An unrestrained proinflammatory M1 macrophage population induced by iron impairs wound healing in humans and mice (original) (raw)

Activated macrophages accumulate and persist in CVUs. Normal wound healing follows a sequence of events involving clotting, inflammation, matrix deposition, and remodeling (4, 37). Within a few hours after injury, first polymorphonuclear neutrophils and later macrophages invade the wound site. Both neutrophils and macrophages produce proteinases and ROS to combat contaminating microorganisms and phagocytose cellular debris. In CVUs, this inflammatory phase is persistent (9, 10). Whether macrophages play a role in maintaining this persistent inflammation is not known.

Biopsies derived from CVU patients presented with high numbers of CD68+ macrophages, representing up to 80% of the cells in the wound margins (Figure 1A and Supplemental Figure 1; supplemental material available online with this article; doi:10.1172/JCI44490DS1). These macrophages showed an increase in classical proinflammatory M1 markers iNOS and TNF-α in all studied biopsies derived from CVU patients (patients 1–5; Figure 1, B and C, and Supplemental Table 1). In contrast, only a transient increase in macrophage numbers expressing both iNOS and TNF-α were observed in AWs at day 2 after wounding, with a subsequent reduction in macrophage numbers to those of normal skin (NS) by day 5 (Figure 1, A–C). iNOS generates microbicidal NO• to combat invading microbes. TNF-α stimulates iNOS synthesis and thus enhances microbicidal defense. However, persistent overproduction of TNF-α, as observed here in CVUs, may be detrimental for tissue repair. In fact, injection of recombinant TNF-α (rTNF-α) in the wound margins of AWs in a mouse model of full-thickness excisional wounds significantly delayed wound healing (Supplemental Figure 2).

Activated macrophages accumulate and persist in CVUs.Figure 1

Activated macrophages accumulate and persist in CVUs. (A) Quantification of CD68+ infiltrating macrophages in NS, AWs at days 1, 2, and 5 after wounding, and CVUs were counted in 10 high-power fields per sample. Results are mean ± SD ratio of CD68+ to total cells counted in the dermis (n = 5). **P < 0.01 versus day-5 AW and NS. (B and C) Representative photomicrographs of skin sections. Activation of macrophages was assessed in NS, AWs, and CVUs by immunostaining of cryosections for the classical activation markers (B) iNOS and (C) TNF-α or the nonclassical activation marker CD163. Nuclei were counterstained with DAPI. Scale bars: 100 μm. Dashed lines indicate the junction between epidermis (e) and dermis. es, eschar; wm, wound margin.

Macrophages mount a distinct activation phenotype in CVUs. We hypothesized that macrophages from CVUs fail to switch from proinflammatory M1 macrophages to the antiinflammatory M2 macrophages required for tissue remodeling and restoration. We studied proinflammatory M1 activation markers, including cytokines (TNF-α and IL-12/IL-23), chemokine receptors involved in monocyte recruitment and activation (CCR2 and CX3CR1), and iNOS, which is involved in phagocytosis and bacterial killing (32, 38). In addition, we included antiinflammatory M2 markers like arginase, which exerts its antiinflammatory effect via degradation of arginine required for iNOS activity, and M2 nonopsonic scavenger (CD36), β-glucan (Dectin-1), and mannose receptors (CD206) (35, 39). We performed double immunostaining for combinations of M1 and M2 activation markers in AWs compared with CVUs using immunohistology and flow cytometric analysis of macrophages from enzymatically digested tissue samples (Figure 2, A–C). In selected experiments, macrophages gated on CD68+ were subsequently regated for M1 and M2 markers in order to demonstrate that M1 and M2 markers are expressed by the same macrophages. An unexpected transient concomitant expression of both proinflammatory M1 (TNF-α, IL-12, CCR2, and CX3CR1) and antiinflammatory M2 markers (CD206, arginase, Dectin-1, IL-10, IL-4Rα, CD36, and CD163) by almost all macrophages was observed in early phases of AWs at day 2 (Figure 2 and data not shown). Virtually all proinflammatory M1 markers were downregulated to basal levels in NS macrophages, and high expression of antiinflammatory M2 markers prevailed in macrophages of AWs at day 5 after wounding (Figure 2). Conversely, persistent coexpression of high levels of proinflammatory M1 markers (TNF-α, IL-12p40, and CCR2) and — apart from high expression of CD163 — intermediate expression levels of M2 markers (arginase, CD206, Dectin-1, IL-10, IL-4Rα, and CD36) was observed in all macrophages in the studied tissue samples from CVU patients (Figure 2, B and C).

The identified macrophage population mounts an unrestrained proinflammatoryFigure 2

The identified macrophage population mounts an unrestrained proinflammatory M1 activation phenotype and accumulates iron in CVUs. (A) Representative photomicrographs with double immunostaining of skin cryosections from AWs and CVUs for M1 and M2 macrophage activation markers TNF-α and CD206 or IL-12 and arginase-1. Nuclei were stained with DAPI. Scale bars: 100 μm. Dashed lines indicate the junction between eschar and wound margin. (B) Flow cytometry analysis of wound macrophages purified from AW tissue 2 and 5 days after wounding and CVUs gated according to side scatter (SSC) and CD68 and regated for CD68 and M1 marker TNF-α, CD68 and M2 marker Dectin-1, or TNF-α and Dectin-1. (C) Expression levels for M1 and M2 activation markers of macrophages isolated from 5 AWs and 6 CVUs by flow cytometry. Results are given in RFU (see Methods). Resident skin macrophages were pooled from NS (n ≥ 5). *P < 0.05, **P < 0.01, Student’s t test. (D) Representative photomicrographs of cryosections from CVU patients and AWs from healthy volunteers stained for iron by Perl Prussian blue and immunostained with CD163 for macrophages. Nuclei were stained with PI. High amounts of iron were identified within the CD163+ macrophages (filled arrows) and extracellular space (open arrows) in CVUs, but not AWs. Scale bars: 150 μm.

These data suggested that a previously undescribed macrophage population in CVUs fails to fully switch from a proinflammatory M1 state to an antiinflammatory M2 program to efficiently promote angiogenesis, connective tissue deposition, and wound repair. These macrophages, with an unrestrained proinflammatory M1 activation state along with highly expressed M2 iron scavenger receptor CD163, are referred to herein as macrophages with unrestrained proinflammatory M1 phenotype.

Unrestrained proinflammatory M1 macrophages accumulate iron in CVUs. Iron deposits in CVUs cause visible brownish skin that always surrounds chronic leg ulcers. The origin of increased iron stores in the skin around CVUs is extravasation of red blood cells in conditions of significantly increased blood pressure and venous stasis due to venous valve insufficiency. The observed enhanced expression of the hemoglobin-haptoglobin scavenger receptor CD163, an important M2 marker, is a consequence of red blood cell extravasation and precedes the engulfment and degradation by tissue macrophages (40). In addition, CD163 expressed on macrophages serves as scavenger receptor for hemoglobin-haptoglobin–bound iron, further increasing intracellular iron stores. Consistent with previous data on tissue iron accumulation in CVUs (18, 19, 25), we found accumulation of ferric iron in macrophages and in the extracellular space in all studied CVU samples (patients 1–5 and 12–16; Figure 2D and Supplemental Table 1). Conversely, iron deposits were absent in AWs. These data indicate that the CD163+ macrophages with unrestrained proinflammatory M1 activation colocalize with iron in CVUs.

Iron causes the induction of the unrestrained proinflammatory M1 macrophage population. In order to assess whether iron can induce the unrestrained proinflammatory activation of the newly defined macrophage population, we established a murine model closely reflecting main pathogenic aspects of CVUs. Repeated treatment of mice with iron-dextran resulted in an abundant deposition of iron in cells within the dermis of the skin (Figure 3A). After wounding, iron accumulated inside the F4/80+ macrophages. Iron loading of macrophages was almost completely prevented when treatment with iron-dextran was followed by injection of the iron chelator desferrioxamine (DFX; Figure 3A). Similar to human macrophages in CVUs, murine macrophages concomitantly expressed both M1 (TNF-α and IL-12) and M2 (CD206 and arginase) activation markers in wound margins from iron-dextran–treated mice in the early (day 2) and late (day 5) inflammatory phases of wound healing (Supplemental Figure 3A). Notably, macrophages isolated from wound margins of iron-dextran–treated mice revealed an activation pattern reminiscent of macrophages isolated from CVUs with a persistent proinflammatory M1 response and intermediate antiinflammatory M2 marker activation (i.e., TNF-αhi, IL-12hi, CCR2hi, Ly6Chi, Dectin-1med, IL-4Rαmed, and CD204med) at day 5 after wounding compared with PBS-dextran–treated control mice (Figure 3B and Supplemental Figure 3B). Even though the nonopsonic scavenger receptors CD206 and CD301 were expressed at comparable levels in iron-dextran– and PBS-dextran–treated control mice and the M2 marker CD163 was highly expressed, the overall antiinflammatory M2 response may not be sufficient to terminate inflammation and promote tissue repair (Figure 3B). Consistent with the phenotype observed in vivo in CVUs and in the murine model, human macrophages mounted an identical activation profile, with dominance of proinflammatory markers when cultured under Fenton reaction–mimicking conditions in the presence of Fe(III) chloride/ascorbate and H2O2 in vitro (Supplemental Figure 4).

Iron is causal for induction of the macrophage population with unrestrainedFigure 3

Iron is causal for induction of the macrophage population with unrestrained proinflammatory M1 phenotype and for impaired wound healing. (A) Representative photomicrographs detects Fe deposition in NS and in F4/80+ macrophages in iron-dextran–treated wounds. Iron deposition was almost completely prevented by coinjection with DFX. Cell nuclei were counterstained with PI. Scale bars: 150 μm. Dashed lines indicate the junction between epidermis and dermis (d). h, hair follicle. (B) Flow cytometry analysis of macrophages isolated from mouse wounds gated for side scatter and F4/80. Expression of M1 and M2 activation markers is shown in RFU (see Methods). Resident skin macrophages were pooled from NS (n = 5). *P < 0.05, **P < 0.01, ***P < 0.001, Student’s t test. (C) Representative macroscopic aspects of wounds at 0, 5, 7, and 10 days after wounding. (D) Statistical analysis of 20 wound areas per group, expressed as percentage of the initial wound size (day 0), for iron-dextran–treated mice (filled symbols) in the presence and absence of DFX or etanercept. Open symbols denote PBS-dextran treatment. Results are mean ± SD (n = 5) representing 1 of 3 independent experiments. *P < 0.05, Mann-Whitney test. (E) Representative photomicrographs of mouse wounds at day 5 after wounding stained for F4/80 and TNF-α, with persistent TNF-α–producing macrophages (yellow), in iron-loaded, but not PBS-dextran– or etanercept-treated, wounds. Nuclei were stained with DAPI. Scale bars: 100 μm. Dashed lines indicate the junction between eschar and wound margin. (F) Statistical analysis of 20 wound areas per group for iron-loaded and PBS-dextran control mice treated with clodronate or PBS liposomes. *P < 0.05, **P < 0.01, Mann-Whitney test.

TNF-α from iron-induced unrestrained proinflammatory M1 activated macrophages is responsible for impaired wound healing. Compared with PBS-dextran–treated mice, iron-loaded mice showed delayed wound healing at days 3, 5, 7, and 10 after wounding (Supplemental Figure 5A). Interestingly, whereas in PBS-treated wounds, macrophages reached a maximal number at day 2 after wounding and continuously decreased thereafter, macrophages accumulated and persisted at significantly higher numbers in iron-overloaded mice compared with control mice at days 3, 4, and 5 after wounding. These data suggest that increased numbers of activated, TNF-α–producing macrophages are most likely responsible for the observed healing defect (Supplemental Figure 5, A and B).

The wound healing phenotype in iron-loaded mice was fully rescued in the presence of the iron chelator DFX (Figure 3, C and D), demonstrating the causal role of iron in delayed wound healing. TNF-α has previously been shown to maintain the activation state of macrophages in an autocrine manner (41, 42). This is in line with our observation that F4/80+ macrophages with unrestrained proinflammatory M1 phenotype highly expressed TNF-α in iron-loaded compared with control mice, and that intracellular TNF-α was reduced in macrophages of iron-loaded mice in the presence of the TNF-α antagonist etanercept (Figure 3E). As TNF-α has previously been shown to maintain the activation state of macrophages in an autocrine manner, neutralization of soluble TNF-α by injection of etanercept most likely resulted in the disruption of this autocrine feedback loop and subsequently in reduced TNF-α synthesis by macrophages and dampened inflammation. Consistent with a causal role of TNF-α in delayed wound healing, etanercept injection at day 3 after wounding rescued impaired wound healing in iron-loaded mice at days 5, 7, and 10 after wounding (Figure 3, C and D), whereas injection of rTNF-α into the margins of AWs in control mice resulted in significant delay of wound healing (Supplemental Figure 2). Given that etanercept injection was performed at day 3, a rescue of the delay in wound healing of day-3 wounds (Figure 3D) cannot be expected.

Virtually all F4/80+ macrophages of enzymatically digested wounds from iron-dextran–treated mice revealed the unrestrained proinflammatory M1 activation pattern. To determine whether the iron-activated proinflammatory TNF-α–releasing macrophages are responsible for delayed wound healing, we depleted the macrophages from wound margins by injections of liposomes encapsulated by dichloromethylene diphosphonate (clodronate) at day 4 after wounding as previously described (41). As expected, 24 hours after injection of clodronate liposomes, macrophages were almost completely depleted (Supplemental Figure 6A), whereas numbers of mature dendritic cells, Langerhans cells, neutrophils, lymphocytes, and mast cells were not affected (Supplemental Figure 6, B–G and refs. 41, 43). Injection of clodronate liposomes, but not control PBS liposomes, in the late inflammatory phase at day 4 after wounding resulted in a significant improvement of impaired wound healing of iron-loaded mice and fully rescued the delayed tissue repair at day 10 after wounding, whereas wound healing of PBS-dextran–treated mice injected with clodronate liposomes was not affected (Figure 3F). These data indicate that macrophages with unrestrained proinflammatory M1 phenotype are important in iron-induced delayed wound healing and imply that TNF-α release from M1 proinflammatory macrophages cause the iron-dependent impaired wound healing phenotype.

Iron-activated macrophages release toxic amounts of OH• and peroxynitrite in vivo. Iron-loaded macrophages, while releasing H2O2, may drive the Fenton reaction and thus cause the generation of highly toxic OH• (9). Also, iron-activated macrophages release O2–• and NO•, which can generate peroxynitrite (ONOO•) (44). To determine whether production of noxious OH• and ONOO• by iron-activated macrophages occurs in vivo, we applied the redox-sensitive dye dihydrorhodamine123 (DHR) on cryosections of NS, CVUs, and skin from PBS-dextran–treated and iron-loaded mouse wounds in the presence and absence of scavengers for distinct ROS. The wound margins of human AWs and CVUs presented with intense green fluorescence (Figure 4A), indicative of a strong oxidative burst. O2–• and H2O2 were primarily responsible for the oxidative burst in AW, as coincubation of the sections with SOD, which detoxifies O2–•, and with the H2O2 scavenger catalase (Cat) clearly quenched the fluorescence intensity. DMSO, the scavenger for OH•, did not quench the signal. Notably, O2–• and OH•, but not H2O2, were the prevailing ROS in all studied biopsies from CVU patients (Figure 4A). Identical results were found in wounds of iron-loaded mice compared with those of control mice (Supplemental Figure 7A). These differences in distinct ROS and concentrations may contribute to tissue damage and the nonhealing state of macrophage iron-overload conditions. In fact, immunostaining of sections with antibodies against 8-oxo-2′-deoxyguanosine (8OHdG), a marker for oxidative DNA damage, identified high numbers of cells with oxidative DNA damage in CVUs, but not in AWs or NS (Figure 4B). Similarly, an antibody against 3-nitrotyrosine (3-NT), a selective marker for protein nitration due to ONOO•, detected enhanced protein nitration in CVUs, but not in AWs or NS (Figure 4C). These data were confirmed in skin lysates from iron-loaded mice compared with PBS-treated mice and CVUs (Figure 4D and Supplemental Figure 8). Reduction of proinflammatory TNF-α released by macrophages with etanercept, or depletion of macrophages with clodronate liposomes, prevented generation of toxic OH• and ONOO• in iron-treated mice (Figure 4D and Supplemental Figure 7B). These data indicate a critical role for unrestrained proinflammatory M1 activated macrophages and enhanced TNF-α release for generation of toxic radicals.

The iron-induced macrophage population with unrestrained proinflammatory M1Figure 4

The iron-induced macrophage population with unrestrained proinflammatory M1 phenotype releases toxic amounts of OH• and ONOO• in situ. (A) Representative photomicrographs of skin cryosections derived from wound margins of AWs 2 days after wounding and CVUs. Oxidative burst was detected in cryosections incubated with DHR, a ROS-sensitive dye; thus, ROS concentrations correlated with green fluorescence. Shown are coincubations of cryosections with DHR and SOD, which scavenges O2–•, with the H2O2 scavenger Cat, and with the OH• scavenger DMSO. Nuclei were stained with DAPI. Scale bars: 150 μm. (B and C) Representative photomicrographs of paraffin-embedded skin sections from AWs 3 days after wounding and CVUs stained with (B) an antibody against 8OHdG and (C) an antibody against 3-NT. Dashed lines indicate the junction between epidermis and dermis. Scale bars: 150 μm; 50 μm (insets). Quantification of positive cells per 10 high-power fields, assessed for 10 different sections of 5 different AW and CVU samples, is shown as mean ± SD ratios of positive to total cells. (B) The brown-colored precipitate is indicative of oxidative DNA damage in CVUs (inset, arrows), but not AWs. (C) Increased protein nitration was observed in CVUs, but not AWs. (D) Nitroblot analysis of wound lysates with an antibody against 3-NT from PBS-treated mice and from iron-loaded mice and iron-loaded mice treated with DFX, etanercept, or clodronate equilibrated to actin. Positive bands indicate increased levels of protein nitration in iron-loaded wounds compared with reduced levels in iron-dextran–loaded mice treated with DFX, etanercept, or clodronate.

Enhanced macrophage-dependent ROS release activates a senescence program. OH• and ONOO• can severely damage macromolecules and can induce double strand breaks and inter-crosslinks in the DNA (44). Such lesions induce a senescence program in resident skin fibroblasts (45, 46) and thus may contribute to impaired wound healing. To test this hypothesis, γH2AX, a phosphorylated histone protein and aging marker (47) with specificity for DNA double strand breaks in vivo (45, 47), was studied. Double staining with antibodies against CD18, the common β chain of β2 integrins that is exclusively expressed on leukocytes, was used to differentiate γH2AX expression in inflammatory leukocytes from resident fibroblasts. In contrast to NS taken in sufficient distance from CVUs of the same patients’ legs, which showed almost no γH2AX expression in nuclei of resident fibroblasts (Figure 5A), a high number of fibroblasts adjacent to CVUs stained positive for γH2AX (Figure 5B). Thus, fibroblasts and not inflammatory cells revealed DNA damage. Western blot analysis of wound lysates confirmed these data (Supplemental Figure 9). Iron-loaded mice, in contrast to PBS-dextran–injected mice, showed enhanced induction of γH2AX that could be completely prevented by either DFX (Figure 5C) by clodronate liposome depletion of macrophages (data not shown). Immunostaining for p16INK4a, a robust in vivo marker for aging in murine and human skin (48, 49), revealed higher p16INK4a expression in CVUs, but not in NS of the patients’ same leg (Figure 5, D and E).

Enhanced ROS release by the unrestrained proinflammatory M1 activated macroFigure 5

Enhanced ROS release by the unrestrained proinflammatory M1 activated macrophage population activates a senescence program in resident wound fibroblasts. (A) Representative photomicrographs of cryosections from wound margins from CVUs and NS with increased numbers of γH2AX+ foci in the cell nuclei of wound margins. Double staining for γH2AX and the leukocyte marker CD18 revealed exclusive γH2AX expression in wound-adjacent fibroblasts. Nuclei were stained with DAPI. Scale bars: 150 μm; 50 μm (inset). (B) Quantification of γH2AX+ and γH2AX+CD18+ cells in 10 high-power fields from CVUs and NS samples (n = 5). Results are mean ± SD percent positive cells relative to total cells. **P < 0.01, Student’s t test. (C) Representative Western blot of wound lysates from iron-loaded and PBS control wounds with and without treatment with DFX, equilibrated to total actin, showing expression of γH2AX in iron-loaded, but not control or DFX-treated, wounds (n = 3). (D) Representative photomicrographs of cryosections of wound margins and NS derived from CVU patients showing increased numbers of p16INK4a in the cell nuclei of ulcer margins. Double staining for p16INK4a and CD18 showed p16INK4a expression exclusively in wound-adjacent fibroblasts, not in CD18+ cells. Nuclei were stained with DAPI. Scale bars: 150 μm. Original magnification of insets, ×400. (E) Quantification of p16INK4a+ and p16INK4a+CD18+ cells in 10 high-power fields from 5 different CVUs and NS samples. Results are mean ± SD percent positive cells relative to total cells. **P < 0.01, Student’s t test.

Collectively, our data indicate that the iron-induced activation of the proinflammatory M1 macrophage population contributes to enhanced DNA damage and senescence in skin resident fibroblasts and, hence, distinctly impairs their capacity for tissue repair.