Chemokine Receptor CX3CR1 Mediates Skin Wound Healing by Promoting Macrophage and Fibroblast Accumulation and Function¹

Repair of tissue damage is an interactive process that involves soluble mediators, extracellular matrix components, resident cells, and infiltrating leukocytes, which participate differentially in three sequential phases: inflammation, proliferation, and maturation (1). Initially, neutrophils accumulate in wound sites, followed by a large influx of macrophages and a small number of T lymphocytes. In the proliferative phase, massive angiogenesis occurs, and fibroblasts migrate into the wound, proliferate, and transform into myofibroblasts, which play a major role in granulation tissue formation. The molecular mediators of wound repair are not fully understood but include cytokines, chemokines, matrix proteins, matrix metalloproteinases, and growth factors (2, 3, 4, 5, 6). This article is focused on the chemokines, a major class of leukocyte chemoattractant proteins, which can be classified into four distinct groups, named C, CC, CXC, and CX3C, based on patterns of conserved cysteines. The chemokines MCP-1/CCL2, MIP-1α/CCL3, and IFN-γ-inducible protein 10/CXCL10, and the chemokine receptor CXCR2 have been identified as regulators of specific leukocyte accumulation at wound sites (3, 7, 8, 9, 10, 11). However, the involvement of most chemokines in wound healing has not yet been addressed. We surveyed expression of inflammatory chemokines in a mouse model of skin wound healing to begin to fill this gap. In so doing, we identified particularly high expression of CX3CL1 (also known as fractalkine), the only known member of the CX3C chemokine subfamily, and its receptor CX3CR1.

CX3CL1 is expressed as a soluble chemokine domain form and as a membrane-bound form on the surface of inflamed endothelial cells, epithelial cells, macrophages, and vascular smooth muscle cells. Interaction between membrane-bound CX3CL1 and CX3CR1 on leukocytes mediates cell-cell adhesion (12, 13, 14). CX3CL1 is expressed in various organs, including skin. CX3CR1 is expressed by monocytes/macrophages, NK cells, a subpopulation of T cells, and smooth muscle cells (14, 15, 16, 17, 18). Accumulating evidence suggests that CX3CL1-CX3CR1 interaction contributes to the development of inflammatory diseases, such as rheumatoid arthritis, glomerulonephritis, atopic dermatitis, psoriasis, Crohn’s disease, and atherosclerosis (15, 19, 20, 21, 22, 23).

Chemokines and their receptors not only may coordinate leukocyte infiltration but also may modulate the function of nonhemopoietic and resident cells in damaged tissue, both directly and indirectly (2, 3). Recently, a profibrotic activity has been suggested for CX3CL1/CX3CR1 based on the finding that patients with systemic sclerosis have increased expression of CX3CL1 and increased CX3CR1⁺ leukocytes in fibrotic skin and lung lesions (16). Moreover, bone marrow (BM)³ cells are known to play important roles in multiple organs in repair/regeneration of injured tissues, through a complex interplay between adhesion molecules, cytokines, and chemokines (3, 24, 25), and CX3CR1 is expressed on most monocytes/macrophages (16, 17, 18, 21, 23). Based on this information, we hypothesized that CX3CL1-CX3CR1 may have multiple roles in wound healing. In the present study, we tested this hypothesis in a mouse model of excisional skin wound healing.

The following mAbs and polyclonal Abs (pAbs) were used in this study: rabbit anti-myeloperoxidase (MPO) pAb (Lab Vision), which reacts with mouse MPO; rat anti-mouse CD31 mAb (BD Biosciences); rat anti-mouse F4/80 mAb (clone BM8; eBioscience); rabbit anti-human CD3 pAb (DakoCytomation), which cross-reacts with mouse CD3; mouse anti-human α-smooth muscle actin (α-SMA) mAb (clone 1A4; DakoCytomation), which cross-reacts with mouse α-SMA; rabbit anti-CX3CR1 pAb (Abcam), which reacts with mouse CX3CR1; goat anti-mouse CCR2 pAb (Abcam); goat anti-mouse CX3CL1 pAb (R&D Systems); cyanine dye 3 (Cy3)-conjugated donkey anti-rat, -mouse, and -goat IgG, FITC-conjugated donkey anti-rabbit and anti-goat IgG, and aminomethylcoumarin acetate (AMCA)-conjugated donkey anti-rat IgG pAbs (Jackson ImmunoResearch Laboratories); and rabbit anti-CX3CR1-neutralizing pAb and control rabbit IgG (Torrey Pines Biolabs).

Male CX3CR1 knockout (KO) mice, generated as previously described (26), were backcrossed onto C57BL/6 mice from Taconic Farms for 10 generations. Control C57BL/6 mice were obtained from Taconic Farms. All animals were used at 8–10 wk of age and housed individually in cages under specific pathogen-free conditions during the experiments. All animals were used under the auspices of a protocol approved by the National Institute of Allergy and Infectious Diseases Animal Care and Use Committee.

Full-thickness wounds were created in the dorsal skin under sterile conditions. Briefly, mice were anesthetized with i.p. administration of ketamine and xylazine. After shaving and cleaning with betadine and 70% ethanol, the dorsal skin was picked up at the midline and punched through two layers of skin with a sterile disposable biopsy punch (4 mm in diameter; Miltex), generating one wound on each side of the midline. The same procedure was repeated three times, generating six wounds. Each wound site was digitally photographed at the indicated time intervals, and wound areas were determined on photographs using Adobe PhotoShop (version 7.0; Adobe Systems). Changes in wound areas over time were expressed as the percentage of the initial wound areas. In another series of experiments, 100 μl of anti-CX3CR1-neutralizing pAb or control IgG was administered i.p. at a concentration of 20 μg/ml into wild-type (WT) mice immediately after wounding as described previously (18, 27, 28). Thereafter, animals received similar injections each day up to 5 days following wound preparation. In some experiments, wounds and their surrounding areas, including the scab and epithelial margins, were cut at the indicated time intervals with a sterile disposable 8-mm diameter biopsy punch.

Wound specimens were fixed in 4% formaldehyde buffered with PBS and then embedded with paraffin. Sections were stained with H&E for histological analysis. Immunohistochemical analyses were also performed using anti-MPO, -F4/80, -CD3, -CD31, -α-SMA, -CX3CR1, or -CX3CL1 Abs, as described previously (5, 29). Deparaffinized sections were immersed in 0.3% H₂O₂ in methanol for 30 min to eliminate endogenous peroxidase activities. The sections were further incubated with PBS containing 1% normal serum corresponding to the secondary Ab and 1% BSA to reduce nonspecific reactions. The sections were incubated with Abs at a concentration of 1 μg/ml at 4°C overnight. Normal rat, rabbit, goat, or mouse IgG was used as negative control. After incubation with biotinylated secondary Ab, immune complexes were visualized using the Catalyzed Signal Amplification System (DakoCytomation) according to the manufacturer’s instructions.

Wound sections were also analyzed by double- or triple-color immunofluorescence microscopy to determine the localization of CX3CL1, CX3CR1, and CCR2 proteins as described previously (30, 31). Briefly, deparaffinized sections were incubated with PBS containing 1% normal donkey serum and 1% BSA to reduce nonspecific reactions. Then, the sections were further incubated with the combination of anti-CX3CR1 and -F4/80, anti-α-SMA or -CD31; anti-CX3CL1 and -F4/80; and anti-F4/80, -CX3CR1, and -CCR2 Abs at 4°C overnight. All Abs were used at a concentration of 1 μg/ml. After incubation with fluorochrome-conjugated secondary Abs (10 μg/ml) at room temperature for 30 min, the sections were observed by fluorescence microscopy.

The wound bed was defined as the area surrounded by uninjured skin and fascia, regenerated epidermis, and eschar (5, 30). The numbers of infiltrating macrophages and T cells within the wound beds were enumerated on 10 randomly chosen visual fields (×200) of the sections stained with anti-F4/80 and -CD3 Ab, respectively, and the average of the selected 10 fields was calculated. Using the freehand tool of PhotoShop, vascular areas, defined as CD31 immunoreactive in the wound beds, were measured and were expressed as the percentage of the entire wound bed area. All sample identifiers were blinded to the observer.

MPO activity was measured to evaluate neutrophil recruitment (5, 30). Skin wound sites were excised using a sterile disposable 8-mm diameter biopsy punch and were then homogenized in 1 ml of 50 mM potassium phosphate buffer solution with 0.5% hexadecyltrimethylammonium bromide (Sigma-Aldrich) and 5 mM EDTA. The samples were freeze-thawed three times and centrifuged at 12,000 × g at 4°C. MPO activities in the supernatants were assayed using the SUMILON peroxidase assay kit (Sumitomo Bekuraito) according to the manufacturer’s instructions. Absorbance was measured at 490 nm and MPO content was determined against a standard curve generated using known concentrations of reagent MPO (Sigma-Aldrich). The data were expressed as units per wound.

Hydroxyproline content was measured as the index of collagen accumulation at the wound sites, as described previously (5, 30). The excised wound samples were dried for 16 h at 120°C and were then hydrolyzed in 6 N HCl at 110°C for 12 h. Fifty-microliter aliquots were added to 1 ml of a solution containing 1.4% chloramine T (Sigma-Aldrich), 10% n-propanol, and 0.5 M sodium acetate at pH 6.0. After 20-min incubation at room temperature, 1 ml of Ehrlich’s solution (1 M p-dimethylaminobenzaldehyde in 70% n-propanol, 20% perchloric acid) was added and the mixture was incubated at 65°C for 15 min. Absorbance was measured at 550 nm and the amount of hydroxyproline was determined against a standard curve generated using known concentrations of reagent hydroxyproline (Sigma-Aldrich). The data were expressed as amount (micrograms) per wound.

Wound samples were homogenized with PBS containing Complete Protease Inhibitor Cocktail (Roche Diagnostics) and centrifuged at 5000 × g for 10 min. Supernatants were used to determine TGF-β1 and vascular endothelial growth factor (VEGF) levels with commercial ELISA kits (R&D Systems) according to the manufacturer’s instructions. Total protein in the supernatants was measured with a commercial kit (Bicinchoninic Acid Protein Assay kit; Pierce). The data were expressed as the target molecule (picograms) per total protein (milligrams) for each sample.

Total RNA was isolated from uninjured and wounded skin samples using TRIzol (Invitrogen Life Technologies) reagent according to the manufacturer’s instructions. Before reverse transcription, RNA was treated with RNase-free DNase I following the manufacturer’s protocol. Three micrograms of total RNA was used for the synthesis of cDNA using a Superscript III RT kit (Invitrogen Life Technologies). Unique primer sets for mouse β-actin, chemokines, chemokine receptors, growth factors, and collagen type I were designed (Table I), based on sequences deposited with the National Center for Biotechnology Information, and were synthesized by Invitrogen Life Technologies. PCR was performed using 1 μl of cDNA in a reaction mix with Taq polymerase (Invitrogen Life Technologies). PCR was found to be linear between 20 and 35 cycles, and PCR conditions were optimized to allow for semiquantitative comparison of the results. Digital images of bands separated on ethidium bromide-stained agarose gels were quantitated using NIH Image Analysis Software. The intensity of each primer product was normalized to the intensity of the β-actin primer product, and expressed relative to the levels found in uninjured skin of WT mice.

The protocol for syngeneic BM chimera mice has been previously described (32). The following BM chimera mice were prepared: male WT BM→female CX3CR1 KO mice, male CX3CR1 KO BM→female WT mice, male WT BM→female WT mice, and male CX3CR1 KO BM→female CX3CR1 KO mice. BM cells were collected from the tibia and femurs of donor mice by aspiration and flushing. Recipient mice received 10 Gy total body irradiation. Then, while under anesthesia, the animals received a suspension of 5 × 10⁶ BM cells i.v. from donor mice in a volume of 200 μl of sterile PBS. For 2 wk after BM transplantation, the mice were housed in sterilized individual cages and were fed normal chow and provided water supplemented with neomycin. To verify successful engraftment, genomic DNA was isolated from peripheral blood and tail tissues of each chimeric mouse 4 and 10 wk after BM transplantation with a NucleoSpin tissue and blood kit (Macherey-Nagel). Then, PCR was performed to detect the Sry gene contained in the Y chromosome (forward primer, 5′-TGACACTGGCCTTTTCTCCT-3′; reverse primer, 5′-CATGGCATGCTGTATTGACC-3′). The amplified PCR products were fractionated on a 2% agarose gel and visualized by ethidium bromide staining. After durable BM engraftment was confirmed, mice were generated with skin wounds as described above.

Before transplantation, BM cells were labeled ex vivo with PKH26 (Sigma-Aldrich), a red fluorescent dye that emits at 567 nm and has a half-life of >100 days. PKH26 is a lipophilic dye that stains the membrane of viable cells and is distributed between cells when mitosis occurs. Labeling was performed according to the manufacturer’s protocol. After labeling, BM cells (5 × 10⁶) were transplanted immediately via tail vein to recipient mice as described above. PKH26-labeled cells were visualized in frozen sections of skin wounds using fluorescence microscopy and appropriate filters.

Means and SEs (SEMs) were calculated for all parameters determined in this study. Statistical significance was evaluated by using ANOVA or the Mann-Whitney U test. A value of p < 0.05 was accepted as statistically significant.

We profiled gene expression for 9 inflammatory chemokines (CXCL1, 2, 9, and 10, and CCL2, 3, 4, and 5, and CX3CL1) and 6 inflammatory chemokine receptors (CXCR2 and 3, CCR1, 2, and 5, and CX3CR1) in a mouse model of excisional skin wound healing. All 15 genes were induced in the model to varying extents and with different kinetics (Fig. 1). The results for CXCL1, 2, 9, and 10, CCL2, 3, and 5, CXCR2, and CCR1 confirm and extend previous reports (2, 3, 5, 6, 7, 8, 9, 10, 11, 29, 33) and the others are novel. In the present report, we focused on CX3CL1 and CX3CR1, because they were particularly highly induced and not previously characterized.

In WT C57BL/6 mice, CX3CR1 mRNA was weakly detectable in normal unwounded skin (Fig. 1). Wounding induced accumulation of receptor mRNA to high levels by day 3 after injury and this was sustained through the end of the 6-day period of observation. CX3CR1 protein was also expressed in wound sites, specifically on infiltrating F4/80⁺ macrophages, α-SMA⁺ fibroblasts, and CD31⁺ endothelial cells (Fig. 2, A–C). Like CX3CR1, CX3CL1 mRNA was also weakly detectable in normal unwounded skin but was increased earlier than the receptor in the wound site (Fig. 1). CX3CL1 mRNA expression increased with linear kinetics throughout the 6-day period of observation after wounding. There were no significant differences in CX3CL1 mRNA expression levels between WT and CX3CR1 KO mice at any time point (data not shown). Immunohistochemical analysis demonstrated CX3CL1 proteins in vascular endothelial cells and infiltrating mononuclear cells in the wound sites (Fig. 3,A). The infiltrating CX3CL1⁺ mononuclear cells were identified as mainly F4/80⁺ macrophages by double immunofluorescence (Fig. 3 B). Induction of both CX3CR1 and CX3CL1 expression in skin by wounding suggested the hypothesis that functional interaction of this ligand/receptor pair may regulate skin wound healing through an autocrine and/or paracrine mechanism.

To test this hypothesis, we first evaluated changes in wound area as a measure of wound closure in WT and CX3CR1 KO mice (Fig. 4). In WT mice, wound areas were reduced with linear kinetics throughout the 14-day period of observation, at the end of which ∼95% closure was observed. In contrast, wound closure was significantly delayed in CX3CR1 KO mice (time to 40% wound closure was 6 vs 10 days in WT vs CX3CR1 KO mice, respectively; p < 0.01). These observations indicated that germline inactivation of CX3CR1 results in delayed wound healing in C57BL/6 mice.

We next examined the effects of CX3CR1 deficiency on leukocyte infiltration into wounds. Because many neutrophils are trapped within the blood clot that is associated with healing wounds, it is difficult to accurately determine neutrophil number. Thus, we measured MPO activity to evaluate this parameter (Fig. 5). MPO activity reached a peak on day 1, declined thereafter, and was similar in magnitude in WT and CX3CR1 KO mice (Fig. 5,B). In contrast, F4/80⁺ macrophages, which were present in small numbers in normal skin, infiltrated the wound sites in large numbers with linear kinetics throughout the 6-day period of observation. Although the kinetics of macrophage infiltration were similar in WT and CX3CR1 KO mice, the magnitudes differed markedly: ∼40% lower in CX3CR1 KO mice at day 6 compared with WT mice (Fig. 5, C and D). Compared with neutrophils and macrophages, CD3⁺ T cells infiltrated the wound site in small numbers after injury; the kinetics and extent of accumulation were similar in CX3CR1 KO and WT mice (Fig. 5, E and F). Thus, the absence of CX3CR1 attenuated recruitment of macrophages, but not neutrophils or CD3⁺ T cells, into the wound sites.

Many CXC and CC chemokines and their receptors were up-regulated after wounding in the model, and typically mRNA expression reached a peak between days 1 and 3 as shown previously in Fig. 1. However, there was no significant difference in expression of any these molecules between WT and CX3CR1 KO mice at any time points (data not shown). These results suggest that CX3CR1 involvement in wound healing may be direct and is not mediated indirectly through effects on expression of other chemokines or chemokine receptors tested.

Circulating mouse monocytes consist of two functionally distinct subpopulations characterized by the chemokine receptor immunophenotypes CX3CR1^highCCR2⁻ and CX3CR1^lowCCR2⁺, which correspond to the CCR2⁻CD16⁺ and CCR2⁺CD16⁻ subsets of human monocytes, respectively (34, 35, 36, 37, 38). Therefore, we tested F4/80⁺ macrophages infiltrating skin wounds for expression of CX3CR1 and CCR2. We detected CX3CR1 protein on most F4/80⁺ macrophages, consistent with Fig. 2,A (Fig. 6, A, B, and D), whereas CCR2 was coexpressed on a minority of F4/80⁺ macrophages (Fig. 6, A, C, and E) at the wound sites 6 days after injury. We found that CX3CR1 expression was more abundant than CCR2 when signals of CX3CR1 and CCR2 were merged (Fig. 6 F). Therefore, expression of CX3CR1 was more prominent than CCR2 on infiltrating macrophages in skin wounds. These observations suggest that CX3CR1-dependent recruitment of macrophages might be an important mechanism for skin wound healing.

To test this, we analyzed the phenotypes of the following BM chimeric mice (donor BM→recipient) in our model: WT BM→WT, WT BM→CX3CR1 KO, CX3CR1 KO BM→CX3CR1 KO, and CX3CR1 KO BM→WT. BM-derived cells from donor mice were labeled ex vivo with PKH26 red dye, and were injected i.v. into lethally irradiated recipients. Skin wounding was performed 10 wk after BM transplantation, and frozen sections of wound sites were prepared and analyzed 6 days later, a time at which macrophages are the predominant BM-derived cell type present in the wound (Fig. 7). Skin sections from transplanted but uninjured control mice contained only a few PKH26-labeled cells in the dermis, whereas wounding caused labeled WT donor BM-derived cells to accumulate in similar large numbers in granulation tissue of both WT and CX3CR1 KO recipients. In contrast, accumulation of labeled cells at the wound site was markedly reduced to a similar extent (∼60%) in both CX3CR1 KO and WT recipient mice when CX3CR1 KO donor BM was transferred. Thus, CX3CR1 expression mediated accumulation of most, but not all, donor BM-derived cells at wound sites 6 days after wounding, suggesting a direct functional role for the receptor on macrophage accumulation in the wound.

Because CX3CR1 in the wound site is located on both macrophages, which originate in BM, and fibroblasts, we next attempted to distinguish the role in repair of receptor expression on each of these cell types. To do so, we analyzed wound closure and collagen deposition in BM-chimeric mice. As expected, both wound closure and hydroxyproline accumulation were significantly delayed over the 14-day period of observation in CX3CR1 KO recipient mice transplanted with CX3CR1 KO donor BM, as compared with the positive control (WT recipients of WT donor BM cells). However, hydroxyproline and wound healing kinetics were also equivalent to the positive control when either the donor BM or the recipient mouse, but not both, were CX3CR1 KO (Fig. 8). Thus, defective wound healing results when CX3CR1 is missing from both a BM-derived cell, probably a macrophage, and a non-BM-derived cell, possibly a fibroblast.

We next directly examined the role of CX3CR1 on collagen accumulation in wounded mice by histopathology. Granulation tissue formation was markedly reduced at the wound sites in CX3CR1 KO mice compared with WT mice at day 6 (Fig. 9,A). In uninjured skin (time = 0), there was no significant difference in collagen type I mRNA expression (Fig. 9, B and C) or hydroxyproline content (Fig. 9,D) between CX3CR1 KO mice and WT mice. Compared with this baseline, skin wounding induced collagen type I mRNA expression in WT mice throughout the 6-day time course of observation after injury, whereas induction was significantly reduced in CX3CR1 KO mice (Fig. 9, B and C). Accordingly, the normal increase in hydroxyproline content at wound sites was significantly decreased in CX3CR1 KO mice throughout the 14-day time course of observation (Fig. 9 D). Thus, CX3CR1 deficiency caused a reduction in collagen deposition and eventually delayed granulation tissue formation in the model.

Because TGF-β1 has been shown to control key steps in granulation tissue formation, such as fibroblast influx, myofibroblast induction, formation of extracellular matrix, and cell proliferation (2, 39, 40, 41), we examined TGF-β1 levels at wound sites as a function of CX3CR1. Normal wound-induced enhancement of TGF-β1 gene expression was significantly attenuated in CX3CR1 KO mice compared with WT mice at days 3 and 6 after injury (Fig. 10, A and B). This result was confirmed at the protein level in wound tissue lysates using an ELISA specific for active TGF-β1 (Fig. 10,C). Consistent with this, immunohistochemical analysis with the myofibroblast marker α-SMA 6 days after injury demonstrated that myofibroblasts were abundant at the wound edges of WT mice, but relatively deficient in the same location of wounds from CX3CR1 KO mice (Fig. 10 D). These results suggest that CX3CR1 activity in vivo promotes TGF-β1 production, which promotes accumulation of α-SMA⁺-activated fibroblasts, which are the direct source of collagen, a key constituent of granulation tissue in skin wound repair.

Angiogenesis, the growth and proliferation of new blood vessels, is another important aspect of the wound healing process, thus we next examined the effect of CX3CR1 deficiency on neovascularization at wound sites. Using immunohistochemical analysis with anti-CD31, no significant difference was observed in the vessel density of uninjured skin (time = 0) between WT and CX3CR1 KO mice (Fig. 11, A and B). Six days after injury, the vessel density within the wound bed was increased in both WT and CX3CR1 KO mice, but was significantly decreased in CX3CR1 KO mice compared with WT mice (Fig. 11, A and B). Consistent with this, VEGF mRNA (Fig. 11, C and D) and protein expression (Fig. 11 E) in the wound sites was significantly decreased in CX3CR1 KO mice compared with WT mice throughout the time course of observation. These data imply that the absence of CX3CR1 may interfere with angiogenesis in skin wound sites, at least partly by reducing VEGF expression.

To clarify the role of CX3CR1 during skin wound healing directly, WT mice were injected with neutralizing anti-CX3CR1 Abs or control IgG. The administration of neutralizing anti-CX3CR1 Abs retarded the wound closure rate in WT mice compared with controls (time to 40% wound closure was 6 vs 9 days in control IgG- vs anti-CX3CR1-injected WT mice, respectively) (Fig. 12, A and B). Moreover, CX3CR1 neutralization significantly diminished hydroxyproline content in the wound sites at days 6 and 10 after injury (Fig. 12 C). These observations indicate that as with CX3CR1 gene disruption, CX3CR1 neutralization impairs collagen deposition, wound closure, and subsequent wound healing.

In the present study, we have demonstrated genetically and by immunologic blockade that the chemokine receptor CX3CR1 regulates skin wound healing in a mouse model. Multiple reparative processes were affected by the receptor, including inflammation, fibrosis, neovascularization, and regeneration of parenchymal cells. The results are consistent with previous reports showing that the CX3CL1-CX3CR1 ligand-receptor pair plays an important role in the initiation and progression of inflammation (12, 13, 14), and is up-regulated in inflammatory diseases (20, 21, 22, 23, 26, 27, 42), including inflammatory diseases of the skin (15, 16, 19). Our data linking CX3CR1 to fibrosis in skin wound healing are consistent with a report of CX3CL1 and CX3CR1 expression in the skin and lung of patients with systemic sclerosis (16), and an in vivo study showing that CX3CR1 KO mice have reduced collagen accumulation and renal fibrosis in a model of ischemia-reperfusion injury (18).

CX3CL1 is the only known ligand for CX3CR1. Expressed as a transmembrane protein on the plasma membrane, CX3CL1 may be subsequently cleaved by proteases such as ADAM10, resulting in a soluble or shed agonist. The membrane form mediates cell-cell adhesion whereas the shed form mediates chemotactic and antiapoptotic responses. This complicates considerations of how CX3CR1 functions at the cellular level in biological contexts. Moreover, consistent with previous reports in other biological contexts (16, 18, 21, 22, 23, 43, 44), in wounded skin both CX3CL1 and CX3CR1 were detectable on macrophages and endothelial cells, and CX3CR1 was detectable on fibroblasts. In this regard, membrane-bound CX3CL1-CX3CR1 interactions could in principle directly mediate homotypic adhesion of macrophages, and/or directly mediate heterotypic adhesion between macrophage and fibroblast, macrophage and endothelial cell, and/or fibroblast and endothelial cell. In addition, shed CX3CL1 could in principle directly recruit/activate macrophages, fibroblasts, and endothelial cells, by acting at CX3CR1. In addition, there is the potential for indirect autocrine and paracrine effects initiated by CX3CR1-dependent activation of macrophages, fibroblasts, and/or endothelial cells.

Among these possibilities, our data provide strong evidence in vivo that CX3CR1 directly mediates macrophage accumulation at the wound site. First, both ligand and receptor were induced at wound sites by wounding, with CX3CL1 induction occurring on cells resident in the skin and CX3CR1 occurring on infiltrating macrophages; moreover, CX3CL1 induction slightly preceded CX3CR1. Second, CX3CR1 was expressed on virtually all macrophages at the wound site. Third, CX3CR1 deficiency markedly reduced macrophage accumulation in wounded skin. Fourth, mRNA levels of various other inflammatory chemokines and chemokine receptors were up-regulated after skin wounding, but were unaffected in CX3CR1-deficient mice. Fifth, labeled BM-derived cells from WT donor mice accumulated in large numbers in wound sites of both recipient WT and CX3CR1 KO mice on day 6 after injury, a time point when macrophages are the major leukocyte subset found in the wound, whereas accumulation was markedly reduced when the donor mouse was CX3CR1 KO. A similar role for CX3CR1 in mediating macrophage accumulation in diseased tissue has been reported in a mouse model of atherosclerosis, and CX3CR1 has been implicated directly in human atherosclerotic cardiovascular disease (26, 45, 46). However, heretofore the chemokine signals regulating macrophage accumulation in healing wounds had not been clearly delineated.

Our study opens up many new questions. The reconstituting BM-derived cell type(s) may be further defined by approaches such as transfer of sorted and labeled BM-derived cells or gene targeting of CX3CR1 in specific lineages. How CX3CR1 works in the model requires further definition: does it only promote monocyte/macrophage influx from the blood into the wound as the transfer experiments suggest, or does it also mediate macrophage retention in the wound? CX3CL1-CX3CR1 signaling has also been reported to maintain cell survival and inhibit Fas-dependent cell death of brain microglia in vitro (47). Perhaps it also regulates survival of dermal macrophages in wounded skin. The reconstitution experiments suggest that BM, which contains inflammatory cell progenitors, mesenchymal stem cells, and multipotent stem cells (48, 49, 50), could be used to promote healing of chronic wounds.

It is important to note that CX3CR1 deficiency did not completely abrogate macrophage accumulation in wounded skin, and that macrophages express other chemoattractant receptors which could act in a coordinated, sequential manner to achieve complete monocyte/macrophage recruitment into the wounds (3, 8, 51). This caveat is particularly important because the adoptive transfer experiments clearly showed that CX3CR1 deficiency is required on both BM-derived cells and non-BM-derived cells to delay wound healing in the model. A good candidate is the important macrophage chemoattractant CCL2. This chemokine acts through the receptor CCR2, which is expressed on a subset of CX3CR1⁺ macrophages in wounded skin. Moreover, it is able to induce CX3CR1 expression on monocytes (52). CCL2 is produced earlier than CX3CL1 in the model and, in functional tests, immunoneutralization of CCL2 reduced macrophage migration into wounded mouse skin (9). However, genetic disruption of CCL2 did not (7), possibly due to development of a compensatory mechanism when CCL2 function is abolished in the germline. In addition to recruiting monocytes from the blood to the wound site, CCR2 might also be required for monocytes to emigrate from the bone marrow to the peripheral circulation, as has been shown in a mouse model of listeriosis (53).

The relevant CX3CR1⁺ non-BM-derived cell type(s) important in wound healing most likely include resident cells such as dermal fibroblasts and endothelial cells, which we showed were both CX3CR1⁺ directly in wounded skin. CX3CL1-CX3CR1 interaction may contribute directly to the accumulation and function of these cell types within wounds. However, our data also implicate indirect paracrine mechanisms involving wound macrophages, which have been shown to initiate the proliferative phase of wound healing by releasing a number of growth factors, including VEGF and TGF-β, which are potent inducers of angiogenesis and fibrosis (2, 54, 55). Consistent with this, both TGF-β1 expression and collagen deposition were decreased at wound sites in CX3CR1 KO mice in association with decreased macrophage accumulation. After migration, fibroblasts may differentiate in a TGF-β1-dependent manner into myofibroblasts, which are specialized cells with high synthetic capacity for extracellular matrix components, including collagen, thereby causing fibrotic changes (39, 40, 41, 56, 57). Consistent with this, α-SMA⁺ myofibroblasts and granulation tissue were diminished at the wound sites in CX3CR1 KO mice.

During wound healing, chemokines exert their regulatory activity on angiogenesis directly or as a consequence of leukocyte infiltration and the induction of growth factor expression (7, 10, 11, 58, 59). Recent studies have implicated CX3CL1 directly in endothelial cell migration and tube formation in vitro and angiogenesis in vivo (43, 44). These observations imply that the CX3CL1-CX3CR1 system may function as an angiogenic mediator in skin wound healing. Consistent with this, our data showed directly by immunohistochemistry that neovascularization was markedly diminished at wound sites of CX3CR1 KO mice. Also, although CX3CL1 directly stimulates angiogenesis, the local microvasculature also depends on several other growth factors, including VEGF and TGF-β, which are produced by macrophages in wounds (60, 61, 62, 63, 64). TGF-β signaling is able to up-regulate VEGF expression in various cell types (65, 66, 67, 68). We also found significant decreases in VEGF gene expression at wound sites in CX3CR1 KO mice. Thus, CX3CL1-CX3CR1 signaling may stimulate angiogenesis directly and/or indirectly during wound healing.

In summary, we have shown that CX3CR1 is important in wound healing in a mouse model. The mechanism includes effects on both CX3CR1⁺ BM-derived cells, most likely monocytes/macrophages, and CX3CR1⁺ non-BM-derived cells, most likely fibroblasts and endothelial cells. We show that CX3CR1 is important for macrophage accumulation in the wound site and that angiogenic and profibrotic macrophage products are reduced in wounded skin in the absence of CX3CR1. This is associated with impaired myofibroblast accumulation, collagen deposition, fibrosis, angiogenesis, and granulation tissue formation. The results suggest that stimulating CX3CR1 in wounds may accelerate healing, and could be beneficial in the context of surgery, chronic ulcers, and other pathologic conditions.

We thank Dr. Toshikazu Kondo (Wakayama Medical University, Japan) for technical advice.

The authors have no financial conflict of interest.