Fibroblast migration is mediated by CD44-dependent TGFβ activation

Injury results in inflammation, which is required to rid the host of the eliciting insult and to clear tissue and cell debris. However, just as critical, are the subsequent resolution of the ensuing inflammatory response and the initiation of tissue repair. This is achieved by a complex series of events that is spatio-temporally coordinated and leads to release of chemokines, pro- and anti-inflammatory mediators including cytokines, migration and activation of inflammatory and mesenchymal cells, and a return to homeostatic tissue function (Clark, 1996; Davidson, 1992). An imbalance in the pro- and anti-inflammatory mediators produced can lead to the development of chronic inflammatory diseases (Weissler, 1989; Selman et al., 2001). In addition to the regulation of the inflammatory response, a transition from an early stage of inflammation to a later stage of repair is required to re-establish normal tissue function following injury. Key elements of this transition include removal of inflammatory cells, and migration and differentiation of mesenchymal cells that contribute to the tissue remodeling process. This `staffing' of the injured tissue with cells suitable for the repair-phase facilitates matrix deposition, repopulation of the epithelium, and return of tissue function. Although the mechanisms that underlie this transition are still poorly understood, injury models have suggested that resolution of inflammation is necessary for the initiation of wound healing. Furthermore, if the fibrotic response associated with healing is not properly controlled, excess fibrosis can lead to the loss of tissue function typical of fibrotic diseases such as pulmonary fibrosis and liver cirrhosis.

Transforming growth factor β (TGFβ) plays an important role in the transition from inflammation to fibrosis (Wahl et al., 1988; Khalil and Greenberg, 1991; Clark and Coker, 1998). During the early phases of tissue injury, TGFβ initiates the inflammatory response by recruiting neutrophils and fibroblasts to the site of injury (Clark and Coker, 1998; Postlethwaite et al., 1987; Battegay et al., 1990). However, TGFβ suppresses acute inflammation and may thereby promote the transition to a reparative phase (Khalil and Greenberg, 1991; Clark and Coker, 1998). The anti-inflammatory effect of TGFβ is indeed demonstrated by severe inflammatory reactions in various organs of TGFβ knockout (KO) mice, including the lungs (Shull et al., 1992; Kulkarni et al., 1993). TGFβ also contributes to the migration of effector cells such as fibroblasts, as well as their growth, differentiation and activation, which are also important to the repair process. TGFβ thus plays a central role in the host response to tissue injury. Given the pleiotropic characteristics of TGFβ, tightly regulated mechanisms have evolved to coordinate its many potential biological effects (Barcellos-Hoff and Dix, 1996; Crawford et al., 1998).

CD44 is a cellular adhesion receptor that is upregulated following tissue injury and implicated in many chronic inflammatory diseases (Holgate, 1997; Mikecz et al., 1995; Jain et al., 1996; Puré and Cuff, 2001). As a principal receptor for hyaluronan (HA), CD44 participates in the activation of leukocytes and parenchymal cells in areas of inflammation, suggesting a role for CD44 in tissue remodeling and fibrosis (Aruffo et al., 1990; Foster et al., 1998). Interactions between CD44 and cytoskeletal components suggest a potential to influence the adhesion and motility of fibroblasts, thereby supporting a role for CD44 in tissue remodeling (Legg et al., 2002; Morrison et al., 2001; Tsukita et al., 1994). Previous studies also implicate CD44 as a co-receptor for α4β1-integrin-mediated cell adhesion (Verfaillie et al., 1994) and trafficking (Kawakami et al., 1999). Interestingly, CD44-deficient mice develop increased lung inflammation following bleomycin-induced lung disease, similarly to the phenotype observed in both the TGFβ and β₆ integrin KO mice (Shull et al., 1992; Munger et al., 1996; Teder et al., 2002). Even more intriguing is the notion that CD44 may regulate activation of TGFβ, thereby localizing its effects to areas of active injury (Teder et al., 2002; Yu and Stamenkovic, 2000). Recent studies also revealed that integrins participate in the activation of TGFβ (Munger et al., 1996; Morris et al., 2003; Mu et al., 2002).

In the studies described here, we investigated the role of CD44 in fibroblast migration in the context of tissue injury/wound healing. Specifically, we used in vitro wound assays to compare the migratory properties of primary lung fibroblasts isolated from CD44WT and CD44 deficient (CD44KO) mice. Our results indicate that CD44 is important in both maintaining the integrity of the actin cytoskeleton and facilitating an organized, directional migratory response to injury. Additionally, we showed that CD44 promotes MMP-dependent activation of TGFβ, and that exogenous TGFβ rescues the morphological phenotype and the velocity of migration of CD44KO fibroblasts. By contrast, exogenous active TGFβ was not able to restore directional migration to CD44KO fibroblasts, which is absolutely dependent on CD44 per se. Our data indicate that CD44 may be critical for the migration of fibroblasts to sites of injury and suggest a mechanism whereby the local effects of TGFβ may depend on the activity of CD44 to control inflammation and initiate the repair process.

To determine whether CD44 impacts cytoskeletal architecture, we compared the distribution of F-actin in primary fibroblasts harvested from the lungs of CD44WT and CD44KO mice. Immunofluorescence analysis of TRITC-phalloidin-labeled cells revealed that CD44WT fibroblasts were larger, more spread out, and had considerably more stress fibers traversing the cytoplasm than CD44KO fibroblasts (Fig. 1A). Furthermore, stress fibers within CD44WT fibroblasts were organized and uniform throughout each cell, whereas CD44KO fibroblasts were characterized by the accumulation of cortical actin. We quantified stress fiber density by incorporating a line profile across the cytoplasm that identified stress fibers by their increased fluorescence relative to areas devoid of stress fibers (Fig. 1B). Sharp, distinct peaks in fluorescence intensity within each line profile represented individual stress fibers crossed by the lines shown in Fig. 1A. Quantification of these peaks in three different regions per cell, in three cells per genotype, showed a three- to fourfold increase in stress fibers in CD44WT fibroblasts compared with levels in CD44KO fibroblasts. These results suggest that CD44 plays an important role in stress fiber formation and in the organization of primary lung fibroblasts.

The reduction we observed in stress fibers raised the possibility that integrin-dependent processes such as cell migration might be altered by the loss of CD44 (Nobes and Hall, 1995; Parsons et al., 2000; Ridley and Hall, 1992; Huttenlocher et al., 1996). To test whether integrin-dependent events were affected in CD44KO fibroblasts, we plated CD44WT and CD44KO cells onto glass coverslips and stained for stress fibers and vinculin-containing focal adhesions – downstream consequences of integrin-dependent adhesion. Although we observed minimal differences in vinculin staining during the initial adhesion and cell spreading (data not shown), after 3 days in culture, we noted distinct patterns of fluorescence between CD44KO and CD44WT fibroblasts stained for vinculin. The reduction in vinculin staining indicated that the maturation or stability of focal adhesion complexes was impaired in CD44KO fibroblasts (Fig. 1C).

To determine whether CD44-dependent cytoskeletal changes translate into altered migration, we compared the motility of CD44WT and CD44KO fibroblasts using a scratch-wound assay. We found that CD44KO fibroblasts were unable to close the wound gap in a time frame comparable with that of CD44WT cells (Fig. 2A, top panel). Tracking of individual cells using live-cell imaging revealed that the migration of CD44KO fibroblasts was disorganized and directionless, whereas CD44WT fibroblasts moved towards the leading edge of the wound, resulting in more efficient wound closure (Fig. 2A, bottom panel). Analysis of the positional coordinates confirmed that the migration of CD44KO fibroblasts was less directional than that of CD44WT fibroblasts (Fig. 2B). Interestingly, despite the inability to close the wound, CD44KO fibroblasts exhibited a significant increase in motility compared with CD44WT fibroblasts (Fig. 2C). Wound assays done in the presence of mitomycin led to similar results, suggesting that the observed CD44KO phenotype is not dependent on the proliferative capacity of the cells (data not shown).

We next examined whether reconstitution with full-length CD44 could reverse the phenotype of CD44KO cells. Transfections were confirmed at the single-cell level by staining with anti-CD44 monoclonal antibody (mAb). Reconstituted fibroblasts exhibited cytoskeletal features resembling that of CD44WT cells, indicating that the defect in stress fibers in the cells from knockout mice was due to the absence of CD44 per se (Fig. 3A). Reconstitution of CD44KO fibroblasts with full-length CD44 also rescued the defect in migration of CD44KO cells when compared with neighboring untransfected cells that did not express recombinant CD44. The loss of directional migration and increase in velocity of locomotion of CD44KO cells, and the capacity of forced expression of CD44 to overcome each of these defects in CD44KO cells, were quantified by comparing the average velocity, the instantaneous velocity and the confinement ratio of tracked CD44WT cells, CD44KO cells and CD44KO cells transfected with full-length CD44 (CD44KO.CD44), respectively. The average velocity of CD44KO.CD44 cells was reduced compared with that of CD44KO cells (Fig. 3B), back to levels typical of CD44WT cells (Fig. 2C). As motile cells do not necessarily migrate continuously or at a constant velocity, the increase in the average velocity of CD44KO cells over the entire tracking period could either be a manifestation of an increase in the velocity at which they move over the entire period, or a shift in the number of cells, or the percentage of time that individual cells move with a greater instantaneous velocity. In other words, the distinct velocity of migration of CD44KO fibroblasts could reflect changes in the evenness with which they migrate. Therefore, we also compared the frequency distribution of the instantaneous velocity of the three groups, which should show a different shape in CD44KO and CD44WT fibroblasts if this parameter is involved in the observed phenotype (Fig. 3C). Indeed, this analysis revealed a significant shift in the instantaneous velocity of CD44KO fibroblasts compared with that of CD44WT cells, and this shift was overcome in CD44KO fibroblasts transfected with CD44.

The confinement ratio is a measure of the final displacement distance relative to the total distance traveled between the two end points such that a confinement ratio closer to 1 suggests more directional migration. We found that the confinement ratio was markedly reduced to below 1 in CD44KO fibroblasts whereas the confinement ratio approached 1 in both CD44WT cells and CD44KO cells transfected with CD44 (Fig. 3D).

These analyses establish that CD44KO cells have increased motility but are less directional, whereas CD44WT cells are less motile but more directional. These data together suggest that CD44WT fibroblasts are more apt to respond to injury in an organized fashion compared with CD44KO fibroblasts.

A recent study demonstrated a pivotal role for CD44 in the resolution of lung inflammation and in promoting the transition to a reparative fibrotic response in a murine model of bleomycin-induced lung fibrosis (Teder et al., 2002). Although multiple mechanisms were involved, the study revealed that CD44KO mice had normal to enhanced levels of latent TGFβ but reduced levels of active TGFβ in broncho-alveolar lavage fluid, suggesting that CD44 may be critical for regulating TGFβ activation.

To investigate the potential effects of wounding and the role of CD44 in TGFβ activation, we compared levels of total and active TGFβ in conditioned media from CD44WT and CD44KO fibroblasts. As determined using a PAI-I promoter-reporter assay, over twofold less active TGFβ was generated by CD44KO fibroblasts when compared with CD44WT cells following wounding. Furthermore, wounding induced the generation of relatively high levels of active TGFβ only by CD44WT cells (Fig. 4A; top panel). This difference was not accounted for by differences in total TGFβ levels because the total levels of TGFβ detected in heat-activated conditioned media generated by CD44KO fibroblasts were comparable with the levels of total TGFβ generated by CD44WT fibroblasts (Fig. 4A; bottom panel). To confirm that the PAI-1 promoter in our reporter cell bioassay was predominantly responding to the TGFβ-signaling pathway and giving a measure of TGFβ levels rather than other signaling pathways, we also performed the bioassay in the presence of a neutralizing anti-TGFβ antibody. As the anti-TGFβ reduced the luciferase signal back to basal levels in all cases, we conclude that the contribution of other signaling pathways in this reporter cell bioassay is minimal, and similar in CD44WT and KO cells (Fig. 4A,B). After taking the non-TGFβ-specific contributions into account, we still observed a greater than 50% decrease in active TGFβ levels in wounded CD44KO fibroblasts when compared with wounded CD44WT fibroblasts.

Finally, to determine whether the defect in TGFβ activation translated into reduced TGFβ signaling in CD44KO fibroblasts, nuclear phospho-Smad2 (Smad2-P) was measured in unwounded and wounded CD44WT and CD44KO cells (Fig. 4B). Under basal (unwounded) conditions, levels of nuclear Smad2-P were low and there was no discernible difference between CD44WT and CDKO fibroblasts. Upon wounding, we observed an approximately threefold increase in nuclear Smad2-P levels in CD44WT cells. Although wounding also induced translocation of Smad2-P to the nucleus in CD44KO fibroblasts, the levels were reduced by ∼20% compared with CD44WT cells based on quantification of densitometric scans of Smad2-P immunoblots when normalized for loading to the signal from anti-retinoblastoma (Rb) immunoblots of each sample. Although the defect in activation of Smad2 was far from complete, the decrease was consistent over several experiments and, interestingly, a similar trend was observed when CD44WT and CD44KO fibroblasts were either treated with latent TGFβ1 alone or in combination with wounding (data not shown). By contrast, the activation of Smad2 was comparable in CD44WT and CD44KO fibroblasts treated with acid-activated TGFβ1. These data indicate that the reduction in nuclear Smad2-P induced by endogenous TGFβ in response to wounding was probably due to the defect in activation of TGFβ observed in CD44KO cells rather than an inherent defect in TGFβ-receptor-mediated signaling in CD44KO cells, or the less likely possibility of a TGFβ-independent defect in Smad2 signaling in CD44KO fibroblasts. Taken together, these results indicate that CD44KO fibroblasts exhibit a defect in TGFβ activation that translates into a decrease in TGFβ-receptor mediated signaling.

To identify a possible mechanism for CD44-dependent TGFβ activation, we next focused on matrix metalloproteinases. Previously, MMP2, MMP9 and MMP14 have been implicated in the activation of TGFβ, and CD44 has been shown to act as a docking molecule for MMP9, localizing MMP9 activity to the cell surface (Yu and Stamenkovic, 2000).

To determine whether MMPs are involved in the CD44-dependent activation of TGFβ, particularly in response to injury, we compared levels of total and active TGFβ in conditioned media from unwounded and wounded CD44WT and CD44KO fibroblasts generated in the presence or absence of a general MMP inhibitor, GM6001 (Fig. 5). Treatment with GM6001 had no effect on the levels of active TGFβ in unwounded CD44WT and CD44KO cells compared with their control-treated counterparts. However, the robust activation of TGFβ induced in CD44WT cells in response to wounding was reduced to near background (uninjured) levels in the presence of the MMP inhibitor. Also, the MMP inhibitor had no effect on the activation of TGFβ in CD44KO fibroblasts, which was already severely attenuated compared with the response of untreated CD44WT cells following wounding, and more similar to the response of CD44WT cells treated with the MMP inhibitor. Again, we detected no significant differences in levels of total TGFβ between CD44KO and CD44WT cells under these conditions (data not shown). These data provide strong evidence for a CD44-dependent, MMP-mediated mechanism of activation of TGFβ, which is induced in fibroblasts in response to injury.

Based on the above data implicating CD44 in the activation of TGFβ and the considerable evidence that TGFβ has a central role in regulating fibroblast function and tissue injury, we investigated whether the addition of exogenous active TGFβ could reverse the phenotype of CD44KO fibroblasts (Clark, 1996; Clark and Coker, 1998; Yu and Stamenkovic, 2000). The addition of exogenous active TGFβ induced an increase in stress fiber formation in CD44KO fibroblasts as determined by staining for F-actin (Fig. 6A). In marked contrast, treatment with exogenous latent TGFβ, had little or no effect on the CD44KO phenotype. Quantification of stress fibers in three different regions per cell, in three cells per genotype, revealed that the number of stress fibers in CD44KO fibroblasts treated with exogenous active TGFβ increased approximately three- to fourfold, to resemble levels in CD44WT cells; whereas stress fibers in CD44KO fibroblasts treated with exogenous latent TGFβ remained unaffected (Fig. 6B). These results confirm our conclusion that CD44KO fibroblasts are severely defective in their capacity to mediate the conversion of endogenous latent TGFβ to active TGFβ and link this defect to the defect in stress fibers.

Addition of active TGFβ also reversed the defect in the velocity of migration of CD44KO fibroblasts, such that their motility resembled that of CD44WT fibroblasts (Fig. 6C,). Quantitative analysis revealed that CD44KO fibroblasts treated with active TGFβ showed a significant shift in their instantaneous velocity, approaching that of the CD44WT phenotype (Fig. 6C). However, exogenous active TGFβ was not sufficient to overcome the defect in the directionality of migration, because the confinement ratio of CD44KO cells was essentially the same in the presence and absence of active TGFβ cells, and in both cases, was significantly reduced compared with the ratio in CD44WT cells (Fig. 6D). These data suggest that CD44-dependent cytoskeletal organization is mediated through the activation of TGFβ, but that establishing and or maintaining the polarity of fibroblasts to create a stable leading edge requires CD44 per se.

Finally, to determine whether endogenous TGFβ mediates stress fiber formation and reorganization in CD44WT fibroblasts, we investigated the effect of anti-TGFβ antibody on CD44WT and CD44KO fibroblasts (Fig. 7A). Blocking TGFβ markedly reduced stress fibers in 75-80% of CD44WT cells only. Moreover, CD44WT fibroblasts resembled the knockout phenotype in that they exhibited an accumulation of cortical actin. Quantification of fluorescent peaks in three different regions per cell, in three cells per genotype, showed an almost 80% decrease in the number of stress fibers in CD44WT fibroblasts treated with anti-TGFβ antibody compared with untreated cells (Fig. 7B). By contrast, CD44KO fibroblasts remained unaffected by the treatment with anti-TGFβ antibody, In summary, endogenous TGFβ mediates CD44-dependent stress fiber formation and cytoskeletal reorganization in CD44WT fibroblasts, and the defect in stress fibers in CD44KO fibroblasts is due to the defective generation of active TGFβ by these cells. These results reinforce the importance of TGFβ and CD44 in cytoskeletal organization.

The mechanisms that control the transition between inflammation and tissue repair are complex and not well understood. Our study provides mechanistic insights into this process. First, the adhesion receptor CD44 plays a role in cytoskeletal organization by impacting stress fiber and focal adhesion complex formation or stability in primary lung fibroblasts. Second, CD44 regulates the migration of fibroblasts. Third, CD44 regulates cytoskeletal remodeling and migration, at least in part, by promoting MMP-mediated activation of TGFβ. Based on these data and evidence from the literature, it appears that CD44 impacts cytoskeletal organization and fibroblast migration, at least in part through a TGFβ-dependent process, and suggests that CD44 has a major role in the resolution of inflammation and transition to reparative fibrosis in response to tissue injury (Nobes and Hall, 1995; Ridley and Hall, 1992; Bhowmick et al., 2003).

The migratory capacity of fibroblasts is dependent upon cytoskeletal reorganization that occurs partially through interaction of cell-adhesion receptors and the surrounding matrix (Huttenlocher et al., 1996; Burridge et al., 1988). CD44 is widely expressed on most cell types (e.g. leukocytes, endothelial cells and parenchymal cells), and is upregulated following tissue injury and inflammation (Puré and Cuff, 2001; Foster et al., 1998). In addition to binding the matrix glycosaminoglycan hyaluronan, specific isoforms of CD44 have been shown to exhibit affinity for matrix proteins such as fibronectin, collagen and osteopontin, all of which contain RGD motifs, which, upon integrin engagement, lead to formation of stress fibers (Bartolazzi et al., 1996; Fitzpatrick et al., 1994; Giachelli et al., 1993; Clark and Brugge, 1995). We provide the first evidence for a link between CD44 and cytoskeletal reorganization and its impact on fibroblast migration by showing that primary CD44WT fibroblasts have more stress fibers than CD44KO fibroblasts. We also found that the loss of CD44-dependent organization of the cytoskeleton was associated with a decrease in the instantaneous velocity with which fibroblasts migrate in response to wounding in vitro. Although others have investigated the impact of CD44 on cell migration, those studies focused on transformed cells plated on hyaluronan-coated surfaces (Peck and Isacke, 1996; Peck and Isacke, 1998). By comparing CD44WT, CD44KO and CD44 reconstituted fibroblasts, we have established that CD44 imparts a slower but more organized migratory phenotype (decreased velocity/increased directionality) compared with CD44-deficient primary lung fibroblasts. This is also consistent with our finding that CD44WT fibroblasts form increased focal adhesions, supporting the notion that focal adhesion formation slows cell motility.

Fibroblasts localize to areas of injury by responding to cytokines and chemokines, and by adhering to local matrix components in an organized fashion. In models of tissue injury and remodeling, it is suggested that disorganized repair leads to fibrosis (Selman et al., 2001). CD44 localizes to the leading edge in fibroblasts following activation of key GTPases, indicating that it may play a role in determining cell polarity through stabilization of F-actin (Shimonaka et al., 2003; Seveau et al., 2001; del Pozo et al., 1999). Previous evidence linking CD44 to the actin cytoskeleton indicates that CD44 may interact with ezrin, radixin and moesin (Legg et al., 2002; Morrison et al., 2001; Tsukita et al., 1994), and CD44 may affect polarization and directed migration of neutrophils (Alstergren et al., 2004). Interestingly, our data shows that despite the lower instantaneous velocity, primary CD44WT fibroblasts closed wounds efficiently, whereas in a similar time frame, CD44-deficient fibroblasts did not. The data presented here indicate that the defect in wound closure observed in CD44KO cells is due to loss of directionality. In a study by Svee and colleagues, treatment with anti-CD44 monoclonal antibodies inhibited the migration of fibroblasts in an acute lung injury model, suggesting that CD44 is capable of mediating fibroblast migration into a provisional matrix following lung injury (Svee et al., 1996). Taken together, these data suggest that CD44 is critical to establishing or maintaining cell polarity so that cells migrate towards the wound rather than randomly, and that directionality is more important than the reduction in velocity with respect to wound closure.

In this context, our findings suggest an intriguing possibility that CD44 not only facilitates migration of neutrophils (Alstergren et al., 2004) to areas of acute inflammation, but also promotes migration of mesenchymal cells to areas of tissue injury, thus placing CD44 at the center of tissue remodeling. Our data emphasize that the two effects of CD44 on migration (velocity and directionality) are mechanistically distinct, because only the velocity phenotype is rescued by TGFβ. Our ability to rescue the migrational velocity without a comparable rescue of the confinement ratio (directionality) following treatment with active TGFβ indicates that the requirement for CD44 can be bypassed by providing active TGFβ with regard to cytoskeletal organization, but that CD44 itself is required to establish or maintain the cell polarity necessary for directional movement. Indeed, directional migration in the context of wounding may be dependent on the development of chemotactic gradients, most notably hyaluronic acid (Alstergren et al., 2004; Asselman et al., 2005). Whether CD44-dependent chemotactic mechanisms play a role in our system remains to be determined.

Previous studies demonstrated that TGFβ signaling may lead to stress fiber formation and expression of α-smooth muscle actin (α-SMA) (Bhowmick et al., 2003; Bhowmick et al., 2001; Serini and Gabbiani, 1999). Interestingly, in addition to the stress fiber phenotype, our analysis revealed decreased basal levels of α-SMA in CD44KO fibroblasts compared with CD44WT fibroblasts (data not shown), suggesting that the presence of CD44 may affect fibroblast activation. The ability to rescue the wild-type phenotype with exogenous active TGFβ1 suggests that there is no inherent defect in the capacity of CD44KO cells to express or assemble the proteomic machinery necessary for the formation of stress fibers, or in their ability to respond to active TGFβ. Thus, our data indicate that the defect in TGFβ-mediated processes in CD44KO fibroblasts is due to the loss of CD44-dependent TGFβ activation, thus providing a mechanism for spatio-temporal control of TGFβ activity following tissue injury through the regulation of CD44 expression and functional activation. This hypothesis is supported by a recent study demonstrating that CD44KO mice had reduced levels of active TGFβ in broncho-alveolar lavage fluid, compared with levels in CD44WT littermate controls, following bleomycin-induced lung injury (Teder et al., 2002). At sites of tissue injury, TGFβ activation can help localize repair efforts (Barcellos-Hoff and Dix, 1996; Crawford et al., 1998; Sato et al., 1990). Indeed, the upregulation of CD44 at sites of injury may facilitate the localized effects of TGFβ, possibly limiting the extent of inflammation and coordinating a transition to remodeling. Taken together with our data, CD44 may provide a critical function in the resolution of lung inflammation and the transition to a reparative fibrotic response following tissue injury, where CD44 is critical for regulating TGFβ activity. In this regard, CD44 offers yet another mechanism by which the activity of TGFβ is tightly regulated to coordinate its many biological effects.

The mechanisms underlying CD44-dependent activation of TGFβ are currently being investigated further, but the data presented here indicate that CD44 promotes activation of TGFβ through an MMP-dependent mechanism. That we detected some activation of TGFβ by CD44KO cells and CD44WT cells treated with GM6001, is consistent with the notion that alternative, CD44-independent and MMP-independent mechanisms of TGFβ activation are also involved. A role for CD44 in MMP-mediated activation of CD44 is also supported by data from Yu and Stamenkovic suggesting that CD44 can act as a docking molecule for MMP9 at the cell surface, thereby locally activating latent TGFβ (Yu and Stamenkovic, 2000). Cell-surface adhesion molecules, such as integrins, have also been shown to play a role in the activation of latent TGFβ (Munger et al., 1996; Annes et al., 2004). Although direct evidence establishing a role for CD44-dependent integrin-mediated adhesion is lacking, studies have implicated CD44 as a co-receptor for integrins in adhesion and cell trafficking (Verfaillie et al., 1994; Kawakami et al., 1999). Our data reveal that CD44 can play a role in integrin-mediated processes – such as the formation of stress fibers and focal adhesions, and migration – and suggest that CD44 and integrins both contribute to activation of TGFβ. Accordingly, we hypothesize that CD44 actively participates in modifying mesenchymal function during the injury response to favor an organized migratory response and wound closure. Whether this putative function of CD44 contributes to the remodeling process following tissue injury is currently under investigation. It is worth noting however, that a link between integrins and tissue injury has already been characterized. β₆-integrin-KO mice develop increased lung inflammation following bleomycin-induced lung injury, which is similar to the phenotype observed in CD44KO mice following bleomycin injury (Munger et al., 1996). Moreover, in vivo models of tissue injury have supported a role for adhesion receptors in the resolution of inflammation, with evidence suggesting a mechanism involving the activation of latent TGFβ (Munger et al., 1996; Morris et al., 2003; Mu et al., 2002). Other studies have revealed that integrin-dependent activation of latent TGFβ requires an intact cytoskeleton (Munger et al., 1996; Annes et al., 2004). Even more intriguing are our data indicating that CD44 also regulates activation of TGFβ, thereby further restricting its effects to areas of active injury and repair. A growing body of evidence suggests that force generation across the cell has a necessary role in the activation of latent TGFβ (Annes et al., 2004; Hinz and Gabbiani, 2003). By affecting the formation of stress fibers, CD44 may certainly modulate cell rigidity, thereby impacting the activation of latent TGFβ. Further studies to determine the extent to which CD44 participates in protease and integrin-mediated activation of latent TGFβ will be of interest.

In conclusion, our data indicate that CD44 plays a role in cytoskeletal organization, thereby impacting fibroblast migration. This mechanism is at least partially TGFβ dependent. We propose that in response to tissue injury, CD44-dependent MMP-mediated activation of latent TGFβ in the proper temporal and cellular context facilitates a coordinated and optimal response to injury. This hypothesis is supported by the following observations: (1) induction of lung injury in a CD44KO mouse results in inefficient activation of latent TGFβ that is associated with unremitting inflammation (Teder et al., 2002); (2) cultured primary CD44KO fibroblasts exhibit altered cytoskeletal architecture and migratory properties inconsistent with an `activated' phenotype; (3) optimal MMP-mediated TGFβ activation is dependent on CD44; (4) treatment of CD44KO fibroblasts with active TGFβ rescues the cytoskeletal organization and velocity of migration; (5) CD44 is absolutely required for directional migration. Taken together, these data indicate that CD44 participates in fibroblast adhesion and migration and does so in a TGFβ-dependent fashion, thereby affecting the functional properties of lung fibroblasts.

Animal protocols were approved by the IACUC of The Wistar Institute. CD44KO mice were backcrossed ten generations onto a C57BL/6 background and C57BL/6 mice were used as CD44WT controls. Lungs were dissected from 8- to 12-week-old mice, minced into 1-3 mm pieces, and cultured in DMEM-10% FBS, at 37°C in an atmosphere of 5% CO₂. Cells were grown to ∼70% confluence (subconfluent) and passaged using trypsin-EDTA. All experiments were performed with primary lung fibroblasts at 3 to 5 passages and different animals were used for each experiment, which was carried out in triplicate. Immunohistochemistry indicated that more than 90% of the cells were vimentin positive (mesenchymal cell), of which 7-10% were α-SMA positive (myofibroblast).

Cells were grown on glass coverslips (Invitrogen, Carlsbad, CA) for the indicated times, washed with PBS, then fixed with 3.7% formaldehyde. In some experiments, cells were grown in the presence or absence of exogenous latent or active TGFβ1 (R&D Biosystems, Minneapolis, MN) or in the presence or absence of an anti-TGFβ antibody (R&D Biosystems). After washing in PBS, ammonium chloride (50 mM, 10 minutes) was added to neutralize residual formaldehyde. Fixed cells were permeabilized with 0.2% Triton X-100 and incubated with Rhodamine-phalloidin (Invitrogen) or anti-vinculin (Sigma, St Louis, MO) and/or anti-CD44 (E.P.), followed by FITC-labeled secondary antibody. Samples were counterstained with DAPI (Invitrogen), and mounted with Slow Fade. A Leica TCS SP2 confocal microscope or Nikon E600 upright microscope was used for imaging.

Observers blinded to the genotype were asked to report qualitative differences in cytoskeletal architecture. In order to quantify the difference in stress fiber phenotype, the number of stress fibers was measured as follows. Using Image Pro Plus software, a line profile was generated for cells within a high-power field where immunofluorescent spikes represented stress fibers. A graphic depiction was then generated where the x-axis represented distance across the cell, and y-axis represented levels of fluorescence. All images were obtained with fixed acquisition parameters and quantifying background fluorescence ensured analytical consistency.

CD44KO fibroblasts in the log phase of growth were transfected with full-length murine CD44 (pRc/CMV/CD44Δ1230) or empty vector (Puré et al., 1995) using Nucleofection (Amaxa Biosystems, Gaithersburg, MD) according to the manufacturer's recommendations. Transfection efficiency as detected by cotransfection with GFP was approximately 30%. Transfected cells were plated and harvested for use within 24-96 hours when optimal expression of CD44 was observed.

Fibroblast migration was analyzed as previously described (DeBiasio et al., 1987). Briefly, fibroblasts were grown in DMEM with 10% FBS serum, penicillin, streptomycin, fungizone, L-glutamine and 10 mM HEPES buffer until confluent and then `scratched' with a pipette tip to produce wounds between 1200 and 1500 μm wide. The wounded monolayer was then washed and incubated for a 30 minute recovery period. Specimens were maintained in 10 mM HEPES-buffered medium in a closed flask environment to control humidity and pH control. Temperature was maintained at 37°C and live-cell migration observed under 4× phase-contrast optics on a Nikon TE300 inverted microscope (Nikon Instruments, Melville, NY) modified for time-lapse acquisition with a Prior motorized focus drive (Prior Scientific, Rockland MA). Images were acquired using an Evolution QEi monochrome camera (Media Cybernetics, Silver Spring, MD) and ImagePro software (Media Cybernetics).

Coordinates of migrating cells were generated with tracking software (Image Pro Plus). Mean migration velocities, cellular displacement and confinement ratios (distance between start and end point divided by total length of track) were calculated over the indicated time periods as described (Mrass et al., 2006). Instantaneous velocity (V_inst) was defined as the velocity of a cell during a 2 hour observation period.

Fibroblasts (1×10⁵) were cultured in six-well plates with 10% FBS/DMEM in triplicate for each condition until confluent and then wounded with either a pipette tip or a sterile comb and then washed. Fresh DMEM with 1% FBS was added and cells were cultured at 37°C, 5% CO₂ for 24 hours. When indicated, cells were pre-incubated with 1 μM GM6001 MMP inhibitor (Calbiochem, EMD Biosciences, San Diego, CA) for 30 minutes, then wounded and cultured in fresh medium containing 1 μM GM6001. After 24 hours, conditioned media were collected, centrifuged to remove cellular debris, and half the conditioned media was treated with 5 μg/ml of anti-TGFβ antibody for 1 hour at 4°C. Total and active-TGFβ was measured using a cell-based bioassay. Briefly, mink lung epithelial cells (MLECs) stably transfected with a truncated plasminogen activator inhibitor 1 promoter fused to a firefly luciferase reporter gene (Abe et al., 1994) (generously provided by Steve Albelda, Philadelphia, PA), were plated at 1.5×10⁴ cells per well in 96-well plates with DMEM-10% FBS and after 5 hours the culture medium was replaced with a 1:1 mixture of conditioned medium and serum-free DMEM-0.1% BSA to determine levels of active TGFβ, or with a 1:3 mixture of conditioned medium, heated for 10 minutes at 80°C, and DMEM-0.1% BSA to determine levels of total TGFβ. Conditioned medium from each triplicate was tested in triplicate. Acid-activated TGFβ1 was used as a positive control and to generate a standard curve for each assay. Serum-free DMEM-0.1% BSA was used as the negative control. To ensure that luciferase induction in MLECs was due to the presence of TGFβ specifically, acid-activated TGFβ (10-50 pg/ml) was combined with a TGFβ neutralizing antibody (5 μg/100 μl) (a generous gift from Steve Albelda). After 18 hours, luciferase expression was quantified with an Enhanced Luciferase Assay Kit (BD Pharmingen, San Diego, CA) and a Wallac Victor2 1420 Multilabel counter. Cells from which the conditioned medium was collected were lysed in M-Per Mammalian Protein Extraction Reagent (Pierce Biotechnology, Rockford, IL) and total protein was quantified using a BCA Protein Assay Kit (Pierce Biotechnology) and used for normalization.

Fibroblasts (1×10⁵) seeded into six-well plates in were grown for 2 days 10% FBS-DMEM and then for 1 day in 0.1% FBS-DMEM before wounding as above or treating with acid-activated TGFβ1 (0.02 nM) and culturing for one additional day. Nuclear proteins were then extracted using a Nuclearbuster Protein Extraction kit (Novagen, EMD Biosciences, San Diego, CA). Protein concentration was determined using a reducing-agent-compatible BCA Protein Assay kit (Pierce Biotechnology). Equal amounts of nuclear protein were resolved by 8% SDS-PAGE and transferred onto PVDF membranes (Life Sciences, PerkinElmer, Boston, MA). Membranes were blocked in 5% dry milk and then probed with 1 μg anti-Smad2-P (pSmad) (Cell Signaling Technology, Danvers, MA) or anti-retinoblastoma (Rb) (Santa Cruz, Santa Cruz, CA) in 5% dry milk/0.1% Tween-PBS. Membranes were probed with horseradish peroxidase (HRP)-conjugated donkey anti-rabbit secondary antibody (Jackson Immunoresearch, West Grove, PA) and ECL Plus western blotting detection reagent (GE Healthcare, Amersham Biosciences, Piscataway, NJ). Bands were visualized by exposure to HyBlot CL autoradiography film (Denville Scientific, Metuchen, NJ), quantified by densitometry using Image-Pro software.

All data were analyzed using the Student's t-test, except instantaneous velocity (V_inst). The Mann-Whitney U test was used to analyze V_inst. P<0.05 was considered statistically significant.

We would like to thank E. A. Hawthorne for technical assistance, F. S. Keeney for assistance with the graphical presentation of imaging data, and A. Whitmore and S. Berliner for help in preparing the manuscript. This work was supported by awards from the PHS [5 PO1 HL062250-08 (R.K.A. and E.P.), RO1 HL65507 (E.P.) and T32 HL07586(P.A.)] and a grant from the Pennsylvania Department of Health. P.M. is supported by a Cancer Research Institute postdoctoral fellow.