Differences in the Regulation of Fibroblast Contraction of Floating Versus Stressed Collagen Matrices*

To learn more about the regulation of contraction of collagen matrices by fibroblasts, we compared the ability of lysophosphatidic acid (LPA) and platelet-derived growth factor (PDGF) to stimulate contraction of floating and stressed collagen matrices. In floating collagen matrices, PDGF and LPA stimulated contraction with similar kinetics, but appeared to utilize complementary signaling pathways since contraction obtained by the combination of growth factors exceeded that observed with saturating concentrations of either alone. The PDGF-simulated pathway was selectively inhibited by the protein kinase inhibitor KT5926. In stressed collagen matrices, PDGF and LPA stimulated contraction with different kinetics, with LPA acting rapidly and PDGF acting only after an ∼1-h lag period. Pertussis toxin, known to block signaling through the Gi class of heterotrimeric G-proteins, inhibited LPA-stimulated contraction of floating but not stressed matrices, suggesting that LPA-stimulated contraction depends on receptors coupled to different G-proteins in floating and stressed matrices. On the other hand, the Rho inhibitor C3 exotransferase blocked contraction of both floating and stressed collagen matrices. These results suggest the possibility that distinct signaling mechanisms regulate contraction of floating and stressed collagen matrices.

Closure of cutaneous wounds involves three processes: epithelialization, connective tissue deposition, and contraction. Wound contraction, which brings the margins of open wounds together (1,2), is believed to be mediated by specialized fibroblasts called myofibroblasts because of their content of actin stress fibers and ␣-smooth muscle actin (3). The myofibroblast phenotype can occur early or late during the wound contraction process depending on the mechanical resistance of surrounding tissue (4). Myofibroblasts have also been implicated in the pathology of wound contractures and fibrotic disease (5,6).
Using several different culture models, we and others have studied the ability of fibroblasts to reorganize and contract collagen matrices in vitro. In the "floating" model (7), a freshly polymerized collagen matrix containing fibroblasts is released from the culture dish and allowed to float in culture medium, and contraction occurs in the absence of external mechanical load and without appearance of actin stress fibers in the cells (8). In the "attached" model, a polymerized collagen matrix containing fibroblasts remains attached to the culture dish during contraction. In this case, mechanical load (i.e. isometric tension) develops during contraction, and cellular stress fibers assemble (9 -11). Finally, the two-step "stressed" model combines an initial period of attached matrix contraction leading to mechanical loading, followed by release of the matrices, resulting in mechanical unloading and further contraction as mechanical stress dissipates (i.e. stress-relaxation) (12).
Contraction of collagen matrices depends on cell binding to collagen through ␣ 2 ␤ 1 integrins (13)(14)(15) and requires stimulation by serum factors (16 -18). Otherwise, the signal transduction mechanisms that regulate contraction are poorly understood. Moreover, previous studies have assumed for the most part interchangeability between different growth factors used to stimulate contraction and different model systems used to measure contraction.
Platelet-derived growth factor was identified as the putative factor in serum that stimulates contraction of floating matrices (13,19). In a previous study (20), we found that PDGF 1 attenuated the ability of sodium vanadate to inhibit serum-dependent contraction of floating matrices, but enhanced the ability of vanadate to inhibit contraction of mechanically stressed matrices. These opposing responses to PDGF raised the possibility that different signaling pathways regulate contraction in floating and stressed matrices.
In this study, we examined the above possibility by comparing floating and stressed matrix contraction stimulated by platelet-derived growth factor and lysophosphatidic acid (21,22). The results of our studies suggest that collagen matrix contraction is determined by complex interrelationships between different growth factors, signal transduction mechanisms, and the mechanical state of the cells, i.e. mechanically loaded or unloaded. Details are reported herein. harvested from monolayer culture with 0.25% trypsin and 1 mM EDTA. Trypsin was neutralized with soybean trypsin inhibitor (3.3 mg/ml) or 10% serum-containing medium. To harvest fibroblasts from collagen matrices (see below), the matrices were incubated at 37°C for 10 min with 0.05% trypsin and 0.53 mM EDTA (200 l/matrix), followed by treatment in a shaking incubator for 20 min with 5 mg/ml collagenase (250 l/matrix) and then 10 mg/ml soybean trypsin inhibitor (100 l).
The procedure for studying contraction of floating and stressed collagen matrices has been described previously (20,23) and is outlined in Fig. 1. Hydrated collagen matrices were prepared from Vitrogen 100 collagen. Neutralized collagen solutions (1.5 mg/ml) contained fibroblasts in DMEM without serum. Aliquots (0.2 ml, 2 ϫ 10 5 cells) of the cell/collagen mixtures were prewarmed to 37°C for 3-4 min and then placed in Corning 24-well culture plates. Each aliquot occupied an area outlined by an 11-mm diameter circular score within a well. Polymerization of collagen matrices required 60 min at 37°C. Subsequently, the polymerized matrices were used immediately or cultured with 1.0 ml of DMEM and 10% FBS and 50 g/ml ascorbic acid for 1-2 days to allow the cells to develop mechanical stress.
To initiate matrix contraction, freshly polymerized or mechanically stressed matrices were gently released from the underlying culture dish with a spatula into ϳ0.5 ml of serum-free DMEM containing 5 mg/ml bovine serum albumin and growth factors as indicated, after which the matrices were incubated at 37°C. For convenience in carrying out the experimental protocols, growth factors usually were added to floating collagen matrices immediately after release and to stressed collagen matrices immediately before release. Reversing the procedure did not change the results, however.
To determine the extent of floating or stressed matrix contraction, samples were fixed with 3% paraformaldehyde in phosphate-buffered saline (150 mM NaCl, 3 mM KCl, 1 mM KH 2 PO 4 , and 6 mM Na 2 HPO 4 , pH 7.2) for 10 min at 22°C. The matrices were washed, placed on a flat surface, and measured with a ruler. For quantitative purposes, contraction data are presented as the change in diameter (starting Ϫ final) measured in millimeters.
All experiments were carried out in duplicate, and every experiment was repeated two or more times. Data points and error bars in the figures represent averages and S.D. Where error bars cannot be seen, the data points overlapped.
Loading with Pertussis Toxin (PTx) and C3 Exotransferase-To load fibroblasts in monolayer culture or mechanically stressed collagen matrices with pertussis toxin, the cells were incubated overnight in DMEM and 10% FBS containing PTx at the concentrations indicated. To load fibroblasts in monolayer cultures with C3 exotransferase (C3), Lipofectin was used as a delivery system (24). Lipofectin/C3 was prepared in 120 l of antibiotic/antimycotic-free DMEM and diluted with additional antibiotic/antimycotic-free DMEM after 1 h at 22°C to give a final concentration of 10 g/ml Lipofectin and C3 as indicated. Subsequently, the cells were incubated with the Lipofectin/C3 mixture or Lipofectin prepared identically except without C3 for 30 min at 37°C. Following treatment with Lipofectin/C3 or Lipofectin alone, the cells were rinsed and further incubated with DMEM and 10% FBS for 60 min at 37°C before harvesting. To load fibroblasts in collagen matrices with C3 exotransferase, cells in stressed matrices were rinsed and incubated with DMEM containing C3 as indicated for 30 min at 37°C. Subsequently, the matrices were released, allowing C3 exotransferase to enter through ϳ4-nm plasma membrane passages that open for Ͻ5 s during basal contraction (23). Afterward, the cells were incubated for an additional 1 h at 37°C. An analogous method was used for loading C3 into cells during embryonic wounding (25). Fig. 2 shows the typical appearance of floating collagen matrices after contraction for 4 h (FMC). The diameter of the matrices was 10.5 mm at the time contraction was initiated. In DMEM alone, little contraction of the matrix occurred, but the matrix diameter decreased markedly when contraction was carried out in the presence of 100 ng/ml PDGF or 10 M LPA. Fig. 3 shows a typical time course of matrix contraction in the presence of LPA or PDGF. The rates of contraction were relatively linear over 4 h, although there was some variation from experiment to experiment whether PDGF or LPA was the better stimulator of contraction.

Contraction of Floating Collagen Matrices in Response to LPA and PDGF and Selective Inhibition of PDGF Stimulation by KT5926 -
Mixing experiments with LPA and PDGF were carried out, and typical results are shown in Fig. 4. We consistently observed faster contraction with LPA and PDGF combined than with saturating concentrations of either growth factor alone, suggesting that complementary signaling mechanisms could activate contraction. One way to distinguish these mechanisms was with the protein kinase inhibitor KT5926. As shown in Fig.  5, KT5926 completely blocked PDGF-stimulated contraction of floating matrices, but had little effect on LPA-stimulated contraction. Another protein kinase inhibitor, staurosporine, also selectively blocked PDGF-stimulated contraction, but appeared to do so by interfering with PDGF receptor phosphorylation (26). KT5926, on the other hand, had no effect on PDGF receptor phosphorylation (data not shown). Fig. 2 also shows the typical appearance of stressed collagen matrices undergoing contraction (SMC). In this case, a substantial decrease in matrix diameter occurred even in DMEM without serum or growth factors (i.e. basal contraction). During the first hour, the extent of contraction was stimulated further by the addition of LPA, but not by PDGF. Fig. 6 shows the time course for contraction of stressed matrices. The basal component of contraction (no growth factor) was complete within 5 min. LPA-stimulated contraction increased rapidly up to 1 h. On the other hand, stimulation of contraction by PDGF was undetectable until after an ϳ1-h lag period. The lag period in PDGF stimulation of stressed matrices was observed at PDGF con- centrations ranging from 12.5 to 200 ng/ml, whereas PDGF at 50 ng/ml was sufficient to maximally stimulate contraction of floating matrices (data not shown).

Differences in the Time Course of Contraction of Stressed Collagen Matrices in Response to LPA and PDGF-
The faster rate of LPA-stimulated contraction of stressed collagen matrices compared with floating matrices might have occurred because cells in stressed matrices have actin stress fibers (12,17). To examine this possibility, the matrices were released in DMEM alone, allowing stress fibers to disassemble in the absence of growth factor-stimulated contraction (23), and then LPA was added. LPA stimulated rapid contraction of stressed matrices even if added 2 h after release (data not shown), suggesting that the presence of the stress fibers at the time of LPA addition was unnecessary for rapid contraction.
LPA Stimulates Contraction of Floating and Stressed Collagen Matrices by Pertussis Toxin-sensitive and -insensitive Pathways, Respectively-Besides the differences in rate of contraction, floating and stressed collagen matrices were observed to respond to markedly different concentrations of LPA. The typical dose-response curve in Fig. 7 shows that 100 nM LPA was sufficient to maximally stimulate contraction of floating collagen matrices, but had little effect on stressed matrices. The latter required 10 -100 M LPA for maximal contraction.
The difference in the LPA concentration dependence and rate of contraction of floating versus stressed collagen matrices raised the possibility that different LPA receptors or receptorlinked signaling pathways were involved. LPA interacts with receptors linked to pertussis toxin-sensitive (G i ) and -insensitive (G q and G 12/13 ) heterotrimeric G-proteins (27). Therefore, experiments were carried out to determine the effects of pertussis toxin on fibroblast contraction of collagen matrices. Fig. 8 shows that preloading fibroblasts with pertussis toxin almost completely blocked the ability of LPA to stimulate contraction of floating matrices. Under the same conditions, PDGF-stimulated contraction was only slightly inhibited. In marked contrast, as shown by the example in Fig. 9, preloading cells with pertussis toxin did not inhibit LPA-stimulated contraction of stressed matrices (SMC). Even when the cells were incubated with 100 ng/ml PTx, there was little effect on contraction of stressed matrices (data not shown). Fig. 9 also shows that when fibroblasts were harvested from stressed collagen matrices and retested for contraction in floating matrices (FMC), contraction of PTx-treated cells was inhibited ϳ50% compared with controls. These results suggest that LPA-stimulated contraction of floating matrices depends on LPA receptors linked to pertussis toxin-sensitive heterotrimeric G-proteins, whereas LPA-stimulated contraction of stressed matrices depends on receptors linked to pertussis toxin-insensitive heterotrimeric G-proteins.
The Rho Inhibitor C3 Exotransferase Blocks Contraction of Floating or Stressed Collagen Matrices-LPA has been identified as the principal factor in serum that promotes assembly of stress fibers and focal adhesions (28) through a contractile process that depends on the small G-protein Rho (29,30). This effect of LPA is pertussis toxin-insensitive (31) and probably involves LPA receptors linked to the heterotrimeric G-protein G 13 (32). Experiments were therefore carried out to test whether collagen matrix contraction was affected by the toxin C3 exotransferase, a specific inhibitor of Rho activity (33). Fig. 10 shows that releasing stressed matrices in the presence of 20 g/ml C3 exotransferase blocked subsequent LPA stimulation of contraction. As would be predicted (23), C3 had no effect if it was added after the matrices were released (data not shown). In control experiments, C3-loaded cells were harvested from collagen matrices and tested for cell spreading on fibronectin-coated substrata after 2 and 24 h. After 2 h, C3loaded cells spread in an arborized morphology on fibronectincoated substrata (31,34), but the cells attained normal morphology within 24 h (data not shown).
To study the effect of C3 exotransferase on contraction of floating collagen matrices, C3 was loaded into fibroblasts in monolayer culture using Lipofectin. Fibroblasts treated with Lipofectin/C3, but not Lipofectin or C3 alone, showed arborized spreading morphology on fibronectin-coated substrata (data not shown). Fig. 11 presents a typical experiment showing that cells treated with Lipofectin/C3 were unable to contract floating collagen matrices regardless of whether LPA or PDGF was used to stimulate contraction, whereas cells treated with Lipofectin or C3 alone were unaffected. Therefore, the small Gprotein Rho appeared to be required not only for contraction of stressed collagen matrices, but also for contraction of floating collagen matrices. DISCUSSION The signal transduction pathways that regulate contraction of collagen matrices by fibroblasts are poorly understood. To learn more about these pathways, we compared the ability of LPA and PDGF to stimulate contraction of floating and stressed collagen matrices. The results suggest that contraction of floating and stressed collagen matrices depends on different signaling mechanisms.
PDGF and LPA appeared to stimulate contraction of floating collagen matrices by complementary signaling pathways since contraction obtained by the combination of growth factors exceeded that observed with saturating concentrations of either alone. One way to distinguish these pathways was with the protein kinase inhibitor KT5926. Staurosporine also selectively blocked PDGF-stimulated contraction, but appeared to do so by interfering with PDGF receptor phosphorylation (26). Previous studies demonstrated that serum-stimulated contraction could be blocked by inhibitors of protein kinase C (35,36), proteintyrosine kinase (36 -38), phospholipase C (39), and phosphatidylinositol 3-kinase (38). None of the other inhibitors used previously, however, was found to preferentially inhibit PDGF- versus LPA-stimulated contraction.
Although we have not determined the precise site at which KT5926 inhibits PDGF-stimulated contraction, it is worth noting other studies suggesting that KT5926 selectively inhibits myosin light chain (MLC) kinase (40), and phosphorylation of MLC has been implicated in matrix contraction, albeit indirectly (21,41). Therefore, preferential inhibition of PDGF-stimulated contraction by KT5926 may reflect a requirement for MLC kinase in MLC phosphorylation during PDGF-but not LPA-stimulated contraction. Consistent with this possibility, LPA can cause MLC phosphorylation without MLC kinase through the activity of Rho kinase (42,43). We cannot exclude the possibility, however, that preferential inhibition of PDGFstimulated contraction by KT5926 occurs at the level of some other protein kinase, and future studies will be required to analyze this point.
Surprisingly, LPA-stimulated contraction of floating matrices was pertussis toxin-sensitive, whereas LPA-stimulated contraction of stressed matrices was pertussis toxin-insensitive. One interpretation of this result is that LPA stimulates contraction of fibroblasts in floating matrices through receptors coupled to G i and stimulates contraction of cells in stressed matrices through receptors coupled to G 13 (although G q cannot be ruled out at present).
LPA receptors coupled to different heterotrimeric G-proteins have been implicated in different aspects of cell motility. LPA stimulation of fibroblast migration and chemotaxis was shown to depend on LPA receptors coupled to G i (44,45). On the other hand, LPA stimulation of stress fiber formation and focal adhesion assembly requires activation of the small G-protein Rho (29,30), a pertussis toxin-insensitive process (31) in which LPA receptors are coupled to G 13 (32). The latter pathway has also been implicated in neurite rounding and retraction (46).
Contraction of floating and stressed collagen matrices may also reflect different aspects of cell motility. Fibroblasts in floating collagen matrices are round to begin with and spread during contraction (Ref. 8; see also Ref. 11), whereas cells in stressed matrices begin in a spread morphology with prominent stress fibers and withdraw their extensions during con-traction (12,17,47). In the former case, contraction probably depends on tractional forces that accompany the protrusion of cell extensions (48). In the latter case, the stress fibers themselves can contract once there is no longer a rigid substratum to maintain isometric tension (49). Given this interpretation, it is not surprising that the Rho inhibitor C3 exotransferase blocks contraction of both floating and stressed collagen matrices since inhibition of Rho with C3 exotransferase can prevent normal spreading of freshly plated cells (34) as well as actin filament assembly and focal adhesion formation in already spread cells (50).
The observation of different regulatory mechanisms for contraction of stressed and floating collagen matrices opens the possibility for future studies to characterize in more detail the signaling mechanisms involved and the interrelationships be- tween different growth factors, signal transduction mechanisms, and the mechanical state of the cells, i.e. mechanically loaded or unloaded. In addition, our results have potentially important clinical implications. In the absence of complications, the process of wound contraction leads to wound closure with little scarring or loss of function. In large wounds and some pathological situations such as hypertrophic scars, however, the consequences of contraction can result in loss of joint motion or major body deformations referred to as contractures (51)(52)(53). Contraction of floating collagen matrices resembles more closely the initial phase of wound contraction dependent on cell motility (8,54), whereas the myofibroblast-like cells in mechanically stressed matrices are more typical of the late phase of wound contraction and contractures (3,4). Consequently, wound contraction and wound contracture may be under the control of different growth factors and signal transduction regulatory mechanisms in vivo and therefore subject to different pharmacological means of clinical intervention.