Safety and Feasibility of Intrastromal Injection of Cultivated Human Corneal Stromal Keratocytes as Cell-Based Therapy for Corneal Opacities

Abstract

Purpose: To evaluate the safety and feasibility of intrastromal injection of human corneal stromal keratocytes (CSKs) and its therapeutic effect on a rodent early corneal opacity model.

Methods: Twelve research-grade donor corneas were used in primary culture to generate quiescent CSKs and activated stromal fibroblasts (SFs). Single and repeated intrastromal injections of 2 to 4 × 10⁴ cells to rat normal corneas (n = 52) or corneas with early opacities induced by irregular phototherapeutic keratectomy (n = 16) were performed, followed by weekly examination of corneal response under slit-lamp biomicroscopy and in vivo confocal microscopy with evaluation of haze level and stromal reflectivity, and corneal thickness using anterior segment optical coherence tomography (AS-OCT). Time-lapse tracing of Molday ION–labelled cells was conducted using Spectralis OCT and label intensity was measured. Corneas were collected at time intervals for marker expression by immunofluorescence, cell viability, and apoptosis assays.

Results: Injected CSKs showed proper marker expression with negligible SF-related features and inflammation, hence maintaining corneal clarity and stability. The time-dependent loss of injected cells was recovered by repeated injection, achieving an extended expression of human proteoglycans inside rat stroma. In the early corneal opacity model, intrastromal CSK injection reduced stromal reflectivity and thickness, resulting in recovery of corneal clarity, whereas noninjected corneas were thicker and had haze progression.

Conclusions: We demonstrated the safety, feasibility, and therapeutic efficacy of intrastromal CSK injection. The cultivated CSKs can be a reliable cell source for potential cell-based therapy for corneal opacities.

The cornea is transparent and consists of a five-layer structure (the outermost corneal epithelium, Bowman's layer, corneal stroma, Descemet's membrane, and the innermost corneal endothelium). The corneal stroma spans approximately 90% of corneal thickness and provides mechanical strength and optical clarity to the cornea due to the singular disposition of collagen fibrils, which form lamellae that run orthogonally to each other, allowing light transmission with minimal diffraction or reflection.¹ Corneal stromal keratocytes (CSKs) reside among collagen lamellae and synthesize and deposit collagens and keratan sulfate proteoglycans (KSPGs; lumican, keratocan, and mimecan) to regulate collagen fibril alignment and interfibrillar spacing, which are crucial for stromal architecture and transparency.^2,3 Upon corneal injury, the quiescent CSKs become activated and change phenotype into stromal fibroblasts (SFs), which are proliferative and produce repair-type extracellular matrix (ECM) components (e.g., fibronectin, proteinases, and α5-integrin) in the event of wound healing.^4–6 SFs can further transform into myofibroblasts resulting in scar formation, and the fibrotic tissues interfere with or obstruct light transmission, resulting in reduced visual acuity and eventually vision loss.^7,8 Trauma, infection, degeneration, and immunologic disorders (e.g., keratoconus), inherited diseases, and/or refractive surgeries can lead to CSK death, and the surviving keratocytes can transit to SFs, causing haze development and even opacification when myofibroblasts are present.^9–11 Corneal opacities/scar are a significant cause of global blindness, affecting over 10 million people worldwide.^12,13 In most situations, surgical removal can restore eyesight.^14,15 Even though the development of eye bank facilities and refinement of surgical procedures (penetrating and lamellar keratoplasty) have substantially improved the treatment outcome of corneal blindness in recent years, widespread accessibility to modern-day surgery is still restricted worldwide, due to the shortage of donor tissue, lack of surgical expertise, postsurgery complications (e.g., graft rejection, immune responses), and adverse side effects (e.g., cataract and/or glaucoma induced by long-term use of corticosteroids).^16,17 Hence, there is an increasing interest in developing new strategies, such as targeted cell therapy using intrastromal injection and/or cell-incorporated bioscaffolds to restore the stromal functions and corneal transparency.^18,19 This demands long-term survival of appropriate stromal cells (i.e., stromal keratocytes) with continuous expression and deposition of stromal crystallins, collagens, and KSPGs. This treatment regimen would aid in recovering stromal architecture and corneal transparency. Even some modifications to the existing scar density could alleviate the need for tissue grafting. 

We have described an ex vivo protocol consisting of soluble amnion stromal extract, Rho-associated coiled-coil containing protein kinase (ROCK) inhibitor Y27632 and insulin-like growth factor-1 (ERI supplement) to propagate human “activated keratocytes.”^20,21 Through media switching to serum-free condition, the culture derives quiescent CSKs expressing specific CSK markers (lumican, keratocan, aldehyde dehydrogenase 3A1 [ALDH3A1]), similar to native stromal tissue. The expanded cultures are negative to CD90/Thy1 and α-smooth muscle actin (SMA) expression, indicating the lack of transition to SFs and myofibroblasts, which has been a major issue concerning CSK culture.^22–25 Use of fibroblast-like cells, such as in stromal cell therapy or tissue engineering, would hamper the proper stromal features as they deposit fibrotic ECM components to increase light scattering, thereby compromising corneal clarity.^8,26–28 

The ability to recruit ex vivo cultivated CSKs in stromal cell therapy has the potential to revolutionize the treatment of corneal opacities and stromal disorders. In order to challenge this hypothesis, we assessed the safety and applicability of injecting human CSKs to rat corneas and studied the effect on corneal clarity and stromal architecture, compared to SFs derived from the same donor source. The cell survival, expression of specific CSK markers, and recipient stromal response were examined. We also elucidated the therapeutic effect of injecting human CSKs in a rat corneal opacity model created by irregular phototherapeutic keratectomy (irrPTK). 

Materials and Methods

Donor Corneas

Twelve research-grade cadaveric corneas (donor age: 36.5 ± 14.1 years old; sex ratio: female/male = 4/8) deemed unsuitable for transplantation (Table 1) were procured from Lions Eye Institute for Transplant and Research, Inc. (Tampa, FL, USA) following institutional review board approval, in accordance with approved guidelines. Consent was obtained at the time of retrieval by next of kin for research use. Corneal tissues preserved in Optisol-GS (Bausch&Lomb Surgical, Irvine, CA, USA) were transported at 4°C to the culture facility. 

Table 1

 

Donor Information for Corneas Used for Primary CSK and SF Cultures

 

Isolation and Culture of Human CSKs and SFs for Injection

The central stroma was processed for primary CSK and SF cultures using ERI protocol.^20,21 Briefly, corneal stroma was digested in 0.2% collagenase I for 3 hours at 37°C and cells were collected for “activated keratocyte” propagation using ERI reagents added with 0.5% fetal bovine serum (FBS) on a collagen I–coated surface until passage 5. Two-thirds of the culture was switched to the serum-free ERI condition for 1 week to generate quiescent CSKs and the rest was placed in 5% FBS to derive SFs. Characterization of both cell types was done using marker expression (including CSK markers keratocan [Kera], lumican [Lum], ALDH3A1; and SF/myofibroblast markers CD90/Thy1, αSMA) prior to animal studies. For cell labelling, Molday ION EverGreen reagent (1:200 dilution; BioPAL, Inc., Worcester, MA, USA) was added to culture for 48 hours followed by extensive washes prior to being harvested for injection. Cell labelling was monitored under fluorescence phase contrast microscopy (Eclipse TS100; Nikon, Melville, NY, USA) and cultures with >95% labelled cells were used for animal injection. 

Rats and Intrastromal Cell Injection

Sprague-Dawley rats (6–8 weeks old, n = 86) were treated according to the guidelines of the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. The experiment protocol was reviewed and approved by the Institutional Animal Care and Use Committee of SingHealth, Singapore (2015/SHS/1033). Rats were anesthetized by intraperitoneal ketamine hydrochloride (80 mg/kg; Parnell Laboratories, Alexandria, NSW, Australia) and xylazine (12 mg/kg; Troy Laboratories, Glendenning, NSW, Australia). The eyes were rinsed with normal saline and anesthetized topically with lignocaine hydrochloride (1%, Pfizer Laboratories, New York, NY, USA). All right eyes were used for experiments and left eyes as normal control. Among them, 70 normal corneas were used for intrastromal cell injection for the safety study and 16 were injured by irrPTK to induce corneal opacities followed by cell injection. An experienced corneal surgeon (JM) performed the intrastromal injection. Prior to injection, a stromal tunnel was created with a 31-gauge (G) needle at the anterior stroma, and 2 × 10⁴ cells in 2 μL sterile PBS (Life Technologies, Carlsbad, CA, USA) were injected through a 31-G blunt needle attached to a Hamilton syringe (Hamilton Company, Reno, NV, USA) (Supplementary Fig. S1). Corneal response (including haze/opacity formation, neovascularization), cell survival, and immunohistochemistry were assessed in all rat corneas (15 injected with CSKs, 25 with SFs, and 12 with PBS only). Another 18 normal corneas were injected with Molday ION EverGreen–labelled cells for cell tracing study. We injected 2 × 10⁴ cells in 2 μL PBS to four corneas and examined cell density changes up to 6 weeks. To replicate the finding, a higher cell dosage (5 × 10⁴ cells in 2 μL PBS) was injected to another four corneas and examined similarly. We also compared the cell density changes between single and repeated cell injections. At week 0, cells at a dosage of 2 × 10⁴ in 2 μL PBS were injected to 10 corneas, of which 5 were randomly selected at week 5 to receive the second injection with same cell dosage, and the corneas were followed up until week 9. After each injection, rat eyes were instilled with tobramycin (1%; Alcon, Fort Worth, TX, USA) four times daily for 3 days. Ophthalmic examinations were performed on anesthetized rats weekly, and rats were euthanized to collect corneas for immunohistochemistry and cell survival studies. 

Rat Corneal Opacity Model Induced by irrPTK

IrrPTK surgery was performed on right eyes of 16 rats as previously described.²⁹ In brief, corneal epithelial debridement was done using a #64 surgical blade (BD, Franklin Lakes, NJ, USA) sparing the limbus, and irrPTK was done within a 3-mm ablation zone on central stroma at an ablation depth of 10 μm using a Technolas 217z excimer laser (Bausch&Lomb). Irregular stromal damage was achieved by placing a fine metal mesh over the laser ablation area after firing 50% of the pulses. The rats received topical tobramycin (1%) four times daily for 3 days. After 1 week, the epithelium was healed and early haze formation was achieved. The corneas then received intrastromal cell injection at the haze region in the anterior stroma at the region of haze formation. The rats were randomly placed into three groups: CSK injection with 4 × 10⁴ cells in 2 μL sterile PBS (n = 5), 2 μL PBS injection (n = 6), and noninjection (n = 5). Thereafter, all injured eyes were instilled with TobraDex (1%, Alcon) four times daily for 7 days and examined weekly. Rats were killed at weekly intervals and all corneas were harvested for immunohistochemistry. 

Ophthalmic Examination and Measurements

All corneal imaging and measurements were performed 3 days prior to injection (preoperative), at the third day (for the resolution of stromal edema and recovery of epithelial wound after injection), and then 1, 2, 3, and 4 weeks after cell injection. Slit-lamp photographs were taken using a Zoom Slit Lamp NS-2D (Righton, Tokyo, Japan). The number of corneas manifesting haze and neovascularization after CSK and SF injections was analyzed by a χ² test for categorical variables. Corneal haze was graded according to the Fantes scale for haze as previously described.³⁰ Two independent observers (GH-FY, MF) graded the corneal clarity, and mean values were obtained for comparative analysis. Corneal cross-section visualization and measurement of central corneal thickness (CCT) were performed using anterior segment optical coherence tomography (AS-OCT, RTVue; Optovue, Inc., Carl Zeiss Meditec, Dublin, CA, USA). Mean CCT was measured as the mean of three measurements taken at the center (0 mm) and at 0.5 mm on either side, respectively.³¹ In vivo confocal microscopy was performed using Heidelberg retinal tomography HRT3 with Rostock corneal module (Heidelberg Engineering GmbH, Heidelberg, Germany). A carbomer gel (Vidisic; Mann Pharma, Berlin, Germany) was applied as immersion fluid. All corneas were examined centrally with at least three z-axis scans from the corneal epithelium to corneal endothelium. Semiquantitative analysis of the stromal reflectivity levels along the injection plane was performed using ImageJ software (National Institutes of Health, Bethesda, MD, USA) after the images were tonal adjusted to map the actual pixel values using PhotoShop CC (Adobe Systems, Inc., San Jose, CA, USA).³¹ For the time-lapse study of Molday ION EverGreen–labelled cells, the imaging was performed 3 days before injection (background reference) and immediately after injection (time 0) and weekly post injection using Spectralis OCT (Heidelberg Engineering GmbH) (with a lens of 30° field of view) and an excitation under fluorescein (FA) mode and intensity setting to 90.³² A mean of 50 consecutive frames were taken for the entire stromal depth. At least five well-focused confocal images were chosen at every examination for the measurement of injected cell intensity using Quantity One 1-D Analysis software (BioRad, Hercules, CA, USA). The grayscale images were measured using area and density modes. After background subtraction (images before injection), the percentages of fluorescence intensity at time intervals were calculated with reference to that at time 0. 

Immunofluorescence

The excised whole rat corneas were fixed in freshly prepared neutral buffered 2% paraformaldehyde (Sigma-Aldrich Corp., St. Louis, MO, USA) and embedded for cryosectioning at a thickness of 6 μm. Sections were treated with ice-cold 50 mM ammonium chloride (Sigma-Aldrich Corp.), saponin permeabilized, and blocked with 2% bovine serum albumin (BSA; Sigma-Aldrich Corp.) and 5% normal goat serum (NGS; Invitrogen, Carlsbad, CA, USA), followed by incubation with primary antibodies (Table 2) for 2 hours at room temperature.³³ For Kera and Lum immunostaining, sections were pretreated with endo-β-galactosidase (1.5 U/mL in 10 mM phosphate buffer pH 7.4; Sigma-Aldrich Corp.) for 30 minutes at 37°C prior to blocking and antibody incubation. After PBS washes, they were labelled with Red-X– or Alexa 488–conjugated IgG secondary antibody (Jackson ImmunoResearch Labs, West Grove, PA, USA) for 1 hour at room temperature, washed, and mounted with Fluoroshield with DAPI (4′,6-diamidino-2-phenylindole; Santa Cruz Biotech, Santa Cruz, CA, USA) and viewed under fluorescence microscopy (Zeiss Axioplan 2; Carl Zeiss, Oberkochen, Germany). 

Table 2

 

Antibodies Used in This Study

 

Quantitative PCR

Total RNA from cultured cells was extracted using RNeasy kit (Qiagen, Singapore) and on-column RNase-free DNase kit (Qiagen), respectively, according to manufacturer's protocol. Reverse transcription of RNA (1 μg) was done with Superscript III RT-PCR kit (Invitrogen) with random hexanucleotide primer (10 ng/mL, Invitrogen). Gene expression was assayed with specific primer pairs (Table 3) by quantitative real-time PCR (qPCR) using Sybr Green Supermix (BioRad) in GFX96 Real-Time System (BioRad).²⁰ Samples were run in quadruplicate, and the relative gene expression level was normalized with the housekeeping glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (CT_GAPDH) and relative fold changes between CSK and SF were analyzed using comparative CT method. Results were expressed as mean and standard deviation (SD) from quadruple runs. 

Table 3

 

Expression Primers

 

Flow Cytometry

Cells were fixed with freshly prepared neutral buffered 2% paraformaldehyde, permeabilized, and blocked by 1% Triton X-100, 2% BSA, 2% NGS. The samples were incubated with rabbit anti-human ALDH3A1 antibody (Proteintech, Rosemont, IL, USA) and isotype-specific IgG (BD Biosciences, Singapore), respectively, followed by Alexa Fluor 488–conjugated IgG secondary antibody and propidium iodide, and the staining signal was analyzed by FACSVerse System (BD Biosciences) with a minimum of 10,000 events in each experiment.³³ 

Cell Viability Assay

The cells grown on coverslips were incubated in calcein AM and ethidium homodimer-1 (EthD-1) (Live/Dead Viability/Cytotoxicity kit, Life Technologies) for 45 minutes. After washes, samples were mounted in Fluoroshield with DAPI (Santa Cruz Biotech) and viewed under fluorescence microscopy (Zeiss Axioplan 2). In a minimum of six random fields captured with a ×20 objective, the number of live (green fluorescence) and dead cells (red fluorescence) was quantified and the percentage of viability was calculated.³⁴ Experiments were done in triplicate. 

Apoptosis Assay

TUNEL (terminal UTP nick-end labelling) reaction was performed on rat corneal cryosections using TMR-In Situ Cell Death Detection kit (Roche, Basel, Switzerland) following the protocol previously described.³⁵ Briefly, the samples were permeabilized with 0.1% Triton X-100 (Sigma-Aldrich Corp.) in 0.1% sodium citrate (Sigma-Aldrich Corp.) for 2 minutes on ice, followed by PBS rinses. They were incubated with complete TUNEL reaction mix for 1 hour at 37°C. The positive control was sections pretreated with 5 U/mL DNase I (Sigma-Aldrich Corp.) for 20 minutes at 20°C to 22°C followed by complete TUNEL reaction. Negative control was DNase-treated sections incubated in TUNEL reaction mix without TdT enzyme. The samples were thoroughly washed, mounted with Fluoroshield with DAPI (Santa Cruz Biotech), and viewed under fluorescence microscopy (Zeiss Axioplan 2). To evaluate the effect of human cell injection on the survival of stromal resident cells, we quantified TUNEL positivity for the injected ION-labelled human cells and resident rat stromal cells, respectively. In a minimum of five sections per cornea collected at different time points, the number of ION-labelled human cells displaying TUNEL signal was counted. Simultaneously, the TUNEL-positive non–ION-labelled cells (rat stromal cells) were quantified inside the stromal region with a boundary set at 200-μm distance from the outermost HuNu-labelled human cells. This was designed to examine the survival of rat stromal cells adjacent to the injection site. For PBS-injected and normal rat corneas, the cells with TUNEL signal were quantified in the central stromal region (1-mm distance on either side of the corneal midline). Percentages of TUNEL-positive cells (apoptosis indices) were compared between the injected human cells and resident rat stromal cells at the same postinjection time point using Mann-Whitney U test for statistical significance. 

Transmission Electron Microscopy (TEM)

Corneal samples at the fourth week post injection were fixed with 3% glutaldehyde (EM Sciences, Hatfield, PA, USA), 1% tannic acid (Sigma-Aldrich Corp.), and 1% aqueous solution of osmium tetroxide (EM Sciences) sequentially and processed for Epon-Aradite 812 embedding. Ultrathin sections (80–90 nm thick) were contrast stained with 3% uranyl acetate (Sigma-Aldrich Corp.) and lead citrate (Sigma-Aldrich Corp.) and examined under TEM (JEOL 2100, Tokyo, Japan). 

Statistical Analysis

All data were expressed as mean ± standard deviation (SD). Statistical analyses were performed with SPSS version 22.0 (IBM, Chicago, IL, USA). CCT changes among CSK-, SF-, and PBS-injected corneas were analyzed using Kruskal-Wallis with Dunn post hoc test for multiple comparisons. Number of rat corneas manifesting with haze presentation after CSK and SF injections was analyzed by a χ² test for categorical variables. Mann-Whitney U test was used to compare label intensity, reflectivity, and haze level data between CSK and SF injections at defined time points. P < 0.05 was considered statistically significant. 

Results

Human CSK and SF Culture and Characterization

Under phase contrast microscopy, the primary human CSKs displayed dendritic/stellate morphology, with cell processes extending to connect with neighboring cells forming a cellular network (Figs. 1A, 1a, 1b). As revealed by immunofluorescence, the majority of cells expressed keratocyte markers (Kera and ALDH3A1) and were devoid of CD90/Thy1 (fibroblast marker) and αSMA (myofibroblast marker) (Figs. 1A, 1c–1e). In contrast, SFs under serum culture showed slender shape (Figs. 1A, 2a, 2b) and negligibly expressed Kera and ALDH3A1, but had positive Thy1 detection (Figs. 1A, 2c, 2d). Both cultures had negative αSMA, indicating that there was minimal myofibroblast transformation (Figs. 1A, 1e, 2e). When SFs were treated with TGF-β2 (10 ng/mL) for 5 to 7 days, αSMA was detected, and this served as the positive control of myofibroblasts (Figs. 1A, 2f). 

Figure 1

 

Cell characterization. (A) Phase contrast microscopy (a to b) of CSKs and SFs cultivated from same donor corneas. CSKs expressed phenotypic markers Kera (1c) and ALDH3A1 (1d) whereas SFs expressed Thy1 (2c, 2d) with negligible αSMA (2e) by immunofluorescence. As a positive control for αSMA expression, SFs were cultured in 10 ng/mL TGF-β2 and 5% FBS condition to induce myofibroblast phenotype (2f). Scale bars: 100 μm. (B) RNA expression of various gene markers by qPCR. SFs negligibly expressed ALDH3A1, Kera, Lum, Col8A2, B3GNT7, and CHST6 while they had upregulated Thy1, when compared to CSK (*P < 0.05, Mann-Whitney U test). (C) Flow cytometry showed ALDH3A1 expressed in >99% of cultured CSK and <1% of SF.

Figure 2

 

Rat corneal changes after injection. (A) Weekly slit-lamp biomicroscopy and AS-OCT examinations showed that CSK-injected and PBS-injected corneas were clear. In contrast, SF-injected corneas developed substantial corneal haze. (B) Mean central corneal thickness (CCT) measured from AS-OCT images illustrated a transient thickening of all corneas after injection. The thickness of CSK- and PBS-injected corneas gradually returned to preinjection level while SF-injected corneas remained thicker, with CCT increased by 30% to 40% compared to preinjection level. The number of corneas recruited for CCT measurement at weekly intervals is indicated below the line graph. *P < 0.05 for significant difference between SF- and CSK-injected corneas, Mann-Whitney U test. (C) Slit-lamp images at 3 to 4 weeks post injection showed clear corneas after CSK (n = 8) and PBS (n = 5) injections. The SF-injected corneas (n = 14) developed corneal opacities and neovascularization (CNV). Numbers of corneas with abnormalities are indicated.

The qPCR study also showed that SF culture had marked reduction of keratocyte marker expression (mean expression level for ALDH3A1 was 1%, Kera 0.7%, Lum 1.9%, Col8A2 10.1%, B3GNT7 18%, and CHST6 11.7%, when compared to CSK from same donor cornea) (all P < 0.05, Mann-Whitney U test), and Thy1 was upregulated by 3.1 ± 1.3-fold (Fig. 1B). Flow cytometry showed the expression of ALDH3A1 in the majority of cultivated CSKs (99.7%), in contrast to approximately 0.1% in SF culture (Fig. 1C). 

Intrastromal Injection of Human CSK or SF to Normal Rat Corneas

Effect of Injection Needle Size on Cell Viability

Cultivated human keratocytes suspended in PBS at 10⁴/μL were delivered through needles (sizes 27, 30, 31, and 33 G) at a rate of 2 μL per second onto a collagen I–coated surface to reproduce the cell injection in vivo. After 6 hours for cell adhesion, a calcein AM assay was performed and results showed that the cells passed through 27 G (lumen diameter 83 μm) (reference from Birmingham gauge standards on hypodermic needles; Med Tube Tech, Inc., Royersford, PA, USA), 30 G (lumen diameter 63 μm), and 31 G (lumen diameter 53 μm) had similar viability compared to cells delivered using a P10 pipette (all had >96% viability). There was a significant reduction of cell viability when cells passed through a 33-G needle (lumen diameter 45 μm) (24 ± 11%) (P < 0.05, Mann-Whitney U test) (Supplementary Fig. S2). This indicated that viable cell delivery required the use of 31-G needle or greater luminal size. 

Intrastromal Injection

Cultivated human CSKs and SFs were intrastromally injected to normal rat corneas (n = 52) at a dose of 2 × 10⁴ cells suspended in 2 μL PBS, respectively. Immediately after injection, a transient bleb was formed in the corneal stroma. This was shown as greater corneal thickness under AS-OCT at the first 1 to 3 days post injection (Fig. 2B). Hence, in order to avoid unclear images due to corneal edema caused by the injection fluid, we performed the first ocular examination at 3 days after injection. 

Corneal Clarity After Human CSK or SF Injection

Slit-lamp micrographs showed that the injection wounds on the corneal surface were invisible after 1 week, without any surface irregularities. Both CSK- and PBS-injected corneas were stable (Fig. 2A). At 3 and 4 weeks post injection, four out of eight CSK-injected corneas (another seven were collected in the first 2 weeks) displayed only mild haze, which was barely visible under direct and diffuse illumination, and no obscured visibility to the underlying iris tissue. The other four CSK-injected corneas were optically clear (Fig. 2C). In the PBS-injected group, only one out of five corneas had mild haze at 3 and 4 weeks post injection. This was significantly fewer than CSK-injected corneas (P = 0.001, χ² test), and this could be due to the xenogeneic cell transplantation. On the other hand, moderate to severe haze was formed together with corneal neovascularization (CNV) in 11 out of 14 SF-injected corneas at 3 and 4 weeks post injection (P = 4 × 10⁻⁵, compared to CSK-injected corneas, χ² test) (Fig. 2C). Greater corneal haze scores were shown in SF corneas (Fig. 3C). The mean CCT gradually decreased after injection but remained at levels significantly greater than in CSK- and PBS-injected corneas from the second week onward (P < 0.05, Mann-Whitney U test) (Fig. 2B). Using multiple comparison with Kruskal-Wallis and Dunn post hoc tests, the temporal CCT changes of SF-injected corneas were significantly different from those of the CSK- and PBS-injected corneas. 

Figure 3

 

(A) In vivo confocal micrographs of rat corneas before and after injection (weeks 2 and 4). Horizontal plane at the injection region in anterior stroma was imaged. Cells with high intensity of reflectivity were detected after CSK and SF injections. PBS-injected corneas had generally fewer reflective nuclei, which were the rat stromal cells. (B) Bar graph showing the mean relative reflectivity levels of five confocal planes of anterior stroma (30- to 50-μm depth from Bowman's layer) in each cornea at pre- and postinjection time points. Five corneas from each injection group were used for the reflectivity measurement. Error bars represent standard deviations. *P < 0.05 comparing among week 4 post injection and **P < 0.05 comparing between preinjection and week 4 post SF injection (Mann-Whitney U test). (C) Clinical grading of corneal clarity on slit-lamp photographs (Fig. 2A) over 4 weeks post injection showed significant haze induction in SF corneas when compared to CSK- or PBS-injected corneas. The number of corneas in each group is indicated in Figure 2B, and error bars represent SD. *P < 0.05 comparing between SF- and CSK-injected corneas (Mann-Whitney U test).

Stromal Changes After Human CSK or SF Injection

In vivo confocal micrographs showed the appearance of highly reflective nuclei and cell processes at the CSK injection plane (Fig. 3A). The stromal reflectance after CSK injection was similar to that prior to injection. In SF-injected corneas, extracellular haze with greater reflectivity was observed among the reflective nuclei. This was likely to be a mix of random ECM components and cell deposition. The reflectivity level was quantified with ImageJ, and the mean intensity levels of CSK- and PBS-injected corneas were approximately 40 pixels while SF-corneas showed increasing reflectivity up to ∼60 pixels at the fourth week post injection (Fig. 3B). There was a significant difference between SF- and CSK-injected corneas at the fourth week as well as in SF corneas between preinjection and the fourth week post injection (P < 0.05, Mann-Whitney U test). 

Marker Expression

To assess if human CSKs maintained keratocyte phenotype after injection into rat stroma, we examined the expression of human KSPGs and collagen I (Col1) by immunostaining using anti-human Kera, Lum, and Col1 antibodies, which were nonreactive to rat antigens. Immunofluorescence showed that human Kera, Lum, and Col1 were detected up to 4 weeks post injection and the signals were distributed around the injected cells (labelled by human-specific HuNu antibody) (Fig. 4A). There was negligible expression of SF markers (Thy1, tenascin-C [TNC], fibronectin [FN]) and myofibroblast marker (αSMA) (Fig. 4B). In contrast, SF-injected corneas had negligible immunoreactivity for human Kera, Lum, and Col1, and they positively expressed SF markers along the injection plane (Figs. 4A, 4B, white arrowhead). The SF marker expression together with αSMA was detected inside rat stroma at the fourth week post injection, even the HuNu-labelled human SFs became indiscernible. The region having positive αSMA and SF marker signals (marked by white brackets in Fig. 4) corresponded to the dense haze appearing under slit-lamp biomicroscopy (Fig. 2C). 

Figure 4

 

Expression of CSK, SF, and inflammatory markers in rat stroma at 2 and 4 weeks after cell injection. (A) CSK markers (Kera, keratocan; Lum, lumican; Col1, collagen I) were expressed predominantly along the CSK-injected region (marked by white arrowhead) as indicated by human-specific HuNu labelling. The staining was absent for SF- and PBS-injected corneas. (B) SF markers (Thy1; tenascin C, TNC; fibronectin, FN) were expressed in SF-injected corneas and the intensity of staining appeared to be more diffuse at 4 weeks compared to 2 weeks after injection. The myofibroblast marker (αSMA) became detectable at the fourth week post injection (region of dense haze illustrated under slit-lamp biomicroscopy marked by white bracket). In contrast, SF and myofibroblast markers were absent in CSK- and PBS-injected corneas. (C) The expression of inflammatory markers (MMP2, CD45) was stronger in SF-injected corneas when compared to the other two groups. The human cells were distinguished by human-specific HuNu staining. PBS-injected corneas were used as control. Nuclei were counterstained using DAPI (blue). Scale bar: 100 μm.

Cell Stability After Injection

Cell Density Changes After Single Injection

We performed a time-lapse study to track the injected cell changes inside the recipient stroma. Immediately after injection of Molday ION–labelled cells, all corneas showed similarly distinct fluorescent signal under FA mode by Spectralis HRA, indicating the injected cell distribution. Signal quantification showed that the label intensities were maintained above 80% in the first 2 weeks as compared to time 0 (set as 100%) (Figs. 5A, 5B). They were then gradually decreased to approximately 50% level at the third week and ∼20% at the sixth week post injection. Similar reduction of signal intensities was observed for two different cell dosages (2 × 10⁴ and 5 × 10⁴ cells per injection) (Fig. 5B). Significant reduction of label intensities was noted after 3 weeks when compared to time 0 (P < 0.05, Mann-Whitney U test). While the ION intensity reduction could be due to signal fading over time, we detected that the ION-labelled injected cells underwent apoptosis as revealed by the TUNEL staining. Some injected human cells with cytoplasmic ION fluorescence labels had nuclear TUNEL signal, indicating the occurrence of apoptosis (Fig. 5C). 

Figure 5

 

Time-lapse cell intensity changes after injection. (A) Intrastromal injection of Molday ION–labelled cells (2 × 10⁴ per injection) and weekly imaging using Spectralis HRA with excitation under FA model. Representative images with good cell focusing were selected from a minimum of 50 consecutive frames taken along the entire stromal depth at each examination time point. (B) Graph showing a time-dependent reduction of label intensities (mean percentages of time 0 level) for both injection dosages (2 × 10⁴ and 5 × 10⁴ cells) until 6 weeks post injection. Approximately ∼50% label intensity remained at around 3 weeks post injection. *P < 0.05, compared to label intensity measured immediately after injection (week 0) (Mann-Whitney U test). (C) TUNEL signal (red) was detected to colocalize with Molday ION EverGreen–labelled cells (green). Sections pretreated with DNase I and incubated with complete TUNEL reagent mix served as positive control whereas similar sections incubated with TUNEL reagent mix without TdT enzyme served as negative control. Nuclei were counterstained using DAPI (blue). Scale bar: 100 μm.

Repeated Cell Injection

To test if the injected cell level could be restored by repeated cell injection, we performed the first intrastromal injection of ION-labelled cells (2 × 10⁴ cells in 2 μL PBS) at day 0 and the second injection after 4 weeks (n = 5 rats). Weekly Spectralis HRA examination and label intensity measurement were conducted for a total of 8 weeks. Similarly to a single injection, the label intensity from the first injection gradually dropped to 37 ± 12% at week 4 (Fig. 6A). After the second injection, the intensity level returned to 126 ± 53% at week 6, from which it dropped again gradually. At week 8, the label intensity stayed at 66 ± 5%. Alternatively, a single injection gave 9.8 ± 0.7% label intensity at the eighth week (n = 5 rats). Immunostaining of human-specific CSK markers for rat corneas at week 9 after double injection showed the expression of human Kera and Lum close to the injected human CSKs (HuNu positive) (Fig. 6B). 

Figure 6

 

Time-lapse cell intensity changes after repeated cell injection. (A) Intrastromal injection of Molday ION–labelled cells was performed at week 0 and week 5, followed by weekly examination using Spectralis HRA. Representative images with good cell focusing were selected at each examination time point. Graph showing label intensity changes for single and double injection groups (each had four corneas) until the eighth week. (B) Immunofluorescence showed the expression of CSK markers (lumican, keratocan) at the ninth week after the reinjection of cells at week 5. Right column shows the magnified image indicated by the inset. The human cells were ION labelled (green fluorescence). Scale bar: 200 μm.

Integration of Injected Cells in Recipient Stroma

A TEM study of rat corneas harvested at 4 weeks after injection showed that the Molday ION-Fe conjugate–labelled cells contained the cytoplasmic inclusions with electron-dense Fe nanoparticles (arrows in Supplementary Fig. S3C). In contrast, native rat stromal cells were devoid of these particles and their cell processes extended in between the stromal lamellae (Supplementary Fig. S3A). The injected labelled cells were also located in proximity to the collagen fibrils (Supplementary Fig. S3B), but narrow gaps were seen between the cell processes and collagen fibrils, indicating that the cell–ECM interaction was not as close as that for the native cells. The fibril distribution adjacent to the injected cells was regularly aligned without distortion. At higher magnification, some mild electron-dense materials were observed between the injected cells and stromal ECM, which suggested that the cell might be involved in stromal remodeling (Supplementary Fig. S3D). 

Recipient Stromal Reaction After Cell Injection

Stromal Inflammation

To address whether cultivated CSKs could be applied for clinical therapy, we examined the host inflammatory response at different time points following injection using anti-rodent MMP2 and CD45 antibodies. We observed negligible metalloproteinase 2 (MMP2) expression at 2 and 4 weeks post CSK injection, similar to observations in PBS-injected corneas (Fig. 4C). However, SF-injected corneas had upregulated MMP2, which might be related to the appearance of neovascularization and ECM remodeling.^36,37 The expression of CD45 (marker of T-lymphocytes and myeloid cells^38,39) was negligible in CSK-, SF-, and PBS-injected rat corneas (Fig. 4C), indicating the absence of graft rejection. 

Effect on Rat Stromal Cells

We performed TUNEL analysis at different time points after single injection to examine the effect of injection on rat stromal resident cell viability. Molday ION–labelled human cells were distinguished from rat cells without ION labels (Supplementary Fig. S4A). Quantification of TUNEL-positive cells in ION-labelled human cells showed that the apoptosis rate post injection was 14.9 ± 0.8% at the second week, 12.8 ± 0.6% at the third week, 12 ± 2.8% at the fourth week, and 10.8 ± 0.5% at the fifth week (Supplementary Fig. S4B). These were significantly higher rates than the apoptosis rate of rat stromal cells, which were maintained at less than 3% (2.1 ± 0.1% for week 2; 3 ± 0.2% for week 3; 2.2 ± 0.4% for week 4; and 1.5 ± 0.1% for week 5). The quantification did not account for all cells inside the stroma as the effect of injection might not be spread to the whole cornea. Indeed, it was conducted within the stromal region measured at 200-μm distance from the ION-labelled human cells at the most peripheral location (Supplementary Fig. S4A). The PBS-injected corneas had 2.8 ± 0.3% cells that had undergone apoptosis inside the stroma (at the second week post injection), and this was similar to what was seen in normal rat corneas (2 ± 0.4%) (Supplementary Fig. S4B). 

Stromal Restoration

We also monitored the stromal changes post injection by examining the stromal pattern of keratan sulfate (KS), which links to proteoglycans. Compared to normal rat corneas showing regular lamellar KS pattern (Supplementary Fig. S5D), corneas after ION-labelled human cell injection had distorted KS lamellae in the first 2 weeks with stromal gaps occurring near to the injected cells (Supplementary Figs. S5A, S5B). After 5 weeks, the stromal KS lamellar pattern was virtually restored and the injected cells, though with reduced number, were closely interactive with the stromal lamellae with a regular KS pattern (Supplementary Fig. S5C) as in normal rat corneas. 

Therapeutic Effect of CSK Injection on a Rat Corneal Opacity Model

Restoration of Corneal Transparency After CSK Injection

To explore the possibility that CSK transplantation can be used as a treatment regimen for corneal opacities, we intrastromally injected human CSKs (4 × 10⁴ cells in 2 μL PBS) to rat eyes (n = 5) with corneal opacities induced by irrPTK. As controls, other groups of irrPTK-injured corneas received PBS injection (n = 6) and no injection (n = 5), respectively. The injection procedure was performed at 1 week after irrPTK, and the injected cell dosage was similar to that previously reported.⁴⁰ Under slit-lamp biomicroscopy with direct and retro-illumination, mild to moderate corneal haze with partial obscuration to the underlying iris tissues was observed at 1 week after irrPTK and continued to manifest CNV in the noninjected group (Fig. 7A). Following injection of human CSKs to the haze region inside the anterior stroma, three out of five corneas had improved transparency with reduced opacities at 3 weeks post injection (Fig. 7A). This matched with the commensurate reduction of stromal reflectivity at approximately 30- to 50-μm depth from the Bowman's layer under in vivo confocal microscopy (Fig. 8A). The detection of hyperreflective nuclei could represent the repopulation by injected CSKs, which remained viable up to 3 weeks (Figs. 8A, e, f). Some cells displayed dendritic morphology at week 3 compared to week 2 post injection, indicating that they had adapted to the new stromal environment. The stromal reflectance at week 3 (Figs. 8A, f) was similar to that before cell injection (Figs. 8A, d), indicating that the progression of stromal haze was suppressed. For the injured corneas injected with PBS (n = 6), only one showed reduced opacities at week 3; the rest had haze progression and were similar to the noninjected injured corneas (Fig. 7A). χ² test for independence showed significant recovery of corneal clarity in CSK-injected corneas (three of five) than seen in PBS-injected corneas (one of six) (P = 0.025). Under confocal microscopy, PBS-injected corneas had increased stromal reflectance over time when compared to that before injection (Figs. 8A, g–i) and the stroma remained acellular. Similar progression was observed in noninjected injured corneas (Figs. 8A, a–c). The percentage of reflectivity was progressively increased during follow-ups for PBS-injected and noninjected corneas (Fig. 8C). Compared to corneas at week 3 post injection, CSK injection resulted in significantly reduced stromal reflectivity than in the PBS and noninjection groups (P < 0.05, Mann-Whitney U test). 

Figure 7

 

Slit-lamp microscopy and anterior segment optical coherence tomography (AS-OCT) images of rat irrPTK-injured corneas with early haze formation followed by intrastromal injection of human CSK and PBS and noninjection. (A) Slit-lamp photographs of injured corneas before injection and at weeks 2 and 3 after injection. The CSK-injected corneas had improved clarity at week 3 whereas PBS-injected and noninjected corneas were hazy and neovascularized. (B) Temporal AS-OCT images of rat irrPTK-injured corneas show thinning of corneas after CSK injection but not in PBS-injected and noninjected corneas. (C) Percentage changes of mean central corneal thickness throughout the examination time points (including post PTK and post injection) compared to preopeative level. CCT of CSK-injected corneas (n = 5) at weeks 2 to 3 was significantly lower than than for PBS-injected and noninjected corneas. Error bars in line graph represent SD. *P < 0.05 denotes statistical significance between CSK-injected (n = 5) and PBS-injected corneas (n = 6).

Figure 8

 

In vivo confocal micrographs of rat irrPTK-injured corneas with intrastromal injection of human CSK and PBS and noninjection. (A) Micrographs were taken at the horizontal plane along the injection level (approximately 30–50 μm depth from Bowman's layer) before injection and at weeks 2 and 3 after injection. A relatively higher light reflective layer was observed after irrPTK. Cell repopulation was observed at weeks 2 and 3 after CSK injection, and this was accompanied by a reduced stromal reflection. PBS-injected and noninjected injured corneas had similar reflection throughout follow-up time points. (B) Dendritic-shaped stromal cells with less reflective stromal ECM in normal rat stroma. (C) Mean stromal reflectivity levels before injection and at weeks 2 and 3 after injection. Error bars represent SD. Bold horizontal line represents the reflectivity level of normal rat corneal stroma. *P < 0.05, Mann-Whitney U test.

Restoration of Corneal Thickness After CSK Injection

AS-OCT scans revealed thick and opaque irrPTK-injured corneas before injection (Fig. 7B). The corneal thickness was restored after CSK injection. The percentage changes of CCT with reference to the preoperative level are presented in a line graph (Fig. 7C). CSK-injected corneas were generally thinner. The mean CCT at week 3 post injection was similar to the preoperative level before irrPTK and was significantly lower than in PBS-injected and noninjected corneas (P < 0.05, Mann-Whitney U test). 

Discussion

In this study, we demonstrated that intrastromal injection of cultivated human CSKs to rat corneas was safe and did not induce haze/opacity formation or stromal structural changes. The injected CSKs behaved like native keratocytes in synthesizing and depositing human KSPGs and Col1 in rat stroma and functionally maintained corneal clarity and stability. There was negligible conversion to human SFs or detection of any SF-related phenotypes. The therapeutic efficacy was demonstrated in a preclinical rat model of corneal opacities using the intrastromal CSK injection approach. Corneal clarity and thickness were restored by CSK injection following irrPTK, a rodent corneal opacity model, whereas injured corneas receiving PBS injection or no injection remained significantly opaque and thicker. Hence, we showed that the transplantation of cultivated CSKs via intrastromal injection could be a safe and feasible tool for potential cell therapy for corneal opacity disorders. 

Corneal stromal reconstruction is targeted to restore corneal transparency and integrity. This strategy may involve either tissue replacement, that is, a transplant procedure, or tissue modulation by replacing dysfunctional stromal keratocytes with functional CSKs to reduce scar formation and progression.¹⁹ Hence it is imperative to use appropriate cells in order to deliver the correct therapeutic outcome. There have been various attempts at stromal cell replacement. Corneal SFs, with their proliferative nature under ex vivo conditions, are unsuitable for stromal cell replacement since they may induce corneal haze. The cells do not produce KSPGs, and there is significant infiltration of inflammatory cells (CD45⁺ and CD3 T-cells).^8,27,28 Our results also showed that SF injection induced CNV, which was significantly worse than with keratocyte injection. Human corneal stromal stem cells (CSSCs) have been shown to generate functional CSKs and produce stroma-like ECM through natural cell doubling.²⁸ However, some mature cells were detected as expressing epithelial-type keratins, giving the possibility of producing CSKs with altered features.^41,42 In vitro differentiation of human embryonic stem cells into CSK-like cells via neural crest progenitors has also been reported,⁴³ but a number of issues, such as induction efficiency and control of cell purity and risk of tumorigenesis, need to be managed before any clinical application. Human bone marrow– and adipose–derived mesenchymal stem cells (MSCs), dental pulp stem cells, and periodontal ligament stem cells, when intrastromally injected into animal corneas, have been shown to assume CSK phenotypes,^33,44–48 suggesting that these cell types could potentially be used in stromal cell therapy. However, natural differentiation of stem cells in a particular niche cannot achieve specific cell fate determination; hence not all differentiated cells can display a consistent profile of CSK features.^49,50 Despite this, a phase I clinical study of five patients with keratoconus underwent intrastromal injection of autologous adipose MSCs. The cells produced new collagen and improved CCT and visual function with stable refraction, intraocular pressure, and endothelial cell density.⁵¹ Cell purity, long-term stability, and adverse effects from nonspecific differentiation have to be unravelled prior to further translational use. Due to these concerns, we assessed if cultivated CSKs per se are the more appropriate functional and stable cell type for the replacement of dysfunctional keratocytes and corneal stromal reconstruction, which could contribute to therapeutic benefits in correcting or reducing corneal opacities. 

In normal corneas, CSKs are quiescent (they are mainly in G0 phase or have a very slow turnover rate) and biosynthesize and deposit collagens and KSPGs (Lum, Kera, and mimecan) for ECM assembly and modeling, and enzymes (such as collagenases) for matrix turnover.⁵ These activities maintain the relative position and alignment of collagen fibrils and regulate fibril growth and interfibrillar spacing, which are important for corneal strength and transparency.^2,52 CSKs create microdomains to regulate stromal matrix assembly to achieve overall and long-range regulation of stromal integrity. In addition, they express stromal crystallins, including aldehyde dehydrogenases (ALDH1A1, 3A1), α-enolase, lactic dehydrogenase, and transketolase, which match the refractive indices between cells and ECM, contributing to corneal transparency.⁵³ Ex vivo propagation of CSKs expressing proper phenotypes has been challenging due to their irreversible transition to SFs under growth factor and serum-supplemented conditions.^5,54 We have described an approach to propagate “activated keratocytes” using a robust ERI protocol added with a low level of serum (∼1%), and the culture has negligible fibroblast phenotype or features (absence of CD90/Thy1, αSMA, and collagen contractibility).²⁰ The cells proliferate at a mitotic index ∼2.4%, and this allows an expansion of 3 to 5 × 10⁴ initial cells to ∼10⁷ CSKs without fibroblast contamination. When returned to serum-free ERI condition, they revert to quiescent CSKs with a re-expression of known CSK markers (Lum, Kera, ALDH1A1, ALDH3A1, Col8A2, B3GNT7, CHST6). In the present study of xenogenic cell transplantation, rat corneas receiving human CSKs showed stromal expression of human Lum, Kera, and Col1 at the fourth week post injection. In addition, they were devoid of SF marker expression (Thy1, αSMA, and tenascin-C), indicating that the injected CSKs had minimal transition to express SF or myofibroblast features. This resulted in corneal clarity in 73% of rat corneas, no significant difference from the PBS-injected group (P > 0.05). The remaining corneas had only mild haze, and this could be due to the minute amount of SFs present in the injected cell suspension. Assayed by flow cytometry, <1% cultivated CSKs were CD90/Thy1 positive (fibroblast gene) even though the rest of cells expressed ALDH3A1. Hence, the cell purity could be improved through the negative sorting of CD90/Thy1. Further quality control could be possible through cell surface CD34 sorting, which will recognize quiescent CSKs but not the activated stromal cells and SFs.^42,55 In contrast, human SF-injected corneas showed increased CNV and significantly more haze and opacities, with only 36% of corneas remaining clear (P = 0.022, compared to CSK-injected corneas). 

We evaluated the functionality of the cultivated human CSKs in a rescue scenario using a rat model of early corneal opacities. In a previous study, human limbal stromal stem cells expanded in culture were suspended in fibrinogen solution and applied to an exposed stromal wound immediately after corneal epithelial debridement, and the authors were able to regenerate transparent native stromal tissue without the expression of fibrotic proteins, hence suppressing scar tissue formation.⁴⁰ However, the mechanism of this regenerative effect remains unknown, and the influence of paracrine signaling on recipient stromal cells repopulating the wound region has to be further investigated. Also, the therapeutic effect on existing scars or corneal haze is still an open question. Our study provided a major advance to demonstrate the rescue effect on early corneal haze. The ability of injected CSKs to restore corneal clarity and thickness by replacing the repair-type ECM with native components (e.g., KSPGs, stromal crystallins) close to the normal stroma is an exciting finding that clearly indicates the important potential for use of these cells in clinical situations to treat human corneal haze. We hypothesize that this new treatment could modulate or reduce the density of opacities. Particularly if the opacities are present on the visual axis, they could be reduced to a level at which visual rehabilitation could be attainable by nonsurgical means, for example, rigid gas-permeable lenses. Hence the need for corneal grafting would be greatly reduced, and the donor pool could be used for treating more chronic dense and developed corneal scars. Future work on injecting CSKs to animal models with mature scar tissues or defective CSKs (such as in transgenic lumican null mice) will be important to elucidate the therapeutic effects on scar resolution and the efficacy of stromal reconstruction. 

The intrastromal cell injection approach is therapeutically practical. It is a relatively simple and minimally invasive procedure, supported by a number of previous studies.^28,47,48 The cells were administered via intrastromal injection, and this method induced only minimal disturbance to the cornea. The injection is painless, easily performed by any ophthalmic surgeon in general clinics, and results in fast and simple rehabilitation.¹⁹ The bleb created at the time of injecting cell suspension might transiently disturb the recipient's vision, but this disappears within hours to days. Previous reports have shown that the injected cells initially stayed at the site of injection and gradually migrated to the corneal periphery, and, in 4 to 6 weeks, spread homogeneously in the entire cornea.⁴⁷ In our study, some rat corneas showed human cells located closer to the peripheral stroma. Hence, the histologic sections displayed an altered stromal cell density (revealed by DAPI staining) when compared to the central stroma. This uncontrolled localization could be a limitation of cell injection if the cells are required to target a specific haze/opacity region. A guided injection would be desirable to deliver cells to the defined locations where they could fully fulfill their function in resolving haze/opacity. In addition, histologic sections of SF-injected corneas had variable cell densities (Fig. 4), which might be related to the haze progression as shown in Figure 2. Our previous proteomics and multiplex cytokine studies had shown the active secretion of human SFs to induce cell proliferation, neurogenesis, healing response, and so on,²¹ and these could relate to stromal resident cell activation and proliferation. In contrast, corneas with more consistent stromal thickness and less variable stromal cell densities were seen after injection of more quiescent CSKs, resulting in the corneal stability. Under TEM, the injected keratocytes interacted closely with the host stromal matrix. The collagen lamellar structure adjacent to the injected cells was nearly unchanged, with regular collagen fibril arrangement. Hence, intrastromal injection is a feasible approach to deliver therapeutically functional CSKs to the target corneas. 

The development of any cell-based therapeutics aiming toward restoring defective tissue function will require the long-term viability of transplanted cells with desired bona fide cellular functions. Several studies have documented injected cells over time after xenogeneic transplantion to murine corneas. DiO-labelled human DPSCs were visualized at the fifth week post injection,⁴⁸ whereas human bone marrow MSCs were detected after 6 weeks, at reduced intensity.⁴⁷ Another study injecting human CSSCs showed cell survival up to 12 weeks in murine stroma.²⁸ Despite these observations, the quantity of cells remaining and the cause of tracing signal loss are not known. In order to quantify the cell loss, we performed an in vivo time-lapse study to critically assess the stability of injected cells inside the corneal stroma. The Molday ION label intensity was stable at ∼90% level (reference to time 0 immediately after injection) for the first 2 weeks, then decreased time-dependently until it was ∼20% at 6 weeks post injection. This cell loss could be due to signal fading over time as described above. Our TUNEL experiment also indicated that the injected cells underwent apoptosis. In order to sustain the functional cell density, we repeated the CSK injection (from the same donor) at week 5, and this restored the cell density and achieved longer-term maintenance of human cells, with a sustained deposition of human KSPGs. In addition, the recipient stromal cells seemed not to be affected by the cell injection as they had a significantly lower apoptosis rate. And the stromal architecture (KS pattern) was shown to be gradually restored after injection. Keratan sulfate links to ∼50% of all stromal glycosaminoglycans forming proteoglycans (Lum, Kera, mimecan) and together with dermatan sulfate proteoglycan (decorin) and collagen I, V, VI, and XII constitutes the ground substance of corneal stroma.^56,57 After all, corneal transparency and stromal thickness after repeated injection appeared normal. Last but not least, to improve treatment efficacy, the number of cells per injection and the number of injections could be adjusted according to the disease type and severity, patient's condition, and target outcome, such as prevention of haze/opacity progression. 

Though our study showed promising results regarding safety and therapeutic efficacy for early corneal haze, whether this cell-based approach using ex vivo propagated CSKs could be applicable to treat existing/mature or clinical corneal opacities needs to be fully established. Based on our results, CSKs propagated for a single donor could represent a reliable cell source for stromal reconstruction. Future studies applying these cells to different animal models with early or developed haze/scar and/or dysfunctional keratocytes will elucidate the therapeutic effects and applicability to managing corneal opacities and stromal disorders. 

Acknowledgments

The authors thank Lion Eye Institute for Transplant and Research (Tampa, FL, USA) for their assistance with procurement of research-grade donor corneas; Singapore Eye Bank, Singapore National Eye Centre, for help in amnion collection (obtaining consent, tissue processing); and SingHealth Advanced Bioimaging Core Facilities for confocal and transmission electron microscopies. 

Supported by SingHealth Foundation Transition Grant SHF/GF585P/2014 and the Singapore National Research Foundation under its Translational and Clinical Research (TCR) program (NMRC/TCR/008-SERI/2013), National Medical Research Council, Ministry of Health, Singapore. 

Disclosure: G.H.-F. Yam, None; M. Fuest, None; N.Z.B.M. Yusoff, None; T.-W. Goh, None; F. Bandeira, None; M. Setiawan, None; X.-Y. Seah, None; N.-C. Lwin, None; T.P. Stanzel, None; H.-S. Ong, None; J.S. Mehta, None