Morphologic Characterization of Organized Extracellular Matrix Deposition by Ascorbic Acid–Stimulated Human Corneal Fibroblasts

Abstract

purpose. To characterize the structure and morphology of extracellular matrix (ECM) synthesized by untransformed, cultured human corneal fibroblasts in long-term cultures.

methods. Human corneal stromal keratocytes were expanded in transwell culture in the presence of fetal bovine serum and a stable derivative of vitamin C. The cells were allowed to synthesize a fibrillar ECM for up to 5 weeks. Constructs were assessed by light (phase-contrast and differential interference-contrast) and transmission (standard and quick freeze/deep etch) microscopy.

results. Electron micrographs revealed stratified constructs with multiple parallel layers of cells and an extracellular matrix comprising parallel arrays of small, polydisperse fibrils (27–51 nm) that often alternate in direction. Differential interference contrast images demonstrated oriented ECM fibril arrays parallel to the plane of the construct, whereas quick-freeze, deep-etch micrographs showed the details of the matrix interaction with fibroblasts through arrays of membrane surface structures.

conclusions. Human keratocytes, cultured in a stable vitamin C derivative, are capable of assembling extracellular matrix, which comprises parallel arrays of ECM fibrils. The resultant constructs, which are highly cellular, are morphologically similar to the developing mammalian stroma, where organized matrix is derived. The appearance of arrays of structures on the cell membranes suggests a role in the local organization of synthesized ECM. This model could provide critical insight into the fundamental processes that govern the genesis of organized connective tissues such as the cornea and may provide a scaffolding suitable for tissue engineering a biomimetic stroma.

Afundamental characteristic of load-bearing collagenous tissue is the high degree of anisotropy in the arrangement of the fibrils. This anisotropy is necessary if the tissue is to support significant loads because collagen fibrils are designed to transmit forces unidirectionally. Understanding how an organized anisotropic collagenous matrix can arise from a population of synthetically active fibroblastic cells remains one of the most important basic science objectives associated with connective tissue research. ^{1 2} Such an endeavor would also support efforts aimed at producing load-bearing constructs for tissue engineering. For the corneal stroma, which has exquisite organization on the nanoscale to allow the transmission of incident light ^{3 4 5} and to resist intraocular pressure, the processes that direct matrix synthesis and organization are not well understood. 

During corneal development, the collagen fibrils are assembled, aligned, and arranged into lamellae of alternating orientation. ^{6 7} How the cells exert control over the natural tendency of assembling collagen fibrils to form random arrays is a matter of conjecture. There appear to be two competing hypotheses concerning collagen fibril assembly. One is that the fibroblastic cells form “surface crypts” within which procollagen is assembled into fibrils and “vectorially” discharged. ^{8 9 10} The other is that the fibroblasts work in concert to produce highly concentrated monomer solutions that gain order through cholesteric effects. ^{11 12}  

Studies describing the organizational changes as a function of time during development have been derived primarily from the study of the embryonic chick ⁷ or rabbit ⁶ cornea. These investigations, though instrumental to our understanding of the development of organized stromal matrix, have limited temporal resolution. The development of an in vitro matrix synthesis model of corneal stromal development in a system that will allow direct observation could provide insight into how organized tissue is produced. 

Tissue engineers have been developing culture systems for use as connective tissue replacements for almost two decades ¹³ (for a review, see Langer and Vacanti ¹⁴ ). The most successful tissue cultured in vitro is skin, which greatly enhances the repertoire of tools available to the clinician. ¹⁵ The success of skin organotypic culture has stimulated tissue engineers to produce corneal constructs from the cells of a variety of different species, including rabbit, bovine, pig, and human. ^{16 17 18 19 20 21} In general, the methods to produce these constructs are focused on the techniques necessary to successfully cultivate the cellular layers onto and within a randomly oriented, self-assembled, collagen-based gel. Tissue-engineered corneas produced in this way have lacked the mechanical strength and transparency required for clinical application. ²² A culture system that induces synthesis of an in vivo-like stromal matrix could provide tissue engineers with a mechanically strong, transparent scaffolding onto which epithelial and endothelial layers could be seeded to produce a functional, organotypic cornea. 

Although vitamin C has been shown to increase the proliferative rate of cultured fibroblasts and to stimulate the synthesis and secretion of appropriate ECM components, ^{23 24 25 26 27} it is not stable for long periods of time when exposed to culture conditions. ^{25 28} In one investigation, primary human corneal fibroblasts were grown in short-term culture in the presence of vitamin C. ²⁹ The fibroblasts produced heterotypic collagenous fibrils containing type I/V monomers and type VI collagen, which are present in normal human corneal stroma. Recently, more stable derivatives of ascorbic acid, L-ascorbic acid 2 phosphate (P-Asc) and 2-O-α-d-glucopyranosyl-L-ascorbic acid (G-Asc), collectively referred to as S-Asc (stabilized ascorbic acid), have been shown to enhance the synthesis and secretion of collagen by fibroblasts in culture. ^{24 25 26 30 31}  

In the current investigation, a new long-term culture system comprising untransformed cultured human corneal fibroblasts was developed to induce the synthesis of an ECM de novo after stimulation with stabilized vitamin C. Our system uses primary human corneal keratocytes that have been converted to fibroblasts by continual exposure to fetal bovine serum. The fibroblasts are then stimulated by S-Asc to synthesize and assemble a supportive ECM. The organization of the secreted extracellular matrix was characterized by light microscopy (phase and differential interference contrast [DIC]) and transmission electron microscopy (standard and quick freeze/deep etch [QFDE]). Indirect immunofluorescence was used to determine whether type V or type VI collagen was synthesized. 

Materials and Methods

All procedures used in these studies adhered to the tenets of the Declaration of Helsinki. Human corneas were obtained from the National Disease Research Interchange (NDRI, Philadelphia, PA). 

Primary Culture of Human Corneal Cells

Corneal epithelium and endothelium were removed from the stroma by scraping with a razor blade. The stromal tissue was cut into small pieces and put into six-well plates (four or five pieces of 2 × 2 mm tissue per well). Explants were allowed to adhere to the bottom of the wells at room temperature for 5 minutes before Eagle minimum essential medium (EMEM; Sigma-Aldrich, St. Louis, MO) containing 10% fetal bovine serum (FBS; ATCC, Manassas, VA) was added. Each 500-mL bottle of medium was supplemented to 1% with antibiotic/antimycotic (Sigma). Care was taken when adding the medium to make sure that the explants remained attached to the wells. After 1 to 2 weeks of cultivation (37°C, 5% CO₂), the fibroblasts were passaged into a T75 flask. The cells were grown to confluence before their use in the culture system. 

Fibroblast-Assembled Extracellular Matrix

Human corneal fibroblasts (HCFs) were plated on transwell six-well plates with membrane inserts with 3.0-μm pores (Costar, Charlotte, NC) at a density of 0.5 × 10⁶ cells/well. Membranes were either polyester or polycarbonate; qualitative comparisons of constructs demonstrated no difference between the surfaces. Fibroblasts were cultured in EMEM with 10% FBS and 1 mM P-Asc or 0.5 mM G-Asc (Wako Chemicals, Richmond, VA). As expected, preliminary investigations demonstrated that the use of either form of ascorbic acid produced similar results with regard to ECM production (thus data using either ascorbic acid was pooled). Medium was changed 3 times a week. Fibroblast cultures were examined and imaged using phase-contrast light microscopy (Zeiss Axiovert 25; Carl Zeiss MicroImaging, Thornwood, NY) equipped with a digital camera (D70; Nikon, Tokyo, Japan). Except for a few constructs that were cryopreserved for QFDE electron microscopy at 3 and 5 weeks, the fibroblast cultures were allowed to generate constructs for 4 weeks. After 4 weeks in culture, the resultant constructs were processed using one or two of the following methods ¹ : fixed in Karnovsky fixative (2% paraformaldehyde, 2.5% glutaraldehyde in cacodylate buffer, pH 7.4) at 4°C for transmission electron microscopy (TEM) ² ; frozen in optimal cutting temperature compound (OCT; Sakura, Torrance, CA) for indirect immunofluorescence microscopy ³ ; slammed (rapidly frozen) for QFDE electron microscopy. 

Light Microscopy

The overall organization of the layering of the fibroblasts from day 1 to 4 weeks was monitored using phase-contrast microscopy. The HCFs were viewed under phase illumination with an inverted microscope (Axiovert 25; Zeiss) and imaged (D70 digital camera; Nikon). Differential interference contrast imaging (DIC) of cultures was performed to estimate the total construct thickness and to observe the orientation of synthesized fibrils relative to the HCF organization. ³² Constructs were gently removed from the underlying transwell membrane and placed in a drop of 1× phosphate-buffered saline on a glass coverslip. A second coverslip was placed over the drop and construct, and the whole preparation was placed on the stage of an inverted microscope (TE2000U; Nikon [Microvideo Instruments, Avon, MA]). Z-scans were used to find the bottom and top of each construct to determine thickness. Within the Z-scans, images were taken to demonstrate the general alignment of the extracellular matrix in the plane of the construct. 

Transmission Electron Microscopy

Constructs were fixed overnight in Karnovsky fixative at 4°C and were processed for TEM as described in Gipson et al. ³³ Briefly, constructs were fixed in half-strength Karnovsky fixative, rinsed in PBS, and processed through postfixation in 2% osmium tetroxide, en bloc staining in 0.5% uranyl acetate, and alcohol dehydration to propylene oxide and were embedded in Epon. Sixty-nanometer sections cut transverse to the plane of the construct with a diamond knife ultramicrotome (LKB, Bromma, Sweden) were viewed and photographed with a transmission electron microscope (410; Philips Electronics NV, Eindhoven, The Netherlands). 

Quick-Freeze/Deep Etch

A trephine was used to cut 3-mm specimens from the fibroblast constructs. These specimens, while still attached to the membrane, were rinsed with PBS to remove any excess medium and then were mounted tissue side up onto specimen carriers using a 2% clay solution (Laponite; Rockwood Additives, Cheshire, UK) as an adhesive and cushioning material. After specimens were mounted, excess PBS was removed with filter paper, and the exposed surface of the tissue was rapidly slam frozen using an electron microscopy workstation (EM CPC Cryoworkstation; Leica, Wetzlar, Germany). The frozen constructs were transferred to a freeze fracture/freeze etch system (CFE-60; Cressington Scientific Instrument, Watford, UK) for replication. During replication, the specimens were superficially fractured and etched at −100°C for 25 minutes. Rotary-shadowed replicas of the etched surfaces of the constructs were created by evaporation of platinum/carbon (for contrast) at a 20° angle onto the rotating construct, followed by evaporation of pure carbon (for replica strength) at a 90° angle. Tissue was digested overnight with household bleach. Clean replicas were picked up on copper 400-mesh grids. Grids were viewed and photographed with a transmission electron microscope (410; Philips). 

Fibril Diameter Estimation/Cell Volume Percentage Estimation

Fibril diameter was estimated directly from the standard TEM images using calibration grids and a graphics editing system (Photoshop; Adobe Systems, San Jose, CA) measurement tools. Fifty fibrils were selected randomly from multiple micrographs and were independent of position in the construct. For fibrils in cross-section, the smallest possible diametric line was chosen. For longitudinal fibrils, the thickest part of the fibril was measured. Data were reported as mean ± SD. 

Cell volume as a percentage of the construct volume was estimated from low-magnification transmission electron micrographs from four representative specimens. To perform the estimate, vertical lines were drawn from the top of the construct to the supporting membrane at three to four equally spaced points on each micrograph. The length of the line that intersected cellular material was divided by the total line length (construct thickness). Data were reported as mean ± SD percentage cellular volume. It is well known that standard TEM processing causes shrinkage in collagen and generally alters the volume of matrix and cells, ³⁴ and the data must be viewed with this caveat in mind. 

Quantitative and Qualitative Demonstration of Organization in Constructs

Quantitative descriptions of organization can be obtained by demonstrating that features of the construct are correlated in some way. For instance, totally random arrays of collagen fibrils would not exhibit correlation in the angle of their long axes with any particular direction. To quantitatively demonstrate that our constructs had an organized structure (i.e., different from a random structure), we examined transverse sections of TEM micrographs and measured the distribution of the angles of fibrils, cell bodies, and cell processes that were parallel to the plane of the section. Fifteen random TEM micrographs were examined. Each micrograph was divided into at least five regions. Within each region, the angle of longest continuous fibril (parallel to the plane of the section) was measured. The angles of the long axis of cells and cell processes in the cross-sections were also measured for comparison to the angles of the fibrils. Results of the examination are presented as differences (mean ± SD) between the orientation angle of the cells and the synthesized collagen fibrils. Qualitatively, organization of the lamellae in the constructs was demonstrated by selecting micrographs in which the collagen in the constructs clearly alternates in its alignment in adjacent lamellae. The presence of arrays of alternating structures, which cannot arise by random synthesis or self-assembly of collagen, should indicate that organizational control of fibrillogenesis is occurring in the cultures. 

Collagen Types V and VI Indirect Immunofluorescence

Collagen type V is thought to play a role in the nucleation and control of collagen fibril diameter ^{35 36 37 38 39} and is found in significant quantities in normal human corneas. To detect type V collagen, fibrils in the construct were first partially disrupted in acetic acid. ⁴⁰ Briefly, sections were washed in 0.1 M PBS three times each and incubated in 10 mM acetic acid followed by two washes in PBS. Disrupted sections were fixed in 4% paraformaldehyde for 15 minutes and washed in PBS three times. Sections were blocked with 5% normal goat serum (NGS) in PBS, incubated in 1:200 mouse anti-collagen type V (Chemicon International, Temecula, CA) in PBS + NGS for 2 hours at room temperature. Sections were then washed three times in PBS and incubated in 1:1000 goat anti-mouse Cy3 (Jackson ImmunoResearch, West Grove, PA) in PBS + NGS for 30 minutes at room temperature. Specimens were washed twice in PBS and coverslipped with mounting media containing DAPI, a marker for cell nuclei (Vectashield; Vector Laboratories, Burlingame, CA). Slides were viewed and imaged under a microscope (Eclipse E800; Nikon) equipped with a digital spot camera (MicroVideo Instruments). Where the primary antibody was omitted, negative controls were performed. 

Collagen type VI is a known component of the normal human corneal stroma. ⁴¹ To determine whether our constructs were producing collagen type VI, we used immunofluorescence methods. Specifically, sections were fixed in 4% paraformaldehyde for 15 minutes followed by three washes in PBS. Sections were transferred to 5% NGS in 0.1 M PBS for 30 minutes and exposed for 2 hours to 1:200 rabbit anti-collagen type VI (Calbiochem, EMD Biosciences, Inc., San Diego, CA) in PBS + NGS at room temperature. Sections were washed in PBS 3 times and incubated in goat anti-rabbit Cy3 (Jackson ImmunoResearch) in PBS + NGS for 30 minutes at room temperature. Sections were then washed two times and viewed as described. 

Results

Early Culture Cellular Organization

After 1 day, the HCFs formed a single layer of randomly oriented cells that exhibited a classic spindle shape (Fig. 1A) . By 1 week, the culture had stratified into multiple layers where the long axes of fibroblasts exhibit orientation changes in different strata. The phase-contrast micrograph (which constitutes an integrated Z-scan) of a 1-week-old construct demonstrated the change in direction of the long axes of cells at different elevations above the transwell membrane. In some locations it appeared that fibroblasts, which were at the same x-y position but at a different elevation, had long axes oriented at approximately 90° angles to one another (Fig. 1B) . Our result confirms those of Newsome et al., ⁴² who showed similar behavior in human corneal fibroblasts. 

Microscopic Organization and Morphology of Mature Cultures

By 4 weeks in culture, the HCFs stratified into multiple layers and produced significant quantities of ECM (Fig. 2) . The average construct total thickness at 4 weeks was 36.3 ± 6.6 μm. Electron micrographs of the constructs in cross-section demonstrated that there were confluent monolayers of cells on the top of the culture and adjacent to the transwell (Fig. 2) . The presence of prominent rough endoplasmic reticulum indicates that the cells were synthetically active (see 3 4 Fig. 5A ). 

The organization of the synthesized ECM, which was contained between bounding cellular monolayers, was similar to native corneal stroma in that arrays of fibrils often alternated in direction (Figs. 3A 3B 3C)and were parallel to both the transwell membrane and the cell body long axes (in the transverse plane). Quantitatively, the average difference in the measured angle between the cell axes and the transverse fibril orientation across all micrographs was 0.65° (± 0.49°; n_fibrils = 90; n_cells = 70). This demonstrates that in each construct, the cells are aligned with the fibrils. If it is assumed that the transverse TEM sections were representative, then at all points in the construct the fibrils and cell bodies may be considered parallel. Figure 3shows TEM micrographs of representative HCF-derived constructs that are qualitatively similar in organization to lamellae found in the normal cornea. 

Nanoscale Collagen Organization/Morphology

Fibril diameters determined from the micrographs measured 38.1 ± 7.4 nm. However, the fibrils in the constructs were polydisperse in their diameter distribution (Fig. 4A) . Periodic banding can be seen on fibrils observed at higher resolution and in longitudinal section (Fig. 4B) . In addition to the larger fibrils, the constructs often exhibited clusters of microfibrils observed in developing mammalian stromas. ^{6 7}  

Cell Density and Cell-Fibril Apposition in Cultures

The ratio of fibroblast-to-ECM volume was high (average cell volume, 45% ± 9%), and the general organization and a number of aspects of the HCF construct were strikingly similar to those of developing mammalian stroma. ⁶ Similarities include the close apposition of cells and cell processes (Fig. 5A) , the prominent rough endoplasmic reticulum (RER; Fig. 5A ), the presence of clusters of microfibrils (Fig. 4C) , and the parallel alignment of fibrils adjacent to the cell membrane (Fig. 5B) . 

DIC Imaging of Planar Organization of Aligned ECM in Constructs

DIC imaging can be used to extract qualitative alignment information about the organization and alignment of extracellular matrix from cell cultures. ³² In our system, we used in-plane DIC imaging to supplement the data obtained in the transverse TEM images (Fig. 6) . 

DIC imaging indicated that the constructs had aligned structures (which we assumed to be arrays of fibrils) parallel to each other over significant distances in the plane of the construct (Fig. 6A) . Because optical imaging has a diffraction limit, it was impossible to determine which structures contributed to the aligned “texture” in the micrograph. However, z-scan flythroughs of the constructs using DIC demonstrated alternating arrays of aligned structures (Figs. 6B 6C)consistent with the TEM images of fibrils in the constructs (Fig. 3) . Taken together, these data suggest qualitatively that the constructs had some degree of organization in their structure transverse to and within the plane of the construct. 

Details of Cell-Matrix Interaction

QFDE images show that the HCFs in the constructs displayed areas of regularly spaced “indentations” on the fibroblast surface (Fig. 7) . Our interpretation is that the structures are always indentations into the cell. Protrusions are observed only when the fibroblast has been “torn” away from the membrane during fracture and we observe indentations from a perspective that was “inside the cell.” The images also demonstrate the presence of aligned small striated fibrils within a fairly dense fibrillar network (Fig. 7B) . Measurement of the fibrils found in the QFDE image (QFDE preserves fibril dimensions) demonstrated that the fibril diameters were small (approximately 22 nm) and polydisperse and that the diameters of individual fibrils varied along their axes. Fibril sizes were not consistent with the diameters of the larger collagen fibrils found elsewhere in the culture by conventional TEM. However, the small diameter and density of fibrillar arrays were consistent with the weakly staining fibrillar structures found throughout the construct by conventional TEM (see 8 Fig. 9 ). Figure 7Cshows a region of amazing regularity in the spacing of the surface indentations. 

In Figure 8 , the fracture plane has fortuitously revealed the intracellular matrix (ICM), the cell membrane (P-face), and the ECM. Some details of the cell/matrix interaction are also elucidated. Within the ICM, it is possible to view the organization of the cell cytoskeleton, which displayed a significant degree of alignment. The images are suggestive that the cell cytoskeleton inserted into the membrane at the base of the indentations (Fig. 8B)where it is possible to observe small, striated moieties (approximately 15 nm in diameter). ECM fibrils, which are in close apposition with and parallel to the cell membrane, appear to “dip” into the indentations on the cell surface (Fig. 8C) . 

The small sizes of most of the fibrils observed in the QFDE images might be explained by closer examination of the standard TEM images. Figure 9shows a dense array of thin fibrils near a collagenous fibril that do not take up stain very well. 

Collagen Types V and VI in Constructs

The constructs in this model were secreted by primary human fibroblasts stimulated with S-Asc. Under similar, but short-term, culture conditions, HCFs have been shown to produce heterotypic types I and V fibrils and type VI collagen. ²⁹ To ascertain whether our long-term cultures contained the collagens (other than type I) usually associated with human corneal stroma (types V and VI ^{41 43} ), immunofluorescence microscopy was used. Collagen type V was detected throughout constructs after disruption of the fibrils by exposure to acetic acid (Fig. 10A) . Type V collagen appeared to stain some sections of the construct more heavily than other sections. Type VI collagen was distributed throughout the construct (Fig. 10B) . 

Discussion

Microscopic Construct Organization

The corneal stroma is one of the most exquisitely organized ECMs in higher-order animals. Thus, within the environment provided during stromal development, prospective corneal fibroblasts exert control over the natural tendency of collagen monomers to form randomly oriented fibrils. ^{6 7} Retention of some of this ability to control collagen organization has been observed on a limited scale in three-dimensional culture systems. ^{44 45} However, attempts to create engineered stromas by seeding degradable scaffoldings with relatively diffuse concentrations of corneal fibroblasts have not resulted in the production of organized, mechanically strong, clear stromal ECM. ^{17 22}  

Gaining a fundamental understanding of the mechanisms by which control over collagen fibrillogenesis is exerted could accelerate our ability to reproduce organized collagenous arrays on the benchtop and enhance our understanding of corneal fibrillogenesis. In the present study, our goal was to investigate whether untransformed donor human corneal fibroblasts, stimulated by a stabilized vitamin C derivative in a scaffold-free system, can reacquire the ability to control collagen fibrillogenesis and assemble an ECM structurally similar to the human stroma (and not to scar tissue). The culture system may constitute a suitable model of stromal development (at least with regard to collagen synthesis) and should afford new opportunities to observe, with high spatial and temporal resolution, those mechanisms that control collagen fibrillogenesis and organization. 

Previous studies ^{24 25 26 27} have shown that the addition of ascorbic acid can increase the proliferative rate of cultured fibroblasts and stimulate the synthesis and secretion of ECM components, such as types I and type III collagen, by acting as a cofactor for the enzymes responsible for hydroxylation of the lysine and proline residues on procollagen. This hydroxylation is required for the proper assembly of procollagen. S-Asc, which extends the longevity of ascorbic acid in culture and has a potent effect on synthesis and secretion of ECM materials, can stimulate the stratification of dermal and corneal fibroblasts and has been shown to stimulate the production of ECM. ^{30 46 47 48}  

We have taken advantage of this effect to produce a corneal fibroblast-synthesized construct, which begins with a single layer of cells in a scaffold-free system. Using the culture system supplemented with stabilized ascorbic acid, fibroblasts are shown to stratify to multiple layers and have produced a collagenous ECM with organized, collagen fibril alignment that alternates in adjacent “lamellae” within 4 weeks of seeding. We assume that the fibrils observed in our cultures were collagenous for several reasons: (1) ascorbic acid–stimulated human corneal fibroblasts are known to produce types I and V heterotypic collagen fibrils under virtually identical culture conditions ²⁹ ; (2) normal corneal collagens (types V and VI) were detected in our cultures by immunofluorescence (see Fig. 10 ); and (3) we are unaware of any striated fibrillar proteins of similar size that are not collagenous. 

In many respects, the gross organization and architecture of the synthesized constructs were strikingly similar to those of a developing rabbit corneal stroma, ⁶ including the high cell-to-matrix volume ratio, the presence of microfibrils, the prominent RER, and the general parallel arrangement of the cells and adjacent ECM fibrils. The presence of the alternating layers of aligned collagen fibrils in “lamellae” produced by the HCFs suggests that mechanisms that control collagen fibrillogenesis and organization are active in our culture system. 

Fibril Diameter Distribution

With regard to the nanostructure of the fibrils themselves, it is important to examine fibril diameters. Corneal transparency depends on the local organization and composition of the stromal ECM. ^{3 4 49 50} With regard to collagen, the spacing, fibril diameter, and fibril polydispersity are important determinants of optical clarity (for a recent review, see Meek and Boote ⁵¹ ). Examination of the standard transmission electron micrographs demonstrated that although organized arrays of aligned collagen fibrils were generated by the HCFs, the fibril diameters were generally larger and more polydisperse (38.1 ± 7.4 nm) than those found by investigators using similar means to image stromal collagen fibrils in adult humans (30.1 ± 2.5 nm ⁵² ) and larger than those found in developing mammalian stromas (approximately 30 nm ⁶ ). For comparison, highly accurate x-ray synchrotron investigations place the fibril diameters in adult human corneas at 31 ± 0.8 nm. ⁵³ The fact that the collagen fibrils in our system were larger and more polydisperse than those found in adult stroma may indicate that the ratio of collagens in the fibrils is incorrect (types I and V heterotypic fibrils, with a ratio of 4:1 ^{35 54} ) or possibly that other molecules thought to control fibril morphology, such as proteoglycans, are not present in appropriate concentrations. ^{55 56 57} It is important to note that our constructs, which were grown freely in a transwell without bounding membranes, are thus in a perpetually “swollen” state during growth. It is likely that some portion of secreted proteoglycans and other soluble molecules were able to diffuse into the medium, where they could not influence the growing collagen fibrils. ⁵⁸  

Cell-Matrix Interaction

QFDE imaging, which has been used to study the fine structure of human, rabbit, chicken, bovine, or developing avian corneal stroma, ^{59 60 61 62 63 64} can reveal the interaction between the ECM and the cell membrane and cytoskeleton in exquisite detail. In our study, it resulted in several observations of interest. The first was the presence of fine fibrils sandwiched between the fibroblasts in an ECM of apparently very high density (Fig. 7A) . The fibrils and microfibrils observed in the QFDE image did not appear to correspond well with the apparent sparsity and fibril size in the collagenous matrix observed in the standard TEM images (Fig. 5A) . This is not surprising because it is well known that QFDE imaging captures more structural information than standard TEM given that it does not rely on stains to absorb electrons and that there is no dehydration step. Instead, any structure that can be coated with platinum is replicated. On closer inspection of the ECM in standard TEM images, however, we did observe a dense matrix of fine fibrils that did not appear to take up the stain very well (Fig. 9) . Weakly staining fibrils of similar size have also been observed in developing rabbit stroma. ⁶  

QFDE imaging also revealed the presence of arrays of shallow indentations on the surfaces of cells, which in some cases are extremely regular in their arrangement (Fig. 7C) . Though arrays of similarly sized indentations on adult human corneal keratocytes in explants have been observed, ⁶⁵ their presence in our cultures of corneal fibroblasts was not expected. The sizes of these indentations were also consistent with those of caveolae found in some fibroblasts (for a review, see Anderson ⁶⁶ ). However, their morphologic structure and apparent interaction were not consistent with those of caveolae. ⁶⁶ Figure 8Asuggests that the indentations compose a network of “binding” sites for fibrils in the ECM that appear to run parallel to the membrane surface. Although this is clearly speculative, the presence of the indentations in organized arrays and the apparent binding of the indentations to fibrils in the ECM are suggestive of a role in the control of matrix organization. 

The fortuitous fracture of the fibroblast cytoskeleton in Figure 8Asuggests that the arrays of indentations are also coupled to the cytoplasmic matrix. The details of the molecular linkage to the cytoskeleton cannot be extracted from the QFDE images without the performance of extensive immunolabeling combined with further QFDE imaging. However, it is possible that the indentations and their coupling molecules comprise a complex ECM attachment system for the fibroblasts. ⁶⁷  

Synthesizing Organized Matrix: Fibripositors and Cholesteric Control

The synthesis of organized collagenous matrix by S-Asc–stimulated HCFs is certainly not a perfect analog for stromal development. Nonetheless, the fact that they are capable of producing significant quantities of aligned collagenous fibrils, packed into arrays that alternate in direction, is proof that the cells in our model had significant control over matrix organization. We evaluated two prevailing theories that explain the deposition of organized collagen by fibroblasts. The first theory was developed primarily in studies on tendon morphogenesis and proposes that fibroblast cells form tubular invaginations, or surface crypts, in their membranes in which collagen monomers are confined and induced to form fibrils that have a particular orientation. ⁸ The fibrils are subsequently “vectorially discharged” from these putative “fibripositors” ^{68 69} into the open extracellular space. Precisely how the cells then “stitch” together an organized matrix is not obvious. It is clear that for the fibripositor model in the cornea to be correct, there must be an organized migration of cells coupled with end-to-end fibril fusion such that the discharged fibrils are laid down in the correct orientation and over the correct distance. That fibroblasts migrate in the direction of aligned features (contact guidance ⁷⁰ ) or along lines of elevated strain or rigidity (durotaxis ⁷¹ ) has been established experimentally. However, our culture system has no preferred orientation, contact guidance features, or regions of enhanced substrate rigidity to cue the cells. Thus, in our constructs, the fibripositor theory would require that the cells move in different directions in adjacent lamellae. 

The second theory postulates that collagen monomers are secreted into the extracellular space in great enough concentration that they become “cholesterically” organized into patterns that reflect the local geometry. ^{72 73} In the developing cornea ^{6 74} and in our constructs, there is a high cell-to-matrix volume ratio, and the cells are aligned and parallel to one another. The cholesteric control of collagen fibrillogenesis theory does not require orchestrated cell movement, just high cell density and high collagen concentration. The relatively small space between cells, the appearance of highly organized collagen within that confined space (which contains dense ECM), and the extremely dense matrix shown by QFDE is consistent with the theory of cholesteric control of matrix organization. The alternating orientation of collagen fibrils in adjacent lamellae, both in our constructs and in developing stroma, is difficult to explain unless the cell surface is capable of influencing the forming fibril orientation. DIC imaging of the confluent monolayer of cells on the transwell membrane suggest that secreted collagen does indeed align along the long axis of spindle-shaped fibroblasts. ⁷⁵ The natural tendency of fibroblasts to change orientation in adjacent confluent layers could explain the alternating collagen orientation if the collagen does align with the body of the cells. 

Conclusions

This is the first investigation to demonstrate that untransformed human corneal fibroblasts in a scaffold-free system can produce a substantial, organized, three-dimensional matrix that bears a qualitative resemblance to the architecture of a developing mammalian corneal stroma. ⁶ It is our opinion that to produce such organized matrix, it is important that the culture system mimic development, where organized matrix is readily and rapidly produced. As in development, we begin with a high density of fibroblastic cells and with no provisional matrix or scaffolding. Using this approach, the fibroblasts, which are in proximity, appear to effectively control the local organization of the matrix they produce. Development is decidedly different from scar resolution, which begins with a provisional matrix or scaffold that is remodeled over long periods of time. ^{76 77} Recent attempts to produce engineered corneas mimic stromal scar resolution and begin by seeding fibroblasts at low density into degradable scaffoldings. ^{16 17 22} The resultant matrix, which is intended to replace the disorganized degrading scaffolding, does not exhibit organized arrays of collagen fibrils. ⁷⁸  

Although our model is not a perfect analog for stromal development, some strong similarities suggest that the same mechanisms may be involved in the syntheses of the organized arrays of collagen fibrils. It is, therefore, a potentially powerful tool in which to study how fibroblasts produce organized matrix. Investigation of this process could lead to better methods to produce organized tissue replacements de novo. 

The organized matrix produced by this scaffold-free system could also provide a suitable starting point for artificial stroma. It is already possible to stack multiple constructs to produce a full-thickness “stromal analog.” However, the complex microarchitecture of a native cornea cannot be easily reproduced in this manner. ^{79 80} Other components of the developmental environment, such as application of in vivo loads ⁸¹ or coculture with other cell populations (Germain L, et al. IOVS 2004;45:ARVO E-Abstract 1451), must be reproduced to refine the matrix morphology. ²¹  

 Supported by National Institutes of Health/National Eye Institute Grant 1R01 EY015500–01.

 Submitted for publication October 10, 2006; revised January 4 and May 30, 2007; accepted July 11, 2007.

 Disclosure: X. Guo, None; A.E.K. Hutcheon, None; S.A. Melotti, None; J.D. Zieske, None; V. Trinkaus-Randall, None; J.W. Ruberti, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

Phase-contrast optical micrographs of the organization of HCFs in transwell. (A) At day 1, the fibroblasts covered the transwell in a monolayer of spindle-shaped cells. (B) At 1 week, the cultures were stratified to multiple layers, and the cells appeared to change orientation as a function of elevation above the transwell membrane. (C) Inset region from (B), where fibroblast orientation change appears orthogonal. Images are representative of five experiments. Scale bar: (A, B) 100 μm; (C) 25 μm.

Figure 2.

 

Transmission electron micrograph of the organization of HCF constructs. Low-magnification TEM of 4-week construct shows 32 μm-thick stratified cell and ECM with confluent cellular monolayers on the top surface and adjacent to the transwell membrane (arrowhead). Bar, 5 μm. Images are representative of five experiments.

Figure 3.

 

Transmission electron micrographs of lamellar-like architecture of the constructs. (A) Low-magnification view of the cells and synthesized arrays of fibrils. Arrows: putative “lamellae” where fibril orientation appears to change direction. Of note is the fact that the lamellae can extend over significant (tens of micrometers) distances. (B) Higher magnification view of the organization of fibrils and their apparent change in direction within the lamellae. Again, arrows indicate the location of changes in fibril orientation. (C) High-magnification view of alternating fibril arrays in the construct. Scale bar: (A, B) 2 μm; (C) 1 μm.

Figure 4.

 

Transmission electron micrographs of fibril morphology in HCF-synthesized constructs. (A) Micrograph showing the diameter polydispersity (end view) of the forming fibrils. Fibril with 28-nm diameter (black arrowhead) and fibril with 45-nm diameter (white arrowhead). (B) Higher magnification micrograph showing the typical banding pattern of the fibrils in the constructs. (C) Micrograph showing the presence of clusters of microfibrils similar to those found in developing mammalian stroma (arrow). Scale bar: (A, C) 500 nm; (B) 200 nm.

Figure 5.

 

Transmission electron micrographs of the architecture of the HCF-synthesized constructs. (A) Dense parallel architecture of cells and cell processes between which the fibril formation is controlled. Note the presence of prominent RER (black arrowheads). (B) Fibrils are often found in close apposition and parallel to the surfaces of the cells in the construct (black arrowheads). Scale bar: (A) 2 μm; (B) 1 μm.

Figure 6.

 

DIC photomicrograph of in plane ECM alignment. (A) Low-magnification DIC imaging qualitatively demonstrates the general alignment of ECM in the middle of a 4-week construct (approximately 25 μm above the transwell membrane). Note the length over which the aligned texture of the image is consistent. Bar, 10 μm. (B) High-magnification DIC showing local alignment in the direction of the arrow. (C) High magnification of same x-y location in (B) but displaced 5 μm more deeply into the construct. Direction of ECM alignment changed relative to the matrix seen in (B). Scale bar: (A–C)10 μm.

Figure 7.

 

QFDE micrograph of 5-week construct cell membranes and extracellular matrix. (A) Low-magnification image of matrix between two fibroblast cells. Cell membranes displayed regular arrays of indentations of unknown function. These regular arrays of indentations on the cell membrane (P-face) had nominal diameters of approximately 70 nm but varied in size. In QFDE, which involves fracturing the specimen, the indentations appeared as protrusions when the cell membrane was viewed from inside the cell (E-face). (B) High magnification of the matrix demonstrates that aligned fibrils were immersed in a dense, less regular ECM. The fibril diameters varied from fibril to fibril and along the axis. (C) Highly organized array of indentations on cell membrane (E-face). The array was closely spaced in one direction (100 nm; white arrows) and more widely spaced in the other (150+ nm; black arrows). Scale bar: (A) 500 nm; (B, C) 100 nm.

Figure 8.

 

QFDE micrographs of cell matrix interaction in 5-week construct. (A) Low magnification of cell interaction with the neighboring ECM. The image shows the intracellular matrix (ICM), the extracellular matrix (ECM), the outside of the cell membrane (P-face), and the details of ECM fibril interaction with the cellular membrane and possibly the cellular cytoskeleton. (B) High magnification of cellular cytoskeletal interaction with the cell membrane where the indentations have been disrupted by the fracture plane. The membrane appears to interact with fibrillar molecules, which span the indentation parallel to the cell surface (arrows). (C) High-magnification image of the indentation morphology and its interaction with an ECM fibril. The indentation was spanned by a fibrillar molecule, which appears to be striated (black arrowhead). At the top of the image, an ECM fibril appears to dip into an indentation (white arrow). (D) Striated fibrils in the matrix presumed to be outside the cell (arrow). The fibrils measure approximately 20 nm in diameter and appear to be branched. Scale bar: (A) 500 nm; (B–D) 50 nm.

Figure 9.

 

Transmission electron micrograph of collagen and associated fibrillar structures. A collagen fibril (white arrow) is associated with numerous faintly stained microfibrils. These microfibrillar structures could account for the high density of fine fibrils in the ECM detected by QFDE imaging. Bar, 100 nm. Image is representative of five experiments

Figure 10.

 

Indirect immunofluorescence micrographs of collagen in HCF-synthesized constructs. (A) Merged type V collagen- (red) and DAPI- stained (blue) construct. Some areas stained more intensely than others, but type V collagen was distributed throughout the extracellular matrix. (B) Merged type VI collagen- (amber) and DAPI-stained (blue) construct. Type VI collagen was found throughout the tissue but appeared more concentrated at the bottom of the construct. Scale bar; (A, B) 50 μm.

The authors thank Pat Pearson for excellent transmission electron microscopy work and Tatfong Ng for the collagen staining and microscopy. They also thank David Birk for many insightful discussions on collagen and ECM structure.