Vitronectin or Fibronectin Is Required for Corneal Fibroblast–Seeded Collagen Gel Contraction

Abstract

purpose. The wound healing process in the corneal stroma involves the activation of corneal keratocytes and the expression of associated phenotypes (fibroblasts and myofibroblasts). One of these phenotypes, the myofibroblasts, synthesizes α-smooth muscle actin in order to affect wound closure by contracting the surrounding matrix. Excessive contraction results in the formation of unresolvable scars that are undesirable in the corneal stroma. The authors tested the effect of vitronectin and fibronectin on the contraction process associated with corneal wound healing.

methods. Collagen gels were prepared and were exposed to different treatments of fetal calf serum (FCS). The FCS used was either depleted of fibronectin and vitronectin or contained a known concentration of fibronectin, vitronectin, or both at 50 μg/ml. Contraction was measured using image analysis and cross sections of contracted gels were examined for α-smooth muscle actin expression using laser confocal microscopy.

results. Fibroblasts seeded in collagen gels paralleled the morphologic characteristics and cell distribution of keratocytes in unwounded cornea. Matrix contraction was dependent on the presence of fibronectin and/or vitronectin where myofibroblasts were present. The cell-mediated contraction process was maximal at 0.5 × 10⁵ fibroblasts/ml.

conclusions. These studies showed that vitronectin or fibronectin is required for the myofibroblast-associated contraction to occur in this in vitro model of stromal wound healing. This model system shows a distinct potential for further studies relating to the corneal wound healing process.

Wound healing of the cornea restores to damaged tissue the ability to perform light refraction. Stromal wound repair involves the migration of corneal fibroblasts (CFs), the transformation of CFs to myofibroblasts which close the wound by contraction, and remodeling of the surrounding tissue. ¹ Abnormal contraction during stromal wound healing invariably results in scar formation. ² This study focuses on the early phase in the process of contraction associated with corneal tissue repair. 

After an injury to the stroma, keratocytes adjacent to the wound are activated to become CFs, which, in addition to other functions, are believed to be responsible for the contraction process. ^{3 4} These activated CFs assemble actin filaments similar to those in smooth muscle cells ⁵ and on this basis, have been named myofibroblasts (MFs). ^{3 4} The transformation of CFs into MFs can be identified by the expression of α-smooth muscle actin (α-SMA). ⁶ Keratocytes in the unwounded normal (quiescent) stroma do not express this protein. ^{3 4}  

The role of MFs in contraction can be effectively studied using established in vitro models. Such models have been validated previously by Minami et al., ⁷ Zieske et al. ⁸ and Dimitrijevich et al. ^{9 10} The human model established by Dimitrijevich et al. ^{9 10} is based on a fibroblast-seeded collagen type I gel (FSCG), which constitutes the stromal element. In all stromal models where fibroblasts are seeded into collagen gels, they are activated and become phenotypically modified to resemble the MFs that are present in the wounded stroma. ¹¹ Hence, the FSCG provides a three-dimensional, in vitro environment that closely resembles the wound healing state found in vivo, thus offering the opportunity to study the CF or MF response during this event. ^{6 11 12} The FSCG model system is supported by studies that have demonstrated that CFs seeded into collagen gels have a similar distribution and appearance to those found in unwounded cornea. ^{13 14 15} As a result, the FSCG system provides a good model for investigating CF/MF responses during wound healing. ^{16 17 18 19 20 21 22}  

There are two schools of thought regarding the tissue contraction process during stromal wound healing. One of these proposes that the presence of a contractile force within cells is a response to the tension produced within the surrounding extracellular matrix. ²³ It is alternatively proposed that contraction is an action brought about by the resident cells (e.g., MFs) as they generate the motive force needed to reorganize the surrounding matrix. ²⁴ In support of the latter view, Welch et al. ²⁵ and Garana et al. ¹ proposed that the contraction is based on the interaction of intracellular stress fibers and extracellular fibronectin (Fn). They concluded that during wound healing, CFs undergo a phenotypic transformation to MFs, which contribute to contraction by pulling in and organizing the extracellular matrix. Studies of molecular linkages between the corneal MFs and the collagen fibers suggest that Fn may act as the connective bridge between resident fibroblasts and the adjacent fibrillar collagen matrix. ²⁶ Later work proposed that fibroblasts active in the wounding process attach to Fn and pull on the surrounding matrix. ²⁵ This suggestion was supported by Garana et al. ¹ who found that serum containing Fn enhanced the contraction process. Furthermore, this group identified the presence of extracellular Fn in the stroma at day 14 after a gape wound to the feline cornea. Because contraction of FSCGs was found to occur in medium containing serum that had been depleted of Fn, ²⁷ this brings to the fore the question of other components within the serum that may act in a similar way to Fn. Schafer et al. ²⁷ proposed that vitronectin (Vn) could replace Fn as a bridging molecule between fibroblasts and collagen. Vn is known to bind to collagen ²⁸ and shares with Fn the cell-binding domain of arginine-glycine-aspartate (RGD). ^{29 30} To date, no evidence has emerged to support a role for Vn in wound contraction, and it remains to be shown definitively what role(s) Vn and Fn play in corneal wound healing. 

This article presents evidence for the direct effect of Vn and/or Fn on fibroblast-induced contraction in a wound healing model of the corneal stroma. An FSCG model was used to determine the extent of contraction and the associated presence of MFs. The results identify an interchangeable requirement for Vn or Fn in the contractile process and provide further insight into the mechanism involved in the course of normal corneal wound healing. 

Materials and Methods

Preparation of CFs

Corneas were excised from bovine eyes received within 2 to 3 hours of slaughter. Stroma, with all endothelium and epithelium peeled off, was isolated from the excised corneas using toothed forceps and a scalpel. Several pieces of stromal tissue were placed in individual wells of a six well tissue culture plate (Corning Costar Corporation, Cambridge, MA) and allowed to attach to the surface for 30 minutes. Five milliliters of 1:1 Dulbecco’s modified Eagles medium (DMEM)/F12 (ICN Biomedicals, OH) supplemented with 10% (vol/vol) fetal calf serum (FCS; P.A. Biologicals, Sydney, Australia) were added to each well, covering the tissue. These were incubated at 37°C in a humidified atmosphere of 5% CO₂ in air. CFs migrated out of stromal tissue and proliferated in culture wells over several days. When the wells were confluent, after approximately 7 days, the monolayers were treated with 0.1% (vol/vol) trypsin (CSL, Melbourne, Australia) for 5 minutes at 37°C, and the harvested cells were sedimented by centrifugation. The cell pellet was washed twice in serum-free DMEM/F12 and held for seeding into the collagen gel mixture (see below). 

FSCG Preparation

Collagen gels were quickly prepared at room temperature with one part 10× Ham’s/F12 medium (ICN Biomedicals), one part reconstitution buffer and eight parts collagen type I (Cellagen 0.3% pepsin solubilized; ICN Biomedicals) as described in Dimitrijevich et al. ¹⁰ This mixture was inoculated with bovine CFs to a final population density of 5.0 × 10⁴ cells/ml (unless otherwise stated in the Results section), and 2 ml was poured into individual semipermeable membrane inserts of 24-mm diameter and 0.4-μm pore size (Corning Costar Corporation). These inserts were placed in the wells of a six-well plate, and the gels were allowed to polymerize for 60 minutes at 37°C. Culture medium was then added to cover the top of the gel, and to the well, so that the nutrients were supplied through the apical and basal surfaces. This assembly was incubated at 37°C in a humidified atmosphere of 5% CO₂ in air. Acellular gels were used as controls for all experiments. 

Culture Media

Gels were maintained in keratinocyte serum-free medium (KSFM; Gibco BRL Life Technologies, Grand Island, NY) and were supplemented with either 2% (vol/vol) FCS or FCS stripped of Fn and Vn, termed double-depleted serum (DD FCS). DD FCS was prepared according to methods described by Engvall et al. ³¹ to remove Fn and Underwood et al. ³² for Vn removal. Fn was removed from FCS using a packed column of gelatin-Sepharose (Pharmacia, Sweden). The Fn-depleted FCS was then fed through a monoclonal anti-bovine Vn affinity column, to form DD FCS. Vn, which was added back to DD FCS to form the treatment known as “Vn,” was derived from the product obtained from stripping intact FCS as described above. Commercially available Fn (Sigma, St. Louis, MO) was used in the experiments in which Fn was added back to DD FCS to form the treatment known as“ Fn.” Protein concentrations for both Vn and Fn were determined using bicinchoninic acid (BCA) protein assays (cat. no. 23223/4; Pierce, Rockford, IL). 

Measurement of Gel Contraction

To determine the degree of gel contraction, an Image Analysis system (Q570; Cambridge Diagnostics, Billerica, MA) was used to measure the surface area of the individual gels. The measured area was subtracted from the original surface area of the collagen gel and expressed as a percentage of initial size. Two experiments were conducted in each case, with three replicate samples per treatment. 

Hematoxylin and Eosin Staining

This technique was based on previously described methods. ^{33 34} FSCGs and intact bovine cornea were embedded in Tissue-Tek OCT compound (Sakura Finetek, CA), snap-frozen, and sectioned at 6-μm thickness onto gelatin-coated slides. Air-dried sections were fixed by immediately immersing them in formalin-acetic-alcohol (10%:85%:5% vol/vol, respectively) solution for 30 seconds. Slides were stained with hematoxylin and eosin, mounted in Gurr mounting medium (BDH Chemicals, Poole, UK), and viewed using a Leica DMLB light microscope (Deerfield, IL). 

Cell Visualization

A Cell Tracker Green fluorescent dye (Molecular Probes, Eugene, OR) was used to visualize the CFs seeded in gels. The washed cell pellet (see above) was resuspended in Cell Tracker Green solution that was diluted in serum-free DMEM/F12 to a final concentration of 25 μM. The final CF concentration was 1.0 × 10⁴ cells/ml. This mixture was placed in a 37°C incubator for 50 minutes and then inoculated into collagen gels as described above. The gels were allowed to set and DMEM/F12 + 10% (vol/vol) FCS was added, followed by incubation at 37°C for 48 hours in a humidified atmosphere of 5% CO₂ in air. Intact gels were viewed by laser confocal microscopy (see below). 

Immunohistochemistry

Collagen gels were embedded in Tissue-Tek OCT Compound at days 9, 20, or 30, snap-frozen, and sectioned at 6-μm thickness onto gelatin-coated slides. Sections were allowed to air dry and were then blocked with 2% (wt/vol) bovine serum albumin in phosphate buffered saline (PBS). The sections were then incubated at room temperature with a monoclonal antibody to α-SMA (cat. no. A2547; Sigma) at a 1:20 dilution at room temperature for 60 minutes. Normal mouse serum was used instead of α-SMA antibody as a negative control. Sections were washed three times with PBS and stained with rabbit anti-mouse fluorescein isothiocyanate–conjugated antibody (DAKO Corporation, Santa Barbara, CA) for 60 minutes at room temperature, during which time they were kept in the dark. Sections were washed three times in PBS and mounted in Fluorosave (Calbiochem, La Jolla, CA). Sections were viewed by laser confocal microscopy (see below). 

Laser Confocal Microscopy

Immunostained sections and Cell Tracker–labeled samples were viewed with a scanning laser confocal microscope (TCS-40; Leica, Heidelberg, Germany) using a 100× objective with a krypton/argon-mixed gas laser and an excitation wavelength of 494 nm. 

Results

The CFs were found to maintain a spatial distribution and morphology in vitro that was similar to that of keratocytes observed in unwounded cornea. Within 24 hours, the CFs in the collagen type I gel changed from a spherical (data not shown) to elongated cell morphology (typically as seen in Fig. 1 ), which resembled that found in unwounded cornea. The distribution of CFs within FSCGs (Fig. 2B ) also was similar to that seen in unwounded cornea (Fig. 2A) . The extended cell processes were a distinguishing feature of cells in FSCGs and a point of similarity to morphology described in previous reports of gel systems. ^{13 14 15}  

Gel contraction was measured on day 20 of culture using quantitative image analysis. Experiments were designed to compare any differences in gel contraction related to the CF densities in FSCGs. Surprisingly, the measured FSCG contraction was not linearly proportional to the CF population seeding density. Figure 3 shows that contraction was optimal when the CF population of the gel was 0.5 × 10⁵ cells/ml. Contraction was reduced at the higher and lower cell densities of 1.0 × 10⁵ and 0.1 × 10⁵ cells/ml, respectively. FSCGs containing 0.5 × 10⁵ cells/ml contracted by more than two times compared with those at 0.1 × 10⁵ cells/ml and approximately 25% of the contraction of FSCGs containing 1.0 × 10⁵ cells/ml. Acellular controls showed no measurable gel contraction (data not shown). 

The role of Fn and/or Vn in FSCG contraction was investigated in gels by comparing contraction in the presence or absence of replacement Fn and/or Vn (the Fn, Vn, and Vn + Fn treatments shown in Table 1 ). The possible effects of other serum factors on gel contraction were reduced by the use of 2% (vol/vol) FCS or DD FCS, a serum level sufficient to maintain CF viability. All FSCGs used in these experiments contained 0.5 × 10⁵ CFs/ml, in keeping with data from cell density studies in this article as well as cell numbers observed in vivo. ³⁵ Acellular controls for each treatment showed no detectable contraction (data not shown). The extent of contraction of FSCGs in the absence of Fn or Vn was significantly reduced (Fig. 4) . In treatments where Fn and/or Vn were replaced, the level of contraction over a 20-day period was restored to that of intact FCS. No significant difference was observed between Fn and/or Vn replacement treatments (which ranged between 10% and 15% gel contraction). Each of these treatments did not differ significantly from the intact FCS-positive controls. 

At day 9, day 20, and day 30, CFs had transformed to MFs when Fn and/or Vn were present (as evaluated by the expression of α-SMA). These results are summarized in Table 1 . In the absence of Fn or Vn (treatments using only DD FCS), no contraction was observed and no MFs could be detected at day 9. The same treatment of FSCGs at day 20 indicated the presence of MFs (Fig. 5) , despite minimal gel contraction (1.70%). MFs were also detected at day 30 when both Fn and Vn were absent. The contraction of FSCGs with this serum treatment also was increased at day 30 (9.7%). 

Preliminary experiments were conducted to examine several additional factors that may have affected the level of measured contraction. These factors included peripheral attachment of FSCGs to the walls of the culture vessel, CF cell passage number, and the final concentration of Fn and/or Vn. The peripheral attachment of FSCGs to culture vessel walls and cell passage number of CFs did not affect the level of contraction measured (data not shown). Furthermore, increasing the concentration of purified Fn and/or Vn from 50 to 300 μg/ml in the replacement experiments, did not increase the measured contraction of FSCGs (data not shown). Treatments containing only DD FCS (with all Fn and Vn removed) were used as negative controls and treatments with intact FCS (with both Fn and Vn) were positive controls. 

Overall, these results showed that either Fn or Vn were involved in the contraction of FSCGs and that Vn was as effective as Fn in mediating this process. 

Discussion

Our results provide direct evidence for the requirement for Fn and/or Vn in the contraction of a wound using an in vitro model system. It is generally accepted that contraction is an important and necessary phase of the wound healing process, yet there is little understanding as to the underlying mechanism involved. The model used in this investigation identified several features involved in collagen gel contraction. We determined that a specific CF density was required for maximum contraction to occur. The phenotypic transformation of CFs into MFs was monitored and using this we showed that MFs were responsible for contraction that occurred if Vn and/or Fn was available. The change in cell morphology from rounded to elongated suggested that the CFs populating the FSCGs were attempting to initiate cell–cell communication. It is not clear if this distribution of CFs was an alignment to optimize contraction or an attempt to mimic the in vivo tissue by forming functional gap junctions. ³⁶ If such a network was established, the FSCG could be a very valuable model, in that it accommodates both cell–cell and cell–matrix communications. 

Our data showed that the presence of either Fn or Vn was essential for the MF-mediated contraction of FSCGs. It is well accepted that serum components play a critical role in fibroblast activation and postwounding contraction. ^{6 12 37 38} However, the role of CF/MFs and their dependence on Fn or Vn in the wound-healing process remains to be determined. ^{16 21 27 38 39 40} Fn is synthesized by CFs as an extracellular matrix component and is known to be present in increasing amounts in the postwounded corneal stroma. ^{1 41} In the present study we have demonstrated that Fn increased the contraction of FSCGs, which was well above the level of contraction shown by the negative controls. We also showed that Vn had the same effect as Fn, supporting the hypothesis proposed by Schafer et al. ²⁷ that Vn can replace Fn in its role as a connective bridge between CFs and the surrounding collagen matrix. This sensitivity to Vn may relate to its localized presence in the corneal stroma. ⁴² The experiments in which FSCGs were treated with both Vn and Fn showed that the contraction achieved was equal to that which occurred when Vn or Fn were used individually. This nonadditive effect was likely to be related to the fact that Vn and Fn competed for the same binding domain on collagen type I (see Gebb et al. ²⁸ ). Thus, when both Fn and Vn were present, Vn interacted preferentially with CFs and their surrounding collagen matrix. As expected, with intact FCS (positive control), the contraction was equal to that observed when Vn, Fn, or both were present. Integrins are likely to be involved with this connective bridge between CFs and their surrounding matrix proteins (Vn/Fn). Previous in vitro studies have in monolayer cultures demonstrated that CFs adhered to Fn and Vn via cell surface integrins. ⁴³ Fibronectin integrin receptors include alpha_v beta_1/3/5, whereas vitronectin has been associated with alpha_v beta_1/5 subunits. ⁴⁴  

It was demonstrated in this study that in the absence of Vn and Fn (the DD FCS treatment), MFs were unable to effectively cause FSCG contraction at day 20. The low level of contraction that did occur at day 20 may be attributed to trace amounts of endogenous Vn and/or Fn synthesized by the resident CFs. This is presumably the case where FSCG contraction increased at day 30. Under the same conditions, detection of α-SMA demonstrated that CFs did transform to the MF phenotype. In the light of results obtained from cell density studies, possible reasons for this are discussed below. 

We found that CF density in a three-dimensional matrix directly affected the contractile capacity of the FSCG (Fig. 3) in the presence of serum or its components (Fn/Vn). This is consistent with reported findings related to CF monolayer cell culture studies, ³⁷ which studied the effect of cell density on the expression of α-SMA in vitro. By monitoring the expression of α-SMA it was found that 70% to 80% of the cell population had transformed to MFs in CF cultures seeded at low density (10³ cells/ml). In CF cultures seeded at high density (10⁵ cells/ml) only 5% to 10% of the cell population was α-SMA positive. ³⁷ The cell density–dependent MF transformation mechanism is thought to be linked to a regulatory role by transforming growth factor beta (TGF-β). ^{12 37} That is, CFs only transformed to MFs when cell–cell contact was limited by low cell density and TGF-β was present. ³⁷ Our studies confirm the suggestion that at specific cell densities, and in the presence of TGF-β (likely to be present in DD FCS or FCS), CFs transformed to MFs, which in the presence of Vn and/or Fn drew on their surrounding matrix and initiated contraction. Hence, cell density and factors such as TGF-β are involved in the transformation of CFs into MFs. ^{12 37} However, the ability of the MFs to undergo contraction of the surrounding matrix only occurred when Vn and/or Fn were present. Conversely in the case when Vn and/or Fn were absent and despite α-SMA being detected in FSCGs, these MFs were unable to bind to collagen type I and contraction did not occur. 

In conclusion, our studies have shown that MFs are dependent on Fn and/or Vn to initiate the contraction process. Considered together with the prior reports, the results of this study are consistent with the following interpretation: TGF-β and a specific cell density are required for MF transformation, but it is only in the presence of Vn and/or Fn that these cells are able to connect to the surrounding type I collagen and initiate contraction. This work has also shown that the transformation of CFs to MFs is dependent on a specific cell density in a three-dimensional FSCG. This finding confirms the significance of the cell density in the process of MF transformation in FSCGs as previously reported for CF monolayer cultures. ³⁷ These results allow a direct comparison of the contracted and noncontracted FSCGs, and support the utility of noncontracted FSCGs, such as those used in this study. These FSCGs may prove valuable for future wound healing studies. 

 Supported by an Australian Postgraduate (Industry) Award (LT).

 Submitted for publication January 7, 1999; revised April 28, 1999; accepted August 17, 1999.

 Commercial relationships policy: C5.

 

Figure 1.

 

A representative CF typically seen within an intact FSCG (seeded with 0.5 × 10⁵ cells/ml) showing fully elongated cell morphology 24 hours after assembly. Confocal image of CF stained with Cell Tracker. White arrows indicate cell extensions. Bar, 100 μm.

Figure 2.

 

The distribution of corneal keratocytes in FSCGs and unwounded corneal stroma. A transverse section of intact bovine corneal stroma (A), and of an FSCG (after 24 hours), which is the stromal component of a corneal model system (B). Note the similarity in distribution of CFs when compared to that of the unwounded cornea. Hematoxylin and eosin sections viewed by phase contrast light microscopy. Bar, 40 μm.

Figure 3.

 

The effect of CF seeding density on the contraction of FSCGs at day 20. Maximum contraction occurred when fibroblasts were seeded into gel at 0.5 × 10⁵ cells/ml in KSFM supplemented with 2% (vol/vol) FCS. Bars, SD of mean.

Table 1.

 

Summary of Results Showing the Measured Contraction of Gels and the Associated MF Expression in Response to Different Treatments

Treatment*	Day 9			Day 20			Day 30
		Gel Contraction (%)	Expression of α-SMA	Gel Contraction (%)	Expression of α-SMA	Gel Contraction (%)	Expression of α-SMA
1. KSFM+ 2% (vol/vol) FCS (FCS)	0	+	13.6	+	16.4	+
2. KSFM+ 2% (vol/vol) DD	0	−	1.7, †	+	9.7, †	+
FCS (DD FCS)
3. KSFM+ 2% (vol/vol) DD	0	+	13.8	+	17.8	+
FCS+ 50 μg/ml of Vn (Vn)
4. KSFM+ 2% (vol/vol) DD	0	+	13.4	+	14.2	+
FCS+ 50 μg/ml of Fn (Fn)
5. KSFM+ 2% (vol/vol) DD	0	+	12.2	+	14.9	+
FCS+ 50 μg/ml of Vn and
FN (Vn+ Fn)

  n = 2. +, positive expression of α-SMA in all resident MFs; −, no expression of alpha-SMA in resident MFs.

*  Abbreviations for treatments in parentheses.

†  Significantly different from all other treatments.

Figure 4.

 

The effect of the presence of Fn and Vn on the contraction of FSCGs at day 20. All FSCGs were seeded with CFs at 0.5 × 10⁵ cells/ml. Depletion of Fn and Vn (DD FCS) reduced the contraction of gel. The ability of the FSCG to contract was restored by adding back either Vn, Fn, or both (Fn + Vn), to the equivalent level in intact serum (FCS). Error bars, SD of mean.

Figure 5.

 

A MF in a contracted FSCG in 2% (vol/vol) FCS with Vn and Fn present at day 20. The MF was identified by the positive expression of α-SMA. Bar, 75 μm.

The authors thank Anne Underwood for her valuable review of this manuscript.