HKCs were isolated from human corneas from patients with Keratoconus defects. These corneas were obtained from Ula Jurkunas (Massachusetts Eye and Ear Infirmary, Boston, MA, USA). HCFs were isolated from normal human corneas obtained from NDRI (National Disease Research Interchange; Philadelphia, PA). Briefly, corneal epithelium and endothelium were removed from the stroma by scraping with a razor blade. The stromal tissue was cut into small pieces (~ 2 × 2mm) and put into 6-well plates (4 or 5 pieces per well). Explants were allowed to adhere to the bottom of the wells and Eagle’s Minimum Essential Medium (EMEM: ATCC; Manassas, VA) containing 10% fetal bovine serum (FBS: ATCC) was added. After 1–2 weeks of cultivation, the cells were passaged into a 100 mm cell culture plate. The cells were allowed to grow to 100% confluence before being used in the culture system. Morphologically, HKCs exhibited characteristic myofibroblastic morphology, such as prominent cytoplasmic actin microfilaments (stress fibers) and high levels of α-smooth muscle actin (SMA) expression (data not shown), that distinguished them from healthy HCFs.

Both HCFs and HKCs were plated on transwell 6-well plates containing polycarbonate membrane inserts with 0.4 μm pores (Costar, Charlotte, NC, USA,) at a density of 10⁶ cells/mL. The protocol followed was identical for both cell types. Cells were cultured in EMEM with 10% FBS and 0.5 mM 2-O-α-D-glucopyranosyl-L-ascorbic acid (VitC, Wako Chemicals USA, Inc.; Richmond, VA, USA). The cultures were allowed to grow for 4 weeks in the presence of 0.1ng/mL TGF-β1 (T1) or TGF-β3 (T3). Cultures without any growth factors served as Controls (C). Culture medium was changed every other day for the duration of the experiment. The morphology of the cultures was examined using transmission electron microscopy (TEM). In addition, indirect-immunofluorescence was used to identify specific markers of stromal components—Smooth muscle actin (SMA), type I collagen (Col I), type III collagen (Col III), type V collagen (Col V) and fibronectin (EDA-Fn).

After 4 weeks in culture, constructs were collected and fixed in 4% paraformaldehyde before being processed for IF, as previously described [9]. Briefly, each sample was incubated at 4 °C overnight with the following primary antibodies diluted in 1% BSA + 0.1% Triton-X: anti-Col I, III, and V (Southern Biotech, Birmingham, AL, USA), anti-EDA-Fn (Sigma Aldrich, St. Louis, MO, USA), and anti-SMA (Dako North America, Carpinteria, CA, USA). The samples were then washed and incubated overnight at 4 °C with the corresponding secondary antibody in 1% BSA + 0.1% Triton-X—donkey anti-goat (Col III), anti-mouse IgM (EDA-Fn), and anti-mouse IgG (SMA, Col I and V). TO-PRO-3 iodide (Life Technologies Corporation, Grand Island, NY, USA) was used as a marker of cell nuclei. Finally, constructs were washed, mounted with Vectashield Mounting Media (Vector Laboratories, Inc.; Burlingame, CA, USA), and then observed and photographed using a confocal TCS-SP2 or TCS-SP5 Leica microscope (Leica Microsystems; Bannockburn, IL, USA). For negative controls, the primary antibody was omitted.

At this time, the construct’s thickness was also measured as previously described [10]. Briefly, using the confocal microscope’s z-scans, measurements began at the top of the construct (first cell visible) and ended at the bottom of the construct (last cell visible). Thicknesses were analyzed (7–11 samples per condition) for significant difference (p < 0.05)

At week 4, constructs were collected and fixed in ½ strength Karnovsky’s fixative at 4 ºC overnight. After which, they were processed for TEM, as described in Gipson et al. [11]. Briefly, constructs were rinsed in PBS, processed through post-fixation in 2% osmium tetroxide, en bloc stained in 0.5% uranyl oxide, dehydrated with alcohol to propylene oxide, and embedded in Embed 812 (Electron Microscopy Sciences; Hatfield, PA, USA). Thin sections, transverse to the plane of the construct, were cut with a diamond knife on an ultramicrotome (LKB, Bromma, Sweden). The sections were viewed and photographed with a Tecnai G² Spirit (FEI Company, Hillsboro, OR, USA). On average, 20 photos were used per condition per cell type. All experiments were repeated at least three times.

FFT is a method to describe the anisotropy of fibers in binary images by using image processing. In this study, we took the images collected from the confocal microscope and quantified the orientation of the collagen fibers and cells at each z-section by applying a graphical FFT. The graphical FFT uses spectral techniques with the aid of 2-dimensional (2D) FFT in the form of vectors in order to graphically quantify the orientation of the fibers or cells [12,13]. Briefly, confocal z-series stack images were converted to grayscale images and the distribution of fibers or cells were captured using our custom MATLAB software. The distribution of orientations was shown on polar plots where the intensity of the 2D FFT profiles was converted into polar coordinates. In each case, the median and the range of the orientations from the top 70% of the population were reported as indicators of the main orientation and extent of alignment at that orientation respectively. Therefore, wider histograms represent a higher range of alignment, which corresponds to more randomly oriented fibers or cells in an image. In order to obtain the values for rotation, the change in the angle of orientation between z-scans were calculated and applied.

Image Pro Plus (Image Pro Plus v.7: Media Cybernetics, Bethesda, MD, USA) is a 2 and 3D analysis software program that was used to analyze and quantitate data from the confocal z-series, as previously described [14]. Briefly, Image Pro Plus was used to count the number of cells in each section (plane-of-focus) of a confocal z-series in order to quantitate the number of cells per construct. At least three confocal z-series were quantitated per condition and their average was analyzed.

All experiments were repeated at least three times and data was analyzed for significant variations (p < 0.05) using the Student’s t-test and Dunnett’s Multiple Comparison test.

In our previous work [14,15], we have shown that different cell types, such as HCF and umbilical cord stem cells, have the ability to secrete and lay down a fibrous collagen ECM in the presence of VitC. We have particularly demonstrated the effect of T1 and T3 in these cultures [14]. In the current study, we examined the ability of HKCs to synthesize a matrix in the presence of VitC ± T1 or T3. Matrix thinning is one of the end results in patients with Keratoconus disease; therefore, the study for matrix secretion and deposition is critical.

After 4 weeks in VitC only, the HKCs’ construct had a mean thickness of 9.5 μm, whereas, the HCFs synthesized a matrix of 17.6 μm thick (Figure 1). When T1 was added to the cultures, the HKC construct’s thickness increased significantly (18.2 μm: p < 0.05; Figure 1); however, this is still less than the HCFs’ construct, where the thickness increased to 52.3 μm (p < 0.001; Figure 1). Upon T3 stimulation, the HKC and the HCF constructs maintained similar thicknesses as with T1 (20.1 μm and 40.2 μm, respectively). Note that with all the conditions, the HKCs secreted significantly less matrix when compared to the HCFs (p < 0.05); however, with the addition of T1 or T3, the HKCs were of comparable thickness as the HCF VitC only constructs. This data suggests that HKCs can be stimulated to produce more matrix, thus partially “rescuing” the HKCs to be more like HCFs.

In order to determine whether this increase in thickness was the result of an increase in the number of cells or an increase in the amount of matrix produced per cell, we examined the total cell number for each construct under all conditions. Results for both cell types showed a 2-3-fold increase in cell number following TGF-β stimulation compared to controls (data not shown). However, the amount of ECM produced per cell remained relatively constant under all conditions.

TEM was used to observe the cell-ECM interaction in each of the HKC constructs at the end of 4 weeks. HCFs were used for comparison (data not shown) [10]. As seen in Figure 2A,B (Control and T1, respectively), areas of organized ECM were apparent, and there was a higher density of collagen fibrils and ECM alignment in the control constructs (Figure 2A) than the T1 (Figure 2B). This data is opposite to that of the HCF, where upon the addition of T1 to the HCF constructs, the collagen fibrils were more organized and compact than in control (data not shown) [10]. The T3-treated HKC constructs (Figure 2C), however, showed the highest density of collagen fibrils and ECM alignment out of all three conditions. Similar results were observed with HCFs (data not shown) [14], where T3 stimulation resulted in highly organized and denser ECM.

Z-series scans obtained with the confocal microscope were imported into a custom-made MATLAB program to calculate and analyze the cell and ECM alignment and orientation for each Z-scan based on FFT analysis. Figure 7 shows representative images used to analyze the data for both cell and ECM orientation. Figure 7A₁ and B₁ are single plane-of-focus (Z-scan) confocal images showing the cell (SMA) and ECM (Type I collagen) staining, respectively. The corresponding FFT analysis is also shown (Figure 7A_2,3 and B_2,3). The global orientation is indicated by the 2D FFT transform of the stained cells (Figure 7A₂) or ECM (Figure 7B₂). To quantify the orientation of the cells or ECM, we plotted polar graphs (or histograms), which indicate the intensity of the FFT transform images at different angles (Figure 7A₃ and B₃). The blue line on the polar graph equals the angle (θ) of the cells or ECM and shows the normalized integral intensity of all the pixels in the 2D FFT image within (0,θ)—the origin of the image—to (200,θ) polar coordinates. The dominant orientation corresponds to the peaks of the polar graph. Furthermore, from the same polar graph, we can calculate how aligned the overall population was by measuring the width of the peaks—the wider the peaks, the more dispersed the distribution of orientations and the more random the alignment.

Using the tools shown above, we quantified the cell orientation for each z-scan under all the conditions using constructs stained with anti-SMA. For analyzing the alignment, we plotted the width of the peak of the polar graph, or the Δ of angle, versus the z-scan position starting at 0 (top of the construct). The step size between z-scans was 0.5 microns. Figure 8 shows the quantification of the cell alignment for both HCFs (Figure 8A) and HKCs (Figure 8B). The lower the value on the Y-axis, the smaller the Δ of angle, thus the more aligned the cells. As seen in Figure 8A, HCFs showed higher cell alignment when stimulated with T1 and T3 when compared to Controls. Controls mainly aligned towards the bottom half of the construct, as indicated by the sudden drop in angle value. On the other hand, in Figure 8B HKCs showed high cellular alignment under all conditions, even under Controls, with T3 showing a trend for the highest alignment between the three conditions. Notably, HKCs were more aligned when compared to HCFs.

In terms of cell rotation, the mean angle on the polar graph, which is equal to the direction that the cells are oriented, was calculated for each z-scan within a construct. The angles for each of the z-scans were then plotted on a graph versus the z-scan position, thus showing the rotation of the oriented cells. Therefore, the wider the range of the angles for a construct, the more cellular rotation is observed. In Figure 9, the cell rotation graphs for both cell types under all conditions are shown. HCFs (Figure 9A) showed cell rotation under all conditions. Controls had a range of ~40°, whereas T1 and T3 had a range of ~55°. HKCs (Figure 9B), on the other hand, showed minimal cell rotation in Control and T3-treated cells; however, T1-treated cells showed a single massive sudden cell rotation of ~110 °C. Overall, from the results in Figure 8 and Figure 9, HKCs were more aligned and had less rotation than HCFs.

For ECM alignment and rotation (Figure 10 and Figure 11, respectively), we used the constructs stained with Type I collagen in order to quantify our data for all conditions. As can be seen in Figure 10A, HCFs secreted a relatively aligned ECM irrespective of the different treatments, with slightly less alignment in the Controls. With the HKCs (Figure 10B), the Controls only secreted a single layer of ECM, which appeared to be aligned (Figure 10B); however, T1 and T3 produced a 15–20 μm-thick ECM, which was equally aligned (Figure 10B).

ECM rotation, and for this specific study, Col I rotation, was seen with both cell types (Figure 11). More specifically, HCFs (Figure 11A) showed ECM rotation throughout the constructs, under all treatments. T1 and T3 exhibited more frequent changes in rotation, whereas, the Controls had a smother rotation/transition. HKC Controls (Figure 11B) only produced a monolayer, therefore rotation could not be judged. T1 and T3 treatments, however, resulted in thicker ECM; therefore, the rotation of the ECM was more evident. For both T1 and T3, the rotation was smooth with no sudden changes.