Role of NK Cells in Skin Wound Healing of Mice

The skin is a complex organ that functions as a protective barrier against environmental and pathogenic threats, and such functions require coordination between different cell types, soluble factors, and extracellular matrix components. After skin injury, diverse populations of immune cells infiltrate the damaged tissue and interact with keratinocytes, fibroblasts, and endothelial cells to prevent infection and promote an efficient healing response, resulting in wound closure and restoration of barrier function (1, 2).

The healing process can be divided in three distinct but overlapping phases: inflammatory, proliferative, and remodeling (3). During the inflammatory phase, proinflammatory cells, including neutrophils and monocytes/macrophages (Mos/Mps), are recruited to the injury site and release proinflammatory factors to prevent infections and clear wound debris (4). During the proliferative phase, Mos/Mps and T cells downregulate proinflammatory factors and release soluble factors that promote proliferation and differentiation of fibroblasts, endothelial cells, and keratinocytes (5). During the remodeling phase, Mos/Mps play a role in reorganization of the provisional matrix, and then immune cells are cleared from the injury site and tissue homeostasis is restored (3, 5, 6). Although the roles of neutrophils, Mos/Mps, and T cells in wound healing have been reasonably well studied, there is only limited information available on the role of another major class of immune cells: NK cells, a member of the innate lymphoid cell family.

NK cells are defined by their “natural killing” function through release of cytotoxic factors. NK cells act in host defense against virus-infected cells and tumor cells, inducing apoptosis of target cells on release of perforin and granzyme granules and through production of cytokines such as IFN-γ and TNF-α (7, 8). In addition to proinflammatory cytokines, NK cells can also produce growth factors such as GM-CSF and TGF-β, and thus can shape immune responses through cross-talk with other immune cells, such as Mos/Mps, T cells, B cells, and dendritic cells, as well as other tissue-resident cells (9). In addition, hyperactivation or dysfunction of NK cells is associated with the pathogenesis of multiple diseases, including lupus, type 1 diabetes, and autoimmune liver disease. Thus, NK cells may play protective or pathogenic roles depending on their activation state/phenotype and tissue context (10, 11).

NK cell–derived cytokines and chemokines participate in the inflammatory response to tissue injury and thus have potential to influence wound healing (12). Although NK cells have been extensively investigated in cancer and viral infection (8, 11, 13, 14), few studies have focused on the role of NK cells in skin wound healing. Of the studies investigating the role of NK cells in wound healing, one reported that NK cells dampen inflammation and play a positive role in healing after corneal epithelial abrasion (15). However, a second reported that NK cell ablation accelerates skin wound healing, suggesting a negative role in healing (16). Therefore, much remains to be learned about the role of NK cells in wound healing.

In this study, we demonstrate that NK cells accumulate in skin wounds and proliferate locally in response to IL-15. Skin wound NK cells exhibit a primarily mature CD11b⁺CD27⁻ and NKG2A⁺NKG2D⁻ phenotype, express Ly49I, and contribute to the expression of proinflammatory cytokines in wounds. Importantly, NK cell depletion enhanced re-epithelization and collagen deposition, suggesting a negative role for these cells in skin wound healing.

C57BL/6J mice were purchased from the Jackson Laboratory (Bar Harbor, ME) and bred at the Association of Assessment and Accreditation of Laboratory Animal Care approved animal facility at the University of Illinois at Chicago. Mice were maintained in controlled environment conditions with 12-h light/dark cycles and water and food (standard diet) ad libitum and were housed in groups of five animals until experimentation. Each experiment was performed with at least two cohorts with a total of n = 6 male mice per group (age 11–14 wk). All animal studies were approved by the Animal Care Committee of the University of Illinois at Chicago and followed the Guide for the Care and Use of Laboratory Animals, in compliance with the U.S. Department of Health and Human Services.

Full-thickness excisional skin wounding was performed as previously described (17, 18). Eight-mm excisional wounds were made on the dorsum of mice with a sterile 8-mm dermal biopsy punch, covered with Tegaderm (3M) held in place with Coban (3M) self-adhesive bandage wrap, as previously described (17). External wound closure was measured in digital images of the dorsum of each animal obtained immediately after wounding procedure (day 0) and before euthanasia on days 3, 5, and 7 postinjury. Wound area was measured using Fiji ImageJ and expressed as percentage reduction of the area compared with day 0. Animals were euthanized by isoflurane overdose followed by cervical dislocation, and samples were collected for various assays as described later.

NK cells were depleted systemically as previously described (16, 19), using i.p. injections of 100 µg of NK1.1 Ab (clone PK136, 108702; BioLegend, San Diego, CA), immediately after wounding (day 0) and days 2, 4, and 6 postinjury. Controls were injected with equivalent doses of IgG2a control Ab (clone MOPC-173, 400202; BioLegend). Ab solutions were diluted in sterile PBS with a total volume of 200 µl/injection. Samples were collected at days 3, 5, and 7 postinjury.

IL-15 was neutralized locally in wounds with intradermal (i.d.) injection of 4 µg of anti–mIL-15 (murine IL-15) Ab (polyclonal, AF447; R&D Systems, Minneapolis, MN) per wound, immediately after wounding (day 0) and days 2 and 4 postinjury. Controls were injected with equivalent doses of IgG control Ab (I-2000-1; Vector Labs, Burlingame, CA). For i.d. injections, the total dose was distributed equally over four sites around the periphery of the wound, as previously described (20). Samples were collected on day 5 postinjury.

Wound cells, spleen cells, femoral bone marrow cells, and peripheral blood cells were collected for cell sorting, quantitative PCR (qPCR), and flow cytometry analysis. For peripheral blood cells, 100 µl of blood per animal was collected after deep isoflurane anesthesia but before euthanasia by cardiac puncture in tubes containing EDTA. Blood samples were subjected to RBC lysis and then filtered, washed, and resuspended in FACS buffer in a single-cell suspension. For wound cells, after blood collection, mice were perfused with PBS through the left ventricle, the pelt was removed from the dorsum of the mouse, and wounds were harvested with an 8-mm biopsy punch and subjected to enzymatic digestion and mechanical disruption to obtain a single-cell suspension in a pool of four wounds per animal, as previously described (20). For bone marrow cells, two femurs per animal were collected and flushed with DMEM. Cells were then filtered, centrifuged, and resuspended in FACS buffer in a single-cell suspension. For spleen (whole organ) and liver (left lobe) cells, samples from each animal were homogenized in DMEM with mechanical disruption and the help of a syringe and a 21G needle. Cells were then filtered, centrifuged, and resuspended in FACS buffer in a single-cell suspension.

NK cells were sorted from single-cell suspensions of blood, bone marrow, wound, or spleen samples using MojoSort magnetic beads system (480020; BioLegend), MojoSort Mouse Anti-allophycocyanin Magnetic Nanobeads (480071; BioLegend), and NK1.1-allophycocyanin Ab (clone PK136, 108710; BioLegend), following the manufacturer’s instructions. Purity and yield of sorting were confirmed by flow cytometry using CD3-PE Ab (clone 145-2C11, 12-0031082; Invitrogen, Waltham, MA), NK1.1-allophycocyanin Ab (clone PK136, 108710; BioLegend), and NK1.1-FITC Ab (clone PK136, 11-594182; Invitrogen).

Single-cell suspensions from wounds, bone marrow, and blood were Fc blocked with anti-CD16/32 Ab (clone 93, 101319; BioLegend) and labeled with CD3-PE (clone 145-2C11, 12-0031082; Invitrogen), NK1.1-FITC (clone PK136, 11-594182; Invitrogen), NK1.1-BV650 (clone PK136, 108735; BioLegend), CD11b-allophycocyanin (clone M1/70, 101212; BioLegend), CD11b-FITC (M1/70, 101206; BioLegend), CD27-PE (clone LG.3A10, 124210; BioLegend), Ly6G-FITC (clone 1A8, 127606; BioLegend), Ly6C-PE (clone HK1.4, 128008; BioLegend), NKG2A-PeCy7 (clone 16A11, 142810; BioLegend), and NKG2D-PE (clone A10, 115605; BioLegend). For wound samples, cells were also stained with Zombie Violet or Aqua (423114 and 423102; BioLegend) to assess cell viability.

For proliferation/cell-cycle analysis, cells were fixed and permeabilized using Cytofix/Cytoperm kit (554722; BD Biosciences, Franklin Lakes, NJ), for intracellular labeling with anti-Ki67 Ab (polyclonal, ab15580; Abcam, Cambridge, MA) followed by AF488 secondary Ab (polyclonal, ab150077; Abcam). Finally, cells were incubated with FxCycle Far Red (F10348; Invitrogen), following the manufacturer’s instructions.

All samples were read in a Cytoflex S (Beckman Coulter) cytometer, and data were analyzed with FlowJo software (FlowJo, Ashland, OR).

Wound histology was assessed essentially as described previously (18). Wounds were embedded in tissue freezing medium and snap frozen in isopentane cooled with liquid nitrogen, followed by sectioning at 10-µm thickness in a Leica CM1520 cryostat. Cryosections taken from the center of the wound (identified by sectioning from one edge of the wound to well past the middle of the wound) were used for H&E staining (re-epithelization), Trichrome staining (granulation tissue and collagen deposition), and immunohistochemistry (NK cell accumulation). For immunohistochemistry, M.O.M. immunodetection kit (BMK-2202; Vector Labs) was used to minimize background staining, and NK cells were labeled with NK1.1 primary Ab (1:50; clone PK136, 108702; BioLegend) followed by staining with (3-amino-9-ethylcarbazole) HRP substrate (SK-4200; Vector Labs) and counterstained with Hematoxylin QS (H3404-100; Vector Labs). Digital images were obtained with Keyence BZ-X710 microscope with 2× (H&E and Trichrome) and 20× (Trichrome and immunohistochemistry) objectives and analyzed with ImageJ.

Percentage of re-epithelialization, length of epithelial tongues, and granulation tissue area were measured in three sections per wound and averaged over sections to provide a representative value for each wound. Collagen deposition was quantified by automated counting of the number of clearly stained pixels based on a selected threshold intensity and normalized by total number of pixels in the image also in three sections per wound. Stained NK cells present in the wound bed were counted and expressed per square millimeter, also in three sections per wound.

Total cells from two separated wounds per animal and sorted NK cells from a pool of four wounds per animal were used for qPCR analysis. Samples were collected in TRIzol reagent (15596026; Invitrogen) and homogenized with a bead homogenizer for mRNA isolation and cDNA synthesis with SuperScript Vilo Kit (11754-050; Invitrogen), following the manufacturer’s instructions. PCR was performed in ABI 7500 Fast Real Time PCR using Fast Sybr Green master mix (4309155; Applied Biosystems, Foster City, CA). Expressions of Il-10, Il-1b, Tnf-a, and Ifn-γ genes were evaluated, and Rpl-4 was used as housekeeping gene. Fold change was calculated using the 2-δ δ cycle threshold method, and values are expressed as fold increase/decrease relative to control group (nonwounded skin).

The phenol phase from mRNA isolation with TRIzol reagent was saved for total protein isolation, following the manufacturer’s instructions. Samples were then diluted, and 10 µg of total protein was used for specific protein detection and quantification with LegendPlex Custom Multiplex (BioLegend) by flow cytometry, following the manufacturer’s instructions. A custom multiplex assay was developed for detecting and quantifying IL-15, CX3CL1, CXCL10, and IFN-γ.

Data are expressed as mean ± SD. Differences between groups were evaluated by t test, one-way ANOVA + Tukey post hoc test, and two-way ANOVA + Sidak post hoc test as indicated. For NK cell depletion experiments, we used ANOVA models with main effects of treatment (two levels: NK1.1 blocking Ab and isotype control Ab) and time point (three levels: days 3, 5, and 7) and treatment-by-time interaction effects. Statistical significance (p < 0.05) is shown in the figures using the hashtag symbol (#) for main effect of treatment from ANOVA results and an asterisk (*) for differences at specific time points from post hoc testing.

We first established the time course of NK cell accumulation in excisional skin wounds of mice. Our immunohistochemistry analysis indicated that NK cells accumulated in the wound bed granulation tissue with a peak on day 5 postinjury in both male (Fig. 1A, 1B) and female mice (Supplemental Fig. 1A, 1B). Flow cytometry analysis corroborated this finding, demonstrating that CD3-NK1.1⁺ cells accumulate in wounds with a peak on day 5 postinjury (Fig. 1D). In blood, CD3-NK1.1⁺ cell levels did not change from day 0 to day 3 postinjury, but significantly decreased on day 5 (Fig. 1E). In bone marrow, there were no significant changes in CD3⁻NK1.1⁺ cell levels, although a trend increase at late time points was observed (Fig. 1F). In liver and spleen, CD3⁻NK1.1⁺ cell levels tended to decrease from day 0 to day 5 (Supplemental Fig. 1C, 1D) (p = 0.06 and p = 0.09, respectively), indicating these tissues may be potential sources of wound NK cells. Interestingly, the ratio of CD3⁻NK1.1⁺ cells in blood and wounds was significantly lower at day 5 than the other time points, indicating a possible redistribution from blood to wounds (Fig. 1G).

As expected (21), CD3⁺NK1.1⁻ T cells were also increased in wounds compared with uninjured skin, persisting from days 3 to 7 postinjury (Supplemental Fig. 1E), whereas very few CD3⁺NK1.1⁺ NKT cells were observed in either uninjured skin or wounds, with no significant changes after wounding (Supplemental Fig. 1F).

Next, we determined whether NK cells proliferate in wounds and whether IL-15, a cytokine previously reported to mediate homeostatic proliferation of NK cells in different tissues (22–24), mediates NK cell proliferation in wounds. We first measured IL-15 protein levels in uninjured skin and in wounds and found that IL-15 was present and demonstrated a nonsignificant trend of an increase after injury (Fig. 2A).

To investigate whether IL-15 mediates NK cell proliferation in wounds, we neutralized IL-15 locally with i.d. injections of IL-15 neutralizing Ab. Neutralizing IL-15 significantly decreased CD3⁻NK1.1⁺ cells in wounds (Fig. 2D) and decreased proliferating CD3⁻NK1.1⁺Ki67⁺FxCycle⁺ cells (Fig. 2E) on day 5 postinjury, when compared with IgG isotype-treated controls. These data indicate that IL-15 promotes NK cell proliferation, contributing to their accumulation in wounds. In contrast, the IL-15 blocking Ab did not have an effect on proliferation of CD3⁻NK1.1⁻ cells (Fig. 2F), indicating the effect was specific to NK cells.

In addition, chemokines, including those in the CXC and CX3C family, have been shown to direct NK cells to sites of inflammation in peripheral tissues (13, 25–27). Two such chemokines, CX3CL1 and CXCL10, were present in wounds (Supplemental Fig. 1G, 1H), but only CX3CL1 was significantly increased over levels in uninjured skin (Supplemental Fig. 1G).

We then began to investigate the phenotypes of NK cells that accumulate in skin wounds. Our flow cytometry data show that the majority (∼70%) of NK1.1⁺ cells sorted from wounds at all time points examined (days 3–7 postinjury) stained positively for CD11b, but not CD27, on their surface, indicating a mature phenotype (28). Double-positive CD11b⁺CD27⁺ staining was also observed for ∼20% of NK1.1⁺ cells, which are considered intermediate maturity, and <3% of NK1.1⁺ cells were CD11b⁻CD27⁺, indicating only a minor contribution from this less mature subset (Fig. 3B). Interestingly, the least mature subset CD11b⁻CD27⁻ represented ∼5–10% of NK1.1⁺ cells sorted from wounds (Fig. 3B). The percentages of these subsets changed significantly at day 7 postinjury, at which point intermediate CD11b⁺CD27⁺ cells increased and mature CD11b⁺CD27⁻ cells decreased, with no significant changes in CD11b⁻CD27⁻ cells (Fig. 3B).

Next, we measured expression of activating NKG2D and inhibitory NKG2A receptors on NK1.1⁺ cells. Interestingly, ∼75% of NK1.1⁺ cells from wounds present an NKG2A⁺NKG2D⁻ phenotype and <5% present NKG2A⁺NKG2D⁺ or NKG2A⁻NKG2D⁺ phenotypes, and there was little change in percentages of cells expressing these receptors from days 3 to 7 postinjury (Fig. 3C). These data indicate that wound NK cells predominantly express the inhibitory receptor NKG2A, but not the activating receptor NKG2D. Another inhibitory receptor, Ly49I, was expressed in ∼70% of CD3⁻NK1.1⁺ cells in uninjured skin and in ∼90% of CD3-NK1.1⁺ cells in day 5 wounds (p = 0.082) (Supplemental Fig. 2F).

In peripheral blood of uninjured mice, NK1.1⁺ cells showed a more mixed distribution of phenotypes with ∼40% mature CD11b⁺CD27⁻ cells, ∼25% intermediate CD11b⁺CD27⁺ cells, and ∼10% less mature CD11b⁻CD27⁺ and CD11b⁻CD27⁻ cells, similar to previous reports of NK cell phenotype distribution in blood (29). After skin wounding, the phenotype of peripheral blood NK1.1⁺ cells significantly changed toward that observed in wounds, with ∼60% of NK1.1⁺ cells presenting a mature CD11b⁺CD27⁻ phenotype, ∼20% intermediate CD11b⁺CD27⁺, and ∼10% immature CD11b⁻CD27⁺ or CD11b⁻CD27⁻ cells (Fig. 3D). Similar to wounds, blood NK1.1⁺ cells predominantly express the inhibitory NKG2A receptor (∼50%), whereas <5% are either NKG2A⁺NKG2D⁺ or NKG2A⁻NKG2D⁺ cells, with little change over the time course analyzed (Fig. 3C, 3E).

In bone marrow, the phenotype of NK1.1⁺ cells differed from that observed in wounds and peripheral blood. Bone marrow NK1.1⁺ cells expressed a less mature phenotype with ∼60% CD11b⁻CD27⁺ cells, ∼40% intermediate CD11b⁺CD27⁺ cells, and <1% of mature CD11b⁺CD27⁻ cells, which did not differ significantly from day 0 to day 7 postinjury (Fig. 3F); these phenotypes are similar to those reported previously for bone marrow NK cells (29). Finally, bone marrow NK1.1⁺ cells exhibited a similar pattern of NKG2 receptor expression as in wounds and blood, with ∼60% of NK1.1⁺ cells expressing inhibitory NKG2A, but not activating NKG2D receptor, and <5% NKG2A⁺NKG2D⁺ or NKG2A⁻NKG2D⁺ cells from day 0 to day 7 postinjury (Fig. 3C, 3E, 3G).

In liver and spleen, CD3⁻NK1.1⁺ cells expressed a generally mature phenotype with ∼50–60% CD11b⁺CD27⁻ cells, whereas <20% of cells expressed intermediate CD11b⁺CD27⁺ and less mature CD11b⁻CD27⁺/CD11b⁻CD27⁻ phenotypes each (Supplemental Fig. 2B, 2C). In addition, the inhibitory Ly49I receptor was expressed in ∼90% and ∼50% of CD3⁻NK1.1⁺ cells of liver and spleen, respectively (Supplemental Fig. 2D, 2E).

Next, we sought to determine the role of NK cells in skin wound healing using a model of systemic NK cell depletion (14, 19). First, we confirmed that serial i.p. injections of the NK1.1 PK136 Ab significantly decreased CD3⁻NK1.1⁺ cells in peripheral blood at each time point postinjury, reducing their levels from ∼5 to 7% of the blood cell population for mice treated with IgG2a isotype controls to ∼1–2% for mice treated with NK1.1 Ab (Fig. 4C). The blocking Ab also decreased CD3⁻NK1.1⁺ cells in skin wounds at each time point, including the time point of peak accumulation (day 5), at which point systemic depletion reduced wound NK cells from ∼8% to ∼ 2% of viable gated cells when compared with mice treated with control Ab (Fig. 4D). Thus, the systemic NK cell depletion successfully reduced NK cell accumulation in skin wounds.

Importantly, NK cell depletion resulted in significantly accelerated wound closure (mean difference of 7.4% over all time points) in digital images of the wound surface (Supplemental Fig. 2G, 2H) and significantly increased histological measurements of re-epithelialization (mean difference of 11.0% over all time points) (Fig. 4E, 4F), as evidenced by a significant main effect of NK1.1 Ab treatment compared with control Ab in ANOVA of each dataset; each dataset also showed a significant main effect of time point, but no interaction effect. NK cell depletion also significantly reduced granulation tissue area (Fig. 4G) as evidenced by a significant main effect of Ab treatment in ANOVA of histological measurements, although the effect size was small (mean difference of 0.8 mm²). Finally, NK cell depletion significantly increased collagen deposition, assessed by Trichrome staining (Supplemental Fig. 2I). The ANOVA for trichrome staining showed a significant main effect of Ab treatment, as well as a significant interaction between treatment and time point. Post hoc testing showed significant differences on days 5 and 7 postinjury; blue collagen staining increased from ∼30 to ∼50% at day 7 postinjury (Fig. 4H). NK cell depletion did not change levels of dead cells isolated from wounds (Supplemental Fig. 2A), providing evidence that NK cells do not kill cells in wounds. In summary, these data indicate a small but significant positive effect of NK cell depletion on wound healing.

Finally, to assess the role of NK cells in the inflammatory response during wound healing, we determined whether NK cell depletion altered the accumulation of other leukocytes, including neutrophils and Mos/Mps. Our data demonstrate that NK cells ablation did not alter the levels of neutrophils (Ly6G⁺CD11b⁺ cells) (Fig. 5B), Mos (Ly6G⁻Ly6C^hiCD11b⁺) (Fig. 5C), or Mps (Ly6G⁻Ly6C^loCD11b⁺ cells) in wounds (Fig. 5D). Nonetheless, NK cell depletion significantly reduced mRNA expression of inflammatory cytokines in wounds, including Ifn-γ (Fig. 5E), Tnf-α (Fig. 5F), and Il-1β (Fig. 5G), as evidenced by a significant main effect of NK1.1 PK136 treatment compared with control Ab treatment in ANOVA. In addition, mRNA expression of Il-1β and Tnf-α showed significant interaction effects, and post hoc analysis showed that Tnf-α was significantly lower at day 5 postinjury (Fig. 5F) and Il-1β was significantly lower at day 3 postinjury (Fig. 5G), compared with control Ab treatment. In contrast, NK cell ablation did not alter expression of Il-10 in wounds (Fig. 5H). The mRNA data were corroborated by protein data showing reduced IFN-γ in wounds of NK-depleted mice compared with controls (Supplemental Fig. 2J). Consistent with the depletion data, NK1.1⁺ cells sorted from wounds expressed elevated levels of Tnf-α and Il-1β, particularly on day 3 postinjury, compared with control NK1.1⁺ cells sorted from spleen (Supplemental Fig. 2K, 2L). Moreover, wound NK cells expressed Ifn-γ, but it did not change from days 3 to 7 postinjury (Supplemental Fig. 2M). Finally, expression of Il-10 was not detectable in sorted NK1.1⁺ cells from spleen or wounds (data not shown). In summary, these data indicate that NK cells contribute to the expression of proinflammatory genes during wound healing.

In this study, we showed that NK cells accumulate in excisional skin wounds, began to characterize their phenotype in wounds, and assessed their role in wound healing using Ab depletion experiments. Our findings showed that NK cell accumulation peaked on day 5 after wounding, and that IL-15–dependent proliferation contributes to their accumulation in wounds. In addition, wound NK cells presented a mature phenotype (primarily CD11b⁺CD27⁻) and contributed to the expression of proinflammatory cytokines in wounds. Finally, NK cell depletion resulted in enhanced re-epithelization and collagen deposition, suggesting a negative role for NK cells in wound healing.

IL-15 is crucial for NK cell development, NK cell proliferation, and survival, not only in bone marrow but also in secondary lymphoid tissues such as spleen (22, 24, 30). Thus, we investigated the role of IL-15 in the accumulation of NK cells in skin wounds. Importantly, we show that IL-15 inhibition in wounds resulted in a significant decrease in both total and proliferating NK cells, indicating that IL-15 contributes to NK cell accumulation in wounds at least partly by inducing their proliferation.

Interestingly, peripheral blood NK cell levels decreased at the time of peak wound NK cell accumulation, suggesting a redistribution from blood to wound sites, and that chemokines likely also play a role in wound NK cell accumulation. In addition, there were trends of decreased NK cell levels in liver and spleen after wounding, suggesting these tissues may be sources of wound NK cells, whereas no differences were observed in bone marrow NK cells. Further investigation is warranted to determine the source of NK cells in skin wounds and mechanisms involved in their trafficking.

Factors that might contribute to NK cell infiltration into the wound include the chemokines CX3CL1 and CXCL10, which are produced by Mo/Mp, fibroblast, endothelial, and dendritic cells (13, 27, 31). We found that CX3CL1 and CXCL10 were expressed in wounds, and CX3CL1 levels increased after wounding, but the time course of this increase did not match that of wound NK cells. Thus, further study is needed to identify the chemokines responsible for NK cell accumulation in skin wounds.

NK cells progress from the least mature CD11b⁻CD27⁻ phenotype, through CD11b⁻CD27⁺ and CD11b⁺CD27⁺ phenotypes, and finally to the most mature CD11b⁺CD27⁻ phenotype (28, 32). As expected, in uninjured mice, bone marrow NK cells presented with phenotypes that were less mature, whereas in blood these cells exhibit primarily a mature phenotype; these results are consistent with previous studies (29, 33, 34). After injury, whereas bone marrow NK cells remained primarily less mature, levels of mature NK cells increased in blood, suggesting that factors produced after skin injury might contribute to NK cells maturation systemically. In wounds, the majority of NK cells present the mature CD11b⁺CD27⁻ phenotype at all time points examined postinjury, with a smaller contribution from intermediate CD11b⁺CD27⁺ cells. As has been found in other studies, the mature CD11b⁺CD27⁻ phenotype of wound NK cells was associated with expression of the inhibitory receptor NKG2A, but not the activating receptor NKG2D (32). However, NK cells can express numerous activating and inhibitor receptors, including NKG2A, NKG2D and Ly49I, and we have only begun characterizing wound NK cells. For example, a study using mass cytometry to simultaneously analyze 35 cell surface Ags, including 28 NK cell receptors, indicated that there can be up to 30,000 distinct NK cell phenotypes in an individual (35).

Two prominent changes in the function of NK cells as they mature are loss of proliferative ability and gain of effector function (32, 34). The proliferative ability of CD11b⁺CD27⁻ and CD11b⁺CD27⁺ NK cells has been reported to be lower than that of CD11b⁻CD27⁺ and CD11b⁻CD27⁻ NK cells; however, the more mature cells retained some proliferative capacity, especially during NK cell replenishment after their depletion (32). These latter data are consistent with results from our study suggesting proliferation of mature NK cells in skin wounds. In addition, NK cell effector functions, including cytokine production, are reported to be enhanced as the cells mature and decrease expression of CD27 receptor (28, 32, 34). Our data demonstrate that NK cells from wounds, which exhibit a more mature phenotype, expressed Tnf-α, Il-1β, and IFN-γ, and that depletion of NK cells reduced expression of these cytokines in wound homogenates, suggesting that NK cells contribute to proinflammatory cytokine expression during the inflammatory stage of healing. In addition, wound NK cells did not appear to contribute to IL-10 expression in wounds, arguing against a regulatory function of these cells, previously described in infection models (36, 37).

Importantly, in our study, depletion of NK cells resulted in enhanced re-epithelization, decreased granulation area, and increased collagen deposition and did not change infiltration of other leukocytes such as neutrophils and Mos/Mps. These findings are consistent with those of a previous study showing that NK cell depletion resulted in accelerated wound closure assessed by external measurements (16). This previous study focused on the role of hypoxia-inducible transcription factor 1α (HIF-1α) in NK cells in regulating a trade-off between wound healing and bacterial defense (16). In this previous study, the lack of HIF-1α in NK cells enhanced neovascularization and accelerated wound closure, but compromised antibacterial defense. Lack of HIF-1α in NK cells also decreased expression of IFN-γ and TNF-α in wounds, similar to our findings with NK cell depletion. Thus, HIF-1α may promote the proinflammatory effects of NK cells observed in both the current and previous studies.

In contrast, a study investigating the role of NK cells in a model of corneal epithelial abrasion suggested that NK cells play a positive role in corneal healing, limiting innate acute inflammation on wounding (15). This study indicated that NK cell ablation resulted in increased neutrophil influx, exacerbated inflammation, inhibited corneal nerve regeneration, and inhibited epithelial healing (15). Importantly, the cornea is associated with so-called immune privilege, which involves tissue-specific tolerance and/or suppression of immune responses. Thus, such tissue-specific differences may contribute to differences observed between skin and cornea (38).

Two additional studies focusing on a different subset of NK cells, namely, NKT cells, reported that mice deficient in NKT cells exhibited accelerated wound closure and enhanced collagen deposition, but no alterations in infiltration or proliferation of other leukocytes, including Mo/Mp and T cells. For identifying NKT cells, the group considered a CD1d dimer⁺Ly49c⁺ population (39, 40). In our study, NKT cells identified as NK1.1⁺CD3⁺ cells were extremely rare; the differences between studies may reflect the different sets of markers used. In contrast, another study reported that mice lacking invariant NKT cells exhibited prolonged neutrophilic inflammation and delayed wound healing, suggesting that invariant NKT cells play a positive role in wound healing (41, 42). The different results of these studies indicate that more research is needed to elucidate the role of different NK cell subsets in wound healing.

Limitations of this study included the use of NK cells from spleen of uninjured mice as a control when evaluating gene expression of NK cells; NK cells in uninjured skin are rare, sometimes undetectable, limiting the assessment of NK cells in such samples. Although like wound NK cells, spleen cells express predominantly a mature CD11b⁺CD27⁻ phenotype, tissue-specific gene expression differences could exist. We also acknowledge that sample size (n = 6) may not have provided sufficient power to detect differences at specific time points for some assays. We performed a power analysis for each assay performed, which indicated power of >80% for detecting main effects of time and treatment, as well as interaction between time and treatment for most assays. A notable exception was the re-epithelialization data in Fig. 4F, which showed >99% power to detect the main effect of time, 81% power to detect the main effect of Ab treatment, but only 17% power to detect time by treatment interactions, thus limiting our ability to detect differences at specific time points. In addition, the IFN-γ expression data in Fig. 5E showed low power for each effect, 36% for main effect of time, 49% for main effect of Ab treatment, and 38% for the interaction effect between time and treatment, again indicating lack of power to detect differences for this assay.

In conclusion, we demonstrate in this study that NK cells accumulate and proliferate in wounds in an IL-15–dependent manner. We also show that NK cells contribute to the expression of proinflammatory cytokines in wounds, including IFN-γ and TNF-α. Finally, NK cell depletion resulted in enhanced re-epithelization and increased collagen deposition, indicating a negative role of NK cells in skin wound healing. Future studies should further investigate the factors that contribute to NK cell accumulation in wounds and their interaction with other wound cells that influence the healing response.

The authors have no financial conflicts of interest.

We thank Dr. Giamila Fantuzzi, University of Illinois at Chicago, for critical comments on a previous draft of the manuscript, and Rachel Lane, Biostatistics Core in the Center for Clinical and Translational Science at University of Illinois at Chicago, for advice on statistical analysis.