CD36 is necessary for the inhibition of endothelial cell migration and tube formation by TSP-1. Soluble GST–CD36 proteins blocked the inhibition of migration by either intact TSP-1 or small TSP-1 peptides only when they contained the TSP-1 binding site. Blocking antibodies against CD36 prevented TSP-1 from inhibiting migration, while an IgM monoclonal antibody against CD36 mimicked TSP-1 and directly inhibited migration. Large vessel endothelial cells that lacked CD36 became sensitive to TSP-1 inhibition of both migration and tube formation after CD36 transfection. The inhibition mediated by CD36 in vitro likely reflects the in vivo situation because inhibition of migration is a consistent predictor of inhibition of neovascularization in vivo (Folkman and Klagsbrun, 1987
; Klagsbrun and D'Amore, 1991
; Bouck et al., 1996
). Furthermore, experiments to be reported elsewhere show that inhibition of corneal neovascularization by TSP-1 but not by angiostatin is defective in CD36 null animals (Febbraio, M., O.V. Volpert, N.P. Bouck, and R.L. Silverstein, unpublished data).
These results indicate that the antiangiogenic activity of TSP-1 is receptor mediated. They are not consistent with TSP-1 inhibiting angiogenesis by sequestering inducers, as has been suggested by the finding that TSP-1 can bind directly to inducers of angiogenesis, such as scatter factor (Lamszus et al., 1996
). In fact, in this study TSP-1 inhibition of scatter factor–induced migration could be completely blocked by an antibody to CD36. Our results are also not consistent with models suggesting that TSP-1 inhibits bFGF-induced angiogenesis by competing with it for binding proteoglycans on endothelial cells, as suggested by experiments with heparin-binding TSP-1 peptides (Vogel et al., 1993
), because migration induced by bFGF was effectively blocked in the absence of TSP-1 by agents that directly activated CD36, and inhibition by TSP-1 was antagonized by antibodies that blocked CD36 engagement.
While CD36 has often been implicated in adhesion and scavenging (Greenwalt et al., 1992
), data presented here show that CD36 expressed on endothelial cells is also a signaling receptor able to trigger a biological response. The response in this case is one that generates an inhibitory signal that blocks a positive response to inducers of angiogenesis. It is reminiscent of receptor-mediated negative signals to which immune cells are particularly sensitive (Scharenberg and Kinet, 1996
). Such signals often require a coreceptor and are mediated by src family kinases. It is not known if CD36 acts alone or in concert with an unidentified coreceptor or other surface molecule, but it does coprecipitate from endothelial cells in association with the src family kinase p59fyn
, and possibly other src family kinases (Bull et al., 1994
). It is also not yet clear how a signal emanating from CD36 might block the variety of stimulatory signaling cascades initiated by many different inducers of angiogenesis. However, the cytoplasmic domain of CD36 has been seen to associate with focal adhesion kinase (Sheibani, N., R. Zhong, and W.A. Frazier, manuscript in preparation), a kinase that can be essential for the stimulation of cell movement (Cary et al., 1996
; Gilmore and Romer, 1996
). If activation of CD36 were to disable focal adhesion kinase and thereby prevent the adhesion-associated tyrosine phosphorylation that may be essential for endothelial cell motility (Williams et al., 1996
), the effectiveness of a wide variety of inducers of angiogenesis would be severely compromised.
Although the CSVTCG motif on TSP-1 has previously been thought to mediate its interactions with CD36 (Asch et al., 1992
; Catimel et al., 1992
; Li et al., 1993
), data presented here show that an additional TSP-1 motif can interact with CD36. The Col overlap peptide lacking CSVTCG was antiangiogenic, was able to displace TSP-1 from CD36-transfected melanoma cells, and was rendered inactive by CD36 fusion proteins. A Mal III variant also lacking CSVTCG was antiangiogenic and able to displace TSP-1 from CD36-expressing cells. These peptides share a central GVQXR sequence that could represent a second CD36 binding motif. Two-step binding of TSP-1 to CD36 has been postulated by others (Leung et al., 1992
). Perhaps CSVTCG facilitates an initial interaction between TSP-1 and CD36 and GVQXR mediates a signaling response. Such a role would fit well with our observation that the Mal III peptide, which contains both motifs, had a higher affinity for CD36 and is a more potent inhibitor of migration (Tolsma et al., 1993
) than Col overlap, which lacks CSVTCG.
The identification of CD36 as a mediator of the inhibitory effects of TSP-1 helps to explain the unusual biphasic dose response curve generated when TSP-1 effects on endothelial cell migration are measured. TSP-1 inhibits endothelial cell migration at low concentrations less than 20 nM, yet stimulates migration at higher concentrations (Taraboletti et al., 1990
; Tolsma et al., 1993
; Gao et al., 1996
), a fact that has led to much confusion about its ultimate activity. Recent work has shown that a blocking antibody against the integrin-associated protein (IAP) prevents higher concentrations of TSP-1 from inducing endothelial cell migration (Gao et al., 1996
). Thus, the biphasic dose response curve for TSP-1 can be explained as resulting from the sum of activities of two distinct receptors, inhibitory CD36 and stimulatory IAP. CD36 mediates the inhibitory effects of TSP-1 at lower concentrations. The ability of CD36 to mediate phagocytosis (Ryeom et al., 1996a
) and the rapid degradation of lipids (Nozaki et al., 1995
) suggest that it may be rapidly cleared from the cell surface upon engagement with TSP-1. Thus, at high concentrations of TSP-1, CD36 may be sufficiently depleted to allow IAP to act unopposed, resulting in the stimulation of endothelial cell migration.
The binding activity of CD36 expressed on platelets has been shown to be regulated by the extracellular phosphorylation state of its ectodomain, with increasing phosphorylation decreasing its binding to TSP-1 and increasing its affinity for collagen and vice versa (Asch et al., 1993
). The in vitro sensitivity of CD36 to both ligands suggests that cultured endothelial cells have both phosphorylated and unphosphorylated CD36 molecules on their surface. In vivo it is possible that changes in CD36 phosphorylation by extracellular phosphatases and kinases could regulate the effect of both TSP-1 and other CD36 ligands on endothelial cells. This may be of particular importance at sites of platelet secretion and inflammation.
A variety of antiangiogenic compounds are now entering clinical trials as anticancer agents (Folkman, 1995a
; Gradishar, 1997
). The identification of CD36 as an inhibitory signaling receptor for TSP-1 could be useful in the design and discovery of additional pharmacologic agents that can inhibit pathologic neovascularization.