Transferrin Promotes Endothelial Cell Migration and Invasion: Implication in Cartilage Neovascularization

Chondrocyte cell culture methods have been extensively described elsewhere (Castagnola et al., 1986; Descalzi Cancedda et al., 1992). Briefly, 6-d-old chick embryo tibiae were digested, and the freshly dissociated chondrocytes were plated on culture dishes. Dedifferentiated cells were maintained in Coon's modified F12 medium with 10% FCS and cultured in adhesion for the following 3 wk. Cells were then transferred to suspension culture on agarose-coated dishes where they resumed differentiation and in 3–4 wk formed a nearly homogeneous population of single isolated hypertrophic chondrocytes. To eliminate residual cell clusters, cells were filtered through a nylon filter. In vitro–formed cartilage was obtained by the culture of these cells in suspension in the presence of 100 μg/ml ascorbic acid and 10 mM β-glycerophosphate.

Alternatively, cells were cultured in a serum-free medium (Quarto et al., 1992). Dedifferentiated cells were suspended in Coon's modified F12 medium supplemented with T₃ (10⁻¹⁰ M), insulin (60 ng/ml), and Dex (10⁻¹⁰ M). Fully differentiated hypertrophic chondrocytes were obtained after 1 wk in culture.

AIDS-KS cells (Albini et al., 1992) were cultured in RPMI 1640/DME (1:1) supplemented with 10% FCS (Gibco Ltd., Paisley, Scotland). EAhy 926 hybrid endothelial cells were the kind gift of Dr. C.J. Edgell (University of North Carolina, Chapel Hill) and were maintained in DME containing 10% FCS (Edgell et al., 1983).

Hypertrophic chondrocytes in different culture conditions were rinsed twice with PBS and maintained in F12 medium for 24 h (5 ml for 100-mm dish) without any addition. This medium was collected and used for chemotaxis assay. Metabolic labeling of the cell culture after incubation with F12 alone confirmed that the protein synthesis remained unchanged in quality and quantity with respect to the previous culture conditions.

As positive controls for angiogenesis, conditioned media from Kaposi sarcoma (KS)¹ cells were used (Albini et al., 1992, 1994; Salahuddin et al., 1988; Thompson et al., 1991). KS cells at 80% confluency were rinsed twice and incubated for 24 h with serum-free medium (8 ml per T75 flask) before conditioned media (CM) collection (KS-CM).

Conditioned media from both cultures were centrifuged for 10 min at 3,000 rpm to eliminate cell debris and stored at −20°C.

DNA content of cultured cells was determined to insure that corresponding concentrations of conditioned media were used. Cell layers were collected in 0.01% SDS in PBS and digested overnight at 50°C with proteinase K (Sigma Chemical Co., St. Louis, MO) at a concentration of 100 μg/ml in 10 mM Tris-HCl, pH 7.8, and 5 mM EDTA.

0.1 ml of each sample was added to 2 ml dye solution (10 mM Tris-HCl, pH 7.6, 1 mM EDTA, 0.1 M NaCl, and 0.1 mg/ml Hoechst 33258 [Sigma Chemical Co.], and DNA determination was performed in a DNA fluorimeter from Hoefer Scientific Instruments (San Francisco, CA).

The Boyden chamber chemoinvasion assay was carried out with slight modifications, as previously described (Albini et al., 1987). Matrigel, a mixture of basement membrane components obtained as an EnglebrethHolm-Swarm murine sarcoma extract (Kleinman et al., 1986), was layered evenly on polycarbonate filters (12-μm pore size, PVP-free; Nucleopore, Concorezzo, Italy) at a concentration of 20–25 μg per filter, air dried, and then reconstituted with medium into a solid gel.

Conditioned media from KS cells or chondrocytes were placed in the lower compartment of Boyden chamber and used as chemoattractants at a concentration proportional to the DNA content of cell layers.

In some experiments, the chemoattractant was transferrin from various sources: apotransferrin from human plasma (Sigma Chemical Co.), ovotransferrin (conalbumin) from chicken egg white (Sigma Chemical Co.), and ovotransferrin purified from chicken chondrocyte-conditioned medium, as previously described (Gentili et al., 1993). Conditioned media used in the different experiments presented in Results were derived from different primary cultures of hypertrophic chondrocytes.

We have previously demonstrated that EAhy 926 hybrid endothelial cells and human umbilical vein endothelial cells behaved similarly when the chemotactic activity of unfractionated medium conditioned by hypertrophic chondrocytes was tested (Descalzi Cancedda et al., 1995). Based on these results, EAhy 926 hybrid endothelial cells were used for the experiments reported in the present work. Endothelial cells (1.3 × 10⁵ cells per chamber) were placed in the Boyden chamber's upper compartment after harvesting with trypsin and washing with serum-free DME.

Standard assay conditions were 6 h at 37°C in 5% CO₂. After mechanical removal of the cells remaining on the upper surface of the filter, cells that migrated to the undersurface were stained (Toluidine blue; Sigma Chemical Co.) and quantitated microscopically. Five to ten random fields were counted on each filter. Assays were performed in triplicate and repeated at least twice. When antibodies were used, chemoattractants with and without antibodies and controls were previously incubated overnight at 4°C.

Chemotaxis assays were performed in the same way as those of chemoinvasion, with the only difference being that the filters were not coated with Matrigel but with gelatin (type A; Sigma Chemical Co.) at a concentration of 5 μg/ml as an attachment substrate.

In some experiments, chemoattractants (conditioned media or transferrins) were mixed with the cells in the upper compartment (checkerboard analysis).

When antibodies were used, chemoattractants with and without antibodies and controls were previously incubated overnight at 4°C.

Cells were labeled with [³⁵S]methionine, as described by Descalzi Cancedda et al. (1992). Briefly, cultured cells were washed in PBS and incubated for 2 h at 37°C in methionine-free Coon's modified F12 medium with 100 μg/ml ascorbic acid. [³⁵S]Methionine was then added and incubation continued for an additional 2 h. Aliquots of culture media were run for protein analysis on 15% SDS-PAGE in nonreducing conditions. Immunoprecipitation of specific proteins was performed as previously described (Descalzi Cancedda et al., 1992).

Cell layers were lysed with 0.1% SDS in PBS. Aliquots of samples were loaded on a 15% SDS polyacrylamide gel and electrophoresis was performed in reducing conditions. After electrophoresis, the gel was blotted to a BA85 nitrocellulose membrane (Schleicher & Schuell GmbH, Dassel, Germany) according to the procedure described by Towbin et al. (1979).

The blot was saturated overnight with 5% skim milk powder (Merck Biochemica, Darmstadt, Germany) in TTBS buffer (20 mM Tris-HCl, pH 7.5, 500 mM NaCl, 0.05% Tween-20) and washed several times with TTBS; it was then incubated (for 3 h at room temperature) with the goat polyclonal antibody directed against the human placental transferrin receptor, made monospecific by adsorption on transferrin to avoid cross-reactivity (a gift from Dr. H.H. Sussman, Stanford University School of Medicine, Stanford, CA).

After further washes, detection was performed by a biotin-conjugated anti–goat IgG (Jackson Immunoresearch Laboratories, Inc., West Grove, PA) and avidin-HRP (Jackson Immunoresearch Laboratories, Inc.), using 4-chloro-1-naphthol (Merck Biochemica) as substrate.

Fertilized White Leghorn chicken eggs (30 per group) were incubated under conditions of constant humidity at a temperature of 37°C. On the third day of incubation, a square window was opened in the egg shell after removal of 2–3 ml of albumen so as to detach the developing chorioallantoic membrane from the shell. The window was sealed with a glass of the same dimension, and the eggs were returned to the incubator.

Embryonic chorioallantoic membranes (CAMs) were treated at day 8 with (a) ovotransferrin purified from chicken hypertrophic chondrocyteconditioned medium (1 μg); (b) ovotransferrin from chicken egg white (Sigma Chemical Co.) (1 μg); (c) apotransferrin from human plasma (Sigma Chemical Co.) (1 μg); and (d) concentrated (3 μl) and not concentrated (3 μl) conditioned medium from cultured hypertrophic chondrocytes adsorbed on 1-mm³ sterilized gelatin sponge (Gelfoam; Upjohn Company, Kalamazoo, MI). Sponges were implanted on top of growing CAM under sterile conditions, within a laminar-flow hood. Sponges containing vehicle alone (PBS) and a stimulator of blood vessel formation, namely basic FGF (bFGF; 1 μg), were used as negative and positive controls, respectively.

CAMs were examined daily until day 12 and photographed in ovo under a stereomicroscope SR equipped with the Camera System MC 63 (Zeiss, Oberkochen, Germany). At day 12, CAMs were processed for light microscopy. Briefly, the embryos and their membranes were fixed in ovo in Bouin's fluid. The sponges and the underlying and immediately adjacent CAM portions were removed and processed for embedding in paraffin. 8-μm serial sections were cut according to a plane parallel to the surface of the CAM and stained with a 0.5% aqueous solution of Toluidine blue (Merck Biochemica) and observed under a Leitz-Dialux 20 light microscope (Leitz, Wetzlar, Germany).

At day 12, some CAMs were also injected intravascularly with India ink according to Olivo et al. (1992).

The angiogenic response was assessed by a planimetric method of point counting (Elias and Hyde, 1983; Ribatti et al., 1995). Briefly, every third section within 30 serial slides from an individual specimen was analyzed by a 144-point mesh inserted in the eyepiece of a Leitz-Dialux 20 photomicroscope. Six randomly chosen microscopic fields of each section were evaluated at ×250 magnification for the total number of the intersection points that were occupied by vessels transversally cut (diameter ranging from 3–10 μm). Mean values ± SD for vessel counts were determined for each analysis.

The vascular density was indicated by the final mean number of the occupied intersection points. The statistical significance of differences between the mean values of the intersection points in the experimental CAMs and control ones was determined by the t test for unpaired data.

The production of transferrins at high levels by differentiating hypertrophic chondrocytes, and the correlation of transferrin with angiogenic activity produced by these cells, led us to test if transferrin could induce migration and invasion of endothelial cells. Commercial human transferrin, purified from serum, and chicken ovotransferrin, purified from chondrocyte-conditioned culture medium, induced chemotactic migration and basement membrane invasion by vascular cells in a Boyden chamber assay (Fig. 1, a and b). A dose-dependent response was observed with a maximal migration at 10 μg/ml. Both forms of commercial human transferrin, holotransferrin and apotransferrin, were tested with identical results.

Human transferrin, chicken egg white ovotransferrin, and ovotransferrin purified from hypertrophic chondrocyte-conditioned culture medium (all at 10 μg/ml) were compared for induction of endothelial cell migration (Fig. 2). Recombinant human bFGF (10 ng/ml), as a positive control, and conditioned medium from hypertrophic chondrocytes in suspension with ascorbic acid for 21 d were also tested. All the factors showed comparable effects on the stimulation of endothelial cell migration (Fig. 2). When ovotransferrin levels in conditioned media from four randomly chosen cultures were determined by a Western blotting assay, values ranging from 1–10 μg/ml were obtained. This is in agreement with the maximal response to purified transferrin observed and the variation in the activity of individual conditioned media preparations.

Checkerboard analysis performed both on human transferrin (Table I) and on ovotransferrin (not shown) revealed a chemotactic activity enhancement. Comparable levels of endothelial cell migration were observed for human transferrin (10/1/0.1 μg/ml) as compared to bFGF (10/1/0.1 ng/ml) by checkerboard analysis (data not shown). Similar results were obtained when checkerboard analysis was performed on unfractionated conditioned medium from hypertrophic chondrocytes grown in suspension in the presence of ascorbic acid (Table II).

Conditioned culture medium from hypertrophic chondrocytes and purified ovotransferrin was tested for chemotactic and chemoinvasive activity in the absence and presence of affinity-purified antibodies against ovotransferrin (Fig. 3, a and b). These antibodies strongly, but not completely, inhibited chemotaxis and chemoinvasion induced by hypertrophic chondrocyte-conditioned medium and completely inhibited both activities induced by the purified ovotransferrin. Nonspecific antibodies (purified Ig from murine myeloma; Sigma Chemical Co.) at equivalent concentrations did not show any inhibitory effect (data not shown). These data indicate that ovotransferrin in the conditioned culture medium plays a major role in inducing migration and invasion of endothelial cells, but that additional factors which contribute to the response are present in the medium.

We have previously reported that in the absence of FCS, dedifferentiated chondrocytes undergo a differentiation in vitro to single hypertrophic chondrocytes when they are transferred in suspension culture in the presence of thyroid hormone, insulin, and dexamethasone (Quarto et al., 1992). Hypertrophic chondrocytes obtained using this serum-free culture system were supplemented with ascorbic acid and split in two subcultures. One subculture was additionally supplemented with 10% FCS, while the other control subculture was not. In both subcultures, due to the presence of ascorbic acid, the hypertrophic chondrocytes organized an extracellular matrix and rapidly aggregated, giving rise to a tissue strongly resembling hypertrophic cartilage as partially previously reported (Descalzi Cancedda et al., 1995). Nevertheless, metabolic labeling established that differences existed in the two subcultures with respect to proteins synthesized and released in the culture medium (Fig. 4). In particular, hypertrophic chondrocytes synthesized and secreted ovotransferrin only when the culture medium was additionally supplemented with FCS.

After 8 d, media in both subcultures were replaced for 24 h with the same fresh medium without ascorbic acid and FCS. These conditioned media from both subcultures were tested for chemotactic activity in the Boyden chamber assay. Both media gave a positive response, but to a different extent. The chemotactic activity in the medium from the subculture not supplemented with FCS (not containing ovotransferrin) showed chemotactic activity that was about one-third of the activity present in the medium from the supplemented subculture (producing ovotransferrin) (Fig. 5). As expected, the activity present in the FCS-free subculture was not significantly affected by the addition of antibodies against ovotransferrin. The activity present in the medium from the supplemented, transferrin-producing subculture was reduced by the addition of the same antibodies to about one-third, i.e., to levels similar to that of the unsupplemented culture.

The presence of transferrin receptors on the membrane of EAhy 926 cells was demonstrated by Western blot analysis. (Fig. 6). Affinity-purified polyclonal antibodies directed against the transferrin receptor identified a single 92-kD band on Western blots of EAhy 926 cell extracts that corresponds to the molecular mass of the known transferrin receptor. When function-blocking mAbs directed against the transferrin receptor (42/6; Trowbridge and Lopez, 1982) were added in the Boyden chamber assay, the chemotactic response of endothelial cells induced by the hypertrophic chondrocyte-conditioned medium was significantly reduced (50%), while the chemotactic response induced by purified transferrins was completely inhibited (Fig. 7). Polyclonal antibodies directed against ovotransferrin gave identical results.

We investigated whether or not transferrins induce endothelial cell proliferation, similar to the effect of other angiogenic growth factors. Immortalized EAhy 926 vascular cell growth was measured by the metabolic MTT assay (Carmichael et al., 1987), according to a published procedure (Albini et al., 1996). Endothelial cells were treated with human transferrin at different concentrations (10/1/0.1 μg/ml) in their culture medium without FCS or supplemented with 0.1% FCS. Transferrin did not enhance endothelial cell proliferation; at each tested culture time (24/ 48/72/96/144 h), the cells exhibited a doubling time similar to that observed with control medium alone. Cells in the presence of 10% FCS actively proliferated (data not shown).

Angiogenic effects of transferrins of various origin and of hypertrophic chondrocyte-conditioned culture medium were investigated in vivo in the CAM assay. Embryonic chorioallantoic membranes at day 8 of incubation were implanted with sponges containing either ovotransferrin purified from chicken hypertrophic chondrocyte-conditioned medium, ovotransferrin from chicken egg white, apotransferrin from human plasma, or conditioned medium from cultured hypertrophic chondrocytes. Vehicle alone (PBS) and bFGF were used as negative and positive controls.

After 12 d of incubation, macroscopically, the majority of the sponges were surrounded by allantoic vessels (Table III); vessels were more numerous in the sponges treated with purified ovotransferrin and developed radially toward the implant in a “spoked wheel” pattern (Fig. 8, A and C). In the specimens treated with PBS, no vascular reaction was detectable around the sponges (Fig. 8, B and D).

At a microscopic level, histological examination of the sponges showed a collagenous matrix containing numerous small blood vessels and fibroblasts localized between the sponge trabeculae (Fig. 9,a) 4 d after implantation. At the boundary between the sponge and the CAM mesenchyme, numerous host capillaries, penetrating in some points the sponge, were recognizable. In contrast, no collagenous matrix, blood vessels, or fibroblasts were present between the sponge trabeculae in the samples treated with PBS (Fig. 9 b).

Table IV shows quantitation of the angiogenic response performed at day 12 of incubation by using the planimetric point-count method (see Materials and Methods). In the experimental series, a higher microvessel density was detectable in the sponges treated with purified ovotransferrin, followed by sponges treated with concentrated conditioned medium, conditioned medium, ovotransferrin from chicken egg white, and apotransferrin from human plasma.

Angiogenesis is a complex process characterized by a cascade of events including activation and migration of endothelial cells, degradation and remodeling of basement membrane and surrounding extracellular matrix, endothelial cell proliferation, and neovessel formation (Brown and Weiss, 1988; Folkman, 1984; Folkman and Shing, 1992). The process is activated by the synthesis and release of angiogenic factors and/or switching off of antiangiogenic factors by the cells.

It is generally believed that cartilage chondrocytes produce migration and growth inhibitors for endothelial cells able to prevent blood vessel invasion even in the presence of angiogenic stimulatory factors. Angiogenic inhibitors have been purified from both resting and proliferative cartilage (Moses et al., 1990, 1992; Pepper et al., 1991). A suppressed synthesis of these antiangiogenic factors together with an activated synthesis of angiogenic factors in hypertrophic chondrocytes could explain the invasion of hypertrophic cartilage by blood vessels. An angiogenic activity has been described to be associated with mineralized hypertrophic chondrocytes in vitro (Brown and McFarland, 1992; Descalzi Cancedda et al., 1995).

We have demonstrated the high levels of ovotransferrin synthesis by hypertrophic chondrocytes both in vitro and in vivo (Gentili et al., 1993, 1994). Transferrins are monomeric, glycosylated, 80-kD proteins reversibly binding iron (de Jong et al., 1990). Hepatocytes are the main producer cells of this factor, but others such as Sertoli (Skinner et al., 1989), oviduct (McKnight et al., 1980), ependymal (Tsutsumi et al., 1989), and oligodendroglial (Bloch et al., 1985) cells synthesize significant amounts of transferrin.

In addition, transferrin differentially stimulates the growth of metastatic variant lines of murine melanoma (Nicolson et al., 1990) and human breast cancer cell lines (Inoue et al., 1993). Transferrin binding, but not growth response, correlates with invasive and metastatic properties of cell lines. This suggests that the transferrin/receptor system may be involved in more than just the support of cell proliferation. Endothelial cells have transferrin receptors (Jefferies et al., 1984) that are involved in transcytosis (Bickel et al., 1994); analogous to other cell types, these receptors could also be involved in endothelial cell proliferation and migration.

In the present study we have shown that transferrin is a chemotactic factor promoting migration and basement membrane invasion by endothelial cells. In the assay used to demonstrate this activity, endothelial cell migration was abolished by polyclonal antibodies directed against transferrin or by blocking antibodies directed against the transferrin receptor. We have previously shown a strong chemotactic and chemoinvasive activity, measured in Boyden chamber assays, in culture medium conditioned by hypertrophic chondrocytes (Descalzi Cancedda et al., 1995). In the presence of the antibodies against transferrin or the transferrin receptor, this activity decreased to 30–50% of the control. This finding indicates that transferrin is a major, but not the only, factor promoting cell migration in hypertrophic chondrocyte-conditioned culture medium and presumably in hypertrophic cartilage in vivo. Taking advantage of a serum-free culture system developed in our laboratory, we have shown that the medium conditioned by hypertrophic chondrocytes cultured in conditions in which they are surrounded by an organized extracellular matrix but do not express transferrin contained only 30% of the chemotactic activity normally observed in the media conditioned by hypertrophic chondrocytes. As expected, this residual activity was not affected by the addition of antibodies directed against the transferrin. From our study it appears that native transferrin in conditioned medium has a higher activity than purified transferrin. A possible synergism between transferrin and other factors present in the medium could explain this finding. An alternative explanation would be the possibility of partial inactivation of transferrin during purification.

In vivo studies on embryonic chorioallantoic membranes showed an angiogenic response to purified transferrins and conditioned culture medium. The angiogenic response of transferrin purified from chondrocyte culture medium and of concentrated conditioned culture medium was comparable to that obtained with bFGF, a well-known angiogenic molecule.

We have shown a role for transferrin in promoting endothelial cell chemotaxis and chemoinvasion in vitro and angiogenesis in an in vivo assay. In evaluating the possible involvement of transferrin in promoting angiogenesis during tissue development, one should consider that the angiogenic process is the result of a balance between angiogenic and antiangiogenic factors. A high expression of transferrin in a given tissue, e.g., the liver, does not necessarily imply implementation of a neoangiogenic program, especially if antiangiogenic or other regulatory factors are also expressed. Transferrin appears to have a unique angiogenic activity, as, unlike other angiogenic factors, it did not promote proliferation in the endothelial cell line we have used for our studies.

Studies to determine the mechanisms of action of transferrin as a chemotactic agent and to identify additional factors acting with transferrin are currently under investigation. It is expected that these studies not only will allow a better understanding of the mechanisms controlling blood vessel invasion in the hypertrophic cartilage during endochondral bone formation, but also will contribute to an understanding of the relationship between control of cell proliferation and angiogenesis in other tissues including metastatic tumors.

We thank Dr. I.S. Trowbridge (Department of Cancer Biology, Salk Institute, San Diego, CA) for blocking mAbs against human transferrin receptor; Prof. M. Presta (University of Brescia, Italy) for kindly providing basic FGF; Mrs. Barbara Minuto for editorial help; and Dr. Douglas Noonan for revision.

This work was supported by a grant from Progetti Finalizzati “Applicazioni cliniche della ricerca oncologica,” Consiglio Nazionale delle Ricerche, Rome, and by funds from the Associazione Italiana per la Ricerca sul Cancro and from Telethon, Italy.

bFGF

basic FGF

CAM

choriallantoic membrane

CM

conditioned medium

KS

Kaposi sarcoma