uPAR-deficient mouse keratinocytes fail to produce EGFR-dependent laminin-5, affecting migration in vivo and in vitro Free

uPAR, the receptor for urokinase (uPA), allows cell-surface conversion of plasminogen to plasmin, thereby increasing pericellular proteolysis and extracellular matrix (ECM) degradation, which are important for cell migration and tissue remodeling (Blasi and Carmeliet, 2002).

In addition to regulating proteolysis, uPAR directly modulates cell adhesion, differentiation, apoptosis, proliferation and migration through non-proteolytic mechanisms (Alfano et al., 2006; Blasi and Carmeliet, 2002). Because uPAR lacks a cytosolic domain, it must transmit intracellular signals through transmembrane proteins, including integrins, epidermal growth factor (EGF) receptor (EGFR), p130Cas (also known as BCAR1) and G-protein coupled receptors (reviewed by Blasi and Carmeliet, 2002). Moreover, uPAR directly and specifically binds to the ECM protein vitronectin (VN), and this interaction is enhanced by uPA. Mutants of uPAR that are impaired in their binding of vitronectin (but not uPA) lose the ability to modify cell shape and to induce cell migration (Madsen et al., 2007).

Deletion of the uPAR gene in mice causes impairment of neutrophil recruitment, abnormal migration and adhesion of peritoneal macrophages (Gyetko et al., 1995; Gyetko et al., 1994; Simon et al., 2000), alteration of kidney membrane permeability, protection from proteinuria, accelerated renal fibrosis in obstructive nephropathy (Wei et al., 2007; Zhang et al., 2003), and bone-homeostasis impairment (Furlan et al., 2007). uPAR-knockout (KO) mice are also deficient in their hematopoietic stem-cell homeostasis (Marc Tjwa and Peter Carmeliet, University of Leuven, Belgium, unpublished).

High levels of uPAR in human cancers is an independent negative prognostic marker (Stephens et al., 1999). The role of uPAR in tumors is still not understood, nor is it understood whether and how uPAR regulates cell growth in vivo. In cancer cell lines, uPAR regulates cell proliferation, activating growth-promoting pathways through integrins and the EGFR (Aguirre Ghiso et al., 1999; Liu et al., 2002). Fast-growing and highly metastatic Hep3 epidermoid carcinoma cells produce large amounts of uPAR. Silencing of uPAR in Hep3 cells induces a dormancy state in vivo, which is reversed by restoring the uPAR levels (Yu et al., 1997). In these cells, uPAR overexpression induces constitutive EGFR signaling, which requires uPAR–integrin-α5β1 interaction (Aguirre Ghiso et al., 1999; Liu et al., 2002). The EGFR might also represent the key molecule linking uPAR to extracellular-signal-regulated kinase (ERK) activation (Jo et al., 2003; Repertinger et al., 2004).

During cutaneous wound healing, growth factors (including EGF), cytokines and chemokines coordinate several processes, such as inflammation, proliferation, migration and angiogenesis, which are all required for appropriate tissue remodeling at the injury site (see Gillitzer and Goebeler, 2001; Werner and Grose, 2003). Resting keratinocytes adjacent to the wound margins are activated and show increased proliferation, dissolution of cell-cell adhesions, detachment from the basement membrane, lateral migration and invasion of the wounded area (Martin, 1997; Singer and Clark, 1999).

We show that, in vivo, uPAR-KO mice have a delayed wound-healing response, and decreased keratinocytes proliferation and migration. In vitro, primary keratinocytes showed deficiencies in adhesion, migration and proliferation. Moreover, in the absence of uPAR, the EGFR and its signaling pathway were not activated, partially explaining the proliferation defect. The deficient EGFR activation impairs laminin-5 (LN5) secretion and deposition both in vitro and in vivo, thus affecting keratinocyte adhesion and migration.

We examined skin wound healing in wild-type (wt) and uPAR-KO littermates as described in the Materials and Methods. The average wound length in uPAR-KO mice was significantly increased at all time points when compared with wt mice (supplementary material Fig. S1 and Table S1). Moreover, whereas the healing time for wt mice was 11.2±1.4 days, in uPAR-KO littermates the value increased up to 15.9±2.1 days, representing a small but significant delay (P=0.00028).

For histological analysis, uPAR-KO and wt skins were wounded and tissue samples collected 1, 3, 5 and 11 days after wounding. Frozen sections were immunostained for keratin 5 and wound diameters measured at each time point. Fig. 1A shows immunostaining 5 days after wounding. The arrows indicate the position of the advancing epithelial edges and define the wound diameter. Measurements at various times after wounding showed that closure of wound margins in uPAR-KO mice was significantly delayed at all time points (Fig. 1B).

During wound healing, keratinocytes migrate and proliferate within the injury (Goldfinger et al., 1999; Martin, 1997; Parks, 1999; Pilcher et al., 1999). At 3 days after wounding in uPAR-wt mice, the epidermis at the margins of the wounds was thickened and a wedge of basal keratinocytes invaded the provisional matrix of the wound bed (Fig. 2A, left panel). This hyperproliferation, however, was almost absent in uPAR-KO wounds, as judged by the number of cell layers (Fig. 2A, right panel).

The measure of this distance and the thickness of the keratinocyte layers give an estimate of the actual distance covered by keratinocytes and of their proliferation rate. As shown in Fig. 2A, the length of the epidermal tongue was strongly reduced in uPAR-KO wounds (202±42 μm; mean value ± s.e.m.) when compared with uPAR-wt wounds (592±89 μm; P<0.001). Indeed, both at 3 (Fig. 2A) and 5 (Fig. 1A) days after injury, in uPAR-KO mice the tip of the epidermal tongue was still located very close to the initial wound edge, in the area just beneath the crust.

As shown in Fig. 2B, the decreased proliferation of uPAR-KO keratinocytes was confirmed by counting the number of PCNA-positive cells 3 days after wounding, from the tip of the epidermal tongue (Fig. 2A, black arrow) to the start of the proliferating keratinocyte zone (Fig. 2A, blue arrow), following immunohistochemistry.

Similar to wt, at day 15 all uPAR-KO wounds were completely re-epithelialized; however, the newly formed epidermis could still be distinguished from normal epidermis by its increased thickness, particularly that of the basal keratinocyte layer (not shown). Indeed, day-15 frozen sections that were stained with anti-keratin-5 antibody showed that, in uPAR-wt skin, the keratinocyte multilayer was thinner than that of uPAR-KO mice, indicating that, whereas wt cells were reacquiring their quiescent state, uPAR-KO keratinocytes were still in their active proliferative state (Fig. 2B). These data indicate that, in vivo, upon skin injury, uPAR-KO keratinocyte migration and proliferation are delayed. These data also indicate that uPAR is necessary for a proper wound-healing time.

The effects observed in uPAR-KO mice did not depend on abnormal acute inflammation or abnormal neoangiogenesis. Quantification of neutrophils, mast cells and macrophages by immunohistochemistry and of microvessel density by immunofluorescence at various time points after wounding failed to show any significant difference between wt and uPAR-KO mice (supplementary material Fig. S2a,b; Fig. S3a,b; Table S2 and S3; and data not shown).

Overall, these results indicate that the wound-healing deficiency observed in uPAR-KO mice is due to abnormal keratinocyte behavior and not to abnormal wound-induced inflammation or angiogenesis.

Our observations in vivo are consistent with previously published data showing that uPAR mRNA is expressed by migrating keratinocytes at the edge of wounds (Romer et al., 1994). To test the role of uPAR in keratinocyte proliferation, primary keratinocytes were isolated from 2-day-old wt and uPAR-KO littermates. The absence of uPAR in cells derived from the KO mice was confirmed by western blot analysis (not shown).

Both uPAR-KO and wt primary keratinocytes were grown in the presence of either 1% or 8% Chelex-treated fetal calf serum (FCS) and their growth monitored every day over a 7-day period. As shown in Fig. 3A, uPAR-expressing keratinocytes grew significantly faster than uPAR-KO cells in the presence of 1% FCS. The difference between the growth rate of wt and uPAR-KO keratinocytes was even higher in the presence of 8% FCS (P<0.001) (Fig. 3B).

To activate primary keratinocytes in vitro, several growth factors, such as EGF, bFGF, PDGF, TGFβ and HGF, were tested. Among these, only EGF differentially affected the proliferation of uPAR-wt vs uPAR-KO keratinocytes (not shown). In the presence of 1% FCS, EGF increased the growth rate of primary uPAR-wt keratinocytes. By contrast, EGF did not promote growth of uPAR-KO cells (Fig. 3C,D). These results suggest that the mitogenic activity of EGF on keratinocytes requires uPAR.

To support these observations, we measured the effect of a specific inhibitor of EGFR tyrosine-kinase activity, AG1478, on the basal and EGF-induced growth of both wt and uPAR-KO keratinocytes. AG1478 did not alter EGFR expression (not shown), but inhibited both basal and EGF-induced proliferation of uPAR-wt keratinocytes, while having no effect on uPAR-KO cells (Fig. 3C,D).

To further investigate the role of uPAR in modulating EGF activity, we performed western blot analysis on total or immunoprecipitated cell extracts of primary keratinocytes grown in the presence or absence of EGF. As shown in Fig. 3E, uPAR-wt and -KO cells had comparable levels of EGFR. However, uPAR-KO keratinocytes showed lower levels of both basal and EGF-induced tyrosine phosphorylation of the EGFR. The two receptors could be co-immunoprecipitated both in the presence and absence of EGF (Fig. 3F). This confirms previous results that showed an EGF-independent interaction between uPAR and the EGFR (Liu et al., 2002; Mazzieri et al., 2006). We therefore conclude that the absence of uPAR does not alter EGFR expression; however, the lack of uPAR-EGFR interaction affects EGFR activation independently of the presence of its natural ligand, EGF.

The absence of uPAR also prevented phosphorylation of ERK1 and ERK2 (also known as MAPK3 and MAPK1, respectively, and collectively referred to as ERK1/2 owing to their high homology) – downstream effectors of EGFR signaling – even in the presence of EGF (Fig. 3G). These data suggest that uPAR is required for EGF-induced ERK1/2 activation in murine keratinocytes.

As shown in Fig. 3H, AG1478 was able to block the basal and EGF-induced activation of both EGFR and ERK1/2 in wt keratinocytes, whereas uPAR co-immunoprecipitation remained unaltered, meaning that uPAR-EGFR interaction is independent from the EGFR activity.

Similar results were obtained in the presence of a MEK1 inhibitor, UO126 (data not shown). Overall, the data indicate that uPAR interacts with the EGFR, and that this interaction modulates EGFR activation and ERK1/2 phosphorylation, which are both required for keratinocyte growth.

To further demonstrate the requirement of uPAR in keratinocyte proliferation and EGFR activation, uPAR-KO keratinocytes were infected with either a murine or human uPAR retroviral expression vector (Fig. 4A). An empty vector was used as a negative control.

As shown in Fig. 4B and C, overexpression of murine uPAR rescued both basal and EGF-induced cell proliferation, as well as EGF-induced ERK1/2 phosphorylation (Fig. 4D). However, human uPAR did not rescue any of the uPAR-mediated phenotypes (Fig. 4D), unless cells were treated with human uPA (supplementary material Fig. S4b). Because the binding of uPA to human and murine uPAR is species specific (Estreicher et al., 1989), this result suggests that endogenous uPA is normally required for the EGF response in murine keratinocytes.

In the skin, the basement membrane underlying keratinocytes is composed of collagen IV (ColIV), LN5 and laminin-10 (LN10). The basement membrane not only supports the skin architecture but is also required for growth, migration and differentiation of keratinocytes (Carter et al., 1990; Frank and Carter, 2004; Hamelers et al., 2005; Nguyen et al., 2000a; Nguyen et al., 2000b; Nguyen et al., 2000c). Moreover, migration of keratinocytes during re-epithelialization of cutaneous wounds is regulated by several ECM components – including collagens, fibronectin (FN) and laminin (Kim et al., 1992; Larjava et al., 1993; MacNeil, 1994); these proteins modulate keratinocyte adhesion, and therefore their migrating and proliferating capabilities.

To compare the adhesion properties of wt and uPAR-KO keratinocytes, cells were seeded on glass coverslips that were coated with either ColIV, FN, VN or LN5. Cell adhesion was quantitated, as described in the Materials and Methods, 8 hours after plating. The adhesion of wt and uPAR-KO keratinocytes on different substrates was not significantly different. However, an important difference was observed in the absence of exogenous substrates (Fig. 5A). Whereas 75% of uPAR-wt cells adhered and spread, only 5-10% of the uPAR-KO keratinocytes adhered to (but did not spread on) a glass coverslip (Fig. 5B).

In the absence of exogenous substrates, adhesion and spreading of keratinocytes depends on the ability of the cells to secrete and deposit their own matrix, in particular LN5 (Frank and Carter, 2004; Hintermann and Quaranta, 2004; Nguyen et al., 2000a; Nguyen et al., 2000b; Nguyen et al., 2000c). Our data therefore suggest that uPAR-KO keratinocytes might have a defect in the production and/or secretion of an ECM substrate.

We also carried out a long-term in vitro wound assay with confluent monolayers of wt and uPAR-KO keratinocytes cultured on ColIV- or LN5-coated surfaces. After scrape wounding, the migration of keratinocytes was measured at different time points. Fig. 5C shows representative examples of long-term migration on exogenous LN5. Wt cells migrated into the scraped area and closed the wound within 24 hours. By contrast, uPAR-KO keratinocytes did not close the wound. Similar results were obtained on ColIV (data not shown). One possible explanation of these results, among others, might be the inefficient production and/or secretion of LN5. Scrape wounding might remove exogenous LN5 and therefore cell migration could depend on newly produced and secreted LN5, which might be deficient in uPAR-KO cells.

Migration of wt and uPAR-KO keratinocytes towards different exogenous substrates was also performed using the modified Boyden chamber assay, as described in the Materials and Methods. uPAR-KO keratinocyte migration was significantly decreased on ColIV, FN and VN, but not on exogenous LN5 (Fig. 5D). The differences were statistically significant (Fig. 5D; *P≤0.051, **P≤0.0013, Student's t-test). These results might be explained by the fact that, even in the presence of exogenous collagen or FN, migration of keratinocytes would still require endogenous LN5, as previously described (DiPersio et al., 1997; Nguyen et al., 2000a; Nguyen et al., 2000b; Zhang and Kramer, 1996). These data further support the hypothesis that uPAR-KO keratinocytes are unable to produce and secrete sufficient amounts of LN5, and therefore would be impaired in adhesion, spreading and migration on uncoated surfaces.

To exclude possible differences in the levels and/or expression of integrins, we measured the levels of integrins for the different substrates used and found no difference between the two genotypes (data not shown). For example, we examined integrin α3β1 – one of the two receptors for LN5 – which is specific for cell-ECM contacts. Western blot analysis on cell lysates revealed similar levels of expression of both α3 and β1 integrin subunits in wt and uPAR-KO cells (data not shown), indicating that the uPAR-KO defect in keratinocyte adhesion and migration was not caused by changes in the expression of the LN5-binding integrin.

Overall, the above data prompted us to investigate the production and secretion of LN5 in uPAR-KO keratinocytes. To verify whether the absence of uPAR affected the intrinsic capacity of keratinocytes to produce LN5, we performed immunoblotting and reverse transcriptase (RT)-PCR analysis on suspended or adherent wt and uPAR-KO keratinocytes. In suspension, i.e. in the absence of adhesive stimuli, an equal amount of LN5 protein (Fig. 6A) and mRNA (Fig. 6B) was found in wt and uPAR-KO keratinocytes, indicating that the loss of uPAR expression did not affect the production of LN5 in the absence of an adhesive stimuli. Infection of uPAR-KO keratinocytes with a murine uPAR cDNA retroviral vector had no effect on LN5 production (Fig. 6A,B).

However, when keratinocytes were analyzed for the deposition of LN5 upon adhesion on ColIV, very different results were obtained. Cells were allowed to adhere by seeding on a ColIV-coated surface for 45 minutes; the secreted matrix was then scraped off the plates, size-separated by SDS-PAGE and immunoblotted with an anti-LN5 (γ2 chain)-specific antibody. As shown in Fig. 6A, densitometric quantification of the immunoblots showed a consistent increase in the amount of deposited LN5-γ2 protein in wt keratinocytes. By contrast, LN5 deposition was very much reduced in uPAR-KO cells. This may well be due to a difference in LN5-γ2 gene activation, as judged by the measurement of the specific mRNA by semi-quantitative RT-PCR (Fig. 6B). Similar data were obtained when cells were allowed to seed on glass (data not shown). Importantly, re-introduction of murine uPAR rescued the uPAR-KO phenotype both at the protein and mRNA level (Fig. 6A,B).

These findings indicate that uPAR is required for the adhesion-induced expression and deposition of LN5.

Adhesion to LN5 is important for proper keratinocyte migration in vitro, and in vivo during re-epithelialization of cutaneous wounds (Frank and Carter, 2004; Hamelers et al., 2005; Nguyen et al., 2000a; Nguyen et al., 2000b; Zhang and Kramer, 1996). To find a link between the EGFR and LN5 phenotypes in uPAR-KO keratinocytes, we analyzed the migration of wt and uPAR-KO keratinocytes on different exogenous substrates in the presence or absence of the specific inhibitor of EGFR tyrosine-kinase activity, AG1478, using a modified Boyden chamber assay. As shown in Fig. 7A, AG1478 inhibited the migration of wt cells seeded on ColIV, FN or VN, but had no effect on uPAR-KO keratinocytes. Moreover, AG1478 did not alter the migration of either wt or uPAR-KO keratinocytes seeded on LN5-coated filters. Similar results were obtained in the presence of a MEK1 inhibitor (data not shown). The lack of inhibition by AG1478 on uPAR-wt keratinocyte migration in the presence of exogenous LN5 contrasts with the inhibitory effect observed on cells seeded on other substrates. Therefore, these data suggest that the exogenous supply of LN5 bypasses the requirement for the EGFR-MAPK pathway in LN5 deposition and, therefore, in cell spreading and migration.

To test this hypothesis, wt and uPAR-KO keratinocytes were first grown in suspension and then seeded on a ColIV-coated surface in the presence or absence of AG1478. After 45 minutes of incubation, cells and secreted matrices were collected and analyzed as described above. RT-PCR analysis (Fig. 7B) showed that AG1478 prevented the upregulation of LN5 mRNA in wt cells, but had no effect in uPAR-KO cells. Likewise, immunoblotting (Fig. 7C) showed that AG1478 drastically decreased the amount of deposited LN5 in wt keratinocytes and had no effect in uPAR-KO keratinocytes. Similar results were obtained in the presence of UO126 (supplementary material Fig. S4c). These findings indicate that, similar to uPAR, the EGFR-MAPK pathway affects expression and deposition of LN5.

We next investigated whether the LN5-deposition defect, observed in uPAR-KO keratinocytes in vitro, was also detectable in vivo. Immunofluorescent staining of intact skin sections with an anti-LN5 (γ2 chain) antibody revealed no differences in LN5 deposition between wt and KO mice (data not shown). However, 3 days after full-thickness skin excision wounds, the same type of immunostaining revealed that LN5 was overexpressed in wt but not in uPAR-KO skin (Fig. 8A).

We also collected wounded skin of wt and uPAR-KO mice at different times after incision, and performed immunoblotting analysis on skin extracts with an anti-LN5 (γ2 chain) antibody (Fig. 8B). Unwounded skin was used as time-zero control (0 days). At 3 and 5 days after wounding, uPAR-wt mice showed induction of LN5 expression, whereas the level of LN5 remained unchanged in uPAR-KO skins. The tissue-extract data confirm the results obtained by immunofluorescence. We can therefore conclude that both in vitro and in vivo induction of LN5 expression and deposition requires uPAR.

Cutaneous wound repair comprises invasion of inflammatory cells and fibroblasts, neo-angiogenesis, proliferation and migration of keratinocytes, contraction, and remodeling of the scar tissue (Martin, 1997; Parks, 1999; Pilcher et al., 1999). In normal skin keratinocytes, expression of mRNA for uPA and its receptor, uPAR, is undetectable by in situ hybridization (Lund et al., 1996). After wounding, uPA and uPAR mRNA are expressed, with uPAR being confined to keratinocytes at the leading edge of the wound (Romer et al., 1994).

Wounding dramatically activates resting epidermal keratinocytes adjacent to the wound margin to a hyperproliferative, migratory and invasive state that allows invasion and successful re-epithelialization of the wound bed. This includes increased proliferation of basal keratinocytes, dissolution of cell-cell adhesion, detachment of keratinocytes from the basement membrane, lateral migration into the wounded area and invasion of the provisional matrix of the wound bed (Martin, 1997; Singer and Clark, 1999).

The EGFR pathway is important in skin wound healing and in skin cancer. Human and mouse squamous skin carcinomas overexpress EGFR ligands and mouse squamous skin tumors also display constitutive activation of the EGFR kinase (Rho et al., 1994; Xian et al., 1995). Following injury, a transient elevation of EGFR and its ligand in the skin and other epithelia is believed to contribute to the migration and proliferation of keratinocytes that are adjacent to wound margins (Nanney et al., 2000; Stoll et al., 1997; Werner and Grose, 2003).

We investigated the role of uPAR in skin wound repair. uPAR-KO mice show delayed wound healing when compared with wt mice (Fig. 1A). The histology of the wounds correlates with their gross appearance, and closure of the wound margins is in fact significantly delayed in the absence of uPAR.

In uPAR-KO epidermis, several events that are crucial to wound healing are altered – in particular, keratinocyte proliferation and migration (Fig. 2A,B). Other processes, such as neutrophil and macrophage recruitment or neo-vessel formation, are not affected (supplementary material Fig. S2 and S3). However, our data show that uPAR contributes to, but is not essential for, wound repair, as wound re-epithelialization is completed in uPAR-KO skins as well (Fig. 1A,B). This suggests a functional overlap between the uPA-uPAR system and other proteases, such as matrix metalloproteases (MMPs), which are expressed by the leading-edge keratinocytes (Martin, 1997; Singer and Clark, 1999). Proteases play a cooperative and crucial role in wound healing. In fact, wound healing is strongly delayed in the absence of plasminogen, but the additional inhibition of all MMPs together prevents closure of the wound (Lund et al., 1999; Romer et al., 1996a; Romer et al., 1996b). The requirement for plasmin and MMPs in wound healing has been ascribed to the need for keratinocytes to dissect their way through the fibrin-rich matrix. The deletion of uPAR, however, does not affect fibrinolysis (Bugge et al., 1996). Therefore, our unexpected delayed wound healing cannot be ascribed simply to deficient proteolysis. Because uPAR is a direct signaling regulator (Blasi and Carmeliet, 2002), these deficiencies probably depend on abnormal signaling.

The in vitro experiments show that uPAR is required for keratinocyte proliferation, the response to exogenous EGF and activation of the EGFR. However, uPAR does not determine EGFR level. Notice that uPAR and EGFR were co-immunoprecipitated in wt keratinocytes and that this interaction was not affected by EGF. uPAR is also required for the activation of the ERK-MAPK pathway, a downstream effector of EGFR signaling, independently of EGF. A similar conclusion has been reached for murine embryonic fibroblasts and MDA-MB 231 breast-cancer cells (Jo et al., 2007).

Thus, uPAR does not alter EGFR expression but affects its activation state. Indeed, EGFR- and MAPK-specific inhibitors block both basal and EGF-induced proliferation in wt but not in uPAR-KO cells, demonstrating that keratinocyte growth and EGF mitogenic activity require both uPAR and ERK1/2 phosphorylation.

Our data suggest that a uPAR-dependent autocrine or paracrine signal activates EGFR and ERK1/2 phosphorylation. Indeed, because uPA-binding to uPAR is species specific (Estreicher et al., 1989), the ability of murine, but not human, uPAR to rescue the mitogenic activity of EGF in uPAR-KO keratinocytes reveals an involvement also of uPA in this effect. The stimulation by human uPA of the growth of uPAR-KO keratinocytes that were infected with retroviral vectors expressing human uPAR further supports the involvement of the uPAR ligand in regulating keratinocyte growth.

What is not clear is the physical connection between uPAR and EGFR. Co-immunoprecipitation of uPAR and EGFR, observed in this and other papers, is not sufficient to demonstrate a direct interaction. Indeed, the co-immunoprecipitation in human monocytes of uPAR with integrins, transmembrane receptors and intracellular signaling molecules such as Src (Bohuslav et al., 1995) suggests that the co-immunoprecipitation represents a subcellular colocalization of various proteins. This is a crucial issue for understanding the nature of the uPAR-dependent activity of several molecules.

Adhesion of basal keratinocytes to the basement membrane (ColIV, LN5 and LN10) not only supports skin architecture but is also required for growth, migration and differentiation (Estreicher et al., 1989; Hintermann and Quaranta, 2004; Nguyen et al., 2000a; Nguyen et al., 2000b; Zhang and Kramer, 1996). LN5 is the major adhesive ligand present in the quiescent epidermal basement membrane and, after injury, is rapidly transcribed (Ryan et al., 1994), translated and deposited into the basement membrane by keratinocytes at the wound edge (Goldfinger et al., 1998; Lampe et al., 1998). This contrasts with the delayed expression of other basement-membrane components, such as type VII collagen (Lampe et al., 1998) and heparin sulphate proteoglycan (Oksala et al., 1995). Some studies also support the idea that LN5 contributes to migration on other substrates, such as collagen or FN (DiPersio et al., 1997; Nguyen et al., 2000a; Nguyen et al., 2000b; Ryan et al., 1994).

We found that uPAR-KO keratinocytes adhered to and spread on various exogenous substrates, but not on uncoated surfaces such as glass, on which the cells had to deposit their own substrate. However, adhesion and motility was rescued in the presence of exogenous LN5. Thus, uPAR-KO keratinocytes might be deficient in LN5 deposition and secretion. Indeed, after initial adhesion on ColIV, whereas wt cells consistently increased the amount of LN5 (γ2 chain) protein and mRNA, uPAR-KO cells did not. These findings indicate that uPAR is required for increased expression and deposition of LN5 upon adhesion of keratinocytes to an exogenous substrate.

Keratinocytes spread on LN5 using α3β1 integrin (Carter et al., 1990; Carter et al., 1991; DiPersio et al., 2000). However, the keratinocyte defect is not explained by a reduced expression of α3β1 integrin in uPAR-KO cell lysates, because the level of integrin α3 and β1 subunits was not decreased. By contrast, in the absence of uPAR, the deficient LN5 deposition can be ascribed to the reduced activation of the EGFR signaling pathway.

LN5 (γ2 chain) is as a marker of invading tumor cells (Ono et al., 2002), and it has been suggested that EGFR influences the invasive activity of tumor cells by stimulating the expression of LN5-γ2 (Katoh et al., 2002; Richter et al., 2005; Veitch et al., 2003). Indeed, a specific EGFR inhibitor, AG1478, inhibits LN5 deposition at the transcriptional and/or post-transcriptional level, and consequently influences LN5-dependent keratinocyte migration. Because the same results were obtained with the MEK1 inhibitor UO126, we propose that, in wt cells, EGFR regulates LN5 deposition (and spreading, migration and proliferation) through ERK1/2.

The results obtained with primary keratinocytes, i.e. decreased proliferation, migration and inability to upregulate LN5 production after wounding, were also observed in vivo. Moreover, wounded uPAR-KO skin displayed aberrant keratinocyte proliferation, revealed by the thinness of the migrating keratinocyte layer, the reduced number of PCNA-positive cells and a delay of the keratinocyte scar to return to normal thickness. Although other signaling systems are certainly involved in wound healing, our results highlight the novel in vivo role of uPAR in this process; this role can be largely explained by the requirement of uPAR for the activation of EGFR and its relative signaling pathways.

The murine polyclonal anti-uPAR antibody was kindly provided by Steven Rosenberg (University of California, Berkeley, CA). MOPC-21 IgG was from Sigma-Aldrich (St Louis, MO); monoclonal anti-murine α3 and β1 from Chemicon International (Temecula, CA); Tyrphostin AG1478 from Calbiochem (San Diego, CA); polyclonal anti-EGFR and anti-laminin γ2 (C-20), and monoclonal anti-total ERK2 antibodies from Santa Cruz Biotechnology (Santa Cruz, CA); monoclonal anti-phosphotyrosine antibody from Upstate Biotechnology (Lake Placid, NY); polyclonal antibody detecting phosphorylated ERK1 and ERK2 from Cell Signaling Technology (Beverly, MA); polyclonal anti-keratin-5 from Covance (Berkeley, CA); and horseradish-peroxidase-conjugated anti-mouse and rabbit IgG antibodies from GE Healthcare.

Primary keratinocytes from 2-day-old newborn wt and uPAR-KO mice (Malliri et al., 2002) were cultured in media containing 0.02 mM CaCl₂ (Hennings et al., 1980). To remove Ca²⁺ from serum, an analytical grade chelating resin, Chelex 100, was used (Bio-Rad, Hercules, CA). Ca²⁺ concentration was then measured on each serum batch with an atomic absorption spectrophotometer, kindly provided by Maria Tringali (University of Bicocca, Milan) (Perkin Elmer, Waltham, MA). Dermis and epidermis were separated overnight by 0.05% trypsin-EDTA (0.02%) and minced in S-MEM (Invitrogen, Milan, Italy) supplemented with 1.2 mM CaCl₂. Cell suspensions were filtered through a 40 μm Cell Strainer (BD Biosciences, Bedford, MA) and distributed in ColIV (Roche, Mannheim, Germany)-coated dishes. Keratinocytes were cultured in S-MEM with 8% or 1% chelexed FCS (Sigma-Aldrich, St Louis, MO), 0.02 mM CaCl₂ and 100 IU/ml penicillin/streptomycin (Invitrogen). For proliferation assays, cells were seeded at 30% confluency on ColIV-coated dishes and grown in 1% chelexed FCS-S-MEM with or without EGF (10 ng/ml) (Cambrex, East Rutherford, NJ), bFGF (Peprotech, Rocky Hill, NJ), HGF (Peprotech, Rocky Hill, NJ), PDGF (Peprotech, Rocky Hill, NJ), TGFβ (Peprotech, Rocky Hill, NJ), AG1478 (10 μM; Calbiochem San Diego, CA) or UO126 (10 μM; Promega, Madison, WI). Growth rate was measured every day over a 7-day period by washing with phosphate-buffered saline (PBS), fixing in 500 μl PBS containing 11% gluteraldehyde for 10 minutes at room temperature, rinsing with water, air-drying and staining with 300 μl 0.1% crystal violet (Sigma) in 200 mM MES, pH 6.0 (Sigma). After extensive rinsing, extraction was performed with 500 μl 10% acetic acid (20 minutes at room temperature) and 650 nm absorbance was measured. Values are expressed as the mean absorbance at 650 nm of triplicate samples ± s.d. Rac-11P cells, kindly provided by Arnoud Sonnenberg (Netherlands Cancer Institute, Amsterdam, The Netherlands), were cultured in DMEM supplemented with 10% bovine calf serum and were seeded 24 hours before use, to obtain a final density of 100%.

pBabe-puro-h-uPAR (human uPAR), or pBabe-puro-m-uPAR (mouse uPAR) were transfected into Phoenix packaging cells, and 24 hours later fresh viral supernatants were collected, filtered (0.45 μm), supplemented with 5 μg/ml polybrene and used to infect uPAR-KO keratinocytes. Cells (5×10⁵) were plated 48 hours before infection and selected for 4 days in puromycin (2 μg/ml; Sigma-Aldrich, St Louis, MO) after four rounds of infection (4 hours each) over a 2-day period. Cells expressing pBabe-puro empty vector were used as control.

All ECM proteins except LN5 were coated to culture dishes overnight at 4°C at 10 μg/ml (FN, Roche; VN, Promega) or 20 μg/ml (ColIV). A LN5 matrix was obtained by culturing Rac-11P cells to confluency (Delwel et al., 1993), after which cells were detached with 10 mM EDTA in PBS containing a mix of protease inhibitors (Complete protease inhibitor cocktail tablets; Roche) at 4°C. Before use, the dishes were washed twice with PBS.

For immunoprecipitation (IP), cells grown in 10-cm-diameter dishes were lysed in 500 μl of buffer containing 50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 2 mM MgCl₂, 10% glycerol, 1% Nonidet P-40 (Sigma) and protease inhibitors. Extracts were centrifuged at 18,000 g for 10 minutes at 4°C, quantitated by Bio-Rad protein assay (Bio-Rad, Hercules, CA) and pre-cleared with lysis buffer-containing protein-G agarose beads for 1 hour at 4°C. Pre-cleared lysates (0.5-1 mg) were incubated with antibodies against murine EGFR (1 μg/ml in lysis buffer) overnight at 4°C. After washing with lysis buffer, the immunoprecipitates were resuspended in SDS-sample buffer and immunoblotted with the indicated antibodies.

For immunoblotting, cell lysates (50-80 μg) from tissue-cultured cells or from mice back skin, or immunoprecipitated proteins were boiled for 5 minutes and resolved by SDS-PAGE. Proteins were electro-transferred to polyvinylidene difluoride membranes (Millipore, Billerica, MA), blocked with Tris-buffered saline (TBS) containing 5% non-fat dried milk and 0.02% Tween 20 for 1 hour at room temperature and probed using the indicated antibodies. For LN5-production analysis, secreted matrix was scraped off the plates, size-separated by SDS-PAGE and immunoblotted with an anti-LN5 (γ2)-specific antibody. Specific binding was detected using an anti-rabbit, anti-goat or anti-mouse horseradish-peroxidase-linked F(ab′)2 fragment (1:5000, GE Healthcare) followed by an enhanced chemiluminescence method according to the manufacturer's instructions (Pierce Chemical, Rockford, IL). When indicated, before protein extraction, cells were serum-starved for 18 hours and pre-incubated for 20 minutes at 37°C in humidified air with 5% CO₂.

24-well microplates were coated overnight at 4°C with ColIV, FN, VN or LN-5 (see section `Coating of dishes with ECM molecules') or PBS. Plates were washed with PBS and saturated with 1% (wt/vol) BSA (Roche), for 2 hours at 37°C to block nonspecific adhesion. Keratinocytes were detached with EDTA and suspended in supplement-free S-MEM. 6×10⁴ cells/well were plated in triplicate and incubated for 6-12 hours at 37°C. Non-adherent cells were removed with PBS, and cell adhesion estimated in a colorimetric assay, using 0.1% crystal violet (Sigma) in 200 mM MES, pH 6.0 (Sigma). Absorbance was measured at 650 nm.

Migration was assessed by using a 48-well microchemotaxis chamber (Neuro Probe, Gaithersburg, MD). 1% chelexed FCS–S-MEM (26 μl) was placed in the bottom compartment and 200 μl of cell suspension (125,000 cells/ml) added to the top compartment. The two compartments were separated by a FN-, ColIV-, VN- or LN5-coated 8-μm-pore-size polycarbonate filter (Neuro Probe) and incubated at 37°C in humidified air with 5% CO₂ for 6 hours. Thereafter, the filter was removed, scraped and stained with Diff-Quick (Dade Diagnostic, Aguada, Puerto Rico), and migrated cells counted by light microscopy. Results, expressed as the mean ± s.d. from triplicate samples represent at least three experiments. The significance of each migration assay was evaluated using the t-test, assuming unequal variances. P-values lower than 0.05 were considered statistically significant. Potential inhibitors were pre-incubated with cells for 30 minutes at 37°C and the assay was performed as described above.

For in vitro scrape-wound assays, keratinocytes were grown to confluency on LN5- or ColIV-coated dishes. Medium was aspirated and the cell-coated surface was scraped with a pipette tip in either a single stripe or a grid pattern. The scrape-wounded surface was washed with PBS to remove debris and then the wounds were allowed to heal in 1% chelexed FCS–S-MEM and photographs taken at various times, as indicated, with the indentation marks aligned.

Total cellular RNA was isolated in vivo from excised wounds and in vitro from keratinocyte suspension or after a 45-minute adhesion on ColIV-coated dishes, using Trizol Reagent (Invitrogen), and cDNA was synthesized by RT-PCR on 1 μg RNA using the Superscript RT-PCR system kit (Invitrogen). Specific transcripts were amplified with the following primers (TIB MOLBIOL, Genova, Italy): LN5 (γ2 chain) (forward, 5′-AACCAGCAAGTGAGTTACGG-3′; reverse, 5′-CCATTGTGACAGGGACATGG-3′); actin (forward, 5′-GGCATCCTGACCCTGAAGT-3′; reverse, 5′-CGGATGTCAACGTCACACTT-3′) (TIB MOLBIOL) using the standard PCR protocol for the EuroTaq kit (Euroclone, Pero, Italy). PCR products were resolved by 1.5% agarose-gel electrophoresis and stained with ethidium bromide.

Adult uPAR-wt and uPAR-KO littermates were anesthetized with isoflurane, shaved and one full-thickness skin excision wound (1 cm in length) was cut with small scissors on either side of the dorsal midline of each mouse. The wounds were left untreated and the mice were killed at different time points after injury. Wounds were either embedded for sectioning or excised with ∼6 mm of the epidermal margins for RNA isolation or protein extraction. For each time point, wound diameters (distance between epithelial rims) were determined in at least five mice per genotype both macroscopically and microscopically. Cryo- and paraffin sections across the middle of the wounds were stained: the cryosections (5 μm) with keratin 5 (K5), LN5 (γ2 chain) and DAPI, and the paraffin sections (7 μm) with hematoxylin and eosin or acidic toluidine blue (Sigma) (to identify mast cells) using standard methods. LN5 (γ2 chain) and K5 primary antibodies were visualized with FITC-labeled secondary antibodies (Molecular Probes, Eugene, Oregon) and images were collected using a Leica microscope. The length of the epidermal tongue was measured blindly in K5-stained sections, at 3 days after incision (eight sections per wound; n=8 per genotype) by computer-assisted morphometry from the tip of the epidermal tongue to the start of the zone of proliferating keratinocytes using Adobe Photoshop software.

Immunohistochemistry was performed using anti-macrophage F4/80 or anti-PCNA antibodies (Serotec, Raleigh, North Carolina), with a biotinylated secondary antibody, an ABC reagent (Vector Laboratories, Burlingame, CA), diaminobenzidine (Sigma-Aldrich) and hematoxylin counterstain. Blood vessels were identified using an anti-CD31 primary antibody (PharmMingen, San Diego, CA), an Alexa-Fluor-488 secondary antibody (Molecular Probes) and DAPI counterstain (Vector Laboratories). Blood vessels were counted in CD31-labeled sections in a ×20 microscopic field centered on the wound, in at least eight wounds per genotype, for each time point. Macrophages, hyperproliferative keratinocytes and mast cells were counted in sections following immunohistochemistry or histochemistry. At least three high-power fields immediately adjacent to the wound site were quantified and averaged in eight wounds per genotype for each time point.

This work was supported by grants from the AICR (Association for International Cancer Research), AIRC (Italian Association for Cancer Research) and the Italian Ministry of Health. The authors thank Roberta Mazzieri (TIGET Institute, San Raffaele, Milan) for her help with the editing of the manuscript and stimulating discussions; Arnoud Sonnenberg from the Netherlands Cancer Institute (Amsterdam, The Netherlands) for providing Rac-11P cells; and Maria Tringali (University of Bicocca, Milan) for her help with Ca²⁺ measurements in cell culture serum.