Abstract
purpose. To study keratocyte activation and cellular apoptosis in transplanted human corneas during the early postoperative period.
methods. Ten human donor corneas preserved for 6 days at 4°C were transplanted into the eyes of 10 adult cats. After confocal and specular microscopy in vivo 1 week after keratoplasty, the cats were killed, and the fixed corneas were examined by TUNEL assay and by scanning (SEM) and transmission electron microscopy (TEM).
results. Abnormal keratocytes, in which portions of cell bodies and processes as well as nuclei were visible, were present in all corneas and occupied the anterior 16 to 562 μm of the stroma. By TEM in the same corneas, these abnormalities represented keratocytes that were activated to a repair phenotype. Only 0% to 1% of all corneal cells were apoptotic by TUNEL assay, except for the donor keratocytes near the wound, where 7% were apoptotic. The midstromal keratocyte density was decreased at 13,936 ± 5,910 cells/mm3 (mean ± SD), and the endothelial cell density was 2,298 ± 688 cells/mm2, representing an endothelial cell loss of 7% ± 16%.
conclusions. Substantial keratocyte activation and low levels of cellular apoptosis occur 1 week after human corneal transplantation. The human-to-cat xenograft model of corneal transplantation demonstrated endothelial cell loss and other clinical findings similar to human allografts. The model will be useful for preclinical testing of new methods of long-term corneal preservation and of donor endothelial cell augmentation, as well as the study of human corneal wound healing and keratocyte replacement during the early postoperative period.
The early wound-healing response of human corneal cells to transplantation has not been studied because of the inability to obtain histologic specimens early after clinical transplantation. Cellular responses such as keratocyte activation and apoptosis after surgical trauma have been studied in the corneas of mice,
1 rats,
2 and rabbits,
3 4 but rarely in human corneas,
5 where the responses may be attenuated, just as the endothelial response to injury is.
6 7 In particular, keratocyte activation and regeneration appear to be decreased in human corneas after photorefractive keratectomy
8 compared with the responses in rabbits. An experimental xenograft model that used human donor tissue would allow the histologic study of the human corneas after transplantation. By studying the transplanted human corneas after surgery, investigators could determine the extent of cellular apoptosis, keratocyte activation, and origin (donor versus recipient) of the cellular response.
Such a model would also be useful to test corneal preservation methods, which continue to be developed in attempts to increase corneal viability and extend storage times before transplantation.
9 10 We have designed and tested a corneal perfusion system that has been used for human cryopreservation studies in vitro.
11 12 The perfusion system, however, did not perfectly match the conditions experienced by preserved corneas in vivo after transplantation. An experimental method for evaluating transplanted human corneas in vivo would be valuable as a final validation for techniques of corneal preservation before undertaking clinical trials.
Thus, we designed a xenograft model to evaluate the short-term effects of both corneal transplantation and corneal preservation on the cells of the human cornea. Immune rejection should be expected with transplants across species lines, and this would affect the results of transplantation. Fortunately, corneal xenografts, unlike vascularized organs, are not rejected hyperacutely.
13 Evidence of immune rejection was not present until 8 days in rat-to-guinea pig corneal xenografts
14 and for more than 4 months in human-to-monkey corneal xenografts.
15 We therefore limited our evaluation to 7 days after surgery to eliminate rejection as a confounding variable.
A xenograft model using human donor corneas eliminates the interspecies differences in corneal preservation that may occur when animal corneas are used to test methods intended for human corneal preservation.
16 17 Such a model requires animals with eyes large enough and corneas thick enough to accommodate transplanted human corneas. It also requires animals that have limited endothelial regeneration capacity, to eliminate the possibility of recipient endothelium’s regenerating and replacing dead or missing donor endothelium on the graft. Because primates are expensive and inappropriate for short-term studies, adult cats are reasonable recipients, because their corneas are large enough to accommodate transplanted human corneas, and they, like primates, have a limited endothelial regenerative capacity.
6 7
We describe a human-to-cat xenograft model and report 1-week postoperative results of corneal thickness, endothelial cell survival, cellular apoptosis, and keratocyte activation in 10 human-to-cat penetrating keratoplasties after standard preservation at 4°C.
This study adhered to both the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and to the Declaration of Helsinki.
Ten human cadaveric eyes from 10 donors aged 60 ± 22 years (mean ± SD; range, 13–89 years) were enucleated 5 ± 2 hours after death (range, 1.5–8 hours), and stored in moist chambers at 4°C for 3 ± 2 hours (range, 1–7 hours). The corneas were then aseptically excised and stored at 4°C in preservative (Optisol-GS; Chiron Vision, Claremont, CA) for 6 ± 2 days (range, 3–8 days) until transplantation. A specular microscope (LSM 2100C; BioOptics Inc., Arlington, MA) was used to examine each donor cornea and obtain baseline endothelial photographs at the beginning of the storage period and corneal thicknesses at the end of the storage period, before transplantation.
Penetrating keratoplasty was performed with aseptic technique in the 10 recipient cats, ages 1.5 to 9 years, by the method of Bahn et al.
18 The animals were fasted for 12 hours and received aspirin (10 mg/kg) orally 8 hours before keratoplasty to prevent clot formation in the anterior chamber during surgery. We induced anesthesia with intramuscular injection of 0.5 mg/kg xylazine and 20 mg/kg ketamine hydrochloride. After intubation, isoflurane gas was used as needed to supplement the anesthesia. We administered 0.06 mg/kg atropine sulfate subcutaneously to reduce secretions and 200 U/kg heparin subcutaneously to prevent fibrin formation in the aqueous humor. Standard microsurgical instruments and an operating microscope were used. We found it essential to position the cat’s head precisely and to retract the nictitating membrane. An 8.0-mm trephine incision was made in the cat recipient cornea to receive an 8.25-mm human donor button punched from the endothelial side. We injected into the anterior chamber both porcine heparin (200 U in preservative-free balanced salt solution) to prevent fibrin formation and 1% sodium hyaluronate to minimize endothelial trauma. The grafts were fixed in place with interrupted 9-0 and 10-0 nylon sutures with buried knots. A subconjunctival injection of triamcinolone (40 mg in 1 mL) and 1% atropine drops were administered at the end of the surgical procedure. No topical medications were given after surgery.
Each cornea was examined with a penlight daily to check for graft clarity, signs of infection, wound dehiscence, inflammation, or other surgical complications. At the 1-week examination, all animals were anesthetized and their corneas were examined by using a slit lamp biomicroscope, a wide-field specular microscope (Keeler Instruments, Inc., Broomall, PA), and a confocal microscope (Tandem Scanning Confocal Microscope; Tandem Scanning Corporation, Reston, VA) with a ×24 (0.6 numeric aperture [NA]) objective lens. Confocal images, acquired from the center of the human graft by continuously advancing the optical section at a constant speed,
19 20 were directly digitized and stored in computer memory. Central graft thickness was measured by using ultrasonic pachometry (Model 1000; DGH Technology, Inc., Frazer, PA) and intraocular pressure was measured by pneumatonometry (model 30 Classic; Mentor O&O, Norwell, MA).
After the 1-week postoperative examinations, the cats were killed by administration of an overdose of pentobarbital sodium. The eyes were then enucleated and immersed in fixative (1% paraformaldehyde, 1% glutaraldehyde in 80 mM cacodylate buffer) while the anterior chambers were infused with fixative at a pressure of 15 mm Hg through a cannula inserted near the limbus. After fixation, the corneas were bisected, with half of each cornea being processed for scanning electron microscopy (SEM) and the other half being bisected radially, leaving a quarter of each cornea for transmission electron microscopy (TEM) and a quarter for assay by terminal deoxynucleotidyl transferase-mediated dUTP nick-end labeling (TUNEL). Corneas were processed for electron microscopy, and the endothelial surface of each cornea was examined with a scanning electron microscope (JSM-6400; JEOL Ltd., Tokyo, Japan). Keratocytes and endothelial cells of five corneas were examined with a transmission electron microscope (model 1200; JEOL Ltd., Tokyo, Japan).
The TUNEL assay was used to assess apoptosis quantitatively in corneal epithelium, stroma, and endothelium.
21 A quarter of each cornea was embedded in paraffin, and 5-μm-thick sections were mounted onto glass slides. The sections were deparaffinized by heating for 20 minutes at 60°C and washing twice for 5 minutes each in xylene. The slides were then transferred through the following solutions: 100% ethanol twice for 5 minutes each, 95% ethanol for 3 minutes, 70% ethanol for 3 minutes, distilled water for 5 minutes, and phosphate-buffered saline (PBS) for 5 minutes. Proteinase K (20 μg/mL; Roche Molecular Biochemicals, Indianapolis, IN) was applied for 20 minutes at room temperature. After two rinses with PBS, the sections were labeled with fluorescein isothiocyanate (FITC) according to the instructions for a commercial kit (In Situ Cell Death Detection Kit, Fluorescein; Roche Molecular Biochemicals). The slides were counterstained with 4′6′-diamidino-2-phenylondole dihydrochloride (DAPI), which binds to double-stranded DNA, thus staining all intact nuclei. Appropriate positive and negative controls were used.
Three sections per cornea were examined using a laser scanning confocal microscope (LSM 510; Carl Zeiss, Inc., Oberkochen, Germany). An argon-krypton laser with a 488-nm excitation wavelength and an argon ion laser with a 351- to 364-nm excitation wavelength were used. The sections were viewed and digitized through a ×10, 0.45 numeric aperture water-immersion objective lens (C-Apochromat; Carl Zeiss) and a 385- to 475-nm emission filter for DAPI and a 505- to 550-nm emission filter for FITC.
Images from the tandem scanning confocal microscope were evaluated for the presence of activated keratocytes (repair fibroblasts).
22 A keratocyte was considered to be activated when the cell body or cell processes, in addition to the nucleus, could be clearly seen.
Stromal keratocyte density was measured by manually counting the number of bright objects (nuclei) in 10 frames from one scan recorded with the camera in automatic-gain mode.
20 Only frames not obscured by movement blur were counted. Each frame was counted by two experienced observers who were masked to the corneal depth of the image, and the results were averaged. The stroma (defined as the frames between the first visible keratocyte and the first endothelial image) was apportioned into five layers proceeding from anterior to posterior: 0% to 10%, 10% to 33%, 34% to 66%, 67% to 90%, and 91% to 100% depth. Two frames were counted in each layer, one from each anteroposterior half. The first countable image was used for the frame in the anterior half of the first layer and the last countable image was used for the frame from the posterior half of the last layer. The thickness of each stromal layer was corrected for stromal depth to account for the nonlinear movement of the focal plane with depth of scan.
20
Three TUNEL assay digital images of each cornea obtained with the laser scanning confocal microscope were analyzed (KS-400 analysis system; Carl Zeiss) and the results were averaged. The epithelium, stroma, and endothelium in each image were manually traced to define each layer. TUNEL-positive nuclei stained with FITC and nuclei stained with DAPI were discriminated and counted in each layer. The percentage of apoptotic cells was estimated in the corneal epithelium, stroma, and endothelium in four regions: the donor cornea within 0.5 mm of the wound, the center of the donor cornea (>1 mm from the wound), the recipient cornea within 0.5 mm of the wound, and the periphery of the recipient cornea (>1 mm from the wound;
Fig. 1 ). Keratocyte density was estimated from the number of DAPI-stained nuclei per measurement area and expressed as cells per square millimeter full-thickness sagittal section for donor and recipient stroma in all four regions.
The specular endothelial photomicrographs of each cornea before and after transplantation were analyzed by digitizing the apices of 100 cells from each cornea with a corneal endothelial analysis system (BioOptics, Inc.) that calculated endothelial cell density, coefficient of variation of cell area, and percentage of hexagonal cells. The magnification of all the photographic systems were calibrated by photographing a micrometer slide. Endothelial cell loss was calculated as the decrease in endothelial cell density from baseline, expressed as a percentage of the baseline value.
Central corneal thickness, endothelial cell density, coefficient of variation of endothelial cell size, and the percentage of hexagonal endothelial cells before and after penetrating keratoplasty were compared using a paired
t-test when the data were normally distributed and a Wilcoxon signed rank test when they were not. The keratocyte density was compared among four defined regions using a one-factor repeated-measures analysis of variance. The percentage of apoptotic cells was compared among the four defined regions using the Friedman test,
23 because these data were not normal. Significant differences among the four regions in keratocyte density or percentage of apoptotic cells were investigated by adjusting for the multiple comparisons, using the Student-Newman-Keuls procedure.
24
After surgery, the activity levels of the cats were normal, without signs of visual deprivation. Slit lamp examinations 1 week after surgery
(Fig. 2) revealed intact corneal epithelium on seven of the grafts, small defects on two grafts, and a large persistent defect on one graft. The stroma was clear in six grafts and mildly edematous in four grafts. The mean central graft thickness was 0.56 ± 0.05 mm (±SD; range, 0.47–0.65). A bright endothelial specular reflection was seen from each graft. Inflammatory cells were noted on the posterior surface of some grafts. A fibrinous membrane was noted in all corneas along the wound margin projecting into the anterior chamber a variable amount, from minimal to moderate. The anterior chambers were deep; there were no severe inflammatory reactions observed or signs of infection or rejection. The intraocular pressure was 17 ± 6 mm Hg (range, 8–26).
Tandem scanning confocal microscopy of each central human donor cornea in vivo at 1 week after penetrating keratoplasty revealed epithelial cells on seven of the grafts that, subjectively, seemed larger than those on normal human corneas. A layer of atypical keratocytes, in which both cell bodies and cell processes, in addition to the nuclei, were visible, was seen in the anterior stroma of all cats
(Fig. 3) . Inflammatory cells may also have been present in this layer. The layer ranged in
z-depth from a minimum of 16 μm to a maximum of 562 μm (mean ± SD, 146 ± 159 μm; median, 108 μm), and extended posteriorly from the most anterior keratocyte layer. In the one graft with a large persistent epithelial defect, activated keratocytes were seen at all depths of the stroma (562 μm). In the middle and posterior third of the stroma in most grafts, cells had a more typical appearance, with bright nuclei in a relatively dark field
(Fig. 4) . Endothelial cells as well as bright scattered objects, presumed to be inflammatory cells, could be seen clearly on the posterior surface of the cornea.
We attempted to count the keratocytes in all layers (10 frames) of each cornea, but in many of the layers with activated keratocytes, individual nuclei could not be discerned well enough to count accurately. Therefore, we did not count cells in frames with activated keratocytes.
Table 1 presents the cell density from the five anteroposterior layers of the stroma for which two frames, one from the anterior half and one from the posterior half of each layer, could be accurately counted. Keratocyte densities in normal subjects are presented for comparison.
Endothelial photographs 1 week after penetrating keratoplasty showed a small decrease in endothelial cell density in the central graft by 7% ± 16% (range, −7%– 36% from baseline), which was not statistically significant
(Table 2) . The coefficient of variation of endothelial cell area was unchanged at 0.27, and the percentage of hexagonal endothelial cells decreased significantly, indicating increased pleomorphism.
The percentage of TUNEL-positive keratocytes in the human donor corneal stroma within 0.5 mm of the wound (10% ± 10%; median 7%), was significantly higher than in the other three regions (
Table 3 ,
Fig. 5 ). Conversely, the human donor keratocyte density was significantly lower near the wound than it was in the center. There were no significant differences in the percentages of TUNEL-positive epithelial or endothelial cells among the four regions. There were no statistically significant correlations between the mean percentage of apoptotic cells for epithelium, stroma, or endothelium and keratocyte density, endothelial cell density, endothelial cell loss, or corneal thickness. There were also no significant correlations between any of the above factors and donor age, death to enucleation time, moist chamber time, or 4°C storage time. No histologic evidence of graft rejection was found.
Examination of the posterior surface of each cornea by SEM showed intact donor and recipient endothelial monolayers. Occasionally, cells that appeared to be inflammatory cells were attached to the endothelium, although a few of them resembled apoptotic endothelial cells
(Fig. 6) . The human endothelial cell borders on the donor grafts were not as distinct as the cat endothelial cell borders on the recipient cornea in most specimens. A fibrinous membrane extended from the posterior cat stroma into the anterior chamber at the wound in each specimen and varied in length from less than 0.5 mm to 2 to 3 mm.
Examination by TEM of the stroma in a region where activated keratocytes had been seen with the confocal microscope showed keratocytes with dilated endoplasmic reticulum and prominent Golgi apparatus
(Fig. 7) . In deeper regions, where normal cell nuclei had been visible by confocal microscopy, normal keratocytes were seen
(Fig. 8) . In the peripheral graft stroma near the wound, occasional apoptotic keratocytes were seen
(Fig. 9) . Most of the endothelial cells appeared normal; a few inflammatory cells were attached to the posterior endothelial surface. Occasional polymorphonuclear cells were visible in the endothelium and stroma of some corneas.
In this study, we successfully developed a human-to-cat xenograft model. All human donor corneas maintained in preservative (Optisol-GS; Chiron) were clear with normal corneal thickness 1 week after penetrating keratoplasty. The epithelium, presumably from the cat, had healed over the grafts in a manner similar to human allografts and was intact by 1 week in most corneas. The keratocyte density 1 week after surgery was similar to that of human transplants in vivo,
25 which is decreased compared with normal corneas.
20 The mean endothelial cell loss at 1 week was 7%, consistent with clinical studies of postoperative cell loss in corneas preserved at 4°C in solutions that contain chondroitin sulfate.
26 27 28 This model will allow the testing of promising long-term corneal storage techniques and methods for endothelial cell augmentation on human corneas before clinical trials with human subjects. It will also allow investigators to study in more detail the characteristics of wound healing in the human cornea, including cellular apoptosis and keratocyte activation.
Apoptosis was infrequent in these corneas 1 week after transplantation. The median percentage of TUNEL-positive cells was either 0% or 1% in all four histologic regions for each cell type, except keratocytes in the human donor near the wound. Apparently, there was little apoptotic cell death in the cornea at the end of the first postoperative week, except for donor keratocytes near the wound. Higher levels of apoptosis may have been present earlier after keratoplasty, however, just as higher levels are often present during storage at 4°C.
21 In 10 human corneas stored for 0 to 21 days (mean, 10 days) in preservative (Optisol-GS; Chiron) at 4°C, the mean percentages of TUNEL-positive epithelial cells (13%), keratocytes (11%), and endothelial cells (8%)
21 were greater than those in the central human donor corneas 1 week after xenotransplantation (2%, 0%, and 2%, respectively;
Table 3 ). After the corneas returned to an in vivo setting for 1 week (i.e., by xenotransplantation), any increased apoptosis present during storage at 4°C apparently resolved. Although the TUNEL assay can at times be positive for cell death by necrosis,
29 30 31 32 in the human cornea the assay detects apoptotic, but not necrotic, cells.
21 Therefore, we assumed that the TUNEL-positive cells in this study were apoptotic. Not all apoptotic cells are detected by the TUNEL assay in the cornea, however; the percentage of apoptotic cells found by histology is higher than that found by the TUNEL assay.
21
Apoptosis of donor keratocytes, however, was significantly increased to a median rate of 7% near the wound. In the remaining three regions of the stroma, keratocyte apoptosis was 1% or less. In particular, almost no keratocyte apoptosis was found in the central graft, where the recently healed epithelium was intact in 7 of the 10 grafts. An increase of donor keratocyte apoptosis near the wound is consistent with the advent of wound healing in this area.
4 Increased apoptosis was not found in the cat keratocytes on the recipient side of the wound. We have no explanation for this finding; one would expect the early wound-healing process to proceed in both donor and recipient tissues near the wound. The highest full-thickness histologic keratocyte density was in the central donor cornea, where the apoptosis rate was the least. Conversely, the lowest keratocyte density was in the peripheral donor cornea near the wound, where the keratocyte apoptosis rate was the highest. The central full-thickness histologic keratocyte density was somewhat less than that observed by Komuro et al.
21 in donor corneas preserved at 4°C, consistent with further loss of donor keratocytes after transplantation. If the keratocytes lost to programmed cell death are not replaced, these findings are consistent with apoptosis’s lowering the keratocyte density in transplanted human corneas.
25 Cell death by necrosis
21 may also contribute, however.
In all corneas, we observed activated keratocytes, recognized by the visibility of their cell bodies and processes (as well as the nuclei, which are normally visible) in confocal microscopy. These cells probably represent repair fibroblasts,
22 33 which appear 3 days after corneal stromal wounding in the rabbit.
7 The cell bodies and processes are thought to become visible because of a change in their refractive index as the protein composition of the cytosol changes.
34 These cells affect corneal transparency, and they have been found in corneal grafts with late endothelial failure.
25 The amount of cellular activation may have been affected by the depo-corticosteroids given at keratoplasty. The cornea with the greatest amount of keratocyte activation, which extended the full thickness, had a large, persistent epithelial defect. The relationship between keratocyte activation and epithelial healing deserves further study. We did not see another type of corneal stromal cell common in healing wounds, the myofibroblast, although we did not specifically stain for α-smooth muscle actin
35 or search for stress fibers to identify these cells.
36 Myofibroblasts appear in the cornea 14 days after wounding in cats and primates,
3 long after the end of this experiment.
In summary, we developed a human-to-cat xenograft model that allows the histologic study of recently transplanted human corneas. One week after surgery, all three corneal cell types in the central cornea experienced a very low rate of apoptosis, lower than the rate often seen during corneal storage at 4°C.
21 The rate of apoptosis of donor keratocytes near the wound was increased, however. All grafts contained activated keratocytes with visible cell bodies and processes. This model will be useful in the development of new long-term corneal preservation techniques and methods for endothelial cell augmentation, postoperative keratocyte repopulation, and wound healing during the early postoperative period, before xenograft rejection occurs.
The authors thank Cheryl Hann for assistance with transmission electron microscopy.