Cytoskeleton in TFG-beta- and bFGF-modulated endothelial monolayer repair

Endothelial cell proliferation and migration in vitro is depressed by transforming growth factor beta (TFG-beta) and enhanced by basic fibroblast growth factor (bFGF) treatment. This study examines interactions between cytoskeletal changes and cell proliferation in regenerating endothelial monolayers treated with bFGF, TFG-beta, and both factors. As previously described by others, monolayer regeneration is enhanced by bFGF and reduced by TFG-beta. Endothelial cell morphology is altered by TFG-beta treatment. Cells lose their cobblestone appearance and assume a pleomorphic shape. Actin microfilament staining is modified in both intact and regenerating TFG-beta-treated monolayers as well. There is a loss of dense peripheral band staining and an enhancement in staining intensity of cytoplasmic stress fibers. No such alterations are seen in bFGF-treated cultures. Cell proliferation at the wound edge, as indicated by bromodeoxyuridine incorporation, is inhibited by TGF-beta. Although monolayer repair is modulated by growth factor treatment, centrosome reorientation and microtubule staining patterns are not altered by either factor. Thus these factors appear to have effects on a mechanism(s) other than centrosome reorientation which may be involved in repair of denuded endothelial monolayers.     

Influence of mitomycin C on endothelial monolayer regeneration in vitro.

This study examines the effect of Mitomycin C, a fungal toxin which inhibits DNA synthesis, on the regeneration of partially denuded large vessel endothelium in vitro. Monolayers of bovine pulmonary artery endothelial cells were treated with Mitomycin C prior to or immediately following partial denudation and were incubated in the continuing presence of Mitomycin C; the effects of this treatment on monolayer repair, cell proliferation, and other aspects of endothelial phenotype were monitored. Cell proliferation, DNA, RNA, and protein synthesis were all reduced in a dose dependent manner in treated cultures. Incubation with Mitomycin C for 48 h or longer resulted in reduced cell spreading, and rounding up and loss of cells from both intact and partially denuded cultures. Effects were less severe with lower doses and shorter incubation times. However, significant reductions in monolayer regeneration occurred within 8 h of incubation, sufficiently early to suggest that Mitomycin C may affect aspects of the regeneration process independent of cell proliferation. Polarization/spreading of cells at the denudation edge was monitored by fluorescence staining for golgi with C5-DMB-ceramide, and for centrioles with antibodies to tubulin. Centrioles and golgi rapidly reoriented to a location at the putative leading edge of control cultures. Mitomycin C treatment had no effect on centriole reorientation, but caused a significant delay in golgi localization. These results suggest that Mitomycin C inhibits endothelial monolayer regeneration by mechanisms independent of cell proliferation and DNA synthesis, perhaps by interfering with cell spreading or translocation at the wound edge.     

The reorganization of microfilaments, centrosomes, and microtubules during in vitro small wound reendothelialization

The repair of small endothelial wounds is an important process by which endothelial cells maintain endothelial integrity. An in vitro wound model system was used in which precise wounds were made in a confluent endothelial monolayer. The repair process was observed by time-lapse cinemicrophotography. Using fluorescence and immunofluorescence microscopy, the cellular morphological events were correlated with the localization and distribution of actin microfilament bundles and vinculin plaques, and centrosomes and their associated microtubules. Single to four-cell wounds underwent closure by cell spreading while wounds seven to nine cells in size closed by initially spreading which was then followed at approximately 1 h after wounding by cell migration. These two processes showed different cytoskeletal patterns. Cell spreading occurred independent of centrosome location. However, centrosome redistribution to the front of the cell occurred as the cells began to elongate and migrate. While the peripheral actin microfilament bundles (i.e., the dense peripheral band) remained intact during cell spreading, they broke down during migration and were associated with a reduction in peripheral vinculin plaque staining. Thus, the major events characterizing the closure of endothelial wounds were precise in nature, followed a specific sequence, and were associated with specific cytoskeletal patterns which most likely were important in maintaining directionality of migration and reducing the adhesion of the cells to their neighbors within the monolayer.     

Microtubules regulate aortic endothelial cell actin microfilament reorganization in intact and repairing monolayers

To understand the role of microtubules and microfilaments in regulating endothelial monolayer integrity and repair, and since microtubules and microfilaments show some co-alignment in endothelial cells, we tested the hypothesis that microtubules organize microfilament distribution. Disruption of microtubules with colchicine in resting confluent aortic endothelial monolayers resulted in disruption of microfilament distribution with a loss of dense peripheral bands, an increase in actin microfilament bundles, and an associated increase of focal adhesion proteins at the periphery of the cells. However, when microfilaments were disrupted with cytochalasin B, microtubule distribution did not change. During the early stages of wound repair of aortic endothelial monolayers, microtubules and microfilaments undergo a sequential series of changes in distribution prior to cell migration. They are initially distributed randomly relative to the wound edge, then align parallel to the wound edge and then elongate perpendicular to the wound edge. When microtubules in wounded cultures were disrupted, dense peripheral bands and lamellipodia formation were lost with increases in central stress fibers. However, following microfilament disruption, microtubule redistribution was not disrupted and the microtubules elongated perpendicular to the wound edge similar to non-treated cultures. Microtubules may organize independently of microfilaments while microfilaments require microtubules to maintain normal organization in confluent and repairing aortic endothelial monolayers.     

Urokinase-type plasminogen activator mediates basic fibroblast growth factor-induced bovine endothelial cell migration independent of its proteolytic activity.

The dependence of urokinase-type plasminogen activator (uPA) induction on endogenous basic fibroblast growth factor (bFGF) activity during endothelial cell migration was investigated utilizing a combination of wounded endothelial cell monolayers and substrate overlay techniques. Purified polyclonal rabbit immunoglobulin G (IgG) against bFGF blocked the appearance of uPA-dependent lytic activity normally observed at the edge of a wounded bovine aortic endothelial (BAE) cell monolayer. Additionally, the migration of cells into the denuded area was inhibited 30-50% by antibodies either to bFGF or to bovine uPA. Incubation of wounded monolayers with either purified bovine uPA or agents able to induce PA activity, such as phorbol myristate acetate (PMA), vanadate, or bFGF, resulted in enhanced migration of cells (28-50%). Anti-bovine uPA IgG blocked a significant fraction (25%) of BAE cell migration induced by exposure to exogenous bFGF. The role of uPA in migration of wounded BAE cells was not dependent on plasmin generation. Furthermore, the amino terminal fragment (ATF) of human recombinant (hr) uPA, which is enzymatically inactive, stimulated endothelial cell movement in the presence of anti-bFGF IgG. These results suggest that BAE cell migration from the edge of a wounded monolayer is dependent upon local increases of uPA mediated by endogenous bFGF. Moreover, the data support the conclusion that migration is stimulated via a signalling mechanism dependent upon occupancy of the uPA receptor but independent of uPA-mediated proteolysis.     

Autocrine activities of basic fibroblast growth factor: regulation of endothelial cell movement, plasminogen activator synthesis, and DNA synthesis

We have found that the spontaneous migration of bovine aortic endothelial cells from the edge of a denuded area in a confluent monolayer is dependent upon the release of endogenous basic fibroblast growth factor (bFGF). Cell movement is blocked by purified polyclonal rabbit IgG to bFGF as well as affinity purified anti-bFGF IgG and anti-bFGF F(ab')2 fragments. The inhibitory effect of the immunoglobulins is dependent upon antibody concentration, is reversible, is overcome by the addition of recombinant bFGF, and is removed by affinity chromatography of the antiserum through a column of bFGF-Sepharose. Cell movement is also reversibly inhibited by the addition of protamine sulfate and suramin; two agents reported to block bFGF binding to its receptor. The addition of recombinant bFGF to wounded monolayers accelerates the movement of cells into the denuded area. Transforming growth factor beta which has been shown to antagonize several other effects of bFGF also inhibits cell movement. The anti-bFGF IgG prevents the movement of bovine capillary endothelial cells, BHK-21, NIH 3T3, and human skin fibroblasts into a denuded area. Antibodies to bFGF, as well as suramin and protamine sulfate also suppress the basal levels of plasminogen activator and DNA synthesis in bovine aortic endothelial cells.     

Microtubule-organizing centers and cell migration: effect of inhibition of migration and microtubule disruption in endothelial cells

We have previously shown that microtubule-organizing centers (MTOC's) become preferentially oriented towards the leading edge of migrating endothelial cells (EC's) at the margin of an experimentally induced wound made in a confluent EC monolayer. To learn more about the mechanism responsible for the reorientation of MTOC's and to determine whether a similar reorientation takes place when cell migration is inhibited, we incubated the wounded cultures with colcemid (C) and cytochalasin B (CB), which disrupt microtubules (MT's) and microfilaments (MF's), respectively. The results obtained showed that the MTOC reorientation can occur independent of cell migration since MTOC's reoriented preferentially toward the wound edge in the CB- treated cultures, even though forward migration of the EC was inhibited. In addition, the MTOC reorientation is inhibited by C, indicating that it requires an intact system of MT's and/or other intracellular structures whose distribution is dependent on that of MT's.     

Centrosome reorientation in wound-edge cells is cell type specific

The reorientation of the microtubule organizing center during cell migration into a wound in the monolayer was directly observed in living wound-edge cells expressing gamma-tubulin tagged with green fluorescent protein. Our results demonstrate that in CHO cells, the centrosome reorients to a position in front of the nucleus, toward the wound edge, whereas in PtK cells, the centrosome lags behind the nucleus during migration into the wound. In CHO cells, the average rate of centrosome motion was faster than that of the nucleus; the converse was true in PtK cells. In both cell lines, centrosome motion was stochastic, with periods of rapid motion interspersed with periods of slower motion. Centrosome reorientation in CHO cells required dynamic microtubules and cytoplasmic dynein/dynactin activity and could be prevented by altering cell-to-cell or cell-to-substrate adhesion. Microtubule marking experiments using photoactivation of caged tubulin demonstrate that microtubules are transported in the direction of cell motility in both cell lines but that in PtK cells, microtubules move individually, whereas their movement is more coherent in CHO cells. Our data demonstrate that centrosome reorientation is not required for directed migration and that diverse cells use distinct mechanisms for remodeling the microtubule array during directed migration.     

Loss of caveolin-1 polarity impedes endothelial cell polarization and directional movement

The ability of a cell to move requires the asymmetrical organization of cellular activities. To investigate polarized cellular activity in moving endothelial cells, human endothelial cells were incubated in a Dunn chamber to allow migration toward vascular endothelial growth factor. Immunofluorescent staining with a specific antibody against caveolin-1 revealed that caveolin-1 was concentrated at the rear of moving cells. Similarly, monolayer scraping to induce random cell walk resulted in relocation of caveolin-1 to the cell rear. These results suggest that posterior polarization of caveolin-1 is a common feature both for chemotaxis and chemokinesis. Dual immunofluorescent labeling showed that, during cell spreading, caveolin-1 was compacted in the cell center and excluded from nascent focal contacts along the circular lamellipodium, as revealed by integrin beta1 and FAK staining. When cells were migrating, integrin beta1 and FAK appeared at polarized lamellipodia, whereas caveolin-1 was found at the posterior of moving cells. Notably, wherever caveolin-1 was polarized, there was a conspicuous absence of lamellipod protrusion. Transmission electron microscopy showed that caveolae, similar to their marker caveolin-1, were located at the cell center during cell spreading or at the cell rear during cell migration. In contrast to its unphosphorylated form, tyrosine-phosphorylated caveolin-1, upon fibronectin stimulation, was associated with the focal complex molecule phosphopaxillin along the lamellipodia of moving cells. Thus, unphosphorylated and phosphorylated caveolin-1 were located at opposite poles during cell migration. Importantly, loss of caveolin-1 polarity by targeted down-regulation of the protein prevented cell polarization and directional movement. Our present results suggest a potential role of caveolin polarity in lamellipod extension and cell migration.     

Rapid and dynamic arginylation of the leading edge β‐actin is required for cell migration

β‐actin plays key roles in cell migration. Our previous work demonstrated that β‐actin in migratory non‐muscle cells is N‐terminally arginylated and that this arginylation is required for normal lamellipodia extension. Here, we examined the function of β‐actin arginylation in cell migration. We found that arginylated β‐actin is concentrated at the leading edge of lamellipodia and that this enrichment is abolished after serum starvation as well as in contact‐inhibited cells in confluent cultures, suggesting that arginylated β‐actin at the cell leading edge is coupled to active migration. Arginylated actin levels exhibit dynamic changes in response to cell stimuli, lowered after serum starvation and dramatically elevating within minutes after cell stimulation by readdition of serum or lysophosphatidic acid. These dynamic changes require active translation and are not seen in confluent contact‐inhibited cell cultures. Microinjection of arginylated actin antibodies into cells severely and specifically inhibits their migration rates. Together, these data strongly suggest that arginylation of β‐actin is a tightly regulated dynamic process that occurs at the leading edge of locomoting cells in response to stimuli and is integral to the signaling network that regulates cell migration.      

Centripetal transport of cytoplasm, actin, and the cell surface in lamellipodia of fibroblasts

Wound healing in Swiss 3T3 cultures was investigated with video-enhanced contrast (VEC) microscopy. The formation of protrusions at the leading edge of cells along wound was investigated in detail during the spreading stage, which usually lasted from 1 to 4 hr postwounding. Lamellipodia exhibited a continuous rearward, or centripetal, transport of a variety of cellular constituents at rates of approximately 0.26 microns/sec from the leading edge. The lamellipodia were also the sites of lateral migration as well as extension and retraction of actin microspikes. Actin fibers oriented transversely to the direction of movement were also observed to transport centripetally at similar rates. These fibers may in part give rise to large actin fibers forming at the interface between the base of the lamellipodia and the lamellae. Beads 0.5 microns in diameter attached to the dorsal surfaces of lamellipodia also transported centripetally at rates of approximately 0.21 microns/sec. Thus there is an apparent correlation between transport of a variety of structures within lamellipodia and with surface movements of lamellipodia.     

SPARC antagonizes the effect of basic fibroblast growth factor on the migration of bovine aortic endothelial cells.

Migration of endothelial cells is requisite to wound repair and angiogenesis. Since the glycoprotein SPARC (secreted protein, acidic and rich in cysteine) is associated with remodeling, cellular migration, and angiogenesis in vitro, we questioned whether SPARC might influence the motility of endothelial cells. In this study we show that, in the absence of serum, exogenous SPARC inhibits the migration of bovine aortic endothelial cells induced by bFGF. Similar results were obtained from two different assays, in which cell migration was measured in a Boyden chamber and in monolayer culture after an experimental wound. Without bFGF, the migration of endothelial cells was unaffected by SPARC. The inhibitory effect of SPARC on cell motility was dose-dependent, required the presence of Ca2+, was mimicked by synthetic peptides from the N- and C-terminal Ca(2+)-binding domains of the protein, and was not seen in the presence of serum. Modulation of the activities of secreted and cell-associated proteases, including plasminogen activators and metalloproteinases, appeared not to be responsible for the effects that we observed on the motility of endothelial cells. Moreover, a molecular interaction between SPARC and bFGF was not detected, and SPARC did not interfere with the binding of bFGF to high-affinity receptors on endothelial cells. Finally, in culture medium that contained serum, SPARC inhibited the incorporation of [3H]-thymidine into newly synthesized DNA, both in the absence and presence of bFGF. However, DNA synthesis was not affected by SPARC when the cells were plated on gelatin or fibronectin in serum-free medium. We propose that the combined action of a serum factor and SPARC regulates both endothelial cell proliferation and migration and coordinates these events during morphogenetic processes such as wound repair and angiogenesis.     

Cep70 contributes to angiogenesis by modulating microtubule rearrangement and stimulating cell polarization and migration

Centrosomal proteins intricately regulate diverse microtubule-mediated cellular activities, including cell polarization and migration. However, the direct participation of these proteins in angiogenesis, which involves vascular endothelial cell migration from preexisting blood vessels, remains elusive. Here we show that the centrosomal protein Cep70 is necessary for angiogenic response in mice. This protein is also required for tube formation and capillary sprouting in vitro from vascular endothelial cells. Wound healing and transwell assays reveal that Cep70 plays a significant role in endothelial cell migration. Depletion of Cep70 results in severe defects in membrane ruffling and centrosome reorientation, indicating a requirement for this protein in cell polarization. In addition, Cep70 is critically involved in microtubule rearrangement in response to the migratory stimulus. Our data further demonstrate that Cep70 is important for Cdc42 and Rac1 activation to promote angiogenesis. These findings thus establish Cep70 as a crucial regulator of the angiogenic process and emphasize the significance of microtubule rearrangement and cell polarization and migration in angiogenesis.     

Cdc42 and noncanonical Wnt signal transduction pathways cooperate to promote cell polarity

Scratch-induced disruption of cultured monolayers induces polarity in front row cells that can be visualized by spatially localized polymerization of actin at the front of the cell and reorientation of the centrosome/Golgi to face the leading edge. We previously reported that centrosomal reorientation and microtubule polarization depend on a Cdc42-regulated signal transduction pathway involving activation of the Par6/aPKC complex followed by inhibition of GSK-3beta and accumulation of the adenomatous polyposis coli (APC) protein at the plus ends of leading-edge microtubules. Using monolayers of primary rodent embryo fibroblasts, we show here that dishevelled (Dvl) and axin, two major components of the Wnt signaling pathway are required for centrosome reorientation and that Wnt5a is required for activation of this pathway. We conclude that disruption of cell-cell contacts leads to the activation of a noncanonical Wnt/dishevelled signal transduction pathway that cooperates with Cdc42/Par6/aPKC to promote polarized reorganization of the microtubule cytoskeleton.     

Regulation of vascular endothelial cell polarization and migration by Hsp70/Hsp90-organizing protein

Hsp70/Hsp90-organizing protein (HOP) is a member of the co-chaperone family, which directly binds to chaperones to regulate their activities. The participation of HOP in cell motility and endothelial cell functions remains largely unknown. In this study, we demonstrate that HOP is critically involved in endothelial cell migration and angiogenesis. Tube formation and capillary sprouting experiments reveal that depletion of HOP expression significantly inhibits vessel formation from endothelial cells. Wound healing and transwell migration assays show that HOP is important for endothelial cell migration. By examination of centrosome reorientation and membrane ruffle dynamics, we find that HOP plays a crucial role in the establishment of cell polarity in response to migratory stimulus. Furthermore, our data show that HOP interacts with tubulin and colocalizes with microtubules in endothelial cells. These findings indicate HOP as a novel regulator of angiogenesis that functions through promoting vascular endothelial cell polarization and migration.     

Pointed-end capping by tropomodulin3 negatively regulates endothelial cell motility

Actin filament pointed-end dynamics are thought to play a critical role in cell motility, yet regulation of this process remains poorly understood. We describe here a previously uncharacterized tropomodulin (Tmod) isoform, Tmod3, which is widely expressed in human tissues and is present in human microvascular endothelial cells (HMEC-1). Tmod3 is present in sufficient quantity to cap pointed ends of actin filaments, localizes to actin filament structures in HMEC-1 cells, and appears enriched in leading edge ruffles and lamellipodia. Transient overexpression of GFP-Tmod3 leads to a depolarized cell morphology and decreased cell motility. A fivefold increase in Tmod3 results in an equivalent decrease in free pointed ends in the cells. Unexpectedly, a decrease in the relative amounts of F-actin, free barbed ends, and actin-related protein 2/3 (Arp2/3) complex in lamellipodia are also observed. Conversely, decreased expression of Tmod3 by RNA interference leads to faster average cell migration, along with increases in free pointed and barbed ends in lamellipodial actin filaments. These data collectively demonstrate that capping of actin filament pointed ends by Tmod3 inhibits cell migration and reveal a novel control mechanism for regulation of actin filaments in lamellipodia.     

Involvement of the Arp2/3 complex and Scar2 in Golgi polarity in scratch wound models

Cell motility and cell polarity are essential for morphogenesis, immune system function, and tissue repair. Many animal cells move by crawling, and one main driving force for movement is derived from the coordinated assembly and disassembly of actin filaments. As tissue culture cells migrate to close a scratch wound, this directional extension is accompanied by Golgi apparatus reorientation, to face the leading wound edge, giving the motile cell inherent polarity aligned relative to the wound edge and to the direction of cell migration. Cellular proteins essential for actin polymerization downstream of Rho family GTPases include the Arp2/3 complex as an actin nucleator and members of the Wiskott-Aldrich Syndrome protein (WASP) family as activators of the Arp2/3 complex. We therefore analyzed the involvement of the Arp2/3 complex and WASP-family proteins in in vitro wound healing assays using NIH 3T3 fibroblasts and astrocytes. In NIH 3T3 cells, we found that actin and Arp2/3 complex contributed to cell polarity establishment. Moreover, overexpression of N-terminal fragments of Scar2 (but not N-WASP or Scar1 or Scar3) interfere with NIH 3T3 Golgi polarization but not with cell migration. In contrast, actin, Arp2/3, and WASP-family proteins did not appear to be involved in Golgi polarization in astrocytes. Our results thus indicate that the requirement for Golgi polarity establishment is cell-type specific. Furthermore, in NIH 3T3 cells, Scar2 and the Arp2/3 complex appear to be involved in the establishment and maintenance of Golgi polarity during directed migration.     

A novel role for 14-kDa phosphohistidine phosphatase in lamellipodia formation

Cell migration involves dynamic regulation of the actin cytoskeleton, which exhibits rapid actin polymerization at the leading edge of migrating cells. This process relies on regulated recruitment of actin nucleators and actin-binding proteins to the leading edge to polymerize new actin filaments. Many of these proteins have been identified, including the actin-related protein (Arp) 2/3 complex, which has emerged as the core player in the initiation of actin polymerization. However, the functional coordination of these proteins is unclear. Previously, we have demonstrated that the 14-kDa phosphohistidine phosphatase (PHP14) is involved in cell migration regulation and affects actin cytoskeleton reorganization. Here, we show that PHP14 may regulate actin remodeling directly and play an important role in dynamic regulation of the actin cytoskeleton. We observed a colocalization of PHP14 with Arp3 and F-actin at the leading edge of migrating cells. Moreover, PHP14 was recruited to the actin remodeling sites in parallel with Arp3 during lamellipodia formation. Furthermore, PHP14 knockdown impaired Arp3 localization at the leading edge of lamellipodia, as well as lamellipodia formation. Most importantly, we found that PHP14 was a novel F-actin-binding protein, displaying an Arp2/3-dependent localization to the leading edge. Collectively, our results indicated a crucial role for PHP14 in the dynamic regulation of the actin cytoskeleton and cell migration.     

The Cdc42 Effector Kinase PAK4 Localizes to Cell-Cell Junctions and Contributes to Establishing Cell Polarity

The serine/threonine kinase PAK4 is a Cdc42 effector whose role is not well understood; overexpression of PAK4 has been associated with some cancers, and there are reports that correlate kinase level with increased cell migration in vitro. Here we report that PAK4 is primarily associated with cell-cell junctions in all the cell lines we tested, and fails to accumulate at focal adhesions or at the leading edge of migrating cells. In U2OS osteosarcoma and MCF-7 breast cancer cell lines, PAK4 depletion did not affect collective cell migration, but affected cell polarization. By contrast, Cdc42 depletion (as reported by many studies) caused a strong defect in junctional assembly in multiple cells lines. We also report that the depletion of PAK4 protein or treatment of cells with the PAK4 inhibitor PF-3758309 can lead to defects in centrosome reorientation (polarization) after cell monolayer wounding. These experiments are consistent with PAK4 forming part of a conserved cell-cell junctional polarity Cdc42 complex. We also confirm β-catenin as a target for PAK4 in these cells. Treatment of cells with PF-3758309 caused inhibition of β-catenin Ser-675 phosphorylation, which is located predominantly at cell-cell junctions.     

Basic fibroblast growth factor and transforming growth factor beta-1: Synergistic mediators of angiogenesis in vitro

We investigated the relative roles of basic fibroblast growth factor (bFGF) and transforming growth factor beta-1 (TGF-b) on bovine aortic endothelial cell mitogenesis and morphogenesis using two-dimensional Petri dish cultures and a threedimensional hydrated collagen gel. bFGF alone stimulated endothelial cell proliferation with an EC₅₀ of 0.5 ng/ml. At bFGF levels greater than 2.5 ng/ml, morphologic alterations in confluent monolayers predominated; cells changed from a cobblestone morphology to an elongated cell pattern and showed enhanced migration into a denuded area of a Petri dish. In the three-dimensional model, exposure of endothelial cell monolayers to high bFGF levels stimulated minor cell migration directly under the monolayer but no invasion into the gel matrix. In combination with bFGF, heparin potentiated morphogenic changes, but not mitogenesis. bFGF, modification of the antiproliferative effect of TGF-b in confluent cultures was evidenced by induction of endothelial cell sprouting in response to 0.5 ng/ml TGF-b and 10–20 ng/ml bFGF in two-dimensional cultures. On collagen gels, endothelial cells migrated into the deep layers of the gel in a dose-dependent manner: invasion was maximal at 0.3–0.7 ng/ml TGF-b with decreased invasion at higher concentrations. The optimal collagen concentration that supported cell invasion was 0.075% collagen with the number of invading cells decreasing with increasing collagen gel density. By scanning electron microscopy, invading endothelial cells assumed a fibroblast-like appearance with slender cell extensions. We concluded that bFGF and TGF-b had independent effects on endothelial cell morphology and mitogenesis in culture. In combination at specific doses, these agents stimulated sprouting in the two-dimensional model and cell invasion in a collagen gel model. Morphogenic changes may be the primary event in determining angiogenesis. © 1993 Wiley-Liss, Inc.