JOURNAL OF CELLULAR PHYSIOLOGY 144:204-215 (1990)

Astrocytes Induce Neural Microvascular Endothelial Cells to Form Capillary-Like Structures In Vitro JOHNLATERRA,* CHRISTOPHER CUERIN, AND GARY W. COLDSTEIN Departments of Neurology (J.l., C.W.C.), Pediatrics (C.W.G.1 and Neurosurgery (C.C.), The Johns Hopkins Medical Institutions, and The Kennedy institute (J.L., C. W.C.), Baltimore, Maryland 2 1205 Astrocytes maintain a unique association with the central nervous system microvasculature and are thought to play a role in neural microvessel formation and differentiation. We investigated the influence of astroglial cells on neural microvascular endothelial differentiation in vitro. Using an astroglial-endothefiai coculture system, rat brain astrocytes and C, cells of astroglial lineage are shown to induce bovine retinal microvascular endothelial (ERE) cells to form capillarylike structures. Light microscopic evidence for endothelial reorganization began within 48 hours and was complete 72-96 hours following the addition of BRE cells to 1-day-old astroglial cultures. The extent of BKE reorganization was quantitated by computer-assisted analysis and shown to be dependent upon the density of both the BRE and C, cells within the cocultures. Coculture conditions in which BRE cells were separated from C, cells by porous membranes failed to generate this endothelial cell change. Likewise, C,-conditioned media and C,endothelial coculture conditioned media did not induce BRE cell reorganization. Extracellular laminin within the C,-endothelial cocultures, identified by indirect inimunofluorescence, was concentrated at the endothelial-astroglial interface of capillary-like structures consistent with incipient basement membrane formation. Astroglial cells accumulated adjacent to capillary-like structures suggesting the presence of bidirectional influences between the reorganized endothelial cells and astroglia. This is the first demonstration of astroglial induction of angiogenesis in vitro and these findings support a functional role for perivascular astrocytes in the vascularization of neural tissue such as retina and brain.

A number of experimental systems are available to study angiogenesis in vivo, the most common being the chick chorioallantoic membrane (Ausprunk et al., 1974) and rabbit corneal pocket assays (Gimbrone et al., 1974). With these systems, steps identified in the process of angiogenesis include endothelial activation, basement membrane dissolution, endothelial migration, and replication t o form cords and tubes, followed by capillary maturation and basement membrane deposition (D’Amore and Thompson, 1987; Folkman, 1986; Furcht, 1986). In addition to in vivo models, isolated endothelial cells form capillary-like structures in vitro after prolonged culture without growth factors (Feder et al., 1983; Folkman and Haudenschild, 1980; Maciag et al., 1982) or in response to specific extracellular matrix molecules (Ingber and Folkman, 198913; Kubota et al., 1988) or certain diffusible factors (Hockel et al., 1987). These and other in vitro models identified the combined roles of the extracellular matrix and soluble factors in modulating endothelial-matrix and endothelial-endothelial adhesive interactions during capillary formation (Herbst et al., 1988; Ingber and Folkman, 1989a; Madri et al., 1988). However, angiogenesis in vivo rarely occurs within an environment free of other parenchymal cells. The influence of different paren0 1990 WILEY-LISS,

INC.

chymal environments on microvessel formation and the expression of tissue-specific endothelial properties is being increasingly realized (Auerbach et al., 1987; Butcher et al., 1980). A special association between astrocytes and the microvasculature of central nervous system tissues has been recognized and recently reviewed (Joo, 1987; Laterra et al., in press). The formation of microvessels in the developing brain occurs in close proximity t o astrocytic cell processes, and vascularization of the retina is temporally preceded by and anatomically associated with the immigration of astrocytes from the optic nerve (Bar and Wolff, 1972; Ling and Stone, 1988; Watanabe

Received February 1, 1990; accepted April 25, 1990.

*To whom correspondenceireprint requests should be addressed at The Kennedy Institute, 707 North Broadway, Baltimore, MD 21205. Abbreviations used: BRE = bovine retinal microvascular endothelial cells; DiI-acyl-LDL = 1-1‘-dioctadecyl-3,3,3‘ ,3’-tetraethyl-indocarbocyanine perchlorate acetylated low density lipoprotein; HBSS = Hanks’ balanced salt solution; NRK = normal rat kidney fibroblasts.

ASTROCYTE INDUCTION OF ANGIOGENESIS I N VITRO

and Raff, 1988).In addition, the microvasculature of the central nervous system is functionally and morphologically distinct from that of other tissues. With few exceptions, central nervous system microvessels express a complex collection of physical, metabolic, and asymmetric transport properties that constitute the bloodbrain barrier. These barrier properties result from secondary differentiation of the microvascular endothelial cells which are nonfenestrated, relatively lacking in vesicular transport systems, and interconnected by a continuous network of complex tight junctions (Betz and Goldstein, 1978;Goldstein and Betz, 1986;Brightman and Reese, 1969; Coomber and Stewart, 1985, 1986; Reese and Karnofsky, 1967;Stewart and Hayakawa, 1987).Within the central nervous system, foot processes of perivascular astrocytes, for which there is no comparable cell type in peripheral organs, ensheathe the microvascular endothelial cells forming a n endothelial-astrocytic complex. The signals that induce endothelial cells to express the blood-brain barrier phenotype are believed to result from specific interactions between capillary endothelial cells and brain parenchyma (Risau e t al., 1986; Stewart and Wiley, 1981)or more specifically between capillary endothelial cells and the ensheathing perivascular astrocytes (Beck et al., 1984;DeBault and Cancilla, 1980;Janzer and Raff, 1987;Tao-Cheng e t al., 1987).Therefore the role of astrocytes in influencing endothelial cell behavior is important to the understanding of neural microvessel formation and function in normal and disease states. This report demonstrates that central nervous system microvascular endothelial cells are induced to form capillary-like structures in response to coculture with rat brain astrocytes or cells of the C, line, which display astrocytic properties (Bissel et al., 1974; Kumar et al., 1986).This in vitro system of “angiogenesis” demonstrates that astrocytes are able to induce endothelial differentiation and provides a model for the study of microvessel morphogenesis and cell-cell interaction relevant to neural microvessel development and blood-brain barrier formation.

MATERIALS AND METHODS

Homologous cell cultures Bovine retinal endothelial (BRE) cells were isolated and cultured by a modification of the methods of Bowman et al. (1982).Briefly, retinas were aseptically removed and placed in iced Hanks’ balanced salt solution containing calcium and magnesium (HBSS), trapped on 183 pm nylon mesh, washed extensively with HBSS, and homogenized with 10 strokes of a 0.2 mm clearance Dounce homogenizer. After centrifugation a t 1,0009 for 5 minutes, the pellet was resuspended in HBSS containing 5 mgiml bovine serum albumin and filtered through a 118 pm nylon mesh. Microvessels within the filtrate were collected on a 53 km mesh, centrifuged at 2009 for 5 minutes, resuspended in Dulbecco’s modified Eagle’s medium (DME) containing 2 mgiml collagenase-dispase [Boehringer Mannheim Biochemicals) and 5 ugiml bovine serum albumin, and orbitally shaken a t 150 rpm, for 45-60 minutes, and the pellet resuspended in minimal essential medium (MEM) plus d-valine containing 20% fetal bovine se-

205

rum, MEM vitamins, nonessential amino acids, 16 U/ ml heparin, 50 ugiml Endothelial Mitogen (Maciag et al, 1979;Biomedical Technologies, Inc.), 100 Uiml penicillin, and 100 pgiml streptomycin (stock endothelial media). Microvessels from 20 retinas were plated onto 50 cm’ of tissue culture plastic pretreated with 20 pg/ ml bovine plasma fibronectin (Sigma) in DME for 1-2 hours, and cultures in humidified 5% C02/95% air a t 37°C.Cells derived from microvessels were cultured under the same conditions in the same media and passaged a t confluence by trypsinization. Cells a t passages 3-10 were used in experiments. These cultures were demonstrated to be essentially 100% endothelial by labeling with 1 ,l‘-dioctadecyl-3,3,3‘,3’-tetramethylindocarbocyanine perchlorate acetylated low density lipoprotein (DiI-acyl-LDL, Biomedical Technologies Inc.) as described below (Netland et al., 1985). Neonatal rat forebrain astrocytes were isolated by a modification of the methods of Franzakis and Kimelberg (1984).Briefly, cerebral hemispheres were removed from 1-2-day-old Sprague-Dawley rat pups, cleaned of meninges and choroid plexus, chopped, and serially digested with 3 mg!ml neutral dispase (Boehringer Mannheim) in DME. Dissociated cells were centrifuged a t 500g for 5 minutes and the pellet resuspended in TIME containing 10% fetal bovine serum and 50 pgiml gentamicin sulfate. Cells were plated at a n approximate density of 30 cm2 per rat brain and grown in humidified 5% CO2195% air at 37°C. Before reaching confluence, cells were treated for 2 days with media containing 2 x lop5 M cytosine arabinoside (Janzer and Raff, 1987).At confluence, oligodendroglia were removed by the orbital shaking method of McCarthy and de Vellis (1980).Astrocytes were then passaged by trypsinization, grown to confluence, and shaken again just before use in experiments. These cultures were approximately 95% glial fibrillary acidic protein (GFAP) positive by indirect immunof luorescence. Stock C, cells (Benda e t al., 1968)obtained from the American Type Culture Collection (ATCC) and normal rat kidney fibroblast cells (NRK, ATCC-CRL 1570) were cultured i n DME containing 10%bovine calf serum and 50 pg/ml gentamicin sulfate in humidified 5% C02/95%air a t 37°C.

Cocultures In most experiments, astrocytes or C, cells were trypsinized, resuspended in stock culture media, and plated a t 10,000 cellsichamber (6,000cellsicm’) into four chamber Lab-Tek slides (Nunc, Inc.j pretreated with 20 pg/ml fibronectin in DME for 1-2 hours. In some experiments the number of astroglial cells plated was varied as indicated in Results. After 24 hours of incubation, the media was removed and replaced with BRE cells suspended in stock endothelial cell media containing 50 pgiml fresh ascorbic acid a t 20,000 cells/ chamber unless otherwise indicated. Control slides lacked either the astroglial or endothelial cell inoculum. Cocultures were returned to the incubator until processed for microscopy. In some experiments, Endothelial Mitogen was excluded, or NRK cells were substituted for astroglial cells. Certain experiments were performed in which media was shared, but the two cell types were physically separated. In these experiments, 10,000 C, cells were

206

LATERRAETAL

plated into 12 mm Millicell-CM culture plate inserts of 0.4 pm dia. pore size (Millipore Corp.) and cultured in fibronectin-treated 12 well plates for 24 hours. Media was then removed and BRE cells plated onto the fibronectin-treated plastic in stock endothelial media containing 50 pg/ml ascorbic acid. The Millicell inserts were replaced and after 3 more days of incubation, the endothelial cells were assessed by phase contrast microscopy for formation of capillary-like structures.

their microvascular origin, develop sprouts that extend above the monolayer but fail to organize into capillarylike structures even if maintained for up to 4 weeks (data not shown). Treatment of these endothelial cell cultures with DiI-acyl-LDL demonstrates their endothelial character (Fig. 1D) and allows their identification within the cocultures described below (Fig. lE,F). Within 3 days following addition to 1-day-old cultures of C, cells, the BRE cells reorganized into linear, curvilinear, and bifurcating arrays to form thick cords and Culture of endothelial cells with thin capillary-like structures (Fig. 1B,E). These astroconditioned media glial-induced structures varied in thickness and length Media from cocultures generating capillary-like but contained numerous 10-20 pm diameter segments structures, or media from solo C, cell cultures, was of up to 2 mm in length, similar to the diameter of stored at -2O"C, and later replenished with fresh microvessels in vivo. Neonatal astrocytes similarly inascorbic acid and 50 ugiml Endothelial Mitogen, and duced endothelial reorganization into capillary-like added to 1-day-old solo cultures of endothelial cells in structures (Fig. lC,F). A high magnification montage fibronectin treated four chamber slides. After 3 days of illustrating the full length of one of these is demonincubation cultures were assessed by phase contrast strated in Figure 2. Nonendothelial cells used in these studies failed to accumulate the endothelial cell microscopy for alterations in endothelial morphology. marker DiI-LDL (data not shown). We evaluated the temporal sequence of morphologic Labeling of cells with fluorescent m a r k e r s change during the formation of these astroglial-inCocultures or controls were rinsed with serum-free duced capillary-like structures. After coculture for 24 media and incubated for 2 hours a t 37°C with DiI-acylhours, the endothelial cells were well spread, cobbleLDL (1:20 dilution in DME). Specimens were then eistoned and similar to endothelial cells cultured alone ther fixed in 3.7% formaldehyde for 15 minutes a t 4°C By 48 hours they began to develop a linear (Fig. 3A,E). or further processed for indirect immunofluorescence organization and although the majority of BRE cells to detect extracellular laminin. The latter specimens continued to remain well spread, some were elongated were incubated sequentially in 1%normal goat serum as and less adherent to the tissue culture substratum (Organon Teknika Corp.), rabbit anti-laminin (1:80; evidenced by a n elevated plane of focus relative to Gibco), and fluorescein isothiocyanate conjugated goat neighboring cells (Fig. 3B,F). At 72 hours the reorgaanti-rabbit immunoglobulin (130; Cooper Biomedical, Inc.) for 1 hour each a t 37"C, fixed in formaldehyde, nizing endothelial cells were transformed from the preand mounted. Controls in which nonimmune serum dominantly well spread morphology (Fig. 3B,F), to linwas substituted for anti-laminin demonstrated mini- early aligned, self-associated forms raised above the plane of the surrounding astroglial monolayer (Fig. mal background staining (data not shown). 3C,G). After 72-96 hours of coculture, mature capillary-like structures composed of numerous cells could Microscopy and computerized image analysis be seen (Fig. 3D,H). This sequence of events suggests Fixed specimens were viewed both by phase contrast a n astroglial-induced transition from preferential enand fluorescence techniques using a Leitz Laborlux mi- dothelial adhesion to the tissue culture substratum to croscope and photographed using Kodak T-Max 400 preferential endothelial self-association into capillaryfilm. Unfixed specimens were assessed by phase con- like structures. trast techniques using a n inverted Olympus CK2 miThe morphological changes associated with astrocroscope. glial-induced capillary-likestructure formation are conQuantitation of the induction of capillary-like struc- sistent with a specific process of endothelial differentitures was performed with the Microcomputer Imaging ation. We wished to confirm, however, that this was not Device (MCID) software package of Imaging Research a nonspecific cell sorting phenomenon in response to Inc. (Brock University, St. Catherines, Ontario Canada the physical-spatial constraints of the coculture sysL2S 3A1), a Sierra Scientific High Resolution CCD tem. When cells of the normal rat kidney fibroblast line camera and Compaq DiskPro 386125 computer. For (NRK) were substituted for astroglial cells, capillaryeach experimental condition a t least four representa- like structures failed to form (Fig. 4).Instead these tive low magnification fields (3.37 mm'/field) were pho- NRK-BRE cocultures resulted in endothelial segregatographed and digitized from the developed negative. tion into well spread, densely populated oval islands. The cummulative length of capillary-like structures per field was then quantitated by manually tracing Requirements for astroglial induction of their course using the the special functions two-point capillary-like structures distance format. The conditions necessary for astroglial induction of RESULTS these capillary-like structures were further defined and the results are presented in Figures 5 and 6 and Induction of capillary-like structures Table 1. Fixed numbers of HRE cells were inoculated BRE cells develop a typical cobblestoned appearance into C, cultures established a t various cell densities 18 when cultured on fibronectin-coated tissue culture hours earlier. The subsequent endothelial response plastic (Fig. 1A,D). Confluent cultures, consistent with was determined by phase and fluorescence microscopy

ASTROCYTE INDUCTION OF ANGIOGENESIS IN VITRO

207

Fig. 1. Induction of capillary-like structures by rat brain astrocytes and C , cells. Stock cultures of rat brain astrocytes, C, cells and bovine retinal microvascular endothelial (ERE) cells were obtained as described in Materials and Methods. On day zero, astrocytes or C, cells were plated on fibronectin-coated cell culture plastic. On day 1, culture media was removed and BRE cells were added to astroglial cultures and to fibronectin-treated surfaces without astroglial cells. On day 3, cultures were incubated with the endothelial specific fluorcscent probe DiI-acyl-LIIL for 2 hours at 37"C, fixed in 3.7% formalde-

hyde, and identical fields photographed under phase (A-C) and fluorescence (D-F) microscopy. A,D: Endothelial cells alone develop a characteristic cobblestoned monolayer with some evidence of sprouting but no capillary-like structures are seen. B,E: Following coculture with C, cells, BRE cells associate into linear cords of varying thickness with numerous capillary-like segments of 10-20 pm diameter (arrows). C,F: Coculture with normal rat brain astrocytes induces similar capillary-like structures (arrows). Bar = 100 pm.

and quantitated by computerized image analysis as described in the Materials and Methods. The extent of BRE cell reorganization is shown in Figure 5 to be dependent upon the density of Ce cells within the cocultures with the greatest cummulative length of capillary-like structures developing in cocultures estab-

lished with C, cell densities of 5,000-10,000 cells/ chamber (3,000-6,000 cells/cm2). In cocultures established with densities of less than or equal to 1,500 C6 cells/cm2, BRE cells predominantly formed islands with cobblestoned morphology similar to that seen in solo BRE cultures. These results indicate that BRE cell

208

LATERRA ET AL.

Fig. 2. Full-length composite of capillary-like structure. Coculture of BRE and C, cells were established, stained with the endothelial specific fluorescent marker DiI-acyl-LDL, and processed as outlined in Figure 1. Composite phase contrast (A) and fluorescence (B) photomicrographs were constructed from consecutive overlapping fields. A: Numerous endothelial cells are organized into a 1.3 mm long capillary-like structure surrounded by C, cells. The endothelial cells

composing this structure frequently change planes relative to the surrounding astroglial monolayer. B: DiI-acyl-LDL staining distinguishes the endothelial cells from surrounding astroglia. This fluorescent marker accumulates around endothelial cell nuclei some of which are indicated by arrows (A and B). Open arrows (A) mark beginning and end of structure. Bar = 50 pm.

differentiation is dependent upon a minimal astroglial density presumably required for the generation of ample amounts of a n astroglial-derived inducer of capillary formation. Conversely, the dependence of capillary-like structure formation upon BRE cell density was also examined and the results presented in Figure 6. Maximum response was observed in cocultures established with 20,000-80,000 BRE celldchamber (12,000-48,000 cellsicm'). The mechanism which leads to this endothelial cell response may require endothelial-astroglial contact, close endothelial-astroglial apposition, or be mediated by diffusible substances t h a t are stable over relatively long distances. To distinguish between these possibilities, BRE and C, cells were separated by a 0.4 pm pore size membrane so that diffusible cell products could equilibrate throughout shared media in the absence of heterologous cell-cell contact. This culture system separates the endothelial and glial cells by approximately 1mm. Under these conditions both cell types were able to grow to confluence and BRE cells demonstrated a

cobblestoned morphology without evidence of morphologic differentiation. This finding indicates that very close endothelial-astroglial approximation is required for induction of capillary-like structures. However, this result is not inconsistent with the possibility that a diffusible inducer of endothelial differentiation results from a n initial close endothelial-astroglial interaction. To address this possibility, conditioned media from a typical capillary-generating coculture was applied to 1-day-old solo endothelial cultures. Under these conditions the endothelial cells reached confluence within 3 days, maintained a cobblestoned morphology, and failed to generate capillary-like structures. All BRE cell stock and cocultures were grown in the presence of Endothelial Mitogen, a preparation of fibroblast growth factor (Maciag et al., 1979). To determine whether this factor is required for endothelial differentiation, cocultures were prepared under identical conditions with or without addition of Endothelial Mitogen. A marked decrease in BRE reorganization was seen in the absence of mitogen. Although this de-

ASTROCPTE INDCCTION OF ANGIOGENESIS IN VITRO

Fig. 3 . Temporal sequence of astroglial induction of capillary-like structures. Cocultures of BRE cells (large arrows) and C, cells (small arrows) were prepared and parallel samples stained with the endothelial specific marker DiI-acyl-LUL. At 24, 48,72, or 96 hours following the addition of BRE cells, the cultures were fixed and identical fields photographed using phase contrast (A-D) and fluorescence microscopy (E-HI. At 24 hours (A and El, BRE cells are well spread and morphologically indistinguishable from endothelial cells of solo cultures (not shown). Astroglial cells lack fluorescence. At 48 hours

209

iB and F), BRE cells are aligned but remain predominantly well spread except for a n occasional cell that has elongated and become raised above the monolayer surface. By 72 hours (C and GI, the BRE cells have elongated and become associated in a n end-to-end orientation predominantly above the astroglial monolayer. At 96 hours (D and H) the BRE cells have completely reorganized into a capillarylike structure consisting of end-to-end overlapping endothelial cells above the nearly confluent astroglial monolayer. Bar = 25 Km.

210

LATERRA ET AL.

Fig. 4. Absence of capillary-like structures in endothelial-normal rat kidney fibroblast (NRK) cocultures. Cocultures were established as described in Figure 1with the exception that NRK cells were used in place of C, cells. A: Phase contrast micrograph demonstrates is-

lands of BRE cells but no capillary-like structures among the surrounding NRK cells (arrows). B: Fluorescent micrograph of identical field distinguishes the segregated BRE cells which stain specifically with DiI-acyl-LDL (arrows). Bar = 125 bm.

pendence on Endothelial Mitogen was most pronounced in cocultures established with 540 x lo3 BRE cells/chamber it was observed for all cell densities examined.

Extracellular matrix deposition Central nervous system capillaries, like those of peripheral tissues, are separated from surrounding parenchyma by a well organized basement membrane that is unique by certain morphologic and biochemical criteria (Schmidley, 1987; Timpl and Dziadek, 1986). Its formation is dependent upon the stoichiometric association of type IV collagen, laminin, heparan proteoglycan, and a large number of other less well characterized proteins. Basement membrane formation is one of the last steps of capillary formation in vivo (Bar and Wolff, 1972; D’Amore and Thompson, 1987; Furcht, 1986). Therefore the presence of basement membrane formation by these in vitro glial-induced capillary-like structures would represent an additional aspect of differentiation relevant to angiogenesis in vivo. Threeday-old solo BRE, solo C,, and cocultures were treated with DiI-Acyl-LDLand then processed prior to fixation for indirect immunofluorescent detection of extracellular laminin. Solo endothelial cultures demonstrated a patchy, flocculent distribution of laminin beneath endothelial cells (Fig. 7A-C) as well as on the tissue culture substratum exposed between cells (not shown). An identical pattern of laminin staining was seen in regions of the astroglial-endothelial cocultures where the endothelial cells had not differentiated but maintained a cobblestone morphology (not shown). Solo C, cultures demonstrated a faint punctate pattern of laminin staining that in subconf luent cultures appeared to be at the margins of cells (Fig. 7D-F, arrowheads within regions negative for DiI-acyl-LDL staining). In contrast, the capillary-like structures within cocultures were frequently associated with a dense linear accumulation of laminin that localized around the endothelial cells with the most intense staining along the en-

T e

2M)o

Y

E

ziE L h

3

loo0

0

1.25

I 2.5

5

10

20

NUYlER OF C6 CELLS ADDED X W . oo1

Fig. 5. Dependence on astroglial density of capillary-like structure formation. C,-BRE cell cocultures were established as previously described using C, densities of 1,250-20,000 cellskhamber and a fixed number (20,0OO/chamber)of BRE cells, stained with the endothelial specific marker DiI-acyl-LDL and fixed. The cumulative lengths of capillary-like structures per low power field were determined by computer assisted image analysis as outlined in the Materials and Methods. Each bar represents the mean of four representative fields t standard error.

dothelial-astroglial interface (Fig. 7D-F, arrows). This accumulation of a major basement membrane-associated protein is consistent with incipient basement

ASTROCYTE IYDUCTION OF ANGIOGENESIS IN VITRO

3000

211

A

T

T 2000

1000

C

L1 5

10

20

80

40

NUMBER OF BRE CELLS ADDED X .001

I 120

Fig. 6 . Dependence on BRE cell density of capillary-like structure formation. C,-BRE cell cocultures were established as previously described using fixed C, densities (10,000 cells/chamber) and BRE densities of 5,000-120,000 cellsichamber, stained with the endothelial specific marker DiI-acyl-LDL and fixed. A The cumulative lengths of capillary-like structures were determined as in Figure 5. Each bar represents the mean of four representative fields f standard error. B:

Fluorescent micrograph depicts the degree of BRE reorganization typically seen in cocultures established with 10,000 C, and 40,000 BRE cells/chamber. C: Fluorescent micrograph depicts the relative lack of capillary-like structure formation seen in cocultures established with 10,000 C, and 5,000 BRE cells/chamber. Bar (B and C) represents 500

TABLE 1. Effect of culture conditions on capillary-like structure formation

Goldstein and Betz, 1986). This anatomical endothelial-astrocyte relationship indicates a specific parenchymal response to neural microvessels that is not seen in peripheral tissues. Figure 8 demonstrates three-dimensional astroglial reorganization in response to capillary-like structures. The C , cells within this coculture have formed a confluent monolayer as demonstrated in Figure 8A (plane-of-focus on monolayer). Above this monolayer is a segment of a longer capillary-like structure consisting of approximately 5-6 endothelial cells aligned end-to-end with a diameter of approximately 12 pm (Fig. 8B,C). DiI-acyl-LDL negative C, cells can be seen to accumulate for up to 20 microns on either side of the endothelial margins in a plane-of-focus higher than the surrounding monolayer (Fig. 8B,C, between hollow arrows). A similar accumulation of astroglia can be seen along segments of the capillary-like structure in Figure 2 and in capillary-like structures induced by rat brain astrocytes (not shown). C6 cells do not accumulate in a similar fashion where they appose BRE cells that have maintained their cobblestoned morphology (not shown). This capillary-associated astroglial reorganization suggests the presence of specific

Endothelial culture condition' Coculture2 with 23000 astroglia/cm2 Coculture with 51500 astroglia/cm2 Coculture with NRK cells Coculture with astroglial-endothelial separation Solo culture with astroglial conditioned media Solo culture with coculture conditioned media Coculture without endothelial mitogen

Induction of capillary-like structures

+-

+ (diminished)

'Bovine retinal microvascular endothelial cells were grown on fibronectin coated matrices in the presence of endothelial mitogen, unless otherwise indicated. 'Coculture implies heterologous cell-cell contact unlcss otherwise stated. Cell Nos. indicate initial No. of cells added per surface area.

membrane formation. Capillary-like structures could be found within the same coculture either associated or unassociated with this extracellular matrix pattern.

Astroglial response to capillary-like structures Cerebral microvessels are ensheathed by foot processes of surrounding astrocytes (Bar and Wolff, 1972;

)Lm.

212

LATERRA ET AL

Fig. 7. Laminin deposition associated with capilllary-like structures. Solo BRE (A-C) and mixed C,-BRE (D-F) cultures were prepared and a t day 3 double labeled for endothelial cells iB and E) and extracellular laminin (C and F) a s described in Materials and Methods. A-C, D-F Identical fields within the ERE solo and C,-BRE cocultures, respectively. Extracellular laminin accumulates in the solo ERE cultures in a patchy flocculent pattern that is predomi-

nantly on the tissue culture surface both beneath (C, open arrow) and on exposed tissue culturc plastic between (not shown) the well spread cells. Within the cocultures, laminin is markedly concentrated at the interface between the C, cells and the segment of a capillary-like structure (D-F, arrows). Solo cultures of C, cells (not shown) demonstrate faint punctate laminin staining identical to that associated with C, cells within the cocultures (F, arrowheads). Bar = 50 pm.

endothelial-astroglial interactions that may represent a n in vilro process relevant to capillary ensheathment by astrocytes in vivo.

cells reorganize into capillary-like structures in response to coculture with normal rat brain astrocytes or C , cells of astroglial lineage. These astroglial-induced capillary-like structures are shown to mimic in vivo cerebral microvessels by light microscopic criteria of size and shape, their associated pattern of extracellular matrix deposition, and their “ensheathment” by adja-

DISCUSSION In this report we describe a novel in vitro system in which monolayer neural microvascular endothelial

ASTROCYTE INDUCTION OF ANGIOGENESIS IN VITRO

Fig. 8. Astroglial response to capillary-like structures. A C,-BRE cell coculture was prepared, labeled at 3 days with the fluorescent endothelial cel I-specific marker DiI-acyl-LDL and processed for microscopy. A: Phase contrast microscopy with plane-of-focus on the confluent C, cell monolayer. The overlying segment of a longer capillary-like structure is faintly visible as unfocused streaks (arrow). B: Phase contrast microscopy of the same field with the plane-of-focus on the overlying capillary-like structure. The lateral margins of this structure are indicated with arrowheads. Surrounding the differentiated endothelial cells are numerous C, cells elevated above the surrounding monolayer and extending as far as 20 microns from the endothelial margins (between open arrows). C: Fluorescence rnicroscopy of the same field as A and B localizes the endothelial cells within this three-dimensional astroglial-endothelial complex. Bar 25 +m. ~

cent astroglia. Furthermore, they closely resemble a t the light microscopic level endothelial structures

213

which have been described in other in vitro systems of “angiogenesis” (Feder et al., 1983; Ingber and Folkman, 1989b; Kubota et al., 1988). Our findings described here support the long-standing belief that perivascular astrocytes have the capacity t o regulate capillary development within the central nervous system. The formation of these capillary-like structures is associated with a transition from preferential endothelial adhesion to the tissue culture substratum to an end-to-end endothelial-endothelial interaction. This suggests the presence of concurrent alterations in endothelial cell-cell and cell-substratum adhesive mechanisms. We propose two possible biochemical explanations for this morphologic transition. Astroglial cell products or astroglial-induced endothelial cell products may modify the fibronectin-containing tissue culture substratum to make it less adhesive, resulting in a passive transition from cobblestoned endothelial morphologies to capillary-like structures. In fact, Ingber and Folkman (198913) have recently demonstrated that capillary-like structure formation in vitro may be modulated by altering the degree of interaction between endothelial cells and the tissue culture surface. Alternatively, the transition to capillary-like forms may result from changes in the expression of endothelial cell surface adhesion molecules. For instance, one cell surface extracellular matrix receptor (e.g., the integrin fibronectin receptor) may predominate on the endothelial cell in the solo cultures, while an endothelial-endothelial cell adhesion molecule may become upregulated as a result of endothelial culture within the astroglial environment. Such modulation of cell adhesion molecules has recently been shown to play a role in the segregation of sarcoma lines in vitro (Friedlander et al., 1989). Uvomorulin, a calcium-dependent adhesion molecule, and desmosome-associatedproteins mediate the formation of epithelial intercellular junctional complexes (Gumbiner et al., 1988; Boyer et al., 1989). Neural microvascular endothelial cells, which resemble epithelia by virtue of their polarity and extensive tight junctions, may use similar adhesion molecules to mediate the rearrangements described in this report. Endogenously synthesized laminin may also play a role in this endothelial reorganization since a specific domain on the laminin B1 chain has been implicated in the formation of capillary-like structures on Matrigel (Grant et al., 1989). However, preliminary experiments with the laminin YIGSR peptide, which blocks Matrigel-induced capillary formation, have not resulted in inhibition in this coculture system. The basis of this difference is presently being investigated. That this endothelial cell response requires a minimal number of cocultured astroglia is consistent with the generation of an astroglial cell-derived “inducer.” The time required for the development of capillary-like structures in this coculture system (48-72 hours) is intermediate between that required for tube formation in solo endothelial cell cultures following withdrawal of growth factors (weeks) or that needed for endothelial cells to form tubes in response to Matrigel (hours) as described by others (Feder et al., 1983; Folkman and Haudenschild, 1980; Kubota et al., 1987). This time frame is also consistent with the endogenous accumulation of inductive factors in these cocultures.

214

LATERRA ET AL.

The inability of astroglial conditioned media, coculture conditioned media or cocultures of physically separated cells to generate capillary-like structures demonstrates that this endothelial response is not likely t o result solely from the actions of a diffusible substance. Astroglial products such a s interleukin-1 (Giulian et al., 1988; Lieberman et al., 1989) o r vascular permeability factor (Keck et al., 1989), both of which have been shown to be angiogenic in vivo, may play a role. Endothelial Mitogen, a preparation of fibroblast growth factor (Maciag et al., 1979) appears to enhance the response. Thus, the influence of mitogen and intimate astroglial-endothelial interaction, if not frank astroglial-endothelial contact, simultaneously influence the formation of capillary-like structures in this experimental system. These results support the findings of Folkman (19861,Madri et al. (19881,and others (Furcht, 1986; Hockel et al., 1987) which implicate the coincident influence of both soluble and insoluble factors during capillary formation. The data presented cannot distinguish between a requirement for direct heterologous cell-cell contact and the effects of astroglial-dependent nondiffusible extracellular matrix molecules. Both perivascular ensheathment by astrocytes and basement membrane formation occur relatively late in the morphogenesis of mature cerebral microvessels (Bar and Wolff, 1972), and these events coincide both temporally and spatially with the expression of bloodbrain barrier endothelial properties in vivo (Coomber and Stewart, 1986: Stewart and Hayakawa, 1987). The deposition of the basement membrane glycoprotein, laminin, is demonstrated in this report to change dramatically in association with the astroglial induction of capillary-like structures. Laminin is shown at the resolution of light microscopy to concentrate around capillary-like structures a t the astroglial-endothelial interface. This is consistent with the incipient organization of basement membrane material. In addition, astroglia are shown to accumulate above the surrounding astroglial monolayer in response to the immediately adjacent capillary-like structures. This astroglial response may result from a haptotactic, chemotactic or mitogenic signal derived from the differentiated endothelial cells. Platelet-derived growth factor of endothelial origin (DiCorleto and Bowen-Pope, 1983), which has been shown by Bressler and colleagues (1985) to be chemotactic to astrocytes, or laminin, which is shown here t o acchmulate at the endothelial-astroglial interface may be responsible in part for this response. This apparent ability of capillary-like structures to influence astroglial behavior is evidence for bidirectional signalling between the differentiated endothelial cells and astroglia and may be relevant to 3-dimensional tissue morphogenesis. In summary, we demonstrate that perivascular astrocytes induce neural microvascular endothelial cells to form capillary-like structures in vitro. These findings indicate that astrocytes are likely to play a n important role in regulating the vascularization of neural tissue such as retina and brain in vivo. Studies to identify the molecular basis of this endothelial response and to define in more detail the degree of endothelial differentiation achieved in this system, with specific reference to lumen formation and expression of neural microvessel-specific endothelial markers, are under-

way. This experimental system of neural “angiogenesis” can be utilized to determine the molecular basis of humoral factors, cell-cell and cell extracellular matrix interactions important to neural microvessel formation and differentiation.

ACKNOWLEDGMENTS The authors thank Ravi Indurti for expert technical assistance, Drs. Amy Fulton and Joseph Bressler for critical discussions, and Susan Cascio for help in preparing the manuscript. This work was supported by National Institutes of Health Research Grants ES02380-10 and EY-03772-10 (G.W. Goldstein, P.I.). J . Laterra is a Clinical Investigator Development Awardee of the National Institute of Neurological Disorders and Stroke iK08-NS-01329-01). c. Guerin is a National Research Service Awardee (CA-09574-03). LITERATURE CITED Auerbach, R., Lu, W.C., Pardon, E., Gumkowski, F., Kaminska, G., and Kaminski, M. (1987) Specificity of adhesion between murine tumor cells and capillary endothelium: an in uitro correlate of preferential metastasis in uivo. Cancer Res., 47:1492-1496. Ausprunk, D.H., Knighton, D.R., and Folkman, J. (1974) Differentiation of vascular endothelium in the chick chorioallantois: a structural and autoradiographic study. Dev. Biol., 38.237-248. Bar, T.H., and Wolff, J.R. (1972) The formation of capillary basement membranes during internal vascularization of the rat’s cerebral cortex. Z. Zelforsch., 133:231-248. Beck, D.W., Vinters, H.V., Hart, M.N., andcancilla, P.A. (1984) Glial cells influence polarity of the blood-brain barrier. J . Neuropathol. Exp. Neurol., 43:219-224. Benda, P., Lightbody, J., Sato, G., Levine, L., and Sweet, W. (1968) Differentiated rat glial cell strain in tissue culture. Science, 161: 370-371. Betz, A.L., and Goldstein, G.W. (1978) Polarity of the blood-brain barrier: neutral amino acid transport into isolated brain capillaries. Science, 202:225-227. Bissel, M.G., Rubinstein, L.J., Bignami, A,, andHerman, M.M. (1974) Characteristics of the rat C6 glioma maintained in organ culture systems. Production of glial fibrillary acidic protein in the absence of gliofibrillogenesis. Brain Res., 82:77-89. Bowman, P.D., Betz, A.L., and Goldstein, G.W. (1982) Primary culture of microvascular endothelial cells from bovine retina: selcctive growth using fibronectin-coated substrate and plasma-derived serum. In Vitro 18t626-632. Boycr, B., Tucker, G.C., Valles, A.M., Franke, W.W., and Thiery, J.P. (1989) Rearrangements of desmosomal and cytoskeletal proteins during the transition from epithelial to fibroblastoid organization in cultured rat bladder carcinoma cells. J. Cell Biol., 109t14951510. Bressler, J.P., Grotendorst, G.R., Levitov, C., and Hjelmeland, L.M. (1985)Chemotaxis of rat brain astrocytes to platelet derived growth factor. Brain Res., 344t249-254. Brightman, M.W., and T.S. Reese. (1969) Junctions between intimately apposed cell membranes in the vertebrate brain. J. Cell Biol., 4Ot648-677. Butcher, E.C., Scollary, R.G., and Weissman, I.L. (1980) Organ specificity of lymphocyte migration: mediation by highly selective lymphocyte interaction with organ-specific determinants on high endothelial venules. Eur. J. Immunol., 20556-561. Coomber, B.L., Stewart, P.A. (1986)Three dimensional reconstruction of vesicles in endothelium of blood-brain barrier versus highly permeable microvessels. Anat. Rec., 215:256-261. Coomber, B.L., and Stewart, P.A. (1985) Morphometric analysis of cns microvascular endothelium. Microvascular Res., 30t99-115. D’Amore, P.A., and Thompson, R.W. (1987) Mechanisms of angiogenesis. Annu. Rev. Physiol., 49:452-464. DeBault, L.E., and Cancilla, P.A. 11980) Gamma-glutamyl transpeptidase in isolated brain endothelial cells: induction by glial cells in uitro. Science 207t653-655. DiCorleto, P.E., and Bowen-Pope, D.F. (1983) Cultured endothelial cells produce a platelet-derived growth factor-like protein. Proc. Natl. Acad. Sci. USA, 80:1919-1923. Feder, J., Marasa, J.C., and Olander, J.V. (1983) The formation of

ASTROCYTE INDUCTION OF ANGIOGENESIS IN VITRO

capillary-like tubes by calf aortic endothelial cells grown in uitro. J. Cell Physiol., II6tl-6. Folkman, J., and Haudenschild, C.C. (1980) Angiogenesis in uitro. Nature, 288t551-556. Folkman, J. (1986) How is blood vessel growth regulated in normal and neoplastic tissue? Cancer Res., 46t467-473. Franzakis, M.V., and Kimelberg, H.K. (1984) Dissociation of neonatal rat brain by dispase for preparation of primary astrocyte cultures. Neurochem Res., 9:1689-1698. Friedlander, D.R., Mege, R.M., Cunningham, B.A., and Edelman, G.M. (1989) Cell sorting-out modulated by both the specificity and amount of different cell adhesion molecules (CAMS! expressed on cell surfaces. Proc. Natl. Acad. Sci. USA, 86:7043-7047. Furcht, L.T. (1986) Critical factors controlling angiogenesis: cell products, cell matrix and growth factors. Lab Invest., 55:505-509. Gimbrone, M.A., Cotran, R.S., Leapman, S.B., and Folkman, J. 11974) Tumor growth and neovascularization: an experimental model using the rabbit cornea. J. Natl. Cancer Inst., 52:413-427. Giulian, D., Woodward, J., Young, D.G., Krebs, J.F., and Lachman, L.B. (1988) Interleukin-1 injected into mammalian brain stimulates astrogliosis and neovascularization. J . Neurosci., 8t24852490. Goldstein, G.W., and Betz, A.L. (1986) The blood-brain barrier. Sci. Am., 254t74-83. Grant, D.S., Tashiro, K.I., Sergui-Real, B., Yamada, Y., Martin, G.R., and Kleinman, H.K. (1989)Two different laminin domains mediate the differentiation of human endothelial cells into capillary-like structures in vitro. Cell, 58t933-943. Gumbiner, B., Stevenson, B., and Grimaldi, A. (1988) The role of the cell adhesion molecule uvomorulin in the formation and maintenance of the epithelial junctional complex. J . Cell Biol., 107:15751587. Herbst, T., McCarthy, J.B., Tsilibary, E.C., and Furcht, L.T. (1988) Differential effects of laminin, intact type IV collagen, and specific domains of type IV collagen on endothelial cell adhesion and migration. J . Cell Biol., 106:1365-1373. Hockel, M., Sasse, J., and Wissler, J.H. (1987) Purified monocytederived angiogenic substance (angiotropin) stimulates migration, phenotypic changes, and “tube formation” but not proliferation of capillary endothelial cell in uitro. J. Cell. Physiol., 133:l-13. Ingber, D.E., and Folkman, J. (1989a) How does extracellular matrix control capillary morphogenesis? Cell, 58t803-805. Ingber, D.E., and Folkman, J . (1989b) Mechanochemical switching between growth and differentiation during fibroblast growth factorstimulated angiogenesis in uitror role of extracellular matrix. J . Cell Biol., 109t317-330. Janzer, R.C., and Raff, M.C. (1987)Astrocytes induce blood-brain barrier properties in endothelial cells. Nature, 325:253-257. Joo, F. (1987) Current aspects of the development of the blood-brain barrier. Int. J. I k v . Neurosci., 5t369-372. Keck, P., Hauser, S.D., Krivi, G., Sanzo, K., Warren, T., Feder, J., and Connolly, D.T. (1989) Vascular permeability factor, a n endothelial cell mitogen related to PDGF. Science, 246:1309-1312. Kubota, Y., Kleinman, H.K., Martin, G.R., and Lawley, T.J. (1988) Role of laminin and basement membrane in the morphological differentiation of human endothelial cells into capillary-like structures. J. Cell B i d , 107:1589-1598.

215

Kumar, S., Holmes, E., Scully, S., Birren, B.W., Wilson, R.H., and deVellis, J . (1986) The hormonal regulation of gene expression of glial markers: glutamine synthetase and glycerol phosphate dehydrogenase in primary cultures of rat brain and in C6 cell lines. J. Neurosci. Res., 16:251-264. Laterra, J., Stewart, P., and Goldstein, G. (in press) Development of the blood-brain barrier. In: Neonatal and Fetal Medicine-Physiology and Pathophysiology. R.A. Polin and W.W. Fox, eds. Saunders, Philadelphia. Lieberman, A.P., Pitha, P.M., Shin, H.S., and Shin, M.L. (1989) Production of tumor necrosis factor and other cytokines by astrocytes stimulated with lipopolysaccharide or a neurotropic virus. Proc. Natl. Acad. Sci. USA, 86t6348-6352. Ling, T., and Stone, J. (1988) The development of astrocyte in the cat retina: evidence of migration from the optic nerve. Dev. Brain Res., 44t73-85. Maciag, T., Cerundolo, J.,Ilsey, S., Kelley, P.R., and Forand, R. (1979) An endothelial cell growth factor from bovine hypothalamus: identification and partial purification. Proc. Natl. Acad. Sci. USA, 761 5674-5678. Maciag, T., Kadish, J . , Wilkins, L., Stemerman, M.B., and Weinstein, R. (1982) Organizational behavior of human umbilical vein endothelial cells. J . Cell Biol., 94:511-520. Madri, J.A., Pratt, B.M., and Tucker, A.M. (1988) Phenotypic modulation of endothelial cells by transforming growth factor-B depends upon the composition and organization of the extracellular matrix. J. Cell Biol., 106:1375-1384. McCarthy, K.D., and de Vellis, J. (1980) Preparation of separate astroglial and oligodendroglial cultures from rat cerebral tissue. J. Cell Biol., 852390-902. Netland, P.A., Zetter, B.R., Via, D.P., and Voyta, J.C. (1985) I n situ labelling of vascular endothelium with fluorescent acetylated low density lipoprotein. Histochem. J., I7t1309-1320. Reese, T.S., and Karnovsky, M.J. (19671 Fine structural localization of a blood-brain barrier to exogenous peroxidase. J. Cell Biol., 34; 207-217. Risau, W., Hallman, R., Albrecht, V., and Henke-Fahle, S. (1986) Brain induces the expression of a n early cell surface marker for blood-brain barrier-specific endothelium. EMBO J., 5:3179-3183. Schmidley, J.W. (1987) Ultrastructural studies of basement membrane of microvessels isolated from rat brain: lack of staining with ruthenium red. Microvasc. Res., 33:417-421. Stewart, P.A., and Hayakawa, E.M. (1987) Interendothelial junctional changes underlie the developmental “tightening” of the blood-brain barrier. Dev. Brain Res., 32:271-281. Stewart, P.A., and Wiley, M.J. (1982) Developing nervous tissue induces formation of blood-brain barrier characteristics in invading endothelial cells: A study using quail-chick transplantation chimeras. Dev. Biol., 84r183-192. Tao-Cheng, J.H., Nagy, Z., and Brightman, M.W. (1987) Tight junctions of brain endothelium in vitro are enhanced by astroglia. J. Neurosci., 7t3293-3299. Timpl, R., and Dziadek, M. (1986) Structure, development and molecular pathology of basement membranes. Int. Rev. Exp. Pathol., 29: 1-112. Watanabe, T., and Raff, M.C. (1988) Retinal astrocytes are immigrants from the optic nerve. Nature, .332;834-837.