A Model of Corneal Graft Rejection in Semi-Inbred NIH Miniature Swine: Significant T-Cell Infiltration of Clinically Accepted Allografts

Abstract

Purpose.: The purpose of our study is to develop a pre-clinical model of corneal graft rejection in the semi-inbred NIH minipig as a model of human rejection.

Methods.: NIH minipigs received corneal allografts with MHC and minor mismatches, or minor mismatches alone. Clinical rejection was monitored, and major subsets of leukocytes and ingress of vessels were quantified post-mortem by automated digital methods. Spectratypes of recipient T-cell receptor β-subunit variable region (TRβV) were analyzed. The capacity of pig corneal endothelial cells to proliferate in vivo was assessed.

Results.: Autografts (n = 5) and SLA^cc to SLA^cc allografts (minor mismatches, n = 5) were not rejected. Median graft survival of SLA^dd and SLA^bb allografts in SLA^cc strain recipients (major and minor mismatches) was 57 (n = 10) and 67 (n = 6) days, respectively. Rejected grafts did not recover clarity in vivo, and corneal endothelial cells did not proliferate in organ culture after cryo-injury. There were significantly more leukocytes in clinically rejected versus accepted grafts (P < 0.0001) and in transplanted versus contralateral eyes (P < 0.0001). Numbers of T-cells were significantly greater in clinically accepted grafts versus autografts and in rejected grafts versus accepted (P < 0.005 for most subsets). There were significant differences in TRβV spectratype between graft groups in cornea, but not in draining lymph node or blood (P < 0.05).

Conclusions.: The NIH minipig offers a robust model of human rejection suitable for immunological or therapeutic studies. In particular, there is limited capacity for corneal endothelial repair in vivo, and histological evidence suggests that allosensitization of the recipient may develop in the absence of clinical rejection.

Introduction

Rodent models have underpinned experimental research into graft rejection since transplantation techniques in rats and mice were perfected more than 20 years ago. Indeed, the use and increasing sophistication of murine gene knock-out and transgenic technologies has resulted in an exponential increase in our understanding of the cellular and molecular interactions governing immunity to many diseases, and has indicated many possible therapeutic intervention targets and strategies. ^1–3 Despite these advances, many novel therapies showing promise in rodents have yet to be translated successfully into clinical practice. Likely contributory factors to this outcome, albeit not exclusively, include the small size, higher metabolic rates and short life spans of rodents. The operation of the immune response, and the effects of drugs and other treatments may be qualitatively different from those in larger species, ⁴ as may be the efficiency of transduction of cells with novel genes and the persistence of those genes in tissues. ⁵ Moreover, in transplantation, inflammatory and healing responses may be modified by differences in surgical techniques, for instance in rodent corneal transplantation, apposition of donor and recipient cornea lacks precision, and suture knots are not “buried” in the corneal stroma. A further compounding issue is that the mitotic capacity of corneal endothelial cells in vivo appears to be greater in small mammals than in man. ^6,7 In view of these considerations, as well as the current increasing possibilities for therapeutic intervention in graft rejection and a clear clinical unmet need, there is an evident demand to develop “pre-clinical” models in larger species that will verify or refute findings from rodent research, and confirm potential therapeutic efficacy before human trials are initiated. 

The most frequently used alternative species for corneal graft rejection studies has been the rabbit. However, as in the rat, rabbit corneal endothelial cells can proliferate, restoring the clarity of a graft after rejection. Cat models, in which corneal endothelial cells essentially are amitotic in vivo, ⁸ and, more recently, sheep and outbred pig models have been reported, ^9,10 where sheep have been used successfully to test the effect on rejection following transduction of the cornea with cytokine genes. ¹¹ However, the supply of reagents for immunological studies in cats and sheep is limited compared to pigs, and no inbred strains are available. The productivity of sheep is limited to one or two offspring per year, whereas pigs have a 3-week estrus cycle, a relatively short (16-week) gestation period, and large litter size, so can be produced in relatively large numbers. There are pig reagents to identify the major subsets of leukocytes and Th1/Th2 cytokines, ¹² and it is the only large animal in which there are inbred strains, permitting the study of immunity to defined major histocompatibility (MHC) mismatches. For these reasons we set up a model of corneal graft rejection in the semi-inbred NIH minipig. We examined graft rejection in two strain combinations bearing histocompatibility mismatches at major and minor loci, and within a single strain exhibiting only minor locus disparities. The model has been validated with respect to clinical features of rejection, immunohistology, T-cell receptor β-subunit variable region (TRβV) spectratyping, and the capacity of corneal endothelial cells to proliferate in vivo. 

Methods

Pigs

SLA^cc and SLA^dd strains of NIH minipigs ¹³ and large (SLA^bb) Babraham pigs ¹⁴ were obtained from the Institute of Animal Health (Compton, UK). All three strains are fully inbred at MHC loci, but retain intra-strain minor histo-incompatibilities. The two minipig herds were derived from a breeding nucleus of 2 boars and 4 gilts of each strain imported from the United States in 1992. Individuals within each strain have since been bred randomly to retain residual genetic heterozygosity However, the extent of heterozygosity is uncertain. The Babraham strain was inbred in the United Kingdom to MHC homozygosity from outbred Large White pigs. All experiments were performed in accordance with the ARVO statement for the Use of Animals in Ophthalmic and Vision Research, with UK Scientific Procedures legislation and Medical Research Council policy to minimize numbers of animals used in scientific research. 

Anesthesia, Surgery, and Peri-Operative Care and Monitoring

Socialization protocols used to prepare pigs for surgery and postoperative intervention, together with anesthesia and sedation protocols, have been described previously. ¹⁵ Preparation of the donor cornea, surgical procedure, and postoperative assessment of corneal inflammation and rejection are described in Supplementary Methods (see Supplementary Material). 

Experimental Protocol

The lymph node draining the cornea was identified first by injection of methylene blue dye (0.5 mL at 3%) into the upper and lower conjunctiva of both eyes of two pigs under deep anesthesia. The conjunctiva was massaged periodically for up to 30 minutes, after which the pig was killed. The neck and throat region was dissected to expose local lymph nodes. Blue coloration was located only in the parotid lymph node, which was sampled thereafter post-mortem from graft recipients for TRβV spectratype analysis. 

Central corneal buttons 7 mm in diameter of SLA^cc or SLA^dd strain minipigs, or Babraham pigs were transplanted orthotopically to SLA^cc strain minipigs, aged 3–5 months (18–40 kg). Graft groups and post-mortem fate of recipients for immunohistology and TRβV spectratyping are shown in Figure 1. No postoperative immunosuppression was administered, and the pattern of allograft rejection was monitored. Six of ten recipients of rejected allografts were killed between 5 and 14 days after rejection onset, while four pigs (three b to c strain and one d to c strain) were retained for longer to determine whether corneal clarity was regained (see below). Recipients of accepted allografts were retained for at least 84 days, with the exception of two pigs, killed at 60 and 64 days, respectively, because the initial terms of the experimental license under United Kingdom legislation did not permit them to be retained for longer than 65 days. Central corneal buttons 8 mm in diameter (to include the donor and graft-host junction) were removed immediately post-mortem from transplanted and contralateral eyes, then bisected. One half was processed for immunohistology and the other half for TRβV spectratyping. Corneas from non-transplanted pigs were prepared likewise, for histology only (Fig. 1). Tonsil samples were removed as positive control tissue for immunohistology. Additional tissues processed for TRβV spectratyping were samples of ipsilateral and contralateral parotid lymph node, blood collected from the jugular vein before transplantation and intracardiac blood collected immediately post-mortem. In vitro studies of corneal endothelial repair were performed on corneas from outbred pigs obtained at slaughter. 

Figure 1.

 

Flow diagram showing fate of transplanted corneas. *Left and right eyes were processed. †Left eye was transplanted.

Three-Color Quantitative Immunofluorescence Histology

Cornea and tonsil tissue was snap-frozen in OCT cryo-embedding compound (Raymond Lamb, Loughborough, UK) and stored at −70°C. Cryosections of cornea and tonsil were processed for immunofluorescence histology, as described in Supplementary Methods (see Supplementary Material). Three combinations of 3 antibodies, linked via immunoglobulin (Ig) isotype- or subclass-specific secondary antibodies to red, green, or blue fluorochromes, were applied to identify antigen presenting cells, T-cells and vessel endothelium. Target antigens were grouped into panels as follows (antibody clones and isotypes in parentheses): 

1.  
   Monocytes/macrophages: CD163 (2A10, ¹⁶ IgG1), CD14 (MIL-2, ¹⁷ IgG2b), and MHC class II (MSA3, ^18,19 IgG2a);
2.  
   T lymphocytes: CD3 (FY1H2, ²⁰ IgG1), CD4 (MIL-17, ¹² IgG2b), and CD8α (MIL-12, ¹² IgG2a); and
3.  
   Dendritic cells and vessels: CD16 (G7, ²¹ IgG1), MHC class II (MSA3), and (MIL-11, ²² an unidentified antigen on vessel endothelium, IgE).

Ten images of stroma per cornea at ×200 magnification, sited immediately beneath the epithelium and excluding the area immediately around sutures, were captured in gray scale. Numbers of pixels above background fluorescence, representing numbers of labeled cells, were counted using ImageJ software (http://rsbweb.nih.gov/ij) ²³ as described in Supplementary Methods (see Supplementary Material). Epithelium and endothelium were examined, but leukocyte infiltration was not quantified. 

TRβV Spectratyping of T-Cells

Samples of cornea and parotid lymph node were collected separately into RNALater (Ambion Ltd., Huntingdon, UK) and stored at −70°C until RNA extraction. Blood samples were collected into PAXgene blood RNA tubes (PreAnalytiX GmBH, Hombrechtikon, Germany) and stored at −20°C until RNA extraction. For sample processing and generation of TRβV spectratype electropherograms, see Supplementary Methods (see Supplementary Material). 

Proliferation of Corneal Endothelial Cells

This was investigated via 2 approaches: (1) four transplants were monitored clinically after rejection to determine whether opacity and edema resolved, and (2) excised corneas were subjected to aseptic freezing injury, and cultured in vitro to monitor physical wound repair, either by silver staining, ²⁴ followed by light microscopy, or by immunofluorescence labeling with an antibody recognizing the replication-associated transcription factor, Ki67, followed by confocal imaging (see Supplementary Methods in Supplementary Material). 

Statistical Analysis

Unless otherwise stated, statistical analysis was performed using PASW statistics 18 (IBM, NY). Heat maps were generated using R (http://www.r-project.org). Data were analyzed by factorial ANOVA, general linear models (GLM) or principal component analysis (PCA). All GLM were fully factorial, followed by post-hoc testing using least significant differences (LSD). PCA was performed for a maximum of 25 iterations for convergence using an unrotated factor solution. Cell infiltrate data was log₁₀ transformed for analysis. 

The fixed factors used in statistical analysis were “eye” (transplanted or contralateral), “graft outcome” (accepted or rejected), “tissue” (TRβV spectratyping only), “graft group,” or “rejection group.” For “graft group,” corneal grafts belonged to one of three allograft groups (SLA^cc recipients of SLA^bb, SLA^dd, or SLA^cc strain grafts), an autograft (SLA^cc) group or non-transplanted corneas (SLA^cc or SLA^dd, histology only). “Rejection group,” consisted of corneas classified into full mismatch accepted (d to c and b to c), full mismatch rejected (d to c and b to c), minor mismatch (c to c, accepted), or autograft groups. 

Results

Clinical Observations

Mild corneal edema and opacity developed during the first week after transplantation, but had resolved when corneas were examined by slit-lamp on day 14. In all recipients, neo-vessels reached the graft-host junction in at least one quadrant (upper or lower) by 14–21 days after transplantation. In accepted grafts, vessels failed to penetrate the donor, frequently curving around the graft scar, while opacity, confined to suture tracks (Fig. 2A, day 89), was identified by histology as a predominantly non-T-cell infiltrate (data not shown). Six of ten SLA^dd strain corneas and 4/6 SLA^bb strain corneas underwent rejection in the form of donor opacification within 90 days of transplantation (medians day 57 and 67, respectively; Table 1 and Fig. 2). All other grafts remained clear until pigs were euthanized as indicated (Table 1). Typically, rejection involved increasing edema and opacity, anterior chamber flare, hyperemia, an epithelial rejection line, and growth of superficial and deep stromal vessels from graft margin to the center of the donor cornea (9/10 rejected corneas, Fig. 2A). The epithelial line formed close to the periphery near the leading edge of superficial vessels, gradually moving across the cornea and frequently forming a complete circle, as has been documented previously in rabbits ²⁵ and rats. ²⁶ Endothelial rejection lines usually were not distinguishable, but there was histological evidence of endothelial rejection (Fig. 3). In the inferior cornea of one pig, an endothelial rejection line was observed between days 48 and 54 after transplantation without other signs of anterior chamber inflammation, hyperemia or epithelial rejection (Fig. 2B). It was accompanied by overlying stromal edema but did not progress across more than 1/3 of the diameter of the cornea before resolving. However, the edema persisted until the pig was euthanized on day 75. One suture became untied in each of 2 corneas, approximately 14 and 34 days after transplantation, respectively. They were removed under sedation, without anterior chamber leakage or other adverse effect. Three graft recipients were excluded from the study (Fig. 1). 

Figure 2.

 

(A) Typical rejection sequence showing development on day 41 after transplantation of inferior opacity, and growth of superficial and deep stromal vessels from inferior graft margin to center of donor cornea, accompanied on days 41 and 50 by occasional hemorrhage over suture tracks. An epithelial rejection line, developing inferiorly on day 41, has progressed to the center by day 50 (arrow) and corneal opacity has increased. By day 90, hyperemia has diminished, but the donor cornea remains opaque. By contrast, an accepted cornea on day 89 remains clear apart from opacity along suture tracks, with vessels growth only to the graft margin. (B) Atypical rejection sequence involving endothelial layer only (1/10 pigs), seen as partial corneal opacity (overlying an inferior endothelial rejection line, which is not visible). Although the rejection line resolved, opacity persisted until the pig was killed on day 75. Note lack of vessel ingress, hyperemia, or epithelial rejection line. Light colored vertical lines on the superior cornea are reflection artefacts from eyelashes.

Table 1.

 

Survival of Corneal Transplants in SLA^cc Strain Minipigs

Donor Strain (mismatch)	Graft Survival (days after transplant)
SLAdd (major and minor)	30, 41, 43, 45, 56, 57, >64,* >84, >84, >90
SLAbb (major and minor)	48,† 53, 62, 72, >89, >89
SLAcc (minor)	>60,* >86, >86, >90, >90
SLAcc autograft (none)	>34, >35, >44, >48, >62, >62

*  See text for explanation for premature sacrifice of recipients of two accepted grafts.

†  Endothelial rejection only.

Figure 3.

 

Immunofluorescence images showing infiltrations of all layers of donor cornea during rejection. (A–C, G, I) T-cell panel (red CD3, green CD4, and blue CD8α). (D, J) APC panel (red CD163, green CD14, and blue MHC class II). (E, F) Dendritic cell/vessel panel, no vessels present (red CD16, green MHC class II). (H, K) Dendritic cell panel with vessels (red CD16, green MHC class II, blue MIL11⁺ vessel endothelium). Superimposed white line (A–D) shows boundary between epithelium and stroma. (A) Autograft showing few stromal T-cells. (B) Accepted graft showing a mixed stromal T-cell infiltrate. (C) Rejected graft showing T-cells in epithelial rejection line. (D) Same rejected graft as in (C), showing APCs in the line and MHC class II expression on corneal epithelium (arrow). Note absence of leukocytes in epithelium of autografts (A) and accepted grafts (B). Corneal endothelium at clinical rejection bears adherent dendritic cells (E, F) and adherent T-cells (G). Note elongated cells expressing MHC class II in E, which may be corneal endothelial cells. (H) MIL 11⁺ neovessels in stroma express MHC class II (cyan, arrows). (I) Positive control section of tonsil shows specificity of labeling of T-cells subset panel. (J) Positive control tonsil for APC subset panel. (K) Positive control tonsil for dendritic cell/vessel panel. (L) Negative control in which primary antibodies were omitted. dm, Descemet's membrane. Bar represents 20 μm (H) and 50 μm in all other micrographs.

Three-Color Quantitative Immunofluorescence Histology

Microscopic examination of labeled sections revealed small numbers of T-cells (Fig. 3A) and other leukocyte subtypes in the stroma of autografts, a substantial increment in numbers of such cells in accepted allografts (Fig. 3B), and large numbers of cells in rejected allografts, consisting of CD3⁺, CD4⁺, and CD8⁺ cells (Fig. 3C); CD163⁺, CD14⁺, and MHC class II⁺ macrophages/dendritic cells (Fig. 3D), and CD16⁺MHCII⁺ dendritic cells (Figs. 3E, 3F). ²⁷ Although leukocytes were quantified only in corneal stroma, both APCs and T-cells were observed in the epithelium of rejected grafts, concentrating in the rejection line (Figs. 3C, 3D), and adhering to endothelium (Figs. 3E–G). MHC class II was expressed on corneal epithelium in the vicinity of the rejection line (Fig. 3D, arrow). and appeared to be expressed on endothelium undergoing rejection (Fig. 3E). Neither leukocytes nor MHC class II expression was observed in the epithelium of autografts (Fig. 3A) or accepted grafts (Fig. 3B), nor associated with their endothelium. MHC class II was expressed on neo-vessel endothelium (Fig. 3H, arrows). Specificity of each antibody panel is illustrated by positive control sections of tonsil (Figs. 3I–K), and omission of primary antibodies yielded minimal labeling of rejected corneas (Fig. 3L). 

Initial statistical analysis of the whole data set showed significant differences in proportions of labeled cells in all three labeling combinations between graft groups (P = <0.0001 for all cell subsets) and between transplanted and contralateral eyes (P = <0.0001 for all cell subsets except CD4⁺CD8⁺ T-cells [P = 0.01], total cells expressing MIL 11 [P = 0.04], and cells expressing MIL 11 in the absence of MHC class II and CD16 [P = 0.9]). There was a significant interaction between “graft group” and “eye” (P < 0.0001 for all cell subsets, except cells expressing MIL 11 in the absence of MHC class II and CD16 [P = 0.003]), likely to result from much larger numbers of leukocytes in transplanted eyes of grafts bearing histocompatibility mismatches with the donor. This is illustrated by profiles of selected subsets of APC, T-cells and vessel endothelium in Figure 4A–D. Further analysis of transplanted eyes (including autografts) showed a significant difference between rejected and accepted grafts (P = <0.0001 for all cell subsets; Figs. 4E–1552H). The extensive network of capillaries in rejected grafts expressed MHC class II (Figs. 3K, 4D, 4H). There are no antibodies that allow unequivocal identification of NK cells or neutrophils in pig tissues, but considerable numbers of CD3⁻CD8α⁺ cells of lymphocyte morphology, likely to be NK cells, were seen (e.g., Figs. 3C, 15523G), and by hematoxylin and eosin staining we observed neutrophils, particularly in the vicinity of sutures, as might be expected (data not shown). 

Figure 4.

 

Comparison of leukocyte infiltration and vessel growth in graft groups (A–D), and rejected versus accepted grafts (E–H). (A–D) Data included corneas from the entire data set (Fig. 1). (E–H) included the four graft groups. We examined 10 images of stroma per cornea as described in the Methods. ▪ transplanted (or left) eye, □ contralateral (or right) eye.

Interestingly, plots of leukocyte numbers in each graft group (Figs. 4A–D) and post-hoc statistical analysis revealed that the stromal infiltrate of c to c (minor mismatched) grafts, none of which was rejected clinically, bore a closer similarity to d to c and b to c grafts, which included rejected grafts, than to autografts. Indeed, there were significantly more of almost all cell subsets in c to c grafts than in autografts, and c to c grafts were not significantly different from d to c grafts in more than half of all subsets. A notable exception were MHC class II⁺ MIL 11⁺ cells, where the numbers of such cells in c to c grafts did not differ from those of autografts, but were significantly fewer than in d to c and b to c grafts. This was consistent with clinical evidence that vessels penetrated the donor cornea only when grafts were rejected clinically. These results implied that there was an allo-specific leukocyte reaction in these clinically accepted minor-mismatched grafts. To determine whether there also was a significant infiltrate in accepted fully mismatched d to c and b to c grafts, a further statistical comparison was made between grafts classified into “rejection group.” This confirmed that both minor mismatched grafts and accepted, fully mismatched grafts contained significantly more cells of most subsets than autografts (again with the important exception of MIL11⁺ MHC class II⁺ vessel endothelium), but significantly fewer of all subsets than rejected grafts (Table S2). Direct comparison between minor mismatched grafts and accepted, fully mismatched grafts revealed a further dichotomy in that these groups had similar numbers of APCs, but minor mismatched grafts contained significantly more T-cells than accepted, fully mismatched grafts (Table S2). This is illustrated graphically in Figure 5, the greatest differences between the four groups being in numbers of T-cells and MHC class II⁺ vessel endothelial cells. 

Figure 5.

 

Comparison of leukocyte infiltration and vessel growth in rejection groups. The four graft groups were included (Fig. 1), and 10 images of stroma were examined per cornea as described in the Methods. Full major and minor mismatches. ▪ CD163⁻ CD14⁺ MHCII⁺, ▪ CD163⁺ CD14⁻ MHCII⁺, ▪ CD3⁺ CD4⁻ CD8⁻, □ CD3⁺ CD4⁺ CD8⁻, ▪ CD3⁺ CD4⁻ CD8⁺, ▪ MIL11⁺ MHCII⁺ CD16⁻.

TRβV Spectratyping

Most TRβV spectratypes from pre- and post-transplant blood and lymph node samples tended towards a normal, Gaussian distribution, suggesting minimal selection by antigen (Fig. S1), although the distribution of some, for example TRβV6 and TRβV11, was skewed variably (Fig. 6). By comparison, transplanted and contralateral corneal spectratypes were more skewed and more variable between animals, with the exception of a single TRβV6S peak (346 base pairs), which occurred regularly in transplanted and contralateral corneas (Fig. 6). Such skewed distributions can occur either as a result of clonal selection by antigen, or when the number of T-cells in a sample is low (as in contralateral corneas), resulting in random selection of clones. Consistent with antigen-selection, transplanted corneas had more spectratype peaks than contralateral corneas. No single spectratype peak was unique to accepted or rejected corneas, but the mean number of peaks always was higher in rejected than accepted corneas (statistically significant in 11/20 TRβV groups/subgroups, Fig. S2). 

Figure 6.

 

Heat map of TRβV spectratypes generated from corneas, parotid lymph nodes, and blood of pigs taken immediately post-mortem, and from blood taken before transplantation. tx, transplantation; contra, contralateral; ipsilat, ipsilateral; LN, lymph node, autograft; c to c (minor mismatch); d to c (full mismatch); b to c (full mismatch); transplant rejected.

Interpretation of the complex data produced by spectratyping requires highly multivariate statistical approaches. Principal component analysis attempts to reduce large data sets, in which many of the variables are correlated partially with each other, to a smaller set of uncorrelated, composite variables. Analysis of the entire data set (i.e., cells derived from blood, lymph nodes, and corneas from transplanted and contralateral eyes) identified principal component 1 (PC1_total, accounting for 11.9% of the variation within the data), which was associated significantly with “tissue” (P < 0.0001: transplanted and contralateral corneal groups were significantly different from each other, and from lymph node and blood groups) and “graft group” (P = 0.009: significant differences between autografts versus c to c grafts, autografts versus d to c grafts, b to c grafts versus c to c grafts, b to c grafts versus d to c grafts). However, when “graft outcome” or “rejection group” were included in a 2-way or 3-way fully factorial GLM, or used in a single factor GLM analysis, they were not associated significantly with PC1 _total. Specifically, this indicated that principal components derived from the full data set could not be associated with graft rejection. 

To remove the effect of the blood and lymph node samples on the analysis, PCA was performed on the subset of corneal data only. In this analysis, the first principal component (PC1_cornea, 11.3% of the variation in the data set) was associated significantly with “eye” (P < 0.0001), “graft group” (P = 0.003: autografts significantly different from all other graft groups) and “graft outcome” (P = 0.038). PC1_cornea also was associated significantly with “eye” (P < 0.0001) and “rejection group” (P = <0.0001: autografts significantly different from all other rejection groups, and full mismatch accepted group significantly different from full mismatch rejected group), and with the interaction between eye and rejection group (P = 0.008). 

Since analysis of corneal data only had identified associations between principal component factors and rejection group, analysis was restricted further to transplanted corneas only. Single fixed factor GLM using only “graft group,” “graft outcome,” or “rejection group” as fixed factors identified PC1_tx-cornea (6.9% of the variation in the data set) as being associated significantly with “graft outcome” only (P = 0.045), and PC2 _tx-cornea (6.7% of the variation in the data set) as associated with “graft outcome” (P = 0.014), “graft group” (P = 0.038: significant differences between autografts versus d to c grafts, and autografts versus b to c grafts) and “rejection group” (P = 0.023: significant difference between autografts and full mismatch rejected groups, Fig. 7). 

Figure 7.

 

Mean principal component 2 values for “rejection group” following PCA on the transplanted cornea data set. Error bars represent 95% confidence intervals. mm, mismatch. *Significant difference P < 0.05.

Proliferation of Corneal Endothelial Cells

Clinical Evidence

Four pigs were monitored after rejection to determine whether corneal clarity was restored. Opacity did not fall below a score of 2.5, and corneas remained edematous until animals were killed 28, 29, 41, and 49 days after rejection onset, respectively, duration of acute rejection in each case being approximately 12 days (Fig. 2). 

Wound Repair

To determine the capacity for wound repair in vitro, excised non-transplanted corneas were subjected to cryo-injury, followed by in vitro culture and silver staining after either 4 (n = 2) or 8 (n = 1) days. By day 4, endothelial cells around the wound periphery had increased as much as 6-fold in size. Enlarged cells extended radially for up to an estimated 0.5 mm on day 8 (Fig. 8A), while the remainder of the lesion remained devoid of cells. Corneal endothelial cells in vivo are reported to be arrested at the G₁ phase of the cell cycle so do not express the transcription factor Ki67. Therefore, corneas in organ culture were examined to determine whether cryo-injury induced early expression of Ki67 as a marker of progression to S phase and mitosis. Ki67⁺ endothelial cells were absent in freshly isolated corneas (n = 5) and on day 1 of culture (n = 2 cryo-injured, n = 1 uninjured), whether corneas were injured or not. On day 2 (n = 4 cryo-injured, n = 2 uninjured) scattered, isolated cells and occasionally small clumps (less than 50 per cornea in total) expressed Ki67, co-localizing with DAPI (Figs. 8B, 8C with associated movie). However, such cells appeared to be endothelial or epithelial cells that had become detached in culture, rather than cells involved in wound repair, as numbers of Ki67⁺ cells were similar in injured and non-injured corneas, they were not located preferentially around the wound margin, and appeared to lie above the cells in the endothelial monolayer (Figs. 8B, 8C, and linked movie). 

Figure 8.

 

Organ culture of non-transplanted corneas. (A) Silver stain showing lack of corneal endothelial cells at the site of a central cryo-injury (lower left). Distal to the wound at upper right, cells are of a relatively uniform small size (similar to the remainder of the cornea), but become increasingly enlarged and morphologically abnormal approaching the denuded area. Day 8 after injury. (B) Main image: one of a series of confocal immunofluorescence images taken at intervals from the anterior chamber surface through the thickness of the endothelial cell layer, showing Ki67⁺ cells in the cell cycle (green) co-localizing with nuclear labeling (blue). Day 2 of culture. Images below and to the right represent digitally reconstructed sections of the cornea in the z plane at the location of horizontal and vertical cross-hairs, respectively, on the main image, and show that the Ki67⁺ cells (arrows) lie above the main endothelial cell layer. (C) Digitally constructed 3D image of the entire endothelial cell layer linked to supplementary 3D movie, which confirms superficial location of Ki67⁺ cells. (D) Lower power image showing no labeling of corneal endothelium when an isotype control antibody was substituted for the primary antibody. (E) Positive control section of thymus showing KI67⁺ cells in B cell follicles co-localizing with cell nuclei. Red is counterstain for actin filaments.

Figure 8C (Movie).  

Discussion

These data, in combination with our previous report documenting preoperative preparation of pigs and anesthesia protocols, ¹⁵ show clearly that corneal transplantation and long-term follow-up are feasible in the NIH minipig. The tempo of rejection was similar in two strain combinations bearing major and minor histocompatibility mismatches. However, interestingly, clinical rejection of grafts bearing only minor mismatches was not observed, although in such grafts there was significant T-cell infiltration. When rejection did occur, the typical response targeted epithelium and endothelium, except in one pig in which rejection resembled closely a typical human endothelial rejection episode, (i.e., corneal edema and an endothelial rejection line, without dense opacity, epithelial rejection line and vessel ingress). To our knowledge, this is the first report in a model species of clinical endothelial rejection without involvement of other cell layers. Premature replacement of donor with recipient epithelial cells might account for this atypical manifestation of rejection. Indeed, absence of donor epithelium is associated with graft acceptance in mice. ²⁸ Alternatively, a minor antigen expressed only on corneal endothelium might have been the specific rejection target in this pig. 

The reasons for the matching discrepancy with rodents, in which minor mismatches result in robust rejection, ^29,30 are unclear. Breeding policy since fixation of the MHC in NIH minipigs has been to maintain genetic heterozygosity, but nevertheless this may be low within the UK lines. For this reason, rejection in the fully mismatched d to c and b to c strain combinations cannot be attributed definitively to the MHC, since there were likely to be more minor mismatches between these strains than within the SLA^cc strain. We currently are analyzing allograft donors and recipients for single nucleotide polymorphisms (SNPs) spanning the entire genome to determine the extent of inbreeding within and between strains, and whether particular SNP mismatches outside the MHC are associated with rejection. 

Not all grafts bearing major and minor mismatches were rejected, even in the b to c combination where donor and recipient parental strains were unrelated, whereas in a previously reported outbred pig model of corneal transplantation 5/5 grafts were rejected within 65 days. ¹⁰ Williams et al. also reported a higher incidence of rejection than we achieved in an outbred sheep model, but their graft size was relatively larger. ⁹ Increasing graft size (from 7 to 8 or 9 mm) or, alternatively, pre-vascularization of the recipient cornea ¹⁰ would be likely to increase the incidence of rejection in this model. 

The persistence of opacity and edema in four pigs retained for a prolonged period after rejection, and the failure of endothelial cells to repopulate a cryo-wound in organ culture or undergo proliferation around the wound margin, even with the addition of growth factors used for human cells, ³¹ is strong evidence that pig endothelium, in common with human endothelium, has a more limited capacity to proliferate than in smaller animal models. ^6,8 Indeed, pig cells proliferated less in vitro than is reported for human cultured cells. One possible explanation for this is that we did not use fetal serum in our cultures (Joyce, personal communication). 

Clearly, the number of pixels of immunofluorescence corresponding to a cell will vary with the size of the cell, ²³ and may lead to discrepancies with manual counting methods, particularly if cells of different size and morphology are being compared. However, since our comparisons are made between cells of the same subset throughout, differences in outcome between pixel-based and manual counting methods are likely to have been minimal. The automated method allowed us to quantify cell subsets in an objective fashion at a level of detail that would not be possible by manual methods, and thereby to uncover novel histological features of rejection. Importantly, both the quantitative immunohistology in Figure 4, and the expansion of repertoire observed in Figures S2 and 7 demonstrate that clinical acceptance of corneal allografts can be associated with significant infiltration of recipient leucocytes, particularly T-cells, although antigen-presenting cell numbers also are increased. We interpret this as evidence of recipient allosensitization, but supporting functional data, such as demonstration of activation markers on infiltrating T-cells or in vitro evidence of donor reactivity of T-cells extracted from grafts, would be required to prove this hypothesis. Interestingly, the increment in numbers of T-cells compared with autografts was more pronounced in grafts bearing only minor mismatches than in accepted grafts bearing MHC mismatches, suggesting a sensitizing effect of minor antigens. A possible explanation for a T-cell infiltrate in accepted grafts would be that some (or all) of these grafts were in an early stage of rejection, as yet not manifest clinically, or that they were undergoing genuine stromal rejection (i.e., rejection of stromal keratocytes) in the absence of involvement of other cell layers. A third possibility is that allo-activated T-cells were being restrained by active immunoregulatory mechanisms and/or the inherent immune regulatory environment of the anterior chamber of the eye. The involvement of Foxp3⁺ regulatory T-cells in graft acceptance has been documented in mice, ³² and availability of antibodies specific for both swine CD25, ³³ and Foxp3³⁴ permit this to be investigated in pigs in future. 

Despite the presence of cell infiltrates in accepted transplants, there were clear differences in the cell infiltrates in rejected compared with accepted corneas, both from the immunohistological and repertoire data. Rejection was associated with an increase in CD4⁺ and CD8⁺ T-cells, but not with recruitment of a repertoire of graft-specific T-cell receptors. This non-specific recruitment is consistent with the appearance of MHC class II (presumably “inflamed”) capillary endothelium observed only in rejected grafts. Despite evident non-specific recruitment, more complex statistical analysis using PCA (Fig. 7) was able to detect changes in repertoire that may be associated specifically with rejection, suggesting that there may be a subpopulation of graft-specific T-cells present in corneas in this group. However, the observation that the principal components that were associated with rejection explained only a small proportion of the variation (6.7% −11.3%) suggests that relatively few of the infiltrating T-cells were selected by corneal antigen. In contrast, no differences indicative of systemic sensitization were observed in spectratype analysis of peripheral blood or draining lymph node T-cells, and further analysis confirmed that IFNγ production by peripheral blood T-cells sampled from recipients of rejected grafts was not increased (data not shown). It is not clear whether this reflects sequestration of the response within the eye, or lack of sensitivity of assays used. 

We described previously the socialization protocol adopted to prepare miniature swine for transplantation, which minimized postoperative stress, and achieved full compliance with application of eye medication and other essential postoperative interventions. ¹⁵ We showed here that these pigs are similar to man in ocular manifestations of rejection and the limited mitotic potential of corneal endothelium. A further important implication of our findings from the human perspective is that a specific immune response, and perhaps immune-mediated damage, might be occurring in what appears to be a “quiet” eye. We conclude that this NIH miniature swine model will be valuable for basic immunological studies of corneal graft rejection and as a pre-clinical model for testing potential therapies. 

Supplementary Materials

pdf S1 

pdf S2 

pdf S3 

avi S4 

Acknowledgments

Stuart Cook provided advice on surgical technique. Tracy Dewey and John Conibear provided photography of clinical rejection. Jane Coghill, University of Bristol Transcriptomics Facility, performed capillary electrophoresis of spectratype PCR products, and Lionel Wheeler assisted with handling pigs. 

Footnotes

 Supported by the Medical Research Council and by a National Eye Research Council Fellowship (SN).

Footnotes

 Disclosure: S.M. Nicholls, None; L.K. Mitchard, None; G.M. Laycock, None; R. Harley, None; J.C. Murrell, None; A.D. Dick, None; M. Bailey, None