In culture, cell confluence generates signals that commit actively growing keratinocytes to exit the cell cycle and differentiate to form a stratified epithelium. Using a comparative proteomic approach, we studied this ‘confluence switch’ and identified a new pathway triggered by cell confluence that regulates basement membrane (BM) protein composition by suppressing the uPA–uPAR–plasmin pathway. Indeed, confluence triggers adherens junction maturation and enhances TGF-β and activin A activity, resulting in increased deposition of PAI-1 and perlecan in the BM. Extracellular matrix (ECM)-accumulated PAI-1 suppresses the uPA–uPAR–plasmin pathway and further enhances perlecan deposition by inhibiting its plasmin-dependent proteolysis. We show that perlecan deposition in the ECM strengthens cell adhesion, inhibits keratinocyte motility and promotes additional accumulation of PAI-1 in the ECM at confluence. In agreement, during wound-healing, perlecan concentrates at the wound-margin, where BM matures to stabilize keratinocyte adhesion. Our results demonstrate that confluence-dependent signaling orchestrates not only growth inhibition and differentiation, but also controls ECM proteolysis and BM formation. These data suggest that uncontrolled integration of confluence-dependent signaling, might favor skin disorders, including tumorigenesis, not only by promoting cell hyperproliferation, but also by altering protease activity and deposition of ECM components.
The epidermis is a continuously renewing tissue generated by the proliferating keratinocytes of the basal layer. Interaction of basal keratinocytes with a specialized underlying extracellular matrix (ECM), called basement membrane (BM), is crucial for skin homeostasis by controlling keratinocyte adhesion, migration, growth, survival and differentiation (Yurchenco and Patton, 2009). Indeed, altered expression of ECM components or their cell surface receptors, such as integrins, affects integrity of the epidermis, resulting in pathological situations as exemplified by bullous diseases (Aberdam et al., 1994; Vidal et al., 1995). The BM is a highly complex supramolecular structure composed of a variety of multifunctional proteins including collagens, laminins, nidogens, heparan sulfate proteoglycan (HSPG), proteases and growth factors. Moreover, BM composition is highly dynamic, evolves over time and contributes to adapt the epithelial cell behavior to different physiological and pathological processes.
ECM remodeling requires both degradation of the existing matrix by proteases and deposition of newly synthesized ECM components. In addition, protease activities modulate cell behavior, such as cell migration and proliferation, by affecting cell adhesion to the ECM and by releasing both bioactive polypeptides and sequestered growth factors and cytokines (Toriseva and Kähäri, 2009). The serine protease of the plasminogen (Plg) activation system plays a crucial role in pericellular ECM proteolysis (Chapman, 1997). Indeed, plasminogen activators uPA and tPA convert Plg into its active form, plasmin, which, in turn, degrades ECM proteins, either directly or indirectly through activation of MMPs. PAI-1 is the major extracellular and physiological inhibitor of uPA and tPA. Inappropriate expression or control of the Plg protease system has been associated with pathological conditions such as chronic wounds and tumor development (Rømer et al., 2001; Wysocki et al., 1999), which underscores the relevance of this proteolytic pathway in the maintenance of skin homeostasis.
In vitro culture of keratinocytes provides insight into the control of the skin homeostatic processes. At low-density, keratinocytes actively grow and migrate, which is reminiscent of the behavior of ‘activated’ keratinocytes in a wound bed. At confluence, maturation of cell–cell contacts generates signals that promote keratinocyte growth-arrest and differentiation (Fuchs, 2007; Müller et al., 2008). Following this ‘confluence switch’, keratinocytes organize into stratified epithelia with a basal layer, in which most of the keratinocytes are growth-inhibited and undifferentiated as observed in normal skin (Kolly et al., 2005).
To investigate the possible involvement of ECM components in the regulation of keratinocyte behavior during the ‘confluence switch’, we used a comparative proteomic approach. We analyzed the molecular composition of the ECM secreted by actively proliferating and migrating keratinocytes versus confluent cultures organized in a growth arrested and differentiated epithelium. Our data demonstrate that cell confluence influences not only keratinocyte growth and differentiation but also controls proteolytic processes and ECM protein deposition. We report that PAI-1 accumulation in the ECM at cell confluence inhibits uPA–uPAR–plasmin proteolysis pathway, which allows perlecan deposition. Perlecan deposition in the ECM improves adhesive strength and decreases migration capacity of keratinocytes. Moreover, presence of perlecan in the ECM allows further PAI-1 association to ECM and Plg cascade inhibition, thereby facilitating BM maturation. Furthermore, we determine that adherens junction maturation and upregulation of TGF-β and activin A cytokines expression by confluence control ECM deposition of PAI-1 and perlecan. Our study unravels an important signaling pathway generated by cell confluence involved in the control of ECM proteolysis and BM formation in skin, which are both essential for the maintenance of epithelial integrity.
Keratinocyte growth arrest and differentiation induced by confluence is associated with modification of the ECM composition
The possible role of the ECM in regulating keratinocyte behavior during the confluence switch was investigated by proteomic analysis. Human keratinocytes (HKs) were plated at the same density and ECM laid down by HKs was collected either in pre- or post-confluent culture conditions. ECM extracts were fractionated by SDS-PAGE electrophoresis followed by in-gel trypsin digestion and LC-MS/MS peptide analysis. Among the 79 identified proteins, 16 were ECM components; the remaining represented cellular contaminants (Fig. 1A). Since spectrum counting is considered quantitative (Zybailov et al., 2005), the number of peptides identified by MS/MS analysis indicated the proteins abundance in pre- and post-confluence cultures. As expected (Carter et al., 1991), Lm-332, composed of β3, γ2 and α3 chains, is the main laminin and the most abundant protein found in the ECM secreted by keratinocytes (Fig. 1A). While incorporation of a subset of proteins (Lm-332, HTRA1, Lm-γ1) in the ECM was unchanged by keratinocyte confluence, others were either reduced (tenascin-C, maspin, thrombospondin-1) or accumulated (perlecan, βig-h3, PAI-1, nephronectin, Lm-β1, collagen α1 (XVIII), Lm-α5, and antithrombin-III) (Fig. 1A). Two of the most strongly accumulated proteins in the ECM at confluence were PAI-1 and perlecan. Because Perlecan is a secreted heparan-sulfate proteoglycan (HSPG) present in most of BM (Iozzo, 2005), which maintains BM integrity and controls growth factor availability, and PAI-1 is the main inhibitor of an important proteolysis pathway, we investigated the functional significance of these molecules during the formation of stratified epithelia.
The mass spectrometry quantitative data were confirmed by western-blot and slot-blot analysis of ECM extracts of pre- to post-confluent HK cultures. The relative amounts of perlecan, PAI-1 and βig-h3 were strongly increased in the ECM of confluent (d7) and post-confluent (d9 and d10) HK cultures compared to Lm-β3, γ2, and vitronectin (Fig. 1B; supplementary material Fig. S1A). Proliferation curve demonstrates that the variation of cell number between pre-confluent (d5) and confluent (d7) is relatively low (around 1.3-fold), indicating that variation of ECM composition is not caused by variation in cell number (supplementary material Fig. S1B,C). Western blotting of cell lysates prepared in parallel showed that cell confluence (d7) triggers accumulation of cell cycle inhibitors, p21WAF1 and p27KIP1, and expression of the suprabasal keratinocyte-specific keratin 10 (Fig. 1C), while Cyclin A, which is crucial during S-phase progression and mitosis, was downregulated (Fig. 1C). Such changes indicate cell cycle arrest and engagement of cell differentiation. In post-confluent cultures, cell growth inhibition and differentiation were confirmed by monitoring BrdU incorporation and analyzing the profile of keratin expression (supplementary material Fig. S1D,E). Interestingly, these data highlight that induction of cell differentiation and growth inhibition at confluence correlates with ECM accumulation of PAI-1 and perlecan.
Cell adhesion and cell movement were analyzed on pre- and post-confluent HK cultures. The attachment capacity of the different cell populations was assessed using a cell detachment kinetic assay (Gagnoux-Palacios et al., 2001). Pre-confluent cultures (d4 and d5) are less resistant to trypsin/EDTA treatment than the confluent (d7) and post-confluent (d9) counterparts (Fig. 1D). To rule out the possibility that trypsin sensibility was due to changes in ECM proteins vulnerability to trypsin digestion, we also use a non-enzymatic detachment assay (supplementary material Fig. S1F) and obtain similar results. Both approaches indicate enhancement of cell attachment by confluence. To analyze the effect of confluence on cell movement, HK cultures were monitored by time-lapse videomicroscopy over 24 hours and movement of individual cells within colonies was determined by manual tracking. As expected, cell confluence reduces dramatically the cell movement (Fig. 1E).
Altogether, these data demonstrate that significant modification of the ECM composition correlates with increased cell adhesion and decreased motility that take place during the ‘confluence switch’, which is known to induce keratinocyte cycle arrest and cell differentiation.
Increased secretion and matrix deposition of PAI-1 inhibits the Plg cascade and enables perlecan deposition
To determine whether increased deposition of PAI-1 at confluence correlates with a modification of the synthesis or/and secretion of the molecule, we analyzed by western blot the presence of PAI-1 in keratinocyte cell lysates and conditioned media of pre-confluent, confluent and post-confluent keratinocytes (Fig. 2A and supplementary material Fig. S2A). PAI-1 levels are strongly increased both in cell lysates and conditioned media of confluent and post-confluent HKs. To further determine the role of protein synthesis in PAI-1 accumulation at post-confluence, pre- and post-confluent keratinocytes were treated with cycloheximide (CHX) to block protein translation and PAI-1 levels were analyzed by western blot (supplementary material Fig. S2B). This experiment shows that PAI-1 protein levels in both pre-and post-confluent HKs are strongly affected by CHX treatment. After removal of CHX and extensive washing, PAI-1 protein reappears more rapidly in post-confluent HK culture, suggesting that confluence is associated with an increase in PAI-1 synthesis. Altogether, these data indicate that induction of PAI-1 expression and secretion contributes to increased levels of PAI-1 in the ECM at confluence.
PAI-1 is the major inhibitor of the Plg activation cascade. Using an enzymatic assay, we measured the activity of the Plg activators (PA) in the media conditioned by keratinocytes. PA activity is strongly inhibited at cell confluence, which correlates with ECM accumulation of PAI-1 (Fig. 2B). Zymography detection of PAs, either secreted or associated with the cell surface, detected a unique 50 kDa band corresponding to uPA which was strongly decreased in cells reaching confluence (Fig. 2C). Western blot analyses against uPA demonstrate that decreased PA activity at confluence correlates with decreased amounts of either secreted or associated enzyme (Fig. 2C). These results indicate the downregulation of the Plg activation cascade when HKs organize into an adhesive and differentiating epithelium.
Perlecan undergoes proteolytic processing by plasmin, which cleaves a short domain interacting with a variety of growth factors (Whitelock et al., 1996). We thus investigated whether the plasmin-dependent-proteolysis block impacts the stability of perlecan within the ECM at confluence. The Plg cascade was either activated by knocking down PAI-1 expression, or inhibited by treating HKs with an uPA inhibitor. As expected, siRNA-mediated knockdown of PAI-1 resulted in significant increase of both secreted and cell surface-bound uPA enzymatic activities (Fig. 2D,E and supplementary material Fig. S2C). We also observed an increase in uPA receptor (uPAR) expression (Fig. 2D), which is consistent with the known role of PAI-1 in down regulation of the uPAR–uPA complex at the cell surface (Nykjaer et al., 1997). Interestingly, PAI-1 knockdown strongly reduced perlecan deposition in the ECM (Fig. 2D). These data suggest that uncontrolled uPA activity prevents accumulation of perlecan in the ECM. Consistent with this hypothesis, addition of increasing amounts of Plg to post-confluent cultures of epithelial A431 cells specifically inhibited perlecan accumulation in the ECM without impairing deposition of other ECM proteins (Fig. 2F). As expected, destabilization of perlecan in the presence of Plg was enhanced by siRNA knockdown of PAI-1 (Fig. 2G). To prove that plasmin degrades perlecan, we also show that addition of purified plasmin on perlecan-enriched matrix triggers formation of degraded forms of perlecan (supplementary material Fig. S2D). Inversely, inhibition of plasmin-dependent proteolysis using an uPA inhibitor stabilized perlecan in the ECM (Fig. 2H). Altogether these results demonstrate that inhibition of plasmin-dependent degradation at confluence is a crucial prerequisite for perlecan accumulation into the ECM.
Perlecan exerts a positive-feedback loop on its own deposition by regulating PAI-1 matrix deposition and uPA/uPAR expression
Because perlecan accumulation in the ECM deposited by confluent HKs relies on inhibition of the Plg activation cascade, we tested whether perlecan could reciprocally regulate the Plg enzymatic pathway. Perlecan expression was silenced in confluent keratinocytes either by RNAi-mediated knockdown or by treating cells with sodium chlorate, a chemical inhibitor of HSPG synthesis (Olwin and Rapraeger, 1992). Inhibition of perlecan expression resulted in reduction of PAI-1 deposition to the ECM (Fig. 3A and supplementary material Fig. S3A). Consequently, the uPA activity (Fig. 3B and supplementary material Fig. S3B) and the uPA association to the cell surface (Fig. 3C) were upregulated. Absence of perlecan also enhanced uPAR levels in HKs (Fig. 3A), as observed in response to PAI-1 knockdown.
Loss of laminin in Drosophila results in abnormal deposition of perlecan (Urbano et al., 2009). To assess whether absence of laminin secretion by HKs inhibits perlecan deposition, and thereby deregulates PAI-1 ECM accumulation, perlecan deposition was investigated in Lm-332 null-keratinocytes. Lm-332-null keratinocytes were isolated from patients affected by junctional epidermolysis bullosa (JEB) caused by absent expression of laminin β3 subunit (Vailly et al., 1998). Slot-blot analysis of ECM extracts failed to detect perlecan in post-confluent JEB keratinocyte cell cultures despite presence of perlecan in the spent culture medium (Fig. 3D). Absence of perlecan in the ECM of JEB keratinocytes correlated with a reduction of PAI-1 deposition and an increase of both uPAR and cell-surface-associated uPA (Fig. 3F,G). Restoration of Lm-332 expression in the JEB keratinocytes upon transduction of a wild-type Lm-β3 cDNA (JEB-Ker-REV) partially restored both perlecan and PAI-1 deposition to the ECM (Fig. 3E).
Altogether these experiments demonstrate that both absence of expression and deposition of perlecan affect PAI-1 targeting to the ECM, which results in abnormal regulation of the Plg activation cascade in HKs.
Perlecan regulates uPA/uPAR mRNA levels and interacts with PAI-1 in the ECM
In order to better characterize the molecular mechanism driving the increase of uPA/uPAR and reduced ECM deposition of PAI-1 in perlecan knockdown cells, we compared their mRNA levels in keratinocytes transfected with control or perlecan siRNA (Fig. 3H). Q-PCR analysis revealed that uPA and uPAR mRNA levels are increased in perlecan knock down cells, while expression of PAI-1 is unchanged. This result suggests that perlecan regulates uPA and uPAR expression through a transcriptional-dependent mechanism.
PAI-1 interaction with the ECM molecule vitronectin (Declerck et al., 1988) has been shown to increase the half-life of the active form of PAI-1 (Mimuro et al., 1987). In light of our results, we hypothesized that perlecan is a functional ECM partner of PAI-1. A possible interaction between PAI-1 and perlecan was investigated using a solid-phase binding assay (Arroyo De Prada et al., 2002). This suggests that soluble PAI-1 binds both perlecan and vitronectin (Fig. 3I). Pre-incubation with uPA blocks binding of PAI-1 to both ECM molecules, demonstrating the specificity of this interaction (supplementary material Fig. S3C). We then investigated whether PAI-1 interaction with perlecan involves the vitronectin binding region present on PAI-1 or its heparin binding domain, since perlecan contains heparan sulfate chains. Using a deficient vitronectin-binding PAI-1 mutant (Q123K) or by pre-incubating PAI-1 with a heparin solution that blocks binding to HS, we prove that PAI-1 interaction with perlecan does not involve the vitronectin binding region and is independent of heparin binding domain (supplementary material Fig. S3C,D). These experiments suggest that perlecan is a molecular partner of PAI-1 and that presence of perlecan in the ECM may trap PAI-1 by direct interaction. Since the binding region of PAI-1 to perlecan does not overlap vitronectin-binding site, our results suggest that perlecan and vitronectin may act in concert to increase PAI-1 concentration and stability in the ECM. Further studies would be required to determine the exact molecular interplay between these three binding partners to modulate the plasmin proteolytic activity.
Taken together, our results suggest that perlecan decreases uPA and uPAR levels at the cell membrane at least through two different mechanisms: (1) by regulating their expression; and (2) by facilitating PAI-1 deposition in the ECM thereby increasing uPA–uPAR complex turn-over. Accumulation of PAI-1 can thus inhibit a plasmin-dependent release of perlecan from the ECM.
Perlecan regulates keratinocyte adhesion and motility
Accumulation of both PAI-1 and perlecan in the ECM at cell confluence correlates with cell growth arrest, reduced cell movement, increased adhesion and differentiation. We used RNA interference on PAI-1 and perlecan to determine their possible involvement in mediating the strengthening of HK adhesion at confluence. A detachment assay disclosed that after PAI-1 or perlecan knockdown, post-confluent keratinocyte cultures detach faster in the presence of trypsin/EDTA than control HKs (Fig. 4A). This suggests that accumulation of PAI-1 and perlecan by confluent cells contributes to adhesion strengthening of the newly formed epithelium.
Possible involvement of PAI-1 and perlecan in inhibition of cell movement was also verified by measuring cell movement of individual PAI-1 or perlecan knockdown keratinocytes in post-confluent cultures. Quantification of the average cell speed demonstrated that both PAI-1 and perlecan knockdowns partially counteracts the cell motility arrest (Fig. 4B,C). In our experimental conditions, we failed to detect any influence of PAI-1 or perlecan deficiency on cell growth, differentiation or apoptosis (not shown). These findings therefore suggest that PAI-1 and perlecan deposition to the ECM improves cell adhesion and reduces cell movement of HKs reaching confluence.
Accumulation of perlecan in the BM at the wound margin correlates with BM maturation and stable cell adhesion
To confirm that perlecan accumulation in the BM requires down-regulation of the Plg activation cascade and maturation of cell–cell junctions, we analyzed deposition of perlecan at the dermal-epidermal junction of wounded skin in full-thickness excision wounds in mice (schematically represented in Fig. 5A). Indeed, it has been reported that hyperproliferating and migrating keratinocytes (HE) at the advancing front express high level of uPA during the re-epithelialization process, whereas uPA expression is down-regulated at the wound margins of the healed epidermis (Grøndahl-Hansen et al., 1988; Rømer et al., 1991).
Deposition of perlecan was strongly up-regulated at the wound margin (arrows) compared to normal unaffected skin (UE) but failed to accumulate at the dermal-epidermal junction underneath proliferative and migrating keratinocytes (HE) (Fig. 5A and supplementary material Fig. S4A). In the HE area, characterized by strong expression of proliferation and stress marker (Ki67 and keratin 6) (supplementary material Fig. S4B,C), signal corresponding to perlecan was more diffuse and stained the granulation tissue. By contrast, Lm-332 expression was increased at the BM all over the wound, as expected owing to its role during migration of keratinocytes on the provisional matrix. These observations were confirmed by quantification of the intensity of signals corresponding to Lm-332 and perlecan in the BM region zone (Fig. 5B). Localization of nidogen, another BM protein, and BP230, a component of stable attachment structures that links the basement membrane to the keratin intermediate filaments, called hemidesmosomes (Litjens et al., 2006), were also analyzed by immunostaining. Both proteins displayed a diffuse and reduced staining in proliferating and migrating keratinocytes (HE) compared to normal control skin (Fig. 5C,D). Interestingly the intense deposition of nidogen and basal localization of BP230 coincides with accumulation of perlecan at the wound margin (Fig. 5C,D, arrows). Since plasmin activation is known to be present in the granulation tissue at the advancing front of the wound (Grøndahl-Hansen et al., 1988; Rømer et al., 1991; Yi et al., 2001), we analyze the activity of plasmin in the different regions of the wound using an in situ activity assay (Fig. 5E). Using this approach, we show that plasmin activity is high at the advancing front of the wound compare to the non-affected skin but decrease significantly at the wound margin region where perlecan accumulates.
Taken together, these results support our in vitro observations. Indeed, wounds immunostaining shows that perlecan failed to accumulate at the dermal epidermal junction in the tip of advancing epidermis where Plg activation cascade is activated. Conversely, reduced plasmin activity at the wound margin confirms accumulation of perlecan. Interestingly, increased BM-accumulation of perlecan specifically in the margin region compared to non-affected skin suggests that this proteoglycan plays an active role in the formation of a new BM and thereby adhesion of basal keratinocytes.
Maturation of cell–cell junctions at confluence is crucial to regulate BM organization
At confluence, formation of a continuous monolayer favors maturation of cell–cell adhesion complexes. Intercellular adhesion structures comprise adherens junctions (AJs), desmosomes, tight junctions and gap junctions. Among these adhesion complexes, it has been postulated that AJs might have a function in sensing epidermal cell density and restricting cell proliferation when cells reach confluence (Lien et al., 2008; Perez-Moreno and Fuchs, 2006). AJs are composed of cadherins, which are transmembrane molecules that indirectly bind actin cytoskeleton by interacting with catenins cytoplasmic adapters. Since AJs are molecular sensors of confluence, we hypothesized that these adhesion structures may impact ECM deposition. We first determined whether in our conditions confluence influences AJ maturation. To achieve this goal, we analyzed the localization of both α- and β-catenins at the cell borders by immunochemistry in pre- and confluent keratinocytes cultures (Fig. 6A). Interestingly, immunostaining of α- and β-catenins is more intense at the cell–cell junction in confluent compared to pre-confluent keratinocytes, suggesting that confluence reinforces AJs. To determine whether confluence affects the expression of these molecules, levels of expression of catenins and connexin 43 (Cx43), a cell–cell adhesion molecule involved in gap junction formation, was analyzed by western blot (Fig. 6B,C). Immunoblotting of total proteins revealed that the expression level of both catenins is not modified by confluence whereas Cx43 is strongly induced (Fig. 6B). Detergent solubility assay indicated that the percentage of α- and β-catenins associated with the detergent-insoluble cytoskeletal fraction is increased at confluence, confirming the immunostaining observation (Fig. 6C). These results indicate that confluence favors maturation of AJs and other cell–cell adhesion structures including gap junction.
We then tested the possible involvement of AJs in BM organization and control of Plg activation cascade by invalidating the expression of α-catenin in keratinocytes. Indeed, by providing a link between cadherin and the cytoskeleton, α-catenin plays a major role in the maintenance of AJs (Vasioukhin et al., 2001). Inhibition of α-catenin expression by RNAi resulted in reduction of PAI-1 and perlecan deposition to the ECM, without significantly affecting Lm-332 deposition (Fig. 6D). Knockdown of α-catenin also enhanced uPA activity detected in the conditioned media of HKs (Fig. 6E). These results indicate that integrity of AJs is crucial for the organization of ECM as well as for regulation of the Plg cascade at confluence.
We also evaluated whether cell–cell junction maturation correlates with BM structuration in vivo (Fig. 6F). The localization of α-catenin was analyzed by immunostaining on 7 day wounds during the wound-healing process. α-catenin signal is lower and fuzzy at the cell–cell junction of the hyperproliferative epithelia (HE), located at the advancing front of the wound, as compared to the wound margin area or to unaffected epidermis (UE). These observations suggest that AJ maturation correlates in vivo with the deposition of BM components, suggesting that maturation of cell–cell junctions is crucial to regulate BM reorganization both in vitro and in vivo.
Activin A and TGF-β regulate perlecan and PAI-1 ECM deposition at confluence
Induction of PAI-1 synthesis at confluence (Fig. 2A) is a crucial event to facilitate shut down of the Plg activation cascade and perlecan ECM accumulation. We investigated the signaling pathway involved in regulation of both PAI-1 and perlecan at confluence. TGF-β is the most prominent factor regulating expression of PAI-1, and other ECM proteins, including Lm-332, βig-h3 and perlecan. This cytokine belongs to a large family comprising three TGF-β isoforms, activins, nodal and BMPs, which play an important role in skin development, wound repair and carcinogenesis (Owens et al., 2008). Since we could verify that PAI-1, perlecan and βig-h3 were up-regulated at the mRNA level by cell confluence (supplementary material Fig. S5A), we assessed the involvement of TGF-β in these inductions. Treatment of HKs with a TGF-β type I receptor inhibitor (SB 431542) resulted in a dose-dependent reduction of perlecan, PAI-1 and βig-h3 deposition in the ECM at confluence, thereby establishing a critical contribution of this signaling pathway the up-regulation of these molecules (Fig. 7A). We then assessed whether cell confluence modulates expression of the TGF-β cytokine family. Expression of the different activins and TGF-β members was monitored by Q-PCR in pre- and post-confluent keratinocytes. Activin βA and, to a lesser extent, the TGF-β1 and β2 isoforms were induced in post-confluent HKs (Fig. 7B), while expression of activin βB, inhibin α, and TGF-β3 was not detected in our conditions (data not shown). Moreover, the steady-state mRNA level of follistatin, which inhibits binding of activin A to its receptor, was dramatically decreased in confluent keratinocytes (Fig. 7B), suggesting that activin A signaling is enhanced.
To further explore involvement of TGF-β1 and activin A in deposition of ECM proteins and inhibition of the Plg activation cascade, pre-confluent HKs were grown in presence of either TGF-β1 or activin A. Addition of TGF-β1 to pre-confluent HKs was sufficient to enhance deposition of PAI-1, βig-h3 and perlecan into the ECM (Fig. 7C). In contrast, activin A, while exerting a minor effect on PAI-1 and βig-h3, triggered a strong increase in perlecan deposition (Fig. 7C). Q-PCR analysis of mRNA isolated from TGF-β1 or activin A treated pre-confluent HKs confirmed the results obtained by western blot (Fig. 7D). The effect of activin A on the regulation of perlecan was verified in vivo using transgenic mice expressing activin βA in the keratinocyte basal layer of the epidermis (Munz et al., 1999). As shown in Fig. 7E, perlecan strongly accumulated at the dermal-epidermal junction of transgenic mice (arrows) compared to control animals. Altogether, these results demonstrate that TGF-β1 and activin A induction at cell confluence act synergistically to efficiently regulate deposition of ECM components and inhibit the Plg activation cascade.
Since activin A is the most efficient cytokine to induce perlecan deposition, we analyzed the impact of activin A treatment on cell adhesion and cell movement capacities of pre-confluent keratinocytes. As expected, activin A treatment significantly improves adhesion of keratinocytes and reduces their motility (supplementary material Fig. S5B). Moreover, these biological effects of activin A are blocked in perlecan knockdown keratinocytes (Fig. 7F), indicating that perlecan mediates, at least in part, the effect of activin A on cell adhesion and cell movement.
Using a comparative proteomic approach, we report here that generation of a stratified epithelium is associated with deep ECM modifications and inhibition of proteolysis pathways. In particular, we prove that accumulation of PAI-1 in the ECM during epithelia formation inhibits the Plg activation cascade, facilitates deposition of ECM proteins, including perlecan, and improves adhesion of keratinocytes.
We show that at confluence perlecan accumulates into the ECM and such accumulation requires down-regulation of the Plg activation pathway both in vitro and in vivo. Presence of perlecan in almost all BM and its ability to interact with other BM components suggest that perlecan is involved in BM assembly. However, invalidation of perlecan expression in mice revealed that perlecan is dispensable for BM formation but is rather required to stabilize BM and provide resistance to physical stress (Costell et al., 1999). Here we show that deposition of perlecan in the ECM increases keratinocyte adhesion and reduces cell migration capacity. Potential role of perlecan in adhesion and BM stabilization is also supported by our immunofluorescence observations showing that perlecan deposition in the BM occurs at the wound margin, where BM matures and favors structuration of hemidesmosomes, which is a hallmark of stable adhesion of the newly formed epithelium. In keratinocytes, perlecan may control cell adhesion by distinct mechanisms, because the large protein core and the heparan-sulfate chains of perlecan take part in numerous interactions with cell surface receptors and other ECM components (Iozzo, 2005). The two identified perlecan receptors, α2β1 integrin and dystroglycan, are expressed in keratinocytes (Herzog et al., 2004; Watt, 2002) and therefore can provide a physical link between the cytoskeleton and the BM. Interestingly, in the skin dystroglycan is known to associate with hemidesmosome (Herzog et al., 2004), probably through its interaction with plectin (Rezniczek et al., 2007). Accordingly to our data, these observations support the idea of a possible contribution of the perlecan/dystroglycan complex to the reinforcement of keratinocyte adhesion.
Cell confluence is associated with down regulation of the plasmin-dependent proteolysis and simultaneous accumulation of perlecan and PAI-1. We demonstrate that perlecan accumulation requires inhibition of plasmin dependent proteolysis. Inversely, we show that perlecan regulates plasmin-dependent proteolysis by controlling ECM deposition of PAI-1 and down-regulation of both uPA and uPAR mRNAs. This idea is consistent with the suggested role of perlecan in control of proteolysis that was inferred from the perlecan-null mouse phenotype, which displays a reduced accumulation of the fibrillar collagen network consequent to excessive ECM degradation (Costell et al., 1999). Our data suggest an intriguing positive-feedback loop mechanism, which promotes perlecan accumulation in the ECM and therefore enhances BM formation. Hence, our results demonstrate that perlecan may interfere with the Plg activation pathway by regulating expression of uPA/uPAR and by stabilizing PAI-1 in the ECM possibly through a direct interaction. How perlecan regulates uPA and uPAR mRNA level and what is the functional significance of PAI-1 interaction with perlecan are still an open question.
We demonstrate in our model that PAI-1 accumulation inhibits cell movement by facilitating Plg cascade inhibition and thereby, perlecan accumulation. The role of PAI-1 in cell migration is a matter of controversy since both stimulatory and inhibitory functions have been reported (Degryse et al., 2004; Degryse et al., 2001). The study of PAI-1 function in migration is complicated since this molecule influences migration through proteolytic and non-proteolytic functions by interacting with vitronectin and LDL receptor-related protein (LRP) (Czekay and Loskutoff, 2004; Czekay et al., 2011). In apparent contradiction with our results, recent studies demonstrate that PAI-1 expression is required to induce optimal migration of keratinocytes (Providence and Higgins, 2004; Providence et al., 2008). This discrepancy may be explained by a dose dependent mechanism since some PAI-1 cellular effects are dose dependent (Bajou et al., 2004; Devy et al., 2002). Devy et al. show that presence of PAI-1 is pro-angiogenic whereas at high concentration it becomes anti-angiogenic (Devy et al., 2002). The role of PAI-1 in cell migration could also be dependent of the cellular context that influences the stoichiometry between PAI-1 and its molecular partners including vitronectin, uPA/uPAR, and LRP.
Our data suggest that establishment of mature AJs at confluence is required to induce deposition of both PAI-1 and perlecan in the ECM and to down regulate Plg dependent proteolysis. In agreement with our results, ablation of α-catenin in mice skin leads to abnormal deposition of Lm-332 (Vasioukhin et al., 2001), suggesting that defect of AJs in vivo leads also to abnormal ECM organization. Moreover, keratinocytes derived from α-catenin-null mice display an invasive phenotype and form carcinoma, suggesting that proteolytic pathways involved in pericellular proteolysis, possibly including the Plg activation cascade, are activated (Kobielak and Fuchs, 2006; Vasioukhin et al., 2001). Loss of another AJ component, p120-catenin, leads also to absence of contact growth inhibition and development of carcinoma in transgenic animals associated with aberrant BM organization (Perez-Moreno et al., 2008). This observation suggests that alteration of other AJ components may also affect deposition of ECM component. An intriguing question is to known whether absence of α-catenin affects ECM organization through alteration of cadherin stabilization or as regulator of cell signaling. Indeed, cadherin-catenin complexes are not only structural linkers but also important signaling centers (Lien et al., 2006; Lien et al., 2008). Altogether, these data support the idea that establishment of mature cell–cell junction maintain tissue integrity by reinforcing tissue cohesiveness and inhibiting proliferative signals and also by facilitating organization of BM to reinforce stable cell-ECM contacts.
Our study suggests that TGF-β and activin A induction at confluence plays a crucial role in down-regulation of the Plg proteolytic cascade and in the BM organization. These two cytokines are known to be regulators of ECM protein synthesis and inducers of keratinocyte growth arrest and differentiation. Deregulation of activin A signaling impairs normal tissue development and healing process in vivo (Munz et al., 1999; Sulyok et al., 2004) and is observed in tumorigenesis (Risbridger et al., 2001). Here we demonstrate that activin A is the prominent inducer of perlecan expression and deposition in the ECM layered down by keratinocytes both in vitro and in vivo. For the first time, our findings involve activin A in the control of the Plg activation cascade and in regulation of the BM organization. This novel function of activin A should help to revisit the role of this cytokine in a variety of physiological and pathological situations. How does cell confluence induce these cytokines remains an open question. Since we demonstrate that cell–cell junction maturation, and in particular AJs, is an important prerequisite to ECM organization it is tempting to speculate that a cell–cell junction dependent mechanism may be involved in the induction of TGF-β family cytokines. Further work will be required to address how these intercellular adhesion complexes are involved in the control of BM assembly via TGF-β family cytokines.
In summary, we demonstrate a novel molecular pathway responsible for suppression of the extracellular proteolytic activity in confluent keratinocytes and formation of a mature BM (Fig. 7G). Indeed, perlecan deposition in the ECM requires maturation of AJs and is driven by upregulation of activin A/TGF-β that occurs upon cell confluence. Accumulation of perlecan is directly involved in regulating the adhesive strength and motility of keratinocytes. We also demonstrate that ECM-bound perlecan facilitates recruitment of PAI-1, which results in suppression of the uPA–uPAR–plasmin cascade. We show that plasmin suppresses perlecan deposition in the ECM. Inversely, increased expression and matrix accumulation of PAI-1 stabilizes perlecan in the ECM, thereby permitting BM formation. This regulatory mechanism may occur during wound healing because accumulation of perlecan in the BM at the wound margin correlates with down-regulation of plasmin activity, restoration of the BM and stable cellular adhesion. Our model suggests that cell confluence, by allowing AJ maturation and regulating cytokines expression, is crucial in the control of BM formation. Alteration of this circuitry may lead to uncontrolled plasmin-dependent proteolysis and BM modification, resulting in pathological outcomes including chronic wounds, bullous diseases and tumorigenesis.
Materials and Methods
Reagents and antibodies
For Western blotting or immunofluorescence, antibodies against laminins (β3, γ1, γ2), PAI-1, vitronectin, p21, p27, cyclin A, Erk-2, GAPDH, Ki67 were purchased from Santa Cruz Biotechnology (Santa Cruz, CA); anti-tubulin from Sigma-Aldrich (Saint-Quentin Fallavier, France); anti-perlecan from Zymed Laboratories (San Francisco, CA, USA) and Thermo Fisher Scientific (Illkirch, France); anti-uPAR from American Diagnostica (Stamford, CT); anti-β-catenin and Connexin-43 (Cx43) from BD transduction laboratories (Franklin Lakes, NJ); anti-uPA from American Diagnostica and Abgent (San Diego, CA, USA); anti-Keratin-14 from Millipore (Billerica, MA); anti-BrdU from Roche Diagnostic (Indianapolis, IN); anti-cytokeratin-6 and anti-α-catenin from Epitomics (Burlingame, CA). Fluorescent and HRP-conjugated secondary antibodies were from Molecular probes and DAKO respectively. Chemical inhibitors used: SB431542 [TGF-β RI inhibitor, Tocris bioscience (Ellisville, MO, USA)], sodium chlorate (HS biosynthesis, Sigma-Aldrich), GGACK (uPA inhibitor), D-Val-Phe-Lys Chloromethyl Ketone [plasmin inhibitor, Merck-Calbiochem (Darmstadt, Germany)] and cycloheximide (CHX) (synthesis inhibitor, Sigma-Aldrich).
Primary human keratinocytes (HKs) or Lm-332-null JEB keratinocytes with null mutations in the genes coding for laminin β3 or γ2 were obtained from full-thickness skin biopsies. Wild-type (Wt-Ker) and JEB keratinocytes (JEB-Ker) were immortalized by transduction with an HPV16 E6E7 cDNA as described elsewhere (Baudoin et al., 2005). Immortalized and non-immortalized keratinocytes were grown on lethally irradiated feeder layers of mice fibroblasts in FAD medium, as previously described (Rheinwald and Green, 1975). A431 cells were maintained in DMEM (Gibco-BRL/Invitrogen, Cergy Pontoise, France) containing 10% fetal calf serum (Perbio/Thermo Fisher Scientific, Illkirch, France) and grown in FAD medium on lethally irradiated feeder layers when indicated. In some experiments, cells were treated with 5 ng/ml of TGF-β1 or activin A (Peprotech, Neuilly-Sur-Seine, France) in FAD medium.
Infection of immortalized JEB-Ker with a retroviral vector expressing Lm-β3 was performed as described (Gagnoux-Palacios et al., 2005). Briefly, infectious supernatants supplemented with polybrene were added to actively growing cells at high MOI followed by spinoculation procedure.
Mass spectroscopy analysis
ECM proteins were prepared as described in immunoblotting section. Equivalent amount (based on cell counting) of ECM samples corresponding to pre- (d5) and post-confluent (d9) HK cultures were separated on one dimension SDS-PAGE. Gels were stained with Biosafe Coomassie stain (Bio-Rad, Marnes-la-Coquette, France), cut manually into 30 slices that were individually digested with trypsin using a standard protocol. Peptides were separated on a nanoHPLC system fitted with an Ion trap (LCQ deca XP, Thermo Fisher Scientific). Mass spectrometry and data analysis were performed as previously described (Chiellini et al., 2008).
siRNA and cell transfection
PAI-1 and perlecan silencing were obtained using individual sequences ON-TARGETplus (Dharmacon/Thermo Fisher Scientific) (for PAI-1: 5′-CGACAUGUUCAGACAGUUU-3′; for perlecan: 5′-GGUGGGAAGUUGCGAUACA-3′) and siRNA targeting luciferase as control (siCtrl: 5′-CGUACGCGGAAUACUUCGA-3′). α-catenin silencing was performed using ON-TARGETplus SMARTpool. For siRNA transfection, cells were transiently transfected using RNAiMax lipofectamine reagent (Invitrogen) according to the manufacturer's procedure.
Immunofluorescence, in situ plasmin activity, BrdU staining and immunoblot analyses
For immunolabeling, cells were grown on glass coverslips, fixed in 3.7% formaldehyde followed by methanol treatment, and then stained as previously described (Gagnoux-Palacios et al., 1996). For α- and β-catenin immunolabeling cells were pretreated for 5 min with PBS containing 0.5% Triton X-100 before fixation. Fresh frozen sections of mouse skin from WT or K14-activin βA mice [previously described by Munz et al. (Munz et al., 1999)] were incubated with indicated antibodies before incubation with fluorescent-conjugated secondary antibodies (Molecular Probes/Invitrogen) and DAPI (Sigma-Aldrich). For BrdU staining, cells were cultured in presence of 10 µM of BrdU, and then stained with anti-BrdU antibody. Quantifications were done using ImageJ software plot profile plugin, signal intensity was measured over a line of 16 µm placed perpendicularly to the dermo-epidermal junction and centered on the junction. For in situ plasmin activity assay, a reaction solution (50 mM Tris-HCl, pH 7.6, 150 mM NaCl) containing a fluorescent plasmin substrate (D-Ala-Leu-Lys-AMC (Bachem) at 1 mg/ml was mixed with a bis/acrylamide solution and ammonium persulfate and Temed (Sigma-Aldrich). Plasmin inhibitor (D-Val-Phe-Lys chloromethyl ketone at 10 µM) was added in control slide to verify the specificity of the reaction. The reaction mixture was poured on 10 µm fresh frozen sections of mouse wounded and covered immediately by a glass coverslip. Slides were incubated in humidified chamber at 37°C and lysis of the substrate was assessed by examination under a fluorescent microscope. For immunoblotting analysis, cells were lysed in lysis buffer (50 mM Tris-HCl, pH 8, 150 mM NaCl, 2 mM EDTA, 1% NP-40, 0.5% deoxycholate) containing phosphatase and protease inhibitors. ECM proteins were prepared by scraping the plastic tissue culture dish in sample buffer under reducing conditions (2% β-mercapto-ethanol). Spent culture medium was precipitated and solubilized in Laemmli buffer. Proteins were separated in SDS-polyacrylamide gel electrophoresis followed by classical immunoblotting procedures. Quantification of perlecan in the ECM was done using a slot-blot procedure.
Detergent solubility assay
Detergent solubility assay on pre- or confluent keratinocytes was performed as described (Gimond et al., 1999). Briefly, cell cultures were lysed 10 min on ice in 1% Triton-X100, 50 mM Tris-HCl pH 7.6, 150 mM NaCl, 2 mM EDTA in the presence of protease inhibitors and cell lysates were centrifuged at 14,000 r.p.m. for 15 min to obtain the soluble fraction of protein. The pellet represents the cytoskeleton-associated protein (insoluble fraction) and was resuspended in Laemmli buffer. 50 µg of soluble protein of each condition and comparable amount of insoluble protein were separated in SDS-polyacrylamide gel electrophoresis followed by classical immunoblotting procedures.
uPA fluorimetric assay and zymography
uPA activity was determined by monitoring the hydrolysis of a chromogenic substrate (H-D-Val-Leu-Leu-Lys-p-nitroanilide, Sigma-Aldrich) and total uPA was quantified by zymography (Mangnall et al., 2004). uPA bound to uPAR at the keratinocytes cell surface were released by incubation of the cells 1 min in presence of acidic buffer (100 mM glycine, pH 3, 150 mM NaCl) followed by zymography analysis.
For solid-phase binding assay (Arroyo De Prada et al., 2002), 96-multiwell ELISA plates were incubated with solutions of purified rat collagen I (Gibco-BRL/Invitrogen), vitronectin (Biosource/Invitrogen), perlecan (Sigma-Aldrich), laminin-111 (Millipore). After washing and saturation of non-specific sites, recombinant purified PAI-1 (Merck-Calbiochem) or PAI-1 Q123K mutant [generously provided by Paul Declerck (Eren et al., 2007; Vleugels et al., 2000)] was added to each well. In competition experiment, uPA (100 U/ml) or heparin (1 U/ml) were pre-incubated with PAI-1 before incubation with ECM proteins. Binding of PAI-1 to the solid phase was detected using a specific PAI-1 rabbit antisera (Calbiochem) followed by incubation with a secondary anti-rabbit-HRP antibody and a colorimetric assay. Values were expressed as OD at 450 nm.
Cell detachment assay
HK cultures were treated with a solution of trypsin/EDTA as reported (Gagnoux-Palacios et al., 2001) or with a non-enzymatic solution containing 0.5 mM-EDTA. The number of cells detached at increasing time of incubation was determined by collecting the cell culture supernatants at different times followed by direct cell counting of triplicates. To allow cell counting, trypsin was added to the cell suspension when cells were detached with EDTA only. The graphs represent means ± s.e.m. of the percentage of detached cells at one time of one representative experiment.
Cell migration assay
To analyze single-cell migration, phase-contrast images of pre- or post-confluent or siRNA transfected HKs were filmed using a videomicroscope (Carl Zeiss, Le Pecq, France). Cell migration was determined by manual tracking of individual cells among the epithelial sheet using the ImageJ software with the ‘manual tracking’ additional plug-in. Average velocity of single-cell migration was calculated from >40 cells per experiment. The graphs represent the means ± s.e.m. of either one representative experiment (out of three independent experiments) or the mean of three experiments as indicated in the figure legend.
RT-PCR and quantitative PCR analysis
For semi-quantitative RT-PCR, PCR products obtained with specific primers were run in agarose gels followed by visualization after ethidium bromide staining. For quantitative RT-PCR (Q-PCR), mRNA expression levels were quantified by real time one-step reverse transcription-PCR using SYBR-Green PCR Master Mix (Applied Biosystems/Life Technologies, Carlsbad, CA) and using RPLP0 mRNA for normalization. Each experiment was repeated three times and a representative experiment or a mean of three experiments is shown as indicated in the figure legends. The oligonucleotide sequences are available upon request.
Data were analyzed using a homoscedastic two-tailed t-test. P<0.05 was considered statistically significant.
We thank John F Marshall for critically reading the manuscript. We are grateful to Sabine Werner and Maria Antsiferova (Department of Biology, Institute of Cell Biology, Zurich, Switzerland) for providing us with skin samples of K14-ActA transgenic mice; to Paul Declerck (Laboratory for Pharmaceutical Biology, Leuven, Belgium) for providing us with recombinant PAI-1 proteins and Agnès Loubat and Isabelle Bourget for technical assistance.
↵‡Present address: CNRS, UMR 5248, GBMN, Université de Bordeaux, Allée de St Hilaire Bât B14, 33600 Pessac, France
This work was supported by the DEBRA Foundation (UK); Agence Nationale de la Recherche [grant number ANR-0378]; and an INSERM-Region fellowship to A.B.
Supplementary material available online at http://jcs.biologists.org/lookup/suppl/doi:10.1242/jcs.096289/-/DC1
- Accepted April 21, 2012.
- © 2012. Published by The Company of Biologists Ltd