SEARCH

SEARCH BY CITATION

Keywords:

  • Human;
  • Cornea;
  • Endothelium;
  • Stem cell;
  • Limbus;
  • Rows;
  • Clusters;
  • Periphery

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION AND SUMMARY
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

The control of corneal transparency depends on the integrity of its endothelial monolayer, which is considered nonregenerative in adult humans. In pathological situations, endothelial cell (EC) loss, not offset by mitosis, can lead to irreversible corneal edema and blindness. However, the hypothesis of a slow, clinically insufficient regeneration starting from the corneal periphery remains debatable. The authors have re-evaluated the microanatomy of the endothelium in order to identify structures likely to support this homeostasis model. Whole endothelia of 88 human corneas (not stored, and stored in organ culture) with mean donor age of 80 ± 12 years were analyzed using an original flat-mounting technique. In 61% of corneas, cells located at the extreme periphery (last 200 μm of the endothelium) were organized in small clusters with two to three cell layers around Hassall-Henle bodies. In 68% of corneas, peripheral ECs formed centripetal rows 830 ± 295 μm long, with Descemet membrane furrows visible by scanning electron microscopy. EC density was significantly higher in zones with cell rows. When immunostained, ECs in the extreme periphery exhibited lesser differentiation (ZO-1, Actin, Na/K ATPase, CoxIV) than ECs in the center of the cornea but preferentially expressed stem cell markers (Nestin, Telomerase, and occasionally breast cancer resistance protein) and, in rare cases, the proliferation marker Ki67. Stored corneas had fewer cell clusters but more Ki67-positive ECs. We identified a novel anatomic organization in the periphery of the human corneal endothelium, suggesting a continuous slow centripetal migration, throughout life, of ECs from specific niches. STEM CELLS2012;30:2523–2534


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION AND SUMMARY
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

Corneal endothelial cells (ECs) are crucial for vision because they control movement of ions and water into the hydrophilic stroma and thus regulate corneal transparency [1]. Derived from the cranial neural crest [2], they form a monolayer of hexagonal cells organized in a tessellated mosaic on their basal membrane, the Descemet's membrane (DM). They uniformly cover the whole posterior corneal surface, and at its periphery reach the trabecular meshwork (TM) in the angle between the cornea and iris root. Central endothelial cell density (ECD, expressed in cells per mm2) physiologically decreases by 0.6% per year during adulthood [3, 4] without impairing corneal transparency, even in centenarians. However, several endothelial dystrophies and certain traumatisms dramatically accelerate this pace: ECD falls below the threshold of 300–500 cells per mm2, where irreversible corneal edema occurs, causing permanent vision loss. Such endothelial dysfunctions represent one-third of the indications for corneal grafts performed yearly worldwide.

In vivo, corneal endothelium has limited wound-healing capacity, using residual ECs which, by enlargement and migration, cover the defects left by lost cells [5] without cell division. Joyce et al. [6, 7] have demonstrated that human corneal ECs in vivo are arrested in the G1 phase of the cell cycle. They express negative regulators of the cell cycle belonging to the CIP/KIP family (p21 and p27), to the INK4 family (p16, p15, and p19), and to the p53 family of proteins (p53, TAp63) [8]. Mitotic inhibition may be caused by the presence of transforming growth factor β (TGF-β) in aqueous humor, by contact inhibition of densely packed ECs in a joint mosaic [9], and by stress-induced premature senescence [10–12].

However, a series of clinical and experimental arguments suggest the existence of endothelial regeneration in the human corneal periphery. Higher ECD in the endothelial periphery was described more than 25 years ago [13–15]. The possibility of centripetal migration of ECs from the periphery is suggested by at least two observations: the reported colonization of donors' DM by recipients' ECs after full-thickness corneal graft [16, 17]; and the demonstrated longer survival of corneal grafts in recipients with high ECD in the endothelial periphery [18, 19], namely in keratoconus, a noninflammatory corneal disease affecting young people and characterized by progressive stromal thinning and ecstasy but an endothelium that remains normal. This second observation suggests that the central endothelium is continuously sustained by the peripheral reserve. Furthermore, the continuous insidious destruction of peripheral ECs by certain artificial lenses implanted during cataract surgery in the 1980s, in contact with the corneal periphery, led 10–15 years later to edematous decompensation of the cornea, thus constituting an iatrogenic demonstration of the importance of peripheral ECs in endothelial and corneal homeostasis. Similarly, there is undoubtedly a rapid decrease in central ECD after corneal graft when no ECs persist in the recipient's corneal periphery, typically in edematous decompensation of the cornea. This suggests that central ECs that die cannot be replaced by peripheral ECs.

In experimental conditions, human corneal ECs retain residual proliferative capacity, even in very old donors [20]. In vitro, peripheral ECs have higher mitotic activity than central ones, as their morphology in primary culture is less typically hexagonal and their intercellular junctions loosen [21]. These differences may be due partly to the higher density of precursors in the endothelial periphery [22]. Whikehart et al. [23] and McGowan et al. [24] suggested that human corneal endothelial stem cells (SCs) may be sequestered in niches located deep in the TM or in the so-called transition zone (TZ) between the endothelial edge and the TM, and supply new cells to heal corneal endothelial wounds, at least during in vitro experiments.

In this study, we present a large series of corroborating observations that combine optical and electron microscopy with immunolocalization done on flat-mounts of whole human endothelia. This provides evidence that EC proliferation exists in the extreme periphery of human corneas in vivo. Cell organization in clusters and radial rows, first described in this article, progresses by one step the hypothesis that ECs continuously migrate centripetically from the extreme periphery, where SCs may nest, to the center of the cornea.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION AND SUMMARY
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

Cornea Characteristics

Three groups of corneas were studied. Group 1 (nonstored corneas) comprised 39 corneas obtained from the Anatomy Department of the Faculty of Medicine of Saint-Etienne thanks to body donations to science and processed immediately after procurement with no immersion in storage medium. Group 2 (stored corneas) comprised 45 corneas studied after storage in our eye bank. These were normal tissues procured after relatives had been informed, as authorized by French bioethics law. They were not suitable for transplantation owing to inconclusive donor serology despite normal endothelial characteristics, in particular central ECD >2,000 cells per mm2, the conventional cutoff value for corneal graft qualification. They were stored in organ culture (OC), the prevalent method used in European eye banks, which maintains cell metabolic activity for at least 5 weeks [25]. Corneas were immersed in a closed vial of 100 ml of a commercial medium containing 2% fetal calf serum (FCS; CorneaMax, Eurobio, Les Ulis, France, http://www.eurobio.fr) at 31°C, renewed after 2 weeks. Forty corneas were studied just before medium renewal, and five 3 days after. Corneas in groups 1 and 2 were procured from donors placed in 4°C mortuary chambers, by in situ excision with 18 mm diameter trephination, thus comprising the whole cornea and a rim of adjacent sclera with TM and iris root, as per the procedure recommended in France for corneas intended for transplantation. Group 3 (fresh corneas) comprised four central corneal buttons (8.25 mm in diameter) obtained, after patients' informed consent, directly in the operating theater during full-thickness graft for keratoconus in Saint-Etienne University Hospital. Time from procurement to fixation was only 15–30 minutes.

In group 1, donor age was (mean ± SD) 84 ± 8 years and death-to-procurement time 20 ± 16 hours. In group 2, donor age was 77 ± 10 years. In group 3, donor ages were 23, 28, 30, and 30 years. Handling of the donor tissues adhered to the tenets of the Declaration of Helsinki of 1975 and its 1983 revision in protecting donor confidentiality.

Observation of Flat-Mounted Corneas (En Face Observation)

Cell Morphology

ECs were stained using Alizarin red to visualize cell borders of the whole endothelial mosaic. Prior to staining, the endothelial side was rinsed with balanced salt solution (BSS, Alcon, Rueil Malmaison, France, http://www.alcon.fr). The dye (0.5% Alizarin red [Sigma, St Louis, MO, http://www.sigmaaldrich.com] dissolved in 0.9% sodium chloride with a pH adjusted to 5.2) was filtered through a 0.2-μm filter and placed in the corneal concavity for 90 seconds. After removal of excess dye, corneas were fixed for 2 minutes in 4% paraformaldehyde (PFA) in phosphate buffered saline (PBS) of pH 7.45. Corneas were then immersed in PBS and shaken until detachment of the dark red deposit that had formed on the endothelial surface. Nuclei were counterstained either with 0.4% trypan blue solution (Sigma) for 2 minutes or Hoechst 33342 (Sigma) 10 μg/ml in PBS for 1 minute. After a brief rinse in PBS, corneas were flat-mounted with four radial cuts. A large coverslip was maintained using adhesive tape to ensure mount stability. In a few corneas, the endothelium was observed using phase contrast combined with fluorescence imaging of nuclei stained with Hoechst but not Alizarin red.

Cell Count

ECD was determined in various endothelial zones (see below) using image analysis with the Cell⁁P software (Olympus Soft Imaging Solutions, Berlin, Germany, http://www.olympus-sis.com). Briefly, cell nuclei were automatically isolated after Hoechst image thresholding. Systematic manual controls prevented false-negative and false-positive cases. Nuclei were then numbered in large regions of interest (ROIs) comprising at least 1,000 nuclei. Half of the nuclei truncated by the ROI edges were taken into account.

Cell Mortality Assessment

In order to assess local differences on the whole endothelium in susceptibility to death after stresses caused by donor death and/or corneal storage, EC viability was determined using trypan blue staining or a live/dead assay with ethidium homodimer/calcein-AM (calcein acetoxymethyl ester) combined with Hoechst 33342, in a subset of corneas randomly chosen from groups 1 and 2. For group 1, corneas were studied within minutes of procurement. Vital staining with trypan blue is routinely used in eye banks during organ culture. We recently optimized triple labeling with Hoechst-ethidium-calcein (HEC) for experimental assessment of endothelial viability over the whole endothelial area [26]. The endothelial side was exposed for 1 minute to 0.4% trypan blue solution (Sigma) or submitted to HEC triple labeling as previously described [26]. After two rinses in BSS to remove excess dye, the cornea was fixed for 2 minutes in 4% PFA. For trypan blue stained corneas, after a brief rinse in PBS, nuclei were counterstained with Hoechst 33342 (Sigma) 10 μg/ml in PBS for 1 minute. All corneas were flat-mounted as previously described.

Histology on Cross-Sections

To compare the en face view and conventional histology on cross-sections, corneas were fixed in 4% PFA for 24 hours, then dehydrated in successive ethanol baths, and embedded in paraffin. Serial 6-μm-thick sections were cut on a microtome, deparaffinized, and stained with hematoxylin eosin safran. On a subset of 10 corneas selected at random, the thickness of DM was measured on ×40 images using the Cell⁁P image analysis software. For each cornea, thickness was measured on three cross-sections selected in different quadrants, and results were averaged. On each section, thickness was measured in the area where DM was deemed the thickest then repeated 0.5, 1, 2, and 5 mm nearer the center. Distance between DM edge and the area of maximum thickness was measured.

Immunostaining on Flat-Mounted Corneas: Proliferation, Differentiation, SC Markers

EC differentiation was studied using the expression pattern of four proteins: hexagonal apical organization of the tight junction protein ZO-1; hexagonal organization of the cytoplasmic F-actin belt; basolateral localization of Na+/K+ ATPase that participates in EC pump functions; and cytoplasmic expression of cytochrome C oxidase (COX) IV, which stains mitochondriae. Stemcellness was studied using Nestin, Telomerase, breast cancer resistance protein (BCRP/ABCG2), Oct-4, and P63-α antibodies. EC proliferation was studied with the ubiquitous proliferation marker Ki67 already validated for human corneal ECs [27].

Specific primary antibodies were IgG from either mouse or rabbit (supporting information Table 1). Nonspecific rabbit IgG (Zymed, Carlsbad, CA, http://www.zymed.com; 02-6102) and mouse IgG (Zymed; 02-6502) were used for negative controls. Secondary antibodies used at 1/500 were Alexa Fluor488 goat anti-mouse IgG (Invitrogen, Carlsbad, CA, http://www.invitrogen.com; A11001) and Alexa Fluor 555 goat anti-rabbit IgG (Invitrogen, A21428). Each cornea was cut radially into four or eight pie-shaped wedges. Fixation was done at room temperature (RT) for 45 minutes either in 0.5% PFA in PBS pH 7.45 or in methanol, depending on the antigen. Fixation was followed by antigen retrieval with 0.5% sodium dodecyl sulfate (SDS) in water for 5 minutes at RT to reveal telomerase. Membrane permeabilization with 1% Triton X-100 in PBS was necessary for PFA-fixed corneas except the corneal wedges treated with SDS. These experimental conditions were optimized after previous experiments recently published (data not shown) [28]. Heat-inactivated goat serum was used to block the nonspecific binding sites for 30 minutes at 37°C. Corneal pieces were immersed in primary antibodies diluted in 2% inactivated goat serum and 2% bovine serum albumin in PBS for 1 hour at 37°C. After three rinses in PBS, pieces were incubated with the secondary antibody for 45 minutes at 37°C. The counterstain was done with Hoechst 33342 10 μg/ml in PBS for 1 minute at RT. Finally, the corneas were flat-mounted in PBS just before observation. All images were captured using an IX81 fluorescence microscope (Olympus, Tokyo, Japan, http://www.olympus-global.com) with Cell⁁P imaging software.

Scanning Electron Microscopy of DM

Five corneas (two from group 1 with death-to-procurement time of 11–66 hours and three from group 2 stored for 10, 20, and 23 days) with mean donor age of 85 ± 19 years (53–100 years) were observed after removal of ECs, to give access to the bare DM. ECs were removed by incubation with 1% SDS in water followed by vigorous washing in distilled water. The corneas were fixed in 0.1 N cacodylate-buffered 2% glutaraldehyde pH 7.4 at 4°C for 24 hours. They were washed with 0.2 N cacodylate and then distilled water. The fixed specimens were dehydrated through ascending concentrations of acetone up to pure acetone. They were then dried using a critical point dryer (E3000, Quarum Technologies, East Sussex, U.K., http://www.quorumtech.com). After coating with gold-palladium in a mini sputter coater (Polaron SC 7620, Quorum Technologies) and mounting on metal stubs, the specimens were observed using a Hitachi S-3000N SEM at an electron accelerating voltage of 5 kV.

Definition of Observation Zones of the Posterior Surface of the Cornea

Cell characteristics were studied systematically in the surgical or anatomic regions of the posterior surface of the cornea, which were defined as follows, from center to periphery (Fig. 1): (a) endothelial center with a 4.0 mm radius, corresponding to the disk usually trephined by the surgeon for corneal graft; (b) endothelial periphery (P) of approximately 1.5–2 mm width (depending on corneal diameter), stopping at the anatomic endothelium edge, that is, the end of the DM, also called Schwalbe's line; (c) for our work, we had to define a supplementary zone inside the periphery corresponding to the last 0.2 mm of the endothelial periphery, which we named endothelial extreme periphery, a zone comprising most of the Hassal-Henle bodies that correspond to age-related Descemetic mushroom-shaped expansions; (d) TZ, 0.15 mm wide, characterized by absence of DM and of fibers typical of the TM. The zone is often confused in the literature with Schwalbe's ring, which is only prominent in 15% of corneas [29]; and (e) finally TM, approximately 0.5 mm wide. Outside these structures, residues of the iris root, gently detached during procurement, remained as a pigmented ring, followed by bare sclera cut at a 9-mm radius.

thumbnail image

Figure 1. Top: Zones of interest on the posterior surface of a human cornea. The sclera has been trephined to a diameter of 18 mm and the iris removed. Bottom: Cell clusters. (A, B): En face view of the corneal endothelial periphery (P). Cell clusters were localized in the EP in contact with the TZ. TM and iris root residues (I) are indicated. A ×10, B ×40. Alizarin red and trypan blue on fixed tissue. (C–F): (×96) Multilayered organization (arrowheads) of cell clusters, always localized in immediate vicinity of Hassall-Henle bodies (asterisk). (C): Cross-section (hematoxylin eosin safran). (D–F): En face view (Alizarin + trypan blue). Abbreviations: EP, extreme periphery; I, iris root residues; P, periphery; TM, trabecular meshwork; TZ, transition zone.

Download figure to PowerPoint

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION AND SUMMARY
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

En Face and Cross-Section Observations

Cell Clusters in the Extreme Periphery

On the flat-mounts, cell clusters were observed in the extreme periphery in nearly two-thirds of cases (61% (n = 23/38): 80% of group 1 corneas (n = 16/20) and 39% of those in group 2 (n = 7/18) (p = .01 (χ2)) (Fig. 1). These clusters were organized between the Hassall-Henle bodies and were not clearly separated and so impossible to count precisely. However, we estimate that on average they occupied one-third of the circumference of each cornea. On the cross-sections, they always comprised two or three cell layers 7–12 μm deep. The cells in the clusters had spherical or ovoid nuclei with smaller diameters than those of the central ECs, which had flat, strictly round nuclei of larger diameter.

Radial EC Rows in the Periphery
General description

In the periphery, EC organization in radial rows was observed in more than two-thirds of cases (68% [n = 26/38]): 67% (n = 12/18) of group 1 corneas and 70% (14/20) of those in group 2 (nonsignificant difference) (Fig. 2). These rows on average occupied 86% ± 24% of the corneal circumference, leaving some zones with none. These rows were variable in length in the same cornea and also between corneas, varying from 350 to 1,465 μm (mean 827 ± 295 μm). There was no difference in donor age or in central ECD between corneas presenting rows and the other corneas (data not shown). ECs of rows consistently had elongated nuclei (compared with those of the central ECs, which were always strictly round), the longitudinal axis of which most often coincided with the axis of the rows. Cells in rows had increased polymorphism and weaker Alizarin red staining than central ECs.

thumbnail image

Figure 2. Radial endothelial rows in the corneal periphery. (A): General view of half cornea flat-mounted thanks to radial cuts. Hoechst 33342 staining, ×4. (B, C): Zoom with ×10 objective on center (B) and periphery (C) where cell rows were clearly present. (D, E): Difference of endothelial cell morphology between center (D) and periphery (E). Alizarine red and Hoechst staining, ×40. (F): Transition between cell rows and more central endothelium in a nonstored cornea. The white line indicates an artificial boundary between the zones. Hoechst staining, ×10.

Download figure to PowerPoint

EC density in radial rows

Of the seven corneas in group 1 (age = 83 ± 13 years, death-to-procurement time = 19 ± 20 hours) picked randomly from among those organized in rows, ECD was significantly higher (by 69% ± 41%) in the cell rows than in the immediately adjacent zone not organized thus, respectively, 3,483 ± 558 versus 2,141 ± 507 EC per mm2 (p < .05). The highest maximum difference observed was 4,344 versus 1,786 cells per mm2, or +143% in a group 1 cornea (donor aged 72 years, death-to-procurement time of 4.5 hours). This difference was less marked in the 10 group-2 corneas (age = 80 ± 14 years, death-to-procurement time = 17 ± 11 hours, time in organ culture 14 ± 6-day): 23% ± 27% higher in the cell rows, respectively, 1,993 ± 698 versus 1,705 ± 750 ECs per mm2.

DM furrows in radial rows

Viewed by light microscopy on flat-mounts, the peripheral DM contained imprints of the linear organization of the ECs, in the form of furrows at the base of which EC nuclei were lodged. On the cross-sections, the DM periphery was thicker (Fig. 3) and more irregular than its center and the furrows were defined by DM excrescences covered by EC cytoplasm (Fig. 4). In scanning electron microscope (SEM), the furrows appeared as the dotted line formed by DM plots (Fig. 4).

thumbnail image

Figure 3. Descemet's membrane thickness from periphery to center. (A): Box plots of measurements performed in the thickest area (M), then 500, 1,000, 2,000, and 5,000 μm more centrally. The thickest area was constantly those starting just after the first Hassal-Henle bodies and was located 121 ± 102 μm after the Descemet edge. (B): Diagram of the measurement method on a histologic cross-section. Hassal-Henle bodies and Descemet excrescents are deliberately hypertrophied. To reduce biais, the mean thickness of each area was determined by image analysis, taking into account the area of Descemet's cross-section visible on a ×40 image of the whole field, that is, a width of 230 μm (area schematized in black).

Download figure to PowerPoint

thumbnail image

Figure 4. Radial Descemet's membrane (DM) furrows in the corneal periphery. Furrows were made visible after partial (A) or complete (C–F) removal of endothelial cells. (A): Flat-mount observed with phase contrast to highlight DM excrescences and with Hoechst 33342 to locate cell nuclei at the base of the furrows. (B): Cross-sections orthogonal to the cell rows, stained with hematoxylin eosin safran in order to highlight position of nuclei with regards to Descemet excrescents. (C–F): Scanning electron microscopy of the bare DM, showing furrows defined by real plots and starting near the TZ. Abbreviations: EP, extreme periphery; P, periphery; TM, trabecular meshwork; TZ, transition zone.

Download figure to PowerPoint

Relationship between cell clusters and radial rows

In all corneas, cell rows were separated from cell clusters by a narrow 150–200 μm band comprising polymorph ECs with an ECD lower than that in cell rows (see above) (Fig. 2; supporting information Fig. 1). Alterations to EC morphology were greater in group-2 corneas. In cross-sections, the DM of this band presented warts more elongated than typical Hassal-Henle bodies.

EC Mortality

In all corneas (groups 1 and 2), small clusters of cells with high cell mortality (>50% of cells) were systematically observed in the extreme periphery, with mortality decreasing very rapidly toward the center (Fig. 5).

thumbnail image

Figure 5. Cell mortality in the extreme periphery compared to the center. (A): Representative example of a nonstored cornea (age 69 years, death-to-procurement time 20 hours). Trypan blue and Hoechst staining. ×40 (B): Representative example of a stored cornea (age 72 years, 14 days of organ culture without medium renewal). Triple labeling with Hoechst, Ethidium homodimer, Calcein-AM. ×10. Dead cells (E+, C-) locally accumulated at the EP and TM, whereas in the periphery (P), endothelial cells remained viable (H+E-C+). Abbreviations: EP, extreme periphery; TM, trabecular meshwork.

Download figure to PowerPoint

Immunolocalization

Differentiation Markers

The cells in the extreme periphery did not have the expression patterns typical of central ECs for ZO-1 (absence of apical organization tight-junctions, hexagonal aspect), Actin (absence of submembranal apical belt, hexagonal aspect), Na+/K+ ATPase (absence of enzyme activity associated with ionic-pump function, located on the basolateral membranes on cell interdigitations, giving a distinctive star aspect), and COX IV (absence of dense mitochondrial network) (Fig. 6).

thumbnail image

Figure 6. Immunolocalization of differentiation and stem cells markers in endothelial cells. The typical expression patterns of ZO-1, Actin, Na/K ATPase, and CoxIV observed in the center were absent in the extreme periphery, in the cell-cluster zone. Telomerase and Nestin were largely expressed in the cell-cluster zone of almost all corneas. By contrast, BCRP/ABCG2 was found only in two nonstored corneas. PHC was used only for the extreme periphery to allow visualization of Hassal Henle bodies (asterisk), which by definition were absent in the center ×40. Abbreviations: BCRP, breast cancer resistance protein; PHC, phase contrast.

Download figure to PowerPoint

SC Markers

Expression of Nestin (nuclear and/or cytoplasmic) and Telomerase (cytoplasmic) was found in ECs in the extreme periphery (Fig. 6; supporting information Table 2). Weak expression of Nestin was observed in the cytoplasm of the scarce isolated ECs elsewhere in the endothelium. Expression of BCRP/ABCG2 was detected not only in a few ECs of the extreme periphery but also in the center of both corneas from one group-1 donor, procured very shortly after death (4.5 hours). Neither Oct-4 nor P63-α was detected (data not shown). There was no difference of expression of any of the five markers between corneas of groups 1 and 2 (data not shown).

Proliferation Marker Ki67

In the nonstored corneas (group 1), scarce Ki67-positive ECs were present but only in clusters in the extreme periphery of the endothelium of two corneas with very short death-to-procurement times (4.5 and 5 hours) (Fig. 7). In the stored corneas (group 2), many Ki67-positive ECs were observed in the periphery and slightly beyond but only in corneas of which the OC medium had been renewed 3 days before observation. In the group-3 corneas, no labeling was observed.

thumbnail image

Figure 7. Immunolocalization of the proliferation marker Ki67. (A, B): Nonstored cornea with death-to-procurement time of 4.5 hours. Only a few isolated endothelial cells (ECs) located in the extreme periphery were positive. (A): Hoechst + Ki67, (B): Merge with phase contrast to highlight Hassall-Henle bodies (asterisk). (C, D): Stored cornea observed 3 days after storage medium renewal. Multiple Ki67-positive ECs were visible, with metaphases indicating active cell cycling. The line in (C) delineated the transition zone.

Download figure to PowerPoint

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION AND SUMMARY
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

The corneal periphery is a complex region that is essential for ocular embryogenesis and for homeostasis of the adult cornea. Superficially, the limbus, a heavily vascularized junction zone between the cornea and sclera, contains epithelial SCs located in specific niches called limbal epithelial crypts (LECs) [30, 31] as well as many transient amplifying cells (TACs) [32]. The epithelium is a multilayered tissue with regular turnover and capable of rapid ad integrum healing even after total mechanical destruction, provided the limbus is intact. Epithelial SCs seem to be activated in case of massive epithelial damage, whereas physiological turnover is thought to be provided by TACs [32, 33]. While the existence of epithelial SCs in the limbus is an old premise, the anatomic description of their specific niches is more recent [30, 34]. These LECs are located at the ends of the palisades of Vogt, which are solid radial cords of cells in the extreme periphery of the corneal epithelium. The periphery of the corneal stroma probably also houses keratocyte progenitors, at greater depth [35]. Finally, the posterior, that is, truly intraocular, face of the cornea is even more complex than its surface, as it contains the posterior limbus (TM, Schlemms'canal, TZ) implicated in the regulation of intraocular pressure and the edges of the corneal endothelium. It may house corneal endothelial progenitors or SCs [22–24]. In the hypothesis of slow endothelial renewal, however, their proliferative capacity remains very limited and in particular, unlike the epithelium, they cannot respond to massive endothelial damage, even when the posterior-limbus structures remain intact. This limitation has yet to be explained.

For the first time, using a large number of human corneas, some of which were studied without storage, and complementary methods of observation, we were able to describe a new microanatomic organization of ECs in the endothelial periphery, which we termed cell clusters and radial cell rows (summarize in supporting information Fig. 1), and which suggests that ECs slowly and continuously migrate from periphery to center throughout life.

EC organization in the extreme periphery of the human cornea, in small clusters of two to three layers between Hassall-Henle bodies, had never been described. Hassall-Henle bodies were described in the late 19th century and are associated with aging of the normal cornea [36]. These mushroom-shaped Descemetic structures are solely observed in the endothelial periphery. They are absent in children and may progressively appear after 30 years, initially near Schwalbe's line and extending centripetically more than 1 mm [37]. They are constant in humans, but their number varies among individuals. Two hypotheses may explain their association with the peripheral cell clusters we describe. First, that Hassall-Henle bodies are synthesized throughout life by the cells of the clusters that perpetually reside in the periphery. Second, that clusters do not initially exist and Hassall-Henle bodies progressively develop during the continuous centripetal migration of cells that emerge from the posterior limbus and synthesize collagen at their emerging site. Secondarily, Hassall-Henle bodies, with aging, become obstacles that progressively slow cell migration and are consequently responsible for cell accumulation in clusters. Further studies on younger donors are necessary to ascertain whether cell clusters are present independently of Hassall-Henle bodies.

Clusters of cells located in Schwalbe's line were described in 1982 by Raviola [38] in seven Macaca mulatta (rhesus monkeys) aged from 3 to 17 years (life span 25 years) and were interpreted as “giving the impression that they originated from the endothelium and successively migrated deeply, toward the meshwork.” Retrospectively, these clusters of tightly packed spherical cells with high nucleus/cytoplasmic ratio located at the tapering end of Descemet's membrane appear identical to the clusters we describe here. Their distribution was discontinuous like ours, but their presence was constant in all specimens studied, independently of animal age.

In 1957, Vrabec noted peculiar fan-shaped groups of ECs in the human corneal periphery that were closely related to the anterior anchorage of the ciliary muscle tendons [39] but did not describe multilayered clusters. As it is very unlikely that endothelial SCs nest at the surface, the finding of a deeper structure, closer to the vascularization, may indicate a potential niche. The apparent smallness and round nuclei of these cells is also consistent with SCs or stem-like cells. These clusters are not directly connected with the cell rows we describe, remaining separated by an irregular zone where Descemet warts do not have the round shape typical of Hassall-Henle bodies. Cells in this zone are not fully differentiated, as we demonstrated using immunostaining, as they do not have structural and functional proteins typical of mature ECs, nevertheless they already form a monolayer which is one feature of ECs. Our staining pattern for SC and TAC markers is consistent with previous reports by Whikehart and McGowan but with certain differences. They described telomerase activity only in ECs located in a zone 4–12 mm from the center and not in the center, TM or TZ [23], whereas protein expression was described in the TM, TZ, and endothelial periphery on cross-sections and only in the TM on flat-mounted whole corneas [24]. This team described the apparition of Oct-4 positive cells in the TM after central trephination followed by subsequent 24–48 hours storage at 4°C and suggested migration of stem-like cells from the TM toward the endothelium. In our series, with immunostaining of flat-mounted whole corneas, telomerase expression seems restricted to the extreme periphery. We found Nestin predominantly in the extreme periphery, whereas it was present only in the TM and TZ (similar on cross-sections and flat-mounts) for Whikehart [23]. BCRP/ABCG2 has been described as a marker for hematopoietic [40] and corneal epithelial SCs [41], among other cell types. We found it expressed occasionally in isolated ECs. These results, associated with negativity for Oct-4 [42, 43], another SC marker, suggest that cells in clusters may be progenitors rather than SCs. Negativity for P63, shown to identify SCs in the basal layer of stratified surface limbal and skin epithelia [44], was consistent with the fact that this marker has never been shown positive in nonepithelial tissues. In any case, both explanations indicate that other markers need to be assessed in this subpopulation of cells. The absence of immunostaining for cells located more deeply in the clusters may be due to our choice of staining only on flat-mounts in order to screen for rare cells among an intact endothelial population [28].

Unlike the LECs described for the corneal epithelial SCs in the limbus [30] and which meet all niche criteria (anatomic protection, specific niche cells, proximity to vascularization, specific microenvironment, adhesion molecules, extracellular matrix components) [45, 46], we did not expect to find similar structures for the endothelium, owing to the absence of clinically significant proliferative capacity. Nevertheless, our microanatomic observations supply fresh arguments in favor of niche-like structures in the extreme periphery of the cornea.

We found overmortality in the extreme periphery, in the zone of the presumed TACs, in nonstored corneas very soon after donor death. Post-mortem ischemia of this heavily vascularized zone probably plays an important role. Increased susceptibility to post-mortem oxidative stress is a hypothesis for these poorly differentiated cells [47], which may have limited protective systems such as, for example, the heat shock proteins in the more central ECs [48]. The literature does not report any particular fragility of SCs or TACs, but, except in corneal-epithelium studies, the other SC types were always procured from beating-heart donors. For the corneal epithelium, however, it has been shown that long death-to-procurement time reduces the proliferative capacity of TACs [49, 50], which may indicate specific mortality of limbic epithelial SCs due to postmortem ischemia. In our study, this apparent higher sensitivity of peripheral cells to cell death soon after donor death may also partly explain the extreme scarcity of Ki67-positive cells in nonstored corneas.

Notably, we also observed overmortality in the extreme periphery of stored corneas, accompanied by lesser frequency of cell clusters and higher EC polymorphism. OC is a storage method derived from cell culture and used since the early 1970s [51]. Contrary to short-term 4°C storage (5–8 days), it lasts up to 5 weeks. Storage-induced cell alterations have been extensively studied in the central cornea, the area of interest for corneal graft. Despite general satisfactory preservation of corneal viability, attested by constant good clinical results identical to 4°C short-term storage [52, 53], death of central ECs increases dramatically during OC, with approximately 1% cell loss per day [54, 55] compared with 0.6% per year throughout life [3, 4]. However, OC-specific alterations of ECs in the extreme periphery have never been described. As OC is known to promote endothelial wound healing [56], we suppose that dead cells in the cell clusters rapidly desquamate and are replaced by neighboring ECs, explaining both why EC clusters were less frequent and why EC polymorphism increased. Moreover, the continuous overmortality that we observed in the extreme periphery may be explained by a specific sensitivity to ischemia reperfusion-like syndrome, as OC conditions restore metabolic activity after prolonged postmortem ischemia. This specific overmortality in corneas of nonstored or stored corneas, not previously reported, may seriously limit further study of EC proliferation ex vivo. The possibility, independently of donor age, of triggering EC division after intercellular junction loosening with EDTA and incubation at 37°C in mitogen-rich medium was demonstrated by Senoo et al. [57]. Cell division was nevertheless not reported to be restricted in the extreme periphery, even if proliferation capacity was higher in the periphery regardless of donor age [58]. In a similar study on 13 corneas (from donors aged 15–68 years) stored for 7 days at 4°C in OptisolGS, subsequently exposed to EDTA, and cultured at 37°C in a cell-culture medium with 10% FCS plus growth factors for 48 and 96 hours to promote cell proliferation, more Ki67-positive ECs were found in the center than in the periphery [59]. The number of Ki67-positive cells was inversely proportional to ECD. No proliferation was observed when initial ECD exceeded 2,000 cells per mm2. Having showed that cell death occurs preferentially in the extreme periphery where TACs may reside, we suppose that, in most previous studies, the initial 4°C storage time may have been deleterious to the peripheral TACs already weakened by donor death-induced cellular stress, precluding observation of a higher number of Ki67-positive ECs in the extreme periphery, even after mitotic stimulation.

Interestingly, we observed numerous proliferating ECs in the periphery of organ-cultured corneas but only shortly after medium renewal. We suppose that proliferation was not higher in the extreme periphery for the reasons set out above. OC media, typically containing 2% FCS, are not known to promote EC division. EC mitosis has only been described in corneas with numerous endothelial lesions due to extended death-to-procurement time and subsequently stored in a specific OC medium supplemented with 8% FCS [60]. In our study, we suppose that the initial storage time triggered cell loss sufficient to locally loosen cell–cell contact and that the rapid increase in nutrients and growth factors during medium renewal stimulated division of several ECs near the local lesions, in a way similar to the experiments reported above with EDTA followed by incubation with mitogens [57].

In our study, the occasional expression of Ki67 in only one nonstored cornea with very short death-to-procurement time (probably with limited ischemia) may indicate that in vivo cell division in this zone concerns only a limited number of cells. Considering cell cycle duration, Ki67 staining of only a few cells is consistent with clinical experience.

To our knowledge, the EC organization in radial rows observed in two-thirds of corneas has never previously been described. In vivo, neither the biomicroscopy used for clinical corneal observation nor the specular microscopy used for endothelial observation allows analysis of the periphery of the posterior face of the cornea. Using SEM, Svedbergh exhaustively described the natural history of the human corneal endothelium of 14 subjects aged 13–88 years without reporting radial rows [61]. His study was done on intact endothelia, whereas Descemetic furrows are only visible after EC removal. Also with SEM, and this time after EC removal by sonication, Inaba studied the periphery of the DM of 87 corneoscleral rims (from donors aged 2 to 100 years), which are postoperative residues of full-thickness corneal graft after central trephination. However, he only examined the TZ and the Hassall-Henle bodies, that is, a zone slightly more peripheral than that of the radial cell rows, and so neither could he describe organization in rows [37]. Indeed, it is likely that the zone comprising the rows was damaged by trepanation. Finally, when the endothelium is observed on histological cross-sections, the most common technique in the literature, rows are not visible and clusters are very hard to see, in the absence of serial cross-sections. Cell clusters and radial rows were frequent but not constant in our series, suggesting interindividual variability. This is not surprising and can be compared with the variability of healing capacity clinically observed at the ocular surface as well as with the success rate of epithelial culture from limbal explants [49]. Organization in radial cell rows is associated with much higher local ECD than in zones not organized thus. This strongly suggests that continuous proliferation in the periphery regularly “pushes” ECs centripetally. The radial cell rows occupy a narrow peripheral band (mean 827 ± 295 μm), which may be relatively well protected by a cornea that is naturally less transparent in this zone, particularly due to the presence of an arcus senilis, frequent in elderly subjects and in our donor population. We also observed by SEM that this likely cell migration throughout life leaves a highly visible imprint on the DM. In particular, we observed plot structures that perfectly match the images obtained by third harmonic multiphoton microscopy which Aptel and colleagues recently published and titled “Descemetic fibrous patches connecting EC to the stroma” [62]. The DM contains a nonbanded layer that is synthesized by ECs throughout life and gradually thickens from 2 to 10 μm between 10 and 80 years [63]. The largest mean thickness found in the periphery, and the association with the plots, indicate increased collagen synthesis by peripheral ECs, linked to their greater number and/or more intense activity. However, the linear organization of the plots, which form clear furrows, reflects the deposit of collagen along the EC path throughout life. The function of these plots, located between the ECs and potentially at the intercellular junctions, has yet to be determined. The heterogeneous distributions of cell-line length indicate that cell migration is heterogeneous in the same cornea and between individuals. Cell lines have never been found in the central endothelium of normal corneas. This indicates that corneal-endothelium regeneration is a very slow process limited to the peripheral endothelium of normal corneas. The age-related decrease in ECD in the central endothelium indicates that this regenerative mechanism may only partially offset cell loss in the central endothelium during adulthood. Again, the absence of young donors in our series prevents us from verifying at what point in life this organization appears.

In the radial cell row zone of nonstored corneas, we found weaker Alizarin red staining and higher polymorphism than in central ECs. These characteristics had already been described years ago using silver cell border staining in ECs of the TZ (although a linear organization was not described) [39]. This may indicate looser intercellular junctions and therefore higher cell-mobility potential than in the center, where the intensity of alizarin red staining and the regular morphometry correlate with the stability of the cell mosaic. Note that both modifications are actually observed in the center during wound healing, when cell migration becomes necessary to cover a defect left by dead cells. The possibility for ECs to easily move in from the extreme periphery is a supplementary argument for slow centripetal migration from stem/progenitor cell niches and may explain why they seem to accumulate in a zone where cells are more uniform and probably more stable, as shown in Figure 3F. Cell nuclei in radial cell rows were always elongated and radially oriented, whereas nuclei of central ECs were perfectly round. This morphological difference between peripheral and central ECs suggests that the endothelium is more stable around the optical axis, and is supplied by a peripheral store of cells that are less differentiated and liable to migrate slowly toward the central zone. In this peripheral zone, which is not needed for vision, the pump function of the endothelium for deturgescence of the corneal stroma, which incidentally is thicker in its periphery than in its center, is probably less efficient.

CONCLUSION AND SUMMARY

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION AND SUMMARY
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

To summarize, we suggest a unifying model of human corneal endothelial homeostasis. This model comprises, in the “non-visual” periphery of the cornea, a renewal zone where ECs divide very slowly and migrate toward the center but probably desquame, while the central ECD in the optical axis remains stable. Cell clusters located in the extreme periphery may be SC niches or at least emerging points for progenitors migrating from deeper niches. Once they make contact with aqueous humor, their dividing capacity, but not their migrating potential, may be partly repressed by cell cycle inhibition by TGF β and by cell contact inhibition as soon as they form a monolayer. These cells may then be continuously pushed toward the center and retain some proliferative capacity in experimental conditions, but as they never divide in vivo after reaching the endothelial monolayer, they do not shorten their telomeres [10, 64]. During migration, they acquire full differentiation and slowly lose their residual proliferative capacity owing to environmental stresses (ultra-violet light, temperature changes, aqueous humor soluble factors) [11, 65] that are probably more pronounced in the corneal center, far from limbal vascularization. These mechanisms may contribute to the stability of the central endothelium, which, as an ionic pump, is devoted solely to maintaining a transparent optical axis.

Acknowledgements

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION AND SUMMARY
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

We thank Jean-Yves Thuret for his comments on the manuscript and fruitful discussions; Isabelle Anselme from the Centre de Microscopie Electronique Stéphanois (CMES) for her contribution to scanning electron microscopy; Sophie Acquart from the Eye Bank of Saint-Etienne and Gérard Guillain from the Anatomy Department at the Faculty of Medicine of Saint-Etienne, both for providing precious human corneas.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION AND SUMMARY
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION AND SUMMARY
  8. Acknowledgements
  9. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  10. REFERENCES
  11. Supporting Information

Additional Supporting Information may be found in the online version of this article.

FilenameFormatSizeDescription
sc-11-1032_sm_SupplFigure1.pdf1400KSupplementary Figure 1. Summary of endothelial and Descemetic centripetal modifications. (A, B): Crosssections. Hematoxylin eosin safran staining demonstrating progressive centripetal alterations of the Descemet membrane (DM). (C): En face view after Hoechst staining of nuclei. (D): Merge Hoechst and phase contrast to highlight DM excrescences and nuclei preferential orientation. (E): Scanning electron microscopy of DM surface structures after endothelial cell removal. All images have been aligned.
sc-11-1032_sm_SupplTable1.pdf80KSupplementary Table 1.
sc-11-1032_sm_SupplTable2.pdf35KSupplementary Table 2.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.