Growth Factor Regulation of Corneal Keratocyte Differentiation and Migration in Compressed Collagen Matrices

Abstract

Purpose.: To evaluate a novel 3D culture model of the corneal stroma and apply it to investigate how key wound-healing growth factors regulate the mechanics of corneal keratocyte migration.

Methods.: Rabbit corneal keratocytes were seeded within collagen matrices that were compacted using external compression. Six-millimeter-diameter buttons were then incubated in media supplemented with 10% FBS, TGFβ1, TGFβ2, platelet-derived growth factor (PDGF), or no growth factor (control). After 1, 3, or 7 days, matrices were labeled with phalloidin and a nucleic acid dye, and were imaged using laser confocal microscopy. To study cell migration, buttons were nested within acellular uncompressed outer collagen matrices before growth factor stimulation.

Results.: Corneal keratocytes in basal media within compressed matrices had a broad, convoluted cell body and thin dendritic processes. In contrast, cells in 10% FBS developed a bipolar fibroblastic morphology. Treatment with TGFβ induced the formation of stress fibers expressing α-smooth muscle actin, suggesting myofibroblast transformation. PDGF induced keratocyte elongation without inducing stress fiber formation. Both 10% FBS and PDGF stimulated significant keratocyte migration through the uncompressed outer matrix, but 10% FBS produced more cell-induced collagen matrix reorganization. TGFβ induced the smallest increase in migration and the greatest matrix reorganization.

Conclusions.: Corneal keratocytes are able to differentiate normally and respond to growth factors within compressed collagen matrices, which provide a high-stiffness, 3D environment, similar to native stromal tissue. In addition, nesting these matrices provides a unique platform for investigating the mechanics of keratocyte migration after exposure to specific wound-healing cytokines.

Stromal keratocytes play a central role in mediating corneal response to injury or refractive surgery. ¹ In the corneal stroma, quiescent keratocytes normally have a dendritic morphology and a cortical distribution of F-actin. ² During wound healing, quiescent corneal keratocytes surrounding the injured region differentiate into fibroblast or myofibroblast phenotypes that mediate cell migration, wound contraction, and matrix remodeling ^3,4 ; these processes have a decisive impact on a patient's corneal clarity and refractive outcome. ^3,5,6 Numerous cell culture studies have identified several key growth factors that play important roles in mediating keratocyte differentiation into specific wound-healing phenotypes. ^7–20 However, these studies have been performed primarily using 2D substrates, and keratocytes reside within a complex 3D extracellular matrix in vivo. Significant differences in cell morphology, adhesion organization, and mechanical behavior have been identified between 2D and 3D culture models, with 3D models generally more closely mimicking in vivo cell behavior. ^21–30 Three-dimensional models also allow investigation of cell-induced matrix degradation and remodeling, key events in corneal wound healing. Thus, an in vitro model that mimics the 3D geometry and mechanical properties of the corneal stroma would be an important new platform for investigating key aspects of the wound-healing process. 

Because collagen is the main ECM component of most tissues, many in vitro models of various tissues involve the use of hydrated 3D fibrillar collagen matrices. The low collagen density and high compliance of these matrices are ideal for assessing how subcellular force generation and cell-induced matrix reorganization are regulated ^31–41 ; however, unlike native corneal stromal tissue, they undergo significant deformation in response to cellular forces, which limits their use for corneal tissue engineering. To overcome this limitation, some investigators have used glutaraldehyde or other chemical cross-linking techniques to increase the rigidity of hydrated collagen matrices. However, these approaches have several limitations: (1) they are often toxic and can only be applied to acellular matrices; (2) repopulating the acellular matrices after cross-linking can only be accomplished using corneal fibroblasts (not quiescent keratocytes); (3) the collagen concentration of these matrices is much lower than the collagen density of the cornea; and (4) they do not allow normal cell-based scaffold remodeling. ^42–45  

Another approach that has been used to generate high-stiffness tissue equivalents in vitro is to plate fibroblasts within standard fibrillar collagen matrices and then stimulate the cells to compact the matrix by contractile force generation. ⁴⁶ This process produces a stiffer matrix with a high density of cells and collagen fibrils (a tissue equivalent). This approach has been used to generate artificial stromas for corneal tissue engineering. ^47,48 However, an important limitation of these models is that the cells have to be activated to pre-contract the collagen matrix, which may alter their subsequent growth factor responses and migratory mechanics. Similarly, collagen sponges and vitrigel collagen membranes do not support the invasion and differentiation of quiescent corneal keratocytes. ^49,50  

More recently, a self-assembly approach has been used to generate more rigid 3D collagen matrices. In this model, corneal fibroblasts are cultured for several weeks in media containing a high concentration of ascorbic acid, which results in the generation of layered cell/matrix constructs that appear to have structural similarity to the in vivo cornea. This elegant approach has the potential to provide important insights into the developmental organization of the corneal stroma. ^51,52 However, it takes several weeks of culture to produce an organized ECM using this method, and these matrices cannot yet be generated using quiescent corneal keratocytes. To study the phenotypic transformation and migratory response of quiescent (i.e., noncontractile) corneal keratocytes to specific wound-healing cytokines and other signals, a model is still needed that does not require the use of activated corneal fibroblasts for its generation. 

Recently, Brown et al. ^53–55 developed a novel approach to dermal and bone tissue engineering based on plastic compression of fibrillar collagen matrices. In this model, standard cell-seeded, hydrated collagen matrices are polymerized. The cell-matrix construct is then immediately compacted by placing a stainless steel block on top and allowing fluid to flow out of the matrix into a paper blotting layer underneath. This process results in the formation of a flat, cell/collagen sheet. Because the excess fluid present in collagen gels is a result of casting and not of an inherent swelling property, the matrix maintains its compressed shape after removing the load (hence, the term plastic compression). These matrices have several unique properties that may be well suited for corneal stromal tissue engineering. In this study, we investigated the behavior of corneal keratocytes within these matrices. We first established that corneal keratocytes survived the matrix compression procedure and appeared to differentiate normally and to respond to growth factors within these matrices. We then used this model to investigate the effects of key wound-healing growth factors on corneal keratocyte migration using a “nested” 3D matrix approach. 

Methods

Cell Culture

Corneal keratocytes (NRK cells) were isolated from rabbit eyes obtained from Pel-Freez (Rogers, AR), as previously described. ³⁷ Cells were cultured in tissue culture flasks with basal medium consisting of Dulbecco's modified Eagle's medium (DMEM; Invitrogen, Carlsbad, CA) supplemented with 1% RPMI vitamin mix (Sigma-Aldrich, St. Louis, MO), 100 μM nonessential amino acids (Invitrogen), 100 μg/mL ascorbic acid, and 1% penicillin/streptomycin amphotericin B (Fungizone; BioWhittaker, Inc., Walkersville, MD) to maintain the keratocyte phenotype. ¹⁵  

Preparation of Compressed Collagen Matrices

Rectangular collagen constructs were prepared as described previously. ^56,57 Briefly, 0.6 mL of 10× DMEM was added to a 5 mL mixture of collagen (Type I Rat Tail; BD Biosciences, San Jose, CA) and acetic acid to achieve a final collagen concentration of 2.5 mg/mL. After dropwise neutralization with 1 M sodium hydroxide, a suspension of 6 × 10⁴ (low-density) or 2 × 10⁶ (high-density) keratocytes in 0.6 mL basal media was added to the collagen mixture. Solution containing the cells and the collagen was poured into a 3.2 × 2 × 1-cm well and allowed to set for 30 minutes at 37°C (Fig. 1A). Matrices were then compacted by a combination of compression and blotting using layers of mesh and paper sheets, as previously described. ^53,58 Briefly, a 165-μm-thick stainless steel mesh (∼300 μm mesh size) and a layer of nylon mesh (∼50 μm mesh size) were placed on a double layer of absorbent paper. The constructs were placed on the nylon mesh, covered with a second nylon mesh, and loaded with a 130-g stainless steel block for 5 minutes at room temperature. This process squeezes media out of the matrix and results in the formation of a flat, cell/collagen sheet (∼150-μm thick) with high mechanical stiffness (Fig. 1B). To determine whether the matrices swell or decompress in culture media, we measured the thickness of acellular compressed matrices 0 and 24 hours after compression. We found a decrease in matrix thickness of only 2.7% ± 2.2% after 24 hours, suggesting that the structure of the compressed collagen matrices remains relatively stable in culture media. 

Figure 1.

 

(A, B) Schematic of the process for constructing compressed matrices. Cells were seeded in rat tail collagen and compressed. (C) For migration studies, a 6-mm button was punched out and placed inside an acellular uncompressed collagen matrix.

After compression, 6-mm-diameter buttons were punched out of the compressed matrix and incubated in basal media for 24 hours to allow cell spreading. Media were then replaced with basal media supplemented with either 10% FBS, TGFβ1 (10 ng/mL), TGFβ2 (10 ng/mL), platelet-derived growth factor (PDGF; 50 ng/mL), or no growth factor (control) for an additional 3 days. Experiments were performed in triplicate (three matrices per condition) and were repeated at least once. 

Preparation of Nested Collagen Matrices

To study cell migration, nested matrices were used. Compressed matrices (high cell density) were prepared as described. Six-millimeter buttons were punched out and nested within acellular uncompressed outer collagen matrices (Fig. 1C). Briefly, hydrated acellular collagen matrices were prepared by mixing neutralized bovine dermal collagen (Purecol; Inamed Corp., Fremont, CA) with 10× DMEM to achieve a final collagen concentration of 2.5 mg/mL. ³⁶ After adjusting the pH, the 100-μL collagen solution was poured onto culture dishes (Bioptechs, Inc., Butler, PA). Bovine collagen was used to allow optimal imaging of matrix organization using confocal reflection microscopy. Then the button was pushed into the middle of the collagen matrices, placed in a humidified incubator for 60 minutes for polymerization, and overlaid with 1.5 mL basal media. After 24 hours, media were replaced with either 10% FBS, TGFβ1 (10 ng/mL), TGFβ2 (10 ng/mL), PDGF (50 ng/mL), or no growth factor (control) in basal media for 3 to 7 days. Experiments were performed in triplicate and repeated twice (n = 3). 

Assessment of Cell Viability

To determine the effects of matrix compression on cell viability, a viability/cytotoxicity assay kit (Live/Dead; Invitrogen) was used. This assay is based on the simultaneous determination of live and dead cells with two probes that measure recognized parameters of cell viability—intracellular esterase activity and plasma membrane integrity. The polyanionic dye calcein is retained within live cells, producing an intense uniform green fluorescence in live cells. Ethidium homodimer enters cells with damaged membranes and undergoes a 40-fold enhancement of fluorescence on binding to nucleic acids, thereby producing a bright red fluorescence in dead cells. After preparing the compressed matrix, assay reagents were added. After incubation for 30 to 45 minutes at room temperature, labeled cells were imaged under a fluorescence microscope. 

F-Actin Labeling and DNA Staining

At the end of each experiment (3 or 7 days), constructs were fixed using 3% paraformaldehyde in phosphate buffer for 10 minutes and permeabilized with 0.5% Triton X-100 in phosphate buffer for 3 minutes. Cells were labeled with Alexa-Fluor 546 phalloidin (1:50; Molecular Probes, Eugene, OR) for 1 hour and then washed in phosphate-buffered saline (PBS; 3 times for 5 minutes). A nucleic acid dye (TOTO-3; 1:200; Molecular Probes, Eugene, OR) was then added to each construct to stain the cell nuclei. Constructs were then incubated for 15 minutes and washed with PBS (3 times for 5 minutes). 

Laser Confocal Microscopy

After labeling F-actin and nuclei, fluorescent light (for F-actin and nuclei) and reflected light (for collagen fibrils) 3D optical section images were acquired using laser confocal microscopy (SP2; Leica, Heidelberg, Germany). An HeNe laser (633 nm) was used for reflected light and excitation of a specific nucleic acid dye (TOTO-3; 1:200; Molecular Probes), and a GreNe laser (543 nm) was used for fluorescent imaging of F-actin. A stack of optical sections (z-series) was acquired for each cell imaged by changing the position of the focal plane in 10-μm steps using a 20× objective (nonimmersion; 0.7 NA; 590 μm free working distance) or 1-μm steps using a 63× objective (water immersion; 1.2 NA; 220 μm free working distance). 

Cell Morphology

Changes in cell morphology were measured using a microscope automation and image analysis package (MetaMorph; Molecular Devices, Sunnyvale, CA). Projected cell length and breadth were calculated by outlining the maximum intensity projection image of a cell (generated from the F-actin z-series), thresholding, and applying the integrated morphometry analysis (IMA) routine. The length is calculated by IMA as the span of the longest chord through the object; the breadth is the caliper width of the object, perpendicular to the longest chord. The height of cells was calculated by measuring the distance between the first and last planes in the z-series in which a portion of the cell was visible. 

Assessment of Cell Migration

Maximum intensity projection images of cell nuclei image stacks were created with the image analysis package (MetaMorph; Molecular Devices) and overlaid with reflected images. Graphics editing software (Photoshop; Adobe, Mountain View, CA) was then used to align these overlapped images, resulting in a 750-μm-wide montage image for each quadrant (Fig. 2). Each montage included the border of the inner matrix (detected by reflected light) and the farthest moving cells. The distance the cells traveled was calculated by drawing a straight line between the interface and the leading-edge cells; an average of 10 cells was used for each montage. The number of cells per 750-μm-wide montage was also counted. 

Figure 2.

 

Schematic showing the montage of 3D image stacks collected in each sample using laser scanning confocal microscopy. Reproduced with permission from: An experimental model for assessing fibroblast migration in 3-D collagen matrices. Karamichos D, Lakshman N, Petroll WM. Cell Motil Cytoskeleton. 66(1):1–9, Copyright © 2009 Wiley-Liss, Inc., A Wiley Company.

Statistical Analysis

All statistical analyses were performed with statistical software (SigmaStat version 3.11; Systat Software Inc., Chicago, IL). One-way ANOVA was used to compare group means. Post hoc multiple comparisons between groups were performed using the Holm-Sidak method. Differences were considered significant if P < 0.05. 

Results

Keratocyte Viability Following Matrix Compression

To first evaluate the viability of keratocytes after matrix compression, two-color fluorescence live-dead staining was performed (Live/Dead; Invitrogen). This probe showed an intense uniform green fluorescence in live cells and a bright red fluorescence in the nuclei of dead cells. As seen in Figure 3, there was no significant difference in the percentage of dead cells in uncompressed and compressed matrices 4 hours after plating (5.41 ± 0.30 vs. 4.50 ± 0.27; n = 3 matrices), which demonstrates that matrix compression did not affect cell viability. 

Figure 3.

 

Viability/cytotoxicity assay staining. There was no significant difference in the percentage of dead cells (5.41 ± 0.30 vs. 4.50 ± 0.27; n = 3 matrices) after compression, which demonstrates that matrix compression did not affect cell viability.

Keratocyte Response to Growth Factors within Compressed Matrices

To assess differentiation and growth factor responses, media were supplemented with either 10% FBS, TGFβ1 (10 ng/mL), TGFβ2 (10 ng/mL), PDGF (50 ng/mL), or no growth factor (control) after 24 hours in basal media and were cultured for an additional 3 days. Growth factor concentrations were determined from preliminary dose-response experiments and represent the lowest concentration to give a maximal effect. NRK cells maintained in basal media within compressed matrices had a broad, convoluted cell body with numerous thin dendritic processes. These processes extended in all directions from the cell body. F-actin was generally limited to the cell cortex, and stress fibers were rarely observed (Fig. 4A). This morphology and cytoskeletal organization is consistent with that of quiescent corneal keratocytes in vivo. ² NRK cells incubated in 10% FBS had a bipolar morphology with pseudopodial processes (Fig. 4B) and did not form broad lamellipodia. Parallel arrays of microfilament bundles (stress fibers) were often observed within the cell body and pseudopodial processes. These results are consistent with previous observations of serum-cultured corneal fibroblasts within 3D uncompressed collagen matrices in vitro ³⁹ and wound-healing fibroblasts in vivo. ^1,3 Cells treated with PDGF BB, which activates Rac, were more elongated and developed a less convoluted cell morphology without inducing stress fiber formation (Fig. 4C), consistent with previous results in uncompressed matrices. ^35,59 When cultured keratocytes were treated with TGFβ1 or TGFβ2, they developed prominent F-actin filament bundles (stress fibers) and lost thin cell processes (Figs. 4D, E), suggesting transformation from a dendritic phenotype to a contractile phenotype. In addition, stress fibers positive for α-smooth muscle actin were observed, consistent with myofibroblast differentiation (Figs. 4F, G). ⁶⁰  

Figure 4.

 

(A–E) Maximum intensity projections of F-actin labeling. (A) F-actin was limited to the cell cortex, and stress fibers were rarely observed in basal media. (B) Keratocytes incubated in 10% FBS developed a bipolar morphology. (C) Cells treated with PDGF appeared more elongated and developed less convoluted cell morphology, without inducing stress fiber formation. (D, E) When cultured keratocytes were treated with TGFβ1 or TGFβ2, they developed prominent F-actin filament bundles (stress fibers) and lost thin cell processes. (F, G) Double-labeling of F-actin and α-smooth muscle actin in a cell treated with TGFβ1. Stress fibers positive for α-smooth muscle actin were observed, consistent with myofibroblast differentiation.

At higher cell density, keratocytes in compressed 3D matrices formed an interconnected network that resembled their organization in vivo. Cell-cell interactions appeared to be mediated by dendritic processes connecting adjacent keratocytes (Fig. 5A). Side views of these matrices (Fig. 5B) revealed the parallel alignment of keratocyte cell bodies, consistent with keratocyte organization in vivo. ² Because little or no cell spreading occurs during the 30-minute incubation period before matrix compression, this cellular alignment is likely a response to compressed matrix geometry, collagen orientation, or both. Note that some slight undulations are observed in the construct when viewed in cross-section (Fig. 5B). 

Figure 5.

 

(A) Maximum intensity projection of F-actin at high keratocyte density shows interconnected cell processes. (B) Side view (90-μm-thick x-z projection) of F-actin reveals the parallel alignment of keratocyte cell bodies, consistent with their organization in vivo.

Quantitative analysis of cell morphology demonstrated that PDGF induced a statistically significant increase in cell length compared with other conditions (Fig. 6A). Cell height was largest after culture in either PDGF or basal media (Fig. 6B). To further assess the differences in cell shape, the length/breadth ratio was compared. This ratio was significantly higher after culture in 10% FBS, confirming a more bipolar morphology (Fig. 6C). Overall, these results confirmed that corneal keratocytes are able to differentiate and respond to growth factors normally within compressed collagen matrices. 

Figure 6.

 

Quantitative analysis of cell morphology in 3D culture. (A) PDGF induced a statistically significant increase in cell length compared with other conditions. (B) Cell height was significantly increased after culture PDGF or basal media. (C) The length/breadth ratio was significantly higher in 10% FBS, confirming a more polarized morphology. A minimum of 11 cells were analyzed for each condition. Mean ± SD are shown. *P < 0.05.

Growth Factor Stimulation of Keratocyte Migration through Collagen Matrix

Experiments were next carried out to study cell migration by using a previously described nesting approach. ^61,62 Six-millimeter-diameter buttons were punched from compressed matrices, nested within acellular uncompressed outer collagen matrices (2.5 mg/mL), and overlaid with basal media. After 24 hours, media were supplemented with 10% FBS, PDGF (50 ng/mL), TGFβ1 (10 ng/mL), TGFβ2 (10 ng/mL), or no growth factor (control) and cultured for up to 7 days. 

Corneal keratocytes migrated from the compressed inner matrix into the uncompressed outer matrix, presumably because of the cell density gradient. Figure 7 shows 750-μm-wide montages of migrating cells after 7 days. Cells were observed in the outer matrix under all conditions, but the number of cells and the distance they traveled varied substantially. To quantify the migration response, the average distance cells traveled was calculated by measuring the distance from the interface of inner-outer matrix to the cells at the leading edge (Fig. 8A). Cells cultured in 10% FBS or PDGF traveled significantly further than those in basal media, TGFβ1, or TGFβ2 (P < 0.05). We also counted the number of cells in the outer matrix after 7 days of migration (Fig. 8B). The largest number of cells in the outer matrix was observed after culture in 10% FBS. Similar results were observed at 3 and 5 days, but the overall number of cells was reduced (not shown). 

Figure 7.

 

Maximum intensity projection images collected after 7 days of culture in (A) basal media, (B) 10% FBS, (C) PDGF, (D) TGFβ1, and (E) TGFβ2. Corneal keratocytes in basal media showed minimal migration into outer matrix. Cells cultured in 10% FBS or PDGF traveled significantly further than those in basal media, TGFβ1, or TGFβ2. The largest number of cells in the outer matrix was observed after culture in 10% FBS.

Figure 8.

 

(A) Quantitative analysis of the distance cells traveled into the outer matrix and (B) average number of cells in outer matrix under different culture conditions at 7 days. (A) Cells in 10% FBS or PDGF traveled significantly further than in other conditions. (B) Culture in 10% FBS resulted in the highest number of cells in the outer matrix. Graphs show mean ± SD of three experiments performed in triplicate. *P < 0.05.

Differential Regulation of Matrix Reorganization during Migration

To assess the local pattern of cell-induced matrix reorganization during migration, confocal reflection microscopy was performed. Reflected light imaging allows detailed visualization of the cells and the fibrillar collagen surrounding them.^41,63,64 Cell-matrix interactions can best be appreciated in movies showing maximum intensity projection images over a range of projection angles (Movies S1–S3). Migrating cells in basal media maintained a stellate morphology with dendritic processes, and collagen fibrils appeared to be randomly oriented without compaction or alignment (Fig. 9A; Movie S1). After culture in 10% FBS, cells developed a bipolar morphology with occasional stress fibers along the along the cell body (Fig. 9B; Movie S2). Collagen fibrils were compacted and aligned parallel to the long axis of pseudopodia, and there was often an enhancement of cell-induced collagen reorganization between the migrating cells. Cells cultured in TGFβ developed a broad morphology, and stress fibers were observed within the cell. Collagen fibrils were compacted around the cells and between the cells (Fig. 9C). Migrating cells in PDGF appeared more elongated with branching processes. Collagen fibrils remained more randomly aligned around the cells, and there was less ECM compaction than with FBS or TGFβ (Fig. 9D; Movie S3). 

Figure 9.

 

Maximum intensity projections of F-actin (green) and collagen fibrils (red) at the interface between the inner and outer matrices. (A) Migrating cells in basal media maintained a stellate morphology with dendritic processes, and collagen fibrils appeared to be randomly oriented without significant compaction or alignment. (B) After culture in 10% FBS, migrating cells developed a bipolar morphology with occasional stress fibers along the cell body. Collagen fibrils were compacted and aligned parallel to the long axes of pseudopodia. (C) After culture in TGFβ2, cells developed a broad morphology, and intracellular stress fibers were observed. Collagen fibrils were compacted both around and between the cells. (D) Migrating cells in PDGF appeared more elongated with branching processes. Collagen fibrils remained more randomly aligned around the cells.

The pattern of cell-induced matrix reorganization at the leading edge of the migration front, where the impact of individual cells on the ECM could be more easily visualized, was also assessed. After culture in basal media or PDGF, collagen fibrils were randomly oriented, and alignment or compaction of fibrils at the ends of cells were not generally observed (Figs. 10A, D). In contrast, collagen fibrils at the ends of cells appeared to be compacted and aligned parallel to the long axis of pseudopodia in 10% FBS, TGFβ1, and TGFβ2 (Figs. 10B-D). Overall, cell morphology was similar to that of the cells at the interface. 

Figure 10.

 

Color overlays of F-actin (green) and collagen fibrils (red) at the leading edge of the migration front after 7 days of culture. (A, E) After culture in basal media and PDGF, collagen fibrils were randomly oriented, and alignment or compaction of fibrils at the end of cells was not generally observed. (B–D) Collagen fibrils at the ends of cells appeared to be compacted and aligned parallel to the long axes of pseudopodia in 10% FBS, TGFβ1, and TGFβ2.

Discussion

Compressed collagen matrices have several unique properties that make them an attractive candidate for corneal stromal tissue engineering. Ultrastructural analysis of compressed collagen matrices demonstrates the formation of parallel collagen lamellae in the x-y plane throughout the depth of the compressed collagen sheet. ⁵³ The thickness of the sheet can be controlled by changing the initial collagen volume, and sheets from 30 μm to 150 μm have been routinely generated. ^53,61 Collagen can be partially aligned along a particular axis within the sheets by prestretching them, and the sheets can be stacked to generate layered constructs. Standard compressed matrices have a type I collagen concentration and mechanical stiffness (1MPa) similar to those of native corneal tissue. ⁵³ Finally, these matrices can be generated rapidly, and the cell density and geometry can be modulated during preparation. ⁵⁸ Most important, because the collagen is compacted by using external compression, cells are not required to be contractile or activated at the start of the experiment. 

In the first part of this study, we assessed the differentiation of quiescent corneal keratocytes when plated within compressed collagen matrix and their response to specific growth factors. Our results show that under basal media conditions, keratocytes survive matrix compression and develop the quiescent (i.e., noncontractile) phenotype normally observed in vivo, as indicated by a cortical F-actin organization and dendritic morphology. ² Exposure to serum leads to fibroblast differentiation, as indicated by the assumption of a bipolar morphology and the formation of intracellular stress fibers. PDGF BB stimulated cell spreading with the formation of numerous dendritic processes but did not induce stress fiber formation. PDGF BB is expressed in tear fluid and has been shown to increase Rac activity and to induce cell spreading in both dermal and corneal fibroblasts. ^35,59,65 In contrast, PDGF AB has been shown to induce fibroblastic transformation of keratocytes on 2D substrates. ¹⁵ When treated with TGFβ, cells developed characteristics of myofibroblasts such as a spread morphology, loss of thin cell processes, and development of prominent stress fibers positive for α-smooth muscle actin. Overall, these data suggest that corneal keratocytes within compressed collagen matrices are able to differentiate normally and to respond as expected to key growth factors expressed during wound healing. ^{9–11,60,66,67}  

We next used the compressed matrix model to investigate the migratory response of corneal keratocytes. For these experiments, 6-mm buttons from compressed matrices were embedded within acellular uncompressed collagen matrices; this is a modification of previously published nesting techniques. ^61,62,68 A unique feature of this approach is that the degree of matrix remodeling produced during cell migration can be assessed. Most important, unlike other 3D migration models, quiescent (noncontractile) cells can be used in the inner matrix. In this study, quiescent keratocytes were allowed to spread for 24 hours in basal media before subsequent culture in 10% FBS, PDGF, or TGFβ. The number of cells in the outer matrix after 7 days was significantly higher in 10% FBS compared with all other conditions studied. These cell counts in the outer matrix likely reflect a combination of both migration from the inner matrix and cell proliferation, particularly under 10% FBS. However, a similar pattern was observed at 3 days, when proliferation would be expected to have less of an impact on the cell numbers. Repopulation of corneal wounds in vivo also involves a combination of migration and proliferation. To better isolate the effects of different treatments on cell translocation, we measured the distance from the inner matrix to the leading edge of the migrating front. Cells in 10% FBS and PDGF migrated a similar distance into the outer matrix; this was significantly greater than cells in either basal media or TGFβ. 

It is widely accepted that the major events of cell migration are extension of a leading edge, establishment of new adhesion sites at the front, cell body contraction, and detachment of adhesions at the cell rear. ⁶⁹ For a cell to migrate, at least two distinct types of force must be generated independently by migrating cells. ⁷⁰ The first is the protrusive force needed to extend lamellipodia or filopodia at the leading edge, and the second is the tractional force needed to overcome the adhesive interactions between the cell and matrix to pull the cell forward. The small GTPases Rho and Rac are prime candidates for regulating the cytoskeletal and mechanical phenotype of fibroblasts during cell migration. In dermal fibroblasts, activation of Rac by PDGF enhances cell spreading and migration within 3D collagen matrices. ^62,71 In contrast, the activation of Rho increases cell contractility through the activation of Rho kinase. ^35,72–77 Although both PDGF and serum stimulated keratocyte migration in this study, the magnitudes of cellular force generation were strikingly different. PDGF stimulated cell migration without the generation of large forces, as indicated by a lack of stress fibers and minimal cell-induced ECM reorganization. In contrast, keratocytes in 10% FBS induced significant compaction and alignment of the ECM during migration. Serum contains factors such as lysophosphatidic acid (LPA) and sphingosine-1-phosphate which activate the Rho/Rho-kinase pathway. ^35,78,79 Interestingly, TGFβ, which also induces significant Rho/Rho-kinase–dependent force generation, ^80–82 produced the largest amount of ECM reorganization and the smallest increase in migration. 

Taken together, the data suggest that the impact corneal keratocytes have on ECM architecture during migration may ultimately depend on the balance of Rho and Rac activation induced by the cytokines to which they are exposed. We previously demonstrated that though most of the large tractional forces generated by corneal fibroblasts are Rho-kinase/myosin II dependent, smaller forces can be generated at the tips of extending pseudopodia that are not produced by actomyosin contraction. ⁵⁹ Thus, it is interesting to speculate that in the present study, keratocyte migration in 10% FBS was mediated primarily by Rho-kinase–dependent actomyosin contractile force generation, whereas migration in PDGF was more dependent on other mechanisms, such as the protrusive forces associated with actin polymerization. ^83–87 Future studies using time-lapse imaging to directly assess dynamic cell-matrix interactions could help to distinguish these underlying mechanisms of cellular force generation. ^59,88  

We previously used a similar nested model to investigate cell migration and matrix remodeling using a corneal fibroblast cell line (HTK cells). ⁶¹ These activated cells migrated into the outer matrix in much higher numbers than the quiescent corneal keratocytes used in the present study. Furthermore, HTK cells cultured in PDGF traveled significantly further than those cultured in 10% FBS. It should be noted that there is a significant basal level of Rho-kinase activity in this corneal fibroblast cell line ⁵⁹ that would not be expected in quiescent corneal keratocytes. ¹⁵ This low level of Rho-kinase activation may enhance cell migration in response to PDGF because of a more optimal balance between contractile and protrusive forces. A similar response is observed when dermal fibroblasts are exposed to PDGF. ⁶² One of the key features of the model used in the present study is that quiescent corneal keratocytes can be used in the inner compressed matrices. Thus the effects of specific growth factors and their downstream signaling pathways can potentially be assessed without preexposure to serum or other factors, which may permanently alter these responses. ¹⁵  

In conclusion, our results indicate that corneal keratocytes are able to differentiate normally and to respond to growth factors within compressed collagen matrices, which provide a high-stiffness, 3D environment similar to native stromal tissue. In addition, nesting these matrices provides a novel platform for investigating the mechanics of corneal keratocyte migration. These initial studies demonstrate that the impact migrating keratocytes have on ECM architecture can vary dramatically in response to different growth factors. This mechanical plasticity may ultimately make it possible to modulate this important aspect of cell behavior during in vivo wound healing. 

Supplementary Materials

Movie S1 - 3.0 MB (QuickTime Movie) 

Maximum intensity projection images of f-actin (green) and collagen fibrils (red) constructed over a range of projection angles. Constructs were cultured is basal media for seven days. Keratocytes migrate out of the inner compressed matrix (left) without inducing significant collagen reorganization. 

Movie S2 - 3.9 MB (QuickTime Movie) 

Maximum intensity projection images of f-actin (green) and collagen fibrils (red) constructed over a range of projection angles. Constructs were cultured is 10% FBS for seven days. Keratocytes in the outer compressed matrix (right) appear fibroblastic, and induce significant compaction and alignment of the ECM. 

Movie S3 - 3.9 MB (QuickTime Movie) 

Maximum intensity projection images of f-actin (green) and collagen fibrils (red) constructed over a range of projection angles. Constructs were cultured is PDGF BB for seven days. Keratocytes in the outer compressed matrix (right) appear elongated, but do not induce significant compaction and alignment of the ECM. 

Footnotes

 Supported in part by National Institutes of Health Grants R01 EY 013322 and R24 EY016664 and by an unrestricted grant and Senior Scientific Investigator Award (WMP) from Research to Prevent Blindness, Inc.

Footnotes

 Disclosure: A. Kim, None; N. Lakshman, None; D. Karamichos, None; W.M. Petroll, None