Identification of Therapeutic Targets of Inflammatory Monocyte Recruitment to Modulate the Allogeneic Injury to Donor Cornea

Abstract

Purpose: We sought to test the hypothesis that monocytes contribute to the immunopathogenesis of corneal allograft rejection and identify therapeutic targets to inhibit monocyte recruitment.

Methods: Monocytes and proinflammatory mediators within anterior chamber samples during corneal graft rejection were quantified by flow cytometry and multiplex protein assays. Lipopolysaccharide or IFN-γ stimulation of monocyte-derived macrophages (MDMs) was used to generate inflammatory conditioned media (CoM). Corneal endothelial viability was tested by nuclear counting, connexin 43, and propidium iodide staining. Chemokine and chemokine receptor expression in monocytes and MDMs was assessed in microarray transcriptomic data. The role of chemokine pathways in monocyte migration across microvascular endothelium was tested in vitro by chemokine depletion or chemokine receptor inhibitors.

Results: Inflammatory monocytes were significantly enriched in anterior chamber samples within 1 week of the onset of symptoms of corneal graft rejection. The MDM inflammatory CoM was cytopathic to transformed human corneal endothelia. This effect was also evident in endothelium of excised human cornea and increased in the presence of monocytes. Gene expression microarrays identified monocyte chemokine receptors and cognate chemokines in MDM inflammatory responses, which were also enriched in anterior chamber samples. Depletion of selected chemokines in MDM inflammatory CoM had no effect on monocyte transmigration across an endothelial blood–eye barrier, but selective chemokine receptor inhibition reduced monocyte recruitment significantly.

Conclusions: We propose a role for inflammatory monocytes in endothelial cytotoxicity in corneal graft rejection. Therefore, targeting monocyte recruitment offers a putative novel strategy to reduce donor endothelial cell injury in survival of human corneal allografts.

Corneal transplantation remains the most commonly performed transplantation worldwide,¹ and 25% of all corneal allografts fail within 5 years, primarily as a result of immune-mediated rejection.² The cellular requirements and the sequence of events in the effector component of the allogeneic response leading to endothelial corneal graft rejection are not fully understood, and much of our information has been obtained from animal models. It is widely believed that CD4⁺ T cells play a critical role in the rejection of rodent orthotopic corneal allografts,^3–5 yet rejection can still occur in CD4 and IFN-γ knockout mice.^3,6 Furthermore, multiple and redundant effector mechanisms have been implicated in graft rejection⁷ and may explain the poor outcomes in respect to rejection in corneal transplant recipients treated only with calcineurin antagonists, which block T-cell clonal expansion.⁸ Several lines of evidence in rodent corneal transplantation suggest monocyte and macrophage involvement in the cell-mediated allogeneic response to transplanted cornea. First, large numbers of macrophages are found in tissue sections at the onset of corneal rejection in mice and rats. At the earliest time points following the onset of corneal rejection in the rat, graft-infiltrating macrophages exceed T cells and NK cells,⁹ and in mice, macrophages were reported among the earliest graft-infiltrating cells, before and after the onset of corneal rejection.¹⁰ Second, local depletion of macrophages by subconjunctival administration with clodronate liposomes of corneal allotransplant recipients significantly prolonged corneal graft survival in treated rats using two different strain combinations.^11,12 Local depletion of macrophages was found to down-regulate infiltration of all alloreactive cell types, down-regulate local and systemic cytotoxic lymphocyte responses, and prevent the generation of antibodies.¹³ Third, earlier pilot investigation of immune cell populations in aqueous humor samples from the eye in patients at presentation with acute transplant rejection indicated a high proportion and selective recruitment of CD14⁺ cells to the anterior chamber of the eye, likely representing mononuclear phagocytic cells,¹⁴ corroborating an earlier report.¹⁵ We sought to reconfirm and extend these data, investigate the mechanisms by which these cells may contribute to the mechanism of human corneal graft failure, and investigate possible strategies to inhibit their recruitment across the blood–eye barrier, as a potential novel therapeutic intervention. 

Materials and Methods

Ethics

Ethics approval was provided by the designated UK National Research Ethics Service committee for anterior chamber sampling from patients with acute corneal graft rejection and use of human corneal specimens in research (Research Ethics Committee reference: 11/LO/1294). Corneas, with healthy endothelium, were excised at surgery in keratoconus patients. Written informed consent was obtained from all participants. This study adhered to the tenets of the Declaration of Helsinki. 

Aqueous Humor Analysis

Aqueous humor samples (100–200 μL), were obtained from 10 patients presenting with corneal allograft rejection (Tables 1, 2). Diagnosis was confirmed by the finding in all patients of active anterior chamber inflammation and keratic precipitates on the donor corneal endothelium at slit-lamp biomicroscope examination. Control aqueous samples were obtained from nine patients undergoing routine cataract surgery (five male and four female; median age, 57 years; range, 3–85 years) without any other ocular disease. Samples were centrifuged at 400g for 5 minutes. The soluble fraction was collected, and the cell pellet was resuspended in 100 μL PBS with 0.5% bovine serum albumin (Sigma, Poole, UK) and 0.01% sodium azide (Sigma). Total cell counts were enumerated with a hemocytometer, and the cells were stained with directly conjugated fluorescent antibodies to CD14 (clone M5E2; Becton Dickinson) and CD16 (clone 3G8; Becton Dickinson, Oxford, UK). Immunostaining was quantified with a FACScalibur flow cytometer (Becton Dickinson) and FlowJo analysis software (version 9.4.3; FlowJo, Ashland, OR, USA). Chemokine concentrations in these samples were measured using a flow cytometric multiplex bead assay kit (HCYTOMAG-60K; Milliplex, Watford, UK). 

Table 1

 

Selection Criteria for Patients With Corneal Rejection

 

Table 2

 

Characteristics of Corneal Rejection Patients

 

Preparation of Monocyte-Derived Macrophage–Conditioned Media

Monocyte-derived macrophages (MDMs) were prepared from human peripheral blood mononuclear cells (PBMCs) as previously described.¹⁶ After being allowed to differentiate for 6 days, the medium was changed to RPMI medium (Sigma) and 10% foetal calf serum (FCS; Biosera, Nuaille, France) with or without 100 ng/mL lipopolysaccharide (LPS) for 6 hours or 10 ng/mL IFN-γ for 24 hours, using 1 mL media/10⁶ MDMs. Conditioned media (CoM) from these cultures were then centrifuged at 10,000g for 5 minutes and stored at −80°C. Residual LPS in LPS-stimulated CoM was neutralized by addition of 10 μg/mL polymyxin B as described previously,¹⁶ and residual IFN-γ was neutralized in IFN-γ–stimulated CoM by addition of 2 μg/mL blocking antibody to IFN-γ. Effective functional neutralization of LPS and IFN-γ under these conditions was confirmed in monocyte transwell migration assays as described below, but without endothelial cells. We confirmed previous reports that LPS inhibits monocyte migration in transwell assays and found that IFN-γ had the same effect, which was reversed by addition of polymyxin B to neutralize LPS or antibody to neutralize IFN-γ (Supplementary Fig. S1). The CoM for each condition were pooled from four separate MDM donors in order to minimize the effects of donor–donor variability. 

Human Corneal Endothelial Cell Toxicity

The immortalized human corneal endothelial cell line (HCEC-B4G12, Ref- ACC 647; DCMZ, Germany)¹⁷ was cultured in human endothelial serum-free medium (Gibco, Paisley, UK) and 10 ng/mL FGF-2 (Gibco) in tissue culture plates precoated with 10 μg/mL laminin (Sigma) and 10 mg/mL chondroitin sulphate (Sigma). At 70% confluence, the media were then changed to 10% CoM from LPS- or IFN-γ–stimulated and unstimulated MDMs before being washed with PBS and stained with calcein-AM (Sigma) and propidium iodide (PI; Sigma) as per the manufacturer's instructions. Cellular fluorescence was imaged in situ on a Leica SPE inverted confocal microscope (Leica Microsystems, Milton Keynes, UK). Following removal of epithelium, freshly excised full-thickness human cornea specimens were cut into quadrants at surgery with a diamond blade, placed in Dulbecco's Modified Eagle's (DME) medium (Gibco) with 10% FCS overnight, and then incubated for 24 hours with CoM from stimulated and unstimulated MDMs. The corneas were then washed in PBS, fixed in 4% paraformaldehyde overnight, and incubated in a blocking solution (PBS, 0.1 M Lysine, and 0.05% Triton X-100) for 1 hour, before immunostaining with a rabbit antibody for Connexin (Cx)43 (Sigma), fluorophore-conjugated secondary antibody, and a bis-benzamide nuclear counterstain. The stained cornea was washed in PBS and mounted in citifluor solution (Citifluor, London, UK) and imaged using a Leica SPE confocal microscope. Sample identifiers were blinded for image analysis. The Cx43 staining was quantified as previously described,¹⁸ and nuclear counting within multiple high power fields was performed manually. 

Transwell Migration Assay Across an Endothelial Barrier

Human cerebral microvascular endothelial cells (hCMEC/D3)¹⁹ cells were obtained as kind gift from PO Couraud, PhD (Institut Cochin, Paris, France). These were cultured in endothelial growth medium-2 (EGM-2; Lonza, Verviers, Belgium) supplemented with 5% FCS, 1% penicillin-streptomycin (Gibco), 1.4 μM hydrocortisone (Sigma), 5 μg/mL ascorbic acid (Sigma), 1% chemically defined lipid concentrate (Gibco), 10 mM HEPES (Gibco), and 1 ng/mL human fibroblast growth factor (Sigma). Cells were seeded at 5 × 10⁴ cells/cm² on the apical side of 0.33-cm² polycarbonate transwell inserts (Corning No. 3421, 6.5-mm diameter, 5.0-μm pores; Corning, Flintshire, UK) precoated with Cultrex rat type 1 collagen 50 μg/mL (R&D Systems, Abingdon, UK). To form a monolayer, hCMEC/D3 cells were maintained in culture for 6 days with media changes after 3 and 6 days¹⁹ and supplemented with 10 mM lithium chloride (Santa Cruz Biotechnology, Heidelberg, Germany) for the entire culture period to generate tight junctions^20,21 confirmed by transendothelial electrical resistance (TEER) measurements using an electrical Volt-Ohm-Meter (EVOM-2; World Precision Instruments, Sarasota, FL, USA). After 6 days, the transwells were then transferred into wells containing 10% stimulated or unstimulated CoM for 24 hours. Cellular transmigration across the endothelial barrier was assessed by addition of 5 × 10⁵ PBMCs obtained from healthy volunteers into the upper chamber of the transwell for 3 hours at 37°C, before collecting cells in the lower chamber. This cell suspension was then stained for CD14 and CD16 and enumerated by flow cytometry using Flow-Check polystyrol fluorospheres (Beckman Coulter, High Wycombe, UK) to standardize cell counting. Monocytes and lymphocytes were discriminated by light scatter properties and CD14/CD16 staining. Transmigration of cells into the lower chamber was expressed as a proportion of the input. 

Chemokine and Chemokine Receptor Expression Data

Data on normalized chemokine receptor expression in monocytes and chemokine expression in LPS- or IFN-γ–stimulated MDMs were obtained from the European Bioinformatics Institute data repository (www.ebi.ac.uk/arrayexpress/) using accession numbers E-TABM-1206 and E-MEXP-2032. A network of interacting chemokines and chemokine receptors was adopted from the Kyoto Encyclopedia of Genes and Genomes (KEGG) cytokine–cytokine receptor interaction reference pathway (www.genome.jp/kegg/kegg2.html, map0460) and constructed in Gephi graph visualization software (version 0.8.2; http://gephi.github.io/). The transcriptomes of hCMEC/D3 cells after 24-hour incubation with CoM from unstimulated MDMs and LPS-stimulated MDMs with Polymixin B (PMB) were also compared by genome-wide expression arrays. Total RNA was purified from cell lysates collected in RLT buffer (Qiagen) using the RNeasy Mini kit (Qiagen). Samples were processed for Agilent microarrays as previously described,²² and local regression (LOESS) normalized data were analyzed using theTM4 microarray software suite MeV (version 4.9). Pathway enrichment analysis of differentially expressed gene lists was performed using the online bioinformatics tools InnateDB²³ and Ingenuity Pathway Analysis (http://www.ingenuity.com/). Microarray data are available from the EBI Array Express repository (http://www.ebi.ac.uk/arrayexpress/) under accession no. E-MTAB-3692. 

Chemokine Depletion and Chemokine Receptor Targeting

To deplete a single or a combination of different monocyte chemotactic chemokines, biotinylated anti-human antibodies against CCL2, CCL3, CCL4, and CCL8 were added at a concentration of 5 μg/mL to the different CoMs. The biotinylated antibodies were incubated with the CoM for 1 hour at room temperature before addition of magnetic streptavidin beads (Dynabeads MyOne Streptavidin T1; Life Technologies, Paisley, UK) using 10⁹ beads/mL. These were incubated for 1 hour with the CoM before removing the beads magnetically. Successful depletion was confirmed by ELISA using paired capture and detection antibodies for each chemokine (eBioscience, Hatfield, UK) according to the manufacturer's instructions and analyzed using a Multiskan absorbance plate reader (Thermo Labsystems, Waltham, MA, USA). For chemokine receptor targeting, 5 × 10⁶ cells/mL PBMCs were incubated with inhibitors of CCR2 (RS 504393 or BMS CCR2 22; both from Tocris, Abingdon, UK), CCR5 (Maraviroc; Tocris), or CXCR4 (AMD 3465 hexahydrobromide; Tocris), used at 10 nM, for 30 minutes. 

Statistical Analysis

Data were analyzed using Graphpad Prism software version 5 (GraphPad Software, Inc., La Jolla, CA, USA). The Mann-Whitney U or t-tests were used to test significance. Values of P < 0.05 were defined as statistically significant. 

Results

Enrichment of Inflammatory Monocytes in the Aqueous Humor of Patients With Acute Corneal Graft Rejection

Samples of aqueous humor from the anterior chamber were obtained from 10 patients with endothelial corneal allograft rejection. The demographic characteristics, primary diagnosis leading to corneal transplantation, number of previous transplants and previous episodes of allograft rejection, and corticosteroid therapy at time of rejection are summarized in Table 2. Total cell counts in the aqueous humor samples revealed that a cellular infiltrate was only detectable in aqueous humor samples from patients presenting within 7 days of the onset of symptoms (Fig. 1A), irrespective of concomitant corticosteroid treatment (Supplementary Fig. S2). Within these samples, we confirmed our previous observation¹⁴ that CD14⁺ cells were significantly enriched in the aqueous humor compared with the peripheral blood (Fig. 1B). We extended these data to show that the enrichment of CD14⁺ cells was entirely due to CD14^hiCD16^low classical inflammatory monocytes, known to be recruited to inflammatory foci (Fig. 1C). 

Figure 1

 

(A) Total cell counts in aqueous humor (AH) samples were compared in patients with more or less than 7 days of symptoms of corneal allograft rejection. In patients for whom a cellular infiltrate was evident (n = 5), the proportion of all CD14⁺ cells (B) and monocyte subsets discriminated by the combination of CD14 and CD16 staining (C) was compared in AH and contemporaneous PBMC samples. The inset dot plot shows the gating strategy used to quantify each of monocyte subsets indicated. *Statistically significant differences (P < 0.05, Mann-Whitney U test). Measurements for individual patient samples are shown in (A) and summarized as median ± interquartile range in (B) and (C).

Monocyte-Derived Macrophages Generate Inflammatory Mediators That Deplete Corneal Endothelial Cells

Classical inflammatory monocytes are extremely short lived, with a half-life of less than 24 hours,²⁴ unless they differentiate into macrophages as a result of environmental signals. Therefore, we reasoned that if monocyte recruitment to the eye contributes to the pathogenesis of corneal transplant rejection, inflammatory mediators generated by MDMs may cause donor corneal endothelial cytotoxicity and depletion. Macrophages produce inflammatory mediators in response to innate immune danger signals or IFN-γ production by lymphocytes. We therefore modeled macrophage inflammatory responses by stimulating MDMs with either LPS or recombinant IFN-γ, and we pooled CoM from stimulated and unstimulated MDM cultures from multiple experiments in order to minimize the confounding of donor to donor variability. In order to focus on the role of the MDM-derived inflammatory response in downstream experiments with these CoM, we neutralized LPS or IFN-γ activity by addition of polymyxin B or blocking antibody to IFN-γ to the relevant samples. We first assessed the effect of MDM CoM on survival of an immortalized human endothelial cell line and found that CoM from LPS-stimulated MDMs induced significant endothelial cell death, indicated by PI staining (Fig. 2A). In order to extend these observations further, we then evaluated the effect of MDM CoM on endothelial cells in excised corneal specimens ex vivo by counting the number of nuclei in the endothelial layer and by quantifying expression of the gap junction protein Cx43 as a surrogate for the integrity of the endothelial layer. In keeping with the effect on the corneal endothelial cell line, we found that CoM from LPS-stimulated MDMs induced significant loss of cells of the corneal endothelium and reduction of detectable Cx43 immunostaining (Figs. 2B–D). In these experiments, we also found depletion of endothelial cells and loss of Cx43 staining as a result of incubation with CoM from IF-Nγ–stimulated MDMs, albeit to a lesser degree than LPS-stimulated CoM (Figs. 2B–D). In addition, we found that the presence of monocytes in this experimental model significantly increased the cytopathic effect of LPS-stimulated CoM as measured by a reduction in Cx43 staining (Fig. 3). 

Figure 2

 

(A) Propidium iodide and calcein staining in a human corneal endothelial cell line incubated with CoM from unstimulated (control) and LPS- or IFN-γ–stimulated MDM was visualized by immunofluoresence microscopy at 30 and 690 minutes. (B) Cx43 and nuclear staining in the endothelial layer of human cornea specimens incubated overnight with CoM from unstimulated and LPS- or IFN-γ–stimulated MDMs was visualized by confocal microscopy. Quantitation of numbers of visible nuclei and positive Cx43 staining per high power field is summarized from four separate experiments (C, D). *Statistically significant differences (P < 0.05, Mann-Whitney U test). Bars represent median ± interquartile ratio. Fluorescence images are representative of four separate experiments in each case.

Figure 3

 

(A) The Cx43 and nuclear staining (DAPI) in the endothelial layer of human cornea specimens incubated ex vivo for 24 hours with CoM from unstimulated (Control) and LPS- or IFN-γ–stimulated MDMs followed by an overnight incubation with CD14 selected human monocytes (Mo) was visualized by confocal microscopy. Quantitation of numbers of visible nuclei and positive Cx43 staining per high power field is summarized from four separate experiments (B, C). *Statistically significant differences (P < 0.05, Mann-Whitney U test). Bars represent mean ± SEM. Fluorescence images are representative of four separate experiments.

Identification of Putative Targets to Modulate Monocyte Recruitment

Our data implicate recruitment of functionally active monocytes in the pathogenesis of corneal allograft rejection. Therapeutic targeting of monocyte recruitment may therefore provide a novel strategy to limit injury to the donor corneal endothelium. Given the importance of chemokine pathways to mediate cell-specific recruitment, we sought to identify the principal chemokine or chemokine receptors that control human monocyte recruitment to inflammatory foci. To do this, we cross-referenced previously published transcriptomic data for chemokine receptor expression in human monocytes²⁵ (Fig. 4A), chemokine expression by MDMs stimulated with LPS or IFN-γ²⁶ (Fig. 4B) with established networks of chemokine and chemokine receptor interactions (Fig. 4C). This analysis identified eight chemokines (CCL2, CCL3, CCL4, CCL5, CCL7, CCL8, CCL13, and CCL18) and the most highly expressed chemokine receptors (CCR1, CCR2, and CCR5) that may participate in amplification of monocyte recruitment to putative inflammatory foci. Importantly, using a multiplex protein assay that included reagents for CCL2, CCL3, and CCL4, we found that each of these chemokines was also detectable at significantly greater levels in anterior chamber samples from patients with acute corneal graft rejection within 7 days of symptom onset compared with samples from patients with greater than 7 days of symptoms or samples from control patients (Fig. 5). These data highlight the most likely monocyte chemokine pathways involved in monocyte recruitment in acute corneal graft rejection. 

Figure 4

 

(A) Relative mRNA expression of chemokine receptors by peripheral blood monocytes from separate healthy volunteer donors and (B) mean fold change of chemokine transcript levels in LPS- or IFN-γ–stimulated MDMs are derived from previously published data sets. (C) A network diagram of known chemokine receptor–ligand interactions for all the chemokine receptors expressed in (A), above the median normalized expression level of 8 (Log₂), in which the size of the nodes is proportional to the number of interactions. Interactions between chemokine receptors (yellow nodes) in this network and chemokines (red nodes) that are up-regulated in LPS- or IFN-γ–stimulated MDMs are highlighted.

Figure 5

 

Comparison of selected chemokine concentrations (CCL2, CCL3, and CCL4) in aqueous humor samples from patients with corneal allograft rejection grouped by duration of symptoms (more or less than 1-week duration) and samples from control patients undergoing cataract surgery. Measurements from individual patient samples and the median of each group are indicated. *Statistically significant differences (P < 0.05, Mann-Whitney U test).

Macrophage Inflammatory Responses Drive Monocyte Recruitment Across a Blood–Eye Barrier

Next, we developed an experimental model for monocyte transmigration across the blood–eye barrier in order to test the hypothesis that the chemokine pathways described above were necessary for monocyte recruitment and could therefore be targeted to effectively reduce monocyte recruitment in corneal allograft rejection. We used a well-characterized human endothelial cell line derived from brain microvasculature to establish an endothelial barrier with tight junctions in transwells (Figs. 6A, 6B). Inflammatory CoM from LPS- or IFN-γ–stimulated MDMs added to the bottom compartment of the transwell apparatus induced significantly greater monocyte transmigration from top to bottom compartment (Figs. 6A, 6C). We used unfractionated PBMCs in order to compare monocyte and lymphocyte recruitment. Consistent with our in vivo finding that monocytes are enriched in the anterior chamber of patients with acute corneal graft rejection, we found significantly greater transmigration of monocytes compared with lymphocytes in this in vitro model (Fig. 6C). 

Figure 6

 

(A) Schematic representation of the transwell model to evaluate PBMC transmigration across a blood–eye barrier. The hCMEC/D3 cell line is cultured in transwell and allowed to form tight junctions, followed by incubation in CoM from stimulated and unstimulated MDMs in the lower chamber and addition of unfractionated PBMCs in the top chamber. (B) Measurement of TEER across hCMEC/D3 cells in culture with and without lithium supplementation to encourage tight junction formation. (C) Quantitation of CD14⁺ (monocytes) and CD14⁻ (lymphocytes) fractions of PBMCs migrating through the hCMEC/D3 barrier as the proportion of the total PBMCs loaded into the top chamber, in response to each of the MDM CoMs indicated in the lower chamber. In (B), bars represent mean ± SEM of at least 10 separate measurements. In (C), bars represent mean ± SEM of six separate experiments. *Statistically significant differences (P < 0.05, t-test).

Chemokine Receptor Targeting to Attenuate Monocyte Transmigration Across a Blood–Eye Barrier

In order to reduce monocyte recruitment in corneal graft rejection to a functionally significant degree, we reasoned that it might be possible to target either the chemokines or chemokine receptors. To test the effect of chemokine targeting we depleted LPS-stimulated MDM CoM of selected chemokines identified in the experiments above individually or all of these chemokines together. We found no attenuation of monocyte transmigration in the endothelial blood–eye barrier model (Fig. 7A). In response to proinflammatory stimuli, endothelial cells up-regulate cell adhesion molecules and chemokines that contribute to leukocyte recruitment.^27–29 Accordingly, genome-wide transcriptional responses in hCMEC/D3 cells to CoM from LPS-stimulated MDMs revealed up-regulation of canonical cell adhesion molecules involved in leukocyte adhesion and diapedesis and significant enrichment of secreted products with chemotactic activity (Supplementary Fig. S3) including CCL2, CCL5, CCL7, and CCL8 (Fig. 7B), thereby supporting monocyte recruitment despite depletion of these chemokines in the CoM from LPS-stimulated MDMs. Therefore, we considered targeting their cognate chemokine receptors instead. A number of small molecule inhibitors of the chemokine receptors CCR2 and CCR5 have already been evaluated in clinical trials. We therefore tested the effect of small molecule inhibitors of these chemokine receptors on monocyte transmigration. We found that targeting CCR2 with inhibitory molecules significantly attenuated monocyte recruitment (Fig. 7C). A CCR5 inhibitor also showed the same effect but did not reach statistical significance in four experimental replicates. In contrast, a small molecule targeting CXCR4 had no effect on monocyte recruitment (Fig. 7C). CXR4 is expressed by monocytes (Fig. 4A), but its ligand, CXCL12 (Fig. 3C), was not up-regulated in LPS- or IFN-γ–stimulated MDMs (Fig. 4B). Of note, CCR2 expression and function are known to be down-regulated as monocytes are differentiated to macrophages,³⁰ although some subsets of macrophages may retain higher CCR2 expression.³¹ This is also reflected in our analysis of the transcriptomes of monocytes and MDMs (Supplementary Fig. S4), even after MDM stimulation with LPS or IFN-γ, and is consistent with the hypothesis that CCR2 has a specific role in tissue recruitment of circulating monocytes. 

Figure 7

 

(A) Monocyte transmigration across hCMEC/D3 cells incubated with CoM from LPS-stimulated MDMs was compared with and without depletion of selected chemokines from the CoM reflected in the final concentration of each of the chemokines indicated in the heat map panel. (B) Gene expression heat map of chemokines up-regulated in hCMEC/D3 cells incubated with CoM from LPS-stimulated MDMs compared with CoM from unstimulated MDMs in four independent experiments. (C) Monocyte transmigration across hCMEC/D3 cells incubated with CoM from LPS-stimulated MDMs was compared in the presence and absence of small molecule inhibitors of the chemokine receptors indicated. Data bars represent mean ± SEM of four separate experiments in each case. *Statistically significant differences (P < 0.05, t-test).

Discussion

In the present manuscript, we show three lines of evidence to support a role for monocytes in corneal rejection. First, classical inflammatory monocytes are specifically enriched in anterior chamber specimens from patients with acute corneal rejection. Second, the proinflammatory mediators of monocyte-derived cells are sufficient to induce cell death of a human corneal endothelial cell line in vitro and death of primary human corneal endothelial cells in corneal buttons ex vivo. Third, the addition of monocytes to proinflammatory mediators significantly enhances corneal endothelial cell death. CD14^hi monocytes comprise approximately 10% of PBMCs in health. Therefore, the finding that 40% to 50% of cells within aqueous humor samples of patients at the time of acute corneal allograft rejection are monocytes clearly indicates selective recruitment, further supported by the finding of elevated levels of chemokines CCL2, CCL3, and CCL4, known to be chemoattractant for monocytes. These were almost entirely CD16^low cells, which is the predominant monocyte subset to be recruited to inflammatory foci and suggests that these cells are likely to be functionally active in the allogeneic tissue injury response. It is of note that the cellular infiltrate was only evident in patients with symptoms for less than 7 days, indicating that these specimens were examined shortly following the onset of the effector phase of the allogeneic response. Of note, this observation was not confounded by presence or absence of corticosteroid treatment, although our sample size was too small for statistically robust subgroup analysis. Additional comparisons in immune correlates of alternative corneal pathologies would be necessary to test the specificity of our findings for acute corneal allograft rejection. 

Monocytes typically survive less than 24 hours or differentiate into tissue-resident macrophages and dendritic cells.²⁴ We speculate that monocyte recruitment to the anterior chamber represents one of the earliest events in the effector phase of corneal allograft rejection and is associated with monocyte differentiation. Hence, the absence of cells later than 7 days following the onset of transplant rejection symptoms may reflect the transition to macrophages that migrate from the anterior chamber or adhere to the transplant surface, as suggested by data from rodent models.⁹ Our study does not address the question of the initial inflammatory trigger that stimulates monocyte recruitment in the first instance. The intersection of coagulation pathways with inflammation and inflammatory responses that arise from so-called danger associated molecular patterns suggest tissue injury may lead to innate immune responses that augment adaptive immune responses to allogeneic antigens.³² In corneal transplantation, the mechanical trauma of surgery may cause significant tissue damage, leading to activation of resident antigen-presenting cells and enhanced immunogenicity, in keeping with the danger model proposed by Matzinger.³³ 

Macrophage infiltration has also long been recognized as a hallmark of acute allograft rejection after heart transplantation,³⁴ and a number of studies suggest that macrophages can promote acute renal allograft rejection.^35–37 The functional consequence of modulating macrophage function in corneal transplantation was most directly shown by prolongation of rat corneal allograft survival following depletion of conjunctival macrophages with clodronate liposomes.¹³ One component of the present study was to assess whether inflammatory responses from MDMs may contribute to loss of donor corneal function in rejection by causing endothelial cell death and thereby compromising the transparency of donor cornea. In two separate models, we found that CoM from LPS-stimulated MDMs, and to a lesser extent, IFN-γ–stimulated MDMs, caused significant corneal endothelial cell loss. Furthermore, monocytes had an additional direct cytotoxic effect on human corneal endothelium ex vivo but only in the context of inflammation. This suggests therefore that the inflammatory cellular microenvironment either drives further proinflammatory cytokine release and subsequent cytotoxicity; in keeping with published evidence showing inflammatory cytokines to promote endothelial cell apoptosis,^38–41 or that the monocytes themselves become activated and release tissue-destructive lysosomal enzymes or free radicals that are directly cytotoxic.^42–44 Further characterization of the fate and phenotype of the MDMs that accumulate during corneal rejection is necessary to test these hypotheses. Due to the limited availability and volume of human aqueous humor sampling, this was not possible in the present study, but requires renewed assessments of tissue specimens from rejected corneal allografts or further experimental studies in animal models. In addition, the mechanism of corneal endothelial cell death is not addressed in our current experiments, but specific cell death pathways, including apoptosis, pyroptosis, and necroptosis, all intersect with inflammatory responses and may therefore be implicated. Elucidating which of these pathways makes the greatest contribution will be important in future studies in order to identify targets to inhibit corneal endothelial cell death. 

Our data suggest a rationale for therapeutic targeting of monocyte recruitment, and we investigated targeting of the monocyte chemotaxis. Cross-reacting specificities of chemokines and chemokine receptors generate a network with potential for significant functional redundancy that may undermine the use of specific inhibitors. We cross-referenced chemokine receptor expression in monocytes and chemokine production in stimulated MDMs with existing databases for chemokine pathways in order to identify those that may contribute to inflammatory monocyte recruitment in our model. Our experimental model simulated changes to chemokine levels associated with acute human corneal graft rejection. We therefore adopted an experimental system in which we could test the role of selected chemokine or chemokine receptors for monocyte migration across endothelial cells that exhibit the tight junction features of the blood–aqueous or blood–brain barrier. Given the functional redundancy within the chemokine network, we were not surprised to find that depletion of individual chemokines in this system had no significant effect on monocyte migration. It was more surprising that depletion of several chemokines that we predicted may play a role also had no effect on chemokine transmigration. However, we found that the endothelial cells modeling the blood–brain barrier also produced these chemokines and can therefore drive monocyte recruitment in response to paracrine activation by macrophages. Therefore, we tested the effect of targeting key monocyte chemokine receptors instead, using CCR2 and CCR5 small molecule inhibitors, which have shown some efficacy in preclinical and clinical models of rheumatoid disease.^45,46 Blockade of CCR2, which interacts with CCL2, CCL7, CCL8, and CCL13, did result in partial inhibition of monocyte transmigration, suggesting that this receptor at least plays a nonredundant role in recruitment of inflammatory monocytes. Additional sampling and analysis of the humoral and cellular components enriched in acute corneal rejection is needed to overcome the limitations of the small aqueous humor sample size in the present study. Nonetheless, our findings add to the evidence for the role of monocyte recruitment early on in the effector phase of corneal transplant rejection and highlights the potential for chemokine receptor blockade to reduce donor endothelial cell injury in rejection. In vivo studies will be required to examine the efficacy of this novel immunomodulatory approach in attenuating allogenic injury to donor endothelium and prolonging human corneal transplant survival. Of note, experimental corneal rejection was ameliorated in mice with targeted deletions of CCR1, but not in mice deleted for CCR2 and CCL3.⁴⁷ Although total mononuclear cell recruitment to the site of rejection was attenuated in CCR1-deficient mice, monocyte recruitment specifically was not evaluated, and cellular recruitment in CCR2/CCL3 deficient mice was not reported. Therefore, genetic and pharmaceutical targeting of chemokine receptors in mouse models merits further evaluation in order to pave the way for human clinical studies. 

Acknowledgments

Supported by a Gertrud-Kusen-Foundation fellowship (TL), the Rosetrees Trust, the Academy of Medical Sciences (SZ), National Institute for Health Research (NIHR) Biomedical Research Centre funding to Moorfields Eye Hospital and UCL Institute of Ophthalmology, and NIHR Biomedical Research Centre funding to University College London Hospitals and UCL. 

Disclosure: T. Lapp, None; S.S. Zaher, None; C.T. Haas, None; D.L. Becker, None; C. Thrasivoulou, None; B.M. Chain, None; D.F.P. Larkin, None; M. Noursadeghi, None