Immune Privilege and Immunogenicity Reside among Different Layers of the Mouse Cornea

Abstract

purpose. To determine the extent to which each layer of the mouse cornea displays alloimmunogenicity or immune privilege.

methods. Intact corneas or individual or combined layers of corneas from normal or cauterized eyes of BALB/c, C57BL/6, and CD95L-deficient B6-gld mice were grafted beneath the kidney capsule of normal BALB/c, B10.D2, BALB.B mice or of BALB/c mice presensitized to donor antigens. Graft fate was assessed clinically and histologically and acquisition of donor-specific delayed hypersensitivity (DH) was assessed at selected intervals after grafting.

results. Full-thickness allogeneic corneas induced vigorous DH and were rejected acutely. Similar results were obtained with allografts of corneal epithelium alone (if supported by syngeneic viable stroma), allografts of epithelium from cauterized corneas (containing Langerhans’ cells), and stromal allografts deprived of endothelium. Grafts comprised of stroma plus endothelium (without epithelium) were not rejected, nor did they induce DH unless the graft had no CD95L expression. If stroma–endothelium grafts had no CD95L expression, DH directed against major histocompatibility complex (MHC), but not minor histocompatibility, alloantigens was induced. Moreover, CD95L expressed on stroma–endothelium grafts protected endothelial cells, but not stromal cells, from rejection in presensitized recipients.

conclusions. When grafted to a heterotopic site, the alloimmunogenicity of the normal cornea resides within its epithelial and stromal layers, whereas immune privilege arises from the endothelium. In normal mice, CD95L-expressing endothelium can inhibit the stroma from inducing immunity directed at MHC alloantigens, but in presensitized mice the endothelium can protect itself only from immune rejection.

Each layer of the cornea has the potential of contributing to the immunogenicity of this tissue as a graft. When corneal allografts are placed orthotopically in eyes of experimental animals, it is difficult to discern the immunogenic potential of the various layers because the graft is placed in an immune privileged site. ^{1 2 3} Any analysis of immune responses to alloantigens expressed on orthotopic corneal grafts is complicated by the immunoregulatory properties of the site itself. ^{4 5} In an attempt to discern the relative contributions to alloimmunization of each layer of the cornea, Khodadoust and Silverstein ⁶ conducted a series of experiments many years ago in which individual layers of the cornea from rabbit A were placed for 4 weeks at orthotopic sites in eyes of allodisparate rabbit B and then removed and grafted back into the fellow eye of rabbit A. All the grafts were rejected, and the conclusion was drawn that all three layers (epithelium, stroma, endothelium) are immunogenic. However, the investigators were unaware at the time of the ability of bone marrow–derived cells (such as Langerhans’ cells) to infiltrate into corneal epithelium when it resurfaces a wound and to migrate into the stroma of a corneal graft. 

In light of recent information that grafted corneal tissue acquires recipient bone marrow–derived cells in both the epithelial and stromal layers, the grafts used by Khodadoust and Silverstein, which were parked for 4 weeks on rabbit B, were surely contaminated by bone marrow–derived cells of rabbit B. Thus, these data do not reveal unequivocally whether each layer of the normal cornea is immunogenic in the absence of bone marrow–derived cells. More recently, corneal tissues have been grafted heterotopically to the cutaneous surface of mice, ⁷ and these studies have found that full-thickness corneal grafts are regularly rejected at this site unless the graft confronts the recipient only with alloantigens encoded by class II genes within the major histocompatibility complex (MHC), H-2. ⁸ The explanation for why class II–only disparate corneal heterografts resist rejection is that the normal cornea has no Langerhans’ cells in the epithelium, and none of the other corneal cells—epithelium, keratocytes, or endothelium—constitutively expresses class II molecules. ⁹  

Bellgrau et al. ¹⁰ have reported recently that allografts of testis survive indefinitely when placed heterotopically beneath the kidney capsule. This is an expression of the inherent immune privilege of testis tissue. In unpublished experiments, these investigators found that testis grafts placed in the skin did not survive, but were rejected (personal communication, Richard Duke, November 1997). This finding suggests that the skin offers unusually stiff barriers to allograft acceptance, barriers that reveal more about the immunobiology of skin than about the tissue graft placed at this heterotopic site. To follow up on this lead, our laboratory has initiated studies of the fate and immunogenicity of corneal tissues grafted heterotopically beneath the kidney capsule. This site has been used for many years by transplantation immunologists as a conventional site for heterotopic grafts. Allografts of skin, kidney, liver, and islets of Langerhans are acutely rejected when placed beneath the kidney capsule. ^{11 12 13}  

From the first series of experiments of this type, we recently reported that epithelium-deprived corneal allografts (stroma plus endothelium) survive indefinitely when placed beneath the kidney capsule in mice, confirming that either or both of these layers of the cornea possess inherent immune privilege. ¹⁴ We have now extended these studies to examine the immunogenic potential of each layer of the allogeneic cornea by implanting full- or partial-thickness corneas beneath the kidney capsule. Our results reveal that both corneal epithelium and stroma when placed beneath the kidney capsule are inherently alloantigenic. These grafts undergo immune rejection and induce donor-specific delayed hypersensitivity (DH). By contrast, corneal endothelium, in part through expression of CD95L, nullifies stromal immunogenicity, prevents allosensitization, and promotes acceptance of stromal plus endothelial heterografts. 

Materials and Methods

Mice and Anesthesia

Male BALB/c (H-2^d) and C57BL/6 (B6, H-2^b) mice were purchased from Taconic Farm (Germantown, NY). Male B6Smn.C3H-Faslgld (B6-gld), B10.D2 (H-2^d), and BALB.B (H-2^b) mice were purchased from Jackson Laboratories (Bar Harbor, ME). All mice were used at 8 to 10 weeks of age and treated according to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Each mouse was anesthetized by intramuscular injection of a mixture of 3.75 mg ketamine and 0.75 mg xylazine before all surgical procedures. 

Preparation of Grafts Comprising One or More Layers of Cornea

Corneal Epithelium and Skin Epidermis.

In some experiments, corneal epithelial sheets (and as control specimens, skin epidermal sheets) were used as grafts. Full-thickness corneas devoid of limbus were harvested from donor eyes. Skin was harvested from donor ears. The epithelium and epidermis were peeled as intact sheets from full-thickness cornea and ear skin by forceps, respectively, after 1 to 1.5 hours’ incubation in 20 mM EDTA at 37°C and washed with phosphate-buffered saline (PBS). 

Corneal Epithelium Containing Langerhans’ Cells.

The corneal surface of mouse eyes was cauterized as described previously. ^{15 16} Briefly, using the tip of a handheld cautery, five burns were applied to the central 50% of the cornea to induce centripetal Langerhans’ cell migration. After this maneuver, the epithelium is known to be resurfaced within 48 hours, and Langerhans’ cells are known to migrate into the epithelium with peak density at 2 weeks after cautery. ¹⁵ Epithelial sheets were removed from cauterized corneas of eyes after 2 weeks. When examined histologically, the sheets contained only the epithelial layer, with little contaminating stroma and no keratocytes (data not shown). 

Grafts of Corneal Stroma Plus Endothelium.

Full-thickness corneas were incubated in EDTA as described. When the epithelium was removed with forceps, the remaining stroma plus endothelium was then used as a graft. 

Grafts of Stroma Devoid of Endothelium.

To produce endothelium-deprived stromal grafts from epithelium-deprived stroma plus endothelium, the corneal endothelium was scraped off the posterior surface using a cotton swab. 

Composite Cornea.

In some experiments, layers of corneal stroma alone were prepared from BALB/c corneas. Epithelial layers from normal C57BL/6 eyes were then prepared and carefully floated onto the stromal layer. Within a few minutes of in vitro incubation, a tight union formed between the layers, and the composite tissue was then used for grafting. 

Heterotopic Corneal Transplantation under the Kidney Capsule

BALB/c, B10.D2, BALB.B, or BALB/c anti-C57BL/6 mice were used as recipients, and BALB/c, C57BL/6, and B6-gld mice were used as donors. Each experimental panel comprised at least 12 recipients. For full-thickness grafts, a 2-mm diameter central portion of the cornea was harvested from normal or cauterized eyes of donors, divided in half (1 × 2 mm), and then grafted beneath the kidney capsule. For partial-thickness grafts, tissues of comparable size were prepared as described and placed beneath the kidney capsule. To serve as positive grafting controls for full-thickness cornea grafts, full-thickness footpad skin (glabrous, without hair follicles and accessory epithelial structures) was used. To place the grafts at the heterotopic site, a skin incision was made in the left flank of recipient mice, and the muscle wall was incised and the kidney exteriorized. A subcapsular pocket was created between the kidney, and the kidney cortex, and the graft was placed into the pocket. The kidney was replaced in the abdominal cavity, and the skin was closed with 7-mm clips. 

Clinical and Histologic Evaluation of Heterotopic Corneal Grafts

Heterotopic corneal graft survival was assessed by visual inspection by a single observer (JH) under the operating microscope at selected time points after implantation. At each time point, graft-bearing mice were anesthetized, the kidney was exteriorized, and a clinical assessment was made, evaluating the graft for evidence of swelling, opacity, and new blood vessel growth into the graft stroma. At the completion of the clinical examination, the graft-bearing kidney was removed for histologic examination. The entire kidney was fixed with 10% formalin, embedded in paraffin, sectioned, and stained with hematoxylin and eosin. Approximately 40 to 60 sections were prepared and examined from each graft-bearing kidney. 

Immunohistochemical Assessment of Heterotopic Corneal Allografts

Immunohistochemical studies for CD45 and I-A^d expression were performed on frozen sections of corneal allografts placed under the kidney capsule. Phykoerythrin (PE)-labeled rat anti-mouse CD45 and fluorescein isothiocyanate (FITC)-labeled rat anti-mouse I-A^d monoclonal antibodies (PharMingen, San Diego, CA) were used as primary antibodies. Graft-bearing kidneys were removed at 7, 14, and 21 days, frozen in optimal cutting temperature (OCT; Miles, Elkhart, IN) compound in acetone-dry ice and stored at −80°C. The frozen specimens were sectioned at 5 μm by cryostat, fixed in acetone, and air dried. After washing with PBS, the sections were incubated in each primary antibody diluted to 4 μg/ml for 2 hours at room temperature. After washing with PBS, the samples were observed by fluorescence microscopy. 

Evaluation of Corneal Endothelial Cell Integrity of Heterotopic Corneal Allografts

To discern the integrity of the endothelial monolayer of corneal grafts under the kidney capsule, immunocytochemistry was performed using a tight junction–associated protein marker, Zonula occludens (ZO)-1 as previously described. ¹⁴ The graft-bearing kidney was removed at 14 days, frozen in OCT compound in acetone-dry ice, and stored at −80°C. The frozen specimens were sectioned at 5 μm by cryostat, fixed in acetone, and air-dried. After incubation with 2% bovine serum albumin (BSA) for 10 minutes to prevent nonspecific binding, the sections were incubated for 2 hours with rabbit polyclonal anti-ZO-1 antibody diluted to 4 μg/ml (Zymed Laboratories, San Francisco, CA). After washing with PBS, the sections were incubated for 1 hour with FITC-conjugated donkey anti-rabbit IgG as a secondary antibody, 6 μg/ml (Jackson ImmunoResearch, West Grove, PA). After washing with PBS, the sections were mounted with medium containing propidium iodide (Vectastain; Vector, Burlingame, CA) and observed under a confocal microscope. The negative control was generated by incubating tissue in secondary antibody alone. 

DH Assessment

At selected times after allogeneic corneal implantation beneath the kidney capsule, 1 × 10⁶ irradiated (2000 R) spleen cells/10 μl from C57BL/6 donors were injected into the right pinnae of recipient mice for an ear-swelling assay. All panels of test mice included five to six animals. As a positive control, a similar number of irradiated spleen cells was injected into the ear pinnae of normal BALB/c mice that had been immunized 1 week previously by subcutaneous (SC) injection of 10 × 10⁶ C57BL/6 spleen cells. As a negative control, 1 × 10⁶ irradiated spleen cells were injected into ear pinnae of naive mice. Twenty-four and 48 hours after ear injection, ear thickness was measured with a low-pressure engineer’s micrometer (Mitsutoyo; MTI, Paramus, NJ). Ear swelling was expressed as follows: Specific swelling = [(24-hour numerical values of right ear − 0 hour numerical values of right ear) − (24 hour numerical values of left ear − 0 hour numerical values of left ear)] × 10⁻³ mm. Ear-swelling responses at 24 hours after ear injection are presented as group means ± SEM. 

Statistical Analyses

Ear-swelling measurements were evaluated statistically by using a two-tailed Student’s t-test. P < 0.05 was deemed significant. 

Results

Fate of Full-Thickness Corneal Allografts

Full-thickness corneal allografts prepared from eyes of normal C57BL/6 or BALB/c mice were placed beneath the kidney capsule of BALB/c recipients. Kidneys bearing syngeneic and allogeneic full-thickness grafts were examined in situ at 2 or 10 weeks after grafting, by visual inspection through an operating microscope. The graft-bearing kidneys were then removed for histologic examination. At 2 weeks after grafting, heterotopic corneal allografts were opaque, and the overlying capsule was neovascularized. At the same time point, syngeneic corneal grafts were also opaque, and accompanied by similar capsular neovessels. When these grafts were examined histologically, the epithelium of syngeneic grafts displayed evidence of epithelial cell proliferation, forming a large keratinized cyst. The stroma of these grafts contained numerous easily identifiable keratocytes without evidence of edema or leukocytic infiltration. An intact endothelium was observed, adjacent to the kidney parenchyma. By contrast, in allografts, only scattered islands of epithelial cells persisted amid innumerable necrotic cells, and a polymorphonuclear infiltrate resided where the epithelial layer should have been. The stroma of full-thickness allogeneic corneal grafts was swollen, infiltrated with numerous leukocytes, and contained projections of new blood vessels, making it difficult to determine whether any viable keratocytes were present. The stromal lamellae were in disarray, and the endothelium was not visible. At 10 weeks, the epithelium of syngeneic grafts formed large, white circular masses that, on histologic examination, were identified as large cysts of keratinized epithelium (Fig. 1A ). At the same time point, no comparable white mass was present at the sites of allografts. Instead, the allografts were represented by flat, hazy masses that, on histologic examination, revealed fragments of stromal elements that were surrounded by scar tissue and neovessels (Fig. 1B) . An inflammatory infiltrate was still present in the overlying capsule, and surrounded the disordered stromal remnants. 

Immunohistochemical analysis of histologic sections of full-thickness corneal heterotransplants showed that leukocytes (CD45⁺ cells) and class II (I-A^d)⁺ presumed antigen-presenting cells (APCs) of recipient origin had infiltrated into the stroma of allogeneic corneal grafts at 14 days.(Figs. 2A 2B ). No similar cellular infiltrate was observed in the stroma of syngeneic cornea heterografts (data not shown). Whereas a linear staining pattern of ZO-1 positive cells (indicating an intact corneal endothelium) was readily observed in 14 day syngeneic corneal grafts, no similarly intact linear layer of ZO-1⁺ cells was detectable in corneal allografts beneath the kidney capsule after 14 days (Fig. 2C) . Together, these results indicate that inclusion of the corneal epithelial layer in corneal allografts placed beneath the kidney capsule leads to acute graft destruction (within 2 weeks), including elimination of viable epithelium, keratocytes, and endothelium. Because no similar fate occurred for heterotopic syngeneic grafts and because allografts were heavily infiltrated with inflammatory cells and neovascularization, the evidence implicates immune rejection as the cause of graft destruction. 

Induction of Donor-Specific DH after Implantation of Full-Thickness Corneal Allografts

To determine whether full-thickness corneal grafts have the capacity to sensitize recipients harboring grafts beneath the kidney capsule, recipients were assayed for the acquisition of donor-specific DH. Full-thickness corneas from normal C57BL/6 donors were implanted beneath the kidney capsule of normal BALB/c mice. As a positive immunizing control, BALB/c mice received a SC injection of 10 × 10⁶ C57BL/6 spleen cells. As a positive grafting control, skin grafts were fashioned from glabrous skin of the footpad of C57BL/6 donors and implanted beneath the kidney capsule of BALB/c mice. At 2 or 4 weeks after grafting, x-irradiated (2000 R) C57BL/6 spleen cells (1 × 10⁶) were injected into the ear pinnae. Ear-swelling responses were assessed 24 and 48 hours later. The data of a representative experiment are presented in Figure 3 . Full-thickness corneal allografts beneath the kidney capsule induced donor-specific DH that was detectable at both 2 and 4 weeks after grafting. The intensity of the DH responses was comparable to that evoked by heterotopic skin allografts. 

Induction of Donor-Specific DH after Implantation of Endothelium-Deprived Corneal Allografts

To begin to delineate the corneal epithelium’s contribution to DH induction, C57BL/6 corneal grafts were prepared so that the endothelium was eliminated by scraping the posterior surface with a cotton swab. Histologic examination of these grafts revealed a total absence of endothelium and Descemet’s membrane (data not shown). Endothelium-deprived grafts were then placed beneath the kidney capsule of BALB/c mice. Positive control BALB/c mice were immunized SC with 10 × 10⁶ C57BL/6 spleen cells. At 1 or 4 weeks after implantation, the ear pinnae of these mice were challenged with irradiated C57BL/6 spleen cells and ear swelling assessed. The results presented in Figure 4 indicate that endothelium-deprived corneal allografts induced DH that was detectable at both 1 and 4 weeks after grafting. Thus, the corneal endothelium was not required for DH induction when allogeneic corneas were placed beneath the kidney capsule. 

Fate of Allogeneic Grafts of Corneal Epithelium Alone

To determine whether corneal epithelium alone is vulnerable to immune rejection, epithelial sheets were prepared from corneas of normal C57BL/6 donors. These epithelial sheets were implanted beneath the kidney capsule of normal BALB/c mice. When inspected at 1 week after grafting, none of the epithelial grafts was visible beneath the kidney capsule. Moreover, histologic examination revealed only necrotic cells at the graft site (data not shown). A similar fate was observed for epithelium-alone grafts prepared from syngeneic BALB/c donors. Because immune rejection cannot be implicated in the disappearance of syngeneic corneal epithelial grafts placed beneath the kidney capsule, we suspected that epithelium grafted in the absence of a supporting stroma was nonviable beneath the kidney capsule. Thus, we were unable to determine whether epithelial allografts alone were vulnerable to immune rejection. 

Induction of Donor-Specific DH after Implantation of Allogeneic Corneal Epithelium Alone

Although we were unable to assess epithelial allograft rejection at this heterotopic site, it was still possible to determine whether allogeneic epithelial grafts were capable of inducing donor-specific DH when implanted beneath the kidney capsule. To test this possibility, epithelial sheets were prepared from corneas of normal C57BL/6 donors and from corneas to which light thermal cautery had been applied 2 weeks previously. These epithelial sheets were implanted beneath the kidney capsule of normal BALB/c mice according to the following plan: group 1, small normal epithelium (1 × 2 mm); group 2, large normal epithelium (3 × 2 mm); group 3, cauterized epithelium (1 × 2 mm); and group 4, ear skin epidermis from C57BL/6 donors, as positive grafting control. Positive immunizing control BALB/c mice received a SC injection of 10 × 10⁶ C57BL/6 spleen cells. The ear pinnae of these mice were challenged with 1 × 10⁶ irradiated C57BL/6 spleen cells at 4 weeks after grafting. The results of a representative experiment, presented in Figure 5 , reveal that both cauterized corneal epithelium and ear skin epidermis induced intense donor-specific DH, whereas normal corneal epithelium (whether small or large) failed to induce DH. These results indicate that corneal epithelium, on its own, has no immunogenicity when implanted at a heterotopic site. However, the possibility exists that the inability of allogeneic Langerhans’ cell–deficient corneal epithelium to sensitize results from the inability of this tissue to survive at the heterotopic graft site. The following experiments addressed this point. 

Induction of Donor-Specific DH after Implantation of Chimeric Corneal Grafts Composed of Allogeneic Epithelium and Syngeneic Stroma

The failure of pure allogeneic corneal epithelial sheets to induce DH beneath the kidney capsule surprised us. Because there is considerable evidence to suggest that the viability of epithelium must be maintained by stromal influences, ¹⁷ we speculated that death of corneal epithelium implanted without stroma occurred so rapidly after implantation that the recipient immune system did not have a chance to detect and become sensitized to donor alloantigens. To explore this issue, we prepared chimeric corneas for implantation composed of pure epithelial sheets from one donor and epithelium-deprived cornea (stroma plus endothelium) from another donor. To create these composite grafts, epithelial layers were carefully placed on the stromal surface of epithelium-deprived corneas, and the combined layers were placed carefully beneath the kidney capsule of BALB/c recipients. In preliminary experiments, we found that chimeric grafts composed of BALB/c epithelium alone layered onto BALB/c epithelium-deprived corneas survived well beneath the kidney capsule. The epithelium of these composite grafts persisted for at least 4 weeks after implantation (data not shown). For the following series of experiments, three experimental groups were created: Group 1 received allogeneic epithelium layered on allogeneic stroma plus endothelium; group 2, allogeneic epithelium layered on syngeneic stroma plus endothelium; and group 3, syngeneic epithelium layered on allogeneic stroma plus endothelium. As before, donor-specific DH was assessed at 4 weeks after grafting. 

The results are presented in Figure 6 . Allogeneic epithelium cotransplanted with allogeneic stroma plus endothelium induced DH. This result resembles the immunogenicity of an intact, allogeneic corneal graft. More important, allogeneic epithelium cotransplanted with syngeneic stroma plus endothelium also induced DH. This finding formally demonstrates that allogeneic epithelium alone is immunogenic, provided that its viability is assured by an inductive stromal layer. To our surprise, chimeric grafts composed of syngeneic epithelium and allogeneic stroma plus endothelium failed to sensitize their recipients. Taken at face value, this result implies that most, or perhaps all, of the alloimmunogenicity of heterotopic corneal grafts resides in the epithelium. 

Fate of Allogeneic of Stroma Alone

To examine the inherent immunogenicity or immune privilege of corneal stroma, corneas were excised from eyes of normal C57BL/6 and BALB/c mice. Grafts were prepared in which the epithelial layer was first removed (after in vitro incubation in EDTA). Subsequently, the endothelium was scraped off the posterior surface of some grafts with a cotton swab. Histologic examination of grafts prepared in this way revealed a complete absence of both epithelium and endothelium, with partial loss of Descemet’s membrane (data not shown). These stromal grafts were placed beneath the kidney capsule of BALB/c mice and examined histologically at 1, 2, 3, and 4 weeks after grafting. When first examined at 7 days (see Fig. 7A ), the grafts contained neither epithelial nor endothelial elements. Moreover, the stroma of allografts was infiltrated with mononuclear cells, disrupting the collagen lamellae. No similar infiltration was observed in syngeneic stromal grafts. By 14 days, the stromal allografts displayed increasingly disarrayed lamellae. Viable keratocytes were no longer identifiable in these grafts. Immunohistochemical analysis revealed that CD45⁺ leukocytes were present, a portion of which were I-A^d positive (presumptive antigen presenting cells). The density of such cells increased in stromal allografts through time (see Fig. 7B ). Syngeneic stromal grafts remained stable and quiescent throughout a similar time interval. The lamellar arrays were maintained, and no infiltrating leukocytes were detected. These results indicate that allogeneic stroma by itself is vulnerable to immune rejection. 

Immunogenicity of Allogeneic Grafts of Stroma Alone

Mice bearing allogeneic grafts of stroma alone beneath the kidney capsule were tested for acquisition of donor-specific DH. The ear pinnae of recipient mice were challenged with 1 × 10⁶ x-irradiated (2000 R) C57BL/6 spleen cells at 1, 2, and 3 weeks after grafting. Positive control BALB/c mice were immunized SC with 10 × 10⁶ C57BL/6 spleen cells 1 week before ear pinnae challenge. The results of a representative experiments are presented in Figure 8 . Mice bearing stroma-alone allografts for 1 week displayed intense ear-swelling responses, indicating the presence of DH. However, stroma alone graft-bearing mice similarly tested at 2 weeks and thereafter showed no significant ear-swelling responses. On the one hand, these results indicate that allogeneic stroma alone was immunogenic—i.e., it induced donor-specific DH. On the other hand, the immunity evoked by stroma was short-lived, whereas immunity evoked by epithelium plus stroma proved to be long lasting (Fig. 4) . Thus, although allogeneic stroma stimulated memory-impaired sensitization, it was sufficient to cause allografts of stroma alone to be rejected. 

Immunogenicity of Allogeneic Grafts of Stroma Plus Endothelium

Epithelium-deprived corneal allografts placed beneath the kidney capsule survive indefinitely, as we have shown in our previous report, ¹⁴ whereas grafts of allogeneic stroma induce DH and succumb to immune rejection. We suspected that this difference in outcome arises from unique properties of the corneal endothelium. In the next experiments, we tested the ability of epithelium-deprived corneal allografts (stroma plus endothelium) placed beneath the kidney capsule to induce DH. C57BL/6 corneas from which epithelium had been removed were placed beneath the kidney capsule of BALB/c mice. Companion mice received heterotopic grafts of allogeneic stroma alone. One week later, both groups of mice, plus positive controls, received ear pinnae challenge with x-irradiated C57BL/6 spleen cells. As revealed by the results of a representative experiment displayed in Figure 9 , DH was detected in mice bearing stroma-alone allografts, but not in mice bearing stroma–endothelium allografts. These findings not only confirm that endothelium protects epithelium-deprived grafts from immune rejection, but it acts by preventing the recipients from acquiring donor-specific DH. It is relevant that endothelium of epithelium-deprived corneas still express CD95L. ¹⁴  

Role of CD95L in Impaired Immunogenicity of Epithelium-Deprived Grafts

Although epithelium-deprived corneal allografts prepared from eyes of normal mice enjoy unlimited survival beneath the kidney capsule, similar grafts prepared from eyes of mice deficient in CD95L expression (B6-gld) experience no privilege at this heterotopic site. Grafts of this type are rejected within 2 weeks. ¹⁴ Our next experiments examined the capacity of CD95L-deficient grafts to induce donor-specific DH. Three recipient strains were used: BALB/c mice that recognize both MHC and minor histocompatibility (H) antigens on C57BL/6 tissues, B10.D2 mice that share the majority of minor antigen alleles with C57BL/6 but recognize the class I and II antigens encoded by the H-2^d chromosome, and BALB.B mice that recognize minor H antigens, but not MHC antigens on C57BL/6 tissues. These mice received epithelium-deprived corneal grafts from B6-gld donors or from wild-type C57BL/6 donors. Four weeks later, the ear pinnae of the recipients, as well as their respective positive controls, were challenged with C57BL/6 spleen cells. The results of these experiments are displayed in Figure 10 . BALB/c mice that received CD95L-deficient grafts acquired donor-specific DH, whereas recipients of similar grafts from normal mice displayed insignificant ear-swelling responses (Fig. 10A) . Similarly, B10.D2 mice that received CD95L-deficient, but not normal, grafts showed development of DH (Fig. 10B) . Alternatively, the ear-swelling responses of BALB.B mice bearing CD95L-deficient grafts were not significantly different from the responses of mice bearing normal grafts, and both were indistinguishable from the negative controls. We conclude that endothelium, presumably through constitutive expression of CD95L, prevents allogeneic stroma from inducing donor-specific DH. Moreover, when CD95L is deficiently expressed by stromal–endothelial allografts, only MHC alloantigens (not minor H antigens) give rise to DH. 

Fate of Epithelium-Deprived Corneal Allografts in Presensitized Mice

To this point, our experimental evidence concerns primarily the capacity of stromal and stromal–endothelial grafts to induce allosensitization when placed heterotopically. The critical role for CD95L revealed by these experiments is in the afferent limb of the immune reflex arc. Yet, CD95L is known primarily for its capacity to induce apoptosis among CD95⁺ T cells that bind to target cells and attempt to kill them—an effect on the efferent limb. ¹⁰ By placing cornea-derived grafts heterotopically in mice presensitized to donor antigens, we hoped to find evidence of an efferent effect of CD95L in this system. Accordingly, BALB/c mice were immunized SC with 10 × 10⁶ C57BL/6 spleen cells. One week later, epithelium-deprived cornea grafts from wild-type and B6-gld donors were placed beneath the kidney capsule of these presensitized mice. Seven days later the grafts were inspected clinically and then removed and examined histologically. CD95L-deficient grafts appeared swollen and opaque, with evidence of neovessels growing at the graft margins. By microscopy, sections of these grafts revealed an intense stromal infiltration with inflammatory cells. No linear array of cells staining positively with anti ZO-1 antibodies (which recognize a molecule associated with tight junctions) was detected (data not shown). Stromal–endothelial grafts prepared from normal eyes also contained an intense stromal infiltrate of inflammatory cells that were CD45⁺ and I-A^d positive (see Fig. 11A ). However, the endothelium of these grafts was preserved, as revealed by a linear pattern of ZO-1 staining (see Fig. 11B ). These findings reveal that CD95L-bearing corneal endothelial cells resist destruction when confronted by primed donor-specific T cells in presensitized hosts. In the absence of CD95L, however, endothelium is destroyed by these immune cells. Finally, in the presensitized state, CD95L-expressing endothelium appears to be incapable of protecting the adjacent stroma from immune damage. 

Discussion

The results indicate that the different layers of the normal cornea display either immunogenicity or immune privilege. Moreover, they indicate that the properties of one layer can influence the properties and fate of another layer. This information may be useful in understanding why corneal allografts placed orthotopically in the eye are sometimes rejected and sometimes accepted. 

Sources of Alloimmunogenicity within the Cornea

The experimental evidence described indicates that the epithelial layer of the cornea is a potent source of immunogenicity in corneal grafts placed heterotopically beneath the kidney capsule. If corneal tissue for grafting was removed from normal eyes of donors and placed as a full-thickness graft beneath the kidney capsule of allogeneic mice, the grafts readily sensitized the recipients to donor alloantigens (DH), and the grafts soon became the targets of a destructive alloimmune rejection reaction. We have previously reported that allogeneic corneas deprived of epithelium and placed beneath the kidney capsule did not undergo immune rejection during a prolonged follow-up interval. ¹⁵ Taken together, these results lead to the reasonable hypothesis that the epithelium is primarily responsible for the alloimmunogenicity of heterotopic corneal grafts. 

Our attempts to prove the validity of this hypothesis by using individual layers of the cornea as heterotransplants revealed relationships among corneal epithelium, stroma, and endothelium that were not particularly obvious at the outset. For example, sheets of pure allogeneic corneal epithelium failed to sensitize their recipients through the kidney capsule. The reason for this failure turned out to be nonimmunologic. Sheets of pure corneal epithelium failed to survive at this heterotopic site, even in syngeneic recipients, and this failure was reversed if the epithelial sheets were cotransplanted with a layer of stroma plus endothelium. When this method was used to preserve the viability of allogeneic corneal sheets (by cotransplanting allogeneic cornea with syngeneic stroma-endothelium), allogeneic corneal epithelium readily induced donor-specific DH. We consider this to be formal proof of the alloimmunogenicity of corneal epithelium at this heterotopic site. 

It is of considerable interest that pure epithelial sheets prepared from corneas cauterized 2 weeks previously also sensitized their recipients when placed beneath the kidney capsule. Epithelial sheets from cauterized corneas differ from sheets removed from normal eyes, in part because of their content of Langerhans’ cells. ¹⁵ Although epithelial sheets from cauterized eyes underwent rapid necrosis beneath the kidney capsule (similar to epithelium from normal eyes; data not shown), epithelial grafts from cauterized eyes nonetheless induced DH in their recipients. Because epithelium from cauterized eyes contains Langerhans’ cells, because Langerhans’ cells are highly mobile bone marrow–derived dendritic cells, and because Langerhans’ cells have been strongly implicated in the induction of alloimmunity after orthotopic skin and corneal grafts, ^{18 19} we speculate that sensitization in this instance occurred because Langerhans’ cells were able to escape from the graft before the epithelial component became nonviable. 

The idea that epithelium is the primary source of alloimmunogenicity in full-thickness corneal grafts is not new. Tuberville et al. ²⁰ championed this concept more than a decade ago and reported that human corneas deprived of epithelium enjoyed better survival in keratoplasty. However, this view was refuted by a similar study in human beings conducted by Stulting et al. ²¹ We suspect, but have no direct information, that the physical method by which epithelium is removed from the donor cornea may have secondary consequences that are deleterious to graft survival. Preliminary experiments in our laboratory revealed that orthotopic corneal allografts deprived of epithelium became rapidly neovascularized and were rejected (J. Hori, unpublished data, 1999). We suspect that these grafts caused rapid sensitization, and we are now testing this possibility. That a clinical entity called epithelial rejection is well described after human keratoplasty indicates that the epithelium itself can often be the direct target of immune destruction. Our studies underline the capacity of corneal epithelium to be both an inducer and a target of allosensitization. 

Our experiments have focused on DH as the measure of sensitization by corneal epithelium beneath the kidney capsule. We are aware that other parameters of sensitization can be assayed (cytotoxic T cells, antibodies, lymphocyte proliferation in vitro). However, at least in the mouse model system, rejection of orthotopic corneal allografts correlates best with the activities of donor-specific CD4⁺ T cells of the type that mediate DH. ^{22 23} Moreover, studies of fragments of allogeneic corneas implanted in the anterior chamber have revealed that donor-specific DH is only induced if the fragment contains an epithelial layer. ²⁴ We believe that there is a direct, but molecularly undefined, link between epithelial cells and the promotion of DH. Wounded epithelial cells are known to secret cytokines that promote angiogenesis, leukocyte migration, and inflammation. ^{25 26} We suspect that one chain in that link is interleukin (IL)-1, because topical treatment of orthotopic corneal allografts with IL-1 receptor antagonist prevents systemic sensitization to donor alloantigens. ^{27 28} Because IL-1α is made in large amounts by traumatized corneal epithelium, ²⁹ this may be the signal that initiates the sensitization cascade. Whether the target of IL-1α is an antigen-presenting Langerhans’ cell, immigrating macrophage, or budding vascular endothelial cells remains to be determined. 

Sources of Immune Privilege in the Cornea

Immune privilege is multifactorial in origin. Unique features of these tissues and sites function passively in creating immune privilege, and molecules expressed either as soluble factors or cell surface molecules actively maintain the privileged status. For so-called immune privileged tissues, the expression of the privileged phenotype is dependent on the site into which the tissue is grafted. When the cornea is placed orthotopically, it benefits from the site itself and displays robust immune privilege. When the corneal graft is placed in or on the skin, it displays few features of immunologic privilege. When placed beneath the kidney capsule, the cornea’s potential to display immune privilege is manifest, and our current results reveal that this property is largely derived from the endothelium. 

The current studies bring to completion our survey of the immune-privileged status of the cornea when placed as an allograft beneath the kidney capsule. To summarize what has been learned, both the epithelium alone and the stroma alone display immunogenic potential. Either on its own is capable of inducing donor-specific DH and of succumbing to immune destruction. Only the endothelium appears to be without these properties. To the contrary, the endothelium prevents allosensitization promoted by the stroma in naive mice, and the endothelium is even able to resist its own elimination when grafted into mice presensitized to donor alloantigens. Thus, at least in this heterotopic grafting model, the immune privilege of the cornea resides solely with the endothelium. However, the power of the endothelium to promote immune privilege of the cornea is overwhelmed if epithelium is included in the graft placed beneath the kidney capsule. In this situation, the overpowering immunogenicity of the epithelium, perhaps residing in its capacity through IL-1α secretion to provoke inflammation, prevents corneal allografts from surviving at the heterotopic site. Because full-thickness corneal allografts often survive indefinitely when placed orthotopically, the potent immunogenicity of corneal epithelium revealed beneath the kidney capsule is at least partially eclipsed in the eye. 

Constitutive expression of CD95L on corneal endothelium is critical to its immune-privileged status. Our evidence confirms that CD95L renders corneal endothelium resistant to immune destruction, revealed by persistence of endothelial cells at heterotopic sites of presensitized mice where stroma is destroyed. We are aware of the demonstration by Tagawa et al. ³⁰ who have shown that presensitized lymphoid cells injected into the anterior chamber of rabbits produces keratic precipitates. Whether these precipitates represent alloimmune destruction is unclear. Moreover, we do not know whether rabbit corneal endothelium expresses CD95L. In addition, our evidence indicates an immunomodulatory role for CD95L in the induction of alloimmunity. Stromal–endothelial allografts only induced systemic donor-specific DH (and their rejection) if the grafts failed to express CD95L. We conclude that CD95L, perhaps by triggering apoptosis in naive alloreactive T cells, prevents allosensitization. 

The capacity of CD95L expression on allografts of stroma-endothelium to prevent sensitization to MHC alloantigens bears further comment. It is pertinent that CD95L-deficient stromal–endothelial grafts did not induce DH to minor histocompatibility antigens. Why would CD95L expression differentially influence sensitization to MHC and minor transplantation antigens? We suspect that the answer is related to the mechanism by which MHC and minor antigens are first detected by alloreactive T cells. MHC alloantigens can be directly recognized on graft cells by certain populations of T cells, through the so-called direct pathway of allorecognition. Minor H antigens cannot be directly recognized by T cells, but must be taken up by recipient APCs and presented as peptides in the context of recipient class I and II molecules. This is called the indirect pathway of allorecognition. In fully allogeneic grafts, such as we used in these experiments, indirect alloreactive T cells can only be sensitized when recipient APCs infiltrate the graft, capture graft antigens, and present them to naive recipient T cells in draining lymph nodes. 

In our experiments, fully allogeneic stromal–endothelial grafts placed beneath the kidney capsule did not induce donor-specific DH unless the grafts failed to express CD95L. When DH emerged in this circumstance, the only alloreactive T cells detected were directed at MHC rather than minor H antigens. Because our corneal grafts were utterly devoid of Langerhans’ cells (or any other donor-derived APCs), and because corneal parenchymal cells are incapable of migrating through lymph to draining lymph nodes, the only site at which MHC-specific, direct alloreactive T cells can become activated is the graft site itself, the subcapsular sinus of the kidney. Sensitization in this manner is called peripheral sensitization and was first postulated as a mechanism of allograft sensitization by Medawar in 1965. ^31. Peripheral sensitization has not gained popularity among transplantation immunologists, primarily because in most experimental systems sensitization to solid tissue allografts occurs predominately through passenger leukocytes. ³² Of course, the cornea has no passenger leukocytes, and therefore central sensitization by corneal allografts must await infiltration of the graft by recipient APCs. We propose that, at least for the cornea, direct alloreactive T cells can be sensitized de novo by stromal–endothelial cell-containing grafts, and that constitutive expression of CD95L on grafts from normal mice eliminates these cells through programmed cell death. For this reason, systemic evidence of sensitization never emerges, unless CD95L is missing. 

 Supported by National Institutes of Health Grant EY10765. JWS is a recipient of a Research to Prevent Blindness Senior Scientific Investigator Award.

 Submitted for publication January 10, 2000; revised April 25, 2000; accepted May 10, 2000.

 Commercial relationships policy: N.

 

Figure 1.

 

Histologic aspects of syngeneic (BALB/c) (A) or allogeneic (C57BL/6) (B) full-thickness corneal graft segments at 10 weeks beneath the kidney capsule of BALB/c mice. Arrows: (A) Epithelial mass shed from the surface of the proliferating epithelium; (B) a neovessel in the kidney capsule adjacent to the graft and a mononuclear cell accumulation adjacent to the graft (large arrowhead) in (B). Inset: High-power image of mononuclear cell accumulation and neovessel. K, kidney; CS, corneal stroma. Hematoxylin and eosin stain; magnification, (A, B) ×33; inset (B) ×100.

Figure 2.

 

Presence of CD45- (A) and I-A^d- (B) positive cells in allogeneic (C57BL/6) full-thickness cornea graft placed beneath the kidney capsule of BALB/c mouse at 14 days. Conventional fluorescence microscope images of cross sections display staining within the same section with anti-CD45 and anti-I-A^d monoclonal antibodies. (C) Immunolocalization of ZO-1 in allogeneic (C57BL/6) full-thickness cornea placed beneath the kidney capsule of BALB/c mouse at 14 days, stained with FITC–anti-ZO-1 antibody, using confocal imaging. Arrow: Site where linear deposit is missing. Inset: ZO-1-positive staining on endothelium of a syngeneic corneal graft at 14 days. Arrows: ZO-1 linear staining. K, kidney; CS, corneal stroma. Magnification, (A, B)× 57; (C) ×144.

Figure 3.

 

Induction of donor-specific DH after implantation of C57BL/6 full-thickness corneal allografts beneath the kidney capsule of BALB/c mice (B6cornea) at 2 weeks (A) or 4 weeks (B). C57BL/6 footpad skin was implanted as a skin grafting control (B6skin). Positive immunization controls (Pos.C) received SC injection of 10 × 10⁶ donor spleen cells 1 week before assay. Ear pinnae received injection of x-irradiated C57BL/6 spleen cells (1 × 10⁶), and ear-swelling responses were assessed 24 and 48 hours later. Negative control received ear pinnae challenge only. Mean 24-hour ear-swelling responses are compared with negative controls (Neg.C). Significantly greater than negative control (* P < 0.05; **P < 0.01).

Figure 4.

 

Induction of donor-specific DH after implantation of C57BL/6 endothelium-deprived corneal allografts beneath the kidney capsule of BALB/c mice at 1 week (St-Epi, 1w) and 4 weeks (St-Epi, 4w). Positive controls (Pos.C), negative controls (Neg.C), and ear pinnae challenge were similar to those described in Figure 3 . Mean ear-swelling responses are compared with negative controls. Significantly greater than negative control (*P < 0.001).

Figure 5.

 

Induction of donor-specific DH after implantation of C57BL/6 corneal epithelium grafts beneath the kidney capsule of BALB/c mice at 4 weeks. Untreated epithelium (naive epi; 1 × 2 mm), large untreated epithelium (large naive epi; 3 × 2 mm), cauterized epithelium (cauterized epi; 1 × 2 mm), and epidermis from ear skin (skin epi) were implanted as a skin grafting control. Positive controls (Pos.C), negative controls (Neg.C), and ear pinnae challenge were similar to those described in Figure 3 . Mean ear-swelling responses are compared with negative controls. Significantly greater than negative control (*P < 0.005, **P < 0.001).

Figure 6.

 

Induction of donor-specific DH after implantation beneath the kidney capsule of chimeric corneal grafts composed of allogeneic epithelium and syngeneic stroma at 4 weeks. Allogeneic (C57BL/6) epithelium layered on allogeneic stroma plus endothelium (allo-Epi, allo-St/Ed), allogeneic epithelium layered on syngeneic (BALB/c) stroma plus endothelium (allo-Epi, syn-St/Ed), and syngeneic epithelium layered on allogeneic stroma plus endothelium (wyn-Epi, allo-St/Ed) were implanted beneath the kidney capsule of BALB/c mice. Positive controls (Pos.C), negative controls (Neg.C), and ear pinnae challenge were similar to those described in Figure 3 . Mean ear-swelling responses are compared with negative controls. Significantly greater than negative control (*P < 0.05, **P < 0.01).

Figure 7.

 

(A) Histologic aspects of an allogeneic (C57BL/6) corneal stroma alone graft beneath the kidney capsule of a BALB/c mouse at 7 days. Arrows: Cellular infiltration, apparently from the surface of kidney into the corneal stroma. (B) Presence of I-A^d-positive cells in allogeneic (C57BL/6) corneal stroma placed beneath the BALB/c kidney capsule at 21 days. Conventional fluorescence microscope images of cross sections. K, kidney; CS, corneal stroma; KC, kidney capsule. Hematoxylin and eosin stain; magnification, (A) ×66, (B) ×114.

Figure 8.

 

Induction of donor-specific DH after implantation of C57BL/6 corneal stroma-alone allografts beneath the kidney capsule of BALB/c mice at 1 week (1w), 2 weeks (2w), or 3 weeks (3w). Positive controls (PosC), negative controls (Neg.C), and ear pinnae challenge are similar to those described in Figure 3 . Mean 24-hour ear-swelling responses are compared with negative controls. Significantly greater than negative control (*P < 0.05).

Figure 9.

 

Induction of donor-specific DH after implantation of C57BL/6 corneal stroma alone (St.alone, 1w) and stroma plus endothelium allografts (St.end, 1w) beneath the kidney capsule of BALB/c mice at 1 week. Positive controls (Pos.C), negative controls (Neg.C), and ear pinnae challenge were similar to those described in Figure 3 . Mean ear-swelling responses are compared with negative controls. Significantly greater than negative control (*P < 0.002).

Figure 10.

 

Induction of donor-specific DH after implantation of corneal stroma plus endothelium of C57BL/6 (B6st.-ed) or B6-gld (B6-gld.-ed), beneath the kidney capsule of BALB/c mice (A) or B10 D2 mice (B) at 4 weeks. Positive controls (Pos.C), negative controls (Neg.C), and ear pinnae challenge were similar to those described in the legend to Figure 3 . Mean ear-swelling responses are compared with negative controls. Significantly greater than negative control (*P < 0.01).

Figure 11.

 

(A) Presence of CD45-positive cells in allogeneic (C57BL/6) corneal stroma plus endothelium grafts placed beneath the kidney capsule of BALB/c mice presensitized to B6 alloantigens, at 14 days after grafting. Arrows: CD45⁺ cells infiltrating into corneal stroma. Conventional fluorescence microscope images of cross sections. (B) Immunolocalization of ZO-1 in graft described in (A) at 14 days, stained with FITC–anti-ZO-1 antibody and using confocal imaging. Arrow: Linear staining pattern of ZO-1 on corneal endothelium. K, kidney; CS, corneal stroma. Magnification, (A, B) ×114.

The authors thank Jacqueline Doherty and Jian Gu for their contribution to our research efforts.