In vitro modulation of endothelial fenestrae: opposing effects of retinoic

acid and transforming grow h factor />

T. LOMBARDI1, R. MONTESANO1, M. B. FURIE2, S. C. SILVERSTEIN3 and L. ORCI1

^Institute of Histology and Embryology, University of Geneva Medical School, 1211 Geneva 4, Switzerlanddepartment of Pathology, State University ofNetv York, Stony Brook, NY 11794, USA^Department of Physiology and Cellular Biophysics, Columbia University, NY 10032, USA

Summary

Cultured endothelial cells isolated from fenestratedcapillaries express many properties characteristicof their in vivo differentiated phenotype, includingthe formation of a limited number of fenestrae. Inthis study, we have investigated whether physio-logical factors that control cell differentiationmight regulate the surface density of fenestrae incapillary endothelial cells. We have found thattreatment of the cultures with retinoic acid (10 fiM)induces a more than threefold increase in thesurface density of endothelial fenestrae, whereas

transforming growth factor /J (TGF/J) (Zcauses a sevenfold decrease in the surface density ofthese structures. These results show that the ex-pression of endothelial fenestrae is susceptible tobidirectional modulation by physiological signals,and suggest that retinoids and TGF/J may par-ticipate in the regulation of fenestral density ofcapillary endothelium in vivo.

Key words: endothelium, retinoids, growth factors.

Introduction

Endothelial fenestrae are unique structural specializ-ations of visceral capillaries that are believed to play a rolein the exchange of substances between blood and tissues(Rhodin, 1962; Maul, 1971; Simionescu et al. 1982;Simionescu, 1983; Bearer & Orci, 1985). The factors thatgovern the formation of these openings in capillaryendothelial cells are poorly understood. It has recentlybeen demonstrated that fenestrations are formed incapillary endothelial cells grown in vitro (Montesano &Orci, 1985; MilicieK al. 1985; Lombardi et al. 1986), andthat their density can be modulated by environmentalsignals (Milici et al. 1985; Lombardi et al. 1986). In thisrespect, we have reported that treatment with the tumourpromoter, 4/3-phorbol 12-myristate 13-acetate (PMA),markedly enhances the formation of fenestrae in culturedcapillary endothelial cells (Lombardi et al. 1986).Although phorbol esters have been shown to modifyprofoundly the differentiation program of a variety of celltypes (Blumberg, 1980; Diamond et al. 1980; Diamond,1984), they are non-physiological signals for endothelialcells. We therefore thought it important to examinewhether physiological factors known to regulate celldifferentiation in other systems might also influence theexpression of fenestrae in endothelial cells. In the presentstudy, we show that two physiological differentiation-modulating agents, retinoids and transforming growth

Journal of Cell Science 91, 313-318 (1988)Printed in Great Britain © The Company of Biologists Limited 1988

factor /3 (TGF/2), have marked, but opposing effects onthe surface density of fenestrae in cultured capillaryendothelial cells.

Materials and methods

Cell cultureBovine microvascular endothelial cells were isolated frombovine adrenal cortex according to Folkman et al. (1979) andcloned as described by Furie et al. (1984). The cells wereroutinely subcultured in gelatin-coated tissue culture flasks(Falcon Labware, Becton, Dickinson & Co., Oxnard, CA) incomplete medium consisting of minimal essential medium,alpha modification (Gibco Laboratories, Grand Island, NY)supplemented with 15 % heat-inactivated donor calf serum(Flow Laboratories, Irvine, Ayrshire, Scotland), penicillin(500i.u. ml"1), and streptomycin (lOO^gml"1). For exper-iments, the endothelial cells were seeded into either 35 mmplastic dishes (Falcon) (for thin-section electron microscopy) or35 mm dishes containing 25 mm round plastic coverslips (Ther-manox, Lux Scientific Inc., Newbury Park, CA) (for freeze-fracture), and grown to confluency before treatment with thevarious agents.

Electron microscopy and quantitative evaluationFor freeze-fracture electron microscopy, cultures grown onplastic coverslips (passages 13-21 after cloning) were fixed with2-5% glutaraldehyde in 0-lM-sodium cacodylate buffer,pH7-4, rinsed in cacodylate buffer, and cut into 2 mm X 2 mm

313

squares. The latter were inverted on Balzers specimen carriers(Balzers High Vacuum Corp., Balzers, Liechtenstein) andfreeze-fractured at — 110°C in a Balzers BAF 301 apparatusaccording to the method of Pauli et al. (1977). Replicas werecleaned by sequential treatment with sodium hypochlorite anddimethylformamide, rinsed in distilled water, recovered oncopper grids (see below) and examined in a Philips EM 300electron microscope.

A quantitative evaluation of the surface density of endothelialfenestrae under the various experimental conditions was carriedout as follows. Three replicas obtained from different regions ofeach culture dish were recovered separately on 400 mesh, squarehole copper grids (mean hole side = 42jt«m) (Tebra IGC 400,Teepe-Brandsma, Hilversum, The Netherlands). In each of thethree grids, micrographs of seven holes containing largeexpanses of plasma membrane fracture faces were taken atrandom at a magnification (X 1200) that permits the entire holearea to be included in a single micrograph (see Fig. 1, below).All clearly identifiable fenestrae present in each grid hole werethen systematically photographed at higher magnification(X9500 to X14000). Negatives were projected on a Tektronix49S3 graphic tablet (Tektronix International AG, Zug, Switzer-land), where both the surface area of the plasma membranefracture faces and the number of individual fenestrae containedin each grid hole were measured with an electronic penconnected to the graphic tablet and to an IBM-XT microproces-sor (IBM Schweiz, Zurich, Switzerland). In this way, the meannumber of fenestrae/100 jum2 of endothelial cell surface couldbe calculated in each experimental condition. Each value givenin the text represents the mean±s.E.M. of three to eightseparate experiments.

ChemicalsAll-/ra«.s-retinoic acid and retinal (Sigma Chemical Co., StLouis, MS) were initially dissolved in absolute ethanol to obtaina 12 mM concentration. These solutions, which were carefullyprotected from light, were then diluted in complete culturemedium with continuous stirring to yield a 10/«M concentrationof retinoid. The final concentration of solvent was 0"08%(control experiments showed that ethanol up to a concentrationof 1 % did not affect the number of fenestrae in endothelialcells). Isobutylmethylxanthine (IBMX), 4/3-phorbol 12-myris-tate 13-acetate (PMA), cortisone acetate, dexamethasone andporcine adrenocorticotropic hormone (ACTH, grade II, codeA6002) were also obtained from Sigma. Sodium butyrate andiV,iV-dimethylformamide were purchased from Merck (MerckSchweiz AG, Zurich, Switzerland), dimethylsulphoxide(DMSO) from Fluka (Fluka AG, Buchs, Switzerland), andepidermal growth factor from Collaborative Research (Wal-tham, MA). l<r,25-dihydroxyvitamin D3 was obtained fromDrs A. Kaiser and W. Meier (Hoffman La Roche, Basel,Switzerland). Transforming growth factor /3 (TGF-/3), purifiedto homogeneity from human platelets as described by Assoian etal. (1983), was a generous gift from Dr M. B. Sporn (NationalCancer Institute, Bethesda, MD), and bovine pituitary basicfibroblast growth factor (bFGF) was kindly provided by Dr A.Baird (The Salk Institute, San Diego, CA).

(caveolae) by their larger diameter, their shallow, flatfloor, and their characteristic occurrence in discreteclusters, as is observed in vivo (Simionescu et al. 1974;Orci & Perrelet, 1975) (Figs 1 and 2). Since clusters offenestrations are relatively infrequent and non-uniformlydistributed in endothelial cells, an extensive quantitativeanalysis of a large surface of endothelial plasma mem-brane fracture faces (2x 105 to 3X105 [im2 per experimen-tal condition) in three to eight separate experiments hasbeen necessary to demonstrate significant differences inthe surface density of individual fenestrae between thevarious experimental conditions. This evaluation showedthat treatment of the cultures with retinoic acid (10/iM)for 5 days induced a more than threefold increase in thedensity of fenestrae/100 jitm2 of cell surface (Table 1)*.Increasing the concentration of retinoic acid up to 100 f-iMfailed to increase further the surface density of endo-thelial fenestrae (data not shown). In a separate set ofexperiments, another retinoid, retinal (lOjiiM), alsoinduced a significant (more than twofold) increase in thesurface density of fenestrae (Table 1). A variety of agentsthat have been shown to promote cell differentiation inother systems were also tested for their possible effect onendothelial fenestrae. These included: sodium butyrate(2-5 mM), la',25-dihydroxyvitamin D3 (5ngml~'), iso-butylmethylxanthine (IBMX, 1-5 mM), and the polarsolvents, dimethylsulphoxide and dimethylformamide(1-5%, v/v). Among these agents, only dimethylforma-mide induced a significant change in the surface densityof fenestrae, with an increase comparable to that elicitedby retinoic acid (control: 5-49 ±0-63; dimethylforma-mide: 18-96 ±2-76; P< 0-005; « = 4).

Over the last few years, evidence has accumulatedindicating that transforming growth factor /J (TGF/3) isan important physiological modulator of cell differen-tiation (for recent reviews, see Sporn et al. 1986, 1987;Keski-Ojae<«/. 1987; Massague", 1987). TGF/3induced adramatic decrease in the surface density of fenestrae inendothelial cells. Confluent cultures incubated for 2 daysin the presence of TGF/3 (2 ng ml~ ) displayed a morethan sevenfold decrease in the surface density of fenestraewith respect to parallel cultures incubated for 2 days incontrol medium (Fig. 3). The density of fenestrae inendothelial cells treated for 2 days with TGF^3 was alsosignificantly lower (P < 0-001) than that measured incultures fixed at the beginning of the experiment (time0), when the endothelial cells had just attained con-fluency (Fig. 3). Thus, TGF/3 not only prevents theincrease in the surface density of fenestrae that occursunder normal conditions after the cells have reachedconfluency, but also causes a decrease in the density ofpre-existing fenestrae. The effect of TGF/3 was revers-

Results

As shown by Lombardi et al. (1986), in freeze-fracturereplicas of cultured capillary endothelial cells, fenestraeare recognized as circular depressions (on the P fractureface) or elevations (on the E fracture face), which caneasily be distinguished from plasmalemmal vesicles

* Throughout the entire series of experiments reported in thisstudy, the surface density of fenestrae in control cultures wasfound to be considerably higher than that measured in aprevious study (Lombardi et al. 1986). The reasons for thisdifference are not clear; the cells used in the two groups ofexperiments were, however, from different frozen stocks ofendothelial cells originally derived from the same clonal popu-lation (cf. Furie et al. 1984).

314 T. Lombardi et al.

Table 1. Surface density of fenestrae in control andretinoid-treated endothelial cells

Total cellsurface

examinedTotal

number offenestrae

Number offenestrae/100 /«n2

ControlRetinoic acid

ControlRetinal

197 863194 651

320225307 329

1175336477

1982141950

5-94 + 0-68*18-74 ± 1-50*

6-19 ±0-64**13-65 ±1-57**

(n = 3)

(» = 5)

Endothelial cells were grown to confluency on plastic coverslipsand further incubated for 5 days in normal medium or in mediumcontaining either retinoic acid or retinal (10f/M). The quantitativeevaluation of the surface density of fenestrae was carried out onfreeze-fracture replicas as described in Materials and methods. Thevalues of number of fenestrae/100ftm2 represent the mean ± S.E.M. ofthree (n — 3) (top) or five (n = 5) (bottom) separate experiments(cultures produced and examined at different times). *P<0-005;**/J<0-01 (Student's/-test).

ible: cultures that had been exposed to TGF/3 for 2 days,subsequently washed, and reincubated in normal culturemedium (in the absence of TGF/3) for a further 5 dayshad approximately the same surface density of fenestraeas controls, whereas fenestrae were very rarely seen inparallel cultures maintained in the continuous presence ofTGF/3 for 7 days (Fig. 3).

Fig. 1. Low-power electronmicrograph of a grid holecontaining a portion of a freeze-fracture replica of culturedcapillary endothelial cells. Thefracture process has exposedlarge expanses of plasmamembrane faces, in whichclusters of tightly packedfenestrae are recognizable(asterisks). The surface densityof fenestrae has been evaluatedby measuring the surface area ofthe plasma membrane fracturefaces and counting the totalnumber of individual fenestraecontained in the grid hole asdescribed in Materials andmethods. The replica is from aculture treated withdimethylformamide for 5 days.X3750; bar, lOftm.

In another group of experiments, endothelial culturesthat were treated simultaneously with both PMA(5ngml~ ) (an agent that stimulates the formation offenestrae; cf. Lombardi et al. 1986) and TGF/3(2ngml~ ) for 3 days, showed very few fenestrae(0-60 ± 0-17). However, cultures treated with both PMAand TGF/3 as above for 3 days, then washed andreincubated with PMA in the absence of TGF/3 for afurther 4 days, contained numerous fenestrae(10-47 ± 1-87), in contrast to parallel cultures maintainedin the presence of both PMA and TGF/3 for 7 days(0-51 ±O06). Thus, TGF/3 can also override, in areversible manner, the effect of a powerful fenestrae-inducing stimulus such as PMA.

In contrast to TGF/3, basic fibroblast growth factor(bFGF) (3 ngml"1) and epidermal growth factor (EGF)(50-500 ngml"1), two polypeptides known to affectother endothelial cell properties (Montesano et al. 1986;Schreiber et al. 1986), did not induce significant changesin the surface density of endothelial fenestrae.

Finally, in view of a recent report indicating thatadrenocorticotropic hormone (ACTH) increases the den-sity of fenestrae in the capillary endothelium of ratadrenals following hypophysectomy (Apkarian & Curtis,1986), we have explored the possible in vitro effect ofACTH (10 units ml"1), cortisone (200 nM-50jUM) and thesynthetic glucocorticoid, dexamethasone (100 nM). None

Modulation of endothelial fenestrae 315

Fig. 2. Freeze-fracture replicaof a capillary endothelial cellculture treated for S days withretinoic acid. Fenestrae occur inclusters and are recognized ascircular depressions on the P-face that are matched bycorresponding elevations on theE-face. X15 000; bar, Zfim.

I..••'**

Time (days)

Fig. 3. Surface density of fenestrae in control and TGF/3-treated endothelial cells. Cultures were fixed either at thebeginning of the experiment, when the endothelial cells hadjust attained confluency (time 0), or after 2 days or 7 days ofincubation in either normal medium ( ) or mediumcontaining 2ngml~ TGF/S ( ). In addition, in eachexperiment, a dish was exposed to TGF/3 for 2 days,subsequently washed and reincubated in normal culturemedium (in the absence of TGF/8) for a further S days( ). The values of the number of fenestrae/100 /im2

represent the mean ± S.E.M. of six separate experiments.Differences in the surface density of fenestrae betweencontrol and TGF/3-treated cultures are highly significant(P< 0-001).

of these hormones significantly modified the surfacedensity of fenestrae in our cultures of adrenal corticalcapillary endothelial cells.

Discussion

Experimental analysis of the factors that might regulatethe formation of endothelial fenestrae and/or their func-tional properties has awaited the development of culturemethods for the isolation and growth of microvascularendothelial cells (Folkman et al. 1979). Recent evidenceindicates that appropriate culture conditions can enhancethe degree of differentiation of capillary endothelial cells,in particular as far as the expression of fenestrae isconcerned: an increased linear density of transendothelialopenings (including fenestrae as well as transendothelialchannels) has been observed in endothelial cells grown onthe extracellular matrix laid down by MDCK epithelialcells (0-157 openings per /.im on MDCK matrix versus0-007 per ^m on plastic) (Milici et al. 1985), and we haverecently demonstrated (Lombardi et al. 1986) an increasein the surface density of fenestrae in endothelial cellstreated with the tumour promoter, PMA (6-13 per100 fim2 in PMA-treated cells versus 1-08 per lOOjitnrincontrol cells). In the present study, we have shown thattwo physiological factors known to influence the differen-tiation state of various cell types have marked butopposite effects on the surface density of fenestrae incultured endothelial cells: retinoic acid increases (18-74per,um2 versus 5-94 per ftm2 in control cells), whereasTGF/3 drastically decreases (0-71 per fim2 versus 5-59per jiim2 in control cells) the density of these structures.(See footnote to p. 314.)

Retinoic acid is a naturally occurring vitamin A metab-

316 T. Lombardi et al.

olite that has been shown to exert profound effects on thedifferentiation of epithelial and mesenchymal cells by stillpoorly understood mechanisms (for reviews, see Lotan,1980; Zile & Cullum, 1983; Sporn & Roberts, 1984;Sporn et al. 1984; Wolf, 1984; Goodman, 1984; Chytil,1984; Sherman, 1986; Shapiro, 1986). This study dem-onstrates that retinoic acid (as well as retinal, anothervitamin A metabolite) also promotes the in vitro differen-tiation of endothelial cells, at least as far as the expressionof fenestrae is concerned. Whether retinoids are involvedin the regulation of the density of endothelial fenestrae inblood capillaries in vivo remains to be established. It isnoteworthy that whereas administration of ACTH in vivoincreases the density of fenestrae in the capillary endo-thelium of rat adrenal cortex following hypophysectomy(Apkarian & Curtis, 1986), neither ACTH nor glucocor-ticoid hormones had any effect on the surface density offenestrae in our in vitro system. Since adrenocorticalhormones increase the serum vitamin A level, acceleratethe rate at which vitamin A is mobilized from the liver,and stimulate the net synthesis of retinol-binding proteinby cultured hepatoma cells (Wang et al. 1954; Clark &Colburn, 1955; McGillivray, 1961; Stoewsand & Scott,1964; Borek et al. 1981), it seems reasonable to suggestthat retinoids mediate the observed in vivo effect ofACTH on endothelial fenestrae.

In marked contrast to the effect of retinoids, TGFySproduced a drastic decrease in the surface density offenestrae, which was reversible after removal of TGF/3from the culture medium. TGF/3 is a polypeptide orig-inally isolated and named on the basis of its capability ofpromoting anchorage-independent growth of non-trans-formed fibroblasts (for recent reviews, see Sporn et al.1986, 1987; Keski-Oja et al. 1987; Massague\ 1987).More recently, however, it has been realized that TGF/3is a multifunctional regulatory molecule: it may eitherstimulate or inhibit cell proliferation, or have numerousother actions often uniquely related to the regulation ofthe specialized, critical function of a particular cell type(Sporn et al. 1986, 1987; Keski-Oja et al. 1987; Massa-gu6, 1987). Thus, the expression of specific phenotypesby cells with differentiating potential is frequently pro-foundly altered by TGF/J. For example, TGF/J promotesthe differentiation of chondrocytes (Seyedin et al. 1986)and epithelial cells (Masui et al. 1986), but inhibits thedifferentiation of myoblasts (Florini et al. 1986; Olson etal. 1986; Massague' et al. 1986) and preadipocytes(Ignotz & Massague\ 1985). TGF/3 has been shown toaffect endothelial cell properties markedly: it inhibitstheir proliferation (Frater-Schroder et al. 1986; Baird &Durkin, 1986; Heimarkef al. 1986; Takehara et al. 1987;Muller et al. 1987) and migration (Heimark et al. 1986;Miiller et al. 1987), decreases their plasminogen activatoractivity (Saksela et al. 1987), and prevents their invasionof collagen matrices in response to PMA (Muller et al.1987; Montesano, unpublished observations quoted byRoberts & Sporn, 1987). The results of this studyindicate that TGF/3 also inhibits endothelial cell differen-tiation, since it markedly decreases the surface density ofspecialized structures characteristic of their in vivophenotype: it prevents the spontaneous appearance of

newly formed fenestrae, as well as that stimulated byPMA (see Results) and retinoic acid (unpublished obser-vations), and induces the removal of preformed fenes-trae. TGF/3 is present in a wide variety of tissues and isespecially abundant in blood platelets (Assoian et al.1983; Assoian & Sporn, 1986), from which it is releasedafter vessel damage (Assoian & Sporn, 1986). It istherefore conceivable that TGF/3 may play a role in theregulation of fenestral density (and thus of the per-meability properties) of capillary blood vessels in vivo.Interestingly, it has been reported that experimentalthrombocytopenia is associated with the formation offenestrae in the normally non-fenestrated endotheliumof muscle capillaries, and that restoration of the numberof circulating platelets to normal levels is accompanied bythe disappearance of newly formed fenestrae (Kitchens &Weiss, 1975).

In conclusion, we have shown that two physiologicaldifferentiation-modulating agents, retinoic acid andTGF/3, have a profound, but opposite effect on thesurface density of fenestrae in cultured capillary endo-thelial cells. This further demonstrates that fenestrae arenot permanent specializations of the endothelial cells, butare labile structures, whose expression is governed byenvironmental signals. Further studies will be required todetermine whether retinoic acid and TGF/3 can regulatethe fenestral density and the permeability of capillaryendothelium in vivo.

We are grateful to Dr M. B. Sporn for generously providingthe TGF/3. We also thank J. Rial, P. Fruleux, P. Sors andM. Bernard for technical assistance, Dr M. Amherdt for helpfuladvice, Drs J.-D. Vassalli and M. S. Pepper for criticallyreading and improving the manuscript, and I. Bernard forsecretarial assistance. This study was supported by the SwissNational Science Foundation, grant no. 3.404.86 and by theJuvenile Diabetes Foundation, grant no. 187464.

References

APKARIAN, R. P. & CURTIS, J. C. (1986). Hormonal regulation ofcapillary fenestrae in the rat adrenal cortex: quantitative studiesusing objective lens staging scanning electron microscopy.Scanning Electron Microscopy 1986, IV, 1381-1393.

ASSOIAN, R. K., KOMORIYA, A., MEYERS, C. A., MILLER, D. M. &

SPORN, M. B. (1983). Transforming growth factor-/} in humanplatelets. J. biol. Chem. 258, 7155-7160.

ASSOIAN, R. K. & SPORN, M. B. (1986). Type /S transforminggrowth factor in human platelets: release during plateletdegranulation and action on vascular smooth muscle cells. J. CellBiol. 102, 1217-1223.

BAIRD, A. & DURKIN, T. (1986). Inhibition of endothelial cellproliferation by type /3 transforming growth factor: interactionswith acidic and basic fibroblast growth factors. Biochem. biophys.Res. Commun. 138, 476-482.

BEARER, E. L. & ORCI, L. (1985). Endothelial fenestral diaphragms:a quick-freeze, deep-etch study. J. Cell Biol. 100, 418-428.

BLUMBERG, P. M. (1980). In vitro studies on the mode of action ofphorbol esters, potent tumor promoters. CRC Crit. Rev. Toxicol. 8,199-234.

BOREK, C , SMITH, J. E., SOPRANO, D. R. & GOODMAN, D. S.

(1981). Regulation of retinol-binding protein metabolism byglucocorticoid hormones in cultured H4IIEC3 liver cells.Endocrinology 109, 386-391.

CHYTIL, F. (1984). Retinoic acid: biochemistry, pharmacology,toxicology, and therapeutic use. Phannac. Rev. 36, 939-1005.

Modulation of endothelial fenestrae 317

CLARK, I. & COLBURN, R. W. (1955). A relationship between vitaminA metabolism and cortisone. Endocrinology 56, 232-238.

DIAMOND, L. (1984). Tumor promoters and cell transformation.Phannac. Ther. 26, 89-145.

DIAMOND, L., O'BRIEN, T. G. & BAIRD, W. M. (1980). Tumor

promoters and the mechanism of tumor promotion. Adv. CancerRes. 32, 1-74.

FLORINI, J. R., ROBERTS, A. B., EWTON, D. Z., FALEN, S. L.,

FLANDERS, K. C. & SPORN, M. B. (1986). Transforming growthfactor-/3. A very potent inhibitor of myoblast differentiation,identical to the differentiation inhibitor secreted by Buffalo ratliver cells. J . biol. Chem. 261, 16509-16513.

FOLKMAN, J., HAUDENSCHILD, C. C. & ZETTER, B. R. (1979). Long

term culture of capillary endothelial cells. Proc. natn. Acad. Sci.U.S.A. 76, 5217-5221.

F RATER-SCHRODER, M., MOLLER, G., BIRCHMEIER, W. & BOHLEN, P.

(1986). Transforming growth factor-beta inhibits endothelial cellproliferation. Biochem. biophys. Res. Commun. 137, 295-302.

FURIE, M. B., CRAMER, E. B., NAPRSTEK, B. L. & SILVERSTEIN, S.

C. (1984). Cultured endothelial cell monolayers that restrict thetransendothelial passage of macromolecules and electrical current.J. Cell Biol. 98, 1033-1041.

GOODMAN, D. S. (1984). Vitamin A and retinoids in health anddisease. Netv Engl.J. Med. 310, 1023-1031.

HEIMARK, R. L., TWARDZIK, D. R. & SCHWARTZ, S. M. (1986).

Inhibition of endothelial regeneration by type-beta transforminggrowth factor from platelets. Science 233, 1078-1080.

IGNOTZ, R. A. & MASSAGUE, J. (1985). Type /? transforming growthfactor controls the adipogenic differentiation of 3T3 fibroblasts.Proc. natn. Acad. Sci. U.S.A. 82, 530-534.

KESKI-OJA, J., LEOF, E. B., LYONS, R. M., COFFEY, R. J. JR &

MOSES, H. L. (1987). Transforming growth factors and control ofneoplastic cell growth. J . Cell Biochem. 33, 95-107.

KITCHENS, C. S. & WEISS, L. (1975). Ultrastructural changes ofendothelium associated with thrombocytopenia. Blood 46, 567-578.

LOMBARDI, T . , MONTESANO, R., FuRIE, M. B. , SlLVERSTEIN, S. C. &

ORCI, L. (1986). Endothelial diaphragmed fenestrae: in vitromodulation by phorbol myristate acetate, jf. Cell Biol. 102,1965-1970.

LOTAN, R. (1980). Effects of vitamin A and its analogs (retinoids) onnormal and neoplastic cells. Biochim. biophys. Ada 605, 33-91.

MASSAGUE, J. (1987). The TGF/3 family of growth anddifferentiation factors. Cell 49, 437-438.

MASSAGUE, J., CHEIFETZ, S., ENDO, T. & NADAL-GINARD, B.

(1986). Type /3 transforming growth factor is an inhibitor ofmyogenic differentiation. Proc. natn. Acad. Sci. U.SA. 83,8206-8210.

MASUI, T., WAKEFIELD, L. M., LECHLER, J. F., LAVECK, M. A.,

SPORN, M. B. & HARRIS, C. C. (1986). Type /3 transforminggrowth factor is the primary differentiation inducing factor fornormal human bronchial epithelial cells. Proc. natn. Acad. Sci.U.SA. 83, 2438-2442.

MAUL, G. G. J. (1971). Structure and formation of pores infenestrated capillaries, jf. Ultrastruct. Res. 36, 768-782.

MCGILLIVRAY, W. A. (1961). Some factors influencing the release ofvitamin A from the liver. Br.J. Nutr. 15, 305-312.

MILICI, A. J., FURIE, M. B. & CARLEY, W. W. (1985). The

formation of fenestrations and channels by capillary endothelium invitro. Proc. natn. Acad. Sci. U.SA. 82, 6181-6185.

MONTESANO, R. & ORCI, L. (1985). Tumor promoting phorbol estersinduce angiogenesis in vitro. Cell 42, 469-477.

MONTESANO, R., VASSALLI, J.-D., BAIRD, A., GUILLEMIN, R. &ORCI, L. (1986). Basic fibroblast growth factor inducesangiogenesis in vitro. Proc. natn. Acad. Sci. U.SA. 83, 7297-7301.

MOLLER, G., BEHRENS, J., NUSSBAUMER, U., BOHLEN, P. &BIRCHMEIER, W. (1987). Inhibitory action of transforming growthfactor /8 on endothelial cells. Proc. natn. Acad. Sci. U.SA. 84,5600-5604.

OLSON, E. N., STERNBERG, E., HU, J. S., SPIZZ, G. & WILCOX, C.

(1986). Regulation of myogenic differentiation by type /3transforming growth factor. J. Cell Biol. 103, 1799-1805.

ORCI, L. & PERRELET, A. (1975). Freeze-etch Histology. Berlin:Springer Verlag.

PAULI, B. U., WEINSTEIN, R. S., SOBLE, L. W. & ALROY, J. (1977).

Freeze-fracture of monolayer cultures. .7. Cell Biol. 72, 763-769.RHODIN, J. A. G. (1962). The diaphragm of capillary endothelial

fenestrations. J . Ultrastmct. Res. 6, 171-185.ROBERTS, A. B. & SPORN, M. B. (1987). Transforming growth

factor-/?: stimulator or inhibitor of angiogenesis? In Angiogenesis:Mechanisms and Pathobiology (ed. D. B. Rifkin & M. Klagsbrun),pp. 69-71. Cold Spring Harbor, NY: Cold Spring HarborLaboratory Press.

SAKSELA, O., MOSCATELLI, D. & RIFKIN, D. B. (1987). The

opposing effects of basic fibroblast growth factor and transforminggrowth factor beta on the regulation of plasminogen activatoractivity in capillary endothelial cells. J . Cell Biol. 105, 957-963.

SCHREIBER, A. B., WlNKLER, M. E. & DERYNCK, R. (1986).Transforming growth factor-cv: a more potent angiogenic mediatorthan epidermal growth factor. Science 232, 1250-1253.

SEYEDIN, S. M., THOMPSON, A. Y., BENTZ, H., ROSEN, D. M.,MCPHERSON, J. M., CONTI, A., SIEGEL, N. R., GALUPPI, G. R. &PIETZ, K. A. (1986). Cartilage-inducing factor-A: apparent identityto transforming growth factor-/! J. biol. Cheiri. 261, 5693-5695.

SHAPIRO, S. S. (1986). Retinoids and epithelial differentiation. InRetinoids and Cell Differentiation (ed. M. I. Sherman), pp. 29-59.Boca Raton, Florida: CRC Press.

SHERMAN, M. I. (ed.) (1986). Retinoids and Cell Differentiation,p. 198. Boca Raton, Florida: CRC Press, Inc.

SlMIONESCU, M. , SlMIONESCU, N . & PALADE, G. (1974).

Morphometric data on the endothelium of blood capillaries. J. CellBiol. 60, 128-152.

SlMIONESCU, M. , SlMIONESCU, N . & PALADE, G. (1982).

Differentiated microdomains on the luminal surface of capillaryendothelium: Distribution of lectin receptors. J. Cell Biol. 94,406-413.

SIMIONESCU, N. (1983). Cellular aspects of transcapillary exchange.Physiol. Rev. 63, 1536-1579.

SPORN, M. B. & ROBERTS, A. (1984). Role of retinoids indifferentiation and carcinogenesis. J. natn. Cancer lust. 73,1381-1387.

SPORN, M. B., ROBERTS, A. B. & GOODMAN, D. S. (1984). The

Retinoids. Orlando: Academic Press.SPORN, M. B., ROBERTS, A. B., WAKEFIELD, L. M. & ASSOIAN, R.

K. (1986). Transforming growth factor-/?: biological function andchemical structure. Science 233, 532-534.

SPORN, M. B., ROBERTS, A. B., WAKEFIELD, L. M. & DE

CROMBRUGGHE, B. (1987). Some recent advances in the chemistryand biology of transforming growth factor beta. J. Cell Biol. 105,1039-1045.

STOEWSAND, G. S. & SCOTT, M. L. (1964). Effects of stress fromhigh protein diets on vitamin A metabolism in chicks. J. Nutrit.82, 188-196.

TAKEHARA, K., LEROY, E. C. & GROTENDORST, G. R. (1987). TGF-P inhibition of endothelial cell proliferation: alteration of EGFbinding and EGF-induced growth regulatory (competence) geneexpression. Cell 49, 415-422.

WANG, P., GLASS, H. L., GOLDENBERG, L., STEARNS, G., KELLY,

H. G. & JACKSON, R. L. (1954). Serum vitamin A and carotenelevels in children with rheumatic fever. Am. J. Dis. Child. 87,659-672.

WOLF, G. (1984). Multiple functions of vitamin A. Phvsiol. Rev. 64,873-937.

ZILE, M. H. & CULLUM, M. E. (1983). The function of vitamin A:current concepts. Ptvc. Soc. exp. Biol. Med. 172, 139-152.

(Received 21 April 1988-Accepted, in revised form, 12 July 1988)

318 T. Lombardi et al.