Reduced migration, altered matrix and enhanced TGFβ1 signaling are signatures of mouse keratinocytes lacking Sdc1

Syndecan-1 (Sdc1) is a single-pass, integral-membrane, heparan-sulfate proteoglycan that is abundantly expressed by epithelial keratinocytes early in development (Bernfield et al et al., 1999). Although the Sdc1-null mouse was found to be fertile and viable, when in vivo wound-healing studies were conducted, phenotypes began to emerge (Stepp et al., 2002; Gotte et al., 2002; Gotte et al., 2005). In cornea and skin of Sdc1-null mice, wound healing was found to be delayed due to slow re-epithelialization and increased inflammation. These mice also showed increased cardiac dysfunction after myocardial infarction associated with increased inflammation, matrix metalloproteinase (MMP) expression and function, and increased collagen disorganization (Vanhoutte et al., 2007). Paradoxically, overexpression of Sdc1 (Sdc/Sdc) in mice was also found to delay skin wound healing because of reduced cell proliferation, granulation tissue accumulation, and poor vascularization (Elenius et al., 2004). Together, these studies show that correct homeostasis of stratified squamous epithelial tissues demands that Sdc1 levels are precisely regulated.

Of the four known syndecans, Sdc1 is the most abundant on epithelial keratinocytes and can mediate cell-to-substrate adhesion by its ability to bind to laminin α chains (Salmivirta et al., 1994; Hoffman et al., 1998; Klass et al., 2000; Utani et al., 2001; Okamoto et al., 2003). In addition, the Sdc1 cytoplasmic domain has been shown to mediate cell spreading and migration (Chakravarti et al., 2005). whereas Sdc4 has been shown to interact with integrins at focal adhesions during wound healing (Alexopoulou et al., 2007), less is known about how Sdc1 mediates migration. Sdc1 has been shown to interact functionally with αv integrins in various cancer cell lines (Beauvais et al., 2004; Beauvais and Rapraeger, 2004; McQuade et al., 2006), and a recent report by Hayashida and colleagues has shown that expression of Sdc1 in epithelial keratinocytes is induced by TGFβ1 through a protein kinase A (PKA)-dependent pathway (Hayashida et al., 2006). In response to injury in vivo, genes for TGFβ1 and Sdc1 are upregulated in both epithelial and mesenchymal cells, and the Sdc1 ectodomain is shed at wound sites where the soluble molecule can regulate chemokine function (Gotte and Echtermeyer, 2003; Tkachenko et al., 2004) and modulate the activation of various MMPs (Kelly et al., 2000; Steffensen et al., 2001; Momota et al., 2005). The role Sdc1 plays in cancer is complex and tissue specific, but recent studies using Sdc1-null mice show that these mice are resistant to mammary carcinogenesis (Alexander et al., 2002; McDermott et al., 2007).

This study was undertaken to investigate the consequences of the loss of Sdc1 on keratinocyte function in vitro, and to identify alterations relevant to the delayed wound-healing phenotype observed in cornea and skin in vivo. Our results show that the loss of Sdc1 on activated keratinocytes enhances the overall amount of several integrins at the keratinocyte surface, increases adhesion, delays migration and causes an increase in the constitutive level of TGFβ1-mediated gene expression. Furthermore, the response of Sdc1-null keratinocytes to TGFβ1 is enhanced compared with that of wild-type (wt) keratinocytes. We also found that the reduced migration of Sdc1-null keratinocytes can be overcome by replating the keratinocytes on permissive substrates such as fibronectin–collagen-I (FNCNI), LN-332, matrix made from wt keratinocytes, or by treating Sdc1-null keratinocytes with TGFβ1 or antibodies that block the function of α6 or αv integrins. Our results identify TGFβ1 signaling and Sdc1 expression as important factors in activation and migration of keratinocytes in vitro and in vivo.

In this study, primary Sdc1-null keratinocytes grew well in culture (Fig. 1A), and growth curves show that the Sdc1-null keratinocytes had equivalent plating efficiencies and reached confluence at the same number of days after seeding (Fig. 1B). However, the Sdc1-null keratinocytes achieved a higher population density than wt keratinocytes due to increased keratinocyte proliferation (Fig. 1C) and increased packing of cells in and around colonies (compare wt and Sdc1-null keratinocytes in Fig. 1A). Since studies using antisense methods (Kato et al., 1993) and tumor cell lines (Bayer-Garner et al., 2001; Kurokawa et al., 2006) have shown that loss of Sdc1 can cause epithelial cells to adopt a mesenchymal phenotype, wt and Sdc1-null keratinocytes were assessed for their localization of epithelial keratins (K-1, K-5, K-6, K-10, and K-14), E-cadherin, vimentin, and F-actin. Shown in Fig. 1D are keratinocytes stained simultaneously with an anti-K-14 antibody, phalloidin (for F-actin) and with the nuclear marker DAPI. The Sdc1-null keratinocytes retained epithelial morphology and continued to express K-14. They also expressed E-cadherin and other epithelial keratins and remained vimentin-negative (data not shown). However, differences in cytoskeletal organization can be seen; Sdc1-null keratinocytes showed thicker cortical actin bundles at cell peripheries compared with wt keratinocytes.

In vivo, re-epithelialization of wounds has been shown to be delayed by 6-8 hours in Sdc1-null mice (Stepp et al., 2002); in the current study, adhesion and migration studies were performed to determine whether reduced migration rates after wounding in vivo correlated with altered properties of the keratinocytes in vitro. After 3 days in culture, equal numbers of wt and Sdc1-null keratinocytes were harvested and allowed to adhere to ECM-coated substrates for 60 minutes (Fig. 2A). Adhesion to all the matrices tested – fibronectin (FN), vitronectin (VN), laminin-111 (LN-111), collagen I (CNI) and collagen IV (CNIV) – was enhanced in the Sdc1-null keratinocytes compared with the wt keratinocytes; increased adhesion ranged from 3.3-fold higher for adhesion to VN to 1.5- to 2.0-fold higher for adhesion to the other matrix ligands. Because adhesion assays measure both cell attachment and spreading, the increased adhesion of the Sdc1-null keratinocytes on purified matrices could be owing to a combination of increased attachment and spreading of the Sdc1-null keratinocytes. When spreading studies were performed at 60 minutes after plating, there were no significant differences in keratinocyte spreading (data not shown). Since the adhesion assays presented in Fig. 2A were performed on keratinocytes 60 minutes after plating, the increased adhesion of the Sdc1-null keratinocytes seen is probably due to increased keratinocyte attachment.

The attachment differences seen between the wt and Sdc1-null keratinocytes could impact the migratory behavior of Sdc1-null keratinocytes. To assess this, keratinocytes were seeded and grown for 5 days, and random movement of keratinocytes within wells in 24-well plates was assessed using time-lapse microscopy. At day 5, the Sdc1-null keratinocytes were found to migrate significantly slower than wt keratinocytes, as seen in the average velocities presented in Fig. 2B and the representative tracks presented in Fig. 2C. When data were evaluated for differences in persistence indices (net migration per total migration), no differences between wt and Sdc1-null keratinocytes emerged.

The slower velocity of the Sdc1-null keratocytes could be caused by altered cytoskeletal dynamics or by their increased attachment to their substrate. To test this we evaluated cell migration after replating keratinocytes onto other matrices; data are presented in Fig. 2D. Keratinocytes were grown for 3 days and replated on purified FN-CNI or the LN-332-rich matrix secreted by 804G cells; as controls, migration rates of keratinocytes were also obtained for keratinocytes at day 4 after initial seeding. Day 3 keratinocytes were also replated onto wells that contained matrix (prepared as described in Materials and Methods) that had been produced and deposited by either wt or Sdc1-null keratinocytes. Triplicate wells were then used to track keratinocyte migration in time-lapse experiments. As expected, the day-4 Sdc1-null keratinocytes migrated more slowly than wt keratinocytes. When Sdc1-null keratinocytes were replated onto FN-CNI, LN-332 or onto the matrix deposited by wt keratinocytes, their migration rate was restored to that of wt keratinocytes; however, when replated onto their own matrix, they continued to migrate more slowly. Wild-type keratinocytes migrated at rates faster than those of control keratinocytes when replated onto day-3 Sdc1-null keratinocyte matrix. Therefore, these data show that the Sdc1-null keratinocyte matrix can support optimal cell migration when keratinocytes express Sdc1 and indicate that there is something distinct about the way the Sdc1-null keratinocytes interact with their matrix that reduces their ability to migrate quickly.

To determine whether there are differences in the organization of the wt and Sdc1-null matrices, we assessed the accumulation of collagen in keratinocyte cultures using a Sirius Red dye binding assay (Heng et al., 2006). Keratinocytes that had been used for tracking studies were immediately fixed and the collagen accumulation within the cultures was assessed. After data normalization for differences in keratinocyte numbers per well, we found that significantly less collagen was present within the wells of the Sdc1-null keratinocytes (Fig. 3A). To determine whether there were differences in the overall profile of proteins present in the matrices deposited by wt or Sdc1-null keratinocytes, we prepared matrices using methods identical to those used for the experiments shown in Fig. 2D, and ran extracts normalized for keratinocyte numbers onto 4-20% SDS-polyacrylamide gels that were then silver stained (Fig. 3B). Since one of the most abundant proteins present in these matrix preparations is LN-332, we also blotted these extracts for detection of LN-332 using the J18 antibody, which recognizes both unprocessed and processed LN-α3, β3, and γ2 chains (Fig. 3C). We did not see any difference in the overall amounts of high molecular weight molecules assembled into matrix at days 5 and 7 by wt or Sdc1-null keratinocytes. We found no difference in the amount of LN-332 present in these matrices; differences in LN-332 processing cannot be quantified without mouse LN-332-subunit-specific antibodies. Next, we looked directly at the LN-332 organization within matrices produced by the wt and Sdc1-null keratinocytes by performing immunolocalization experiments on matrix preparations using the J18 antibody (Fig. 3D). Data confirmed that the wt and Sdc1-null keratinocytes deposited similar amounts of LN-332-enriched matrix. However, Sdc1-null keratinocytes deposited LN-332 into more highly ordered (arrowhead) arrays oriented in the direction of keratinocyte migration, whereas LN-332 left behind by wt keratinocytes was organized into cloud-like aggregates (arrowheads, Fig. 3D).

Thus far, our data show that Sdc1-null keratinocytes migrated more slowly than wt keratinocytes owing to factors involving their ability to interact with the matrix they deposit. α6β4 integrin is known to interact with LN-332 and has been implicated, together with α3β1 integrin, in mediating keratinocyte migration (Belkin and Stepp, 2000). To test directly whether increased activity of integrins on the Sdc1-null keratinocytes contributes to their reduced migration rate, we repeated time-lapse experiments using integrin-function-blocking antibodies: GoH3 for α6, 9EG7 for β1 integrins and RMV-7 for αv integrins. Data are presented in Fig. 4A as the fold change in velocity relative to untreated wt keratinocytes. Antagonizing the activity of either α6 or αv integrins fully restores velocity of Sdc1-null keratinocytes to the same or slightly higher rate than that of wt keratinocytes. Antagonizing β1 integrins slows down migration rates of both wt and Sdc1-null keratinocytes but the affect is more profound in wt keratinocytes where migration rates were reduced by just over 60% of the rates seen in untreated or control-IgG-treated wt keratinocytes. For Sdc1-null keratinocytes, the β1 antagonist decreased migration rates by ∼30%, not significant compared with that of untreated Sdc1-null keratinocytes. Thus, Sdc1-null keratinocytes migrate at slower rates due to reduced β1 integrin activity and/or increased α6β4 and/or αv integrin activity.

We went on to examine localization of α3 and β4 integrin in the wt and Sdc1-null keratinocytes. Finding only subtle differences in α3β1 localization (data not shown), we focused on colocalization studies of LN-332 and β4 integrin in wt and Sdc1-null keratinocytes. Whereas there are a wide variety of keratinocyte morphologies and β4 integrin and LN-332 localization profiles in primary keratinocyte cultures, differences in β4 integrin localization and cell morphology emerged when we compared localization of these two proteins in actively migrating wt and Sdc1-null keratinocytes surrounded by prominent LN-332 trails (Fig. 4B). Migrating wt keratinocytes were smaller and less uniform in shape compared with Sdc1-null keratinocytes, and there was less β4 integrin present at cell peripheries. LN-332 and β4 integrin are present beneath the cell nucleus in both migrating and stationary wt and Sdc1-null keratinocytes. In migrating Sdc1-null keratinocytes, we were frequently able to see close association of β4 integrin with LN-332 at cell peripheries, and Sdc1-null keratinocytes were less uniformly round in shape. For both the wt and Sdc1-null keratinocytes, those that were not actively moving, as evidenced by the absence of LN-332 trails, were more round with less β4 integrin present at cell peripheries. β4 integrin was present on the basal surface and surrounding the perinuclear region and, although it appeared more organized in the Sdc1-null keratinocytes, there was significant variability in this phenotype.

αv integrins, especially αvβ6 and αvβ8, are involved in mediating the activation of TGFβ1 signaling in epithelial cells (Sheppard, 2005). Little is known about how the activity of αv integrins might affect keratinocyte migration. TGFβ1 has long been known to alter the surface expression of integrins, including those containing the αv and β1 subunits (Gailit et al., 1994; Decline et al., 2003). Furthermore, Sdc2 has recently been shown to facilitate binding and activity of TGFβ1 on cell surfaces (Chen et al., 2004), and TGFβ1 signaling has been shown to induce Sdc1 expression through a PKA-dependent pathway (Hayashida et al., 2006). To determine whether TGFβ1 played a role in mediating the migration rates of the wt and Sdc1-null keratinocytes, we treated keratinocytes with a neutralizing antibody against TGFβ1 (TGFβ1NA) or with exogenous addition of TGFβ1, and measured keratinocyte migration rates using time-lapse microscopy. Data are presented in Fig. 5A,B. Addition of the TGFβ1NA to both wt and Sdc1-null keratinocytes reduced cell migration rates of both genotypes significantly; control IgGs at the same concentration had no affect on the rate of cell migration (data not shown). Despite the fact that TGFβ1NA reduced the migration rates of both wt and Sdc1-null keratinocytes, it had a more profound affect on migration rates of wt keratinocyte because it reduced wt migration to ∼55% that of untreated wt control keratinocytes and reduced migration of Sdc1-null cells by ∼30% compared with untreated Sdc1-null keratinocytes (Fig. 5A). These data suggest that TGFβ1 signaling plays a role in regulating the overall velocity of keratinocytes in primary culture, and further suggests that differences between wt and Sdc1-null keratinocytes in TGFβ1 signaling exist. Consistent with these data, addition of 0.25 ng/ml TGFβ1 to both wt and Sdc1-null keratinocytes increased their migration rates. Whereas wt keratinocytes were no longer able to increase their migration rates in response to higher doses (2.5 ng/ml) of TGFβ1, Sdc1-null keratinocytes migrated faster than untreated keratinocytes in response to the higher dosage of growth factor (Fig. 5B).

To get a better idea about how TGFβ1 affects migration of wt and Sdc1-null keratinocytes, we visualized β4 integrin and LN-332 in wt and Sdc1-null keratinocytes that had been treated either with TGFβ1NA or with 0.25 ng/ml TGFβ1 (Fig. 5C). TGFβ1NA-treated wt and Sdc1-null keratinocytes were similar in overall morphology and β4 integrin localization compared with untreated keratinocytes (compare Fig. 5C with Fig. 4B) and both genotypes showed increased close association of LN-332 with β4 integrin. The migrating TGFβ1-treated wt keratinocytes (Fig. 5Ce,f) were more spread out than untreated wt keratinocytes (Fig. 4Bb), whereas the migrating TGFβ1-treated Sdc1-null keratinocytes (Fig. 5Cg,h) were less well spread out compared with migrating untreated Sdc1-null keratinocytes (Fig. 4Be,f). Also, the close association of β4 integrin and LN-332 at the keratinocyte peripheries seen in migrating untreated Sdc1-null cells was decreased. These data suggest that treatments that restore Sdc1-null keratinocyte migration rates to levels similar to those of wt keratinocytes reduce the close association of β4 integrin with LN-332, whereas treatments that reduce wt and Sdc1-null keratinocyte migration enhance the close association of β4 integrin with LN-332.

The increased attachment and reduced migration rates observed in the Sdc1-null keratinocytes, and the changes in migration that accompany activation of TGFβ1 signaling by growth factor treatment could be due to altered expression of integrins in keratinocytes lacking Sdc1. To test this, we assessed total integrin expression in wt and Sdc1-null keratinocytes and found that, for each of the eight keratinocyte integrin subunits assessed by immunoblotting and after normalization for total protein and/or actin, there were no differences in expression in wt versus Sdc1-null keratinocytes (see supplementary material Fig. S1). Expression of α9 integrin in both wt and Sdc1-null keratinocytes was downregulated after keratinocytes were placed in culture and, therefore, could not be detected by immunoblotting.

Integrins are present within intracellular compartments as well as on the cell surface. To analyze integrin surface expression on wt and Sdc1-null keratinocytes, we initially used flow cytometry. Data are presented in Fig. 6A for β1 and β4 integrins and indicate 1.2 and 1.6 times more, respectively, of both integrins on the surface of untreated Sdc1-null keratinocytes compared with untreated wt keratinocytes. Since there are few antibodies available that detect extracellular epitopes of mouse integrins, we measured surface integrins by surface-labeling keratinocytes in suspension at 4°C using biotinylation, followed by immunoprecipitation (IP) with integrin antibodies and detection of biotin-labeled integrin heterodimers using horseradish peroxidase (HRP)-conjugated avidin. After demonstrating that the biotinylation efficiency of the Sdc1-null keratinocytes was similar to that of the wt keratinocytes – using a dot-blot technique and by assessing biotinylation of total protein extracts from both wt and Sdc1-null keratinocytes (Fig. 6B) – extracts were normalized based on equal amounts of total protein, and IP was performed. Although under the conditions used for the IPs both α and β subunits were pulled down, biotin labels primary amine groups and the numbers of biotins added per integrin molecule vary between the different subunits. As a result, frequently only one of the two subunits was detected in the unreduced mini-gels as shown in Fig. 6B; data were quantified and are shown in Fig. 6C.

Comparing integrin surface expression in wt and Sdc1-null keratinocytes, we show that Sdc1-null keratinocytes expressed significantly (1.3 to 1.5 times) more β4, αv, β6 and β8 integrins than wt keratinocytes. Because the difference between the surface expression of integrins on the wt and Sdc1-null keratinocytes was less than twofold, these experiments were repeated seven times to determine statistical significance. The increase in α2 integrin was not significant and α3 integrin expression was not elevated in the Sdc1-null keratinocytes (Fig. 6C). In wt keratinocytes but not in Sdc1-null keratinocytes, TGFβ1 significantly enhanced surface expression of several integrins including β4, α2, αv, β6 and β8, with increases ranging from 1.5 times higher (α2) to over 2 times higher (αv). Only α3 integrin expression at the keratinocyte surface remained at similar levels on both control and TGFβ1-treated wt and Sdc1-null keratinocytes. Adding 2.5 ng/ml TGFβ1 to day-3 cultures of wt and Sdc1-null keratinocytes yielded differences in integrin expression similar to those seen for 0.25 ng/ml (data not shown). These data show that adding TGFβ1, which increases migration in both wt and Sdc1-null keratinocytes, also increased integrin surface expression in wt keratinocytes. However, Sdc1-null keratinocytes showed elevated levels of several integrins before treatment with TGFβ1, and those levels were not altered in response to TGFβ1.

All of the integrins whose surface expression in wt keratinocytes was altered by addition of 0.25 ng/ml TGFβ1, namely α2, αv, β4, β6 and β8 integrins, were also surface-elevated in the Sdc1-null keratinocytes prior to TGFβ1 treatment. These data, together with the data showing different migratory responses to high concentrations of TGFβ1, suggest the possibility that the Sdc1-null keratinocytes have alterations in TGFβ1 signaling.

To assess the possibility of defective TGFβ1 signaling, we investigated the ability of increasing TGFβ1 concentrations (0.015 ng to 1 ng/ml) to inhibit DNA synthesis. Like wt keratinocytes, Sdc1-null keratinocytes ceased proliferation with a similar dose response when treated with increasing concentrations of TGFβ1 (Fig. 7A). We then looked in more detail at TGFβ1-induced gene expression by using a dual reporter assay, we assessed the ability of TGFβ1 to induce transcription of Smad4-dependent promoters. Data are presented in Fig. 7B for keratinocytes assayed 20 hours after TGFβ1 treatment and expressed as levels of TGFβ1-induced gene expression after controlling for differences in transfection efficiency. Similar results were obtained 6 hours after TGFβ1 treatment (data not shown). Both wt and Sdc1-null keratinocytes had detectable TGFβ1-mediated gene expression prior to the addition of TGFβ1; however, the constitutive level of TGFβ1-mediated gene expression in the Sdc1-null keratinocytes (6.7-fold) was significantly greater than that in wt keratinocytes (3.9-fold). The increase in levels of TGFβ1-induced gene expression measured after addition of 0.5 ng/ml TGFβ1 was 47-fold in Sdc1-null keratinocytes compared with 30-fold in wt keratinocytes. Increasing the concentration of TGFβ1 in the medium from 0.25 to 0.50 ng/ml increased TGFβ1-induced gene expression in Sdc1-null keratinocytes but had no affect on wt keratinocytes.

The data from the dual-reporter assay show that, (1) Sdc1-null keratinocytes have increased constitutive signaling through the TGFβ1 pathway and, (2) Sdc1-null keratinocytes respond to levels of TGFβ1 above those that elicit a transcriptional or migratory response in wt keratinocytes. We confirmed the data regarding increased constitutive signaling by assessing Smad2 phosphorylation in wt and Sdc1-null keratinocytes. Prior to the addition of exogenous TGFβ1, Sdc1-null keratinocytes showed elevated levels of phosphorylated Smad2 (P-Smad2) compared with those in wt keratinocytes (Fig. 7C). Addition of TGFβ1 induced a reproducible increase in P-Smad2 in wt keratinocytes within 15 minutes, which was sustained until at least 60 minutes later. By contrast, the Sdc1-null keratinocytes showed no change in P-Smad2 levels in response to the addition of TGFβ1.

Next, we performed semi-quantitative RT-PCR to assay the levels of mRNAs known to be upregulated by TGFβ1 in wt keratinocytes using untreated day-3 wt and Sdc1-null keratinocytes. Data are presented numerically in Table 1; supplementary material Fig. S2 shows the gels. After normalizing against the mRNA levels in wt keratinocytes, we saw increased levels of several TGFβ1-inducible mRNAs, including proteoglycans-like biglycan (1.8 times) and lumican (1.8 times), as well as matrix molecules including collagens α1-I (1.8 times) and α2-VI (1.9 times) despite the fact that we had not given the keratinocytes TGFβ1. Thrombospondin and Mmp9 mRNAs also showed a modest (1.5 times and 1.3 times, respectively) increase in Sdc1-null keratinocytes. These mRNA studies support the conclusion that the Sdc1-null keratinocytes were engaged in an elevated level of constitutive TGFβ1 signaling.

We next considered whether the overall production of TGFβ1 by Sdc1-null keratinocytes was greater than that of wt keratinocytes. Evaluating total TGFβ1 accumulation in conditioned media we observed, at the earliest time point detectable, that Sdc1-null keratinocytes secreted ∼1.3 times more total TGFβ1 per cell than wt keratinocytes. By day 15, the Sdc1-null keratinocytes had secreted about two times more total TGFβ1 than had the wt keratinocytes (Fig. 7D). When the amount of active TGFβ1 in the conditioned medium was also assessed by using a standard cell assay with luciferase-tagged transfected mink lung epithelial cells (TMLCs) (Fig. 7E), 1.6 times more active TGFβ1 per 10⁶ cells was seen in the medium of Sdc1-null keratinocytes than in medium of wt keratinocytes. In addition, we also assessed the amount of active TGFβ1 present in keratinocyte extracts obtained from the wt and Sdc1-null keratinocytes (Fig. 7E). At day 14, the extracts obtained from the Sdc1-null keratinocytes showed amounts of active TGFβ1 similar to those from wt keratinocytes.

In this study, we show that cultured Sdc1-null keratinocytes migrate more slowly than wt keratinocytes, seemingly because of factors related to their interaction with and the assembly of their extracellular matrix. We further show that Sdc1-null keratinocytes have elevated constitutive TGFβ1 signaling and also respond to concentrations of TGFβ1 above those that elicit responses in wt keratinocytes. By 3 days in culture, the Sdc1-null keratinocytes expressed higher levels of integrins on their surface than wt keratinocytes, despite the fact that both wt and Sdc1-null keratinocytes isolated directly from wt and Sdc1-null mouse skin showed similar levels of integrins on their surfaces. When we evaluated the migratory phenotypes of wt and Sdc1-null cells after treatment of cells with antibodies that block α6, β1 and αv integrins, we were able to show involvement of αv-family integrins in mediating α6β4 activity.

When we inhibited β1 integrin function the rate of migration was reduced for both the wt and Sdc1-null keratinocytes, but the difference between wt and Sdc1-null keratinocyte velocities after blocking β1-family integrin function was still significant: Sdc1-null keratinocytes still migrated slower than wt cells when they were forced to use αv-family integrins and α6β4 as their primary cell-to-substrate adhesions. This result implicates either αv-family integrins or α6β4 as causative in the delayed migration rates of Sdc1-null keratinocytes. Blocking α6 function restored migration rates of Sdc1-null keratinocytes to the same levels as in wt cells; thus, when Sdc1-null keratinocytes are forced to use αv-family and β1-family integrins, they no longer experience any delay in their migration rate. For α6β4 to reduce Sdc1-null keratinocyte migration, the activity of αv family integrins is needed. Blocking αv function restored keratinocyte migration rates in Sdc1-null cells but, unlike α6 integrin, the α-integrin antagonist reversed the phenotype so that the Sdc1-null keratinocytes were migrating faster than wt keratinocytes. Thus, when Sdc1-null keratinocytes depend on α6β4 and β1-family integrins for their migration, they migrate faster than similarly treated wt cells. Taken together, these data implicate αv-family integrins as positive regulators of the adhesive functions of α6β4. In wt keratinocytes, Sdc1 cooperates with αv integrins to decrease the adhesion and increase the migratory activity of α6β4.

Our data show that Sdc1 on keratinocytes modulates the function of α6β4 integrin by facilitating detachment from LN-332 matrices during migration and it does so by indirectly interacting with αv-family integrins. A model summarizing these results is presented in Fig. 8. Incorporated into the model is our knowledge that in normal and migrating epithelial tissues in skin and cornea, Sdc1 is not localized at the basal membrane of the basal keratinocyte surface where α6β4 is found but, rather, is present on basolateral and apical membranes and in endosomal compartments (Stepp et al., 2002). Although during wound healing the localization of α6β4 extends to the basolateral membrane surfaces, it remains enhanced at the basal surface where Sdc1 is absent. Thus, any effect that Sdc1 has in mediating keratinocyte adhesion and migration would have to be a regulatory one.

Although other studies have suggested that α6β4 integrin affects migration by altering LN-332 matrix assembly (Rabinovitz et al., 2001; Sehgal et al., 2006), our study is the first to implicate TGFβ1 and Sdc1 in this process. Conversion of α6β4 integrin from a state that mediates migration to one that mediates firm adhesion has been shown to be regulated by phosphorylation of the integrin β4 subunit by the EGF receptor (Mainiero et al., 1996; Mariotti et al., 2001). Sdc1 is known to associate with LN α chains (Hoffman et al., 1998), but recently a study by Ogawa and colleagues has shown that Sdc1 also associates with a fragment derived from the LN-332 γ2 chain and this complex inhibits phosphorylation of α6β4 integrin (Ogawa et al., 2007). Interaction between Sdc1 and the LN γ2 fragment causes integrin α6β4 to promote adhesion over migration. These data are consistent with previous data showing that proteolytic processing of LN-332 favors stable adhesion, whereas unprocessed forms of the molecule favor migration (Goldfinger et al., 1999). Ogawa and colleagues have also indicated that the interaction between Sdc1 and integrin α6β4 is indirect (Ogawa et al., 2007). Realizing that integrin α6β4 and Sdc1 generally exist in separate domains on the plasma membrane, we propose that Sdc1 interacts with αv-family integrins on basolateral and apical membranes such that both molecules are sequestered from α6β4 leaving it available to exist in its migratory state, shown by others to be phosphorylated by the EGF receptor. In Sdc1-null keratinocytes, αv-family integrins do not associate with Sdc1, are not sequestered apart from α6β4 and their presence at the basal surface of the cell may block phosphorylation of α6β4 integrin.

Studies have shown that deposition of LN-332 is polarized, and that α2β1 and α3β1 integrins regulate persistent migration in keratinocytes (Nguyen et al., 2000; Frank and Carter, 2004), a process that also involves formation of Rac1 gradients towards the forward- or leading-edge during directed cell migration (Pankov et al., 2005; Choma et al., 2004; Sehgal et al., 2006). Whereas Sdc1-null cells show differences in migration rate, we found no differences in their persistence indices on any of the substrates tested.

In a model for the study of coronary infarcts, Sdc1-null mice showed reduced deposition and increased disorder of collagens after wounding, which was reversed when Sdc1-null mice were infected with an adenovirus expressing Sdc1 before wounding. Disorganized collagen contributed to cardiac dilatation and reduced coronary function after infarcts in Sdc1-null mice (Vanhoutte et al., 2007). Here, we show that keratinocytes isolated from Sdc1-null mice also show reduced matrix assembly. These are the same cultures at the same time point that we also showed were synthesizing more of several different collagen mRNAs. Defective matrix assembly in Sdc1-null keratinocytes could be due to several factors; Vanhoutte and colleagues (Vanhoutte et al., 2007) have suggested in their heart model that elevated levels of MMP9 secreted by inflammatory cells contribute to collagen degradation before newly synthesized collagens had assembled into mature fibers. Sdc1 has been shown to bind to and sequester proteases after wounding in the skin (Bernfield et al., 1999) and, therefore, defective proteolytic balance could contribute to matrix destruction in vivo. Our data in vitro in the absence of inflammatory cells suggests that the matrix assembly defect is inherent in epithelial cells derived from the Sdc1-null mouse. If shed Sdc1 extracellular domain serves to protect nascent collagen molecules from destruction, its lack could lead to denaturation and unfolding, which could alter the ability of the matrix to support keratinocyte adhesion and migration.

The fact that the LN-332 tracks are better organized in the Sdc1-null keratinocytes may result from the enhanced adhesion-promoting activity of α6β4 integrin in these cells. Whereas the LN-332 deposited by Sdc1-null cells was better organized, the Sdc1-null cell matrix itself contained similar amounts of LN-332 and was able to support robust migration when wt keratinocytes were plated on it. TGFβ1 treatment of wt keratinocytes increased cell-surface expression of integrins but induced only modest increases in keratinocyte migration; similar treatment of Sdc1-null keratinocytes had a marked effect on migration rate, enhancing it significantly without altering the overall levels of integrins on the Sdc1-null keratinocyte surfaces. The mechanism whereby the Sdc1-null keratinocytes increase their migration rate above those of untreated wt cells after TGFβ1 treatment remains a subject of ongoing investigation. We have shown here that the Sdc1-null keratinocytes cease proliferating after TGFβ1 treatment with the same dose-response as wt control keratinocytes but we do not know whether or how long the wt and Sdc1-null cells remain viable after they cease proliferating.

Previous studies have shown conflicting results relating to the affect of TGFβ1 signaling on epithelial cell migration. In human keratinocytes, TGFβ1 increased migration rates (Decline et al., 2003). In vivo skin-wound-healing experiments using Smad3-null mice have shown accelerated wound healing, suggesting that endogenous TGFβ1 signaling impedes re-epithelialization after wounding (Ashcroft et al., 1999). Transgenic mice overexpressing Smad2 have shown delayed wound healing, again supporting the idea that elevated TGFβ1 signaling delays healing in vivo (Hosokawa et al., 2005). Other data have shown that neutralizing TGFβ1 in a wound-healing model that allows keratinocytes to migrate as sheets after injury, accelerates sheet movement (Neurohr et al., 2006). Experimental models for the study of epithelial cell migration in vitro via sheet movement are limited, but it appears that the differing mechanisms used by αvβ6 and αvβ8 integrins to activate TGFβ1 at the cell surface play important roles in relaying TGFβ1 signals from outside to inside keratinocytes during sheet movement (Sheppard, 2005).

By 3 days in culture, Sdc1-null keratinocytes have elevated levels of αvβ6, αvβ8 and α6β4 on their surfaces, the same integrins whose surface expression is enhanced when wt cells are treated with TGFβ1. How epithelial integrins accumulate at the surface of Sdc1-null keratinocytes is unclear. They could accumulate as a result of the elevated constitutive TGFβ1 signaling, which acts to enhance integrin surface expression on keratinocytes. Sdc-2 has recently been shown to mediate TGFβ-induced fibrosis in a kidney cell culture model via a mechanism that involved binding of Sdc2 with betaglycan, one of the TGFβ receptors present on kidney cell surfaces (Chen et al., 2004). The molecular mechanisms underlying syndecan-induced TGFβ affects are likely to vary in cell-type specific ways, especially in epithelial and mesenchymal cells. Several studies have implicated the Sdc1 cytoplasmic domain in mediating endocytosis of cytokine receptors (Fuki et al., 2000; Chen et al., 2004; Zimmermann et al., 2005), making it possible that the lack of Sdc1 alters integrin and growth-factor-receptor-mediated endocytosis (Caswell and Noman, 2006). Such differences could account for aspects of both the cell migration defects and enhanced responsiveness of the Sdc1-null keratinocytes to TGFβ1.

Our data on Sdc1-null keratinocytes grown in vitro are consistent with the hypothesis that the delayed corneal and skin wound healing we reported previously in vivo (Stepp et al., 2002) results from (1) the migrating Sdc1-null keratinocytes being more adherent to their underlying matrix due to increased adhesion promoting activity mediated by α6β4 integrin and, (2) altered responsiveness of the activated Sdc1-null keratinocytes to TGFβ1 in their environment, which increases surface expression of integrins and alters matrix synthesis and deposition. The Sdc1-null keratinocytes produce and secrete more active TGFβ1 in vitro. At the site of a wound, TGFβ1 can be released by keratinocytes, mesenchymal cells and by the inflammatory cells that are known to be present in elevated numbers after wounding in Sdc1-null mice (Stepp et al., 2002: Gotte and Echtermeyer, 2003; Neurohr et al., 2006). Altered responsiveness of keratinocytes to TGFβ1 could affect Sdc1-null keratinocyte migration rates by altering matrix remodeling and reassembly. Additional studies on the effects of the depletion of Sdc1 on signal transduction networks in vivo in skin and cornea will shed light on the mechanisms underlying the wound-healing defects induced by loss of this important proteoglycan and, in doing so, provide insight into the roles played by Sdc1 in forming and maintaining epithelial tissues in health and disease.

For immunoblots and immunoprecipitations, we used the following antibodies against: actin (Chemicon International, Temecula, CA; MAB1501R), αv integrin (Chemicon; AB1930), β5 integrin (Chemicon; AB1926), β6 integrin (Chemicon; MAB2076Z), β8 integrin (Santa Cruz Biotechnology, Santa Cruz, CA; sc-10817), α2 integrin (Chemicon; AB1936), α5 integrin [BD Pharmingen, Franklin Lakes, NJ; 5H10-27 (MFR5)], and LN-332 (Jonathan Jones, Northwestern University, Chicago, IL). The β1, β4, α3 and α9 integrin antibodies were rabbit polyclonals against cytoplasmic domain peptides (Sta Iglesia et al., 2000). For function blocking of integrins, 25 μg/ml each of the rat anti-mouse monoclonal antibodies 9EG7 for β1 integrin, GoH3 for α6 integrin and RMV-7 for αv integrins were used. These, along with an isotype-specific control IgGs were obtained from BD Pharmingen. For immunofluorescence microscopy localization of integrins, the same antibodies used for biochemical analyses listed above were used. For F-actin localization, we used Alexa-Fluor-488-labeled phalloidin (Molecular Probes/Invitrogen, Carlesbad, CA; A-12379), and for keratin-14, we used a rabbit polyclonal against mouse keratin-14 (Covance Research Products, Princeton, NJ; PRB-155P). For flow-cytometry analysis, we used the following antibodies: α6-FITC (BioLegend, San Diego, CA; 313605), β1-phycoerythrin (PE) (BioLegend; 102207), and β4-PE (Santa Cruz; sc-18883). For TGFβ1 neutralization studies, the antibody was obtained from R&D Systems (Minneapolis, MN; AB-101-NA) and was used at 1 μg/ml; a chicken polyclonal anti-vimentin antibody (Novus Biologicals, Littleton, CO 80120) was used at the same dilution as a control.

Wild-type (wt) mice (Balb/C) were obtained from NCI-Frederick (Frederick, MD). Tissue culture media, stocks, and buffers were obtained from Gibco/Invitrogen (Carlesbad, CA) unless otherwise indicated. Construction of Sdc1-deficient mice has been described previously (Stepp et al., 2002); mice have been backcrossed into a Balb/C genetic background (McDermott et al., 2007; Alexander et al., 2000). Primary mouse keratinocytes were isolated from skin of newborn Balb/C or Sdc1-null mice as described (Dlugosz et al., 1995), resuspended in freezing media (S-MEM with 8% fetal calf serum (FCS), 1.4 mM calcium, 10% dimethylsulfoxide, 10 mM Hepes, pH 7.3) and stored in liquid nitrogen until use. For each experiment, primary keratinocytes were grown in regular low-Ca²⁺ media (S-MEM and 8% FCS with calcium concentration of 0.05 mM) for the times indicated. Tissue culture plates were routinely coated with a mixture of human plasma fibronectin and collagen I (FNCNI; 10 μg/ml human plasma FN (BD Pharmingen, San Jose, CA); 1% Vitrogen (v:v) and 100 μg/ml bovine serum albumen (BSA) in S-MEM) for 15 minutes at 37°C prior to plating keratinocytes. For studies involving growth curves, data are presented for adherent keratinocytes only.

Wt or Sdc1-null keratinocytes were cultured for 4 days. Medium was removed and the keratinocytes were washed three times with PBS. Then, 250 μl M-Per protein extraction reagent (Pierce Chemical Company, Rockland, IL; 78503) with proteinase inhibitor (1:100 dilution) (Pierce Chemical Co.; inhibitor cocktail, 78415) was added to each of the 10-cm cell culture dishes, and the keratinocytes were harvested by scraping. A total of 10 μg protein from each extract was loaded to the 4-20% gel (Invitrogen, EC6025BOX) and SDS-PAGE electrophoresis was performed at 140 V. Proteins were transferred to PVDF membrane (Millipore, Billerica, MA; IPVH15150) at 300 mA for 1.5 hours, and the blot was then blocked in blocking solution [Tris-buffered saline (TBS) with 0.1% Tween 20 (TBST) and 10% milk] overnight at 4°C. Blots were subjected to enhanced chemiluminescence (ECL) reaction (Amersham/GE Healthcare Services, Piscataway NJ; RPN2132), and chemiluminescence was detected using X-ray film. When appropriate, data were quantified using NIH ImageJ software, v1.345 (available as a free download at http://rsb.info.nih.gov/ij/).

For-flow cytometry, keratinocytes were trypsinized and resuspended in serum-containing media, and concentrations were adjusted to normalize the cell counts for wt and Sdc1-null keratinocytes. Per antibody tested, 200,000 keratinocytes were spun down and resuspended in blocking buffer [PBS supplemented with 3% BSA containing 1 μl Fc-receptor (AbD Serotec, Raleigh, NC; BUF041A)]. Antibodies used were conjugated directly with phycoerythrin (PE) and controls included keratinocytes incubated with isotype-matched PE-conjugated antibodies, as well as keratinocytes incubated in blocking buffer alone. For quantitation, FloJo software (Windows Version 7.1.2, Tree Star, Inc., Ashland, OR) was used; median values for fluorescence intensity were obtained for each experimental and control antibody, and the ratios of experimental to control values obtained. Each determination was performed a minimum of three times on three different cell preparations, and data were tested for significance by Student's t-test.

For biotinylation of cell surfaces, keratinocytes were grown in standard FNCNI-coated tissue culture dishes with standard 0.05 mM Ca²⁺ medium and harvested as for the flow cytometry studies described above. Biotinylation was performed on keratinocytes in suspension using EZ-Link NHS-Biotin (Pierce Chemical Co.; 20217) as recommended for labeling adherent keratinocytes, with the exceptions that the biotin concentration was 0.5 mg/ml with PBS, and the reaction took place at 4°C for 2 hours. After labeling, keratinocytes were washed twice in PBS, their proteins extracted, and integrins were immunoprecipitated as described in Stepp et al., (Stepp et al., 1990). The efficiency of biotinylation was followed by running 50 and 100 ng extract of total biotinylated protein from wt and Sdc1-null cells, transferring the proteins to filters, and detecting biotin using horse radish peroxidase (HRP)-conjugated avidin (Bio-rad, Hercules, CA; 170-6528). For all of the data presented, the biotinylation efficiencies were confirmed to be similar between wt and Sdc1-null keratinocytes.

For immunofluorescence microscopy, keratinocytes were routinely grown on glass well chamber slides (Lab Tek II Chamber Slide System, 154526, Nalge Nunc International Corp., Naperville IL) that had been precoated with FNCN1 for 3 hours at 37°C as described above. After 3 days, the keratinocytes were fixed in ice-cold 50% methanol and 50% acetone for 2 minutes followed by 20 minutes in 100% methanol and, storage in PBS. Immunofluorescence was performed as described previously (Stepp et al., 2002).

For cell adhesion and spreading studies, s wt and Sdc1-null keratinocytes that had been cultured for 3-4 day were released by incubation in trypsin for 8 minutes. Trypsin activity was inhibited by addition of serum, and keratinocytes were resuspended in 0.2% serum-containing low-Ca²⁺ (0.05 mM) medium. Cell adhesion studies were performed using the CytoMatrix Screening Kit (Chemicon, ECM205), as recommended by the manufacturer. Cell spreading was determined at 15 and 45 minutes after plating keratinocytes onto either FNCNI or LN5/LN-332 (Aumailley et al., 2005) obtained from conditioned media of 804G cells (Hormia et al., 1995; Baker et al., 1996), a gift from Jonathan C. Jones (Northwestern University, Chicago, IL). Keratinocytes were fixed with 2% paraformaldehyde for 15 minutes, permeabilized with 0.1% Triton X-100 for 20 minutes, and stained with an antibody against actin. After visualization of the actin-stained cells using Alexa-Fluor-488-conjugated secondary antibody (Molecular Probes/Invitrogen), low-magnification images were obtained. Areas of no fewer than 100 keratinocytes were determined using Image Pro Plus software (Media Cybernetics, Silver Spring, MD), experiments were performed three times, and data expressed in arbitrary units.

Keratinocytes were seeded on 24-well plates and allowed to grow for 3 days before imaging on an Olympus IX81 research microscope (Olympus America, Melville, NY), equipped with a Proscan motorized stage (Prior Scientific Instruments Ltd., Rockland, MA) and placed in a temperature- and CO₂-controlled chamber (LiveCell Incubation System, Neue Biosciences, Camp Hill, PA). Using relief-contrast optics, 10× images were taken of each well every 10 minutes (for 16 hours, 40 minutes) until 100 images were captured. For each variable, triplicate wells were tracked and images managed using Slidebook 4.0 Digital Microscopy Solutions software (Intelligent Imaging Inovations, Denver, CO). Images were transferred to a workstation equipped with Metamorph image analysis software (Molecular Devices Corp., Chicago, IL), where velocities were calculated using the track cell module. Per well, 10-20 keratinocytes were tracked over the entire time period. A Visual Basic program was written by one of the co-authors (LJ) to allow us to choose the keratinocytes to track randomly from each field and to assist in data analysis. From each cell tracked, an average velocity was calculated. To verify that there was no change in velocity over time for each experiment, velocity over time was routinely assessed for each cell tracked. For experiments involving measurement of cell velocity on prepared wt and Sdc1-null cell matrices, wt and Sdc1-null keratinocytes were initially plated out onto both 100-mm dishes (for harvesting) and 24-well plates (for tracking). After 3 days, keratinocytes were harvested from 100-mm dishes by trypsinization and replated in triplicate onto 24-well plates coated with FNCNI or LN-332. For matrix swapping experiments, triplicate wells of wt or Sdc1-null keratinocytes grown in 24 well plates were lysed for 10 minutes at room temperature using 0.02 M ammonium hydroxide as described by deHart and colleagues (deHart et al., 2003), wells were washed three times with PBS to neutralize the solution and remove cellular debris, and matrices were seeded with the harvested keratinocytes. Whenever matrix swapping was performed, control wells containing wt and Sdc1-null keratinocytes grown for 3 days but not trypsinized were also tracked to provide internal controls for each experiment. After replating, keratinocytes were allowed to adhere for 1 hour. Non-attached keratinocytes were then removed and medium was added. Keratinocytes were then allowed to acclimate in the 37°C incubator for 2 hours prior to initiation of tracking. For studies using TGFβ1 neutralizing antibodies, keratinocytes were grown overnight in media supplemented with the antibody and tracked the next day. For studies using integrin-function-blocking antibodies, antibodies were added and keratinocytes tracked overnight; however, velocity data were derived from keratinocytes during the first 6 hours after integrin-neutralizing antibodies were added.

This assay was performed as described by Heng and colleagues (Heng et al., 2006). After cell tracking, cells within the wells of 24-well plates were fixed with Bouin's Fixative (75 ml of a saturated solution of picric acid (Sigma-Aldrich cat. number 319287), 20 ml formaldehyde, and 5 ml glacial acetic acid) for 1 hour, washed, and wells dried in a 42°C oven for 15 minutes; 0.5 ml Sirius Red staining solution was added to each well [0.25 gm Sirius Red (direct red 80, Sigma Aldrich cat. number 365548) in 250 ml picric acid] and cells incubated in the hood for 60 minutes while shaking gently. Wells were washed with water, washed again with acid water (0.01 N HCl), and dried in a 42°C oven for 15 minutes. Bound Sirius Red dye was eluted with 250 μl of 0.1N NaOH and collagen accumulation per well we determined by measuring the OD of the eluted dye at 550 nm. After elution of the Sirius Red dye, wells were washed once again, dried, and cell counts determined using a crystal violet dye binding assay to allow the collagen accumulated in each well to be normalized for differences in cell number.

For determination of total TGFβ1 in conditioned media, keratinocytes were seeded onto 100-mm culture dishes and cultured for the indicated times, after which they were washed twice with PBS and then incubated with serum-free S-MEM containing 0.05 mM Ca²⁺, for 24 hours. Cell debris was removed by centrifugation and conditioned medium, and stored at –70°C. After conditioned medium was harvested, the number of keratinocytes present per dish was counted. TGFβ1 levels in the conditioned medium were determined after acid activation by sandwich ELISA (R&D Systems; TGFβ1-Quantikine-DB100B).

For determination of TGFβ1 activity, transfected mink lung epithelial cells (TMLCs) carrying the of promoter plasminogen activator inhibitor 1 (PAI-1) driving expression of firefly luciferase were used as described (Mazzieri et al., 2000). TMLCs were plated into 24-well plates and, after reaching confluence, were treated with the conditioned medium harvested above. Relative luciferase expression in the TMLCs was determined using the Luciferase Assay System (Promega, Madison, WI; E1500).

For semi-quantitative RT-PCR, total RNA was purified from cultured wt or Sdc1-null keratinocytes using Trizol (Invitrogen, cat. number 15596-026). RNA pellets were re-suspended in DEPC-treated water and concentrations determined. Reverse transcription was performed using SuperScript First-Strand Synthesis System (Invitrogen, cat. number 11904-018) according to the manufacturer's instructions. The optimal number of cycles for a particular mRNA was determined so that the reaction product was in the linear range of the product-cycle curve to ensure that the amount of the PCR product reflected the amount of the cDNA and, thus, the mRNA. cDNAs were amplified using the following primer sequences: mouse biglycan Fw 5′-GACAACCGTATCCGCAAAGT-3′, Re 5′-GTGGTCCAGGTGAAGTTCGT-3′; CN4 α1 Fw 5′ ATTTCCCCGCCGTCTCTGCTGGTC-3′, Re 5′-GGTCCCCTCGTCCCTTTTGTGC-3′; CN4 α2 Fw 5′-GCCAGGCCGCAGTGTGAG-3′, Re 5′-CAGCCGGGCAGTGGGGTAT-3′; CN5 α1 Fw 5′-GGCCGCTCCCCTGTCTTC-3′, Re 5′-AGTCGGGGCTGTAGTGCTCA-3′; mouse CN6α2 Fw 5′-AGAAGCGCTGTGGTGCCCTGGATG-3′, Re 5′-CTGGCGCCGGCTTTCTTTGATGAG-3′; mouse decorin Fw 5′-TGAGCTTCAACAGCATCACC-3′, Re 5′-AAGTCATTTTGCCCAACTGC-3′; mouse EGF Receptor I Fw 5′-CGCGAGAACCACACTGCTGGTGT-3′, Re 5′-TAGTATCCATATTGCAGAGGATG-3′; fibromodulin Fw 5′-CTGCTTCTGCAGTTGCTCTG-3′, Re 5′-GGTGTGGGTGTCTCCTGACT-3′; mouse lumican: Fw 5′-AAGAGTGTGCCAATGGTTCC-3′, Re 5′-GGACTCGGTCAGGTTGTTGT-3′; mouse smad7 Fw 5′-TGGCATACTGGGAGGAGAAGAC-3′, Re 5′-CACGTCTAGTCGAAACACTTCC-3′; mouse thrombospondin Fw 5′-GAACCAGCTGAGCAAGAACC-3′, Re 5′-CAGGTGAGCAGGTGATCTGA-3′; mouse TGF-β1: Fw 5′-GAAAGCGGCAACCAAATC-3′, Re 5′-TGACATCAAAGGACAGCCAC-3′; mouse TGF β Receptor I Fw 5′-ACCTTCTGATCCATCGGTTG-3′, Re 5′-TTCCTGTTGGCTGAGTTGTG-3′; mouse TGFβ Receptor II Fw 5′-GCAAGTTTTGCGATGTGAGA-3′, Re 5′-GGCATCTTCCAGAGTGAAGC-3′.

For the Smad4 dual-reporter assay, wt and Sdc1-null keratinocytes were grown in 24-well plates and, after 3 days in culture, were transiently transfected simultaneously with one plasmid driving expression of firefly luciferase under a SMAD4 promoter (pSBE4-luc) and a second plasmid driving expression of a Renilla (sea pansy) luciferase under a transketolase promoter (pRL-TK-luc) to serve as an internal control for transfection efficiency. After 2 days, wt and Sdc1-null keratinocytes were treated in triplicate with different concentrations of TGFβ1 for 6 or 24 hours. The ratio of firefly luciferase to Renilla luciferase was determined using Stop and Glo Dual-Luciferase Reporter Assay System (Promega, E1910) and the fold increase in TGFβ1-mediated transcription was calculated as a function of the concentration of TGFβ1 added to the keratinocytes in culture.

The cessation of cell proliferation after TGFβ1 addition was determined as previously described (Vijayachandra et al., 2003). In brief, at both 4 and 7 days in culture, tritiated thymidine was added to keratinocytes after 24-hour treatment with varying concentrations of TGFβ1, ranging from 0-1.0 ng/ml. The incorporation of radioactivity into DNA was assessed. Data are expressed as the percentage of radioactive counts incorporated per well after addition of TGFβ1, relative to incorporation of radioactivity in keratinocytes that had no exogenous TGFβ1 added.

For experiments involving the use of added TGFβ1 for time-lapse or Smad phosphorylation studies, TGFβ1 was added at a concentration of 0.25 ng/ml unless otherwise indicated. The antibody used to detect Smad2 phosphorylation was directed against the Ser465-Ser467 residue of Smad2 (cat. number 3101, Cell Signaling Technology, Inc., Danvers, MA 01923).

We thank Caroline Alexander, Christophe Cataisson, Michael Gerdes, Adam Glick, Ulrike Lichti, Kwang Suh, Barbara Taylor, and Kinnimulki Vijayachandra for advice and assistance at various stages of this project. This work was initiated during a sabbatical from GWU Medical School Department of Anatomy and Cell Biology for M.A.S., and was supported in part by the Intramural Research Program of the NIH, National Cancer Institute, Center for Cancer Research. In addition, this work was funded by NIH RO1-EY08512-17 and RO1-EY13559-05 (M.A.S.).