The α2β1 integrin is a collagen-binding protein with very high affinity for collagen I. It also binds several other collagens and laminins and it is expressed by many cells, including keratinocytes and fibroblasts in the skin. In the past, α2β1 integrin was suggested to be responsible for cell attachment, spreading and migration on monomeric collagen I and contraction of three-dimensional collagen lattices. In view of these functions, normal development and fertility in integrin α2-deficient mice, which we generated by targeting the integrin α2 gene, came as a surprise. This suggested the existence of compensatory mechanisms that we investigate here using primary fibroblasts and keratinocytes isolated from wild-type and α2-deficient mice, antibodies blocking integrin function and downregulation of integrin α2 expression. The results show that the α2β1 integrin is absolutely required for keratinocyte adhesion to collagens whereas for fibroblasts other collagen-binding integrins partially back-up the lack of α2β1 in simple adhesion to collagen monomers. A prominent requirement for α2β1 integrins became apparent when fibroblasts executed mechanical tasks of high complexity in three-dimensional surroundings, such as contracting free-floating collagen gels and developing isometric forces in tethered lattices. The deficits observed for α2-deficient fibroblasts appeared to be linked to alterations in the distribution of force-bearing focal adhesions and deregulation of Rho-GTPase activation.
The collagen-rich matrix of the skin provides a mechanical scaffold for cell adhesion and a source of biological cues controlling cell behavior and signaling. Earlier studies with tissue-extracted collagens found in the skin have demonstrated that collagen I (Grinnell and Bennett, 1981; Mauch et al., 1986; Santoro, 1986), collagen IV (Aumailley and Timpl, 1986), collagen V (Ruggiero et al., 1994) and collagen VI (Aumailley et al., 1989; Pfaff et al., 1993) are excellent cell-adhesion substrates. Cellular interactions with these collagens under native conformation are mediated by integrins of the β1 family, including the rather ubiquitous α1β1 and α2β1 integrins (Wayner and Carter, 1987; Takada et al., 1988; Kramer and Marks, 1989) and the more recently discovered α10β1 and α11β1 with more restricted distribution (Camper et al., 1998; Velling et al., 1999). These integrins belong to a subclass of the β1 family characterized by an inserted domain (I domain) with homology to von Willebrand factor A domain and interacting with the ligand (Emsley et al., 2000; White et al., 2004). In vitro studies with purified integrins or their recombinant I domain and collagen monomers showed that α1β1 integrin has higher affinity for collagen IV than collagen I and, conversely, α2β1 and α11β1 integrins bind collagen I more efficiently than collagen IV (Kern et al., 1993; Tiger et al., 2001; Tulla et al., 2001). Collagens are, however, not the only extracellular matrix (ECM) ligands for α1β1 and α2β1 integrins and both bind laminin 1 (Elices and Hemler, 1989; Goodman et al., 1991; Pfaff et al., 1994; Ettner et al., 1998; Colognato et al., 1997). Additional interactions between α1β1 integrin and matrilin-1 (Makihira et al., 1999) and between α2β1 integrin and decorin (Guidetti et al., 2002), chondroadherin (Camper et al., 1997) and laminin 5 (Orian-Rousseau et al., 1998; Decline and Rousselle, 2001) have been reported.
Integrins are heterodimeric cell-surface receptors transducing outside-in and inside-out biological and mechanical cues between the ECM and the cell machinery, thereby regulating diverse cellular activities such as adhesion, migration, differentiation, apoptosis and expression of specific genes. Integrins execute these functions by establishing a physical link between the ECM and the actin cytoskeleton at specific membrane locations, the focal adhesions (Ingber, 1997; Grinnell, 2003; Danen and Sonnenberg, 2003; Humphries et al., 2004). The dynamics of both the actin cytoskeleton and focal adhesions is under the control of binary molecular switches of the Rho subfamily of small GTPases, in particular the trio RhoA, Rac1 and Cdc42 (Hall, 1992). By switching between soluble, inactive, GDP-bound and membrane-anchored, active conformations, RhoA, Rac1 and Cdc42 control actin polymerization and hence the cytoskeleton stiffness, allowing it to modulate and counterbalance forces applied to, or generated by cells during movement or matrix remodeling (Hall, 1992; Chrzanowska-Wodnicka and Burridge, 1996; Nobes and Hall, 1999; Wiesner et al., 2005).
Based on cell culture experiments, it has been suggested that optimal fibroblast proliferation requires α1β1 integrins (Pozzi et al., 1998) whereas α2β1 integrins are involved in several aspects of matrix remodeling, including formation of collagen fibrils (Klein et al., 1991; Li et al., 2003; Jokinen et al., 2004), contraction of collagen matrices (Schiro et al., 1991; Langholz et al., 1995; Jenkins et al., 1999) and induction of collagenase activity (Schiro et al., 1991; Langholz et al., 1995), which agrees with upregulation of the latter during scar contraction (Wu et al., 2003). In addition, both α1β1 and α2β1 integrins regulate collagen synthesis, but in opposite ways (Langholz et al., 1995; Riikonen et al., 1995). Less is known about α11β1 integrin function. It may also regulate collagen fibril organization as well as fibroblast migration (Tiger et al., 2001; Popova et al., 2004). Thus α1β1, α2β1 and α11β1 integrins may partially compensate each other, as underscored by the subtle phenotypes of mice with targeted deletion of either α1 or α2 subunits (Gardner et al., 1996; Holtkötter et al., 2002; Chen et al., 2002). Indeed, skin fibroblasts co-express α1β1, α2β1 and α11β1 integrins (Voigt et al., 1995; Wu and Santoro, 1996; Tiger et al., 2001) and the contribution of each integrin to interactions with different ECM ligands may overlap. By contrast, the repertoire of collagen-binding integrins expressed by keratinocytes is restricted to α2β1 (Watt, 2002). Using primary fibroblasts and keratinocytes isolated from wild-type and α2-deficient mice, integrin function-blocking antibodies and downregulation of integrin α2 expression by small-interfering RNAs (siRNAs), we have evaluated the contribution of α2β1 integrin when cells are performing tasks of different complexity, ranging from simple adhesion to a planar surface of ECM proteins to generation of forces in three-dimensional (3D) surroundings.
The α2β1 integrin is dispensable for fibroblast adhesion to collagen I
Dermal fibroblasts isolated from wild-type and α2-deficient mice and grown as cell monolayers on rigid tissue culture support did not reveal any obvious differences under phase-contrast microscopy (not shown). The two cell strains were compared for adhesion to different ECM proteins. In 30-minute-long assays, both type of fibroblasts adhere equally well to laminin 1, laminin 5 and collagen IV, whereas adhesion of α2-deficient fibroblasts to monomeric collagen I was about 35% lower than that of wild-type counterparts (Fig. 1A). Examination of cell spreading on different ECM substrates by phase contrast microscopy showed no obvious difference between α2-deficient and wild type fibroblasts (shown for collagen I, Fig. 1B,C). To get independent evidence that α2β1 integrins are dispensable for fibroblast adhesion to collagen I, we used RNA interference to downregulate integrin α2 subunit in wild-type mouse fibroblasts. Immunoblotting detection of integrin α2 subunit in cell lysates showed that two (siRNA2 and siRNA3) out of three tested siRNAs induced a gradual decrease of integrin α2 subunit expression over time to reach a minimum (70% of the control) 76 hours post-transfection (not shown). More pronounced downregulation of integrin α2 subunit was obtained by transfecting the fibroblasts concomitantly with siRNA2 and siRNA3 once (Fig. 2A, downregulation to 26% of control level) or twice successively (Fig. 2B, downregulation to 19% of control level). Under these conditions, adhesion of the siRNA-transfected fibroblasts to collagen I was decreased proportionally to the downregulation of the integrin subunit (Fig. 2, compare adhesion curves A and B), with a maximal decrease of 30% compared with controls following two successive rounds of transfection (Fig. 2B). Thus downregulation of integrin α2 subunit by at least 90% impaired adhesion of fibroblasts to collagen I monomers by only 30%, consistent with the adhesion results obtained with fibroblasts lacking the integrin α2 subunit.
As collagen-binding integrins co-expressed by fibroblasts may not come into play simultaneously, adhesion of fibroblasts from α2-knockout mice, characterized by complete absence of the α2 subunit, were investigated in more detail by varying the length of the assay and the concentration of coated collagen I (Fig. 3). In short-term assays (15 minutes), adhesion of α2-deficient fibroblasts was markedly reduced at each concentration of collagen compared with wild-type cells (Fig. 3, 15 minutes). In longer assays (30 and 60 minutes), adhesion of mutant cells to collagen I was decreased by about 30% compared with wild-type fibroblasts (Fig. 3). Together these results indicate that α2β1 integrins are important in the very early steps of fibroblast adhesion to collagen I and that later on other receptors are involved and partially compensate for the lack of α2β1 integrins.
The α1β1 integrin participates in fibroblast adhesion to collagen I, collagen IV and laminin 1
The contribution of α1β1 integrins to fibroblast adhesion to different ECM proteins was examined by using specific integrin function-blocking antibodies. Adhesion of wild-type and α2-deficient fibroblasts to collagen I, collagen IV, laminin 1 and laminin 5 was inhibited by antibodies against integrin β1 subunit (Fig. 4), indicating that only integrins of the β1 family mediate adhesion of primary fibroblasts to the ECM proteins tested. Antibodies against the integrin α2 subunit had no or marginal effects on wild-type or α2-deficient fibroblast adhesion to all four substrates (Fig. 4). By contrast, function-blocking antibodies against integrin α1 subunit completely inhibited adhesion of wild-type and α2-deficient fibroblasts to collagen IV and laminin 1 (Fig. 4), indicating that α1β1 integrin is the major receptor mediating fibroblast adhesion to collagen IV and laminin 1. Function-blocking antibodies against integrin α1 subunit did not prevent adhesion of wild-type or α2-deficient fibroblasts to laminin 5 (Fig. 4), which agrees with previous studies showing that cellular interactions with laminin 5 are mediated by α3β1 integrins (Carter et al., 1991; Rousselle and Aumailley, 1994). There was also no effect of the α1 integrin function-blocking antibody on wild-type fibroblast adhesion to collagen I (Fig. 4). However, the antibody induced a slight and reproducible inhibition of α2-deficient fibroblast adhesion to collagen I (Fig. 4), indicating the need of both α1β1 and α2β1 integrins for optimal fibroblast adhesion to collagen I.
The α2β1 integrin is required for keratinocyte adhesion to collagen I and IV
The integrin repertoire varies according to cell type. For example, keratinocytes are known to express α2β1, but not α1β1 integrins (Watt, 2002). We therefore isolated keratinocytes from the epidermis of wild-type and α2-deficient mice. Phase-contrast microscopy revealed that the overall morphology of wild-type and α2-deficient primary keratinocytes was similar (not shown). We then compared keratinocyte expression of collagen-binding integrins to that of fibroblasts by immunoblotting cell lysates with antibodies against the integrin α1, α2, and α11 subunits (Fig. 5). As expected, the α2 subunit was expressed by wild-type keratinocytes and fibroblasts and was absent in cells from α2-deficient mice (Fig. 5B). The integrin α1 and α11 subunits were expressed in fibroblasts but not in keratinocytes (Fig. 5A,C). Thus keratinocytes represent the model of choice to investigate whether the α2β1 integrin mediates adhesion to collagens when it is the only collagen-binding integrin present in a cell. Wild-type and α2-deficient keratinocytes adhered equally well to laminin 1 and laminin 5 (Fig. 5D). By contrast, whereas wild-type cells were able to adhere and spread on collagen I and collagen IV, there was almost no adhesion of α2-deficient keratinocytes to either collagen (Fig. 5D). Moreover, adhesion of wild-type mouse keratinocytes to collagen I and IV was completely inhibited by a function-blocking antibody against the α2 subunit (Fig. 5E). The antibody, however, did not inhibit keratinocyte adhesion to laminin 1 (Fig. 5E), which is a strong ligand for α6-subunit-containing integrins expressed by keratinocytes (Sonnenberg et al., 1990; Rousselle and Aumailley, 1994) and used here as a control.
The α2β1 integrin is important for transmission of forces in fibroblasts
In addition to mediating cell adhesion to the ECM, integrins are important for outside-in and inside-out transmission of mechanical forces, such as during cell migration and matrix remodeling (Ingber, 1997; Palecek et al., 1997; Eastwood et al., 1998; Grinnell, 2003). The mechanical consequences resulting from the absence of α2β1 integrins were first analysed by time-lapse videomicroscopy recording of cell movement on collagen I monomers. Wild-type and α2-deficient fibroblasts displayed roughly similar speed (0.32±0.13 μm/minute versus 0.29±0.10 μm/minute) and processive indexes (0.53±0.20 versus 0.42±0.22), indicating that the absence of α2β1 integrin does not grossly impair fibroblast locomotion on collagen I. We next examined the properties of the α2-deficient fibroblasts in assays more specific of fibroblast function in 3D surroundings, i.e. in collagen gels lacking mechanical load and in tethered collagen lattices. In a first set of experiments, wild-type and α2-deficient mouse fibroblasts were seeded in free-floating collagen gels to test their capacity to exert tractional forces as reflected by gel contraction. Contraction of the free-floating gels by α2-deficient fibroblasts was delayed compared with wild-type cells and it did not reach the plateau values observed for the controls (Fig. 6A,B). In a second set of experiments, the fibroblasts were cast into tethered collagen lattices and their ability to generate isometric forces against collagen fibrils was monitored over time. The characteristic bi-phasic curves representing force kinetics displayed by the cells within the tethered collagen lattices revealed that α2-deficient fibroblasts developed forces of significantly lower magnitude than the wild-type cells (Fig. 6C).
Alteration of focal adhesions and RhoA/Cdc42 activation in α2-deficient fibroblasts
Appropriate inside-out and outside-in force transmission by cells requires integrity of both focal adhesions and actin cytoskeleton and finely tuned modulation by Rho GTPases (Hall, 1992; Wiesner et al., 2005). To examine whether absence of the integrin α2 subunit had any effect on the actin cytoskeleton and focal adhesions, we stained fibrillar actin and vinculin, a marker of focal adhesions. Wild-type and α2-deficient fibroblasts adhering to collagen I displayed a well-developed actin cytoskeleton with numerous stress fibers running across the cell body (Fig. 7). Vinculin-rich focal adhesions were also present in both cell strains, however, with different distributions. In wild-type fibroblasts vinculin patches were observed mostly at the cell periphery at the end of actin stress fibers (Fig. 7). By contrast, vinculin-containing focal adhesions were mainly concentrated at the ventral surface of α2-deficient fibroblasts (Fig. 7).
Because assembly and turnover of focal adhesions are controlled by the Rho subfamily of GTPases, we analysed activation of these proteins upon fibroblast adhesion to collagen I in the absence of α2β1 integrins. Serum-starved fibroblasts were seeded on collagen I for various periods of time ranging from 60 to 120 minutes and GTP-bound RhoA, Cdc42 and Rac1 were analysed by pull-down assays. At 60 and 90 minutes, the levels of active, GTP-bound RhoA and Cdc42 were higher in α2-deficient fibroblasts than in wild-type fibroblasts (Fig. 8). By contrast, active, GTP-bound Rac1 levels were similar in wild-type and α2-deficient fibroblasts (Fig. 8). Thus in the absence of α2β1 integrins, there is a deregulation of Rho GTPase activation in fibroblasts adhering to collagen I associated with diminished generation of tractional and isometric forces.
Skin wounds close normally in α2-deficient mice
It has been suggested that α2β1 integrin could be important for wound contraction and closure because it mediates contraction of collagen gels by fibroblasts and migration of keratinocytes on collagen (Schiro et al., 1991; Pilcher et al., 1997; Nguyen et al., 2001). We therefore compared closure of skin wounds in wild-type and α2-deficent mice. Measurement of scar areas at successive time points did not reveal any differences between both strains of mice (Fig. 9). This confirms the results obtained with another strain of α2-deficient mice (Chen et al., 2002).
Ablation of α2β1 integrin in mice does not impair normal development and lifespan, and has so far revealed only a subtle phenotype, with partially defective branching morphogenesis and haemostasis (Holtkötter et al., 2002; Chen et al., 2002; Grüner et al., 2004). Similarly, deletion of integrin α1 permits normal murine development and connective tissue defects are apparent only after experimental injuries (Gardner et al., 1996; Gardner et al., 1999; Pozzi et al., 2000; Chen et al., 2004). This suggests that there is sufficient back-up between these integrins and other receptors in vivo. To address the existence of compensatory mechanisms, we subjected primary skin cells, fibroblasts and keratinocytes, from wild-type and α2-deficient mice to a panel of functional assays that had previously suggested important roles for α2β1 integrins. Cultivated keratinocytes and fibroblasts provide two excellent models of normal cells differing by their repertoire in collagen-binding integrins because keratinocytes express α2β1 integrin only and fibroblasts express α1β1, α2β1 and α11β1 integrins. Moreover, although we cannot exclude the fact that culture conditions modulate expression levels, the qualitative integrin profiles on cultivated cells are consistent with the integrin mRNA profiles in dermis and epidermis (B.E., unpublished observation).
The α2β1 integrin was initially identified as a cell-surface collagen-binding protein by ligand-affinity chromatography and by showing that antibodies against the integrin partially or completely inhibited adhesion of different cell types to native collagens (Wayner and Carter, 1987; Kramer and Marks, 1989; Elices and Hemler, 1989). The collagen-binding property of α2β1 integrin was confirmed by in vitro affinity studies with purified proteins (Kern et al., 1993; Xu et al., 2000; Tiger et al., 2001; Tulla et al., 2001). Together these observations led to the current belief that α2β1 integrin is required for cell adhesion and migration on collagens. Our results show that indeed this holds true for primary keratinocytes which express solely α2β1 integrins as collagen receptors, however, only to some extent for primary skin fibroblasts with a broader repertoire of collagen-binding integrins. Indeed keratinocytes from mutant mice lacking α2β1 integrins are incapable of adhering to collagen I and IV and adhesion of wild-type keratinocytes to these collagens is completely inhibited by function-blocking antibodies against the integrin α2 subunit. Thus, keratinocytes absolutely rely on α2β1 integrins to adhere to collagens and, at least for this task, they lack a back-up mechanism to substitute for α2β1 integrin deficiency. A different and more complex picture emerged for fibroblasts, where adhesion to collagen I was severely impaired in the initial steps and partially rescued at later time points. As blockade of β1 integrins by antibodies abolished fibroblast adhesion to collagen I, either α1β1 or α11β1 integrins or both are compensating for the absence of α2β1 integrins at late time points of adhesion. In contrast to complete inhibition of keratinocyte adhesion to collagens with a function-blocking antibody against the integrin α2 subunit, this antibody did not inhibit adhesion of wild-type fibroblast to collagen I, although knockout or knockdown of the integrin α2 subunit reduced fibroblast adhesion to collagen I by 30%. As shown recently for platelets and Chinese hamster ovary cells (Van de Walle et al., 2005), the conformation of α2β1 integrin may differ on fibroblasts and keratinocytes, explaining different reactivity towards the antibody. Alternatively, although function-blocking antibodies directly or sterically hinder integrin binding to the ligand, outside-in signaling may still proceed as previously suggested (Chen et al., 2002), allowing intracellular cross-talk between collagen-binding integrins expressed by fibroblasts. By contrast, when the α2β1 integrin is absent, as in α2-deficient fibroblasts, no antibody-triggered outside-in signaling can proceed or when α2β1 is the sole collagen-binding integrin, as in wild-type keratinocytes, antibody-triggered outside-in signaling cannot lead to cross-talk between collagen-binding integrins.
Partial inhibition of α2-deficient fibroblast adhesion to collagen I by antibodies against the integrin α1 subunit indicates that α1β1 compensates, at least in part, the absence of α2β1 integrins and we also anticipate similar functional compensation by α11β1 integrins. However, the reverse is not true, i.e. compensation by α2β1 integrin for the lack of α1β1, because antibodies against integrin α1 subunit completely inhibit both wild-type and α2-deficient fibroblast adhesion to collagen IV. These results strongly suggest that neither α2β1 nor α11β1 integrins substitute for α1β1-mediated adhesion of fibroblasts to collagen IV and laminin 1. This is in line with results obtained with embryonic fibroblasts from α1-deficient mice that displayed defective spreading on collagen IV (Gardner et al., 1996) and with studies establishing that α1β1 integrin binds better to collagen IV than to collagen I (Kern et al., 1993; Tiger et al., 2001; Tulla et al., 2001). It could be that the amounts of α2β1 and α11β1 integrins expressed by fibroblasts are not sufficient to substitute for α1β1 integrin binding to collagen IV. By contrast, although binding of α1β1 integrin to collagen I is lower than that of α2β1 integrin (Kern et al., 1993; Tiger et al., 2001; Tulla et al., 2001), it does participate in α2-deficient fibroblast adhesion to collagen I.
In addition to its contribution to fibroblast adhesion, the α2β1 integrin is required to generate and counter-balance mechanical forces in 3D networks of collagen fibrils, but not for fibroblast migration on 2D collagen I coats. The ability of α2-deficient fibroblasts to contract free-floating collagen gels was delayed and did not reach wild-type levels, indicating that α2β1 integrins are required for exerting tractional forces. It agrees with previous observations showing that antibodies against α2β1 integrin impede collagen gel contraction (Schiro et al., 1991; Klein et al., 1991; Langholz et al., 1995). Impairment of fibroblast mechanical properties is further underscored by the observation that α2-deficient fibroblasts generate isometric forces of lower magnitude than wild-type fibroblasts in tethered collagen lattices. We had previously observed that expression of integrin α2 subunit is upregulated in fibroblasts embedded in contracting collagen I lattices and reaches a plateau level 12-24 hours after seeding (Klein et al., 1991), suggesting that α2β1 integrin is important in the initial phases of matrix remodeling. These results led us to surmise that no compensatory activity is exerted by α1β1 or α11β1 integrins for tractional force generation. We cannot, however, exclude expression or activation of discoidin domain receptors 1 and 2 at later time points after fibroblast seeding, as these receptors are characterized by an unusually long lag time between collagen binding and activation (Vogel et al., 1997; Shrivastava et al., 1997).
It may appear surprising that movement of α2-deficient fibroblasts on collagen-I-coated surfaces was not altered in comparison to control cells because migration also requires transmission of forces. However, the mechanisms governing cell motility on 2D surfaces and remodeling of 3D networks are different in terms of activation of motor proteins (Meshel et al., 2005) and magnitude of forces (Eastwood et al., 1998). We anticipate that the decreased efficiency of α2-deficient fibroblasts to develop tension relates to the atypical distribution of the force-bearing focal adhesions and de-synchronization of Rho-GTPase activation as both are involved in coordinating mechanotransduction (Bershadsky et al., 2003). Based on pharmacological approaches, previous works suggested involvement of Rho GTPases in collagen gel contraction by fibroblasts (Abe et al., 2003; Parizi et al., 2000). Our recent studies show more directly that a tight regulation of RhoA activation controls collagen gel contraction and distribution of focal adhesions in fibroblasts (Zhang et al., 2006). Locking RhoA in its active conformation results in a reduced ability of a clonal population of fibroblasts to contract collagen gels and to an altered distribution of focal adhesions (Zhang et al., 2006), two features closely resembling those of fibroblasts lacking α2 expression and activating RhoA. This parallel strongly suggests that collagen gel contraction, activity of α2 integrin and regulation of RhoA activation are linked together and that a precise control of the turnover between active and inactive RhoA is crucial for force transmission. This is particularly relevant for dermal fibroblasts which are embedded in 3D surroundings rich in collagen fibrils in vivo.
This is not the case for keratinocytes, which are polarized and at the most contact collagens in a 2D manner.
In conclusion, it holds true that the α2β1 integrin has the ability to mediate cell adhesion to collagens but it fulfils this task in a cell-type-specific manner. For keratinocytes, it is the only receptor executing the task in vitro. Yet in vivo, the function of α2β1 in epidermal cells remains elusive. It was suggested to be important for keratinocyte adhesion and migration on the provisional matrix during skin wound healing (Pilcher et al., 1997; Nguyen et al., 2001). However, it is apparently not needed as wounds heal normally in mice deficient in α2β1 (Chen et al., 2002). Furthermore, using primary fibroblasts from α2-knockout animals allowed us to establish that they are equipped with sufficient back-up by other collagen-binding integrins that gradually compensate cell adhesion and force transmission. Compensation is an important biological phenomenon that could not be assessed in previous works using cells with a restricted repertoire in collagen-binding integrins and it helps to explain the subtle phenotype of α2-deficient mice. Whether compensation for the lack of α2β1 integrin in fibroblasts is due to redundancy or to an increase in the functional capacity of α1β1 and α11β1 collagen-binding integrins may be addressed in future work by generating fibroblast models with double- and triple-integrin deficiency.
Materials and Methods
Isolation and culture of mouse skin primary cells
Mice deficient for integrin α2 subunit (Holtkötter et al., 2002) and corresponding wild-type littermates were generated by heterozygous matings of animals that had been backcrossed into a C57Bl/6 background for more than six generations. Newborn animals were killed by decapitation. The entire trunk skin was removed and incubated overnight with 0.1% trypsin, 0.02% EDTA in PBS, followed by mechanical separation of epidermis from dermis. To isolate fibroblasts, the dermis was diced finely and incubated with 400 U/ml of collagenase I (Cell Systems, St Katharinen, Germany) for 1 hour at 37°C. Tissue debris was removed by centrifugation and collected fibroblasts were cultured in Dulbecco's modified Eagle's medium (DMEM, Invitrogen, Karlsruhe, Germany) containing 10% fetal calf serum (FCS, PAA Laboratories, Cölbe, Germany), 2 mM glutamine and antibiotics (Seromed-Biochrom, Berlin, Germany), and 50 μg/ml sodium ascorbate. The fibroblasts were used between passage three and five for experiments. Keratinocytes were isolated according to Rheinwald and Green (Rheinwald and Green, 1975) with some modifications. Finely minced epidermis was incubated in DMEM/Ham's F12 (3:1 v/v; Seromed-Biochrom) supplemented with 50 μM Ca2+, 1.8×10-4 M adenine, 0.5 μg/ml hydrocortisone, 5 μg/ml insulin, 1×10-10 M cholera enterotoxin, 10 ng/ml EGF, antibiotics (all from Sigma-Aldrich, Schnelldorf, Germany) and 10% FCS (PAA Laboratories) depleted of calcium by treatment with Chelex (BioRad, Munich, Germany). Cells were dissociated by agitation and seeded onto rat plasma fibronectin-coated (30 μg/ml; Sigma) plates together with growth-arrested 3T3 J2 feeder cells. Cultures were maintained in DMEM/Ham's with supplements at 32°C and used for experiments at passage two to four. Mouse neonatal primary keratinocytes (CellnTec, Bern, Switzerland) were subcultured in serum-free mouse keratinocyte growth medium (CellnTec).
Cell adhesion assays
Tissue culture wells (96-well plates, Costar) were coated with collagen I (Seromed-Biochrom), collagen IV (kindly provided by K. Kühn, Max-Planck Institute for Biochemistry, Martinsried, Germany), laminin-nidogen complex, referred to hereafter as laminin 1 (kindly provided by R. Timpl, Max-Planck Institute for Biochemistry, Martinsried, Germany) and laminin 5 (Rousselle and Aumailley, 1994). After saturation of the wells with 1% BSA (Fraction V, Serva, Heidelberg, Germany), equal number of cells were seeded in triplicate wells for 30 minutes when not otherwise indicated. For inhibition experiments with integrin function-blocking antibodies, suspended cells were mixed with appropriate dilutions of antibodies before seeding on coated wells. We used hamster monoclonal antibodies against rat and cross-reacting with mouse integrin β1 (Ha2/5), α1 (Ha31/8) and α2 (Ha1/29) subunits (PharMingen, Heidelberg, Germany). At the end of the experiments, adherent cells were quantified as previously reported (Aumailley et al., 1989) and photographed with a digital camera (PowerShot G5, Canon, Tokyo, Japan) mounted on a phase-contrast microscope (Axiovert 100, Carl Zeiss, Göttingen, Germany).
siRNA-mediated silencing of the α2 subunit
Double-stranded RNA oligonucleotides specifically targeting mouse integrin α2 subunit were obtained from Eurogentec (Sart-Tilman, Belgium) and the sense sequences were as follows: 5′-GGAGACAUCUCCAGUUCUUTT-3′ (siRNA1), 5′-GUCCAGACUUUCAGUUCUUTT-3′ (siRNA2), 5′-GCAUGGCAUUGGUGACUAUTT-3′ (siRNA3), and 5′-AUACUUACGCACGCUCCAATT-3′ (control siRNA). Semi-confluent wild-type mouse fibroblasts were transfected by calcium phosphate co-precipitation of one or two siRNAs (100 nM each) and further grown up to 76 hours. In some experiments the fibroblasts were transfected a second time after 36 hours. Expression of integrin α2 subunit was determined by SDS-PAGE fractionation of cell lysates and immunoblotting.
SDS-PAGE and immunoblotting
Confluent cell monolayers were lysed and homogenized in 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% Triton X-100, 1 mM each MgCl2, MnCl2, CaCl2, 1 μg/ml each leupeptin and pepstatin, 0.5 mM PMSF. Cell lysates were clarified by centrifugation and aliquots were fractionated by SDS-PAGE on 7% acrylamide gels under reducing conditions followed by immunoblotting with rabbit antibodies raised against the intracellular domain of integrin α1 (AB1934, Chemicon Europe, Hofheim, Germany), α2 and α11 subunits (peptide synthesis, conjugation to keyhole limpet hemocyanin and immunization were performed at Innovagen, Lund, Sweden) and secondary antibodies coupled to horseradish peroxidase (DakoCytomation). Reactive bands were visualized by ECL (Amersham Biosciences, Freiburg, Germany).
Cell migration assays
Cell migration was monitored by time-lapse video-microscopy. Briefly, equal numbers of cells were seeded in the centre of wells (10-μl drop/well, 24-well tissue culture plates), in DMEM supplemented with 10% fetal calf serum to allow rapid cell adhesion. After 1 hour, the cells were washed with PBS and the wells filled with fresh DMEM containing collagen I (40 μg/ml). Cell movement was recorded on a thermally controlled chamber (37°C, 5% CO2) placed on an inverted microscope (Axiovert S100TV, Carl Zeiss) equipped with a digital CCD camera (Xillix MicroImager, Richmond, British Columbia, Canada). Phase-contrast photographs were automatically captured every 5 minutes for 800 minutes and stored with Openlab software system (Improvision, Coventry, England). The sequences of images were converted to Quick Time movies and migration tracks of at least 20 cells were analysed using Dynamic Image Analysis System software (Solltech, Oakdale, IA). Extracted migration parameters included cell velocity (cell speed in μm/minute) and processive index defined by the ratio between the linear and the absolute distances covered by a cell during the time of recording.
Contraction of collagen gels
Mouse fibroblasts were seeded at a density of 1.5×105 cells/ml into 32 mm bacteriological plates (2 ml/dish) in DMEM supplemented with 10% fetal calf serum, sodium ascorbate (50 μg/ml), antibiotics and containing 0.3 mg/ml of acid-extracted collagen I from newborn calf skin (Institut für Biomedizinische Forschung, Leipzig, Germany) as previously described (Kessler et al., 2001). The cultures were placed at 37°C to allow collagen polymerization and gradual lattice contraction was monitored by measuring gel diameter of triplicate setups at successive time points up to 96 hours.
Force measurement in tethered collagen lattices
Mouse fibroblasts were suspended in supplemented DMEM (1.8×105 cells/ml) containing collagen I (1.75 mg/ml; rat tail collagen; First Link, Brierley Hill, UK) and placed in rectangular moulds (17×22×7 mm) in which lattice shrinkage is mechanically prevented by gel attachment to polyethylene bars located at the long ends of the moulds. The polyethylene bars were connected to a culture force transducer allowing isometric tension recording (KG 7A with Bridge-Amplifier DUBAM 7C, Scientific Instruments, Heidelberg, Germany). Force output was digitized at 0.23 Hz (Analog to Digital Converter μMeter4 and Software Nextview light from BMC Systeme, Maisach, Germany) and recorded on a personal computer. Quadruplicate lattices were used in each experiment.
Wound healing analysis
Wild-type and α2-deficient mice (females, 3-5 month-old) were anaesthetized by intraperitoneal injection of ketamine (10 g/l)/xylazine (8 g/l) solution (10 μl/g body weight). Two full-thickness wounds of 6 mm diameter excising the skin and panniculus carnosus were created using biopsy punches on both sides of the dorsal midline on the shaved back (n⩾8 mice per time point and genotype) and left uncovered. Wounds were digitally photographed at day 0, 2, 3, 5, 7 and 10 after injury. Wound areas were calculated using Photoshop (Adobe Systems, San Jose, CA) and expressed as percentage of initial (day 0) wound area.
Fibroblasts were seeded on glass coverslips coated with collagen I (40 μg/ml) for 60 minutes, fixed with 2% paraformaldehyde in PBS for 15 minutes, permeabilized with 0.2% Triton X-100 and processed for immunofluorescence staining. Mouse monoclonal antibody F-VII against vinculin (a gift from M. Glukhova, Institut Curie, Paris, France) was used as primary antibody followed by Cy3-conjugated secondary antibodies (Dianova, Hamburg, Germany) applied together with FITC-conjugated phalloidin (Sigma-Aldrich). The coverslips were mounted on histoslides in DAKO medium (DakoCytomation, Hamburg, Germany) and observed by laser-scanning confocal microscopy (Leica Instruments, Heidelberg, Germany). Confocal images were captured with single channel excitation, stored with the microscope internal software and merged using Photoshop.
GTPase pull-down assays
Fibroblasts were serum-starved for 24 hours, resuspended in serum-free medium and seeded on tissue culture plates coated with collagen I (40 μg/ml). After 60, 90 or 120 minutes, adherent cells were lysed in ice-cold buffer containing 1% Triton X-100, 25 mM HEPES pH 7.3, 150 mM NaCl, 4% glycerol, 4 μg/ml aprotinin, 1 μg/ml leupeptin and 1 μg/ml pepstatin. Lysates were cleared by centrifugation and supernatants (500 μl) were used to pull down Cdc42/Rac1 and RhoA with GST-PBD and GST-RBD fusion proteins, respectively, as previously described (Servotte et al., 2006; Zhang et al., 2006). Lysate (40 μl aliquots) and pull-down fractions were separated by SDS-PAGE on 15% acrylamide gels under reducing conditions and immunoblotted with mouse monoclonal primary antibodies against RhoA (clone 26C4; Santa Cruz Biotechnology, Heidelberg, Germany), Rac1 (clone 23A8; Biozol diagnostika, Eching, Germany) and Cdc42 (clone 44; Becton Dickinson, Heidelberg, Germany).
We are grateful to Ingo Haase and Semra Frimpong for assisting with keratinocyte culture, to Marina Glukhova, Klaus Kühn and Rupert Timp for gifts of reagents, to Monika Pesch and Kerstin Elias for excellent technical assistance and to Andreas Woeste for providing plenty of primary wild-type mouse fibroblasts to set up siRNA transfections. The work was supported by the Deutsche Forschungsgemeinschaft (SFB 589), the Deutsche Zentrum für Luft- und Raumfahrt (50WB0321), the Center for Molecular Medicine Cologne (TV80) and the Medical Faculty of the University of Cologne. M.A. is a researcher from the Centre National de la Recherche Scientifique.
- Accepted January 25, 2006.
- © The Company of Biologists Limited 2006