NK cells are necessary for recovery of corneal CD11c⁺ dendritic cells after epithelial abrasion injury

The recovery of dendritic cells in healing corneal epithelial wounds depends on NK cells, IFNγ, and ICAM-1.

Abstract

Mechanisms controlling CD11c⁺ MHCII⁺ DCs during corneal epithelial wound healing were investigated in a murine model of corneal abrasion. Selective depletion of NKp46⁺ CD3− NK cells that normally migrate into the cornea after epithelial abrasion resulted in >85% reduction of the epithelial CD11c⁺ MHCII⁺ DCs, normally present during and after epithelial wound closure. Transfer (i.v.) of spleen NK cells into NK cell-depleted mice significantly restored levels of corneal epithelial DCs (P<0.01). Immigrated NK cells were predominately positive for IFN-γ, and topical corneal anti-IFN-γ reduced epithelial DCs by 79% (P<0.01). IFN-γ^−/− mice had 69% fewer DCs than WT controls (P<0.01), and topical rIFN-γ applied to NK cell-depleted corneas increased epithelial DCs significantly (P<0.01). The contribution of ICAM-1, an adhesion molecule involved in leukocyte migration, expressed on healing corneal epithelium, was evaluated. ICAM-1^−/− mice exhibited >70% reduction in epithelial DC recovery in the first 48 h after epithelial abrasion (P<0.01). These interventions reveal an early turnover of DCs in the epithelium after injury, and ICAM-1, NK cells, and IFN-γ are necessary for the immigration phase of this turnover.

Introduction

Healing of the epithelium and subbasal nerve plexus after corneal abrasion depends, in part, on an inflammatory response that is sufficient to promote tissue repair and prevent infection. In animal models, corneal epithelial wounds induce a rapid inflammatory response in the limbal vasculature with vasodilatation, neutrophil [1–4], and platelet [5, 6] extravasation into the stroma and migration of γδ T cells [7] and DCs [8–10] into the healing epithelium. This innate inflammatory response contributes directly to healing, as evidenced by cell-selective depletion experiments that result in delayed wound closure and diminished epithelial and nerve recovery [4, 5, 7, 8, 11–13].

In the present study, we analyze CD11c⁺ MHCII⁺ DCs in the cornea after epithelial injury. Although typically viewed as APCs [14], the DCs in the epithelium appear to be involved directly in corneal wound healing [8]. Selective depletion of the CD11c⁺ cells in the cornea delays wound closure, reduces the survival of epithelial cells, and reduces epithelial expression of inflammatory genes, such as IL-1β, CXCL10, and TSLP [8, 15]. We found recently that NKp46⁺ NKG2D⁺ NK1.1⁺ RORγt⁺ IL-22− CD3− NK cells accumulate in the corneal stroma in response to epithelial injury. Given the well-known crosstalk between NK cells and DCs [16–20], we sought to determine if the immigrating NK cells influence the recovery of CD11c⁺ DCs in the corneal epithelium after wounding.

NK cells are known to modulate inflammatory responses [21–27]. Haworth et al. [22] demonstrated that resolution of an adaptive immune inflammatory response in the lung is substantially dependent on NK cells. Others have also found similar NK cell-dependent modulation of adaptive immune inflammatory responses in the CNS [23–26] and in inflammatory arthritis [27]. With regard to the innate inflammatory response to tissue injury, we reported that depletion of NK cells increases neutrophil influx significantly into the abraded cornea, prolongs residence of the inflammatory cells in the stroma, and delays wound closure and nerve recovery [21]. In the innate and adaptive experimental settings, adoptive transfer of isolated NK cells into NK cell-depleted mice restores normal kinetics of inflammatory resolution, an effect prominently dependent on NKG2D receptors on the transferred NK cells [21, 22].

In the murine model of corneal injury used in this study, epithelial wound closure is rapid, usually complete within 24 h [5, 28], and is associated with epithelial expression of the adhesion molecule, ICAM-1 (CD54) [28], the influx of NK cells into the stroma [21], and migration of CD11c⁺ DCs in the epithelium [8]. Here, we provide evidence that recovery of the DCs in the epithelium after epithelial abrasion is dependent on NK cells, IFN-γ, and ICAM-1.

MATERIALS AND METHODS

Animals

Female TCRδ^−/− and IFN-γ^−/− mice on the C57BL/6 background and C57BL/6 WT mice were purchased from The Jackson Laboratory (Bar Harbor, ME, USA). Male C57BL/6 WT mice from The Jackson Laboratory were used for one set of experiments. All animals used in this study were 8–12 weeks old, bred, and housed in our facility, and all experiments were performed according to the guidelines described in the Association for Research in Vision and Ophthalmology Statement for the Use of Animals in Vision and Ophthalmic Research and Baylor College of Medicine Animal Care and Use Committee policy.

Corneal epithelial wound model and experimental interventions

The central corneal abrasion was performed as described previously [21]. In brief, mice were anesthetized by i.p. injection of sodium pentobarbital (Nembutal; 40 mg/kg body weight). The central epithelium was delineated with a 2-mm trephine and removed using a golf club spud for refractive surgery (Accutome, Malvern, PA, USA) under a dissecting microscope. While under anesthesia, the ocular surface was protected from drying by topical administration of sterile-buffered balanced salt solution (PBS) immediately following injury. To assess the rate of wound closure (as described previously [5]), the corneas were examined by fluorescein staining every 6 h after wounding and photographed with a digital camera. The size of epithelial defect was expressed as a percentage of the original wound area. An additional parameter of epithelial healing, quantitation of dividing basal epithelial cells, was also used as described previously [5].

To examine the contribution of NK cells to CD11c+ cell migration in wounded corneas, mice were injected i.p. with purified antiasialo-GM1 (Wako Pure Chemical Industries, Osaka, Japan) or purified anti-NK1.1 (mouse IgG2a, PK136 clone; eBioscience, San Diego, CA, USA), as published previously [21]. The dose of each antibody was 50 μg in 200 μl PBS, given 24 h before corneal abrasion to deplete NK cells. Negative controls for the depletion experiments used rabbit IgG, 50 μg in 200 μl PBS. To analyze the local effects of IFN-γ on corneal inflammation after wounding, mice received topically on the cornea 5 μl PBS containing 1 μg purified anti-IFN-γ antibody (functional grade purified goat anti-mouse IFN-γ from eBioscience) or 5 μl PBS containing 0.5 μg mouse rIFN-γ (R&D Systems, Minneapolis, MN, USA), dissolved in PBS every 4 h for 24 h. The control animals received an equal concentration of goat IgG solution or PBS alone.

Immunofluorescence and deconvolution imaging

Wounded corneas with the complete limbus were dissected, fixed (2% formaldehyde for 1 h), permeabilized (0.1% Triton X-100), and then incubated with the following labeled mAb: FITC- or PE-conjugated anti-CD11c (HL3 clone; BD PharMingen, San Diego, CA, USA) for DCs; PE-conjugated anti-CD11b (M1/70; BD PharMingen) for macrophages; FITC- or PE-conjugated anti-NKp46 (29A1.4 clone; R&D Systems) for NK cells; FITC- or allophycocyanin-conjugated anti-CD31 (MEC 13.3 clone; BD PharMingen) for limbal vessel endothelium; FITC- or PE-conjugated anti-MHCII (M5/114.15.2; eBioscience). Each step was followed by three washes with PBS. Controls using isotype- and species- matched antibodies were, in all cases, negative. Radial cuts were made in the cornea so that it could be flattened by a coverslip, and the cornea was mounted in Celvol 205 (Sekisui Specialty Chemical Company, Dallas, TX, USA), containing 1 μg/ml DAPI (Sigma-Aldrich, St. Louis, MO, USA), to assess nuclear morphology. Image analysis and quantification of wounded corneas were performed using DeltaVision (Applied Precision, Issaquah, WA, USA). Whole mounts were evaluated using a 40× oil immersion lens to assess each field of view (field-of-view diameter, 0.53 mm) across the cornea from limbus to limbus. The limbus was defined as the peripheral field containing limbal vessels (limbal vessels stained with anti-CD31) and evident in all whole mounts as a reference for counting leukocytes. The graphical values were obtained by counting the total number of a selected corneal layer in each of nine, 40× fields of view comprising the diameter of a cornea.

Isolation of NK cells for adoptive transfer

NK cells were isolated from the spleens of WT mice, as published previously [21]. Briefly, the spleens were first teased into pieces using a pair of forceps, minced, and filtered through a 40-μm pore cell strainer (Falcon; BD Biosciences, San Jose, CA, USA). After preparing a single-cell suspension of mouse spleens, NK cells were isolated using a NK cell isolation kit (Miltenyi Biotec, Auburn, CA, USA). To assess the effect of adoptive transferring, isolated NK cells were labeled with PKH67 cell linker with green fluorescence (5 μM; Sigma-Aldrich) for general cell-membrane labeling, followed by washing twice with DMEM, supplemented with 10% FCS and tissue-culture antibiotics (streptomycin and penicillin). Twenty-four hours after NK cell depletion with antiasialo-GM1, corneal abrasion was performed in recipient mice, followed by reconstitution of NK cells (i.v.) with ∼2 × 10⁶ freshly isolated donor NK cells, as we and others have published previously [21, 22]. The number of CD11c⁺ DCs in the cornea was recorded at 48 h after wounding.

Flow cytometry analysis of inflammatory cells in the cornea

For flow cytometric analysis, the corneas were cut with a 4-mm trephine and then collected and digested using collagenase (1 mg/ml; Sigma-Aldrich) at 37°C for 30 min. The digested samples were filtered with a 100-μm cell strainer and then incubated with fluorescent-labeled antibodies on ice for 20 min, washed twice with cold PBS, and resuspended in PBS containing 1% paraformaldehyde. Fluorescence of surface markers was recorded from 30,000 events occurring within standardized flow and fluorescence parameters using a FACS LSRII (BD Biosciences) and analyzed with FlowJo 7.5 (TreeStar, Ashland, OR, USA) software. Cell viability was assessed with 7-amino actinomycin D (eBioscience), and labeled, nonbinding, isotype-matched antibodies served as controls.

Data analysis

Data analysis was performed using ANOVA and pairwise multiple comparisons using Bonferroni test. A value of P < 0.05 was considered significant. Data are expressed as means ± sd.

RESULTS

Changes in CD11c⁺ cells after epithelial abrasion

Without injury, CD11c⁺ cells are known to be most evident in the peripheral cornea and limbus in the epithelium and stroma, with sparse distribution toward the central cornea [29–31]. In whole-mount preparations in the current experiments, clear distinction could be seen between the CD11c⁺ cells in the epithelium in the plane of the basal cells and those in the stroma below the basal epithelial cells. Those in the epithelium exhibited a typical dendritic morphology, with fine processes inserted between basal epithelial cells and into the stratified epithelium, consistent with reports from other investigators [29, 32], and these cells were mostly negative for MHCII, as noted by others [29]. In the stroma, CD11c⁺ cells were distributed most abundantly in the region of the limbal blood vessels, with very few cells evident in the paralimbal region, and these cells were neither dendritic in appearance nor positive for MHCII.

Changes in CD11c⁺ cells with dendritic morphology (DCs) in the epithelium (Fig. 1A) were analyzed after central corneal abrasion. The total number of these cells counted in nine fields of view across the cornea from limbus to limbus did not vary significantly over the first 36 h after epithelial abrasion (Fig. 1C), a time-frame that extends ∼12 h beyond epithelial wound closure in this model [5]. In the uninjured corneas, the number of DCs counted (using the analysis pattern described in Materials and Methods) was 39.1 ± 8 (mean and sd), and at 36 h after abrasion, the number was 46.4 ± 6.2. In addition, DCs were very rare (less than one cell/field of view) in the center, paracenter, and parawound regions of the cornea throughout this time period. At 48 h after injury, DCs increased to 107 ± 7.2 (P<0.001) compared with uninjured (n=4) across the cornea (Fig. 1C), evident in all regions of the epithelium, from the limbus to the center (Fig. 1D). As mice exhibit some sexual dimorphism [33], we analyzed DCs in male mice and found a generally consistent pattern with the female data, although with higher baseline numbers (72.5±8.7), some dip at 24 h (55.5±6.2) and greater numbers at 48 h after injury (181.3±15.9, n=3, P<0.01). Migration of DCs into the central region of the cornea was not evident at 24 h but was evident at 48 h. All subsequent data were collected in female mice.

Figure 1. CD11c⁺ cells in the epithelium and stroma of murine corneas after central epithelial abrasion.

(A) Representative images of CD11c⁺ DCs in the paralimbus of the corneal epithelium at 48 h after wounding. (Left panel) Anti-CD11c-PE-stained cells (red); (right panel) anti-MHCII-FITC (green); DAPI stain of epithelial nuclei (blue). (B) Representative images of CD11c⁺ cells in the limbal stroma at 48 h after wounding. (Left panel) Anti-CD11c-allophycocyanin (pseudocolored orange) and anti-CD31-PE (red); (right panel) anti-MHCII-FITC (green). Arrows indicate MHCII⁺ cells that were CD11c-negative. (C) Total epithelial CD11c⁺ cells (DCs) across one diameter of the cornea at different times after epithelial abrasion, and (D) the distribution of these cells at the 48-h time-point were plotted for the limbus (L), paralimbus (PL), original wound margin (WM), paracenter (PC), and center (C) of the cornea on the same graph as the distribution of DCs in unwounded corneas (n=4, mean±sd). (E) Total stromal CD11c⁺ cells across one diameter of the cornea were determined at different times after abrasion, and (F) the distribution of these cells at 48 h after injury was plotted from the limbus to the center of the cornea on the same graph as stromal CD11c⁺ cells in unwounded corneas (n=4, mean±sd). (G) Representative image of a CD11c⁺ cell in the epithelium at 12 h after wounding showing costaining for MHCII. (H) Representative image (left panel) of CD11c⁺ cells in the epithelium at 96 h after wounding, with failure to stain with anti-MHCII-FITC (right panel; position indicated by arrows). (I) Images of the limbal stroma at 48 h after injury showing CD11c⁺ cells (pseudocolor orange; left panel, arrows) that were not stained by anti-CD11b-FITC (right panel), although other cells in the field of view were positive for CD11b. **P < 0.01; ***P < 0.001.

CD11c⁺ cells were analyzed in the stroma of the abraded cornea. CD11c⁺ cells in the stroma were not dendritic in morphology (Fig. 1B), like those in the epithelium (Fig. 1A). The stromal CD11c⁺ cell number increased at 36 h from 75.7 ± 7.9 across the corneal stroma in uninjured mice to 180.9 ± 24.6 after injury (Fig. 1E) and was highest at the 48-h analysis time (283.4±27.5, n=4, P<0.001). The distribution was limited primarily to the region of the limbus extending to the original wound margin (Fig. 1F), with few CD11c⁺ cells in the stroma at the center of the cornea (average of two/field of view). At 96 h after injury, CD11c⁺ cell counts in the stroma remained elevated (154.8±31.1) over that of uninjured corneas (Fig. 1E). This was in contrast to the DCs of the epithelium, which returned to baseline levels at 96 h (Fig. 1C).

Changes in phenotypic characteristics of the epithelial DC and the CD11c⁺ cells in the stroma were analyzed. Without injury, the epithelial DCs were negative for MHCII, and <6% of the CD11c⁺ cells were positive for F4/80, a marker for macrophages. However, by 12 h after injury, most of the epithelial DCs and stromal CD11c⁺ cells were positive for MHCII, although the intensity of staining appeared less than at 48 h (Fig. 1G), and F4/80-positive cells were very rare in this population. The expression of MHCII by the epithelial DCs was transient, and by 96 h, they were rarely positive for MHCII (Fig. 1H). In contrast, most of the stromal CD11c⁺ cells were positive for MHCII at 96 h, with staining intensities similar to those seen at 48 h. At peak accumulation, the epithelial DCs and stromal CD11c⁺ cells were negative for CD11b (Fig. 1I), as well as F4/80, CD80, CD4, NKp46, and Ly6G (data not shown), although other cells in the cornea were positive for each of these markers. Thus, the CD11c⁺ cells, accumulating at 48 h in response to the abrasion injury, could be distinguished from macrophages, neutrophils, T cells, or NK cells. Epithelial abrasion, therefore, induced a transient increase of CD11c⁺ cells within the epithelium, highest at the 48 h time-point and waning by 96 h, with activation of MHCII expression evident within 12 h and waning in the epithelial DCs by 96 h but sustained in the stromal CD11c⁺ cells.

NK cells influence the accumulation of CD11c⁺ cells in the cornea

The peak influx of NK cells into the corneal stroma occurs at 24–30 h after abrasion injury [21], thus preceding by 18–24 h the highest observed accumulation of epithelial DCs and stromal CD11c⁺ cells (Fig. 1). To assess the possible contributions of NK cells to the changes of CD11c⁺ cells in the epithelium and stroma, we used two distinct approaches. The first was depletion of NK cells, and the second was i.v. transfer of isolated and labeled splenic NK cells into NK cell-depleted mice (using techniques that we [21] and others [22] have reported). The results from these experiments are described below and presented in Fig. 2.

Figure 2. Effects of depletion and restoration of NK cells.

WT mice received anti-NK1.1, antiasialo-GM1, or IgG, 24 h before central epithelial abrasion, or antiasialo-GM1, 24 h before abrasion and adoptive transfer of spleen NK cells at the time of epithelial abrasion. At 48 h after injury, CD11c⁺ cells were determined in the epithelium (A) and stroma (B) and expressed as total across two diameters of each cornea. (C) Representative images of CD11c⁺ DCs in the epithelium of IgG (upper image)- and antiasialo-GM1 (lower)-treated mice were labeled with anti-CD11c-PE (red) and DAPI (blue); original bars, 20 μm. (D) CD11c⁺ DCs in the epithelium were evaluated in uninjured corneas or 24 h-injured corneas at 24 h after i.p. injection of antiasialo-GM1. The sum of eight randomly selected fields in the paralimbus is plotted for controls (IgG-injected) and NK cell-depleted mice (n=4, mean±sd). (E) Representative images of CD11c⁺ DC in the stroma of IgG (upper image)- and antiasialo-GM1 (lower)-treated mice were labeled with anti-CD11c-FITC (green) and anti-CD31-PE (red; to label limbal vessels, out of focus in these fields of view); original bars, 20 μm. (F) Images of NKp46+ PKH67+ cells in the paralimbal epithelium and limbal stroma at 24 h after epithelial abrasion and adoptive transfer of PKH67-labeled spleen NK cell preparations. (G) The distribution of CD11c⁺ cells in the epithelium and (H) stroma was plotted from the limbus to the center of the cornea, showing the effects of depletion of NK cells by antiasialo-GM1 antibody and i.v. adoptive transfer of spleen NK cell preparations into NK cell-depleted mice compared with control mice injected with IgG (n=4–6, mean±sd). (I) Representative histograms from flow cytometry of digested corneas collected at 24 h after epithelial abrasion. Suspended cells were stained with anti-NKp46, anti-IFN-γ, and isotype and fluorochrome-matched, nonbinding IgG. Of the gated cells in the stained suspension, 34.1% were positive for NKp46, and of these, 82.9% were positive for IFN-γ. PE-A, PE-area. For (A, B, and D) *P < 0.05; ***P < 0.001. For (G and H), ***P < 0.001 comparing IgG with Anti-GM1; #P < 0.05 and ###P < 0.001 comparing Anti-GM1 with adoptive transfer.

NK cell depletion induced by antiasialo-GM1 or anti-NK1.1, given 24 h prior to corneal abrasion, reduced the number of CD11c⁺ cells in the epithelium but increased CD11c⁺ cells in the stroma. Data at 48 h after injury are shown in Fig. 2A and B. Depletion of NK cells using antiasialo-GM1 resulted in a 79% reduction in epithelial CD11c⁺ DCs at 24 h (n=4, P<0.001) after abrasion and as shown in Fig. 2A, a 91% reduction at 48 h. The CD11c⁺ cells remaining in the peripheral epithelium of NK cell-depleted mice retained the DC morphology (Fig. 2C). Antiasialo-GM1 did not alter the number (Fig. 2D) or MHCII negativity of resident CD11c⁺ cells in the uninjured epithelium. In contrast to the epithelium, CD11c⁺ cells in the stroma accumulated in the limbus in the NK cell-depleted corneas and concentrated between the limbal vessels and the basal epithelium (Fig. 2B and E). Thus, NK cell depletion altered the normal patterns of CD11c⁺ cells in response to corneal epithelial abrasion.

Adoptive transfer of NK cells into NK cell-depleted mice was performed, and their migration into the inflamed tissue was confirmed by detection in the tissues of PKH67⁺ cells that could be stained with anti-NKp46 (Fig. 2F), as we [21] and others [22] have published previously. Twenty-four hours after NK cell depletion with antiasialo-GM1, corneal abrasion was performed in WT recipient mice, followed immediately by i.v. injection of ∼2 × 10⁶ freshly isolated donor NK cells (preparations of spleen NK cells contained 74±5.2% NKp46⁺ cells, 95.3±1.3% of which labeled with PKH67, and 97±1% were viable). Localization in corneal stroma was confirmed by the detection of NKp46⁺ PKH67⁺ cells in the tissue. Compared with the control group (NK cell-depleted mice receiving PBS), CD11c⁺ DCs in the epithelium at 48 h after epithelial abrasion increased significantly (Fig. 2A), and their distribution increased in the limbus and paralimbus with extention toward the central cornea in mice receiving spleen NK cells (Fig. 2G). Additionally, with adoptive transfer of NK cells, the increased accumulation of CD11c⁺ cells in the stroma that was evident with NK cell depletion failed to occur (Fig. 2B and H).

IFN-γ and the response of CD11c⁺ cells to epithelial abrasion

NK cells have been reported to be a source of IFN-γ in infected murine corneas [20]. We assessed the possibility that this was true following corneal epithelial abrasion. Of cells collected from digested corneas at 24 h after epithelial abrasion, 28.1 ± 13.3% (n=3) were positive for NKp46, as analyzed by flow cytometry. Gating on this population revealed that 81 ± 15.7% (n=3) were positive for intracellular IFN-γ. Representative histograms are shown in Fig. 2I.

To determine if CD11c⁺ cell recovery was influenced by IFN-γ, anti-IFN-γ was applied topically at 4-h intervals between 12 and 36 h after epithelial abrasion. Corneas were then collected at 48 h, the peak time of CD11c⁺ cell accumulation in epithelium and stroma. Compared with the WT and IgG controls, anti-IFN-γ significantly inhibited the migration of CD11c⁺ cells into the epithelium and stroma (Fig. 3A and B). Distribution of cells into the central epithelium and central stroma was significantly inhibited by anti-IFN-γ (Fig. 3C and D). IFN-γ^−/− mice were also evaluated. Compared with WT mice, IFN-γ^−/− mice had significantly reduced numbers of CD11c⁺ cells in the epithelium and in the stroma (Fig. 3E–H), consistent with the effects of anti-IFN-γ. To determine if IFN-γ can promote accumulation of CD11c⁺ DCs in the corneal epithelium, topical application of rIFN-γ to the wounded corneas of NK cell-depleted mice followed the same treatment schedule as with the topical anti-IFN-γ experiments. Compared with the NK cell-depleted mice, rIFN-γ significantly increased the CD11c⁺ cells in the epithelium and promoted distribution of these cells toward the center of the cornea (Fig. 4A and B). Topical rIFN-γ also reduced marginally the number of CD11c⁺ cells in the stroma (Fig. 4C and D).

Figure 3. Effects of IFN-γ depletion on CD11c⁺ cell accumulation in abraded corneas.

(A–D) Topical application of anti-IFN-γ or equivalent IgG in PBS every 4 h between 12 and 30 h after central epithelial abrasion. At 48 h after injury, CD11c⁺ cells were counted in the epithelium (A) and stroma (B) across two diameters of each cornea, and the distribution of these cells in the epithelium (C) and stroma (D) was plotted from the limbus to the center of the cornea (n=4–6, mean±sd). (E–H) Corneas from WT and IFN-γ^−/− mice were analyzed at 48 h after central epithelial abrasion for accumulation of CD11c⁺ cells, counted in the epithelium (E) and stroma (F) across two diameters of each cornea. The distribution of CD11c⁺ cells in the epithelium (G) and stroma (H) was plotted from the limbus to the center of the cornea (n=4–6, mean±sd). *P < 0.05; **P < 0.01; ***P < 0.001.

Figure 4. Effects of rIFN-γ on CD11c⁺ cell accumulation in abraded corneas of WT mice depleted of NK cells.

(A–D) Corneas from WT mice, depleted of NK cells by antiasialo-GM1, exposed topically to PBS with and without rIFN-γ, every 4 h between 12 and 30 h after central epithelial abrasion, were analyzed at 48 h after abrasion for accumulation of CD11c⁺ cells, counted in the epithelium (A) and stroma (C) across two diameters of each cornea. The distribution of CD11c⁺ cells in the epithelium (B) and stroma (D) was plotted from the limbus to the center of the cornea (n=4–6, mean±sd). *P < 0.05; **P < 0.01; ***P < 0.001.

To determine if IFN-γ played a significant role in epithelial healing, IFN-γ^−/− mice were evaluated for wound closure and found to have delayed closure evident by 12 h after wounding (Fig. 5A). In addition, topical application of anti-IFN-γ every 6 h after wounding resulted in 42% reduction in dividing basal epithelial cells and 50% reduction in wound closure analyzed at 18 h after epithelial abrasion (Fig. 5B).

Figure 5. Corneal epithelial wound healing.

(A) Following 2-mm diameter corneal epithelial abrasion in WT and IFN-γ^−/− mice, the open-wound area was revealed by the application of fluorescein solution at 0, 6, 12, 18, 24, 30, and 36 h after wounding, calculated, and plotted as percent of the initial wound area (n=3). (B) Following 2-mm diameter corneal epithelial abrasions in WT mice, the corneas were topically treated every 4 h with anti-IFN-γ or control IgG. At 18 h after wounding, the open-wound area was determined, and corneas were collected, prepared for whole mounts, and analyzed microscopically for cell division in the paralimbal region of the cornea (n=3, mean±sd). *P < 0.05; **P < 0.01.

ICAM-1 and the response of CD11c⁺ cells to epithelial abrasion

A possible mechanism by which IFN-γ promotes accumulation of CD11c⁺ cells in the epithelium is sustained epithelial expression of ICAM-1, given that ICAM-1 may be necessary for migration of CD11c⁺ cells within this tissue. ICAM-1^−/− mice were analyzed for the number and distribution of CD11c⁺ cells in the cornea and found to be comparable with strain and sex-matched controls when uninjured (epithelial DCs in ICAM-1^−/−, 74 ± 27.6, and WT, 42.6±15.8; stromal CD11c⁺ cells in ICAM-1^−/−, 81 ± 5.8, and WT, 75.9±7.9, n=6, P=NS). However, after central epithelial abrasion, ICAM-1^−/− mice failed to sustain the level of CD11c⁺ DCs in the epithelium over a 48-h observation (Fig. 6A), and the distribution of the diminished population, remaining at 48 h after injury in the ICAM-1^−/− mice, was limited to the limbal epithelium and stroma (Fig. 6B and C). Thus, it appears that ICAM-1 is necessary for CD11c⁺ cell influx into the corneal epithelium and stroma in response to epithelial abrasion.

Figure 6. ICAM-1 and CD11c⁺ cells in the epithelium.

(A) Changes in the number of CD11c⁺ cells in the epithelium of WT and IFN-γ^−/− mice at 24 and 48 h after epithelial abrasion, plotted relative to the number of CD11c⁺ cells in the uninjured corneas. The values for the ICAM-1^−/− mice at 24 and 48 h were significant (n=4, P<0.01). The distribution of CD11c⁺ cells in the epithelium (B) and stroma (C) was plotted from the limbus to the center of the cornea (n=4–6, mean±sd). (D) Representative images of corneal basal epithelium at 48 h after central epithelial abrasion, labeled with anti-ICAM-1-PE showing two regions: paralimbus and central cornea. Each pair of images was captured at the same exposure time and shows increased labeling of the cornea in mice depleted of NK cells. The accompanying graph shows quantitation of fluorescence intensity from anti-ICAM-1-PE staining of the epithelium in the limbus, paralimbus, and center and vessels in the limbus. Arbitrary density values, means ± sd for randomly sampled areas in corneas, are plotted. *P < 0.05; ***P < 0.001.

To assess the possible influence of NK cell influx on ICAM-1 levels in the epithelium, the intensity of epithelial labeling with anti-ICAM-1 was analyzed and in contrast to expectations, was found to be increased in mice depleted of NK cells rather than reduced (Fig. 6D). Although ICAM-1 expression is necessary for CD11c⁺ cell migration into the epithelium, it is not sufficient. Our earlier studies revealed that NK cell depletion significantly increases the acute inflammatory response in abraded corneas [21], possibly accounting for the apparent enhancement of ICAM-1 expression.

DISCUSSION

In the current study, we present evidence that NK cells may provide a positive influence on corneal healing. They appear to be necessary for influx of CD11c⁺ MHCII⁺ DCs into the epithelium. Evidence supporting this interpretation includes markedly reduced numbers of epithelial DCs in the healing epithelium in NK cell-depleted mice. Adoptive transfer of splenic NK cell preparations into NK cell-depleted mice prevented this effect of NK cell depletion. The potential significance of these observations relates to recent evidence that DCs contribute to epithelial healing and tissue repair [8]. It appears that DCs, beyond their well-known role in adaptive immunity [14, 31, 34–38], interact with the corneal epithelial cells to promote expression of chemokines CXCL10 and TSLP and epithelial wound closure [8]. We have shown previously that CXCL10 is critical for efficient accumulation of NK cells in the cornea after epithelial abrasion [21], consistent with the hypothesis that resident CD11c⁺ cells initiate a cascade, including NK cells, that maintains and enhances the number of epithelial DCs during wound healing.

CD11c⁺ DCs migrate in the epithelium toward the wound center within the healing epithelium after an abrasion injury [8] or other types of surface damage to the cornea [39]. TNF-α and IL-1, as well as the chemokine receptor CCR5, contribute to this phenomenon [38, 40], and DCs increase in the central cornea. The mechanisms that maintain and increase DCs during corneal wound healing are not understood. The local environment may retain resident DCs with expansion of the population through immigration or cell division. However, under some experimental conditions, the epithelial DCs are known to leave the cornea through lymphatics to the regional LNs [41]. If this occurs during the healing of epithelial abrasion, immigration of new DCs would be necessary to sustain or expand the epithelial population. Here, we consider the possibility of CD11c⁺ DCs arriving in the healing epithelium by migration from the stroma in the limbal and paralimbal regions. Support for this hypothesis comes from the following experimental observations. NK cell depletion not only inhibited normal increases in the epithelial CD11c⁺ cell numbers but also resulted in a >85% reduction in these cells within the first 24 h after wounding. The reduction in the epithelium was associated with significant increases in CD11c⁺ MHCII⁺ cells in the limbal stroma at a position between the limbal vessels and the basal epithelium. Transfer (i.v.) of spleen NK cells into NK-depleted mice resulted in significant increases in epithelial CD11c⁺ MHCII⁺ DCs and decreases in stromal CD11c⁺ MHCII⁺ cells.

Three interventions in the current investigation dysregulated the patterns of CD11c⁺ cells in the healing epithelium after corneal surface abrasion. ICAM-1^−/−, NK cell depletion, and IFN-γ^−/− each resulted in markedly low levels of CD11c⁺ DCs in the epithelium within the first 48 h after injury. Furthermore, we have shown previously that ICAM-1^−/− [28] and NK cell depletion [21] significantly delay epithelial healing, and in the current study, we show that IFN-γ ^−/− delays epithelial healing significantly. As the normal response is to maintain and increase the CD11c⁺ DCs within this time period after injury, the >70% reduction in these cells at 24 and 48 h after injury as a result of these interventions is particularly interesting. It may reveal a substantial, early turnover of DCs in the healing epithelium, consistent with the concept that DCs leave the cornea after injury [41], and ICAM-1, NK cells, and IFN-γ are necessary for the immigration phase of this turnover. ICAM-1 is an adhesion molecule, up-regulated rapidly on the corneal epithelium and limbal vessel endothelium after injury [28]. As its expression is necessary for migration of γδ T cells into the healing epithelium [28] and migration of NK cells into the stroma after epithelial injury [21], its role in maintaining epithelial CD11c⁺ cells is complex. ICAM-1 may serve as an adhesive ligand for integrins on the CD11c⁺ cells, supporting their migration into the epithelium. It is also necessary for an inflammatory cascade, as γδ T cells are required for accumulation of NK cells in the stroma [21], which in turn, are required for recovery of epithelial DCs after epithelial injury.

The mechanisms of the NK cell contributions to corneal wound healing are not clearly defined. In the current study, we provide evidence that NK cells obtained from digested corneas are predominantly positive by flow cytometry for IFN-γ and that topical rIFN-γ increases DC recovery significantly in NK cell-depleted mice. This suggests a role for NK cell-derived IFN-γ, but a weakness in the current study is that we were unable to consistently detect IFN-γ by high-sensitivity ELISA in extracts of whole wounded corneas (unpublished results). The significant effect of topical anti-IFN-γ and the results in IFN-γ^−/− mice indicate an important role for this cytokine but fail to localize the source. In addition, in an earlier publication [21], we presented evidence that NK cells limit the acute innate inflammatory response to corneal abrasion. In that study, NK cell depletion resulted in a significant increase in the acute inflammatory response, as indicated by an enhanced and sustained influx of neutrophils into the corneal stroma. Wound healing was also delayed significantly by NK cell depletion, as evidenced by delayed wound closure, reduced epithelial cell division, and diminished sub-basal nerve regeneration. These results are consistent with the concept that NK cells limit the innate acute inflammatory response in the injured cornea, much like they were reported to limit an adaptive immune inflammatory response in other tissues [22–27], and that in the cornea [21] and lung [22], this function is dependent on NKG2D, an activating receptor for NK cell-dependent cytotoxicity and cytokine production. This newly recognized function of NK cells in the cornea thus may serve to protect the healing tissues from excessive inflammation. We have shown previously that excessive neutrophil accumulation is associated with very poor corneal wound healing [13]. Thus, the mechanisms by which NK cells support corneal wound healing are obviously complex, limiting neutrophils influx into the stroma and promoting DC influx into the epithelum. NK cells apparently intersect with several steps in the inflammatory cascade important for healing.

In summary, we found that depletion of NK cells significantly reduced the recovery of DCs in corneal epithelium during healing from epithelial abrasion, and i.v. transfer of spleen NK cell preparations reversed the effect of NK cell depletion. Topical administration of anti-IFN-γ to the wounded corneas also significantly reduced DC recovery, and IFN-γ^−/− mice exhibited significantly reduced DC recovery. NK cells in wounded corneas of WT mice were predominately positive for IFN-γ, and topical administration of rIFN-γ significantly increased DC recovery in NK cell-depleted mice. Depletion or absence of IFN-γ resulted in delayed epithelial wound healing. ICAM-1, an adhesion molecule expressed by injured corneal epithelium, was important, as ICAM-1^−/− mice exhibited markedly reduced DC recovery after epithelial wounding. DC recovery after corneal epithelial wounding is apparently dependent on NK cells, IFN-γ, and ICAM-1, each playing multifaceted roles in the inflammatory cascade necessary for efficient corneal wound healing.