Interleukin-4 Receptor α Signaling in Myeloid Cells Controls Collagen Fibril Assembly in Skin Repair

IL-4Rα Signaling in Macrophages Is Critical for Skin-Wound Healing

To investigate the role of IL-4 and IL-13 in skin repair, full-thickness-excision skin wounds were inflicted on the backs of control mice (Il4ra^fl/− or Il4ra^fl/fl) and mice with complete (Il4ra⁻/⁻) or myeloid-cell-restricted gene deletion (Il4ra^fl/−Lyz2-cre). Signaling for both cytokines is dependent upon the IL-4Rα receptor chain. To characterize the functional impact of IL-4 and IL-13 during the entire time course of the healing response, we performed wound-tissue analysis during the early (4 days post injury [dpi]), middle (7 dpi), and late phases (14 dpi) of repair.

First, we analyzed expression of IL-4Rα in wound tissues during the entire time course of the healing response. Double stainings of wound-tissue sections for IL-4Rα and F4/80 (macrophages), CD31 (vascular cells), or αSMA (myofibroblasts) identified macrophages and myofibroblasts as the major cell populations expressing IL-4Rα during skin repair (Figures 1A–1C).

To examine whether IL-4Rα was expressed in macrophages infiltrating the wound site, we sorted single cell suspensions of CD11b⁺F4/80⁺ cells by flow cytometry from wound tissue of control mice and subjected them to gene-expression analysis by qRT-PCR (Figure S1A). Blood monocytes recruited to the site of skin injury showed a dynamic increase in Il4ra expression, peaking with a 3.8 ± 0.6-fold increase over circulating blood monocytes during the middle stage of repair (Figure 1D). qRT-PCR analysis of macrophages isolated from wound tissue and T and B cells isolated from spleen tissue in Il4ra^fl/−Lyz2-cre mice confirmed efficient myeloid-specific gene deletion during the healing response (Figure 1D; Figure S1B). In addition, disrupted IL-4 and IL-13 responsiveness in macrophages from Il4ra^fl/−Lyz2-cre and Il4ra^−/− mice was confirmed with functional in vitro assays of macrophage polarization (Figures S1C and S1D).

Morphological analysis by H&E staining of wound-tissue sections during the diverse healing stages revealed major differences in the repair response in control mice versus in IL-4Rα-deficient mice. Whereas control mice developed a vascularized and cellular granulation tissue by 7 dpi, in mice with complete or myeloid-cell-restricted Il4ra gene deletion, the granulation tissue was highly hemorrhagic (4.3 ± 1.5-fold and 3.7 ± 1.5-fold increase of area covered by erythrocytes in Il4ra^fl/−Lyz2-cre and Il4ra^−/− mice, respectively, in comparison to area covered in controls), indicative of defective granulation-tissue formation and a failure of the repair response (Figures 1E and 1F). Hemorrhages largely resolved during the time course of repair (Figure S2). The disturbed healing response in complete and myeloid-cell-restricted IL-4Rα-deficient mice was also evident by a transient delay in wound epithelialization at 4 dpi in these mice when they were compared to controls (Figure 1F). The overall amount of granulation tissue was similar in wounds of control and IL-4Rα-deficient mice (Figure 1F).

Analysis of single cell suspensions of wounded tissues by flow cytometry revealed a reduction in the absolute and relative numbers of macrophages (CD11b⁺F4/80⁺ cells) present during the early and middle phases of repair in Il4ra^fl/−Lyz2-cre mice in comparison to those present in controls (Figure 1G). BrdU pulse labeling of wounded mice showed that in wounds of control mice, 6.9% ± 1.4% of macrophages proliferated during the early stage of repair and that this process is 27.7% ± 21.5% attenuated in mutant mice, suggesting that macrophage proliferation during skin repair is partially controlled by IL-4Rα signaling (Figure 1H; Figure S3A). The overall cell number was similar in wounds of control and mutant mice (Figures S3B and S3C). There were no signs of increased bacterial infection in mutant versus control wounds (Figures S3D–S3F).

IL-4Rα Signaling in Macrophages Controls Collagen Fibril Formation and Vascular Structures

To investigate the molecular basis for the hemorrhagic phenotype in IL-4Rα-deficient mice, we assessed the formation and structure of vascular networks within the granulation tissue by double staining for critical components of vascular integrity (Lindahl et al., 1997; Davis and Senger, 2005), including CD31, perivascular cells (Desmin), and several matrix components of the basement membrane (collagen IV, laminin 1 γ chain, laminin 4 α chain). Whereas the overall quantity of CD31 staining within the granulation tissue in wounds of control and IL-4Rα-deficient mice at 7 dpi was comparable (Figure S3G), 3D confocal imaging analysis revealed pronounced differences in the vascular morphology. Whereas, in control mice, the formation of endothelial tubes in network-like structures (vessels) was easily detectable throughout the entire granulation tissue covering up to 79.8% ± 5.8% of the CD31⁺ area, in IL-4Rα-deficient mice, CD31⁺ cells preferentially accumulated in clusters covering up to 54.4% ± 11.3% in Il4ra^fl/−Lyz2-cre and 44.8% ± 4.6% in Il4ra⁻/⁻ mice in the granulation tissue (Figure 2A; Movies S1, S2, and S3). In the gene-deficient mice, tube-like structures were only occasionally detectable toward the wound edge (Figure 2A). In addition, vascular alterations in mutant mice were paralleled by significantly reduced association of CD31⁺ cells with Desmin⁺ perivascular cells in mutant mice when they were compared to controls (Figure 2B). Of note, qRT-PCR analysis of wounded and unwounded skin in Il4ra^fl/−Lyz2-cre and control mice revealed a reduction of 51.3% ± 20.5% in Pdgfb expression in myeloid-cell-specific Il4ra mutants, a factor known to be critical for recruitment of perivascular cells that regulate vascular stability (Figure 2C) (Lindahl et al., 1997). Thus, the formation of functional vascular networks was perturbed in granulation tissue of Il4ra mutants.

The morphology of vascular alterations in mutant mice strongly resembled the perturbed vascular phenotype recently described in embryoid bodies generated from β1-integrin or integrin-linked kinase (ILK)-deficient endothelial cells, which reflects perturbed endothelial-cell-matrix interactions (Malan et al., 2010; Malan et al., 2013). Hence, to examine whether vascular alterations might be associated with structural matrix defects in mutant mice, we performed ultrastructural analysis of wound tissues. Transmission electron microscopy (TEM) analysis revealed profound differences between mouse strains in the morphology of collagen fibrils in wound tissue at 7 and 14 dpi. First, in mutant mice, collagen was less orderly and less densely packed than in control mice (Figure 2D). Second, there was great variability in the shape and size of fibrils in mutant mice; individual fibrils exhibited irregular outlines in cross-section, in contrast to the circular outlines in controls. In some instances, the cross-sectional profiles of collagen fibrils reached very large diameters up to 280–300 nm and showed irregular edges (Figure 2D, asterisk) that were entirely absent in control mice. Third, there was the coexistence of extremely thin collagen fibrils (Figure 2D, arrowheads) intermingled with large ones in mutant mice, whereas in control mice the average size was quite uniform. Quantification of collagen fibril diameters demonstrated that mutant mice exhibited a wider range, with profiles varying from 40–300 nm, whereas control mice contained profiles ranging from 40–200 nm. Collagen fibril morphology was similar in unwounded skin of control and mutant mice. Thus, the absence of IL-4Rα signaling in macrophages perturbed collagen fibril formation, which might hamper proper endothelial-cell-matrix interactions, leading to impaired vascular tube formation and, ultimately, hemorrhages. Finally, numbers of hemosiderin-laden macrophages (indicative of red blood cell [RBC] uptake) were comparable in wounds of control and Il4ra^fl/−Lyz2-cre mice, making it unlikely that the hemorrhagic phenotype was primarily caused by a RBC removal defect (Figure S3H).

Earlier studies have demonstrated that increased collagen I fibril diameter is related to an enhanced tensile strength in skin and repaired tissues (Bradshaw et al., 2003; Doillon et al., 1988). To examine whether altered collagen fibril architecture in IL-4Rα-deficient mice might also impact the biomechanical properties of the scar, we determined tensile strength at 16 dpi. Notably, significantly more force (26.4% ± 14.4%) was required to rupture scar tissue derived from mutants versus that from controls (Figure 2E).

IL-4Rα Signaling in Macrophages Directs Plod2 Expression and Collagen Cross-Links in Skin Wounds

On the basis of the disturbed collagen fibrillogenesis, we hypothesized that IL-4Rα signaling might control the structural and biochemical assembly of the collagen matrix. We thus examined the expression of collagen-modifying enzymes implicated in this process. Lysyl hydroxylases (LHs, encoded by Plod genes) are essential for hydroxylating lysine residues, a process that directs the nature of cross-links of collagen fibers that ultimately define biochemical and structural stability of collagenous tissues (Myllyharju and Kivirikko, 2004). To examine the expression of diverse LH isoforms in skin-wound healing, we subjected wounded and unwounded skin of Il4ra^fl/−Lyz2-cre and control mice to qRT-PCR analysis. Whereas wounding of control and mutant mice did not lead to significant induction of LH1 (Plod1) and LH3 (Plod3) expression, LH2 (long) (Plod2) expression in control mice peaked with a 215 ± 115-fold induction during the middle stage of repair (Figure 3A). Notably, Plod2 expression in wounds of Il4ra^fl/−Lyz2-cre mice was 76.1% ± 11.5% and 47.7% ± 11.7% reduced at 7 and 14 dpi, respectively, in comparison to expression in control mice (Figure 3A). Attenuated Plod2 expression in wounds of Il4ra^fl/−Lyz2-cre mice was confirmed by weak immunohisto-chemical staining for LH2 in granulation tissue of mutant mice in comparison to a strong staining signal in control mice (Figures 3B and 3C).

LH2 is the only enzyme known that hydroxylates lysine residues specifically in the telopeptides of collagen polypeptides, a process that is essential to direct dihydroxy lysinonorleucine (DHLNL) cross-links of collagen fibrils (van der Slot et al., 2003; Pornprasertsuk et al., 2004; Pornprasertsuk et al., 2005). DHLNL collagen cross-links are a typical feature of mechanically stiff tissues such as bone and cartilage and are virtually absent in human soft connective tissues such as normal skin, which is characterized by lysine-aldehyde-derived collagen cross-links (hydroxylysinonorleucine, HLNL) (Myllyharju and Kivirikko, 2004). In human skin, identification of DHLNL collagen cross-links has been limited to pathological conditions such as lipodermato-sclerosis (skin fibrosis associated with chronic venous insufficiency) or scleroderma (Brinckmann et al., 1999; van der Slot et al., 2004). Here, we have shown that in unwounded and wounded skin of control and mutant mice, the concentration of HLNL per collagen was similar and not altered during the repair response (Figure 3D). In contrast, DHLNL collagen cross-links were detected at low amounts in unwounded skin of control mice and increased 1.9 ± 0.3-fold and 3.1 ± 0.5-fold in granulation tissue 5 and 10 dpi (Figure 3D), respectively. In mutant mice, DHLNL cross-links were also detectable in unwounded skin, in comparable amounts to those detected in controls (Figure 3D). However, DHLNL collagen cross-links were significantly attenuated in wound tissue of mutant mice when compared to that in controls (Figure 3D). Collectively, we found that IL-4 and IL-13 responsiveness in macrophages controlled Plod2 expression in skin wounds and determined the nature of collagen cross-links.

Attenuated Alternative Macrophage Activation in Late-Stage Skin-Wound Healing in Il4ra^fl/−Lyz2-cre Mice

To investigate whether impaired granulation tissue formation and defective collagen fibrillogenesis in Il4ra^fl/−Lyz2-cre mice is associated with altered macrophage activation during healing progression, CD11b⁺F4/80⁺ cells from wound tissue were sorted by flow cytometry and subjected to qRT-PCR analysis. Gene expression of wound macrophages was correlated with gene expression in circulating blood monocytes (SSC^lowCD11b⁺) from unwounded mice and sorted by flow cytometry. Consistent with earlier reports from our group and others (Lucas et al., 2010; Willenborg et al., 2012), early and late-stage wound macrophages are characterized by a distinct gene-expression profile reflecting classical and alternative activation during the early and late phase of repair, respectively (Figure 4A). In control mice, expression of genes characterizing alternative activation, including Cd163, Retlna, Cd206, and Il10, was low during the early stage of repair (4 dpi) and dynamically upregulated toward 14 dpi (Figure 4A). Furthermore, early stage wound macrophages were characterized by a 27.7 ± 18.1-fold induction of Nos2 expression (characterizing classical activation) that declined over the time course of healing (Figure 4A). In contrast, in Il4ra^fl/−Lyz2-cre mice, alternative polarization of late-stage wound macrophages was significantly attenuated in comparison to polarization in control mice, whereas in early stage wound macrophages, Nos2 expression was significantly upregulated and prolonged in comparison to expression in controls (4 dpi, 5.1 ± 2.2-fold increase and 7 dpi, 8.1 ± 4.7-fold increase, respectively, over controls). Of note, other genes characterizing early stage classical activation, such as Il1b, Tnf, and Il6, were not differentially expressed in control or mutant mice (Figure S4).

Relm-α Derived from Middle- and Late-Stage-Activated Wound Macrophages Directs Plod2 Expression in Fibroblasts

To explore the hypothesis that defective collagen fibril cross-linking in Il4ra^fl/−Lyz2-cre mice is mediated by an altered cross-talk between macrophages and fibroblasts, we used an in vitro co-culture system, in which the cell types have no direct contact. Plod2 expression was assessed by qRT-PCR in skin fibroblasts after exposure to classically (IFN-γ and LPS) or alternatively (IL-4 and IL-13) activated macrophages. Plod2 expression in fibroblasts (isolated from Il4ra^−/− mice) was 1.4 ± 0.1-fold upregulated upon exposure to IL-4- and IL-13-activated macrophages in comparison to expression in fibroblasts not exposed to macrophages or exposed to macrophages cultured in vehicle or activated with IFN-γ and LPS (Figure 4B). Consistently, Plod2 expression in fibroblasts was not detected when co-cultured with IL-4- and IL-13-treated macrophages isolated from IL-4Rα-deficient mice (Figure 4B).

We next examined candidate macrophage-derived signals that might induce Plod2 expression in fibroblasts. Although in vitro studies have reported that TGF-β1 can induce Plod2 expression in fibroblasts (van der Slot et al., 2005a), Tgfb1 expression was not upregulated in IL-4- and IL-13-activated macrophages in our model (data not shown). In contrast, Retnla, the gene encoding a small cysteine-rich secreted molecule that is a hallmark of alternatively activated macrophages (Gordon, 2003) and has been associated with experimental fibrosis (Liu et al., 2004) and pro-fibrotic conditions in human diseases (Angelini et al., 2009; Kushiyama et al., 2013; Fang et al., 2015), was continuously upregulated in wound macrophages during the time course of repair (Figure 4A). Therefore, we tested whether Relm-α released by IL-4- and IL-13-activated macrophages controlled Plod2 expression in fibroblasts. We first showed that Retlna could be effectively silenced by siRNA in macrophages, stimulated with IL-4 and IL-13, and co-cultured with fibroblasts (Figure 4C). Notably, IL-4- and IL-13-stimulated macrophages transfected with Relm-α siRNA failed to induce Plod2 expression in fibroblasts, whereas control macrophages transfected with scrambled RNA induced expression of Plod2 in fibroblasts (Figure 4C). To corroborate the finding that Relm-α induced Plod2 expression in fibroblasts, we exposed fibroblasts to recombinant murine Relm-α (rRelm-α). Stimulation of fibroblasts with rRelm-α induced Plod2 expression up to 1.6 ± 0.2-fold in comparison to expression in the control (Figure 4D).

Local Application of rRelm-α Rescues Wound-Healing Pathology and Attenuated Plod2 Expression in Il4ra^fl/−Lyz2-cre Mice

We next investigated whether myeloid-cell-restricted IL-4Rα signaling is critical for Relm-α-mediated orchestration of granulation tissue formation and collagen fibril assembly. For this purpose, wound tissue of Il4ra^fl/−Lyz2-cre and control mice was subjected to qRT-PCR analysis to quantify Retnla expression. Whereas in control mice Retnla expression was upregulated during the time course of healing, the expression was significantly reduced in wounds of Il4ra^fl/−Lyz2-cre mice at 7 and 14 dpi, respectively (Figure 5A). Consistent with reduced Retnla expression in purified wound macrophages in Il4ra^fl/−Lyz2 mice (Figure 4A), immunohistochemical staining of wound tissues showed a reduced number of Relm-α⁺F4/80⁺ double-positive cells within the granulation tissue of Il4ra^fl/−Lyz2-cre mice in comparison to that in control mice (Figures 5B and 5C).

To address whether local application of rRelm-α could rescue impaired granulation tissue formation in Il4ra^fl/−Lyz2-cre mice, we injected the edges of wounds in Il4ra^fl/−Lyz2-cre mice twice (3 and 5 dpi) with rRelm-α (1 μg/wound) or vehicle, and the repair response was assessed. H&E staining of wound-tissue sections at 7 dpi revealed that rRelm-α treatment normalized hemorrhages within the granulation tissue of Il4ra^fl/−Lyz2-cre mice (Figures 1E and 1F, Figures 5D and 5E). The number of wound macrophages (CD68⁺ cells) was comparable in granulation tissue in vehicle- or rRelm-α-injected Il4ra^fl/−Lyz2-cre mice, suggesting that the resolution of the hemorrhages is not due to increased macrophage numbers (Figure 5F).

We then tested whether the normalized wound phenotype in rRelm-α-treated wounds in Il4ra^fl/−Lyz2-cre mice was associated with increased local Plod2 expression. qRT-PCR analysis of complete wound tissue revealed that, indeed, Plod2 expression was 9.7 ± 3.7-fold upregulated in comparison to expression in vehicle-treated wounds in Il4ra^fl/−Lyz2-cre mice. In fact, Plod2 expression was restored to an amount similar to that in Il4ra^fl/− mice at 7 dpi (Figure 3A, Figure 5G).

The Disturbed Healing Phenotype Is Similar in Il4ra^fl/−Lyz2-cre and Retnla-Deficient Mice

To substantiate a direct role of Relm-α in skin repair, we characterized the wound-healing response in Retnla^−/− mice. Notably, we detected intriguing parallels regarding morphological, structural, and biochemical alterations of the wound-healing response in Retnla^−/− and Il4ra^fl/−Lyz2-cre mice (Figure 6). As such, the repair response in Retnla^−/− mice was characterized by a highly hemorrhagic granulation tissue (Figures 6A and 6B), transient delay in wound epithelialization (Figure 6C), altered assembly and size of collagen fibrils (Figure 6D), reduced DHLNL collagen cross-links (Figure 6E), 39.0% ± 30.7% reduced expression of Plod2 (Figure 6F) and 22.4% ± 18.8% increased tensile strength (Figure 6G). Expression of inflammatory mediators including Il6, Tnf, and Il1b, numbers of inflammatory cells (F4/80⁺ or Gr1⁺), and hemosiderin laden macrophages were comparable in wound tissue in control and Retnla^−/− mice (Figure S5). Collectively, these findings demonstrate a direct role of Relm-α in rapid and effective restoration of tissue integrity and support our hypothesis that impaired wound healing in Il4ra^fl/−Lyz2-cre mice is at least in part mediated by a reduction of Relm-α.

To investigate whether initiation of pro-fibrotic pathways by Relm-α might also apply to humans, we examined the effect of rRelm-β (1 μg/mL, 24 hr), the human homolog of murine Relm-α (Holcomb et al., 2000), on gene expression in human dermal fibroblasts. rRelm-β induced a 30.0% ± 18.5% increased expression of PLOD2 and 22.6% ± 15.1% increased expression of ACTA2 (α-smooth muscle actin) (Figure 6H). In addition, we quantified expression of RETNLB and of PLOD2 in tissue samples of lipodermatosclerosis (excessive skin fibrosis associated with chronic venous insufficiency) and healthy human skin by qRT-PCR analysis (Figure 6I). Indeed, expression of both genes was 2.8 ± 2.2-fold and 4.2 ± 1.9-fold upregulated in fibrotic skin in comparison to expression in normal skin, suggesting a potential relevance in the pathology of this human fibrotic skin condition.

Animals

Myeloid-cell-specific IL-4Rα-deficient mice (Il4ra^fl/−Lyz2-cre) and Retnla^−/− mice (BALB/c background, 10–12 weeks) were generated and genotyped as previously described (Figure S6 and Supplemental References). All procedures were approved by the North Rhine-Westphalian State Agency for Nature, Environment, and Consumer Protection and the University of Cologne.

Cell Culture

Primary dermal fibroblasts were isolated from the back skin of mice and from the skin of two different human donors. Macrophages were isolated from the peritoneal cavity by plastic adhesion or from the bone marrow (BMDM). Stimulation of cells and the co-culture system are described in the Supplemental Experimental Procedures.

Relm-α Silencing by siRNA

BMDM were transfected via the Nucleofector Kit (Lonza) with either Relm-α targeted siRNA or scrambled siRNA (Thermo Scientific Dharmacon). Details are described in the Supplemental Experimental Procedures.

Wounding

Wounding, preparation of wound tissue for histology, and morphometric analysis was performed as described previously (Willenborg et al., 2012). For local treatment of wounds with rRelm-α, wound edges were injected either with rRelm-α (1 μg in 50 μL PBS) or vehicle (50 μL PBS) at 3, 4, and 5 dpi.

Lipodermatosclerosis

Human tissue samples were obtained from patients presenting with lipodermatosclerosis. After collection, tissues were snap-frozen in liquid nitrogen or stored in RNAlater (Life Technologies). RNA was isolated with the Fibrous Tissue Mini Kit (QIAGEN). The study was approved by the Ethics Committee of the Medical Faculty of the University of Cologne and informed consent of patients was received.

Immunohistochemistry

Immunohistochemical stainings (10 μm and 40 μm cryosections) were performed as described previously (Willenborg et al., 2012). Confocal microscopy for 3D imaging of the endothelial tubes and clusters was performed as described previously (Malan et al., 2013) by performing z-stacks and subsequent use of 3D projection function of the LSM 510 Meta (Zeiss). Antibodies and analysis of stained sections, Gram staining, and iron staining are described in the Supplemental Experimental Procedures.

Electron Microscopy

Tissue was fixed and processed according to a standard protocol. After tissue was embedded in Epon-Araldite, ultrathin sections (50–100 nm) were analyzed via a Zeiss 902A Transmission Electron Microscope (Zeiss) and TEM Imaging Platform iTEM Software (Soft Imaging Systems).

Flow Cytometry

Flow cytometry and cell sorting were performed as described previously (Willenborg et al., 2012). FACS analysis and cell sorting was performed with the BD FACS Canto II flow cytometer, the BD FACS Aria III cell sorter, and the BD FACS Diva Software. Antibodies used are listed in the Supplemental Experimental Procedures.

Real-time PCR Analysis

RNA from complete wound tissue was isolated with the Fibrous Tissue Mini Kit, and RNA from single cell suspensions was isolated with the RNeasy Mini or Micro Kits (QIAGEN). Reverse transcription and qRT-PCR analysis are described in the Supplemental Experimental Procedures.

Collagen Cross-Link Analysis

Collagen cross-link analysis was performed as previously described (Brinckmann et al., 2005). Tissue specimens (about 20 mg wet weight) were reduced, denatured, and digested. Analysis was performed with an amino acid analyzer (Biochrom 20). Details are described in the Supplemental Experimental Procedures.

Recombinant Relm-α

Recombinant murine rRelm-α was expressed in HEK293-EBNA cells and purified from cell culture supernatant. The identity of the purified recombinant protein was confirmed by western blot with anti Relm-α (Peprotech). Details are described in the Supplemental Experimental Procedures.

Tensile Strength Test

Incisional wounds (length: 2 cm) were placed on the backs of mutant mice. At 16 dpi, two identical rectangle-shaped (1 cm × 2.5 cm) pieces of the back skin with scar tissue located in the center were excised and stored in isotonic NaCl solution at −20°C. Thawed specimens underwent a tensile strength test via a material testing machine (Z2.5/TN1S, Zwick). After preloading (0.05 N, 0.1 mm/s), the scar tissue was stretched until failure with a cross-head speed of 15 mm/min.

Statistical Analysis

Statistical analysis was performed with the Student’s paired or unpaired two-tailed t test or ANOVA one-way test analysis with Dunnett’s Multiple Comparison Test. In case of not assumed equal variances, a Welch correction was performed. Data are presented as the mean and, where indicated, with ± SD, and significance levels are indicated in each figure.