Latent and Active Transforming Growth Factor b1 Releasedfrom Genetically Modi®ed Keratinocytes ModulatesExtracellular Matrix Expression by Dermal Fibroblasts in aCoculture System

Barbara S. Bauer, Edward E. Tredget,² Yvonne Marcoux, Paul G. Scott,* and Aziz GhaharyDepartment of Surgery, Wound Healing Research Group; *Department of Biochemistry, and ²Division of Plastic and Reconstructive Surgery,

Division of Critical Care, University of Alberta, Edmonton, Canada

Transforming growth factor b1 is a multifunctionalcytokine involved in many aspects of wound healing.Here we report the effects of both latent and activetransforming growth factor b1 released from genetic-ally modi®ed keratinocytes on extracellular matrixexpression by dermal ®broblasts in a coculture sys-tem. Human keratinocytes were genetically modi®edwith adenovirus containing cDNA for latent trans-forming growth factor b1 (AdTGF-b1) or activetransforming growth factor b1 (AdTGF-b1223/225) orLacZ and cultured with human dermal ®broblasts.Northern blotting for mRNA con®rmed that kerati-nocytes were successfully transduced with the adeno-viruses as the cDNA transcripts are smaller thannative transforming growth factor b1 mRNA. Anenzyme-linked immunosorbent assay speci®c fortransforming growth factor b1 demonstrated that thetransforming growth factor b1 produced by thegenetically modi®ed keratinocytes was able to passthrough the membrane separating the two cell layers.Levels of transforming growth factor b1 were signi®-cantly higher for both latent (p < 0.0001) and active(p < 0.0001) transforming growth factor b1 compared

to the LacZ control. Without acid activation ofsamples, keratinocytes transduced with the activetransforming growth factor b1 construct exhibitedsigni®cantly higher levels of transforming growthfactor b1 than either the latent construct or the LacZcontrol (p < 0.0001). The transforming growth factorb1 produced was biologically active, as shown by theplasminogen activator inhibitor assay (p < 0.0001).To demonstrate that transforming growth factor b1had an effect on underlying ®broblasts, mRNA wasextracted and analyzed using Northern analysis.Latent transforming growth factor b1 signi®cantlyincreased the expression of type I collagen mRNA(p < 0.05) but did not signi®cantly affect collagenasemRNA. Active transforming growth factor b1signi®cantly increased type I collagen mRNA(p <0.005) while also decreasing collagenase mRNA(p <0.05). These results illustrate the ability ofincreased levels of transforming growth factor b1 tooverride the effects of normal keratinocytes on thebehavior of dermal ®broblasts. Key words: adenovirus/coculture/®broblasts/keratinocytes/transforming growth fac-tor b1. J Invest Dermatol 119:456±463, 2002

Transforming growth factor b1 (TGF-b1) is a multi-functional growth factor important in the process ofwound healing due to its ®brogenic effects. FiveTGF-b isoforms have been identi®ed to date, withTGF-b1, TGF-b2, and TGF-b3 being found in

mammals. During wound healing, TGF-b1 stimulates extracellularmatrix (ECM) synthesis and deposition by inducing ®broblasts tosynthesize collagen, ®bronectin, and glycosaminoglycans (Ignotzand Massague, 1986; Sporn et al, 1987). TGF-b1 also modulates theexpression of proteases and their inhibitors (Edwards et al, 1987;Overall et al, 1989), acts as a chemoattractant for monocytes

(Wake®eld et al, 1987) and ®broblasts (Postlethwaite et al, 1987),and enhances neovascularization (Allen et al, 1993).

TGF-b1 is secreted as a latent complex (LTGF-b1) consisting ofa 25 kDa dimeric mature protein and an N-terminal pro-proteincalled the latency-associated peptide (LAP) (Tsuji et al, 1990). It hasbeen suggested that either a conformational change of the latentcomplex (Ghahary et al, 1999) or dissociation of LAP from themature protein is required for activation of TGF-b1 (Dallas et al,1994; Okazaki et al, 1995) as the TGF-b1 receptors do notrecognize native LTGF-b1 (Wake®eld et al, 1987). Although themechanisms by which TGF-b1 is activated in vivo have not beenfully elucidated, plasmin and cathepsin D are capable of activatingTGF-b1 by cleaving LAP from the mature protein in theN-terminal region (Lyons et al, 1988, 1990; Sato and Rifkin,1989; Sato et al, 1990). In vitro, TGF-b1 can be activated byextremes of pH and heat (80°C for 10 min) and by detergents(Gleizes et al, 1997). Latency is very important in the regulation ofTGF-b and previous experiments have shown very different

Manuscript received July 10, 2001; revised February 14, 2002; acceptedfor publication February 27, 2002.

Reprint requests to: Dr. Aziz Ghahary, 161 HMRC, Wound HealingResearch Group, Department of Surgery, University of Alberta,Edmonton, Canada T6G 2E1.

0022-202X/02/$15.00 ´ Copyright # 2002 by The Society for Investigative Dermatology, Inc.

456

outcomes depending on the state of the TGF-b1 (Sellheyer et al,1993; Sime et al, 1997).

Many growth factors require prolonged exposure to the woundenvironment before eliciting an effect. Lynch et al (1989)demonstrated that, in a partial-thickness dermatome wound inpigs, single topical applications of epidermal growth factor (EGF),TGF-a, basic ®broblast growth factor (bFGF), insulin-like growthfactor 1 (IGF-1), and platelet-derived growth factor BB did notaccelerate healing. It has also been reported that daily injections ofEGF into polyvinyl alcohol sponges implanted subcutaneously intorats produced only a small increase in granulation tissue whereasusing slow-release pellets carrying EGF resulted in a signi®cantincrease in granulation tissue volume and organization (Buckley etal, 1985). The concentration of recombinant growth factorsrequired to stimulate DNA synthesis or cell migration in vivo is1000 times higher than that required in vitro (Bennett and Schultz,1993). This may be due to proteolytic degradation and/or diffusionof the growth factor from the wound. Thus a potential solution tothis problem seems to be the use of a genetically modi®ed skinsubstitute through which a wound healing promoting factor such asTGF-b1 is released at the wound site. This study was conducted toachieve the following objectives: (i) to examine the expression ofboth latent and active TGF-b1 from genetically modi®ed humankeratinocytes; (ii) to demonstrate that the TGF-b1 produced wasbiologically active using a novel sensitive bioassay for TGF-b1activity, the plasminogen activator inhibitor/luciferase (PAI/L)assay (Abe et al, 1994); (iii) to study the effect of latent and activeTGF-b1 on the expression of type I collagen and collagenase inhuman dermal ®broblasts cocultured with TGF-b1 geneticallymodi®ed keratinocytes.

METHODS AND MATERIALS

Adenoviral vectors expressing active TGF-b1, latent TGF-b1, or LacZ asa control were used in this study. Two recombinant adenoviruses werekindly provided by Dr. Jack Gauldie (Department of Biology, HealthSciences Center, McMaster University) (Sime et al, 1997). The AdLacZcontrol was obtained from Quantum Biotechnologies, Chaval, Quebec,Canada. Brie¯y, both recombinant TGF-b1 adenoviruses contain thecDNA of the coding region of full-length porcine TGF-b1 as describedby Sime et al (1997). The cDNA fragments for latent and active TGF-b1were 1.5 and 1.1 kb, respectively. Latent TGF-b1 was expressed by theAdTGF-b1 construct and active TGF-b1 was expressed by the AdTGF-b1223/225 construct. TGF-b1 was expressed as the spontaneously activeprotein by using a mutant in which Cys223 and Cys225 in the TGF-b1pro-peptide were converted to Ser. This results in the dissociation of thepro-peptide and secretion of bioactive TGF-b1 (Brunner et al, 1989;Samuel et al, 1992; Sime et al, 1997). The adenoviruses were ampli®edin 293 A cells and titers were estimated from the tissue culture infectiousdose50 (TCID50), so that the multiplicity of infection (MOI) could beestablished.

Cell cultures and gene transduction The procedure of Rheinwaldand Green (1975) was used for cultivation of human foreskinkeratinocytes using serum-free keratinocyte medium (KSFM) (Gibco,Grand Island, NY) supplemented with bovine pituitary extract (25 mgper ml) and EGF (0.5 ng per ml). Primary cultured keratinocytes atpassages 3±5 were used. To establish the ®broblast cultures normal skinpunch biopsies, obtained from patients undergoing electivereconstructive surgery, were established in Dulbecco's modi®ed Eagle'smedium (DMEM) supplemented with 10% fetal bovine serum (FBS), aspreviously described (Ghahary et al, 1994). Strains of dermal ®broblasts atpassages 4±6 were used in this study. Keratinocytes were then grownalone in the upper chamber on a permeable Millicell culture plate insert(Millipore, Bedford, MA). These inserts have a pore size of 0.4 mm andare coated with 0.3 mg bovine type I collagen. Dermal ®broblasts weregrown in the lower chamber of the coculture system.

Keratinocytes were plated at a density of 4 3 105 cells per insert andincubated 24 h. An MOI of 0.5 was used to transduce the keratinocyteswith the adenoviruses overnight. The keratinocytes were washed and theinserts were placed into empty six-well plates or plates containing 1 3105 ®broblasts per well and 4 ml fresh medium (49% KSFM, 49%DMEM, 2% FBS) was added. The cell cultures were then incubated for48 h and conditioned medium was collected and tested for the presenceof active and latent TGF-b1.

Fibroblast proliferation Cocultures were established as stated aboveusing two strains of keratinocytes each in combination with two strainsof ®broblasts (four separate experiments). Keratinocytes were transducedwith LacZ, TGF-b1, or AdTGF-b1223/225 prior to coculturing with®broblasts. Each condition was tested in triplicate. Following a 48 hincubation, ®broblasts were counted using a Coulter counter (ModelZM, Coulter Electronics, Bedfordshire, U.K.).

Enzyme-linked immunosorbent assay (ELISA) for TGF-b1 Todetermine the concentration of TGF-b1, a sandwich ELISA was usedbased on the procedure reported by Danielpour et al (1989). Brie¯y, 96-well plates were coated with 100 ml per well of mouse monoclonalantibody to human TGF-b (R&D Systems, Minneapolis, MN) at aconcentration of 1 mg per ml in phosphate-buffered saline (PBS). Theplates were incubated for 3 h at room temperature followed by 16 h at4°C. After washing twice with PBS±Tween-20 (PBS-T), the plates wereblocked with 1% bovine serum albumin (BSA, crystallized, Sigma) for60 min at room temperature and washed twice with PBS-T. To activateTGF-b1, the culture medium was acidi®ed with 12 ml per 500 mlconditioned medium of 3 N HCl for 15 min at room temperature andneutralized with 35 ml of 1 M HEPES/5 N NaOH (5:2). One hundredmicroliters of the acidi®ed/neutralized samples was added to each welland the plates were then incubated at 37°C for 1 h. After washing withPBS-T, the plates were incubated with 100 ml per well of chickenantihuman TGF-b (R&D Systems) at a concentration of 2.5 mg per mlfor 1 h. After washing ®ve times with PBS-T, the plates were incubatedwith alkaline phosphatase-conjugated rabbit antichicken IgG (Sigma) atroom temperature for 1 h followed by washing ®ve times with PBS-T.After addition of the substrate (o-nitrophenyl phosphate, 1.5 mg per ml,Sigma), the plates were incubated at room temperature for 1 h and theoptical density was read using a THERMOmax (Molecular Devices,Menlo Park, CA) microplate reader at a wavelength of 405 nm. Serialdilutions (0, 125, 250, 625, 1250, to 2500 pg per ml) of recombinanthuman TGF-b1 (R&D Systems) were used to prepare a standard curve.

TGF-b1 time course Keratinocytes were plated at 1.5 3 106 cells per75 cm2 ¯ask. Cells were transduced with an MOI of 0.5 for eachadenovirus or incubated in medium alone. Conditioned medium wascollected every 2 d and replaced with new medium. The TGF-b1ELISA was performed for each time point. Total TGF-b1 was measuredfollowing acidi®cation and active TGF-b1 was evaluated by measuringnonacidi®ed samples.

RNA extraction and hybridization Following 48 h incubation,culture medium was removed and cell layers were lyzed in 2 ml ofguanidinium thiocyanate (GITC) as previously described (Ghahary et al,1995). Total RNA was extracted by the GITC/CsCl procedure ofChirgwin et al (1979); 10 mg of total mRNA from each treatment wasseparated by electrophoresis and blotted onto nitrocellulose ®lters. Filterswere then baked under vacuum for 2 h at 80°C and prehybridized in asolution containing 50% formamide, 0.3 M sodium chloride, 20 mMTris±HCl (pH 8.0), 1 mM ethylenediamine tetraacetic acid, 1 3Denhardt's solution (0.02% BSA, Ficoll, and polyvinylpyrrolidone),0.005% salmon sperm DNA, and 0.005% poly(A) for 3±4 h at 45°C.Hybridization was performed in the same solution at 45°C for 16±20 husing cDNA probes for human TGF-b1, type I collagen, collagenase, or18S ribosomal RNA. The probe for 18S rRNA was used to ensureequal loading of samples. The probes were labeled with 32P-g-CTP(DuPont Canada, Streetsville, Mississauga, Ontario, Canada) by nicktranslation. Filters were initially washed at room temperature with 2 3SSC (SSC = 0.15 M sodium chloride, 0.015 M sodium citrate) and 0.1%sodium dodecyl sulfate (SDS) for 30 min, and then for 20 min at 65°Cin 0.1 3 SSC and 0.1% SDS solution. Autoradiography was performedby exposing Kodak X-Omat ®lm to the nitrocellulose ®lters at ±70°C inthe presence of an enhancing screen. The cDNA probes for 18Sribosomal RNA and collagenase were obtained from the American TypeCulture Collection (Rockville, MD). The other cDNA probes weregifts: TGF-b (Dr. G.I. Bell, Howard Hughes Medical Institute,Department of Biochemistry and Molecular Biology and Medicine,University of Chicago, IL) and type I procollagen (Drs G. Tromp, H.Kuivaniemi, and L. Ala-Kokko, Department of Biochemistry andMolecular Biology, Jefferson Institute of Molecular Medicine,Philadelphia, PA).

Luciferase assay for TGF-b1 Conditioned medium was collectedafter 48 h and assayed for TGF-b1 bioactivity using mink lung cellsbearing the PAI/L construct (Abe et al, 1994). Mink lung epithelial cellstransfected with this construct (a generous gift from Dr. Daniel Rifkin,New York University Medical Center, New York, NY) were cultured

VOL. 119, NO. 2 AUGUST 2002 LATENT AND ACTIVE TGF-b1 IN A COCULTURE SYSTEM 457

in DMEM supplemented with 10% FBS, L-glutamine (2 mM), penicillinG (100 U per ml), streptomycin sulfate (100 mg per ml), andamphotericin B (250 ng per ml). Samples were acid activated (pH 3) for15 min followed by neutralization. Conditioned medium collected fromTGF-b1 and TGF-b1223/225 transduced cells was also analyzed withoutacid activation to determine the relative amount of spontaneously activeTGF-b1. Samples known to contain high concentrations of TGF-b1were diluted in 0.1% pyrogen-poor BSA (wt/vol) (Pierce, Rockford, IL)in serum-free DMEM. A standard curve was constructed (0±500 pg perml) using recombinant human TGF-b1 (R&D Systems) diluted in 0.1%BSA in DMEM.

Transfected mink lung epithelial cells were plated at a density of 1.63 104 cells per well in 96-well tissue culture plates. The cells wereallowed to attach for 3 h at 37°C in an atmosphere of 5% CO2. Themedium was replaced with conditioned media or standards and incubatedfor 14 h at 37°C. Cells were then washed with PBS and lysed using50 ml of 1 3 cell lysis buffer (Analytical Luminescence, San Diego, CA)for 20 min at room temperature on a rotating platform. The cell lysates(45 ml) were then transferred to clear polypropylene scintillation vialsinto which 100 ml of both substrate A and B (Analytical Luminescence)were added and samples were immediately analyzed using the BeckmanLS 6000 TA scintillation counter.

Statistical analysis For ELISA and PAI/L data, the differencesbetween the LacZ transduced control cells and either AdTGF-b1 orAdTGF-b1223/225 transduced cells were evaluated using a two-tailednonparametric test (alternative Welch t test). Samples were measured intriplicate. Data are presented as means 6 1 standard deviation. Student'st test was used for statistical analysis of mRNA results. p-values < 0.05were considered signi®cant for all statistical tests.

RESULTS

The kinetics of TGF-b1 expression were studied over 12 d.Conditioned medium was collected every 48 h for ELISAmeasurement of active and total TGF-b1. Signi®cantly higherlevels of TGF-b1 were released from keratinocytes transduced withAdTGF-b1 or AdTGF-b1223/225 compared to the LacZ control atall times for at least 12 d (Fig 1A). The level of constitutively activeTGF-b1 released from AdTGF-b1223/225 in nonacidi®ed sampleswas also signi®cantly higher than with AdTGF-b1 at all times(Fig 1B).

To determine whether TGF-b1 produced by the transducedkeratinocytes was able to pass through the insert membrane toin¯uence dermal ®broblasts grown in the lower chamber, mediumwas collected from the top and bottom chambers and analyzedfollowing acid activation using the TGF-b1 ELISA (data notshown). There was no signi®cant difference in the concentration ofTGF-b1 in the upper and lower chambers. This ®nding shows thatdiffusion of TGF-b1 is not impeded by the insert.

TGF-b1 released from keratinocytes was shown to increase®broblast proliferation. Keratinocytes transduced with eitherAdTGF-b1 or AdTGF-b1223/225 signi®cantly increased the prolif-eration of ®broblasts in the lower chamber compared to the LacZcontrol (p < 0.0001) (Fig 2).

To demonstrate that cells transduced with the AdTGF-b1223/225

construct produced enhanced levels of active TGF-b1, whereasthose transduced with the AdTGF-b1 construct did not, activeTGF-b1 was measured by ELISA in samples that were not acidi®ed(Fig 3). The conditioned medium derived from keratinocytestransduced with AdTGF-b1223/225 had a high mean level of activeTGF-b1 (731 6 198 and 818 6 265 pg per ml for keratinocytesalone and keratinocytes cocultured with ®broblasts, respectively)compared to keratinocytes transduced with AdTGF-b1 (31 6 29and 36 6 39 pg per ml alone and with ®broblasts, respectively).The difference between the level of active TGF-b1 from cellstransduced with AdTGF-b1223/225 and those transduced withAdTGF-b1 was found to be extremely signi®cant (p < 0.0001).

To examine whether the level of protein production is consistentwith TGF-b1 gene expression, TGF-b1 mRNA was analyzedusing northern analysis (Fig 4). Doublet bands were seen in lanescontaining keratinocytes transduced with either AdTGF-b1 orAdTGF-b1223/225. An increase in endogenous TGF-b1 was also

noted in ®broblasts that were cocultured with the geneticallymodi®ed keratinocytes.

The size of the mRNA transcript is smaller for porcine AdTGF-b1 and AdTGF-b1223/225 compared to endogenous human TGF-b1. The autoradiograph shows that there are two transcripts presentin keratinocytes that have been transduced with either AdTGF-b1(appears as a doublet) or AdTGF-b1223/225, whereas only onetranscript was seen in the controls and in ®broblasts (endogenousTGF-b1 mRNA). This indicates that the keratinocytes weresuccessfully transduced with the porcine cDNA TGF-b1 constructs

Figure 1. Time course of release of TGF-b1 from geneticallymodi®ed keratinocytes. (A) Conditioned medium was collected every2 d and total TGF-b1 was measured using an ELISA followingacidi®cation. Compared to the LacZ control (squares), signi®cantly moreTGF-b1 was released from AdTGF-b1 (circles) and AdTGF-b1223/225

(triangles) (*p < 0.0001, **p < 0.001). (B) Constitutively active TGF-b1was measured in nonacidi®ed samples. Signi®cantly more active TGF-b1was released from AdTGF-b1223/225 (black bar) than AdTGF-b1 (open bar)at all times (p < 0.0001, n = 3).

Figure 2. TGF-b1 released from genetically modi®edkeratinocytes increases the proliferation of dermal ®broblasts.Four separate experiments were performed with different cell strains.LacZ was normalized (LacZ = 1) for all experiments. There was asigni®cant increase in ®broblast proliferation in the presence of AdTGF-b1 or AdTGF-b1223/225 compared to the LacZ control (*p < 0.0001).

458 BAUER ET AL THE JOURNAL OF INVESTIGATIVE DERMATOLOGY

and that the ®broblasts in coculture with the transducedkeratinocytes were not subsequently infected by adenovirus. Eachblot was probed with a cDNA speci®c for 18S rRNA as a loadingcontrol.

To compare the levels of total TGF-b1 released by thekeratinocytes transduced by the latent and active constructs,samples were acidi®ed and total TGF-b1 was measured in mediumfrom keratinocytes alone (Fig 5A) and those cocultured withdermal ®broblasts (Fig 5B).

The concentration of total TGF-b1 in conditioned mediumfrom AdTGF-b1 and AdTGF-b1223/225 transduced keratinocytesboth alone and cocultured with ®broblasts was signi®cantly higherthan that from the AdLacZ transduced keratinocyte control (p< 0.0001). Keratinocytes alone transduced with AdTGF-b1 andcocultured with ®broblasts exhibited signi®cantly higher levels oftotal TGF-b1 compared to those transduced with AdTGF-b1223/

225.In order to examine whether the TGF-b1 released from

transduced keratinocytes was biologically active, the PAI/L assay

was performed. Following acid activation, the relative concentra-tions of total TGF-b1 followed the same trends as those seen byELISA, both for keratinocytes alone (Fig 6A) and for cocultures ofkeratinocytes with ®broblasts (Fig 6B).

In general, the level of total TGF-b1 released from latentTGF-b1 transduced cells was markedly higher than from AdTGF-b1223/225 transduced keratinocytes. As shown in Fig 7, however,without acidi®cation, the biologic activity of AdTGF-b1223/225

transduced keratinocytes was signi®cantly higher than both theLacZ control (p < 0.0001) and AdTGF-b1 (p < 0.0001).

To demonstrate whether TGF-b1 could affect the ®broblasts inthe lower chamber, total RNA was extracted and analyzed bynorthern analysis. The results show that ®broblasts cultured withlatent TGF-b1 transduced keratinocytes exhibited a signi®cantincrease in type I collagen mRNA (p < 0.05) (Fig 8A, B)compared to the LacZ control. The expression of collagenasemRNA was not signi®cantly altered in response to latent TGF-b1(Fig 9A, B).

To quantify the expression of collagenase and type I collagenmRNA, the intensity of signals was evaluated by densitometryusing 18S rRNA as a loading control. The level of mRNAbetween experiments was normalized using the LacZ control andresults for each experimental group were expressed as a percentageof the LacZ control. On the other hand, ®broblasts cocultured withkeratinocytes transduced with the active TGF-b1 construct hadsigni®cantly higher levels of type I collagen mRNA (p < 0.005)(Fig 8A, B) as well as signi®cantly lower levels of collagenasemRNA (p < 0.05) (Fig 9A, B) compared to the LacZ control.

Figure 3. TGF-b1 released from keratinocytes transduced withAdTGF-b1223/225 was constitutively active. Active TGF-b1 wasmeasured in conditioned medium, which had not been acid activated.Signi®cantly more TGF-b1 was released from keratinocytes transducedwith the active construct (AdTGF-b1223/225) than from keratinocytestransduced with the latent construct (AdTGF-b1) (*p < 0.0001). Thisdifference was observed both for keratinocytes cultured alone and forkeratinocytes cocultured with dermal ®broblasts (four separateexperiments).

Figure 4. Keratinocytes transduced with either AdTGF-b1 orAdTGF-b1223/225 express a smaller TGF-b1 mRNA transcriptcompared to endogenous TGF-b1 mRNA. Representative northernanalysis for TGF-b1 mRNA: keratinocytes cultured alone, lanes 1±4;keratinocyte mRNA following coculture with ®broblasts, lanes 5±8; and®broblast mRNA cultured alone and following coculture withkeratinocytes, lanes 9±13. 18S rRNA was used as a loading control. TheTGF-b1 mRNA resulting from the adenoviral constructs is smaller thannative mRNA. Two bands were observed in keratinocytes transducedwith AdTGF-b1 (lanes 3 and 7) and AdTGF-b1223/225 (lanes 4 and 8).Untreated keratinocytes (lanes 1 and 5), keratinocytes transduced withthe LacZ control (lanes 2 and 6), and ®broblasts under all conditions(lanes 9±13; F, F/K, F/K + LacZ, F/K + latent, F/K + active) showedonly one TGF-b1 mRNA band.

Figure 5. Total TGF-b in conditioned medium followingacidi®cation. (A) Keratinocytes cultured alone were either untreated(K), or transduced with LacZ, AdTGF-b1, or AdTGF-b1223/225. (B)Fibroblasts alone and cocultured with keratinocytes that were eitheruntreated (F) or transduced with LacZ, AdTGF-b1, or AdTGF-b1223/

225. Keratinocytes transduced with either AdTGF-b1 or AdTGF-b1223/

225 had signi®cantly higher levels of TGF-b1 compared to the LacZcontrol (*p < 0.0001). Keratinocytes transduced with AdTGF-b1had signi®cantly higher levels of total TGF-b1 compared to AdTGF-b1223/225 (**p < 0.0001) (four separate experiments).

VOL. 119, NO. 2 AUGUST 2002 LATENT AND ACTIVE TGF-b1 IN A COCULTURE SYSTEM 459

DISCUSSION

The purpose of this study was to establish a keratinocyte/®broblastcoculture system in which keratinocytes release either active orlatent TGF-b1. In this system, the effects of latent and activeTGF-b1 on underlying ®broblasts can be examined. Replicationde®cient adenoviruses were used to transduce human keratinocytes.Adenovirus was chosen because of its capacity to transduce bothreplicating and nonreplicating cells and the consequent highef®ciency of gene transfer. Unlike retroviruses, which may cause

insertional mutagenesis, adenoviral replication occurs episomally,i.e., outside the host cell nucleus. Gene therapy with adenovirusesmay be preferred to recombinant TGF-b1 protein therapy due tothe relative expense and short half-life of the protein (Bowen-Popeet al, 1984; Wake®eld et al, 1990). Keratinocytes turn over rapidlyand, if an undesirable outcome occurs in vivo, the epidermis can beremoved (Vogel, 1993). Also, adenoviruses do not integrate intothe host cell genome and are eventually lost through cellreplication. The transient nature of this system is preferred as thegene product will be expressed during wound healing and will beundetectable as wound closure is completed.

Numerous studies have used TGF-b1 in the context of woundhealing. TGF-b1 has been applied as a recombinant protein(Roberts et al, 1986; Puolakkainen et al, 1995), naked cDNA (Bennet al, 1996), in viral vectors (Sime et al, 1997; Ghahary et al, 1998),and expressed by transgenic mice (Sellheyer et al, 1993; Chui et al,1995; Chan et al, 2002). The system used for delivery of TGF-b1will have a signi®cant effect on the level and duration of proteinexpression. This in turn will determine the extent of the in¯uenceon formation of ECM during wound healing.

Figure 6. The biologic activity of total TGF-b1 was measuredusing the PAI/L assay following acidi®cation of conditionedmedium. (A) Keratinocytes cultured alone (K, transduced with LacZ,AdTGF-b1, or AdTGF-b1223/225). (B) Fibroblasts alone or coculturedwith keratinocytes. Keratinocytes transduced with both AdTGF-b1 andAdTGF-b1223/225 had signi®cantly higher levels of biologically activeTGF-b1 compared to the LacZ control (*p < 0.0001) (two separateexperiments).

Figure 7. The biologic activity of active TGF-b1 was measuredusing the PAI/L assay without acid activation. Keratinocytestransduced with AdTGF-b1223/225 and cultured alone and in coculturewith ®broblasts released signi®cantly higher levels of functional TGF-b1than the LacZ controls (* p < 0.0001). Keratinocytes transduced withAdTGF-b1 and cultured alone also had a signi®cantly higher level offunctional TGF-b1 than keratinocytes transduced with LacZ(**p < 0.001) (two separate experiments).

Figure 8. Expression of type I collagen mRNA by dermal®broblasts cocultured with genetically modi®ed keratinocytes. (A)Total RNA was extracted and probed for type I procollagen mRNA bynorthern analysis. The autoradiogram shown is representative of fourseparate experiments, demonstrating the pattern of 5.8 kb and 4.8 kbtranscripts (top) corresponding to pro-a(I) procollagen mRNA. Thebottom autoradiogram shows the pro®le of the 18S ribosomal RNA usedas a control for loading, obtained by rehybridization of the same blot.(B) Densitometry was performed and the relative intensity of pro-a(I)procollagen mRNA to 18S rRNA was calculated; the graph showscombined data from four experiments. Results for each treatment wereexpressed as a percentage of the LacZ control (lane 3). Lanes 1 and 2 inpanel A correspond to F and F/K. Fibroblasts exposed to AdTGF-b1(lane 4) and AdTGF-b1223/225 (lane 5) had signi®cantly higher levels ofpro-a(I) procollagen compared to the LacZ control (*p < 0.05 and**p < 0.005, respectively; n = 4).

460 BAUER ET AL THE JOURNAL OF INVESTIGATIVE DERMATOLOGY

The concentration of TGF-b1 required to stimulate collagen andelastin production in vitro has been established as 10 ng per ml(Davidson et al, 1993). It is expected that in vivo at least 10 times thisconcentration will be required to get a similar rise in ECM proteinsdue to the potential lack of bioavailability of TGF-b1 in thiscomplex system. This very high concentration of growth factorwould be very dif®cult to sustain over a period of days or perhapsweeks until the wound is healed. An adenoviral vector mightenable high levels of protein to be expressed, however, and theappropriate concentrations could be determined by adjusting theinfectious dose of the virus.

We have demonstrated that the adenoviruses are capable oftransducing the human keratinocytes without infecting the under-lying dermal ®broblasts. We were able to show that both the latentand active forms of TGF-b1 were produced and were biologicallyactive. The TGF-b1 released by keratinocytes was able to traversethe membrane insert in the coculture system and affect theunderlying ®broblasts. Interestingly, in addition to changes in®broblast collagen and collagenase mRNAs, we also notedautoinduction of TGF-b1 mRNA in ®broblasts cocultured withthe genetically modi®ed keratinocytes (Fig 4). Others have shownthat TGF-b1 can induce its own positive feedback loop (Van

Obberghen-Schilling et al, 1988; Kim et al, 1990; Lin et al, 1995;Seder et al, 1998).

The effects of normal keratinocytes on dermal ®broblasts havebeen reported. Keratinocytes and keratinocyte-conditioned mediahave been shown to increase ®broblast proliferation whiledecreasing collagen synthesis. It has been postulated that keratino-cytes may either decrease ®broblast production of collagen orincrease production of collagenase or another metalloproteinase, orthat keratinocytes may decrease the production of a proteinaseinhibitor (Garner, 1998).

TGF-b1 has been shown to increase collagen synthesis (Robertset al, 1986; Ghahary et al, 1998) by dermal ®broblasts in culture. It isnot clear, however, whether TGF-b1 can increase the synthesis ofECM proteins in the presence of other cytokines and growthfactors released by keratinocytes. Here we demonstrate thatculturing ®broblasts with keratinocytes reduces the level of type Icollagen mRNA compared to ®broblasts alone, and that the levelscan be restored by TGF-b1. We have also observed signi®cantchanges in collagenase mRNA with active TGF-b1. Fibroblastsalone express very little collagenase mRNA, but after coculturewith keratinocytes, collagenase mRNA is greatly upregulated.With active TGF-b1 treatment, levels are signi®cantly decreasedalthough not to the level seen in ®broblasts alone. This may be dueto the in¯uence of keratinocyte derived cytokines, which TGF-b1alone may not be able to abrogate. Edwards et al (1987) had shownthat TGF-b1 was capable of reducing the induction of collagenaseexpression by EGF or bFGF to control levels, where the effect wasmost pronounced in the presence of other growth factors asopposed to TGF-b1 alone.

The extent to which TGF-b1 is able to elicit an effect on thewound depends on its activation. Latent TGF-b1 cannot usuallybind to its receptor unless it is ®rst released from the LAP or aconformational change occurs. As many cell types possess TGF-b1receptors, activation of latent TGF-b1 is an important regulatorystep for the function of TGF-b1 in vivo. In vitro, it has beendemonstrated that the close proximity of two different cell typescan result in activation (Dennis and Rifkin, 1990). Latent TGF-b1has been shown to bind to M6P/IGF-II receptors on cell surfacesvia the phosphorylated glycosylation sites on the LAP and thisbinding is inhibited by M6P (Purchio et al, 1988). Ghahary et al(1999) demonstrated that the ability of different cell types toactivate latent TGF-b1 varies with the number of M6P/IGF-IIreceptors. By using 125I-labeled IGF-II, the M6P/IGF-II receptorson ®broblasts and mink lung epithelial cells were quanti®ed.Signi®cantly more were present on ®broblasts and only the®broblasts and not the Mu1Lu cells responded to latent TGF-b1.

In this study, we performed four separate groups of experiments,each with a different ®broblast cell strain. Each of the cell strainsshowed an increase in the expression of type I collagen mRNAcompared to the LacZ control in response to both active and latentTGF-b1. Although only a very small proportion of latent TGF-b1was activated in culture (Fig 3), this may have been suf®cient toelicit an increase in type I collagen mRNA. The differences in the®brogenic outcomes of these two growth factor states was apparentin a study of lung ®brosis using both active and latent TGF-b1(Sime et al, 1997). The transgene was introduced to the rat lungsusing adenoviral vectors; the TGF-b1 protein peaked at day 7 anddeclined rapidly by day 14. Overexpression of active TGF-b1 andnot latent TGF-b1 resulted in prolonged interstitial ®brosis. ActiveTGF-b1 caused the deposition of collagen, ®bronectin, and elastinas well as the differentiation of ®broblasts into myo®broblasts. Simeet al (1997) did not see an increase in hydroxyproline synthesis withthe latent construct, which may be due to the increased complexityof their in vivo system compared to our in vitro culture system andalso to the different cell types involved. The latent TGF-b1 thatwas activated in vivo may diffuse from the wound area, be degradedby proteinases or interact with TGF-b1 binding proteins such asa2-macroglobulin, and be cleared from the tissue.

Chan et al (2002) developed a transgenic mouse model bytargeting latent TGF-b1 to the epithelium using a K14 promoter.

Figure 9. Expression of collagenase mRNA by dermal ®broblastscocultured with genetically modi®ed keratinocytes. (A) TotalRNA was extracted and probed for collagenase mRNA by northernanalysis. The autoradiogram shown is representative of three separateexperiments, demonstrating the 2.1 kb transcript (top) corresponding tocollagenase mRNA. The bottom autoradiogram shows the pro®le of the18S ribosomal RNA used as a control for loading, obtained byrehybridization of the same blot. (B) Densitometry was performed andthe relative intensity of pro-a(I) procollagen mRNA to 18S rRNA wascalculated. The graph shows combined data from three experiments,with the results of each treatment expressed as a percentage of the LacZcontrol (lane 3). Lanes 1 and 2 correspond to F and F/K. Fibroblastsexposed to AdTGF-b1 (lane 4) did not have signi®cantly lower levels ofcollagenase mRNA compared to the control, whereas those exposed toAdTGF-b1223/225 (lane 5) did (p < 0.05*).

VOL. 119, NO. 2 AUGUST 2002 LATENT AND ACTIVE TGF-b1 IN A COCULTURE SYSTEM 461

The study demonstrated a signi®cantly higher level of type Icollagen mRNA in both homozygous and heterozygous animalsversus normal controls at day 9 following a full-thickness punchbiopsy wound. This difference coincided with signi®cantly higherlevels of total TGF-b1 as measured by the PAI/L assay. One of themost obvious differences in healing in the transgenic animals wasthe delay in reepithelialization compared to normal animals, whichmay be due to TGF-b1 directly inhibiting keratinocyte prolifer-ation (Shipley et al, 1986). Reepithelialization may also be affectedas a result of TGF-b1 diminishing keratinocyte migration bydecreasing the expression of collagenase. Keratinocytes rely uponthe release of metalloproteinases in order to break down ECM sothat they may migrate across granulation tissue and dissect escharfrom viable tissue (Pilcher et al, 1997).

In summary, the model we used allows for a direct comparison ofthe effects of latent and active TGF-b1 on dermal ®broblastscocultured with genetically modi®ed keratinocytes. As the woundenvironment contains elements required for the activation of latentTGF-b1, the use of latent TGF-b1 genetically modi®ed keratino-cytes in the form of either sheets of keratinocytes or a keratinocyte/®broblast skin substitute should be advantageous over the use ofactive TGF-b1. In this way, the activation of TGF-b1 is limited tothe wound area and the risk of development of ®brosis in otherorgans due to possible leakage of latent TGF-b1 is minimal.

The authors would like to thank Dr. Liju Yang for her assistance with the PAI/L

assay. This work was supported by the Canadian Institute of Health Research

(CIHR) (E. Tredget and A. Ghahary) and a CIHR studentship (B. Bauer).

REFERENCES

Abe M, Harpel JG, Metz CN, Nunes I, Loskutoff DJ, Rifkin DB: An assay fortransforming growth factor-b using cells transfected with a plasminogenactivator inhibitor-1 promoter±luciferase construct. Anal Biochem 216:276±284, 1994

Allen JB, Wong HL, Costa P, Simon GL, Bienkowski MJ, Wahl SM: Suppression ofmonocyte function and differential regulation of IL-1 and IL-4 contribute toresolution of experimental arthritis. J Immunol 151:4344±4351, 1993

Benn SI, Whitsitt JS, Broadley KN, et al: Particle-mediated gene transfer withtransforming growth factor-b1 cDNAs enhances wound repair in rat skin. JClin Invest 98:2894±2902, 1996

Bennett NT, Schultz GS: Growth factors and wound healing: part II. Role in normaland chronic wound healing. Am J Surg 166:74±81, 1993

Bowen-Pope DF, Malpass TW, Foster DM, Ross R: Platelet-derived growth factorin vivo: levels, activity, and rate of clearance. Blood 64:458±469, 1984

Brunner AM, Marquardt H, Malacko AR, Lioubin MN, Purchio AF: Site-directedmutagenesis of cysteine residues in the pro region of the transforming growthfactor b1 precursor: expression and characterization of mutant proteins. J BiolChem 264:13660±13664, 1989

Buckley A, Davidson JM, Kamerath CD, Wolt TB, Woodward SC: Sustained releaseof epidermal growth factor accelerates wound repair. Proc Natl Acad Sci USA82:7340±7344, 1985

Chan T, Ghahary A, Demare J, Yang L, Iwashina T, Scott PG, Tredget EE:Development, characterization and wound healing of the keratin-14 promotedtransforming growth factor-b1 transgenic mice, Wound Repair Regeneration10:177±187, 2002

Chirgwin JM, Przybyla AE, MacDonald RJ, Rutter WJ: Isolation of biologicallyactive ribonucleic acid from sources enriched in ribonucleases. Biochemistry18:5294±5299, 1979

Chui W, Fowlis DJ, Cousins FM, Duf®e E, Bryson S, Balmain A, Akhurst RJ:Concerted action of TGF-b1 and its type II receptor in control of epidermalhomeostasis in transgenic mice. Gene Dev 9:945±955, 1995

Dallas SL, Park-Snyder S, Miyazono K, Twardzik D, Mundy GR, Bonewald LF.Characterization and autoregulation of latent transforming growth factor beta(TGF-beta) complexes in osteoblast-like cell line: production of a latentcomplex lacking the latent TGF-b binding protein. J Biol Chem 269:6815±6821, 1994

Danielpour D, Dart LL, Flanders KC, Roberts AB, Sporn MB: Immunodetection ofquantitation of the two forms of transforming growth factor-beta (TGF-b1 andTGF-b2) secreted by cells. J Cell Physiol 138:79±86, 1989

Davidson JM, Zoia O, Liu J-M: Modulation of transforming growth factor-b1stimulated elastin and collagen production and proliferation in porcine vascularsmooth muscle cells and skin ®broblasts by basic ®broblast growth factor-a, andinsulin-like growth factor-1. J Cell Phys 155:149±156, 1993

Dennis PA, Rifkin DB: Cellular activation of latent transforming growth factor brequires binding to the cation-independent mannose-6-phosphate/insulin-likegrowth factor type II receptor. Proc Natl Acad Sci USA 88:580±584, 1990

Edwards DR, Murphy G, Reynolds JJ, Whitman SE, Docherty AJP, Angel P, HeathJK: Transforming growth factor-beta stimulates the expression of collagenaseand metalloproteinase inhibitor. EMBO J 6:1899±1904, 1987

Garner WL: Epidermal regulation of dermal ®broblast activity. Plast Reconstr Surg102:135±139, 1998

Ghahary A, Shen YJ, Scott PG, Tredget EE: Expression of mRNA for transforminggrowth factor-b1 is reduced in hypertrophic scar and normal dermal ®broblastsfollowing serial passage in vitro. J Invest Dermatol 103:684±686, 1994

Ghahary A, Shen YJ, Nedelec B, Scott PG, Tredget EE: Interferon gamma andalpha-2b differentially regulate the expression of collagenase and TIMP-1mRNA in human hypertrophic and normal dermal ®broblasts. Wound Rep Reg3:13±21, 1995

Ghahary A, Tredget EE, Chang LJ, Scott PG, Shen Q: Genetically modi®ed dermalkeratinocytes express high levels of transforming growth factor-b1. J InvestDermatol 110:800±805, 1998

Ghahary A, Tredget EE, Shen Q: Insulin-like growth factor-II/mannose 6 phosphatereceptors facilitate the matrix effects of latent transforming growth factor-b1released from genetically modi®ed keratinocytes in a ®broblast/keratinocyteco-culture system. J Cell Physiol 180:61±70, 1999

Gleizes PE, Mungler JS, Nunes I, Harpel JG, Mazzieri R, Noguera I, Rifkin DB:TGF-b latency: biological signi®cance and mechanism of activation. Stem Cells15:190±197, 1997

Ignotz RA, Massague J: Transforming growth factor-b1 stimulates the expression of®bronectin and collagen and their incorporation into the extracellular matrix. JBiol Chem 261:4337±4345, 1986

Kim SJ, Angel P, Lafyatis R, et al: Autoinduction of transforming growth factor b1 ismediated by the AP-1 complex. Mol Cell Biol 10:1492±1497, 1990

Lin RY, Sullivan KM, Argenta PA, Meuli M, Lornez HP, Adzick NS: Exogenoustransforming growth factor-beta ampli®es its own expression and induces scarformation in a model of human fetal skin repair. Ann Surg 222:146±154, 1995

Lynch SE, Colvin RB, Antoniades HN: Growth factors in wound healing. J ClinInvest 84:640±646, 1989

Lyons RM, Oja JK, Moses HL: Proteolytic activation of latent transforming growthfactor-b from ®broblast conditioned medium. J Cell Biol 106:1659±1665, 1988

Lyons RM, Gentry LE, Purchio AF, Moses HL: Mechanism of activation of latentrecombinant transforming growth factor-b1 by plasmin. J Cell Biol 110:1361±1367, 1990

Okazaki R, Durham SK, Riggs BL, Conver CA: Transforming growth factor andforskolin increase all classes of insulin-like growth factor-1 transcripts in normaland human osteoblast-like cells. Biochem Biophys Res Commun 207:963±970, 1995

Overall C, Wrana JL, Sodeck J: Independent regulation of collagenase, 72 kDaprogelatinase, and metalloendoproteinase inhibitor expression in human®broblasts by transforming growth factor-b. J Biol Chem 264:1860±1869, 1989

Pilcher BK, Dumin JA, Sudbeck BD, Krane SM, Welgus HG, Parks WC: Theactivity of collagenase-1 is required for keratinocyte migration of a type Icollagen matrix. J Cell Biol 137:1445±1457, 1997

Postlethwaite AE, Keski-Oja J, Moses HL, Kang AH: Stimulation of the chemotacticmigration of human ®broblasts by transforming growth factor beta. J Exp Med165:251±256, 1987

Puolakkainen PA, Twardzik DR, Ranchalis JE, Pankey SC, Reed MJ, GombotzWR: The enhancement in wound healing by transforming growth factor-b1depends on the topical delivery system. J Surg Res 58:321±329, 1995

Purchio AF, Cooper JA, Brunner AM, et al: Identi®cation of mannose 6-phosphatein 2 asparagine-linked sugar chains of recombinant transforming growth factor-b-1 precursor. J Biol Chem 263:14211±14215, 1988

Rheinwald JG, Green H: Serial cultivation of strains of human epidermalkeratinocytes: the formation of keratinizing colonies from single cells. Cell6:331±343, 1975

Roberts AB, Sporn MB, Assoian RK, et al: Transforming growth factor type b: rapidinduction of ®brosis and angiogenesis in vivo and stimulation of collagenformation in vitro. Proc Natl Acad Sci USA 83:4167±4171, 1986

Samuel SK, Hurta RAR, Kondaiah P, Khalil N, Turley EA, Wright J, GreenbergAH: Autocrine induction of tumor protease production and invasion by ametallothionein-regulated TGF-b1. EMBO J 11:1599±1605, 1992

Sato Y, Rifkin DB: Inhibition of endothelial cell movement by pericytes and smoothmuscle cells: activation of a latent transforming growth factor-1-like moleculeby plasmin. J Cell Biol 109:309±315, 1989

Sato Y, Tsuboi R, Lyons RM, Moses HL, Rifkin DB: Characterization of theactivation of latent TGF-b by co-cultures of endothelial cells and pericytes orsmooth muscle cells: a self-regulating system. J Cell Biol 111:757±763, 1990

Seder RA, Marth T, Sieve MC, Strober W, Letterio JJ, Roberts AB, Kelsall B:Factors involved in the differetiation of TGF-b-producing cells from niaveCD4+ T cells: IL-4 and IFN-g have opposing effects, while TGF-b positivelyregulates its own production. J Immunol 160:5719±5728, 1998

Sellheyer K, Bickenbach JR, Rothnagel JA, et al: Inhibition of skin development byoverexpression of transforming growth factor b1 in the epidermis of transgenicmice. Proc Natl Acad Sci USA 90:5237±5241, 1993

Shipley GD, Pittelkow MR, Wille JJ, Scott RE, Moses HL: Reversible inhibition ofnormal human prokeratinocyte proliferation by type beta transforming growthfactor±growth inhibitor in serum-free medium. Cancer Res 46:2068±2071, 1986

Sime PJ, Xing Z, Graham FL, Csaky KG, Gauldie J: Adenovector-mediated genetransfer of active transforming growth factor-b1 induces prolonged severe®brosis in rat lung. J Clin Invest 100:768±776, 1997

Sporn MB, Roberts AB, Wake®eld LM, de Crombrugghe B: Some recent advancesin the chemistry and biology of transforming growth factor-b. J Cell Biol105:1039±1045, 1987

Tsuji T, Okada F, Yamaguchi K, Nakamura T: Molecular cloning of the large

462 BAUER ET AL THE JOURNAL OF INVESTIGATIVE DERMATOLOGY

subunit of transforming growth factor type b masking protein andexpression of the mRNA in various rat tissues. Proc Natl Acad Sci USA87:8835±8839, 1990

Van Obberghen-Schilling E, Roche NS, Flanders KC, Sporn MB, Roberts AB:Transforming growth factor b1 positively regulates its own expression innormal and transformed cells. J Biol Chem 263:7741±7746, 1988

Vogel JC: Keratinocyte gene therapy. Arch Dermatol 129:1478±1483, 1993

Wake®eld LM, Smith DM, Masui T, Harris CC, Sporn MB: Distribution andmodulation of the cellular receptor for transforming growth factor-beta. J CellBiol 105:964±975, 1987

Wake®eld LM, Winoker TS, Hollands RS, Christopherson K, Levinson AD,Sporn MB: Recombinant latent TGF-b1 has a longer half life in rats thanactive TGF-b1 and a different tissue distribution. J Clin Invest 86:1976±1984, 1990

VOL. 119, NO. 2 AUGUST 2002 LATENT AND ACTIVE TGF-b1 IN A COCULTURE SYSTEM 463