American Journal of Pathology, Vol. 138, No. 1, January 1991 Copyright C American Association ofPathologists

Vascular Cells Respond Differentially to Transforming Growth Factors Beta1 and Beta2 In Vitro

June Rae Merwin,* Walter Newman,t L. Dawson BeaIl,t Adeline Tucker,* and Joseph Madri* From the Department of Pathology, Yale University, New Haven, Connecticut*; and Otsuka Pharmaceutical

Company, Rockville, Marylandt

Transforming growth factors 4,1 (TGF-p,) and 42 (TGF-132) are equipotent in many cell systems studies thus far. Recent data, however, show different effects elicited by these two growth factors in specific biologic systems. This investigation compares the effects of TGF-#3, and TGF-32 bovine aortic endothelial cells (BAECs), rat epididymal fat pad microvascular endothelium (RFCs), and bovine aortic smooth muscle cells (BASMCs). In two-dimensional cultures, proliferation of BAECs, BASMCs, and RFCs were all inhibited by TGF-f,, while in response to TGF-,42, BASMCs were fully inhibited, RFCs were modestly inhibited, and BAECs were unaffected. Bovine aortic endothelial cell migration was significantly inhibited by TGF-#,,, but only slightly inhibited by TGF-432. In contrast, BASMC migration was enhanced by TGF-3, and was not affected by TGF-f32. In three-dimensional cultures, RFCs were stimulated to undergo in vitro angiogenesis in response to TGF-f,, and TGF-p2 at 10-fold higher concentrations. Three distinct receptor assays demonstrated the presence of type I and type II TGF-fp, cell-surface-binding proteins on BAECs, BASMCs, and RFCs. Labeled TGF-43, was competed off completely with 100-fold molar excess unlabeled TGF,B,, but only partially with equivalent excess unlabeled TGF-/32. Furthermore the ratios of type I to type II TGF-f3 receptors in these three vascular cell types vary from 1:1 in BAECs to 1.5:1 in RFCs to 3:1 in BASMCs and can be correlated with the differences noted in cellular responses to TGF-fp, and TGF-,12 in proliferation, migration, and in vitro angiogenic assays. Thesefindings support the hypothesis that there are different responses to the TGF-I3s, depending on the cell type and experimental con-

ditions as well as the TGF-( concentration and isoform used. (Am JPathol 1991, 138:37-51)

The transforming growth factors beta (TGF-43) are polypeptides that act hormonally to regulate differentiation and proliferation of a variety of cell types, depending on the microenvironment.1' 2 Presently, there are six known TGF, homologs: TGF-431, TGF-432, TGF-,1.2, TGF-433, TGF-#4, and TGF-435. TGF-434 and TGF-435, the most recently discovered isoforms of the TGF-,B family, have been cloned,3 and preliminary investigations are beginning. TGF-433 is of mesoderm origin and has been identified by cDNA characterization.4 TGF-431.2 results from a heterodimeric combination of a subunit of TGF-431 and TGF-432.5 TGF-431, the polypeptide originally described as TGF-43,6 exhibits a 72% amino acid N-terminal identity with TGF-f2.7 A wide variety of cell types respond to the TGF-43s. In comparative studies, TGF-f31 and TGF-#2 appear to have indistinguishable activities in most assays.1 TGF-431 and TGF-132 are equipotent at stimulating growth of mesenchymal cells in soft agar,8 inhibiting normal and tumorderived epithelial cell proliferation9 and altering differentiation.10 Recent experimentation, however, suggests that TGF-431 and TGF-432 may not be interchangeable in vivo, with their functions differing spatially and temporally.11 Such studies show that TGF-431, but not TGF-432, is an inhibitor of hematopoietic progenitor cell proliferation;'2 TGF-432, but not TGF-(31, is active in mesoderm induction, as shown by increased muscle-specific a-actin mRNA13; a variety of cultured cells secrete predominantly TGF-431 or TGF-#2, while others secrete both peptides in nearly equal amounts14; and that both TGF-#s have differing abilities to interact with a family of TGF-,B receptors.15-17 Preliminary studies by Jennings et al18 suggest that TGF-431 is more effective than TGF-#2 in inhibiting DNA synthesis Supported in part by Joshua Macy Pre-doctoral Fellowship to June R. Merwin and USPHS Grants ROI-HL-28373 and POI DK 38979 to Joseph A. Madri. Accepted for publication August 22, 1990. Address reprint requests to June Rae Merwin, Pathology Department, Li 11, Yale University School of Medicine, 310 Cedar St., New Haven, CT

06510.

37

38

Merwin et al

AJPJanuary 1991, Vol. 138, No. I

of large vessel endothelial cells. More studies are needed, however, to sort out the complex responses of vascular cells to the TGF-1s. Endothelial cells of the various vascular beds exhibit a range of diverse functions and appearances along with their shared features of nonthrombogenicity, polarity, metabolic, and transport functions.19-' Neovascularization is a common response, yet it varies depending on whether the endothelium is derived from large vessels or the microvasculature. Large-vessel endothelial cells form a flat sheet lining the vessel and respond to injury by sheet migration and proliferation; whereas microvascular endothelial cells express a significant arc of curvature necessary for the lumen formation by a single endothelial cell and respond to injury by migration into a three-dimensional matrix followed by the tube formation.21 Beneath the luminal endothelium, smooth muscle cells and their surrounding extracellular matrix (ECM) are major structural components of the vascular wall. Components of this pericellular matrix are known to modulate smooth muscle and endothelial cell proliferation that is normally seen after vessel injury.2325 In addition, smooth muscle cells and the overlying endothelia influence each other' platelet-derived growth factor (PDGF)271 and TGF-31,,9 as well as by the ECM components they each synthesize. Thus after tissue damage, TGF-,31, a factor that has been shown to modulate both endothelial and smooth muscle cell behavior, is likely to be derived from these vascular cell populations as well as from platelets.231 Vascular cell production, secretion, and activation of TGF- are likely to be important mechanisms involved in autocrine and paracrine modulation of specific vascular cell populations in such diverse processes as angiogenesis and vasculogenesis. This, coupled with the possibility of multiple TGF-,3 isoforms being produced by a variety of cell types during different times in vasculogenesis or angiogenesis, raises the likelihood of this family of growth factors being of major importance in modulating vascular cell populations. Our rationale for this study involves the necessity to unravel the complex reactions that are associated with vascular cell responses to selected TGF-,B isoforms. By combining in vitro data showing differing effects elicited by TGF-i31 and TGF-,32 in various vascular cell types, we addressed questions concerning the roles of these growth factors and vascular cells in processes such as proliferation, migration, and angiogenesis. Because the dynamic vessel response after injury is not always advantageous to the host, these investigations have also begun to elucidate possible roles of TGF-,B isoforms in modulating the beneficial reactions such as re-endothelialization after denudation injury versus detrimental responses such as vascular smooth muscle migration and proliferation subsequent to such injuries. These studies also further char-

acterize roles of the TGF-p isoforms in the process of angiogenesis.

Materials and Methods

Cell Cultures Smooth muscle cells were grown from explants of bovine aortic media in complete Dulbecco's modified Eagle's medium (DME) supplemented with 10% fetal calf serum (FCS) as described.',' After several days in culture, bovine aortic smooth muscle cells (BASMCs) migrate out from the medial explants. After the development of confluency, the cells were trypsinized, passaged, and used for various studies. These a-smooth muscle actin-positive cells formed the hill-and-valley pattern typical of cultured smooth muscle cells and were used between passage 2 and 5. Capillary endothelial cells were isolated and cultured from rat epididymal fat pads as described by Madri and Williams.' Bovine calf aortic endothelial cells were isolated and cultured according to Madri et al.32 Three-dimensional rat epididymal fat pad microvascular endothelium (RFC) cultures were composed of acid-soluble calf dermis collagen type 15 and prepared using the method of Madri et al.21 To address the possibility of differences due to the use of two species; proliferative, migratory, and angiogenic studies were performed using bovine adrenal cortex microvascular endothelial cells, with the results being equivalent to those obtained with RFCs (Merwin and Madri, unpublished data).

Growth Factors TGF-,31 prepared as described by Assoian et al,6 iodinated TGF-#,1 as described by Frolik et al,36 and TGF-i32 as described by Cheifetz et al15 were the gifts of Drs. A. Roberts and M. Sporn, Laboratory of Chemoprevention, National Cancer Institute, National Institutes of Health, Bethesda, Maryland. Epidermal growth factor (EGF) was purchased from Collaborative Research Inc., Bedford, Massachusetts, and basic fibroblast growth factor (bFGF) was the gift of Drs. M. Klagsbrun and J. Folkman. Biotinylated TGF-f3, was prepared as previously described, repurified to remove any nonbiotinylated TGF-f31, and shown to be fully active in three bioassays.37 All factors were added at concentrations determined to be optimum experimentally as previously21 shown or using morphologic assays of angiogenesis.

Matrix Components and Coating Protocol Bovine type and human type Ill collagens were isolated and purified as described.35 839Tissue culture dishes and

Vascular Cell Responses to TGF-41 and 2

39

AJPJanuary 1991, Vol. 138, No. 1

flasks were coated with collagen type at a concentration of 12.5 ,ug/ml as described.'

scope and attached Olympus C35 AD camera (Olympus Corporation of America, New Hyde Park, NY). Photographs were taken on ILFORD XPI 400 film (Ilford Limited, Mobberley, Cheshire, England).

Proliferation Assays Collagen type I-coated dishes were washed in phosphatebuffered saline (PBS) before the addition of cell suspensions (1.4 X 104 cells/dish). After 6 hours, samples were counted to determine starting cell numbers on substrate. At this time, fresh medium ± factors (TGF-f1, TGF-42, EGF, bFGF) were added to the cultures. The medium and factors were replaced once again on day 3. Cell numbers were determined by lifting the cells off the culture dishes with trypsin/ethylenediaminetetra-acetic acid (EDTA) and counting quadruplicate samples using a Coulter counter (Coulter Electronics, Inc., Hialeah, FL). The mean number of cells per dish for each factor addition was then calculated. Alternatively for three-dimensional cultures, DNA was quantitated using DAPI (4'6"-diaminidno-2-phenylindole: Hoechst No. 33258) in a fluorescence assay as described.41 Total micrograms of DNA were correlated with cell number as previously described.21

Migration Assay Stimulus for cell migration was accomplished by release from contact inhibition of confluent cultures using a stainless steel fence device as previously described.42 After confluency was achieved in the center wells of the fences (- 6 hours), migration was induced by removal of the fences and cells were observed to migrate outward in a radial fashion. Cultures were fed once again on day 3 ± growth factors. After 6 days, the cultures were washed with PBS and fixed with 10% neutral buffered formalin. Dishes used in assessing migration were stained with Harris hematoxylin. Net increase in surface area covered was assessed using an overhead projector, and areas calculated using a computerized graphics tablet and a MacMeasure program (Yale Shareware, Yale University, New Haven, CT).

Receptor-binding Assay for Electron Microscopy Millicell-HA chambers, (Millipore Products Division, Bedford, MA) placed in a Costar 24-well cluster tissue culture dish (Costar Corporation, Cambridge, MA), were coated overnight at 40C with 12.5 jig/ml collagen type 1. The chambers were rinsed twice with PBS and once with the medium of choice. Cells were seeded onto the filters, allowed to adhere for -4 hours, and excess cells rinsed off with PBS before new medium was added with equal levels inside and outside the millicell. At confluency, the cells were washed three times for 5 minutes at room temperature (RT) with PBS containing magnesium and calcium, incubated 1 hour at RT with binding buffer (DME, 25 mmol/l [millimolar] Hepes, 0. 1% bovine serum albumin [BSA], pH 7.4) and washed as above. The cultures were incubated for varying times (1, 5, 15, 30, and 60 minutes) and temperatures (RT and 4 C) with 0.5 ng/ml biotinylated TGF-#1 (B-TGF-f31) washed rapidly with PBS containing magnesium and calcium, and immediately placed in periodate-lysine-paraformaldehyde fixative' for 1 hour at RT. The monolayers were incubated with streptavidin 5- or 10-nm gold-conjugated gamma G immunoglobulin (IgG; Janssen Life Science Prod., Piscataway, NJ), diluted 1: 50 in binding buffer for 1 hour at RT, with washings before and after. Additional conditions included streptavidin gold alone for background and competition with 1 00-fold molar excess of nonbiotinylated TGF-,31. The samples were treated with formaldehyde-glutaraldehyde fixative for 2 hours at RT and placed in 10% sodium cacodylate holding buffer until processed and analyzed.44

Electron Microscopy

Tube Formation Assay

Samples were posffixed in 1 % osmium tetraoxide buffered in 0.2 mol/l (molar) s-collidine for 1 hour at 4 C. After rinsing three times for 10 minutes with 0.1 mol/I s-collidine,

Three-dimensional cultures were incubated ± growth factors at varying concentrations for 4 days. The cultures were rinsed three times with PBS, removed from the millicell, and quickly frozen in Optimal Cufting Temperature (OCT) embedding compound (Miles Scientific Co., Kankakee, IL). Eight-micron cryosections were placed on potassium dichromate-coated glass slides, acetone fixed for 1 minute at -20 C, air dried, and viewed under Hoffman Interference Light Microscopy with an Olympus IM micro-

the samples were stained with uranyl acetate/oxalic acid, pH 7.4, for 1 hour at 4°C and rinsed as above. The millicell filters were cut into 2-mm pieces and dehydrated in graded concentrations of 70% to 100% ethyl alcohol (ETOH), followed by 100% propylene oxide before embedding in Epon-812 (E. Fullam, Inc., Latham, NY). Ultrathin sectioning was done on an LKB 111 8800 Ultramicrotome (LKB Produkter AB, Bromma, Sweden) and viewed on a Zeiss EM 10B Electron Microscope (Carl Zeiss, Inc., Oberkochen,

40

Merwin et al

AJPJanuary 1991, Vol. 138, No. 1

FRG). Photographs were taken on Kodak (Eastman Kodak, Rochester, NY) film.

Ligand Blots The procedure according to Towbin et al45 was modified. Samples were subjected to 10% sodium dodecyl sulfatepolyacrylamide gel electrophoresis (SDS-PAGE)4 and transferred to nitrocellulose as described by Basson et al.47

Receptor Binding Autoradiography Assay The assay was a modification of Segarini et al."7 Briefly, confluent monolayers grown on 35-mm, 1.5% gelatincoated tissue culture dishes were rinsed three times with PBS containing magnesium and calcium and then incubated with 0.5 ml 60 pmol 1251-TGF-f3, (3.19 ,C/pmol) in binding buffer (DME, 25 mmol/l Hepes, pH 7.4; 0.1% BSA) for 3 hours at 4°C on a rotating plafform. For competition assays, 1 00-fold molar excess unlabeled TGF-3,1 or TGF2 was included in the medium. The unbound material was removed and 0.5 ml fresh binding buffer added containing 0.3 mmol/l dissuccinimidyl suberate (DSS; Pierce Chemical Company, Rockford, IL). A 1 OOX stock solution of DSS was dissolved in 100 ,ul dimethyl sulfoxide (DMSO) just before incubation. After 15 minutes at 4°C, the medium was removed, monolayers were washed three times with 250 mmol/l sucrose, 10 mmol/l TRIS pH 7.4, 1 mmol/ EDTA, and solubilized with 200 ,ul 1% Triton X-1 00 containing 1 mmol/l phenyl methyl sulfonyl fluoride (PMSF). Laemmli loading buffer (70 ul of a 4X solution) and 30 ,u 1 mmol/l dithiothreitol (DTT) were added to each extract, boiled 3 minutes, and electrophoresed according to Laemmli.46 Linear gradient resolving gels of 5% to 10% polyacrylamide were constructed with a 3.5% stacking gel. The gels were dried and exposed to Kodak Diagnostic X-Omat-AR film in Dupont Cronex screens (Dupont Corp., Wilmington, DE) at -700C. The films were developed using a Kodak X-Omat M20 Processor. Densitometric scanning was performed on a Hoefer GS 300 Transmittance/Reflectance Scanning Densitometer using the GS350 H Data System (Hoefer Scientific Instruments, San Francisco, CA).

Flow Cytometry Flow cytometry was performed on a Coulter model 541 equipped with a quartz flow cell and an argon laser. Data analysis was performed on an EPICS Easy 88 computer workstation. Detection of TGF-,13 receptors on bovine aortic endothelial cells (BAECs) was performed after de-

tachment of cells by mild trypsinization and incubation for 4 hours at 37°C in DME, 1% BSA (radioimmunoassay [RIA] grade, Sigma Chemical Co., St. Louis, MO). This incubation period allowed for re-expression of the trypsinsensitive TGF-j31 receptor.37 Rat epididymal fat pad microvascular endothelium and BASMCs were detached by brief treatment with PBS containing 2 mmol/l EDTA. To prepare cells for flow cytometry, 50-gl aliquots of cells at 1.5 X 1 06/ml in Dulbecco's phosphate-buffered saline (DPBS), 1% BSA, and 0.02% azide (wash buffer) were added to wells of a round bottom 96-well plate that contained 10 ng B-TGF-#13, in a total volume of 100 ,ul, or in wells containing B-TGF-,13 plus a 100-fold molar excess of nonbiotinylated TGF-,13 or TGF-,32. After 45 minutes' incubation at 40C, cells were washed three times and resuspended in 100 gl wash buffer containing 2 Ag/ml fluorescein isothiacyanate (FITC)-avidin (Vector Labs, Burlingame, CA). After an additional 30 minutes' incubation, cells were washed three times and then analyzed directly. The background level of fluorescence was determined by incubation with FITC-avidin alone. The percent inhibition of binding of B-TGF-f31 by TGF-#13 or TGF-132 was calculated by setting the maximum binding in the absence of competitor at 100%. This represents the area under the histogram of B-TGF-f31 that is greater than the lowest 95% fluorescence intensity of the FITC-avidin-only histogram. The percent inhibition is the proportion of this fluorescence intensity that is competable by nonbiotinylated ligand.

Results

Transforming Growth Factor-beta Cellsurface-Binding Proteins Located on Vascular Cells To investigate the presence of cell-surface-binding proteins for the TGF-,Bs, flow cytometry was done using FITCavidin and biotinylated TGF-j31 (B-TGF-f31). This resulted in the identification of a cell-surface-binding protein for the TGF-,13 isoform on all three cell types studies. Representative analyses of each of the vascular cell types studied are illustrated in Figure 1. Saturating concentrations of B-TGF-f31 showed that the highest level of receptors was on RFCs, the lowest on BAECs; BASMCs had intermediate levels. The majority of the binding of B-TGF-,13 to BAECs, BASMCs, and RFCs is competable with 100fold molar excess TGF-,31 (60%, 56%, and 83%, respectively), but only a modest inhibition was noted in the presence of the same molar excess of TGF-f32 (6%, 17%, and 25%, respectively), suggesting some difference in the binding characteristics of TGF-f13 and TGF-132 to the cell surface, consistent with previous reports.12,153748 Although previous studies2547 have shown that after trypsinization

BAECs

A 175 150 CC LL

w

125

m

100

-J

75

D z

w LU 0

50 25 0

CHANNEL NUMBEIR

BASMCs

B 175 Figure 1. Flow-cytometric studies of TGF-g,1 cell-surface binding proteins on vascular cells: Cell suspensions ofBAECs (A), BASMCs (B) and RFCs (C) were treated with 10 ng/ ml B-TGF-3,1 as described in Materials and Methods. Background fluorescence with FITC-avidin alone (1); B-TGF-,3, + FITCavidin (2); B-TGF-f3, + 100-fold molar excess TGF-(,3 + FITC-avidin (3); B-TGF-f3, + 100-fold molar excess TGF-,2 + FITC-avidin (4). Fluorescence intensity, shown in arbitrary units, is displayed in log scale such that every 25 channels represents an approximate doubling of fluorescence. The cell number is shown in arbitrary units on the ordinal. Each histogram represents a cell-by-cell analysis of 5000 cells.

150 w m

2

100

.j

75

D Z W 0

25

0

CHANNEL NUMBE.R

C

RFCs 17'

154 cr w

12!

m

104

z

-X -

w 0

7!5 6 2 100

125

150

CHANNEL NUMBER

175

200

42

Merwin et al

AJPJanuary 1991, Vol. 138, No. I

Figure 2. EM studies of TGF-3,1 cell-surface bindingproteins on BAEC cultures: A: BAEC confluent monolayers treated for 15 minutes with 05 ng/ml B-TGF-f, and streptavidin 10 nm gold conjugates (1:50) showed positive cell surface binding. B: Treatment with 100-fold molar excess TGF-f,% competed off 100% ofthe binding. C: Nonspecific binding after treatment with streptavidin gold IgG alone was essentially nonexistent. Bar:A = 0.1 It;B+ C= 1.0g.

various endothelial cell-surface-binding proteins are maximally re-expressed in less than 2 hours, the authors caution that the results may, in fact, be an underestimate of the true receptor number. Having established the presence of TGF-f31 binding proteins on the cell surfaces by fluorescence-activated cell-sorting (FACS) analysis, we embarked on localization

studies using the B-TGF-f3, probe. The results illustrated in Figure 2 showed specific, competable binding at room temperature (RT) for 5- and 15-minute incubations, with no apparent difference between confluent and subconfluent monolayer cultures. Figure 2A illustrates modest BAEC cell-surface binding of B-TGF-f1 at 0.5 ng/ml. There was complete competition with 1 00-fold molar excess un-

Vascular Cell Responses to TGF-,1 and (2

43

AJPJanuary 1991, Vol. 138, No. 1

A 1

B 3

2

4

C 6

5

7

9

8

mw

kD

220 -212

-

-

170 116

-97 -

-

85kD -* 7OkD -

-

76

-53

-43 a

I

I

Figure 3. Autoradiographic studies sbowing iodinated TGF-f31 cell-surface binding: Cell monolayers were incubated with 6OpM '25I TGF-P,j and DSS cross-linked. The Triton X-100 cell lysate was run on reduced SDS-PAGE. A: BAEC, B: BASMC, C: RFC. The resultant autoradiographs revealed Type I (- 70 kd) and Type II (- 85 kd) receptors on all three cell types (Lanes 1, 4, 7). '251-TGF-P, binding was competed off with unlabeled TGF-3,1 (Lanes 2, 5, 8), but only partial competition was seen after incubation with unlabeled TGF-fi2 (Lanes 3, 6, 9). MW markers included: 220 kdfibronectin, 212 kd myosin, 170 kd a2-macroglobulin, 116 kdfi-galactosidase, 97 kdphosphorylase b, 76 kd transferrin, 53 kd glutamic debydrogenase, and 43 kd ovalbumin.

derivatized TGF-i31 (Figure 2B). Rat epididymal fat pad microvascular endothelium and BASMC cultures incubated with B-TGF-,31 under the same conditions gave similar results (data not shown). Electron microscopic (EM) analyses demonstrated no apparent binding on cells incubated with the B-TGF-j31 probe for 30- or 60-minute periods at RT. In each condition, approximately 10 mm of membrane from cell monolayers was scanned on the electron microscope, disclosing modest but consistent cell-surface binding of the biotinylated probe/avidin gold complex, while background using streptavidin gold alone was essentially nonexistent (Figure 2C). Similar results were obtained when cells were incubated with B-TGF-fl, for either 1 or 3 hours at 4°C. To further analyze vascular cell-surface binding of TGFf31, ligand blotting was performed. Lysates of cell monolayers were run on 5% to 10% linear gradient SDS-PAGE and blotted onto nitrocellulose sheets. The blots were then incubated with B-TGF-#1 followed by an avidin-alkaline phosphatase conjugate and a chromogenic substrate. A doublet was seen at -65 kd (data not shown). This method was not thought to be as sensitive as previously described methods utilizing iodinated probes, however; therefore, we used iodinated TGF-f31 (1251-TGF-f31), which had been used successfully by others to elucidate TGFp1-binding proteins. 15,17.49 Our autoradiographs showed specifically labeled, competable bands of -70 kd and -85 kd, consistent with TGF-,.1 type and type 11 receptors on all three cell types studied (Figure 3, lanes 1, 4, 7).

There was complete competition with 100-fold molar exunlabeled TGF-f31 (Figure 3, lanes 2, 5, 8); however 1 00-fold molar excess TGF-,32 only partially competed off the iodinated TGF-,31 in all three cell types (Figure 3, lanes 3, 6, 9). Similar results were observed at both a 15-minute RT incubation and 1 hour at 40C. Densitometric scanning (n = 5 for RFCs, n = 6 for BAECs and BASMCs) of the autoradiographs showed type I to type 11 receptor ratios that varied with the cell type; BAECs expressed 1:1; RFCs 1.5:1 and BASMCs 3:1 ratios. In addition to type I and type 11 receptor proteins, all three cell types expressed trace amounts of an apparent type IlIl receptor band at 280 kd and a 130- to 1 40-kd band that was observed on longer exposure times (data not shown). Figure 4 shows cross-linking analysis of 125I-TGF-fl1 binding to a subclone of TGF-#1-insensitive keratinocytes derived from chemically induced mouse papilloma (pI17) as a negative control (Figure 4, lane 1) and to the TGF-#,1-sensitive parent line5" as a positive control (Figure 4, lane 3). cess

Large-vessel Proliferative Responses to TGF-1, and TGF-f2 When BAECs were treated with TGF-f31, there was a dosedependent inhibition of proliferation with the optimum response of 70% at 0.5 ng/ml. However, when TGF-#2 was added to subconfluent BAEC monolayers, there was no significant inhibition at any concentration used, 0.05, 0.5 or 5.0 ng/ml (Figure 5A). To investigate a possible syn-

44 Merwin et al AJPJanuary 1991, Vol. 138, No. 1

B

A 1

2

3

4

-85!kD 7OkD

-bi

and only elicited a 20% inhibition at 5.0 ng/ml. As noted recently by Bell and Madri,51 TGF-,B1 elicited a 50% increase in migration of BASMCs in a concentration-dependent manner (Figure 7B); however, it has now been shown that the TGF-f32 isoform responded differently by showing no significant alteration from controls at all concentrations tested. To analyze whether the combination of TGF- f1 and TGF-,f2 had either a synergistic or competitive effect in BASMCs, we repeated the migration assay using 0.5 ng/ml TGF-,31 and a 10-fold increase in TGF-fl2 concentration (5.0 ng/ml). There was no change in the TGF-fl1-modulated increase in migration when TGF,32 was included in the medium in the presence of TGFf31 (data not shown). Migration of RFCs is nominal in a monolayer culture and was not influenced by TGF-f31 nor

TGF-#2. Figure 4. Autoradiographic analyses ofTGF-,3, sensitive parent and insensitive daughter keratinocyte cell lines: Cell monolayers were incubated with 60pmol/1 '25I-TGF-f,i and DSS crosslinked. The Triton X- 100 cell lysate was run on reduced SDSPAGE A: TGF-,3,-insensitive subclone and B: TGF-#% -sensitive parent line. The resultant autoradiographs demonstrated band for type I and type H TGF-f31 receptors on the sensitive parent line (Lane 3), which was competed off with 100-fold molar excess unlabeled TGF-f,% (Lane 4). The insensitive subclone showed only a trace amount of TGF-j,% cell surface binding protein (Lane 1), which was competed off with unlabeled TGF#I (Lane 2). MW markers included: 220 kdfibronectin, 212 kd myosin, 170 kd a2-macroglobulin, 116 kd ,3-galactosidase, 97 kd phosphorylase b, 76 kd transferrin, 53 kd glutamic dehydrogenase, and 43 kd ovalbumin.

ergistic effect or competition for TGF-f3-binding proteins between TGF-41 and TGF-32, a concentration range of TGF-32 (0.05 to 5.0 ng/ml) was added to the medium containing a constant amount of TGF-f31 (0.5 ng/ml) (Figure 5B). The combination of TGF-,81 with TGF-,32 showed no significant difference from the results obtained with TGF-f31 alone. Bovine aortic smooth muscle cell proliferation responses to TGF-f31 and TGF-f32 also were examined. The dose-dependent inhibition of subconfluent monolayers of BASMCs was optimum at 0.5 ng/ml TGFf1 (60%; Figure 6). In contrast to BAECs, BASMCs were inhibited by TGF-f32 to approximately the same levels as noted for TGF-31.

Large-vessel Migratory Responses to TGF-f3, and TGF-f2 To investigate further responses of the large vessel endothelium and smooth muscle cells to TGF-j31 and TGFf2, a migration assay was used. TGF-f13 was observed to inhibit BAEC migration in a dose-dependent fashion and at 5.0 ng/ml gave a 35% to 40% inhibition (Figure 7A). In contrast, TGF-fB2 was not inhibitory at 0.05 nor 0.5 ng/ml

Microvascular Endothelial Response to TGF-f!, and TGF-f2 We also examined the effects of TGF-,31 and TGF-132 on rat epididymal fat pad microvascular endothelial cell monolayers. As noted previously,21 TGF-j31 inhibited RFC proliferation in a dose-dependent fashion in two-dimentional cultures. To obtain a similar degree of inhibition by the two TGF-3 isoforms, it was necessary to use approximately a 1 0-fold higher concentration of TGF-g2 compared with TGF-f31. The dose-dependent response was calculated after 5 days in culture (Figure 8A). In contrast, when RFCs were grown in a three-dimentional culture system the cells responded differently to the TGF-,Bs. In a twodimentional environment, TGF-f31 elicited up to a 75% inhibition of RFC proliferation; however, when grown in a three-dimentional culture, the RFCs expressed no significant difference in proliferation from control cultures as measured by DNA analysis in response to any of the growth factors used (bFGF, EGF, TGF-f31, or TGF-i32, all at 0.5 ng/ml; Figure 8B).

To investigate further the potential roles of TGF-,B isoforms in the phenomena of in vitro angiogenesis, we grew RFCs in a three-dimentional culture with selected growth factors at varying concentrations. Epidermal growth factor, at all three concentrations, (0.05, 0.5, and 5.0 ng/ml) did not elicit an angiogenic response. Basic fibroblast growth factor induced complex tubular formations at both 0.5(Figure 9B) and 5.0-ng/ml concentrations. In contrast, TGF31 at 0.5 ng/ml (Figure 9C) elicited an angiogenic response equivalent to bFGF, while the higher concentration (5.0 ng/ml) appeared to have a negative effect on in vitro angiogenesis (data not shown). TGF-#2 was unable to elicit an angiogenic response at 0.05 or 0.5 ng/ml (Figure 9D); however, at 5.0 ng/ml the growth factor stimulated the formation of complex tubelike structures (Figure 9E).

Vascular Cell Responses to TGF-#1 and #2 45 AJPJanuary 1991, Vol. 138, No. 1

A

I

SO] OTGF4j

Ia

*TGF4Z

7

so-

I

2

a0, 0 =

Ic'c

20

-r

aS-- aEN6m 6000mi IMMMML. l-om

0

Ua mmoomm.

SAO

n [nWmq

Co

B so

c

so Figure 5. TGF-4, but not TGF-fi2 inhibits aortic endothelial proliferation: A: Fiveday, BAEC monolayers were grown in DME plus 10% FCS (control); with additions of bFGF, EGF, TGF-4, or TGF-g_. TGF4l, treated cultures were growth inhibited, while cultures treated with bFGF, EGF, or TGF-fi2 did not differ significantly from control B: Samples treated with TGF-#, gave an 80% inhibition ofproliferation, while TGF-f2 was ineffective. Cultures treated with a constant dose of TGF-,S, (0.5 ng/ml) and varying amounts of TGF-j2 (0. 05, 0.5 or5. 0 ng/ml) showed no differencefrom cultures treated with TGF-f,i alone. SE =