Here, we report evidence that uPAR directly associates with the head domains of integrin α5β1, modifies α5β1 conformation, and creates an additional binding site for Fn, likely within the second Fn heparin-binding domain. Complexes of uPAR and α5β1 are functionally relevant because uPAR binding promotes α5β1-dependent Fn matrix assembly and migration. Importantly, our observations are not based strictly on transfected cells because these features of uPAR–α5β1 interaction could be demonstrated in several tumor cell lines expressing endogenous uPAR. Collectively, these functional changes imply α5β1 activation by uPAR binding, as suggested previously by Aguirre-Ghiso et al. (1999)
. Our studies solidify α5β1 as a binding partner of uPAR and further define the uPAR-binding region on the integrin. The positioning of the uPAR-binding site near the integrin RGD-binding site also reveals a potential mechanism whereby uPAR–α5β1 complex formation empowers PAI-1–dependent cell detachment from Fn: while the integrin RGD-binding site remains intact in uPAR–α5β1 (), concurrent binding of urokinase and PAI-1 to uPAR now displaces intact Fn or the Fn cell–binding domain (containing RGD) from the integrin (), presumably by steric hindrance.
The consequences of uPAR–α5β1 complex formation contrast with that of other pathways of integrin activation. Available evidence indicates that integrin activation involves a global change in integrin conformation, at least part of which is a change in the orientation of the α and β head domains to better accommodate ligand binding. Several lines of evidence also support a model in which the very “bent” integrin conformation found in the αVβ3 crystal structure extends to point the head domains away from the cell under activating conditions (Takagi et al., 2002
). However, the full range of conformational changes a ligand-bound integrin may assume is uncertain (Mould and Humphries, 2004
). It is especially difficult to envision a fully extended conformation of activated integrins acting in cis to engage the much smaller GPI-anchored uPAR at the integrin upper surface. Rather, our findings suggest uPAR–α5β1 complexes exhibit an activation state involving a modified bent integrin with distinct functional properties. Similarities and differences between models of “extended” integrin activation and a model consistent with results reported here are summarized in . The model raises the more general possibility that some version of an angled integrin configuration, rather than being inactive, actually functions to promote integrin binding to cis-acting membrane ligands, such as uPAR, which coordinate integrin function with specific cellular needs.
Figure 8. Model for regulation of α5β1 integrin conformation and function by uPAR binding. Model proposes three basic forms of α5β1 exist on the cell surface: (1) A bent inactive form in the absence of integrin ligand; (2) (more ...)
The notion that uPAR activates and stabilizes α5β1 has been previously proposed by Ossowski and colleagues based on their studies of human epidermoid carcinoma cell lines. Enhanced adhesion to Fn of tumorigenic (T-HEp-3) over dormant (D-HEp-3) epidermoid cells was directly related to uPAR levels (Aguirre-Ghiso et al., 1999
). The high levels of uPAR in human epidermoid carcinoma cells (Hep-3) resulted in increased α5β1-dependent signaling to ERK as well as increased formation of Fn fibrils (Aguirre-Ghiso et al., 2001
). Both α5β1-dependent ERK signaling and Fn matrix assembly were decreased in the presence of a uPAR-binding peptide, P25, which blocks uPAR–integrin association, indicating that in these cells ligation of uPAR with P25 inhibited α5β1 function. Similar to these findings, we observed that induction of uPAR expression with Tet in an epithelial cell line increased Fn fibril formation ( B). Conversely, suppression of uPAR expression by RNA interference in many tumor cell lines decreased adhesion ( A, Fig. S1). Additional observations reported here help provide a physical rationale for these and prior functional studies. The current data indicate that uPAR directly binds and changes the conformation and matrix-binding properties of α5β1. The capacity of short β1-chain peptides to block uPAR–α5β1 functions without affecting α5β1-mediated Fn binding itself point to important conformational differences between free and uPAR-bound α5β1. The finding that β1P1 does not simply convert the function of uPAR–α5β1 to that of free α5β1 by dissociating uPAR also suggests that the conformational change incurred by complex formation with uPAR is distinct and perhaps not readily reversible. This is consistent with the hypothesis that bent and extended conformations of α5β1 can function as distinct activation states ().
This hypothesis is supported by our results, which reveal that suppressing uPAR expression induces ligand-induced binding site (LIBS) epitopes in HT1080, MDA-MB-231, and Skov-3 cells ( A). A recent report also documented increased β1-chain LIBS epitopes on human skin fibroblasts exposed to a peptide that disrupts uPAR–integrin interactions (Wei et al., 1996
; Monaghan et al., 2004
). The LIBS antibodies (HUTS-21, 9EG7) used here map to sites near the hinge region of the integrin but far away from the uPAR interaction site. Although these LIBS antibodies are thought to recognize the “active” conformational state of the β1 subunit that can be induced by ligand binding (e.g., Fn, RGD peptides, by activating antibody TS2/16, or Mn2+
), their binding is more sensitive to conformational changes in the hinge, knee, or leg domains than changes near the ligand-binding pocket (Bazzoni et al., 1995
; Luque et al., 1996
; Mould and Humphries, 2004
). We postulate that in uPAR-expressing cells the lower LIBS antibody binding reflects integrin angulation resulting from uPAR–α5β1 complex formation. This occurs in spite of “activation” of the integrin as judged by enhanced adhesion and Fn matrix assembly, further supporting the idea that activated integrins could exist in grossly different conformational states depending on the nature of the ligand ().
Previous studies have shown that initial cell attachment and spreading on Fn is mediated by the interaction of the RGD-containing Fn cell-binding domain (type III repeats 9–10) with α5β1 (Mould et al., 2000
; Redick et al., 2000
; Takagi et al., 2003
), but that further progression of the cytoskeletal response requires additional signals (Hocking et al., 1998
; Tarui et al., 2003
). Additional binding sites for cells on Fn provide the necessary signals. For example, interaction of cells with the Fn NH2
-terminal region can trigger integrin-mediated intracellular signals that are distinct from those generated in response to ligation with the RGD sequence (Forsyth et al., 2002
). However, in our assays, adhesion of epithelial cells to this 70-kD fragment was not influenced by uPAR expression and not inhibited by α5β1 blocking antibodies (unpublished data). Signals for cytoskeletal reorganization may also be provided by the interaction of Fn fragments containing the heparin-binding domain (Hep II) (type III repeats 12–14) with cell surface proteoglycans (Huang et al., 2001
). In fibroblasts, this response requires two cooperative signals provided by interactions of the RGD sequence with α5β1 integrin and the heparin-binding domain with syndecan-4 (Kim et al., 2001
). Our data show that both cells with uPAR or without uPAR adhere to Fn III 9-11 in an RGD-dependent manner, whereas only cells bearing uPAR adhere to Fn III 12-15. The latter cannot be blocked by RGD peptides, but can be blocked by β1 peptides that disrupt uPAR–β1 integrin interaction ( C). In most uPAR-expressing cells there are likely to be pools of α5β1 both free and bound to uPAR, suggesting that the incorporation of the heparin binding domain into the uPAR–α5β1 complex results in distinct signals that lead to enhanced integrin function, as our data show (). We cannot be sure whether uPAR–α5β1 complexes possess both Fn-binding sites or binding to both sites in Fn requires free and uPAR-complexed integrin. Future studies may distinguish between these possibilities.
We have previously reported that uPAR expression in kidney embryonic 293 cells both promotes Vn adhesion through association of uPAR with α3β1 and impairs Fn adhesion mediated by α5β1 (Wei et al., 1996
). Impairment of Fn adhesion in 293 cells appears anomalous with respect to all other cells expressing uPAR examined here and by others (Aguirre-Ghiso et al., 1999
). Consistent with this difference, expression of uPAR in 293 cells did not decrease binding of HUTS-21 and 9EG7 antibodies (unpublished data), implying that for some reason uPAR interacts, but not in the same manner, with α5β1 in 293 cells as that seen in other transformed cells. The molecular basis for the anomalous behavior of 293 cells remains to be defined.
The discovery of the capacity of α5β1 to undergo a phenotypic switch (i.e., RGD vs. β1P1 dependent; ), in Fn attachment may be relevant to attempts to regulate inflammation or tumor progression through integrin inhibition in vivo. uPAR is up-regulated in both inflammatory cells and many tumor cells with a metastatic phenotype. Indeed, uPAR expression is an independent risk factor for tumor metastasis in several clinical studies. RGD-based compounds or peptides have been shown to inhibit integrin function in vivo, but our data imply that one limitation in their use is the complete resistance of β1 integrins complexed with uPAR from RGD-dependent ligand binding. Vn adhesion mediated by uPAR–α3β1 complexes is also RGD-resistant (Wei et al., 1994
). Instead, uPAR-bound β1 integrins are sensitive to β1 peptides that map to the region of uPAR–integrin interaction. As these β1 peptides block cell adhesion ( B) and migration of various tumor cells (), it is possible that these reagents, perhaps coupled with RGD-based compounds, have therapeutic potential for suppression of tumor progression.