Stromal Niche Controls the Plasticity of Limbal and Corneal Epithelial Differentiation in a Rabbit Model of Recombined Tissue

Abstract

purpose. The adult rabbit limbal basal epithelium contains corneal epithelial stem cells, which have been characterized by a negative expression of keratin-3 (K3) and a lower expression of connexin 43 (Cx43). This study was conducted to determine whether the limbal stroma dictates the plasticity of limbal and corneal epithelial differentiation.

methods. Viable epithelial sheets of the central cornea and the pigmented limbus were isolated from Dutch belted rabbits by incubation of 50 mg/mL of dispase II in supplemental hormonal epithelial medium (SHEM) for 18 hours at 4°C. The cleavage plane was studied by immunostaining with antibodies against K3, Cx43, integrin β1, and collagen IV. Viability of single cells derived from these sheets was assessed by a live-dead assay. Such limbal (L) and corneal (K) epithelial sheets were recombined with either limbal (Ls) or corneal (Ks) stroma, and cultured in SHEM for 10 days before lifting to the air–fluid interface for 1 week. The resultant epithelial phenotype was determined by histology and immunostaining to K3 and Cx43, and apoptosis was investigated by Hoechst and TUNEL nuclear staining.

results. Viability of isolated limbal and corneal epithelial sheets was determined to be 91.1% ± 2.9%. The basal epithelium of isolated limbal epithelial sheets was positive for integrin β1, negative for K3, but weakly positive for Cx43, and still retained patches of collagen IV. All recombinants showed stratified epithelia, with intraepithelial cysts with desquamated debris noted only in K/Ks, and epithelial outgrowth onto the insert filter from L/Ls. As expected, expression of K3 was negative in the basal layer of L/Ls, but positive in that of K/Ks. Expression of K3 was sporadically positive in the basal layer of L/Ks but largely negative in that of K/Ls. Expression of Cx43 was uniformly expressed in the basal layer of the K/Ks, but weak in that of L/Ls, K/Ls, and L/Ks. A higher apoptosis index was only noted in intraepithelial cysts of K/Ks.

conclusions. These results strongly indicate that the limbal stroma modulates epithelial differentiation, proliferation and apoptosis in the direction favoring stemness, whereas the corneal stroma promotes differentiation. Further investigation into elements constituting such a niche should help unveil the secrecy whereby stemness is controlled.

The adult corneal epithelium is continuously regenerated from stem cells (SCs) located at the basal layer of the limbal epithelium. The limbal epithelial SCs differ from corneal transient amplifying cells (TACs) located at the basal layer of the corneal epithelium in their lack of expression of cornea-specific differentiation keratins (K3/K12), ^{1 2 3} connexin 43 (Cx43)-mediated gap junction intercellular communication, ^{4 5 6} cell cycle length, ⁷ and label-retaining property. ⁸  

One important mechanism that modulates the aforementioned characteristics—that is “stemness” of the limbal SCs—is that the limbal stroma provides a unique microenvironment or niche. Although the salient features of such a niche remain elusive, it is well known that the limbus is strategically protected by heavy pigmentation, forms palisades of Vogt, and is highly innervated and vascularized. ^{6 9} Furthermore, damage to such a niche is sufficient to give rise to limbal SC deficiency in a number of corneal diseases that carry the hallmark of conjunctivalization. ¹⁰  

Recently, we have developed a novel method of isolating an intact and viable limbal epithelium from human corneoscleral tissues by enzymatic cleaving of adhesive basement membrane components. ¹¹ This advance allows us to separate the epithelium from the underlying stroma without damaging either tissue. In this study, we sought to demonstrate whether the “SC niche” is inherently present in the limbal stroma by recombining isolated limbal and corneal epithelial tissues with either limbal or corneal stroma. Our results showed that the limbal stroma indeed differs from the corneal stroma in modulating epithelial differentiation, proliferation, and apoptosis in the direction favoring stemness. The significance of this finding is further discussed. 

Materials and Methods

The following reagents and chemicals including fluorescence-conjugated goat anti-mouse antibody, fluorescence-conjugated rabbit anti-goat antibody, bovine serum albumin (BSA), Hoechst 33342 stain, mouse-derived epidermal growth factor, cholera-toxin (subunit A), dimethylsulfoxide, hydrocortisone, insulin-transferrin-sodium selenite (ITS) media supplement and sorbitol were purchased from Sigma-Aldrich (St. Louis, MO). Dispase II powder was obtained from Roche (Indianapolis, IN). 4′,6′-Diamino-2-phenylindole (DAPI) mounting medium (Vectashield) was from Vector Laboratories (Burlingame, CA). Antibodies were obtained to the following antigens: keratin 3 (K3; ICN, Aurora, OH), integrin β₁ (Sigma-Aldrich), collagen IV (Southern Biotechnology, Birmingham, AL), and Cx43 (Chemicon International, Temecula, CA). An apoptosis detection kit (DeadEnd Fluorometric TUNEL System) and DNAse were from Promega (Madison, WI). A live-dead assay was obtained from Molecular Probes (Eugene, OR). 

Enzymatic Isolation of Limbal and Corneal Epithelial Sheets

Animals used in this study were handled according to the guidelines described in ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Seven Dutch belted rabbits, 5 months of age, were killed and their globes enucleated. An entire anterior corneoscleral segment was removed from the globe by cutting the sclera 5-mm posterior to the limbus with Wescott scissors. A central cornea button and a corneoscleral rim were then obtained with an 8.0-mm disposable Barron trephine and immediately transferred to supplemented hormonal epithelial medium (SHEM), which is made of an equal volume of HEPES-buffered DMEM and Ham’s F12 containing bicarbonate, 0.5% dimethyl sulfoxide, 2 ng/mL mouse epidermal growth factor (EGF), 5 μg/mL insulin, 5 μg/mL transferrin, 5 ng/mL sodium selenite, 0.5 μg/mL hydrocortisone, 30 ng/mL cholera toxin A subunit, 5% fetal bovine serum (FBS), 50 μg/mL gentamicin, and 1.25 μg/mL amphotericin B. Corneoscleral rims were rubbed off the uveal tissue with a cotton tip and cut by a razor blade into four equal segments, each spanning three clock hours starting from 12 o’clock. Central cornea buttons were cut into two equal segments. Epithelial sheets were isolated by our recently published method. ¹¹ In short, all segments were incubated for 18 hours at 4°C in SHEM containing 50 mg/mL dispase II and 100 mM sorbitol. Under a dissecting microscope, limbal and central corneal epithelial sheets, already loose, were separated by a noncutting flat stainless-steel spatula and transferred to a 32-mm plastic dish containing SHEM (Fig. 1) . 

Cell Viability

Four isolated limbocorneal epithelial sheets from four different donor rims were incubated at 37°C for 5 minutes in Hanks’ balanced salt solution (HBSS) containing 0.05% trypsin and 0.53 mM EDTA. After brief pipetting, resultant single cells were centrifuged at 800g for 5 minutes and resuspended in PBS containing 2 μM calcein AM and 4 μM ethidium homodimer, according to the manufacturer of the live-dead cell assay (Molecular Probes) for 45 minutes before cell counting under a fluorescence microscope. A mean percentage of live cells was calculated by counting both dead (red fluorescence) and live (green fluorescence) cells at 10 different locations of a plastic dish under 400× magnification. Cultured human corneal epithelial cells expanded from limbal explants that were exposed to methanol for 1 hour were used as a positive control as dead cells. 

Tissue Recombination

Immediately after epithelial sheet removal, remaining limbal or central corneal stroma was rinsed three times in HBSS to inactivate dispase II and placed on the filter membrane of transwell-COL insert. The isolated limbal (L) epithelial or corneal (K) sheets in SHEM were transferred by aspiration into a Pasteur glass pipette, maintaining the sheets in the tip of the pipette to make sure that the orientation was not lost, to either the limbal (Ls) or corneal (Ks) stromal surface. This resulted in four different tissue recombinations: L/Ls, L/Ks, K/Ls, and K/Ks. They were cultured for 10 days in SHEM and lifted to an air–fluid interface for 1 week. 

Immunostaining

To characterize the cleavage plane after epithelial sheet isolation, a separated epithelial sheet including the limbus and the peripheral cornea was subjected to frozen sectioning and subsequent immunostaining. All recombined tissues were cut into frozen sections. All sections were fixated in cold acetone for 10 minutes at −20°C. Immunofluorescence staining was performed using primary antibodies against K3 (1:100), connexin 43 (1:100), collagen IV (1:50), and integrin β1 (1:100). Specific binding was detected by an FITC-conjugated secondary antibody and the sections were mounted in anti-fading solution containing 4′,6′-diamino-2-phenylindole (DAPI; Vectashield; Vector Laboratories, Burlingame, CA) and photographed with a epi-fluorescence microscope (Te-2000u Eclipse; Nikon, Tokyo, Japan). 

Apoptosis Assay

Epithelial apoptosis was evaluated by the terminal deoxynucleotidyl transferase-mediated dUTP nick-end labeling (TUNEL) assay according to the manufacturer’s protocol. In brief, sections were fixed in 4% formaldehyde in PBS, and cell membranes were permeabilized with proteinase K. The samples were then incubated with TdT enzyme and a nucleotide mixture labeled with fluorescein for 1 hour. After quenching the reaction, sections were mounted in anti-fade medium (DAPI-Vectashield; Vector Laboratories). Apoptosis was also determined by the cells’ nuclear morphology, by fixing sections in 4% formaldehyde and incubating them in 1 μg/mL Hoechst 33342 in PBS for 5 minutes. An apoptotic index was obtained by measuring the percentage of apoptotic cells (fluorescent cells) from a total number of cells counterstained by DAPI in seven different representative sections of each recombinant group at 400× magnification. 

Results

Enzymatic Isolation of Limbal and Central Corneal Epithelial Sheets

Consistent with our recent study of human tissues, ¹¹ after prolonged digestion in dispase II, the rabbit limbal epithelial sheet also became loose. Under microscopic examination, a stainless-steel spatula was easily inserted into a plane between an already-loose limbal or central corneal epithelium and the respective underlying stroma (Figs. 1A 1B) . Such isolated limbal epithelial sheets contained a high degree of pigmentation (Fig. 1C) , whereas in the remaining stromal tissue, there was no pigmentation but there was an imprint of palisades on the surface (Fig. 1D) . No epithelial cells were left after isolation (results not shown). Using such a technique, we successfully isolated intact limbal and central corneal epithelial sheets from 14 eyes, demonstrating the procedure’s feasibility and reproducibility (Figs. 2A 2B , respectively). 

Characterization of Isolated Limbal Epithelial Sheets and Viability

Hematoxylin staining of isolated limbal and central corneal epithelial sheets revealed the same morphologic characteristics as found in vivo. They maintained compact stratified epithelial layers, with superficial cells being large and squamous, whereas basal cells were small and round. Pigmented cells were observed predominantly in the basal area of the limbal epithelium with the basal cell membrane of the basal layer showing projections and extensions (Fig. 2C) . Immunostaining showed that K3 was expressed throughout the full thickness of the peripheral corneal epithelium (Fig. 2D , right of the arrow), whereas in the suprabasal layers of the limbal epithelium the basal epithelial cells were totally negative (Fig. 2D , left of the arrow). Immunostaining to Cx43 showed a pericellular punctate pattern in the intercellular location of the corneal basal epithelium, and much less such staining, with sporadic positivity, in the limbal epithelium (Fig. 2E) . Immunostaining for collagen IV was negative on the surface of the remaining stroma (Fig. 3B) , but was patchy in a linear pattern on the basal surface of the isolated epithelial sheet (Fig. 3C) . These results collectively support that the cleavage plane is within the lamina densa, a finding consistent with our recent report in human tissue. ¹¹ Immunostaining for integrin-β1 showed a pericellular pattern throughout the entire epithelial layer of isolated limbal sheets (Fig. 3D) . 

We then rendered isolated limbal and corneal epithelial sheets (Fig. 3E) into single cells by trypsin and EDTA. The viability of resultant single cells was measured to be 91.1% ± 2.9% (ranging from 86.7% to 93.4%; n = 4; Fig. 3F ). The positive control showed 100% dead cells (Fig. 3F , inset). 

Characterization of Epithelial Phenotype of Recombined Tissues

Hematoxylin staining of sections confirmed that all recombined tissues maintained multiple stratified epithelial layers with the basal epithelial cells being more cuboidal or columnar in shape (Fig. 4) . The number of epithelial stratification varied among different groups. The number of cell layers of K/Ks was 10.7 ± 2.1. There was an intraepithelial cyst with epithelial degeneration in two samples (Fig. 4A , stars). The number of cell layers in L/Ls was 9.7 ± 3.5 (Fig. 4B) , in K/Ls was 5.7 ± 3.9 (Fig. 4C) , and in L/Ks was 5.6 ± 2.6 (Fig. 4D) . The number of cell layers of the stratified epithelium on lateral and inferior surfaces of the tissue explant was less than five for all recombined tissues. Only epithelial cells of L/Ls continued to expand over to the filter membrane from the tissue explant. Fibroblasts were found in the entire stroma of recombined tissues in a density comparable with that of in vivo samples. 

K3 was expressed by all suprabasal epithelial layers of all recombined tissues. Nevertheless, expression of K3 in the basal layer varied according to the underlying stroma. As expected that the basal epithelium of K/Ks uniformly expressed K3 (Fig. 5A) , and that of L/Ls was entirely negative (Fig. 5B) . Of interest, expression of K3 was largely positive in the basal layer of L/Ks (Fig. 5C) and largely negative in that of K/Ls (Fig. 5D) . These results collectively indicated that the limbal stroma facilitated epithelial differentiation in the direction of the limbal epithelium, whereas the corneal stroma favored epithelial differentiation in the direction of the corneal epithelium. 

Cx43 expression by the basal epithelium was then examined. As expected strong continuous punctate staining in the intercellular space was noted in K/Ks (Fig. 6A) , whereas a sporadic positive punctate staining was noted in L/Ls (Fig. 6B) . Again, the staining pattern of Cx43 of L/Ks and K/Ls were influenced by the underlying stroma. L/Ks showed an increased expression of Cx43 (Fig. 6C) , in contrast to the limbal stroma (K/Ls) in which Cx43 expression decreased (Fig. 6D) . 

Disparities in the number and location of apoptotic cells were observed among the four recombinant groups. Staining by Hoechst 33342 showed apoptotic cells to be localized in the superficial desquamating epithelial layers in all recombinant groups; however, apoptotic cells with fragmented nuclei were observed only within the epithelium in the K/Ks group corresponding to the intraepithelial cyst, which was filled with desquamating cells (Figs. 7A 7B 7C) . This finding was further confirmed by TUNEL staining (Figs. 7D 8) . Differences in the apoptotic index were noted in these four groups: K/K (9.7 ± 3.4; Fig. 8A ), L/L (2.7 ± 1; Fig. 8B ), K/L (2.4 ± 0.7; Fig. 8C ), and L/K (2.5 ± 0.7; Fig. 8D ). These differences were statistically significant between K/K and the other groups (P < 0.001 for L/Ls versus K/Ks, K/Ls versus KK, and L/Ks versus K/Ks; ANOVA). However, the differences between the indexes in L/Ls, L/Ks, and K/Ls were not statistically significant. 

Discussion

Among all epithelial tissues, the corneal epithelium is most unique in having the TACs residing in the central and peripheral cornea, whereas the SCs are located exclusively in the limbus. ¹² As a result, these two types of epithelial progenitor cells with different proliferative and differentiative potentials are compartmentalized at two separate anatomic locations. This unique feature allowed us to explore the respective role of the limbal and corneal stroma in influencing epithelial proliferation and differentiation. In this article, we provide strong experimental evidence supporting the concept that the underlying stroma may indeed dictate the plasticity of epithelial differentiation, at least in a rabbit model. 

The notion that the stroma influences the direction of epithelial differentiation has been demonstrated in several in vitro studies of recombined-tissue models. Embryonic and adult epithelia from different locations can be transdifferentiated into a different epithelial phenotype according to the type of embryonic or neonatal stroma recombined. ^{13 14 15 16} These type of experiments challenge the concept that epithelial differentiation is irreversibly determined during embryonic and early postnatal life. The stromal tissues tested thus far that exert such a dominant influence on the direction of epithelial differentiation include those obtained from skin, ¹⁶ cornea, ^{17 18 19} mammary glands, ²⁰ vagina, ¹⁴ and urinary tract. ^{14 15 21} In the cornea, when embryonic ^{17 18} or adult ¹⁹ rabbit corneal epithelium is recombined with embryonic murine dermis and subsequently transplanted into nude mice, expression of the cornea-specific K3/K12 pair is changed to that of the epidermis-specific K1/K10 pair with formation of hair follicles or sweat glands. Such a dramatic transdifferentiation from corneal to epidermal differentiation is clearly dictated by embryonic dermal stroma. 

Unlike the aforementioned studies, ours is the first addressing whether adult limbal or corneal stroma may similarly modulate epithelial differentiation within a committed lineage, especially between SCs and TACs. Herein, we demonstrated that the limbal stroma indeed facilitated epithelial differentiation in the direction of being less differentiated with less apoptosis, whereas the corneal stroma promoted more epithelial differentiation and apoptosis. In theory, the stroma may influence epithelial differentiation through epithelial–mesenchymal interactions mediated by soluble factors, matrix components, or cell adhesion molecules. We believe that stromal fibroblasts were alive in both limbal and corneal stromas as shown by the presence of cells in the light microscopy and DAPI nuclear staining in sections. Furthermore, immunostaining showed positive bromodeoxyuridine (BrdU) uptake by stromal fibroblasts (Espana EM, unpublished observation, 2003). However, we do not know whether the modulating effect of the limbal stroma requires live mesenchymal cells. As shown in the current study, the limbal stroma is sufficient to downregulate the expression of K3 and Cx43 and intraepithelial desquamation and apoptosis of the corneal epithelium in the direction of SCs. Unfortunately, because there was no consensus of positive markers for limbal SCs, we could not confirm whether indeed TACs were dedifferentiated to SCs. Because the basement membrane components such as collagen IV (Fig. 3) ¹¹ and laminin 5 ¹¹ are digested by dispase II before recombination, we do not think that different components of the basement membrane exert a major modulating effect. 

The notion that the limbal stroma plays a dominant role in maintaining the characteristics of limbal SCs (i.e., stemness) is supported by the clinical findings that dysfunction of the limbal stroma by diverse causes may lead to limbal stem cell deficiency (LSCD). ¹⁰ The tissue recombination model described herein may be used to examine how a variety of pathologic insults such as inflammation, neurotrophic and hormonal deficiencies, and developmental anomalies, such as aniridia, may alter the limbal stroma that serves as a niche for supporting limbal SCs. 

 Presented at the annual meeting of the Association for Research in Vision and Ophthalmology, Fort Lauderdale, Florida, May 2003.

 Supported by National Eye Institute Grant EY06819 to Tissue Tech, Inc. (SCGT) and an unrestricted grant from the Ocular Surface Research and Education Foundation, Miami, Florida.

 Submitted for publication June 10, 2003; revised July 21, 2003; accepted July 23, 2003.

 Disclosure: E.M. Espana, TissueTech (E); T. Kawakita, TissueTech (E); A. Romano, None; M. Di Pascuale, TissueTech (E); R. Smiddy, None; C.-y. Liu, None; S.C.G. Tseng, TissueTech (E)

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

 

Figure 1.

 

Removal of intact limbal and central corneal epithelial sheets by dispase II. (A) A pigmented limbal epithelium from a Dutch belted rabbit. (B) After dispase digestion, a stainless-steel spatula was easily inserted into a plane between an already loose limbal epithelium (☆) and the stroma. (C) A limbal epithelial sheet (★ and ☆) with pigmentation was isolated. (D) The remaining stromal tissue was without pigmentation but revealed the imprint of palisades (arrows) on the surface.

Figure 2.

 

Characterization of isolated limbal and central corneal epithelial sheets. (A) Isolated limbal sheet showed high pigmentation in the limbal epithelium. (B) Central corneal sheet similarly isolated but without pigmentation. (C) Hematoxylin staining of a limbal sheet showed the presence of pigmentation in basal cells. K3 immunostaining was negative in limbal basal cells (left of arrowhead) but was positive (D) in the full thickness of the layer of corneal cells (right of arrowhead). (E) Cx43 immunostaining showed an interrupted pericellular punctate pattern (arrowheads) in the limbal epithelium. Bar, 50 μm.

Figure 3.

 

Characterization of the cleavage plane and cell viability. (A) Immunostaining for collagen IV showed linear staining corresponding to the basement membrane of the normal corneolimbal epithelium. (B) After dispase digestion, immunostaining was negative on the surface of the remaining stroma. (C) Linear patchy staining was present on the basal surface of the isolated epithelial sheet. (D) Immunostaining for β1 integrin showed a pericellular pattern throughout the entire isolated limbal epithelial layers. Predominant green fluorescent (viable) cells in isolated sheets were revealed by a live-dead cell assay before (E) and after (F) sheets were rendered into single cells. Inset: as a positive control, dead cells were stained with red fluorescence. Bar, 50 μm.

Figure 4.

 

Hematoxylin staining of the four recombined tissues. Stratified epithelial layers were noted, and fibroblasts were present in the entire stroma. K/Ks showed multiple intraepithelial cysts with epithelial degeneration (A, ☆). In L/Ls, epithelial cells continued to expand to the filter membrane from the tissue explant. L/Ks (C) and K/Ls (D) had a similar degree of stratification. Bar, 50 μm.

Figure 5.

 

Expression of K3. Immunostaining for K3 was positive in all suprabasal layers of four recombinant tissues. As expected, the basal epithelium of K/Ks uniformly expressed K3 (A) and that of L/Ls was entirely negative (B), resembling that known to be present in vivo. The basal epithelium of L/Ks was sporadically positive (C, arrows), whereas that of K/Ls was largely negative (D). Bar, 50 μm.

Figure 6.

 

Expression of Cx43. As expected, strong, continuous punctate staining was noted in the intercellular space of K/Ks (A), whereas a sporadically positive punctate staining was noted in L/Ls (B). Arrows: areas of cells with negative staining. L/Ks showed an increased expression of Cx43 (C), but K/Ls showed decreased Cx43 expression (D). Arrows: areas of cells with negative staining. Bar, 50 μm.

Figure 7.

 

Characterization of an intraepithelial cyst in K/Ks. Hematoxylin staining showed pyknotic nuclei in the cyst and cells adjacent to the cyst (A). Immunostaining to K3 showed cells within the cyst were densely packed with this intermediate filament in the cytoplasm, similar to desquamated cells (B). Morphologic analysis of apoptotic nuclei by Hoechst 33342 (C). Apoptotic nuclei were detected by TUNEL staining (D). Bar, 50 μm.

Figure 8.

 

Apoptotic nuclei detected by TUNEL staining. Intraepithelial apoptotic nuclei were found only in K/Ks, corresponding to an intraepithelial cyst (A). Positive fragmented nuclei corresponding to apoptotic cells were observed in superficial layers of the four recombined tissues (B) L/Ls, (C) K/Ls, and (D) L/Ks, respectively. Bar, 50 μm.