|Home | About | Journals | Submit | Contact Us | Français|
The urokinase receptor (uPAR) is linked to cellular migration through its capacity to promote pericellular proteolysis, regulate integrin function, and mediate cell signaling in response to urokinase (uPA) binding. The mechanisms for these activities remain incompletely defined, although uPAR was recently identified as a cis-acting ligand for the β2 integrin CD11b/CD18 (Mac-1). Here we show that a major β1 integrin partner for uPAR/uPA signaling is α3. In uPAR-transfected 293 cells uPAR complexed (>90%) with α3β1 and antibodies to α3 blocked uPAR-dependent vitronectin (Vn) adhesion. Soluble uPAR bound to recombinant α3β1 in a uPA-dependent manner (Kd < 20 nM) and binding was blocked by a 17-mer α3β1 integrin peptide (α325) homologous to the CD11b uPAR-binding site. uPAR colocalized with α3β1 in MDA-MB-231 cells and uPA (1 nM) enhanced spreading and focal adhesion kinase phosphorylation on fibronectin (Fn) or collagen type I (Col) in a pertussis toxin- and α325-sensitive manner. A critical role of α3β1 in uPA signaling was verified by studies of epithelial cells from α3-deficient mice. Thus, uPAR preferentially complexes with α3β1, promoting direct (Vn) and indirect (Fn, Col) pathways of cell adhesion, the latter a heterotrimeric G protein-dependent mechanism of signaling between α3β1 and other β1 integrins.
The heterodimeric α and β subunits of the integrin family of adhesion proteins have no intrinsic signaling capacity. Therefore, transduction of information into cells, after engagement of ligands by integrins, is dependent on the dynamic assembly of signaling complexes around their transmembrane and cytoplasmic integrin tails (Diamond and Springer, 1994 ; Schwartz et al., 1995 ; Burridge and Chrzanowska-Wodnicka, 1996 ). These dynamic aspects of integrin function are regulated in part by the interaction of integrins with neighboring nonintegrin membrane-associated proteins, including tetraspan-4-superfamily 1 members (CD9, CD81, CD151, and others) (Berditchevski et al., 1996 ; Maecker et al., 1997 ), integrin-associated protein (CD47) (Cooper et al., 1995 ), caveolin (Wei et al., 1999 ), and the glycosylphosphatidylinositol-anchored urokinase receptor (uPAR, CD87) (Wei et al., 1996 ). In all cases reported to date integrin-associated proteins in some way promote integrin signaling, although there is considerable mechanistic diversity. CD47 associates preferentially with αvβ3, promoting signaling through a heterotrimeric G protein-coupled pathway (Frazier et al., 1999 ). CD151, in contrast, preferentially associates with the β1 integrin partners α3 and α6 and promotes association of a cytoplasmic lipid kinase with these integrins (Berditchevski et al., 1996 ). Caveolin also associates with a set of β1 integrins, promoting their association with Src family kinases, probably by concentrating cholesterol-rich membrane “rafts” containing these kinases around integrins (Wary et al., 1998 ; Wei et al., 1999 ).
The influence of uPAR on integrin function appears complex. In experimental models either high levels of expressed recombinant uPAR or soluble uPAR have been reported to impair ligand binding by integrins and their adhesive functions (Wei et al., 1996 ). On the other hand, in most cells bearing endogenously expressed uPAR, uPAR, like other integrin-associated proteins, promotes integrin function. For example, we and others have recently reported evidence that signaling through the Fn receptor α5β1, and cell migration on Fn, was promoted by the association of this integrin with uPAR (Aguirre Ghiso et al., 1999 ; Wei et al., 1999 ). In one study soluble uPAR was found to promote signaling through α5β1 (Aguirre Ghiso et al., 1999 ). This is consonant with abundant, more circumstantial observations linking the expression of uPAR with cell migration important to inflammation and tumor metastasis (Bianchi et al., 1996 ; Andreasen et al., 1997 ; Ferrero et al., 2000 ; Huang et al., 2000 ). Whether the association of uPAR and α5β1 is direct or indirect is unclear because there has been no structural evidence to explain how uPAR might affect ligand engagement or signaling through integrins.
On the basis of homology with G protein-coupled receptors, Springer (1997) has proposed that the N-terminal region (~450 amino acids) of integrin α subunits folds into a seven-bladed β-propeller. In this model repeating units (W1-W7) of antiparallel β sheets connected by surface loops (~60 aa/unit) arrange into a torus around a small central cavity. The upper surface loops are thought to contain the major ligand-binding sites, which synergize with binding sites on the β chain to define the specificity and affinity of interactions of integrins with their ligands. We have recently identified a linear sequence within the α chain of CD11b (Mac-1) (αM424–440) that is a critical site for direct interaction between Mac-1 and uPAR (Simon et al., 2000 ). In the β-propeller model, this sequence comprises the entire upper loop sequence of the W4 repeat and extends into the third β strand of this repeat, indicating that uPAR is an atypical integrin ligand, at least for CD11b. We now extend these findings to α chain partners of β1 integrins, identifying α3β1 as a preferential uPAR-binding integrin.
Prourokinase was a kind gift of Dr. Jack Henkin (Abbott Laboratories, Abbott Park, IL). Human soluble uPAR with or without biotinylation and murine uPA were kindly provided by Dr. Steven Rosenberg (Chiron Corporation, Emeryville, CA). Integrin α3β1 was purified as described (Eble et al., 1998 ). Purified integrin α5β1 was a gift from Dr. Sarah C. Bodary (Genentech, South San Francisco, CA). Human fibronectin (Fn), collagen type I (Col), pertussis toxin, and goat anti-mouse IgG secondary antibody were purchased from Sigma (St. Louis, MO) and vitronectin (Vn) was from BD Biosciences (San Jose, CA). Monoclonal antibodies to integrin α2 (P1E6), α3 (P1B5), α5 (P1D6), and α5β1 (HA5), and polyclonal anti-β1 (AB1937) were obtained from Chemicon (Temecula, CA). A monoclonal antibody (mAb) against integrin β1 (JB1A) was a kind gift from Dr. John Wilkins (University of Manitoba, Winnipeg, MB, Canada). The polyclonal antibody to a Gαi subunit of heterotrimeric G proteins (Gαi-3) and the polyclonal antibody to Src family kinases (Src2) were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). The mAb to focal adhesion kinase (FAK) and the mAb to phospho-FAK were obtained from Transduction Laboratories (Lexington, KY). Rabbit anti-uPAR polyclonal antibody was purchased from American Diagnostica (Greenwich, CT). Purified mouse anti-human human leukocyte antigen-A,B,C mAb was obtained from BD PharMingen (San Diego, CA). Cy3 conjugated goat anti-rabbit IgG secondary antibody was from Zymed Laboratories (South San Francisco, CA). Monoclonal antibodies to integrin α2 (A2IIE10) and α3 (A3x8) and polyclonal antibody to integrin α3 were raised in Dr. Martin E. Hemler's lab, and the first two were conjugated with fluorescein isothiocyanate (FITC) with the use of a kit from Molecular Probes (Madison, WI). Peptides α325 (PRHRHMGAVFLLSQEAG), scα325 (HQLPGAHRGVEARFSML), α525 (PKGNLTYGYVTILNGSD), α625 (PRANHSGAVVLLKRDMK), αv25 (PRA-ARTL GMVYIYDGKN), 25, and M25 were synthesized and purified by Quality Controlled Biochemicals (Framingham, MA).
Human embryonic kidney cell line 293 and human carcinoma cell line MDA-MB-231 were obtained from American Type Culture Collection (Rockville, MD). All these cells were grown in DMEM (Invitrogen, Carlsbad, CA) supplemented with 10% fetal bovine serum (Hyclone, Logan, UT). Urokinase receptor transfected 293 cells were cultured in DMEM complete medium containing 0.9 mg/ml geneticin (G418) (Invitrogen). Immortalized epithelial cells from α3β1 integrin-deficient kidney (B12) and human α3-transfected B12 cells (R10) were obtained and cultured as described (Wang et al., 1999 ).
Cells were seeded in fibronectin- (5 μg/ml), collagen type I- (5 μg/ml), or vitronectin (1 μg/ml)-coated 96-well tissue culture plates to assess adhesion or spreading to these matrix proteins. The cell adhesion assays were performed as previously described (Wei et al., 1996 ). Briefly, 5 × 104/ml cells suspended in 100 μl of DMEM/0.1% bovine serum albumin (BSA) were seeded in triplicate on protein-coated 96-well plates and incubated for 1 h at 37°C, followed by three washes of phosphate-buffered saline (PBS). When performing inhibition assays, integrin α3 mAb (P1B5) and α5 mAb (P1D6) (5 μg/ml), pertussis toxin (100 ng/ml), or peptide α325 or scrambled peptide α325 (10–200 μM) were used. In some experiments, human pro-uPA (1 nM) was added to MDA-MB-231 cells or murine uPA (10 nM) to R10 and B12 cells. Cells attached to each plate were fixed with methanol and then stained with Giemsa. The data were quantified by measuring absorbance at a wavelength of 550 nm. When performing spreading assays, round and spread cells visualized by phase microscopy were counted from three different areas in each of triplicate wells after incubating with various peptides, antibodies, or urokinase.
Wild-type or uPAR-transfected 293 cells were detached and incubated with PBS containing 0.1% BSA and primary antibodies to integrins α3 (P1B5) or α5β1 (HA5) on ice for 30 min. After washing, cells were incubated with fluorescein isothiocyanate-conjugated goat anti-mouse IgG (Sigma) and analyzed on a flow cytometer (FACScan; BD Biosciences).
Cells (5 × 106) expressing uPAR were lysed on ice for 30 min in 1.5 ml of RIPA buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% deoxycholate, 0.1% SDS, 1% Triton X-100, 1 mM sodium orthovanadate, 1 mM phenylmethylsulfonyl fluoride, and 10 μg/ml leupeptin) or Triton lysis buffer (50 mM HEPES, pH 7.5, 150 mM NaCl, 1% Triton X-100), supplemented with protease inhibitors. After preclearing with protein A agarose, lysates were incubated with antibodies to integrins β1 (JB1A), α3 (P1B5), or α5β1 (HA5) at 4°C overnight. In some experiments, 100 μM peptides α325, scα325, or 25 was added to the lysates. The immunoprecipitates were blotted for uPAR (399R) or Gαi. In some cases, the membranes were stripped and reblotted for Src family kinases or β1 integrins. Initial experiments indicated that >95% of the total uPAR was solubilized by both 1% Triton and RIPA buffer. However, ~10% of the total cellular uPAR was coimmunoprecipitated with α3β1 in either 1% Triton or RIPA buffer.
Nunc high-binding microtiter plates were coated with purified α3β1 or α5β1 (2 μg/ml) and blocked with 10 mg/ml BSA. Biotinylated soluble uPAR (suPAR) (1–200 nM) with or without equimolar amounts of pro-uPA was added to each well in PBS/1 mg/ml BSA, and the plates were incubated for 1 h at 25°C. After washing, bound suPAR was quantified with avidin-peroxidase as described (Wei et al., 1994 ). To test specificity of binding, 100-fold molar excess nonbiotinylated suPAR was added. Data were expressed as specific binding, i.e., total binding minus the binding observed in the presence of excess unlabeled suPAR, which accounted for <20% of the total. Binding to wells coated with BSA alone accounted for <10% of the total.
The binding of biotinylated suPAR to peptide α325 was performed as described (Simon et al., 2000 ). In brief, Nunc microtiter plates were coated with α325 (20 μg/ml) in PBS overnight at 37°C and blocked with 1% BSA. Biotinylated suPAR (100 nM) without or with α325, scα325, α525, or αv25 (1–50 μM) was then added to each well for 1.5 h at 25°C. After washing, avidin peroxidase was added and biotinylated suPAR was quantified as described above. Relative binding was calculated as the ratio of binding in the presence of peptide to binding in the absence of peptide.
To visualize integrin and uPAR clustering, human breast cancer cells (MDA-MB-231) were trypsinized, recovered in suspension at 37°C for 1 h to allow reexpression of surface proteins, washed with serum-free DMEM, and incubated with antibodies to α3 (P1B5) and control HLA or FITC-conjugated monoclonal antibodies to integrins α2 (A2IIE10) and α3 (A3x8) at 4°C for 30 min. After washing, cells in suspension were incubated without or with goat anti-mouse secondary antibodies for 1 h at 37°C, immobilized on 50 μg/ml polylysine-coated glass coverslips for 30 min, and then fixed 20 min in 3.7% paraformaldehyde. Fixed cells were blocked in 10% goat serum for 1 h and incubated with rabbit polyclonal antibody to uPAR for 1 h at room temperature then incubated with Cy3-conjugated secondary antibodies and coverslips mounted in Prolong (Molecular Probes). Fluorescence staining was analyzed by Zeiss microscope or confocal laser (model MRC1024; Bio-Rad Laboratories, Hercules, CA) attached to a Zeiss microscope (model Axiovert S100) with the use of separate filters for each fluorochrome. Ventral planes were imported into Adobe Photoshop (Adobe Systems, Mountain View, CA) and processed.
To analyze FAK activity, cells were seeded on fibronectin- or collagen type I-coated 24-well plates. After incubating with peptides α325 or scα325 and antibodies to integrins α2 or α3, cells were lysed in RIPA buffer (150 mM NaCl, 50 mM Tris-HCl, pH 7.5, 1% deoxycholate, 0.1% SDS, 1% Triton X-100) supplemented with protease inhibitors. Lysates were immunoblotted for phospho-FAK and total FAK.
The putative structural organization of integrin α chains is indicated in Figure Figure1.1. Although no crystal structure for an integrin α chain has been reported, several lines of evidence favor a seven-bladed β-propeller folding pattern for their amino-terminal, ligand-binding region (Irie et al., 1997 ; Springer, 1997 ). The position and sequence of an uPAR/CD11b (αM) interaction site within the fourth blade (W4 repeat) of the β-propeller is also shown in Figure Figure11 (Simon et al., 2000 ). Surprisingly, comparison of the Mac-1 sequence with sequences of all other integrin α chains in the GenBank database reveals the two integrin α chains with the closest homology to Mac-1 at this site are α3 and α6 (each 40% identical), two integrin chains not previously recognized to be physically associated with uPAR. Figure Figure11 shows the aligned sequences of the W4 repeat region of α3 and α6 along with the that of two integrins for which indirect evidence has favored a physical association with uPAR (Xue et al., 1997 ; Aguirre Ghiso et al., 1999 ). As is evident, the primary sequences of α5 and αv in this region are less homologous than either α3 or α6. Based on this information, we initiated a series of experiments to determine whether the α3 and α6 sequences, termed α325 and α625, respectively, are functionally analogous to the previously reported M25 and whether α3β1 is a major signaling partner of uPAR.
To explore whether uPAR physically associates with α3β1 as predicted by sequence homologies (Figure (Figure1),1), coprecipitation experiments were performed in uPAR-transfected 293 cells. Previous studies have shown that 293 cells only express uPAR after expression by transfection (Wei et al., 1996 ). We verified that 293 cells express more α5β1 than α3β1 by fluorescence-activated cell sorter (FACS) analysis (Figure (Figure2A).2A). Lysates of uPAR/293 cells were immunoprecipitated sequentially with α3, α5, and β1 antibodies and the precipitates immunoblotted for uPAR. As is evident in Figure Figure2B,2B, the bulk of β1-associated uPAR coprecipitates with α3, ~90% by densitometric analysis. A small but consistent fraction of uPAR (~10%) was not removed with α3 antibodies but was precipitated with α5 antibodies. After sequential α3 and α5 immunoprecipitations, antibodies to β1 integrin chains recovered little or no uPAR, indicating little uPAR associated with other β1 integrins. Reversing the order of sequential immunoprecipitations (α5 then α3) verified the finding that uPAR preferentially associates with α3β1 in these cells.
Consistent with prior studies, 293 cells were found to adhere avidly to Fn with the use of the classic Fn receptor α5β1. Antibodies to α5 but not α3 completely block adhesion (Figure (Figure3).3). However, 293 cells expressing high levels of uPAR adhere poorly to fibronectin and instead adhere avidly to vitronectin, with the use of the vitronectin-binding site on uPAR. This adhesion is not blocked by EDTA or Arg-Gly-Asp (RGD) peptides (Wei et al., 1994 ) or enhanced by urokinase (Wei, unpublished observation). Given the finding that uPAR predominantly associates with α3β1 in these cells, we asked whether uPAR-dependent adhesion was blocked by antibodies to α3 (P1B5). These antibodies are reported to inhibit α3β1 function, although they do not block association of uPAR with α3 because P1B5 was used to coimmunoprecipitate uPAR and α3 (Figure (Figure2).2). Antibodies to α3, but not α5, completely blocked uPAR-dependent adhesion to vitronectin, consistent with the finding that uPAR requires an associated integrin to mediate adhesion and that in 293 cells at least this integrin is predominantly α3β1.
Although a loop sequence in α3 (Figure (Figure1)1) is most homologous to the previously identified interaction site for uPAR in CD11b (M25), the α3 sequence is not very homologous to the original phage display peptide sequence, peptide 25, used to identify M25 in the first place (Simon et al., 2000 ). This raises questions as to whether the α3 sequence, termed α325, is really involved in uPAR/integrin interactions and whether the association of uPAR with α3β1 (Figure (Figure2)2) is even direct. To address these issues, we examined binding between purified, soluble α3β1 and purified, suPAR under defined conditions in vitro. In this assay α3β1 was immobilized on plastic and the binding of biotinylated, soluble uPAR was measured. As indicated in Figure Figure4A,4A, suPAR binding to α3β1 was dependent upon uPA. In the presence of uPA, suPAR bound to α3β1 in a dose-dependent, saturable manner and with high affinity (Kd < 20 nM) (Figure (Figure4B).4B). The uPA/suPAR binding to α3β1 was almost completely abrogated by α325 but not by scrambled α325 or homologous peptides from either α5 (Figure (Figure4A)4A) or αv. To determine whether α325 itself affects uPA binding to uPAR we measured the capacity of α325 and scrambled α325 to inhibit binding of suPAR to immobilized uPA (residues 1–48, the receptor binding domain of uPA). In concentrations up to 50 μM, the highest concentration tested, neither α325 nor scrambled α325 affected binding of suPAR to uPA (Wei, unpublished observation).
We also tested whether suPAR interacts directly with the α325 peptide. The α325 peptide was immobilized on plastic and binding of suPAR to the peptide measured in the presence of increasing concentrations of α325, scrambled α325, or additional peptides as indicated in Figure Figure4C.4C. suPAR binding to peptide was detectable and blocked only by α325 (IC50 = ~5 μM) and not the other peptides tested. Interestingly, unlike suPAR binding to intact α3β1 (Figure (Figure4A),4A), the binding of suPAR to the α325 peptide alone was not influenced by the presence or absence of uPA.
To explore further the direct interaction of uPAR with β1 integrins, the binding of suPAR to immobilized α3β1 was compared with α5β1. suPAR/uPA binding was greater to immobilized α3β1 than to α5β1 (Figure (Figure4D),4D), and only binding to α3β1 was blocked by α325. A limitation of this analysis is that, although as judged by micro bicinchoninic acid protein assay (Pierce, Rockford, IL) equivalent integrins bound to the plastic, α5β1 was purified from tissues, whereas purified α3β1 had been expressed in soluble form. The folding on plastic of these two proteins could be different. Nonetheless, together these data indicate that uPAR directly binds α3β1 in a uPA-dependent manner and that this binding, like that to CD11b, involves a W4 loop peptide (α325) in the β-propeller region. The lack of effect of uPA on the direct interaction between uPAR and the α325 peptide suggests that this loop sequence is only a part of the overall interaction between uPAR and α3β1.
Consistent with these in vitro data, α325 was found to have similar functional and biochemical properties to those previously described for 25 and M25, the homologous sequence in CD11b (Simon et al., 2000 ). The 17-mer α3 peptide (α325) blocked uPAR-dependent adhesion to Vn in uPAR-transfected 293 cells in a dose-dependent manner, with and IC50 value of ~25 μM (Figure (Figure5A)5A) and at 100 μM blocked coprecipitation of Src family kinases with α3β1 integrins in these cells (Figure (Figure5B).5B). Neither peptides from α5 or αv nor scrambled versions of α3 had any activities in these assays. Thus, the physical and functional effects of α325 appear nearly identical to that of M25 and peptide 25, confirming that the W4 repeat of α3 is probably an interaction site for uPAR paralleling that described for Mac-1.
We next asked whether uPAR/α3β1 interactions regulate integrin function in nontransfected cells. MDA-MB-231 cells are known to express uPAR (Solberg et al., 1994 ) and a set of β1 integrins, including α2, α3, α5, α6, and αv (Morini et al., 2000 ; Gui et al., 1995 ; Lundstrom et al., 1998 ; Meyer et al., 1998 ). These cells migrate in vitro on the expected extracellular matrix proteins and grow and metastasize in vivo (Holst-Hansen et al., 1999 ; Kruger et al., 2000 ). Because the level of expression of uPAR in these cells is much lower than that of transfected 293 cells, we explored a possible interaction between uPAR and α3β1 functionally rather than biochemically. The α3β1 on MDA-MB-231 cells was clustered with secondary antibodies and then the distribution of uPAR determined by confocal microscopy. Clustering of α2β1 and MHC class I molecules served as controls. Clustering of α3β1 but not α2β1 resulted in dramatic coclustering of uPAR (Figure (Figure6).6). Clustered HLA class I molecules also had no effect on uPAR distribution. These data confirm that uPAR associates preferentially with α3β1 in MDA-MB-231 cells.
The influence of uPAR and α3β1 on spreading of MDA-MB-231 cells on various extracellular matrix-coated surfaces was next tested. Although MDA-MB-231 cells attach and spread on Fn and Col with the use of α5β1 and α2β1, respectively, spreading on Fn occurs relatively slowly >2 h at 37°C. The addition of recombinant human urokinase (pro-urokinase), enhanced spreading of MDA-MB-231 cells on both Fn and Col (Figure (Figure7A).7A). uPA-stimulated spreading was blocked by α325 but not controls. As expected, antibodies to α5 and α2 caused cellular detachment from Fn and Col, respectively. When cells were plated on vitronectin, uPA did not enhance spreading. Remarkably, antibodies (P1B5) to α3 but not antibodies to α2 or α5 were also found to enhance the rate of spreading of MDA-MB-231 cells on Fn or Col (Figure (Figure7A).7A). The enhancing effect of α3 ligation with antibodies was again abrogated by the addition of α325 in a dose dependent manner. These functional effects of α325 were mirrored by biochemical effects on FAK phosphorylation (Figure (Figure7B).7B). The addition of uPA (Figure (Figure7B)7B) or α3 antibodies caused enhanced tyrosine phosphorylation of FAK as measured 30 or 15 min after addition of uPA or antibodies to MDA-MB-231 cells plated on either Fn or Col at 37°C. Again α325, but not scrambled α325, abrogated increased FAK phosphorylation on either surface, suggesting the enhancing effects of uPA or α3β1 ligation on Fn and Col spreading require association of uPAR and indicating that α3β1 regulates the spreading response to engagement of α5β1 and α2β1 by their cognate ligands. Consistent with these observations, MDA-MB-231 cells exposed to uPA in suspension, after plating on poly-lysine-coated surfaces, or after plating on vitronectin failed to increase FAK phosphorylation (Wei, observation). In MDA-MB-231 cells the major vitronectin adhesion receptor appears to be αvβ5 rather than a β1 integrin (Meyer et al., 1998 ).
To test further whether uPAR interactions with α3β1 regulate the function of other β1 integrins, immortalized kidney epithelial cells derived from α3-deficient mice were examined (Wang et al., 1999 ). The influence of uPA on the Fn spreading response of both α3−/− cells (B12) and α3−/− cells reconstituted with human α3 (R10) was tested. Of note, the baseline spreading response of the α3−/− cells to Fn was at least twofold greater than that of α3-reconstituted cells, consistent with prior studies (Lichtner et al., 1998 ). The addition of 10 nM uPA clearly induced spreading of the α3β1-reconstituted cells, whereas uPA had no effect on spreading of α3−/− cells. Murine uPA enhanced spreading of the α3β1-reconstituted cells 80–160% on Fn and 50–140% on Col within 120 min of plating (Figure (Figure8A).8A). Accordingly, uPA increased FAK phosphorylation >2-fold in R10 cells at 120 min and this effect was blocked by the α325 peptide but not control (Figure (Figure8B).8B). The addition of uPA had no effect on the FAK phosphorylation state of B12 cells, which consistently had higher baseline phosphorylated FAK. Comparable amounts of uPAR were detected in B12 and R10 cells by semiquantitative PCR and by FACs analysis with the use of murine uPA-FITC.
The enhancing effect of urokinase on Fn or Col spreading, but not basal adhesion, is completely blocked by the addition of pertussis toxin (Figure (Figure9A).9A). Spreading induced by α3β1 ligation was also pertussis toxin sensitive (Wei, observation). These results suggests that a heterotrimeric G protein is required for “cross talk” between α3β1/uPAR complexes and α5β1 or α2β1. This is not completely unexpected because prior studies have indicated that signaling through β1 integrins is promoted by the presence of caveolin-1 (Wary et al., 1998 ; Wei et al., 1999 ). Caveolin-1 localizes to cholesterol-rich regions of cell membranes and has been demonstrated to associate with heterotrimeric G proteins such as Gαi. Gi proteins, like Src family kinases, are both myristylated and palmitoylated near their N terminus, providing a driving force for localization to cholesterol-rich membrane rafts (Li et al., 1996 ; Harder and Simons, 1997 ). uPAR also localizes to membrane rafts via its glycolipid membrane anchor, the integrin/uPAR protein interaction then promoting accumulation of rafts around integrins (Wei et al., 1999 ). To determine whether Gαi is in fact associated with α3β1, coprecipitation experiments were again performed. In both uPAR/293 cells and MDA-MB-231 cells (Figure (Figure9B),9B), antibodies to α3β1 precipitated not only the integrin but also both Gαi and Src family kinases. Similar results were obtained whether coimmunoprecipitation was done with cells lysed in either 1% Triton or RIPA buffer. The coprecipitation of both sets of these signaling molecules was blocked by the addition of 100 μM α325 but not scrambled α325 to the cell lysates. Treating intact cells with peptide α325 before lysis gave similar results. Identical experiments in nontransfected 293 cells revealed detectable src family kinases coprecipitating with α3 antibodies but little or no Gαi. In the absence of uPAR, the α325 peptide had no effect on association of src family kinases with α3β1 (Figure (Figure9B).9B). Similar expression of Gαi and Src family kinases in nontransfected and uPAR transfected 293 cells was detected by Western blotting of cell lysates. Although these results cannot be viewed as quantitative, the findings indicate that the presence of uPAR alters qualitatively the complement of signaling partners associated with α3β1.
Coupling of cellular adhesiveness with proteolytic cascades is an increasingly recognized paradigm for coordinating focal attachment and detachment important to cell migration (Werb, 1997 ). The uPAR is a prototypical example of this strategy (Chapman, 1997 ). By localizing with integrins and binding urokinase, uPAR focuses plasmin activation at or near sites of focal contact between the cell surface and extracellular matrix proteins (Blasi et al., 1987 ). Plasmin activates cascades involving both matrix metalloproteases and growth factors in the pericellular milieu (Carmeliet et al., 1997 ). Prior studies have also shown that the complexes uPAR forms with integrins are important to binding and adhesion of hematopoietic cells to matrix vitronectin, plasmin cleavage of vitronectin reversing this attachment (Wei et al., 1996 ; Waltz et al., 1997 ). Results reported here further develop this paradigm by elucidating the specificity of interaction between uPAR and β1 integrins. Our results indicate that uPAR preferentially interacts with α3β1 and that this interaction has two important functional consequences: 1) uPAR/α3β1 complexes enable a pathway of cellular adhesion to Vn, especially in cells with little or no αvβ3; and 2) these complexes initiate a signaling pathway promoting the function of α5β1 and α2β1. Both pathways of signaling and enhanced adhesion are activated by concurrent binding of urokinase to uPAR. The pathways are nonetheless distinct because pertussis toxin only blocks the cross talk between α3β1/uPAR and other β1 integrins and not α3β1/uPAR-dependent Vn adhesion. The observation that urokinase signals through uPAR/α3β1 complex formation is consistent with a recent report that urokinase induces metalloproteinases in oral keratinocytes through an α3β1-dependent mechanism (Ghosh et al., 2000 ). Thus, the intricate connections between expression of proteases and function of the adhesive machinery of cells is epitomized by the reorganization of membrane partners induced by uPAR expression and its association with α3β1.
Our current findings may help clarify previously reported, apparently contradictory observations regarding the influence of uPAR on the function of the Fn receptor α5β1. In 293 cells, high levels of uPAR expression impair the function of Fn receptors (Wei et al., 1996 ). Yet our data (Figure (Figure2)2) indicate that the majority uPAR in these cells is associated with α3β1 and not the Fn receptor. This finding suggests that the inhibition of Fn receptor function by uPAR is probably indirect. Because caveolin-1 is important to β1 integrin signaling and preferentially associates with uPAR/integrin complexes, and because 293 cells express relatively low levels of caveolin-1, overexpression of uPAR may enrich α3β1 complexes with caveolin and at the same time deplete Fn receptors of caveolin. This may explain why impaired Fn receptor function in uPAR-transfected 293 cells is reversed by overexpression of caveolin-1 (Wei et al., 1999 ). In contrast, physiological levels of uPAR expression in most cells appear to promote rather than impair Fn receptor function. Our data suggest that this operates, at least in part, through signals derived from uPAR/α3β1 complexes. Although Fn receptors do not require such signals for adhesion, the presence of these signals accelerates FAK phosphorylation and cell spreading on Fn, and therefore may promote Fn receptor-dependent cell migration. We have previously reported that peptides that disrupt uPAR/integrin association impairs smooth muscle cell migration on Fn (Wei et al., 1999 ). We postulate that this may explain the recently observed requirement for uPAR expression in Fn receptor-dependent tumor invasion (Aguirre Ghiso et al., 1999 ). This pathway may also underlie the requirement of uPAR for αvβ5-dependent migration of pancreatic carcinoma cells on vitronectin even though uPAR was not required for vitronectin adhesion of these cells (Yebra et al., 1996 ).
A series of recent studies by Blasi and colleagues have defined a pathway of urokinase- and uPAR-mediated chemotaxis (Fazioli et al., 1997 ; Degryse et al., 1999 ). Urokinase stimulates chemotaxis of uPAR-bearing cells in a pathway involving Src kinase activation and sensitive to heterotrimeric G protein inactivation with pertussis toxin. The requirements for FAK and Src kinase activation for this migration favor integrin activation as a critical feature of urokinase-dependent chemotaxis. Data reported here may shed light on these observations. We find urokinase, by promoting uPAR/α3β1 interactions, promotes FAK activation and spreading of MDA-MB-231 cells on either fibronectin or collagen in a G protein-dependent manner (Figures (Figures77 and and9).9). This signaling is blocked by peptides that dissociate uPAR and Gαi-3 from α3β1, increasing the possibility that urokinase is chemotactic for cells because urokinase enables ligand-dependent G protein activation through an integrin. It remains to be determined how conformational changes in uPAR or α3β1 induced by urokinase could mediate Gα or Gβγ activation. Although the mechanism is not defined, our observations are conceptually similar to recent reports that the integrin-associated protein CD47 promotes association of αvβ3 with heterotrimeric G proteins and that this is important to spreading mediated by this integrin (Frazier et al., 1999 ; Green et al., 1999 ). The finding of two distinct examples of coupling of integrins to heterotrimeric G proteins by integrin-associated proteins suggest this may be a common adaptive mechanism of cells to link matrix attachment to cell migration.
Prior studies have indicated that in addition to binding laminin-5, the integrin α3β1 regulates the function of other β1 integrins (DiPersio et al., 1995 ; Fukushima et al., 1998 ; Hodivala-Dilke et al., 1998 ). Antibodies (P1B5) to α3β1 that block laminin-5 attachment promote spreading and migration of cells on Col and/or Fn (Kubota et al., 1997 ; Lichtner et al., 1998 ), consistent with findings reported here (Figure (Figure7A).7A). Furthermore, epithelial cells from mice deficient in α3 show altered organization of integrin focal contacts and enhanced spreading on Fn, suggesting an inhibitory role for α3β1 on Fn and Col integrin receptors (Lichtner et al., 1998 ). Our finding that urokinase, mimicking P1B5, evokes α3β1-dependent signals promoting activation of several β1 integrins indicates uPA-dependent association of uPAR with α3β1 attenuates and even reverses the dominant negative function of α3β1 on other β1 integrins. This observation raises the possibility that prior observations of “integrin activation” by soluble uPAR may operate through its association with α3β1 (Aguirre Ghiso et al., 1999 ). In addition, our observations may also shed some light on the possible molecular basis for such cross talk. The association of α3β1 with uPAR appears to be required for coprecipitation of Gα and Src family kinases with this integrin. Such complexes may complement the binding of the same integrin to CD151, a tetraspan family member that associates specifically with α3β1 and that has been recently linked to signaling and migration of cells via this integrin (Yauch et al., 1998 ; Berditchevski and Odinstova, 1999 ). Antibodies to CD151 coprecipitate uPAR in uPAR-transfected 293 cells. However, peptides α325 and M25, which disrupt uPAR/integrin interactions, have no effect on α3β1/CD151 complexes (Wei and Hemler, unpublished observations), consistent with the mapping of the interaction site between CD151 and α3β1 to the membrane proximal region of the α chain and the mapping of the uPAR/integrin interaction site to the β-propeller (Simon et al., 2000 ; Yauch et al., 2000 ). We postulate that multimeric complexes involving CD151, uPA/uPAR, and α3β1 have distinct signaling capacity promoting integrin signaling and migration on multiple matrix ligands for β1 integrins. How these complexes organize and whether other membrane adaptor proteins contribute importantly to their signaling function remains to be determined. It is important to reiterate that the discovery that uPAR associates preferentially with α3β1 is based on sequence homology with a previously defined integrin interaction site on CD11b/CD18 (Mac-1). The α6 amino acid sequence in the same region is also quite homologous and we show here that a α6 peptide based on this sequence also disrupts uPAR/integrin coprecipitation and uPAR-dependent adhesion, whereas peptides of the identical region of α5 (Figure (Figure5A)5A) or αv were inactive. The possible functional significance of uPAR/α6β1 complexes in cells expressing both of these receptors remains to be defined.
Finally, our observations that uPAR associates preferentially with integrin α chains mediating laminin-5 binding may provide an explanation for prior findings that uPAR colocalizes with the distribution of laminin-5 in vivo at sites of tumor cell invasion (Pike et al., 1995 ). Laminin-5 is a major basement membrane matrix protein that is breeched during the invasion of metastatic cells into or out of blood vessels. The finding that uPAR associates preferentially with the laminin-5 binding β1 integrins supports the hypothesis that invasive tumor cells have exploited the advantage of coordinate signaling of integrins, proteases, and protease receptors embodied by uPAR/integrin interactions to promote invasion and metastasis. This is also supported by studies correlating uPAR expression with metastatic capacity and poor prognosis of breast cancer patients (Solberg et al., 1994 ). If so, our studies identifying a critical site for interaction between uPAR and laminin-5 binding integrins may be a site for intervention in the invasive process.
We thank Dr. Christopher S. Stipp for labeling antibodies with FITC, and Dr. Martin E. Hemler for critical comments on this article and for antibodies to α3β1. This work was supported by National Institutes of Health grants HL-44712 awarded to H.A.C. and GM-38903 awarded to Dr. Martin E. Hemler.