteinswere expressed in inclusion bodies, the bacterial pelletswere treated with lysozyme(Sigma) and then dissolvedwith 6 M guanidine-hydrochloride. Utilizing non-reduc-ing conditions favored the dissolution of non-intermo-lecularly cross-linked recombinant proteins withretention of most dimeric and polymeric protein in theresidual pellet. The extracted, denatured recombinantproteins were refolded by dialysis against aerated 0.1 Msodium borate buffer, pH 10.0 for 2 h at 208C (Lucket al., 1992). Residual guanidinium salt was removedby extensive dialysis against chromatography buffer(100 mM NaP , 0.5 M NaCl, pH 8.0) at 4 8C.i

The three FN-derived recombinant proteins were puri-fied by Zn -chelate chromatography. After sample2q

loading and extensive washes in chromatography buffer,non-specifically bound bacterial proteins were removedfirst with 1.0 M NaCl and then by elution with 35 mMimidazole in chromatography buffer. Specifically boundrecombinant protein was then eluted at high purity with100 mM imidazole. Because rI6–I7, but not rII1–II2 orr14AA, was found to bind to gelatin-Sepharose duringinitial optimization of the purification protocol, pooledeluates of rI6–I7 from the Zn -chelate column were2q

applied to a gelatin-Sepharose 4B column as a secondlevel affinity purification. Non-specifically bound pro-tein was eluted with 1 M NaCl and the gelatin-bindingrI6–I7 was eluted with 7%(vyv) DMSO in 50 mMTris, pH 7.4. All proteins were dialyzed against 50 mMTris, pH 7.4, quantitated by the BCA assay(Pierce),snap frozen in liquid N , and stored aty80 8C.2

2.5. SDS-PAGE, Western blotting, and enzymography

Proteins were separated by SDS-polyacrylamide gelelectrophoresis(SDS-PAGE) according to Laemmli(1970). Protein samples were analyzed without reduc-tion or in the presence of 65 mM dithiothreitol(DTT)followed by heating at 958C for 5 min. Gels werestained with Coomassie Brilliant Blue R-250 and bandsquantitated with the K1D electrophoresis software(Kodak) after image captures using a DC120 camera(Kodak). Molecular mass determinations from migrationin SDS-PAGE gels utilized high molecular mass stan-dards(Gibco).

For Western blotting, proteins were transferred toImmobilon-P polyvinylidine difluoride membranes(Mil-lipore) following separation by SDS-PAGE. Proteinswere reacted with affinity purifieda-His or a-FN6

antibodies at dilutions of 1:1000 and then detected withhorseradish peroxidase-conjugated goat anti-rabbit anti-body and enhanced chemiluminescence reagents(Pierce) and BioMax film (Kodak) as described earlier(Steffensen et al., 1995).Samples analyzed by enzymography were separated

under non-reducing conditions on 10%(wyv) polya-


Table 2Corrections to the published DNA sequences coding for the FN col-a

lagen binding domain and predicted amino acids

Moduleb Sequencec Amino acid change

I6 290 294Q W L K T SilentCAG TGG C

¯TG AAG ACA

II1 322 326G N S

¯N G L to S

GGC AAC TC¯A AAT GGA

II1 343 347C T T E G SilentTGC ACC ACA

¯GAA GGG

II2 378 382Q T R

¯G G Q to R

CAG ACT CG¯A GGA GGA

Confirmed in four cell lines as described in Section 2. Reviseda

sequence is accessible from GenBank under accession�AF313299.Designates FN modules.b

Protein and DNA sequences given with corrections underlined.c

Amino acid number is shown.

crylamide gels containing 100mgyml heat-denaturedacid-soluble type I gelatin(BioRad). Gels were proc-essed as described previously(Steffensen et al., 1995).After incubation in collagenase assay buffer(50 mMTris, 200 mM NaCl, 5 mM CaCl , pH 7.2) at 37 8C,2

counterstaining with Coomassie Brilliant Blue revealedcleared areas on a blue background corresponding togelatinolytic activities. The relative activities were quan-titated from reversed digitized gel images with the K1Dsoftware(Kodak) as described above.

2.6. Mass spectrometry

The predicted masses of the recombinant domainswere further confirmed for rI6–I7 by nano-electrospraymass spectrometry using a Finnigan LCQ Mass Spec-trometer after desalting the samples on C18 ZipTips(Millipore) and for rII1–II2 and r14AA by matrix-assisted laser desorption ionization time-of-flight massspectrometry(1 pmol protein in sinapinic acid matrix)on a Voyager-DE STR Mass Spectrometer(ABI) at theUniversity of Texas Health Science Center InstitutionalMass Spectrometry Laboratory.

2.7. Affinity chromatography

Potential interactions of the recombinant FN frag-ments with type I collagen and heparin were determinedusing gelatin- or heparin-Sepharose affinity resins. Toassess binding to elastin, affinity columns containedinsoluble grains of elastin in a 1:1(vyv) mixture withSephadex G-10. Chromatography on mini columns(Overall et al., 1989) used conditions detailed previously(Steffensen et al., 1995). Briefly, standard amounts ofrecombinant protein(25 mg) were loaded onto theaffinity mini columns(V 25–100ml) in chromatogra-t

phy buffer(50 mM Tris, pH 7.4). Elution included stepgradients of NaCl from 0.1 to 1.0 M and of DMSOfrom 1 to 10% (vyv) in 50 mM Tris, pH 7.4. Thecollected column fractions were analyzed by SDS-PAGEat constant concentrations relative to the loaded samplevolumes. The affinity column experiments were repeatedat least three times. The absence of binding to thenegatively charged CM-Sepharose served as a controlfor specific binding to heparin.The requirement for intact disulfide bonds for binding

to type I gelatin was analyzed after bond disruption bythe reducing agent DTT. Proteins were incubated with10 mM DTT for 30 min at 378C. Additional sampleswere reduced and then incubated with 5 mM iodoace-tamide for 30 min at 378C to chemically block thethiol groups. Control protein samples were incubated at37 8C for the same period of time. Gelatin binding ofthe treated and control proteins was analyzed by affinitychromatography as described above, with the exceptionthat all buffers and elution solutions contained 10 mMDTT.

2.8. Competitive affinity chromatography and fluores-cence spectroscopy

Competition experiments first employed gelatin- andheparin-Sepharose affinity columns to assess whethertype II module-containing domains from FN(rI6–I7)and MMP-2(rCBD) competed for binding sites on theseligands. Following saturation binding of one protein(typically 50 mg, 1.5 nmol) to the affinity column(Vt

25 ml for gelatin-Sepharose and 100ml for heparin-Sepharose), unbound protein was removed by thoroughrinses(20=V ), and competing protein was loaded(50t

mg, 1.5 nmol). Columns were again rinsed to removecompetitively displaced and unbound protein. Finally,all column-bound rI6–I7 and rCBD were eluted with10% DMSO from gelatin-Sepharose or with 1 M NaClfor elution from heparin-Sepharose columns. Thereleased proteins in the column fractions were monitoredby SDS-PAGE and quantitated from digitized images asdescribed above.After initial observation of competition between rI6–

I7 and rCBD for gelatin binding by competitive affinitychromatography, fluorescence spectroscopy was appliedto verify this result. To label rI6–I7 and rCBD, 5mMof the recombinant proteins were incubated with afivefold molar excess(25 mM) of fluorescein-5-isothio-cyanate(FITC) (Sigma) in 25 mM Hepes, pH 7.5, for1 h in the dark at room temperature(RT). Glycine wasthen added to 100mM and the reaction incubated for 1h at 4 8C to bind free FITC. The FITC-labeled proteinwas separated from free FITC-glycine using a G-25desalting column(BioRad) equilibrated with 25 mMHepes, pH 7.5). To first assess interactions of rI6–I7and rCBD individually with gelatin, 0.2mM of FITC-


labeled rI6–I7 or rCBD in 25 mM Hepes, pH 7.0, wereincubated with increasing concentrations(0–12.5mM)of gelatin at 258C. Using an SPF-500C Spectrofluoro-meter(SLM Instruments, Inc. Urbana, Illinois) with anexcitation wavelength of 495 nm, emission spectra wererecorded from 510 to 550 nm. After verifying interac-tions for each recombinant protein with gelatin, weanalyzed the potential for rCBD to displace rI6–I7complexed with gelatin. First, 0.2mM FITC-labeledrI6–I7 was incubated with 5mM gelatin in 25 mMHepes, pH 7.0 for approximately 5 min at 258C and afluorescence quenching of;70% was achieved. Sub-sequently, to assure saturation of the reaction, a broadconcentration range of 0–5mM unlabeled rCBD wasadded in increments while monitoring recovery of thefluorescent signal indicating the presence of unbound,labeled rI6–I7, displaced from gelatin by the competingunlabeled rCBD.

2.9. Microwell substrate binding assay

Interactions of FN-derived fragments with ECM mol-ecules were also analyzed using the microwell proteinbinding assay as detailed previously(Steffensen et al.,1995) for types I, II, III, IV, V, VI and X collagens innative and denatured form, Matrigel , and laminin. BSA�

served as a negative control protein in these assays. Inbrief, 96-microwell plates were coated with protein(typically 0.5–2mg, 10 pmol per well). Matrigel and�

type I collagen were also prepared as three-dimensionalgels according to the manufacturers instructions. Afterblocking non-specific binding sites on coated plates with2.5% (wyv) BSA in phosphate-buffered saline(PBS),serially diluted recombinant protein in 50 mM Tris, pH7.4 (400–0.125 pmolywell) was added. Recombinantprotein bound to the coated ECM molecules was detect-ed with affinity-purified polyclonal antibody raisedagainst the His fusion tag(Wallon and Overall, 1997)6

followed by an alkaline phosphatase-conjugated goatanti-rabbit polyclonal antibody(BioRad). Quantitationwas carried out at 405 nm in an Opsys MR plate reader(Dynex) using p-nitrophenyl phosphate disodium(Sig-ma) as the substrate. Negative controls consisted of thereaction mixture minus recombinant protein, primaryantibody, or secondary antibody. The data were graphedand analyzed with SigmaPlot graphing software(SPSSCorp.) using a four-parameter algorithm,ys((ayd)y(1q(xyc) ))qd, where xsconcentration of recombi-b

nant protein added,ysbound recombinant protein,bsslope,csconcentration at inflection point, apparentK ,dasminimum binding anddsabsorbance at saturation).All binding experiments were performed at least twicein duplicate and, for reliable comparisons, all recombi-nant proteins were analyzed concurrently on the sameplates.

2.10. Cell attachment assay

Cell attachment was analyzed using 96 microwellplate assays as described previously(Steffensen et al.,1998). In short, plates were coated with serially dilutedrII1–II2, r14AA, or rI6–I7 (typically 25–0.5, and 0mgyml) and blocked with heat-denatured BSA prior toseeding human fibrosarcoma HT1080 cells or humangingival fibroblasts at 4=10 cellsywell in serum-free4

a-minimal essential medium(a-MEM). After a 90-minincubation, attached cells were fixed with 4% formal-dehyde in PBS and stained with crystal violet(Keunget al., 1989; Steffensen et al., 1998). The number ofattached cells was quantified from the optical density ofcell-bound dye dissolved with 10% acetic acid andmeasured at 590 nm. Positive control wells were coatedwith FN or rCBD (Steffensen et al., 1998). Attachmentto BSA-blocked wells served for correction of non-specific binding. Experiments were repeated in triplicateand comparisons between proteins were carried outwithin the same plates.In addition, cells were seeded on rI6–I7 in the

presence of blocking monoclonal antibody mAB13(0.6–5mgyml) to theb -integrin subunit(kindly pro-1

vided by Dr K. Yamada, NIDCR, National Institutes ofHealth). Affinity purified antibody to rCBD served ascontrol in the 90-min incubation.

2.11. Enzyme activation assay

Surfaces of 24-well tissue culture plates were coatedwith 25 mgyml laminin, FN, or the recombinant FNfragments rII1–II2, r14AA, or rI6–I7 in PBS overnightat 4 8C. Non-specific binding sites were blocked with10 mgyml heat-denatured BSA for 1 h at 208C andthen rinsed with PBS. Uncoated plastic wells served ascontrol surfaces. HT-1080 cells(500ml) were seeded at2=10 cellsyml and maintained in serum-freea-MEM5

for 48 h. Conditioned medium was collected and ana-lyzed by gelatin enzymography as described above. Theconversion of latent to intermediate and fully activeforms of MMP-2 and MMP-9 was measured from theintensity of the protein bands on reversed digitized gelimages as described above.

3. Results

3.1. Characterization of recombinant FN type II modules

The three FN constructs(rII1–II2, r14AA, rI6–I7)expressed well inE. coli using an expression systemthat has previously provided high yield of functionallyfolded recombinant collagen binding domain fromMMP-2 (Steffensen et al., 1995). Purified, reduced rII1–II2, r14AA, and rI6–I7 migrated as single protein bandson SDS-PAGE with apparent molecular masses of 17.2,


Fig. 1. Recombinant FN fragments require intact disulfide bonds forinteraction with denatured type I collagen. Three recombinant FNfragments (rII1–II2, r14AA, rI6–I7) were electrophoresed(1mgylane) on 15% cross-linked polyacrylamide minislab gels underreducing (qDTT) or non-reducing(yDTT) conditions. Separatedproteins were stained with Coomassie Brilliant Blue(panel a) ortransferred to PVDF membranes and reacted with an anti-His poly-6

clonal antibody(panel b). Panels c–e: samples(S) of recombinantFN fragments(rI6–I7 shown) (25 mg, 0.9–1.1 nmol) were loadedonto mini-columns of gelatin-Sepharose and chromatographed asdescribed in Section 2. Elution strategies included first 1.0 M NaCl(N) and then a step gradient of DMSO(%DMSO) from 1 to 4% inchromatography buffer. Column fractions were analyzed on 15%SDS-PAGE gels under reducing conditions. The rI6–I7(panel c)bound specifically to gelatin-Sepharose as shown by minor amountsof rI6–I7 in the unbound(U) and wash fractions(W) and absencefrom the 1.0 M NaCl eluate(N). Peak elution of rI6–I7 was at 1–2% (vyv) DMSO. Two other FN fragments, rII1–II2 and r14AA, didnot bind gelatin-Sepharose(not shown). To assess the requirement forintact disulfide bonds, rI6–I7 binding to gelatin-Sepharose was thenanalyzed after disruption of disulfide bonds. Contrary to non-reducedrI6–I7 (panel c), both reduction with 10 mM DTT alone(30 min at37 8C) (panel d) and reduction followed by treatment with 5 mMiodoacetamide for 30 min at 378C to chemically block the exposedthiol groups(panel e) abolished the gelatin binding properties of rI6–I7. Under these conditions, all loaded protein was recovered in theunbound and wash fractions. The positions of the kDa of protein stan-dard(Mr) and recombinant proteins(rI6–I7) are indicated.

19.4, and 26.4 kDa, respectively(Fig. 1a). Under non-reducing conditions, the three recombinant proteinsmigrated at 15.5, 17.6 and 24.5 kDa, respectively(Fig.1a). The migration of the non-reduced proteins at;1.7–1.9 kDa lower apparent masses indicated that the pro-teins were folded and contained intact disulfide bonds(Kaderbhai and Austen, 1985). Only r14AA demonstrat-ed a small proportion(-10%) of dimeric proteinmigrating at 32 kDa. These masses corresponded tothose predicted from the amino acid sequences.Mass spectrometry gave masses of 24 628 Da for

rI6–I7, 16 661 Da for r14AA, and 15 174 Da for rII1–II2, which corresponded precisely to the masses predict-ed for the NH -terminal methionine-processed forms of2

the recombinant proteins(not shown). Further, affinity-purified anti-peptide antibody(aHis ) raised against the6

His of the NH -terminal fusion peptide reacted specif-6 2

ically with all three purified recombinant proteins inWestern blot analysis(Fig. 1b).

3.2. Characterization of interactions with denatured typeI collagen

The interactions of the FN-derived recombinant frag-ments with denatured type I collagen in solution wereinvestigated using gelatin-Sepharose mini columns. TherI6–I7 bound specifically to gelatin-Sepharose(Fig. 1c).Stepwise elution with up to 1.0 M NaCl was ineffectivein eluting rI6–I7 showing that ionic interactions alonewere not responsible for the binding. However, boundrI6–I7 characteristically was eluted from the gelatin-Sepharose columns with a peak at;1–2% DMSO inchromatography buffer(Fig. 1c) indicating that hydro-phobic interactions are central to the interaction. Therewas no binding by rII1–II2 and r14AA to gelatin-Sepharose(not shown). The specific interaction of rI6–I7 with gelatin mirrored the interaction of full-lengthFN that also bound specifically to gelatin Sepharose andwas eluted with a major peak at;2% DMSO and alesser peak at 10% DMSO(not shown).Since each type II module contains four cysteines

forming pairwise disulfide bonds, the two FN type IImodules that are contained in each of our three recom-binant FN fragments may form a total of four disulfidebonds. Isaacs et al.(1989) showed that reduction byitself was sufficient to disrupt gelatin binding of largerproteolytically-generated FN fragments containing typeII modules. In accordance with that observation, ourexperiments demonstrated that the smallest fragment ofFN with collagen binding properties, rI6–I7, also didnot bind gelatin after reduction with 10 mM DTT alone(Fig. 1d) or after combined reduction with DTT andchemical modification to block the thiol groups withiodoacetamide(Fig. 1e). In comparison, reduction with65 mM DTT plus carboxymethylation were required todisrupt rCBD binding to gelatin(Steffensen et al.,


Fig. 2. FN fragments interact with the native and denatured forms of several collagens. Equal amounts of either native or denatured types I, II,III, IV, V, VI and X collagen (typically 0.5mg, 10 pmol) were coated as films on 96-well microtiter plates as detailed in Section 2. After blockingnon-specific binding sites with 2.5% BSA in PBS, serially diluted recombinant proteins(400–0.125, and 0 pmolywell) were added and incubatedfor 1 h at 22 8C. After rinses to remove unbound recombinant protein, bound FN fragments were reacted with an affinity-purified polyclonalanti-His antibody and detected using an alkaline phosphatase-conjugated goat anti-rabbit antibody.p-Nitrophenyl phosphate disodium served as6

substrate for quantitation and the reaction was read at 405 nm. Analyses in parallel of the collagen interactions by rCBD permitted comparisonbetween type II modules in FN and MMP-2. Graphs and corresponding apparent dissociation constants(K ) are presented only for specific andd

saturable protein interactions. Data points are representative of two to four experiments in duplicate.

1995). Thus, intact disulfide cross-links are required tomaintain a functional conformation of the FN type IImodules.

3.3. Interactions with other collagen types

The interactions of recombinant FN fragments withnative and denatured types I, II, III, IV, V, VI and Xcollagen were analyzed subsequently by microwell pro-tein binding assays. Type I and III collagens were uniqueamong the analyzed collagen types in that rI6–I7 boundboth the native and denatured forms of these collagens(Fig. 2a,c). The saturation level binding of rI6–I7 totype I collagen was lower than that observed for rCBDwhereas the binding to denatured type III collagen wascomparable for the two recombinant domains. The rI6–I7 bound type II, IV, V and X collagens, but only thedenatured forms(Fig. 2b,d,e,g). For these collagens, thesaturation level binding for rCBD consistently exceededthat for rI6–I7 and the strength of interaction wasconsistently stronger for rCBD than rI6–I7 as shown by3–20-fold lower apparentK values. Importantly, andd

not previously reported, the rCBD of MMP-2 reacted

avidly with both native and denatured forms of typesIII, V and X collagen (Fig. 2c,e,g). The apparentKd

values for these interactions ranged from 0.06 to0.52=10 M. The rI6–I7 and rCBD did not bind typey7

VI collagen(Fig. 2f).In agreement with results from other laboratories,

neither the II1–II2 nor the 14AA segments of FN wereable to bind type I collagen(Skorstengaard et al., 1994).In addition, we demonstrated that these shorter segmentsof the collagen binding domain of FN do not bindseveral other collagen types either(Fig. 2a–g).

3.4. Interactions with other extracellular matrixmolecules

Elastin, which is a central component of elastic fibers,has similarity to collagens in that one-third of theresidues are glycine and it contains a high percentageof proline (Mecham and Heuser, 1991). Because wepreviously found that elastin is bound by the MMP-2CBD (Steffensen et al., 1995), and others demonstratedthat this domain in MMP-2 is required for elastinolysis(Shipley et al., 1996), our hypothesis was that the type


Fig. 3. Interactions of FN fragments and rCBD with reconstitutedbasement membrane, Matrigel , laminin, and heparin. In microwell�

protein binding assays, none of the FN fragments proteins or theMMP-2 rCBD bound native Matrigel films(panel a) or Matrigel� �

in gel form (not shown). However, heat denaturation of Matrigel�

prior to film coating resulted in specific binding of both rI6–I7 andrCBD, but not rII1–II2 or r14AA (panel a). The major basementmembrane component laminin also did not support binding of eitherof the FN fragments or rCBD(panel b). Interactions between the FNfragments and heparin were analyzed by affinity chromatography anddetected by SDS-PAGE(panels c–e). Samples(S) of rII1–II2,r14AA, and rI6–I7 loaded in 50 mM Tris, pH 7.4, bound to the hep-arin affinity matrix and were eluted with a peak at 0.25 M NaCl inchromatography buffer(NaCl, 1) after extensive washes(W, 1–3)(panel c, rI6–I7 shown). No additional protein was eluted with 10%DMSO (D). The absence of binding of the recombinant FN fragmentsto the negatively charged CM-Sepharose(CM) confirmed the speci-ficity of the interactions(panel, rI6I7 shown). However, when loadedin the presence of 150 mM NaCl, none of the FN proteins boundheparin-Sepharose(panel, rI6I7 shown) suggesting that the observedinteraction is not physiologically important. The positions of rI6–I7and the protein molecular weight markers(kDa=10 ) are indicated.y3

II modules of FN would confer elastin binding to FN.Surprisingly, none of the three recombinant FN frag-ments bound elastin in affinity chromatography assays(not shown).Since FN is a component of the basement membrane

and may contribute to the integrity of this structure viainteractions with other basement membrane molecules,we first investigated interactions with reconstituted base-ment membrane(Matrigel ) as a film and as a gel. The�

recombinant FN fragments did not bind Matrigel films�

(Fig. 3a) or gels (not shown). However, upon heattreatment of Matrigel under conditions that denature�

fibrillar collagens, we found specific, albeit low levelbinding of rI6–I7 to Matrigel films(Fig. 3a). This�

treatment of Matrigel may have exposed cryptic bind-�

ing sites on one or more of the constituent moleculessuch as type IV collagen that bound rI6–I7 only afterdenaturation(Fig. 2d). Likewise, the rCBD, which wepreviously reported to not bind native Matrigel films�

or gels(Steffensen et al., 1995), also bound the dena-tured Matrigel films(Fig. 3a).�

Because Matrigel binding sites for the FN fragments�

and rCBD might be masked by inter-molecular com-plexes, we subsequently examined the interactions withindividual basement membrane components. There wasno specific binding of any FN fragments to the majorMatrigel component, laminin, in microwell plate assays�

(Fig. 3b). This agreed with our prior result for therCBD (Steffensen et al., 1995). In affinity chromatog-raphy experiments exploring interactions with heparin—a structural and functional analogue of heparan sulfate—we detected weak interaction by rII1–II2, r14AA andrI6–I7 (Fig. 3c, rI6–I7 shown). Absence of proteinbinding to the negatively charged CM-Sepharose controlcolumns (Fig. 3e) ruled out non-specific interactionswith heparin. However, when the recombinant proteinswere loaded onto heparin-Sepharose columns in thepresence of 150 mM NaCl, no binding was observed(Fig. 3d) indicating that the interactions of the collagenbinding fragments with heparin are not physiologicallyimportant.

3.5. The rI6–I7 and rCBD compete for ligandinteractions

There is mounting evidence that proteolytic fragmentsof FN may alter biological functions, including celladhesion(Grinnell et al., 1992) and MMP expression(Werb et al., 1989; Huhtala et al., 1995; Kapila et al.,1996). To understand the potential for competitive ligandinteractions mediated by type II module-containing seg-ments of FN and MMP-2, we analyzed the ability ofrI6–I7 and rCBD to compete for binding to denaturedtype I collagen and heparin.In initial competition affinity chromatography assays,

rCBD consistently displaced a significant amount of

bound rI6–I7 from gelatin-Sepharose leaving a finalbound ratio of the two proteins of 21:79(rI6–I7:rCBD),as quantitated from digitized SDS-PAGE gels(Fig.4a,A1). In comparison, rI6–I7 added to rCBD-saturatedgelatin-Sepharose columns displaced only limitedamounts of rCBD resulting in a rI6–I7:rCBD of 34:66in the final DMSO eluate(Fig. 4a,A2). Competitionstudies on interactions with heparin, which are ionic in


Fig. 4. rI6–I7 and rCBD compete for binding sites on gelatin. Competition between the FN fragment rI6–I7 and rCBD from MMP-2 for bindingto gelatin-Sepharose was determined by competition affinity chromatography assays followed by analysis of fractions using SDS-PAGE(panela) and by fluorescent spectroscopy(panels b–d). In chromatography assays competing bound rI6I7 with rCBD, rI6–I7(S) was loaded first(50mg, 1.5 nmol) on gelatin-Sepharose affinity columns(panel a,A1). Saturation of the affinity matrix was confirmed by the presence of protein inthe unbound(U) and early wash(W, 1–4) fractions. Subsequently, an equal molar amount of rCBD was loaded. Unbound rCBD(U2) andreleased rI6–I7 were monitored during additional washes(W, 5–7) and, finally, all bound rI6–I7 and rCBD were eluted with chromatographybuffer containing 10% DMSO(D, panel a1) and quantified from digitized SDS-PAGE gel images as detailed in Section 2. In assays using identicalprocedures, but with the reverse order of protein loading, bound rCBD was released from gelatin by rI6–I7(panel a,A2). The positions of therecombinant proteins(rI6–I7 and rCBD) and mass markers(Mr) are given. The interactions of rI6–I7 and rCBD with gelatin were confirmed influorescence quantum yield assays. A concentration range of gelatin(0, 2.5–15.0=10 M) was added to reactions containing 2=10 M ofy6 y7

FITC-labeled rI6–I7(panel b) or rCBD (panel c) and the fluorescent spectra recorded as detailed in Section 2. The smoothened spectra showndemonstrate specific interactions reflected by a concentration-dependent quenching of the signal. In a competitive binding assays(panel d), FITC-labeled rI6–I7 at 2=10 M (BL) was first reacted with 5=10 M gelatin, achieving fluorescence quenching of;70% (Gelatin). Subsequenty7 y6

addition of 0.2–5=10 M unlabeled rCBD resulted in a concentration-dependent recovery of the fluorescent signal reflecting displacement ofy6

rI6–I7 bound to gelatin by rCBD. No similar change was observed by addition of lysozyme, as a control protein, in the same concentration range.

nature, indicated competitive events(not shown) similarto those seen for gelatin. However, in light of the weakheparin binding observed under physiological salt con-centration(Fig. 3d), the biological significance of thisobservation is questionable.Fluorescent spectroscopy was applied as a second

analytical approach to confirm the competition betweenrI6–I7 and rCBD for gelatin binding(Fig. 4b–d). FITC-

labeled rI6–I7 and rCBD both showed a concentration-dependent reduction in quantum yield upon addition ofgelatin, indicative of complex formation(Fig. 4b,c).Subsequently, when unlabeled rCBD was added atincreasing concentration to FITC-rI6–I7ygelatin com-plexes with a quenched signal, the fluorescence wasrecovered(Fig. 4d), indicating displacement of rI6–I7by rCBD.


Fig. 5. FN collagen binding domain supportsb -integrin-dependent cell attachment, but does not induce MMP-2 activation. Cell attachment was1

assessed in 96-well plates coated with serially diluted recombinant FN fragments(50–0.5, and 0mgyml). HT-1080 fibrosarcoma cells were seeded(4=10 cellsywell) and allowed to attach for 90 min. Cell attachment was quantitated from redissolved cell-bound crystal violet stain as detailed4

in Section 2. Full-length FN served as positive and BSA-blocked wells as negative controls. rI6–I7 and r14AA supported cell attachment withhalf-maximal binding at 2.7 and 1.7mM, respectively, whereas cells did not attach to rII1–II2(panel a). In comparison, the half-maximal bindingto FN was at 10 nM(panel b). The cell attachment to rI6–I7-coated plates(30 mgyml) was inhibited in a concentration-dependent manner inthe presence of 0–5mgyml Mab13, ab1-integrin blocking antibody(panel c). In contrast, no reduction in cell attachment was detected in thepresence of the same concentration range of an rCBD-specific control antibody(a-rCBD). Data from cell attachment studies are representativeof three to five separate experiments. Subsequently, to assess the contribution of the FN fragments to MMP-2 activation, HT-1080 cells wereseeded at 10 cellsywell in 12-well plates coated with 25 mgyml laminin (LM), FN (FN), or the FN fragments rII1–II2, r14AA, or rI6–I7(rII1–5

II2, r14AA, rI6–I7) (panel d). After incubation for 48 h in serum-freea-MEM, conditioned medium was collected and analyzed by gelatinenzymography as described in Section 2. Quantitative analysis of the zymograms revealed significantly greater proportions of partly and fullyactivated forms of MMP-2(62 and 59 kDa) and MMP-9(83 kDa) from cells cultured on full-length FN(FN) compared to media from cells onrecombinant FN fragments, laminin, or uncoated plastic(panel d). Positions of masses(kDa) of latent and processed forms of MMP-2(MMP-2) and MMP-9(MMP-9) are indicated.

3.6. FN fragments support cell attachment via b -1

integrins and reduce MMP-2 activation

The major RGD-containing cell attachment site of FNresides in module I10(Pierschbacher and Ruoslahti,1984; Pytela et al., 1985) but additional segments withcell attachment properties have been located, includingthe CS-1 alternatively spliced region(Wayner et al.,1989). Therefore, we examined the potential of the threerecombinant FN-derived fragments to support cellattachment. HT1080 fibrosarcoma cells and human gin-gival fibroblasts (not shown) attached to rI6–I7 andr14AA in a specific, saturable manner, but not to rII1–II2 (Fig. 5a). As it is also the case for the cell attachmentfor RGD peptides(Pytela et al., 1987) half-maximalcell attachment to the fragments of FN required signifi-cantly higher molar concentrations of coated protein(r14AA ;1.7 mM; rI6–I7 ;2.7 mM) (Fig. 5a) com-

pared to full-length FN(;10 nM) (Fig. 5b). Themaximal number of cells bound to the FN fragmentscorresponded to approximately one-fourth of the cellnumber on FN. No cells attached to negative controlsurfaces coated with BSA(not shown).A role for b -integrins in rI6–I7-mediated cell attach-1

ment was demonstrated using mAb13, an anti-b1-inte-grin blocking monoclonal antibody. At 0.4mgyml, morethan 50% of the cells attachment to rI6–I7-coated wellswas inhibited (Fig. 5c). At )2 mgyml mAb, theinhibition was virtually complete. In contrast, affinity-purified a-rCBD antibodies had no significant blockingeffects at these concentrations(Fig. 5c).Pro-MMP-2 from HT-1080 fibrosarcoma cells cul-

tured on full-length FN, but not on laminin, undergoesactivation (Stanton et al., 1998). Because MMP regu-lation is also altered in cells exposed to several FNpeptides and domains(Huhtala et al., 1995), we exam-


ined whether type II module-containing recombinant FNfragments modified cellular MMP-2 activation. MMP-2in the conditioned media from HT-1080 cells on surfacescoated with rI6–I7, r14AA or rII1–II2 showed a signif-icantly reduced level of conversion from the latent formto the active species compared to cells cultured in wellscoated with full-length FN(Fig. 5d). Control experi-ments confirmed the absence of activation on lamininand plastic(Fig. 5d). In HT-1080 cultures on FN, theactive forms of MMP-2(62-kDa intermediate and 59-kDa fully processed) and the 83-kDa active form ofMMP-9 accounted for 74 and 47% of total enzymeactivity, respectively. For HT1080 cells on rI6–I7, rII1–II2, and r14AA, the proportions of activated the enzymewere significantly lower, namely, 21, 20 and 22% forMMP-2, and 19, 12 and 10% for MMP-9, respectively.Importantly, whereas the fully activated 59-kDa form ofMMP-2 constituted;32% in cultures on full-lengthFN, this form accounted for less than 3% on any of thethree recombinant FN fragment. Thus, in addition tomodifying cell attachment, the type II module-containingFN fragments reduced MMP-2 and MMP-9 activationin HT-1080 cells.

4. Discussion

In this report, we demonstrate for the first time thatthe shortest segment of FN with collagen binding prop-erties, modules I6–II1–II2–I7(rI6–I7), interacts spe-cifically not only with type I collagen, but also withcollagen types II, III, IV, V and X. The collagen bindingdomain had only weak heparin binding and there wasno detectable interaction with elastin or reconstitutedbasement membrane, Matrigel , and its major constitu-�

ent, laminin. Of potentially important biological signif-icance, two FN collagen binding domain fragmentscontaining type II modules, rI6–I7 and r14AA, support-ed cell attachment. When HT-1080 fibrosarcoma cellswere seeded on these fragments they displayed signifi-cantly reduced activation of MMP-2 and MMP-9 com-pared to cells seeded on native full-length FN.Our observation that both flanking type I modules, I6

and I7, are required for binding of the type II modulesto any of several tested collagens, while no binding wasobserved for the type II modules alone(rII1–II2),concurs with and broaden the understanding of the FNcollagen binding site modules presented by Skorsten-gaard et al.(1994), who focused on interactions withtype I gelatin only. Together the two sources of resultsadd to the concept that the smallest recombinant segmentof FN to retain gelatin binding properties is I6–II1–II2–I7.Inter-modular interactions between the type I and type

II modules, as detected in studies of the 42-kDa prote-olytic gelatin-binding segment of FN(Litvinovich et al.,1991), may explain the requirement for the I6 and I7

modules in addition to II1–II2 for achieving gelatinbinding. Such inter-modular interactions contribute tostructural stability in FN and MMPs. For example, thestability of the FN type II2 and I7 modules in the 42-kDa fragment is greatly enhanced by interactions thatmost likely involve modules II1 and I6, respectively(Litvinovich et al., 1991). Likewise, thermal stabilitystudies of FN type II-like modules in MMP-2 revealedthat the central module II2 has a significantly highermelting temperature in the tri-modular configuration(II1–II2–II3) than as an individual module(Banyai etal., 1996). Evidently, the module interactions in MMP-2 have functional implications as neither of the individ-ual modules bound gelatin as strongly as did thetri-modular domain of the enzyme(Banyai et al., 1994).FN-like type II modules have been identified in

several molecules, including the 72-kDa and 92-kDagelatinases(MMP-2 and -9) (Collier et al., 1988;Wilhelm et al., 1989). In spite of high levels of sequencesimilarity between the II modules of FN and these twoMMPs, only the isolated type II modules from MMP-2and MMP-9 have the capacity to bind collagen. Thisdifference between FN and the MMPs does not resultmerely from MMP-2 and MMP-9 containing three ratherthan two type II modules. Indeed, others and we havedemonstrated that truncated CBDs consisting of one ortwo of the three type II-like modules from MMP-2(Abbey et al., 1997; Overall et al., 2000) and MMP-9(Collier et al., 1992) retain at least part of their collagenbinding property. Furthermore, module II3 has beenfound to provide most of the collagen binding in MMP-9 (Collier et al., 1992).The available information about the precise nature of

the interactions between collagen and the FN type IImodules remains limited. On type I collagen, a majorFN binding site has been located in a 36-residue peptideof the cyanogen-bromide fragmenta1(I)-CB7, whichalso contains the collagenase cleavage site(Kleinmanet al., 1976, 1978). A second binding site was foundnear the NH -terminal end of the molecule(Guidry et2

al., 1990). The region of the main binding site lacksproline and hydroxyproline but contains a number ofhydrophobic residues within a 12-residue sequence.Together these features may destabilize the collagentriple helix locally and, thereby, allow FN binding at thesite. This model also is supported by the observationthat introduction of hydroxyprolines into this sequencein mutated collagens reduced FN binding accordingly(Dzamba et al., 1993).The understanding of the structure of the collagen

binding sites on FN type II modules has been enhancedby NMR analyses of module II2 from PDC-109 andmodule II1 from FN. These data point to a potentialcollagen binding site involving a large hydrophobicsurface that is formed by solvent-exposed aromaticresidues(Constantine et al., 1992; Pickford et al., 1997)


and contains a putative binding pocket for collagena1(I)-CB7 leucine and isoleucine side chains(Constan-tine et al., 1992; Pickford et al., 1997). In support ofthis proposed binding site, FN fragments bound toimmobilized gelatin can be eluted by reagents thatdisrupt hydrophobic interactions(Isaacs et al., 1989)(see also Fig. 1c). In addition, site-specific substitutionsof single conserved residues in the MMP-9 module II2hydrophobic surface with alanines decreased gelatinbinding of the module(Collier et al., 1992). Since ouranalyses demonstrated that FN type II module interac-tions with several collagens other than type I rely on ahydrophobic mode of binding, the binding may utilizethe same or similar residues on the FN type II modulesand segments of the collagens that bear structural simi-larity to the FN binding region on type I collagen.Recently, the structure of MMP-2 was solved by X-

ray crystallography(Morgunova et al., 1999). The inves-tigators showed that the hydrophobic pockets of thethree FN-like type II modules are arranged in a ‘three-pronged fishhook’ configuration. This orientation con-curs with our earlier functional observation that rCBDcan bind at least two different molecules of collagenconcurrently(Steffensen et al., 1995). Yet, the reducedaffinity and saturation level binding of collagen bytruncated, two-modular forms of the MMP-2 rCBDindicate that cooperation between the three type IImodules of the full-length CBD contribute to the strong-er collagen binding in the native molecule(Abbey etal., 1997; Overall et al., 2000).It is of special interest that rCBD competed bound

rI6–I7 from gelatin as demonstrated first by affinitychromatography and then verified by fluorescence spec-troscopy(Fig. 4). The rI6–I7 also displaced rCBD, butto a lesser extent. These observations provide initialevidence that rCBD and rI6–I7 utilize binding sites onthese ligands that are either identical or closely posi-tioned so that binding by one molecule may stericallyblock or compete interactions by the other. The lowerapparentK for gelatin interactions of rCBD relative tod

rI6–I7 could explain the preferential binding of rCBDto gelatin over rI6–I7. Alternatively, the tri-modularstructure of CBD may provide an enhanced binding siteconfiguration compared to rI6–I7.Cell attachment to FN is primarily mediated by

interactions of thea b integrin receptor with the Arg–5 1

Gly–Asp (RGD) sequence in module I10(Pierschbach-er and Ruoslahti, 1984; Pytela et al., 1985). However,additional segments of FN confer cell attachment prop-erties, including the CS-1 alternatively spliced module,which binds specifically to thea b receptor(Wayner4 1

et al., 1989). In addition, an NH -terminal 70-kDa2

proteolytic fragment of FN binds to attached fibroblasts(McKeown-Longo and Mosher, 1985). Our previousobservation that fibronectin-like type II modules fromMMP-2 support cell attachment via ab -integrin-1

dependent mechanism(Steffensen et al., 1998) suggest-ed the presence of a similar role in cell adhesion for thetype II modules in FN. Here we demonstrate thatHT1080 fibrosarcoma cells attach specifically to rI6–I7and r14AA, but not to rII1–II2. As anticipated from theevidence for reduced cell adhesion to FN RGD peptidesand proteolytic fragments(Ginsberg et al., 1985; Ruos-lahti, 1996), a likely result of differences in conforma-tion and peptide exposure(Ruoslahti, 1996), cellattachment to the recombinant collagen binding domainvariants in the present study also was less efficientcompared to full-length FN. We are currently character-izing in more detail the cell attachment mechanism andbehavior on these proteins.Activation of MMP-2 and MMP-9, which is observed

in HT1080 cells cultured on FN(Stanton et al., 1998),was significantly reduced when the cells were seededon the FN fragments containing type II modules.Although the membrane-type-1 MMP(MT1-MMP),which is central to the cell surface-mediated activationof pro-MMP-2, was processed to a 45-kDa shorter formduring the MMP-2 activation on FN(Stanton et al.,1998), the precise role of FN in the activation ofproMMP-2 remains, in part, unresolved. Cell attachmentand enzyme activation by cells cultured on FN involvea b integrin-mediated signals(Stanton et al., 1998).5 1

Corresponding to cell attachment via the MMP-2 CBD(Steffensen et al., 1998), a monoclonal antibody specificfor the b -integrin subunit blocked cell attachment to1

the isolated FN collagen binding domain pointing to acentral contribution ofb -integrins to cell attachment1

on this region of FN. It is noteworthy that an Arg–Gly–Asp sequence does not occur in the FN type IImodules. Interactions of theb -integrin with a different1

recognition sequence could explain the reduced cell-mediated MMP-2 activation. Therefore, the study ofdifferences in induction of MMP-2 and -9 activationbetween the full-length FN and the collagen binding FNfragments could provide a tool for use in mapping theactivation pathway.The potential biological impact resulting from modi-

fied cellular behavior on FN degradation fragments isconsiderable. First, such FN fragments have been detect-ed at increased levels in several pathological conditionsincluding chronic inflammation of arthritis(Carsons etal., 1985; Xie et al., 1992), poorly healing chronic ulcers(Wysocki and Grinnell, 1990; Grinnell et al., 1992),and periodontal disease(Talonpoika et al., 1993). Ourdemonstration of reduced cell attachment to FN frag-ments and inhibition of the MMP activation complementresults from other investigators who also showed alteredMMP expression in cells exposed to larger gelatinbinding FN fragments(Homandberg et al., 1992; Kapilaet al., 1996). Therefore, characterization of the molecu-lar basis for the interaction of cells and other matrixmolecules with FN type II modules could provide an


avenue for targeted control of cell behavior as well asenzyme activation and regulation.

Acknowledgments

We gratefully acknowledge Dr S. Weintraub, Univer-sity of Texas Health Science Center at San AntonioInstitutional Mass Spectrometry Laboratory, for perform-ing the electrospray mass spectrometry analyses. DrC.M. Overall, University of British Columbia, Vancou-ver, BC, Canada, Dr F. Grinnell, Southwest MedicalCenter, Dallas, TX, and Dr K. Yamada, the NationalInstitute of Dental and Craniofacial Research, NIH,Bethesda, MD, kindly provided antibodies, and Dr T.Kirsch generously made type X collagen available forour experiments. This work was supported by grants�DE12818 and� DE14236 from the NIDCR, NIH,Bethesda, MD, and a grant from the South Texas HealthResearch Center.

References

Abbey, R.S., Steffensen, B., Overall, C.M., 1997. Protein engineeringof the 72-kDa gelatinase collagen binding domain. J. Dent. Res.76, 300.

Banyai, L., Tordai, H., Patthy, L., 1994. The gelatin-binding site ofhuman 72 kDa type IV collagenase(gelatinase A). Biochem. J.298, 403–407.

Banyai, L., Tordai, H., Patthy, L., 1996. Structure and domain-domaininteractions of the gelatin-binding site of human 72-kilodalton typeIV collagenase(gelatinase A, matrix metalloproteinase 2). J. Biol.Chem. 271, 12003–12008.

Banyai, L., Trexler, M., Koncz, S., Gyenes, M., Sipos, G., Patthy, L.,1990. The collagen-binding site of type-II units of bovine seminalfluid protein PDC-109 and fibronectin. Eur. J. Biochem. 193,801–806.

Carsons, S., Lavietes, B.B., Diamond, H.S., Kinney, S.G., 1985. Theimmunoreactivity, ligand, and cell binding characteristics of rheu-matoid synovial fluid fibronectin. Arthritis. Rheum. 28, 601–612.

Chomczynski, P., Sacchi, N., 1987. Single-step method of RNAisolation by acid guanidinum thiocyanate-phenol-chloroformextraction. Anal. Biochem. 162, 156–159.

Collier, I.E., Krasnov, P.A., Strongin, A.Y., Birkedal-Hansen, H.,Goldberg, G.I., 1992. Alanine scanning mutagenesis and functionalanalysis of the fibronectin-like collagen-binding domain fromhuman 92-kDA type IV. J. Biol. Chem. 267, 6776–6781.

Collier, I.E., Wilhelm, S.M., Eisen, A.Z., et al., 1988. H-ras oncogene-transformed human bronchial epithelial cells(TBE-1) secrete asingle metalloprotease capable of degrading basement membranecollagen. J. Biol. Chem. 263, 6579–6587.

Constantine, K.L., Madrid, M., Banyai, L., Trexler, M., Patthy, L.,Llinas, M., 1992. Refined solution structure and ligand-bindingproperties of PDC-109 domain b. A collagen-binding type IIdomain. J. Mol. Biol. 223, 281–298.

Dzamba, B.J., Wu, H., Jaenisch, R., Peters, D.M., 1993. Fibronectinbinding site in type I collagen regulates fibronectin fibril formation.J. Cell Biol. 121, 1165–1172.

Eltis, L.D., Iwagami, S.G., Smith, M., 1994. Hyperexpression of asynthetic gene encoding a high potential iron sulfur protein. ProteinEng. 7, 1145–1150.

Esch, F.S., Ling, N.C., Bohlen, P., Ying, S.Y., Guillemin, R., 1983.Primary structure of PDC-109, a major protein constituent ofbovine seminal plasma. Biochem. Biophys. Res. Commun. 113,861–867.

Ginsberg, M.H., Pierschbacher, M., Ruoslahti, E., Marguerie, G.,Plow, E.F., 1985. Inhibition of fibronectin binding to platelets byproteolytic fragments and synthetic peptides which support fibro-blast adhesion. J. Biol. Chem. 260, 3931–3936.

Grinnell, F., Ho, C.-H., Wysocki, A.B., 1992. Degradation of fibro-nectin and vitronectin in chronic wound fluid: Analysis by cellblotting, immunoblotting, and cell adhesion assays. J. Invest.Dermatol. 98, 410–416.

Guidry, C., Miller, E.J., Hook, M., 1990. A second fibronectin-binding region is present in collagen alpha chains. J. Biol. Chem.265, 19230–19236.

Guillemette, J.G., Matsushima-Hibiya, Y., Atkinson, T., Smith, M.,1991. Expression inE. coli of a synthetic gene coding for horseheart myoglobin. Protein Eng. 4, 585–592.

Hackett, A.J., Smith, H.S., Springer, E.L., et al., 1977. Two synergisticcell lines from human breast tissue: The aneuploid mammaryepithelial (Hs578T) and the diploid myoepithelial(Hs578st) celllines. J. Natl. Cancer Inst. 58, 1795–1806.

Hahn, S.-H.E., Yamada, K.M., 1979. Isolation and biological char-acterization of active fragments of the adhesive glycoproteinfibronectin. Cell 18, 1043–1051.

Homandberg, G.A., Meyers, R., Xie, D.-L., 1992. Fibronectin frag-ments cause chondrolysis of bovine articular cartilage slices inculture. J. Biol. Chem. 267, 3597–3604.

Huhtala, P., Humphries, M.J., McCarthy, J.B., Tremble, P.M., Werb,Z., Damsky, C.H., 1995. Cooperative signaling by alpha 5 beta 1and alpha 4 beta 1 integrins regulates metalloproteinase geneexpression in fibroblasts adhering to fibronectin. J. Cell Biol. 129,867–879.

Hynes, R.O., 1990. Fibronectins. Springer-Verlag, New York.Ingham, K.C., Brew, S.A., Isaacs, B.S., 1988. Interaction of fibronec-

tin and its gelatin-binding domains with fluorescent-labeled chainsof type I collagen. J. Biol. Chem. 263, 4624–4628.

Ingham, K.C., Brew, S.A., Migliorini, M.M., 1989. Further localiza-tion of the gelatin-binding determinants within fibronectin. Activefragments devoid of type II homologous repeat modules. J. Biol.Chem. 264, 16977–16980.

Isaacs, B.S., Brew, S.A., Ingham, K.C., 1989. Reversible unfoldingof the gelatin-binding domain of fibronectin: Structural stability inrelation to function. Biochemistry 28, 842–850.

Kaderbhai, M.A., Austen, B.M., 1985. Studies on the formation ofintrachain disulphide bonds in newly biosynthesized bovine prolac-tin. Role of protein-disulphide isomerase. Eur. J. Biochem. 153,167–178.

Kapila, Y.L., Kapila, S., Johnson, P.W., 1996. Fibronectin and fibro-nectin fragments modulate the expression of proteinases andproteinase inhibitors in human periodontal ligament cells. MatrixBiol. 15, 251–261.

Kapila, Y.L., Wang, S., Johnson, P.W., 1999. Mutations in the heparinbinding domain of fibronectin in cooperation with the V regioninduce decreases in pp125FAK levels plus proteoglycan-mediatedapoptosis via caspases. J. Biol. Chem. 274, 30906–30913.

Keung, W., Silber, E., Eppenberger, U., 1989. Quantification of cellscultured on 96-well plates. Anal. Biochem. 182, 16–19.

Kleinman, H.K., McGoodwin, E.B., Klebe, R.J., 1976. Localizationof the cell attachment region in Types I and II collagens. Biochem.Biophys. Res. Commun. 72, 426–432.

Kleinman, H.K., McGoodwin, E.B., Martin, G.R., Klebe, R.J.,Fietzek, P.P., Woolley, D.E., 1978. Localization of the binding


sitefor cell attachment in the a1(I) chain of collagen. J. Biol.Chem. 253, 5642–5646.

Kornblihtt, A.R., Umezawa, K., Vibe-Pedersen, K., Baralle, F.E.,1985. Primary structure of human fibronectin: differential splicingmay generate at least 10 polypeptides from a single gene. EMBOJ. 4, 1755–1759.

Laemmli, U.K., 1970. Cleavage of structural proteins during theassembly of the head of bacteriophage T4. Nature 227(259),680–685.

Litvinovich, S.V., Strickland, D.K., Medved, L.V., Ingham, K.C.,1991. Domain structure and interactions of the type I and type IImodules in the gelatin-binding region of fibronectin. All sixmodules are independently folded. J. Mol. Biol. 217, 563–575.

Lobel, P., Dahms, N.M., Breitmeyer, J., Chirgwin, J.M., Kornfeld, S.,1987. Cloning of the bovine 215-kDa cation-independent mannose6-phosphate receptor. Proc. Natl. Acad. Sci. USA 84, 2233–2237.

Luck, D.N., Gout, P.W., Sutherland, E.R., Fox, K., Huyer, M., Smith,M., 1992. Analysis of disulphide bridge function in recombinantbovine prolactin using site-specific mutagenesis and renaturationunder mild alkaline conditions: a crucial role for the centraldisulphide bridge in the mitogenic activity of the hormone. ProteinEng. 5, 559–567.

MacMullen, B.A., Fujikawa, K., 1985. Amino acid sequence of theheavy chain of human alpha- factor XIIa(activated Hagemanfactor). J. Biol. Chem. 260, 5328–5341.

McKeown-Longo, P.J., Mosher, D.F., 1985. Interaction of the 70,000-mol-wt amino-terminal fragment of fibronectin with the matrix-assembly receptor of fibroblasts. J. Cell Biol. 100, 364–374.

Mecham, R.P., Heuser, J.E., 1991. The elastic fiber. In: Hay, E.D.(Ed.), Cell Biology of Extracellular Matrix, Vol. 2. Plenum Press,New York, pp. 79–109.

Morgan, D.O., Edman, J.C., Standring, D.N., et al., 1987. Insulin-like growth factor II receptor as a multifunctional binding protein.Nature 329, 301–307.

Morgunova, E., Tuuttila, A., Bergmann, U., et al., 1999. Structure ofhuman pro-matrix metalloproteinase-2: Activation mechanismrevealed. Science 284, 1667–1670.

Overall, C.M., Wallon, U.M., Steffensen, B., DeClerk, Y., Tschesche,H., Abbey, R.S., 2000. Substrate and TIMP interactions with humangelatinase A recombinant COOH-terminal hemopexin-like andfibronectin type II-like domains: Both the N- and C-domains ofTIMP-2 bind the C-domain of gelatinase A. In: Edwards, D., et al.(Eds.), Inhibitors of Metalloproteinases in Development and Dis-ease. Gordon & Breach, Amsterdam, Holland, pp. 57–69.

Overall, C.M., Wrana, J.L., Sodek, J., 1989. Independent regulationof collagenase, 72-kDa progelatinase, and metalloendoproteinaseinhibitor expression in human fibroblasts by transforming growthfactor-b. J. Biol. Chem. 264, 1860–1869.

Owens, R.J., Baralle, F.E., 1986. Mapping the collagen-binding siteof human fibronectin by expression inEscherichia coli. EMBO J.5, 2825–2830.

Pickford, A.R., Potts, J.R., Bright, J.R., Phan, I., Campbell, I.D.,1997. Solution structure of a type 2 module from fibronectin:Implications for the structure and function of the gelatin-bindingdomain. Structure 5, 359–370.

Pierschbacher, M.D., Ruoslahti, E., 1984. Cell attachment activity offibronectin can be duplicated by small synthetic fragments of themolecule. Nature 309, 30–33.

Piez, K.A., 1967. Soluble collagen and the components resulting fromits denaturation. In: Ramaclandran, G.N.(Ed.), Treatise on colla-gen. Academic Press, London, pp. 207–252.

Pytela, R., Pierschbacher, M.D., Argraves, S., Suzuki, S., Ruoslahti,E., 1987. Arginine–glycine–aspartic acid adhesion receptors. Meth-ods Enzymol. 144, 475–489.

Pytela, R., Pierschbacher, M.D., Ruoslahti, E., 1985. Identificationand isolation of a 140 kD cell surface glycoprotein with propertiesexpected of a fibronectin receptor. Cell 40, 191–198.

Ruoslahti, E., 1996. RGD and other recognition sequences forintegrins. Annu. Rev. Cell Dev. Biol. 12, 697–715.

Sanger, F., Coulson, A.R., Friedman, T., et al., 1978. The nucleotidesequence of bacteriophage ÆX174. J. Mol. Biol. 125, 225–246.

Seidah, N.G., Manjunath, P., Rochemont, J., Sairam, M.R., Chretien,M., 1987. Complete amino acid sequence of BSP-A3 from bovineseminal plasma. Homology to PDC-109 and to the collagen-bindingdomain of fibronectin. Biochem. J. 243, 195–203.

Shipley, J.M., Doyle, G.A.R., Fliszar, C.J., et al., 1996. The structuralbasis for the elastolytic activity of the 92-kDa and 72-kDa gelati-nases. Role of the fibronectin type II-like repeats. J. Biol. Chem.271, 4335–4341.

Skorstengaard, K., Holtet, T.L., Etzerodt, M., Thogersen, H.C., 1994.Collagen-binding recombinant fibronectin fragments containingtype II domains. FEBS Lett. 343, 47–50.

Skorstengaard, K., Jensen, M.S., Sahl, P., Petersen, T.E., Magnusson,S., 1986. Complete primary structure of bovine plasma fibronectin.Eur. J. Biochem. 161, 441–453.

Stanton, H., Gavrilovic, J., Atkinson, S.J., et al., 1998. The activationof proMMP-2 (gelatinase A) by HT1080 fibrosarcoma cells ispromoted by culture on a fibronectin substrate and is concomitantwith an increase in processing of MT1-MMP(MMP-14) to a 45kDA form. J. Cell. Sci. 111, 2789–2798.

Steffensen, B., Bigg, H.F., Overall, C.M., 1998. The involvement ofthe fibronectin type II-like modules of human gelatinase A in cellsurface localization and activation. J. Biol. Chem. 273,20622–20628.

Steffensen, B., Wallon, U.M., Overall, C.M., 1995. Extracellularmatrix binding properties of recombinant fibronectin type II-likemodules of human 72-kDa gelatinaseytype IV collagenase. Highaffinity binding to native type I collagen but not native type IVcollagen. J. Biol. Chem. 270, 11555–11566.

Talonpoika, J., Paunio, K., Soderling, E., 1993. Molecular forms andconcentration of fibronectin and fibrin in human gingival crevicularfluid before and after periodontal treatment. Scand. J. Dent. Res.101, 375–381.

Taylor, M.E., Conary, J.T., Lennartz, M.T., Stahl, P.D., Drickamer,K., 1990. Primary structure of the mannose receptor containsmultiple motifs resembling carbohydrate-recognition domains. J.Biol. Chem. 265, 12156–12162.

Wallon, U.M., Overall, C.M., 1997. The COOH-terminal hemopexin-like domain of human gelatinase A(MMP-2) requires Ca for2q

fibronectin and heparin binding: Binding properties of recombinantgelatinase A C-domain to extracellular matrix and basementmembrane components. J. Biol. Chem. 272, 7473–7481.

Wayner, E.A., Garcia-Pardo, A., Humphries, M.J., McDonald, J.A.,Carter, W.G., 1989. Identification and characterization of the Tlymphocyte adhesion receptor for an alternative cell attachmentdomain(CS-1) in plasma fibronectin. J. Cell Biol. 109, 1321–1330.

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.

Wilhelm, S.M., Collier, I.E., Marmer, B.L., Eisen, A.Z., Grant, G.A.,Goldberg, G.I., 1989. SV40-transformed human lung fibroblastssecrete a 92-kDa type IV collagenase which is identical to thatsecreted by normal human macrophages. J. Biol. Chem. 264,17213–17221.

Wysocki, A.B., Grinnell, F., 1990. Fibronectin profiles in normal andchronic wound fluid. Lab. Invest. 63, 825–831.


Xie, D.-L., Meyers, R., Homandberg, G.A., 1992. Fibronectin frag-ments in osteoarthritic synovial fluid. J. Rheumatol. 19, 1448–1452.

Zardeneta, G., Milam, S.B., Schmitz, J.P., 2000. Iron-dependentgeneration of free radicals: plausible mechanisms in the progressivedeterioration of the temporomandibular joint. J. Oral Maxillofac.Surg. 58, 302–308.

Zardi, L., Carnemolla, B., Balza, E., et al., 1985. Elution of fibronectinproteolytic fragments from a hydroxyapatite chromatography col-umn. A simple procedure for the purification of fibronectindomains. Eur. J. Biochem. 146, 571–579.

human fibronectin and mmp-2 collagen binding domains compete

Documents